Re: [ovs-dev] [PATCH] INSTALL.Windows.md: Update the minimum required compiler.

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

It would be good if we could verify that the driver is compiled and run when 
compiled with VS Update 4 and Winddk 8.1.
Thanks
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Tuesday, July 14, 2015 11:57 AM
To: dev@openvswitch.org
Cc: Gurucharan Shetty
Subject: [ovs-dev] [PATCH] INSTALL.Windows.md: Update the minimum required 
compiler.

MSVC 2013 update 4 was released in Nov 2014. Its release notes says that it has 
fixed the problem wherein using designated initializers to initialize unions 
within structs would fail to compile.

Using designated initializers is useful in OVS and so this commit updates the 
minimum required compiler for Windows.

Signed-off-by: Gurucharan Shetty gshe...@nicira.com
---
 INSTALL.Windows.md |6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md index 6d870ed..889bb0b 
100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -28,9 +28,9 @@ the following entry in /etc/fstab - 'C:/MinGW /mingw'.
 * Install the latest Python 2.x from python.org and verify that its path is  
part of Windows' PATH environment variable.
 
-* You will need at least Visual Studio 2013 to compile userspace binaries. In 
-addition to that, if you want to compile the kernel module you will also need 
to -install Windows Driver Kit (WDK) 8.1 Update.
+* You will need at least Visual Studio 2013 (update 4) to compile 
+userspace binaries.  In addition to that, if you want to compile the 
+kernel module you will also need to install Windows Driver Kit (WDK) 8.1 
Update.
 
 It is important to get the Visual Studio related environment variables and to  
have the $PATH inside the bash to point to the proper compiler and linker. One
--
1.7.9.5

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=j7iHpo0VgVdmCO_wv2COlEnWbdn-uyYnVKITELN9pcws=zM2m3mB4cNPSETWfrgDi3I-HnJJgvSwU8eNRJrCXZMce=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Realease dispatch lock after processing

[Eitan Eliahu]

Alin, it seems that RtlUnicodeStringToAnsiString() can't be called in an 
elevated IRQL.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, July 13, 2015 10:06 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Realease dispatch lock after 
processing

Release the switch context dispatch lock after all possible processing is done 
over the vport.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Vport.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/Vport.c b/datapath-windows/ovsext/Vport.c 
index 4315464..7cbe59e 100644
--- a/datapath-windows/ovsext/Vport.c
+++ b/datapath-windows/ovsext/Vport.c
@@ -1563,7 +1563,7 @@ OvsGetExtInfoIoctl(POVS_VPORT_GET vportGet,
 extInfo-vmUUID[0] = 0;
 extInfo-vifUUID[0] = 0;
 }
-NdisReleaseRWLock(gOvsSwitchContext-dispatchLock, lockState);
+
 if (doConvert) {
 status = OvsConvertIfCountedStrToAnsiStr(vport-portFriendlyName,
  extInfo-name, @@ -1593,6 
+1593,7 @@ OvsGetExtInfoIoctl(POVS_VPORT_GET vportGet,
  */
 status = STATUS_SUCCESS;
 }
+NdisReleaseRWLock(gOvsSwitchContext-dispatchLock, lockState);
 
 ext_info_done:
 return status;
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=GAWhBhugfvlx6aF4gdmQcZ32qoLSuu3_v0x8Ms7qdXos=hLdaxR10BjtU410yb92nRFmN3i3nTsFc_5lgrjnzia4e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: update extension information

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com 
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, July 13, 2015 10:18 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: update extension information

This patch sets additional information about the driver used by various

applications.



Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com

---

 datapath-windows/.gitignore|1 +

 datapath-windows/ovsext/ovsext.rc  |  104 ++-

 datapath-windows/ovsext/ovsext.vcxproj |1 +

 datapath-windows/ovsext/resource.h | 1576 

 4 files changed, 1662 insertions(+), 20 deletions(-)

 create mode 100644 datapath-windows/ovsext/resource.h



diff --git a/datapath-windows/.gitignore b/datapath-windows/.gitignore

index 356a6da..a3ed49b 100644

--- a/datapath-windows/.gitignore

+++ b/datapath-windows/.gitignore

@@ -3,3 +3,4 @@

 *.sdf

 /ovsext/x64/

 /x64

+/ovsext/ovsext.aps

diff --git a/datapath-windows/ovsext/ovsext.rc 
b/datapath-windows/ovsext/ovsext.rc

index 8f6e121..6446590 100644

--- a/datapath-windows/ovsext/ovsext.rc

+++ b/datapath-windows/ovsext/ovsext.rc

@@ -1,23 +1,87 @@

+// Microsoft Visual C++ generated resource script.

 //

-// Copyright (c) Microsoft Corporation. All Rights Reserved.

+#include resource.h

+/

+// English (United States) resources

+

+#if !defined(AFX_RESOURCE_DLL) || defined(AFX_TARG_ENU)

+LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US

+#pragma code_page(1252)

+

+/

+//

+// Version

+//

+

+VS_VERSION_INFO VERSIONINFO

+ FILEVERSION 6,3,9600,17298

+ PRODUCTVERSION 6,3,9600,17298

+ FILEFLAGSMASK 0x3fL

+#ifdef _DEBUG

+ FILEFLAGS 0x9L

+#else

+ FILEFLAGS 0x8L

+#endif

+ FILEOS 0x40004L

+ FILETYPE 0x3L

+ FILESUBTYPE 0x6L

+BEGIN

+BLOCK StringFileInfo

+BEGIN

+BLOCK 04b0

+BEGIN

+VALUE CompanyName, Open vSwitch

+VALUE FileDescription, Open vSwitch Extension

+VALUE FileVersion, 6.3.9600.17298

+VALUE InternalName, OVSExt.SYS

+VALUE LegalCopyright, � Licensed under the Apache License, 
Version 2.0 (the License)

+VALUE OriginalFilename, OVSExt.SYS

+VALUE ProductName, Open vSwitch 8/8.1 DDK driver

+VALUE ProductVersion, 6.3.9600.17298

+END

+END

+BLOCK VarFileInfo

+BEGIN

+VALUE Translation, 0x0, 1200

+END

+END

+

+

+#ifdef APSTUDIO_INVOKED

+/

+//

+// TEXTINCLUDE

 //

 

-#include windows.h

-#include ntverp.h

-

-/*---*/

-/* the following lines are specific to this file */

-/*---*/

-

-/* VER_FILETYPE, VER_FILESUBTYPE, VER_FILEDESCRIPTION_STR

- * and VER_INTERNALNAME_STR must be defined before including COMMON.VER

- * The strings don't need a '\0', since common.ver has them.

- */

-#defineVER_FILETYPEVFT_DRV

-#defineVER_FILESUBTYPE VFT2_DRV_NETWORK

-#define VER_FILEDESCRIPTION_STR VMWare OVS Extension

-#define VER_INTERNALNAME_STROVSExt.SYS

-#define VER_ORIGINALFILENAME_STROVSExt.SYS

-#define VER_LANGNEUTRAL

-

-#include common.ver

+1 TEXTINCLUDE

+BEGIN

+resource.h\0

+END

+

+2 TEXTINCLUDE

+BEGIN

+\0

+END

+

+3 TEXTINCLUDE

+BEGIN

+\r\n

+\0

+END

+

+#endif// APSTUDIO_INVOKED

+

+#endif// English (United States) resources

+/

+

+

+

+#ifndef APSTUDIO_INVOKED

+/

+//

+// Generated from the TEXTINCLUDE 3 resource.

+//

+

+

+/

+#endif// not APSTUDIO_INVOKED

diff --git a/datapath-windows/ovsext/ovsext.vcxproj 
b/datapath-windows/ovsext/ovsext.vcxproj

index 7050015..3367ca6 100644

--- a/datapath-windows/ovsext/ovsext.vcxproj

+++ b/datapath-windows/ovsext/ovsext.vcxproj

@@ -90,6 +90,7 @@

 ClInclude Include=PacketIO.h /

 ClInclude Include=PacketParser.h /

 ClInclude Include=precomp.h /

+ClInclude Include=resource.h /

 ClInclude Include=Stt.h /

 ClInclude Include=Switch.h /

 ClInclude Include=Tunnel.h /

diff --git a/datapath-windows/ovsext/resource.h 
b/datapath-windows/ovsext/resource.h

new file mode 100644

index 000..3294ef6

--- /dev/null

+++ b/datapath-windows/ovsext/resource.h

@@ -0,0 +1,1576 @@

+//{{NO_DEPENDENCIES}}

+// Microsoft Visual C++ generated include file.

+// Used by ovsext.rc

+//

+#define SW_HIDE 0

+#define HIDE_WINDOW

Re: [ovs-dev] [PATCH] datapath-windows: Modified dump start message memory representation

[Eitan Eliahu]

Alin, it is a different one.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, July 13, 2015 11:58 AM
To: Ben Pfaff; Sorin Vinturis
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: Modified dump start message 
memory representation

I am biased about the patch since this patch is already in:

https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs_commit_216e1c149a5c6d83884db447e16d30e063e1275dd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SNEISd9_Eld6TNkho7w3ebxXpWc_qJRQUxswVpcxOMIs=hiz_0swvd7FH-mSqSjuAuEduyuI7C6654LfexQhiWmIe=
 



Alin.



-Mesaj original-

De la: dev [mailto:dev-boun...@openvswitch.org] În numele Ben Pfaff

Trimis: Monday, July 13, 2015 8:54 PM

Către: Sorin Vinturis

Cc: dev@openvswitch.org

Subiect: Re: [ovs-dev] [PATCH] datapath-windows: Modified dump start message 
memory representation



On Fri, Jul 03, 2015 at 03:36:37PM +, Sorin Vinturis wrote:

 Changed the dump start message memory representation to be static.

 

 Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com



It's been over a week without a review.  I hope that someone looks at this soon.

___

dev mailing list

dev@openvswitch.org

https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SNEISd9_Eld6TNkho7w3ebxXpWc_qJRQUxswVpcxOMIs=77JFN63KVczTH9Qw9T0Vu-6-WXCqFhHYqKsljgZATDwe=
 

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SNEISd9_Eld6TNkho7w3ebxXpWc_qJRQUxswVpcxOMIs=77JFN63KVczTH9Qw9T0Vu-6-WXCqFhHYqKsljgZATDwe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Modified dump start message memory representation

[Eitan Eliahu]

Hi Sorin,
Thank you for addressing my comments. Can we please use a flag which indicates 
that the a dump process is in progress rather than setting the whole buffer to 
zero?
Besides of that, everything is good.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, July 03, 2015 8:37 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Modified dump start message memory 
representation

Changed the dump start message memory representation to be static.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Datapath.c | 53 +-
 datapath-windows/ovsext/Datapath.h |  8 +++---
 datapath-windows/ovsext/Flow.c |  4 +--
 datapath-windows/ovsext/Util.h |  5 
 datapath-windows/ovsext/Vport.c|  9 ---
 5 files changed, 40 insertions(+), 39 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index 4af909c..603008d 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -347,21 +347,16 @@ NDIS_SPIN_LOCK ovsCtrlLockObj;  PNDIS_SPIN_LOCK 
gOvsCtrlLock;
 
 NTSTATUS
-InitUserDumpState(POVS_OPEN_INSTANCE instance,
-  POVS_MESSAGE ovsMsg)
+SetUserDumpState(POVS_OPEN_INSTANCE instance,
+ POVS_MESSAGE ovsMsg)
 {
 /* Clear the dumpState from a previous dump sequence. */
-ASSERT(instance-dumpState.ovsMsg == NULL);
+ASSERT(OvsBufferIsEmpty((char*)instance-dumpState.ovsMsg,
+ sizeof(instance-dumpState.ovsMsg)));
 ASSERT(ovsMsg);
 
-instance-dumpState.ovsMsg =
-(POVS_MESSAGE)OvsAllocateMemoryWithTag(sizeof(OVS_MESSAGE),
-   OVS_DATAPATH_POOL_TAG);
-if (instance-dumpState.ovsMsg == NULL) {
-return STATUS_NO_MEMORY;
-}
-RtlCopyMemory(instance-dumpState.ovsMsg, ovsMsg,
-  sizeof *instance-dumpState.ovsMsg);
+RtlCopyMemory(instance-dumpState.ovsMsg, ovsMsg,
+  sizeof instance-dumpState.ovsMsg);
 RtlZeroMemory(instance-dumpState.index,
   sizeof instance-dumpState.index);
 
@@ -369,13 +364,9 @@ InitUserDumpState(POVS_OPEN_INSTANCE instance,  }
 
 VOID
-FreeUserDumpState(POVS_OPEN_INSTANCE instance)
+ClearUserDumpState(POVS_OPEN_INSTANCE instance)
 {
-if (instance-dumpState.ovsMsg != NULL) {
-OvsFreeMemoryWithTag(instance-dumpState.ovsMsg,
- OVS_DATAPATH_POOL_TAG);
-RtlZeroMemory(instance-dumpState, sizeof instance-dumpState);
-}
+RtlZeroMemory(instance-dumpState, sizeof instance-dumpState);
 }
 
 VOID
@@ -588,7 +579,6 @@ OvsRemoveOpenInstance(PFILE_OBJECT fileObject)
 OvsReleaseCtrlLock();
 ASSERT(instance-eventQueue == NULL);
 ASSERT (instance-packetQueue == NULL);
-FreeUserDumpState(instance);
 OvsFreeMemoryWithTag(instance, OVS_DATAPATH_POOL_TAG);  }
 
@@ -839,12 +829,14 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
  *
  * In the absence of a dump start, return 0 bytes.
  */
-if (instance-dumpState.ovsMsg == NULL) {
+if (OvsBufferIsEmpty((char*)instance-dumpState.ovsMsg,
+ sizeof(instance-dumpState.ovsMsg))) {
 replyLen = 0;
 status = STATUS_SUCCESS;
 goto done;
 }
-RtlCopyMemory(ovsMsgReadOp, instance-dumpState.ovsMsg,
+
+RtlCopyMemory(ovsMsgReadOp, instance-dumpState.ovsMsg,
   sizeof (ovsMsgReadOp));
 
 /* Create an NL message for consumption. */ @@ -1263,11 +1255,12 @@ 
HandleGetDpDump(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 } else {
 NL_BUFFER nlBuf;
 NTSTATUS status;
-POVS_MESSAGE msgIn = instance-dumpState.ovsMsg;
+POVS_MESSAGE msgIn = instance-dumpState.ovsMsg;
 
 ASSERT(usrParamsCtx-devOp == OVS_READ_DEV_OP);
 
-if (instance-dumpState.ovsMsg == NULL) {
+if (OvsBufferIsEmpty((char*)instance-dumpState.ovsMsg,
+ sizeof(instance-dumpState.ovsMsg))) {
 ASSERT(FALSE);
 return STATUS_INVALID_DEVICE_STATE;
 }
@@ -1285,7 +1278,7 @@ HandleGetDpDump(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 
 if (status != STATUS_SUCCESS) {
 *replyLen = 0;
-FreeUserDumpState(instance);
+ClearUserDumpState(instance);
 return status;
 }
 
@@ -1293,8 +1286,8 @@ HandleGetDpDump(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 instance-dumpState.index[0] = 1;
 *replyLen = msgOut-nlMsg.nlmsgLen;
 
-/* Free up the dump state, since there's no more data to continue. */
-FreeUserDumpState(instance);
+/* Clear the dump state, since there's no more data to continue. */
+ClearUserDumpState(instance);
 }
 
 return STATUS_SUCCESS;
@@ 

Re: [ovs-dev] [PATCH] datapath-windows: Update VXLAN header information

[Eitan Eliahu]

Hi Alin,
Can you please explain why the FilterID is used to maintain the number of 
transmitted bytes?
The FilterID is sent to WFP when we add a filter so the filter can be deleted 
later by specifying this ID.  Once the ID is modified the removal of the filter 
will fail.
I think we want to use as separate variable for it (e.g. STT code).
Thank you,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, July 13, 2015 8:39 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Update VXLAN header information

Use tunnel key information on the IP header preceding the VXLAN header.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Vxlan.c | 13 -
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/datapath-windows/ovsext/Vxlan.c b/datapath-windows/ovsext/Vxlan.c 
index f43805a..16c80b9 100644
--- a/datapath-windows/ovsext/Vxlan.c
+++ b/datapath-windows/ovsext/Vxlan.c
@@ -15,6 +15,7 @@
  */
 
 #include precomp.h
+#include Atomic.h
 #include NetProto.h
 #include Switch.h
 #include Vport.h
@@ -263,11 +264,13 @@ OvsDoEncapVxlan(POVS_VPORT_ENTRY vport,
 ipHdr = (IPHdr *)((PCHAR)ethHdr + sizeof *ethHdr);
 
 ipHdr-ihl = sizeof *ipHdr / 4;
-ipHdr-version = IPV4;
-ipHdr-tos = 0;
+ipHdr-version = IPPROTO_IPV4;
+ipHdr-tos = tunKey-tos;
 ipHdr-tot_len = htons(NET_BUFFER_DATA_LENGTH(curNb) - sizeof *ethHdr);
-ipHdr-id = 0;
-ipHdr-frag_off = IP_DF_NBO;
+ipHdr-id = (uint16)atomic_add64(vportVxlan-filterID,
+ NET_BUFFER_DATA_LENGTH(curNb));
+ipHdr-frag_off = (tunKey-flags  OVS_TNL_F_DONT_FRAGMENT) ?
+  IP_DF_NBO : 0;
 ipHdr-ttl = tunKey-ttl ? tunKey-ttl : VXLAN_DEFAULT_TTL;
 ipHdr-protocol = IPPROTO_UDP;
 ASSERT(tunKey-dst == fwdInfo-dstIpAddr); @@ -279,7 +282,7 @@ 
OvsDoEncapVxlan(POVS_VPORT_ENTRY vport,
 
 /* UDP header */
 udpHdr = (UDPHdr *)((PCHAR)ipHdr + sizeof *ipHdr);
-udpHdr-source = htons(tunKey-flow_hash | 32768);
+udpHdr-source = htons(tunKey-flow_hash | MAXINT16);
 udpHdr-dest = htons(vportVxlan-dstPort);
 udpHdr-len = htons(NET_BUFFER_DATA_LENGTH(curNb) - headRoom +
 sizeof *udpHdr + sizeof *vxlanHdr);
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Gka-eilVbOezK04WXSr9wUJYN7f8vbr2ZVKTk4ck4eos=-IPN21a76q0_LvUuzMjrtWNsLoKW1CQxqCzWu1QlB_Ae=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Solved memory leak in OVS datapath

[Eitan Eliahu]

Hi Sorin,
Please consider to have the dump state  as part of the usermode instance 
(regardless if dump initiated or not).  This would save the whole housekeeping 
around this insignificant dynamic memory  allocation.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Thursday, July 02, 2015 12:24 AM
To: Sorin Vinturis; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: Solved memory leak in OVS 
datapath

Acked-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com





-Mesaj original-

De la: dev [mailto:dev-boun...@openvswitch.org] În numele Sorin Vinturis

Trimis: Thursday, July 2, 2015 9:53 AM

Către: dev@openvswitch.org

Subiect: [ovs-dev] [PATCH] datapath-windows: Solved memory leak in OVS datapath



When closing opened instances, make sure the user dump state is cleared.



Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_90d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=y9vDcTc0RXTftcLjlqfZ62P-E9kaIrToeowiiE1Xsx4s=2ysWHleKrgA8nBthYFZPPmIcC0WCECIbOSO4Am6j_M8e=
 

---

This patch should be applied both on master and branch 2.4.

---

 datapath-windows/ovsext/Datapath.c | 1 +

 1 file changed, 1 insertion(+)



diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c

index d8024c8..4af909c 100644

--- a/datapath-windows/ovsext/Datapath.c

+++ b/datapath-windows/ovsext/Datapath.c

@@ -588,6 +588,7 @@ OvsRemoveOpenInstance(PFILE_OBJECT fileObject)

 OvsReleaseCtrlLock();

 ASSERT(instance-eventQueue == NULL);

 ASSERT (instance-packetQueue == NULL);

+FreeUserDumpState(instance);

 OvsFreeMemoryWithTag(instance, OVS_DATAPATH_POOL_TAG);  }

 

--

1.9.0.msysgit.0

___

dev mailing list

dev@openvswitch.org

https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=y9vDcTc0RXTftcLjlqfZ62P-E9kaIrToeowiiE1Xsx4s=ZlPfSZTKWL7e49wkW7YE6wpMtK_43vjgKepE_68sKqoe=
 

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=y9vDcTc0RXTftcLjlqfZ62P-E9kaIrToeowiiE1Xsx4s=ZlPfSZTKWL7e49wkW7YE6wpMtK_43vjgKepE_68sKqoe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Failed initialization for datapath tunnel ports

[Eitan Eliahu]

Hi Sorin,
Can you explain which change broke this code.
I recall validating connectivity on STT and I believe Nithin validated on STT 
and on VXLAN channel as well.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 12:52 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Failed initialization for datapath 
tunnel ports

Tunnel ports are not initialized with the corresponding default port.
The newly allocated vport is not yet initialized and the ovsType member does 
not reflect the correct tunnel port type, thus the transport port destination 
won't be correctly initialized.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_88d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=pfesL1KVpNw4qmTq5MZIQqau5e5kp7SsBE2NKZPDZ-cs=SNm8Qtz_cxNIYVDKYh4l_LI6uE5u2TiHBhM3JAF_WA4e=
---
 datapath-windows/ovsext/Vport.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/Vport.c b/datapath-windows/ovsext/Vport.c 
index 9139545..bff3ff6 100644
--- a/datapath-windows/ovsext/Vport.c
+++ b/datapath-windows/ovsext/Vport.c
@@ -2160,7 +2160,7 @@ OvsNewVportCmdHandler(POVS_USER_PARAMS_CONTEXT 
usrParamsCtx,
 if (OvsIsTunnelVportType(portType)) {
 UINT16 transportPortDest = 0;
 
-switch (vport-ovsType) {
+switch (portType) {
 case OVS_VPORT_TYPE_VXLAN:
 transportPortDest = VXLAN_UDP_PORT;
 break;
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=pfesL1KVpNw4qmTq5MZIQqau5e5kp7SsBE2NKZPDZ-cs=yyR0IBVadQHVJleSVrBVbjjppum2LP-iZVNNO2aYa_Ue=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Duplicate tag for STT pools

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 1:13 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Duplicate tag for STT pools

All pool allocations should have unique tags.
OVS_STT_POOL_TAG matches the OVS_MEMORY_TAG.

Changed the OVS_STT_POOL_TAG to make it unique.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Util.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index ee676fa..e3f9ede 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -33,7 +33,7 @@
 #define OVS_SWITCH_POOL_TAG 'SSVO'
 #define OVS_USER_POOL_TAG   'USVO'
 #define OVS_VPORT_POOL_TAG  'PSVO'
-#define OVS_STT_POOL_TAG'TSVO'
+#define OVS_STT_POOL_TAG'RSVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag);
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=5eunmaQkFeujFUE0-chSa0AEbfthYm7XlQqDiPWgNFEs=OJZprgyI70qvYIG-r2y36Z1d_cz_uXa_K5Sjehmpa6we=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 1/2] OVS_UNUSED is defined indirectly

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks,
Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, June 30, 2015 9:27 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 1/2] OVS_UNUSED is defined indirectly

Currently OVS_UNUSED is defined in compiler.h since syslog.h is a standalone 
wrapper remove it from the parameters.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 include/windows/syslog.h | 7 +++
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/include/windows/syslog.h b/include/windows/syslog.h index 
242bfc4..d59cd3d 100644
--- a/include/windows/syslog.h
+++ b/include/windows/syslog.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2013 Cloudbase Solutions Srl
+ * Copyright 2013, 2015 Cloudbase Solutions Srl
  *
  * Licensed under the Apache License, Version 2.0 (the License); you may
  * not use this file except in compliance with the License.You may obtain @@ 
-50,13 +50,12 @@
 #define LOG_LOCAL7  (233) /* reserved for local use */
 
 static inline void
-openlog(const char *ident OVS_UNUSED, int option OVS_UNUSED,
-int facility OVS_UNUSED)
+openlog(const char *ident, int option, int facility)
 {
 }
 
 static inline void
-syslog(int priority OVS_UNUSED, const char *format OVS_UNUSED, ...)
+syslog(int priority, const char *format, ...)
 {
 }
 
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=UiNKiAj_AxA1655axYT2tbB9LVY1qVFbj0fQGC-kbC0s=mkpAIljmIxKyESdXcIQxCwMOaRwtZ2qEBaRW2QX2FJEe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Avoid double incrementation of tx VXLAN packet statistics

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 1:04 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Avoid double incrementation of tx 
VXLAN packet statistics

VXLAN tunnel tx packet stats are incorrectly incremented.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_89d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=wJFMOfNAUyMKMAuFe66_D9sWneOoHJXXkynabNoC8aks=3LRuqX08bbZZE9doRqrXDD7cLKgbcUlrJfIo0sn8c9Ie=
 
---
 datapath-windows/ovsext/Actions.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/datapath-windows/ovsext/Actions.c 
b/datapath-windows/ovsext/Actions.c
index e5c122f..c8de7c5 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -313,7 +313,6 @@ OvsDetectTunnelPkt(OvsForwardingContext *ovsFwdCtx,
 ovsActionStats.txStt++;
 break;
 }
-ovsActionStats.txVxlan++;
 ovsFwdCtx-tunnelTxNic = dstVport;
 }
 
-- 
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=wJFMOfNAUyMKMAuFe66_D9sWneOoHJXXkynabNoC8aks=4kmTXoMnlAIIege3jgxvVtD3AZ-KeZYoKWrnRxrHQA0e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Failed initialization for datapath tunnel ports

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com 
Thanks,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Wednesday, July 01, 2015 8:27 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [PATCH] datapath-windows: Failed initialization for datapath 
tunnel ports

Eitan,

This code was changed by the STT patch. If the created tunnel has specified the 
tunnel destination port, then the issue does not occur. But if the created 
tunnel does not have a specified tunnel port, then the default port is not set 
and the tunnel destination port will be zero, thus the problem.

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Wednesday, 1 July, 2015 18:21
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [PATCH] datapath-windows: Failed initialization for datapath 
tunnel ports

Hi Sorin,
Can you explain which change broke this code.
I recall validating connectivity on STT and I believe Nithin validated on STT 
and on VXLAN channel as well.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 12:52 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Failed initialization for datapath 
tunnel ports

Tunnel ports are not initialized with the corresponding default port.
The newly allocated vport is not yet initialized and the ovsType member does 
not reflect the correct tunnel port type, thus the transport port destination 
won't be correctly initialized.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_88d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=pfesL1KVpNw4qmTq5MZIQqau5e5kp7SsBE2NKZPDZ-cs=SNm8Qtz_cxNIYVDKYh4l_LI6uE5u2TiHBhM3JAF_WA4e=
---
 datapath-windows/ovsext/Vport.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/Vport.c b/datapath-windows/ovsext/Vport.c 
index 9139545..bff3ff6 100644
--- a/datapath-windows/ovsext/Vport.c
+++ b/datapath-windows/ovsext/Vport.c
@@ -2160,7 +2160,7 @@ OvsNewVportCmdHandler(POVS_USER_PARAMS_CONTEXT 
usrParamsCtx,
 if (OvsIsTunnelVportType(portType)) {
 UINT16 transportPortDest = 0;
 
-switch (vport-ovsType) {
+switch (portType) {
 case OVS_VPORT_TYPE_VXLAN:
 transportPortDest = VXLAN_UDP_PORT;
 break;
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=pfesL1KVpNw4qmTq5MZIQqau5e5kp7SsBE2NKZPDZ-cs=yyR0IBVadQHVJleSVrBVbjjppum2LP-iZVNNO2aYa_Ue=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 2/2] Undefined symbol make_unix_socket

[Eitan Eliahu]

Hi Alin,
Should we implement OVS system logging for Windows through the use of 
Application Event Logging?
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, June 30, 2015 9:28 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 2/2] Undefined symbol make_unix_socket

syslog_direct_create defined in (lib/syslog-direct.c) uses make_unix_socket 
which is currently undefined on the windows build.

We either can remove the new file from the chain but this patch proposes to 
define a wrapper to make_unix_socket in which we return EINVAL.

This will avoid this kind of problems in the future.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 lib/socket-util.h | 7 +++
 1 file changed, 7 insertions(+)

diff --git a/lib/socket-util.h b/lib/socket-util.h index 1178fb8..a0a5c36 100644
--- a/lib/socket-util.h
+++ b/lib/socket-util.h
@@ -95,6 +95,13 @@ int af_inet_ifreq_ioctl(const char *name, struct ifreq *,  
#endif
 
 #ifdef _WIN32
+static inline int make_unix_socket(int style, bool nonblock,
+   const char *bind_path,
+   const char *connect_path) {
+return EINVAL;
+}
+
 /* Windows defines the 'optval' argument as char * instead of void *. */  
#define setsockopt(sock, level, optname, optval, optlen) \
 rpl_setsockopt(sock, level, optname, optval, optlen)
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=y_igwxCxxtgun7X4BmzxDG2ek0C_h9frh1L9ukvLkucs=zcKFCr4PuGr2y2JPn60J1jdtIFKCtjqoPxtLm4kw0gYe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS ports

[Eitan Eliahu]

Hi Sorin,
Thank you for debugging this issue.
I think that the NBL context signature is used for code validation only (think 
about a signature we put in an object for the purpose of validating this object 
when executing in a different context).
Can you think about another way to fix this issue (e.g. a status/flag)?
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 7:02 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS ports

This BSOD occurred in the context of a packet (NBL) with multiple
NET_BUFFER(s) (NBs). The reason for the BSOD is due to the marking of NBLs 
created by OVS as being external and wrongly completing them.

This patch should be applied both on master and branch 2.4.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_82d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=yx7HeX13135X7eeWKg3s73H3l711NpzfUes52_NsBqQs=3pshLMQB4SoTOb8sZncux7Bi6G5ADQwH7I0TlH9Is2Ue=
---
 datapath-windows/ovsext/BufferMgmt.c | 8 
 1 file changed, 8 insertions(+)

diff --git a/datapath-windows/ovsext/BufferMgmt.c 
b/datapath-windows/ovsext/BufferMgmt.c
index 83d6cde..d36dfa2 100644
--- a/datapath-windows/ovsext/BufferMgmt.c
+++ b/datapath-windows/ovsext/BufferMgmt.c
@@ -558,6 +558,12 @@ OvsInitExternalNBLContext(PVOID ovsContext,
 NDIS_STATUS status;
 UINT16 flags;
 
+ctx = (POVS_BUFFER_CONTEXT)NET_BUFFER_LIST_CONTEXT_DATA_START(nbl);
+if (ctx  ctx-magic == OVS_CTX_MAGIC) {
+OVS_LOG_INFO(nbl context already initialized/allocated by OVS.);
+return ctx;
+}
+
 poolHandle = NdisGetPoolFromNetBufferList(nbl);
 
 if (poolHandle == context-ovsPool.ndisHandle) { @@ -801,6 +807,7 @@ 
OvsPartialCopyNBL(PVOID ovsContext,
   OVS_DEFAULT_PORT_NO);
 
 InterlockedIncrement((LONG volatile *)srcCtx-refCount);
+
 #ifdef DBG
 OvsDumpNetBufferList(nbl);
 OvsDumpForwardingDetails(nbl);
@@ -808,6 +815,7 @@ OvsPartialCopyNBL(PVOID ovsContext,
 OvsDumpNetBufferList(newNbl);
 OvsDumpForwardingDetails(newNbl);
 #endif
+
 OVS_LOG_LOUD(Partial Copy new NBL: %p, newNbl);
 return newNbl;
 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=yx7HeX13135X7eeWKg3s73H3l711NpzfUes52_NsBqQs=M6NSK31Vva4-yAxaCNBMGVI5d9vESZHrNxsIVtrjJEEe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, June 30, 2015 9:39 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

Since commit:
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs_commit_2e460098bff351b9fddcb917447caa3b97a35d86d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=uAgTXNKjBK-e8P2evuJh9E4v1c6DHvqSWZL7ljeVFm8s=mvXIWGGUYLF3C5iDvNdcIGD-IZlT2ZufDJGcdvQx1sce=
a new packet attribute was introduced.

This patch adds OVS_PACKET_ATTR_PROBE to nlPktExecPolicy in datapath-windows 
and ignores it for the moment to maintain binary compatibility.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 datapath-windows/ovsext/User.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/User.c b/datapath-windows/ovsext/User.c 
index 9f462cf..56f2f7c 100644
--- a/datapath-windows/ovsext/User.c
+++ b/datapath-windows/ovsext/User.c
@@ -332,7 +332,8 @@ OvsNlExecuteCmdHandler(POVS_USER_PARAMS_CONTEXT 
usrParamsCtx,
 [OVS_PACKET_ATTR_ACTIONS] = {.type = NL_A_UNSPEC, .optional = FALSE},
 [OVS_PACKET_ATTR_USERDATA] = {.type = NL_A_UNSPEC, .optional = TRUE},
 [OVS_PACKET_ATTR_EGRESS_TUN_KEY] = {.type = NL_A_UNSPEC,
-.optional = TRUE}
+.optional = TRUE},
+[OVS_PACKET_ATTR_PROBE] = { .type = NL_A_FLAG, .optional = TRUE 
+ }
 };
 
 RtlZeroMemory(execute, sizeof(OvsPacketExecute));
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=uAgTXNKjBK-e8P2evuJh9E4v1c6DHvqSWZL7ljeVFm8s=Ck0jQ06QL47kPSKcCUPm1YJccmz6EL65hwB80BQIMWke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS ports

[Eitan Eliahu]

Sorin,
I do understand the issue but the use of the signature to solve it is not 
clean. The signature field in the context structure is used for validation 
during debug. It could be removed and the code should still work.
We should not use the signature to identify self-allocated NBLs. 
Please consider to use the pool handle (ovsPool.ndisHandle) instaed.
Thanks,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Wednesday, July 01, 2015 8:50 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS 
ports

Eitan,

Consider the scenario below.

We receive a packet (NBL), that has multiple NBs, in OvsStartNBLIngress(). The 
NBL is not allocated by OVS and the NBL context is allocated and initialized in 
OvsInitExternalNBLContext(). After this the OvsCreateNewNBLsFromMultipleNBs() 
function is called and the NBL with multiple NBs is splitted into multiple NBLs 
with single NB, the latter NBLs being allocated by OVS. The execution continues 
with the first newly allocated NBL with single NB and all goes well until the 
loop goes to the next NBL with single NB. For this second newly created NBL 
with single NB the OvsInitExternalNBLContext() is called. Now we need to 
identify the NBL as being allocated by OVS, so we look at the NBL context 
signature.

I think this is a clean way to identify OVS NBLs, when calling 
OvsInitExternalNBLContext().

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Wednesday, 1 July, 2015 18:12
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS 
ports

Hi Sorin,
Thank you for debugging this issue.
I think that the NBL context signature is used for code validation only (think 
about a signature we put in an object for the purpose of validating this object 
when executing in a different context).
Can you think about another way to fix this issue (e.g. a status/flag)?
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, July 01, 2015 7:02 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when adding OVS ports

This BSOD occurred in the context of a packet (NBL) with multiple
NET_BUFFER(s) (NBs). The reason for the BSOD is due to the marking of NBLs 
created by OVS as being external and wrongly completing them.

This patch should be applied both on master and branch 2.4.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_82d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=yx7HeX13135X7eeWKg3s73H3l711NpzfUes52_NsBqQs=3pshLMQB4SoTOb8sZncux7Bi6G5ADQwH7I0TlH9Is2Ue=
---
 datapath-windows/ovsext/BufferMgmt.c | 8 
 1 file changed, 8 insertions(+)

diff --git a/datapath-windows/ovsext/BufferMgmt.c 
b/datapath-windows/ovsext/BufferMgmt.c
index 83d6cde..d36dfa2 100644
--- a/datapath-windows/ovsext/BufferMgmt.c
+++ b/datapath-windows/ovsext/BufferMgmt.c
@@ -558,6 +558,12 @@ OvsInitExternalNBLContext(PVOID ovsContext,
 NDIS_STATUS status;
 UINT16 flags;
 
+ctx = (POVS_BUFFER_CONTEXT)NET_BUFFER_LIST_CONTEXT_DATA_START(nbl);
+if (ctx  ctx-magic == OVS_CTX_MAGIC) {
+OVS_LOG_INFO(nbl context already initialized/allocated by OVS.);
+return ctx;
+}
+
 poolHandle = NdisGetPoolFromNetBufferList(nbl);
 
 if (poolHandle == context-ovsPool.ndisHandle) { @@ -801,6 +807,7 @@ 
OvsPartialCopyNBL(PVOID ovsContext,
   OVS_DEFAULT_PORT_NO);
 
 InterlockedIncrement((LONG volatile *)srcCtx-refCount);
+
 #ifdef DBG
 OvsDumpNetBufferList(nbl);
 OvsDumpForwardingDetails(nbl);
@@ -808,6 +815,7 @@ OvsPartialCopyNBL(PVOID ovsContext,
 OvsDumpNetBufferList(newNbl);
 OvsDumpForwardingDetails(newNbl);
 #endif
+
 OVS_LOG_LOUD(Partial Copy new NBL: %p, newNbl);
 return newNbl;
 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=yx7HeX13135X7eeWKg3s73H3l711NpzfUes52_NsBqQs=M6NSK31Vva4-yAxaCNBMGVI5d9vESZHrNxsIVtrjJEEe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 2/2] Undefined symbol make_unix_socket

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: Alin Serdean [mailto:aserd...@cloudbasesolutions.com] 
Sent: Wednesday, July 01, 2015 11:49 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [PATCH 2/2] Undefined symbol make_unix_socket

Hi Eitan,



It would be a nice idea for a future feature. I think this issue is related to 
that: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_74d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=mnhiuiHZ3Anzhv-e-D4lRdTb1bqvx_i7JKZ6mjDQ7Hgs=U_DHjCEmSBPo2PVDGEy_8cew0YmY3QVld4tLY1Hrqg4e=
 .



Alin.



-Mesaj original-

De la: Eitan Eliahu [mailto:elia...@vmware.com] 

Trimis: Wednesday, July 1, 2015 6:03 PM

Către: Alin Serdean; dev@openvswitch.org

Subiect: RE: [PATCH 2/2] Undefined symbol make_unix_socket





Hi Alin,

Should we implement OVS system logging for Windows through the use of 
Application Event Logging?

Thanks,

Eitan



-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean

Sent: Tuesday, June 30, 2015 9:28 AM

To: dev@openvswitch.org

Subject: [ovs-dev] [PATCH 2/2] Undefined symbol make_unix_socket



syslog_direct_create defined in (lib/syslog-direct.c) uses make_unix_socket 
which is currently undefined on the windows build.



We either can remove the new file from the chain but this patch proposes to 
define a wrapper to make_unix_socket in which we return EINVAL.



This will avoid this kind of problems in the future.



Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com

---

 lib/socket-util.h | 7 +++

 1 file changed, 7 insertions(+)



diff --git a/lib/socket-util.h b/lib/socket-util.h index 1178fb8..a0a5c36 100644

--- a/lib/socket-util.h

+++ b/lib/socket-util.h

@@ -95,6 +95,13 @@ int af_inet_ifreq_ioctl(const char *name, struct ifreq *,  
#endif

 

 #ifdef _WIN32

+static inline int make_unix_socket(int style, bool nonblock,

+   const char *bind_path,

+   const char *connect_path) {

+return EINVAL;

+}

+

 /* Windows defines the 'optval' argument as char * instead of void *. */  
#define setsockopt(sock, level, optname, optval, optlen) \

 rpl_setsockopt(sock, level, optname, optval, optlen)

--

1.9.5.msysgit.0

___

dev mailing list

dev@openvswitch.org

https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=y_igwxCxxtgun7X4BmzxDG2ek0C_h9frh1L9ukvLkucs=zcKFCr4PuGr2y2JPn60J1jdtIFKCtjqoPxtLm4kw0gYe=
 

___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

[Eitan Eliahu]

Hi Ben,
In general it doesn't seem that if the datapath does not support an attribute 
it would fail. (In other words the parsing validation code will not fail if an 
attribute is not included in the policy array).
I am not sure if the addition of this specific attribute to the policy masks an 
issue or not.
Perhaps Alin can tell.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
Sent: Wednesday, July 01, 2015 2:53 PM
To: Alin Serdean
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

On Wed, Jul 01, 2015 at 06:57:50PM +, Alin Serdean wrote:
 Since commit:
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvs
 witch_ovs_commit_2e460098bff351b9fddcb917447caa3b97a35d86d=BQIGaQc=S
 qcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e
 3geGYp56lkcH-5cLCYm=wSN1s5IDTOoE-dDGQWtmN87LEkQnYq71hKWcn7d3SJks=8_A
 Va78P_KXRWsNoVVtFSjNbBrJjyie8OjjjH4U953Ue=
 a new packet attribute was introduced.
 
 This patch adds OVS_PACKET_ATTR_PROBE to nlPktExecPolicy in 
 datapath-windows and ignores it for the moment to maintain binary 
 compatibility.
 
 Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
 Acked-by: Eitan Eliahu elia...@vmware.com
 ---
 This patch should be applied on master and branch-2.4
 v2: add acked-by

This patch makes me worry a bit.  The Linux kernel and OVS userspace 
implementations of Netlink ignore attributes that they do not recognize, which 
allows for some kinds of compatibility.  To me, this patch implies that the 
Windows datapath reports an error when it sees an unrecognized attribute.  Is 
that correct?  If so, then that should be changed, because otherwise it will 
cause compatibility problems in the future.
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=wSN1s5IDTOoE-dDGQWtmN87LEkQnYq71hKWcn7d3SJks=TZgQzcgvt8NDu-lf_W0nI9BkcWxoExr-9T2PSSonAbke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

[Eitan Eliahu]

Yes, this would make the parse function to fail and in fact it will access the 
policy array beyond its boundary.
Thanks Alin and Ben,
Eitan

-Original Message-
From: Alin Serdean [mailto:aserd...@cloudbasesolutions.com] 
Sent: Wednesday, July 01, 2015 5:13 PM
To: Ben Pfaff; Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

Sorry was relocating home.

Thanks for raising the question, it made me realize another problem :)

To answer your question directly: no this patch just ignores the attribute 
completely it does not do any additional operations.

I introduced the attribute because it was failing at 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs_blob_master_datapath-2Dwindows_ovsext_User.c-23L341d=BQIFBAc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=j9PYXgmBEpEQ5TpurOxWZwzwvN4XAFYcfCY5m02h7dks=7JHTeOT7zkl1tqjWKBXmNXZrAB6YtqNxiYOyTN6vedoe=
 , and to be more direct because the optional parameter was zero but apparently 
this is just the symptom.

Due to your question I relooked over the code and found the actual problem: 
when we are calling NlAttrParse we are calling it with array size of the 
__OVS_PACKET_ATTR_MAX(the one defined in openvswitch.h) when we should call it 
with the array size of the policy itself.

I will look over the windows datapath code to see if we slipped in that mistake 
in other parts as well and will try to send the patches asap.

Regarding the patch it is semi dead weight because it just introduces the 
attribute but no real logic is behind (it will work with and without it).

Alin.

-Mesaj original-
De la: Ben Pfaff [mailto:b...@nicira.com]
Trimis: Thursday, July 2, 2015 2:36 AM
Către: Eitan Eliahu
Cc: Alin Serdean; dev@openvswitch.org
Subiect: Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE

Alin?

On Wed, Jul 01, 2015 at 11:04:57PM +, Eitan Eliahu wrote:
 Hi Ben,
 In general it doesn't seem that if the datapath does not support an attribute 
 it would fail. (In other words the parsing validation code will not fail if 
 an attribute is not included in the policy array).
 I am not sure if the addition of this specific attribute to the policy masks 
 an issue or not.
 Perhaps Alin can tell.
 Thanks,
 Eitan
 
 -Original Message-
 From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
 Sent: Wednesday, July 01, 2015 2:53 PM
 To: Alin Serdean
 Cc: dev@openvswitch.org
 Subject: Re: [ovs-dev] [PATCH] datapath-windows: OVS_PACKET_ATTR_PROBE
 
 On Wed, Jul 01, 2015 at 06:57:50PM +, Alin Serdean wrote:
  Since commit:
  https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_open
  vs
  witch_ovs_commit_2e460098bff351b9fddcb917447caa3b97a35d86d=BQIGaQc
  =S
  qcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ
  4e
  3geGYp56lkcH-5cLCYm=wSN1s5IDTOoE-dDGQWtmN87LEkQnYq71hKWcn7d3SJks=8
  _A Va78P_KXRWsNoVVtFSjNbBrJjyie8OjjjH4U953Ue=
  a new packet attribute was introduced.
  
  This patch adds OVS_PACKET_ATTR_PROBE to nlPktExecPolicy in 
  datapath-windows and ignores it for the moment to maintain binary 
  compatibility.
  
  Signed-off-by: Alin Gabriel Serdean 
  aserd...@cloudbasesolutions.com
  Acked-by: Eitan Eliahu elia...@vmware.com
  ---
  This patch should be applied on master and branch-2.4
  v2: add acked-by
 
 This patch makes me worry a bit.  The Linux kernel and OVS userspace 
 implementations of Netlink ignore attributes that they do not recognize, 
 which allows for some kinds of compatibility.  To me, this patch implies that 
 the Windows datapath reports an error when it sees an unrecognized attribute. 
  Is that correct?  If so, then that should be changed, because otherwise it 
 will cause compatibility problems in the future.
 ___
 dev mailing list
 dev@openvswitch.org
 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_ma
 ilman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-
 uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=wSN1s5IDTOoE-dDGQW
 tmN87LEkQnYq71hKWcn7d3SJks=TZgQzcgvt8NDu-lf_W0nI9BkcWxoExr-9T2PSSonAb
 ke=
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows datapath

[Eitan Eliahu]

Alin,
We just need to add:
[OVS_FLOW_ATTR_PROBE] = { .type = NLA_FLAG },
to the flow policy. This is not a hack.
If it is time consuming please go ahead with the uder mode change.
Thanks,
Eitan


-Original Message-
From: Alin Serdean [mailto:aserd...@cloudbasesolutions.com] 
Sent: Thursday, June 25, 2015 7:30 AM
To: Nithin Raju; Eitan Eliahu
Cc: Ben Pfaff; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

Both are hacks IMO. 

We could respond to DPIF_FP_PROBE but I prefer to do it in userspace until we 
address the bigger problem in the datapath.

Alin.

-Mesaj original-
De la: Nithin Raju [mailto:nit...@vmware.com]
Trimis: Thursday, June 25, 2015 4:51 PM
Către: Eitan Eliahu
Cc: Ben Pfaff; Alin Serdean; dev@openvswitch.org
Subiect: Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

 On Jun 24, 2015, at 12:57 PM, Eitan Eliahu elia...@vmware.com wrote:
 
 
 I'm wondering if we could fail this call in the datapath level by examining 
 the DPIF_FP_PROBE bit in the flags.
 If it is too hard we can still live with user mode code change.
 Thanks,
 Eitan

That’s a good suggestion. We can go with Alin’s solution for 2.4 since we 
exactly know what code we will be disabling, and work on the flow validation 
logic for master.

thanks,
-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] [PATCH v3][branch-2.4] dpif_probe_feature not suported in windows datapath

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Thursday, June 25, 2015 10:19 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] [PATCH v3][branch-2.4] dpif_probe_feature not 
suported in windows datapath

This patch disables features which are not currently supported in the windows 
datapath.

Unfortunately we have to do it in userspace because dpif_probe_feature is not 
treated accordingly in the windows the datapath.

I opened the issue to track the feature for later implementations:
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_85d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=HJQhy35_OR1JNLBJwkqAnyF4JP3rIFj7afqs8puK-5Qs=DQfR4oXdznIeaYj4XE1I8Np2yJchVsAJQbB2paQyOf8e=
 

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
v3: Add Sign-off-by
v2: Rebase
---
---
 ofproto/ofproto-dpif.c | 17 +
 1 file changed, 17 insertions(+)

diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c index 
369e0b9..6cdaa32 100644
--- a/ofproto/ofproto-dpif.c
+++ b/ofproto/ofproto-dpif.c
@@ -1010,8 +1010,14 @@ check_recirc(struct dpif_backer *backer)
 
 ofpbuf_use_stack(key, keybuf, sizeof keybuf);
 odp_flow_key_from_flow(key, flow, NULL, 0, true);
+#ifdef _WIN32
+/* XXX Force disable of datapath recirculation from userspace until the
+ * dpif_probe_feature is properly implemented in the windows datapath */
+enable_recirc = false;
+#else
 enable_recirc = dpif_probe_feature(backer-dpif, recirculation, key,
NULL);
+#endif
 
 if (enable_recirc) {
 VLOG_INFO(%s: Datapath supports recirculation, @@ -1045,7 +1051,13 
@@ check_ufid(struct dpif_backer *backer)
 odp_flow_key_from_flow(key, flow, NULL, 0, true);
 dpif_flow_hash(backer-dpif, key.data, key.size, ufid);
 
+#ifdef _WIN32
+/* XXX Force disable of datapath recirculation from userspace until the
+ * dpif_probe_feature is properly implemented in the windows datapath */
+enable_ufid = false;
+#else
 enable_ufid = dpif_probe_feature(backer-dpif, UFID, key, ufid);
+#endif
 
 if (enable_ufid) {
 VLOG_INFO(%s: Datapath supports unique flow ids, @@ -1151,6 +1163,11 
@@ check_max_mpls_depth(struct dpif_backer *backer)
 
 ofpbuf_use_stack(key, keybuf, sizeof keybuf);
 odp_flow_key_from_flow(key, flow, NULL, 0, false);
+#ifdef _WIN32
+/* XXX Force disable of datapath recirculation from userspace until the
+ * dpif_probe_feature is properly implemented in the windows datapath 
*/
+break;
+#endif
 if (!dpif_probe_feature(backer-dpif, MPLS, key, NULL)) {
 break;
 }
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=HJQhy35_OR1JNLBJwkqAnyF4JP3rIFj7afqs8puK-5Qs=qV4ezZ0iFbhztu1BiqvQsHBK0NjZa-WCQ0gQpUm4TvIe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows datapath

[Eitan Eliahu]

If it gets too involved have your user mode hack instead.
Thanks Alin.
Eitan

-Original Message-
From: Alin Serdean [mailto:aserd...@cloudbasesolutions.com] 
Sent: Thursday, June 25, 2015 8:38 AM
To: Eitan Eliahu; Nithin Raju
Cc: Ben Pfaff; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

Eitan,

Ok I will add it in the datapath.

Why I am considering it as a hack, because if we see it in the datapath we just 
reply with an error, there will be no logic behind it at the moment. But maybe 
it makes some sense to have it for further user.

Alin.

-Mesaj original-
De la: Eitan Eliahu [mailto:elia...@vmware.com] 
Trimis: Thursday, June 25, 2015 6:01 PM
Către: Alin Serdean; Nithin Raju
Cc: Ben Pfaff; dev@openvswitch.org
Subiect: RE: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath



Alin,
We just need to add:
[OVS_FLOW_ATTR_PROBE] = { .type = NLA_FLAG }, to the flow policy. This 
is not a hack.
If it is time consuming please go ahead with the uder mode change.
Thanks,
Eitan


-Original Message-
From: Alin Serdean [mailto:aserd...@cloudbasesolutions.com]
Sent: Thursday, June 25, 2015 7:30 AM
To: Nithin Raju; Eitan Eliahu
Cc: Ben Pfaff; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

Both are hacks IMO. 

We could respond to DPIF_FP_PROBE but I prefer to do it in userspace until we 
address the bigger problem in the datapath.

Alin.

-Mesaj original-
De la: Nithin Raju [mailto:nit...@vmware.com]
Trimis: Thursday, June 25, 2015 4:51 PM
Către: Eitan Eliahu
Cc: Ben Pfaff; Alin Serdean; dev@openvswitch.org
Subiect: Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

 On Jun 24, 2015, at 12:57 PM, Eitan Eliahu elia...@vmware.com wrote:
 
 
 I'm wondering if we could fail this call in the datapath level by examining 
 the DPIF_FP_PROBE bit in the flags.
 If it is too hard we can still live with user mode code change.
 Thanks,
 Eitan

That’s a good suggestion. We can go with Alin’s solution for 2.4 since we 
exactly know what code we will be disabling, and work on the flow validation 
logic for master.

thanks,
-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows datapath

[Eitan Eliahu]

I'm wondering if we could fail this call in the datapath level by examining the 
DPIF_FP_PROBE bit in the flags.
If it is too hard we can still live with user mode code change.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
Sent: Wednesday, June 24, 2015 11:05 AM
To: Alin Serdean
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in windows 
datapath

I guess this solution is OK with me for 2.4.  Nithin (or anyone), do you agree 
that it's the best solution for now?

On Fri, Jun 19, 2015 at 04:41:47PM +, Alin Serdean wrote:
 To do it the proper way yes.
 
 I really hate to do it this way but given the timeframe for 2.4 we could nuke 
 it later on.
 
 Alin.
 
 -Mesaj original-
 De la: Nithin Raju [mailto:nit...@vmware.com]
 Trimis: Friday, June 19, 2015 7:36 PM
 Către: Alin Serdean
 Cc: dev@openvswitch.org
 Subiect: Re: [ovs-dev] [PATCH v2] dpif_probe_feature not suported in 
 windows datapath
 
 hi Alin,
 Do you know what is required to complete the implementation of 
 dpif_probe_feature() in the Windows datapath? Is it the flow validation logic?
 
 thanks,
 -- Nithin
 
 
  On Jun 18, 2015, at 6:47 PM, Alin Serdean aserd...@cloudbasesolutions.com 
  wrote:
  
  This patch disables features which are not currently supported in 
  the windows datapath.
  
  Unfortunately we have to do it in userspace because 
  dpif_probe_feature is not treated accordingly in the windows the datapath.
  
  I opened the issue to track the feature for later implementations:
  https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_open
  vs
  witch_ovs-2Dissues_issues_85d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeA
  w- 
  YihVMNtXt-uEsr=pNHQcdr7B40b4h6Yb7FIedI1dnBsxdDuTLBYD3JqV80m=H9ymRG
  wE 
  SDpjH0qHp5EifPK6QimZgzvHEIyjM3QzRwgs=-rOo7BG2BoxcGi2BlSQASDz8v5Hkhp
  4y
  MoGbPZ-zQuse=
  ---
  v2: Rebase
  ---
  ofproto/ofproto-dpif.c | 17 +
  1 file changed, 17 insertions(+)
  
  diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c index 
  0de8686..55d004e 100644
  --- a/ofproto/ofproto-dpif.c
  +++ b/ofproto/ofproto-dpif.c
  @@ -1014,8 +1014,14 @@ check_recirc(struct dpif_backer *backer)
  
  ofpbuf_use_stack(key, keybuf, sizeof keybuf);
  odp_flow_key_from_flow(odp_parms, key);
  +#ifdef _WIN32
  +/* XXX Force disable of datapath recirculation from userspace until the
  + * dpif_probe_feature is properly implemented in the windows datapath 
  */
  +enable_recirc = false;
  +#else
  enable_recirc = dpif_probe_feature(backer-dpif, recirculation, key,
 NULL);
  +#endif
  
  if (enable_recirc) {
  VLOG_INFO(%s: Datapath supports recirculation, @@ -1052,7
  +1058,13 @@ check_ufid(struct dpif_backer *backer)
  odp_flow_key_from_flow(odp_parms, key);
  dpif_flow_hash(backer-dpif, key.data, key.size, ufid);
  
  +#ifdef _WIN32
  +/* XXX Force disable of datapath recirculation from userspace until the
  + * dpif_probe_feature is properly implemented in the windows datapath 
  */
  +enable_ufid = false;
  +#else
  enable_ufid = dpif_probe_feature(backer-dpif, UFID, key, 
  ufid);
  +#endif
  
  if (enable_ufid) {
  VLOG_INFO(%s: Datapath supports unique flow ids, @@ 
  -1161,6
  +1173,11 @@ check_max_mpls_depth(struct dpif_backer *backer)
  
  ofpbuf_use_stack(key, keybuf, sizeof keybuf);
  odp_flow_key_from_flow(odp_parms, key);
  +#ifdef _WIN32
  +/* XXX Force disable of datapath recirculation from userspace 
  until the
  + * dpif_probe_feature is properly implemented in the windows 
  datapath */
  +break;
  +#endif
  if (!dpif_probe_feature(backer-dpif, MPLS, key, NULL)) {
  break;
  }
  --
  1.9.5.msysgit.0
  ___
  dev mailing list
  dev@openvswitch.org
  https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_
  ma
  ilman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtX
  t-
  uEsr=pNHQcdr7B40b4h6Yb7FIedI1dnBsxdDuTLBYD3JqV80m=H9ymRGwESDpjH0qH
  p5 
  EifPK6QimZgzvHEIyjM3QzRwgs=voVGofZZJTP055uwJnpQBjFDxPbFyaiHEtEtt140
  MO
  Ee=
 
 ___
 dev mailing list
 dev@openvswitch.org
 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_ma
 ilman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-
 uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=VgcETFFbbuqnE2agha
 rVPMWHqSyP7L2kf73r940PsMgs=HMaPgTvdxvCtlAQc-yi3VXOLpNatDI43OgRlwifibU
 Me=
___
dev mailing list
dev@openvswitch.org

Re: [ovs-dev] [PATCH v2] datapath-windows: Return success for already existing WFP objects

[Eitan Eliahu]

Hi Sorin,
Although the patch by itself can't hurt but I would suggest that we would root 
cause the issue of not freeing WFP object on error conditions.
Thank you,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Monday, June 22, 2015 10:26 AM
To: Eitan Eliahu
Cc: Nithin Raju; dev@openvswitch.org
Subject: RE: [PATCH v2] datapath-windows: Return success for already existing 
WFP objects

Hi Eitan/Nithin,

In normal execution flow all WFP objects are released after the extension is 
disabled. The patch was meant to address the cases when the extension was not 
properly disabled due to some unforeseen error, which does not deletes all WFP 
objects like the OVS sublayer or callout. Many times I have seen an improper 
activation of the OVS extension because the sublayer or callout already existed.

However I have not been able to reproduce the problem with the latest code from 
the master branch. The extension is much more stable with the latest patches 
and I cannot provide reproduction steps for this issue.

-Sorin


-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Saturday, 20 June, 2015 15:35
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [PATCH v2] datapath-windows: Return success for already existing 
WFP objects

Hi Sorin,
Could you please let us know what are the cases where these object are already 
existed? I just want to be sure that we free all these objects when stopping 
the extension.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, June 18, 2015 11:37 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Return success for already 
existing WFP objects

There are cases when the WFP callout or sublayer, being persistent objects, 
already exists when we try to register the OVS callout. In this cases, when 
trying to add again these WFP objects the return code is 
STATUS_FWP_ALREADY_EXISTS, which we are interpreting as an error.
This is incorrect and this patch changes that.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_84d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=2m9yV1dqLYwW4Lrbg7_G7J56pKu7tJ_TE125ybtwKTAs=bUUf7r2FXP4oMOFNjo2zKOSwbuJHfFjLSq2Pe8yz0sMe=
---
 datapath-windows/ovsext/TunnelFilter.c | 13 +++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c
index 08cc13f..231750e 100644
--- a/datapath-windows/ovsext/TunnelFilter.c
+++ b/datapath-windows/ovsext/TunnelFilter.c
@@ -414,7 +414,12 @@ OvsTunnelRegisterDatagramDataCallouts(const GUID *layerKey,
 
 status = FwpmCalloutAdd(gEngineHandle, mCallout, NULL, NULL);
 if (!NT_SUCCESS(status)) {
-goto Exit;
+if (STATUS_FWP_ALREADY_EXISTS != status) {
+OVS_LOG_ERROR(Failed to add WFP callout, status: %x.,
+  status);
+goto Exit;
+}
+status = STATUS_SUCCESS;
 }
 
 Exit:
@@ -459,7 +464,11 @@ OvsTunnelRegisterCallouts(VOID *deviceObject)
 
 status = FwpmSubLayerAdd(gEngineHandle, OvsTunnelSubLayer, NULL);
 if (!NT_SUCCESS(status)) {
-goto Exit;
+if (STATUS_FWP_ALREADY_EXISTS != status) {
+OVS_LOG_ERROR(Failed to add WFP sublayer, status: %x.,
+  status);
+goto Exit;
+}
 }
 
 /* In order to use this callout a socket must be opened. */
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=2m9yV1dqLYwW4Lrbg7_G7J56pKu7tJ_TE125ybtwKTAs=1h_FXA4x7kLJqZuawSWu_SwiMnBCMlk29lqJ5ldr1Tke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] datapath-windows: Return success for already existing WFP objects

[Eitan Eliahu]

Hi Sorin,
Could you please let us know what are the cases where these object are already 
existed? I just want to be sure that we free all these objects when stopping 
the extension.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, June 18, 2015 11:37 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Return success for already 
existing WFP objects

There are cases when the WFP callout or sublayer, being persistent objects, 
already exists when we try to register the OVS callout. In this cases, when 
trying to add again these WFP objects the return code is 
STATUS_FWP_ALREADY_EXISTS, which we are interpreting as an error.
This is incorrect and this patch changes that.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_84d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=2m9yV1dqLYwW4Lrbg7_G7J56pKu7tJ_TE125ybtwKTAs=bUUf7r2FXP4oMOFNjo2zKOSwbuJHfFjLSq2Pe8yz0sMe=
---
 datapath-windows/ovsext/TunnelFilter.c | 13 +++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c
index 08cc13f..231750e 100644
--- a/datapath-windows/ovsext/TunnelFilter.c
+++ b/datapath-windows/ovsext/TunnelFilter.c
@@ -414,7 +414,12 @@ OvsTunnelRegisterDatagramDataCallouts(const GUID *layerKey,
 
 status = FwpmCalloutAdd(gEngineHandle, mCallout, NULL, NULL);
 if (!NT_SUCCESS(status)) {
-goto Exit;
+if (STATUS_FWP_ALREADY_EXISTS != status) {
+OVS_LOG_ERROR(Failed to add WFP callout, status: %x.,
+  status);
+goto Exit;
+}
+status = STATUS_SUCCESS;
 }
 
 Exit:
@@ -459,7 +464,11 @@ OvsTunnelRegisterCallouts(VOID *deviceObject)
 
 status = FwpmSubLayerAdd(gEngineHandle, OvsTunnelSubLayer, NULL);
 if (!NT_SUCCESS(status)) {
-goto Exit;
+if (STATUS_FWP_ALREADY_EXISTS != status) {
+OVS_LOG_ERROR(Failed to add WFP sublayer, status: %x.,
+  status);
+goto Exit;
+}
 }
 
 /* In order to use this callout a socket must be opened. */
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=2m9yV1dqLYwW4Lrbg7_G7J56pKu7tJ_TE125ybtwKTAs=1h_FXA4x7kLJqZuawSWu_SwiMnBCMlk29lqJ5ldr1Tke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] datapath-windows: BSOD when disabling the extension

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks Sorin.
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Tuesday, June 16, 2015 2:36 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] datapath-windows: BSOD when disabling the 
extension

Hi Eitan,

Please see below the stacktrace of the BSOD. The FilterDetach routine was 
called while the requests were being processed and the gOvsSwitchContext global 
pointer was set to NULL. In this case the gOvsSwitchContext was not released, 
but only the gOvsSwitchContextRefCount reference count was decreased.

-Sorin

***
* *
*Bugcheck Analysis*
* *
***

SYSTEM_SERVICE_EXCEPTION (3b)
An exception happened while executing a system service routine.
Arguments:
Arg1: c005, Exception code that caused the bugcheck
Arg2: f800023e685b, Address of the instruction which caused the bugcheck
Arg3: d000236adad0, Address of the context record for the exception that 
caused the bugcheck
Arg4: , zero.

Debugging Details:
--


KERNEL32.DLL was not found in the image list.
Debugger will attempt to load KERNEL32.DLL at given base `.

Please provide the full image name, including the extension (i.e. kernel32.dll) 
for more reliable results.Base address and size overrides can be given as 
.reload image.ext=base,size.
Unable to add module at `

EXCEPTION_CODE: (NTSTATUS) 0xc005 - The instruction at 0x%08lx referenced 
memory at 0x%08lx. The memory could not be %s.

FAULTING_IP: 
OVSExt!OvsNewVportCmdHandler+27b 
[c:\1.data\cloudbase\work\git\ovs\datapath-windows\ovsext\vport.c @ 2136]
f800`023e685b 488b4870mov rcx,qword ptr [rax+70h]

CONTEXT:  d000236adad0 -- (.cxr 0xd000236adad0;r)
rax= rbx=e33517a0 rcx=e39688e4
rdx=d000236ae584 rsi=e2d93c90 rdi=e33517a0 
rip=f800023e685b rsp=d000236ae500 rbp=d000236aeb80
 r8=  r9=f800023f0a50 r10=d00020b02f80
r11=d00020afec30 r12= r13=0001
r14=e33518b8 r15=e2e53920
iopl=0 nv up ei pl zr na po nc
cs=0010  ss=0018  ds=002b  es=002b  fs=0053  gs=002b efl=00010246
OVSExt!OvsNewVportCmdHandler+0x27b:
f800`023e685b 488b4870mov rcx,qword ptr [rax+70h] 
ds:002b:`0070=
Last set context:
rax= rbx=e33517a0 rcx=e39688e4
rdx=d000236ae584 rsi=e2d93c90 rdi=e33517a0 
rip=f800023e685b rsp=d000236ae500 rbp=d000236aeb80
 r8=  r9=f800023f0a50 r10=d00020b02f80
r11=d00020afec30 r12= r13=0001
r14=e33518b8 r15=e2e53920
iopl=0 nv up ei pl zr na po nc
cs=0010  ss=0018  ds=002b  es=002b  fs=0053  gs=002b efl=00010246
OVSExt!OvsNewVportCmdHandler+0x27b:
f800`023e685b 488b4870mov rcx,qword ptr [rax+70h] 
ds:002b:`0070=
Resetting default scope

DEFAULT_BUCKET_ID:  WIN8_DRIVER_FAULT

BUGCHECK_STR:  0x3B

PROCESS_NAME:  ovs-vswitchd.e

CURRENT_IRQL:  0

ANALYSIS_VERSION: 6.3.9600.17237 (debuggers(dbg).140716-0327) amd64fre

LAST_CONTROL_TRANSFER:  from f800023d67f6 to f800023e685b

STACK_TEXT:  
d000`236ae500 f800`023d67f6 : d000`236ae7c0 d000`236ae728 
e000` `0010 : OVSExt!OvsNewVportCmdHandler+0x27b 
[c:\1.data\cloudbase\work\git\ovs\datapath-windows\ovsext\vport.c @ 2136]
d000`236ae630 f800`023f6aff : d000`236ae7c0 f800`023f2220 
d000`236ae728 e000`039688c0 : OVSExt!InvokeNetlinkCmdHandler+0x106 
[c:\1.data\cloudbase\work\git\ovs\datapath-windows\ovsext\datapath.c @ 1003]
d000`236ae6b0 f800`0073bc18 : e000`02d93c90 e000`033517a0 
e000`02e53920 e000`033517a0 : OVSExt!OvsDeviceControl+0x98f 
[c:\1.data\cloudbase\work\git\ovs\datapath-windows\ovsext\datapath.c @ 912]
d000`236ae840 f803`8ce4f395 : e000`033517a0 `0001 
e000`02e53920 `000e : NDIS!ndisDummyIrpHandler+0x88
d000`236ae870 f803`8ce4fd2a : e32b`7f20ffbd 000c`001f0003 
`0001 ` : nt!IopXxxControlFile+0x845
d000`236aea20 f803`8cbe08b3 : ` ` 
`0001 f803` : nt!NtDeviceIoControlFile+0x56
d000`236aea90 `77a22772 : `77a22371 0023`77a6b63c 
`0023 `00ff : nt!KiSystemServiceCopyEnd+0x13
`00f1e8b8 

Re: [ovs-dev] [PATCH v2] datapath-windows: BSOD when disabling the extension

[Eitan Eliahu]

Hi Sorin,
Can you please forward stack trace?
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Monday, June 15, 2015 7:49 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: BSOD when disabling the 
extension

When the filter detach routine is called while there are packets still in 
processing, the OvsUninitSwitchContext function call will decrement the switch 
context reference count without releasing the switch context structure. This 
behaviour is correct and expected, but the BSOD is caused in this case because 
the gOvsSwitchContext variable is set to NULL, which is wrong.

The gOvsSwitchContext global variable must be set to NULL only when the switch 
context structure is actually released.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_80d=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=fxSjeLfMN40XNZsRUW1pOUC4BaXKLNmvM0IzVTxe32os=PNqZC6M2Nbl0WZJOWoEUmJTabsU0eJulnLzdB9mcYkke=
Acked-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Switch.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index f877854..99a306d 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -201,6 +201,7 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 status = OvsInitSwitchContext(switchContext);
 if (status != NDIS_STATUS_SUCCESS) {
 OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
+switchContext = NULL;
 goto create_switch_done;
 }
 
@@ -240,7 +241,6 @@ OvsExtDetach(NDIS_HANDLE filterModuleContext)
 }
 OvsDeleteSwitch(switchContext);
 OvsCleanupIpHelper();
-gOvsSwitchContext = NULL;
 /* This completes the cleanup, and a new attach can be handled now. */
 
 OVS_LOG_TRACE(Exit: OvsDetach Successfully); @@ -495,6 +495,7 @@ 
OvsReleaseSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 
 if (ref == 1) {
 OvsDeleteSwitchContext(switchContext);
+gOvsSwitchContext = NULL;
 }
 }
 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=fxSjeLfMN40XNZsRUW1pOUC4BaXKLNmvM0IzVTxe32os=OgmOQuSMC-PwPV_FBD6LjMrxl7Ze1VPrMIRnwXiXNVIe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH ovn] ofctrl: Don't use designated initializers.

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Monday, June 15, 2015 12:52 PM
To: dev@openvswitch.org
Cc: Gurucharan Shetty
Subject: [ovs-dev] [PATCH ovn] ofctrl: Don't use designated initializers.

MSVC 2013 does not like designated initializers when structs are initialized 
inside structs.
Apparently it has been fixed in MSCV 2015.

Signed-off-by: Gurucharan Shetty gshe...@nicira.com
---
 ovn/controller/ofctrl.c |   41 +++--
 1 file changed, 19 insertions(+), 22 deletions(-)

diff --git a/ovn/controller/ofctrl.c b/ovn/controller/ofctrl.c index 
bc4e8d4..ec0d868 100644
--- a/ovn/controller/ofctrl.c
+++ b/ovn/controller/ofctrl.c
@@ -401,12 +401,11 @@ ofctrl_update_flows(void)
 if (!d) {
 /* Installed flow is no longer desirable.  Delete it from the
  * switch and from installed_flows. */
-struct ofputil_flow_mod fm = {
-.match = i-match,
-.priority = i-priority,
-.table_id = i-table_id,
-.command = OFPFC_DELETE_STRICT,
-};
+struct ofputil_flow_mod fm;
+fm.match = i-match;
+fm.priority = i-priority;
+fm.table_id = i-table_id;
+fm.command = OFPFC_DELETE_STRICT;
 queue_flow_mod(fm);
 ovn_flow_log(i, removing);
 
@@ -416,14 +415,13 @@ ofctrl_update_flows(void)
 if (!ofpacts_equal(i-ofpacts, i-ofpacts_len,
d-ofpacts, d-ofpacts_len)) {
 /* Update actions in installed flow. */
-struct ofputil_flow_mod fm = {
-.match = i-match,
-.priority = i-priority,
-.table_id = i-table_id,
-.ofpacts = d-ofpacts,
-.ofpacts_len = d-ofpacts_len,
-.command = OFPFC_MODIFY_STRICT,
-};
+struct ofputil_flow_mod fm;
+fm.match = i-match;
+fm.priority = i-priority;
+fm.table_id = i-table_id;
+fm.ofpacts = d-ofpacts;
+fm.ofpacts_len = d-ofpacts_len;
+fm.command = OFPFC_MODIFY_STRICT;
 queue_flow_mod(fm);
 ovn_flow_log(i, updating);
 
@@ -446,14 +444,13 @@ ofctrl_update_flows(void)
 struct ovn_flow *d;
 HMAP_FOR_EACH_SAFE (d, next, hmap_node, desired_flows) {
 /* Send flow_mod to add flow. */
-struct ofputil_flow_mod fm = {
-.match = d-match,
-.priority = d-priority,
-.table_id = d-table_id,
-.ofpacts = d-ofpacts,
-.ofpacts_len = d-ofpacts_len,
-.command = OFPFC_ADD,
-};
+struct ofputil_flow_mod fm;
+fm.match = d-match;
+fm.priority = d-priority;
+fm.table_id = d-table_id;
+fm.ofpacts = d-ofpacts;
+fm.ofpacts_len = d-ofpacts_len;
+fm.command = OFPFC_ADD;
 queue_flow_mod(fm);
 ovn_flow_log(d, adding);
 
--
1.7.9.5

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=EOQPdoae3PS5ck_aP1mpldzMKugHnzj-denjPqV1-gws=roWEsifMlJdcrB8QZHp1XshtiLitj46fru_pRIeVrrke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Correctly complete the original NBL with multiple NBs

[Eitan Eliahu]

No need to apologize Sorin. We will get everything addressed. Sorry if you felt 
like we pushed you.
Thanks for the patch,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, June 12, 2015 5:27 AM
To: Nithin Raju
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: Correctly complete the 
original NBL with multiple NBs

Nithin,

The previous patch you are referring was one of a series I had pushed when 
working at the Multiple NBLs issue. At that time I had some trouble when 
generating packets large enough for the a NBL with multiple NBs to be 
generated. Because Eitan and Ankur, that also worked on this issue, pushed me 
to commit my changes so they can see how things were handled, I committed the 
patch series without thoroughly testing it.

The current patch solves a bug I have found when managing to generate large 
enough packets. I hope this satisfies your curiosity.

-Sorin

-Original Message-
From: Nithin Raju [mailto:nit...@vmware.com]
Sent: Thursday, 11 June, 2015 18:56
To: Sorin Vinturis
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: Correctly complete the 
original NBL with multiple NBs

 On Jun 11, 2015, at 5:52 AM, Sorin Vinturis 
 svintu...@cloudbasesolutions.com wrote:
 
 OvsCreateNewNBLsFromMultipleNBs function failed to correctly complete 
 the original NBL with multiple NBs after creating multiple NBLs with 
 single NB.
 
 Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Sorin,
Thanks for the fix. I am curious how this was not caught during testing for the 
previous patch.

Acked-by: Nithin Raju nit...@vmware.com

thanks,
-- Nithin
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=g0bgvrDEAVQy7ywNIgRZcfydL28BvzrFcBWVmwhkslos=uX0jWkHs1N9iNx-Oz9CvmAGFqpxNeBqYzDR7ZG5RLEQe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

[ovs-dev] [PATCH v3] datapath-windows: Stateless TCP Tunnelling protocol - Initial implementation

[Eitan Eliahu]

This change include an initial implementable of STT.

The following should be added:
[1] Checksum offload (SW and HW)
[2] LSO (SW and HW)
[3] IP layer WFP callout for IP segments

Added support for multiple (per TCP port) STT ports

Testing: link layer connection through ping works. File transfer.

Signed-off-by: Eitan Eliahu elia...@vmware.com
Co-authored-by: Saurabh Shah ssaur...@vmware.com
Signed-off-by: Saurabh Shah ssaur...@vmware.com
---
 INSTALL.Windows.md |  40 ++--
 datapath-windows/automake.mk   |   2 +
 datapath-windows/ovsext/Actions.c  |  64 --
 datapath-windows/ovsext/Debug.h|   1 +
 datapath-windows/ovsext/Stt.c  | 369 +
 datapath-windows/ovsext/Stt.h  |  89 
 datapath-windows/ovsext/Switch.h   |   2 -
 datapath-windows/ovsext/Tunnel.c   |   3 +-
 datapath-windows/ovsext/Util.h |   1 +
 datapath-windows/ovsext/Vport.c|  52 -
 datapath-windows/ovsext/Vport.h|  47 -
 datapath-windows/ovsext/Vxlan.c|  16 +-
 datapath-windows/ovsext/Vxlan.h|   9 +-
 datapath-windows/ovsext/ovsext.vcxproj |   4 +-
 14 files changed, 634 insertions(+), 65 deletions(-)
 create mode 100644 datapath-windows/ovsext/Stt.c
 create mode 100644 datapath-windows/ovsext/Stt.h

diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md
index 0ec0af0..6ec9f29 100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -386,29 +386,31 @@ Hyper-Vs.  The following examples demonstrate how it can 
be done:
 % ovs-vsctl add-port br-int ovs-port-a tag=900
 % ovs-vsctl add-port br-int ovs-port-b tag=900
 
-Steps to add VXLAN tunnels
+Steps to add tunnels
 --
-The Windows Open vSwitch implementation support VXLAN tunnels.  To add VXLAN
+The Windows Open vSwitch implementation support VXLAN and STT tunnels. To add
 tunnels, the following steps serve as examples.
 
 Note that, any patch ports created between br-int and br-pif MUST be beleted
-prior to adding VXLAN tunnels.
-
-01 Add the vxlan port between 172.168.201.101 - 172.168.201.102
-% ovs-vsctl add-port br-int vxlan-1
-% ovs-vsctl set Interface vxlan-1 type=vxlan
-% ovs-vsctl set Interface vxlan-1 options:local_ip=172.168.201.101
-% ovs-vsctl set Interface vxlan-1 options:remote_ip=172.168.201.102
-% ovs-vsctl set Interface vxlan-1 options:in_key=flow
-% ovs-vsctl set Interface vxlan-1 options:out_key=flow
-
-02 Add the vxlan port between 172.168.201.101 - 172.168.201.105
-% ovs-vsctl add-port br-int vxlan-2
-% ovs-vsctl set Interface vxlan-2 type=vxlan
-% ovs-vsctl set Interface vxlan-2 options:local_ip=172.168.201.102
-% ovs-vsctl set Interface vxlan-2 options:remote_ip=172.168.201.105
-% ovs-vsctl set Interface vxlan-2 options:in_key=flow
-% ovs-vsctl set Interface vxlan-2 options:out_key=flow
+prior to adding tunnels.
+
+01 Add the tunnel port between 172.168.201.101 - 172.168.201.102
+% ovs-vsctl add-port br-int tun-1
+% ovs-vsctl set Interface tun-1 type=port-type
+% ovs-vsctl set Interface tun-1 options:local_ip=172.168.201.101
+% ovs-vsctl set Interface tun-1 options:remote_ip=172.168.201.102
+% ovs-vsctl set Interface tun-1 options:in_key=flow
+% ovs-vsctl set Interface tun-1 options:out_key=flow
+
+02 Add the tunnel port between 172.168.201.101 - 172.168.201.105
+% ovs-vsctl add-port br-int tun-2
+% ovs-vsctl set Interface tun-2 type=port-type
+% ovs-vsctl set Interface tun-2 options:local_ip=172.168.201.102
+% ovs-vsctl set Interface tun-2 options:remote_ip=172.168.201.105
+% ovs-vsctl set Interface tun-2 options:in_key=flow
+% ovs-vsctl set Interface tun-2 options:out_key=flow
+
+   Where port-type is the string stt or vxlan
 
 
 Requirements
diff --git a/datapath-windows/automake.mk b/datapath-windows/automake.mk
index 9324b3c..a4f5a57 100644
--- a/datapath-windows/automake.mk
+++ b/datapath-windows/automake.mk
@@ -56,6 +56,8 @@ EXTRA_DIST += \
datapath-windows/ovsext/Vport.c \
datapath-windows/ovsext/Vport.h \
datapath-windows/ovsext/Vxlan.c \
+   datapath-windows/ovsext/Stt.h \
+   datapath-windows/ovsext/Stt.c \
datapath-windows/ovsext/Vxlan.h \
datapath-windows/ovsext/ovsext.inf \
datapath-windows/ovsext/ovsext.rc \
diff --git a/datapath-windows/ovsext/Actions.c 
b/datapath-windows/ovsext/Actions.c
index 79e464c..bfcf02e 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -23,6 +23,7 @@
 #include NetProto.h
 #include Flow.h
 #include Vxlan.h
+#include Stt.h
 #include Checksum.h
 #include PacketIO.h
 
@@ -35,6 +36,8 @@
 typedef struct _OVS_ACTION_STATS {
 UINT64 rxVxlan;
 UINT64 txVxlan;
+UINT64 rxStt;
+UINT64 txStt;
 UINT64 flowMiss;
 UINT64 flowUserspace;
 UINT64 txTcp;
@@ -184,9 +187,6 @@ OvsInitForwardingCtx(OvsForwardingContext *ovsFwdCtx,
 }
 
 /*
- * XXX

Re: [ovs-dev] [PATCH] datapath-windows: Correctly complete the original NBL with multiple NBs

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, June 11, 2015 5:53 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Correctly complete the original 
NBL with multiple NBs

OvsCreateNewNBLsFromMultipleNBs function failed to correctly complete the 
original NBL with multiple NBs after creating multiple NBLs with single NB.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
---
 datapath-windows/ovsext/PacketIO.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/datapath-windows/ovsext/PacketIO.c 
b/datapath-windows/ovsext/PacketIO.c
index 7b2c0c8..d41335b 100644
--- a/datapath-windows/ovsext/PacketIO.c
+++ b/datapath-windows/ovsext/PacketIO.c
@@ -524,11 +524,12 @@ OvsCreateNewNBLsFromMultipleNBs(POVS_SWITCH_CONTEXT 
switchContext,
 }
 lastNbl-Next = *nextNbl;
 *nextNbl = newNbls-Next;
-*curNbl = newNbls;
-(*curNbl)-Next = NULL;
 
 OvsCompleteNBL(switchContext, *curNbl, TRUE);
 
+*curNbl = newNbls;
+(*curNbl)-Next = NULL;
+
 error = FALSE;
 } while (error);
 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=BQIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SJYuCd-gcL6KClc0-g9-5X1vUkr8h1it_Ebv1yzOeAEs=QvI9jn40yGR5GqNe3Wqmw8gEFFIFtmu9Jl1Dti3gjewe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] datapath-windows: Stateless TCP Tunnelling protocol - Initial implementation

[Eitan Eliahu]

Yes, will do.
Eitan

-Original Message-
From: Ben Pfaff [mailto:b...@nicira.com] 
Sent: Saturday, June 06, 2015 1:52 PM
To: Eitan Eliahu
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v2] datapath-windows: Stateless TCP Tunnelling 
protocol - Initial implementation

On Wed, May 27, 2015 at 09:11:42AM -0700, Eitan Eliahu wrote:
 This change include an initial implementable of STT.
 The following should be added:
 [1] Checksum offload (SW and HW)
 [2] LSO (SW and HW)
 [3] IP layer WFP callout for IP segments
 
 Testing: link layer connection through ping works. File transfer.
 
 Signed-off-by: Eitan Eliahu elia...@vmware.com
 Co-authored-by: Saurabh Shah ssaur...@vmware.com
 Signed-off-by: Saurabh Shah ssaur...@vmware.com

It's been a few days and I guess we have all the reviews we're going to 
initially get.  I get a few patch rejects trying to apply this, and at least 
some of them seemed to require a little thinking to resolve, so will you rebase 
and post v3?

Thanks,

Ben.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2 3/4] netlink: support for asynchronous NL transactions

[Eitan Eliahu]

Nithin,
The driver is never blocked and it always executes in asynchronous fashion. A 
driver can immediately complete an I/O request by returning a status success 
code or an error. If the I/O is not completed the driver returns a pending 
status and the I/O will be completed later from a different thread context 
(this enables asynchronous/overlapped I/O for user mode process).
Regardless of the driver specific I/O implementation, User mode thread can opt 
to wait on a I/O device request or not. If an overlapped structure  is sent in 
the DeviceIoControl, the DeviceIOControl returns immediately regardless if the 
I/O has been completed or not.
If a NULL sent rather than an overlapped structure, the thread which calls 
DeviceIOControl is blocked until the I/O is completed by the driver.

With respect to WFP related I/O I don't think we need to return pending status 
from the driver. It should be fine to block the user mode thread during the WFP 
transaction processing.
Thanks,
Eitan


-Original Message-
From: Nithin Raju 
Sent: Wednesday, May 20, 2015 11:04 AM
To: Eitan Eliahu; Sorin Vinturis; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v2 3/4] netlink: support for asynchronous NL 
transactions

Sorin/Eitan,
Do we need this patch or not?

Eitan tells me that if return status is STATUS_PENDING, then the context is 
blocked in the kernel. If that is true, we probably don’t need this patch since 
control won’t return to userspace at all, till the IRP is complete. This is 
especially true since we are passing NULL for the ‘lpOverlapped’ parameter to 
DeviceIoControl().

Pls. see the following documentation on MSDN:
==
https://msdn.microsoft.com/en-us/library/windows/desktop/aa363216%28v=vs.85%29.aspx

lpOverlapped [in, out, optional]:
For overlapped operations, DeviceIoControl returns immediately, and the event 
object is signaled when the operation has been completed. Otherwise, the 
function does not return until the operation has been completed or an error 
occurs.
==

Based on this, my belief is that the call to DeviceIoControl() would block 
indefinitely until the IRP is completed.

In your testing, did you think this patch was necessary?

thanks,
-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] ARP lookup and next hop functionality on windows (updated version)

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, May 19, 2015 10:02 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] ARP lookup and next hop functionality on windows

This patch implements two functionalities needed for an active manager:
1. ARP lookup
2. Next hop

The first functionality relies on the internal Windows API:
https://urldefense.proofpoint.com/v2/url?u=https-3A__msdn.microsoft.com_en-2Dus_library_windows_desktop_aa365956-2528v-3Dvs.85-2529.aspxd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=tQfMAkM2SjeoDNgJA5lzr3VUDfYRDmjOR3QrymbowLcs=AjthuTJGRGkQqO4iaxxvp5uVrWnSngSiynyZFrmI4fMe=

The second one:
https://urldefense.proofpoint.com/v2/url?u=https-3A__msdn.microsoft.com_en-2Dus_library_windows_desktop_aa365915-2528v-3Dvs.85-2529.aspxd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=tQfMAkM2SjeoDNgJA5lzr3VUDfYRDmjOR3QrymbowLcs=cYblpE7EGRgk2wDKdIc9MFjjG9U-91iBWzp2rstlaH4e=

Both API's are found in the Iphlpapi library. We need to add this library when 
compiling.

Documentation and appveyor config has been updated to match the use of the new 
library.

Tested using opendaylight.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_63d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=OzBp2rjLgq1hrRQaHib0MxIu5M4HlH1EvhmW2aCEJ0Qs=fZsrIJis7ZxmwWfZLrnf1UhsdtE02hnUYMyLpIHcNEMe=
---
v2: call GetIpNetTable and GetAdaptersAddresses with a zero length buffer.
add pretty printing
---
---
 INSTALL.Windows.md   |  25 +--
 lib/netdev-windows.c | 114 +++
 2 files changed, 127 insertions(+), 12 deletions(-)

diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md index 78af0a1..0ec0af0 
100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -62,9 +62,10 @@ or from a distribution tar ball.
   the right compiler, linker, libraries, Open vSwitch component installation
   directories, etc. For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
-  --prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
-  --sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread
+% ./configure CC=./build-aux/cccl LD=`which link` \
+  LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+  --localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+   --with-pthread=C:/pthread
 
 By default, the above enables compiler optimization for fast code.
 For default compiler optimization, pass the --with-debug configure @@ 
-114,10 +115,10 @@ Note down the directory where OpenSSL is installed (e.g.: 
C:/OpenSSL-Win32).
 * While configuring the package, specify the OpenSSL directory path.
 For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
---prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
---sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread \
---enable-ssl --with-openssl=C:/OpenSSL-Win32
+% ./configure CC=./build-aux/cccl LD=`which link`  \
+LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+--localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+--with-pthread=C:/pthread --enable-ssl --with-openssl=C:/OpenSSL-Win32
 
 * Run make for the ported executables.
 
@@ -131,11 +132,11 @@ level 'make' will invoke building the kernel datapath, if 
the  '--with-vstudioddk' argument is specified while configuring the package.
 For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
---prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
---sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread \
---enable-ssl --with-openssl=C:/OpenSSL-Win32 \
---with-vstudioddk=WDK to use
+% ./configure CC=./build-aux/cccl LD=`which link` \
+LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+--localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+--with-pthread=C:/pthread --enable-ssl \
+--with-openssl=C:/OpenSSL-Win32 --with-vstudioddk=WDK to use
 
 Possible values for WDK to use are:
 Win8.1 Debug, Win8.1 Release, Win8 Debug and Win8 Release.
diff --git a/lib/netdev-windows.c b/lib/netdev-windows.c index 1fc1da7..1eb8727 
100644
--- a/lib/netdev-windows.c
+++ b/lib/netdev-windows.c
@@ -17,6 +17,7 @@
 #include stdlib.h
 #include config.h
 #include errno.h
+#include iphlpapi.h
 
 #include net/if.h
 
@@ -373,6 +374,117 @@ netdev_windows_update_flags(struct netdev *netdev_,
 return 0;
 }
 
+/* Looks up in the ARP table entry for a given 'ip'. If it is found

Re: [ovs-dev] [PATCH] ARP lookup and next hop functionality on windows

[Eitan Eliahu]

Hi Alin,
It looks good besides some minor comments.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, May 04, 2015 8:35 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] ARP lookup and next hop functionality on windows

This patch implements two functionalities needed for an active manager:
1. ARP lookup
2. Next hop

The first functionality relies on the internal Windows API:
https://urldefense.proofpoint.com/v2/url?u=https-3A__msdn.microsoft.com_en-2Dus_library_windows_desktop_aa365956-2528v-3Dvs.85-2529.aspxd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=tQfMAkM2SjeoDNgJA5lzr3VUDfYRDmjOR3QrymbowLcs=AjthuTJGRGkQqO4iaxxvp5uVrWnSngSiynyZFrmI4fMe=
 

The second one:
https://urldefense.proofpoint.com/v2/url?u=https-3A__msdn.microsoft.com_en-2Dus_library_windows_desktop_aa365915-2528v-3Dvs.85-2529.aspxd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=tQfMAkM2SjeoDNgJA5lzr3VUDfYRDmjOR3QrymbowLcs=cYblpE7EGRgk2wDKdIc9MFjjG9U-91iBWzp2rstlaH4e=
 

Both API's are found in the Iphlpapi library. We need to add this library when 
compiling.

Documentation and appveyor config has been updated to match the use of the new 
library.

Tested using opendaylight.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_63d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=tQfMAkM2SjeoDNgJA5lzr3VUDfYRDmjOR3QrymbowLcs=L3JZ3eRbqh48NNmq_mcC2elPy7pBrI0pLWhhDEnepuke=
---
 INSTALL.Windows.md   |  25 ++-
 appveyor.yml |   2 +-
 lib/netdev-windows.c | 125 +++
 3 files changed, 139 insertions(+), 13 deletions(-)

diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md index 78af0a1..0ec0af0 
100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -62,9 +62,10 @@ or from a distribution tar ball.
   the right compiler, linker, libraries, Open vSwitch component installation
   directories, etc. For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
-  --prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
-  --sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread
+% ./configure CC=./build-aux/cccl LD=`which link` \
+  LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+  --localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+   --with-pthread=C:/pthread
 
 By default, the above enables compiler optimization for fast code.
 For default compiler optimization, pass the --with-debug configure @@ 
-114,10 +115,10 @@ Note down the directory where OpenSSL is installed (e.g.: 
C:/OpenSSL-Win32).
 * While configuring the package, specify the OpenSSL directory path.
 For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
---prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
---sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread \
---enable-ssl --with-openssl=C:/OpenSSL-Win32
+% ./configure CC=./build-aux/cccl LD=`which link`  \
+LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+--localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+--with-pthread=C:/pthread --enable-ssl --with-openssl=C:/OpenSSL-Win32
 
 * Run make for the ported executables.
 
@@ -131,11 +132,11 @@ level 'make' will invoke building the kernel datapath, if 
the  '--with-vstudioddk' argument is specified while configuring the package.
 For example,
 
-% ./configure CC=./build-aux/cccl LD=`which link` LIBS=-lws2_32 \
---prefix=C:/openvswitch/usr --localstatedir=C:/openvswitch/var \
---sysconfdir=C:/openvswitch/etc --with-pthread=C:/pthread \
---enable-ssl --with-openssl=C:/OpenSSL-Win32 \
---with-vstudioddk=WDK to use
+% ./configure CC=./build-aux/cccl LD=`which link` \
+LIBS=-lws2_32 -liphlpapi --prefix=C:/openvswitch/usr \
+--localstatedir=C:/openvswitch/var --sysconfdir=C:/openvswitch/etc \
+--with-pthread=C:/pthread --enable-ssl \
+--with-openssl=C:/OpenSSL-Win32 --with-vstudioddk=WDK to use
 
 Possible values for WDK to use are:
 Win8.1 Debug, Win8.1 Release, Win8 Debug and Win8 Release.
diff --git a/appveyor.yml b/appveyor.yml index a14f0fc..863b561 100644
--- a/appveyor.yml
+++ b/appveyor.yml
@@ -39,5 +39,5 @@ build_script:
 - C:\MinGW\msys\1.0\bin\bash -lc cp 
/c/pthreads-win32/Pre-built.2/dll/x86/*.dll /c/openvswitch/.
 - C:\MinGW\msys\1.0\bin\bash -lc mv /bin/link.exe /bin/link_copy.exe
 - C:\MinGW\msys\1.0\bin\bash -lc cd /c/openvswitch  ./boot.sh
-- C:\MinGW\msys\1.0\bin\bash -lc cd /c/openvswitch  ./configure 
CC=build-aux/cccl LD=\`which link`\ LIBS=-lws2_32 

Re: [ovs-dev] [PATCH] datapath-windows: Fix warning from the powershell module

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks Alin.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Monday, May 04, 2015 9:45 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Fix warning from the powershell 
module

This patch fixes the warning when datapath-windows/misc/OVS.psm1 is imported.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-by: Hemanth Kumar Mantri man...@nutanix.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_69d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=28OaspoKKvRf7sMd_OuP7WGNu0Y8Gm_uyKtsixeU1gAs=bOxLzQumIA-Cglwy6ZOdkBhODOusq4Q40MFG982V3Loe=
---
 datapath-windows/misc/OVS.psm1 | 8 +---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/datapath-windows/misc/OVS.psm1 b/datapath-windows/misc/OVS.psm1 
index d6b6b00..a8ffcae 100644
--- a/datapath-windows/misc/OVS.psm1
+++ b/datapath-windows/misc/OVS.psm1
@@ -1,5 +1,5 @@
 #
-Copyright 2014 Cloudbase Solutions Srl
+Copyright 2014, 2015 Cloudbase Solutions Srl
 
 Licensed under the Apache License, Version 2.0 (the License);  you may not 
use this file except in compliance with the License.
@@ -57,7 +57,7 @@ function Set-VMNetworkAdapterOVSPort
 $retVal = $vsms.ModifyResourceSettings(@($sd.GetText(1)))
 try
 {
-Check-WMIReturnValue $retVal
+CheckWMIReturnValue $retVal
 }
 catch
 {
@@ -142,7 +142,7 @@ function Get-VMNetworkAdapterWithOVSPort
 }
 }
 
-function Check-WMIReturnValue($retVal)
+function CheckWMIReturnValue($retVal)
 {
 if ($retVal.ReturnValue -ne 0)
 {
@@ -206,3 +206,5 @@ function Set-VMNetworkAdapterOVSPortDirect
 $vnic[0] | Set-VMNetworkAdapterOVSPort -OVSPortName $OVSPortName
 }
 }
+
+Export-ModuleMember -function Set-*, Get-*
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=28OaspoKKvRf7sMd_OuP7WGNu0Y8Gm_uyKtsixeU1gAs=TUgRV-w8Mpu1iAnHHa09XxMdrFF24pcdea9Mde0Gnw8e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v4 2/4] Sample to automate build installer

[Eitan Eliahu]

Hi Alin,
I'm wondering if you addressed the issue of installation on Windows Server 2012.
Unless we install the 8.0 driver, the driver load would fail.
Thanks,
Eitan

 Begin forwarded message:
 
 From: Alin Serdean aserd...@cloudbasesolutions.com
 To: dev@openvswitch.org dev@openvswitch.org
 Date: April 28, 2015 at 3:36:28 PM PDT
 Subject: [ovs-dev] [PATCH v4 2/4] Sample to automate build installer
 
 This commit is a POC on how to build the installer from the command line.
 
 One could issue the following commands to build the binaries, driver 
 and installer.
 
 ./boot.sh;./configure CC=./build-aux/cccl LD=`which link` 
 LIBS=-lws2_32 \  --prefix=C:/openvswitch/usr 
 --localstatedir=C:/openvswitch/var \  
 --sysconfdir=C:/openvswitch/etc \  --with-pthread=C:/pthread/ 
 --with-vstudioddk=Win8.1Release; \  make clean  make  make 
 installer
 
 
 Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
 Co-authored-by: Nithin Raju nit...@vmware.com
 ---
 v3: Fix pthread libraries
 v2: Respin
 Makefile.am | 20 ++
 datapath-windows/ovsext.sln | 50 ++---
 m4/openvswitch.m4   |  8 
 3 files changed, 49 insertions(+), 29 deletions(-)
 
 diff --git a/Makefile.am b/Makefile.am index 8bc431b..9a2ffd6 100644
 --- a/Makefile.am
 +++ b/Makefile.am
 @@ -354,6 +354,25 @@ if LINUX_ENABLED
   cd datapath/linux  $(MAKE) modules_install endif
 
 +if WIN32
 +PTHREAD_TEMP_DIR=`echo $(PTHREAD_LDFLAGS) | sed 
 +'s|^.\(.*\).$:\1||'`
 +installer: $(BUILT_SOURCES) $(ovsext_make)
 + cp -f $(top_srcdir)/datapath-windows/misc/OVS.psm1 
 windows/ovs-windows-installer/Services/OVS.psm1
 + cp -f $(top_srcdir)/vswitchd/vswitch.ovsschema 
 windows/ovs-windows-installer/Services/vswitch.ovsschema
 + cp -f $(top_srcdir)/vswitchd/ovs-vswitchd.exe 
 windows/ovs-windows-installer/Services/ovs-vswitchd.exe
 + cp -f $(top_srcdir)/ovsdb/ovsdb-server.exe 
 windows/ovs-windows-installer/Services/ovsdb-server.exe
 + cp -f 
 $(top_srcdir)/datapath-windows/x64/$(VSTUDIO_CONFIG)/package/ovsext.cat 
 windows/ovs-windows-installer/Driver/ovsext.cat
 + cp -f 
 $(top_srcdir)/datapath-windows/x64/$(VSTUDIO_CONFIG)/package/ovsext.inf 
 windows/ovs-windows-installer/Driver/ovsext.inf
 + cp -f 
 $(top_srcdir)/datapath-windows/x64/$(VSTUDIO_CONFIG)/package/OVSExt.sys 
 windows/ovs-windows-installer/Driver/OVSExt.sys
 + cp -f $(top_srcdir)/utilities/*.exe 
 windows/ovs-windows-installer/Binaries/
 + cp -f $(top_srcdir)/utilities/*.pdb 
 windows/ovs-windows-installer/Symbols/
 + cp -f $(top_srcdir)/ovsdb/ovsdb-client.exe 
 windows/ovs-windows-installer/Binaries/ovsdb-client.exe
 + cp -f $(top_srcdir)/ovsdb/ovsdb-tool.exe 
 windows/ovs-windows-installer/Binaries/ovsdb-tool.exe
 + cp -f $(top_srcdir)/ovsdb/*.pdb windows/ovs-windows-installer/Symbols/
 + cp -f $(PTHREAD_TEMP_DIR)/../../dll/x86/*.dll 
 windows/ovs-windows-installer/Binaries/
 + cp -f /c/Program Files (x86)/Common Files/Merge 
 Modules/Microsoft_VC120_CRT_x86.msm 
 windows/ovs-windows-installer/Redist/Microsoft_VC120_CRT_x86.msm
 + MSBuild.exe windows/ovs-windows-installer.sln /target:Build 
 /property:Configuration=Release
 +endif
 dist-docs:
   VERSION=$(VERSION) $(srcdir)/build-aux/dist-docs $(srcdir) $(docs)
 .PHONY: dist-docs
 @@ -377,3 +396,4 @@ include tutorial/automake.mk include 
 vtep/automake.mk include datapath-windows/automake.mk include 
 datapath-windows/include/automake.mk
 +include windows/automake.mk
 diff --git a/datapath-windows/ovsext.sln b/datapath-windows/ovsext.sln 
 index 9cb767d..60e9318 100644
 --- a/datapath-windows/ovsext.sln
 +++ b/datapath-windows/ovsext.sln
 @@ -1,6 +1,6 @@
 Microsoft Visual Studio Solution File, Format Version 12.00 # Visual 
 Studio 2013 -VisualStudioVersion = 12.0.21005.1
 +VisualStudioVersion = 12.0.31101.0
 MinimumVisualStudioVersion = 10.0.40219.1
 Project({2150E333-8FDC-42A3-9474-1A3956D46DE8}) = Package, Package, 
 {6BA8554E-AE50-49B0-9C98-4592447FEF8D}
 EndProject
 @@ -12,32 +12,32 @@ Project({8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}) 
 = ovsext, ovsext\ovsext.vcx EndProject Global
   GlobalSection(SolutionConfigurationPlatforms) = preSolution
 - Win8 Debug|x64 = Win8 Debug|x64
 - Win8 Release|x64 = Win8 Release|x64
 - Win8.1 Debug|x64 = Win8.1 Debug|x64
 - Win8.1 Release|x64 = Win8.1 Release|x64
 + Win8.1Debug|x64 = Win8.1Debug|x64
 + Win8.1Release|x64 = Win8.1Release|x64
 + Win8Debug|x64 = Win8Debug|x64
 + Win8Release|x64 = Win8Release|x64
   EndGlobalSection
   GlobalSection(ProjectConfigurationPlatforms) = postSolution
 - {911D7389-3E61-449F-B8F3-14AD7EE9A0F2}.Win8 Debug|x64.ActiveCfg 
 = Win8 Debug|x64
 - {911D7389-3E61-449F-B8F3-14AD7EE9A0F2}.Win8 Debug|x64.Build.0 = 
 Win8 Debug|x64
 - {911D7389-3E61-449F-B8F3-14AD7EE9A0F2}.Win8 
 

[ovs-dev] [PATCH] datapath-windows: Stateless TCP Tunnelling protocol - Initial implementation

[Eitan Eliahu]

This change include an initial implementable of STT.
The following should be added:
[1] Checksum offload (SW and HW)
[2] LSO (SW and HW)
[3] IP layer WFP callout for IP segments

Testing: link layer connection through ping works. File transfer.

Signed-off-by: Eitan Eliahu elia...@vmware.com
Co-authored-by: Saurabh Shah ssaur...@vmware.com
---
 datapath-windows/automake.mk   |   2 +
 datapath-windows/ovsext/Actions.c  |  13 ++
 datapath-windows/ovsext/Debug.h|   1 +
 datapath-windows/ovsext/Stt.c  | 363 +
 datapath-windows/ovsext/Stt.h  |  87 
 datapath-windows/ovsext/Switch.h   |   3 +-
 datapath-windows/ovsext/Tunnel.c   |   3 +-
 datapath-windows/ovsext/Util.h |   1 +
 datapath-windows/ovsext/Vport.c|  18 ++
 datapath-windows/ovsext/Vport.h|  22 ++
 datapath-windows/ovsext/Vxlan.c|   4 +-
 datapath-windows/ovsext/ovsext.vcxproj |   4 +-
 12 files changed, 515 insertions(+), 6 deletions(-)
 create mode 100644 datapath-windows/ovsext/Stt.c
 create mode 100644 datapath-windows/ovsext/Stt.h

diff --git a/datapath-windows/automake.mk b/datapath-windows/automake.mk
index 9324b3c..a4f5a57 100644
--- a/datapath-windows/automake.mk
+++ b/datapath-windows/automake.mk
@@ -56,6 +56,8 @@ EXTRA_DIST += \
datapath-windows/ovsext/Vport.c \
datapath-windows/ovsext/Vport.h \
datapath-windows/ovsext/Vxlan.c \
+   datapath-windows/ovsext/Stt.h \
+   datapath-windows/ovsext/Stt.c \
datapath-windows/ovsext/Vxlan.h \
datapath-windows/ovsext/ovsext.inf \
datapath-windows/ovsext/ovsext.rc \
diff --git a/datapath-windows/ovsext/Actions.c 
b/datapath-windows/ovsext/Actions.c
index a93fe03..39f36b4 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -23,6 +23,7 @@
 #include NetProto.h
 #include Flow.h
 #include Vxlan.h
+#include Stt.h
 #include Checksum.h
 #include PacketIO.h
 
@@ -207,6 +208,10 @@ OvsDetectTunnelRxPkt(OvsForwardingContext *ovsFwdCtx,
 flowKey-ipKey.l4.tpDst == VXLAN_UDP_PORT_NBO) {
 tunnelVport = ovsFwdCtx-switchContext-vxlanVport;
 ovsActionStats.rxVxlan++;
+} else if (!flowKey-ipKey.nwFrag 
+flowKey-ipKey.nwProto == IPPROTO_TCP 
+flowKey-ipKey.l4.tpDst == STT_DST_PORT_NBO) {
+tunnelVport =ovsFwdCtx-switchContext-sttVport;
 }
 
 // We might get tunnel packets even before the tunnel gets initialized.
@@ -632,6 +637,10 @@ OvsTunnelPortTx(OvsForwardingContext *ovsFwdCtx)
ovsFwdCtx-switchContext,
(VOID *)ovsFwdCtx-completionList,
ovsFwdCtx-layers, newNbl);
+break;
+case OVS_VPORT_TYPE_STT:
+status = OvsEncapStt(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
+ ovsFwdCtx-tunKey, ovsFwdCtx-layers, newNbl);
 break;
 default:
 ASSERT(! Tx: Unhandled tunnel type);
@@ -697,6 +706,10 @@ OvsTunnelPortRx(OvsForwardingContext *ovsFwdCtx)
 status = OvsDoDecapVxlan(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
 ovsFwdCtx-tunKey, newNbl);
 break;
+case OVS_VPORT_TYPE_STT:
+status = OvsDecapStt(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
+ ovsFwdCtx-tunKey, newNbl);
+break;
 default:
 OVS_LOG_ERROR(Rx: Unhandled tunnel type: %d\n,
   tunnelRxVport-ovsType);
diff --git a/datapath-windows/ovsext/Debug.h b/datapath-windows/ovsext/Debug.h
index a0da5eb..4b7b526 100644
--- a/datapath-windows/ovsext/Debug.h
+++ b/datapath-windows/ovsext/Debug.h
@@ -40,6 +40,7 @@
 #define OVS_DBG_OTHERS   BIT32(21)
 #define OVS_DBG_NETLINK  BIT32(22)
 #define OVS_DBG_TUNFLT   BIT32(23)
+#define OVS_DBG_STT  BIT32(24)
 
 #define OVS_DBG_RESERVED BIT32(31)
 //Please add above OVS_DBG_RESERVED.
diff --git a/datapath-windows/ovsext/Stt.c b/datapath-windows/ovsext/Stt.c
new file mode 100644
index 000..a1dae21
--- /dev/null
+++ b/datapath-windows/ovsext/Stt.c
@@ -0,0 +1,363 @@
+/*
+ * Copyright (c) 2015 VMware, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the License);
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an AS IS BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include precomp.h
+#include NetProto.h
+#include Switch.h
+#include Vport.h
+#include Flow.h
+#include stt.h
+#include IpHelper.h
+#include Checksum.h
+#include User.h
+#include PacketIO.h
+#include

[ovs-dev] [PATCH] datapath-windows: Stateless TCP Tunnelling protocol - Initial implementation

[Eitan Eliahu]

This change include an initial implementable of STT.
The following should be added:
[1] Checksum offload (SW and HW)
[2] LSO (SW and HW)
[3] IP layer WFP callout for IP segments

Testing: link layer connection through ping works. File transfer.

Signed-off-by: Eitan Eliahu elia...@vmware.com
Co-authored-by: Saurabh Shah ssaur...@vmware.com
Signed-off-by: Saurabh Shah ssaur...@vmware.com
---
 datapath-windows/automake.mk   |   2 +
 datapath-windows/ovsext/Actions.c  |  13 ++
 datapath-windows/ovsext/Debug.h|   1 +
 datapath-windows/ovsext/Stt.c  | 363 +
 datapath-windows/ovsext/Stt.h  |  87 
 datapath-windows/ovsext/Switch.h   |   3 +-
 datapath-windows/ovsext/Tunnel.c   |   3 +-
 datapath-windows/ovsext/Util.h |   1 +
 datapath-windows/ovsext/Vport.c|  18 ++
 datapath-windows/ovsext/Vport.h|  22 ++
 datapath-windows/ovsext/Vxlan.c|   4 +-
 datapath-windows/ovsext/ovsext.vcxproj |   4 +-
 12 files changed, 515 insertions(+), 6 deletions(-)
 create mode 100644 datapath-windows/ovsext/Stt.c
 create mode 100644 datapath-windows/ovsext/Stt.h

diff --git a/datapath-windows/automake.mk b/datapath-windows/automake.mk
index 9324b3c..a4f5a57 100644
--- a/datapath-windows/automake.mk
+++ b/datapath-windows/automake.mk
@@ -56,6 +56,8 @@ EXTRA_DIST += \
datapath-windows/ovsext/Vport.c \
datapath-windows/ovsext/Vport.h \
datapath-windows/ovsext/Vxlan.c \
+   datapath-windows/ovsext/Stt.h \
+   datapath-windows/ovsext/Stt.c \
datapath-windows/ovsext/Vxlan.h \
datapath-windows/ovsext/ovsext.inf \
datapath-windows/ovsext/ovsext.rc \
diff --git a/datapath-windows/ovsext/Actions.c 
b/datapath-windows/ovsext/Actions.c
index a93fe03..39f36b4 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -23,6 +23,7 @@
 #include NetProto.h
 #include Flow.h
 #include Vxlan.h
+#include Stt.h
 #include Checksum.h
 #include PacketIO.h
 
@@ -207,6 +208,10 @@ OvsDetectTunnelRxPkt(OvsForwardingContext *ovsFwdCtx,
 flowKey-ipKey.l4.tpDst == VXLAN_UDP_PORT_NBO) {
 tunnelVport = ovsFwdCtx-switchContext-vxlanVport;
 ovsActionStats.rxVxlan++;
+} else if (!flowKey-ipKey.nwFrag 
+flowKey-ipKey.nwProto == IPPROTO_TCP 
+flowKey-ipKey.l4.tpDst == STT_DST_PORT_NBO) {
+tunnelVport =ovsFwdCtx-switchContext-sttVport;
 }
 
 // We might get tunnel packets even before the tunnel gets initialized.
@@ -632,6 +637,10 @@ OvsTunnelPortTx(OvsForwardingContext *ovsFwdCtx)
ovsFwdCtx-switchContext,
(VOID *)ovsFwdCtx-completionList,
ovsFwdCtx-layers, newNbl);
+break;
+case OVS_VPORT_TYPE_STT:
+status = OvsEncapStt(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
+ ovsFwdCtx-tunKey, ovsFwdCtx-layers, newNbl);
 break;
 default:
 ASSERT(! Tx: Unhandled tunnel type);
@@ -697,6 +706,10 @@ OvsTunnelPortRx(OvsForwardingContext *ovsFwdCtx)
 status = OvsDoDecapVxlan(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
 ovsFwdCtx-tunKey, newNbl);
 break;
+case OVS_VPORT_TYPE_STT:
+status = OvsDecapStt(ovsFwdCtx-switchContext, ovsFwdCtx-curNbl,
+ ovsFwdCtx-tunKey, newNbl);
+break;
 default:
 OVS_LOG_ERROR(Rx: Unhandled tunnel type: %d\n,
   tunnelRxVport-ovsType);
diff --git a/datapath-windows/ovsext/Debug.h b/datapath-windows/ovsext/Debug.h
index a0da5eb..4b7b526 100644
--- a/datapath-windows/ovsext/Debug.h
+++ b/datapath-windows/ovsext/Debug.h
@@ -40,6 +40,7 @@
 #define OVS_DBG_OTHERS   BIT32(21)
 #define OVS_DBG_NETLINK  BIT32(22)
 #define OVS_DBG_TUNFLT   BIT32(23)
+#define OVS_DBG_STT  BIT32(24)
 
 #define OVS_DBG_RESERVED BIT32(31)
 //Please add above OVS_DBG_RESERVED.
diff --git a/datapath-windows/ovsext/Stt.c b/datapath-windows/ovsext/Stt.c
new file mode 100644
index 000..a1dae21
--- /dev/null
+++ b/datapath-windows/ovsext/Stt.c
@@ -0,0 +1,363 @@
+/*
+ * Copyright (c) 2015 VMware, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the License);
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an AS IS BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include precomp.h
+#include NetProto.h
+#include Switch.h
+#include Vport.h
+#include Flow.h
+#include stt.h
+#include IpHelper.h
+#include Checksum.h

Re: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink socket like

[Eitan Eliahu]

Hi Nithin,
Sorry, I come late on this one, change looks good.

We need to check if the IRP output buffer is big enough to fit the transaction 
message before copying it. Otherwise, we need to return error and break the 
NetLink connection.
Thanks,
Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Tuesday, April 28, 2015 2:36 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink 
socket like

In this patch, we make changes to usersapce as well as kernel datapath on 
hyperv to make it more netlink socket like. Previously, the kernel datapath did 
not distinguish between transport errors and other errors. Netlink semantics 
dictate that netlink functions should only return an error only in the case of 
a transport error
which is generally something fatal. Eg. failure to communicate with the OVS 
module, or an invalid command altogether. Other errors such as an unsupported 
action, or an invalid flow key is not considered a transport error, and in 
such cases, netlink functions are to return success with a 'struct nlmsgerr' 
populated in the output buffer.

This patch implements these semantics.

Signed-off-by: Nithin Raju nit...@vmware.com
Acked-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_72d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=gnFVocxivsB-5iJX4WfGUf-eI54hnHjhLPFrIhaeSqUs=4gTRd-4F3b_yYXbvXPMWy1sqnpt68ZC-3iT_GkKoVqce=
---
v2: addressed Sorin's comments
v3: rebase to HEAD

 datapath-windows/ovsext/Datapath.c |  61 +++---
 datapath-windows/ovsext/Datapath.h |   2 -
 lib/netlink-socket.c   | 102 ++---
 3 files changed, 114 insertions(+), 51 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index 1dead33..7646f0a 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -78,10 +78,11 @@ typedef struct _NETLINK_CMD {
 /* A netlink family is a group of commands. */  typedef struct _NETLINK_FAMILY 
{
 CHAR *name;
-UINT32 id;
+UINT16 id;
 UINT8 version;
-UINT8 pad;
+UINT8 pad1;
 UINT16 maxAttr;
+UINT16 pad2;
 NETLINK_CMD *cmds;  /* Array of netlink commands and handlers. */
 UINT16 opsCount;
 } NETLINK_FAMILY, *PNETLINK_FAMILY;
@@ -143,12 +144,12 @@ NETLINK_CMD nlControlFamilyCmdOps[] = {
 },
 { .cmd = OVS_CTRL_CMD_EVENT_NOTIFY,
   .handler = OvsReadEventCmdHandler,
-  .supportedDevOp = OVS_READ_EVENT_DEV_OP,
+  .supportedDevOp = OVS_READ_DEV_OP,
   .validateDpIndex = FALSE,
 },
 { .cmd = OVS_CTRL_CMD_READ_NOTIFY,
   .handler = OvsReadPacketCmdHandler,
-  .supportedDevOp = OVS_READ_PACKET_DEV_OP,
+  .supportedDevOp = OVS_READ_DEV_OP,
   .validateDpIndex = FALSE,
 }
 };
@@ -799,12 +800,17 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 inputBufferLen = 0;
 
 ovsMsg = ovsMsgReadOp;
-ovsMsg-nlMsg.nlmsgType = OVS_WIN_NL_CTRL_FAMILY_ID;
+RtlZeroMemory(ovsMsg, sizeof *ovsMsg);
+ovsMsg-nlMsg.nlmsgLen = sizeof *ovsMsg;
+ovsMsg-nlMsg.nlmsgType = nlControlFamilyOps.id;
 ovsMsg-nlMsg.nlmsgPid = instance-pid;
+
 /* An artificial command so we can use NL family function table*/
 ovsMsg-genlMsg.cmd = (code == OVS_IOCTL_READ_EVENT) ?
   OVS_CTRL_CMD_EVENT_NOTIFY :
   OVS_CTRL_CMD_READ_NOTIFY;
+ovsMsg-genlMsg.version = nlControlFamilyOps.version;
+
 devOp = OVS_READ_DEV_OP;
 break;
 
@@ -895,8 +901,8 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 }
 
 /*
- * For read operation, the netlink command has already been validated
- * previously.
+ * For read operation, avoid duplicate validation since 'ovsMsg' is either
+ * artificial or was copied from a previously validated 'ovsMsg'.
  */
 if (devOp != OVS_READ_DEV_OP) {
 status = ValidateNetlinkCmd(devOp, instance, ovsMsg, nlFamilyOps); @@ 
-982,7 +988,9 @@ done:
 
 /*
  * --
- * Function to invoke the netlink command handler.
+ * Function to invoke the netlink command handler. The function also 
+ stores
+ * the return value of the handler function to construct a 'NL_ERROR' 
+ message,
+ * and in turn returns success to the caller.
  * --
  */
 static NTSTATUS
@@ -1004,6 +1012,43 @@ InvokeNetlinkCmdHandler(POVS_USER_PARAMS_CONTEXT 
usrParamsCtx,
 }
 }
 
+/*
+ * Netlink socket semantics dictate that the return value of the netlink
+ * function should be an error ONLY under fatal conditions. If the 

Re: [ovs-dev] [PATCH v2 1/3] Windows installer

[Eitan Eliahu]

6.40 filter will fail to register itself  on Windows Server 2012 and will get 
unloaded.
Eitan

-Original Message-
From: Nithin Raju 
Sent: Wednesday, April 29, 2015 3:12 PM
To: Eitan Eliahu
Cc: Alin Serdean; Gurucharan Shetty; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v2 1/3] Windows installer

 On Apr 29, 2015, at 3:04 PM, Eitan Eliahu elia...@vmware.com wrote:
 
 Actually, both version should be available for the installer. It should query 
 the OS type and install the appropriate driver.
 
 But, unless there is a good reason for registering as 6.40 we could register 
 as 6.30 with the 8.1 driver.
 This way we can get reed of the 8.0 driver and the installer has a single 
 driver to install for Server and Server R2.

There are some calls to functions that are available only in 6.40. How does it 
work if such a driver gets installed on Windows 2012 (not r2)?

-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink socket like

[Eitan Eliahu]

Besides that LG.
Thanks,
Eitan

-Original Message-
From: Nithin Raju 
Sent: Wednesday, April 29, 2015 7:38 AM
To: Eitan Eliahu
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink 
socket like

Thanks for the review.

 On Apr 29, 2015, at 6:06 AM, Eitan Eliahu elia...@vmware.com wrote:
 
 
 Hi Nithin,
 Sorry, I come late on this one, change looks good.
 
 We need to check if the IRP output buffer is big enough to fit the 
 transaction message before copying it. Otherwise, we need to return error and 
 break the NetLink connection.

That’s a good point. There are some checks before hand when we validate the 
ioctl command, where we check for a “minimum size” for the output buffer. The 
parameter #3 to MapIrpOutputBuffer() is a “required length”. Pls. let me know 
if I need to add more checks.

---
 case OVS_IOCTL_READ_EVENT:
 case OVS_IOCTL_READ_PACKET:
 /*
  * Output buffer is mandatory. These IOCTLs are used to read events and
  * packets respectively. It is convenient to have separate ioctls.
  */
 if (outputBufferLen != 0) {
 status = MapIrpOutputBuffer(irp, outputBufferLen,
 sizeof *ovsMsg, outputBuffer);
---

thanks,
-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink socket like

[Eitan Eliahu]

We will fix the transaction error message copy in the kernel in a separate 
change.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
Sent: Wednesday, April 29, 2015 7:45 AM
To: Nithin Raju
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v3] ovs-hyperv: make kernel return values netlink 
socket like

On Tue, Apr 28, 2015 at 02:35:37PM -0700, Nithin Raju wrote:
 In this patch, we make changes to usersapce as well as kernel datapath 
 on hyperv to make it more netlink socket like. Previously, the kernel 
 datapath did not distinguish between transport errors and other 
 errors. Netlink semantics dictate that netlink functions should only 
 return an error only in the case of a transport error
 which is generally something fatal. Eg. failure to communicate with 
 the OVS module, or an invalid command altogether. Other errors such as 
 an unsupported action, or an invalid flow key is not considered a 
 transport error, and in such cases, netlink functions are to return 
 success with a 'struct nlmsgerr' populated in the output buffer.
 
 This patch implements these semantics.
 
 Signed-off-by: Nithin Raju nit...@vmware.com
 Acked-by: Sorin Vinturis svintu...@cloudbasesolutions.com
 Reported-at: 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvs
 witch_ovs-2Dissues_issues_72d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-
 YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Ea0P825C
 d9r9LO_WuiBF5qIYzvUpud9ODF9pApcXxhIs=VSYDudaje5XFAmpzjcYHIgg3wPtfoM3V
 b_Je1q23iske=
 ---
 v2: addressed Sorin's comments
 v3: rebase to HEAD

I applied this, thanks!

This assert in nl_sock_transact_multiple__() will fire (killing the
process) whenever the if condition is entered.  Is it really desirable?

if (request_nlmsg-nlmsg_seq != reply_nlmsg-nlmsg_seq) {
ovs_assert(request_nlmsg-nlmsg_seq == reply_nlmsg-nlmsg_seq);
VLOG_DBG_RL(rl, mismatched seq request %#PRIx32
, reply %#PRIx32, request_nlmsg-nlmsg_seq,
reply_nlmsg-nlmsg_seq);
break;
}
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Ea0P825Cd9r9LO_WuiBF5qIYzvUpud9ODF9pApcXxhIs=c85g8L1ozRygk-9THMEsnsmjkIdb8oeeVbqaPzgvLOge=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

As I understand the semantics, we need to return an error and bail out only 
when there a system level error which equivalent to a transport error or 
disconnection in a standard IPC. All transaction based errors should return a 
success by the I/O control and an NL error message should be  built in the 
reply buffer.
Incrementing the *done variable would be closer to what we want to achieve but 
until we aligned the driver with this policy this change would not be complete. 
For now, we just want to prevent the state where we enter an infinite loop.
Thanks,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Wednesday, April 15, 2015 9:10 PM
To: Nithin Raju; Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

Returning error on driver request failure would exit the while loop no matter 
if the done counter is incremented or not.
So either we return error and exit immediately the while loop, or increment the 
done counter and continue to discuss to an unresponsive driver until the 
transactions are consumed.

-Original Message-
From: Nithin Raju [mailto:nit...@vmware.com] 
Sent: Thursday, 16 April, 2015 00:22
To: Eitan Eliahu
Cc: Sorin Vinturis; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

 On Apr 15, 2015, at 1:49 PM, Eitan Eliahu elia...@vmware.com wrote:
 
 If we remove the increment of the transaction than the whole sequence of 
 transaction will be lost.

This is not true. We increment ‘*done’ after each successful transaction:
48 /* Count the number of successful transactions. */ 
49 (*done)++; 

 This should be done only after we change the driver to never fail in case the 
 transaction fail. The current driver fails also when there are transaction 
 level erros and when there are some other temporary errors. I would prefer 
 to move to the next transaction in that case rather than dump the whole 
 sequence. 

Yes, the driver needs to be updated to return STATUS_SUCCESS in most cases, 
unless there are issues with the genetlink header itself. If there’s such an 
error value returned by the kernel, instead of checking 'error = EINVAL’, we 
should check for GetLastError(). Otherwise, nl_sock_transact_multiple__() 
should behave as though no error occurred.

-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

Nithin, the only issue I have with that is that currently the driver returns an 
I/O error in some cases where the I/O succeeds but the transaction fails (e.g. 
when OVS sends down a port delete NL message and there is no such a port). In 
this case, the current user mode will dump all the other transactions in the 
same batch.

Thanks you for working on the driver side. I thought I will start to work on it 
but please go ahead.
Eitan

-Original Message-
From: Nithin Raju 
Sent: Thursday, April 16, 2015 9:06 AM
To: Eitan Eliahu
Cc: Sorin Vinturis; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

Eitan/Sorin,
This is the code on Linux:

/* Receive a reply. */  
error = nl_sock_recv__(sock, buf_txn-reply, false);
if (error) {
if (error == EAGAIN) {  
nl_sock_record_errors__(transactions, n, 0);
*done += n; 
error = 0;  
}   
break;  
}   

I am reading this as “If there’s an error, break out of the loop since the 
current transaction failed, and it is highly unlikely that future transactions 
in this batch will succeed”. At this point, we are not counting the current 
transaction that failed (with error other than EAGAIN), and hence we are not 
incrementing ‘*done’. I am advocating for the same logic to be followed on 
Windows also, in addition to breaking out of the loop.

There’s special handling for EAGIN which implies that there’s no reply queued 
up in the socket in the kernel. It is not an error as such. So, record EAGAIN 
in the transactions, and return as though the transaction completed and let the 
caller handle EAGAIN.

If you are particular that we should increment ‘*done’ on error, I am fine with 
it. I am working on the patch to make the kernel changes to return success 
during regular errors too. We’ll revisit this at that point if need be, since 
we’ll have the concrete kernel patch and come up with whatever is good 
end-to-end.

Sorin’s patch is good to go in, IMO.

thanks,
-- Nithin


 On Apr 16, 2015, at 7:06 AM, Eitan Eliahu elia...@vmware.com wrote:
 
 
 As I understand the semantics, we need to return an error and bail out only 
 when there a system level error which equivalent to a transport error or 
 disconnection in a standard IPC. All transaction based errors should return a 
 success by the I/O control and an NL error message should be  built in the 
 reply buffer.
 Incrementing the *done variable would be closer to what we want to achieve 
 but until we aligned the driver with this policy this change would not be 
 complete. For now, we just want to prevent the state where we enter an 
 infinite loop.
 Thanks,
 Eitan
 
 -Original Message-
 From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
 Sent: Wednesday, April 15, 2015 9:10 PM
 To: Nithin Raju; Eitan Eliahu
 Cc: dev@openvswitch.org
 Subject: RE: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop 
 when EINVAL is returned
 
 Returning error on driver request failure would exit the while loop no matter 
 if the done counter is incremented or not.
 So either we return error and exit immediately the while loop, or increment 
 the done counter and continue to discuss to an unresponsive driver until the 
 transactions are consumed.
 
 -Original Message-
 From: Nithin Raju [mailto:nit...@vmware.com] 
 Sent: Thursday, 16 April, 2015 00:22
 To: Eitan Eliahu
 Cc: Sorin Vinturis; dev@openvswitch.org
 Subject: Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop 
 when EINVAL is returned
 
 On Apr 15, 2015, at 1:49 PM, Eitan Eliahu elia...@vmware.com wrote:
 
 If we remove the increment of the transaction than the whole sequence of 
 transaction will be lost.
 
 This is not true. We increment ‘*done’ after each successful transaction:
 48 /* Count the number of successful transactions. */ 
 49 (*done)++; 
 
 This should be done only after we change the driver to never fail in case 
 the transaction fail. The current driver fails also when there are 
 transaction level erros and when there are some other temporary errors. I 
 would prefer to move to the next transaction in that case rather than dump 
 the whole sequence. 
 
 Yes, the driver needs to be updated to return STATUS_SUCCESS in most cases, 
 unless there are issues with the genetlink header itself. If there’s such an 
 error value returned by the kernel, instead of checking 'error = EINVAL’, we 
 should check for GetLastError(). Otherwise, nl_sock_transact_multiple__() 
 should behave as though

Re: [ovs-dev] [PATCH] INSTALL.Windows: external.1 should be part of 'br-pif'

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com 

Thanks Nithin for the prompt fix.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Thursday, April 16, 2015 9:23 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] INSTALL.Windows: external.1 should be part of 
'br-pif'

Fixing a minor typo introduced in a previous commit.

Signed-off-by: Nithin Raju nit...@vmware.com
---
 INSTALL.Windows.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md index 0dd5cda..78af0a1 
100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -299,9 +299,9 @@ as a special name to refer to that adapter.
 Port br-pif
 Interface br-pif
 type: internal
-Bridge br-int
 Port external.1
 Interface external.1
+Bridge br-int
 Port br-int
 Interface br-int
 type: internal
--
1.9.4.msysgit.2

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=PiO3Ll20LEqf8kGr70PDSOcDz5aHQBaPhNgxOXZbj6Is=r4uLERoAQXfDhHJTTEoNXcLZ7qVyr8g551-391kHePQe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com 

Thanks Sorin,
Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, April 15, 2015 2:31 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

The nl_sock_transact_multiple function enters in an infinite loop, when invalid 
error, EINVAL, is returned by nl_sock_transact_multiple__.
EINVAL is the error returned by the latter function when a driver request fails.

v2: Any error returned by nl_sock_transact_multiple__, except EAGAIN, causes 
the rest of NL transactions to be aborted.
An error response from the driver is considered a successful transaction and 
the 'done' transactions counter is incremented.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_57d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=1N6hdMturzIaFRO2ATP9dnq9fQT12iWRqfJVO_bmUvws=VP1AHCbeR7vgEFNWbe3wBmThujk0NNfiKYbP2VZdkose=
---
 lib/netlink-socket.c | 9 +
 1 file changed, 9 insertions(+)

diff --git a/lib/netlink-socket.c b/lib/netlink-socket.c index fab2a66..b73d843 
100644
--- a/lib/netlink-socket.c
+++ b/lib/netlink-socket.c
@@ -797,8 +797,12 @@ nl_sock_transact_multiple__(struct nl_sock *sock,
  txn-request-size,
  reply_buf, sizeof reply_buf,
  reply_len, NULL)) {
+/* The transaction was successfully sent but failed. */
+(*done)++;
 /* XXX: Map to a more appropriate error. */
 error = EINVAL;
+VLOG_DBG_RL(rl, fatal driver failure: %s,
+ovs_lasterror_to_string());
 break;
 }
 
@@ -909,6 +913,11 @@ nl_sock_transact_multiple(struct nl_sock *sock,
 } else if (error) {
 VLOG_ERR_RL(rl, transaction error (%s), ovs_strerror(error));
 nl_sock_record_errors__(transactions, n, error);
+if (error != EAGAIN) {
+/* A fatal error have occured. Abort the rest of
+ * transactions. */
+break;
+}
 }
 }
 }
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=1N6hdMturzIaFRO2ATP9dnq9fQT12iWRqfJVO_bmUvws=7NAcY0ICyK5QVOqfrFbaGe177U7Qu-az4hmbcyeo3U4e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2 1/2] datapath-windows: Enable extension after rrestart

[Eitan Eliahu]

Sorin,
As we discussed before we need to communicate an error from the BFE state 
change callback to the driver. Until we have this code could you please add an 
assert here:

+if (FWPM_SERVICE_RUNNING == bfeState) {
+status = OvsTunnelFilterInitialize(driverObject);
+if (!NT_SUCCESS(status)) {

ASSERT(ndisStatus == NDIS_STATUS_SUCCESS);  

+OVS_LOG_ERROR(
+Failed to initialize tunnel filter, status: %x.,
+status);
+}

Besides of that it looks good.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, April 15, 2015 3:49 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2 1/2] datapath-windows: Enable extension after 
rrestart

The extension failed to be activated during booting due to the failure to 
initialize tunnel filter. This happened because the Base Filtering Engine (BFE) 
is not started and no session to the engine could be acquired.

The solution for this was to registered a BFE notification callback that is 
called whenever the BFE's state changes. Only if the BFE's state is running the 
tunnel filter is initialized.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_77d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=PAwE2zArQ3gt-yDkBA-CMNQoYXUy5GYf0rdyR8g_jFUs=963Q2_2uIQVM9YXNkCj4eUMrsA19hfjr-BdszYb2BR0e=
---
 datapath-windows/ovsext/Switch.c   |  11 ++--
 datapath-windows/ovsext/TunnelFilter.c | 117 -
 datapath-windows/ovsext/TunnelIntf.h   |   4 +-
 3 files changed, 105 insertions(+), 27 deletions(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index 61a4531..2b9a713 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -41,6 +41,7 @@ UINT64 ovsTimeIncrementPerTick;  extern PNDIS_SPIN_LOCK 
gOvsCtrlLock;  extern NDIS_HANDLE gOvsExtDriverHandle;  extern NDIS_HANDLE 
gOvsExtDriverObject;
+extern PDEVICE_OBJECT gOvsDeviceObject;
 
 static NDIS_STATUS OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
POVS_SWITCH_CONTEXT *switchContextOut); @@ 
-202,12 +203,8 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 goto create_switch_done;
 }
 
-status = OvsTunnelFilterInitialize(gOvsExtDriverObject);
-if (status != NDIS_STATUS_SUCCESS) {
-OvsUninitSwitchContext(switchContext);
-OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
-goto create_switch_done;
-}
+OvsInitTunnelFilter(gOvsExtDriverObject, gOvsDeviceObject);
+
 *switchContextOut = switchContext;
 
 create_switch_done:
@@ -261,7 +258,7 @@ OvsDeleteSwitch(POVS_SWITCH_CONTEXT switchContext)
 if (switchContext)
 {
 dpNo = switchContext-dpNo;
-OvsTunnelFilterUninitialize(gOvsExtDriverObject);
+OvsUninitTunnelFilter(gOvsExtDriverObject);
 OvsClearAllSwitchVports(switchContext);
 OvsUninitSwitchContext(switchContext);
 OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG); diff --git 
a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c
index 4b879c0..e54ecff 100644
--- a/datapath-windows/ovsext/TunnelFilter.c
+++ b/datapath-windows/ovsext/TunnelFilter.c
@@ -111,7 +111,8 @@ DEFINE_GUID(
 PDEVICE_OBJECT gDeviceObject;
 
 HANDLE gEngineHandle = NULL;
-HANDLE gBfeSubscriptionHandle = NULL;
+HANDLE gTunnelProviderBfeHandle = NULL; HANDLE gTunnelInitBfeHandle = 
+NULL;
 UINT32 gCalloutIdV4;
 
 
@@ -547,8 +548,8 @@ Exit:
 }
 
 VOID NTAPI
-OvsBfeStateChangeCallback(PVOID context,
-  FWPM_SERVICE_STATE bfeState)
+OvsTunnelProviderBfeCallback(PVOID context,
+ FWPM_SERVICE_STATE bfeState)
 {
 HANDLE handle = NULL;
 
@@ -564,19 +565,19 @@ OvsBfeStateChangeCallback(PVOID context,  }
 
 NTSTATUS
-OvsSubscribeBfeStateChanges(PVOID deviceObject)
+OvsSubscribeTunnelProviderBfeStateChanges(PVOID deviceObject)
 {
 NTSTATUS status = STATUS_SUCCESS;
 
-if (!gBfeSubscriptionHandle) {
+if (!gTunnelProviderBfeHandle) {
 status = FwpmBfeStateSubscribeChanges(deviceObject,
-  OvsBfeStateChangeCallback,
+  
+ OvsTunnelProviderBfeCallback,
   NULL,
-  gBfeSubscriptionHandle);
+  
+ gTunnelProviderBfeHandle);
 if (!NT_SUCCESS(status)) {
 OVS_LOG_ERROR(
-Failed to open subscribe BFE state change callback, status: 
%x.,
-status);
+

Re: [ovs-dev] [PATCH v2 2/2] datapath-windows: Removed assert from FilterNetPnPEvent handler

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com 

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, April 15, 2015 3:42 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2 2/2] datapath-windows: Removed assert from 
FilterNetPnPEvent handler

Removed an inappropriate assert from the FilterNetPnPEvent routine, 
OvsExtNetPnPEvent. When NDIS calls the FilterNetPnPEvent routine, the extension 
is in paused state and, obviously, the switch is not active. The switch becomes 
active after FilterRestart routine is called and the restart is successfully 
complete.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Switch.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index 2b9a713..cfbb5a0 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -523,7 +523,6 @@ OvsExtNetPnPEvent(NDIS_HANDLE filterModuleContext,
 switchContext-isActivateFailed = TRUE;
 } else {
 ASSERT(switchContext-isActivated == FALSE);
-ASSERT(switchActive == TRUE);
 if (switchContext-isActivated == FALSE  switchActive == TRUE) {
 status = OvsActivateSwitch(switchContext);
 OVS_LOG_TRACE(OvsExtNetPnPEvent: activated switch: %p 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=CSXu-iIaADJIxM63f-MXIdiak6jSiIUo2oyeS7uqPJss=kjRXgtYuj-q0wGQqabL0U0U-XMev-5H0lYo5y2rU2LYe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

If we remove the increment of the transaction than the whole sequence of 
transaction will be lost. This should be done only after we change the driver 
to never fail in case the transaction fail. The current driver fails also when 
there are transaction level erros and when there are some other temporary 
errors. I would prefer to move to the next transaction in that case rather than 
dump the whole sequence. 
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, April 15, 2015 1:22 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v4] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

The nl_sock_transact_multiple function enters in an infinite loop, when invalid 
error, EINVAL, is returned by nl_sock_transact_multiple__.
EINVAL is the error returned by the latter function when a driver request fails.

v2: Any error returned by nl_sock_transact_multiple__, except EAGAIN, causes 
the rest of NL transactions to be aborted.
An error response from the driver is considered a successful transaction and 
the 'done' transactions counter is incremented.

v3: Removed the incrementation of the transactions counter in case of driver 
request failure.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_57d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SoJBySQ9Ab0iZ66kD5oQW1lNQi2mdEyB-hkYI_568PYs=qlfxf-x6rCbgRjcbZqeI1d7oSV9FpZDsQr1jdgHz_wIe=
Acked-by: Eitan Eliahu elia...@vmware.com
Acked-by: Nithin Raju nit...@vmware.com
---
 lib/netlink-socket.c | 9 +
 1 file changed, 9 insertions(+)

diff --git a/lib/netlink-socket.c b/lib/netlink-socket.c index fab2a66..b73d843 
100644
--- a/lib/netlink-socket.c
+++ b/lib/netlink-socket.c
@@ -797,8 +797,12 @@ nl_sock_transact_multiple__(struct nl_sock *sock,
  txn-request-size,
  reply_buf, sizeof reply_buf,
  reply_len, NULL)) {
 /* XXX: Map to a more appropriate error. */
 error = EINVAL;
+VLOG_DBG_RL(rl, fatal driver failure: %s,
+ovs_lasterror_to_string());
 break;
 }
 
@@ -909,6 +913,11 @@ nl_sock_transact_multiple(struct nl_sock *sock,
 } else if (error) {
 VLOG_ERR_RL(rl, transaction error (%s), ovs_strerror(error));
 nl_sock_record_errors__(transactions, n, error);
+if (error != EAGAIN) {
+/* A fatal error has occured. Abort the rest of
+ * transactions. */
+break;
+}
 }
 }
 }
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=SoJBySQ9Ab0iZ66kD5oQW1lNQi2mdEyB-hkYI_568PYs=CJsqfRiyJVs7TlbH5t9Cuugwakz_9cloFcKkSdXxxKke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

Yes, some cases where the transaction is completed with an error we need to 
increase the done counter, This case would be different than the case where 
the transaction didn't go through.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, April 14, 2015 2:19 PM
To: Ben Pfaff; Sorin Vinturis
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

In nl_sock_transact_multiple__ we do the following:



if (!DeviceIoControl(sock-handle, OVS_IOCTL_TRANSACT,

 txn-request-data,

 txn-request-size,

 reply_buf, sizeof reply_buf,

 reply_len, NULL)) {

/* XXX: Map to a more appropriate error. */

error = EINVAL;

break;

}



We map every failure to EINVAL that is why it did not pop out into Linux.



We should definitely log the error using ovs_lasterror_to_string before setting 
it to EINVAL.



Maybe we should just increase the number of transactions in some 
situations(i.e. STATUS_INVALID_PARAMETER) as an idea to allow the rest of the 
transactions to be processed.



Alin.

-Mesaj original-

De la: dev [mailto:dev-boun...@openvswitch.org] În numele Ben Pfaff

Trimis: Tuesday, April 14, 2015 11:42 PM

Către: Sorin Vinturis

Cc: dev@openvswitch.org

Subiect: Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction loop when 
EINVAL is returned



On Tue, Apr 14, 2015 at 08:25:59PM +, Sorin Vinturis wrote:

 The nl_sock_transact_multiple function enters in an infinite loop, 

 when invalid error, EINVAL, is returned by nl_sock_transact_multiple__.

 EINVAL is the error returned by the latter function when a driver 

 request fails.

 

 Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com

 Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com

 Reported-at: 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_57d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=YVmZuxZuPYrnWO888OUuBR52fT8FJiccWxGc1MINlzos=TyVETpKzpDFZ0fNq_BkMdUZ2ZWIiKOCnNjGFbftWyp0e=
  



I see that this fixes a bug, even on Linux.  Thank you.  It's actually a pretty 
serious bug (given the infinite loop), but I guess that it must not occur in 
any normal circumstances, otherwise we would have heard about it over the years.



However, I want to make sure of something before I commit it.

nl_sock_transact_multiple__() should only return an error in the case of a 
transport error, that is, of some problem communicating with the datapath 
(e.g. the kernel module has been removed or something similarly fatal).  It 
should not return an error in cases where some message asks the datapath to do 
something erroneous (e.g. to add a flow that the datapath doesn't understand, 
to delete a vport that doesn't exist, ...).  This is because only in the former 
case should all of the transactions be aborted; in the latter case, any 
remaining transactions should still be processed.

___

dev mailing list

dev@openvswitch.org

https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=YVmZuxZuPYrnWO888OUuBR52fT8FJiccWxGc1MINlzos=UmlNVfETCHyYzoXJ3xE5k5-JXC_h_V1gLAT1hKo6bbke=
 

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=YVmZuxZuPYrnWO888OUuBR52fT8FJiccWxGc1MINlzos=UmlNVfETCHyYzoXJ3xE5k5-JXC_h_V1gLAT1hKo6bbke=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction loop when EINVAL is returned

[Eitan Eliahu]

The DeviceIOControl() error should correspond to a transport error in Linux  or 
to a system error.
We need to transfer all transaction failures through the reply buffer 
(nlmsghdr+nlmsgerr) and return success for the DeviceIOControl().
I am aware that we didn't follow the exact  semantics in the driver and we 
sometimes fail the I/O request in the case where the transaction fails. It 
should be fixed in the driver.
Thanks,
Eitan

 

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
Sent: Tuesday, April 14, 2015 2:31 PM
To: Alin Serdean
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction loop when 
EINVAL is returned

Does every error returned by DeviceIoControl() correspond to a transport error 
(with the meaning that I explained earlier)?
nl_sock_transact_multiple__() should report that are not transport errors by 
constructing a nlmsghdr reply with an mlmsgerr as its payload.

Here is the high level situation.  With a Linux kernel, it is more efficient to 
send multiple Netlink requests to the kernel in a batch using a single system 
call.  The results are equivalent to sending the requests one-by-one.  As an 
optimization, if there is a severe error that means that no requests are going 
to succeed at all, then the code may skip sending the remaining requests.  That 
is the only case that this is meant to handle:

} else if (error) {
VLOG_ERR_RL(rl, transaction error (%s), ovs_strerror(error));
nl_sock_record_errors__(transactions, n, error);
}

This code will *not* handle ordinary errors short of this kind of severe error 
properly, and should not be invoked in that case.  (It is also buggy in that 
case, and I see that the patch fixes that bug, but that is not my point here.)  
If in fact on Windows it is getting invoked in cases other than a severe error, 
then
nl_sock_record_errors__() should be modified to report ordinary errors 
differently (probably by constructing an nlmsghdr+nlmsgerr).

Thanks,

Ben.

On Tue, Apr 14, 2015 at 09:19:10PM +, Alin Serdean wrote:
 In nl_sock_transact_multiple__ we do the following:
 
 if (!DeviceIoControl(sock-handle, OVS_IOCTL_TRANSACT,
  txn-request-data,
  txn-request-size,
  reply_buf, sizeof reply_buf,
  reply_len, NULL)) {
 /* XXX: Map to a more appropriate error. */
 error = EINVAL;
 break;
 }
 
 We map every failure to EINVAL that is why it did not pop out into Linux.
 
 We should definitely log the error using ovs_lasterror_to_string before 
 setting it to EINVAL.
 
 Maybe we should just increase the number of transactions in some 
 situations(i.e. STATUS_INVALID_PARAMETER) as an idea to allow the rest of the 
 transactions to be processed.
 
 Alin.
 -Mesaj original-
 De la: dev [mailto:dev-boun...@openvswitch.org] ??n numele Ben Pfaff
 Trimis: Tuesday, April 14, 2015 11:42 PM
 C??tre: Sorin Vinturis
 Cc: dev@openvswitch.org
 Subiect: Re: [ovs-dev] [PATCH] netlink-socket: Exit NL transaction 
 loop when EINVAL is returned
 
 On Tue, Apr 14, 2015 at 08:25:59PM +, Sorin Vinturis wrote:
  The nl_sock_transact_multiple function enters in an infinite loop, 
  when invalid error, EINVAL, is returned by nl_sock_transact_multiple__.
  EINVAL is the error returned by the latter function when a driver 
  request fails.
  
  Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
  Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
  Reported-at: 
  https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_open
  vswitch_ovs-2Dissues_issues_57d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXV
  eAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=7b
  qSnC4PRttbbFBZrlSXsx-0THTQOfNE_D4bKpkkChws=2zCC8GZSSTlR-cjRseO-dFWa
  CIQRAb8jm-uXX-zOa5Ee=
 
 I see that this fixes a bug, even on Linux.  Thank you.  It's actually a 
 pretty serious bug (given the infinite loop), but I guess that it must not 
 occur in any normal circumstances, otherwise we would have heard about it 
 over the years.
 
 However, I want to make sure of something before I commit it.
 nl_sock_transact_multiple__() should only return an error in the case of a 
 transport error, that is, of some problem communicating with the datapath 
 (e.g. the kernel module has been removed or something similarly fatal).  It 
 should not return an error in cases where some message asks the datapath to 
 do something erroneous (e.g. to add a flow that the datapath doesn't 
 understand, to delete a vport that doesn't exist, ...).  This is because only 
 in the former case should all of the transactions be aborted; in the latter 
 case, any remaining transactions should still be processed.
 ___
 dev mailing list
 dev@openvswitch.org
 

Re: [ovs-dev] [PATCH] datapath-windows: extension fails to be enabled

[Eitan Eliahu]

Hi Sorin, I am somewhat confused about this change. Have you concluded that the 
filter could not be initialized unless a provider is already registered?
I am trying to understand the relationships between this change and the change 
you check in before (also, to address the issue that the BFE had not started 
yet).
On another issue: I don't see how the filter failure to initialize is 
communicated to the driver. I think we need to log a message in the system log 
and also to put the driver in a failure state.
I'm concerned that a callback will be registered twice (On Driverentry as weel 
on Attach).
Please let us know which issue you are trying to address with this change.
Thanks,
Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, April 09, 2015 5:20 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: extension fails to be enabled

The extension failed to be activated during booting due to the failure to 
initialize tunnel filter. This happened because the Base Filtering Engine
(BFE) is not started and no session to the engine could be acquired.

The solution for this was to registered a BFE notification callback that is 
called whenever the BFE's state changes. Only if the BFE's state is running the 
tunnel filter is initialized.

Also, I have removed an inappropriate assert from the FilterNetPnPEvent 
routine, OvsExtNetPnPEvent. When NDIS calls the FilterNetPnPEvent routine, the 
extension is in paused state and, obviously, the switch is not active. The 
switch becomes active after FilterRestart routine is called and the restart is 
successfully complete.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_77d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=dBFLkhnGUaFy8lz8z0hjyZlLKcbd5GfUsaJezXD6AYws=Z1vNsPvCGNTMUH4_qkVTuNa-Xt23vnYCHeLHJ--o44oe=
---
 datapath-windows/ovsext/Switch.c   |  12 ++--
 datapath-windows/ovsext/TunnelFilter.c | 117 -
 datapath-windows/ovsext/TunnelIntf.h   |   4 +-
 3 files changed, 105 insertions(+), 28 deletions(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index 61a4531..cfbb5a0 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -41,6 +41,7 @@ UINT64 ovsTimeIncrementPerTick;  extern PNDIS_SPIN_LOCK 
gOvsCtrlLock;  extern NDIS_HANDLE gOvsExtDriverHandle;  extern NDIS_HANDLE 
gOvsExtDriverObject;
+extern PDEVICE_OBJECT gOvsDeviceObject;
 
 static NDIS_STATUS OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
POVS_SWITCH_CONTEXT *switchContextOut); @@ 
-202,12 +203,8 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 goto create_switch_done;
 }
 
-status = OvsTunnelFilterInitialize(gOvsExtDriverObject);
-if (status != NDIS_STATUS_SUCCESS) {
-OvsUninitSwitchContext(switchContext);
-OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
-goto create_switch_done;
-}
+OvsInitTunnelFilter(gOvsExtDriverObject, gOvsDeviceObject);
+
 *switchContextOut = switchContext;
 
 create_switch_done:
@@ -261,7 +258,7 @@ OvsDeleteSwitch(POVS_SWITCH_CONTEXT switchContext)
 if (switchContext)
 {
 dpNo = switchContext-dpNo;
-OvsTunnelFilterUninitialize(gOvsExtDriverObject);
+OvsUninitTunnelFilter(gOvsExtDriverObject);
 OvsClearAllSwitchVports(switchContext);
 OvsUninitSwitchContext(switchContext);
 OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG); @@ -526,7 
+523,6 @@ OvsExtNetPnPEvent(NDIS_HANDLE filterModuleContext,
 switchContext-isActivateFailed = TRUE;
 } else {
 ASSERT(switchContext-isActivated == FALSE);
-ASSERT(switchActive == TRUE);
 if (switchContext-isActivated == FALSE  switchActive == TRUE) {
 status = OvsActivateSwitch(switchContext);
 OVS_LOG_TRACE(OvsExtNetPnPEvent: activated switch: %p 
diff --git a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c
index 4b879c0..e54ecff 100644
--- a/datapath-windows/ovsext/TunnelFilter.c
+++ b/datapath-windows/ovsext/TunnelFilter.c
@@ -111,7 +111,8 @@ DEFINE_GUID(
 PDEVICE_OBJECT gDeviceObject;
 
 HANDLE gEngineHandle = NULL;
-HANDLE gBfeSubscriptionHandle = NULL;
+HANDLE gTunnelProviderBfeHandle = NULL; HANDLE gTunnelInitBfeHandle = 
+NULL;
 UINT32 gCalloutIdV4;
 
 
@@ -547,8 +548,8 @@ Exit:
 }
 
 VOID NTAPI
-OvsBfeStateChangeCallback(PVOID context,
-  FWPM_SERVICE_STATE bfeState)
+OvsTunnelProviderBfeCallback(PVOID context,
+ FWPM_SERVICE_STATE bfeState)
 {
 HANDLE handle = NULL;
 
@@ 

Re: [ovs-dev] [PATCH v2] datapath-windows: Solved BSOD when uninstalling the driver (race condition)

[Eitan Eliahu]

Hi Sorin,
This is much better (it is hard to protect on the object using the object 
itself).
Thank you!
Eitan

Acked-by: Eitan Eliahu elia...@vmware.com Thanks, Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, April 03, 2015 8:41 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Solved BSOD when uninstalling 
the driver (race condition)

The BSOD occurred because the FilterAttach routine released the switch context, 
while there were IRPs in processing.

The solution was to add a reference count to prevent premature deallocation of 
the global switch context structure, gOvsSwitchContext.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_58d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=JPNUC1kAlOusH_OvuwUMUDbLcGYZ-gCLHEDUfC60ht4s=0-K6UGrJBAH522ntyDezqkI54X6R93wF7wQAJCVzhxIe=
---
 datapath-windows/ovsext/Datapath.c | 12 ++--
 datapath-windows/ovsext/Switch.c   | 56 ++
 datapath-windows/ovsext/Switch.h   |  6 +++-
 3 files changed, 71 insertions(+), 3 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index 8eb13f1..b692225 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -701,8 +701,13 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 
 /* Check if the extension is enabled. */
 if (NULL == gOvsSwitchContext) {
-status = STATUS_DEVICE_NOT_READY;
-goto done;
+status = STATUS_NOT_FOUND;
+goto exit;
+}
+
+if (!OvsAcquireSwitchContext()) {
+status = STATUS_NOT_FOUND;
+goto exit;
 }
 
 /* Concurrent netlink operations are not supported. */ @@ -874,6 +879,9 @@ 
OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 status = InvokeNetlinkCmdHandler(usrParamsCtx, nlFamilyOps, replyLen);
 
 done:
+OvsReleaseSwitchContext(gOvsSwitchContext);
+
+exit:
 KeMemoryBarrier();
 instance-inUse = 0;
 
diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index a228d8e..94565a6 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -42,6 +42,12 @@ extern PNDIS_SPIN_LOCK gOvsCtrlLock;  extern NDIS_HANDLE 
gOvsExtDriverHandle;  extern NDIS_HANDLE gOvsExtDriverObject;
 
+/*
+ * Reference count used to prevent premature deallocation of the global 
+switch
+ * context structure, gOvsSwitchContext.
+ */
+volatile LONG  gOvsSwitchContextRefCount = 1;
+
 static NDIS_STATUS OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
POVS_SWITCH_CONTEXT *switchContextOut);  
static NDIS_STATUS OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext); @@ 
-420,6 +426,7 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 switchContext-isActivateFailed = FALSE;
 switchContext-dpNo = OVS_DP_NUMBER;
 ovsTimeIncrementPerTick = KeQueryTimeIncrement() / 1;
+
 OVS_LOG_TRACE(Exit: Succesfully initialized switchContext: %p,
   switchContext);
 return NDIS_STATUS_SUCCESS;
@@ -428,6 +435,12 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)  
static VOID  OvsUninitSwitchContext(POVS_SWITCH_CONTEXT switchContext)  {
+OvsReleaseSwitchContext(switchContext);
+}
+
+VOID
+OvsDeleteSwitchContext(POVS_SWITCH_CONTEXT switchContext) {
 OVS_LOG_TRACE(Enter: Delete switchContext:%p, switchContext);
 
 /* We need to do cleanup for tunnel port here. */ @@ -450,6 +463,49 @@ 
OvsUninitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 OVS_LOG_TRACE(Exit: Delete switchContext: %p, switchContext);  }
 
+VOID
+OvsReleaseSwitchContext(POVS_SWITCH_CONTEXT switchContext) {
+LONG ref = 0;
+LONG newRef = 0;
+LONG icxRef = 0;
+
+do {
+ref = gOvsSwitchContextRefCount;
+newRef = (0 == ref) ? 0 : ref - 1;
+icxRef = InterlockedCompareExchange(gOvsSwitchContextRefCount,
+newRef,
+ref);
+} while (icxRef != ref);
+
+if (ref == 1) {
+OvsDeleteSwitchContext(switchContext);
+}
+}
+
+BOOLEAN
+OvsAcquireSwitchContext(VOID)
+{
+LONG ref = 0;
+LONG newRef = 0;
+LONG icxRef = 0;
+BOOLEAN ret = FALSE;
+
+do {
+ref = gOvsSwitchContextRefCount;
+newRef = (0 == ref) ? 0 : ref + 1;
+icxRef = InterlockedCompareExchange(gOvsSwitchContextRefCount,
+newRef,
+ref);
+} while (icxRef != ref);
+
+if (ref != 0) {
+ret = TRUE;
+}
+
+return ret

Re: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when uninstalling the driver (race condition)

[Eitan Eliahu]

Hi Sorin,

Can you please explain how the case where  FilterDetach is called on a 
different core concurrently with the Release and between switchContext is 
checked for NULL and the swicthContext-refCount is referenced ? (where it is 
marked )
It seems that FilterDetach would set gOvsSwitchContext to NULL even if there 
are outstanding IRPs and the switch context has not been released as of yet.

+do {
+if (NULL == switchContext) {
+break;
+}

+if (0 == InterlockedCompareExchange(
+(LONG volatile *)switchContext-refCount, 0, 0)) {
+break;
+}

Can you inline the acquisition and release functions?
Thanks!
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, March 20, 2015 10:28 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when uninstalling the 
driver (race condition)

The BSOD occurred because the FilterAttach routine released the switch context, 
while there were IRPs in processing.

The solution was to add a reference count in the switch context to prevent 
premature deallocation of the switch context structure.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_58d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=bKJV7k8Y9bG2Z1Hczw1uCkihZsQUivQ4nMAjrAWy20Is=JgFv3Pq9CjJRPHRfSRgXB4ibNGyMlSl8a6kKOfNQ2_ke=
---
 datapath-windows/ovsext/Datapath.c | 10 +
 datapath-windows/ovsext/Switch.c   | 44 ++
 datapath-windows/ovsext/Switch.h   | 12 ++-
 3 files changed, 61 insertions(+), 5 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index 8eb13f1..8b561a3 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -699,10 +699,9 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 outputBufferLen = irpSp-Parameters.DeviceIoControl.OutputBufferLength;
 inputBuffer = irp-AssociatedIrp.SystemBuffer;
 
-/* Check if the extension is enabled. */
-if (NULL == gOvsSwitchContext) {
-status = STATUS_DEVICE_NOT_READY;
-goto done;
+if (!OvsAcquireSwitchContext(gOvsSwitchContext)) {
+status = STATUS_NOT_FOUND;
+goto exit;
 }
 
 /* Concurrent netlink operations are not supported. */ @@ -874,6 +873,9 @@ 
OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 status = InvokeNetlinkCmdHandler(usrParamsCtx, nlFamilyOps, replyLen);
 
 done:
+OvsReleaseSwitchContext(gOvsSwitchContext);
+
+exit:
 KeMemoryBarrier();
 instance-inUse = 0;
 
diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index a228d8e..8366944 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -420,6 +420,9 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 switchContext-isActivateFailed = FALSE;
 switchContext-dpNo = OVS_DP_NUMBER;
 ovsTimeIncrementPerTick = KeQueryTimeIncrement() / 1;
+
+switchContext-refCount = 1;
+
 OVS_LOG_TRACE(Exit: Succesfully initialized switchContext: %p,
   switchContext);
 return NDIS_STATUS_SUCCESS;
@@ -428,6 +431,12 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)  
static VOID  OvsUninitSwitchContext(POVS_SWITCH_CONTEXT switchContext)  {
+OvsReleaseSwitchContext(switchContext);
+}
+
+VOID
+OvsDeleteSwitchContext(POVS_SWITCH_CONTEXT switchContext) {
 OVS_LOG_TRACE(Enter: Delete switchContext:%p, switchContext);
 
 /* We need to do cleanup for tunnel port here. */ @@ -447,9 +456,44 @@ 
OvsUninitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 switchContext-pidHashArray = NULL;
 OvsDeleteFlowTable(switchContext-datapath);
 OvsCleanupBufferPool(switchContext);
+switchContext-refCount = 0;
 OVS_LOG_TRACE(Exit: Delete switchContext: %p, switchContext);  }
 
+VOID
+OvsReleaseSwitchContext(POVS_SWITCH_CONTEXT switchContext) {
+if (1 == InterlockedCompareExchange(
+(LONG volatile *)switchContext-refCount, 0, 1)) {
+OvsDeleteSwitchContext(switchContext);
+}
+else {
+InterlockedDecrement((LONG volatile *)switchContext-refCount);
+}
+}
+
+BOOLEAN
+OvsAcquireSwitchContext(POVS_SWITCH_CONTEXT switchContext) {
+BOOLEAN ret = FALSE;
+
+do {
+if (NULL == switchContext) {
+break;
+}
+
+if (0 == InterlockedCompareExchange(
+(LONG volatile *)switchContext-refCount, 0, 0)) {
+break;
+}
+
+InterlockedIncrement((LONG volatile *)switchContext-refCount);
+ret = TRUE;
+} while (!ret);
+
+return ret;
+}
+
 /*
  * 

Re: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when uninstalling the driver (race condition)

[Eitan Eliahu]

Hi Sorin,
Per your example: suppose there is another IRP which is currently executing and 
almost completed. This thread will call release and free the memory pointed by 
switchContext. As you said the switchContext of the first thread maintains its 
value. Then, reference to - switchContext -refCount will generate a memory 
violation.

Also, I think gOvsSwitchContext need to be memory synchronized with a memory 
barrier.
Thanks,
Eitan


-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Thursday, April 02, 2015 3:12 PM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [PATCH] datapath-windows: Solved BSOD when uninstalling the driver 
(race condition)

Hi Eitan,

I will try to explain with an example.

Considering that switchContext != NULL and switchContext-refCount == 1, we 
have the following scenario:

On thread 1, a new packet is received and OvsAcquireSwitchContext() is called. 
The first condition checks the switchContext against NULL and it is FALSE. The 
thread 1 execution is stalled and the context is switched to thread 2.

On thread 2, filter detach is called which later calls 
OvsReleaseSwitchContext(). The first condition checks to see if 
switchContext-refCount == 1. The condition is TRUE and switchContext-refCount 
is set to 0 (zero). Then the gOvsSwitchContext is released and set to NULL. 
Thread 1 resumes execution.

On thread 1, the execution is resumed and the second condition is checked. The 
second condition checks to see if switchContext-refCount == 0. The condition 
is TRUE and the function returns FALSE, whithout increasing the 
switchContext-refCount. Here, please observe that switchContext is the 
argument passed to the OvsAcquireSwitchContext() function and it points to the 
address of gOvsSwitchContext. If the latter is released and set to NULL, the 
former retains its value. Thus, the value at switchContext-refCount is valid 
and zero. Subsequent calls of OvsAcquireSwitchContext() will return FALSE, 
because the switchContext is NULL.

I hope the above explanation answers your question.

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Friday, 3 April, 2015 00:21
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [PATCH] datapath-windows: Solved BSOD when uninstalling the driver 
(race condition)



Hi Sorin,

Can you please explain how the case where  FilterDetach is called on a 
different core concurrently with the Release and between switchContext is 
checked for NULL and the swicthContext-refCount is referenced ? (where it is 
marked ) It seems that FilterDetach would set gOvsSwitchContext to NULL 
even if there are outstanding IRPs and the switch context has not been released 
as of yet.

+do {
+if (NULL == switchContext) {
+break;
+}

+if (0 == InterlockedCompareExchange(
+(LONG volatile *)switchContext-refCount, 0, 0)) {
+break;
+}

Can you inline the acquisition and release functions?
Thanks!
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, March 20, 2015 10:28 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Solved BSOD when uninstalling the 
driver (race condition)

The BSOD occurred because the FilterAttach routine released the switch context, 
while there were IRPs in processing.

The solution was to add a reference count in the switch context to prevent 
premature deallocation of the switch context structure.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_58d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=bKJV7k8Y9bG2Z1Hczw1uCkihZsQUivQ4nMAjrAWy20Is=JgFv3Pq9CjJRPHRfSRgXB4ibNGyMlSl8a6kKOfNQ2_ke=
---
 datapath-windows/ovsext/Datapath.c | 10 +
 datapath-windows/ovsext/Switch.c   | 44 ++
 datapath-windows/ovsext/Switch.h   | 12 ++-
 3 files changed, 61 insertions(+), 5 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index 8eb13f1..8b561a3 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -699,10 +699,9 @@ OvsDeviceControl(PDEVICE_OBJECT deviceObject,
 outputBufferLen = irpSp-Parameters.DeviceIoControl.OutputBufferLength;
 inputBuffer = irp-AssociatedIrp.SystemBuffer;
 
-/* Check if the extension is enabled. */
-if (NULL == gOvsSwitchContext) {
-status = STATUS_DEVICE_NOT_READY;
-goto done;
+if (!OvsAcquireSwitchContext(gOvsSwitchContext)) {
+status = STATUS_NOT_FOUND;
+goto exit;
 }
 
 /* Concurrent netlink operations are not supported. */ @@ -874,6 +873,9

Re: [ovs-dev] [PATCH 06/11] datapath-windows: Added specific pool tag for vxlan code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 06/11] datapath-windows: Added specific pool tag for 
vxlan code

All memory allocations within vxlan code have 'XSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=s8ZVryFVQEEQTOvSHLwkW76YNVil3iX9H-yeapC06-4s=ExQaobyK6h03xd4tE566aGpa_R7mN5F0Bkfdh2QcU28e=
---
 datapath-windows/ovsext/Util.h  | 1 +
 datapath-windows/ovsext/Vxlan.c | 5 +++--
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index 907cd3c..c153acf 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -27,6 +27,7 @@
 #define OVS_DATAPATH_POOL_TAG   'DSVO'
 #define OVS_EVENT_POOL_TAG  'ESVO'
 #define OVS_FLOW_POOL_TAG   'LSVO'
+#define OVS_VXLAN_POOL_TAG  'XSVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag); diff --git 
a/datapath-windows/ovsext/Vxlan.c b/datapath-windows/ovsext/Vxlan.c index 
1ce5af2..8c57185 100644
--- a/datapath-windows/ovsext/Vxlan.c
+++ b/datapath-windows/ovsext/Vxlan.c
@@ -59,7 +59,8 @@ OvsInitVxlanTunnel(POVS_VPORT_ENTRY vport,  {
 POVS_VXLAN_VPORT vxlanPort;
 
-vxlanPort = OvsAllocateMemory(sizeof (*vxlanPort));
+vxlanPort = OvsAllocateMemoryWithTag(sizeof (*vxlanPort),
+ OVS_VXLAN_POOL_TAG);
 if (vxlanPort == NULL) {
 return STATUS_INSUFFICIENT_RESOURCES;
 }
@@ -86,7 +87,7 @@ OvsCleanupVxlanTunnel(POVS_VPORT_ENTRY vport)
 return;
 }
 
-OvsFreeMemory(vport-priv);
+OvsFreeMemoryWithTag(vport-priv, OVS_VXLAN_POOL_TAG);
 vport-priv = NULL;
 }
 
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=s8ZVryFVQEEQTOvSHLwkW76YNVil3iX9H-yeapC06-4s=dq-mcIzIbOb9sCusAMy3xoUKhKPed-M1X1ye0SLpLdEe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 02/11] datapath-windows: Added specific pool tag for datapath code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 02/11] datapath-windows: Added specific pool tag for 
datapath code

All memory allocations within datapath code have 'DSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=DZ9NxwHwM2-ju5h2ApoXI4ixxtWhE5XOWxehunLNePAs=03pH_zMsgx4mHjve75q_KTCcDggcvsnovQkvAGLLWg0e=
---
 datapath-windows/ovsext/Datapath.c | 38 +++---
 datapath-windows/ovsext/Datapath.h | 31 +++
 datapath-windows/ovsext/Util.h |  1 +
 3 files changed, 39 insertions(+), 31 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index c6fe89e..888c6ef 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -349,6 +349,37 @@ extern POVS_SWITCH_CONTEXT gOvsSwitchContext;  
NDIS_SPIN_LOCK ovsCtrlLockObj;  PNDIS_SPIN_LOCK gOvsCtrlLock;
 
+NTSTATUS
+InitUserDumpState(POVS_OPEN_INSTANCE instance,
+  POVS_MESSAGE ovsMsg)
+{
+/* Clear the dumpState from a previous dump sequence. */
+ASSERT(instance-dumpState.ovsMsg == NULL);
+ASSERT(ovsMsg);
+
+instance-dumpState.ovsMsg =
+(POVS_MESSAGE)OvsAllocateMemoryWithTag(sizeof(OVS_MESSAGE),
+   OVS_DATAPATH_POOL_TAG);
+if (instance-dumpState.ovsMsg == NULL) {
+return STATUS_NO_MEMORY;
+}
+RtlCopyMemory(instance-dumpState.ovsMsg, ovsMsg,
+  sizeof *instance-dumpState.ovsMsg);
+RtlZeroMemory(instance-dumpState.index,
+  sizeof instance-dumpState.index);
+
+return STATUS_SUCCESS;
+}
+
+VOID
+FreeUserDumpState(POVS_OPEN_INSTANCE instance) {
+if (instance-dumpState.ovsMsg != NULL) {
+OvsFreeMemoryWithTag(instance-dumpState.ovsMsg,
+ OVS_DATAPATH_POOL_TAG);
+RtlZeroMemory(instance-dumpState, sizeof instance-dumpState);
+}
+}
 
 VOID
 OvsInit()
@@ -497,7 +528,8 @@ OvsAddOpenInstance(POVS_DEVICE_EXTENSION ovsExt,
PFILE_OBJECT fileObject)  {
 POVS_OPEN_INSTANCE instance =
-(POVS_OPEN_INSTANCE) OvsAllocateMemory(sizeof (OVS_OPEN_INSTANCE));
+(POVS_OPEN_INSTANCE)OvsAllocateMemoryWithTag(sizeof(OVS_OPEN_INSTANCE),
+ 
+ OVS_DATAPATH_POOL_TAG);
 UINT32 i;
 
 if (instance == NULL) {
@@ -508,7 +540,7 @@ OvsAddOpenInstance(POVS_DEVICE_EXTENSION ovsExt,
 
 if (ovsNumberOfOpenInstances = OVS_MAX_OPEN_INSTANCES) {
 OvsReleaseCtrlLock();
-OvsFreeMemory(instance);
+OvsFreeMemoryWithTag(instance, OVS_DATAPATH_POOL_TAG);
 return STATUS_INSUFFICIENT_RESOURCES;
 }
 RtlZeroMemory(instance, sizeof (OVS_OPEN_INSTANCE)); @@ -559,7 +591,7 @@ 
OvsRemoveOpenInstance(PFILE_OBJECT fileObject)
 OvsReleaseCtrlLock();
 ASSERT(instance-eventQueue == NULL);
 ASSERT (instance-packetQueue == NULL);
-OvsFreeMemory(instance);
+OvsFreeMemoryWithTag(instance, OVS_DATAPATH_POOL_TAG);
 }
 
 NTSTATUS
diff --git a/datapath-windows/ovsext/Datapath.h 
b/datapath-windows/ovsext/Datapath.h
index ba31ece..863afc4 100644
--- a/datapath-windows/ovsext/Datapath.h
+++ b/datapath-windows/ovsext/Datapath.h
@@ -129,35 +129,10 @@ InitUserParamsCtx(PIRP irp,
 usrParamsCtx-outputLength = outputLength;  }
 
-static __inline NTSTATUS
-InitUserDumpState(POVS_OPEN_INSTANCE instance,
-  POVS_MESSAGE ovsMsg)
-{
-/* Clear the dumpState from a previous dump sequence. */
-ASSERT(instance-dumpState.ovsMsg == NULL);
-ASSERT(ovsMsg);
-
-instance-dumpState.ovsMsg =
-(POVS_MESSAGE) OvsAllocateMemory(sizeof (OVS_MESSAGE));
-if (instance-dumpState.ovsMsg == NULL) {
-return STATUS_NO_MEMORY;
-}
-RtlCopyMemory(instance-dumpState.ovsMsg, ovsMsg,
-  sizeof *instance-dumpState.ovsMsg);
-RtlZeroMemory(instance-dumpState.index,
-  sizeof instance-dumpState.index);
-
-return STATUS_SUCCESS;
-}
+NTSTATUS InitUserDumpState(POVS_OPEN_INSTANCE instance,
+   POVS_MESSAGE ovsMsg);
 
-static __inline VOID
-FreeUserDumpState(POVS_OPEN_INSTANCE instance) -{
-if (instance-dumpState.ovsMsg != NULL) {
-OvsFreeMemory(instance-dumpState.ovsMsg);
-RtlZeroMemory(instance-dumpState, sizeof instance-dumpState);
-}
-}
+VOID FreeUserDumpState(POVS_OPEN_INSTANCE instance);
 
 NTSTATUS OvsSetupDumpStart(POVS_USER_PARAMS_CONTEXT usrParamsCtx

Re: [ovs-dev] [PATCH 03/11] datapath-windows: Added specific pool tag for buffermgmt code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 03/11] datapath-windows: Added specific pool tag for 
buffermgmt code

All MDL memory allocations within buffermgmt code have 'BSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=FMyEju7S1mdlucSXaZFurH5KBLKl_lSvnoefnsnhLWEs=ZmuOe6qU6VuCnDa4PWKxosVp3gACVWtZaQlPzzRgW2ce=
---
 datapath-windows/ovsext/BufferMgmt.c | 6 +++---
 datapath-windows/ovsext/Util.h   | 1 +
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/datapath-windows/ovsext/BufferMgmt.c 
b/datapath-windows/ovsext/BufferMgmt.c
index e0377c1..572b298 100644
--- a/datapath-windows/ovsext/BufferMgmt.c
+++ b/datapath-windows/ovsext/BufferMgmt.c
@@ -433,14 +433,14 @@ OvsAllocateMDLAndData(NDIS_HANDLE ndisHandle,
 PMDL mdl;
 PVOID data;
 
-data = OvsAllocateMemory(dataSize);
+data = OvsAllocateMemoryWithTag(dataSize, OVS_MDL_POOL_TAG);
 if (data == NULL) {
 return NULL;
 }
 
 mdl = NdisAllocateMdl(ndisHandle, data, dataSize);
 if (mdl == NULL) {
-OvsFreeMemory(data);
+OvsFreeMemoryWithTag(data, OVS_MDL_POOL_TAG);
 }
 
 return mdl;
@@ -454,7 +454,7 @@ OvsFreeMDLAndData(PMDL mdl)
 
 data = MmGetMdlVirtualAddress(mdl);
 NdisFreeMdl(mdl);
-OvsFreeMemory(data);
+OvsFreeMemoryWithTag(data, OVS_MDL_POOL_TAG);
 }
 
 
diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index 0f32654..6a9c169 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -23,6 +23,7 @@
 #define OVS_NBL_ONLY_POOL_TAG   'OSVO'
 #define OVS_NET_BUFFER_POOL_TAG 'NSVO'
 #define OVS_OTHER_POOL_TAG  'MSVO'
+#define OVS_MDL_POOL_TAG'BSVO'
 #define OVS_DATAPATH_POOL_TAG   'DSVO'
 
 VOID *OvsAllocateMemory(size_t size);
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=FMyEju7S1mdlucSXaZFurH5KBLKl_lSvnoefnsnhLWEs=DhACiQI8gD1DUydSnQANO5k_RZW7x-xVAHFY47P-4Zwe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 10/11] datapath-windows: Added specific pool tag for user code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 10/11] datapath-windows: Added specific pool tag for 
user code

All memory allocations within user code have 'USVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Bnsc6KJDUzV5MvIKhovWz3pmaTxrHmiYkHtZGS-RS9os=JKwXbeFASLrQohiIc2ziFw8gbBtT-LzRNIZ0o5xhJYAe=
---
 datapath-windows/ovsext/User.c | 18 ++  
datapath-windows/ovsext/Util.h |  1 +
 2 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/datapath-windows/ovsext/User.c b/datapath-windows/ovsext/User.c 
index d8a657e..03f0377 100644
--- a/datapath-windows/ovsext/User.c
+++ b/datapath-windows/ovsext/User.c
@@ -85,7 +85,7 @@ OvsPurgePacketQueue(POVS_USER_PACKET_QUEUE queue,
 LIST_FORALL_SAFE(tmp, link, next) {
 RemoveEntryList(link);
 elem = CONTAINING_RECORD(link, OVS_PACKET_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_USER_POOL_TAG);
 }
 }
 
@@ -132,13 +132,13 @@ OvsCleanupPacketQueue(POVS_OPEN_INSTANCE instance)
 LIST_FORALL_SAFE(tmp, link, next) {
 RemoveEntryList(link);
 elem = CONTAINING_RECORD(link, OVS_PACKET_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_USER_POOL_TAG);
 }
 if (irp) {
 OvsCompleteIrpRequest(irp, 0, STATUS_SUCCESS);
 }
 if (queue) {
-OvsFreeMemory(queue);
+OvsFreeMemoryWithTag(queue, OVS_USER_POOL_TAG);
 }
 
 /* Verify if gOvsSwitchContext exists. */ @@ -170,7 +170,8 @@ 
OvsSubscribeDpIoctl(PVOID instanceP,
 OvsReleasePidHashLock();
 
 } else if (instance-packetQueue == NULL  join) {
-queue = (POVS_USER_PACKET_QUEUE) OvsAllocateMemory(sizeof *queue);
+queue = (POVS_USER_PACKET_QUEUE) OvsAllocateMemoryWithTag(
+sizeof *queue, OVS_USER_POOL_TAG);
 if (queue == NULL) {
 return STATUS_NO_MEMORY;
 }
@@ -248,7 +249,7 @@ OvsReadDpIoctl(PFILE_OBJECT fileObject,
 }
 
 *replyLen = len;
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_USER_POOL_TAG);
 }
 return STATUS_SUCCESS;
 }
@@ -762,7 +763,7 @@ OvsQueuePackets(PLIST_ENTRY packetList,
 while (!IsListEmpty(dropPackets)) {
 link = RemoveHeadList(dropPackets);
 elem = CONTAINING_RECORD(link, OVS_PACKET_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_USER_POOL_TAG);
 num++;
 }
 
@@ -1060,7 +1061,8 @@ OvsCreateQueueNlPacket(PVOID userData,
 dataLen + extraLen);
 
 allocLen = sizeof (OVS_PACKET_QUEUE_ELEM) + nlMsgSize;
-elem = (POVS_PACKET_QUEUE_ELEM)OvsAllocateMemory(allocLen);
+elem = (POVS_PACKET_QUEUE_ELEM)OvsAllocateMemoryWithTag(allocLen,
+
+ OVS_USER_POOL_TAG);
 if (elem == NULL) {
 ovsUserStats.dropDuetoResource++;
 return NULL;
@@ -1163,6 +1165,6 @@ OvsCreateQueueNlPacket(PVOID userData,
 
 return elem;
 fail:
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_USER_POOL_TAG);
 return NULL;
 }
diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index 2cebe6f..a8eed91 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -31,6 +31,7 @@
 #define OVS_IPHELPER_POOL_TAG   'HSVO'
 #define OVS_OID_POOL_TAG'ASVO'
 #define OVS_SWITCH_POOL_TAG 'SSVO'
+#define OVS_USER_POOL_TAG   'USVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag);
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Bnsc6KJDUzV5MvIKhovWz3pmaTxrHmiYkHtZGS-RS9os=WWItM_TDlmXjtpG-LzyxSUcpA2l9t7q__gRTEiHLi6Ye=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 08/11] datapath-windows: Added specific pool tag for oid code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 08/11] datapath-windows: Added specific pool tag for 
oid code

All memory allocations within oid code have 'ASVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Zgt7ajB9JBKvuIzeflV2OMfV7prmxLskkHDn7l32FC4s=lMXyQUZgk7Kfxo_HiYMgnid6d11UVX1RmJf2IO23uzwe=
---
 datapath-windows/ovsext/Oid.c  | 41 +
 datapath-windows/ovsext/Oid.h  |  3 +++  datapath-windows/ovsext/Util.h |  1 +
 3 files changed, 33 insertions(+), 12 deletions(-)

diff --git a/datapath-windows/ovsext/Oid.c b/datapath-windows/ovsext/Oid.c 
index 83fa1e3..c94d17e 100644
--- a/datapath-windows/ovsext/Oid.c
+++ b/datapath-windows/ovsext/Oid.c
@@ -605,7 +605,7 @@ OvsIssueOidRequest(POVS_SWITCH_CONTEXT switchContext,
 NDIS_STATUS status;
 PNDIS_OID_REQUEST oidRequest;
 POVS_OID_CONTEXT oidContext;
-ULONG OvsExtOidRequestId = 'ISVO';
+ULONG OvsExtOidRequestId =  'ISVO';
 
 DBG_UNREFERENCED_PARAMETER(inputSize);
 DBG_UNREFERENCED_PARAMETER(oidInputBuffer);
@@ -617,15 +617,17 @@ OvsIssueOidRequest(POVS_SWITCH_CONTEXT switchContext,
 ASSERT(oidOutputBuffer == NULL || outputSize != 0);
 ASSERT(KeGetCurrentIrql() == PASSIVE_LEVEL);
 
-oidRequest = OvsAllocateMemory(sizeof *oidRequest);
+oidRequest = OvsAllocateMemoryWithTag(sizeof *oidRequest,
+  OVS_OID_POOL_TAG);
 if (!oidRequest) {
 status = NDIS_STATUS_RESOURCES;
 goto done;
 }
 
-oidContext = OvsAllocateMemory(sizeof *oidContext);
+oidContext = OvsAllocateMemoryWithTag(sizeof *oidContext,
+  OVS_OID_POOL_TAG);
 if (!oidContext) {
-OvsFreeMemory(oidRequest);
+OvsFreeMemoryWithTag(oidRequest, OVS_OID_POOL_TAG);
 status = NDIS_STATUS_RESOURCES;
 goto done;
 }
@@ -684,8 +686,8 @@ OvsIssueOidRequest(POVS_SWITCH_CONTEXT switchContext,
 status = oidContext-status;
 ASSERT(status != NDIS_STATUS_PENDING);
 
-OvsFreeMemory(oidRequest);
-OvsFreeMemory(oidContext);
+OvsFreeMemoryWithTag(oidRequest, OVS_OID_POOL_TAG);
+OvsFreeMemoryWithTag(oidContext, OVS_OID_POOL_TAG);
 
 done:
 OVS_LOG_TRACE(Exit: status %8x., status); @@ -710,7 +712,8 @@ 
OvsQuerySwitchActivationComplete(POVS_SWITCH_CONTEXT switchContext,
 OVS_LOG_TRACE(Enter: switchContext: %p, switchActive: %p,
   switchContext, switchActive);
 
-switchParams = OvsAllocateMemory(sizeof *switchParams);
+switchParams = OvsAllocateMemoryWithTag(sizeof *switchParams,
+OVS_OID_POOL_TAG);
 if (!switchParams) {
 status = NDIS_STATUS_RESOURCES;
 goto done;
@@ -741,7 +744,7 @@ OvsQuerySwitchActivationComplete(POVS_SWITCH_CONTEXT 
switchContext,
 *switchActive = switchParams-IsActive;
 }
 
-OvsFreeMemory(switchParams);
+OvsFreeMemoryWithTag(switchParams, OVS_OID_POOL_TAG);
 
 done:
 OVS_LOG_TRACE(Exit: status %8x, switchActive: %d., @@ -769,7 +772,7 @@ 
OvsGetPortsOnSwitch(POVS_SWITCH_CONTEXT switchContext,
 do {
 UINT32 reqdArraySize;
 
-portArray = OvsAllocateMemory(arraySize);
+portArray = OvsAllocateMemoryWithTag(arraySize, 
+ OVS_OID_POOL_TAG);
 if (!portArray) {
 status = NDIS_STATUS_RESOURCES;
 goto done;
@@ -794,7 +797,7 @@ OvsGetPortsOnSwitch(POVS_SWITCH_CONTEXT switchContext,
 break;
 }
 
-OvsFreeMemory(portArray);
+OvsFreeMemoryWithTag(portArray, OVS_OID_POOL_TAG);
 arraySize = reqdArraySize;
 if (status != NDIS_STATUS_INVALID_LENGTH) {
 break;
@@ -827,7 +830,7 @@ OvsGetNicsOnSwitch(POVS_SWITCH_CONTEXT switchContext,
 do {
 UINT32 reqdArraySize;
 
-nicArray = OvsAllocateMemory(arraySize);
+nicArray = OvsAllocateMemoryWithTag(arraySize, 
+ OVS_OID_POOL_TAG);
 if (!nicArray) {
 status = NDIS_STATUS_RESOURCES;
 goto done;
@@ -852,7 +855,7 @@ OvsGetNicsOnSwitch(POVS_SWITCH_CONTEXT switchContext,
 break;
 }
 
-OvsFreeMemory(nicArray);
+OvsFreeMemoryWithTag(nicArray, OVS_OID_POOL_TAG);
 arraySize = reqdArraySize;
 if (status != NDIS_STATUS_INVALID_LENGTH) {
 break;
@@ -863,3 +866,17 @@ done:
 OVS_LOG_TRACE(Exit: status %8x., status);
 return status;
 }
+
+VOID OvsFreeSwitchPortsArray

Re: [ovs-dev] [PATCH 01/11] datapath-windows: Support for allocating/releasing memory with tag

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 01/11] datapath-windows: Support for 
allocating/releasing memory with tag

Added functions for allocating and releasing memory with specified tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=iiMwE46zODzOOZShsrIWxf9yaByBOWDSCc0ipPeRGNEs=y-EsykCkI02UvBkUz93hyEZlqpJKnek7iW26IstuqK0e=
---
 datapath-windows/ovsext/Util.c | 15 +++  
datapath-windows/ovsext/Util.h |  2 ++
 2 files changed, 17 insertions(+)

diff --git a/datapath-windows/ovsext/Util.c b/datapath-windows/ovsext/Util.c 
index 2dfba8e..65dd132 100644
--- a/datapath-windows/ovsext/Util.c
+++ b/datapath-windows/ovsext/Util.c
@@ -24,6 +24,21 @@
 
 extern NDIS_HANDLE gOvsExtDriverHandle;
 
+VOID*
+OvsAllocateMemoryWithTag(size_t size, ULONG tag) {
+OVS_VERIFY_IRQL_LE(DISPATCH_LEVEL);
+return NdisAllocateMemoryWithTagPriority(gOvsExtDriverHandle,
+(UINT32)size, tag, NormalPoolPriority); }
+
+VOID
+OvsFreeMemoryWithTag(VOID *ptr, ULONG tag) {
+ASSERT(ptr);
+NdisFreeMemoryWithTagPriority(gOvsExtDriverHandle, ptr, tag); }
+
 VOID *
 OvsAllocateMemory(size_t size)
 {
diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index e752209..0303f46 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -25,8 +25,10 @@
 #define OVS_OTHER_POOL_TAG  'MSVO'
 
 VOID *OvsAllocateMemory(size_t size);
+VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag);
 VOID *OvsAllocateAlignedMemory(size_t size, UINT16 align);  VOID 
OvsFreeMemory(VOID *ptr);
+VOID OvsFreeMemoryWithTag(VOID *ptr, ULONG tag);
 VOID OvsFreeAlignedMemory(VOID *ptr);
 
 #define LIST_FORALL(_headPtr, _itemPtr) \
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=iiMwE46zODzOOZShsrIWxf9yaByBOWDSCc0ipPeRGNEs=VSyJYABI0h0fPUGH2BAojznY4kw8yJgEmhTxtzHeLJMe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 04/11] datapath-windows: Added specific pool tag for event code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 04/11] datapath-windows: Added specific pool tag for 
event code

All memory allocations within event code have 'ESVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=6Gb5VwwMRFwFfedKK4urpaqWEAnbw1nMYSGGBHiupaYs=xAwytzEw4DMilFMr41BPZ_MP9GiVW0BtFBNK-Lev0b4e=
---
 datapath-windows/ovsext/Event.c | 18 ++  
datapath-windows/ovsext/Util.h  |  1 +
 2 files changed, 11 insertions(+), 8 deletions(-)

diff --git a/datapath-windows/ovsext/Event.c b/datapath-windows/ovsext/Event.c 
index 00f4616..cca9575 100644
--- a/datapath-windows/ovsext/Event.c
+++ b/datapath-windows/ovsext/Event.c
@@ -96,9 +96,9 @@ OvsCleanupEvent(POVS_OPEN_INSTANCE instance)
 
 LIST_FORALL_SAFE(queue-elemList, link, next) {
 elem = CONTAINING_RECORD(link, OVS_EVENT_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_EVENT_POOL_TAG);
 }
-OvsFreeMemory(queue);
+OvsFreeMemoryWithTag(queue, OVS_EVENT_POOL_TAG);
 }
 }
 
@@ -139,7 +139,8 @@ OvsPostEvent(UINT32 portNo,
 portNo == OVS_DEFAULT_PORT_NO) {
 queue-pollAll = TRUE;
 } else {
-elem = (POVS_EVENT_QUEUE_ELEM)OvsAllocateMemory(sizeof(*elem));
+elem = (POVS_EVENT_QUEUE_ELEM)OvsAllocateMemoryWithTag(
+sizeof(*elem), OVS_EVENT_POOL_TAG);
 if (elem == NULL) {
 queue-pollAll = TRUE;
 } else {
@@ -158,7 +159,7 @@ OvsPostEvent(UINT32 portNo,
 LIST_FORALL_SAFE(queue-elemList, curr, next) {
 RemoveEntryList(curr);
 elem = CONTAINING_RECORD(curr, OVS_EVENT_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_EVENT_POOL_TAG);
 }
 queue-numElems = 0;
 }
@@ -243,7 +244,8 @@ OvsSubscribeEventIoctl(PFILE_OBJECT fileObject,
 }
 
 if (request-subscribe) {
-queue = (POVS_EVENT_QUEUE)OvsAllocateMemory(sizeof (OVS_EVENT_QUEUE));
+queue = (POVS_EVENT_QUEUE)OvsAllocateMemoryWithTag(
+sizeof(OVS_EVENT_QUEUE), OVS_EVENT_POOL_TAG);
 if (queue == NULL) {
 status = STATUS_NO_MEMORY;
 OVS_LOG_WARN(Fail to allocate event queue); @@ -284,9 +286,9 @@ 
done_event_subscribe:
 }
 LIST_FORALL_SAFE(queue-elemList, link, next) {
 elem = CONTAINING_RECORD(link, OVS_EVENT_QUEUE_ELEM, link);
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_EVENT_POOL_TAG);
 }
-OvsFreeMemory(queue);
+OvsFreeMemoryWithTag(queue, OVS_EVENT_POOL_TAG);
 } else {
 OvsReleaseEventQueueLock();
 }
@@ -446,7 +448,7 @@ OvsRemoveEventEntry(POVS_OPEN_INSTANCE instance,
 elem = (POVS_EVENT_QUEUE_ELEM)RemoveHeadList(queue-elemList);
 entry-portNo = elem-portNo;
 entry-status = elem-status;
-OvsFreeMemory(elem);
+OvsFreeMemoryWithTag(elem, OVS_EVENT_POOL_TAG);
 queue-numElems--;
 status = STATUS_SUCCESS;
 }
diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index 6a9c169..7811288 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -25,6 +25,7 @@
 #define OVS_OTHER_POOL_TAG  'MSVO'
 #define OVS_MDL_POOL_TAG'BSVO'
 #define OVS_DATAPATH_POOL_TAG   'DSVO'
+#define OVS_EVENT_POOL_TAG  'ESVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag);
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=6Gb5VwwMRFwFfedKK4urpaqWEAnbw1nMYSGGBHiupaYs=Y7bHH0cVHbhX-ZoGJwNH3z83HqkXVJhUJUQ9ZoOo4Cge=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 09/11] datapath-windows: Added specific pool tag for switch code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 09/11] datapath-windows: Added specific pool tag for 
switch code

All memory allocations within switch code have 'SSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=W0_d9rMWpp0MtU-vOBgF_wnxPFPQj235wBoskWDsrWws=tITgO0DRBiI7OFICgSVEdreSYadm0PhtWdDcpEhDopce=
---
 datapath-windows/ovsext/Switch.c | 53 +++-
 datapath-windows/ovsext/Util.h   |  1 +
 2 files changed, 31 insertions(+), 23 deletions(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index a228d8e..61a4531 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -168,8 +168,8 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 
 OVS_LOG_TRACE(Enter: Create switch object);
 
-switchContext =
-(POVS_SWITCH_CONTEXT) OvsAllocateMemory(sizeof(OVS_SWITCH_CONTEXT));
+switchContext = (POVS_SWITCH_CONTEXT) OvsAllocateMemoryWithTag(
+sizeof(OVS_SWITCH_CONTEXT), OVS_SWITCH_POOL_TAG);
 if (switchContext == NULL) {
 status = NDIS_STATUS_RESOURCES;
 goto create_switch_done;
@@ -187,7 +187,7 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 if (status != NDIS_STATUS_SUCCESS) {
 OVS_LOG_ERROR(OvsExtAttach: Extension is running in 
   non-switch environment.);
-OvsFreeMemory(switchContext);
+OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
 goto create_switch_done;
 }
 
@@ -198,14 +198,14 @@ OvsCreateSwitch(NDIS_HANDLE ndisFilterHandle,
 
 status = OvsInitSwitchContext(switchContext);
 if (status != NDIS_STATUS_SUCCESS) {
-OvsFreeMemory(switchContext);
+OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
 goto create_switch_done;
 }
 
 status = OvsTunnelFilterInitialize(gOvsExtDriverObject);
 if (status != NDIS_STATUS_SUCCESS) {
 OvsUninitSwitchContext(switchContext);
-OvsFreeMemory(switchContext);
+OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
 goto create_switch_done;
 }
 *switchContextOut = switchContext;
@@ -264,7 +264,7 @@ OvsDeleteSwitch(POVS_SWITCH_CONTEXT switchContext)
 OvsTunnelFilterUninitialize(gOvsExtDriverObject);
 OvsClearAllSwitchVports(switchContext);
 OvsUninitSwitchContext(switchContext);
-OvsFreeMemory(switchContext);
+OvsFreeMemoryWithTag(switchContext, OVS_SWITCH_POOL_TAG);
 }
 OVS_LOG_TRACE(Exit: deleted switch %p  dpNo: %d, switchContext, dpNo);  
} @@ -358,14 +358,14 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 switchContext-dispatchLock =
 NdisAllocateRWLock(switchContext-NdisFilterHandle);
 
-switchContext-portNoHashArray = (PLIST_ENTRY)
-OvsAllocateMemory(sizeof(LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE);
-switchContext-ovsPortNameHashArray = (PLIST_ENTRY)
-OvsAllocateMemory(sizeof (LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE);
-switchContext-portIdHashArray= (PLIST_ENTRY)
-OvsAllocateMemory(sizeof (LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE);
-switchContext-pidHashArray = (PLIST_ENTRY)
-OvsAllocateMemory(sizeof(LIST_ENTRY) * OVS_MAX_PID_ARRAY_SIZE);
+switchContext-portNoHashArray = (PLIST_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE, OVS_SWITCH_POOL_TAG);
+switchContext-ovsPortNameHashArray = 
(PLIST_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE, OVS_SWITCH_POOL_TAG);
+switchContext-portIdHashArray= (PLIST_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE, OVS_SWITCH_POOL_TAG);
+switchContext-pidHashArray = (PLIST_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(LIST_ENTRY) * OVS_MAX_PID_ARRAY_SIZE, 
+ OVS_SWITCH_POOL_TAG);
 status = OvsAllocateFlowTable(switchContext-datapath, switchContext);
 
 if (status == NDIS_STATUS_SUCCESS) { @@ -381,17 +381,20 @@ 
OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 NdisFreeRWLock(switchContext-dispatchLock);
 }
 if (switchContext-portNoHashArray) {
-OvsFreeMemory(switchContext-portNoHashArray);
+OvsFreeMemoryWithTag(switchContext-portNoHashArray,
+ OVS_SWITCH_POOL_TAG);
 }
 if (switchContext-ovsPortNameHashArray) {
-OvsFreeMemory(switchContext-ovsPortNameHashArray

Re: [ovs-dev] [PATCH 11/11] datapath-windows: Added specific pool tag for vport code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 11/11] datapath-windows: Added specific pool tag for 
vport code

All memory allocations within vport code have 'PSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=b1xAj6ZEHsXvzzx0ZlDUIGvDt0cjUMsSy8CKK0QRbTos=ZdMEPrK8nOi1frMSHgKkOSuttJG6-NsOHjVLV0XMA4we=
---
 datapath-windows/ovsext/Util.h  |  1 +
 datapath-windows/ovsext/Vport.c | 31 ---
 2 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index a8eed91..9a01242 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -32,6 +32,7 @@
 #define OVS_OID_POOL_TAG'ASVO'
 #define OVS_SWITCH_POOL_TAG 'SSVO'
 #define OVS_USER_POOL_TAG   'USVO'
+#define OVS_VPORT_POOL_TAG  'PSVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag); diff --git 
a/datapath-windows/ovsext/Vport.c b/datapath-windows/ovsext/Vport.c index 
c9dfaea..12751be 100644
--- a/datapath-windows/ovsext/Vport.c
+++ b/datapath-windows/ovsext/Vport.c
@@ -306,7 +306,7 @@ HvCreateNic(POVS_SWITCH_CONTEXT switchContext,
 OvsInitPhysNicVport(vport, virtExtVport, nicParam-NicIndex);
 status = InitHvVportCommon(switchContext, vport, TRUE);
 if (status != NDIS_STATUS_SUCCESS) {
-OvsFreeMemory(vport);
+OvsFreeMemoryWithTag(vport, OVS_VPORT_POOL_TAG);
 goto add_nic_done;
 }
 }
@@ -658,7 +658,7 @@ OvsFindVportByHvNameA(POVS_SWITCH_CONTEXT switchContext,
 SIZE_T wstrSize = length * sizeof(WCHAR);
 UINT i;
 
-PWSTR wsName = OvsAllocateMemory(wstrSize);
+PWSTR wsName = OvsAllocateMemoryWithTag(wstrSize, 
+ OVS_VPORT_POOL_TAG);
 if (!wsName) {
 return NULL;
 }
@@ -666,7 +666,7 @@ OvsFindVportByHvNameA(POVS_SWITCH_CONTEXT switchContext,
 wsName[i] = name[i];
 }
 vport = OvsFindVportByHvNameW(switchContext, wsName, wstrSize);
-OvsFreeMemory(wsName);
+OvsFreeMemoryWithTag(wsName, OVS_VPORT_POOL_TAG);
 return vport;
 }
 
@@ -703,7 +703,8 @@ POVS_VPORT_ENTRY
 OvsAllocateVport(VOID)
 {
 POVS_VPORT_ENTRY vport;
-vport = (POVS_VPORT_ENTRY)OvsAllocateMemory(sizeof (OVS_VPORT_ENTRY));
+vport = (POVS_VPORT_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(OVS_VPORT_ENTRY), OVS_VPORT_POOL_TAG);
 if (vport == NULL) {
 return NULL;
 }
@@ -1073,7 +1074,7 @@ OvsRemoveAndDeleteVport(POVS_SWITCH_CONTEXT switchContext,
 ASSERT(switchContext-numPhysicalNics == 0);
 switchContext-virtualExternalPortId = 0;
 switchContext-virtualExternalVport = NULL;
-OvsFreeMemory(vport);
+OvsFreeMemoryWithTag(vport, OVS_VPORT_POOL_TAG);
 if (vportDeallocated) {
 *vportDeallocated = TRUE;
 }
@@ -1151,7 +1152,7 @@ OvsRemoveAndDeleteVport(POVS_SWITCH_CONTEXT switchContext,
 } else {
 switchContext-numNonHvVports--;
 }
-OvsFreeMemory(vport);
+OvsFreeMemoryWithTag(vport, OVS_VPORT_POOL_TAG);
 if (vportDeallocated) {
 *vportDeallocated = TRUE;
 }
@@ -1189,19 +1190,20 @@ OvsAddConfiguredSwitchPorts(POVS_SWITCH_CONTEXT 
switchContext)
  OvsInitVportWithPortParam(vport, portParam);
  status = InitHvVportCommon(switchContext, vport, TRUE);
  if (status != NDIS_STATUS_SUCCESS) {
- OvsFreeMemory(vport);
+ OvsFreeMemoryWithTag(vport, OVS_VPORT_POOL_TAG);
  goto cleanup;
  }
 }
+
 cleanup:
 if (status != NDIS_STATUS_SUCCESS) {
 OvsClearAllSwitchVports(switchContext);
 }
 
-if (portArray != NULL) {
-OvsFreeMemory(portArray);
-}
+OvsFreeSwitchPortsArray(portArray);
+
 OVS_LOG_TRACE(Exit: status: %x, status);
+
 return status;
 }
 
@@ -1248,7 +1250,7 @@ OvsInitConfiguredSwitchNics(POVS_SWITCH_CONTEXT 
switchContext)
 nicParam-NicIndex);
 status = InitHvVportCommon(switchContext, vport, TRUE);
 if (status != NDIS_STATUS_SUCCESS) {
-OvsFreeMemory(vport);
+OvsFreeMemoryWithTag(vport, OVS_VPORT_POOL_TAG);
 vport = NULL;
 }
 }
@@ -1268,9 +1270,8 @@ OvsInitConfiguredSwitchNics(POVS_SWITCH_CONTEXT 
switchContext

Re: [ovs-dev] [PATCH 0/11] datapath-windows: Created multiple tags for memory pools

[Eitan Eliahu]

Just fyi, memory leaks can be detected when Verifier is installed. The system 
would Bugcheck and the stack could be examined.  But, it is still useful to 
have different tags for different memory allocations. 
Thanks Sorin.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 0/11] datapath-windows: Created multiple tags for 
memory pools

This patch series adds support for multiple memory pool tags for OVS extension.

Sorin Vinturis (11):
  [PATCH 01/11] datapath-windows: Support for allocating/releasing
  [PATCH 02/11] datapath-windows: Added specific pool tag for datapath code
  [PATCH 03/11] datapath-windows: Added specific pool tag for buffermgmt code
  [PATCH 04/11] datapath-windows: Added specific pool tag for event code
  [PATCH 05/11] datapath-windows: Added specific pool tag for flow code
  [PATCH 06/11] datapath-windows: Added specific pool tag for vxlan code
  [PATCH 07/11] datapath-windows: Added specific pool tag for iphelper code
  [PATCH 08/11] datapath-windows: Added specific pool tag for oid code
  [PATCH 09/11] datapath-windows: Added specific pool tag for switch code
  [PATCH 10/11] datapath-windows: Added specific pool tag for user code
  [PATCH 11/11] datapath-windows: Added specific pool tag for vport code

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=fjpaEkAVy-i-Q4snTVJPySG-gFvJHelrwgIrxhC35xYs=w2tyno-srOue5Qhi6H2R9Sw1mQ-uXE3ffnqiW1C834oe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 07/11] datapath-windows: Added specific pool tag for iphelper code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 07/11] datapath-windows: Added specific pool tag for 
iphelper code

All memory allocations within iphelper code have 'HSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=mynH1qoskm_UUJgHQKINg-7UcfKnF8Ufv72fECx0t_4s=WYhs4tALpBZ2RGUFOXJkZCaddy0XvDFzGYmRHrc1CsMe=
---
 datapath-windows/ovsext/IpHelper.c | 75 +-
 datapath-windows/ovsext/Util.h |  1 +
 2 files changed, 42 insertions(+), 34 deletions(-)

diff --git a/datapath-windows/ovsext/IpHelper.c 
b/datapath-windows/ovsext/IpHelper.c
index 0c64be4..f3e9658 100644
--- a/datapath-windows/ovsext/IpHelper.c
+++ b/datapath-windows/ovsext/IpHelper.c
@@ -14,6 +14,11 @@
  * limitations under the License.
  */
 
+#ifdef OVS_MEMORY_TAG
+#undef OVS_MEMORY_TAG
+#endif
+#define OVS_IPHELPER_POOL_TAG  'HSVO'
+
 #include precomp.h
 #include IpHelper.h
 #include Switch.h
@@ -777,7 +782,8 @@ OvsCreateIPNeighEntry(PMIB_IPNET_ROW2 ipNeigh)
 UINT64 timeVal;
 
 ASSERT(ipNeigh != NULL);
-entry = (POVS_IPNEIGH_ENTRY)OvsAllocateMemory(sizeof (OVS_IPNEIGH_ENTRY));
+entry = (POVS_IPNEIGH_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(OVS_IPNEIGH_ENTRY), OVS_IPHELPER_POOL_TAG);
 if (entry == NULL) {
 return NULL;
 }
@@ -802,8 +808,8 @@ OvsCreateIPForwardEntry(PMIB_IPFORWARD_ROW2 ipRoute)
 
 ASSERT(ipRoute);
 
-entry =
-   (POVS_IPFORWARD_ENTRY)OvsAllocateMemory(sizeof (OVS_IPFORWARD_ENTRY));
+entry = (POVS_IPFORWARD_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(OVS_IPFORWARD_ENTRY), OVS_IPHELPER_POOL_TAG);
 if (entry == NULL) {
 return NULL;
 }
@@ -823,7 +829,8 @@ OvsCreateFwdEntry(POVS_FWD_INFO fwdInfo)  {
 POVS_FWD_ENTRY entry;
 
-entry = (POVS_FWD_ENTRY)OvsAllocateMemory(sizeof (OVS_FWD_ENTRY));
+entry = (POVS_FWD_ENTRY)OvsAllocateMemoryWithTag(
+sizeof(OVS_FWD_ENTRY), OVS_IPHELPER_POOL_TAG);
 if (entry == NULL) {
 return NULL;
 }
@@ -855,7 +862,7 @@ OvsRemoveFwdEntry(POVS_FWD_ENTRY fwdEntry)
 if (ipf-refCount == 0) {
 ASSERT(IsListEmpty(ipf-fwdList));
 RemoveEntryList(ipf-link);
-OvsFreeMemory(ipf);
+OvsFreeMemoryWithTag(ipf, OVS_IPHELPER_POOL_TAG);
 }
 
 if (ipn-refCount == 0) {
@@ -864,10 +871,10 @@ OvsRemoveFwdEntry(POVS_FWD_ENTRY fwdEntry)
 NdisAcquireSpinLock(ovsIpHelperLock);
 RemoveEntryList(ipn-slink);
 NdisReleaseSpinLock(ovsIpHelperLock);
-OvsFreeMemory(ipn);
+OvsFreeMemoryWithTag(ipn, OVS_IPHELPER_POOL_TAG);
 }
 
-OvsFreeMemory(fwdEntry);
+OvsFreeMemoryWithTag(fwdEntry, OVS_IPHELPER_POOL_TAG);
 }
 
 
@@ -886,7 +893,7 @@ OvsRemoveIPForwardEntry(POVS_IPFORWARD_ENTRY ipf)
 ASSERT(ipf-refCount == 1);
 
 RemoveEntryList(ipf-link);
-OvsFreeMemory(ipf);
+OvsFreeMemoryWithTag(ipf, OVS_IPHELPER_POOL_TAG);
 }
 
 
@@ -908,7 +915,7 @@ OvsRemoveIPNeighEntry(POVS_IPNEIGH_ENTRY ipn)
 NdisAcquireSpinLock(ovsIpHelperLock);
 RemoveEntryList(ipn-slink);
 NdisReleaseSpinLock(ovsIpHelperLock);
-OvsFreeMemory(ipn);
+OvsFreeMemoryWithTag(ipn, OVS_IPHELPER_POOL_TAG);
 }
 }
 
@@ -1041,7 +1048,7 @@ OvsCleanupIpHelperRequestList(VOID)
STATUS_DEVICE_NOT_READY,
NULL);
 }
-OvsFreeMemory(request);
+OvsFreeMemoryWithTag(request, OVS_IPHELPER_POOL_TAG);
 }
 }
 
@@ -1076,8 +1083,8 @@ OvsInternalAdapterUp(UINT32 portNo,
 RtlCopyMemory(ovsInternalNetCfgId, netCfgInstanceId, sizeof (GUID));
 RtlZeroMemory(ovsInternalRow, sizeof (MIB_IF_ROW2));
 
-request =
-  (POVS_IP_HELPER_REQUEST)OvsAllocateMemory(sizeof 
(OVS_IP_HELPER_REQUEST));
+request = (POVS_IP_HELPER_REQUEST)OvsAllocateMemoryWithTag(
+sizeof(OVS_IP_HELPER_REQUEST), OVS_IPHELPER_POOL_TAG);
 if (request == NULL) {
 OVS_LOG_ERROR(Fail to initialize Internal Adapter);
 return;
@@ -1103,7 +1110,7 @@ OvsHandleInternalAdapterUp(POVS_IP_HELPER_REQUEST request)
 MIB_UNICASTIPADDRESS_ROW ipEntry;
 GUID *netCfgInstanceId = ovsInternalNetCfgId;
 
-OvsFreeMemory(request);
+OvsFreeMemoryWithTag(request, OVS_IPHELPER_POOL_TAG);
 
 status = OvsGetIfEntry(ovsInternalNetCfgId, ovsInternalRow);
 
@@ -1161,7 +1168,7 @@ OvsEnqueueIpHelperRequest(POVS_IP_HELPER_REQUEST request)
 if (ovsInternalPortNo == OVS_DEFAULT_PORT_NO ||
 ovsInternalIPConfigured == FALSE

Re: [ovs-dev] [PATCH 05/11] datapath-windows: Added specific pool tag for flow code

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, March 25, 2015 1:14 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 05/11] datapath-windows: Added specific pool tag for 
flow code

All memory allocations within flow code have 'LSVO' pool tag.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_56d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=iipaKZG9kjHd2qkNoo44q-ZVMb7RcJZlg8gJxrOVcPks=ycyG9BI8rKzkpCXyIThfDMW7j6Td_BoIEizTO49styQe=
---
 datapath-windows/ovsext/Flow.c | 11 ++-  
datapath-windows/ovsext/Util.h |  1 +
 2 files changed, 7 insertions(+), 5 deletions(-)

diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c 
index d3de8cc..97220b4 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -1512,7 +1512,7 @@ OvsDeleteFlowTable(OVS_DATAPATH *datapath)
 }
 
 DeleteAllFlows(datapath);
-OvsFreeMemory(datapath-flowTable);
+OvsFreeMemoryWithTag(datapath-flowTable, OVS_FLOW_POOL_TAG);
 datapath-flowTable = NULL;
 NdisFreeRWLock(datapath-lock);
 
@@ -1534,8 +1534,8 @@ OvsAllocateFlowTable(OVS_DATAPATH *datapath,
 PLIST_ENTRY bucket;
 int i;
 
-datapath-flowTable = OvsAllocateMemory(OVS_FLOW_TABLE_SIZE *
-sizeof (LIST_ENTRY));
+datapath-flowTable = OvsAllocateMemoryWithTag(
+OVS_FLOW_TABLE_SIZE * sizeof(LIST_ENTRY), OVS_FLOW_POOL_TAG);
 if (!datapath-flowTable) {
 return NDIS_STATUS_RESOURCES;
 }
@@ -1976,7 +1976,7 @@ VOID
 FreeFlow(OvsFlow *flow)
 {
 ASSERT(flow);
-OvsFreeMemory(flow);
+OvsFreeMemoryWithTag(flow, OVS_FLOW_POOL_TAG);
 }
 
 NTSTATUS
@@ -2259,7 +2259,8 @@ OvsPrepareFlow(OvsFlow **flow,
 
 do {
 *flow = localFlow =
-OvsAllocateMemory(sizeof(OvsFlow) + put-actionsLen);
+OvsAllocateMemoryWithTag(sizeof(OvsFlow) + put-actionsLen,
+ OVS_FLOW_POOL_TAG);
 if (localFlow == NULL) {
 status = STATUS_NO_MEMORY;
 break;
diff --git a/datapath-windows/ovsext/Util.h b/datapath-windows/ovsext/Util.h 
index 7811288..907cd3c 100644
--- a/datapath-windows/ovsext/Util.h
+++ b/datapath-windows/ovsext/Util.h
@@ -26,6 +26,7 @@
 #define OVS_MDL_POOL_TAG'BSVO'
 #define OVS_DATAPATH_POOL_TAG   'DSVO'
 #define OVS_EVENT_POOL_TAG  'ESVO'
+#define OVS_FLOW_POOL_TAG   'LSVO'
 
 VOID *OvsAllocateMemory(size_t size);
 VOID *OvsAllocateMemoryWithTag(size_t size, ULONG tag);
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=iipaKZG9kjHd2qkNoo44q-ZVMb7RcJZlg8gJxrOVcPks=DLuBxSaXBI450vCDtn3ZzfmOso1IDrrcGBsEazM3l2Ue=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider handling

[Eitan Eliahu]

Acked-by: Eitan Eliahu  elia...@vmware.com

Thanks Sorin.
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Tuesday, March 24, 2015 12:29 AM
To: Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling

Eitan,

The BFE is not started when the machine is booted and OVS driver is 
initialized. This happens all the time. Due to this, the WFP system provider is 
not added.

When the OS is started and the userspace tools are executed the BFE is running 
and all WFP objects are correctly added. At this point, if the commands 'net 
stop ovsext' + 'net start ovsext' are executed all OVS initializations are 
successfully performed, including the addition of the WFP system provider.

So this patch is trying to address the scenario when the WFP system provider 
addition is performed at boot time.

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Tuesday, 24 March, 2015 06:37
To: Sorin Vinturis
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling

Thanks Sorin,
What would be the case where BFE is not started before the driver initializes?
When BFE is not started does the driver continue with WFP rule setiings?
Eitan



-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Monday, March 23, 2015 4:40 PM
To: Sorin Vinturis; Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling

As I explained in the patch description, I have added this change because, at 
driver initialization phase, the system provider failed to be added due to the 
fact that the Base Filtering Engine (BFE) is not started and no session to the 
engine could be acquired, i.e. OvsTunnelEngineOpen() fails.



-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis

Sent: Tuesday, 24 March, 2015 01:01

To: Eitan Eliahu

Cc: dev@openvswitch.org

Subject: Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Hi Eitan,



In this case the second call will enter this code path:



 status = FwpmProviderAdd(handle,

  provider,

  NULL);

 if (!NT_SUCCESS(status)) {

-OVS_LOG_ERROR(Fail to add WFP provider, status: %x., status);

-break;

+if (STATUS_FWP_ALREADY_EXISTS != status) {

+OVS_LOG_ERROR(Failed to add WFP provider, status: %x.,

+  status);

+break;

+}

 }



Thanks,

Sorin



-Original Message-

From: Eitan Eliahu [mailto:elia...@vmware.com] 

Sent: Monday, 23 March, 2015 18:13

To: Sorin Vinturis

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Sorin,

Ok on [1]. Thanks.



On [2], I can see a scenario when the provider can added twice from 
OvsRegisterSystemProvider() as well as from the callback



if (FWPM_SERVICE_RUNNING == FwpmBfeStateGet()) {    Callback is 
called from a different thread context !

 OvsTunnelEngineOpen(handle);





How do you prevent it?



Also, can you please explain which issue you ran into which triggered this 
change?

Thank you,

Eitan



-Original Message-

From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com]

Sent: Monday, March 23, 2015 1:37 AM

To: Eitan Eliahu

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Hi Eitan,



Thanks for reviewing my patch.

[1] The OvsTunnelEngineClose(HANDLE *handle) is verifying the handle it 
receives before closing it.

[2] In order to retrieve the current state of the filter engine I am using the 
FwpmBfeStateGet function. But before calling FwpmBfeStateGet, the callout 
driver MUST call the FwpmBfeStateSubscribeChanges function. That is the reason 
I am registering the callback no mater of the state of the filter engine. 

Only if the engine is running the provider is added by opening the engine 
handle first and closing it after the provider is added, then unsubscribes the 
callback.

If the engine is not running the callback is called and the provider is added 
only for the service running notification.



Hope this answers your comments.



Thanks,

Sorin



-Original Message-

From: Eitan Eliahu [mailto:elia...@vmware.com]

Sent: Monday, 23 March, 2015 02:32

To: Sorin Vinturis

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling





Hi Sorin,

Thank you for posting this change.

Here are few comments:

[1] Please call OvsTunnelEngineClose(handle) only if the handle is not zero. 
(you might want to add an assertion if it zero

Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider handling

[Eitan Eliahu]

Sorin,
Ok on [1]. Thanks.

On [2], I can see a scenario when the provider can added twice from 
OvsRegisterSystemProvider() as well as from the callback

if (FWPM_SERVICE_RUNNING == FwpmBfeStateGet()) {
   Callback is called from a different thread context !
 OvsTunnelEngineOpen(handle);


How do you prevent it?

Also, can you please explain which issue you ran into which triggered this 
change?
Thank you,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Monday, March 23, 2015 1:37 AM
To: Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling

Hi Eitan,

Thanks for reviewing my patch.
[1] The OvsTunnelEngineClose(HANDLE *handle) is verifying the handle it 
receives before closing it.
[2] In order to retrieve the current state of the filter engine I am using the 
FwpmBfeStateGet function. But before calling FwpmBfeStateGet, the callout 
driver MUST call the FwpmBfeStateSubscribeChanges function. That is the reason 
I am registering the callback no mater of the state of the filter engine. 
Only if the engine is running the provider is added by opening the engine 
handle first and closing it after the provider is added, then unsubscribes the 
callback.
If the engine is not running the callback is called and the provider is added 
only for the service running notification.

Hope this answers your comments.

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Monday, 23 March, 2015 02:32
To: Sorin Vinturis
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling


Hi Sorin,
Thank you for posting this change.
Here are few comments:
[1] Please call OvsTunnelEngineClose(handle) only if the handle is not zero. 
(you might want to add an assertion if it zero).
[2] I'm somewhat confused about the registration for engine state change. As I 
understand you check the state of the engine and if it is running you open the 
engine handle (in the context of the foreground thread). However, you subscribe 
to state change anyway and you open the handle to the engine from the callback 
function too.
Did I miss anything?
Thanks,
Eitan


-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis

Sent: Monday, 16 March, 2015 20:08

To: dev@openvswitch.org

Subject: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



If the Base Filtering Engine (BFE) is not started, the WFP system provider 
failed to be added because no session to the engine could be acquired.



The solution for this was to registered a BFE notification callback that is 
called whenever the BFE's state changes. Only if the BFE's state is running the 
WFP system provider is added.



Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_65d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=MVriWGIsZGTQ5f_haLS0X5H8R3YdfikbqnWKRpKksKUs=1iWqFul95i3o_mD41jKjZNdLnK-df_QWRzSY1614RTwe=
 

---

 datapath-windows/ovsext/Datapath.c | 20 ++-

 datapath-windows/ovsext/TunnelFilter.c | 99 --

 datapath-windows/ovsext/TunnelIntf.h   |  8 +--

 3 files changed, 100 insertions(+), 27 deletions(-)



diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c

index 8eb13f1..c6fe89e 100644

--- a/datapath-windows/ovsext/Datapath.c

+++ b/datapath-windows/ovsext/Datapath.c

@@ -353,35 +353,19 @@ PNDIS_SPIN_LOCK gOvsCtrlLock;  VOID

 OvsInit()

 {

-HANDLE handle = NULL;

-

 gOvsCtrlLock = ovsCtrlLockObj;

 NdisAllocateSpinLock(gOvsCtrlLock);

 OvsInitEventQueue();

-

-OvsTunnelEngineOpen(handle);

-if (handle) {

-OvsTunnelAddSystemProvider(handle);

-}

-OvsTunnelEngineClose(handle);

 }

 

 VOID

 OvsCleanup()

 {

-HANDLE handle = NULL;

-

 OvsCleanupEventQueue();

 if (gOvsCtrlLock) {

 NdisFreeSpinLock(gOvsCtrlLock);

 gOvsCtrlLock = NULL;

 }

-

-OvsTunnelEngineOpen(handle);

-if (handle) {

-OvsTunnelRemoveSystemProvider(handle);

-}

-OvsTunnelEngineClose(handle);

 }

 

 VOID

@@ -448,6 +432,8 @@ OvsCreateDeviceObject(NDIS_HANDLE ovsExtDriverHandle)

 if (ovsExt) {

 ovsExt-numberOpenInstance = 0;

 }

+} else {

+OvsRegisterSystemProvider((PVOID)gOvsDeviceObject);

 }

 

 OVS_LOG_TRACE(DeviceObject: %p, gOvsDeviceObject); @@ -471,6 +457,8 @@ 
OvsDeleteDeviceObject()

 NdisDeregisterDeviceEx(gOvsDeviceHandle);

 gOvsDeviceHandle = NULL;

 gOvsDeviceObject = NULL

Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider handling

[Eitan Eliahu]

Thanks Sorin,
What would be the case where BFE is not started before the driver initializes?
When BFE is not started does the driver continue with WFP rule setiings?
Eitan



-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Monday, March 23, 2015 4:40 PM
To: Sorin Vinturis; Eitan Eliahu
Cc: dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling

As I explained in the patch description, I have added this change because, at 
driver initialization phase, the system provider failed to be added due to the 
fact that the Base Filtering Engine (BFE) is not started and no session to the 
engine could be acquired, i.e. OvsTunnelEngineOpen() fails.



-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis

Sent: Tuesday, 24 March, 2015 01:01

To: Eitan Eliahu

Cc: dev@openvswitch.org

Subject: Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Hi Eitan,



In this case the second call will enter this code path:



 status = FwpmProviderAdd(handle,

  provider,

  NULL);

 if (!NT_SUCCESS(status)) {

-OVS_LOG_ERROR(Fail to add WFP provider, status: %x., status);

-break;

+if (STATUS_FWP_ALREADY_EXISTS != status) {

+OVS_LOG_ERROR(Failed to add WFP provider, status: %x.,

+  status);

+break;

+}

 }



Thanks,

Sorin



-Original Message-

From: Eitan Eliahu [mailto:elia...@vmware.com] 

Sent: Monday, 23 March, 2015 18:13

To: Sorin Vinturis

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Sorin,

Ok on [1]. Thanks.



On [2], I can see a scenario when the provider can added twice from 
OvsRegisterSystemProvider() as well as from the callback



if (FWPM_SERVICE_RUNNING == FwpmBfeStateGet()) {    Callback is 
called from a different thread context !

 OvsTunnelEngineOpen(handle);





How do you prevent it?



Also, can you please explain which issue you ran into which triggered this 
change?

Thank you,

Eitan



-Original Message-

From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com]

Sent: Monday, March 23, 2015 1:37 AM

To: Eitan Eliahu

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



Hi Eitan,



Thanks for reviewing my patch.

[1] The OvsTunnelEngineClose(HANDLE *handle) is verifying the handle it 
receives before closing it.

[2] In order to retrieve the current state of the filter engine I am using the 
FwpmBfeStateGet function. But before calling FwpmBfeStateGet, the callout 
driver MUST call the FwpmBfeStateSubscribeChanges function. That is the reason 
I am registering the callback no mater of the state of the filter engine. 

Only if the engine is running the provider is added by opening the engine 
handle first and closing it after the provider is added, then unsubscribes the 
callback.

If the engine is not running the callback is called and the provider is added 
only for the service running notification.



Hope this answers your comments.



Thanks,

Sorin



-Original Message-

From: Eitan Eliahu [mailto:elia...@vmware.com]

Sent: Monday, 23 March, 2015 02:32

To: Sorin Vinturis

Cc: dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling





Hi Sorin,

Thank you for posting this change.

Here are few comments:

[1] Please call OvsTunnelEngineClose(handle) only if the handle is not zero. 
(you might want to add an assertion if it zero).

[2] I'm somewhat confused about the registration for engine state change. As I 
understand you check the state of the engine and if it is running you open the 
engine handle (in the context of the foreground thread). However, you subscribe 
to state change anyway and you open the handle to the engine from the callback 
function too.

Did I miss anything?

Thanks,

Eitan





-Original Message-



From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis



Sent: Monday, 16 March, 2015 20:08



To: dev@openvswitch.org



Subject: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling







If the Base Filtering Engine (BFE) is not started, the WFP system provider 
failed to be added because no session to the engine could be acquired.







The solution for this was to registered a BFE notification callback that is 
called whenever the BFE's state changes. Only if the BFE's state is running the 
WFP system provider is added.







Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com



Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com



Reported-at: 
https://urldefense.proofpoint.com/v2/url

Re: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider handling

[Eitan Eliahu]

Hi Sorin,
Thank you for posting this change.
Here are few comments:
[1] Please call OvsTunnelEngineClose(handle) only if the handle is not zero. 
(you might want to add an assertion if it zero).
[2] I'm somewhat confused about the registration for engine state change. As I 
understand you check the state of the engine and if it is running you open the 
engine handle (in the context of the foreground thread). However, you subscribe 
to state change anyway and you open the handle to the engine from the callback 
function too.
Did I miss anything?
Thanks,
Eitan


-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis

Sent: Monday, 16 March, 2015 20:08

To: dev@openvswitch.org

Subject: [ovs-dev] [PATCH] datapath-windows: Updated WFP system provider 
handling



If the Base Filtering Engine (BFE) is not started, the WFP system provider 
failed to be added because no session to the engine could be acquired.



The solution for this was to registered a BFE notification callback that is 
called whenever the BFE's state changes. Only if the BFE's state is running the 
WFP system provider is added.



Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-by: Sorin Vinturis svintu...@cloudbasesolutions.com

Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_65d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=MVriWGIsZGTQ5f_haLS0X5H8R3YdfikbqnWKRpKksKUs=1iWqFul95i3o_mD41jKjZNdLnK-df_QWRzSY1614RTwe=
 

---

 datapath-windows/ovsext/Datapath.c | 20 ++-

 datapath-windows/ovsext/TunnelFilter.c | 99 --

 datapath-windows/ovsext/TunnelIntf.h   |  8 +--

 3 files changed, 100 insertions(+), 27 deletions(-)



diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c

index 8eb13f1..c6fe89e 100644

--- a/datapath-windows/ovsext/Datapath.c

+++ b/datapath-windows/ovsext/Datapath.c

@@ -353,35 +353,19 @@ PNDIS_SPIN_LOCK gOvsCtrlLock;  VOID

 OvsInit()

 {

-HANDLE handle = NULL;

-

 gOvsCtrlLock = ovsCtrlLockObj;

 NdisAllocateSpinLock(gOvsCtrlLock);

 OvsInitEventQueue();

-

-OvsTunnelEngineOpen(handle);

-if (handle) {

-OvsTunnelAddSystemProvider(handle);

-}

-OvsTunnelEngineClose(handle);

 }

 

 VOID

 OvsCleanup()

 {

-HANDLE handle = NULL;

-

 OvsCleanupEventQueue();

 if (gOvsCtrlLock) {

 NdisFreeSpinLock(gOvsCtrlLock);

 gOvsCtrlLock = NULL;

 }

-

-OvsTunnelEngineOpen(handle);

-if (handle) {

-OvsTunnelRemoveSystemProvider(handle);

-}

-OvsTunnelEngineClose(handle);

 }

 

 VOID

@@ -448,6 +432,8 @@ OvsCreateDeviceObject(NDIS_HANDLE ovsExtDriverHandle)

 if (ovsExt) {

 ovsExt-numberOpenInstance = 0;

 }

+} else {

+OvsRegisterSystemProvider((PVOID)gOvsDeviceObject);

 }

 

 OVS_LOG_TRACE(DeviceObject: %p, gOvsDeviceObject); @@ -471,6 +457,8 @@ 
OvsDeleteDeviceObject()

 NdisDeregisterDeviceEx(gOvsDeviceHandle);

 gOvsDeviceHandle = NULL;

 gOvsDeviceObject = NULL;

+

+OvsUnregisterSystemProvider();

 }

 }

 

diff --git a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c

index e0adc37..4b879c0 100644

--- a/datapath-windows/ovsext/TunnelFilter.c

+++ b/datapath-windows/ovsext/TunnelFilter.c

@@ -111,6 +111,7 @@ DEFINE_GUID(

 PDEVICE_OBJECT gDeviceObject;

 

 HANDLE gEngineHandle = NULL;

+HANDLE gBfeSubscriptionHandle = NULL;

 UINT32 gCalloutIdV4;

 

 

@@ -173,17 +174,20 @@ OvsTunnelAddSystemProvider(HANDLE handle)

 provider.displayData.name = OVS_TUNNEL_PROVIDER_NAME;

 provider.displayData.description = OVS_TUNNEL_PROVIDER_DESC;

 /*

-* Since we always want the provider to be present, it's easiest to add

-* it as persistent object during driver load.

-*/

+ * Since we always want the provider to be present, it's easiest to add

+ * it as persistent object during driver load.

+ */

 provider.flags = FWPM_PROVIDER_FLAG_PERSISTENT;

 

 status = FwpmProviderAdd(handle,

  provider,

  NULL);

 if (!NT_SUCCESS(status)) {

-OVS_LOG_ERROR(Fail to add WFP provider, status: %x., status);

-break;

+if (STATUS_FWP_ALREADY_EXISTS != status) {

+OVS_LOG_ERROR(Failed to add WFP provider, status: %x.,

+  status);

+break;

+}

 }

 

 status = FwpmTransactionCommit(handle); @@ -541,3 +545,88 @@ Exit:

 

 return status;

 }

+

+VOID NTAPI

+OvsBfeStateChangeCallback(PVOID context,

+  FWPM_SERVICE_STATE bfeState) {

+HANDLE handle = NULL;

Re: [ovs-dev] OVS Windows MSI installer

[Eitan Eliahu]

I think we can drop 32bit support once the 64bit is there.
Thanks!
Eitan

-Original Message-
From: Nithin Raju 
Sent: Monday, March 16, 2015 11:10 PM
To: Alessandro Pilotti
Cc: Eitan Eliahu; dev@openvswitch.org
Subject: Re: [ovs-dev] OVS Windows MSI installer

 On Mar 15, 2015, at 7:54 PM, Alessandro Pilotti 
 apilo...@cloudbasesolutions.com wrote:
 
 
 
 
 On 16 mar. 2015, at 03:30, Eitan Eliahu elia...@vmware.com wrote:
 
 
 Alessandro,
 Yes, no problem of having the installer separate.  But, I am confused as I 
 see the two services are there but the driver and the INF file are not. I 
 believe that the MSI file includes everything.
 Do you think you can add the OS installer script?
 Will we have already built MSI file (x86 and x86-64)?
 
 We'll commit the x64 MSI build code as soon as we have the x64 OVS build 
 merged upstream.
 
 One thing that we should decide is if it makes sense to keep a separate x86 
 build at all at that point. 
 
 Since Hyper-V is limited to x64, the only use case for a x86 build is for a 
 32bit client OS build where a user wants to install the CLI tools only.
 Given how limited this use case is, I'd go with a single x64 build.

This is fine. If the x64 binaries can be built reliably and pass the tests, I 
think we can flip the switch on the default “mode” and just go with x64 
binaries. I maybe wrong, but there’s no real usecase for 32 bit binaries on 
Hyper-V.

-- Nithin
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] INSTALL.Windows.md: Mention the easy way to get into bash.

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Tuesday, March 17, 2015 3:31 PM
To: dev@openvswitch.org
Cc: Gurucharan Shetty; Hemanth Kumar Mantri
Subject: [ovs-dev] [PATCH] INSTALL.Windows.md: Mention the easy way to get into 
bash.

Just typing 'bash' from MSVC developers command prompt _can_ cause MSYS to use 
the Windows version of 'sort' instead of the GNU sort.

Reported-by: Hemanth Kumar Mantri man...@nutanix.com
Signed-off-by: Gurucharan Shetty gshe...@nicira.com
---
 AUTHORS|1 +
 INSTALL.Windows.md |6 --
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/AUTHORS b/AUTHORS
index d7925db..008f6f1 100644
--- a/AUTHORS
+++ b/AUTHORS
@@ -246,6 +246,7 @@ Gur Stavi   gst...@mrv.com
 Hari Sasank Bhamidipalli hbham...@cisco.com
 Hassan Khan hassan.k...@seecs.edu.pk
 Hector Oron hector.o...@gmail.com
+Hemanth Kumar Mantriman...@nutanix.com
 Henrik Amrenhen...@nicira.com
 Hiroshi Tanaka  htan...@nicira.com
 Hiroshi Miyata  miyahiro.d...@gmail.com
diff --git a/INSTALL.Windows.md b/INSTALL.Windows.md index 258e2fb..e84013a 
100644
--- a/INSTALL.Windows.md
+++ b/INSTALL.Windows.md
@@ -35,11 +35,13 @@ install Windows Driver Kit (WDK) 8.1 Update.
 It is important to get the Visual Studio related environment variables and to  
have the $PATH inside the bash to point to the proper compiler and linker. One  
easy way to achieve this is to get into the Developer Command prompt for 
visual -studio and through it enter into the bash shell available from msys.
+studio and through it enter into the bash shell available from msys by 
+typing 'bash --login'.
 
 If after the above step, a 'which link' inside MSYS's bash says,  
/bin/link.exe, rename /bin/link.exe to something else so that the -Visual 
studio's linker is used.
+Visual studio's linker is used. You should also see a 'which sort' 
+report /bin/sort.exe.
 
 * For pthread support, install the library, dll and includes of pthreads-win32 
 project from
--
1.7.9.5

___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=FvYUU6lQk0zG7NvgUA9l1JyOC00f4vxHTB9Q7eugHYAs=EqNY4ilrR619-1aWCH2-U0UXQx_lO3UlKNOr_nFxqRAe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] OVS Windows MSI installer

[Eitan Eliahu]

I don't have a strong opinion but once the installer  is included in the OVS 
tree all directory structure is fixed so there is no ambiguity on how or where 
the installer collects the files. It will also avoid having more than one copy 
of the same file.

It would be nice also if the installer and the MSI file generated by the 
installer could be built as part of OVS user mode. 

That said, the MSI file itself should be available to users w/o the need to 
build anything and would be related to an OVS release. 

We need get the driver signed and officially pass WHQL (We can discuss it 
later).

Thanks,
Eitan.



-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Ben Pfaff
Sent: Tuesday, March 17, 2015 8:26 AM
To: Nithin Raju
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] OVS Windows MSI installer

On Tue, Mar 17, 2015 at 06:06:47AM +, Nithin Raju wrote:
 Regarding where the installer code should be located, my vote would be 
 to keep this within the OVS repo. A good reason to keep it outside is 
 if this code gets reused for installing or packaging non-OVS 
 executables. At the Openstack summit we had discussed that the 
 Cloudbase Neutron plugin installer can also install OVS in the future. 
 If we are talking of such an installer, then it makes sense to put it 
 in a different repo. Since this installer is specific to OVS, I think 
 we should keep it within the OVS repo. (I am probably restating what 
 Ben already said).

I'd like to merge it into the OVS tree, too, after looking at it.  That way we 
don't start to splinter packaging off into other trees and make it more 
difficult for users to find a set of code that works well together.
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=aCQkLNl5h_C4Zsh3AoB2MbzKkws4kvcT99BiiUgzYHYs=qhwi87O8u3mnEl1sFfbCr2Gtbw4uYPN3tArQdqXiov0e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 3/3] netlink: support for asynchronous NL transactions

[Eitan Eliahu]

Hi Sorin,
It would be probably better to make this call synchronous as the thread has 
nothing else to do. The WFP transaction in the driver is guaranteed to being 
completed so we should be ok here.
Thanks,
Eitan 

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, March 12, 2015 5:43 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 3/3] netlink: support for asynchronous NL transactions

Modified 'nl_sock_transact_multiple__' function to send asynchronous I/O 
requests to the kernel driver.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_64d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=mnPxbs5kxqNdJzi-qc-vofl6JQrEuyx3Tb5Uxj5l7YAs=vGvK7gc_nZkPLpeiZVXNdRHky8DjHd9aT46oNKn8lmIe=
---
 lib/netlink-socket.c | 26 ++
 1 file changed, 22 insertions(+), 4 deletions(-)

diff --git a/lib/netlink-socket.c b/lib/netlink-socket.c index e4f153f..64f93ff 
100644
--- a/lib/netlink-socket.c
+++ b/lib/netlink-socket.c
@@ -832,10 +832,28 @@ nl_sock_transact_multiple__(struct nl_sock *sock,
  ofpbuf_data(txn-request),
  ofpbuf_size(txn-request),
  reply_buf, sizeof reply_buf,
- reply_len, NULL)) {
-/* XXX: Map to a more appropriate error. */
-error = EINVAL;
-break;
+ reply_len, sock-overlapped)) {
+if (GetLastError() != ERROR_IO_PENDING) {
+/* XXX: Map to a more appropriate error. */
+VLOG_DBG_RL(rl, nl_transact operation failed - %s,
+ovs_strerror(error));
+error = EINVAL;
+break;
+} else {
+/* Operation has been queued and will complete in the future. 
*/
+/* Wait for the operation to complete before continuing. */
+if (!GetOverlappedResult(sock-handle,
+sock-overlapped,
+reply_len,
+TRUE)) {
+/* Operation has completed, but it failed. */
+/* XXX: Map to a more appropriate error. */
+VLOG_DBG_RL(rl, nl_transact asynchronous operation 
failed - %s,
+ovs_strerror(error));
+error = EINVAL;
+break;
+}
+}
 }
 
 if (reply_len  sizeof *reply_nlmsg) {
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=mnPxbs5kxqNdJzi-qc-vofl6JQrEuyx3Tb5Uxj5l7YAs=Ea6-zNGdcWXl55yngXjvxDyuRn1v9WCsX8B3jJj5mtAe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2 2/3] datapath-windows: Support for multiple VXLAN tunnels

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, March 12, 2015 5:43 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2 2/3] datapath-windows: Support for multiple VXLAN 
tunnels

At the moment the OVS extension supports only one VXLAN tunnel that is cached 
in the extension switch context. Replaced the latter cached pointer with an 
array list that contains all VXLAN tunnel vports.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_64d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=GmpbZWnQYzaDAxKvmJ05dBJ8JYOtGdxMBDiXWBzMytEs=IwHmCeVaHOwt0p6FccEexn_7ay5hwkPhDo4yRbp-cEke=
---
 datapath-windows/ovsext/Actions.c | 13   
datapath-windows/ovsext/Switch.c  | 16 ++-  
datapath-windows/ovsext/Switch.h  |  6 --  datapath-windows/ovsext/Tunnel.c 
 |  3 ++-
 datapath-windows/ovsext/Vport.c   | 43 +++
 datapath-windows/ovsext/Vport.h   | 21 +--
 datapath-windows/ovsext/Vxlan.c   |  2 +-
 datapath-windows/ovsext/Vxlan.h   |  4 +---
 8 files changed, 67 insertions(+), 41 deletions(-)

diff --git a/datapath-windows/ovsext/Actions.c 
b/datapath-windows/ovsext/Actions.c
index a93fe03..79e464c 100644
--- a/datapath-windows/ovsext/Actions.c
+++ b/datapath-windows/ovsext/Actions.c
@@ -184,6 +184,9 @@ OvsInitForwardingCtx(OvsForwardingContext *ovsFwdCtx,  }
 
 /*
+ * XXX: When we search for the tunnelVport we also need to specify the
+ * tunnelling protocol or the L4 protocol as key as well, because there 
+ are
+ * different protocols that can use the same destination port.
  * --
  * OvsDetectTunnelRxPkt --
  * Utility function for an RX packet to detect its tunnel type.
@@ -203,16 +206,17 @@ OvsDetectTunnelRxPkt(OvsForwardingContext *ovsFwdCtx,
  * packets only if they are at least VXLAN header size.
  */
 if (!flowKey-ipKey.nwFrag 
-flowKey-ipKey.nwProto == IPPROTO_UDP 
-flowKey-ipKey.l4.tpDst == VXLAN_UDP_PORT_NBO) {
-tunnelVport = ovsFwdCtx-switchContext-vxlanVport;
-ovsActionStats.rxVxlan++;
+flowKey-ipKey.nwProto == IPPROTO_UDP) {
+UINT16 dstPort = htons(flowKey-ipKey.l4.tpDst);
+tunnelVport = OvsFindTunnelVportByDstPort(ovsFwdCtx-switchContext,
+  dstPort);
 }
 
 // We might get tunnel packets even before the tunnel gets initialized.
 if (tunnelVport) {
 ASSERT(ovsFwdCtx-tunnelRxNic == NULL);
 ovsFwdCtx-tunnelRxNic = tunnelVport;
+ovsActionStats.rxVxlan++;
 return TRUE;
 }
 
@@ -1318,6 +1322,7 @@ OvsExecuteSetAction(OvsForwardingContext *ovsFwdCtx,
status = OvsTunnelAttrToIPv4TunnelKey((PNL_ATTR)a, tunKey);
 ASSERT(status == NDIS_STATUS_SUCCESS);
 tunKey.flow_hash = (uint16)(hash ? *hash : OvsHashFlow(key));
+tunKey.dst_port = key-ipKey.l4.tpDst;
 RtlCopyMemory(ovsFwdCtx-tunKey, tunKey, sizeof ovsFwdCtx-tunKey);
 
 break;
diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index cf5e3c4..665aff4 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -366,6 +366,8 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 OvsAllocateMemory(sizeof (LIST_ENTRY) * OVS_MAX_VPORT_ARRAY_SIZE);
 switchContext-pidHashArray = (PLIST_ENTRY)
 OvsAllocateMemory(sizeof(LIST_ENTRY) * OVS_MAX_PID_ARRAY_SIZE);
+switchContext-tunnelVportsArray = (PLIST_ENTRY)
+OvsAllocateMemory(sizeof(LIST_ENTRY) * 
+ OVS_MAX_VPORT_ARRAY_SIZE);
 status = OvsAllocateFlowTable(switchContext-datapath, switchContext);
 
 if (status == NDIS_STATUS_SUCCESS) { @@ -376,7 +378,8 @@ 
OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 switchContext-portNoHashArray == NULL ||
 switchContext-ovsPortNameHashArray == NULL ||
 switchContext-portIdHashArray== NULL ||
-switchContext-pidHashArray == NULL) {
+switchContext-pidHashArray == NULL ||
+switchContext-tunnelVportsArray == NULL) {
 if (switchContext-dispatchLock) {
 NdisFreeRWLock(switchContext-dispatchLock);
 }
@@ -394,6 +397,10 @@ OvsInitSwitchContext(POVS_SWITCH_CONTEXT switchContext)
 OvsFreeMemory(switchContext-pidHashArray);
 }
 
+if (switchContext-tunnelVportsArray) {
+OvsFreeMemory(switchContext-tunnelVportsArray);
+}
+
 OvsDeleteFlowTable(switchContext-datapath);
 OvsCleanupBufferPool(switchContext);
 
@@ -403,12 +410,9

Re: [ovs-dev] OVS Windows MSI installer

[Eitan Eliahu]

Hi Alessandro,
I couldn't find the driver binary as part of the installer. Also, can you 
explain how uninstlaation is done and the case where a driver is already loaded.
Thank you!
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alessandro Pilotti
Sent: Sunday, March 15, 2015 5:33 PM
To: dev@openvswitch.org
Subject: [ovs-dev] OVS Windows MSI installer

Hi guys,



As discussed on IRC, here's the Windows MSI installer project that we’d like to

submit for upstream inclusion:



https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_cloudbase_ovs-2Dwindows-2Dinstallerd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=ssYlF73rox7Qc5jV0EuVGd31T3gjOjJrGilWxhpfCQUe=
 



The included README.rst provides information on how to generate the MSI and

I’ll also be happy to answer any question and review.



In terms of location of the code, we could IMO either include the installer as

a separate project on 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitchd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=j7vZQ7TETCqJcuj8GlWvxlDQj3eaZDowoeugOoF62m4e=
  or as a subfolder under

the 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovsd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=64IUwcJqsI-IqUHcDpffloCd02L8ksBeDjCy2OB09Soe=
  project itself.

My vote is for the first option, but if you guys prefer the second I’ll be

happy to send the full patchset on the ML for review. 



Thanks,



Alessandro



___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=Q9DuaqvXtQI-5E0ozVMau9cLqIUtEVFPGxNzhW5FhtUe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] OVS Windows MSI installer

[Eitan Eliahu]

Alessandro,
Yes, no problem of having the installer separate.  But, I am confused as I see 
the two services are there but the driver and the INF file are not. I believe 
that the MSI file includes everything.
Do you think you can add the OS installer script?
Will we have already built MSI file (x86 and x86-64)?
Thanks,
Eitan 

-Original Message-
From: Alessandro Pilotti [mailto:apilo...@cloudbasesolutions.com] 
Sent: Sunday, March 15, 2015 6:11 PM
To: Eitan Eliahu
Cc: dev@openvswitch.org
Subject: Re: OVS Windows MSI installer

Hi Eitan,



 On 16 Mar 2015, at 02:57, Eitan Eliahu elia...@vmware.com wrote:

 

 

 Hi Alessandro,

 I couldn't find the driver binary as part of the installer. Also, can you 
 explain how uninstlaation is done and the case where a driver is already 
 loaded.



As explained in the README, the actual binaries are produced by the OVS build, 
so it does not make sense to include them in this project, otherwise we’d need 
to commit the binaries at every OVS build.



During uninstall the driver is removed. The general rule for Windows installers 
is that whatever gets installed, must be removed on uninstall (except user 
generated content, e.g. configuration files).

I didn’t include the WMI / PowerShell call for deregistering the extension from 
Hyper-V switches yet, since infamously that operation generated a significant 
amount of crashes until very recently (we still have a pending patch on this 
topic AFAIK).



Here is where the driver is uninstalled: 

https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_cloudbase_ovs-2Dwindows-2Dinstaller_blob_master_ovs-2Dwindows-2Dinstaller_Product.wxs-23L48-2DL49d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=WmmTcAbYsoZf0kKnbhjcKx6VU2dw-WDIcSF-V8Gtd94s=ksjvS04wlV-fvUUTh7wLch51SUmIkKEpjAxRNZtebAwe=
 





Alessandro





 Thank you!

 Eitan

 

 -Original Message-

 From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alessandro Pilotti

 Sent: Sunday, March 15, 2015 5:33 PM

 To: dev@openvswitch.org

 Subject: [ovs-dev] OVS Windows MSI installer

 

 Hi guys,

 

 

 

 As discussed on IRC, here's the Windows MSI installer project that we’d like 
 to

 

 submit for upstream inclusion:

 

 

 

 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_cloudbase_ovs-2Dwindows-2Dinstallerd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=ssYlF73rox7Qc5jV0EuVGd31T3gjOjJrGilWxhpfCQUe=
  

 

 

 

 The included README.rst provides information on how to generate the MSI and

 

 I’ll also be happy to answer any question and review.

 

 

 

 In terms of location of the code, we could IMO either include the installer as

 

 a separate project on 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitchd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=j7vZQ7TETCqJcuj8GlWvxlDQj3eaZDowoeugOoF62m4e=
   or as a subfolder under

 

 the 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovsd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=64IUwcJqsI-IqUHcDpffloCd02L8ksBeDjCy2OB09Soe=
   project itself.

 

 My vote is for the first option, but if you guys prefer the second I’ll be

 

 happy to send the full patchset on the ML for review. 

 

 

 

 Thanks,

 

 

 

 Alessandro

 

 

 

 ___

 dev mailing list

 dev@openvswitch.org

 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=Q9DuaqvXtQI-5E0ozVMau9cLqIUtEVFPGxNzhW5FhtUe=
  



___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] OVS Windows MSI installer

[Eitan Eliahu]

Alessandro, How the MSI solution find files from the OVS build that need to be 
included?
Thanks,
Eitan
 

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Eitan Eliahu
Sent: Sunday, March 15, 2015 6:30 PM
To: Alessandro Pilotti
Cc: dev@openvswitch.org
Subject: Re: [ovs-dev] OVS Windows MSI installer



Alessandro,

Yes, no problem of having the installer separate.  But, I am confused as I see 
the two services are there but the driver and the INF file are not. I believe 
that the MSI file includes everything.

Do you think you can add the OS installer script?

Will we have already built MSI file (x86 and x86-64)?

Thanks,

Eitan 



-Original Message-

From: Alessandro Pilotti [mailto:apilo...@cloudbasesolutions.com] 

Sent: Sunday, March 15, 2015 6:11 PM

To: Eitan Eliahu

Cc: dev@openvswitch.org

Subject: Re: OVS Windows MSI installer



Hi Eitan,







 On 16 Mar 2015, at 02:57, Eitan Eliahu elia...@vmware.com wrote:



 



 



 Hi Alessandro,



 I couldn't find the driver binary as part of the installer. Also, can you 
 explain how uninstlaation is done and the case where a driver is already 
 loaded.







As explained in the README, the actual binaries are produced by the OVS build, 
so it does not make sense to include them in this project, otherwise we’d need 
to commit the binaries at every OVS build.







During uninstall the driver is removed. The general rule for Windows installers 
is that whatever gets installed, must be removed on uninstall (except user 
generated content, e.g. configuration files).



I didn’t include the WMI / PowerShell call for deregistering the extension from 
Hyper-V switches yet, since infamously that operation generated a significant 
amount of crashes until very recently (we still have a pending patch on this 
topic AFAIK).







Here is where the driver is uninstalled: 



https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_cloudbase_ovs-2Dwindows-2Dinstaller_blob_master_ovs-2Dwindows-2Dinstaller_Product.wxs-23L48-2DL49d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=WmmTcAbYsoZf0kKnbhjcKx6VU2dw-WDIcSF-V8Gtd94s=ksjvS04wlV-fvUUTh7wLch51SUmIkKEpjAxRNZtebAwe=
 











Alessandro











 Thank you!



 Eitan



 



 -Original Message-



 From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alessandro Pilotti



 Sent: Sunday, March 15, 2015 5:33 PM



 To: dev@openvswitch.org



 Subject: [ovs-dev] OVS Windows MSI installer



 



 Hi guys,



 



 



 



 As discussed on IRC, here's the Windows MSI installer project that we’d like 
 to



 



 submit for upstream inclusion:



 



 



 



 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_cloudbase_ovs-2Dwindows-2Dinstallerd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=ssYlF73rox7Qc5jV0EuVGd31T3gjOjJrGilWxhpfCQUe=
  



 



 



 



 The included README.rst provides information on how to generate the MSI and



 



 I’ll also be happy to answer any question and review.



 



 



 



 In terms of location of the code, we could IMO either include the installer as



 



 a separate project on 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitchd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=j7vZQ7TETCqJcuj8GlWvxlDQj3eaZDowoeugOoF62m4e=
   or as a subfolder under



 



 the 
 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovsd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=64IUwcJqsI-IqUHcDpffloCd02L8ksBeDjCy2OB09Soe=
   project itself.



 



 My vote is for the first option, but if you guys prefer the second I’ll be



 



 happy to send the full patchset on the ML for review. 



 



 



 



 Thanks,



 



 



 



 Alessandro



 



 



 



 ___



 dev mailing list



 dev@openvswitch.org



 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=utkZq_G_Z0gpFYbAClV8e6AD4sXWL1oC5IbDSJwm-Dks=Q9DuaqvXtQI-5E0ozVMau9cLqIUtEVFPGxNzhW5FhtUe=
  







___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=E8G40dNCxrZ38oxXQmMUMaHyuhmF76rDGyoDLu_wGKEs=ia3tRWFMlcRaiMrcYWBFlaohQfmUS0l1hHhy7z_lNSoe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman

Re: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN tunnel port

[Eitan Eliahu]

Hi Ben,
Sorin plans to address some issues once he is back.
Thanks,
Eitan


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Eitan Eliahu
Sent: Wednesday, January 07, 2015 1:16 PM
To: Sorin Vinturis; dev@openvswitch.org
Subject: Re: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN 
tunnel port



Hi Sorin,

On the default port, it is also created through the user mode using the NL 
interface so if we create all other ports dynamically I would prefer to create 
the default port  using the same method. (unless you see an issue here).



On the concurrency issue: The spec reads as follows:

 Each client session can have only one transaction in progress at a time

If we don't serialize transactions we cannot guaranty that transactions are not 
overlapped . You are right the second transaction will return with IN_PROGRESS 
but is should have really completeed successfully if we were serializing the 
transactions.

I realize that it involves a significant work to serialize the transactions and 
to initiate them from another thread which  runs in a lower IRQL. Let me get 
back to you on that ( iad this code before but removed it). Perhaps we could 
retry if it fails on with FWP_E_TXN_IN_PROGRESS.

Thank you,

Eitan



-Original Message-

From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 

Sent: Wednesday, January 07, 2015 11:59 AM

To: Eitan Eliahu; dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN 
tunnel port



Hi Eitan,



I think we should keep the default port filter creation at the initialization 
phase because most users are using the default port.



Regarding the new filter creation for other tunnel port, the 
OvsTunnelAddFilterEx function does not open the session to the filtering 
engine. The session is already opened when the OvsTunnelAddFilterEx function is 
called. It is opened during tunnel initialization phase at filter attach and is 
closed at filter detach.



The FwpmTransactionBegin0 function begins an explicit transaction within the 
current session. The function cannot be called from within an already started 
transaction. It will fail with FWP_E_TXN_IN_PROGRESS error. So, a subsequent 
call to OvsTunnelAddFilterEx would be unsuccessful if a transaction is in 
progress. I don't think that this scenario is common. I can add an error log in 
case the function fails. What do you think?



Thanks,

Sorin



-Original Message-

From: Eitan Eliahu [mailto:elia...@vmware.com] 

Sent: Wednesday, 7 January, 2015 17:04

To: Sorin Vinturis; dev@openvswitch.org

Subject: RE: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN 
tunnel port



Hi Sorin, thank you for working on this.

Do you thing the default port filter had to be created during initialization ?



On another thing,  since OvsTunnelAddFilterEx is called with IRQL = PASSIVE and 
no lock is handled it may be called from multiple thread contexts. I am not 
sure if opening. Closing and aborting the transaction engine would be thread 
safe. We might need to serialize the calls and to invoke the engine from a 
different thread context running on PASSIVE_IRQL.

Thanks,

Eitan



-Original Message-

From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis

Sent: Wednesday, January 07, 2015 6:24 AM

To: dev@openvswitch.org

Subject: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN tunnel 
port



The kernel datapath supports only port 4789 for VXLAN tunnel creation.

Added support in order to allow for the VXLAN tunnel port to be configurable to 
any port number set by the userspace.



The patch also checks to see if an existing WFP filter, for the necessary UDP 
tunnel port, is already created before adding a new one.

This is a double check, because currently the userspace also verifies this, but 
in my opinion is necessary to avoid future inconsistencies.



Custom VXLAN tunnel port requires the addition of a new WFP filter with the new 
UDP tunnel port. The cleanest way for the creation or removal of the new WFP 
filters is to add them in OvsInitVxlanTunnel or OvsCleanupVxlanTunnel functions.

But the latter functions are running at IRQL = DISPATCH_LEVEL, due to the NDIS 
RW lock acquisition, and all WFP calls must be running at IRQL = PASSIVE_LEVEL. 
This is why the WFP filter creation/removal is made when the spinlock is 
released and the IRQL level is lowered at PASSIVE_LEVEL.



The default VXLAN tunnel port 4789 behaviour is unchanged. The WFP filter for 
UDP port 4789 is added when the extension is enabled and is removed when the 
extension is disabled. No other NL command is able to remove it.



Now it is necessary that OvsTunnelFilterUninitialize function is called after 
OvsClearAllSwitchVports in order to allow for the added WFP filters to be 
removed. OvsTunnelFilterUninitialize function closes the global engine handle 
used by most of the WFP calls, including

Re: [ovs-dev] [PATCH v2] tests: Enable running parallel unit tests for Windows.

[Eitan Eliahu]

Thanks!

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Wednesday, February 11, 2015 6:00 PM
To: dev@openvswitch.org
Cc: Gurucharan Shetty
Subject: [ovs-dev] [PATCH v2] tests: Enable running parallel unit tests for 
Windows.

testsuite uses mkfifo in its job dispatcher that manages parallel unit tests. 
MinGW does not have a mkfifo. This results in unit tests running serially on 
Windows. Right now it takes up to approximately 40 minutes to run all the unit 
tests on Windows.

This commit provides a job dispatcher for MinGW that uses temporary files 
instead of mkfifo to manage parallel jobs.
With this commit, on a Windows machine with 4 cores and with
8 parallel unit test sessions, it takes approximately 8 minutes to finish a 
unit test run.

Signed-off-by: Gurucharan Shetty gshe...@nicira.com
---
 INSTALL.md|2 ++
 tests/automake.mk |7 +++--
 tests/testsuite.patch |   76 +
 3 files changed, 83 insertions(+), 2 deletions(-)  create mode 100644 
tests/testsuite.patch

diff --git a/INSTALL.md b/INSTALL.md
index 94c25f7..273093b 100644
--- a/INSTALL.md
+++ b/INSTALL.md
@@ -41,6 +41,8 @@ you will need the following software:
 
   - Python 2.x, for x = 4.
 
+  - patch (The utility that is used to patch files).
+
 On Linux, you may choose to compile the kernel module that comes with  the 
Open vSwitch distribution or to use the kernel module built into  the Linux 
kernel (version 3.3 or later).  See the [FAQ.md] question diff --git 
a/tests/automake.mk b/tests/automake.mk index 2e8d213..50d8ad2 100644
--- a/tests/automake.mk
+++ b/tests/automake.mk
@@ -6,7 +6,8 @@ EXTRA_DIST += \
$(KMOD_TESTSUITE) \
tests/atlocal.in \
$(srcdir)/package.m4 \
-   $(srcdir)/tests/testsuite
+   $(srcdir)/tests/testsuite \
+   $(srcdir)/tests/testsuite.patch
 
 COMMON_MACROS_AT = \
tests/ovsdb-macros.at \
@@ -87,6 +88,7 @@ KMOD_TESTSUITE_AT = \
tests/kmod-traffic.at
 
 TESTSUITE = $(srcdir)/tests/testsuite
+TESTSUITE_PATCH = $(srcdir)/tests/testsuite.patch
 KMOD_TESTSUITE = $(srcdir)/tests/kmod-testsuite  DISTCLEANFILES += 
tests/atconfig tests/atlocal
 
@@ -196,8 +198,9 @@ clean-local:
test ! -f '$(TESTSUITE)' || $(SHELL) '$(TESTSUITE)' -C tests --clean
 
 AUTOTEST = $(AUTOM4TE) --language=autotest
-$(TESTSUITE): package.m4 $(TESTSUITE_AT) $(COMMON_MACROS_AT)
+$(TESTSUITE): package.m4 $(TESTSUITE_AT) $(COMMON_MACROS_AT) 
+$(TESTSUITE_PATCH)
$(AM_V_GEN)$(AUTOTEST) -I '$(srcdir)' -o $@.tmp $@.at
+   patch -p0 $@.tmp $(TESTSUITE_PATCH)
$(AM_V_at)mv $@.tmp $@
 
 $(KMOD_TESTSUITE): package.m4 $(KMOD_TESTSUITE_AT) $(COMMON_MACROS_AT) diff 
--git a/tests/testsuite.patch b/tests/testsuite.patch new file mode 100644 
index 000..e0c6bb3
--- /dev/null
+++ b/tests/testsuite.patch
@@ -0,0 +1,76 @@
+--- testsuite  2015-02-11 17:19:21.654646439 -0800
 testsuite  2015-02-11 17:15:03.810653032 -0800
+@@ -4669,6 +4669,73 @@
+   fi
+   exec 6-
+   wait
++elif test $at_jobs -ne 1 
++ test $IS_WIN32 = yes; then
++  # FIFO job dispatcher.
++  trap 'at_pids=
++  for at_pid in `jobs -p`; do
++at_pids=$at_pids $at_job_group$at_pid
++  done
++  if test -n $at_pids; then
++at_sig=TSTP
++test ${TMOUT+set} = set  at_sig=STOP
++kill -$at_sig $at_pids 2/dev/null
++  fi
++  kill -STOP $$
++  test -z $at_pids || kill -CONT $at_pids 2/dev/null' TSTP
++
++  echo
++  # Turn jobs into a list of numbers, starting from 1.
++  running_jobs=`pwd`/tests/jobdispatcher
++  mkdir -p $running_jobs
++  at_joblist=`$as_echo $at_groups | sed -n 1,${at_jobs}p`
++
++  set X $at_joblist
++  shift
++  for at_group in $at_groups; do
++$at_job_control_on 2/dev/null
++(
++  # Start one test group.
++  $at_job_control_off
++  touch $running_jobs/$at_group
++  trap 'set +x; set +e
++  trap  PIPE
++  echo stop  $at_stop_file
++  rm -f $running_jobs/$at_group
++  as_fn_exit 141' PIPE
++  at_fn_group_prepare
++  if cd $at_group_dir 
++   at_fn_test $at_group 
++   . $at_test_source
++  then :; else
++  { $as_echo $as_me:${as_lineno-$LINENO}: WARNING: unable to parse 
++test group: $at_group 5 $as_echo $as_me: WARNING: unable to parse test 
group: $at_group 2;}
++  at_failed=:
++  fi
++  rm -f $running_jobs/$at_group
++  at_fn_group_postprocess
++) 
++$at_job_control_off
++shift # Consume one token.
++if test $# -gt 0; then :; else
++  while [ `ls -l $running_jobs 2/dev/null | wc -l` -gt $at_jobs ]; do
++sleep 0.1
++  done
++  set x $*
++fi
++test -f $at_stop_file  break
++  done
++  # Read back the remaining ($at_jobs - 1) tokens.
++  set X $at_joblist
++  shift
++  if test $# -gt 0; then
++shift
++while [ `ls -l

Re: [ovs-dev] [PATCH] appveyor: Build windows kernel datapath.

[Eitan Eliahu]

Guru, driver build is very short (less than a min), It uses precompiled 
headers. Let's have the debug build until we will have it as an option.

For user mode to take advantage on precompiled header, there is work to be dome 
in arranging the header files.

Thank you for taking the time!
Eitan  

-Original Message-
From: Gurucharan Shetty [mailto:shet...@nicira.com] 
Sent: Friday, February 06, 2015 7:39 AM
To: Eitan Eliahu
Cc: Nithin Raju; dev@openvswitch.org; Gurucharan Shetty
Subject: Re: [ovs-dev] [PATCH] appveyor: Build windows kernel datapath.

 Would be nice to have an option for Release or debug build. User mode code 
 too.
Right now, we are not deploying the created executables anywhere.
Appveyor does have that option. We can probably enhance it once we have a 
in-tree windows installer.

On the release/debug question, the only option is to do both - which will 
increase the build time. Increase of build time means that pre-checkin check 
through appveyor will be slow and people may not actually use it.

On the question of the pre-compiled headers, it is likely not going to help the 
userspace build because it is done one file at a time. I will nevertheless 
investigate.


 -Original Message-
 From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin 
 Raju
 Sent: Thursday, February 05, 2015 10:39 PM
 To: Gurucharan Shetty
 Cc: dev@openvswitch.org; Gurucharan Shetty
 Subject: Re: [ovs-dev] [PATCH] appveyor: Build windows kernel datapath.

 +- C:\MinGW\msys\1.0\bin\bash -lc cd /c/openvswitch  ./configure 
 CC=build-aux/cccl LD=\`which link`\ LIBS=-lws2_32 
 --with-pthread=C:/pthreads-win32/Pre-built.2 --with-openssl=C:/OpenSSL-Win32 
 --with-vstudioddk=\Win8.1 Release\”



 Thanks for the patch.



 Only comment I had was to was to prefer to use Win8.1 Debug” rather than 
 “Win8.1 Release”. It doesn’t make any practical difference though, but it 
 might be best to do “Debug” builds rather than “Release” in a build-test 
 setup.



 Acked-by: Nithin Raju nit...@vmware.com 
 ___
 dev mailing list
 dev@openvswitch.org
 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=MQliY0lervWD1h0IE-vbsFegcFKLa9zmypMkSUENsKUs=xwhEJXSZhbRRLtHYTw6veobMVUDNCK01EUqRgvoeOlwe=
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] appveyor: Build windows kernel datapath.

[Eitan Eliahu]

Would be nice to have an option for Release or debug build. User mode code too.

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Thursday, February 05, 2015 10:39 PM
To: Gurucharan Shetty
Cc: dev@openvswitch.org; Gurucharan Shetty
Subject: Re: [ovs-dev] [PATCH] appveyor: Build windows kernel datapath.

 +- C:\MinGW\msys\1.0\bin\bash -lc cd /c/openvswitch  ./configure 
 CC=build-aux/cccl LD=\`which link`\ LIBS=-lws2_32 
 --with-pthread=C:/pthreads-win32/Pre-built.2 --with-openssl=C:/OpenSSL-Win32 
 --with-vstudioddk=\Win8.1 Release\”



Thanks for the patch.



Only comment I had was to was to prefer to use Win8.1 Debug” rather than 
“Win8.1 Release”. It doesn’t make any practical difference though, but it might 
be best to do “Debug” builds rather than “Release” in a build-test setup.



Acked-by: Nithin Raju nit...@vmware.com 
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=MQliY0lervWD1h0IE-vbsFegcFKLa9zmypMkSUENsKUs=xwhEJXSZhbRRLtHYTw6veobMVUDNCK01EUqRgvoeOlwe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 2/2] appveyor: Provide a autobuild service for Windows.

[Eitan Eliahu]

Testing on a single CPU machine is problematic (even if it would have run 
fast), but the build itself  has its own advantage. The driver build should be 
much quicker.

Do we use precompiled header option?
https://msdn.microsoft.com/en-us/library/2yzw0wyd.aspx
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Thursday, February 05, 2015 1:27 PM
To: Ben Pfaff
Cc: dev; Gurucharan Shetty
Subject: Re: [ovs-dev] [PATCH 2/2] appveyor: Provide a autobuild service for 
Windows.


 Oh!

 Do you have an idea of what takes up the bulk of the time for the OVS 
 unit tests in Windows?
I think the main reasons are:
* Creation of new processes is slow on Windows. So some unit tests that run 
hundreds of processes causes the unit tests to go very slow (for e.g., bfd and 
cfm).
* I think 'configure' is also slow on Windows for the same reason as above.
* autoconf tests on mingw does not have the infrastructure to be executed in 
parallel. (I think it is because of no mknod. Probably it can be worked around, 
but the fix will likely be needed upstream in the code that generates testsuite)
* Appveyor itself provides only one CPU. So I cannot run 'make -jx'
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=GX0hr7CUW9JW_B5yLYnNDaWDoO-5-MOBGFO8vOZDujUs=Cw_TEP7eHYIGXE0sNkC55mer7x-Wa10Uxrm9m0XrVC0e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH 2/2] appveyor: Provide a autobuild service for Windows.

[Eitan Eliahu]

Looks like user mode does not use pre-compiled headers, but the driver does.


-Original Message-
From: Eitan Eliahu 
Sent: Thursday, February 05, 2015 1:39 PM
To: 'Gurucharan Shetty'; Ben Pfaff
Cc: dev; Gurucharan Shetty
Subject: RE: [ovs-dev] [PATCH 2/2] appveyor: Provide a autobuild service for 
Windows.

Testing on a single CPU machine is problematic (even if it would have run 
fast), but the build itself  has its own advantage. The driver build should be 
much quicker.

Do we use precompiled header option?
https://msdn.microsoft.com/en-us/library/2yzw0wyd.aspx
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Gurucharan Shetty
Sent: Thursday, February 05, 2015 1:27 PM
To: Ben Pfaff
Cc: dev; Gurucharan Shetty
Subject: Re: [ovs-dev] [PATCH 2/2] appveyor: Provide a autobuild service for 
Windows.


 Oh!

 Do you have an idea of what takes up the bulk of the time for the OVS 
 unit tests in Windows?
I think the main reasons are:
* Creation of new processes is slow on Windows. So some unit tests that run 
hundreds of processes causes the unit tests to go very slow (for e.g., bfd and 
cfm).
* I think 'configure' is also slow on Windows for the same reason as above.
* autoconf tests on mingw does not have the infrastructure to be executed in 
parallel. (I think it is because of no mknod. Probably it can be worked around, 
but the fix will likely be needed upstream in the code that generates testsuite)
* Appveyor itself provides only one CPU. So I cannot run 'make -jx'
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=GX0hr7CUW9JW_B5yLYnNDaWDoO-5-MOBGFO8vOZDujUs=Cw_TEP7eHYIGXE0sNkC55mer7x-Wa10Uxrm9m0XrVC0e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: accommodate to UFID changes

[Eitan Eliahu]

Alin, looks good only minor thing : 
+rc = STATUS_SUCCESS
Is redundant
Thanks,
Eitan

Acked-by: Eitan Eliahu elia...@vmware.com

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, February 03, 2015 11:25 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: accommodate to UFID changes

Current flow commands: new, set, get, del need to respond with a NETLINK error 
in the case OVS_FLOW_ATTR_KEY is missing.

OVS_FLOW_ATTR_KEY is now an optional attribute.

Also add OVS_FLOW_ATTR_UFID attribute  to the kernel for further use.

Signed-off-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
---
 datapath-windows/ovsext/Flow.c | 26 +++---  
datapath-windows/ovsext/Flow.h |  6 ++
 2 files changed, 29 insertions(+), 3 deletions(-)

diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c 
index d3de8cc..b849666 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -95,7 +95,7 @@ static NTSTATUS OvsDoDumpFlows(OvsFlowDumpInput *dumpInput,
 
 /* For Parsing attributes in FLOW_* commands */  const NL_POLICY 
nlFlowPolicy[] = {
-[OVS_FLOW_ATTR_KEY] = {.type = NL_A_NESTED, .optional = FALSE},
+[OVS_FLOW_ATTR_KEY] = {.type = NL_A_NESTED, .optional = TRUE},
 [OVS_FLOW_ATTR_MASK] = {.type = NL_A_NESTED, .optional = TRUE},
 [OVS_FLOW_ATTR_ACTIONS] = {.type = NL_A_NESTED, .optional = TRUE},
 [OVS_FLOW_ATTR_STATS] = {.type = NL_A_UNSPEC, @@ -103,7 +103,9 @@ const 
NL_POLICY nlFlowPolicy[] = {
  .maxLen = sizeof(struct ovs_flow_stats),
  .optional = TRUE},
 [OVS_FLOW_ATTR_TCP_FLAGS] = {NL_A_U8, .optional = TRUE},
-[OVS_FLOW_ATTR_USED] = {NL_A_U64, .optional = TRUE}
+[OVS_FLOW_ATTR_USED] = {NL_A_U64, .optional = TRUE},
+[OVS_FLOW_ATTR_UFID] = { .type = NL_A_UNSPEC, .optional = TRUE,
+ .minLen = sizeof(OvsU128) }
 };
 
 /* For Parsing nested OVS_FLOW_ATTR_KEY attributes.
@@ -308,6 +310,11 @@ OvsFlowNlCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
goto done;
 }
 
+if (nlAttrs[OVS_FLOW_ATTR_KEY] == NULL) {
+nlError = NL_ERROR_INVAL;
+goto done;
+}
+
 if ((rc = _MapNlToFlowPut(msgIn, nlAttrs[OVS_FLOW_ATTR_KEY],
  nlAttrs[OVS_FLOW_ATTR_ACTIONS], nlAttrs[OVS_FLOW_ATTR_CLEAR],
  mappedFlow))
@@ -420,7 +427,7 @@ _FlowNlGetCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 PNL_MSG_HDR nlMsgOutHdr = NULL;
 UINT32 attrOffset = NLMSG_HDRLEN + GENL_HDRLEN + OVS_HDRLEN;
 PNL_ATTR nlAttrs[__OVS_FLOW_ATTR_MAX];
-
+NL_ERROR nlError = NL_ERROR_SUCCESS;
 OvsFlowGetInput getInput;
 OvsFlowGetOutput getOutput;
 NL_BUFFER nlBuf;
@@ -455,6 +462,11 @@ _FlowNlGetCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 goto done;
 }
 
+if (nlAttrs[OVS_FLOW_ATTR_KEY] == NULL) {
+nlError = NL_ERROR_INVAL;
+goto done;
+}
+
 keyAttrOffset = (UINT32)((PCHAR) nlAttrs[OVS_FLOW_ATTR_KEY] -
 (PCHAR)nlMsgHdr);
 
@@ -532,6 +544,14 @@ _FlowNlGetCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 *replyLen += NlMsgSize(nlMsgOutHdr);
 
 done:
+if (nlError != NL_ERROR_SUCCESS) {
+POVS_MESSAGE_ERROR msgError = (POVS_MESSAGE_ERROR)
+usrParamsCtx-outputBuffer;
+NlBuildErrorMsg(msgIn, msgError, nlError);
+*replyLen = msgError-nlMsg.nlmsgLen;
+rc = STATUS_SUCCESS;
+}
+
 return rc;
 }
 
diff --git a/datapath-windows/ovsext/Flow.h b/datapath-windows/ovsext/Flow.h 
index e817bcf..02a72de 100644
--- a/datapath-windows/ovsext/Flow.h
+++ b/datapath-windows/ovsext/Flow.h
@@ -37,6 +37,12 @@ typedef struct _OvsFlow {
 NL_ATTR actions[1];
 } OvsFlow;
 
+typedef union _OvsU128{
+uint32_t u32[4];
+struct {
+uint64_t lo, hi;
+} u64;
+} OvsU128;
 
 typedef struct _OvsLayers {
 UINT32 l3Ofs; // IPv4, IPv6, ARP, or other L3 header.
--
1.9.5.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=h-xnHaFVXtokcPzoZ2splmejx62S-xmfF3kpRTH09E4s=jDEwGj_nNoRyr5TpRU2IUyAXEIYfICDpxyn7X_UJVJMe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN tunnel port

[Eitan Eliahu]

Hi Sorin,
Thank you for addressing these issues. Please find some minor comments below.
The issue I see now is that the IOCTL completion is not synchronized with the 
engine execution and if the transaction fails it is not propagated to user mode 
. I realize that the thread exists when the transaction fails but we might just 
fail the transaction.
I think once we have the IOCTL synced we can go ahead with this challenging 
change.
Eitan

-Original Message-
From: Eitan Eliahu 
Sent: Tuesday, January 20, 2015 9:07 AM
To: 'Sorin Vinturis'
Subject: RE: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN 
tunnel port


Hi Sorin, once you comment out the call to OvsDetectTunnelRxPkt() the driver 
should ignore the received packet and to indicate it up to the host stack. This 
means that any packet (not just fragmented) will be indicated to the host.
We can deal with adding the opening of the socket in vswitchd later. For 
purpose of testing this code you can run any other application which open a 
VXLAN UDP port (concurrently with vswitchd.exe) .
Thanks,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com]
Sent: Monday, January 19, 2015 7:04 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN 
tunnel port

Hi Eitan,

I have not manage to successfully shortcut fast path. I have changed the 
netlink library to create the necessary socket in order for the packets to go 
through the filter, but the resulted vswitchd binary hangs when I execute it. 
Thus could not perform the test that shortcuts the fast path.

For testing the patch I have created a script that adds vxlan tunnels with 
different destination ports. Each add vxlan tunnel command, that triggers a 
OVS_VPORT_CMD_NEW command, is followed by a set command, that triggers first a 
OVS_VPORT_CMD_NEW command for the new vport with the new destination port, and 
an OVS_VPORT_CMD_DELETE command for deleting the first vport. 

Below is an excerpt of the script I used:

start .\rundb.bat
Start-Sleep -s 2
start .\runvswitch.bat
.\ovs-vsctl.exe --db=tcp:127.0.0.1:8777 add-br br0 .\ovs-vsctl.exe 
--db=tcp:127.0.0.1:8777 add-port br0 vxlan1 -- set interface vxxlan1 type=vxlan 
options:remote_ip=10.1.1.112 options:dst_port=5000 .\ovs-vsctl.exe 
--db=tcp:127.0.0.1:8777 -- set interface vxlan1 type=vxlan 
options:remote_ip=10.1.1.112 options:dst_port=5001

--
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com]
Sent: Friday, 16 January, 2015 17:38
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN 
tunnel port

Hi Sorin,
Thank you for addressing the issues, specifically,  executing the engine from a 
thread running in lower IRQL.
Can you please add some comments on the testing you performed? Where you able 
to shortcut the fast path and to open the VXLAN UDP port in user mode so 
packets will go through the filter?

Eitan 

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, January 15, 2015 1:11 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN tunnel 
port

The kernel datapath supports only port 4789 for VXLAN tunnel creation.
Added support in order to allow for the VXLAN tunnel port to be configurable to 
any port number set by the userspace.

The patch also checks to see if an existing WFP filter, for the necessary UDP 
tunnel port, is already created before adding a new one.
This is a double check, because currently the userspace also verifies this, but 
it is necessary to avoid future issues.

Custom VXLAN tunnel port requires the addition of a new WFP filter with the new 
UDP tunnel port. The creation of a new WFP filter is triggered in 
OvsInitVxlanTunnel function and the removal of the WFP filter in 
OvsCleanupVxlanTunnel function.
But the latter functions are running at IRQL = DISPATCH_LEVEL, due to the NDIS 
RW lock acquisition, and all WFP calls must be running at IRQL = PASSIVE_LEVEL. 
This is why I have created a system thread which records all filter 
addition/removal requests into a list for later processing by the system 
thread. The ThreadStart routine processes all received requests at IRQL = 
PASSIVE_LEVEL, which is the required IRQL for the necessary WFP calls for 
adding/removal of the WFP filters.

The WFP filter for the default VXLAN port 4789 is not added anymore at filter 
attach. All WFP filters for the tunnel ports are added when the tunnel ports 
are initialized and are removed at cleanup.

It is necessary that OvsTunnelFilterUninitialize function is called after 
OvsClearAllSwitchVports in order to allow for the added WFP filters to be 
removed. OvsTunnelFilterUninitialize function closes the global engine handle 
used by most of the WFP calls, including filter removal.

Signed-off-by: Sorin Vinturis

Re: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN tunnel port

[Eitan Eliahu]

Hi Sorin,
Thank you for addressing the issues, specifically,  executing the engine from a 
thread running in lower IRQL.
Can you please add some comments on the testing you performed? Where you able 
to shortcut the fast path and to open the VXLAN UDP port in user mode so 
packets will go through the filter?

Eitan 

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Thursday, January 15, 2015 1:11 PM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Support for custom VXLAN tunnel 
port

The kernel datapath supports only port 4789 for VXLAN tunnel creation.
Added support in order to allow for the VXLAN tunnel port to be configurable to 
any port number set by the userspace.

The patch also checks to see if an existing WFP filter, for the necessary UDP 
tunnel port, is already created before adding a new one.
This is a double check, because currently the userspace also verifies this, but 
it is necessary to avoid future issues.

Custom VXLAN tunnel port requires the addition of a new WFP filter with the new 
UDP tunnel port. The creation of a new WFP filter is triggered in 
OvsInitVxlanTunnel function and the removal of the WFP filter in 
OvsCleanupVxlanTunnel function.
But the latter functions are running at IRQL = DISPATCH_LEVEL, due to the NDIS 
RW lock acquisition, and all WFP calls must be running at IRQL = PASSIVE_LEVEL. 
This is why I have created a system thread which records all filter 
addition/removal requests into a list for later processing by the system 
thread. The ThreadStart routine processes all received requests at IRQL = 
PASSIVE_LEVEL, which is the required IRQL for the necessary WFP calls for 
adding/removal of the WFP filters.

The WFP filter for the default VXLAN port 4789 is not added anymore at filter 
attach. All WFP filters for the tunnel ports are added when the tunnel ports 
are initialized and are removed at cleanup.

It is necessary that OvsTunnelFilterUninitialize function is called after 
OvsClearAllSwitchVports in order to allow for the added WFP filters to be 
removed. OvsTunnelFilterUninitialize function closes the global engine handle 
used by most of the WFP calls, including filter removal.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_66d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=VmREdmxxsERc3TdRTLuqj-tDd42IJUEuvcXW-vWM5Mcs=M_OrcxSLPFtZNI5KVfqaF7oSnM_vU3_Sj2E0GAESvqce=
---
 datapath-windows/ovsext/Switch.c   |   2 +-
 datapath-windows/ovsext/Tunnel.h   |   6 -
 datapath-windows/ovsext/TunnelFilter.c | 484 -
 datapath-windows/ovsext/TunnelIntf.h   |   4 +
 datapath-windows/ovsext/Vport.c|  13 +-
 datapath-windows/ovsext/Vxlan.c|  53 +++-
 datapath-windows/ovsext/Vxlan.h|   2 +-
 7 files changed, 473 insertions(+), 91 deletions(-)

diff --git a/datapath-windows/ovsext/Switch.c b/datapath-windows/ovsext/Switch.c
index a228d8e..cf5e3c4 100644
--- a/datapath-windows/ovsext/Switch.c
+++ b/datapath-windows/ovsext/Switch.c
@@ -261,8 +261,8 @@ OvsDeleteSwitch(POVS_SWITCH_CONTEXT switchContext)
 if (switchContext)
 {
 dpNo = switchContext-dpNo;
-OvsTunnelFilterUninitialize(gOvsExtDriverObject);
 OvsClearAllSwitchVports(switchContext);
+OvsTunnelFilterUninitialize(gOvsExtDriverObject);
 OvsUninitSwitchContext(switchContext);
 OvsFreeMemory(switchContext);
 }
diff --git a/datapath-windows/ovsext/Tunnel.h b/datapath-windows/ovsext/Tunnel.h
index 2978bb3..2c45e35 100644
--- a/datapath-windows/ovsext/Tunnel.h
+++ b/datapath-windows/ovsext/Tunnel.h
@@ -32,12 +32,6 @@ typedef struct OVS_TUNNEL_PENDED_PACKET_
FWPS_CLASSIFY_OUT *classifyOut;
 } OVS_TUNNEL_PENDED_PACKET;
 
-/* Shared global data. */
-
-extern UINT16 configNewDestPort;
-
-extern UINT32 gCalloutIdV4;
-
 //
 // Shared function prototypes
 //
diff --git a/datapath-windows/ovsext/TunnelFilter.c 
b/datapath-windows/ovsext/TunnelFilter.c
index e0adc37..b4289e6 100644
--- a/datapath-windows/ovsext/TunnelFilter.c
+++ b/datapath-windows/ovsext/TunnelFilter.c
@@ -63,9 +63,6 @@
 /* The session name isn't required but it's useful for diagnostics. */
 #define OVS_TUNNEL_SESSION_NAME LOVS tunnel session
 
-/* Configurable parameters (addresses and ports are in host order) */
-UINT16   configNewDestPort = VXLAN_UDP_PORT;
-
 /*
  * Callout and sublayer GUIDs
  */
@@ -105,14 +102,57 @@ DEFINE_GUID(
 0xa5, 0x36, 0x1e, 0xed, 0xb9, 0xe9, 0xba, 0x6a
 );
 
+KSTART_ROUTINE  OvsTunnelFilterThreadProc;
+NTSTATUSOvsTunnelFilterThreadInit();
+VOIDOvsTunnelFilterThreadStop();
+VOIDOvsTunnelFilterThreadCleanup();
+
 /*
- * Callout driver global 

Re: [ovs-dev] OVN architecture

[Eitan Eliahu]

No problem, my point is that it is doable to have a dynamic linking model. 
Unless there is a custom installer and mechanism to enforce versioning  it is 
better to package the static libraries with the executable.
Thanks,
Eitan

-Original Message-
From: Gurucharan Shetty [mailto:shet...@nicira.com] 
Sent: Thursday, January 15, 2015 9:07 AM
To: Eitan Eliahu
Cc: Thomas Graf; dev; YAMAMOTO Takashi
Subject: Re: [ovs-dev] OVN architecture

On Thu, Jan 15, 2015 at 8:55 AM, Eitan Eliahu elia...@vmware.com wrote:
 Shared libraries on Windows could be implemented as DLLs (Dynamic Link 
 Libraries) with some custom support for versioning.
 This means also that updates of new versions of the DLLs should  controlled 
 by a custom installer.
I did not intend to mean that Windows does not support shared libraries. Sorry 
about that. What I was saying is that autoconf/automake in openvswitch does not 
allow shared library building on Windows right now (it does on Linux).




 -Original Message-
 From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Thomas 
 Graf
 Sent: Thursday, January 15, 2015 7:19 AM
 To: Gurucharan Shetty
 Cc: dev; YAMAMOTO Takashi
 Subject: Re: [ovs-dev] OVN architecture

 On 01/15/15 at 07:04am, Gurucharan Shetty wrote:
 On Thu, Jan 15, 2015 at 4:06 AM, Thomas Graf tg...@noironetworks.com wrote:
  On 01/15/15 at 03:36pm, YAMAMOTO Takashi wrote:
   On Thu, Jan 15, 2015 at 10:38:45AM +0900, YAMAMOTO Takashi wrote:
ovn-controller
--
  
   neutron ofagent agent has a similar design to ovn-controller.
   you might be able to reuse at least some of code if python+ryu 
   is acceptable.
  
   https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com
   _ 
   openstack_neutron_tree_stable_juno_neutron_plugins_ofagentd=Aw
   I 
   GaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6Ex
   L
   XY798tmo3LJ4e3geGYp56lkcH-5cLCYm=BOjy7uXWvQ7_zay-Yzf-Z2zPIMCBc
   5 
   NiTCjTNxYAdYUs=bQdLaLSSbfeFOSxAm6ke5xfCzrQA39wkiG0g_ZcR04Ue=
  
   I didn't know that there was an existing local controller.  I'll 
   learn something about the design.
 
  similar:
 
  - it's a local OpenFlow controller running on each nodes
 
  - it has ARP suppression feature implemented with packet-ins
(called local arp responder there)
 
  different:
 
  - ofagent doesn't have a layer equivalent to OVN database.
it obtains the necessary info from its CMS (neutron) directly
 
  Another local controller to look at is the OpFlex OVS agent which 
  runs locally as well and uses libopenvswitch and libofproto to talk 
  OF/OVSDB to OVS. (I think OVN should be built on top of the now 
  exposed shared libraries as well).
 A point to be noted while taking a decision is that there is 
 currently no capability to build shared libraries on Windows platform.

 Good point. I don't know enough about packaging on Windows to know 
 whether fall back to static linking for Windows would impose any 
 packaging/distribution problems. On Linux, the shared library 
 dependency would simplify the packaging in Linux distributions. (A bug 
 in OF code must be pushed to both OVS and OVN packages.) 
 ___
 dev mailing list
 dev@openvswitch.org
 https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_ma
 ilman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-
 uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=BOjy7uXWvQ7_zay-Yz
 f-Z2zPIMCBc5NiTCjTNxYAdYUs=UyabyhKPWJeZTHnkqZ2RX5jEpNn_edhyd3ljIMS0sN
 4e=
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] OVN architecture

[Eitan Eliahu]

Shared libraries on Windows could be implemented as DLLs (Dynamic Link 
Libraries) with some custom support for versioning.
This means also that updates of new versions of the DLLs should  controlled by 
a custom installer.


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Thomas Graf
Sent: Thursday, January 15, 2015 7:19 AM
To: Gurucharan Shetty
Cc: dev; YAMAMOTO Takashi
Subject: Re: [ovs-dev] OVN architecture

On 01/15/15 at 07:04am, Gurucharan Shetty wrote:
 On Thu, Jan 15, 2015 at 4:06 AM, Thomas Graf tg...@noironetworks.com wrote:
  On 01/15/15 at 03:36pm, YAMAMOTO Takashi wrote:
   On Thu, Jan 15, 2015 at 10:38:45AM +0900, YAMAMOTO Takashi wrote:
ovn-controller
--
  
   neutron ofagent agent has a similar design to ovn-controller.
   you might be able to reuse at least some of code if python+ryu 
   is acceptable.
  
   https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_
   openstack_neutron_tree_stable_juno_neutron_plugins_ofagentd=AwI
   GaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExL
   XY798tmo3LJ4e3geGYp56lkcH-5cLCYm=BOjy7uXWvQ7_zay-Yzf-Z2zPIMCBc5
   NiTCjTNxYAdYUs=bQdLaLSSbfeFOSxAm6ke5xfCzrQA39wkiG0g_ZcR04Ue=
  
   I didn't know that there was an existing local controller.  I'll 
   learn something about the design.
 
  similar:
 
  - it's a local OpenFlow controller running on each nodes
 
  - it has ARP suppression feature implemented with packet-ins
(called local arp responder there)
 
  different:
 
  - ofagent doesn't have a layer equivalent to OVN database.
it obtains the necessary info from its CMS (neutron) directly
 
  Another local controller to look at is the OpFlex OVS agent which 
  runs locally as well and uses libopenvswitch and libofproto to talk 
  OF/OVSDB to OVS. (I think OVN should be built on top of the now 
  exposed shared libraries as well).
 A point to be noted while taking a decision is that there is currently 
 no capability to build shared libraries on Windows platform.

Good point. I don't know enough about packaging on Windows to know whether fall 
back to static linking for Windows would impose any packaging/distribution 
problems. On Linux, the shared library dependency would simplify the packaging 
in Linux distributions. (A bug in OF code must be pushed to both OVS and OVN 
packages.) ___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=BOjy7uXWvQ7_zay-Yzf-Z2zPIMCBc5NiTCjTNxYAdYUs=UyabyhKPWJeZTHnkqZ2RX5jEpNn_edhyd3ljIMS0sN4e=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN tunnel port

[Eitan Eliahu]

Hi Sorin,
On the default port, it is also created through the user mode using the NL 
interface so if we create all other ports dynamically I would prefer to create 
the default port  using the same method. (unless you see an issue here).

On the concurrency issue: The spec reads as follows:
 Each client session can have only one transaction in progress at a time
If we don't serialize transactions we cannot guaranty that transactions are not 
overlapped . You are right the second transaction will return with IN_PROGRESS 
but is should have really completeed successfully if we were serializing the 
transactions.
I realize that it involves a significant work to serialize the transactions and 
to initiate them from another thread which  runs in a lower IRQL. Let me get 
back to you on that ( iad this code before but removed it). Perhaps we could 
retry if it fails on with FWP_E_TXN_IN_PROGRESS.
Thank you,
Eitan

-Original Message-
From: Sorin Vinturis [mailto:svintu...@cloudbasesolutions.com] 
Sent: Wednesday, January 07, 2015 11:59 AM
To: Eitan Eliahu; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN 
tunnel port

Hi Eitan,

I think we should keep the default port filter creation at the initialization 
phase because most users are using the default port.

Regarding the new filter creation for other tunnel port, the 
OvsTunnelAddFilterEx function does not open the session to the filtering 
engine. The session is already opened when the OvsTunnelAddFilterEx function is 
called. It is opened during tunnel initialization phase at filter attach and is 
closed at filter detach.

The FwpmTransactionBegin0 function begins an explicit transaction within the 
current session. The function cannot be called from within an already started 
transaction. It will fail with FWP_E_TXN_IN_PROGRESS error. So, a subsequent 
call to OvsTunnelAddFilterEx would be unsuccessful if a transaction is in 
progress. I don't think that this scenario is common. I can add an error log in 
case the function fails. What do you think?

Thanks,
Sorin

-Original Message-
From: Eitan Eliahu [mailto:elia...@vmware.com] 
Sent: Wednesday, 7 January, 2015 17:04
To: Sorin Vinturis; dev@openvswitch.org
Subject: RE: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN 
tunnel port

Hi Sorin, thank you for working on this.
Do you thing the default port filter had to be created during initialization ?

On another thing,  since OvsTunnelAddFilterEx is called with IRQL = PASSIVE and 
no lock is handled it may be called from multiple thread contexts. I am not 
sure if opening. Closing and aborting the transaction engine would be thread 
safe. We might need to serialize the calls and to invoke the engine from a 
different thread context running on PASSIVE_IRQL.
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Wednesday, January 07, 2015 6:24 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Support for custom VXLAN tunnel 
port

The kernel datapath supports only port 4789 for VXLAN tunnel creation.
Added support in order to allow for the VXLAN tunnel port to be configurable to 
any port number set by the userspace.

The patch also checks to see if an existing WFP filter, for the necessary UDP 
tunnel port, is already created before adding a new one.
This is a double check, because currently the userspace also verifies this, but 
in my opinion is necessary to avoid future inconsistencies.

Custom VXLAN tunnel port requires the addition of a new WFP filter with the new 
UDP tunnel port. The cleanest way for the creation or removal of the new WFP 
filters is to add them in OvsInitVxlanTunnel or OvsCleanupVxlanTunnel functions.
But the latter functions are running at IRQL = DISPATCH_LEVEL, due to the NDIS 
RW lock acquisition, and all WFP calls must be running at IRQL = PASSIVE_LEVEL. 
This is why the WFP filter creation/removal is made when the spinlock is 
released and the IRQL level is lowered at PASSIVE_LEVEL.

The default VXLAN tunnel port 4789 behaviour is unchanged. The WFP filter for 
UDP port 4789 is added when the extension is enabled and is removed when the 
extension is disabled. No other NL command is able to remove it.

Now it is necessary that OvsTunnelFilterUninitialize function is called after 
OvsClearAllSwitchVports in order to allow for the added WFP filters to be 
removed. OvsTunnelFilterUninitialize function closes the global engine handle 
used by most of the WFP calls, including filter removal.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Alin Gabriel Serdean aserd...@cloudbasesolutions.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_66d=AwIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=fVUQFtLj8jLS3oqKrhXRCrc7OwBHhpU4eYO-t9Nbm-gs

Re: [ovs-dev] [PATCH 1/2] datapath-windows: refactor BuildReplyMsgFromMsgIn BuildErrorMsg

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks!
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Monday, December 08, 2014 9:43 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 1/2] datapath-windows: refactor 
BuildReplyMsgFromMsgIn  BuildErrorMsg

In this patch, we consolidate code in Netlink.c.

Signed-off-by: Nithin Raju nit...@vmware.com
---
 datapath-windows/ovsext/Datapath.c|2 +-
 datapath-windows/ovsext/Flow.c|4 +-
 datapath-windows/ovsext/Netlink/Netlink.c |   40 -
 datapath-windows/ovsext/Netlink/Netlink.h |6 +---
 datapath-windows/ovsext/User.c|2 +-
 datapath-windows/ovsext/Vport.c   |   28 +---
 6 files changed, 30 insertions(+), 52 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index f6e6e50..af8d818 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -1324,7 +1324,7 @@ cleanup:
 POVS_MESSAGE_ERROR msgError = (POVS_MESSAGE_ERROR)
 usrParamsCtx-outputBuffer;
 
-BuildErrorMsg(msgIn, msgError, nlError);
+NlBuildErrorMsg(msgIn, msgError, nlError);
 *replyLen = msgError-nlMsg.nlmsgLen;
 }
 
diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c 
index f47d469..f50bb39 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -355,7 +355,7 @@ done:
 if (nlError != NL_ERROR_SUCCESS) {
 POVS_MESSAGE_ERROR msgError = (POVS_MESSAGE_ERROR)
usrParamsCtx-outputBuffer;
-BuildErrorMsg(msgIn, msgError, nlError);
+NlBuildErrorMsg(msgIn, msgError, nlError);
 *replyLen = msgError-nlMsg.nlmsgLen;
 rc = STATUS_SUCCESS;
 }
@@ -387,7 +387,7 @@ OvsFlowNlGetCmdHandler(POVS_USER_PARAMS_CONTEXT 
usrParamsCtx,
 (usrParamsCtx-outputBuffer)) {
 POVS_MESSAGE_ERROR msgError = (POVS_MESSAGE_ERROR)
usrParamsCtx-outputBuffer;
-BuildErrorMsg(msgIn, msgError, nlError);
+NlBuildErrorMsg(msgIn, msgError, nlError);
 *replyLen = msgError-nlMsg.nlmsgLen;
 status = STATUS_SUCCESS;
 goto done;
diff --git a/datapath-windows/ovsext/Netlink/Netlink.c 
b/datapath-windows/ovsext/Netlink/Netlink.c
index 7633f2f..0d48d08 100644
--- a/datapath-windows/ovsext/Netlink/Netlink.c
+++ b/datapath-windows/ovsext/Netlink/Netlink.c
@@ -102,41 +102,23 @@ NlFillNlHdr(PNL_BUFFER nlBuf, UINT16 nlmsgType,
 return writeOk ? STATUS_SUCCESS : STATUS_INVALID_BUFFER_SIZE;  }
 
-static VOID
-BuildMsgOut(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut, UINT16 type,
-UINT32 length, UINT16 flags)
-{
-msgOut-nlMsg.nlmsgType = type;
-msgOut-nlMsg.nlmsgFlags = flags;
-msgOut-nlMsg.nlmsgSeq = msgIn-nlMsg.nlmsgSeq;
-msgOut-nlMsg.nlmsgPid = msgIn-nlMsg.nlmsgPid;
-msgOut-nlMsg.nlmsgLen = length;
-
-msgOut-genlMsg.cmd = msgIn-genlMsg.cmd;
-msgOut-genlMsg.version = msgIn-genlMsg.version;
-msgOut-genlMsg.reserved = 0;
-}
-
 /*
- * XXX: should move out these functions to a Netlink.c or to a OvsMessage.c
- * or even make them inlined functions in Datapath.h. Can be done after the
- * first sprint once we have more code to refactor.
+ * 
+ --
+ -
+ * Prepare a 'OVS_MESSAGE_ERROR' message.
+ * 
+ --
+ -
  */
 VOID
-BuildReplyMsgFromMsgIn(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut, UINT16 flags)
+NlBuildErrorMsg(POVS_MESSAGE msgIn, POVS_MESSAGE_ERROR msgError, UINT 
+errorCode)
 {
-BuildMsgOut(msgIn, msgOut, msgIn-nlMsg.nlmsgType, sizeof(OVS_MESSAGE),
-flags);
-}
+NL_BUFFER nlBuffer;
 
-VOID
-BuildErrorMsg(POVS_MESSAGE msgIn, POVS_MESSAGE_ERROR msgOut, UINT errorCode) -{
-BuildMsgOut(msgIn, (POVS_MESSAGE)msgOut, NLMSG_ERROR,
-sizeof(OVS_MESSAGE_ERROR), 0);
+NlBufInit(nlBuffer, (PCHAR)msgError, sizeof *msgError);
+NlFillNlHdr(nlBuffer, NLMSG_ERROR, 0,
+msgIn-nlMsg.nlmsgSeq, msgIn-nlMsg.nlmsgPid);
 
-msgOut-errorMsg.error = errorCode;
-msgOut-errorMsg.nlMsg = msgIn-nlMsg;
+msgError-errorMsg.error = errorCode;
+msgError-errorMsg.nlMsg = msgIn-nlMsg;
+msgError-nlMsg.nlmsgLen = sizeof(OVS_MESSAGE_ERROR);
 }
 
 /*
diff --git a/datapath-windows/ovsext/Netlink/Netlink.h 
b/datapath-windows/ovsext/Netlink/Netlink.h
index 18e40b1..a7a53e0 100644
--- a/datapath-windows/ovsext/Netlink/Netlink.h
+++ b/datapath-windows/ovsext/Netlink/Netlink.h
@@ -94,10 +94,8 @@ NTSTATUS NlFillNlHdr(PNL_BUFFER nlBuf,
  UINT16 nlmsgType, UINT16 nlmsgFlags,
  UINT32 nlmsgSeq, UINT32 nlmsgPid);
 
-VOID BuildReplyMsgFromMsgIn(POVS_MESSAGE msgIn

Re: [ovs-dev] [PATCH 2/2] datapath-windows: return bool from NlFillOvs[Msg/Hdr]

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks for addressing my comment.
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Monday, December 08, 2014 9:43 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 2/2] datapath-windows: return bool from 
NlFillOvs[Msg/Hdr]

Per review comment, in this patch, we update the return values of
NlFillOvsMsg() and NlFillOvsHdr() from NTSTATUS to BOOLEAN to make them 
consistent with the Nl* functions.

Signed-off-by: Nithin Raju nit...@vmware.com
---
 datapath-windows/ovsext/Flow.c|   31 ++--
 datapath-windows/ovsext/Netlink/Netlink.c |8 +++---
 datapath-windows/ovsext/Netlink/Netlink.h |   12 +-
 datapath-windows/ovsext/User.c|   12 +++---
 datapath-windows/ovsext/Vport.c   |   30 ++--
 5 files changed, 53 insertions(+), 40 deletions(-)

diff --git a/datapath-windows/ovsext/Flow.c b/datapath-windows/ovsext/Flow.c 
index f50bb39..93f37a0 100644
--- a/datapath-windows/ovsext/Flow.c
+++ b/datapath-windows/ovsext/Flow.c
@@ -247,6 +247,7 @@ OvsFlowNlCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 UINT32 *replyLen)
 {
 NTSTATUS rc = STATUS_SUCCESS;
+BOOLEAN ok;
 POVS_MESSAGE msgIn = (POVS_MESSAGE)usrParamsCtx-inputBuffer;
 POVS_MESSAGE msgOut = (POVS_MESSAGE)usrParamsCtx-outputBuffer;
 PNL_MSG_HDR nlMsgHdr = (msgIn-nlMsg); @@ -258,7 +259,6 @@ 
OvsFlowNlCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 OvsFlowStats stats;
 struct ovs_flow_stats replyStats;
 NL_ERROR nlError = NL_ERROR_SUCCESS;
-
 NL_BUFFER nlBuf;
 
 RtlZeroMemory(mappedFlow, sizeof(OvsFlowPut)); @@ -294,13 +294,14 @@ 
OvsFlowNlCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
   usrParamsCtx-outputLength);
 
 /* Prepare nl Msg headers */
-rc = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, 0,
+ok = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, 0,
   nlMsgHdr-nlmsgSeq, nlMsgHdr-nlmsgPid,
   genlMsgHdr-cmd, OVS_FLOW_VERSION,
   ovsHdr-dp_ifindex);
-
-if (rc == STATUS_SUCCESS) {
+if (ok) {
 *replyLen = msgOut-nlMsg.nlmsgLen;
+} else {
+rc = STATUS_INVALID_BUFFER_SIZE;
 }
}
 
@@ -330,13 +331,15 @@ OvsFlowNlCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
   usrParamsCtx-outputLength);
 
 /* Prepare nl Msg headers */
-rc = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, 0,
+ok = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, 0,
   nlMsgHdr-nlmsgSeq, nlMsgHdr-nlmsgPid,
   genlMsgHdr-cmd, OVS_FLOW_VERSION,
   ovsHdr-dp_ifindex);
-
-if (rc != STATUS_SUCCESS) {
+if (!ok) {
+rc = STATUS_INVALID_BUFFER_SIZE;
 goto done;
+} else {
+rc = STATUS_SUCCESS;
 }
 
 /* Append OVS_FLOW_ATTR_STATS attribute */ @@ -586,15 +589,20 @@ 
_FlowNlDumpCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 
 /* Done with Dump, send NLMSG_DONE */
 if (!(dumpOutput.n)) {
+BOOLEAN ok;
+
 OVS_LOG_INFO(Dump Done);
 
 nlMsgOutHdr = (PNL_MSG_HDR)(NlBufAt(nlBuf, NlBufSize(nlBuf), 0));
-rc = NlFillNlHdr(nlBuf, NLMSG_DONE, NLM_F_MULTI,
+ok = NlFillNlHdr(nlBuf, NLMSG_DONE, NLM_F_MULTI,
  nlMsgHdr-nlmsgSeq, nlMsgHdr-nlmsgPid);
 
-if (rc != STATUS_SUCCESS) {
+if (!ok) {
+rc = STATUS_INVALID_BUFFER_SIZE;
 OVS_LOG_ERROR(Unable to prepare DUMP_DONE reply.);
 break;
+} else {
+rc = STATUS_SUCCESS;
 }
 
 NlMsgAlignSize(nlMsgOutHdr); @@ -603,17 +611,18 @@ 
_FlowNlDumpCmdHandler(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 FreeUserDumpState(instance);
 break;
 } else {
+BOOLEAN ok;
 
 hdrOffset = NlBufSize(nlBuf);
 nlMsgOutHdr = (PNL_MSG_HDR)(NlBufAt(nlBuf, hdrOffset, 0));
 
 /* Netlink header */
-rc = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, NLM_F_MULTI,
+ok = NlFillOvsMsg(nlBuf, nlMsgHdr-nlmsgType, NLM_F_MULTI,
   nlMsgHdr-nlmsgSeq, nlMsgHdr-nlmsgPid,
   genlMsgHdr-cmd, genlMsgHdr-version,
   ovsHdr-dp_ifindex);
 
-if (rc != STATUS_SUCCESS) {
+if (!ok) {
 /* Reset rc to success so that we can
  * send already added messages to user space. */
 rc = STATUS_SUCCESS;
diff --git a/datapath-windows/ovsext/Netlink/Netlink.c 
b/datapath-windows/ovsext/Netlink/Netlink.c
index 0d48d08..589e3a1 100644
--- a/datapath-windows/ovsext/Netlink/Netlink.c
+++ b

Re: [ovs-dev] [PATCH 1/3] datapath-windows: Move Build*Msg() to Netlink.c

[Eitan Eliahu]

Acked-by: Eitan Eliahu elia...@vmware.com
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Nithin Raju
Sent: Wednesday, December 03, 2014 7:56 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH 1/3] datapath-windows: Move Build*Msg() to Netlink.c

Moving the functions that build netlink messages to Netlink.c from Vport.c

Signed-off-by: Nithin Raju nit...@vmware.com
---
 datapath-windows/ovsext/Datapath.h|   14 
 datapath-windows/ovsext/Netlink/Netlink.c |   37 ++
 datapath-windows/ovsext/Netlink/Netlink.h |   13 
 datapath-windows/ovsext/Vport.c   |   47 +++-
 4 files changed, 55 insertions(+), 56 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.h 
b/datapath-windows/ovsext/Datapath.h
index e436699..2e3bac5 100644
--- a/datapath-windows/ovsext/Datapath.h
+++ b/datapath-windows/ovsext/Datapath.h
@@ -25,14 +25,6 @@
 #define __DATAPATH_H_ 1
 
 /*
- * Structure of an error message sent as a reply from kernel.
- */
-typedef struct _OVS_MESSAGE_ERROR {
-NL_MSG_HDR nlMsg;
-NL_MSG_ERR errorMsg;
-} OVS_MESSAGE_ERROR, *POVS_MESSAGE_ERROR;
-
-/*
  * Device operations to tag netlink commands with. This is a bitmask since it
  * is possible that a particular command can be invoked via different device
  * operations.
@@ -98,12 +90,6 @@ NTSTATUS OvsCompleteIrpRequest(PIRP irp, ULONG_PTR infoPtr, 
NTSTATUS status);  VOID OvsAcquireCtrlLock();  VOID OvsReleaseCtrlLock();
 
-/* XXX: Move this to netlink.[ch] eventually. */ -VOID 
BuildReplyMsgFromMsgIn(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut,
-UINT16 flags);
-VOID BuildErrorMsg(POVS_MESSAGE msgIn, POVS_MESSAGE_ERROR msgOut,
-   UINT errorCode);
-
 /*
  * Utility structure and functions to collect in one place all the parameters
  * passed during a call from userspace.
diff --git a/datapath-windows/ovsext/Netlink/Netlink.c 
b/datapath-windows/ovsext/Netlink/Netlink.c
index ae10a87..7633f2f 100644
--- a/datapath-windows/ovsext/Netlink/Netlink.c
+++ b/datapath-windows/ovsext/Netlink/Netlink.c
@@ -102,6 +102,43 @@ NlFillNlHdr(PNL_BUFFER nlBuf, UINT16 nlmsgType,
 return writeOk ? STATUS_SUCCESS : STATUS_INVALID_BUFFER_SIZE;  }
 
+static VOID
+BuildMsgOut(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut, UINT16 type,
+UINT32 length, UINT16 flags) {
+msgOut-nlMsg.nlmsgType = type;
+msgOut-nlMsg.nlmsgFlags = flags;
+msgOut-nlMsg.nlmsgSeq = msgIn-nlMsg.nlmsgSeq;
+msgOut-nlMsg.nlmsgPid = msgIn-nlMsg.nlmsgPid;
+msgOut-nlMsg.nlmsgLen = length;
+
+msgOut-genlMsg.cmd = msgIn-genlMsg.cmd;
+msgOut-genlMsg.version = msgIn-genlMsg.version;
+msgOut-genlMsg.reserved = 0;
+}
+
+/*
+ * XXX: should move out these functions to a Netlink.c or to a 
+OvsMessage.c
+ * or even make them inlined functions in Datapath.h. Can be done after 
+the
+ * first sprint once we have more code to refactor.
+ */
+VOID
+BuildReplyMsgFromMsgIn(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut, UINT16 
+flags) {
+BuildMsgOut(msgIn, msgOut, msgIn-nlMsg.nlmsgType, sizeof(OVS_MESSAGE),
+flags);
+}
+
+VOID
+BuildErrorMsg(POVS_MESSAGE msgIn, POVS_MESSAGE_ERROR msgOut, UINT 
+errorCode) {
+BuildMsgOut(msgIn, (POVS_MESSAGE)msgOut, NLMSG_ERROR,
+sizeof(OVS_MESSAGE_ERROR), 0);
+
+msgOut-errorMsg.error = errorCode;
+msgOut-errorMsg.nlMsg = msgIn-nlMsg; }
+
 /*
  * ---
  * Adds Netlink Header to the NL_BUF.
diff --git a/datapath-windows/ovsext/Netlink/Netlink.h 
b/datapath-windows/ovsext/Netlink/Netlink.h
index 438d857..18e40b1 100644
--- a/datapath-windows/ovsext/Netlink/Netlink.h
+++ b/datapath-windows/ovsext/Netlink/Netlink.h
@@ -32,6 +32,14 @@ typedef struct _OVS_MESSAGE {
 /* Variable length nl_attrs follow. */  } OVS_MESSAGE, *POVS_MESSAGE;
 
+/*
+ * Structure of an error message sent as a reply from kernel.
+ */
+typedef struct _OVS_MESSAGE_ERROR {
+NL_MSG_HDR nlMsg;
+NL_MSG_ERR errorMsg;
+} OVS_MESSAGE_ERROR, *POVS_MESSAGE_ERROR;
+
 /* Netlink attribute types. */
 typedef enum
 {
@@ -86,6 +94,11 @@ NTSTATUS NlFillNlHdr(PNL_BUFFER nlBuf,
  UINT16 nlmsgType, UINT16 nlmsgFlags,
  UINT32 nlmsgSeq, UINT32 nlmsgPid);
 
+VOID BuildReplyMsgFromMsgIn(POVS_MESSAGE msgIn, POVS_MESSAGE msgOut,
+UINT16 flags); VOID 
+BuildErrorMsg(POVS_MESSAGE msgIn, POVS_MESSAGE_ERROR msgOut,
+   UINT errorCode);
+
 /* Netlink message accessing the payload */  PVOID NlMsgAt(const PNL_MSG_HDR 
nlh, UINT32 offset);
 UINT32 NlMsgSize(const PNL_MSG_HDR nlh); diff --git 
a/datapath-windows/ovsext/Vport.c b/datapath-windows/ovsext/Vport.c index 
fc905b1..35f95bc 100644
--- a/datapath-windows/ovsext/Vport.c
+++ b/datapath-windows/ovsext/Vport.c
@@ -1631,44 +1631,6 @@ OvsWaitActivate(POVS_SWITCH_CONTEXT switchContext

[ovs-dev] [PATCH] datapath-windows: Add PID Number attribute for port notification / Fix logic

[Eitan Eliahu]

[1] User mode OVS expects to have the PID number in the port state
notification command
[2] Fix logic error-for-Windows-user-mode

Signed-off-by: Eitan Eliahu elia...@vmware.com
---
 datapath-windows/ovsext/Datapath.c | 14 --
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/datapath-windows/ovsext/Datapath.c 
b/datapath-windows/ovsext/Datapath.c
index b9d97f8..0a8ddc5 100644
--- a/datapath-windows/ovsext/Datapath.c
+++ b/datapath-windows/ovsext/Datapath.c
@@ -2148,7 +2148,7 @@ OvsPortFillInfo(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 PNL_BUFFER nlBuf)
 {
 NTSTATUS status;
-BOOLEAN rc;
+BOOLEAN ok;
 OVS_MESSAGE msgOutTmp;
 PNL_MSG_HDR nlMsg;
 POVS_VPORT_ENTRY vport;
@@ -2177,8 +2177,8 @@ OvsPortFillInfo(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 }
 msgOutTmp.ovsHdr.dp_ifindex = gOvsSwitchContext-dpNo;
 
-rc = NlMsgPutHead(nlBuf, (PCHAR)msgOutTmp, sizeof msgOutTmp);
-if (!rc) {
+ok = NlMsgPutHead(nlBuf, (PCHAR)msgOutTmp, sizeof msgOutTmp);
+if (!ok) {
 status = STATUS_INVALID_BUFFER_SIZE;
 goto cleanup;
 }
@@ -2189,10 +2189,12 @@ OvsPortFillInfo(POVS_USER_PARAMS_CONTEXT usrParamsCtx,
 goto cleanup;
 }
 
-rc = NlMsgPutTailU32(nlBuf, OVS_VPORT_ATTR_PORT_NO, eventEntry-portNo) ||
- NlMsgPutTailU32(nlBuf, OVS_VPORT_ATTR_TYPE, vport-ovsType) ||
+ok = NlMsgPutTailU32(nlBuf, OVS_VPORT_ATTR_PORT_NO, eventEntry-portNo) 
+ NlMsgPutTailU32(nlBuf, OVS_VPORT_ATTR_TYPE, vport-ovsType) 
+ NlMsgPutTailU32(nlBuf, OVS_VPORT_ATTR_UPCALL_PID,
+ vport-upcallPid) 
  NlMsgPutTailString(nlBuf, OVS_VPORT_ATTR_NAME, vport-ovsName);
-if (!rc) {
+if (!ok) {
 status = STATUS_INVALID_BUFFER_SIZE;
 goto cleanup;
 }
-- 
1.9.4.msysgit.2

___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Fix BSOD when uninstalling driver

[Eitan Eliahu]

Hi Alin,
Thank you for fixing this issue. Do you think we can acquire the lock in 
OvsDeleteFlowTable() ?
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Alin Serdean
Sent: Tuesday, November 25, 2014 9:09 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Fix BSOD when uninstalling driver

Add an additional check to see if the flowTable is not NULL.

kd k
Child-SP  RetAddr   Call Site
d000`26166af8 f802`dde5e7c6 nt!DbgBreakPointWithStatus
d000`26166b00 f802`dde5e0d7 nt!KiBugCheckDebugBreak+0x12
d000`26166b60 f802`51a4 nt!KeBugCheck2+0x8ab
d000`26167270 f802`ddde0be9 nt!KeBugCheckEx+0x104
d000`261672b0 f802`f43a nt!KiBugCheckDispatch+0x69
d000`261673f0 f800`024cb4d4 nt!KiPageFault+0x23a
d000`26167580 f800`024cc3ef OVSExt!OvsDoDumpFlows+0xa0 
[c:\work\ovs\datapath-windows\ovsext\flow.c @ 2015]
d000`261675e0 f800`024d134c OVSExt!_FlowNlDumpCmdHandler+0x197 
[c:\work\ovs\datapath-windows\ovsext\flow.c @ 590]
d000`26167740 f800`024e128f OVSExt!InvokeNetlinkCmdHandler+0x6c 
[c:\work\ovs\datapath-windows\ovsext\datapath.c @ 952]
d000`26167770 f800`0053bc18 OVSExt!OvsDeviceControl+0x263 
[c:\work\ovs\datapath-windows\ovsext\datapath.c @ 862]
d000`26167840 f802`de04f395 NDIS!ndisDummyIrpHandler+0x88
d000`26167870 f802`de04fd2a nt!IopXxxControlFile+0x845
d000`26167a20 f802`ddde08b3 nt!NtDeviceIoControlFile+0x56
d000`26167a90 `775a2772 nt!KiSystemServiceCopyEnd+0x13
`009eee88 `775a2371 wow64cpu!CpupSyscallStub+0x2
`009eee90 `775c323a wow64cpu!DeviceIoctlFileFault+0x31
`009eef40 `775c317e wow64!RunCpuSimulation+0xa
`009eef90 7ffb`c1ca6bd0 wow64!Wow64LdrpInitialize+0x172
`009ef4d0 7ffb`c1ca6aa6 ntdll!_LdrpInitialize+0xd8
`009ef540 ` ntdll!LdrInitializeThunk+0xe
kd !analyze -v
***
* *
*Bugcheck Analysis*
* *
***

DRIVER_IRQL_NOT_LESS_OR_EQUAL (d1)
An attempt was made to access a pageable (or completely invalid) address at an 
interrupt request level (IRQL) that is too high.  This is usually caused by 
drivers using improper addresses.
If kernel debugger is available get stack backtrace.
Arguments:
Arg1: , memory referenced
Arg2: 0002, IRQL
Arg3: , value 0 = read operation, 1 = write operation
Arg4: f800024cb4d4, address which referenced memory

Debugging Details:
--


KERNEL32.DLL was not found in the image list.
Debugger will attempt to load KERNEL32.DLL at given base `.

Please provide the full image name, including the extension (i.e. kernel32.dll) 
for more reliable results.Base address and size overrides can be given as 
.reload image.ext=base,size.
Unable to add module at `

READ_ADDRESS:  

CURRENT_IRQL:  2

FAULTING_IP:
OVSExt!OvsDoDumpFlows+a0 [c:\work\ovs\datapath-windows\ovsext\flow.c @ 2015]
f800`024cb4d4 488b18  mov rbx,qword ptr [rax]

DEFAULT_BUCKET_ID:  WIN8_DRIVER_FAULT

BUGCHECK_STR:  AV

PROCESS_NAME:  ovs-vswitchd.e

ANALYSIS_VERSION: 6.3.9600.17237 (debuggers(dbg).140716-0327) amd64fre

TRAP_FRAME:  d000261673f0 -- (.trap 0xd000261673f0)
NOTE: The trap frame does not contain all registers.
Some register values may be zeroed or incorrect.
rax= rbx= rcx=
rdx=d000261675e0 rsi= rdi=
rip=f800024cb4d4 rsp=d00026167580 rbp=
 r8=d00026167601  r9= r10=c00d
r11=d000261677b0 r12= r13=
r14= r15=
iopl=0 nv up ei pl zr na po nc
OVSExt!OvsDoDumpFlows+0xa0:
f800`024cb4d4 488b18  mov rbx,qword ptr [rax] 
ds:`=
Resetting default scope

LAST_CONTROL_TRANSFER:  from f802dde5e7c6 to f802bc90

STACK_TEXT:
d000`26166af8 f802`dde5e7c6 : ` ` 
d000`26166c60 f802`ddd83654 : nt!DbgBreakPointWithStatus
d000`26166b00 f802`dde5e0d7 : `0003 d000`26166c60 
f802`ddde3070 `00d1 : nt!KiBugCheckDebugBreak+0x12
d000`26166b60 f802`51a4 : ` `0001 
f6fb` d000`26e0 : nt!KeBugCheck2+0x8ab
d000`26167270 f802`ddde0be9 : `000a ` 
`0002 ` : nt!KeBugCheckEx+0x104

Re: [ovs-dev] [PATCH] datapath-windows: Update OVSEXT VS project to support 6.40

[Eitan Eliahu]

Hi Sorin,
Can we have Win8.1 DDK for both NDIS 6.40 and NDIS 6.30 drivers?
We don't want to have 6.30 driver built with not current DDK.
Thanks,
Eitan

Acked-by: Eitan Eliahu elia...@vmware.com


-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Monday, November 24, 2014 8:18 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Update OVSEXT VS project to 
support 6.40

Added support for creating OVS extension driver NDIS 6.40 compliant.

Currently the OVSEXT Visual Studio project has four build configurations,
'Win8 Release', 'Win8 Debug', 'Win8.1 Release' and 'Win8.1 Debug'. All of them 
are creating a binary that is NDIS 6.30 compliant. I have changed the
Win8.1 build configurations in order to create a binary that is NDIS 6.40 
compliant.

In this way, the OVSEXT project is able to create a release/debug binary that 
is NDIS 6.30 compliant, using the 'Win8 Release' and 'Win8 Debug'
build configurations, as well as a release/debug binary that is NDIS 6.40 
compliant, using the 'Win8.1 Release' and 'Win8.1 Debug' build configurations.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
---
 datapath-windows/ovsext/ovsext.vcxproj | 35 +-
 1 file changed, 34 insertions(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/ovsext.vcxproj 
b/datapath-windows/ovsext/ovsext.vcxproj
index 88c9122..693bc50 100644
--- a/datapath-windows/ovsext/ovsext.vcxproj
+++ b/datapath-windows/ovsext/ovsext.vcxproj
@@ -102,7 +102,7 @@
   PropertyGroup
 TargetNameOVSExt/TargetName
   /PropertyGroup
-  ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8 
+ Release|x64'
 ClCompile
   
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
 /ClCompile
@@ -113,6 +113,39 @@
   
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
 /ResourceCompile
   /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8.1 
Release|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8 
Debug|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8.1 
Debug|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
   ItemDefinitionGroup
 Link
   
AdditionalDependencies%(AdditionalDependencies);$(DDK_LIB_PATH)\ndis.lib;$(DDK_LIB_PATH)\fwpkclnt.lib;$(SDK_LIB_PATH)\uuid.lib;$(DDK_LIB_PATH)\netio.lib/AdditionalDependencies
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AAIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=Vvf3Ian_SaQZWBuLksgcYMJrE9Z1Gpp3lbgkr1MkdbAs=x19l9o7TC5LKuchoj9q1j83Kk8gulY8hOKnaWhpDGsYe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH v2] datapath-windows: Update OVSEXT VS project to support 6.40

[Eitan Eliahu]

Sorin, did you remove Win8 project? Can we have 8.1 WInDDK libraries for both 
6.30 and 6.40 drivers?
Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Monday, November 24, 2014 9:03 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH v2] datapath-windows: Update OVSEXT VS project to 
support 6.40

Added support for creating OVS extension driver NDIS 6.40 compliant.

Currently the OVSEXT Visual Studio project has four build configurations,
'Win8 Release', 'Win8 Debug', 'Win8.1 Release' and 'Win8.1 Debug'. All of them 
are creating a binary that is NDIS 6.30 compliant. I have changed the
Win8.1 build configurations in order to create a binary that is NDIS 6.40 
compliant.

In this way, the OVSEXT project is able to create a release/debug binary that 
is NDIS 6.30 compliant, using the 'Win8 Release' and 'Win8 Debug'
build configurations, as well as a release/debug binary that is NDIS 6.40 
compliant, using the 'Win8.1 Release' and 'Win8.1 Debug' build configurations.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Acked-by: Eitan Eliahu elia...@vmware.com
---
 datapath-windows/ovsext/ovsext.vcxproj | 35 +-
 1 file changed, 34 insertions(+), 1 deletion(-)

diff --git a/datapath-windows/ovsext/ovsext.vcxproj 
b/datapath-windows/ovsext/ovsext.vcxproj
index 88c9122..693bc50 100644
--- a/datapath-windows/ovsext/ovsext.vcxproj
+++ b/datapath-windows/ovsext/ovsext.vcxproj
@@ -102,7 +102,7 @@
   PropertyGroup
 TargetNameOVSExt/TargetName
   /PropertyGroup
-  ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8 
+ Release|x64'
 ClCompile
   
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
 /ClCompile
@@ -113,6 +113,39 @@
   
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
 /ResourceCompile
   /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8.1 
Release|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8 
Debug|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
+  ItemDefinitionGroup Condition='$(Configuration)|$(Platform)'=='Win8.1 
Debug|x64'
+ClCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ClCompile
+Midl
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/Midl
+ResourceCompile
+  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640=1/PreprocessorDefinitions
+/ResourceCompile
+  /ItemDefinitionGroup
   ItemDefinitionGroup
 Link
   
AdditionalDependencies%(AdditionalDependencies);$(DDK_LIB_PATH)\ndis.lib;$(DDK_LIB_PATH)\fwpkclnt.lib;$(SDK_LIB_PATH)\uuid.lib;$(DDK_LIB_PATH)\netio.lib/AdditionalDependencies
--
1.9.0.msysgit.0
___
dev mailing list
dev@openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_devd=AAIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=LrfnoGch9G6_TPkJ5ZHRsNTI3exzEFBiWaD6dUH_zFQs=n4bOIxBT0f4aoMlS1R30ZavdqgdgJLZAj9W5yK21nAMe=
 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Re: [ovs-dev] [PATCH] datapath-windows: Support for hybrid forwarding

[Eitan Eliahu]

Hi Sorin,
This change applies only for 6.40 drivers. Unless we bump the driver version 
this code has no effect.
You may want to bump the minor version.

#define NDIS_FILTER_MAJOR_VERSION 6
#define NDIS_FILTER_MINOR_VERSION 30   40

driverChars.MajorNdisVersion = NDIS_FILTER_MAJOR_VERSION;
driverChars.MinorNdisVersion = NDIS_FILTER_MINOR_VERSION;

Thanks,
Eitan

-Original Message-
From: dev [mailto:dev-boun...@openvswitch.org] On Behalf Of Sorin Vinturis
Sent: Friday, November 21, 2014 8:16 AM
To: dev@openvswitch.org
Subject: [ovs-dev] [PATCH] datapath-windows: Support for hybrid forwarding

In a hybrid forwarding environment, there are two types of packets that enter 
and leave the Hyper-V extensible switch: NVGRE packets and non-NVGRE packets. 
Hybrid forwarding involves filtering the incoming traffic based on packet type.

Thus, we must split the incoming traffic into NVGRE and non-NVGRE packets.
All non-NVGRE traffic is forwarded by the OVS extension and processed as usual, 
and the NVGRE traffic is passed to NDIS to be handled by the HNV module.

Currently, the OVS extension supports only NDIS 6.30, but hybrid forwarding 
involves supporting NDIS 6.40. Thus, I have changed the necessary compiler 
settings to reflect this.

More details about hybrid forwarding and the necessary NDIS 6.40 support is 
provided in issue #52.

Signed-off-by: Sorin Vinturis svintu...@cloudbasesolutions.com
Reported-by: Nithin Raju nit...@vmware.com
Reported-at: 
https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_openvswitch_ovs-2Dissues_issues_52d=AAIGaQc=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEsr=CWsgHUxi6ExLXY798tmo3LJ4e3geGYp56lkcH-5cLCYm=IF9N0kgUMpt8gOjYSq3_AJkSJSQlF85oyI1sxDSezaEs=KskXnDJSk-dizBlmk1XJga2EXqtmL6-7g7HTZF0rEloe=
---
 datapath-windows/ovsext/PacketIO.c | 33 +++--
 datapath-windows/ovsext/ovsext.vcxproj |  6 +++---
 2 files changed, 34 insertions(+), 5 deletions(-)

diff --git a/datapath-windows/ovsext/PacketIO.c 
b/datapath-windows/ovsext/PacketIO.c
index 1af391b..d022c45 100644
--- a/datapath-windows/ovsext/PacketIO.c
+++ b/datapath-windows/ovsext/PacketIO.c
@@ -193,6 +193,10 @@ OvsStartNBLIngress(POVS_SWITCH_CONTEXT switchContext,
 LIST_ENTRY missedPackets;
 UINT32 num = 0;
 OvsCompletionList completionList;
+PNET_BUFFER_LIST extForwardedNbls = NULL;
+PNET_BUFFER_LIST nativeForwardedNbls = NULL;
+PNET_BUFFER_LIST *nextExtForwardNbl = extForwardedNbls;
+PNET_BUFFER_LIST *nextNativeForwardedNbl = nativeForwardedNbls;
 
 dispatch = NDIS_TEST_SEND_AT_DISPATCH_LEVEL(SendFlags)?
 NDIS_RWL_AT_DISPATCH_LEVEL : 0; @@ 
-201,8 +205,27 @@ OvsStartNBLIngress(POVS_SWITCH_CONTEXT switchContext,
 
 InitializeListHead(missedPackets);
 OvsInitCompletionList(completionList, switchContext, sendCompleteFlags);
-
-for (curNbl = netBufferLists; curNbl != NULL; curNbl = nextNbl) {
+
+/*
+ * Split NBL list into NBLs to be forwarded by us, and those that require
+ * native forwarding. */
+for (curNbl = netBufferLists; curNbl != NULL; curNbl = nextNbl) {
+nextNbl = curNbl-Next;
+curNbl-Next = NULL;
+fwdDetail = NET_BUFFER_LIST_SWITCH_FORWARDING_DETAIL(curNbl);
+
+if (fwdDetail-NativeForwardingRequired) {
+*nextNativeForwardedNbl = curNbl;
+nextNativeForwardedNbl = (curNbl-Next);
+}
+else
+{
+*nextExtForwardNbl = curNbl;
+nextExtForwardNbl = (curNbl-Next);
+}
+}
+
+for (curNbl = extForwardedNbls; curNbl != NULL; curNbl = nextNbl) {
 POVS_VPORT_ENTRY vport;
 UINT32 portNo;
 OVS_DATAPATH *datapath = switchContext-datapath; @@ -315,6 +338,12 
@@ dropit:
 }
 }
 
+if (nativeForwardedNbls) {
+/* This is NVGRE encapsulated traffic and is forwarded to NDIS
+ * in order to be handled by the HNV module. */
+OvsSendNBLIngress(switchContext, nativeForwardedNbls, SendFlags);
+}
+
 /* Queue the missed packets. */
 OvsQueuePackets(missedPackets, num);
 OvsFinalizeCompletionList(completionList);
diff --git a/datapath-windows/ovsext/ovsext.vcxproj 
b/datapath-windows/ovsext/ovsext.vcxproj
index 88c9122..2a34f5d 100644
--- a/datapath-windows/ovsext/ovsext.vcxproj
+++ b/datapath-windows/ovsext/ovsext.vcxproj
@@ -104,13 +104,13 @@
   /PropertyGroup
   ItemDefinitionGroup
 ClCompile
-  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+  
+ PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640
+ =1/PreprocessorDefinitions
 /ClCompile
 Midl
-  
PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS630=1/PreprocessorDefinitions
+  
+ PreprocessorDefinitions%(PreprocessorDefinitions);NDIS_WDM=1;NDIS640
+ =1/PreprocessorDefinitions
 /Midl
 ResourceCompile
-