s
OVS with conntrack enabled: 3.9-4.0Gbps
After optimization:
Native: 6.0Gbps
OVS: 5.1-5.75Gbps
OVS with conntrack enabled:: 4.2-4.4Gbps
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-nat.c | 7 +-
datapath-windows/ovsext/Conntrack-related.c | 17 ++--
datapath-windows/ovsext/C
on:
Native: 6.0Gbps
OVS: 5.1-5.75Gbps
OVS with conntrack enabled: 3.9-4.0Gbps
After optimization:
Native: 6.0Gbps
OVS: 5.1-5.75Gbps
OVS with conntrack enabled:: 4.2-4.4Gbps
Tested by loading/unloading driver with driver verifier enabled
Signed-off-by: Anand Kumar
---
v1->v2: Update c
Hi Shashank,
Thanks for the review. I will address your comments and send out a V3.
Regards,
Anand Kumar
From: Shashank Ram
Date: Friday, June 8, 2018 at 10:11 AM
To: Anand Kumar
Cc: "d...@openvswitch.org"
Subject: Re: [ovs-dev] [PATCH v2] datapath-windows: Optimize conntrack
p
iperf3.
Before optimization:
Native: 6.0Gbps
OVS: 5.1-5.75Gbps
OVS with conntrack enabled: 3.9-4.0Gbps
After optimization:
Native: 6.0Gbps
OVS: 5.1-5.75Gbps
OVS with conntrack enabled:: 4.2-4.4Gbps
Tested by loading/unloading driver with driver verifier enabled
Signed-off-by: Anand Kumar
---
Abandoning this change. I have couple of patches to optimize conntrack, will be
sending it out as a series.
Thanks,
Anand Kumar
On 6/11/18, 11:39 AM, "Anand Kumar" wrote:
- Use spinlock instead of read/write lock for conntrack entry.
- Use NdisInterlockedXX wrapper api
This patch series is primarily to refactor conntrack code for
better throughput with conntrack.
With this patch series TCP throughput with conntrack increased
by ~50%.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Anand Kumar (4
Conntrack 5-tuple consists of src address, dst address, src port,
dst port and protocol which will be unique to a ct session.
Use this information along with zone to compute hash.
Also re-factor conntrack code related to parsing netlink attributes.
Signed-off-by: Anand Kumar
---
datapath
kedXX.
- Update 'ctTotalRelatedEntries' using interlocked functions.
- Move conntrack lock out of NAT module.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-nat.c | 7 +-
datapath-windows/ovsext/Conntrack-related.c | 17 ++--
datapath-windows/ovsext/Conntrack.
TCP traffic.
- Optimize conntrack look up.
- Remove 'bucketlockRef' member from conntrack entry structure.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-ftp.c | 4 +-
datapath-windows/ovsext/Conntrack-tcp.c | 15 ++---
datapath-windows/ovsext/Conntrack.c
The 'ovsNatTable' and 'ovsUnNatTable' tables are shared
between cleanup threads and packet processing thread.
In order to protect these two tables use a spinlock.
Also introduce counters to track number of nat entries.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/
Hi Shashank,
Thanks for the review.
Please find my response inline.
Thanks,
Anand Kumar
From: Shashank Ram
Date: Monday, June 18, 2018 at 11:27 AM
To: Anand Kumar , "d...@openvswitch.org"
Subject: Re: [ovs-dev] [PATCH v4 1/4] datapath-windows: Use spinlock instead of
RW lock fo
Hi Shashank,
Thanks for the review. Please find my response inline.
Thanks,
Anand Kumar
On 6/18/18, 11:54 AM, "Shashank Ram" wrote:
On 06/17/2018 10:37 PM, Anand Kumar wrote:
> This patch primarily gets rid of NdisRWLock in conntrack for NAT
> function
Hi Shashank,
I will address this in next version of the patch.
Thanks,
Anand Kumar
On 6/18/18, 2:36 PM, "Shashank Ram" wrote:
This patch should be combined with the patch where NAT lock is removed
from CT. Keeping this separate will cause the previous patches in this
kedXX.
- Update 'ctTotalRelatedEntries' using interlocked functions.
- Move conntrack lock out of NAT module.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Calculate the dispatch level only
This patch series is primarily to refactor conntrack code for
better throughput with conntrack.
With this patch series TCP throughput with conntrack increased
by ~50%.
Anand Kumar (3):
datapath-windows: Use spinlock instead of RW lock for ct entry
datapath-windows: Implement locking in
with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Updated commit message to include testing done.
---
datapath-windows/ovsext/Conntrack.c | 228 ++--
datapath-windows/ovsext/Conntrack.h | 2 -
2 files changed,
x27;bucketlockRef' member from conntrack entry structure.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Merge patch 2 and 3 so that NAT locks related changes are in a
single pa
This patch series is primarily to refactor conntrack code for
better throughput with conntrack.
With this patch series TCP throughput with conntrack increased
by ~50%.
Anand Kumar (3):
datapath-windows: Use spinlock instead of RW lock for ct entry
datapath-windows: Implement locking in
kedXX.
- Update 'ctTotalRelatedEntries' using interlocked functions.
- Move conntrack lock out of NAT module.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Calculate the dispatch level only
x27;bucketlockRef' member from conntrack entry structure.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Merge patch 2 and 3 so that NAT locks related changes are in a
with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Updated commit message to include testing done.
v2->v3: No change.
---
datapath-windows/ovsext/Conntrack.c | 228 ++--
datapath-windows/ovsext/Conntrack.h |
This patch series is primarily to refactor conntrack code for
better throughput with conntrack.
With this patch series TCP throughput with conntrack increased
by ~50%.
Anand Kumar (3):
datapath-windows: Use spinlock instead of RW lock for ct entry
datapath-windows: Implement locking in
x27;bucketlockRef' member from conntrack entry structure.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Merge patch 2 and 3 so that NAT locks related changes are in a
single pa
kedXX.
- Update 'ctTotalRelatedEntries' using interlocked functions.
- Move conntrack lock out of NAT module.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Calculate the dispatch level only
with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
---
v1->v2: Updated commit message to include testing done.
v2->v3: No change
v3->v4: No change
---
datapath-windows/ovsext/Conntrack.c | 228 ++--
datapath-windo
Hi Alin,
Thanks for running the code analysis on the patch series.
As discussed in the Hyper-V meeting, I have addressed them and send out a v4 of
my patch series.
Thanks,
Anand Kumar
On 6/19/18, 9:14 AM, "Alin Serdean" wrote:
Thanks a lot for the series and the ben
Hi Alin,
Thanks for reviewing the patch series.
I will send out a v5 to address this issue and retain the acks.
Thanks,
Anand Kumar
On 6/22/18, 7:18 AM, "Alin Serdean" wrote:
> -Mesaj original-
> De la: ovs-dev-boun...@openvswitch.org boun...@openvswitch.org&
This patch series is primarily to refactor conntrack code for
better throughput with conntrack.
With this patch series TCP throughput with conntrack increased
by ~50%.
Anand Kumar (3):
datapath-windows: Use spinlock instead of RW lock for ct entry
datapath-windows: Implement locking in
kedXX.
- Update 'ctTotalRelatedEntries' using interlocked functions.
- Move conntrack lock out of NAT module.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
v1->
with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
v1->v2: Updated commit message to include testing done.
v2->v3: No change
v3->v4: No change
v4->v5: No change
---
datapath-windows/ovsext/Connt
x27;bucketlockRef' member from conntrack entry structure.
Testing:
Verified loading/unloading the driver with driver verified enabled.
Ran TCP/UDP and ICMP traffic.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
v1->v2: Merge patch 2 and 3 so that NAT locks related changes
Hi Alin,
Thanks for applying the patch.
Yes, we need this on branch 2.8 as well.
Thanks,
Anand Kumar
On 10/16/17, 7:33 AM, "Alin Serdean" wrote:
It looks good and I managed to test it, sorry for the delay.
I had some issues while disabling the offloads on the hw
Set the geneve flags OVS_TNL_F_GENEVE_OPT and OVS_TNL_F_CRT_OPT
in OvsDecapGeneve, so that windows behavior is similiar to linux
https://github.com/openvswitch/ovs/blob/master/datapath/linux/compat/geneve.c#L242
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Geneve.c | 8
1
Hi Shashank,
I have set the tunnel key flags OVS_TNL_F_GENEVE_OPT by default to match the
linux behavior
@https://github.com/openvswitch/ovs/blob/master/datapath/linux/compat/geneve.c#L242
Thanks,
Anand Kumar
On 10/18/17, 9:07 AM, "Shashank Ram" wrote:
On 10/17/1
Hi Sairam,
Sure, I will add the check to see if there are any geneve tunnel options and
set OVS_TNL_F_GENEVE_OPT accordingly and send out v2 version of the patch.
Thanks,
Anand Kumar
On 10/19/17, 8:13 AM, "Sairam Venugopal" wrote:
Anand,
Looks like Linux does make a ch
OVS_TNL_F_GENEVE_OPT
if the packet has geneve options.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Geneve.c | 9 +
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/datapath-windows/ovsext/Geneve.c b/datapath-windows/ovsext/Geneve.c
index 43374e2..6dca69b 100644
--- a
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 11/1/17, 10:19 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Gabriel Serdean" wrote:
_set_output_format is deprecated ang no longer required
starting from MSC_VER 1900 (VS 2015):
https://urldefense.proofpoint.co
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 11/1/17, 10:19 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Gabriel Serdean" wrote:
Starting from WDK 10 the structure `timespec` is defined in .
This patch adds a check for the structure to make aware of it,
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 11/1/17, 10:19 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Gabriel Serdean" wrote:
For some unclear and accidental reasons, the Windows 10 SDK
renamed _Interlocked* functions to _InlineInterlocked* (al
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 11/1/17, 10:19 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Gabriel Serdean" wrote:
Unfortunately starting from VS 2015, the "C" definition for `offsetof`
has been changed. Please see:
https://urldefen
This patch series replaces existing one RW lock implemenation in
conntrack with two RW locks in conntrack and one RW lock in NAT.
Anand Kumar (3):
datapath-windows: Refactor conntrack code.
datapath-windows: Add a global level RW lock for NAT
datapath-windows: Optimize conntrack lock
Some of the functions and code are refactored
so that new conntrack lock can be implemented
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-nat.c | 11 +--
datapath-windows/ovsext/Conntrack.c | 170 ++--
datapath-windows/ovsext/Conntrack.h
Currently NAT module relies on the existing conntrack lock.
This patch provides a basic lock implementation for NAT module
in conntrack.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack.c | 36 ++--
1 file changed, 34 insertions(+), 2 deletions
reference(bucketLockRef)
to the corresponding OvsCtBucketLock of conntrack table.
We need this reference to retrieve ovsCtBucketLock from ct entry
for delete operation.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-nat.c | 6 +
datapath-windows/ovsext/Conntrack.c | 231
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 11/20/17, 3:06 PM, "ovs-dev-boun...@openvswitch.org on behalf of Shashank
Ram" wrote:
Decap functions for tunneling protocols do not compute
the packet header offsets correctly when there is a VLAN
tag in the L2 header. This
since conntrack 5-tuple is an optional parameter.
Anand Kumar (1):
datapath-windows: Add support for deleting conntrack entry by 5 tuple.
datapath-windows/ovsext/Conntrack.c | 146 +---
1 file changed, 134 insertions(+), 12 deletions(-)
--
2.9.3.windows.1
To delete a conntrack entry specified by 5-tuple pass an additional
conntrack 5-tuple parameter to flush-conntrack.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack.c | 146 +---
1 file changed, 134 insertions(+), 12 deletions(-)
diff --git a
This patch series replaces existing one RW lock implemenation in
conntrack with two RW locks in conntrack and one RW lock in NAT.
---
v1->v2:
- Patch 3, address review comments
---
Anand Kumar (3):
datapath-windows: Refactor conntrack code.
datapath-windows: Add a global level RW lock for
Some of the functions and code are refactored
so that new conntrack lock can be implemented
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack-nat.c | 11 +-
datapath-windows/ovsext/Conntrack.c | 174 ++--
datapath
Currently NAT module relies on the existing conntrack lock.
This patch provides a basic lock implementation for NAT module
in conntrack.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack.c | 36 ++--
1 file changed
reference(bucketLockRef)
to the corresponding OvsCtBucketLock of conntrack table.
We need this reference to retrieve ovsCtBucketLock from ct entry
for delete operation.
Signed-off-by: Anand Kumar
---
v1->v2: Address potential memory leak in conntrack initialization.
---
datapath-windows/ovs
This patch series replaces existing one RW lock implemenation in
conntrack with two RW locks in conntrack and one RW lock in NAT.
---
v1->v2:
- Patch 3, address review comments
v2->v3:
- Patch 3, fix invalid memory access after deleting ct entry
---
Anand Kumar (3):
datapath-windows: Re
Some of the functions and code are refactored
so that new conntrack lock can be implemented
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack-nat.c | 11 +-
datapath-windows/ovsext/Conntrack.c | 174 ++--
datapath
reference(bucketLockRef)
to the corresponding OvsCtBucketLock of conntrack table.
We need this reference to retrieve ovsCtBucketLock from ct entry
for delete operation.
Signed-off-by: Anand Kumar
---
v1->v2: Address potential memory leak in conntrack initialization.
v2->v3: Fix invalid memory
Currently NAT module relies on the existing conntrack lock.
This patch provides a basic lock implementation for NAT module
in conntrack.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack.c | 36 ++--
1 file changed
Currently NAT module relies on the existing conntrack lock.
This patch provides a basic lock implementation for NAT module
in conntrack.
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack.c | 36 ++--
1 file changed
analyzer warning message
---
Anand Kumar (3):
datapath-windows: Refactor conntrack code.
datapath-windows: Add a global level RW lock for NAT
datapath-windows: Optimize conntrack lock implementation.
datapath-windows/ovsext/Conntrack-nat.c | 17 +-
datapath-windows/ovsext/Conntrack.c
Some of the functions and code are refactored
so that new conntrack lock can be implemented
Signed-off-by: Anand Kumar
Acked-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Conntrack-nat.c | 11 +-
datapath-windows/ovsext/Conntrack.c | 174 ++--
datapath
reference(bucketLockRef)
to the corresponding OvsCtBucketLock of conntrack table.
We need this reference to retrieve ovsCtBucketLock from ct entry
for delete operation.
Signed-off-by: Anand Kumar
---
v1->v2: Address potential memory leak in conntrack initialization.
v2->v3: Fix invalid memory
Hi Alin,
Thanks for the review. I’ll address the warning and send out the patch.
Thanks,
Anand Kumar
On 1/29/18, 5:20 AM, "Alin Serdean" wrote:
Trimming the patch a bit.
Just one small nit from the static analyzer inlined.
Rest looks good.
Acked-by: Al
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-icmp.c | 1 +
datapath-windows/ovsext/Conntrack-tcp.c | 4
datapath-windows/ovsext/Conntrack.c | 4
3 files changed, 9 insertions(+)
diff --git a/datapath-windows/ovsext/Conntrack-icmp.c
b/datapath-windows/ovsext
Hi Alin,
Thanks for the quick review. I will expand log messages to “Invalid XXX packet
detected”.
I’m not sure if I follow your comment about “s/syn/SYN/g”.
Do you want me to change it to Uppercase?
Thanks,
Anand Kumar
On 2/2/18, 2:06 PM, "Alin Serdean" wrote:
Looks good ju
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-icmp.c | 1 +
datapath-windows/ovsext/Conntrack-tcp.c | 4
datapath-windows/ovsext/Conntrack.c | 6 ++
3 files changed, 11 insertions(+)
diff --git a/datapath-windows/ovsext/Conntrack-icmp.c
b/datapath-windows
Hi,
My thoughts are with Shashank on this, it makes sense to have 1 configure and 1
make command to build a particular target, instead of having flexibility to
specify multiple targets.
Thanks,
Anand Kumar
On 2/8/18, 10:56 AM, "ovs-dev-boun...@openvswitch.org on behalf of Shashank
Acked-by: Anand Kumar
Thanks,
Anand Kumar
From: Shashank Ram
Sent: Tuesday, February 6, 2018 1:29 PM
To: d...@openvswitch.org
Cc: Shashank Ram
Subject: [PATCH
Currently nbl information is getting dumped whenever a nbl is copied or
allocated, since OVS_DBG_DEFAULT is set to OVS_DBG_INFO for debug builds,
which affects the ovs performance. Instead dump nbl information only when
OVS_DBG_DEFAULT is set to OVS_LOG_LOUD
Signed-off-by: Anand Kumar
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 2/21/18, 6:57 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin Gabriel
Serdean" wrote:
Use key->ct.mark instead of key->ct.zone when generating the hash
over the mark.
Signed-off-by: Alin
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 2/27/18, 11:57 AM, "ovs-dev-boun...@openvswitch.org on behalf of Shashank
Ram" wrote:
Adds support to selectively compile kernel driver for
target versions. This is useful when environments to
compile for all targets mi
Currently nbl information is getting dumped whenever a nbl is copied or
allocated, since OVS_DBG_DEFAULT is set to OVS_DBG_INFO for debug builds,
which affects the ovs performance. Instead dump nbl information only when
OVS_DBG_DEFAULT is set to OVS_LOG_LOUD
Signed-off-by: Anand Kumar
: Anand Kumar
---
datapath-windows/ovsext/IpFragment.c | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/datapath-windows/ovsext/IpFragment.c
b/datapath-windows/ovsext/IpFragment.c
index 3d5277a..da9d33a 100644
--- a/datapath-windows/ovsext/IpFragment.c
+++ b/datapath-windows
Thanks for the review.
MIN_FRAGMENT_SIZE is used to determine maximum number of fragments that are
allowed for an IP datagram.
I will update the macro MAX_FRAGMENTS to compute the value based on
MIN_FRAGMENT_SIZE.
Regards,
Anand Kumar
On 3/6/18, 5:43 AM, "aserd...@ovn.org" wro
: Anand Kumar
---
datapath-windows/ovsext/IpFragment.c | 9 +++--
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/datapath-windows/ovsext/IpFragment.c
b/datapath-windows/ovsext/IpFragment.c
index 3d5277a..d59d7cf 100644
--- a/datapath-windows/ovsext/IpFragment.c
+++ b/datapath
Hi Alin,
Thanks for reviewing the patch.
Can we get this patch applied?
Thanks,
Anand Kumar
On 3/7/18, 3:42 AM, "ovs-dev-boun...@openvswitch.org on behalf of
aserd...@ovn.org" wrote:
Acked-by: Alin Gabriel Serdean
Alin.
-Mesaj original-
De la: ov
Acked-by: Anand Kumar
On 4/20/17, 7:43 PM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
Until now we only needed to compute software checksums on net buffer lists
containing a single net buffer.
This patch allows the software checksums to be ap
This patch adds support for tracking Ipv4 fragments in conntrack module.
Individual fragments are not tracked and are consumed by the
fragmentation/reassembly. Only the reassembled Ipv4 datagram is tracked and
treated as a single ct entry.
Signed-off-by: Anand Kumar
---
v6->v7: Made changes
This patch introduces a new field MRU(Maximum Recieved Unit) in the
_OVS_BUFFER_CONTEXT and it is used only for Ip Fragments to retain MRU for
the reassembled IP datagram when the packet is forwarded to userspace.
Signed-off-by: Anand Kumar
---
v6->v7: Increase value size in OVS_BUFFER_CONT
This patch adds support for Fragmenting NBL based on the MRU value.
MRU value is updated only for Ipv4 fragments, if it is non zero, then
fragment the NBL.
Signed-off-by: Anand Kumar
---
v6->v7: Fragment the NBL for tunnel packets
v5->v6: No Change
v4->v5:
- Use MRU informati
Ipv4 fragments.
Signed-off-by: Anand Kumar
---
v6->v7: Added new memory pool tag and align function description
v5->v6: No Change
v4->v5:
- Modified Patch 3 to retain MRU in _OVS_BUFFER_CONTEXT instead of
using it in ovsForwardingContext with minor changes in rest of the
With this patch, OvsTcpSegmentNBL not only supports fragmenting NBL
to TCP segments but also Ipv4 fragments.
To reflect the new changes, renamed function name from OvsTcpSegmentNBL
to OvsFragmentNBL and created a wrapper for OvsTcpSegmentNBL.
Signed-off-by: Anand Kumar
---
v6->v7: As per Al
ch 4 updated to make it compile for release mode.
---
Anand Kumar (5):
datapath-windows: Added a new file to support Ipv4 fragments.
datapath-windows: Added Ipv4 fragments support in Conntrack
datapath-windows: Retain MRU value in the _OVS_BUFFER_CONTEXT.
datapath-windows: Updated OvsTcp
messages, summary, fixed alignment issues.
v1->v2:
- Patch 4 updated to make it compile for release mode.
---
Anand Kumar (5):
datapath-windows: Added a new file to support Ipv4 fragments.
datapath-windows: Added Ipv4 fragments support in Conntrack
datapath-windows: Retain MRU value in th
This patch introduces a new field MRU(Maximum Recieved Unit) in the
_OVS_BUFFER_CONTEXT and it is used only for Ip Fragments to retain MRU for
the reassembled IP datagram when the packet is forwarded to userspace.
Signed-off-by: Anand Kumar
---
v7->v8: No change
v6->v7: Increase value s
This patch adds support for Fragmenting NBL based on the MRU value.
MRU value is updated only for Ipv4 fragments, if it is non zero, then
fragment the NBL and send out the new NBL to the vnic.
Signed-off-by: Anand Kumar
---
v7->v8: No change
v6->v7: Fragment the NBL for tunnel packets
v5-&
With this patch, OvsTcpSegmentNBL not only supports fragmenting NBL
to TCP segments but also Ipv4 fragments.
To reflect the new changes, renamed function name from OvsTcpSegmentNBL
to OvsFragmentNBL and created a wrapper for OvsTcpSegmentNBL.
Signed-off-by: Anand Kumar
---
v7->v8: No change
Ipv4 fragments.
Signed-off-by: Anand Kumar
---
v7->v8: Address locking issues found by inspection that could resolve the
assertion observed by Alin in OvsLookupIPFrag() with verifier enabled.
v6->v7: Added new memory pool tag and align function description
v5->v6: No Chan
This patch adds support for tracking Ipv4 fragments in conntrack module.
Individual fragments are not tracked and are consumed by the
fragmentation/reassembly. Only the reassembled Ipv4 datagram is tracked and
treated as a single ct entry.
Signed-off-by: Anand Kumar
---
v7->v8: No change
v6-
Acked-by: Anand Kumar kumaran...@vmware.com
Thanks,
Anand Kumar
On 5/17/17, 6:49 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
Update the key used by the actions which follow up defragmentation, with
no fragment set in the IP header.
F
Thanks for identifying and fixing this.
Acked-by: Anand Kumar
Regards,
Anand Kumar
On 5/17/17, 6:54 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
Found using static analysis tools.
Signed-off-by: Alin Gabriel Serdean
---
datapa
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 5/17/17, 5:57 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
Found by inspection.
Signed-off-by: Alin Gabriel Serdean
---
datapath-windows/ovsext/Actions.c | 24
Hi Alin,
Thank you for identifying and fixing issue in fragmenatation.
Acked-by: Anand Kumar
Regards,
Anand Kumar
On 5/17/17, 6:43 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
When a conntrack action is applied over an IP fragment we pend th
Thank you for the patches.
Acked-By: Anand Kumar
Thanks,
Anand Kumar
On 5/23/17, 10:59 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
The removal is mandatory to use the VStudio 2013 static code analyzer.
The only function that was used from t
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 5/23/17, 10:59 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
Add two new build targets: 'Win8Analyze' and 'Win8.1Analyze'.
The new build targets have the static code analyzer (buil
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 5/23/17, 10:59 AM, "ovs-dev-boun...@openvswitch.org on behalf of Alin
Serdean" wrote:
This patch adds the new make target 'datapath_windows_analyze' (static
analysis over the windows datapath code) to the appveyor bu
- Minimum valid fragment size is 400 bytes, any fragment smaller
is likely to be intentionally crafted.
- Validate maximum length of an Ip datagram
- Added counters to keep track of number of fragments for a given
Ip datagram.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Actions.c
Thanks for refactoring it.
Acked-by: Anand Kumar
Thanks,
Anand Kumar
On 7/24/17, 3:31 PM, "ovs-dev-boun...@openvswitch.org on behalf of Shashank
Ram" wrote:
Previously, the function would take the curNbl and nextNbl
as inputs, and modify the linked list, and merge
once.
Testing Done:
- Verified that ICMP requests are no longer treated as replies
in Conntrack.
Change-Id: I826a164cfb9137e2167c404ff5c9bfd9dfaa33ad
Co-authored-by: Sairam Venugopal
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack.c | 19 ++-
1 file change
issue by not modifying icmp_id field during
SNAT/DNAT only for ICMP traffic
The current NAT module doesn't take the ICMP type/id/code into account
during the lookups. Fix this to make it similar with the other conntrack
module.
Signed-off-by: Anand Kumar
---
datapath-windows/ovsext/Conntrack-
- Update the CT comparison function to compare individual fields instead of
NdisEqualMemory.
- Add in some padding for the ct_endpoint's union.
- Update the Orig Tuple to use ICMP Type and Code instead of Port for ICMP
Co-authored-by: Sairam Venugopal
Signed-off-by: Anand Kumar
---
dat
Hi Alin,
Thanks for reviewing the patch. Please find my responses inline.
Will send out a v2 patch addressing the review comments.
Thanks,
Anand Kumar
On 8/14/17, 8:14 PM, "Alin Serdean" wrote:
We should revisit how we do hashes and compares over the 'OVS_CT_KEY&
Hi Alin,
Thanks for the review. Please find my responses inline.
Will address the review comments and send out v2 version of the patch.
Thanks,
Anand Kumar
On 8/14/17, 8:04 PM, "Alin Serdean" wrote:
Hi Sai and Anand,
Thanks a lot for the patch. I have a few questions
1 - 100 of 207 matches
Mail list logo