[ovs-dev] [PATCH] mpls: Fix MPLS restoration after patch port and group bucket.

Morin <thomas.mo...@orange.com> Suggested-by: Takashi YAMAMOTO <yamam...@ovn.org> Fixes: 8bfd0fdac ("Enhance userspace support for MPLS, for up to 3 labels.") Fixes: 1b035ef20 ("mpls: Allow l3 and l4 actions to prior to a push_mpls action") Signed-off-by: Jarno Ra

Re: [ovs-dev] assertion failing in commit_set_ipv4_action(), flow->nw_proto != base_flow->nw_proto

> On Nov 30, 2016, at 8:50 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Wed, Nov 30, 2016 at 06:58:57PM -0800, Jarno Rajahalme wrote: >> >>> On Nov 30, 2016, at 8:41 AM, Ben Pfaff <b...@ovn.org> wrote: >>> >>> On Wed, Nov 30, 201

Re: [ovs-dev] [PATCH] ofproto: Support reset_counts on flow mod message.

> On Nov 29, 2016, at 3:46 PM, Tony van der Peet > wrote: > > If the reset_counts flag is set on a flow modification message, the > flow counters must be cleared, even if the flow does not already have > the reset_counts flag set. And the flow modification

Re: [ovs-dev] [PATCH net-next v2] ofproto: Honor OFPFF_RESET_COUNTS flag in flow modify message.

> On Nov 30, 2016, at 9:06 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Wed, Nov 30, 2016 at 05:51:12PM -0800, Jarno Rajahalme wrote: >> While a flow modify must keep the original flow's flags, it must reset >> counts if (and only if) the reset_counts flag is present

Re: [ovs-dev] assertion failing in commit_set_ipv4_action(), flow->nw_proto != base_flow->nw_proto

> On Nov 30, 2016, at 8:41 AM, Ben Pfaff wrote: > > On Wed, Nov 30, 2016 at 12:05:46PM +0100, Thomas Morin wrote: >> Hi Ben, >> >> 2016-11-30, Ben Pfaff: >>> Do you have any idea what in your OpenFlow pipeline might do that, >>> i.e. is there anything especially tricky in the

Re: [ovs-dev] relationship between dpcls_rule and dp_netdev_flow

> On Nov 15, 2016, at 3:35 PM, Joo Kim wrote: > > Hello, > > In this OVS (2.6) code below, > It seems that a dp_netdev_flow obj contains a dpcls_rule struct, and given > a dpcls_rule, it gets the corresponding dp_netdev_flow via > dp_netdev_flow_cast(). > Does it mean

Re: [ovs-dev] Request to work on the Extensions - EXT-320 / EXT-46 : OF Version 1.5 ref.

> On Oct 26, 2016, at 12:08 AM, Pratyushaw P/HYD/TCS > wrote: > > Hello, > > It's a pleasure to contribute to the to the OF version 1.5 Extensions for the > upcoming releases. > > We have planned and picked up few of the below Extensions from the OpenFlow > 1.5

Re: [ovs-dev] [PATCH 2/2] configure: Support compiling with Linux 4.8.

> On Oct 20, 2016, at 2:55 PM, Pravin Shelar <pshe...@ovn.org> wrote: > > On Tue, Oct 18, 2016 at 5:01 PM, Jarno Rajahalme <ja...@ovn.org> wrote: >> Datapath should now compile and work with Linux 4.8. >> >> Signed-off-by: Jarno Rajahalme <ja...@

Re: [ovs-dev] [PATCH 1/2] datapath: Support a fixed size of 128 distinct labels.

> On Oct 20, 2016, at 2:55 PM, Pravin Shelar <pshe...@ovn.org> wrote: > > On Tue, Oct 18, 2016 at 5:01 PM, Jarno Rajahalme <ja...@ovn.org > <mailto:ja...@ovn.org>> wrote: >> Port upstream change in conntrack labels extension. Add a new >> conf

Re: [ovs-dev] [PATCH v3 07/12] dpif-netdev: Cache align netdev_flow_keys.

> On Oct 19, 2016, at 10:01 AM, Joe Stringer wrote: > > On 14 October 2016 at 07:37, Bhanuprakash Bodireddy > wrote: >> Aligning the 'keys' array seems to have positive performance impact. >> >> Signed-off-by: Bhanuprakash Bodireddy

Re: [ovs-dev] [PATCH] Makefile.am: Add DocumentationStyle.rst to distribution.

> On Oct 19, 2016, at 7:25 AM, Russell Bryant <russ...@ovn.org> wrote: > > > > On Tue, Oct 18, 2016 at 7:59 PM, Jarno Rajahalme <ja...@ovn.org > <mailto:ja...@ovn.org>> wrote: > Compile fails otherwise. > > Signed-off-by: Jarno R

[ovs-dev] [PATCH 2/2] configure: Support compiling with Linux 4.8.

Datapath should now compile and work with Linux 4.8. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- acinclude.m4 | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/acinclude.m4 b/acinclude.m4 index a3c95f5..2c500a8 100644 --- a/acinclude.m4 +++ b/acinclude.m4 @@ -

[ovs-dev] [PATCH 1/2] datapath: Support a fixed size of 128 distinct labels.

r.org> Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- acinclude.m4 | 2 ++ datapath/conntrack.c | 14 -- 2 files changed, 14 insertions(+), 2 deletions(-) diff --git a/acinclude.m4 b/acinclude.m4 index 353519d..a3c95f5 100644 --- a/acinclude.m4 +++ b/acinclude.m4 @

[ovs-dev] [PATCH] Makefile.am: Add DocumentationStyle.rst to distribution.

Compile fails otherwise. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- Makefile.am | 1 + 1 file changed, 1 insertion(+) diff --git a/Makefile.am b/Makefile.am index a4842c1..dc74886 100644 --- a/Makefile.am +++ b/Makefile.am @@ -69,6 +69,7 @@ docs = \ CONTRIBUT

Re: [ovs-dev] [PATCH v3 01/12] dpcls: Use 32 packet batches for lookups.

if (cnt != NETDEV_MAX_BURST) { > +keys_map >>= NETDEV_MAX_BURST - cnt; /* Clear extra bits. */ ‘keys_maps’ has MAP_BITS bits set, and after this is should only have the ‘cnt’ LSBs set. The assert above allows NETDEV_MAX_BURST to be less than MAP_BITS, hence MAP_BITS must be used here

Re: [ovs-dev] [PATCH v3 02/12] flow: Add comments to mf_get_next_in_map().

to mf_get_next_in_map() to make it more >> comprehensible. >> >> Signed-off-by: Jarno Rajahalme <ja...@ovn.org <mailto:ja...@ovn.org>> >> Acked-by: Bhanuprakash Bodireddy <bhanuprakash.bodire...@intel.com >> <mailto:bhanuprakash.bodire...@intel.com&g

Re: [ovs-dev] [PATCH v3 03/12] flow: Skip invoking expensive count_1bits() with zero input.

> On Oct 17, 2016, at 2:10 AM, Fischetti, Antonio <antonio.fische...@intel.com> > wrote: > > Thanks Jarno, one question below. > > Antonio > >> -Original Message- >> From: dev [mailto:dev-boun...@openvswitch.org >> <mailto:dev-boun...@o

Re: [ovs-dev] [PATCH v3 03/12] flow: Skip invoking expensive count_1bits() with zero input.

-by: Bhanuprakash Bodireddy <bhanuprakash.bodire...@intel.com> > Co-authored-by: Antonio Fischetti <antonio.fische...@intel.com> > Signed-off-by: Antonio Fischetti <antonio.fische...@intel.com> > Acked-by: Jarno Rajahalme <ja...@ovn.org> > --- > lib/flow.h | 15 ++

Re: [ovs-dev] [PATCH 06/13] netdev-dpdk: Do not abort if out of hugepage memory.

> On Oct 4, 2016, at 6:22 PM, Daniele Di Proietto > wrote: > > We can run out of hugepage memory coming from rte_*alloc() more easily > than heap coming from malloc(). > > Therefore: > > * We should use hugepage memory if we're going to access it only in > the slow

[ovs-dev] [PATCH] ofproto: Return the OFPC_BUNDLES bit in switch features reply.

Add definitions for the OpenFlow 1.5 specific capabilities bits OFPC15_BUNDLES and OFPC15_FLOW_MONITORING. Return the bundles capability bit in switch features reply. Reported-by: Andrej Leitner <andrej.leit...@pantheon.tech> Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- incl

Re: [ovs-dev] [PATCH] meta-flow: Fix the NXM_NX_* names of xxreg2 and xxreg3.

> On Oct 11, 2016, at 8:33 AM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Oct 10, 2016 at 11:38:21AM -0700, Jarno Rajahalme wrote: >> >>> On Oct 7, 2016, at 5:54 PM, Justin Pettit <jpet...@ovn.org> wrote: >>> >>> >>>>

Re: [ovs-dev] [PATCH 09/12] dpif: Reorder elements in dpif_upcall structure.

/* typedef MARKER */ void * cacheline0[0]; /* > 64 0 */ > > } > struct nlattr *key; > > . > . > } > > I tried to pack this hole by moving other members in to this space.

Re: [ovs-dev] [PATCH] meta-flow: Fix the NXM_NX_* names of xxreg2 and xxreg3.

> On Oct 7, 2016, at 5:54 PM, Justin Pettit <jpet...@ovn.org> wrote: > > >> On Oct 7, 2016, at 5:41 PM, Jarno Rajahalme <ja...@ovn.org> wrote: >> >> xxreg2 and xxreg3 had the same NXM_NX_* names as xxreg0 and xxreg1, >> correspondingly. >>

Re: [ovs-dev] [PATCH] meta-flow: Fix the NXM_NX_* names of xxreg2 and xxreg3.

Should we also reserve some range of NXM_NX numbers after 114 for potential future registers, say 115-118 at least? Other register types do not have this problem as they have their own classes of numbers. Jarno > On Oct 7, 2016, at 5:41 PM, Jarno Rajahalme <ja...@ovn.org> wrote: &g

[ovs-dev] [PATCH] meta-flow: Fix the NXM_NX_* names of xxreg2 and xxreg3.

xxreg2 and xxreg3 had the same NXM_NX_* names as xxreg0 and xxreg1, correspondingly. Found by inspection. CC: Justin Pettit <jpet...@ovn.org> Fixes: b23ada8eecfd ("Introduce 128-bit xxregs.") Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- include/openvswitch/meta-f

Re: [ovs-dev] [PATCH 12/12] timeval: Reorder elements in clock structure.

I would leave the ‘stopped’ member below the comment. Also, the 2nd cacheline is only ever accessed during unit tests, so this should not have real performance impact. Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > <bhan

Re: [ovs-dev] [PATCH 11/12] netlink-socket: Reorder elements in nl_dump structure.

Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > <bhanuprakash.bodire...@intel.com> wrote: > > By reordering the elements in nl_dump structure, pad bytes can be > reduced there by saving a cache line. > > Before

Re: [ovs-dev] [PATCH 10/12] ovsdb: Reorder elements in ovsdb_table_schema structure.

Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > <bhanuprakash.bodire...@intel.com> wrote: > > By reordering the elements in ovsdb_table_schema structure, pad bytes > can be reduced and also a cache line is saved. >

Re: [ovs-dev] [PATCH 09/12] dpif: Reorder elements in dpif_upcall structure.

CodingStyle.md instructs to group struct members into related groups. Also, changing the relative order of pointers should not make any difference. Could you achieve the same by reordering just the members above the ‘DPIF_UC_ACTION only.’ comment? Jarno > On Oct 7, 2016, at 9:17 AM,

Re: [ovs-dev] [PATCH 08/12] dpif-netdev: Reorder elements in dp_netdev_port structure.

Would equivalent packing be achieved by moving the line down before the bool instead? If yes, it would be preferable. Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > <bhanuprakash.bodire...@intel.com> wrote: > > By reor

Re: [ovs-dev] [PATCH 07/12] dpif-netdev: Cache align netdev_flow_keys.

> On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > > Aligning the 'keys' array seems to positively impact performance. > > Signed-off-by: Bhanuprakash Bodireddy > Signed-off-by: Antonio Fischetti

Re: [ovs-dev] [PATCH 06/12] cmap: Remove prefetching in cmap_find_batch().

> On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > > prefetching the data in to the caches isn't improving the performance in > cmap_find_batch(). Moreover its found that there is slight improvement > in performance with out prefetching. > I

Re: [ovs-dev] [PATCH 05/12] dpif-netdev: Clear flow batches inside packet_batch_execute.

Daniele had a comment on this, I believe? > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > > There is a slight negative performance impact, by zeroing out the flow > batch pointers in dp_netdev_input__ ahead of packet_batch_execute(). The >

Re: [ovs-dev] [PATCH 04/12] hash: Skip invoking mhash_add__() with zero input.

> On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > > mhash_add__() is expensive and should be only called with valid input. > This patch will validate the input before invoking the mhash_add__ and > there by saving some cpu cycles. > >

Re: [ovs-dev] [PATCH 03/12] flow: Skip invoking expensive count_1bits() with zero input.

With the nit below, Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > <bhanuprakash.bodire...@intel.com> wrote: > > This patch checks if trash is non-zero and only then resets the flowmap > bit and increment the p

Re: [ovs-dev] [PATCH 02/12] flow: Add comments to mf_get_next_in_map()

> On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > A brief commit message should be included here. E.g.: This patch adds comments to mf)get_next_in_map() to make it more comprehensible. > Signed-off-by: Bhanuprakash Bodireddy

Re: [ovs-dev] [PATCH 01/12] dpcls: Use 32 packet batches for lookups.

> On Oct 7, 2016, at 9:17 AM, Bhanuprakash Bodireddy > wrote: > > This patch increases the number of packets processed in a batch during a > lookup from 16 to 32. Processing batches of 32 packets improves > performance and also one of the internal loops can be

Re: [ovs-dev] [PATCH] ofproto: Always delete rules before deleting a meter.

> On Oct 4, 2016, at 2:32 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Fri, Sep 30, 2016 at 11:08:19AM -0700, Jarno Rajahalme wrote: >> When deleting a bridge it is currently possible to delete a mater >> without deleting the rules using the meter first. Fix this by

Re: [ovs-dev] [PATCH] Destroy rules before destroying meters

Petr, While deleting all the rules before deleting al the meters is more efficient, I think we should make the code more future proof by making sure we never delete a meter without deleting the rules referring to it. Would you try the following patch to see if it fixes the problem you face:

[ovs-dev] [PATCH] ofproto: Always delete rules before deleting a meter.

When deleting a bridge it is currently possible to delete a mater without deleting the rules using the meter first. Fix this by moving the meter's rule deletion to meter_delete(). Signed-off-by: Jarno Rajahalme <ja...@ovn.org> Reported-by: Petr Machata <pe...@mellanox.com> --- ofpro

Re: [ovs-dev] Deferring ofproto_class::destruct vs. ovs-appctl exit

bug in OVS 2.5(.1) that needs to be fixed, or if this is something that affects your development. Thanks, Jarno > On Sep 29, 2016, at 12:54 PM, Jarno Rajahalme <ja...@ovn.org> wrote: > > This may not be the cleanest solution, but how about changing the last line > of close_dpi

Re: [ovs-dev] Deferring ofproto_class::destruct vs. ovs-appctl exit

This may not be the cleanest solution, but how about changing the last line of close_dpif_backer() in ofproto/ofproto-dpif.c like this: diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c index 83dcc9c..5b42b7e 100644 --- a/ofproto/ofproto-dpif.c +++ b/ofproto/ofproto-dpif.c @@ -864,7

Re: [ovs-dev] [PATCH v2 3/3] ofp-actions: Always consider inconsistent CT actions as an error.

Thanks for the fast review, Joe! Series pushed to master, branch-2.6, and branch-2.5. Jarno > On Sep 27, 2016, at 2:38 PM, Joe Stringer <j...@ovn.org> wrote: > > On 27 September 2016 at 14:32, Jarno Rajahalme <ja...@ovn.org > <mailto:ja...@ovn.org>> wrote:

Re: [ovs-dev] [PATCH] system-traffic: Collapse FTP NAT tests.

Acked-by: Jarno Rajahalme <ja...@ovn.org> > On Sep 7, 2016, at 4:34 PM, Joe Stringer <j...@ovn.org> wrote: > > Previously we had the following tests: > * FTP with NAT > * FTP with NAT (seq-adj) > * FTP with NAT 2 > > Tests 1 and 2 share everything, except

[ovs-dev] [PATCH v2 1/3] ofp-actions: Style fixes.

Replace a tab by a space and remove an unnecessary variable. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- include/openvswitch/ofp-actions.h | 2 +- lib/ofp-actions.c | 8 +++- 2 files changed, 4 insertions(+), 6 deletions(-) diff --git a/include/openvswit

[ovs-dev] [PATCH v2 2/3] ofp-actions: Check that 'alg=ftp' matches on TCP.

Datapath flow setup fails when setting the FTP helper on an unsupported IP protocol. It is better to fail at the OpenFlow rule set-up time instead. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- lib/ofp-actions.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/l

[ovs-dev] [PATCH v2 3/3] ofp-actions: Always consider inconsistent CT actions as an error.

We can't downgrade to OF1.0 and expect inconsistent CT actions be silently disgarded. Instead, datapath flow install fails, so it is better to flag inconsistent CT actions as hard errors. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- lib/ofp-actions.c | 7 +-- tests/o

Re: [ovs-dev] [PATCH] ofp-actions: Error on conntrack action inconsistency.

> On Sep 27, 2016, at 11:06 AM, Joe Stringer <j...@ovn.org> wrote: > > On 26 September 2016 at 18:46, Jarno Rajahalme <ja...@ovn.org > <mailto:ja...@ovn.org>> wrote: >> Setting up a datapath flow that has a conntrack action with 'alg=ftp', >>

Re: [ovs-dev] 2.6 release update

> On Sep 27, 2016, at 8:36 AM, Ben Pfaff wrote: > > The previous plan was to release last Thursday. When I surveyed the > team at VMware, they said there were a few more bug fixes they wanted to > get into 2.6. I think that the last of these is this fix from Jarno: >

Re: [ovs-dev] Any updates on the "userspace meter" patches ?

> On Sep 27, 2016, at 4:57 AM, Enas Ahmad <enas.ah...@kaust.edu.sa> wrote: > > Hi, > I know that meters are still not implemented in the OVS master branch. > However, I would like to enquiry about the "userspace meter" patches > submitted a while ago by Jar

Re: [ovs-dev] [PATCH v2] upcall: Don't start new revalidation round too soon after the last one.

> On Sep 26, 2016, at 5:39 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Tue, Sep 20, 2016 at 11:42:45AM -0700, Jarno Rajahalme wrote: >> The execution time of 'ovs-ofctl add-flows' with a large number of >> flows can be more than halved if revalidators are not run

[ovs-dev] [PATCH] ofp-actions: Error on conntrack action inconsistency.

dependent. Correspondingly, when translating a conntrack action, do not issue any datapath actions if the packet being translated is not an IP packet, as conntrack can operate only on IP packets and a datapath flow set-up otherwise fails anyway. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- i

Re: [ovs-dev] [PATCH 3/3] seq: Add support for initial delay before waking up.

Thanks for the review. I sent a v2 with an implementation that just does the right thing (I hope) in the revalidator thread itself. Jarno > On Sep 19, 2016, at 10:32 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Fri, Sep 16, 2016 at 04:10:51PM -0700, Jarno Rajahalme wrote: >&g

Re: [ovs-dev] [PATCH 2/3] token-bucket: Add token_bucket_wait_at().

> On Sep 19, 2016, at 5:19 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Fri, Sep 16, 2016 at 04:10:50PM -0700, Jarno Rajahalme wrote: >> Having the caller of token_bucket_wait() indicated in the log messages >> makes debugging easier. >> >> Signed

Re: [ovs-dev] [PATCH 1/3] ofproto: Do not signal revalidation for group mods twice.

> On Sep 20, 2016, at 8:20 AM, Ben Pfaff <b...@ovn.org> wrote: > > On Fri, Sep 16, 2016 at 04:10:49PM -0700, Jarno Rajahalme wrote: >> The new group mod implementation signals revalidation through >> '->set_tables_version()', so the separate '->group

[ovs-dev] [PATCH v2] upcall: Don't start new revalidation round too soon after the last one.

core on a four core machine. In testing the 'ovs-ofctl add-flows' execution time is not significantly improved from this even if the revalidators are not notified about the flow table changes at all. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- ofproto/ofproto-dpif-upcall.

Re: [ovs-dev] [PATCH] ofproto-dpif-xlate: Fix memory leak in execute_controller_action.

> On Sep 20, 2016, at 8:35 AM, Ryan Moats wrote: > > commit df70a7731 ("ofproto-dpif-xlate: Allow translating > without side-effects.") created a memory leak by removing the > dp_packet_delete statement in execute_controller_action that > freed the earlier cloned packet.

Re: [ovs-dev] [PATCH v2 4/5] treewide: Use WC_MASK_FIELD() and FLOW_WC_GET_AND_MASK_WC().

Unfortunately this and the last patch of the series does not apply to master, so you’d need to rebase these. Jarno > On Aug 30, 2016, at 6:47 PM, Daniele Di Proietto > wrote: > > I think it makes the code more readable. No functional change. > > Signed-off-by:

Re: [ovs-dev] [PATCH v2 3/5] tnl-neigh-cache: Unwildcard flow members before inspecting them.

Acked-by: Jarno Rajahalme <ja...@ovn.org> IMO this and the earlier (2/5) one are bug fixes that should be cherry-picked to branches 2.6 and 2.5, if applicable. Jarno > On Aug 30, 2016, at 6:47 PM, Daniele Di Proietto <diproiet...@vmware.com> > wrote: > > t

Re: [ovs-dev] [PATCH v2 2/5] ofproto-dpif-xlate: Adjust generated mask for fragments.

Acked-by: Jarno Rajahalme <ja...@ovn.org> Jarno > On Aug 30, 2016, at 6:47 PM, Daniele Di Proietto <diproiet...@vmware.com> > wrote: > > It's possible to install an OpenFlow flow that matches on udp source and > destination ports without matching on fragme

[ovs-dev] [PATCH 3/3] seq: Add support for initial delay before waking up.

-ofctl add-flows' execution time is not significantly improved from this even if the revalidators are not notified about the flow table changes at all. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- lib/seq.c | 50 ++- lib

[ovs-dev] [PATCH 1/3] ofproto: Do not signal revalidation for group mods twice.

The new group mod implementation signals revalidation through '->set_tables_version()', so the separate '->group_modify()' is no longer needed. The ofproto-provider API is changed to allow 'group_modify' to be NULL. Fixes: 5d08a275cd ("ofproto: Make groups versioned.") Sign

[ovs-dev] [PATCH 2/3] token-bucket: Add token_bucket_wait_at().

Having the caller of token_bucket_wait() indicated in the log messages makes debugging easier. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- include/openvswitch/token-bucket.h | 5 - lib/token-bucket.c | 7 --- 2 files changed, 8 insertions(+), 4 deletions(-)

[ovs-dev] [PATCH 0/3] Fixes for OVS 2.6

These are fixes that would still be nice to get to OVS 2.6 release. The second patch is a nice to have, but is not durectly used by the other patches in the series. Jarno Rajahalme (3): ofproto: Do not signal revalidation for group mods twice. token-bucket: Add token_bucket_wait_at(). seq

Re: [ovs-dev] [PATCH v2] xlate: Use dp_hash for select groups.

Thu, Sep 15, 2016 at 06:43:39PM -0700, Jarno Rajahalme wrote: >> Thanks for the fix! >> >> While I was working with tightening the parsing, I found that I had earlier >> introduced a bug that crashes ovs-ofctl when a parsing error is found after >> parsin

Re: [ovs-dev] [PATCH v3 2/3] ofp-parse: Harden checking with group selection_method.

Thanks! Pushed to master and branch-2.6. Jarno > On Sep 15, 2016, at 7:28 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Thu, Sep 15, 2016 at 06:40:20PM -0700, Jarno Rajahalme wrote: >> Only allow fields when "selection_method=hash". Only allow >>

Re: [ovs-dev] [PATCH v3 1/3] ofp-parse: Remove double uninit of group mod if parsing fails.

Thanks for the review, pushed to master and branch-2.6. Jarno > On Sep 15, 2016, at 7:27 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Thu, Sep 15, 2016 at 06:40:19PM -0700, Jarno Rajahalme wrote: >> Double ofputil_uninit_group_mod() used to be harmless, but leads to >&g

Re: [ovs-dev] [PATCH v2] xlate: Use dp_hash for select groups.

t;>> generic resulting in less upcalls to userspace, but adds recirculation >>> prior to group selection. >>> >>> Signed-off-by: Jarno Rajahalme <ja...@ovn.org> >>> --- >>> v2: Rebase and documentation. >> >> Thanks for adding the

[ovs-dev] [PATCH v3 3/3] xlate: Use dp_hash for select groups.

r to group selection. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: Further improved documentation, and moved the NEWS piece to post-2.6. NEWS | 8 + lib/ofp-util.c | 16 -- ofproto/ofproto-d

[ovs-dev] [PATCH v3 1/3] ofp-parse: Remove double uninit of group mod if parsing fails.

Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: New patch for v3. lib/ofp-parse.c | 4 tests/ofproto-dpif.at | 4 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/lib/ofp-parse.c b/lib/ofp-parse.c index 0568fc7..92c4693 100644 --- a/lib/ofp-parse.c +++ b

[ovs-dev] [PATCH v3 2/3] ofp-parse: Harden checking with group selection_method.

Only allow fields when "selection_method=hash". Only allow selection_method_param when a non-nil selection_method is given. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: New patch for v3. lib/ofp-parse.c | 12 tests/ofproto-dpif.at| 10 +

Re: [ovs-dev] [PATCH 1/2] ofproto: Remove double reporting from bundles.

> On Sep 15, 2016, at 2:35 PM, Jarno Rajahalme <ja...@ovn.org> wrote: > >> >> On Sep 14, 2016, at 8:55 PM, Ben Pfaff <b...@ovn.org> wrote: >> >> On Wed, Sep 14, 2016 at 07:25:34PM -0700, Jarno Rajahalme wrote: >>> Patch b0d38b2f17 unified flo

Re: [ovs-dev] [PATCH v3 03/13] ofproto: Add a fixed bundle idle timeout of 10 seconds.

> On Sep 13, 2016, at 5:14 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Tue, Sep 13, 2016 at 03:01:42PM -0700, Jarno Rajahalme wrote: >> >>> On Sep 13, 2016, at 11:14 AM, Ben Pfaff <b...@ovn.org> wrote: >>> >>> On Mon, Sep 12, 2016 at 01:52:3

Re: [ovs-dev] [PATCH v3 01/13] ofproto: Don't use connmgr after destruction.

> On Sep 13, 2016, at 5:13 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Tue, Sep 13, 2016 at 02:45:54PM -0700, Jarno Rajahalme wrote: >> >>> On Sep 13, 2016, at 10:56 AM, Ben Pfaff <b...@ovn.org> wrote: >>> >>> On Mon, Sep 12, 2016 at 01:52

[ovs-dev] [PATCH] NEWS: Move bundle timeout to 2.6.

Move the news for bundle timeouts to OVS 2.6, as we decided to backport it there. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- NEWS | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/NEWS b/NEWS index 6cc01eb..343f7f1 100644 --- a/NEWS +++ b/NEWS @@ -4,8 +4,6 @

Re: [ovs-dev] [PATCH 2/2] NEWS: Memory efficiency improvements.

Pushed to master, 2.6 to follow shortly, thanks! Jarno > On Sep 14, 2016, at 8:55 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Wed, Sep 14, 2016 at 07:25:35PM -0700, Jarno Rajahalme wrote: >> Mention both flow table and bundle memory efficiency improvements. >> >&g

Re: [ovs-dev] [PATCH 1/2] ofproto: Remove double reporting from bundles.

> On Sep 14, 2016, at 8:55 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Wed, Sep 14, 2016 at 07:25:34PM -0700, Jarno Rajahalme wrote: >> Patch b0d38b2f17 unified flow mod reporting in ofproto for both >> stand-alone flow mods and bundle flow mods, but left b

[ovs-dev] [PATCH 1/2] ofproto: Remove double reporting from bundles.

("ofproto: Report flow mods also from bundles.") Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- ofproto/bundles.c | 20 +++- ofproto/bundles.h | 2 +- ofproto/connmgr.c | 4 ++-- ofproto/ofproto.c | 2 +- 4 files changed, 11 insertions(+), 17 deletions(-) diff

[ovs-dev] [PATCH 2/2] NEWS: Memory efficiency improvements.

Mention both flow table and bundle memory efficiency improvements. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- NEWS | 8 1 file changed, 8 insertions(+) diff --git a/NEWS b/NEWS index a1ca864..6cc01eb 100644 --- a/NEWS +++ b/NEWS @@ -29,6 +29,14 @@ v2.6.0 - xx xx

Re: [ovs-dev] [PATCH v3 13/13] ofproto: Support packet_outs in bundles.

Series now pushed to master, Jarno > On Sep 14, 2016, at 2:57 PM, Jarno Rajahalme <ja...@ovn.org> wrote: > >> >> On Sep 13, 2016, at 5:09 PM, Ben Pfaff <b...@ovn.org <mailto:b...@ovn.org>> >> wrote: >> >> On Mon, Sep 12, 2016 at 01

Re: [ovs-dev] [PATCH v3 13/13] ofproto: Support packet_outs in bundles.

> On Sep 13, 2016, at 5:09 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:43PM -0700, Jarno Rajahalme wrote: >> Add support for OFPT_PACKET_OUT messages in bundles. >> >> While ovs-ofctl already has a packet-out command, w

Re: [ovs-dev] [PATCH v3 12/13] ofproto: Refactor packet_out handling.

> On Sep 13, 2016, at 4:44 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:42PM -0700, Jarno Rajahalme wrote: >> Refactor handle_packet_out() to prepare for bundle support for packet >> outs in a later patch. >> >> Two new callbacks a

Re: [ovs-dev] [PATCH v3 09/13] ofproto: Use ofproto_flow_mod for learn execution from xlate cache.

> On Sep 13, 2016, at 1:04 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:39PM -0700, Jarno Rajahalme wrote: >> Use ofproto_flow_mod with a reference to an existing or new rule >> instead of ofputil_flow_mod for learn action execution from xlat

Re: [ovs-dev] [PATCH v3 10/13] ofproto-dpif-xlate: Allow translating without side-effects.

> On Sep 13, 2016, at 1:06 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:40PM -0700, Jarno Rajahalme wrote: >> Extend 'may_learn' attribute to also control the treatment of >> FIN_TIMEOUT action and asynchronous messages (packet ins, >

Re: [ovs-dev] [PATCH v3 07/13] ofproto-dpif-xlate: Expose xlate cache.

> On Sep 13, 2016, at 12:58 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:37PM -0700, Jarno Rajahalme wrote: >> Later patches will need to create xlate cache entries from different >> modules. This patch refactors the xlate cache code in

Re: [ovs-dev] [PATCH v3 05/13] connmgr: Make connmgr_wants_packet_in_on_miss() lock-free.

> On Sep 13, 2016, at 12:38 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:35PM -0700, Jarno Rajahalme wrote: >> Make connmgr_wants_packet_in_on_miss() use an atomic int instead of a >> list traversal taking the 'o

Re: [ovs-dev] [PATCH v3 03/13] ofproto: Add a fixed bundle idle timeout of 10 seconds.

> On Sep 13, 2016, at 11:14 AM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:33PM -0700, Jarno Rajahalme wrote: >> Timing out idle bundles frees memory that would effectively be leaked >> if a long standing OpenFlow connection would fail to co

Re: [ovs-dev] [PATCH v3 01/13] ofproto: Don't use connmgr after destruction.

> On Sep 13, 2016, at 10:56 AM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Sep 12, 2016 at 01:52:31PM -0700, Jarno Rajahalme wrote: >> Set ofproto's connmgr pointer to NULL after the connmgr has been >> destructed, and check for NULL when sending a flow removed >&

Re: [ovs-dev] [PATCH v2 10/15] ofproto-dpif-xlate: Add xlate cache type XC_TABLE.

> On Aug 29, 2016, at 2:57 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Aug 22, 2016 at 04:31:36PM -0700, Jarno Rajahalme wrote: >> Xlate cache entry type XC_TABLE is required for the table stats >> (number of misses and matches) to be correctly attributed. >>

Re: [ovs-dev] [PATCH v2 03/15] types.h: Move mirror_mask_t here.

> On Aug 29, 2016, at 2:11 PM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Aug 22, 2016 at 04:31:29PM -0700, Jarno Rajahalme wrote: >> Move mirror_mask_t from ofproto/ofproto-dpif-mirror.h to >> openvswitch/types.h to avoid including function definitions w

Re: [ovs-dev] [RFC PATCH] xlate: Use dp_hash for select groups.

I added a NEWS flash as well, Jarno > On Sep 12, 2016, at 3:53 PM, Ben Pfaff <b...@ovn.org> wrote: > > Apparently ;-) > > On Mon, Sep 12, 2016 at 02:59:16PM -0700, Jarno Rajahalme wrote: >> I’m rebasing this now. Adding ovs-ofctl documentation should be sufficien

[ovs-dev] [PATCH v2] xlate: Use dp_hash for select groups.

r to group selection. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v2: Rebase and documentation. NEWS | 8 + lib/ofp-parse.c | 2 ++ lib/ofp-util.c | 16 -- ofproto/ofproto-d

Re: [ovs-dev] [RFC PATCH] xlate: Use dp_hash for select groups.

I’m rebasing this now. Adding ovs-ofctl documentation should be sufficient, right? Jarno > On Apr 22, 2016, at 8:27 AM, Ben Pfaff <b...@ovn.org> wrote: > > On Mon, Apr 18, 2016 at 05:42:53PM -0700, Jarno Rajahalme wrote: >> Add a new select group selection method

[ovs-dev] [PATCH v3 12/13] ofproto: Refactor packet_out handling.

ut defers all OpenFlow-visible side-effects (stats, learn actions, actual packet output, etc.) to be explicitly executed with the ->packet_execute() call. Adds a new ofproto_rule_reduce_timeouts__() that must be called with 'ofproto_mutex' held. This is used in the next patch. Signed-off-by: Jar

[ovs-dev] [PATCH v3 13/13] ofproto: Support packet_outs in bundles.

bundle is successful. A failing packet-out translation may also make the whole bundle to fail. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: Rebase. NEWS| 1 + include/openvswitch/ofp-parse.h | 5 ++ include/openvswitch/ofp-util.h | 1 + l

[ovs-dev] [PATCH v3 01/13] ofproto: Don't use connmgr after destruction.

("ofproto: Postpone sending flow removed messages.") Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: New patch for v3. ofproto/connmgr.c | 16 ofproto/connmgr.h | 6 -- ofproto/fail-open.c| 8 +--- ofproto/fail-open.h

[ovs-dev] [PATCH v3 03/13] ofproto: Add a fixed bundle idle timeout of 10 seconds.

timeout to be 10 seconds. We do not limit the number of messages in a bundle, so it does not make sense to limit the number of bundles either, especially now that idle bundles are timed out. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: New patch for v3. NEWS

[ovs-dev] [PATCH v3 11/13] coverage: Rename init functions to avoid symbol collisions.

ofproto now uses various *_init() functions, so use something else for coverage constructors. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: Use '_init_coverage' suffix instead of '_init__'. lib/coverage.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git

[ovs-dev] [PATCH v3 10/13] ofproto-dpif-xlate: Allow translating without side-effects.

support for one-time asynchronous messages to xlate cache, so that all side-effects of the translation may be executed at a later stage. This will be useful for bundle commits. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: Rebase. ofproto/ofproto-dpif-xlate-cache.

[ovs-dev] [PATCH v3 04/13] ofproto: Change rule's 'removed' member to a tri-state 'state'.

As a rule may not be re-inserted to ofproto data structures, it is cleaner to have three states for the rule, rather than just two. This will be useful for managing learned flows in later patches. Signed-off-by: Jarno Rajahalme <ja...@ovn.org> --- v3: Do not change the locking requir

<    1   2   3   4   5   6   7   8   9   10   >