Re: [ovs-dev] [PATCH 2/2] openflow: Add extension to flush CT by generic match

On Mon, Dec 5, 2022 at 10:38 PM Ilya Maximets wrote: > On 12/1/22 14:19, Ales Musil wrote: > > + > > +static enum ofperr > > +handle_nxt_ct_flush(struct ofconn *ofconn, const struct ofp_header > *oh) > > +{ > > +struct ofproto *ofproto = ofconn_get_ofproto(ofconn); > >

Re: [ovs-dev] [PATCH 2/2] openflow: Add extension to flush CT by generic match

On 12/1/22 14:19, Ales Musil wrote: > + > +static enum ofperr > +handle_nxt_ct_flush(struct ofconn *ofconn, const struct ofp_header *oh) > +{ > +    struct ofproto *ofproto = ofconn_get_ofproto(ofconn); > +    struct ofputil_ct_match match = {0}; > +    uint16_t

Re: [ovs-dev] [RFC PATCH v2] dpdk: Update to use v22.11.

> > On Thu, Dec 1, 2022 at 3:24 PM Ilya Maximets wrote: > Frode, do you know the approximate timeline on when we could expect > >>> > >>> As explained above, without landing this ... never :-) > >>> But speaking of timeline - we'd prefer to get this resolved for > >>> Debian/Ubuntu before

[ovs-dev] [PATCH v4] dpdk: Update to use v22.11.1.

This commit add support to for DPDK v22.11.1, it includes the following changes. 1. ci: Reduce DPDK compilation time. 2. system-dpdk: Update vhost tests to be compatible with DPDK 22.07. http://patchwork.ozlabs.org/project/openvswitch/list/?series=316528 3. system-dpdk: Update vhost tests to

Re: [ovs-dev] [PATCH ovn 0/7] OVN IC bugfixes & proposals/questions

On 12/5/22 17:40, Dumitru Ceara wrote: > On 12/2/22 18:31, Vladislav Odintsov wrote: >> Hi, >> >> we’ve met with an issue, where it was possible to create multiple similar >> routes within LR (same ip_prefix, nexthop, and route table). >> >> Initially the problem stared after OVN upgrade. We use

Re: [ovs-dev] [PATCH v3] dpdk: Update to use v22.11.

> Hi Ian, > > On Wed, Nov 30, 2022 at 4:32 PM Ian Stokes wrote: > > > > This commit add support to for DPDK v22.11, it includes the following > > changes. > > > > 1. ci: Reduce DPDK compilation time. > > 2. system-dpdk: Update vhost tests to be compatible with DPDK 22.07. > > > >

Re: [ovs-dev] [PATCH ovn 6/7] northd: limit ECMP group by 1024 members

Saying more, the OVS register, which is used to store the bucket ID is a 16-bit reg8[16..31], which was introduced by Han. @Han, can you clarify if it was planned to have 2^16 ECMP paths within one group (route)? Or maybe this is not needed to have such a bit IDs space and we can give more

Re: [ovs-dev] [PATCH ovn 6/7] northd: limit ECMP group by 1024 members

It’s a good idea. But one thing is that this is not the only one place where the buckets are created. Also they’re created in LBs. Should we just put some common function, which returns the current configured (or default MAX) and use it in every place? Regards, Vladislav Odintsov > On 5 Dec

Re: [ovs-dev] [PATCH ovn 5/7] ic: minor code improvements

Regards, Vladislav Odintsov > On 5 Dec 2022, at 19:37, Dumitru Ceara wrote: > > On 12/2/22 18:31, Vladislav Odintsov wrote: >> 1. Remove excess nbrec_logical_router variable. >> 2. Remove excess call to add_static_to_routes_ad(). >> 3. Remove double route_table check in ic_route_fin(). > >

Re: [ovs-dev] [PATCH ovn 4/7] actions: limit possible OF group bucket count

Do you mean that northd should have some kind of common code, which is used to generate buckets? Initially I wanted to prevent the possibility even to send such a big OF group to OVS. For all the cases: - a new possible functionality which doesn’t limit buckets count; - some kind of northd bug.

Re: [ovs-dev] [PATCH ovn 1/7] ic: move routes_ad hmap insert to separate function

Hi, Okay, I’ll split these patches in two series and squash patch #1 with patch #7. Regards, Vladislav Odintsov > On 5 Dec 2022, at 20:00, Numan Siddique wrote: > > On Mon, Dec 5, 2022 at 11:37 AM Dumitru Ceara > wrote: >> >> On 12/2/22 18:31, Vladislav Odintsov

Re: [ovs-dev] [PATCH ovn 2/7] ic: remove orphan ovn interconnection routes

Hi Dumitru, please, see answer inline. Regards, Vladislav Odintsov > On 5 Dec 2022, at 19:37, Dumitru Ceara wrote: > > On 12/2/22 18:31, Vladislav Odintsov wrote: >> Before this patch if one deletes transit switch through which there were >> routes in ICSB:Route table, such routes were left

Re: [ovs-dev] [PATCH ovn 1/7] ic: move routes_ad hmap insert to separate function

On Mon, Dec 5, 2022 at 11:37 AM Dumitru Ceara wrote: > > On 12/2/22 18:31, Vladislav Odintsov wrote: > > This change will be useful in next commit. > > > > Signed-off-by: Vladislav Odintsov > > --- > > Hi Vladislav, > > This looks OK to me but I think I'd squash it in the patch that actually >

Re: [ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

On 12/5/22 16:22, Lorenzo Bianconi wrote: >> On 12/5/22 16:16, Lorenzo Bianconi wrote: For the case when multiple LBs (same VIP but different port) share the same subset of backends we need to differentiate between them by also matching on the L4 port. Without that affinity

Re: [ovs-dev] [PATCH ovn 0/7] OVN IC bugfixes & proposals/questions

On 12/2/22 18:31, Vladislav Odintsov wrote: > Hi, > > we’ve met with an issue, where it was possible to create multiple similar > routes within LR (same ip_prefix, nexthop, and route table). > > Initially the problem stared after OVN upgrade. We use python ovsdbapp > library, > and we found a

Re: [ovs-dev] [PATCH ovn 6/7] northd: limit ECMP group by 1024 members

On 12/2/22 18:31, Vladislav Odintsov wrote: > This patch is intended to show that currently it's possible to build > ECMP group of 65k buckets. > > Signed-off-by: Vladislav Odintsov > --- > northd/northd.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/northd/northd.c

Re: [ovs-dev] [PATCH ovn 5/7] ic: minor code improvements

On 12/2/22 18:31, Vladislav Odintsov wrote: > 1. Remove excess nbrec_logical_router variable. > 2. Remove excess call to add_static_to_routes_ad(). > 3. Remove double route_table check in ic_route_fin(). Nit: s/route_table/nexthop/ > 4. Move variable declarations out of loop. > > Signed-off-by:

Re: [ovs-dev] [PATCH ovn 4/7] actions: limit possible OF group bucket count

On 12/2/22 18:31, Vladislav Odintsov wrote: > It is possible to send OpenFlow group_mod message to OVS to create a > group with any number of buckets: > > ovs-ofctl dump-groups br-int > NXST_GROUP_DESC reply (xid=0x2): > >

Re: [ovs-dev] [PATCH ovn 3/7] ic: lookup southbound port_binding only if needed

On 12/2/22 18:31, Vladislav Odintsov wrote: > Signed-off-by: Vladislav Odintsov > --- Acked-by: Dumitru Ceara Thanks! ___ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Re: [ovs-dev] [PATCH ovn 1/7] ic: move routes_ad hmap insert to separate function

On 12/2/22 18:31, Vladislav Odintsov wrote: > This change will be useful in next commit. > > Signed-off-by: Vladislav Odintsov > --- Hi Vladislav, This looks OK to me but I think I'd squash it in the patch that actually uses the new way of calling ic_route_find(). Thanks, Dumitru >

Re: [ovs-dev] [PATCH ovn 2/7] ic: remove orphan ovn interconnection routes

On 12/2/22 18:31, Vladislav Odintsov wrote: > Before this patch if one deletes transit switch through which there were > routes in ICSB:Route table, such routes were left forever in the DB. > > Now we validate that each ICSB:Route has an appropriate transit switch. > > Signed-off-by: Vladislav

Re: [ovs-dev] [PATCH ovn] controller: Restore MAC and vlan for DVR scenario

On 9/30/22 12:43, Dumitru Ceara wrote: > On 9/20/22 22:18, Mark Michelson wrote: >> Thanks Ales, >> >> Acked-by: Mark Michelson >> > > I applied this to the main branch and backported it to all stable > branches down to branch-22.03. > Hi all, There was an internal request from within Red Hat

[ovs-dev] mlx5 rte_eth_dev_info.reta_size value

Hi Ori, While working on a patch for OvS[1], I have tried to reconfigure the redirection table using the code examples that are layout around in testpmd and other places. [1]: http://patchwork.ozlabs.org/project/openvswitch/patch/20221021145308.141933-1-rja...@redhat.com/ Here is a stripped

Re: [ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

> On 12/5/22 16:16, Lorenzo Bianconi wrote: > >> For the case when multiple LBs (same VIP but different port) share the > >> same subset of backends we need to differentiate between them by also > >> matching on the L4 port. Without that affinity configuration from one > >> load balancer might be

Re: [ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

Looks good to me, thanks. Acked-by: Ales Musil On Mon, Dec 5, 2022 at 2:04 PM Dumitru Ceara wrote: > For the case when multiple LBs (same VIP but different port) share the > same subset of backends we need to differentiate between them by also > matching on the L4 port. Without that affinity

Re: [ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

On 12/5/22 16:16, Lorenzo Bianconi wrote: >> For the case when multiple LBs (same VIP but different port) share the >> same subset of backends we need to differentiate between them by also >> matching on the L4 port. Without that affinity configuration from one >> load balancer might be

Re: [ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

> For the case when multiple LBs (same VIP but different port) share the > same subset of backends we need to differentiate between them by also > matching on the L4 port. Without that affinity configuration from one > load balancer might be incorrectly applied to another. > > Adapt the unit and

Re: [ovs-dev] [PATCH v2] rhel: move conf.db to /var/lib/openvswitch, using symlinks

On 12/4/22 09:23, Roi Dayan wrote: > > > On 30/11/2022 17:55, Ilya Maximets wrote: >> On 11/14/22 20:41, Timothy Redaelli wrote: >>> conf.db is by default at /etc/openvswitch, but it should be at >>> /var/lib/openvswitch like on Debian or like ovnnb_db.db and ovnsb_db.db. >>> >>> If conf.db

Re: [ovs-dev] [PATCH v3] dpdk: Update to use v22.11.

Hi Ian, On Wed, Nov 30, 2022 at 4:32 PM Ian Stokes wrote: > > This commit add support to for DPDK v22.11, it includes the following > changes. > > 1. ci: Reduce DPDK compilation time. > 2. system-dpdk: Update vhost tests to be compatible with DPDK 22.07. > >

Re: [ovs-dev] [PATCH v1 2/2] tests: add unit tests to rculist

On Mon, Dec 5, 2022 at 3:41 AM Adrian Moreno wrote: > > Low test coverage on this area caused some errors to remain unnoticed. > Add basic functional test of rculist. > > Signed-off-by: Adrian Moreno Looks good! Triggers a warning if the preceding patch hasn't been applied as intended.

Re: [ovs-dev] [PATCH v1 1/2] rculist: use rculist_back_protected to access prev

On Mon, Dec 5, 2022 at 3:41 AM Adrian Moreno wrote: > > The .prev member of a rculist should not be used directly by users > because it's not rcu-safe. A convenient fake mutex (rculist_fake_mutex) > helps ensuring that in conjunction with clang's thread safety > extensions. > > Only writers with

[ovs-dev] [PATCH ovn] northd: Include VIP port in LB affinity learn flow matches.

For the case when multiple LBs (same VIP but different port) share the same subset of backends we need to differentiate between them by also matching on the L4 port. Without that affinity configuration from one load balancer might be incorrectly applied to another. Adapt the unit and system

Re: [ovs-dev] [PATCH v1] ovs-ctl: Allow inclusion of hugepages in coredumps

On 12/5/22 13:46, David Marchand wrote: > On Fri, Dec 2, 2022 at 7:00 PM Mike Pattrick wrote: >>> Did you consider madvise()? >>> >>> MADV_DONTDUMP (since Linux 3.4) >>>Exclude from a core dump those pages in the range >>> specified by addr and length.

Re: [ovs-dev] [PATCH v1] ovs-ctl: Allow inclusion of hugepages in coredumps

On Fri, Dec 2, 2022 at 9:14 PM Mike Pattrick wrote: > > On Fri, Dec 2, 2022 at 1:40 PM Ilya Maximets wrote: > > > > On 12/2/22 18:59, Mike Pattrick wrote: > > > On Fri, Dec 2, 2022 at 11:59 AM Ilya Maximets wrote: > > >> > > >> On 12/2/22 11:36, Maxime Coquelin wrote: > > >>> > > >>> > > >>> On

Re: [ovs-dev] [PATCH v1] ovs-ctl: Allow inclusion of hugepages in coredumps

On 12/2/22 21:14, Mike Pattrick wrote: > On Fri, Dec 2, 2022 at 1:40 PM Ilya Maximets wrote: >> >> On 12/2/22 18:59, Mike Pattrick wrote: >>> On Fri, Dec 2, 2022 at 11:59 AM Ilya Maximets wrote: On 12/2/22 11:36, Maxime Coquelin wrote: > > > On 12/2/22 11:09, David Marchand

Re: [ovs-dev] [PATCH v1] ovs-ctl: Allow inclusion of hugepages in coredumps

On Fri, Dec 2, 2022 at 7:00 PM Mike Pattrick wrote: > > Did you consider madvise()? > > > > MADV_DONTDUMP (since Linux 3.4) > > Exclude from a core dump those pages in the range > > specified by addr and length. This is useful in applications that

Re: [ovs-dev] [v6] odp-execute: Add ISA implementation of set_masked IPv6 action

On 12/5/22 12:53, Finn, Emma wrote: > > >> -Original Message- >> From: Ilya Maximets >> Sent: Friday 2 December 2022 14:22 >> To: Finn, Emma ; d...@openvswitch.org >> Cc: i.maxim...@ovn.org; Van Haaren, Harry ; >> echau...@redhat.com; Stokes, Ian >> Subject: Re: [v6] odp-execute: Add

Re: [ovs-dev] [v6] odp-execute: Add ISA implementation of set_masked IPv6 action

> -Original Message- > From: Ilya Maximets > Sent: Friday 2 December 2022 14:22 > To: Finn, Emma ; d...@openvswitch.org > Cc: i.maxim...@ovn.org; Van Haaren, Harry ; > echau...@redhat.com; Stokes, Ian > Subject: Re: [v6] odp-execute: Add ISA implementation of set_masked IPv6 > action

[ovs-dev] [PATCH v1 2/2] tests: add unit tests to rculist

Low test coverage on this area caused some errors to remain unnoticed. Add basic functional test of rculist. Signed-off-by: Adrian Moreno --- tests/automake.mk| 1 + tests/library.at | 5 ++ tests/test-rculist.c | 205 +++ 3 files changed, 211

[ovs-dev] [PATCH v1 1/2] rculist: use rculist_back_protected to access prev

The .prev member of a rculist should not be used directly by users because it's not rcu-safe. A convenient fake mutex (rculist_fake_mutex) helps ensuring that in conjunction with clang's thread safety extensions. Only writers with exclusive access to the rculist should access .prev via some of