[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <80b08b65-804b-2c83-c953-67def27ee656@ovn.org>
Date: Wed, 11 Oct 2023 14:22:40 +0200
From: Ilya Maximets <i.maximets@....org>
To: Nicholas Piggin <npiggin@...il.com>, netdev@...r.kernel.org
Cc: dev@...nvswitch.org, Pravin B Shelar <pshelar@....org>,
Aaron Conole <aconole@...hat.com>, Eelco Chaudron <echaudro@...hat.com>,
Flavio Leitner <fbl@...hat.com>, i.maximets@....org,
Simon Horman <horms@....org>
Subject: Re: [PATCH 0/7] net: openvswitch: Reduce stack usage
On 10/11/23 05:43, Nicholas Piggin wrote:
> Hi,
>
> I'll post this out again to keep discussion going. Thanks all for the
> testing and comments so far.
Hi, Nicholas. This patch set still needs performance evaluation
since it touches very performance-sensitive parts of the stack.
Did you run any performance tests with this version?
IIRC, Aaron was still working on testing for the RFC. I think,
we should wait for his feedback before re-spinning a new version.
>
> Changes since the RFC
> https://lore.kernel.org/netdev/20230927001308.749910-1-npiggin@gmail.com/
>
> - Replace slab allocations for flow keys with expanding the use
> of the per-CPU key allocator to ovs_vport_receive.
While this is likely to work faster than a dynamic memory allocation,
it is unlikley to be on par with a stack allocation. Performance
evaluation is necessary.
>
> - Drop patch 1 with Ilya's since they did the same thing (that is
> added at patch 3).
The patch is already in net-next, so should not be included in this set.
For the next version (please, hold) please rebase the set on the
net-next/main and add the net-next to the subject prefix of the patches.
They are not simple bug fixes, so should go through net-next, IMO.
You may also see in netdev+bpf patchwork that CI failed trying to guess
on which tree the patches should be applied and no tests were executed.
>
> - Change push_nsh stack reduction from slab allocation to per-cpu
> buffer.
I still think this change is not needed and will only consume a lot
of per-CPU memory space for no reason, as NSH is not a frequently
used thing in OVS and the function is not on the recursive path and
explicitly not inlined already.
Best regards, Ilya Maximets.
P.S. Please use my ovn.org email instead.
>
> - Drop the ovs_fragment stack usage reduction for now sinc it used
> slab and was a bit more complicated.
>
> I posted an initial version of the per-cpu flow allocator patch in
> the RFC thread. Since then I cleaned up some debug code and increased
> the allocator size to accommodate the additional user of it.
>
> Thanks,
> Nick
>
> Ilya Maximets (1):
> openvswitch: reduce stack usage in do_execute_actions
>
> Nicholas Piggin (6):
> net: openvswitch: generalise the per-cpu flow key allocation stack
> net: openvswitch: Use flow key allocator in ovs_vport_receive
> net: openvswitch: Reduce push_nsh stack usage
> net: openvswitch: uninline action execution
> net: openvswitch: uninline ovs_fragment to control stack usage
> net: openvswitch: Reduce stack usage in ovs_dp_process_packet
>
> net/openvswitch/actions.c | 208 +++++++++++++++++++++++--------------
> net/openvswitch/datapath.c | 56 +++++-----
> net/openvswitch/flow.h | 3 +
> net/openvswitch/vport.c | 27 +++--
> 4 files changed, 185 insertions(+), 109 deletions(-)
>
Powered by blists - more mailing lists