netdev - Re: [PATCH v2 0/4] cpumask: improve on cpumask_local

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 15 Nov 2022 10:32:31 -0800
From:   Yury Norov <yury.norov@...il.com>
To:     Valentin Schneider <vschneid@...hat.com>
Cc:     linux-kernel@...r.kernel.org,
        "David S. Miller" <davem@...emloft.net>,
        Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
        Barry Song <baohua@...nel.org>,
        Ben Segall <bsegall@...gle.com>,
        haniel Bristot de Oliveira <bristot@...hat.com>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Gal Pressman <gal@...dia.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Heiko Carstens <hca@...ux.ibm.com>,
        Ingo Molnar <mingo@...hat.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Jason Gunthorpe <jgg@...dia.com>,
        Jesse Brandeburg <jesse.brandeburg@...el.com>,
        Jonathan Cameron <Jonathan.Cameron@...wei.com>,
        Juri Lelli <juri.lelli@...hat.com>,
        Leon Romanovsky <leonro@...dia.com>,
        Mel Gorman <mgorman@...e.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Rasmus Villemoes <linux@...musvillemoes.dk>,
        Saeed Mahameed <saeedm@...dia.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Tariq Toukan <tariqt@...dia.com>,
        Tariq Toukan <ttoukan.linux@...il.com>,
        Tony Luck <tony.luck@...el.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        linux-crypto@...r.kernel.org, netdev@...r.kernel.org,
        linux-rdma@...r.kernel.org
Subject: Re: [PATCH v2 0/4] cpumask: improve on cpumask_local_spread()
 locality

On Tue, Nov 15, 2022 at 05:24:56PM +0000, Valentin Schneider wrote:
> Hi,
> 
> On 12/11/22 11:09, Yury Norov wrote:
> > cpumask_local_spread() currently checks local node for presence of i'th
> > CPU, and then if it finds nothing makes a flat search among all non-local
> > CPUs. We can do it better by checking CPUs per NUMA hops.
> >
> > This series is inspired by Tariq Toukan and Valentin Schneider's "net/mlx5e:
> > Improve remote NUMA preferences used for the IRQ affinity hints"
> >
> > https://patchwork.kernel.org/project/netdevbpf/patch/20220728191203.4055-3-tariqt@nvidia.com/
> >
> > According to their measurements, for mlx5e:
> >
> >         Bottleneck in RX side is released, reached linerate (~1.8x speedup).
> >         ~30% less cpu util on TX.
> >
> > This patch makes cpumask_local_spread() traversing CPUs based on NUMA
> > distance, just as well, and I expect comparabale improvement for its
> > users, as in case of mlx5e.
> >
> > I tested new behavior on my VM with the following NUMA configuration:
> >
> > root@...ian:~# numactl -H
> > available: 4 nodes (0-3)
> > node 0 cpus: 0 1 2 3
> > node 0 size: 3869 MB
> > node 0 free: 3740 MB
> > node 1 cpus: 4 5
> > node 1 size: 1969 MB
> > node 1 free: 1937 MB
> > node 2 cpus: 6 7
> > node 2 size: 1967 MB
> > node 2 free: 1873 MB
> > node 3 cpus: 8 9 10 11 12 13 14 15
> > node 3 size: 7842 MB
> > node 3 free: 7723 MB
> > node distances:
> > node   0   1   2   3
> >   0:  10  50  30  70
> >   1:  50  10  70  30
> >   2:  30  70  10  50
> >   3:  70  30  50  10
> >
> > And the cpumask_local_spread() for each node and offset traversing looks
> > like this:
> >
> > node 0:   0   1   2   3   6   7   4   5   8   9  10  11  12  13  14  15
> > node 1:   4   5   8   9  10  11  12  13  14  15   0   1   2   3   6   7
> > node 2:   6   7   0   1   2   3   8   9  10  11  12  13  14  15   4   5
> > node 3:   8   9  10  11  12  13  14  15   4   5   6   7   0   1   2   3
> >
> 
> Is this meant as a replacement for [1]?

No. Your series adds an iterator, and in my experience the code that
uses iterators of that sort is almost always better and easier to
understand than cpumask_nth() or cpumask_next()-like users.

My series has the only advantage that it allows keep existing codebase
untouched.
 
> I like that this is changing an existing interface so that all current
> users directly benefit from the change. Now, about half of the users of
> cpumask_local_spread() use it in a loop with incremental @i parameter,
> which makes the repeated bsearch a bit of a shame, but then I'm tempted to
> say the first point makes it worth it.
> 
> [1]: https://lore.kernel.org/all/20221028164959.1367250-1-vschneid@redhat.com/

In terms of very common case of sequential invocation of local_spread()
for cpus from 0 to nr_cpu_ids, the complexity of my approach is n * log n,
and your approach is amortized O(n), which is better. Not a big deal _now_,
as you mentioned in the other email. But we never know how things will
evolve, right?

So, I would take both and maybe in comment to cpumask_local_spread()
mention that there's a better alternative for those who call the
function for all CPUs incrementally.

Thanks,
Yury