linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Bobby Eshleman <bobbyeshleman@gmail.com>
To: "David S. Miller" <davem@davemloft.net>,
	 Eric Dumazet <edumazet@google.com>,
	Jakub Kicinski <kuba@kernel.org>,
	 Paolo Abeni <pabeni@redhat.com>, Simon Horman <horms@kernel.org>,
	 Kuniyuki Iwashima <kuniyu@google.com>,
	 Willem de Bruijn <willemb@google.com>,
	Neal Cardwell <ncardwell@google.com>,
	 David Ahern <dsahern@kernel.org>,
	Mina Almasry <almasrymina@google.com>
Cc: Stanislav Fomichev <sdf@fomichev.me>,
	netdev@vger.kernel.org,  linux-kernel@vger.kernel.org,
	Bobby Eshleman <bobbyeshleman@meta.com>
Subject: [PATCH net-next v5 0/4] net: devmem: improve cpu cost of RX token management
Date: Thu, 23 Oct 2025 13:58:19 -0700	[thread overview]
Message-ID: <20251023-scratch-bobbyeshleman-devmem-tcp-token-upstream-v5-0-47cb85f5259e@meta.com> (raw)

This series improves the CPU cost of RX token management by replacing
the xarray allocator with an niov array and a uref field in niov.

Improvement is ~13% cpu util per RX user thread.
    
Using kperf, the following results were observed:

Before:
	Average RX worker idle %: 13.13, flows 4, test runs 11
After:
	Average RX worker idle %: 26.32, flows 4, test runs 11

Two other approaches were tested, but with no improvement. Namely, 1)
using a hashmap for tokens and 2) keeping an xarray of atomic counters
but using RCU so that the hotpath could be mostly lockless. Neither of
these approaches proved better than the simple array in terms of CPU.

The sysfs /proc/sys/net/core/devmem_autorelease is added to opt-out of
the optimization, but give users the performance gain by default.

Note that prior revs reported only a 5% gain. This lower gain was
measured with cpu frequency boosting (unknowingly) disabled. A
consistent ~13% is measured for both kperf and nccl workloads with cpu
frequency boosting on.

Signed-off-by: Bobby Eshleman <bobbyeshleman@meta.com>
---
Changes in v5:
- add sysctl to opt-out of performance benefit, back to old token release
- Link to v4: https://lore.kernel.org/all/20250926-scratch-bobbyeshleman-devmem-tcp-token-upstream-v4-0-39156563c3ea@meta.com

Changes in v4:
- rebase to net-next
- Link to v3: https://lore.kernel.org/r/20250926-scratch-bobbyeshleman-devmem-tcp-token-upstream-v3-0-084b46bda88f@meta.com

Changes in v3:
- make urefs per-binding instead of per-socket, reducing memory
  footprint
- fallback to cleaning up references in dmabuf unbind if socket
  leaked tokens
- drop ethtool patch
- Link to v2: https://lore.kernel.org/r/20250911-scratch-bobbyeshleman-devmem-tcp-token-upstream-v2-0-c80d735bd453@meta.com

Changes in v2:
- net: ethtool: prevent user from breaking devmem single-binding rule
  (Mina)
- pre-assign niovs in binding->vec for RX case (Mina)
- remove WARNs on invalid user input (Mina)
- remove extraneous binding ref get (Mina)
- remove WARN for changed binding (Mina)
- always use GFP_ZERO for binding->vec (Mina)
- fix length of alloc for urefs
- use atomic_set(, 0) to initialize sk_user_frags.urefs
- Link to v1: https://lore.kernel.org/r/20250902-scratch-bobbyeshleman-devmem-tcp-token-upstream-v1-0-d946169b5550@meta.com

---
Bobby Eshleman (4):
      net: devmem: rename tx_vec to vec in dmabuf binding
      net: devmem: refactor sock_devmem_dontneed for autorelease split
      net: devmem: use niov array for token management
      net: add per-netns sysctl for devmem autorelease

 include/net/netmem.h       |   1 +
 include/net/netns/core.h   |   1 +
 include/net/sock.h         |   8 +++-
 net/core/devmem.c          |  57 +++++++++++++++-------
 net/core/devmem.h          |  13 ++++-
 net/core/net_namespace.c   |   1 +
 net/core/sock.c            | 115 +++++++++++++++++++++++++++++++++++++--------
 net/core/sysctl_net_core.c |   9 ++++
 net/ipv4/tcp.c             |  69 ++++++++++++++++++++-------
 net/ipv4/tcp_ipv4.c        |  12 +++--
 net/ipv4/tcp_minisocks.c   |   3 +-
 11 files changed, 229 insertions(+), 60 deletions(-)
---
base-commit: 61b7ade9ba8c3b16867e25411b5f7cf1abe35879
change-id: 20250829-scratch-bobbyeshleman-devmem-tcp-token-upstream-292be174d503

Best regards,
-- 
Bobby Eshleman <bobbyeshleman@meta.com>


             reply	other threads:[~2025-10-23 21:00 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-10-23 20:58 Bobby Eshleman [this message]
2025-10-23 20:58 ` [PATCH net-next v5 1/4] net: devmem: rename tx_vec to vec in dmabuf binding Bobby Eshleman
2025-10-23 20:58 ` [PATCH net-next v5 2/4] net: devmem: refactor sock_devmem_dontneed for autorelease split Bobby Eshleman
2025-10-23 20:58 ` [PATCH net-next v5 3/4] net: devmem: use niov array for token management Bobby Eshleman
2025-10-23 20:58 ` [PATCH net-next v5 4/4] net: add per-netns sysctl for devmem autorelease Bobby Eshleman

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20251023-scratch-bobbyeshleman-devmem-tcp-token-upstream-v5-0-47cb85f5259e@meta.com \
    --to=bobbyeshleman@gmail.com \
    --cc=almasrymina@google.com \
    --cc=bobbyeshleman@meta.com \
    --cc=davem@davemloft.net \
    --cc=dsahern@kernel.org \
    --cc=edumazet@google.com \
    --cc=horms@kernel.org \
    --cc=kuba@kernel.org \
    --cc=kuniyu@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=ncardwell@google.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=sdf@fomichev.me \
    --cc=willemb@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).