Linux Trace Kernel
 help / color / mirror / Atom feed
From: Kunhai Dai <daikunhai@didiglobal.com>
To: <tj@kernel.org>
Cc: <axboe@kernel.dk>, <josef@toxicpanda.com>, <yukuai@fygo.io>,
	<cui.tao@linux.dev>, <cgroups@vger.kernel.org>,
	<linux-block@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
	<rostedt@goodmis.org>, <mhiramat@kernel.org>,
	<mathieu.desnoyers@efficios.com>,
	<linux-trace-kernel@vger.kernel.org>
Subject: [PATCH v2 0/2] blk-iocost: fix hweight cache publication and donation fail-safe
Date: Mon, 19 Oct 2026 08:54:10 -0400	[thread overview]
Message-ID: <20261019125412.3123895-1-daikunhai@didiglobal.com> (raw)

This series fixes the blk-iocost hweight cache publication race and adds a
fail-safe for invalid donation weights.

Previous version:
  https://lists.openwall.net/linux-kernel/2026/10/15/1954
Replies:
  https://lists.openwall.net/linux-kernel/2026/10/16/1431
  https://lists.openwall.net/linux-kernel/2026/10/16/2798

Changes since v1:
- split into two patches: the donation fail-safe and the atomic cache;
- add the Fixes tags agreed on the list (e08d02aa5fc9, 7caa47151ab2) and
  Cc stable on the fail-safe;
- drop the donating > 1 guard, which only changed the value printed by the
  warning;
- do not add max(after, 1): the no-race path has not been shown to be
  reachable and the total budget invariant has not been re-validated;
- use GENMASK/FIELD_PREP/FIELD_GET helpers for the packed cache;
- have tracepoints decode the packed cache without refreshing it, and only
  evaluate that read when the tracepoint is enabled;
- explain the activation-time cache update;
- update iocost_monitor.py;
- document the AI-assisted work.

The no-race invalid-weight path is still under investigation: a hand-built
input can trigger the divide at the function level, but runtime
reachability has not been demonstrated.  The first patch is therefore a
fail-safe, not a claim that the path is reachable without a race.

Kunhai Dai (2):
  blk-iocost: abort weight transfer on invalid donation weights
  blk-iocost: publish hweights as one atomic value

 block/blk-iocost.c             | 111 +++++++++++++++++++++++++++------
 include/trace/events/iocost.h  |   7 ++-
 tools/cgroup/iocost_monitor.py |   7 ++-
 3 files changed, 101 insertions(+), 24 deletions(-)

-- 
2.27.0


             reply	other threads:[~2026-10-19 12:57 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-19 12:54 Kunhai Dai [this message]
2026-10-19 12:54 ` [PATCH v2 1/2] blk-iocost: abort weight transfer on invalid donation weights Kunhai Dai
2026-10-19 12:54 ` [PATCH v2 2/2] blk-iocost: publish hweights as one atomic value Kunhai Dai
2026-10-19 23:12 ` [PATCH v2 0/2] blk-iocost: fix hweight cache publication and donation fail-safe Tejun Heo
2026-10-20  9:16 ` Tao Cui

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=20261019125412.3123895-1-daikunhai@didiglobal.com \
    --to=daikunhai@didiglobal.com \
    --cc=axboe@kernel.dk \
    --cc=cgroups@vger.kernel.org \
    --cc=cui.tao@linux.dev \
    --cc=josef@toxicpanda.com \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-trace-kernel@vger.kernel.org \
    --cc=mathieu.desnoyers@efficios.com \
    --cc=mhiramat@kernel.org \
    --cc=rostedt@goodmis.org \
    --cc=tj@kernel.org \
    --cc=yukuai@fygo.io \
    /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