From: Hui Su <sh_def@163.com>
To: elver@google.com, dvyukov@google.com
Cc: kasan-dev@googlegroups.com, corbet@lwn.net,
skhan@linuxfoundation.org, rdunlap@infradead.org,
workflows@vger.kernel.org, linux-doc@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: [RFC PATCH 4/4] docs: kcsan: document targeted watchpoint sampling
Date: Wed, 14 Oct 2026 18:52:38 +0900 [thread overview]
Message-ID: <20261014095238.2132580-5-sh_def@163.com> (raw)
Describe the KCSAN debugfs target interface, including address and kallsyms
symbol forms, module symbols, disabling the target, and how it relates to
report filters. Clarify CONFIG_KALLSYMS_ALL requirements and that a
symbol target selects only the symbol start address. Warn about
hot-target overhead and clear the target before unloading a module that
owns the address.
Signed-off-by: Hui Su <sh_def@163.com>
---
Documentation/dev-tools/kcsan.rst | 17 +++++++++++++++++
1 file changed, 17 insertions(+)
diff --git a/Documentation/dev-tools/kcsan.rst b/Documentation/dev-tools/kcsan.rst
index 8575178aa87f..53cc3e5bc1f0 100644
--- a/Documentation/dev-tools/kcsan.rst
+++ b/Documentation/dev-tools/kcsan.rst
@@ -169,6 +169,23 @@ The file ``/sys/kernel/debug/kcsan`` provides the following interface:
can be used to silence frequently occurring data races; the whitelist feature
can help with reproduction and testing of fixes.
+* Writing ``target=<address>`` or ``target=<symbol>`` to
+ ``/sys/kernel/debug/kcsan`` enables targeted watchpoint sampling for one
+ address. An instrumented access is eligible for watchpoint setup when its
+ byte range contains the selected address. While targeted sampling is active,
+ non-target accesses do not consume the per-CPU ``kcsan.skip_watch`` interval.
+ Symbol targets are resolved using kallsyms. Naming vmlinux data symbols
+ requires ``CONFIG_KALLSYMS_ALL``, and module symbols use ``module:symbol``.
+ A symbol target selects the symbol's start address, not the full object.
+ Marked atomic accesses remain ineligible for watchpoint setup. This controls
+ watchpoint setup and is independent of report whitelist/blacklist filtering.
+ Write ``target=off`` to restore normal sampling.
+ Matching accesses bypass the normal skip interval, so targeting a hot
+ address can significantly increase KCSAN overhead. Use targeted sampling to
+ focus debugging on a known suspicious location.
+ The target does not hold a module reference; clear it before unloading a
+ module that owns the selected address.
+
Tuning performance
~~~~~~~~~~~~~~~~~~
--
2.55.0
reply other threads:[~2026-10-14 9:53 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
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=20261014095238.2132580-5-sh_def@163.com \
--to=sh_def@163.com \
--cc=corbet@lwn.net \
--cc=dvyukov@google.com \
--cc=elver@google.com \
--cc=kasan-dev@googlegroups.com \
--cc=linux-doc@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=rdunlap@infradead.org \
--cc=skhan@linuxfoundation.org \
--cc=workflows@vger.kernel.org \
/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