* [PATCH v2] coccinelle: locks: Add must_hold.cocci
@ 2026-11-20 9:26 Marco Elver
2026-11-20 18:14 ` Bart Van Assche
0 siblings, 1 reply; 2+ messages in thread
From: Marco Elver @ 2026-11-20 9:26 UTC (permalink / raw)
To: elver
Cc: Julia Lawall, Nicolas Palix, linux-kernel, cocci, kasan-dev,
Bart Van Assche
Clang's context analysis treats lockdep_assert_held*() as an assumption
that the lock is held. A function that asserts instead of being
annotated with __must_hold() therefore passes the analysis, but its
callers are never checked.
Add a script that reports functions which assert on all paths that a
lock is held, where the lock can be named in a function attribute, and
which are not annotated with __must_hold*(), __releases*(),
__context_unsafe() or __no_context_analysis. A lock can be named if all
identifiers in the lock expression are parameters or member names, or if
it is a global of the form &name or &name.member. Functions that mention
the lock before asserting it, e.g. to acquire it, are skipped.
lockdep_assert_held() passes for any mode, so for it the script cannot
tell __must_hold() from __must_hold_shared() and reports both.
For example, on v7.4-rc2:
kernel/locking/rwsem.c:396:1-20: rwsem_del_waiter() asserts &sem->wait_lock is held but lacks __must_hold(&sem->wait_lock) or __must_hold_shared(&sem->wait_lock)
kernel/futex/pi.c:919:1-20: fixup_pi_state_owner() asserts q->lock_ptr is held but lacks __must_hold(q->lock_ptr) or __must_hold_shared(q->lock_ptr)
drivers/nvme/host/fabrics.c:650:1-20: nvmf_lookup_transport() asserts &nvmf_transports_rwsem is held but lacks __must_hold(&nvmf_transports_rwsem) or __must_hold_shared(&nvmf_transports_rwsem)
Limitations:
- Annotations on a prototype in another file are not seen; non-static
functions are reported with "check prototype".
- Annotations are detected textually and are not matched against the
asserted lock.
- Functions that require a lock without asserting it are not found.
- Only lockdep_assert_held{,_write,_read,_once}() and
assert_{raw_,}spin_locked() are matched; wrappers such as
lockdep_assert_wiphy() are not.
- Functions that test the lock expression before the assertion, e.g.
__futex_unqueue(), are skipped along with those that acquire it.
- &name.member is assumed to be a global.
Cc: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/all/CANpmjNNM=qBJMvBkRu-WbhKA0hnkb_Bk8fyF9DTTsVmENDvZuQ@mail.gmail.com/
Signed-off-by: Marco Elver <elver@google.com>
---
v2:
* Report both __must_hold() and __must_hold_shared() for
lockdep_assert_held() (Bart).
* Fix detection of annotations on prototypes in the same file (Bart).
* Make "forall" explicit (Bart).
* Skip functions that mention the lock before asserting it (Bart).
* Require all identifiers in the lock expression to be parameters or
member names; accept &name.member as a global (Bart).
* Match parameter names and "static" in SmPL instead of with regular
expressions (Bart).
* Also match lockdep_assert_held_once() and assert_{raw_,}spin_locked()
(Bart).
* Document skipped attributes and further limitations.
---
scripts/coccinelle/locks/must_hold.cocci | 131 +++++++++++++++++++++++
1 file changed, 131 insertions(+)
create mode 100644 scripts/coccinelle/locks/must_hold.cocci
diff --git a/scripts/coccinelle/locks/must_hold.cocci b/scripts/coccinelle/locks/must_hold.cocci
new file mode 100644
index 000000000000..0acec2680aa6
--- /dev/null
+++ b/scripts/coccinelle/locks/must_hold.cocci
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/// Find functions that unconditionally assert that a lock is held, but lack
+/// the corresponding __must_hold() annotation. The compiler's context
+/// analysis does not check the callers of such functions.
+///
+/// Only locks that can be named in a function attribute are reported: either
+/// all identifiers in the lock expression are parameters or member names, or
+/// the lock is a global of the form &name or &name.member. Functions that
+/// mention the lock before asserting it (e.g. to acquire it) are skipped, as
+/// are functions annotated with __must_hold*(), __releases*(),
+/// __context_unsafe() or __no_context_analysis.
+//
+// Confidence: Moderate
+// Comments: Annotations on a prototype in another file are not seen; such
+// functions are reported with "check prototype". Annotations are
+// detected textually and are not matched against the asserted lock.
+// Wrappers like lockdep_assert_wiphy() are not handled.
+// Options: --no-includes --include-headers
+
+virtual report
+virtual org
+
+@initialize:python@
+@@
+
+import re
+
+ATTRS = re.compile(r'__(must_hold|must_hold_shared|releases|releases_shared|'
+ r'context_unsafe|no_context_analysis)\b')
+params = {}
+static = set()
+
+def key(pf):
+ return (pf[0].file, pf[0].line)
+
+def missing_must_hold(f, a, E, pf, p):
+ e = re.sub(r'\s+', '', E)
+
+ # Identifiers that are neither member names nor called functions.
+ free = re.findall(r'(?<![\w.>])[A-Za-z_]\w*(?!\w*\()', e)
+ if not (all(i in params.get(key(pf), ()) for i in free) or
+ re.fullmatch(r'&\w+(\.\w+)*', e)):
+ return None # Cannot be named in an attribute.
+
+ with open(pf[0].file, errors='replace') as fp:
+ src = fp.read()
+
+ # Attributes on the definition, or on a prototype in the same file.
+ hdr = ''.join(src.splitlines(True)[int(pf[0].line) - 1:]).split('{')[0]
+ if ATTRS.search(hdr):
+ return None
+ for d in re.finditer(r'\b%s\s*\([^;{}]*;' % re.escape(f), src):
+ if ATTRS.search(d.group(0)):
+ return None
+
+ if a == 'lockdep_assert_held_read':
+ want = "__must_hold_shared(%s)" % e
+ elif a == 'lockdep_assert_held':
+ want = "__must_hold(%s) or __must_hold_shared(%s)" % (e, e)
+ else:
+ want = "__must_hold(%s)" % e
+ return "%s() asserts %s is held but lacks %s%s" % (
+ f, e, want, "" if key(pf) in static else "; check prototype")
+
+@prm@
+identifier f, x;
+type T;
+position pf;
+@@
+
+f@pf(..., T x, ...) { ... }
+
+@script:python@
+x << prm.x;
+pf << prm.pf;
+@@
+
+params.setdefault(key(pf), set()).add(x)
+
+@st@
+identifier f;
+position pf;
+@@
+
+static f@pf(...) { ... }
+
+@script:python@
+pf << st.pf;
+@@
+
+static.add(key(pf))
+
+@r forall@
+identifier f;
+identifier a = {lockdep_assert_held, lockdep_assert_held_write,
+ lockdep_assert_held_read, lockdep_assert_held_once,
+ assert_spin_locked, assert_raw_spin_locked};
+expression E;
+position pf, p;
+@@
+
+f@pf(...)
+{
+ ... when != E
+ a@p(E);
+ ...
+}
+
+@script:python depends on report@
+f << r.f;
+a << r.a;
+E << r.E;
+pf << r.pf;
+p << r.p;
+@@
+
+msg = missing_must_hold(f, a, E, pf, p)
+if msg:
+ coccilib.report.print_report(p[0], msg)
+
+@script:python depends on org@
+f << r.f;
+a << r.a;
+E << r.E;
+pf << r.pf;
+p << r.p;
+@@
+
+msg = missing_must_hold(f, a, E, pf, p)
+if msg:
+ coccilib.org.print_todo(p[0], msg)
--
2.56.1.750.gf70dc682cf-goog
^ permalink raw reply [flat|nested] 2+ messages in thread* Re: [PATCH v2] coccinelle: locks: Add must_hold.cocci
2026-11-20 9:26 [PATCH v2] coccinelle: locks: Add must_hold.cocci Marco Elver
@ 2026-11-20 18:14 ` Bart Van Assche
0 siblings, 0 replies; 2+ messages in thread
From: Bart Van Assche @ 2026-11-20 18:14 UTC (permalink / raw)
To: Marco Elver; +Cc: Julia Lawall, Nicolas Palix, linux-kernel, cocci, kasan-dev
On 11/20/26 1:26 AM, Marco Elver wrote:
> +ATTRS = re.compile(r'__(must_hold|must_hold_shared|releases|releases_shared|'
> + r'context_unsafe|no_context_analysis)\b')
Should __cond_acquires and __cond_releases be included in this
expression?
> +def missing_must_hold(f, a, E, pf, p):
> + e = re.sub(r'\s+', '', E)
re.sub(r'\s+', '', E) turns (struct foo *)p into (structfoo*)p. Can this
cause a silent false negative and garbled report text? Collapsing
whitespace instead of removing it would keep tokens separate.
Additionally, missing_must_hold() re-reads the file, runs splitlines()
on all of it and regex-scans it for every match. Has caching per file
been considered? Would this help on a full-tree make coccicheck?
Thanks,
Bart.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-11-20 18:14 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-20 9:26 [PATCH v2] coccinelle: locks: Add must_hold.cocci Marco Elver
2026-11-20 18:14 ` Bart Van Assche
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®