uprobes: avoid false lockdep splat in uprobe timer callback
Subsystems (4)
Andrii Nakryiko
maintainerPATCH v2 1/1 uprobes: avoid false lockdep splat in uprobe timer callback
Avoid a false-positive lockdep warning in PREEMPT_RT configuration when using write_seqcount_begin() in uprobe timer callback by using raw_write_* APIs. Uprobe's use of timer callback is guaranteed to not race with itself for a given uprobe_task, and as such seqcount's insistence on having preemption disabled on the writer side is irrelevant. So switch to raw_ variants of seqcount API instead of disabling preemption unnecessarily.
Also, point out in the comments more explicitly why we use seqcount despite our reader side being rather simple and never retrying. We favor well-maintained kernel primitive in favor of open-coding our own memory barriers.
---v1->v2:
- fix comment style and s/hardirqs/preemption/ (Sebastian);
- improved added comment based on Sebastian's suggestions.
kernel/events/uprobes.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)
kernel/events/uprobes.c +13 -2
@@ -1944,6 +1944,9 @@ static void free_ret_instance(struct uprobe_task *utask,* to-be-reused return instances for future uretprobes. If ri_timer() * happens to be running right now, though, we fallback to safety and * just perform RCU-delated freeing of ri. +* Admittedly, this is a rather simple use of seqcount, but it nicely +* abstracts away all the necessary memory barriers, so we use +* a well-supported kernel primitive here. */ if (raw_seqcount_try_begin (&utask->ri_seqcount, seq)) {/* immediate reuse of ri without RCU GP is OK */
@@ -2004,12 +2007,20 @@ static void ri_timer(struct timer_list *timer)/* RCU protects return_instance from freeing. */ guard (rcu)(); -write_seqcount_begin (&utask->ri_seqcount); +/* +* See free_ret_instance() for notes on seqcount use. +* We also employ raw API variants to avoid lockdep false-positive +* warning complaining about enabled preemption. The timer can only be +* invoked once for a uprobe_task. Therefore there can only be one +* writer. The reader does not require an even sequence count to make +* progress, so it is OK to remain preemptible on PREEMPT_RT. +*/ +raw_write_seqcount_begin (&utask->ri_seqcount);for_each_ret_instance_rcu (ri, utask->return_instances)hprobe_expire (&ri->hprobe, false); -write_seqcount_end (&utask->ri_seqcount); +raw_write_seqcount_end (&utask->ri_seqcount); }static struct uprobe_task *alloc_utask (void )
Signature · 1 line
2.47.1
Quoted 11 lines · Andrii Nakryiko
Also, point out in the comments more explicitly why we use seqcount despite our reader side being rather simple and never retrying. We favor well-maintained kernel primitive in favor of open-coding our own memory barriers.
Link https://lore.kernel.org/bpf/CAADnVQLLOHZmPO4X_dQ+cTaSDvzdWHzA0qUqQDhLFYL3D6xPxg@mail.gmail.com/ open hereReported-by Alexei Starovoitov <[email protected]>Suggested-by Sebastian Siewior <[email protected]>Fixes 8622e45b5da1 ("uprobes: Reuse return_instances between multiple uretprobes within task")Signed-off-by Andrii Nakryiko <[email protected]>
LGTM. FWIW,
[tip: perf/urgent] uprobes: Avoid false-positive lockdep splat on CONFIG_PREEMPT_RT=y in the ri_timer() uprobe timer callback, use raw_write_seqcount_*()
In reply to Andrii NakryikoThe following commit has been merged into the perf/urgent branch of tip:
The following commit has been merged into the perf/urgent branch of tip:
Commit-ID: 0cd575cab10e114e95921321f069a08d45bc412e Gitweb: https://git.kernel.org/tip/0cd575cab10e114e95921321f069a08d45bc412e Author: Andrii Nakryiko <[email protected]> AuthorDate: Fri, 04 Apr 2025 12:48:48 -07:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Mon, 07 Apr 2025 20:15:16 +02:00
uprobes: Avoid false-positive lockdep splat on CONFIG_PREEMPT_RT=y in the ri_timer() uprobe timer callback, use raw_write_seqcount_*()
Avoid a false-positive lockdep warning in the CONFIG_PREEMPT_RT=y configuration when using write_seqcount_begin() in the uprobe timer callback by using raw_write_* APIs.
Uprobe's use of timer callback is guaranteed to not race with itself for a given uprobe_task, and as such seqcount's insistence on having preemption disabled on the writer side is irrelevant. So switch to raw_ variants of seqcount API instead of disabling preemption unnecessarily.
Also, point out in the comments more explicitly why we use seqcount despite our reader side being rather simple and never retrying. We favor well-maintained kernel primitive in favor of open-coding our own memory barriers.
--- kernel/events/uprobes.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)
kernel/events/uprobes.c +13 -2
@@ -1956,6 +1956,9 @@ static void free_ret_instance(struct uprobe_task *utask,* to-be-reused return instances for future uretprobes. If ri_timer() * happens to be running right now, though, we fallback to safety and * just perform RCU-delated freeing of ri. +* Admittedly, this is a rather simple use of seqcount, but it nicely +* abstracts away all the necessary memory barriers, so we use +* a well-supported kernel primitive here. */ if (raw_seqcount_try_begin (&utask->ri_seqcount, seq)) {/* immediate reuse of ri without RCU GP is OK */
@@ -2016,12 +2019,20 @@ static void ri_timer(struct timer_list *timer)/* RCU protects return_instance from freeing. */ guard (rcu)(); -write_seqcount_begin (&utask->ri_seqcount); +/* +* See free_ret_instance() for notes on seqcount use. +* We also employ raw API variants to avoid lockdep false-positive +* warning complaining about enabled preemption. The timer can only be +* invoked once for a uprobe_task. Therefore there can only be one +* writer. The reader does not require an even sequence count to make +* progress, so it is OK to remain preemptible on PREEMPT_RT. +*/ +raw_write_seqcount_begin (&utask->ri_seqcount);for_each_ret_instance_rcu (ri, utask->return_instances)hprobe_expire (&ri->hprobe, false); -write_seqcount_end (&utask->ri_seqcount); +raw_write_seqcount_end (&utask->ri_seqcount); }static struct uprobe_task *alloc_utask (void )