Re: [PATCH v14 06/14] HWBP: Add modify_local_hw_breakpoint_addr() API

From
sashiko-bot@kernel.org
Date
2026-08-30 14:40:57
Also in
linux-perf-users
Reply to
Masami Hiramatsu (Google), 2026-08-30
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Lockless updates to `bp->attr.bp_addr` and `counter_arch_bp(bp)->address` lack `WRITE_ONCE()`, constituting a formal data race under the Linux Kernel Memory Model.
--

commit 0c79300d09e43eac52a027f6cb480808e7f3902f
Author: Masami Hiramatsu (Google) [off-list ref]

HWBP: Add modify_local_hw_breakpoint_addr() API

This commit adds the modify_local_hw_breakpoint_addr() API to update the
watched address of an installed hardware breakpoint on the local CPU without
releasing and reserving its hardware slot.

Link: https://lore.kernel.org/all/f9c49dfa49bdc57ba8c0574bc9981c1e581acf92.1785067572.git.wangjinchao600@gmail.com/ (local)
quoted hunk ↗ jump to hunk
diff --git a/kernel/events/hw_breakpoint.c b/kernel/events/hw_breakpoint.c
index 789add0c185a0..124074c3d3bc2 100644
--- a/kernel/events/hw_breakpoint.c
+++ b/kernel/events/hw_breakpoint.c
[ ... ]
quoted hunk ↗ jump to hunk
+int modify_local_hw_breakpoint_addr(struct perf_event *bp,
+				    unsigned long addr)
+{
+	unsigned long old_addr;
+	int ret;
+
+	lockdep_assert_irqs_disabled();
+
+	old_addr = bp->attr.bp_addr;
+	counter_arch_bp(bp)->address = addr;
+	bp->attr.bp_addr = addr;
[Severity: Low]
The function intentionally updates bp->attr.bp_addr locklessly so NMI handlers
can observe the new address during the transition. However, without WRITE_ONCE()
annotations, does this constitute a formal data race under the Linux Kernel
Memory Model? 

While a torn store on 32-bit architectures might only overwrite the upper 32
bits with zero (leaving the address functionally intact), could this still
trigger KCSAN warnings when an NMI handler concurrently reads bp->attr.bp_addr
via perf_bp_event()?
quoted hunk ↗ jump to hunk
+
+	ret = arch_modify_local_hw_breakpoint_addr(bp, addr);
+	if (ret) {
+		counter_arch_bp(bp)->address = old_addr;
+		bp->attr.bp_addr = old_addr;
[Severity: Low]
Similarly, should the rollback path also use WRITE_ONCE() to ensure concurrent
accesses by NMI handlers avoid data race warnings from KCSAN?
quoted hunk ↗ jump to hunk
+		return ret;
+	}
+
+	return 0;
+}
-- 
Sashiko AI review · https://sashiko.dev/#/patchset/178810001186.64882.2161016469449127450.stgit@devnote2?part=6
Thread (36 messages) 36 messages, 2 authors, last reply 2026-09-06
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help