KCSAN: data-race in __perf_event_overflow / perf_pending_task

Status: upstream: reported on 2026/11/19 10:54
Subsystems: perf
Labels: prio:low race:harmful
[Documentation on labels]
Reported-by: syzbot+a480b581a0bf5f5e8b0d@syzkaller.appspotmail.com
First crash: 2d09h, last: 1d18h
✨ AI-generated root cause summary:

In kernel/events/core.c, perf_pending_task() clears event->pending_work to 0 before calling perf_sigtrap(), which reads event->pending_addr. Because perf_pending_task() runs in task context with interrupts enabled, a concurrent interrupt (such as an hrtimer interrupt calling __perf_event_overflow()) on the same CPU observes event->pending_work == 0, considers the pending slot available, re-enqueues event->pending_task via task_work_add(), and overwrites event->pending_addr while perf_sigtrap() is concurrently reading it.

▶ ▼ ✨ Full AI root cause analysis
Based on kernel commit 666539eecdf4:

Technical Explanation of Root Cause

1. Participating Entities and Locations


2. Mechanism and Lifecycle Violation

When a perf_event with attr.sigtrap set triggers an overflow in __perf_event_overflow(), it schedules SIGTRAP signal delivery using task work:

if (!event->pending_work &&
    !task_work_add(current, &event->pending_task, notify_mode)) {
	event->pending_work = pending_id;
	local_inc(&event->ctx->nr_no_switch_fast);
	WARN_ON_ONCE(!atomic_long_inc_not_zero(&event->refcount));

	event->pending_addr = 0;
	if (valid_sample && (data->sample_flags & PERF_SAMPLE_ADDR))
		event->pending_addr = data->addr;
...

Here, event->pending_work functions as an exclusivity/ownership guard: while non-zero, subsequent overflows are prevented from overwriting event->pending_addr or adding another event->pending_task to the task's work list.

On return to user mode (e.g., via exit_to_user_mode_loop() -> task_work_run()), the task runs queued task works. In task_work_run() (kernel/task_work.c:231-236), the callback entry head = &event->pending_task is already detached from task->task_works prior to calling work->func(work) (perf_pending_task()).

In perf_pending_task():

static void perf_pending_task(struct callback_head *head)
{
	struct perf_event *event = container_of(head, struct perf_event, pending_task);
	int rctx;

	rctx = perf_swevent_get_recursion_context();

	if (event->pending_work) {
		event->pending_work = 0;
		perf_sigtrap(event);
		local_dec(&event->ctx->nr_no_switch_fast);
	}
	put_event(event);

	if (rctx >= 0)
		perf_swevent_put_recursion_context(rctx);
}

perf_pending_task() executes in process/task context with hardware interrupts enabled. Crucially:

  1. event->pending_work is reset to 0 before calling perf_sigtrap(event) and reading event->pending_addr.
  2. An interrupt (such as an hrtimer interrupt driving perf_swevent_hrtimer) can arrive on the current CPU after event->pending_work = 0 is executed and while perf_sigtrap() is reading event->pending_addr.
  3. In interrupt context, __perf_event_overflow() evaluates !event->pending_work as true (since pending_work was prematurely reset to 0).
  4. Because event->pending_task was already detached from the task's work list by task_work_run(), task_work_add(current, &event->pending_task, notify_mode) succeeds, and __perf_event_overflow() resets event->pending_addr = 0 (and event->pending_addr = data->addr).
  5. This write to event->pending_addr races directly with the interrupted task reading event->pending_addr at kernel/events/core.c:7676.
▶ ▼ ✨ AI Jobs (2)
ID Workflow Result Correct Bug Created Started Finished Revision
be1f2b3d assessment-security DenialOfService: ❌ Exploitable: ❌ FilesystemTrigger: ❌ FuzzingArtifact: ❌ MemoryCorruption: ❌ NetworkNeedsSetup: ❌ NetworkTrigger: ❌ PeripheralTrigger: ❌ RemoteTrigger: ❌ Unprivileged: ✅ UserNamespace: ✅ VMGuestTrigger: ❌ VMHostTrigger: ❌ ❓ KCSAN: data-race in __perf_event_overflow / perf_pending_task 2026/11/19 10:38 2026/11/19 10:38 2026/11/19 10:46 2006bdcc
d5d7dd9b assessment-kcsan Benign: ❌ ❓ KCSAN: data-race in __perf_event_overflow / perf_pending_task 2026/11/19 10:29 2026/11/19 10:29 2026/11/19 10:49 2006bdcc
▶ ▼ Discussions (1)
Title Replies (including bot) Last reply
[syzbot] [perf?] KCSAN: data-race in __perf_event_overflow / perf_pending_task 0 (1) 2026/11/19 10:54

Sample crash report:
==================================================================
BUG: KCSAN: data-race in __perf_event_overflow / perf_pending_task

write to 0xffff88811b623050 of 8 bytes by interrupt on cpu 1:
 __perf_event_overflow+0x7f1/0x950 kernel/events/core.c:10959
 perf_event_overflow kernel/events/core.c:11004 [inline]
 perf_swevent_hrtimer+0x293/0x320 kernel/events/core.c:12424
 __run_hrtimer kernel/time/hrtimer.c:2079 [inline]
 __hrtimer_run_queues+0x1f8/0x540 kernel/time/hrtimer.c:2136
 hrtimer_interrupt+0x27b/0x870 kernel/time/hrtimer.c:2255
 local_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1051 [inline]
 __sysvec_apic_timer_interrupt+0x5f/0x1f0 arch/x86/kernel/apic/apic.c:1068
 instr_sysvec_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1062 [inline]
 sysvec_apic_timer_interrupt+0x6f/0x80 arch/x86/kernel/apic/apic.c:1062
 asm_sysvec_apic_timer_interrupt+0x1a/0x20 arch/x86/include/asm/idtentry.h:674
 native_pause arch/x86/include/asm/vdso/processor.h:13 [inline]
 delay_tsc+0x43/0xa0 arch/x86/lib/delay.c:76
 udelay include/asm-generic/delay.h:64 [inline]
 delay_access kernel/kcsan/core.c:332 [inline]
 kcsan_setup_watchpoint+0x255/0x3f0 kernel/kcsan/core.c:622
 perf_sigtrap kernel/events/core.c:7676 [inline]
 perf_pending_task+0x16a/0x220 kernel/events/core.c:7780
 task_work_run+0x145/0x1a0 kernel/task_work.c:233
 resume_user_mode_work+0x85/0xa0 include/linux/resume_user_mode.h:50
 __exit_to_user_mode_loop kernel/entry/common.c:70 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:101 [inline]
 __exit_to_user_mode_prepare include/linux/irq-entry-common.h:207 [inline]
 irqentry_exit_to_user_mode_prepare include/linux/irq-entry-common.h:244 [inline]
 irqentry_exit_to_user_mode include/linux/irq-entry-common.h:315 [inline]
 irqentry_exit+0x170/0x640 kernel/entry/common.c:152
 sysvec_apic_timer_interrupt+0x44/0x80 arch/x86/kernel/apic/apic.c:1062
 asm_sysvec_apic_timer_interrupt+0x1a/0x20 arch/x86/include/asm/idtentry.h:674

read to 0xffff88811b623050 of 8 bytes by task 4399 on cpu 1:
 perf_sigtrap kernel/events/core.c:7676 [inline]
 perf_pending_task+0x16a/0x220 kernel/events/core.c:7780
 task_work_run+0x145/0x1a0 kernel/task_work.c:233
 resume_user_mode_work+0x85/0xa0 include/linux/resume_user_mode.h:50
 __exit_to_user_mode_loop kernel/entry/common.c:70 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:101 [inline]
 __exit_to_user_mode_prepare include/linux/irq-entry-common.h:207 [inline]
 irqentry_exit_to_user_mode_prepare include/linux/irq-entry-common.h:244 [inline]
 irqentry_exit_to_user_mode include/linux/irq-entry-common.h:315 [inline]
 irqentry_exit+0x170/0x640 kernel/entry/common.c:152
 sysvec_apic_timer_interrupt+0x44/0x80 arch/x86/kernel/apic/apic.c:1062
 asm_sysvec_apic_timer_interrupt+0x1a/0x20 arch/x86/include/asm/idtentry.h:674

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 UID: 0 PID: 4399 Comm: syz.0.427 Not tainted syzkaller #0 PREEMPT(lazy) 
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 11/06/2026
==================================================================

Crashes (3):
Time Kernel Commit Syzkaller Config Log Report Syz repro C repro VM info Assets (help?) Manager Title
2026/11/17 17:23 upstream 666539eecdf4 122f94d5 .config console log report info [disk image] [vmlinux] [kernel image] ci2-upstream-kcsan-gce KCSAN: data-race in __perf_event_overflow / perf_pending_task
2026/11/17 10:28 upstream 666539eecdf4 43af6cac .config console log report info [disk image] [vmlinux] [kernel image] ci2-upstream-kcsan-gce KCSAN: data-race in __perf_event_overflow / perf_pending_task
2026/11/17 02:05 upstream 666539eecdf4 a3e64cfc .config console log report info [disk image] [vmlinux] [kernel image] ci2-upstream-kcsan-gce KCSAN: data-race in __perf_event_overflow / perf_pending_task
* Struck through repros no longer work on HEAD.