linux-kernel - Re: [PATCH] perf/core: Handling the race between exit

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20250605110815.GQ39944@noisy.programming.kicks-ass.net>
Date: Thu, 5 Jun 2025 13:08:15 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Mark Rutland <mark.rutland@....com>
Cc: Baisheng Gao <baisheng.gao@...soc.com>, Ingo Molnar <mingo@...hat.com>,
	Arnaldo Carvalho de Melo <acme@...nel.org>,
	Namhyung Kim <namhyung@...nel.org>,
	Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
	Jiri Olsa <jolsa@...nel.org>, Ian Rogers <irogers@...gle.com>,
	Adrian Hunter <adrian.hunter@...el.com>,
	"reviewer:PERFORMANCE EVENTS SUBSYSTEM" <kan.liang@...ux.intel.com>,
	"open list:PERFORMANCE EVENTS SUBSYSTEM" <linux-perf-users@...r.kernel.org>,
	"open list:PERFORMANCE EVENTS SUBSYSTEM" <linux-kernel@...r.kernel.org>,
	cixi.geng@...ux.dev, hao_hao.wang@...soc.com
Subject: Re: [PATCH] perf/core: Handling the race between exit_mmap and perf
 sample

On Wed, Jun 04, 2025 at 05:08:23PM +0100, Mark Rutland wrote:
> That looks good to me!

I now haz the below patch.

---
Subject: perf: Fix sample vs do_exit()
From: Peter Zijlstra <peterz@...radead.org>
Date: Thu Jun 5 12:31:45 CEST 2025

Baisheng Gao reported an ARM64 crash, which Mark decoded as being a
synchronous external abort -- most likely due to trying to access
MMIO in bad ways.

The crash further shows perf trying to do a user stack sample while in
exit_mmap()'s tlb_finish_mmu() -- i.e. while tearing down the address
space it is trying to access.

It turns out that we stop perf after we tear down the userspace mm; a
receipie for disaster, since perf likes to access userspace for
various reasons.

Flip this order by moving up where we stop perf in do_exit().

Additionally, harden PERF_SAMPLE_CALLCHAIN and PERF_SAMPLE_STACK_USER
to abort when the current task does not have an mm (exit_mm() makes
sure to set current->mm = NULL; before commencing with the actual
teardown). Such that CPU wide events don't trip on this same problem.

Fixes: c5ebcedb566e ("perf: Add ability to attach user stack dump to sample")
Reported-by: Baisheng Gao <baisheng.gao@...soc.com>
Suggested-by: Mark Rutland <mark.rutland@....com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@...radead.org>
---
 kernel/events/core.c |    7 +++++++
 kernel/exit.c        |   17 +++++++++--------
 2 files changed, 16 insertions(+), 8 deletions(-)

--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -7439,6 +7439,10 @@ perf_sample_ustack_size(u16 stack_size,
 	if (!regs)
 		return 0;
 
+	/* No mm, no stack, no dump. */
+	if (!current->mm)
+		return 0;
+
 	/*
 	 * Check if we fit in with the requested stack size into the:
 	 * - TASK_SIZE
@@ -8150,6 +8154,9 @@ perf_callchain(struct perf_event *event,
 	const u32 max_stack = event->attr.sample_max_stack;
 	struct perf_callchain_entry *callchain;
 
+	if (!current->mm)
+		user = false;
+
 	if (!kernel && !user)
 		return &__empty_callchain;
 
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -944,6 +944,15 @@ void __noreturn do_exit(long code)
 	taskstats_exit(tsk, group_dead);
 	trace_sched_process_exit(tsk, group_dead);
 
+	/*
+	 * Since sampling can touch ->mm, make sure to stop everything before we
+	 * tear it down.
+	 *
+	 * Also flushes inherited counters to the parent - before the parent
+	 * gets woken up by child-exit notifications.
+	 */
+	perf_event_exit_task(tsk);
+
 	exit_mm();
 
 	if (group_dead)
@@ -959,14 +968,6 @@ void __noreturn do_exit(long code)
 	exit_task_work(tsk);
 	exit_thread(tsk);
 
-	/*
-	 * Flush inherited counters to the parent - before the parent
-	 * gets woken up by child-exit notifications.
-	 *
-	 * because of cgroup mode, must be called before cgroup_exit()
-	 */
-	perf_event_exit_task(tsk);
-
 	sched_autogroup_exit_task(tsk);
 	cgroup_exit(tsk);