perf: Fix perf_pending_task() UaF
authorPeter Zijlstra <peterz@infradead.org>
Thu, 24 Nov 2022 11:49:12 +0000 (12:49 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 19 Dec 2022 11:36:43 +0000 (12:36 +0100)
commit8bffa95ac19ff27c8261904f89d36c7fcf215d59
tree6ccc7ba83ab089520cc4de45a23c7a499224c662
parent825bd2af4227b35017a0e0be19dcc308b32f126c
perf: Fix perf_pending_task() UaF

[ Upstream commit 517e6a301f34613bff24a8e35b5455884f2d83d8 ]

Per syzbot it is possible for perf_pending_task() to run after the
event is free()'d. There are two related but distinct cases:

 - the task_work was already queued before destroying the event;
 - destroying the event itself queues the task_work.

The first cannot be solved using task_work_cancel() since
perf_release() itself might be called from a task_work (____fput),
which means the current->task_works list is already empty and
task_work_cancel() won't be able to find the perf_pending_task()
entry.

The simplest alternative is extending the perf_event lifetime to cover
the task_work.

The second is just silly, queueing a task_work while you know the
event is going away makes no sense and is easily avoided by
re-arranging how the event is marked STATE_DEAD and ensuring it goes
through STATE_OFF on the way down.

Reported-by: syzbot+9228d6098455bb209ec8@syzkaller.appspotmail.com
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Marco Elver <elver@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
kernel/events/core.c