perf: Paper over the hw.target problems

euler inclusion
category: bugfix
bugzilla: 9513/11006/11050
CVE: NA
--------------------------------------------------

[ Cheng Jian
HULK-Syzkaller reported a problem which has been reported to mainline(lkml)
by syzbot early, this patch comes from the reply form lkml.
v1	https://lkml.org/lkml/2019/2/28/529
v2	https://lkml.org/lkml/2019/3/8/206
we merged v1 first but cause bugzilla #11050, it was because :
we also use perf_remove_from_context() in perf_event_open() when we move
events from a SW context to a HW context, so we can't destroy the event
here.
now v2 will not exhibit that warning.
it's same to another patch at https://lkml.org/lkml/2019/3/8/536.
but more clear than it.]

First, we have a race between perf_event_release_kernel() and
perf_free_event(), which happens when parent's event is released while the
child's fork fails (because of a fatal signal, for example), that looks
like this:

cpu X                            cpu Y
-----                            -----
                                 copy_process() error path
perf_release(parent)             +->perf_event_free_task()
+-> lock(child_ctx->mutex)       |  |
+-> remove_from_context(child)   |  |
+-> unlock(child_ctx->mutex)     |  |
|                                |  +-> lock(child_ctx->mutex)
|                                |  +-> unlock(child_ctx->mutex)
|                                +-> free_task(child_task)
+-> put_task_struct(child_task)

Technically, we're still holding a reference to the task via
parent->hw.target, that's not stopping free_task(), so we end up poking at
free'd memory, as is pointed out by KASAN in the syzkaller report (see Link
below). The straightforward fix is to drop the hw.target reference while
the task is still around.

Therein lies the second problem: the users of hw.target (uprobe) assume
that it's around at ->destroy() callback time, where they use it for
context. So, in order to not break the uprobe teardown and avoid leaking
stuff, we need to call ->destroy() at the same time.

This patch fixes the race and the subsequent fallout by doing both these
things at remove_from_context time.
Signed-off-by: NAlexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://syzkaller.appspot.com/bug?extid=a24c397a29ad22d86c98Signed-off-by: NCheng Jian <cj.chengjian@huawei.com>
Reviewed-by: NXie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: NYang Yingliang <yangyingliang@huawei.com>

kernel/events/core.c

@@ -2104,6 +2104,28 @@ static void perf_remove_from_context(struct perf_event *event, unsigned long fla
 
 	event_function_call(event, __perf_remove_from_context, (void *)flags);
 
+	/*
+	 * This is as passable as any hw.target handling out there;
+	 * hw.target implies task context, therefore, no migration.
+	 * Which together with DETACH_GROUP means that this is the
+	 * final remove_from_context of a task event.
+	 */
+	if (event->hw.target && (flags & DETACH_GROUP)) {
+		/*
+		 * Now, the problem with, say uprobes, is that they
+		 * use hw.target for context in their ->destroy()
+		 * callbacks. Supposedly, they may need to poke at
+		 * its contents, so better call it while we still
+		 * have the task.
+		 */
+		if (event->destroy) {
+			event->destroy(event);
+			event->destroy = NULL;
+		}
+		put_task_struct(event->hw.target);
+		event->hw.target = NULL;
+	}
+
 	/*
 	 * The above event_function_call() can NO-OP when it hits
 	 * TASK_TOMBSTONE. In that case we must already have been detached