[PATCH] bpf: sync pending IRQ work before freeing ring buffer

[neqbal <nooraineqbal@gmail.com>]

#syz test

Fix a possible vmalloc out-of-bounds access caused by pending IRQ work

Reported-by: syzbot+2617fc732430968b45d2@syzkaller.appspotmail.com
Signed-off-by: neqbal <nooraineqbal@gmail.com>
---
 kernel/bpf/ringbuf.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/kernel/bpf/ringbuf.c b/kernel/bpf/ringbuf.c
index 719d73299397..a6597417412f 100644
--- a/kernel/bpf/ringbuf.c
+++ b/kernel/bpf/ringbuf.c
@@ -216,6 +216,12 @@ static struct bpf_map *ringbuf_map_alloc(union bpf_attr *attr)
 
 static void bpf_ringbuf_free(struct bpf_ringbuf *rb)
 {
+	/* Ensure all pending IRQ work completes before freeing.
+	 * In PREEMPT_RT, IRQ work runs in thread context and can
+	 * race with ring buffer destruction.
+	 */
+	irq_work_sync(&rb->work);
+
 	/* copy pages pointer and nr_pages to local variable, as we are going
 	 * to unmap rb itself with vunmap() below
 	 */
-- 
2.51.0