bpf: fix lockdep false positive in stackmap
authorAlexei Starovoitov <ast@kernel.org>
Sun, 10 Feb 2019 20:52:35 +0000 (12:52 -0800)
committerDaniel Borkmann <daniel@iogearbox.net>
Mon, 11 Feb 2019 15:36:24 +0000 (16:36 +0100)
commit3defaf2f15b2bfd86c6664181ac009e91985f8ac
treee14f5b1fb6e2fb4447518ad12b7ef38618e8f5a1
parente6762c8bcf982821935a2b1cb33cf8335d0eefae
bpf: fix lockdep false positive in stackmap

Lockdep warns about false positive:
[   11.211460] ------------[ cut here ]------------
[   11.211936] DEBUG_LOCKS_WARN_ON(depth <= 0)
[   11.211985] WARNING: CPU: 0 PID: 141 at ../kernel/locking/lockdep.c:3592 lock_release+0x1ad/0x280
[   11.213134] Modules linked in:
[   11.214954] RIP: 0010:lock_release+0x1ad/0x280
[   11.223508] Call Trace:
[   11.223705]  <IRQ>
[   11.223874]  ? __local_bh_enable+0x7a/0x80
[   11.224199]  up_read+0x1c/0xa0
[   11.224446]  do_up_read+0x12/0x20
[   11.224713]  irq_work_run_list+0x43/0x70
[   11.225030]  irq_work_run+0x26/0x50
[   11.225310]  smp_irq_work_interrupt+0x57/0x1f0
[   11.225662]  irq_work_interrupt+0xf/0x20

since rw_semaphore is released in a different task vs task that locked the sema.
It is expected behavior.
Fix the warning with up_read_non_owner() and rwsem_release() annotation.

Fixes: bae77c5eb5b2 ("bpf: enable stackmap with build_id in nmi context")
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
kernel/bpf/stackmap.c