[PATCH AUTOSEL 4.20 08/60] bpf: fix lockdep false positive in stackmap

[Sasha Levin]

From: Alexei Starovoitov <a...@kernel.org>

[ Upstream commit 3defaf2f15b2bfd86c6664181ac009e91985f8ac ]

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 <a...@kernel.org>
Signed-off-by: Daniel Borkmann <dan...@iogearbox.net>
Signed-off-by: Sasha Levin <sas...@kernel.org>
---
 kernel/bpf/stackmap.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
index d43b14535827..950ab2f28922 100644
--- a/kernel/bpf/stackmap.c
+++ b/kernel/bpf/stackmap.c
@@ -44,7 +44,7 @@ static void do_up_read(struct irq_work *entry)
        struct stack_map_irq_work *work;
 
        work = container_of(entry, struct stack_map_irq_work, irq_work);
-       up_read(work->sem);
+       up_read_non_owner(work->sem);
        work->sem = NULL;
 }
 
@@ -338,6 +338,12 @@ static void stack_map_get_build_id_offset(struct 
bpf_stack_build_id *id_offs,
        } else {
                work->sem = &current->mm->mmap_sem;
                irq_work_queue(&work->irq_work);
+               /*
+                * The irq_work will release the mmap_sem with
+                * up_read_non_owner(). The rwsem_release() is called
+                * here to release the lock from lockdep's perspective.
+                */
+               rwsem_release(&current->mm->mmap_sem.dep_map, 1, _RET_IP_);
        }
 }
 
-- 
2.19.1