2022-06-16 14:18:41

by Chen Jun

Subject: [PATCH 1/1] lockdep: pass curr_inner to print_lock_invalid_wait_context

Same information (task_wait_context(curr)) is printed twice.
curr_inner in check_wait_context is what we need.

Signed-off-by: Chen Jun <[email protected]>
---
kernel/locking/lockdep.c | 10 +++-------
1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 81e87280513e..2eb9802df7b9 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -4676,10 +4676,8 @@ static inline short task_wait_context(struct task_struct *curr)

static int
print_lock_invalid_wait_context(struct task_struct *curr,
- struct held_lock *hlock)
+ struct held_lock *hlock, short curr_inner)
{
- short curr_inner;
-
if (!debug_locks_off())
return 0;
if (debug_locks_silent)
@@ -4695,9 +4693,7 @@ print_lock_invalid_wait_context(struct task_struct *curr,
print_lock(hlock);

pr_warn("other info that might help us debug this:\n");
-
- curr_inner = task_wait_context(curr);
- pr_warn("context-{%d:%d}\n", curr_inner, curr_inner);
+ pr_warn("context-{%d:%d}\n", curr_inner, task_wait_context(curr));

lockdep_print_held_locks(curr);

@@ -4763,7 +4759,7 @@ static int check_wait_context(struct task_struct *curr, struct held_lock *next)
}

if (next_outer > curr_inner)
- return print_lock_invalid_wait_context(curr, next);
+ return print_lock_invalid_wait_context(curr, next, curr_inner);

return 0;
}
--
2.17.1

2022-06-17 13:59:00

by Peter Zijlstra

Subject: Re: [PATCH 1/1] lockdep: pass curr_inner to print_lock_invalid_wait_context

On Thu, Jun 16, 2022 at 01:40:42PM +0000, Chen Jun wrote:
> Same information (task_wait_context(curr)) is printed twice.

Yes, because at that point the inner and outer context are the same.

> curr_inner in check_wait_context is what we need.

IIRC it simply prints the task_wait_context as a starting point, the
lockdep_print_held_locks() should include the contexts for each of
those, after which you can compute the resulting value.

IOW, the information is complete.

2022-06-20 08:53:14

by Chen Jun

Subject: Re: [PATCH 1/1] lockdep: pass curr_inner to print_lock_invalid_wait_context

$B:_(B 2022/6/17 21:22, Peter Zijlstra $B<LF;(B:
> On Thu, Jun 16, 2022 at 01:40:42PM +0000, Chen Jun wrote:
>> Same information (task_wait_context(curr)) is printed twice.
>
> Yes, because at that point the inner and outer context are the same.
>
>> curr_inner in check_wait_context is what we need.
>
> IIRC it simply prints the task_wait_context as a starting point, the
> lockdep_print_held_locks() should include the contexts for each of
> those, after which you can compute the resulting value.
>
> IOW, the information is complete.
>
>
>

Thanks for pointing this out.

--
Regards
Chen Jun