linux-kernel - Re: [PATCH] locking/lockdep: unlikely bfs error check

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <9c60a4a9-c241-73de-57b5-c5fc45720677@redhat.com>
Date:   Wed, 16 Jun 2021 11:11:21 -0400
From:   Waiman Long <llong@...hat.com>
To:     Xiongwei Song <sxwjean@...com>, Waiman Long <llong@...hat.com>
Cc:     peterz@...radead.org, mingo@...hat.com, will@...nel.org,
        boqun.feng@...il.com, linux-kernel@...r.kernel.org,
        Xiongwei Song <sxwjean@...il.com>
Subject: Re: [PATCH] locking/lockdep: unlikely bfs error check

On 6/16/21 10:59 AM, Xiongwei Song wrote:
>
>> On Jun 16, 2021, at 10:48 PM, Waiman Long <llong@...hat.com> wrote:
>>
>> On 6/16/21 10:42 AM, Xiongwei Song wrote:
>>> From: Xiongwei Song <sxwjean@...il.com>
>>>
>>> The error from graph walk is small probability event, so unlikely
>>> bfs_error can improve performance a little bit.
>>>
>>> Signed-off-by: Xiongwei Song <sxwjean@...il.com>
>>> ---
>>>   kernel/locking/lockdep.c | 12 ++++++------
>>>   1 file changed, 6 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
>>> index 074fd6418c20..af8c9203cd3e 100644
>>> --- a/kernel/locking/lockdep.c
>>> +++ b/kernel/locking/lockdep.c
>>> @@ -2646,7 +2646,7 @@ static int check_irq_usage(struct task_struct *curr, struct held_lock *prev,
>>>   	bfs_init_rootb(&this, prev);
>>>     	ret = __bfs_backwards(&this, &usage_mask, usage_accumulate, usage_skip, NULL);
>>> -	if (bfs_error(ret)) {
>>> +	if (unlikely(bfs_error(ret))) {
>>>   		print_bfs_bug(ret);
>>>   		return 0;
>>>   	}
>>> @@ -2664,7 +2664,7 @@ static int check_irq_usage(struct task_struct *curr, struct held_lock *prev,
>>>   	bfs_init_root(&that, next);
>>>     	ret = find_usage_forwards(&that, forward_mask, &target_entry1);
>>> -	if (bfs_error(ret)) {
>>> +	if (unlikely(bfs_error(ret))) {
>>>   		print_bfs_bug(ret);
>>>   		return 0;
>>>   	}
>>> @@ -2679,7 +2679,7 @@ static int check_irq_usage(struct task_struct *curr, struct held_lock *prev,
>>>   	backward_mask = original_mask(target_entry1->class->usage_mask);
>>>     	ret = find_usage_backwards(&this, backward_mask, &target_entry);
>>> -	if (bfs_error(ret)) {
>>> +	if (unlikely(bfs_error(ret))) {
>>>   		print_bfs_bug(ret);
>>>   		return 0;
>>>   	}
>>> @@ -2998,7 +2998,7 @@ check_prev_add(struct task_struct *curr, struct held_lock *prev,
>>>   	 * Is the <prev> -> <next> link redundant?
>>>   	 */
>>>   	ret = check_redundant(prev, next);
>>> -	if (bfs_error(ret))
>>> +	if (unlikely(bfs_error(ret)))
>>>   		return 0;
>>>   	else if (ret == BFS_RMATCH)
>>>   		return 2;
>>> @@ -3911,7 +3911,7 @@ check_usage_forwards(struct task_struct *curr, struct held_lock *this,
>>>     	bfs_init_root(&root, this);
>>>   	ret = find_usage_forwards(&root, usage_mask, &target_entry);
>>> -	if (bfs_error(ret)) {
>>> +	if (unlikely(bfs_error(ret))) {
>>>   		print_bfs_bug(ret);
>>>   		return 0;
>>>   	}
>>> @@ -3946,7 +3946,7 @@ check_usage_backwards(struct task_struct *curr, struct held_lock *this,
>>>     	bfs_init_rootb(&root, this);
>>>   	ret = find_usage_backwards(&root, usage_mask, &target_entry);
>>> -	if (bfs_error(ret)) {
>>> +	if (unlikely(bfs_error(ret))) {
>>>   		print_bfs_bug(ret);
>>>   		return 0;
>>>   	}
>> I think it is better to put the unlikely() directly into the bfs_error() inline function instead of sprinkling it all over the place.
> Sounds good. Thank you for the suggestion. I will update the patch.

Another nit. It is a bit odd that sent out two patches separately though 
they do seem to have a bit of dependency. I think you should post them 
as a 2-patch series.

Cheers,
Longman