Re: [PATCH 1/4] audit: refactor queue full checks

[Rinat Gadelshin <rgadelsh@gmail.com>]

On 10.05.2023 10:17, Eiichi Tsukata wrote:
>
>> On May 10, 2023, at 15:54, Rinat Gadelshin <rgadelsh@gmail.com> wrote:
>>
>> Hi Eiichi!
>>
>> Just one one for your patch.
>>
>> On 08.05.2023 10:58, Eiichi Tsukata wrote:
>>> Currently audit queue full checks are done in multiple places.
>>> Consolidate them into one audit_queue_full().
>>>
>>> Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com>
>>> ---
>>>   kernel/audit.c | 21 +++++++++++----------
>>>   1 file changed, 11 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/kernel/audit.c b/kernel/audit.c
>>> index 9bc0b0301198..c15694e1a76b 100644
>>> --- a/kernel/audit.c
>>> +++ b/kernel/audit.c
>>> @@ -341,6 +341,12 @@ static inline int audit_rate_check(void)
>>>    return retval;
>>>   }
>>>   +static inline int audit_queue_full(const struct sk_buff_head *queue)
>>> +{
>>> + return audit_backlog_limit &&
>>> +        (skb_queue_len(queue) > audit_backlog_limit);
>> It seems that we should use `>=` here.
> Hi Rinat
>
> Could you provide the detailed reason?
>
> Currently queue full checks are done with ‘>’,
> on the other hand queue NOT full checks are done with ‘<‘.
>
> Looking into other similar checks in the kernel, unix_recvq_full() is using ‘>’.
Was (OR statement): `if (!audit_backlog_limit || 
skb_queue_len(&audit_retry_queue) < audit_backlog_limit)
For AND-statement it should be `if (audit_backlog_limit && 
(skb_queue_len(&audit_retry_queue) >= audit_backlog_limit))
Otherwise we get false for case `(skb_queue_len(&audit_retry_queue) == 
audit_backlog_limit)` which was true for the old implementation.
>
> Paul, how do you think about it?
>
> Eiichi
>
>