Re: [RFC][PATCH] fsnotify: optimize the case of no access event watchers

[Jens Axboe <axboe@kernel.dk>]

On 1/9/24 1:12 PM, Jens Axboe wrote:
> On 1/9/24 12:48 PM, Amir Goldstein wrote:
>> Commit e43de7f0862b ("fsnotify: optimize the case of no marks of any type")
>> optimized the case where there are no fsnotify watchers on any of the
>> filesystem's objects.
>>
>> It is quite common for a system to have a single local filesystem and
>> it is quite common for the system to have some inotify watches on some
>> config files or directories, so the optimization of no marks at all is
>> often not in effect.
>>
>> Access event watchers are far less common, so optimizing the case of
>> no marks with access events could improve performance for more systems,
>> especially for the performance sensitive hot io path.
>>
>> Maintain a per-sb counter of objects that have marks with access
>> events in their mask and use that counter to optimize out the call to
>> fsnotify() in fsnotify access hooks.
>>
>> Signed-off-by: Amir Goldstein <amir73il@gmail.com>
>> ---
>>
>> Jens,
>>
>> You may want to try if this patch improves performance for your workload
>> with SECURITY=Y and FANOTIFY_ACCESS_PERMISSIONS=Y.
> 
> Ran the usual test, and this effectively removes fsnotify from the
> profiles, which (as per other email) is between 5-6% of CPU time. So I'd
> say it looks mighty appealing!

Tried with an IRQ based workload as well, as those are always impacted
more by the fsnotify slowness. This patch removes ~8% of useless
overhead in that case, so even bigger win there.

-- 
Jens Axboe