svn commit: r314043 - in head/sys: compat/linuxkpi/common/include/linux compat/linuxkpi/common/src conf modules/linuxkpi

[Adrian Chadd]

Sweet! Thanks!

(God I'd like this for native FreeBSD drivers actually...)


-adrian

On 22 February 2017 at 01:22, Hans Petter Selasky <hps at selasky.org> wrote:
> On 02/22/17 00:18, Adrian Chadd wrote:
>>
>> Hiya,
>>
>> My understanding of tasklets is that they run on the CPU that they
>> were scheduled on, rather than there being a single tasklet thread?
>>
>> Is that the direction you're thinking of heading in, or?
>>
>
> Hi Adrian,
>
> The plan is to use grouptaskqueue in the end. From what I can see all the
> pieces are in place for that. Then the tasklets will run on the same CPU
> that they were scheduled on.
>
> Some patches needs to go into the grouptaskqueue first:
>
> 1) grouptaskqueue needs to support LinuxKPI module unload. There is current
> missing/unimplemented drain logic in grouptaskqueue APIs.
>
> 2) Needs to expose internal gtaskqueues, so that we don't create unneccesary
> threads to handle a single IRQ, which is the case currently.
>
> I've already CC'ed a few people on this, but not action yet. If you can
> help, would be great.
>
> BTW:
> TASKQGROUP_DECLARE(net);
>
> in sys/gtaskqueue.h seems non-existing and should be removed.
>
> --HPS
>