Re: Call for Testers: Convert watchdog spinlock into a sleepable mutex in if_ndis

Works fine for me. No regressions encountered yet.

On 6/10/08, Coleman Kane <cokane_at_freebsd.org> wrote:
> On Tue, 2008-06-10 at 14:56 +0200, Paul B. Mahol wrote:
>> On 6/10/08, Coleman Kane <cokane_at_freebsd.org> wrote:
>> > On Tue, 2008-06-10 at 05:16 +0200, Paul B. Mahol wrote:
>> >> On 6/9/08, Coleman Kane <cokane_at_freebsd.org> wrote:
>> >> > Ignore the previous patch (#2) and try this one instead:
>> >> >       *
>> >> > http://people.freebsd.org/~cokane/patches/if_ndis-spinlock-to-mtx3.patch
>> >> >
>> >>
>> >> Fine for me, LOR appear any more.
>> >
>> > Thanks for the report.
>> >
>> >>
>> >> BTW: I'm dont like first line showing over and over again. It would be
>> >> nice
>> >> if that clueless message goes away.
>> >
>> > Is this a new message introduced with my changes?
>> No, it is not. Thanks for patch(es).
>>
>
> I've got one more for you to try:
>       *
> http://people.freebsd.org/~cokane/patches/if_ndis-spinlock-to-mtx4.patch
>
> It removes the locking inside the ndis_tick function altogether, and
> just relies upon ndis_ticktask, ndis_resettask, and ndis_starttask to
> perform the appropriate locking (which they appear to have been written
> to do already). This might improve concurrency even more (and probably
> get rid of some mutex recursions).
>
> --
> Coleman Kane
>