Re: [PATCH v4 2/3] locking: Implement an algorithm choice for Wound-Wait mutexes

From: Rob Clark
Date: Thu Jan 17 2019 - 07:47:50 EST


On Wed, Jan 16, 2019 at 11:49 AM Thomas Hellstrom <thellstrom@xxxxxxxxxx> wrote:
>
> Hi,
>
> On Wed, 2019-01-16 at 09:24 -0500, Rob Clark wrote:
> > So, I guess this is to do w/ the magic of merge commits, but it looks
> > like the hunk changing the crtc_ww_class got lost:
>
> So what happened here is that this commit changed it to
> DEFINE_WD_CLASS
> and the following commit changed it back again to
> DEFINE_WW_CLASS
> so that the algorithm change and only that came with the last commit.
> Apparently the history of that line got lost when merging since the
> merge didn't change it; git blame doesn't show it changed, but when
> looking at the commit history in gitk it's all clear. I guess that's a
> feature of git.
>
> The modeset locks now use true wound-wait rather than wait-die.

ahh, ok, that makes sense. Thanks for clearing up my confusion.

BR,
-R