On Wed, 2016-03-16 at 18:48 +0100, Maarten Lankhorst wrote: > Op 16-03-16 om 17:19 schreef Ander Conselvan De Oliveira: > > On Mon, 2016-03-14 at 09:27 +0100, Maarten Lankhorst wrote: > > > With async modesets this is no longer protected with connection_mutex, > > > so ensure that each pll has its own lock. The pll configuration state > > > is still protected; it's only the pll updates that need locking against > > > concurrency. > > I think I need to look at your async branch, since I'm not really sure how > > async > > will work. But locking the individual plls might fail in SKL with the > > current > > code. The register DPLL_CTRL1 controls all 4 plls, and currently it is > > updated > > with a read-modify-write in the enable hook, so we can't update two plls > > concurrently. > > > Would making the dpll lock global help? I don't think in practice the locks > will be contended much, > it's not a performance sensitive path.
Yeah, I think that should be enough. Ander _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
