On Tue, Jul 22, 2025, at 12:29, Chris Bainbridge wrote:> On Mon, Jul 21, 2025 at 08:22:48AM -0400, Satadru Pramanik wrote: >> Hello all, >> >> I suspect this commit in 6.16-rc7 has broken acceleration with Mesa's >> nouveau drivers on my machine. >> >> glxinfo -B reports that I'm using llvmpipe. >> >> Reverting this in 6.16-rc7 restores nouveau acceleration, and glxinfo >> then reports: "OpenGL renderer string: NVE7" > > I also bisected an issue to this commit. On my laptop, this commit > results in an intermittent desktop crash (Xorg segfault) when changing > display scale, which can be more reliably reproduced with: > > for x in {1..100}; do > xrandr --output eDP-1 --mode 2560x1600 --scale 0.5 --filter nearest > xrandr --output eDP-1 --mode 2560x1600 --scale 1 --filter nearest > done >I won't have time to work on fixing my patch before the merge window, let's just revert it. Arnd
Danilo Krummrich
2025-Jul-22 11:45 UTC
[PATCH] drm/nouveau: check ioctl command codes better
On 7/22/25 12:52 PM, Arnd Bergmann wrote:> On Tue, Jul 22, 2025, at 12:29, Chris Bainbridge wrote: >> On Mon, Jul 21, 2025 at 08:22:48AM -0400, Satadru Pramanik wrote: >>> Hello all, >>> >>> I suspect this commit in 6.16-rc7 has broken acceleration with Mesa's >>> nouveau drivers on my machine.Thanks for the report! Please make sure to keep maintainers in the loop, for some reason I was removed from the recipient list for this regression report.>>> >>> glxinfo -B reports that I'm using llvmpipe. >>> >>> Reverting this in 6.16-rc7 restores nouveau acceleration, and glxinfo >>> then reports: "OpenGL renderer string: NVE7" >> >> I also bisected an issue to this commit. On my laptop, this commit >> results in an intermittent desktop crash (Xorg segfault) when changing >> display scale, which can be more reliably reproduced with: >> >> for x in {1..100}; do >> xrandr --output eDP-1 --mode 2560x1600 --scale 0.5 --filter nearest >> xrandr --output eDP-1 --mode 2560x1600 --scale 1 --filter nearest >> done >> > > I won't have time to work on fixing my patch before the merge window, > let's just revert it.@Arnd: Yes, given the short timeframe I think that's the best. Can you please send the revert?