r300: Regression in f424ef18010 breaks wayland on RS480M

[Valtteri Koskivuori]

Hello all.

I run Arch Linux on a 2006 Fujitsu S2110 laptop that has sentimental 
value to me. Earlier in 2024, I observed a regression that broke 
wayland-based compositors, such as Sway and Cage. The framebuffer shows 
up garbled, and the following error is emitted to the kernel log:

[drm:radeon_crtc_do_set_base [radeon]] *ERROR* trying to scanout microtiled buffer

I spent the past weekend tracking this down, and managed to set up for a 
bisect, which lead me to that specific commit, which was a part of merge 
request !28209 in March.
Mesa seems to be setting the tiling flag RADEON_TILING_MICRO, which then 
later gets checked in the Linux driver, giving that error above.

TL;DR: reverting f424ef18010 on the latest HEAD resolves this issue for me, 
though I'm not familiar enough with the code to determine if reverting 
it results in adverse effects on other hardware. More details below for 
the interested.

GPU: [AMD/ATI] RS480M [Mobility Radeon Xpress 200]
CPU: AMD Turion(tm) 64 Mobile Technology MT-37
Linux arch 6.11.1-arch1-1 #1 SMP PREEMPT_DYNAMIC Mon, 30 Sep 2024 23:49:50 +0000 x86_64 GNU/Linux
Mesa 24.2.3-1 (arch package)
Sway and Cage were broken, but i3 seems to be unaffected.

I hope this information is helpful. I'd be happy to help with testing a 
potential fix, if needed.

Thank you all for the work you do.
Br,
Valtteri Koskivuori
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20241008/0add33b3/attachment.sig>