AMDGPU and kernel 5.19
Mario Limonciello
mario.limonciello at amd.com
Fri Jan 20 03:28:12 UTC 2023
Hi,
There is a problem in amdgpu that if not all of IP blocks are supported
or not all of the firmware is present then the framebuffer is destroyed
and the screen freezes. In more recent kernels the amdgpu driver loads
for all PCI VGA class AMD devices.
This effectively means that unless you have all the pieces you need to
support a GPU then the installer freezes unless you do nomodeset.
This problem is to be fully fixed in kernel 6.3 with a ~47 patch series
that is currently in drm-next. This does two basic things:
1) If the IP blocks isn't supported, don't destroy the framebuffer.
2) If firmware for any IP blocks aren't present, don't destroy the
framebuffer.
The whole patch series still needs more time to back in drm-next, but
the most important part of the series is the first patch which
accomplishes "1".
This patch went out to stable 6.1.y as well:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/drivers/gpu/drm/amd?h=linux-6.1.y&id=dcfeba477b3e3df526e0f543b58fa71c045dff8b
As I recently saw a headline that the 22.04 5.19 based HWE stack was
delayed, I strongly think it's worth bringing this patch into that HWE
stack.
My reasoning is that it will ensure that if someone picks up a newer GPU
such as Ryzen 7000 desktop or RDNA3 based they can at least
install Ubuntu without needing to use "Safe Graphics Mode" at the installer.
Sure; they won't have hardware acceleration without the rest of the
kernel and framework, but that's a separate problem to a basic display.
So would you please consider to take the patch linked above into 5.19
with the time afforded by the HWE release delay?
Thanks,
More information about the kernel-team
mailing list