Bug#990279: linux-image-5.10.0-7-powerpc64le: Upgrade to 5.10.0-7-powerpc64le from 5.10.0-0.bpo.5-powerpc64le breaks amdgpu

[Nathaniel Filardo]

I just now noticed Timothy Pearson's note about reverting the
alignment check patch.  An excellent thread to start pulling on, but I
don't think that is the correct fix in itself.  If the addresses
involved are not 64K aligned but merely 4K aligned, then the IOMMU
will have to open an overly wide aperture to include the misaligned
first (and possibly last) pages.

This may suggest that a fix is needed in userspace to ensure proper
alignment, or the misaligned addresses may be coming from somewhere
else in the kernel itself.  I suspect adding some WARN_ON (or perhaps
WARN_ON_ONCE, just in case) will prove informative.  If time permits,
I'll do that, but please don't wait on me.

Cheers,
--nwf;

Bug#990279: linux-image-5.10.0-7-powerpc64le: Upgrade to 5.10.0-7-powerpc64le from 5.10.0-0.bpo.5-powerpc64le breaks amdgpu

[Nathaniel Wesley Filardo]

Package: src:linux
Version: 5.10.40-1
Severity: normal
X-Debbugs-Cc: nwfila...@gmail.com

Dear Maintainer,

Upgrading from 5.10.0-0.bpo.5-powerpc64le=5.10.24-1~bpo10+1 to
5.10.0-7-powerpc64le=5.10.40-1 has broken my X server's ability to use GL, and
has, indeed, lead to markedly poor performance even in things like rxvt.
Quoting /var/log/Xorg.0.log:

[19.084] (II) LoadModule: "glx"
[19.087] (II) Loading /usr/lib/xorg/modules/extensions/libglx.so
[19.102] (II) Module glx: vendor="X.Org Foundation"
[19.143] (II) Loading sub module "glamoregl"
[19.143] (II) LoadModule: "glamoregl"
[19.143] (II) Loading /usr/lib/xorg/modules/libglamoregl.so
[19.159] (II) Module glamoregl: vendor="X.Org Foundation"
[19.573] (II) modeset(0): glamor X acceleration enabled on AMD Radeon (TM) 
Pro WX 7100 Graphics (POLARIS10, DRM 3.40.0, 5.10.0-7-powerpc64le, LLVM 11.0.1)
[19.573] (II) modeset(0): glamor initialized
[19.647] (II) Initializing extension GLX
[19.656] (II) AIGLX: Loaded and initialized radeonsi
[19.656] (II) GLX: Initialized DRI2 GL provider for screen 0
[20.985] (EE) glamor0: GL error: GL_OUT_OF_MEMORY in glTexSubImage
[21.018] (WW) glamor: Failed to allocate 48x48 FBO due to GL_OUT_OF_MEMORY.
[21.018] (WW) glamor: Expect reduced performance.

Things work just fine with the same machine on the old kernel, so I'm holding
back for now, but this isn't a great long-term solution.  This is a Talos II
workstation, making it a little niche and therefore unsurprising that
regressions have crept in, but it'd be nice to get them resolved.  Any
suggestions, advice, requests for further information, &c are all appreciated.
I can rebuild kernels and bisect if pointed at the right git repository and
tags, if that'd be a useful thing.

FWIW, the same thing happened, I believe, with 5.10.0-6-powerpc64le, previously,
but I no longer have it installed to check again.

Thanks much in advance,
--nwf;

-- Package-specific info:
** Version:
Linux version 5.10.0-7-powerpc64le (debian-kernel@lists.debian.org) (gcc-10 
(Debian 10.2.1-6) 10.2.1 20210110, GNU ld (GNU Binutils for Debian) 2.35.2) #1 
SMP Debian 5.10.40-1 (2021-05-28)

** Command line:
root=ZFS=rpool/ROOT/debian ro quiet

** Tainted: POE (12289)
 * proprietary module was loaded
 * externally-built ("out-of-tree") module was loaded
 * unsigned module was loaded

** Kernel log:
[5.306593] [drm] amdgpu kernel modesetting enabled.
[5.309579] amdgpu 0030:01:00.0: vgaarb: deactivate vga console
[5.309753] amdgpu 0030:01:00.0: enabling device (0140 -> 0142)
[5.309767] amdgpu 0030:01:00.0: amdgpu: Trusted Memory Zone (TMZ) feature 
not supported
[5.644486] amdgpu 0030:01:00.0: amdgpu: Fetched VBIOS from ROM BAR
[5.644492] amdgpu: ATOM BIOS: 113-C9540101-103
[5.647802] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_mc.bin
[5.647819] amdgpu 0030:01:00.0: BAR 2: releasing [mem 
0x620001000-0x62000101f 64bit pref]
[5.647822] amdgpu 0030:01:00.0: BAR 0: releasing [mem 
0x62000-0x62fff 64bit pref]
[5.647869] amdgpu 0030:01:00.0: BAR 0: assigned [mem 
0x62000-0x62001 64bit pref]
[5.647879] amdgpu 0030:01:00.0: BAR 2: assigned [mem 
0x62002-0x62002001f 64bit pref]
[5.647913] amdgpu 0030:01:00.0: amdgpu: VRAM: 8192M 0x00F4 - 
0x00F5 (8192M used)
[5.647916] amdgpu 0030:01:00.0: amdgpu: GART: 256M 0x00FF - 
0x00FF0FFF
[5.648123] [drm] amdgpu: 8192M of VRAM memory ready
[5.648128] [drm] amdgpu: 8192M of GTT memory ready.
[5.649983] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_pfp_2.bin
[5.650921] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_me_2.bin
[5.651856] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_ce_2.bin
[5.652822] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_rlc.bin
[5.654748] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_mec_2.bin
[5.657197] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_mec2_2.bin
[5.659428] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_sdma.bin
[5.661056] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_sdma1.bin
[5.661159] amdgpu: hwmgr_sw_init smu backed is polaris10_smu
[5.663358] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_uvd.bin
[5.666276] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_vce.bin
[5.668251] amdgpu 0030:01:00.0: firmware: direct-loading firmware 
amdgpu/polaris10_smc.bin
[5.727666] snd_hda_intel 0030:01:00.1: bound 0030:01:00.0 (ops 
amdgpu_dm_audio_component_bind_ops [amdgpu])
[5.946750] amdgpu 0030:01:00.0: amdgpu: SE 4, SH per SE 1, CU per SH 9, 
active_cu_number 36
[6.091370] amdgpu 0