Currently amdgpu calls drm_sched_fini() from the fence driver sw fini
routine - such function is expected to be called only after the
respective init function - drm_sched_init() - was executed successfully.
Happens that we faced a driver probe failure in the Steam Deck
recently, and the function drm_sched_fini() was called even without
its counter-part had been previously called, causing the following oops:
amdgpu: probe of 0000:04:00.0 failed with error -110
BUG: kernel NULL pointer dereference, address: 0000000000000090
PGD 0 P4D 0
Oops: 0002 [#1] PREEMPT SMP NOPTI
CPU: 0 PID: 609 Comm: systemd-udevd Not tainted 6.2.0-rc3-gpiccoli #338
Hardware name: Valve Jupiter/Jupiter, BIOS F7A0113 11/04/2022
RIP: 0010:drm_sched_fini+0x84/0xa0 [gpu_sched]
[...]
Call Trace:
<TASK>
amdgpu_fence_driver_sw_fini+0xc8/0xd0 [amdgpu]
amdgpu_device_fini_sw+0x2b/0x3b0 [amdgpu]
amdgpu_driver_release_kms+0x16/0x30 [amdgpu]
devm_drm_dev_init_release+0x49/0x70
[...]
To prevent that, check if the drm_sched was properly initialized for a
given ring before calling its fini counter-part.
Notice ideally we'd use sched.ready for that; such field is set as the latest
thing on drm_sched_init(). But amdgpu seems to "override" the meaning of such
field - in the above oops for example, it was a GFX ring causing the crash, and
the sched.ready field was set to true in the ring init routine, regardless of
the state of the DRM scheduler. Hence, we ended-up using sched.ops as per
Christian's suggestion [0], and also removed the no_scheduler check [1].
[0] https://lore.kernel.org/amd-gfx/984ee981-2906-0eaf-ccec-9f80975cb136@amd.com/
[1] https://lore.kernel.org/amd-gfx/cd0e2994-f85f-d837-609f-7056d5fb7231@amd.com/
Fixes: 067f44c8b4 ("drm/amdgpu: avoid over-handle of fence driver fini in s3 test (v2)")
Suggested-by: Christian König <christian.koenig@amd.com>
Cc: Guchun Chen <guchun.chen@amd.com>
Cc: Luben Tuikov <luben.tuikov@amd.com>
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
The pid field corresponds to the result of gettid() in userspace.
However, userspace cannot reliably attribute PTE events to processes
with just the thread id. This patch allows userspace to easily
attribute PTE update events to specific processes by comparing this
field with the result of getpid().
For attributing events to specific threads, the thread id is also
contained in the common fields of each trace event.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Friedrich Vock <friedrich.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Currently amdgpu calls drm_sched_fini() from the fence driver sw fini
routine - such function is expected to be called only after the
respective init function - drm_sched_init() - was executed successfully.
Happens that we faced a driver probe failure in the Steam Deck
recently, and the function drm_sched_fini() was called even without
its counter-part had been previously called, causing the following oops:
amdgpu: probe of 0000:04:00.0 failed with error -110
BUG: kernel NULL pointer dereference, address: 0000000000000090
PGD 0 P4D 0
Oops: 0002 [#1] PREEMPT SMP NOPTI
CPU: 0 PID: 609 Comm: systemd-udevd Not tainted 6.2.0-rc3-gpiccoli #338
Hardware name: Valve Jupiter/Jupiter, BIOS F7A0113 11/04/2022
RIP: 0010:drm_sched_fini+0x84/0xa0 [gpu_sched]
[...]
Call Trace:
<TASK>
amdgpu_fence_driver_sw_fini+0xc8/0xd0 [amdgpu]
amdgpu_device_fini_sw+0x2b/0x3b0 [amdgpu]
amdgpu_driver_release_kms+0x16/0x30 [amdgpu]
devm_drm_dev_init_release+0x49/0x70
[...]
To prevent that, check if the drm_sched was properly initialized for a
given ring before calling its fini counter-part.
Notice ideally we'd use sched.ready for that; such field is set as the latest
thing on drm_sched_init(). But amdgpu seems to "override" the meaning of such
field - in the above oops for example, it was a GFX ring causing the crash, and
the sched.ready field was set to true in the ring init routine, regardless of
the state of the DRM scheduler. Hence, we ended-up using sched.ops as per
Christian's suggestion [0], and also removed the no_scheduler check [1].
[0] https://lore.kernel.org/amd-gfx/984ee981-2906-0eaf-ccec-9f80975cb136@amd.com/
[1] https://lore.kernel.org/amd-gfx/cd0e2994-f85f-d837-609f-7056d5fb7231@amd.com/
Fixes: 067f44c8b4 ("drm/amdgpu: avoid over-handle of fence driver fini in s3 test (v2)")
Suggested-by: Christian König <christian.koenig@amd.com>
Cc: Guchun Chen <guchun.chen@amd.com>
Cc: Luben Tuikov <luben.tuikov@amd.com>
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The pid field corresponds to the result of gettid() in userspace.
However, userspace cannot reliably attribute PTE events to processes
with just the thread id. This patch allows userspace to easily
attribute PTE update events to specific processes by comparing this
field with the result of getpid().
For attributing events to specific threads, the thread id is also
contained in the common fields of each trace event.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Friedrich Vock <friedrich.vock@gmx.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The path for the "mod_info_packet.h" header file is
incomplete, so add its location to the header search path
in the amdgpu Makefile.
See on ARCH=alpha (275 times in one build).
In file included from ../drivers/gpu/drm/amd/amdgpu/amdgpu.h:90,
from ../drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c:43:
../drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.h:62:10: fatal error: mod_info_packet.h: No such file or directory
62 | #include "mod_info_packet.h"
| ^~~~~~~~~~~~~~~~~~~
compilation terminated.
Fixes: 5b49da02dd ("drm/amd/display: Enable Freesync over PCon")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Signed-off-by: Sung Joon Kim <sungkim@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 1a65327a84.
This should be resolved so we can re-enable this. Also,
the AGP apeture was bring programmed to 0 on MMHUB 3.0.1
since agp_start and end were not being set.
Acked-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Accessing GRBM_GFX_CNTL in full access time has risk when VF is doing MMIO attacking.
Therefore, VF writing GRBM_GFX_CNTL are blocked by L1 Policy.
For RLCG interface, RLCG use SCRATCH_REG2 which is copied from GRBM_GFX_CNTL.
[How]
Remove writing GRBM_GFX_CNTL in amdgpu_virt_rlcg_reg_rw.
v2:
Remove directly writing GRBM_GFX_INDEX in amdgpu_virt_rlcg_reg_rw
as RLCG interface no need to use it.
Signed-off-by: Yifan Zha <Yifan.Zha@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We allow sending PSP messages LOAD_ASD and UNLOAD_TA without
acquiring a lock in drm_dev_enter during driver unload
because we must call drm_dev_unplug as the beginning
of unload driver sequence.
Added WARNING if other PSP messages are sent without a lock.
After this commit, the following commands would work
-sudo modprobe -r amdgpu
-sudo modprobe amdgpu
Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Reviewed-by Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
SQ_WAVE_INST_DW0 isn't present on gfx11 compared to gfx10, so update
wave data type to signify a difference.
Signed-off-by: Graham Sider <Graham.Sider@amd.com>
Reviewed-by: Mukul Joshi <Mukul.Joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We guard the suspend entry code from running unless we have proper
BIOS support for either S3 mode or s0ix mode.
If a user's system doesn't support either of these modes the kernel
still does offer s2idle in `/sys/power/mem_sleep` so there is an
expectation from users that it works even if the power consumption
remains very high.
Rafael Ávila de Espíndola reports that a system of his has a
non-functional graphics stack after resuming. That system doesn't
support S3 and the FADT doesn't indicate support for low power idle.
Through some experimentation it was concluded that even without the
hardware s0i3 support provided by the amd_pmc driver the power
consumption over suspend is decreased by running amdgpu's s0ix
suspend routine.
The numbers over suspend showed:
* No patch: 9.2W
* Skip amdgpu suspend entirely: 10.5W
* Run amdgpu s0ix routine: 7.7W
As this does improve the power, remove some of the guard rails in
`amdgpu_acpi.c` for only running s0ix suspend routines in the right
circumstances.
However if this turns out to cause regressions for anyone, we should
revert this change and instead opt for skipping suspend/resume routines
entirely or try to fix the underlying behavior that makes graphics fail
after resume without underlying platform support.
Reported-by: Rafael Ávila de Espíndola <rafael@espindo.la>
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2364
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit fac53471d0.
The following change: move the drm_dev_unplug call after
amdgpu_driver_unload_kms in amdgpu_pci_remove. The reason is
the following: amdgpu_pci_remove calls drm_dev_unregister
and it should be called first to ensure userspace can't access the
device instance anymore. If we call drm_dev_unplug after
amdgpu_driver_unload_kms then we observe IGT PCI software unplug
test failure (kernel hung) for all ASICs. This is how this
regression was found.
After this revert, the following commands do work not, but it would
be fixed in the next commit:
- sudo modprobe -r amdgpu
- sudo modprobe amdgpu
Signed-off-by: Vitaly Prosyak <vitaly.prosyak@amd.com>
Reviewed-by Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Due to holidays we started -next with more -fixes in-flight than
usual, and people have been asking where they are. Backmerge to get
things better in sync.
Conflicts:
- Tiny conflict in drm_fbdev_generic.c between variable rename and
missing error handling that got added.
- Conflict in drm_fb_helper.c between the added call to vgaswitcheroo
in drm_fb_helper_single_fb_probe and a refactor patch that extracted
lots of helpers and incidentally removed the dev local variable.
Readd it to make things compile.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
AV1 is only supported on the first instance.
Added vcn_v4_0_enc_find_ib_param() to help search for an IB param.
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This allows to correlate the infos printed by
/sys/kernel/debug/dri/n/amdgpu_gem_info to the ones found
in /proc/.../fdinfo and /sys/kernel/debug/dma_buf/bufinfo.
Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Laptops with APUs from a variety of manufacturers and generations
show a warning about a missing PSP runtime database.
As it's not required for PSP to dump this database into framebuffer,
decrease messages about it missing to debug.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
