When amdgpu_ctx_mgr_fini() calls amdgpu_ctx_mgr_entity_fini() it contains
the exact same "context still alive" check as it will do next. Remove the
duplicated copy.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Function amdgpu_ctx_mgr_entity_fini() only has a single local caller so
lets make it local.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Don't enable BACO when in passthrough. PCI resets don't work
correctly when in BACO.
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
xgmi_port_num interface reports external link number for port number. To
be consistent, use the external link number for reporting other XGMI
link data also.
v2: For invalid link number return -EINVAL (Kevin)
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Yang Wang <kevinyang.wang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add sysfs nodes to provide compute paritition specific data.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The current cleanup order during file descriptor close can lead to
a race condition where the eviction fence worker attempts to access
a destroyed mutex from the user queue manager:
[ 517.294055] DEBUG_LOCKS_WARN_ON(lock->magic != lock)
[ 517.294060] WARNING: CPU: 8 PID: 2030 at kernel/locking/mutex.c:564
[ 517.294094] Workqueue: events amdgpu_eviction_fence_suspend_worker [amdgpu]
The issue occurs because:
1. We destroy the user queue manager (including its mutex) first
2. Then try to destroy eviction fences which may have pending work
3. The eviction fence worker may try to access the already-destroyed mutex
Fix this by reordering the cleanup to:
1. First mark the fd as closing and destroy eviction fences,
which flushes any pending work
2. Then safely destroy the user queue manager after we're certain
no more fence work will be executed
The copy in amdgpu_driver_postclose_kms() needs to be removed (Christian)
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Lock and refer to the eviction fence before the eviction fence
schedules work queue tries to signal it.
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
irq_domain_add_linear() is going away as being obsolete now. Switch to
the preferred irq_domain_create_linear(). That differs in the first
parameter: It takes more generic struct fwnode_handle instead of struct
device_node. Therefore, of_fwnode_handle() is added around the
parameter.
Note some of the users can likely use dev->fwnode directly instead of
indirect of_fwnode_handle(dev->of_node). But dev->fwnode is not
guaranteed to be set for all, so this has to be investigated on case to
case basis (by people who can actually test with the HW).
[ tglx: Fix up subject prefix ]
Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20250319092951.37667-19-jirislaby@kernel.org
The eviction fence destroy path incorrectly calls dma_fence_put() on
evf_mgr->ev_fence after releasing the ev_fence_lock. This introduces a
potential use-after-unlock or race because another thread concurrently
modifies evf_mgr->ev_fence.
Fix this by grabbing a local reference to evf_mgr->ev_fence under the
lock and using that for dma_fence_put() after waiting.
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
These structures are basically ported from MMSCH v4_0
The structures are the same as v4_0 except for the
init header
Signed-off-by: fanhuang <FangSheng.Huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Compatible NPS modes for a partition mode are exposed through xcp_config
interface. To determine if a compute partition mode is valid, check if
the current NPS mode is part of compatible NPS modes.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If the MEC firmware supports chaining runlists of XNACK+/XNACK-
processes, set SQ_CONFIG1 chicken bit and SET_RESOURCES bit 28.
When the MEC/HWS supports it, KFD checks the XNACK+/XNACK- processes mix
happens or not. If it does, enter over-subscription.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Reviewed-by: Philip Yang <Philip.Yang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On VCN v4.0.5 there is a race condition where the WPTR is not
updated after starting from idle when doorbell is used. Adding
register read-back after written at function end is to ensure
all register writes are done before they can be used.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Tested-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 07c9db090b)
Cc: stable@vger.kernel.org
Expose the debugfs file node for user space to dump the IFWI image
on spirom.
For one transaction between PSP and host, it will read out the
images on both active and inactive partitions so a buffer with two
times the size of maximum IFWI image (currently 16MByte) is needed.
v2: move the vbios gfl macros to the common header and rename the
bo triplet struct to spirom_bo for this specific usage (Hawking)
v3: return directly the result of last command execution (Lijo)
Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
As the userq resource was already freed at the drm_release
early phase, it should avoid freeing userq resource again
at the later kms postclose callback.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
A circular locking dependency was detected between the global
`adev->userq_mutex` and per-file `userq_mgr->userq_mutex` when
creating user queues. The issue occurs because:
1. `amdgpu_userq_suspend()` and `amdgpu_userq_resume` take `adev->userq_mutex` first, then
`userq_mgr->userq_mutex`
2. While `amdgpu_userq_create()` takes them in reverse order
This patch resolves the issue by:
1. Moving the `adev->userq_mutex` lock earlier in `amdgpu_userq_create()`
to cover the `amdgpu_userq_ensure_ev_fence()` call
2. Releasing it after we're done with both queue creation and the
scheduling halt check
v2: remove unused adev->userq_mutex lock (Prike)
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On VCN v4.0.5 there is a race condition where the WPTR is not
updated after starting from idle when doorbell is used. Adding
register read-back after written at function end is to ensure
all register writes are done before they can be used.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528
Signed-off-by: David (Ming Qiang) Wu <David.Wu3@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Tested-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On GFX1151, the reported MALL cache size reflects only
half of its actual size; this adjustment corrects the discrepancy.
Signed-off-by: Tim Huang <tim.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 0a5c060b59)
Cc: stable@vger.kernel.org
After process exit to unmap csa and free GPU vm, if signal is accepted
and then waiting to take vm lock is interrupted and return, it causes
memory leaking and below warning backtrace.
Change to use uninterruptible wait lock fix the issue.
WARNING: CPU: 69 PID: 167800 at amd/amdgpu/amdgpu_kms.c:1525
amdgpu_driver_postclose_kms+0x294/0x2a0 [amdgpu]
Call Trace:
<TASK>
drm_file_free.part.0+0x1da/0x230 [drm]
drm_close_helper.isra.0+0x65/0x70 [drm]
drm_release+0x6a/0x120 [drm]
amdgpu_drm_release+0x51/0x60 [amdgpu]
__fput+0x9f/0x280
____fput+0xe/0x20
task_work_run+0x67/0xa0
do_exit+0x217/0x3c0
do_group_exit+0x3b/0xb0
get_signal+0x14a/0x8d0
arch_do_signal_or_restart+0xde/0x100
exit_to_user_mode_loop+0xc1/0x1a0
exit_to_user_mode_prepare+0xf4/0x100
syscall_exit_to_user_mode+0x17/0x40
do_syscall_64+0x69/0xc0
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 7dbbfb3c17)
Cc: stable@vger.kernel.org
This patch enables per-queue and per-pipe reset functionality for
GFX IP v9.5.0 when using MEC firmware version 21 (0x15) or later.
This change:
1. Refactors the pipe reset support check in gfx_v9_4_3_pipe_reset_support()
to use the compute_supported_reset flags instead of hardcoding
version checks.
2. Adds support for GFX9.5.0 (IP 9.5.0) with MEC firmware version >= 21
to enable per-queue and per-pipe reset capabilities.
v2: Replaced mec version check with !!(adev->gfx.compute_supported_reset & AMDGPU_RESET_TYPE_PER_PIPE) (Lijo)
Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set vram type so we can take different actions according to the type.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make the code more general, user doesn't need to pay attention to the
detail of flip bits setting.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The number of newly added de counts and the number of
newly added error addresses remain consistent
Signed-off-by: Ce Sun <cesun102@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
During driver load, RAS event manager may not be initialized. This will
cause any ATHUB event during driver load to be skipped in dmesg log. Log
the error in dmesg log for easier diagnosis.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This resolves a deadlock between user queue management and GPU reset
paths by enforcing consistent lock ordering.
The deadlock occurred when:
1. Process exit path (amdgpu_userq_mgr_fini) would:
- Take uqm->userq_mutex
- Then try to take adev->userq_mutex for list operations
2. GPU reset path (amdgpu_userq_pre_reset) would:
- Take adev->userq_mutex first (for list traversal)
- Then take uqm->userq_mutex
The solution establishes a strict top-down locking order:
1. Always take adev->userq_mutex before any uqm->userq_mutex
2. Maintain this order consistently across all code paths
Changes made:
- Reordered locking in amdgpu_userq_mgr_fini() to take device lock first
- Kept existing proper order in amdgpu_userq_pre_reset()
- Simplified the fini flow by removing redundant operations
This prevents circular dependencies while maintaining thread safety
during both normal operation and GPU reset scenarios.
Fixes: 4ce60dbada ("drm/amdgpu: store userq_managers in a list in adev")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Get RAS retire flip bits for HBM with different types in various NPS modes.
Also set flip row bit and MCA R13 bit in PA in different NPS modes.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The RAS bad page retire flip bits can be set per vram type,
vram vendor and NPS mode.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add the general interface to get flip bits for RAS bad page retirement.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Per UMC address conversion algorithm, the high row bits of UMC MCA
address are changed when they're converted into normalized address
on specific ASICs.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Support new version of HBM.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
there is only MCA records in V3, no need to care about PA records.
recalculate the value of ras_num_bad_pages when parsing failed and
go on with the left records instead of quit.
Signed-off-by: ganglxie <ganglxie@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On GFX1151, the reported MALL cache size reflects only
half of its actual size; this adjustment corrects the discrepancy.
Signed-off-by: Tim Huang <tim.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
To resolve the warning regarding the missing error code 'r' in
amdgpu_userq_wait_ioctl(), assign the value 'r = -EINVAL'.
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/r/202505080458.rnV8YfiY-lkp@intel.com/
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
After process exit to unmap csa and free GPU vm, if signal is accepted
and then waiting to take vm lock is interrupted and return, it causes
memory leaking and below warning backtrace.
Change to use uninterruptible wait lock fix the issue.
WARNING: CPU: 69 PID: 167800 at amd/amdgpu/amdgpu_kms.c:1525
amdgpu_driver_postclose_kms+0x294/0x2a0 [amdgpu]
Call Trace:
<TASK>
drm_file_free.part.0+0x1da/0x230 [drm]
drm_close_helper.isra.0+0x65/0x70 [drm]
drm_release+0x6a/0x120 [drm]
amdgpu_drm_release+0x51/0x60 [amdgpu]
__fput+0x9f/0x280
____fput+0xe/0x20
task_work_run+0x67/0xa0
do_exit+0x217/0x3c0
do_group_exit+0x3b/0xb0
get_signal+0x14a/0x8d0
arch_do_signal_or_restart+0xde/0x100
exit_to_user_mode_loop+0xc1/0x1a0
exit_to_user_mode_prepare+0xf4/0x100
syscall_exit_to_user_mode+0x17/0x40
do_syscall_64+0x69/0xc0
Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Except HDP v5.2 all use a common logic for HDP flush. Use a generic
function. HDP v5.2 forces NO_KIQ logic, revisit it later.
Reapply after fixing up an HDP regression.
v2: merge the fix (Alex)
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com> (v1)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
APU doesn't have second IH ring, so re-routing action here is a no-op.
It will take a lot of time to wait timeout from PSP during the
initialization. So remove the function in psp v12.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the s3/s0ix and s4 flags in the pm notifier so that we can skip
the resource evictions properly in pm prepare based on whether
we are suspending or hibernating. Drop the eviction as processes
are not frozen at this time, we we can end up getting stuck trying
to evict VRAM while applications continue to submit work which
causes the buffers to get pulled back into VRAM.
v2: Move suspend flags out of pm notifier (Mario)
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4178
Fixes: 2965e6355d ("drm/amd: Add Suspend/Hibernate notification callback support")
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 06f2dcc241)
Cc: stable@vger.kernel.org
This reverts commit 3a9626c816.
This breaks S4 because we end up setting the s3/s0ix flags
even when we are entering s4 since prepare is used by both
flows. The causes both the S3/s0ix and s4 flags to be set
which breaks several checks in the driver which assume they
are mutually exclusive.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3634
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit ce8f7d9589)
Cc: stable@vger.kernel.org
VCN1_AON_SOC_ADDRESS_3_0 offset varies on different
VCN generations, the issue in vcn4.0.5 is caused by
a different VCN1_AON_SOC_ADDRESS_3_0 offset.
This patch does the following:
1. use the same offset for other VCN generations.
2. use the vcn4.0.5 special offset
3. update vcn_4_0 and vcn_5_0
Acked-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 5c89ceda99)
Cc: stable@vger.kernel.org
If there is a problem requiring a reset of the VCN engine, it is better to
reset the VCN engine rather than the entire GPU.
Add a reset callback for the ring which will stop and start VCN if an
issue happens.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20250506204948.12048-4-mario.limonciello@amd.com
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If there is a problem requiring a reset of the VCN engine, it is better to
reset the VCN engine rather than the entire GPU.
Add a reset callback for the ring which will stop and start VCN if an
issue happens.
Link: https://lore.kernel.org/r/20250506204948.12048-3-mario.limonciello@amd.com
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>
There is a problem occurring on VCN 4.0.5 where in some situations a job
is timing out. This triggers a job timeout which then causes a GPU
reset for recovery. That has exposed a number of issues with GPU reset
that have since been fixed. But also a GPU reset isn't actually needed
for this circumstance. Just restarting the ring is enough.
Add a reset callback for the ring which will stop and start VCN if the
issue happens.
Link: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12528
Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3909
Link: https://lore.kernel.org/r/20250506204948.12048-2-mario.limonciello@amd.com
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>
This reverts commit 18a878fd8a.
Revert this temporarily to make it easier to fix a regression
in the HDP handling.
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since secure firmware is more stable than bring up phase, I believe we
don't need such mdelays any more before wait PSP response on PSP v12.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Trigger Huang <Trigger.Huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If applications unmap the memory before destroying the userptr, it needs
trigger a segfault to notify user space to correct the free sequence in
VM debug mode.
v2: Send gpu access fault to user space
v3: Report gpu address to user space, remove unnecessary params
v4: update pr_err into one line, remove userptr log info
Signed-off-by: Shane Xiao <shane.xiao@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In VM debug mode, it is desirable to notify the application
to correct the freeing sequence by unmapping the memory before
destroying the userptr in the old userptr path. Add a bitmask
to decide whether to send gpu vm fault to the applition.
Signed-off-by: Shane Xiao <shane.xiao@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It requires unlocking the reserved gem BO before returning from
attaching the eviction fence error.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The driver doesn't want to implicitly sync on the DMA_RESV_USAGE_BOOKKEEP
usage fences, and the BOOKEEP fences should be synced explicitly. So, as
the VM implicit syncing only need to return and sync the dependent read
fences.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This is an optional TA which is only available on
certain embedded systems. Mark it as optional to avoid
user confusion. This mirrors what we already do for
other optional TAs.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4181
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the s3/s0ix and s4 flags in the pm notifier so that we can skip
the resource evictions properly in pm prepare based on whether
we are suspending or hibernating. Drop the eviction as processes
are not frozen at this time, we we can end up getting stuck trying
to evict VRAM while applications continue to submit work which
causes the buffers to get pulled back into VRAM.
v2: Move suspend flags out of pm notifier (Mario)
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4178
Fixes: 2965e6355d ("drm/amd: Add Suspend/Hibernate notification callback support")
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This notification may arrive in VF mailbox while polling for response from
another event.
This patches covers the following scenarios:
- If VF is already in RMA state, then do not attempt to contact the host.
Host will ignore the VF after sending the notification.
- If the notification is detected during polling, then set the RMA status,
and return error to caller.
- If the notification arrives by interrupt, then set the RMA status and
queue a reset. This reset will fail and VF will stop runtime services.
Reviewed-by: Shravan Kumar Gande <Shravankumar.Gande@amd.com>
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Host may stop runtime services after reaching a bad page threshold.
This notification will indicate to the VF that it no longer has
access to the GPU.
Reviewed-by: Shravan Kumar Gande <Shravankumar.Gande@amd.com>
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 3a9626c816.
This breaks S4 because we end up setting the s3/s0ix flags
even when we are entering s4 since prepare is used by both
flows. The causes both the S3/s0ix and s4 flags to be set
which breaks several checks in the driver which assume they
are mutually exclusive.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3634
Cc: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
VCN1_AON_SOC_ADDRESS_3_0 offset varies on different
VCN generations, the issue in vcn4.0.5 is caused by
a different VCN1_AON_SOC_ADDRESS_3_0 offset.
This patch does the following:
1. use the same offset for other VCN generations.
2. use the vcn4.0.5 special offset
3. update vcn_4_0 and vcn_5_0
Acked-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The dma_resv_add_fence() already refers to the added fence.
So when attaching the evciton fence to the gem bo, it needn't
refer to it anymore.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Host will send a notification when new bad pages are available.
Uopn guest request, the first 256 bad page addresses
will be placed into the PF2VF region.
Guest should pause the PF2VF worker thread while
the copy is in progress.
Reviewed-by: Shravan Kumar Gande <Shravankumar.Gande@amd.com>
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently VFs rely on poison consumption interrupt from HW
to kick off the bad page retirement process. Part of this process
includes a VF reset.
This patch adds the following:
1) Host Bad Pages notification message.
2) Guest request bad pages message.
When combined, VFs are able to reserve the pages early, and potentially
avoid future poison consumption that will disrupt user services
from consequent FLR.
Reviewed-by: Shravan Kumar Gande <Shravankumar.Gande@amd.com>
Signed-off-by: Victor Skvortsov <victor.skvortsov@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add some random documentation associated with the ring buffer
manipulations and writeback.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
-----BEGIN PGP SIGNATURE-----
iQFSBAABCgA8FiEEq68RxlopcLEwq+PEeb4+QwBBGIYFAmgX1CgeHHRvcnZhbGRz
QGxpbnV4LWZvdW5kYXRpb24ub3JnAAoJEHm+PkMAQRiGxiIH/A7LHlVatGEQgRFi
0JALDgcuGTMtMU1qD43rv8Z1GXqTpCAlaBt9D1C9cUH/86MGyBTVRWgVy0wkaU2U
8QSfFWQIbrdaIzelHtzmAv5IDtb+KrcX1iYGLcMb6ZYaWkv8/CMzMX1nkgxEr1QT
37Xo3/F17yJumAdNQxdRhVLGy2d3X5rScecpufwh97sMwoddllMCDs2LIoeSAYpG
376/wzni09G2fADa8MEKqcaMue4qcf0FOo/gOkT8YwFGSZLKa6uumlBLg04QoCt0
foK2vfcci1q4H4ZbCu3uQESYGLQHY0f2ICDCwC3m25VF9a81TmlbC3MLum3vhmKe
RtLDcXg=
=xyaI
-----END PGP SIGNATURE-----
BackMerge tag 'v6.15-rc5' into drm-next
Linux 6.15-rc5, requested by tzimmerman for fixes required in drm-next.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Keep only the latest fences to reduce the number of values
given back to userspace
v2: - Export this code from dma-fence-unwrap.c(by Christian).
v3: - To split this in a dma_buf patch and amd userq patch(by Sunil).
- No need to add a new function just re-use existing(by Christian).
v4: Export dma_fence_dedub_array function and used it(by Christian).
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Adjusted the enforce isolation setting handling to include the ability
to disable the cleaner shader without affecting isolation between tasks.
v2: Updated enforce isolation documentation and parameters. (Alex)
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
update the functions in amdgpu_userqueues.c from
DRM_DBG_DRIVER to drm_dbg_driver so multi gpu instance
can be logged in.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
change the DRM_ERROR and drm_err to drm_file_err
to add process name and pid to the logging.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drm_file will be used in usermode queues code to
enable better process information in logging and hence
add drm_file part of the userq_mgr struct.
update the drm_file pointer in userq_mgr for each
amdgpu_driver_open_kms.
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For vcn5.0.1 only, enable DPG PAUSE to avoid DPG resets.
Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 3e5f86c14c)
APUs in passthrough mode use HDP flush. 0x7F000 offset used for
remapping HDP flush is mapped to VPE space which could get power gated.
Use another unused offset in BIF space.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit d8116a32cd)
Cc: stable@vger.kernel.org
If peer memory is XGMI-accessible, we should never access it through PCIe
P2P DMA mappings. PCIe P2P is slower, has different coherence behaviour,
limited or no support for atomics, or may not work at all. Fail with a
warning if DMABUF mappings of such memory are attempted.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit dbe4c63689)
The "ticket" pointer points to in the middle of the &exec struct so it
can't be NULL. Remove the check.
Reviewed-by: Christian König <christian.koenig@amd.com>
Acked-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This error path should call amdgpu_bo_unreserve() before returning.
Fixes: d8675102ba ("drm/amdgpu: add vm root BO lock before accessing the vm")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When kernel queues are disabled, all GC vmids are available
for the scheduler. MM vmids are still managed by the driver
so make all 16 available.
Also fix gmc 10 vs 11 mix up in
commit 1f61fc28b9 ("drm/amdgpu/mes: make more vmids available when disable_kq=1")
v2: Properly handle pre-GC 10 hardware
Fixes: 1f61fc28b9 ("drm/amdgpu/mes: make more vmids available when disable_kq=1")
Cc: Arvind Yadav <Arvind.Yadav@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the KIQ pipe for kernel queues and the SCHED pipe for
user queues.
Fixes: 2408b0272b ("drm/amdgpu/mes: consolidate on a single mes reset callback")
Cc: Michael Chen <Michael.Chen@amd.com>
Cc: Shaoyun Liu <Shaoyun.Liu@amd.com>
Reviewed-by: Michael Chen <michael.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the legacy one as it covers both kernel queues and
user queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
These were leftover from mes bring up and are unused.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If it needs a long wait for completion of bootloader execution, report
the status in between. That helps to know if there is some issue during
bootloader execution.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Register access print missed device info.
[How]
Using dev_xxx instead of DRM_xxx to indicate which device
of a hive is the message for.
Signed-off-by: Yifan Zha <Yifan.Zha@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Keep the register offsets as per link order for querying XGMI v6.4.1
link status.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Tested-by: Mangesh Gadre <Mangesh.Gadre@amd.com>
Fixes: 6dee64e765 ("drm/amdgpu: Fix xgmi v6.4.1 link status reporting")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The mes_v11_0_reset_hw_queue and mes_v12_0_reset_hw_queue functions were
using the wrong union type (MESAPI__REMOVE_QUEUE) when getting the offset
for api_status. Since these functions handle queue reset operations, they
should use MESAPI__RESET union instead.
This fixes the polling of API status during hardware queue reset operations
in the MES for both v11 and v12 versions.
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-By: Shaoyun.liu <Shaoyun.liu@amd.com>
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For vcn5.0.1 only, enable DPG PAUSE to avoid DPG resets.
Signed-off-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
DRM_AMDGPU_NAVI3X_USERQ config support is not required for
usermode queue.
v2: rebase.
Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a description for the get_csb_buffer callback, update the glossary,
and add some extra information about RB, which is associated with CSB
configuration.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove duplications from gfx_v6_0_get_csb_buffer by using CSB helpers.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Instead of having the hardcoded values for the CSB buffer in
gfx_v7_0_get_csb_buffer, use the values calculated in previous steps by
accessing raster_config and raster_config_1.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The evf fence name can clearly identify the userq usage.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
APUs in passthrough mode use HDP flush. 0x7F000 offset used for
remapping HDP flush is mapped to VPE space which could get power gated.
Use another unused offset in BIF space.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This debug dump will help on debugging the evf attached gem obj fence
related issue.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If peer memory is XGMI-accessible, we should never access it through PCIe
P2P DMA mappings. PCIe P2P is slower, has different coherence behaviour,
limited or no support for atomics, or may not work at all. Fail with a
warning if DMABUF mappings of such memory are attempted.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use CSB helpers to remove code duplication from gfx_v7_0_get_csb_buffer.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove code duplication from gfx_v8_0_get_csb_buffer by using CSB
helpers.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Eliminate code duplication in gfx_v9_0_get_csb_buffer by using CSB
helpers.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove duplicate code by using CSB helpers.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Part of the code in gfx_v11_0_get_csb_buffer can be removed in favor of
some GFX CSB helpers. This commit removes the duplicated part for the
GFX 11 CSB function.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
From GFX6 to GFX11, there is a function for getting the CSB buffer to be
put into the hardware. Three common parts are duplicated in all of these
GFX functions:
1. Prepare the CSB preamble.
2. Parser the CS data.
3. End the CSB preamble.
This commit creates helpers to be used from GFX6 to GFX11.
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There is a spelling mistake with the array utcl2_rounter_str, it
appears it should be utcl2_router_str. Fix it.
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
GCC really does not want to consider NULL (or near-NULL) addresses as
valid, so calculations based off of NULL end up getting range-tracked into
being an offset wthin a 0 byte array. It gets especially mad about this:
if (vbios_str == NULL)
vbios_str += sizeof(BIOS_ATOM_PREFIX) - 1;
...
if (vbios_str != NULL && *vbios_str == 0)
vbios_str++;
It sees this as being "sizeof(BIOS_ATOM_PREFIX) - 1" byte offset from
NULL, when building with -Warray-bounds (and the coming
-fdiagnostic-details flag):
In function 'atom_get_vbios_pn',
inlined from 'amdgpu_atom_parse' at drivers/gpu/drm/amd/amdgpu/atom.c:1553:2:
drivers/gpu/drm/amd/amdgpu/atom.c:1447:34: error: array subscript 0 is outside array bounds of 'unsigned char[0]' [-Werror=array-bounds=]
1447 | if (vbios_str != NULL && *vbios_str == 0)
| ^~~~~~~~~~
'amdgpu_atom_parse': events 1-2
1444 | if (vbios_str == NULL)
| ^
| |
| (1) when the condition is evaluated to true
......
1447 | if (vbios_str != NULL && *vbios_str == 0)
| ~~~~~~~~~~
| |
| (2) out of array bounds here
In function 'amdgpu_atom_parse':
cc1: note: source object is likely at address zero
As there isn't a sane way to convince it otherwise, hide vbios_str from
GCC's optimizer to avoid the warning so we can get closer to enabling
-Warray-bounds globally.
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Kees Cook <kees@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reject queries to get current partition modes during reset. Also, don't
accept sysfs interface requests to switch compute partition mode while
in reset.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Asad Kamal <asad.kamal@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
It turns out that there are no platforms that have PCI but don't have an
MMU, so adding a Kconfig dependency on CONFIG_PCI simplifies build testing
kernels for those platforms a lot, and avoids a lot of inadvertent build
regressions.
Add a dependency for CONFIG_PCI and remove all the ones for PCI specific
device drivers that are currently marked not having it.
There are a few platforms that have an optional MMU, but they usually
cannot have PCI at all. The one exception is Coldfire MCF54xx, but this is
mainly for historic reasons, and anyone using those chips should really use
the MMU these days.
Link: https://lore.kernel.org/lkml/a41f1b20-a76c-43d8-8c36-f12744327a54@app.fastmail.com/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> # SCSI
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patch.msgid.link/20250423202215.3315550-1-arnd@kernel.org
If peer memory is accessible through XGMI, allow leaving it in VRAM
rather than forcing its migration to GTT on DMABuf attachment.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 372c8d72c3)
Pinning of VRAM is for peer devices that don't support dynamic attachment
and move notifiers. But it requires that all such peer devices are able to
access VRAM via PCIe P2P. Any device without P2P access requires migration
to GTT, which fails if the memory is already pinned for another peer
device.
Sharing between GPUs should not require pinning in VRAM. However, if
DMABUF_MOVE_NOTIFY is disabled in the kernel build, even DMABufs shared
between GPUs must be pinned, which can lead to failures and functional
regressions on systems where some peer GPUs are not P2P accessible.
Disable VRAM pinning if move notifiers are disabled in the kernel build
to fix regressions when sharing BOs between GPUs.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 05185812ae)
When determining the domains for pinning DMABufs, filter allowed_domains
and fail with a warning if VRAM is forbidden and GTT is not an allowed
domain.
Fixes: f5e7fabd1f ("drm/amdgpu: allow pinning DMA-bufs into VRAM if all importers can do P2P")
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 3940796a6e)
log context:seqno of the fence during timeout rather
than logging fence pointer.
Reviewed-by: Arvind Yadav <Arvind.Yadav@amd.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch is load usermode queue based on FW support for gfx12.
CP Ucode FW Vesion: [PFP = 2840, ME = 2780, MEC = 3050, MES = 123]
v2: Addressed review comments from Alex
- Just check the firmware versions directly.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Shashank Sharma <shashank.sharma@amd.com>
Cc: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch is load usermode queue based on FW support for gfx11.
CP Ucode FW version: [PFP = 2530, ME = 2390, MEC = 2600, MES = 120]
v2: Addressed review comments from Alex.
- Just check the firmware versions directly.
v3: Firmware version checks only for Navi3x(by Alex).
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Shashank Sharma <shashank.sharma@amd.com>
Cc: Sunil Khatri <sunil.khatri@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Arvind Yadav <Arvind.Yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
s/userqueue/userq/
1. remove the mix of amdgpu_userqueue and amdgpu_userq
2. to be consistent with other amdgpu_userq_fence.c
3. it's shorter
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
suspend/resume -> evict/restore
Rename to avoid confusion with the system suspend
and resume helpers.
v2: update error messages
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Need to wait for the last fence before unmapping. This
also fixes a memory leak in amdgpu_userqueue_cleanup()
when the fence isn't signalled.
Fixes: b0db33c8c5 ("drm/amdgpu/userq: rework front end call sequence")
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This was missed when the map and unmap were split out
of the mqd create and destroy functions.
Fixes: b0db33c8c5 ("drm/amdgpu/userq: rework front end call sequence")
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Track the state of the queue rather than simple active vs
not. This is needed for other states (hung, preempted, etc.).
While we are at it, move the state tracking into the user
queue front end code.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If user queues are disabled for all IPs in the case
of suspend and resume and for gfx/compute in the case
of enforce isolation, we can return early.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Print a kernel message when the scrub bit of status register is set to
indicate that errors are being logged by the scrub.
Signed-off-by: Xiang Liu <xiang.liu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a helper to get a mask of IPs which support user queues.
Use this in the INFO IOCTL to get the IP mask to replace
the current code.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add queue id support to the user queue wait IOCTL
drm_amdgpu_userq_wait structure.
This is required to retrieve the wait user queue and maintain
the fence driver references in it so that the user queue in
the same context releases their reference to the fence drivers
at some point before queue destruction.
Otherwise, we would gather those references until we
don't have any more space left and crash.
v2: Modify the UAPI comment as per the mesa and libdrm UAPI comment.
Libdrm MR: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/408
Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34493
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
So that we initialize the MQD as a secure queue.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If peer memory is accessible through XGMI, allow leaving it in VRAM
rather than forcing its migration to GTT on DMABuf attachment.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pinning of VRAM is for peer devices that don't support dynamic attachment
and move notifiers. But it requires that all such peer devices are able to
access VRAM via PCIe P2P. Any device without P2P access requires migration
to GTT, which fails if the memory is already pinned for another peer
device.
Sharing between GPUs should not require pinning in VRAM. However, if
DMABUF_MOVE_NOTIFY is disabled in the kernel build, even DMABufs shared
between GPUs must be pinned, which can lead to failures and functional
regressions on systems where some peer GPUs are not P2P accessible.
Disable VRAM pinning if move notifiers are disabled in the kernel build
to fix regressions when sharing BOs between GPUs.
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Tested-by: Hao (Claire) Zhou <hao.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set up TMZ for queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set up TMZ for queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When determining the domains for pinning DMABufs, filter allowed_domains
and fail with a warning if VRAM is forbidden and GTT is not an allowed
domain.
Fixes: f5e7fabd1f ("drm/amdgpu: allow pinning DMA-bufs into VRAM if all importers can do P2P")
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use this to track the whether we want TMZ for queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Update the minimum firmware version for the Cleaner Shader in the
gfx_v10_0_sw_init function.
This change adjusts the minimum required firmware version for the MEC
firmware from 152 to 151, allowing for broader compatibility with
GFX10.1 GPUs.
Fixes: 25961bad92 ("drm/amdgpu/gfx10: Add cleaner shader for GFX10.1.10")
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch removes the deprecated SDMA reset callback mechanism, which was previously used to register pre-reset and post-reset callbacks for SDMA engine resets.
The callback mechanism has been replaced with a more direct and efficient approach using `stop_queue` and `start_queue` functions in the ring's function table.
The SDMA reset callback mechanism allowed KFD and AMDGPU to register pre-reset and post-reset functions for handling SDMA engine resets.
However, this approach added unnecessary complexity and was no longer needed after the introduction of the `stop_queue` and `start_queue` functions in the ring's function table.
1. **Remove Callback Mechanism**:
- Removed the `amdgpu_sdma_register_on_reset_callbacks` function and its associated data structures (`sdma_on_reset_funcs`).
- Removed the callback registration logic from the SDMA v4.4.2 initialization code.
2. **Clean Up Related Code**:
- Removed the `sdma_v4_4_2_set_engine_reset_funcs` function, which was used to register the callbacks.
- Removed the `sdma_v4_4_2_engine_reset_funcs` structure, which contained the pre-reset and post-reset callback functions.
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add context and seqno of the fence in error logging
rather than printing fence ptr.
Reviewed-by: Christian König <christian.koenig@amd.com>
Suggested-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Suggested-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch refactors the SDMA v5.2 queue reset and stop logic to improve
code readability, maintainability, and performance. The key changes include:
1. **Generalized `sdma_v5_2_gfx_stop` Function**:
- Added an `inst_mask` parameter to allow stopping specific SDMA instances
instead of all instances. This is useful for resetting individual queues.
2. **Simplified `sdma_v5_2_reset_queue` Function**:
- Removed redundant loops and checks by directly using the `ring->me` field
to identify the SDMA instance.
- Reused the `sdma_v5_2_gfx_stop` function to stop the queue, reducing code
duplication.
v1: The general coding style is to declare variables like "i" or "r" last. E.g. longest lines first and short lasts. (Chritian)
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch refactors the SDMA v5.0 queue reset and stop logic to improve
code readability, maintainability, and performance. The key changes include:
1. **Generalized `sdma_v5_0_gfx_stop` Function**:
- Added an `inst_mask` parameter to allow stopping specific SDMA instances
instead of all instances. This is useful for resetting individual queues.
2. **Simplified `sdma_v5_0_reset_queue` Function**:
- Removed redundant loops and checks by directly using the `ring->me` field
to identify the SDMA instance.
- Reused the `sdma_v5_0_gfx_stop` function to stop the queue, reducing code
duplication.
v1: The general coding style is to declare variables like "i" or "r" last. E.g. longest lines first and short lasts. (Chritian)
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Register stop/start/soft_reset queue functions for SDMA IP versions v5.2.
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Register stop/start/soft_reset queue functions for SDMA IP versions v5.0.
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch introduces a new function `amdgpu_sdma_soft_reset` to handle SDMA soft resets directly,
rather than relying on the DPM interface.
1. **New `amdgpu_sdma_soft_reset` Function**:
- Implements a soft reset for SDMA engines by directly writing to the hardware registers.
- Handles SDMA versions 4.x and 5.x separately:
- For SDMA 4.x, the existing `amdgpu_dpm_reset_sdma` function is used for backward compatibility.
- For SDMA 5.x, the driver directly manipulates the `GRBM_SOFT_RESET` register to reset the specified SDMA instance.
2. **Integration into `amdgpu_sdma_reset_engine`**:
- The `amdgpu_sdma_soft_reset` function is called during the SDMA reset process, replacing the previous call to `amdgpu_dpm_reset_sdma`.
v2: r should default to an error (Alex)
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Replace old callback mechanism with direct calls to stop/start functions.
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable users to create queues at different priority levels.
The highest level is restricted to drm master.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Handle the queue priority set by the user.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
So we can track this when we create user queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Convert driver priority levels to MES11 priority levels.
At the moment they are the same, but they may not always
be.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Convert driver priority levels to MES11 priority levels.
At the moment they are the same, but they may not always
be.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reuse the _pad field for flags.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Jesse.Zhang <Jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enforce isolation serializes access to the GFX IP. User
queues are isolated in the MES scheduler, but we still
need to serialize between kernel queues and user queues.
For enforce isolation, group KGD user queues with KFD user
queues.
v2: split out variable renaming, add config guards
v3: use new function names
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since they will be used for both KFD and KGD user queues,
rename them from kfd to userq. No intended functional
change.
Acked-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This will be used to stop/start user queue scheduling for
example when switching between kernel and user queues when
enforce isolation is enabled.
v2: use idx
v3: only stop compute/gfx queues
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Track this to align with KFD for enforce isolation
handling.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For VF, it doesn't have the permission to clear overflow, clear the bit
by reset.
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Replace disable_kq parameter with user_queue parameter.
The parameter has the following logic:
-1 = auto (ASIC specific default)
0 = user queues disabled
1 = user queues enabled and kernel queues enabled (if supported)
2 = user queues enabled and kernel queues disabled
The default behavior (-1) is currently the same as 0 for current
ASICs. To enable user queues (in addition to kernel queues) set
user_queue=1. To enable user queues and disable kernel queues
(to make all resources available to user queues), set user_queue=2.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to take a reference to the interrupts to make
sure they stay enabled even if the kernel queues have
disabled them.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We need to take a reference to the interrupts to make
sure they stay enabled even if the kernel queues have
disabled them.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable doorbell for JPEG5_0_1 and adjust index for VCN5_0_1.
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Increase vcn doorbell range for gfx950 to 11.
Signed-off-by: Shiwu Zhang <shiwu.zhang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Regardless of whether we disable kernel queues, we need
to take an extra reference to the pipe interrupts for
user queues to make sure they stay enabled in case we
disable them for kernel queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Regardless of whether we disable kernel queues, we need
to take an extra reference to the pipe interrupts for
user queues to make sure they stay enabled in case we
disable them for kernel queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
1) Checkpatch complains if we print an error message for kzalloc()
failure. The kzalloc() failure already has it's own error messages
built in. Also this allocation is small enough that it is guaranteed
to succeed.
2) Return directly instead of doing a goto free_fence_drv. The
"fence_drv" is already NULL so no cleanup is necessary.
Reviewed-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The goto frees "fence_drv" so this is a double free bug. There is no
need to call amdgpu_seq64_free(adev, fence_drv->va) since the seq64
allocation failed so change the goto to goto free_fence_drv. Also
propagate the error code from amdgpu_seq64_alloc() instead of hard coding
it to -ENOMEM.
Fixes: e7cf21fbb2 ("drm/amdgpu: Few optimization and fixes for userq fence driver")
Reviewed-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Pull it out of the MES code and into the generic code.
It's not MES specific and needs to be applied to all user
queues regardless of the backend.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Shaoyun.liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch introduces new function pointers in the amdgpu_sdma structure
to handle queue stop, start and soft reset operations. These will replace
the older callback mechanism.
The new functions are:
- stop_kernel_queue: Stops a specific SDMA queue
- start_kernel_queue: Starts/Restores a specific SDMA queue
- soft_reset_kernel_queue: Performs soft reset on a specific SDMA queue
v2: Update stop_queue/start_queue function paramters to use ring pointer instead of device/instance(Chritian)
v3: move stop_queue/start_queue to struct amdgpu_sdma_instance and rename them. (Alex)
v4: rework the ordering a bit (Alex)
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
since we loop through the queues |= the errors.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Unmap user queues on suspend and map them on resume.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add helpers to unmap and map user queues on suspend and
resume.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If userq creation fails, we need to properly unwind and free the
user queue fence driver.
v2: free idr as well (Sunil)
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move some userq fence handling code into amdgpu_userq_fence.c.
This matches the other code in that file.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Split out the queue map from the mqd create call and split
out the queue unmap from the mqd destroy call. This splits
the queue setup and teardown with the actual enablement
in the firmware.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Rename to map and umap to better align with what is happening
at the firmware level and remove the extra level of indirection
in the MES userq code.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There are a few prechecks made before HDP flush like a flush is not
required on APU bare metal. Using hdp callback directly bypasses those
checks. Use amdgpu_device_flush_hdp which takes care of prechecks.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 1d9bff4cf8)
JPEG is not supported on Vega only.
Fixes: 0a6e7b06bd ("drm/amdgpu: Remove JPEG from vega and carrizo video caps")
Signed-off-by: David Rosca <david.rosca@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 0f4dfe86fe)
Cc: stable@vger.kernel.org
On systems that default to 'deep' some userspace software likes
to try to suspend in 'deep' first. If there is a failure for any
reason (such as -ENOMEM) the failure is ignored and then it will
try to use 's2idle' as a fallback. This fails, but more importantly
it leads to graphical problems.
Forbid this behavior and only allow suspending in the last state
supported by the system.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4093
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20250408180957.4027643-1-superm1@kernel.org
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 2aabd44aa8)
Otherwise triggering sysfs multiple times without other submissions in
between only runs the shader once.
v2: add some comment
v3: re-add missing cast
v4: squash in semicolon fix
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 8b2ae7d492)
drm-misc-next for v6.16-rc1:
UAPI Changes:
- Add ASAHI uapi header!
- Add apple fourcc modifiers.
- Add capset virtio definitions to UAPI.
- Extend EXPORT_SYNC_FILE for timeline syncobjs.
Cross-subsystem Changes:
- Adjust DMA-BUF sg handling to not cache map on attach.
- Update drm/ci, hlcdc, virtio, maintainers.
- Update fbdev todo.
- Allow setting dma-device for dma-buf import.
- Export efi_mem_desc_lookup to make efidrm build as a module.
Core Changes:
- Update drm scheduler docs.
- Use the correct resv object in TTM delayed destroy.
- Fix compiler warning with panic qr code, and other small fixes.
- drm/ci updates.
- Add debugfs file for listing all bridges.
- Small fixes to drm/client, ttm tests.
- Add documentation to display/hdmi.
- Add kunit tests for bridges.
- Dont fail managed device probing if connector polling fails.
- Create Kconfig.debug for drm core.
- Add tests for the drm scheduler.
- Add and use new access helpers for DPCPD.
- Add generic and optimized conversions for format-helper.
- Begin refcounting panel for improving lifetime handling.
- Unify simpledrm and ofdrm sysfb, and add extra features.
- Split hdmi audio in bridge to make DP audio work.
Driver Changes:
- Convert drivers to use devm_platform_ioremap_resource().
- Assorted small fixes to imx/legacy-bridg, gma500, pl111, nouveau, vc4,
vmwgfx, ast, mxsfb, xlnx, accel/qaic, v3d, bridge/imx8qxp-ldb, ofdrm,
bridge/fsl-ldb, udl, bridge/ti-sn65dsi86, bridge/anx7625, cirrus-qemu,
bridge/cdns-dsi, panel/sharp, panel/himax, bridge/sil902x, renesas,
imagination, various panels.
- Allow attaching more display to vkms.
- Add Powertip PH128800T004-ZZA01 panel.
- Add rotation quirk for ZOTAC panel.
- Convert bridge/tc358775 to atomic.
- Remove deprecated panel calls from synaptics, novatek, samsung panels.
- Refactor shmem helper page pinning and accel drivers using it.
- Add dmabuf support to accel/amdxdna.
- Use 4k page table format for panfrost/mediatek.
- Add common powerup/down dp link helper and use it.
- Assorted compiler warning fixes.
- Support dma-buf import for renesas
Signed-off-by: Dave Airlie <airlied@redhat.com>
# Conflicts:
# include/drm/drm_kunit_helpers.h
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://lore.kernel.org/r/e147ff95-697b-4067-9e2e-7cbd424e162a@linux.intel.com
This commit modifies the gfx_v9_0_ring_emit_cleaner_shader function
to use a switch statement for cleaner shader emission based on the
specific GFX IP version.
The function now distinguishes between different IP versions, using
PACKET3_RUN_CLEANER_SHADER_9_0 for the versions 9.0.1, 9.1.0,
9.2.1, 9.2.2, 9.3.0, and 9.4.0, while retaining
PACKET3_RUN_CLEANER_SHADER for version 9.4.2.
v2: Simplify logic (Alex).
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit introduces the PACKET3_RUN_CLEANER_SHADER_9_0 definition,
which is a command packet utilized to instruct the GPU to execute the
cleaner shader for the GFX9.0 graphics architecture.
The cleaner shader is a piece of GPU code that is responsible for
clearing or initializing essential GPU resources, such as Local Data
Share (LDS), Vector General Purpose Registers (VGPRs), and Scalar
General Purpose Registers (SGPRs). Properly clearing these resources is
vital for ensuring data isolation and security between different
workloads executed on the GPU.
When the GPU receives this packet, it fetches and runs the cleaner
shader instructions from the specified location in the packet. Thus by
preventing data leaks and ensuring that previous job states do not
interfere with subsequent workloads.
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Fix an array index out of bounds warning in the DMA IP case of
amdgpu_hw_ip_info() where it was incorrectly checking
adev->gfx.gfx_ring[i].no_user_submission instead of
adev->sdma.instance[i].ring.no_user_submission.
The mismatch caused UBSAN to report an array bounds violation since
it was accessing the GFX ring array with SDMA instance indices.
Fixes: 4310acd446 ("drm/amdgpu: add ring flag for no user submissions")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
There are a few prechecks made before HDP flush like a flush is not
required on APU bare metal. Using hdp callback directly bypasses those
checks. Use amdgpu_device_flush_hdp which takes care of prechecks.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
With adding sriov_vf check, we directly return EOPNOTSUPP in
ras_reset_error_count as we should not do anything on VF to reset RAS error
count.
This also fixes the issue that loading guest driver causes register
violations.
Reviewed-by: Ahmad Rehman <Ahmad.Rehman@amd.com>
Signed-off-by: Ellen Pan <yunru.pan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Except HDP v5.2 all use a common logic for HDP flush. Use a generic
function. HDP v5.2 forces NO_KIQ logic, revisit it later.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set RAS EEPROM table version to v3 for umc v12_5.
Signed-off-by: Candice Li <candice.li@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add support for per-queue reset on SDMA v4.4.2 when running with:
1. MEC firmware version 17 or later
2. DPM indicates SDMA reset is supported
v2: Fixed supported firmware versions (Lijo)
Signed-off-by: Jesse.Zhang <Jesse.zhang@amd.com>
Reviewed-by: Tim Huang <tim.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable the cleaner shader for additional GFX11.5.2/11.5.3 series GPUs to
ensure data isolation among GPU tasks. The cleaner shader is tasked with
clearing the Local Data Store (LDS), Vector General Purpose Registers
(VGPRs), and Scalar General Purpose Registers (SGPRs), which helps avoid
data leakage and guarantees the accuracy of computational results.
This update extends cleaner shader support to GFX11.5.2/11.5.3 GPUs,
previously available for GFX11.0.3. It enhances security by clearing GPU
memory between processes and maintains a consistent GPU state across KGD
and KFD workloads.
Cc: Mario Sopena-Novales <mario.novales@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Switch from a bool to an enum and allow more options
for enforce isolation. There are now 3 modes of operation:
- Disabled (0)
- Enabled (serialization and cleaner shader) (1)
- Enabled in legacy mode (no serialization or cleaner shader) (2)
This provides better flexibility for more use cases.
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the local setting rather than the global parameter.
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the local setting rather than the global parameter.
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
JPEG is not supported on Vega only.
Fixes: 0a6e7b06bd ("drm/amdgpu: Remove JPEG from vega and carrizo video caps")
Signed-off-by: David Rosca <david.rosca@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Implement the GFX12 KCQ pipe reset, and disable the GFX12
kernel compute queue until the CPFW fully supports it.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Checking hive is more readable.
The following smatch warning:
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:6820 amdgpu_pci_slot_reset()
warn: iterator used outside loop: 'tmp_adev'
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Ce Sun <cesun102@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Kernel doorbell BOs needs to be freed before ttm_fini.
Fixes: 54c30d2a8d ("drm/amdgpu: create kernel doorbell pages")
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: ZhenGuo Yin <zhenguo.yin@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
disable ASPM with some ASICs on some specific platforms.
required from PCIe controller owner.
Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The MES queue deactivation and active status are already set in
mes_userq_unmap|map(), so the caller needn't set the queue_active
bit again.
Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add function for vcn queue reset to make driver to
do fine-grained reset instead of the whole gpu reset.
Reviewed-by: Sonny Jiang <sonny.jiang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Ruili Ji <ruiliji2@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Guest is reading/writing to snoop register which is a security violation
We moved the code to the host driver
And also added a validation on the guest side to check if it's guest
Signed-off-by: Masha Grinman <Masha.Grinman@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On systems that default to 'deep' some userspace software likes
to try to suspend in 'deep' first. If there is a failure for any
reason (such as -ENOMEM) the failure is ignored and then it will
try to use 's2idle' as a fallback. This fails, but more importantly
it leads to graphical problems.
Forbid this behavior and only allow suspending in the last state
supported by the system.
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4093
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20250408180957.4027643-1-superm1@kernel.org
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Otherwise triggering sysfs multiple times without other submissions in
between only runs the shader once.
v2: add some comment
v3: re-add missing cast
v4: squash in semicolon fix
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
amdgpu:
- MES FW version caching fixes
- Only use GTT as a fallback if we already have a backing store
- dma_buf fix
- IP discovery fix
- Replay and PSR with VRR fix
- DC FP fixes
- eDP fixes
- KIQ TLB invalidate fix
- Enable dmem groups support
- Allow pinning VRAM dma bufs if imports can do P2P
- Workload profile fixes
- Prevent possible division by 0 in fan handling
amdkfd:
- Queue reset fixes
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQQgO5Idg2tXNTSZAr293/aFa7yZ2AUCZ/akUAAKCRC93/aFa7yZ
2HtpAP9ptJj5RKxOcdh8lbiOZN3RGjxsqw6wXk7LUnw3A2npVAD5AXoaTis4S84e
0NIY7bd1Q1njP2akA4O5AKykkpgVtgw=
=Ha4H
-----END PGP SIGNATURE-----
Merge tag 'amd-drm-fixes-6.15-2025-04-09' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes
amd-drm-fixes-6.15-2025-04-09:
amdgpu:
- MES FW version caching fixes
- Only use GTT as a fallback if we already have a backing store
- dma_buf fix
- IP discovery fix
- Replay and PSR with VRR fix
- DC FP fixes
- eDP fixes
- KIQ TLB invalidate fix
- Enable dmem groups support
- Allow pinning VRAM dma bufs if imports can do P2P
- Workload profile fixes
- Prevent possible division by 0 in fan handling
amdkfd:
- Queue reset fixes
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20250409165238.1180153-1-alexander.deucher@amd.com
Don't fetch it again if we already have it. It seems the
registers don't reliably have the value at resume in some
cases.
Fixes: 785f0f9fe7 ("drm/amdgpu: Add mes v12_0 ip block support (v4)")
Reviewed-by: Shaoyun.liu <Shaoyun.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 9e7b08d239)
Cc: stable@vger.kernel.org # 6.12.x
This is normally handled in the gfx IP suspend callbacks, but
for S0ix, those are skipped because we don't want to touch
gfx. So handle it in device suspend.
Fixes: b9467983b7 ("drm/amdgpu: add dynamic workload profile switching for gfx10")
Fixes: 963537ca23 ("drm/amdgpu: add dynamic workload profile switching for gfx11")
Fixes: 5f95a15495 ("drm/amdgpu: add dynamic workload profile switching for gfx12")
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 906ad45167)
Cc: stable@vger.kernel.org
When the parameter is set, disable user submissions
to kernel queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When the parameter is set, disable user submissions
to kernel queues.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For SDMA, we still need kernel queues for paging so
they need to be initialized, but we no not want to
accept submissions from userspace when disable_kq
is set.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Plumb in support for disabling kernel queues.
v2: use ring counts per Felix' suggestion
v3: fix stream fault handler, enable EOP interrupts
v4: fix MEC interrupt offset (Sunil)
v5: clean up after removing extra sched.ready settings
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Plumb in support for disabling kernel queues in
GFX11. We have to bring up a GFX queue briefly in
order to initialize the clear state. After that
we can disable it.
v2: use ring counts per Felix' suggestion
v3: fix stream fault handler, enable EOP interrupts
v4: fix MEC interrupt offset (Sunil)
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
If we don't have kernel queues, the vmids can be used by
the MES for user queues.
Acked-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make all resources available to user queues.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Suggested-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add proper checks for disable_kq functionality in
gfx helper functions. Add special logic for families
that require the clear state setup.
v2: use ring count as per Felix suggestion
v3: fix num_gfx_rings handling in amdgpu_gfx_graphics_queue_acquire()
v4: fix error code (Alex)
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This would be set by IPs which only accept submissions
from the kernel, not userspace, such as when kernel
queues are disabled. Don't expose the rings to userspace
and reject any submissions in the CS IOCTL.
v2: fix error code (Alex)
Reviewed-by: Sunil Khatri<sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
On chips that support user queues, setting this option
will disable kernel queues to be used to validate
user queues without kernel queues.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Similar to KFD, prevent runtime pm while user queues are active.
Reviewed-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
So we can iterate across them when we need to manage
all user queues.
v2: add uq_mgr to adev list in amdgpu_userq_mgr_init
Reviewed-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add an INFO query to check if user queues are supported.
v2: switch to a mask of IPs (Marek)
v3: move to drm_amdgpu_info_device (Marek)
Cc: marek.olsak@amd.com
Cc: prike.liang@amd.com
Cc: sunil.khatri@amd.com
Cc: yogesh.mohanmarimuthu@amd.com
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a separate switch statement for the userq callback
assignment so that we can assign the callbacks for each
asic as the firmware becomes available.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
With the ME details fixed, we can now consolidate
this state. Also split out the userq setup into a separate
switch statement so that we can set them per IP version
when the firmwares are ready.
Reviewed-by: Sunil Khatri <sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The display is freezing because the amdgpu_userq_wait_ioctl()
is waiting for a non-user queue fence(specifically, the PT update fence).
RootCause:
The resume_work is initiated by both amdgpu_userq_suspend and
amdgpu_userqueue_ensure_ev_fence at same time. The amdgpu_userq_suspend
signals a dma-fence and subsequently triggers the resume_work, which is
intended to replace the existing fence by creating new dma-fence. However,
following this, the amdgpu_userqueue_ensure_ev_fence schedules another
resume_work that generates a new dma-fence, thereby replacing the one
created by amdgpu_userq_suspend. Consequently, the original fence will
never be signaled.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Shashank Sharma <shashank.sharma@amd.com>
Cc: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Warn if the number of pipes exceeds what the MES supports.
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Move it to amdgpu_mes to align with the compute and
sdma hqd masks. No functional change.
v2: rebase on new changes
v3: misc optimizations
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Sunil Khatri<sunil.khatri@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This was leftover from MES bring up when we had MES
user queues in the kernel. It's no longer used so
remove it.
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Leftover from the MES self tests that were removed previously.
Reviewed-by: Mukul Joshi <mukul.joshi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Make sure these are set properly to ensure compatibility if
we ever update the IOCTL interface.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Encountering a taint issue during the unloading of gpu_sched
due to the fence not being released/put. In this context,
amdgpu_vm_clear_freed is responsible for creating a job to
update the page table (PT). It allocates kmem_cache for
drm_sched_fence and returns the finished fence associated
with job->base.s_fence. In case of Usermode queue this finished
fence is added to the timeline sync object through
amdgpu_gem_update_bo_mapping, which is utilized by user
space to ensure the completion of the PT update.
[ 508.900587] =============================================================================
[ 508.900605] BUG drm_sched_fence (Tainted: G N): Objects remaining in drm_sched_fence on __kmem_cache_shutdown()
[ 508.900617] -----------------------------------------------------------------------------
[ 508.900627] Slab 0xffffe0cc04548780 objects=32 used=2 fp=0xffff8ea81521f000 flags=0x17ffffc0000240(workingset|head|node=0|zone=2|lastcpupid=0x1fffff)
[ 508.900645] CPU: 3 UID: 0 PID: 2337 Comm: rmmod Tainted: G N 6.12.0+ #1
[ 508.900651] Tainted: [N]=TEST
[ 508.900653] Hardware name: Gigabyte Technology Co., Ltd. X570 AORUS ELITE/X570 AORUS ELITE, BIOS F34 06/10/2021
[ 508.900656] Call Trace:
[ 508.900659] <TASK>
[ 508.900665] dump_stack_lvl+0x70/0x90
[ 508.900674] dump_stack+0x14/0x20
[ 508.900678] slab_err+0xcb/0x110
[ 508.900687] ? srso_return_thunk+0x5/0x5f
[ 508.900692] ? try_to_grab_pending+0xd3/0x1d0
[ 508.900697] ? srso_return_thunk+0x5/0x5f
[ 508.900701] ? mutex_lock+0x17/0x50
[ 508.900708] __kmem_cache_shutdown+0x144/0x2d0
[ 508.900713] ? flush_rcu_work+0x50/0x60
[ 508.900719] kmem_cache_destroy+0x46/0x1f0
[ 508.900728] drm_sched_fence_slab_fini+0x19/0x970 [gpu_sched]
[ 508.900736] __do_sys_delete_module.constprop.0+0x184/0x320
[ 508.900744] ? srso_return_thunk+0x5/0x5f
[ 508.900747] ? debug_smp_processor_id+0x1b/0x30
[ 508.900754] __x64_sys_delete_module+0x16/0x20
[ 508.900758] x64_sys_call+0xdf/0x20d0
[ 508.900763] do_syscall_64+0x51/0x120
[ 508.900769] entry_SYSCALL_64_after_hwframe+0x76/0x7e
v2: call dma_fence_put in amdgpu_gem_va_update_vm
v3: Addressed review comments from Christian.
- calling amdgpu_gem_update_timeline_node before switch.
- puting a dma_fence in case of error or !timeline_syncobj.
v4: Addressed review comments from Christian.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Shashank Sharma <shashank.sharma@amd.com>
Cc: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This can be enabled now. We have the firmware checks
in place.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently disabled until the firmwares are officially
released.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The feature is not navi3x specific at this point.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
I'd swear this was already fixed, but I guess the patch never
landed. Add it now.
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Take a reference when we create a queue and drop it
when we destroy the queue. We need to keep the device
active while user queues are active.
v2: squash in fix from Sunil
v3: squash in fix from Prike
Reviewed-by: Prike Liang <Prike.Liang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Use the IP type to look up the userq functions rather
than hardcoding it.
Reviewed-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The seq64 VM cache policy should be set to UC (Uncached) to
match with userqueue fence address kernel mapped memory's
cache settings.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
VCN and VPE have different offset range, update the doorbell
offset range repsectively.
Doorbell size for VCN and VPE is 32bit.
v1 : add gfx switch case and fix checkpatch warnings (Shashank)
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Introduce db_info structure to the populate the doorbell
information that is required to be mapped.
Made changes to the doorbell mapping func more generic,
by taking parameters that vary based on IPs and/or usecase
into db_info structure.
v2 - Fix space alignment and checkpatch warnings(Shashank)
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
That needs to be done after grabbing the lock, not before.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
When applications closes, it triggers the drm_file_free
function which subsequently releases all allocated buffer
objects. Concurrently, the resume_worker thread will attempt
to map the usermode queue. However, since the wptr buffer
object has already been deallocated, this will result in
an Illegal opcode error being raised in the command stream.
Now replacing drm_release() with a new function
amdgpu_drm_release(). This function will set the flag to
prevent the scheduling of any new queue resume/map, stop
all queues and then call drm_release().
V2:
- Replace drm_release with amdgpu_drm_release(Christian).
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Apply sign extension to seq64 va address.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Modify the MES process va end limit to max pfn.
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add the correct fences count variable [num_fences] in the fences
array iteration to handle the userq / non-userq fences.
v2:(Christian)
- All fences in the array either come from some reservation object
or drm_syncobj. If any of those are NULL then there is a bug
somewhere else.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add mqd for userq compute queue for gfx11/gfx12
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch enables attachment and detachment of eviction fences.
This is just a fork of eviction fence enabling code from the first
patch of the series so that the CI testing can happen on fully
fledged code.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The basic idea in this redesign is to add an eviction fence only in UQ
resume path. When userqueue is not present, keep ev_fence as NULL
Main changes are:
- do not create the eviction fence during evf_mgr_init, keeping
evf_mgr->ev_fence=NULL until UQ get active.
- do not replace the ev_fence in evf_resume path, but replace it only in
uq_resume path, so remove all the unnecessary code from ev_fence_resume.
- add a new helper function (amdgpu_userqueue_ensure_ev_fence) which
will do the following:
- flush any pending uq_resume work, so that it could create an
eviction_fence
- if there is no pending uq_resume_work, add a uq_resume work and
wait for it to execute so that we always have a valid ev_fence
- call this helper function from two places, to ensure we have a valid
ev_fence:
- when a new uq is created
- when a new uq completion fence is created
v2: Worked on review comments by Christian.
v3: Addressed few more review comments by Christian.
v4: Move mutex lock outside of the amdgpu_userqueue_suspend()
function (Christian).
v5: squash in build fix (Alex)
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
- Add a field in struct amdgpu_mqd_prop for userqueue
secure sem fence address since now we have a generic
file for mes_userqueue.c
- Add secure sem fence address mqd support to gfx12 into
their corresponding init functions.
- Enable secure semaphore IRQ handling
V2: Address review comment from Alex:
Use fence_address instead of fenceaddress (Shashank)
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch enables Usermode queue support across GFX, Compute
and SDMA IPs on GFX12/SDMA7. It typically reuses Navi3X userqueue
IP functions to create and destroy MQDs.
v2: rebase on proposed changes (Alex)
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Arvind Yadav <arvind.yadav@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Now that all of the IP specific code has been moved into
the IP specific functions, we can make this code generic.
V2: Fixed build errors and porting logics (Shashank)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This can all be handled by in the IP specific mpd init
code.
V2: Removed setting of gds_va, which was removed during UAPI
review (Shashank)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the addresses for the UQ metadata.
V2: Fix lower offset mask (Shashank)
V2: Use lower_32_bits for mqd objects(Alex)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the addresses for the UQ metadata.
V2: Fix lower address mask (Shashank)
V3: Use lower_32_bits for MQD objects (Alex)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the addresses for the UQ metadata.
V2: Fix lower address mask (Shashank)
V3: Use lower_32_bits() for MQD objects (Alex)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch fixes one of the regressions in eviction fence code with
IGT tests.
Reviewed-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Amaranath Somalapuram <amaranath.somalapuram@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Set the addresses for the UQ metadata.
V2: Fix lower address (Shashank)
V3: Restore lower_32_bits() for MQD addresses (Alex)
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
These are needed to make userqueue infrastructure generic.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The eviction process can get into a race condition between the eviction
fence suspend work (which replaces the old fence with new) and kms_close
(which destroys the fence and doesn't expect a new one).
This patch:
- adds a flag to indicate that fd is closing, so fence replacement is
not required (evf_mgr->fd_closing)
- adds a flush_work() during the ev_fence_destroy routine
V2: Addressed review comments from Christian:
- Do not use mutex to sync
- Use flush_work and wait for suspend_work to be done
V3: Fixed state machine for queue->active, which adds into race between
suspend/resume and queue ops
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds support for userqueue resume. What it typically does is
this:
- adds a new delayed work for resuming all the queues.
- schedules this delayed work from the suspend work.
- validates the BOs and replaces the eviction fence before resuming all
the queues running under this instance of userq manager.
V2: Addressed Christian's review comments:
- declare local variables like ret at the bottom.
- lock all the object first, then start attaching the new fence.
- dont replace old eviction fence, just attach new eviction fence.
- no error logs for drm_exec_lock failures
- no need to reserve bos after drm_exec_locked
- schedule the resume worker immediately (not after 100 ms)
- check for NULL BO (Arvind)
V5: Rebased wrt changes in suspend patch
- moved amdgpu_userqueue_validate_vm_bo in this patch
- initialized ret in resume_all
V6: Rebase
V7: Addressed review comments from Christian
- Do not use list_for_each_safe() with vm->invalidated, its not
correct way
V8: Fixed the race condition between suspend/close/fence
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Acked-by: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds suspend support for gfx userqueues. It typically does
the following:
- adds an enable_signaling function for the eviction fence, so that it
can trigger the userqueue suspend,
- adds a delayed work to handle suspending of the eviction_fence
- adds a suspend function to handle suspending of userqueues which
suspends all the queues under this userq manager and signals the
eviction fence,
- adds a function to replace the old eviction fence with a new one and
attach it to each of the objects,
- adds reference of userq manager in the eviction fence container so
that it can be used in the suspend function.
V2: Addressed Christian's review comments:
- schedule suspend work immediately
V4: Addressed Christian's review comments:
- wait for pending uq fences before starting suspend, added
queue->last_fence for the same
- accommodate ev_fence_mgr into existing code
- some bug fixes and NULL checks
V5: Addressed Christian's review comments (gitlab)
- Wait for eviction fence to get signaled in destroy,
don't signal it
- Wait for eviction fence to get signaled in replace fence,
don't signal it
V6: Addressed Christian's review comments
- Do not destroy the old eviction fence until we have it replaced
- Change the sequence of fence replacement sub-tasks
- reusing the ev_fence delayed work for userqueue suspend as well
(Shashank).
V7: Addressed Christian's review comments
- give evf_mgr as argument (instead of fpriv) to replace_fence()
- save ptr to evf_mgr in ev_fence (instead of uq_mgr)
- modify suspend_all_queues logic to reflect error properly
- remove the garbage drm_exec_lock section in wait_for_signal
- grab the userqueue mutex before starting the wait for fence
- remove the unrelated gobj check from signal_ioctl
V8: Added race condition fixes
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Acked-by: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds userqueue suspend/resume functions at
core MES V11 IP level.
V2: use true/false for queue_active status (Christian)
added Christian's R-B
V3: reset/set queue status in mqd.create and mqd.destroy
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds basic eviction fence framework for the gfx buffers.
The idea is to:
- One eviction fence is created per gfx process, at kms_open.
- This fence is attached to all the gem buffers created
by this process.
- This fence is detached to all the gem buffers at postclose_kms.
This framework will be further used for usermode queues.
V2: Addressed review comments from Christian
- keep fence_ctx and fence_seq directly in fpriv
- evcition_fence should be dynamically allocated
- do not save eviction fence instance in BO, there could be many
such fences attached to one BO
- use dma_resv_replace_fence() in detach
V3: Addressed review comments from Christian
- eviction fence create and destroy functions should be called
only once from fpriv create/destroy
- use dma_fence_put() in eviction_fence_destroy
V4: Addressed review comments from Christian:
- create a separate ev_fence_mgr structure
- cleanup fence init part
- do not add a domain for fence owner KGD
V5: Addressed review comments from Christian:
- drop the dma_fence_is_signaled check
- use a local variable to access evf_mgr->ev_fence under the
spin_lock() multiple places
- remove the vm->is_compute_ctx check to attach gfx eviction fence,
in gem_object_open
V6: Addressed review comments from Christian:
- drop the return value from eviction_fence_signal
- reserve_fence should be the first thing inside the
attach_eviction_fence function, also keep the resv_add_fence inside
the lock
- remove the unwanted ev_fence check inside detach function
- fix wrong variable check in eviction_fence_init function
- return the error value of eviction_fence_init to the caller, dont
keep it void.
- fail gem_object_open if attaching of eviction_fence fails
- detach the eviction fence only when amdgpu_vm_is_bo_always_valid
is not true.
V7: Addressed review comments from Christian:
- Do not add a uq_mgr ptr in ev_fence, rather add evf_mgr
V8: Move eviction fence enabling into separate patch for CI
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian Koenig <christian.koenig@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add argument description for the input argument
gpu_addr for amdgpu_seq64_alloc.
Fixes the warning raised by the compiler:
drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c:168:
warning: Function parameter or struct member 'gpu_addr' not described in 'amdgpu_seq64_alloc
Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Signed-off-by: Sunil Khatri <sunil.khatri@amd.com>
Reviewed-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds a new subquery (AMDGPU_INFO_UQ_FW_AREAS) in
AMDGPU_INFO_IOCTL to get the size and alignment of shadow
and csa objects from the FW setup. This information is
required for the userqueue consumers.
V2: Added Alex's suggestions and addressed review comments:
- make this query IP specific (GFX/SDMA etc)
- give a better title (AMDGPU_INFO_UQ_METADATA)
- restructured the code as per sample code shared by Alex
V3: Split the UAPI patch from shadow_size_fn modifications
V4: Addressed review comments from UAPI review (Marek/Pierre-Eric)
- Change the query name to AMDGPU_INFO_UQ_FW_AREAS
- remove unused inpur parameter for AMDGPU_HW_IP*
UAPI link: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/400/
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Arvind Yadav <arvind.yadav@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This callback gets the size and alignment requirements
for the gfx shadow buffer for preemption.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Modify kernel UAPI userq signal/wait struct field names and
description corresponding to the libdrm UAPI review comments.
libdrm MR: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/392
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Currently, the shadow FW space size and alignment information is
protected under a flag (adev->gfx.cp_gfx_shadow) which gets set
only in case of SRIOV setups.
if (amdgpu_sriov_vf(adev))
adev->gfx.cp_gfx_shadow = true;
But we need this information for GFX Userqueues, so that user can
create these objects while creating userqueue. This patch series
creates a method to get this information bypassing the dependency
on this check.
This patch:
- adds a new input parameter flag to the gfx.funcs->get_gfx_shadow_info
fptr definition, so that it can accommodate the information without the
check (adev->gfx.cp_gfx_shadow) on request.
- updates the existing definition of amdgpu_gfx_get_gfx_shadow_info to
adjust with this new flag.
Next patch in the series is adding a UAPI which will consume this info.
V2: split this patch from the new UAPI patch
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Arvind Yadav <arvind.yadav@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch fixes some of the pending UAPI review comments
from the libDRM/UAPI review process.
- It updates some outdated comments in the userqueue UAPI header
highlighted during the libdrm UAPI review.
- It removes the GDS BO support which was found unused.
- It also removes the unused flags parameter from the UAPI.
- It also adds a padding variables in userqueue in/out structures.
(Pierre-Eric and Marek)
- clarify comments on top of drm_amdgpu_userq_in
- clarify comment for queue_id (in)
- clarify comment for mqd
- clarify comment for compute MQD size
- clarify comment for queue_id (out)
- remove GDB object from BO object list
- remove the unused flags parameter
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 6be2ad4f00.
This patch was to block userspace to use doorbell manager UAPI
until usermode queue UAPI gets approved. UQ UAPI got approved in the
following MR:
https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/392
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch adds input fences to VM_IOCTL for buffer object.
The kernel will map/unmap the BO only when the fence is signaled.
The UAPI for the same has been approved here:
https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/392
V2: Bug fix (Arvind)
V3: Bug fix (Arvind)
V4: Rename UAPI objects as per UAPI review (Marek)
V5: Addressed review comemnts from Christian
- function should return error.
- Add 'TODO' comment
- The input fence should be independent of the operation.
V6: Addressed review comemnts from Christian
- Release the memory allocated by memdup_user().
V7: Addressed review comemnts from Christian
- Drop the debug print and add "return r;" for the error handling.
V11: Rebase
v12: Fix 32-bit holes issue in sturct drm_amdgpu_gem_va.
v13: Fix deadlock issue.
v14: Fix merge conflict.
v15: Fix review comment by renaming syncobj handles.
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Keep the user queue fence signal and wait IOCTLs in the
kernel config CONFIG_DRM_AMDGPU_NAVI3X_USERQ.
v2(Christian):
- Remove the userq specific config added for kernel queues fence init
function.
v3(Alex):
- It will be better to return an error(-ENOTSUPP) in these cases.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add gpu address support to seq64 alloc function.
v1:(Christian)
- Add the user of this new interface change to the same
patch.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Drop AMDGPU_USERQ_BO_WRITE as this should not be a global option
of the IOCTL, It should be option per buffer. Hence adding separate
array for read and write BO handles.
v2(Marek):
- Internal kernel details shouldn't be here. This file should only
document the observed behavior, not the implementation .
v3:
- Fix DAL CI clang issue.
v4:
- Added Alex RB to merge the kernel UAPI changes since he has
already approved the amdgpu_drm.h changes.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Suggested-by: Marek Olšák <marek.olsak@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add a vm root BO lock before accessing the userqueue VM.
v1:(Christian)
- Keep the VM locked until you are done with the mapping.
- Grab a temporary BO reference, drop the VM lock and acquire the BO.
When you are done with everything just drop the BO lock and
then the temporary BO reference.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add the missing error handling for xa_store() call in the function
amdgpu_userq_fence_driver_alloc().
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Few optimization and fixes for userq fence driver.
v1:(Christian):
- Remove unnecessary comments.
- In drm_exec_init call give num_bo_handles as last parameter it would
making allocation of the array more efficient
- Handle return value of __xa_store() and improve the error handling of
amdgpu_userq_fence_driver_alloc().
v2:(Christian):
- Revert userq_xa xarray init to XA_FLAGS_LOCK_IRQ.
- move the xa_unlock before the error check of the call xa_err(__xa_store())
and moved this change to a separate patch as this is adding a missing error
handling.
- Removed the unnecessary comments.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Remove MES self test as this conflicts the userqueue fence
interrupts.
v2:(Christian)
- remove the amdgpu_mes_self_test() function and any now unused code.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch updates the VM_IOCTL to allow userspace to synchronize
the mapping/unmapping of a BO in the page table.
The major changes are:
- it adds a drm_timeline object as an input parameter to the VM IOCTL.
- this object is used by the kernel to sync the update of the BO in
the page table during the mapping of the object.
- the kernel also synchronizes the tlb flush of the page table entry of
this object during the unmapping (Added in this series:
https://patchwork.freedesktop.org/series/131276/ and
https://patchwork.freedesktop.org/patch/584182/)
- the userspace can wait on this timeline, and then the BO is ready to
be consumed by the GPU.
The UAPI for the same has been approved here:
https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/392
V2:
- remove the eviction fence coupling
V3:
- added the drm timeline support instead of input/output fence
(Christian)
V4:
- made timeline 64-bit (Christian)
- bug fix (Arvind)
V5: GLCTS bug fix (Arvind)
V6: Rename syncobj_handle -> timeline_syncobj_out
Rename point -> timeline_point_in (Marek)
V7: Addressed review comments from Christian:
- do not send last_update fence in case of vm_clear_freed, instead
return the fence from gen_va_update_vm
- move the functions to update bo_mapping to amdgpu_gem.c
- do not use amdgpu_userq_update_vm anymore in userq_create()
V8: Addressed review comments from Christian:
- Split amdgpu_gem_update_bo_mapping function.
- amdgpu_gem_va_update_vm should return stub for error.
V9: Addressed review comments from Christian:
- Rename the function amdgpu_gem_update_timeline_node.
- amdgpu_gem_update_timeline_node should be void function.
- when timeline_point is zero don't allocate a chain and
call drm_syncobj_replace_fence() instead of
drm_syncobj_add_point().
V11: rebase
V12: Fix 32-bit holes issue in sturct drm_amdgpu_gem_va.
V13: Fix the review comment by renaming timeline syncobj (Marek)
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add support to handle the userqueue protected fence signal hardware
interrupt.
Create a xarray which maps the doorbell index to the fence driver address.
This would help to retrieve the fence driver information when an userq fence
interrupt is triggered. Firmware sends the doorbell offset value and
this info is compared with the queue's mqd doorbell offset value.
If they are same, we process the userq fence interrupt.
v1:(Christian):
- use xa_load to extract the fence driver.
- move the amdgpu_userq_fence_driver_process call within the xa_lock
as there is a chance that fence_drv might be freed.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Add user fence wait IOCTL timeline syncobj support.
v2:(Christian)
- handle dma_fence_wait() return value.
- shorten the variable name syncobj_timeline_points a bit.
- move num_points up to avoid padding issues.
v3:(Christian)
- Handle timeline drm_syncobj_find_fence() call error
handling
- Use dma_fence_unwrap_for_each() in timeline fence as
there could be more than one fence.
v4:(Christian)
- Drop the first num_fences since fence is always included in
the dma_fence_unwrap_for_each() iteration, when fence != f
then fence is most likely just a container.
v5: Added Alex RB to merge the kernel UAPI changes since he has
already approved the amdgpu_drm.h changes.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch introduces new IOCTL for userqueue secure semaphore.
The signal IOCTL called from userspace application creates a drm
syncobj and array of bo GEM handles and passed in as parameter to
the driver to install the fence into it.
The wait IOCTL gets an array of drm syncobjs, finds the fences
attached to the drm syncobjs and obtain the array of
memory_address/fence_value combintion which are returned to
userspace.
v2: (Christian)
- Install fence into GEM BO object.
- Lock all BO's using the dma resv subsystem
- Reorder the sequence in signal IOCTL function.
- Get write pointer from the shadow wptr
- use userq_fence to fetch the va/value in wait IOCTL.
v3: (Christian)
- Use drm_exec helper for the proper BO drm reserve and avoid BO
lock/unlock issues.
- fence/fence driver reference count logic for signal/wait IOCTLs.
v4: (Christian)
- Fixed the drm_exec calling sequence
- use dma_resv_for_each_fence_unlock if BO's are not locked
- Modified the fence_info array storing logic.
v5: (Christian)
- Keep fence_drv until wait queue execution.
- Add dma_fence_wait for other fences.
- Lock BO's using drm_exec as the number of fences in them could
change.
- Install signaled fences as well into BO/Syncobj.
- Move Syncobj fence installation code after the drm_exec_prepare_array.
- Directly add dma_resv_usage_rw(args->bo_flags....
- remove unnecessary dma_fence_put.
v6: (Christian)
- Add xarray stuff to store the fence_drv
- Implement a function to iterate over the xarray and drop
the fence_drv references.
- Add drm_exec_until_all_locked() wrapper
- Add a check that if we haven't exceeded the user allocated num_fences
before adding dma_fence to the fences array.
v7: (Christian)
- Use memdup_user() for kmalloc_array + copy_from_user
- Move the fence_drv references from the xarray into the newly created fence
and drop the fence_drv references when we signal this fence.
- Move this locking of BOs before the "if (!wait_info->num_fences)",
this way you need this code block only once.
- Merge the error handling code and the cleanup + return 0 code.
- Initializing the xa should probably be done in the userq code.
- Remove the userq back pointer stored in fence_drv.
- Pass xarray as parameter in amdgpu_userq_walk_and_drop_fence_drv()
v8: (Christian)
- Move fence_drv references must come before adding the fence to the list.
- Use xa_lock_irqsave_nested for nested spinlock operations.
- userq_mgr should be per fpriv and not one per device.
- Restructure the interrupt process code for the early exit of the loop.
- The reference acquired in the syncobj fence replace code needs to be
kept around.
- Modify the dma_fence acquire placement in wait IOCTL.
- Move USERQ_BO_WRITE flag to UAPI header file.
- drop the fence drv reference after telling the hw to stop accessing it.
- Add multi sync object support to userq signal IOCTL.
V9: (Christian)
- Store all the fence_drv ref to other drivers and not ourself.
- Remove the userq fence xa implementation and replace with
kvmalloc_array.
v10: (Christian)
- Add a comment for the userq_xa xarray
- drop the if check of userq_fence->fence_drv_array
- use the i variable to initialize userq_fence->fence_drv_array_count
- drop the fence reference before you free the array in the error handling,
otherwise it could be that some references leaked.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
