| Age | Commit message (Collapse) | Author |
|---|
|
This patch adds robust reset handling for user queues (userq) to improve
recovery from queue failures. The key components include:
1. Queue detection and reset logic:
- amdgpu_userq_detect_and_reset_queues() identifies failed queues
- Per-IP detect_and_reset callbacks for targeted recovery
- Falls back to full GPU reset when needed
2. Reset infrastructure:
- Adds userq_reset_work workqueue for async reset handling
- Implements pre/post reset handlers for queue state management
- Integrates with existing GPU reset framework
3. Error handling improvements:
- Enhanced state tracking with HUNG state
- Automatic reset triggering on critical failures
- VRAM loss handling during recovery
4. Integration points:
- Added to device init/reset paths
- Called during queue destroy, suspend, and isolation events
- Handles both individual queue and full GPU resets
The reset functionality works with both gfx/compute and sdma queues,
providing better resilience against queue failures while minimizing
disruption to unaffected queues.
v2: add detection and reset calls when preemption/unmaped fails.
add a per device userq counter for each user queue type.(Alex)
v3: make sure we hold the adev->userq_mutex when we call amdgpu_userq_detect_and_reset_queues. (Alex)
warn if the adev->userq_mutex is not held.
v4: make sure we have all of the uqm->userq_mutex held.
warn if the uqm->userq_mutex is not held.
v5: Use array for user queue type counters.(Alex)
all of the uqm->userq_mutex need to be held when calling detect and reset. (Alex)
v6: fix lock dep warning in amdgpu_userq_fence_dence_driver_process
v7: add the queue types in an array and use a loop in amdgpu_userq_detect_and_reset_queues (Lijo)
v8: remove atomic_set(&userq_mgr->userq_count[i], 0).
it should already be 0 since we kzalloc the structure (Alex)
v9: For consistency with kernel queues, We may want something like:
amdgpu_userq_is_reset_type_supported (Alex)
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>
|
|
Remove amdgpu_asic_flush_hdp & amdgpu_asic_invalidate_hdp functions and
directly use the mapped ones
Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Update asic_invalidate_hdp and asic_flush_hdp function to check if ip
function exist, if not return void
v2: Use else/if (Kevin)
Update function name (Lijo)
Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Suggested-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
This commit refactors the AMDGPU userqueue management subsystem to replace
IDR (ID Allocation) with XArray for improved performance, scalability, and
maintainability. The changes address several issues with the previous IDR
implementation and provide better locking semantics.
Key changes:
1. **Global XArray Introduction**:
- Added `userq_doorbell_xa` to `struct amdgpu_device` for global queue tracking
- Uses doorbell_index as key for efficient global lookup
- Replaces the previous `userq_mgr_list` linked list approach
2. **Per-process XArray Conversion**:
- Replaced `userq_idr` with `userq_mgr_xa` in `struct amdgpu_userq_mgr`
- Maintains per-process queue tracking with queue_id as key
- Uses XA_FLAGS_ALLOC for automatic ID allocation
3. **Locking Improvements**:
- Removed global `userq_mutex` from `struct amdgpu_device`
- Replaced with fine-grained XArray locking using XArray's internal spinlocks
4. **Runtime Idle Check Optimization**:
- Updated `amdgpu_runtime_idle_check_userq()` to use xa_empty
5. **Queue Management Functions**:
- Converted all IDR operations to equivalent XArray functions:
- `idr_alloc()` → `xa_alloc()`
- `idr_find()` → `xa_load()`
- `idr_remove()` → `xa_erase()`
- `idr_for_each()` → `xa_for_each()`
Benefits:
- **Performance**: XArray provides better scalability for large numbers of queues
- **Memory Efficiency**: Reduced memory overhead compared to IDR
- **Thread Safety**: Improved locking semantics with XArray's internal spinlocks
v2: rename userq_global_xa/userq_xa to userq_doorbell_xa/userq_mgr_xa
Remove xa_lock and use its own lock.
v3: Set queue->userq_mgr = uq_mgr in amdgpu_userq_create()
v4: use xa_store_irq (Christian)
hold the read side of the reset lock while creating/destroying queues and the manager data structure. (Chritian)
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add ras module ip block to amdgpu discovery.
Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-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 already a helper for telling if a block is valid, but if
IP handling wants to check if it's HW is enabled no such helper
exists.
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add amdgpu_discovery_info structure to keep all discovery related
information.
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>
|
|
amdgpu_device_ip_suspend() doesn't have a caller outside of
amdgpu_device.c. Make it static.
No intended functional changes.
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>
|
|
The Constant Engine found on gfx6-gfx10 HW has been a notorious source of
problems.
RADV never used it in the first place, radeonsi only used it for a few
releases around 2017 for gfx6-gfx9 before dropping support for it as
well.
While investigating another problem I just recently found that submitting
to the CE seems to be completely broken on gfx9 for quite a while.
Since nobody complained about that problem it most likely means that
nobody is using any of the affected radeonsi versions on current Linux
kernels any more.
So to potentially phase out the support for the CE and eliminate another
source of problems block submitting CE IBs unless it is enabled again
using a debug flag.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Define a utility macro for defining capabilities and their attributes.
Capability attributes are read-only, write-only, read-write.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Acked-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
This code isn't needed anymore as we collect the same information
into pm_display_cfg instead.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add mmio_remap bookkeeping to amdgpu_device and introduce
amdgpu_ttm_mmio_remap_bo_init()/fini() to manage a kernel-owned,
one-page (4K) BO in AMDGPU_GEM_DOMAIN_MMIO_REMAP.
Bookkeeping:
- adev->rmmio_remap.bo : kernel-owned singleton BO
The BO is allocated during TTM init when a remap bus address is available
(adev->rmmio_remap.bus_addr) and PAGE_SIZE <= AMDGPU_GPU_PAGE_SIZE (4K),
and freed during TTM fini.
v2:
- Check mmio_remap bus address (adev->rmmio_remap.bus_addr) instead of
rmmio_base. (Alex)
- Skip quietly if PAGE_SIZE > AMDGPU_GPU_PAGE_SIZE or no bus address
(no warn). (Alex)
- Use `amdgpu_bo_create()` (not *_kernel) - Only with this The object
is stored in adev->mmio_remap.bo and will later be exposed to
userspace via a GEM handle. (Christian)
v3:
- Remove obvious comment before amdgpu_ttm_mmio_remap_bo_fini() call.
(Alex)
v4:
- Squash bookkeeping into this patch (Christian)
Suggested-by: Christian König <christian.koenig@amd.com>
Suggested-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Remove the unnecessary use of volatile in some of the amdgpu.h and
amdgpu_ih.h headers.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
None of the pointer operations handled by the ring file requires
volatile, for this reason, this commit removes all occurrences of
volatile associated with rings.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Rodrigo Siqueira <siqueira@igalia.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
warnings
-Wflex-array-member-not-at-end was introduced in GCC-14, and we are
getting ready to enable it, globally.
Move the conflicting declarations to the end of the corresponding
structures. Notice that `struct dev_pagemap` is a flexible structure,
this is a structure that contains a flexible-array member.
struct dev_pagemap always has room for at least one range. amdgpu only
uses a single range. Therefore no change are needed to the allocation
of struct amdgpu_device.
Fix 283 of the following type of warnings:
283 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h:111:28: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Felix Kuehling <felix.kuehling@amd.com>
Reviewed-by: Felix Kuehling <felix.kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
AMDGPU_HAS_VRAM is redundant with is_app_apu, as both refer to
APUs with no carve-out. Since AMDGPU_HAS_VRAM only occurs once,
remove AMDGPU_HAS_VRAM definition. The tmr allocation can be covered
with AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_VRAM in both vram and
non vram ASICs.
Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
During a DPC error kernel waits for the link to be active before
notifying downstream devices. On certain platforms with Broadcom switch
in synthetiic mode, switch responds with values even though the link is
not fully ready. The config space restoration done by pcie port driver
for SWUS/DS of dGPU is thus not effective as the switch is still doing
internal enumeration.
As a workaround, save state of SWUS/DS device in driver. Add additional
check to see if link is active and restore the values during DPC error
callbacks.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Yang Wang <kevinyang.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add a struct to store unique id information for each type. Add helper
to fetch the unique id.
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>
|
|
Used to to set the MQD appropriately for each queue type.
Kernel queues have additional privileges.
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.16.x
|
|
Create amd_isp_capture MFD device with swnode initialized to
isp specific software_node part of fwnode graph in amd_isp4
x86/platform driver. The isp driver use this swnode handle
to retrieve the critical properties (data-lanes, mipi phyid,
link-frequencies etc.) required for camera to work on AMD ISP4
based targets.
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Pratap Nirujogi <pratap.nirujogi@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Pass amdgpu device context instead of drm device context to some
amdgpu_device_* functions. DRM device context is not required in those
functions. No functional change.
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>
|
|
Fixes error with dev_info_once usage in amdgpu_device_asic_has_dc_support.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202506281140.mXfWT3EN-lkp@intel.com/
Fixes: a3e510fd69c3 ("drm/amdgpu: Convert from DRM_* to dev_*")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
It is not necessary to be ip generation specific
v2: rename the helper to is_multi_aid (Lijo)
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Just use kmalloc for the fences in the rare case we need
an independent fence.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
In case of unexpected errors, check if device is alive on the bus.
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>
|
|
complete() callbacks are supposed to handle reversing anything
that occurred during prepare() callbacks. They'll be called on every
power state transition, and will also be called if the sequence is
failed (such as an aborted suspend).
Add support for IP blocks to support this action.
Reviewed-by: Alex Hung <alex.hung@amd.com>
Link: https://lore.kernel.org/r/20250602014432.3538345-2-superm1@kernel.org
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add debug mask to disable kernel logs of RAS correctable errors,
including both ACA and CE error counter kernel messages.
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>
|
|
ISP is a child device to GFX, and its device specific information
is not available in ACPI. Adding the 2 GPIO resources required for
ISP_v4_1_1 in amdgpu_isp driver.
- GPIO 0 to allow sensor driver to enable and disable sensor module.
- GPIO 85 to allow ISP driver to enable and disable ISP RGB streaming mode.
Signed-off-by: Pratap Nirujogi <pratap.nirujogi@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@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>
|
|
This reverts commit 3a9626c816db901def438dc2513622e281186d39.
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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>
|
|
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 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>
|
|
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>
|
|
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>
|
|
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>
|
|
This patch adds IP independent skeleton code for amdgpu
usermode queue. It contains:
- A new files with init functions of usermode queues.
- A queue context manager in driver private data.
V1: Worked on design review comments from RFC patch series:
(https://patchwork.freedesktop.org/series/112214/)
- Alex: Keep a list of queues, instead of single queue per process.
- Christian: Use the queue manager instead of global ptrs,
Don't keep the queue structure in amdgpu_ctx
V2:
- Reformatted code, split the big patch into two
V3:
- Integration with doorbell manager
V4:
- Align the structure member names to the largest member's column
(Luben)
- Added SPDX license (Luben)
V5:
- Do not add amdgpu.h in amdgpu_userqueue.h (Christian).
- Move struct amdgpu_userq_mgr into amdgpu_userqueue.h (Christian).
V6: Rebase
V9: Rebase
V10: Rebase + Alex's R-B
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Reviewed-by: Christian König <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>
|
|
Add a new parameter to disable BAR resizing. Note that this
only disables the driver from attempting to resize the BAR,
The BIOS may have resized the BAR at boot.
Some teams have found this useful in debugging P2P DMA
issues on systems where the available MMIO space did not allow
for all of the GPUs present to resize their BARs.
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add support for DPC recover based on refactored code
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>
|
|
Split amdgpu_device_gpu_recover into the following stages:
halt activities,asic reset,schedule resume and amdgpu resume.
The reason is that the subsequent addition of dpc recover
code will have a high similarity with gpu reset
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>
|
|
KIQ invalidate_tlbs request has been seen to marginally exceed the
configured 100 ms timeout on systems under load.
All other KIQ requests in the driver use a 10 second timeout. Use a
similar timeout implementation on the invalidate_tlbs path.
v2: Poll once before msleep
v3: Fix return value
Signed-off-by: Jay Cornwall <jay.cornwall@amd.com>
Cc: Kent Russell <kent.russell@amd.com>
Reviewed-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Limiting the number of available VMIDs to enforce isolation causes some
issues with gang submit and applying certain HW workarounds which
require multiple VMIDs to work correctly.
So instead start to track all submissions to the relevant engines in a
per partition data structure and use the dma_fences of the submissions
to enforce isolation similar to what a VMID limit does.
v2: use ~0l for jobs without isolation to distinct it from kernel
submissions which uses NULL for the owner. Add some warning when we
are OOM.
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
