This WA needs to be applied to graphics GT when the media version
is 2000. Currently, media version 2000 is always paired with
graphics version 2004 which will result in writing same register
with same bits twice. We can't add optional rule in rtp
framework and also writing same register with same bits gives
warning.
Currently, media version 2000 is always paired with graphics version
2004, so just checking the latter is sufficient.
V2(Lucas):
- Add more detail in commit message
- Improve code comment to follow guideline
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2016
Fixes: 131328aa56 ("drm/xe/xe2lpm: Add permanent Wa_14020756599")
Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240607122528.1048610-1-tejas.upadhyay@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
xe_force_wake_assert_held() is designed to confirm a particular
forcewake domain's wakefulness; it doesn't verify the wakefulness of
multiple domains. Make sure the caller doesn't input multiple
domains(XE_FORCEWAKE_ALL) as a parameter.
v2
- use domain != XE_FORCEWAKE_ALL (Michal)
v3
- Add kernel-doc
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Badal Nilawar <badal.nilawar@intel.com>
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240607125741.1407331-1-himal.prasad.ghimiray@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
If possible, we should prefer xe_gt_err() over drm_err().
While at it, improve and fix some of the error messages.
Also drop unnecessary "xe_gt_sysfs.h" include.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240609181931.1724-4-michal.wajdeczko@intel.com
We are converting xe_guc_pc to xe_guc_ct few times already.
Add simple helper function to avoid code duplication.
While at it, simplify other helper functions and fix order
of local variables to match the guideline.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240609181931.1724-3-michal.wajdeczko@intel.com
We already have a dedicated file for GuC SLPC ABI definitions.
Move definition of the SETUP_PC_GUCRC action and related enum
to that file, rename them to match format of other new ABI
definitions and add simple kernel-doc.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240609181931.1724-2-michal.wajdeczko@intel.com
The Local Memory (aka VRAM) is only available on DGFX platforms.
We shouldn't attempt to provision VFs with LMEM or attempt to
update the LMTT on non-DGFX platforms. Add missing asserts that
would enforce that and fix release code that could crash on iGFX
due to uninitialized LMTT.
Fixes: c063cce7df ("drm/xe/pf: Update the LMTT when freeing VF GT config")
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240607153155.1592-1-michal.wajdeczko@intel.com
Clean up laying violation of setting q->flags EXEC_QUEUE_FLAG_BANNED bit
in GuC backend. Move banned to GuC owned bit and report banned status to
upper layers via reset_status vfunc. This is a slight change in behavior
as reset_status returns true if wedged or killed bits set too, but in
all of these cases submission to queue is no longer allowed.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240604184700.1946918-1-matthew.brost@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
- Remove unused bit definitions.
- Driver uses BIT(0) for waking/sleeping the domain and since the
registers are masked respective mask bit BIT(16) needs to be set. Use
defines for these bits and use them in domain initialization.
v3
- Use defines within domain_init
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Badal Nilawar <badal.nilawar@intel.com>
Suggested-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Reviewed-by: Badal Nilawar <badal.nilawar@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240606085720.1327152-1-himal.prasad.ghimiray@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Add helper function to implement VMA (user binding) page faults, remove
unnecessary userptr.invalidate_link list del operation, retry on memory
pressure, remove unnecessary xe_vma_userptr_check_repin after rebinding,
remove unnecessary TLB invalidation, and always use vm->lock in write
mode. Changes help facilitate SVM page faults.
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Jagmeet Randhawa <jagmeet.randhawa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240607015212.2190106-1-matthew.brost@intel.com
During early initialization, in the xe_guc_min_load_for_hwconfig()
function, we are successfully enabling CTB communication, but it
will only allow us to send non-blocking H2G messages, as due to
not yet enabled IRQs, including G2H IRQs, we will not notice any
new G2H message sent by the GuC, including replies to our blocking
H2G request messages. And those successful replies are mandatory
for the VF drivers to continue normal operations.
As attempt to workaround this driver initialization ordering issue,
introduce special safe-mode CTB worker, that will periodically
trigger G2H processing, like original IRQ handler, in case no
MSI/MSIX IRQs were enabled on the driver yet. Once we detect that
IRQ were enabled, we will stop this worker.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240606130639.1504-3-michal.wajdeczko@intel.com
In the next patch we will want to perform the same steps that
g2h worker function is doing but from the different worker.
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240606130639.1504-2-michal.wajdeczko@intel.com
disable c6 called in guc_pc_fini_hw is unreachable.
GuC PC init returns earlier if skip_guc_pc is true and never
registers the finish call thus making disable_c6 unreachable.
move this call to gt idle.
v2: rebase
v3: add fixes tag (Himal)
Fixes: 975e4a3795 ("drm/xe: Manually setup C6 when skip_guc_pc is set")
Signed-off-by: Riana Tauro <riana.tauro@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240606100842.956072-3-riana.tauro@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
We currently have debugfs support that allows the userspace to initiate
an asynchronous gt reset on command. However, userspace may also wish
to wait for the completion of the gt reset before performing any
additional work. To that end, add a version of the force_reset gt
debugfs function that operates synchronously.
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/1068
Suggested-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
CC: John Harrison <john.c.harrison@intel.com>
CC: Stuart Summers <stuart.summers@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240605150828.2736396-1-jonathan.cavitt@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
job->fence is not assigned until xe_sched_job_arm(), check for
job->fence in xe_sched_job_seqno() so any usage of this function (trace
points) do not result in NULL ptr dereference. Also check job->fence
before assigning error in job trace points.
Fixes: 0ac7a2c745 ("drm/xe: Don't initialize fences at xe_sched_job_create()")
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240605055041.2082074-1-matthew.brost@intel.com
Only few steps from the GT restart phase are applicable for the
VF drivers, as initialization of PAT, WOPCM, MOCS or CCS mode can
be done only by the native or PF drivers. Use custom GT restart
function if running in VF mode.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240604212231.1416-5-michal.wajdeczko@intel.com
The VF drivers can't trigger real GuC firmware reset using GDRST
register, but for the VF drivers it is sufficient to send VF_RESET
message to reset any VF specific state maintained by the GuC.
Use our existing VF bootstrap function as VF_RESET is part of it.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240604212231.1416-4-michal.wajdeczko@intel.com
VF drivers can't modify WOPCM registers nor upload firmwares to
GuC, HuC or GSC. Modify xe_uc initialization functions to skip
those steps if running in the VF mode, or defer to a new custom
helper function that would not include those steps.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240604212231.1416-3-michal.wajdeczko@intel.com
Overmapping the identity VRAM mapping is triggering hardware bugs on
certain platforms. Use 2M pages for the last unaligned (to 1G) VRAM
chunk.
v2:
- Always use 2M pages for last chunk (Fei Yang)
- break loop when 2M pages are used
- Add assert for usable_size being 2M aligned
v3:
- Fix checkpatch
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Fei Yang <fei.yang@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Fei Yang <fei.yang@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240603181824.1927675-1-matthew.brost@intel.com
Tests show that user fence signalling requires kind of write barrier,
otherwise not all writes performed by the workload will be available
to userspace. It is already done for render and compute, we need it
also for the rest: video, gsc, copy.
Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240605-fix_user_fence_posted-v3-2-06e7932f784a@intel.com
This reverts commit 38007fa964.
Signaling user-fence after seqno write does not seem to be good solution.
Instead of changing order separate barrier should be put before user-fence,
this will be done in separate patch.
v2: added fixes tag in case reverted patch gets backported to stable
Fixes: 38007fa964 ("drm/xe: flush gtt before signalling user fence on all engines")
Signed-off-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240605-fix_user_fence_posted-v3-1-06e7932f784a@intel.com
The if condition !A || A && B can be simplified to !A || B.
Fixes the following Coccinelle/coccicheck warning reported by
excluded_middle.cocci:
WARNING !A || A && B is equivalent to !A || B
Compile-tested only.
Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240603180005.191578-1-thorsten.blum@toblux.com
For xe2_lpm Wa_14020756599 is applied to all steppings and
when RCS is present on graphics GT.
V5(MattR):
- Add more comments about new API
V4:
- Make it part of lrc wa
- Check for RCS as rtp rule
V3(MattR):
- Rename rtp api name
- Use MEDIA_VERx100
V2:
- Remove engine filter video decode
- Fix typo GRAPHICS/MEDIA/s - Himal
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240603104951.705603-2-tejas.upadhyay@intel.com
Kernel VM do not have an Xe file. Include a check for Xe file in the VM
before trying to get pid from VM's Xe file when taking a devcoredump.
Fixes: b10d0c5e9d ("drm/xe: Add process name to devcoredump")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530203341.1795181-1-matthew.brost@intel.com
The LMTT must be updated whenever we change the VF LMEM configuration.
We missed that step when freeing the whole VF GT config, which could
result in stale PTE in LMTT or LMTT PT object leaks. Fix that.
Fixes: ac6598aed1 ("drm/xe/pf: Add support to configure SR-IOV VFs")
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240527115408.1064-1-michal.wajdeczko@intel.com
The initialization order of GT topology, MCR, PAT and GuC HWconfig
as done today by native/PF driver, can't be followed as-is by the
VF driver, since fuse registers used in GT topology discovery will
be obtained by the VF driver from the GuC in HWconfig step.
While native/PF drivers need to program the HW PAT table prior to
loading the GuC, this requires only multicast writes support from
the MCR code, which could be initialized separately from the full
MCR support that requires the GT topology to setup steering data.
Split MCR initialization into two steps to avoid introducing VF
specific code paths. This also fixes duplicated spin_lock inits.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530115814.1284-1-michal.wajdeczko@intel.com
There is no point in mixing register access and VRAM code in the
same file. Move and rename the VRAM probe function to a new file
(there are no other changes other then new simple kernel-doc).
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530133527.1328-5-michal.wajdeczko@intel.com
Rename static functions to align with our typical coding style.
While at it, downgrade the existing kernel-doc for internal
function to normal comment.
Suggested-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530133527.1328-4-michal.wajdeczko@intel.com
We should keep all hardware definitions separated from the driver
code. Move LMEM_BAR definition to new regs/xe_bars.h file and also
add there GTTMMADR_BAR definition to avoid using magic 0 resource.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530133527.1328-3-michal.wajdeczko@intel.com
We should not define registers directly in the code while we have
dedicated files for all register definitions. Move XEHP_MTCFG_ADDR
to regs/xe_regs.h
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530133527.1328-2-michal.wajdeczko@intel.com
This reverts commit cd506a33b0.
The gt_remove function was explicitly added as part of the remove flow
instead of using drmm/devm automatic cleanup due to it being illegal
to remove a component after the driver has been detached from the pci
device; the GSC proxy component is removed as part of gt_remove, so we
need to do it in the pci cleanup flow. The function already has a
comment above it to explain this.
Note that the change to use the devm also caused an invalid pointer
deref in the gsc_proxy unbind function, but I didn't bother to debug
which pointer was bad since we shouldn't be calling the unbind that
late anyway and this revert fixes it.
Both issue were not seen in CI because the GSC loading is temporarily
disabled due to a critical bug, which means we're not binding the
component.
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240528182354.1200424-1-daniele.ceraolospurio@intel.com
Using snprintf() with a format string from task->comm is a bit
dangerous since the string may be controlled by unprivileged
userspace:
drivers/gpu/drm/xe/xe_devcoredump.c: In function 'devcoredump_snapshot':
drivers/gpu/drm/xe/xe_devcoredump.c:184:9: error: format not a string literal and no format arguments [-Werror=format-security]
184 | snprintf(ss->process_name, sizeof(ss->process_name), process_name);
| ^~~~~~~~
In this case there is no reason for an snprintf(), so use a simpler
string copy.
Fixes: b10d0c5e9d ("drm/xe: Add process name to devcoredump")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240528133251.2310868-1-arnd@kernel.org
Decouple xe_lrc from xe_exec_queue and reference count xe_lrc.
Removing hard coupling between xe_exec_queue and xe_lrc allows
flexible design where the user interface xe_exec_queue can be
destroyed independent of the hardware/firmware interface xe_lrc.
v2: Fix lrc indexing in wq_item_append()
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240530032211.29299-1-niranjana.vishwanathapura@intel.com
Do not hold xef->exec_queue.lock mutex while parsing the xarray
xef->exec_queue.xa in xe_file_close() as it is not needed and
will cause an unwanted dependency between this lock and the vm->lock.
This lock protects the exec queue lookup and reference taking which
doesn't apply to this code path. When FD is closing, IOCTLs presumably
can't be modifying the xarray.
v2: Update commit text (Matt Brost)
v3: Add more code comment (Rodrigo Vivi)
v4: Further expand code comment (Rodirgo Vivi)
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Jagmeet Randhawa <jagmeet.randhawa@intel.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240529221639.23117-1-niranjana.vishwanathapura@intel.com
We don't have to use exact names of the enumerators as the
potentially user-facing platform names. When constructing
platform descriptor fields, use the unique platform tag and
add the XE_ prefix only to the generated enum field.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240521142257.756-3-michal.wajdeczko@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Add engine name to the engine reset and cat error log
which should be useful while debugging.
v2: Add logical mask and engine class(Matt)
Use xe_gt_{info|dbg} (Michal)
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240528101445.27688-1-nirmoy.das@intel.com
Signed-off-by: Nirmoy Das <nirmoy.das@intel.com>
