On current platforms with multiple GTs, all of the GT IDs are
consecutive; as a result we know that the GT IDs range from 0 to
gt_count-1 and can determine if a GT ID is valid by comparing against
the count. The consecutive nature of GT IDs may not hold true on future
platforms if/when we have platforms that are both multi-tile and have
multiple GTs within each tile. Once such platforms exist, it's quite
possible that we could wind up with something like a GT list composed of
IDs 0, 2, and 3 with no GT 1 (which would be a 2-tile platform with
media only on the second tile).
To future-proof the code we should stop comparing against the GT count
to determine whether a GT ID is valid or not. Instead we should do an
actual lookup of the ID to determine whether the GT exists. This also
means that our GT loop macro should not end at the GT count, but should
rather examine the entire space up to (# of tiles) * (max GT per tile)
to ensure it doesn't stop prematurely.
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://lore.kernel.org/r/20250701201320.2514369-15-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Sometimes it's useful to load the driver with a smaller set of engines
to allow more targeted debugging, particularly on early enabling.
Besides checking what is fused off in hardware, add similar logic to
disable engines in software. This will use configfs to allow users
to set what engine to disable, so already add prepare for that. The
exact configfs interface will be added later.
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20250528-engine-mask-v4-3-f4636d2a890a@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
There's a mismatch on API: while xe_rtp_process_to_sr() processes
entries until an entry without name, the active tracking with
xe_rtp_process_ctx_enable_active_tracking() needs to use the number of
elements. The number of elements is taken everywhere using ARRAY_SIZE(),
but that will have one entry too many. This leads to the following
warning, as reported by lkp:
drivers/gpu/drm/xe/xe_tuning.c: In function 'xe_tuning_dump':
>> include/drm/drm_print.h:228:31: warning: '%s' directive argument is null [-Wformat-overflow=]
228 | drm_printf((printer), "%.*s" fmt, (indent), "\t\t\t\t\tX", ##__VA_ARGS__)
| ^~~~~~
drivers/gpu/drm/xe/xe_tuning.c:226:17: note: in expansion of macro 'drm_printf_indent'
226 | drm_printf_indent(p, 1, "%s\n", engine_tunings[idx].name);
| ^~~~~~~~~~~~~~~~~
That's because it will still process the last entry when tracking the
active tunings. The same issue exists in the WAs. Change
xe_rtp_process_to_sr() to also take the number of elements so the empty
entry can be removed and the warning should go away. Fixing on the
active-tracking side would more fragile as the it would need a `- 1`
everywhere and continue to use a different approach for number of
elements.
Aside from the warning, it's a non-issue as there would always be enough
bits allocated and the last entry would never be active since
xe_rtp_process_to_sr() stops on the sentinel.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202503021906.P2MwAvyK-lkp@intel.com/
Cc: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250306-fix-print-warning-v1-1-979c3dc03c0d@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Instead of handling the whitelist directly in the GuC ADS
initialization, make it follow the same logic as other engine registers
that are save-restored. Main benefit is that then the SW tracking then
shows it in debugfs and there's no risk of an engine workaround to write
to the same nopriv register that is being passed directly to GuC.
This means that xe_reg_whitelist_process_engine() only has to process
the RTP and convert them to entries for the hwe. With that all the
registers should be covered by xe_reg_sr_apply_mmio() to write to the HW
and there's no special handling in GuC ADS to also add these registers
to the list of registers that is passed to GuC.
Example for DG2:
# cat /sys/kernel/debug/dri/0000\:03\:00.0/gt0/register-save-restore
...
Engine
rcs0
...
REG[0x24d0] clr=0xffffffff set=0x1000dafc masked=no mcr=no
REG[0x24d4] clr=0xffffffff set=0x1000db01 masked=no mcr=no
REG[0x24d8] clr=0xffffffff set=0x0000db1c masked=no mcr=no
...
Whitelist
rcs0
REG[0xdafc-0xdaff]: allow read access
REG[0xdb00-0xdb1f]: allow read access
REG[0xdb1c-0xdb1f]: allow rw access
v2:
- Use ~0u for clr bits so it's just a write (Matt Roper)
- Simplify helpers now that unused slots are not written
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241209232739.147417-6-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
When we decide to kill a job, (from guc_exec_queue_timedout_job), we could
end up with 4 possible scenarios at this starting point of this decision:
1. the guc-captured register-dump is already there.
2. the driver is wedged.mode > 1, so GuC-engine-reset / GuC-err-capture
will not happen.
3. the user has started the driver in execlist-submission mode.
4. the guc-captured register-dump is not ready yet so we force GuC to kill
that context now, but:
A. we don't know yet if GuC will be successful on the engine-reset
and get the guc-err-capture, else kmd will do a manual reset later
OR B. guc will be successful and we will get a guc-err-capture
shortly.
So to accomdate the scenarios of 2 and 4A, we will need to do a manual KMD
capture first(which is not be reliable in guc-submission mode) and decide
later if we need to use that for the cases of 2 or 4A. So this flow is
part of the implementation for this patch.
Provide xe_guc_capture_get_reg_desc_list to get the register dscriptor
list.
Add manual capture by read from hw engine if GuC capture is not ready.
If it becomes ready at later time, GuC sourced data will be used.
Although there may only be a small delay between (1) the check for whether
guc-err-capture is available at the start of guc_exec_queue_timedout_job
and (2) the decision on using a valid guc-err-capture or manual-capture,
lets not take any chances and lock the matching node down so it doesn't
get re-claimed if GuC-Err-Capture subsystem is running out of pre-cached
nodes.
Signed-off-by: Zhanjun Dong <zhanjun.dong@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241004193428.3311145-6-zhanjun.dong@intel.com
The xe_guc_exec_queue_snapshot is not really a GuC internal thing and
is definitely not a GuC CT thing. So give it its own section heading.
The snapshot itself is really a capture of the submission backend's
internal state. Although all it currently prints out is the submission
contexts. So label it as 'Contexts'. If more general state is added
later then it could be change to 'Submission backend' or some such.
Further, everything from the GuC CT section onwards is GT specific but
there was no indication of which GT it was related to (and that is
impossible to work out from the other fields that are given). So add a
GT section heading. Also include the tile id of the GT, because again
significant information.
Lastly, drop a couple of unnecessary line feeds within sections.
v2: Add GT section heading, add tile id to device section.
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Reviewed-by: Julia Filipchuk <julia.filipchuk@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241003004611.2323493-4-John.C.Harrison@Intel.com
A xe_hw_engine_group is a group of hw engines. Two hw engines belong to
the same xe_hw_engine_group if one hw engine cannot make progress while
the other is stuck on a page fault.
Typically, hw engines of the same group share some resources such as EUs,
but this really depends on the hardware configuration of the platforms.
The simple engines partitioning proposed here might be too conservative
but is intended to work for existing platforms. It can be optimized later
if more sets of independent engines are identified.
The hw engine groups are intended to be used in the context of faulting
long-running jobs submissions.
v2: Move to own files, improve error handling (Matt Brost)
v3: Fix build issue reported by CI, improve commit message (Matt Roper)
v4: Fix kernel doc
v5: Add switch case for XE_ENGINE_CLASS_OTHER
Signed-off-by: Francois Dugast <francois.dugast@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/20240809155156.1955925-2-francois.dugast@intel.com
Enable feature to allow memory reads to take a priority memory path.
This will reduce latency on the read path, but may introduce read after
write (RAW) hazards as read and writes will no longer be ordered.
To avoid RAW hazards, SW can use the MI_MEM_FENCE command or any other
MI command that generates non posted memory writes. This will ensure
data is coherent in memory prior to execution of commands which read
data from memory. RCS,BCS and CCS support this feature.
No pattern identified in KMD that could lead to a hazard.
v2: Modify commit message, enable priority mem read feature for media,
modify version range, modify bspec detail (Matt Roper)
v3: Rebase, fix cramped line-wrapping (jcavitt)
v4: Rebase
v5: Media does not support Priority Mem Read. Modify commit
to reflect the same.
v6: Rebase
Bspec: 60298, 60237, 60187, 60188
Signed-off-by: Pallavi Mishra <pallavi.mishra@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Acked-by: José Roberto de Souza <jose.souza@intel.com>
Acked-by: Carl Zhang <carl.zhang@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240731195622.1868401-1-pallavi.mishra@intel.com
Per client engine utilization uses RING_TIMESTAMP to return
drm-total-cycles to the user. Current code uses XE_FW_GT to read this
register on the first available engine in a GT. When testing on DG2, it
is observed that this value is 0 when running test on some engines. To
resolve that, get the hwe domain specific FW for reading the engine
timestamp.
v2:
- update commit message
- use domain specific FW (Matt)
v3:
- Drop check for hwe in the helper (Matt, Michal)
v4:
- checkpatch fixes
v5: Rebase
Fixes: 188ced1e0f ("drm/xe/client: Print runtime to fdinfo")
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240627235105.2631135-1-umesh.nerlige.ramappa@intel.com
The flags stored in the BO grew over time without following
much a naming pattern. First of all, get rid of the _BIT suffix that was
banned from everywhere else due to the guideline in
drivers/gpu/drm/i915/i915_reg.h that xe kind of follows:
Define bits using ``REG_BIT(N)``. Do **not** add ``_BIT`` suffix to the name.
Here the flags aren't for a register, but it's good practice to keep it
consistent.
Second divergence on names is the use or not of "CREATE". This is
because most of the flags are passed to xe_bo_create*() family of
functions, changing its behavior. However, since the flags are also
stored in the bo itself and checked elsewhere in the code, it seems
better to just omit the CREATE part.
With those 2 guidelines, all the flags are given the form
XE_BO_FLAG_<FLAG_NAME> with the following commands:
git grep -le "XE_BO_" -- drivers/gpu/drm/xe | xargs sed -i \
-e "s/XE_BO_\([_A-Z0-9]*\)_BIT/XE_BO_\1/g" \
-e 's/XE_BO_CREATE_/XE_BO_FLAG_/g'
git grep -le "XE_BO_" -- drivers/gpu/drm/xe | xargs sed -i -r \
-e 's/XE_BO_(DEFER_BACKING|SCANOUT|FIXED_PLACEMENT|PAGETABLE|NEEDS_CPU_ACCESS|NEEDS_UC|INTERNAL_TEST|INTERNAL_64K|GGTT_INVALIDATE)/XE_BO_FLAG_\1/g'
And then the defines in drivers/gpu/drm/xe/xe_bo.h are adjusted to
follow the coding style.
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240322142702.186529-3-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Starting on Xe2, the GSCCS engine reset is a 2-step process. When the
driver or the GuC hits the GDRST register, the CS is immediately reset
and a success is reported, but the GSC shim continues its reset in the
background. While the shim reset is ongoing, the CS is able to accept
new context submission, but any commands that require the shim will
be stalled until the reset is completed. This means that we can keep
submitting to the GSCCS as long as we make sure that the preemption
timeout is big enough to cover any delay introduced by the reset; since
the GSC preempt timeout is not tunable at runtime, we only need to check
that the value set in kconfig is big enough (and increase it if it
isn't).
When the shim reset completes, a specific CS interrupt is triggered,
in response to which we need to check the GSCI_TIMER_STATUS register
to see if the reset was successful or not.
Note that the GSCI_TIMER_STATUS register is not power save/restored,
so it gets reset on MC6 entry. However, a reset failure stops MC6,
so in that scenario we're always guaranteed to find the correct value.
Since we can't check the register within interrupt context, the
existing GSC worker has been updated to handle it.
The expected action to take on ER failure is to trigger a driver FLR,
but we still don't support that, so for now we just print an error. A
comment has been added to the code to keep track of the FLR requirement.
v2: Add a check for the initial timeout value (Alan)
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240304145634.820684-1-daniele.ceraolospurio@intel.com
