drm/xe: make xe_ttm_funcs const

Place the function pointers in rodata. Also drop the extra declaration
while at it.

Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117122044.1544174-1-jani.nikula@intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/xe: Modify the cfb size to be page size aligned for FBC

drm_gem_private_object_init expect the object size be page size
aligned. The xe_bo create functions do not update the size for
any alignment requirements. So align cfb size to be page size
aligned in xe stolen memory handling.

Signed-off-by: Vinod Govindapillai <vinod.govindapillai@intel.com>
Reviewed-by: Jouni Högander <jouni.hogander@intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240112092803.61664-2-vinod.govindapillai@intel.com

drm/xe: correct the calculation of remaining size

In function write_pgtable, the calculation of chunk in the do-while
loop is wrong, we should always compare against remaining size instead
of the total size update->qwords.

Signed-off-by: Fei Yang <fei.yang@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240116223709.652585-2-fei.yang@intel.com

drm/xe/gsc: add support for GSC proxy interrupt

The GSC notifies us of a proxy request via the HECI2 interrupt. The
interrupt must be enabled both in the HECI layer and in our usual gt irq
programming; for the latter, the interrupt is enabled via the same enable
register as the GSC CS, but it does have its own mask register. When the
interrupt is received, we also need to de-assert it in both layers.

The handling of the proxy request is deferred to the same worker that we
use for GSC load. New flags have been added to distinguish between the
init case and the proxy interrupt.

v2: rename irq define, fix include ordering (Alan)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117182621.2653049-3-daniele.ceraolospurio@intel.com

drm/xe/gsc: Initialize GSC proxy

The GSC uC needs to communicate with the CSME to perform certain
operations. Since the GSC can't perform this communication directly on
platforms where it is integrated in GT, the graphics driver needs to
transfer the messages from GSC to CSME and back. The proxy flow must be
manually started after the GSC is loaded to signal to GSC that we're
ready to handle its messages and allow it to query its init data from
CSME.

Note that the component must be removed before the pci_remove call
completes, so we can't use a drmm helper for it and we need to instead
perform the cleanup as part of the removal flow.

v2: add function documentation, more targeted memory clear, clearer logs
and variable names (Alan)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
Cc: Suraj Kandpal <suraj.kandpal@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117182621.2653049-2-daniele.ceraolospurio@intel.com

drm/xe/guc: Add kernel-doc for xe_guc_ct_send_recv()

Add initial documentation for recently updated xe_guc_ct_send_recv().

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/20240112102554.761-2-michal.wajdeczko@intel.com

drm/xe/guc: Return CTB HXG response DATA0 if no buffer provided

Most of the synchronous GuC HXG action responses are defined in
such a way that only mandatory DATA0 from the HXG header is used
and only in few cases it is more than MBZ (must be zero).

For those cases where HXG action returns just DATA0, return that
value if caller didn't provide buffer for the full response.

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/20240112102554.761-1-michal.wajdeczko@intel.com

drm/xe: Mark internal gmdid mappings as const

The mapping between HW IP version and its description is const, so
mark it as such.

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240116120207.1133-1-michal.wajdeczko@intel.com

drm/xe/guc: Enable WA 14018913170

The GuC handles the WA, the KMD just needs to set the flag to enable
it on the appropriate platforms.

v2:
  - Fixed CI checkpatch warning, alignment should match open parenthesis.
  - Fixed GUC FW version check to use XE_UC_FW_VER_RELEASE which points to
    current GUC FW version instead of XE_UC_FW_VER_COMPATIBILITY which
    holds GUC FW I/F version (Badal).
v3:
  - Removed extra character in debug print.

Signed-off-by: Karthik Poosa <karthik.poosa@intel.com>
Reviewed-by: Badal Nilawar <badal.nilawar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117055035.2417711-1-karthik.poosa@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Use a NULL pointer instead of 0.

The last argument of xe_pcode_read() is a pointer. Use NULL instead of 0.

Fixes: 92d44a422d0d ("drm/xe/hwmon: Expose card reactive critical power")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Francois Dugast <francois.dugast@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117134048.165425-6-thomas.hellstrom@linux.intel.com

drm/xe/dmabuf: Make xe_dmabuf_ops static

It is not referenced outside of the xe_dma_buf.c source file.

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Francois Dugast <francois.dugast@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240117134048.165425-2-thomas.hellstrom@linux.intel.com

drm/xe: Group normal kunit tests in a single module

Creating one module for each compilation unit to be tested seems
excessive as the number of tests increase. Group them all in a single
kunit test module called xe_test.ko.

The tests requiring the physical device, aka "live" tests, are still
kept in separate modules since they are normally triggered via igt,
and not via kunit.py. After igt is converted, those can be merged in
a single module as well.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231122203147.988021-2-lucas.demarchi@intel.com

drm/xe: Use kstrdup while creating snapshot

There is no need to copy string step by step, use existing helper.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://lore.kernel.org/r/20240112160652.893-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: display support should not depend on EXPERT

Remove the DRM_XE_DISPLAY config dependency on EXPERT. I can only
presume the idea was only experts should be able to disable it, but the
effect is the opposite.

Reported-by: Eero Tamminen <eero.t.tamminen@intel.com>
Reviewed-by: Francois Dugast <francois.dugast@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240111104716.3548744-1-jani.nikula@intel.com

drm/xe: Fix bounds checking in __xe_bo_placement_for_flags()

Requesting all memory regions on PVC will fill bo->placements up to
XE_BO_MAX_PLACEMENTS. The subsequent call to try_add_stolen() will trip
over the bounds checking even though XE_PL_STOLEN is not expected to
be used in this case.

This is hit with igt@xe_exec_fault_mode@once-basic-prefetch:
    xe 0000:8c:00.0: [drm] Assertion `*c < (sizeof(bo->placements) / sizeof((bo->placements)[0]) + ((int)(sizeof(struct { int:(-!!(__builtin_types_compatible_p(typeof((bo->placements)), typeof(&(bo->placements)[0])))); }))))` failed!
    WARNING: CPU: 30 PID: 6161 at drivers/gpu/drm/xe/xe_bo.c:203 __xe_bo_placement_for_flags+0x218/0x240 [xe]

Is fixed here by moving the bounds checks closer to where we actually
write into the bo->placement array.

Fixes: 8c54ee8a8606 ("drm/xe: Ensure that we don't access the placements array out-of-bounds")
Link: https://patchwork.freedesktop.org/patch/msgid/20240111002111.10190-1-brian.welty@intel.com
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Brian Welty <brian.welty@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe/migrate: Cap PTEs written by MI_STORE_DATA_IMM to 510

Although MI_STORE_DATA_IMM's "length" field is 10-bits, 0x3FE is
considered the largest legal value accepted.  Since that instruction
field is always encoded in (val-2) format, this translates to 0x400
dwords for the true maximum length of the instruction.  Subtracting the
instruction header (1 dword) and address (2 dwords), that leaves 0x3FD
dwords (i.e., 0x1FE qwords) for PTE values.

Bspec: 60246, 45753
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20240111220238.1467572-2-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

drm/xe: Fix potential deadlock in __fini_dbm

If Doorbell Manager is in unclean state during fini phase, for
debug purposes we try to print it's state, but we missed the fact
that we are already holding a lock so the xe_guc_db_mgr_print()
will deadlock since it also attempts to grab the same lock.

Fixes: 587c73343ac7 ("drm/xe: Introduce GuC Doorbells Manager")
Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240111185603.673-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Allow to exclude part of GGTT from allocations

Soon we will be required to exclude some of the GGTT addresses
from the allocations, since on some platforms running the SR-IOV VF
mode, we will be able to use only selected range of the GGTT space.

Add helper functions to manage such GGTT range exclusions, and
follow the naming from the similar concept used by GVT-g.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240111182559.629-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Use HXG definitions on HXG messages

While parsing and processing CTB G2H messages we should extract
underlying HXG message and use HXG definitions on such message.
Using outer CTB layer message in HXG definitions require use of
shifted dword index, which might be confusing:

FIELD_GET(GUC_HXG_MSG_0_xxx, msg[1])

instead of:

FIELD_GET(GUC_HXG_MSG_0_xxx, hxg[0])

Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20240111210632.717-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Return CTB response length

Not all CTB responses from the GuC are fixed size and we need to
pass response length to the caller, if there was a response_buffer.
Easiest solution is to return it as positive value from all
xe_guc_ct_send_recv() functions.  The CTB response length is always
between 1 and 254 (ie. GUC_HXG_MSG_MIN_LEN and GUC_CTB_MAX_DWORDS
- GUC_HXG_MSG_MIN_LEN).

Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20240111152724.497-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Treat non-response message after BUSY as unexpected

Once GuC replied with GUC_HXG_TYPE_NO_RESPONSE_BUSY message then
we may expect that only RESPONSE_SUCCESS or FAILURE message will
be sent, anything else is a violation of the HXG protocol.

Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20240111154838.541-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Split GuC communication initialization

Soon we will be trying to communicate with the GuC firmware very
early during VF driver probe, before we finish normal init steps.
Split GuC communication initialization code so the GuC MMIO based
communication xe_guc_mmio_send() functions will work where needed.

Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20240111162051.585-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/migrate: Fix CCS copy for small VRAM copy chunks

Since the migrate code is using the identity map for addressing VRAM,
copy chunks may become as small as 64K if the VRAM resource is fragmented.

However, a chunk size smaller that 1MiB may lead to the *next* chunk's
offset into the CCS metadata backup memory may not be page-aligned, and
the XY_CTRL_SURF_COPY_BLT command can't handle that, and even if it could,
the current code doesn't handle the offset calculaton correctly.

To fix this, make sure we align the size of VRAM copy chunks to 1MiB. If
the remaining data to copy is smaller than that, that's not a problem,
so use the remaining size. If the VRAM copy cunk becomes fragmented due
to the size alignment restriction, don't use the identity map, but instead
emit PTEs into the page-table like we do for system memory.

v2:
- Rebase
v3:
- Future proof somewhat by taking into account the real data size to
  flat CCS metadata size ratio. (Matt Roper)
- Invert a couple of if-statements for better readability.
- Fix support for 4K-granularity VRAM sizes. (Tested on DG1).
v4:
- Fix up code comments
- Fix debug printout format typo.
v5:
- Add a Fixes: tag.

Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Fixes: e89b384cde62 ("drm/xe/migrate: Update emit_pte to cope with a size level than 4k")
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240110163415.524165-1-thomas.hellstrom@linux.intel.com

drm/xe: unlock on error path in xe_vm_add_compute_exec_queue()

Drop the "&vm->lock" before returning.

Fixes: 24f947d58fe5 ("drm/xe: Use DRM GPUVM helpers for external- and evicted objects")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe/selftests: Fix an error pointer dereference bug

Check if "bo" is an error pointer before calling xe_bo_lock() on it.

Fixes: d6abc18d6693 ("drm/xe/xe2: Modify xe_bo_test for system memory")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe/device: clean up on error in probe()

This error path should clean up before returning.

Smatch detected this bug:
  drivers/gpu/drm/xe/xe_device.c:487 xe_device_probe() warn: missing unwind goto?

Fixes: 4cb12b71923b ("drm/xe/xe2: Determine bios enablement for flat ccs on igfx")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Invert access counter queue head / tail

Convention for queues in Linux is the producer moves the head and
consumer moves the tail. Fix the access counter queue to conform to
this convention.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Add build on bug to assert access counter queue works

If ACC_QUEUE_NUM_DW % ACC_MSG_LEN_DW != 0 then the access counter queue
logic does not work when wrapping occurs. Add a build bug on to assert
ACC_QUEUE_NUM_DW % ACC_MSG_LEN_DW == 0 to enforce this restriction and
document the code.

v2:
- s/NUM_ACC_QUEUE/ACC_QUEUE_NUM_DW (Brian)

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Invert page fault queue head / tail

Convention for queues in Linux is the producer moves the head and
consumer moves the tail. Fix the page fault queue to conform to this
convention.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Add build on bug to assert page fault queue works

If PF_QUEUE_NUM_DW % PF_MSG_LEN_DW != 0 then the page fault queue logic
does not work when wrapping occurs. Add a build bug on to assert
PF_QUEUE_NUM_DW % PF_MSG_LEN_DW == 0 to enforce this restriction and
document the code.

v2:
- s/NUM_PF_QUEUE/PF_QUEUE_NUM_DW (Brian)

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Remove set_job_timeout_ms() from exec_queue_ops

This function is no longer used as the job_timeout is now
updated prior to calling queue_ops.init().

Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Finish refactoring of exec_queue_create

Setting of exec_queue user extensions is moved from the end of the ioctl
function earlier, into __xe_exec_queue_alloc().
This fixes bug in that the USM attributes for access counters were being
applied too late, and effectively were ignored.

However, in order to apply user extensions this early, we can no longer
call q->ops functions.  Instead, make it more efficient. The user extension
functions can simply update the q->sched_props values and they will be
applied by the backend during q->ops->init().

v2: minor changes for readability (Matt)

Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Add exec_queue.sched_props.job_timeout_ms

The purpose here is to allow to optimize exec_queue_set_job_timeout()
in follow-on patch.  Currently it does q->ops->set_job_timeout(...).
But we'd like to apply exec_queue_user_extensions much earlier and
q->ops cannot be called before __xe_exec_queue_init().

It will be much more efficient to instead only have to set
q->sched_props.job_timeout_ms when applying user extensions. That value
will then be used during q->ops->init().

Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Refactor __xe_exec_queue_create()

Split __xe_exec_queue_create() into two functions, alloc and init.

We have an issue in that exec_queue_user_extensions are applied too late.
In the case of USM properties, these need to be set prior to xe_lrc_init().
Refactor the logic here, so we can resolve this in follow-on. We only need
the xe_vm_lock held during the exec_queue_init function.

Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Fix build bug for GCC 11

Building drivers/gpu/drm/xe/xe_gt_pagefault.c with GCC 11 results
in the following build errors:

./include/linux/fortify-string.h:57:33: error: writing 16 bytes into a region of size 0 [-Werror=stringop-overflow=]
   57 | #define __underlying_memcpy     __builtin_memcpy
      |                                 ^
./include/linux/fortify-string.h:644:9: note: in expansion of macro ‘__underlying_memcpy’
  644 |         __underlying_##op(p, q, __fortify_size);                        \
      |         ^~~~~~~~~~~~~
./include/linux/fortify-string.h:689:26: note: in expansion of macro ‘__fortify_memcpy_chk’
  689 | #define memcpy(p, q, s)  __fortify_memcpy_chk(p, q, s,                  \
      |                          ^~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/xe/xe_gt_pagefault.c:340:17: note: in expansion of macro ‘memcpy’
  340 |                 memcpy(pf_queue->data + pf_queue->tail, msg, len * sizeof(u32));
      |                 ^~~~~~
In file included from drivers/gpu/drm/xe/xe_device_types.h:17,
                 from drivers/gpu/drm/xe/xe_vm_types.h:16,
                 from drivers/gpu/drm/xe/xe_bo.h:13,
                 from drivers/gpu/drm/xe/xe_gt_pagefault.c:16:
drivers/gpu/drm/xe/xe_gt_types.h:102:25: note: at offset [1144, 265324] into destination object ‘tile’ of size 8
  102 |         struct xe_tile *tile;
      |                         ^~~~

Fix these by removing -Wstringop-overflow from drm/xe builds.

Closes: https://lore.kernel.org/all/45ad1d0f-a10f-483e-848a-76a30252edbe@paulmck-laptop/
Fixes: 7a8bc11782d3 ("drm/xe: Enable W=1 warnings by default")
Suggested-by: Stephen Rothwell <sfr@rothwell.id.au>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
[ This particular warning is broken on GCC11. In future changes it will
  be moved to the normal C flags in the top level Makefile (out of
  Makefile.extrawarn), but accounting for the compiler support. Just
  remove it out of xe's forced extra warnings for now ]
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Check skip_guc_pc before setting SLPC flag

Don't set SLPC GuC feature ctl flag if skip_guc_pc is true.

v2: Skip the freq related sysfs creation as well (Badal)
v3: Remove unnecessary parenthesis (Lucas)

Fixes: 975e4a3795d4 ("drm/xe: Manually setup C6 when skip_guc_pc is set")
Fixes: bef52b5c7a19 ("drm/xe: Create a xe_gt_freq component for raw management and sysfs")
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://lore.kernel.org/r/20240108225842.966066-1-vinay.belgaumkar@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Add vram frequency sysfs attributes

Add vram frequency sysfs attributes under the below hierarchy;

/device/tile#/memory/freq0
|-max_freq
|-min_freq

v2: Drop "vram" from attribute names (Rodrigo)

v3: Add documentation for new sysfs (Riana)
    Drop prefix from XEHP_PCODE_FREQUENCY_CONFIG (Riana)

v4: Create sysfs under tile#/freq0 after removal of
    physical_memsize attrbute

v5: Revert back to creating sysfs under tile#/memory/freq0
    Remove definition of GT_FREQUENCY_MULTIPLIER (Rodrigo)

v6: Rename attributes to max/min_freq (Anshuman)
    Fix review comments (Rodrigo)

v7: Make documentation more verbose
    Move sysfs to separate file (Anshuman)

v8: Fix platform specific conditions and add kernel doc (Anshuman)
    Fix typos and remove redundant headers (Riana)

v9: Fix typo (Riana)
    Change function name to include "sysfs" (Lucas)

Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
Reviewed-by: Anshuman Gupta <anshuman.gupta@intel.com>
Link: https://lore.kernel.org/r/20240109110418.2065101-1-sujaritha.sundaresan@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix modifying exec_queue priority in xe_migrate_init

After exec_queue has been created, we cannot simply modify q->priority.
This needs to be done by the backend via q->ops.  However in this case,
it would be more efficient to simply pass a flag when creating the
exec_queue and set the desired priority upfront during queue creation.

To that end: new flag EXEC_QUEUE_FLAG_HIGH_PRIORITY is introduced.
The priority field is moved to be with other scheduling properties and
is now exec_queue.sched_props.priority. This is no longer set to initial
value by the backend, but is now set within __xe_exec_queue_create().

Fixes: b4eecedc75c1 ("drm/xe: Fix potential deadlock handling page faults")
Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe: Fix guc_exec_queue_set_priority

We need to set q->priority prior to calling guc_exec_queue_add_msg() as
that will call init_policies() and sets the scheduling properties to those
stored in the exec_queue.

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Signed-off-by: Brian Welty <brian.welty@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>

drm/xe/xe2_lpg: Add Wa_16018610683

Force max 128KB SLM during WMTP PASS1 Restore.

BSpec: 70202
Signed-off-by: Shekhar Chauhan <shekhar.chauhan@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20240109055550.679289-1-shekhar.chauhan@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

drm/xe: Annotate xe_ttm_stolen_mgr::mapping with __iomem

The pointer points to IO memory, but the __iomem annotation was
incorrectly placed. Annotate it correctly, update its usage accordingly
and fix the corresponding sparse error.

Fixes: d8b52a02cb40 ("drm/xe: Implement stolen memory.")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240109112405.108136-5-thomas.hellstrom@linux.intel.com

drm/xe: Annotate multiple mmio pointers with __iomem

There are a couple of pointers pointing to MMIO space. Annotate them
with __iomem and fix the corresponding sparse warnings.

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Fixes: 3b0d4a557996 ("drm/xe: Move register MMIO into xe_tile")
Fixes: 399a13323f0d ("drm/xe: add 28-bit address support in struct xe_reg")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Koby Elbaz <kelbaz@habana.ai>
Cc: Ofir Bitton <obitton@habana.ai>
Cc: Moti Haimovski <mhaimovski@habana.ai>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240109112405.108136-4-thomas.hellstrom@linux.intel.com

drm/xe: Annotate xe_mem_region::mapping with __iomem

The pointer points to IO memory, but the __iomem annotation was
incorrectly placed. Annotate it correctly, update its usage accordingly
and fix the corresponding sparse error.

Fixes: 0887a2e7ab62 ("drm/xe: Make xe_mem_region struct")
Cc: Oak Zeng <oak.zeng@intel.com>
Cc: Michael J. Ruhl <michael.j.ruhl@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240109112405.108136-3-thomas.hellstrom@linux.intel.com

drm/xe: Use __iomem for the regs pointer

The regs pointer points to IO memory. Annotate it properly and
fix the corresponding sparse warning.

Fixes: a4e2f3a299ea ("drm/xe: refactor xe_mmio_probe_tiles to support MMIO extension")
Cc: Koby Elbaz <kelbaz@habana.ai>
Cc: Ofir Bitton <obitton@habana.ai>
Cc: Moti Haimovski <mhaimovski@habana.ai>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240109112405.108136-2-thomas.hellstrom@linux.intel.com

drm/xe/vm: Fix an error path

If using the VM_BIND_OP_UNMAP_ALL without any bound vmas for the
vm, we will end up dereferencing an uninitialized variable and leak a
bo lock. Fix this.

v2:
- Updated commit message (Lucas De Marchi)

Reported-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Closes: https://lore.kernel.org/intel-xe/jrwua7ckbiozfcaodx4gg2h4taiuxs53j5zlpf3qzvyhyiyl2d@pbs3plurokrj/
Suggested-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Fixes: b06d47be7c83 ("drm/xe: Port Xe to GPUVA")
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20231222175904.16732-1-thomas.hellstrom@linux.intel.com

drm/xe/guc: Only take actions in CT irq handler if CTs are enabled

Protect entire IRQ handler by CT being enabled rather than just G2H
handler.

v2: Return on not enabled in CT irq handler (Michal)

Suggested-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Fix exec IOCTL long running exec queue ring full condition

The intent is to return -EWOULDBLOCK to the user if a long running exec
queue is full during the exec IOCTL. -EWOULDBLOCK aliases to -EAGAIN
which results in the exec IOCTL doing a retry loop. Fix this by ensuring
the retry loop is broken when returning -EWOULDBLOCK.

Fixes: 8ae8a2e8dd21 ("drm/xe: Long running job update")
Reported-by: Sai Gowtham Ch <sai.gowtham.ch@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Brian Welty <brian.welty@intel.com>

drm/xe/exec: reserve fence slot for CPU bind

Looks possible to switch from CPU binding to GPU binding mid exec, and
if that happens for the same dma-resv we might use two fence slots, once
for the dummy fence, and another for the actual GPU bind.

References: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/698
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/xe/exec: move fence reservation

We currently assume that we can upfront know exactly how many fence
slots we will need at the start of the exec, however the TTM bo_validate
can itself consume numerous fence slots, and due to how the
dma_resv_reserve_fences() works it only ensures that at least that many
fence slots are available. With this it is quite possible that TTM
steals some of the fence slots and then when it comes time to do the vma
binding and final exec stage we are lacking enough fence slots, leading
to some nasty BUG_ON(). A simple fix is to reserve our own fences later,
after the validate stage.

References: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/698
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Tested-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/xe/dgfx: Release mmap mappings on rpm suspend

Release all mmap mappings for all vram objects which are associated
with userfault such that, while pcie function in D3hot, any access
to memory mappings will raise a userfault.

Upon userfault, in order to access memory mappings, if graphics
function is in D3 then runtime resume of dgpu will be triggered to
transition to D0.

v2:
  - Avoid iomem check before bo migration check as bo can migrate
    to system memory (Matthew Auld)
v3:
  - Delete bo userfault link during bo destroy
  - Upon bo move (vram-smem), do bo userfault link deletion in
    xe_bo_move_notify instead of xe_bo_move (Thomas Hellström)
  - Grab lock in rpm hook while deleting bo userfault link (Matthew Auld)
v4:
  - Add kernel doc and wrap vram_userfault related
    stuff in the structure (Matthew Auld)
  - Get rpm wakeref before taking dma reserve lock (Matthew Auld)
  - In suspend path apply lock for entire list op
    including list iteration (Matthew Auld)
v5:
  - Use mutex lock instead of spin lock
v6:
  - Fix review comments (Matthew Auld)

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Badal Nilawar <badal.nilawar@intel.com>
Acked-by: Thomas Hellström <thomas.hellstrom@linux.intel.com> #For the xe_bo_move_notify() changes
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://lore.kernel.org/r/20240104130702.950078-1-badal.nilawar@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe/xe2: synchronise CS_CHICKEN1 with WMTP support

Recommendation is to read FUSE4 register to check if WMTP has been
enabled/disabled by HW. If enabled we don't need to do anything special,
however if disabled recommendation is to also disable the WMTP mode in
the FF_SLICE_CS_CHICKEN2 register, falling back to thread-group and
mid-batch preemption only. However on Linux, the per-context CS_CHICKEN1
is how userspace controls pre-emption, so instead use the default lrc to
disable WMTP using CS_CHICKEN1, if disabled by HW. Userspace is still
free to set CS_CHICKEN1 to whatever they want later.

v2: remove redundant version check and also add descriptive name(Matt)
v3: remove usage of REG_FIELD_GET(Matt)

Cc: Matt Roper <matthew.d.roper@intel.com>
Co-developed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20240104182615.21327-1-nirmoy.das@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

drm/xe/kunit: Drop xe_wa tests for pre-production DG2

As workarounds for pre-production DG2 were dropped in
commit 707d1b992cfe ("drm/xe/dg2: Drop pre-production workarounds"),
there's no point running the kunit tests for them. Drop
those steppings from kunit.

Cc: Gustavo Sousa <gustavo.sousa@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231221163213.3849523-1-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe: Fix spelling mistake "gueue" -> "queue"

There is a spelling mistake in a drm_info message. Fix it.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20240102092014.3347566-1-colin.i.king@gmail.com

drm/xe: Fix compilation without CONFIG_KUNIT

It looks that declaration of kunit_get_current_test() was
available in xe_guc_relay.c only with CONFIG_KUNIT enabled.
If CONFIG_KUNIT is disabled we fail with:

In file included from ../include/linux/build_bug.h:5,
                 from ../include/linux/bitfield.h:10,
                 from ../drivers/gpu/drm/xe/xe_guc_relay.c:6:
../drivers/gpu/drm/xe/xe_guc_relay.c: In function ‘xe_guc_relay_process_guc2vf’:
../drivers/gpu/drm/xe/xe_guc_relay.c:863:52: error: implicit declaration of function ‘kunit_get_current_test’ [-Werror=implicit-function-declaration]
  863 |  if (unlikely(!IS_SRIOV_VF(relay_to_xe(relay)) && !kunit_get_current_test()))
      |                                                    ^~~~~~~~~~~~~~~~~~~~~~
../include/linux/compiler.h:77:42: note: in definition of macro ‘unlikely’
   77 | # define unlikely(x) __builtin_expect(!!(x), 0)
      |                                          ^

Reported-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Fixes: 811fe9f556fc ("drm/xe/guc: Introduce Relay Communication for SR-IOV")
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Tested-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://lore.kernel.org/r/20240105171947.321-1-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Add mocs reset kunit

This kunit verifies the mocs registers content
with the KMD programmed values before and after
GT reset.

v2: Remove extra blank lines between the local variables
definitions (Matt Roper)

Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Ruthuvikas Ravikumar <ruthuvikas.ravikumar@intel.com>
Signed-off-by: Janga Rahul Kumar <janga.rahul.kumar@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231222192352.927101-1-janga.rahul.kumar@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

drm/xe/guc: Start handling GuC Relay event messages

GuC Relay infrastructure is ready, start handling relay messages
from the GuC to unblock testing on the live system.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-11-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Add GuC Relay kunit tests

Add few tests to make sure that some negative and normal use
scenarios of the GuC Relay are implemented correctly.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-10-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Allow to replace xe_guc_ct_send_recv() with stub

We want to use replacement functions in upcoming kunit tests.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-9-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Introduce Relay Communication for SR-IOV

There are scenarios where SR-IOV Virtual Function (VF) driver will
need to get additional data that is not available over VF MMIO BAR
nor could be queried from the GuC firmware and must be obtained
from the Physical Function (PF) driver.

To allow such communication between VF and PF drivers, GuC supports
set of H2G and G2H actions which allows relaying embedded messages,
that are otherwise opaque for the GuC.

To allow use of this communication mechanism, provide functions for
sending requests and handling replies and placeholder where we will
put handlers for incoming requests.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-8-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Add Relay Communication ABI definitions

The communication between Virtual Function (VF) drivers and
Physical Function (PF) drivers is based on the GuC firmware
acting as a proxy (relay) agent.

Add related ABI definitions that we will be using in upcoming
patches with our GuC Relay implementation.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-7-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Update few GuC CTB ABI definitions

In upcoming new GuC ABI definitions we will want to refer to max
number of dwords that could fit into CTB HXG message. Add explicit
definition named as GUC_CTB_MAX_DWORDS and start using it.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-6-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/guc: Add helpers for HXG messages

In addition to MMIO and CTB communication between the host driver
and the GUC firmware, we will start using GuC HXG message protocol
in communication between SR-IOV VFs and PF. Define helpers related
to HXG message protocol to minimize code duplication.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-5-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Introduce GT-oriented SR-IOV logging macros

To simplify logging and help identify SR-IOV specific messages
related to the GT, define set of helper macros that will add
prefix to the messages based on the current SR-IOV mode.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-4-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Define Virtual Function Identifier

According to the PCI Express specification, the SR-IOV Virtual
Functions (VFs) are numbered starting with 1 (VF1, VF2, ...).
Additionally, both driver and GuC will refer to Physical Function
(PF) as VF0. Define helper macro to represent VFn and PF.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-3-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Allocate dedicated workqueue for SR-IOV workers

We plan to use several workers where we might be running long
operations. Allocate dedicated workqueue to avoid undesired
interaction with non-virtualized workers.

Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20240104222031.277-2-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Use intel_wakeref_t in intel_runtime_pm functions

Now intel_wakeref_t is a unsigned long and Xe KMD version of those
functions should use the same type, so changing from bool to
intel_wakeref_t.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>

drm/xe: Fix definition of intel_wakeref_t

i915 defines it as unsigned long so Xe should do the same to avoid
compilation warnings:

  CC [M]  drivers/gpu/drm/i915/i915_gem.o
  CC [M]  drivers/gpu/drm/xe/i915-display/intel_display_power_well.o
In file included from ./include/drm/drm_mm.h:51,
                 from drivers/gpu/drm/xe/xe_bo_types.h:11,
                 from drivers/gpu/drm/xe/xe_bo.h:11,
                 from ./drivers/gpu/drm/xe/compat-i915-headers/gem/i915_gem_object.h:11,
                 from ./drivers/gpu/drm/xe/compat-i915-headers/i915_drv.h:15,
                 from drivers/gpu/drm/i915/display/intel_display_power.c:8:
drivers/gpu/drm/i915/display/intel_display_power.c: In function ‘print_async_put_domains_state’:
drivers/gpu/drm/i915/display/intel_display_power.c:408:29: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 5 has type ‘int’ [-Wformat=]
  408 |         drm_dbg(&i915->drm, "async_put_wakeref %lu\n",
      |                             ^~~~~~~~~~~~~~~~~~~~~~~~~
  409 |                 power_domains->async_put_wakeref);
      |                 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                              |
      |                              int
./include/drm/drm_print.h:410:39: note: in definition of macro ‘drm_dev_dbg’
  410 |         __drm_dev_dbg(NULL, dev, cat, fmt, ##__VA_ARGS__)
      |                                       ^~~
./include/drm/drm_print.h:510:33: note: in expansion of macro ‘drm_dbg_driver’
  510 | #define drm_dbg(drm, fmt, ...)  drm_dbg_driver(drm, fmt, ##__VA_ARGS__)
      |                                 ^~~~~~~~~~~~~~
drivers/gpu/drm/i915/display/intel_display_power.c:408:9: note: in expansion of macro ‘drm_dbg’
  408 |         drm_dbg(&i915->drm, "async_put_wakeref %lu\n",
      |         ^~~~~~~
drivers/gpu/drm/i915/display/intel_display_power.c:408:50: note: format string is defined here
  408 |         drm_dbg(&i915->drm, "async_put_wakeref %lu\n",
      |                                                ~~^
      |                                                  |
      |                                                  long unsigned int
      |                                                %u
  CC [M]  drivers/gpu/drm/i915/i915_gem_evict.o
  CC [M]  drivers/gpu/drm/i915/i915_gem_gtt.o
  CC [M]  drivers/gpu/drm/xe/i915-display/intel_display_trace.o
  CC [M]  drivers/gpu/drm/xe/i915-display/intel_display_wa.o
  CC [M]  drivers/gpu/drm/i915/i915_query.o

Fixes: 44e694958b95 ("drm/xe/display: Implement display support")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915: don't make assumptions about intel_wakeref_t type

intel_wakeref_t is supposed to be a mostly opaque cookie to its
users. It should only be checked for being non-zero and set to
zero. Debug logging its actual value is meaningless. Switch to just
debug logging whether the async_put_wakeref is non-zero.

The issue dates back to much earlier than
commit b49e894c3fd8 ("drm/i915: Replace custom intel runtime_pm tracker
with ref_tracker library"), but this is the one that brought about a
build failure due to the printf format.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Closes: https://lore.kernel.org/r/20240102111222.2db11208@canb.auug.org.au
Fixes: b49e894c3fd8 ("drm/i915: Replace custom intel runtime_pm tracker with ref_tracker library")
Cc: Andrzej Hajda <andrzej.hajda@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240104164600.783371-1-jani.nikula@intel.com

drm/xe/xe2: Add workaround 16020183090

Graphics version 20.04, used in Lunar Lake, needs WA 16020183090 for
steppings A*. Set ENABLE_SEMAPHORE_POLL_BIT in INSTPM(RENDER_RING_BASE)
and whitelist CSBE_DEBUG_STATUS for userspace to be able to use it
and complement the workaround.

Cc: Haridhar Kalvala <haridhar.kalvala@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231207175117.2334022-1-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe/uapi: Remove DRM_XE_VM_BIND_FLAG_ASYNC comment left over

This is a comment left over of commit d3d767396a02
("drm/xe/uapi: Remove sync binds").

Fixes: d3d767396a02 ("drm/xe/uapi: Remove sync binds")
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>

drm/xe: Disable 32bits build

Add a dependency on CONFIG_64BIT since currently the xe driver doesn't
build on 32bits. It may be enabled again after all the issues are fixed.

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://lore.kernel.org/r/20231221222809.4123220-2-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

drm/xe/guc: Use FAST_REQUEST for non-blocking H2G messages

We're currently sending non-blocking H2G messages using the EVENT type,
which suppresses all CTB protocol replies from the GuC, including the
failure cases. This might cause errors to slip through and manifest as
unexpected behavior (e.g. a context state might not be what the driver
thinks it is because the state change command was silently rejected by
the GuC). To avoid this kind of problems, we can use the FAST_REQUEST
type instead, which suppresses the reply only on success; this way we
still get the advantage of not having to wait for an ack from the GuC
(i.e. the H2G is still non-blocking) while still detecting errors.
Since we can't escalate to the caller when a non-blocking message
fails, we need to escalate to GT reset instead.

Note that FAST_REQUEST failures are NOT expected and are usually a sign
that the H2G was either malformed or requested an illegal operation.

v2: assign fence values to FAST_REQUEST messages, fix abi doc, use xe_gt
printers (Michal).

v3: fix doc alignment, fix and improve prints (Michal)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com> #v2
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/xelpg: Extend Wa_14019877138 for Graphics 12.70/71

Wa_14019877138 is also needed for xe_lpg graphics 12.70/71

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>

drm/xe/kunit: Add GuC Doorbells Manager tests

Add few tests to make sure that basic usage scenarios of the GuC
Doorbells Manager are implemented correctly.

Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-11-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Enable CONFIG_LOCKDEP in tests

Tests might use locking, better to have LOCKDEP enabled.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-10-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Use xe kunit helper in WA test

Use xe helper code to allocate fake xe device.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-9-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Use xe kunit helper in RTP test

Replace drm_kunit_helper_alloc_drm_device with xe helper.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-8-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Restore test->priv when done with fake xe device

Current KUnit implementation does not reset test->priv in case of
parametrized tests and that may lead to wrongly treat our output
pointer to fake xe_device from first call as input pointer with
xe_pci_fake_data on subsequent calls.  Restore test->priv to
original value to avoid invalid access.

Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-7-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Define helper functions to allocate fake xe device

There will be more KUnit tests added that will require fake device.
Define generic helper functions to avoid code duplications.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-6-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/kunit: Set SR-IOV mode of the fake device

We want to add code that will check the driver's SR-IOV mode.

Update xe_pci_fake_device_init() and struct xe_pci_fake_data to
either explicitly specify desired SR-IOV mode of the fake device
or fallback to the default bare-metal mode.

Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-5-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Introduce GuC Doorbells Manager

The GFX doorbell solution provides a mechanism for submission of
workload to the graphics hardware by a ring3 application without
the penalty of ring transition for each workload submission.

This feature is not currently used by the Linux drivers, but in
SR-IOV mode the doorbells are treated as shared resource and the
PF driver must be able to provision exclusive range of doorbells
IDs across all enabled VFs.

Introduce simple GuC doorbell ID manager that will be used by the
PF driver for VFs provisioning and can later be used by submission
code once we are ready to switch from H2G based notifications to
doorbells mechanism.

Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-4-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Report TLB timeout using GT oriented functions

We track TLB invalidation seqno per GT and we have GT oriented
message helpers, so it's better to use GT oriented log functions.

Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-3-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Add GT oriented drm_printers

Instead of using generic drm_err_printer() that just adds static
prefix, or drm_info_printer() that outputs only device name, add
new helpers that create dedicated drm_printers that use our GT
oriented xe_gt_err() and xe_gt_info() functions.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231218190629.502-2-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/vf: Add VF specific interrupt handler

There are small differences in handling of the register based
interrupts on the VF driver as some registers are not accessible
to the VF driver. Additionally VFs must support Memory Based
Interrupts. Add VF specific interrupt handler for this.

Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-11-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/vf: Setup memory based interrupts in GuC

When Memory Based Interrupts are used, the VF driver must provide
to the GuC references to the Source and Status Report Pages.

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-10-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/vf: Update LRC with memory based interrupts data

When Memory Based Interrupts are used, the VF driver must provide
in the LRC the references to the Source and Status Report Pages.

Update the LRC according to the requirements.

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-9-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/vf: Introduce Memory Based Interrupts Handler

The register based interrupts infrastructure does not scale
efficiently to allow delivering interrupts to a large number
of virtual machines. Memory based interrupt reporting provides
an efficient and scalable infrastructure.

Define handler to read and dispatch memory based interrupts.
We will use this handler in upcoming patch.

Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-8-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Add XE_BO_NEEDS_UC flag to force UC mode instead WB

When we map BO in GGTT, then by default we are using PAT index
that corresponds to XE_CACHE_WB but ppcoming feature will require
use of the PAT index of the XE_CACHE_UC.  Define new BO flag that
could be used during BO creation to force alternate caching.

Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-7-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Define IRQ offsets used by HW engines

When interrupts are delivered using memory based mechanism, engines
will write status to the report page at the offset (in bytes) that
corresponds to their interrupt bit from the GT_INTR_DW register.

Add engine interrupt offset definitions to engine info as we will
need this to process memory based interrupts.

Bspec: 46149, 50829, 50844
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-6-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Update definition of GT_INTR_DW

Add bits definitions that we will be using in upcoming patch.

Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-5-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Update LRC context layout definitions

The new memory based interrupt processing uses additional entries
in the context.  Add required definitions.

Bspec: 45585, 60184
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-4-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Define registers used by memory based irq processing

The RING_INT_SRC_RPT_PTR register points to a cacheline in memory
to which an engine must report as source of interrupt prior to
generating an interrupt to the host.

The RING_INT_STATUS_RPT_PTR register points to the first cacheline
of the Interrupt Status Report (ISR) page (4KB) in graphics memory
to which all engines report their interrupt status.

The RING_IMR register has the interrupt enables and interrupt masks
for an engine.

We will refer to these registers shortly.

Bspec: 45963, 45964, 45965
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-3-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe: Add command MI_LOAD_REGISTER_MEM

We will need this shortly during context state preparation.

Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://lore.kernel.org/r/20231214185955.1791-2-michal.wajdeczko@intel.com
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>

drm/xe/dg2: Drop pre-production workarounds

Pre-production hardware is anything before C0 (for DG2-G10), before B1
(for DG2-G11), or before A1 (for DG2-G12).  Workarounds specific to such
hardware was already removed from i915 in commit eaeb4b361452
("drm/i915/dg2: Drop pre-production GT workarounds") and there's even
less value keeping these around in the Xe driver.

v2:
 - Drop Wa_14011441408 from xe_mocs.c.  (Gustavo)
 - Drop Wa_14010648519, Wa_14010198302, and Wa_1608949956 which were
   mis-implemented; they were only supposed to apply to early steppings
   of DG2-G10, but were being applied unconditionally on all DG2.
   (Gustavo)
 - Drop reference to Wa_16011620976; the implementation stays because it
   still matches Wa_22015475538.  (Gustavo)

Cc: Gustavo Sousa <gustavo.sousa@intel.com>
Reviewed-by: Gustavo Sousa <gustavo.sousa@intel.com>
Link: https://lore.kernel.org/r/20231215214531.2576215-2-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>

drm/xe: Fix warning on impossible condition

Having a different value for op is not possible: this is already kept
out of user-visible warning by the check in xe_wait_user_fence_ioctl()
if op > MAX_OP. The warning is useful as if this switch() is not update
when a new op is added, it should be triggered.

Fix warning as reported by 0-DAY CI Kernel:

drivers/gpu/drm/xe/xe_wait_user_fence.c:46:2: warning: variable 'passed' is used uninitialized whenever switch default is taken [-Wsometimes-uninitialized]

Closes: https://lore.kernel.org/oe-kbuild-all/202312170357.KPSinwPs-lkp@intel.com/
Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://lore.kernel.org/r/20231218163301.3453285-1-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix UBSAN splat in add_preempt_fences()

add_preempt_fences() calls dma_resv_reserve_fences() with num_fences ==
0 resulting in the below UBSAN splat. Short circuit add_preempt_fences()
if num_fences == 0.

[   58.652241] ================================================================================
[   58.660736] UBSAN: shift-out-of-bounds in ./include/linux/log2.h:57:13
[   58.667281] shift exponent 64 is too large for 64-bit type 'long unsigned int'
[   58.674539] CPU: 2 PID: 1170 Comm: xe_gpgpu_fill Not tainted 6.6.0-rc3-guc+ #630
[   58.674545] Hardware name: Intel Corporation Tiger Lake Client Platform/TigerLake U DDR4 SODIMM RVP, BIOS TGLSFWI1.R00.3243.A01.2006102133 06/10/2020
[   58.674547] Call Trace:
[   58.674548]  <TASK>
[   58.674550]  dump_stack_lvl+0x92/0xb0
[   58.674555]  __ubsan_handle_shift_out_of_bounds+0x15a/0x300
[   58.674559]  ? rcu_is_watching+0x12/0x60
[   58.674564]  ? software_resume+0x141/0x210
[   58.674575]  ? new_vma+0x44b/0x600 [xe]
[   58.674606]  dma_resv_reserve_fences.cold+0x40/0x66
[   58.674612]  new_vma+0x4b3/0x600 [xe]
[   58.674638]  xe_vm_bind_ioctl+0xffd/0x1e00 [xe]
[   58.674663]  ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe]
[   58.674680]  drm_ioctl_kernel+0xc1/0x170
[   58.674686]  ? __pfx_xe_vm_bind_ioctl+0x10/0x10 [xe]
[   58.674703]  drm_ioctl+0x247/0x4c0
[   58.674709]  ? find_held_lock+0x2b/0x80
[   58.674716]  __x64_sys_ioctl+0x8c/0xb0
[   58.674720]  do_syscall_64+0x3c/0x90
[   58.674723]  entry_SYSCALL_64_after_hwframe+0x6e/0xd8
[   58.674727] RIP: 0033:0x7fce4bd1aaff
[   58.674730] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00
[   58.674731] RSP: 002b:00007ffc57434050 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[   58.674734] RAX: ffffffffffffffda RBX: 00007ffc574340e0 RCX: 00007fce4bd1aaff
[   58.674736] RDX: 00007ffc574340e0 RSI: 0000000040886445 RDI: 0000000000000003
[   58.674737] RBP: 0000000040886445 R08: 0000000000000002 R09: 00007ffc574341b0
[   58.674739] R10: 000055de43eb3780 R11: 0000000000000246 R12: 00007ffc574340e0
[   58.674740] R13: 0000000000000003 R14: 00007ffc574341b0 R15: 0000000000000001
[   58.674747]  </TASK>
[   58.674748] ================================================================================

Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://lore.kernel.org/r/20231215230203.719244-1-matthew.brost@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Remove uninitialized variable from warning

"err" is not initialized when failing to create and add the freq0 sysfs
file. Remove it from the message. This fixes the following warning with
clang:

../drivers/gpu/drm/xe/xe_gt_freq.c:202:30: error: variable 'err' is uninitialized when used here [-Werror,-Wuninitialized]
 kobject_name(gt->sysfs), err);
  ^~~

Fixes: bef52b5c7a19 ("drm/xe: Create a xe_gt_freq component for raw management and sysfs")
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://lore.kernel.org/r/20231220161923.3740489-1-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Remove ci-only GuC FW definitions

As part of the FW definitions, we declare each blob as required via the
MODULE_FIRMWARE() macro. This causes the initramfs update (or equivalent
process) to look for the blobs on disk when the kernel is installed;
therefore, we need to make sure that all FWs we define are available in
linux-firmware.

We currently don't plan to push the PVC blob to linux-firmware, while the
LNL one will only be pushed once we have machines in CI to test it, so we
need to remove them from the list for now.

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/xe: Fix build without CONFIG_FAULT_INJECTION

Ideally this header could be included without the CONFIG_FAULT_INJECTION
and it would take care itself for the includes it needs.
So, let's temporary workaround this by moving this below and including
only when CONFIG_FAULT_INJECTION is selected to avoid build breakages.

Another solution would be us including the linux/types.h as well, but
this creates unnecessary cases.

Reference: https://lore.kernel.org/all/20230816134748.979231-1-himal.prasad.ghimiray@intel.com/
Cc: Himal Prasad Ghimiray <himal.prasad.ghimiray@intel.com>
Cc: Oded Gabbay <ogabbay@kernel.org>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>

MAINTAINERS: Updates to Intel DRM

Introduce the Maintainers of the new drm/xe driver for upcoming
Intel GPUs.

Since it has a shared display with drm/i915, let's also create a
dedicated block to group display related files. But without any
substantial change to the i915 side. The display patches will
continue to flow through i915 from drm-intel-next branches for now.

Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Acked-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Oded Gabbay <ogabbay@kernel.org>
Acked-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>