After successful call to drm_suballoc_manager_init() we should
make sure to call drm_suballoc_manager_fini() as it may include
some cleanup code even if we didn't start using it for real.
As we can abort init() early due to kvzalloc() failure, we should
either explicitly call drm_suballoc_manager_fini() or, even better,
postpone drm_suballoc_manager_init() once we finish all other
preparation steps, so we can rely on fini() that will do cleanup.
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/20241220194205.995-2-michal.wajdeczko@intel.com
VFs need to communicate with the GuC to obtain the GMDID value
and existing GuC functions used for that assume that the GT has
it's MMIO members already setup. However, due to recent refactoring
the gt->mmio is initialized later, and any attempt by the VF to use
xe_mmio_read|write() from GuC functions will lead to NPD crash due
to unset MMIO register address:
[] xe 0000:00:02.1: [drm] Running in SR-IOV VF mode
[] xe 0000:00:02.1: [drm] GT0: sending H2G MMIO 0x5507
[] BUG: unable to handle page fault for address: 0000000000190240
Since we are already tweaking the id and type of the primary GT to
mimic it's a Media GT before initializing the GuC communication,
we can also call xe_gt_mmio_init() to perform early setup of the
gt->mmio which will make those GuC functions work again.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Piotr Piórkowski <piotr.piorkowski@intel.com>
Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250114211347.1083-1-michal.wajdeczko@intel.com
In order to avoid having userspace to use MI_MEM_FENCE,
we are adding a mechanism for userspace to generate a
PCI memory barrier with low overhead (avoiding IOCTL call
as well as writing to VRAM will adds some overhead).
This is implemented by memory-mapping a page as uncached
that is backed by MMIO on the dGPU and thus allowing userspace
to do memory write to the page without invoking an IOCTL.
We are selecting the MMIO so that it is not accessible from
the PCI bus so that the MMIO writes themselves are ignored,
but the PCI memory barrier will still take action as the MMIO
filtering will happen after the memory barrier effect.
When we detect special defined offset in mmap(), We are mapping
4K page which contains the last of page of doorbell MMIO range
to userspace for same purpose.
For user to query special offset we are adding special flag in
mmap_offset ioctl which needs to be passed as follows,
struct drm_xe_gem_mmap_offset mmo = {
.handle = 0, /* this must be 0 */
.flags = DRM_XE_MMAP_OFFSET_FLAG_PCI_BARRIER,
};
igt_ioctl(fd, DRM_IOCTL_XE_GEM_MMAP_OFFSET, &mmo);
map = mmap(NULL, size, PROT_WRITE, MAP_SHARED, fd, mmo);
IGT : b2dbc6f228
UMD : https://github.com/intel/compute-runtime/pull/772
V7:
- Dgpu filter added
V6(MAuld)
- Move physical mmap to fault handler
- Modify kernel-doc and attach UMD PR when ready
V5(MAuld)
- Return invalid early in case of non 4K PAGE_SIZE
- Format kernel-doc and add note for 4K PAGE_SIZE HW limit
V4(MAuld)
- Add kernel-doc for uapi change
- Restrict page size to 4K
V3(MAuld)
- Remove offset defination from UAPI to be able to change later
- Edit commit message for special flag addition
V2(MAuld)
- Add fault handler with dummy page to handle unplug device
- Add Build check for special offset to be below normal start page
- Test d3hot, mapping seems to be valid in d3hot as well
- Add more info to commit message
Cc: Matthew Auld <matthew.auld@intel.com>
Acked-by: Michal Mrozek <michal.mrozek@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Tejas Upadhyay <tejas.upadhyay@intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250113114201.3178806-1-tejas.upadhyay@intel.com
The GuC log snapshot code would complain loudly if there was no GuC
log to take a snapshot of or if the snapshot alloc failed. Originally,
this code was only called on demand when a user (or developer)
explicitly requested a dump of the log. Hence an error message was
useful.
However, it is now part of the general devcoredump file and is called
for any GPU hang. Most people don't care about GuC logs and GPU hangs
do not generally mean a kernel/GuC bug. More importantly, there are
valid situations where there is no GuC log, e.g. SRIOV VFs.
So drop the error message.
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3958
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250113194405.2033085-1-John.C.Harrison@Intel.com
Enable DCC (Duty Cycle Control) in Lunar Lake.
DCC is the SLPC task that tries to keep
the GT from operating inefficiently when thermally constrained.
Although the recommendation is to enable it, LNL GuC is leaving
it disabled by default on LNL.
It would minimize the GT frequency oscillation on throttled
scenarios, which could potentially reduce latencies.
v2: Move set_policies call after wait for running state, so
we ensure it is not overwritten. (Vinay)
v3: Fix English in the commit message (Jonathan)
v4: Also set disable to 0 so DCC can really get into effect.
v5: Avoid lnl_ prefix (Vinay)
v6: Finish renaming...
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com> #v3
Link: https://patchwork.freedesktop.org/patch/msgid/20250115145053.1142023-2-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
This is a follow up fix for
https://patchwork.freedesktop.org/patch/msgid/20241203021929.1919730-1-oak.zeng@intel.com
The overall goal is to fail vm_bind when there is memory pressure. See more
details in the commit message of above patch. Abbove patch fixes the issue
when user pass in a vm_id parameter during gem_create. If user doesn't pass
in a vm_id during gem_create, above patch doesn't help.
This patch further reject BO eviction (which could be triggered by bo validation)
if BO is bound to the current VM. vm_bind could fail due to the eviction failure.
The BO to VM reverse mapping structure is used to determine whether BO is bound
to VM.
v2:
Move vm_bo definition from function scope to if(evict) clause (Thomas)
Further constraint the condition by adding ctx->resv (Thomas)
Add a short comment describe the change.
Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Oak Zeng <oak.zeng@intel.com>
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/20250110210137.3181576-1-oak.zeng@intel.com
The "mmio_ext" and 'REG_EXT" code is currently unused on any existing
platform. Going forward, this also isn't the design we want to use for
any future platforms/features either, so we should just go ahead and
remove the dead code to avoid confusion.
mmio_ext was originally added in an attempt to hack around the early
(mis)design of the Xe driver, which used xe_gt as the target for all
register MMIO access, even those completely unrelated to the GT subunit
of the hardware. With the introduction of commit 34953ee349 ("drm/xe:
Create dedicated xe_mmio structure") and its follow-up patches, that
misdesign has been corrected and access to register MMIO regions
specific to hardware units is now done through xe_mmio structures which
encapsulate an iomap, region size, and some other metadata.
Although all of the registers used by the driver today happen to fall
within one specific PCI BAR region, and thus re-use a single device-wide
iomap, there's no requirement that this stay true for future platforms
or features. I.e., if a future platform adds a new 'foo' hardware unit
that exists at a different area in the BAR, or even in a completely
different BAR, then that would be handled by doing a separate iomap of
that unit's register region and wrapping it in its own 'struct xe_mmio
foo_regs' structure. The pointer to the new 'foo_regs' could be placed
within the xe_device, xe_tile, xe_gt, etc., according to where the new
hardware unit falls within the current hardware hierarchy.
This effectively reverts the following commits, although parts of these
commits had already vanished or changed with the earlier xe_mmio
refactor work:
- commit 399a13323f ("drm/xe: add 28-bit address support in struct
xe_reg")
- commit fdef72e02e ("drm/xe: add a flag to bypass multi-tile config
from MTCFG reg")
- commit 866b2b1764 ("drm/xe: add MMIO extension support flags")
- commit ef29b390c7 ("drm/xe: map MMIO BAR according to the num of
tiles in device desc")
- commit a4e2f3a299 ("drm/xe: refactor xe_mmio_probe_tiles to support
MMIO extension")
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Koby Elbaz <kelbaz@habana.ai>
Acked-by: Maciej Patelczyk <maciej.patelczyk@intel.com>
Reviewed-by: Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Stuart Summers <stuart.summers@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250106234312.2986065-2-matthew.d.roper@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
If GuC fails to load, the driver wedges, but in the process it tries to
do stuff that may not be initialized yet. This moves the
xe_gt_tlb_invalidation_init() to be done earlier: as its own doc says,
it's a software-only initialization and should had been named with the
_early() suffix.
Move it to be called by xe_gt_init_early(), so the locks and seqno are
initialized, avoiding a NULL ptr deref when wedging:
xe 0000:03:00.0: [drm] *ERROR* GT0: load failed: status: Reset = 0, BootROM = 0x50, UKernel = 0x00, MIA = 0x00, Auth = 0x01
xe 0000:03:00.0: [drm] *ERROR* GT0: firmware signature verification failed
xe 0000:03:00.0: [drm] *ERROR* CRITICAL: Xe has declared device 0000:03:00.0 as wedged.
...
BUG: kernel NULL pointer dereference, address: 0000000000000000
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: Oops: 0000 [#1] PREEMPT SMP NOPTI
CPU: 9 UID: 0 PID: 3908 Comm: modprobe Tainted: G U W 6.13.0-rc4-xe+ #3
Tainted: [U]=USER, [W]=WARN
Hardware name: Intel Corporation Alder Lake Client Platform/AlderLake-S ADP-S DDR5 UDIMM CRB, BIOS ADLSFWI1.R00.3275.A00.2207010640 07/01/2022
RIP: 0010:xe_gt_tlb_invalidation_reset+0x75/0x110 [xe]
This can be easily triggered by poking the GuC binary to force a
signature failure. There will still be an extra message,
xe 0000:03:00.0: [drm] *ERROR* GT0: GuC mmio request 0x4100: no reply 0x4100
but that's better than a NULL ptr deref.
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3956
Fixes: 7dbe8af13c ("drm/xe: Wedge the entire device")
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250103001111.331684-2-lucas.demarchi@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
This is a set of squashed commits to facilitate smooth applying to
stable. Each commit message is retained for reference.
1) Allow a GGTT mapped batch to be submitted to user exec queue
For a OA use case, one of the HW registers needs to be modified by
submitting an MI_LOAD_REGISTER_IMM command to the users exec queue, so
that the register is modified in the user's hardware context. In order
to do this a batch that is mapped in GGTT, needs to be submitted to the
user exec queue. Since all user submissions use q->vm and hence PPGTT,
add some plumbing to enable submission of batches mapped in GGTT.
v2: ggtt is zero-initialized, so no need to set it false (Matt Brost)
2) xe/oa: Use MI_LOAD_REGISTER_IMMEDIATE to enable OAR/OAC
To enable OAR/OAC, a bit in RING_CONTEXT_CONTROL needs to be set.
Setting this bit cause the context image size to change and if not done
correct, can cause undesired hangs.
Current code uses a separate exec_queue to modify this bit and is
error-prone. As per HW recommendation, submit MI_LOAD_REGISTER_IMM to
the target hardware context to modify the relevant bit.
In v2 version, an attempt to submit everything to the user-queue was
made, but it failed the unprivileged-single-ctx-counters test. It
appears that the OACTXCONTROL must be modified from a remote context.
In v3 version, all context specific register configurations were moved
to use LOAD_REGISTER_IMMEDIATE and that seems to work well. This is a
cleaner way, since we can now submit all configuration to user
exec_queue and the fence handling is simplified.
v2:
(Matt)
- set job->ggtt to true if create job is successful
- unlock vm on job error
(Ashutosh)
- don't wait on job submission
- use kernel exec queue where possible
v3:
(Ashutosh)
- Fix checkpatch issues
- Remove extra spaces/new-lines
- Add Fixes: and Cc: tags
- Reset context control bit when OA stream is closed
- Submit all config via MI_LOAD_REGISTER_IMMEDIATE
(Umesh)
- Update commit message for v3 experiment
- Squash patches for easier port to stable
v4:
(Ashutosh)
- No need to pass q to xe_oa_submit_bb
- Do not support exec queues with width > 1
- Fix disabling of CTX_CTRL_OAC_CONTEXT_ENABLE
v5:
(Ashutosh)
- Drop reg_lri related comments
- Use XE_OA_SUBMIT_NO_DEPS in xe_oa_load_with_lri
Fixes: 8135f1c09d ("drm/xe/oa: Don't reset OAC_CONTEXT_ENABLE on OA stream close")
Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com> # commit 1
Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241220171919.571528-2-umesh.nerlige.ramappa@intel.com
Fix the false-positive "Missing outer runtime PM protection" warning
triggered by
release_async_domains() -> intel_runtime_pm_get_noresume() ->
xe_pm_runtime_get_noresume()
during system suspend.
xe_pm_runtime_get_noresume() is supposed to warn if the device is not in
the runtime resumed state, using xe_pm_runtime_get_if_in_use() for this.
However the latter function will fail if called during runtime or system
suspend/resume, regardless of whether the device is runtime resumed or
not.
Based on the above suppress the warning during system suspend/resume,
similarly to how this is done during runtime suspend/resume.
Suggested-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241217230547.1667561-1-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
tbt-alt mode is missing uhbr rates 10G and 20G. This requires
requires pll clock rates 312.5 MHz and 625 MHz to be added,
respectively. The uhbr rates are supported only form PTL+
platforms.
v2: Add drm_WARN_ON() to check if port clock is not supported by
the platform (Imre)
Combine forward ungate with mask parameter (Imre)
Rename XE3LPDP_* to XE3D_* (Imre)
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241217143440.572308-1-mika.kahola@intel.com
The DUAL_QUEUE_WA tells the GuC to not allow concurrent submissions
on RCS and CCSes with different address spaces, which on DG2 is
required as a WA for an HW bug. On newer platforms, this block has
been moved in HW at the CS level, by stalling the RCS/CCS context
switch when one of the other RCS/CCSes is busy with a different
address space. While functionally correct, having a submission
stalled on the HW limits the GuC ability to shuffle things around and
can cause complications if the non-stalled submission runs for a long
time, because the GuC doesn't know that the stalled submission isn't
actually running and might declare it as hung. Therefore, we enable
the DUAL_QUEUE_WA on all newer platforms to move management back to
the GuC.
Note that the GuC specs also recommend enabling this for all platforms
starting from MTL that have a CCS.
v2: only apply the WA on GTs that have CCS engines
v3: split comment (Jonathan)
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Cc: Jesus Narvaez <jesus.narvaez@intel.com>
Cc: Jonathan Cavitt <jonathan.cavitt@intel.com>
Reviewed-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241213181012.2178794-1-daniele.ceraolospurio@intel.com
