Nirmoy Das [Tue, 22 Oct 2024 10:35:55 +0000 (12:35 +0200)]
drm/xe: Don't restart parallel queues multiple times on GT reset
In case of parallel submissions multiple GuC id will point to the
same exec queue and on GT reset such exec queues will get restarted
multiple times which is not desirable.
v2: don't use exec_queue_enabled() which could race,
do the same for xe_guc_submit_stop (Matt B)
Michal Wajdeczko [Mon, 21 Oct 2024 20:15:06 +0000 (22:15 +0200)]
drm/xe/pf: Show VFs LMEM provisioning summary over debugfs
While we can already view individual VF LMEM provisioning using
lmem_quota debugfs attribute, we want to have unified way to show
summary across all VFs, like we do for GGTT or GuC doorbells/IDs.
Fei Yang [Thu, 17 Oct 2024 16:27:10 +0000 (09:27 -0700)]
drm/xe: enable lite restore
The lite restore is a performance improvement feature which avoids
unnecessary context switch (flush, save and restore) if the incoming
context has a ContextID matching that of the outgoing context. The
scheduling is done by the GuC firmware, so on the driver side it's
just a matter of setting corresponding GUC_CTL_FEATURE flag.
This is supposed to be enabled by default, thus the flag is set
unconditionally.
Shuicheng Lin [Thu, 17 Oct 2024 22:15:47 +0000 (22:15 +0000)]
drm/xe: Handle unreliable MMIO reads during forcewake
In some cases, when the driver attempts to read an MMIO register,
the hardware may return 0xFFFFFFFF. The current force wake path
code treats this as a valid response, as it only checks the BIT.
However, 0xFFFFFFFF should be considered an invalid value, indicating
a potential issue. To address this, we should add a log entry to
highlight this condition and return failure.
The force wake failure log level is changed from notice to err
to match the failure return value.
v2 (Matt Brost):
- set ret value (-EIO) to kick the error to upper layers
v3 (Rodrigo):
- add commit message for the log level promotion from notice to err
v4:
- update reviewed info
Badal Nilawar [Thu, 17 Oct 2024 11:14:10 +0000 (16:44 +0530)]
drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout
In case if g2h worker doesn't get opportunity to within specified
timeout delay then flush the g2h worker explicitly.
v2:
- Describe change in the comment and add TODO (Matt B/John H)
- Add xe_gt_warn on fence done after G2H flush (John H)
v3:
- Updated the comment with root cause
- Clean up xe_gt_warn message (John H)
drm/xe: forcewake debugfs open fails on xe_forcewake_get failure
A failure in xe_force_wake_get() no longer increments the domain's
refcount. Therefore, if xe_force_wake_get() fails during forcewake
debugfs open, return an error. This ensures there are no valid file
descriptors to close via forcewake debugfs, preventing refcount
mismanagement.
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
v5
- return unsigned int from xe_force_wake_get()
v6
- Use helper xe_force_wake_ref_has_domain()
to determine the status of the call.
drm/xe/vram: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be escalated/considered as
probing error. It internally WARNS on domain ack failure.
drm/xe/query: Update handling of xe_force_wake_get return
With xe_force_wake_get() now returning the refcount-incremented
domain mask, a non-zero return value in the case of XE_FORCEWAKE_ALL
does not necessarily indicate success. Use xe_force_wake_ref_has_domain()
to determine the status of the call.
Modify the return handling of xe_force_wake_get() accordingly and
pass the return value to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
v5
- return unsigned int from xe_force_wake_get()
v6
- Use helper Use xe_force_wake_ref_has_domain()
drm/xe/xe_reg_sr: Update handling of xe_force_wake_get return
With xe_force_wake_get() now returning the refcount-incremented
domain mask, a non-zero return value in the case of XE_FORCEWAKE_ALL does
not necessarily indicate success. Use xe_force_wake_ref_has_domain()
to determine the status of the call.
Modify the return handling of xe_force_wake_get() accordingly and
pass the return value to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
drm/xe/gt_tlb_invalidation_ggtt: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
v5
- return unsigned int from xe_force_wake_get()
- remove redundant warns
drm/xe/pat: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
- don't use xe_assert() to report HW errors (Michal)
v5
- return unsigned int from xe_force_wake_get()
- remove redundant warns
drm/xe/oa: Handle force_wake_get failure in xe_oa_stream_init()
With xe_force_wake_get() now returning the refcount-incremented domain
mask, a non-zero return value in the case of XE_FORCEWAKE_ALL does not
necessarily indicate success. use xe_force_wake_ref_has_domain ()
to determine the status of the call.
Modify the return handling of xe_force_wake_get() accordingly and pass
the return value to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
drm/xe/huc: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
drm/xe/guc: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Use helper xe_force_wake_ref_has_domain to
verify all domains are initialized or not. Update the return handling of
xe_force_wake_get() to reflect this behavior, and ensure that the return
value is passed as input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
v5
- return unsigned int from xe_force_wake_get()
- Remove redundant xe_gt_WARN_ON
drm/xe/xe_gt_debugfs: Update handling of xe_force_wake_get return
With xe_force_wake_get() now returning the refcount-incremented
domain mask, a non-zero return value in the case of XE_FORCEWAKE_ALL does
not necessarily indicate success. Use xe_force_wake_ref_has_domain()
determine the status of the call.
Modify the return handling of xe_force_wake_get() accordingly and
pass the return value to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
drm/xe/xe_drm_client: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Use helper xe_force_wake_ref_has_domain to
verify all domains are initialized or not. Update the return handling of
xe_force_wake_get() to reflect this behavior, and ensure that the return
value is passed as input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
drm/xe/mocs: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- don't use xe_assert() to report HW errors (Michal)
v5
- return unsigned int from xe_force_wake_get()
- Remove redundant warn
With xe_force_wake_get() now returning the refcount-incremented domain
mask, a return value of 0 indicates failure for single domains.
Change assert condition to incorporate this change in return and
pass the return value to xe_force_wake_put()
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
drm/xe/devcoredump: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Use helper xe_force_wake_ref_has_domain to
verify all domains are initialized or not. Update the return handling of
xe_force_wake_get() to reflect this behavior, and ensure that the return
value is passed as input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
drm/xe/xe_gt_idle: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
v4
- Rebase fix
v5
- return unsigned int for xe_force_wake_get()
- Remove reudandant WARN calls.
drm/xe/gt: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Use helper xe_force_wake_ref_has_domain to verify
all domains are initialized or not. Update the return handling of
xe_force_wake_get() to reflect this behavior, and ensure that the return
value is passed as input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be checked. It internally
WARNS on domain ack failure.
v4
- Rebase fix
v5
- return unsigned int for xe_force_wake_get()
- remove redundant XE_WARN_ON()
v6
- use helper for checking all initialized domains are awake or not.
drm/xe/gsc: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
v5
- return unsigned int for xe_force_wake_get()
- No need to WARN from caller in case of forcewake get failure.
drm/xe/hdcp: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
drm/xe/device: Update handling of xe_force_wake_get return
xe_force_wake_get() now returns the reference count-incremented domain
mask. If it fails for individual domains, the return value will always
be 0. However, for XE_FORCEWAKE_ALL, it may return a non-zero value even
in the event of failure. Update the return handling of xe_force_wake_get()
to reflect this behavior, and ensure that the return value is passed as
input to xe_force_wake_put().
v3
- return xe_wakeref_t instead of int in xe_force_wake_get()
- xe_force_wake_put() error doesn't need to be escalated/considered as
probing error. It internally WARNS on domain ack failure.
drm/xe: Modify xe_force_wake_put to handle _get returned mask
Instead of calling xe_force_wake_put on all domains that were input to
xe_force_wake_get, call _put only on the domains whose reference counts
were successfully incremented by the _get call. Since the return value
of _get can be a mask that does not match any specific value in the enum
xe_force_wake_domains, change the input parameter of _put to unsigned int.
v3
- Move WARN to this patch (Badal)
- use xe_gt_WARN instead of XE_WARN (Michal)
- Stop using xe_force_wake_domains for non enum values.
- Remove kernel-doc from this patch (Badal)
-v5
- Fix global awake_domain
-v6
- put all initialized domains in case of FORCEWAKE_ALL.
- Modify ret variable name (Michal)
- Modify input var name (Michal)
- Modify commit message and warn (Badal)
If an acknowledgment timeout occurs for a forcewake domain awake
request, do not increment the reference count for the domain. This
ensures that subsequent _get calls do not incorrectly assume the domain
is awake. The return value is a mask of domains that got refcounted,
and these domains need to be provided for subsequent xe_force_wake_put
call.
While at it, add simple kernel-doc for xe_force_wake_get()
v3
- Use explicit type for mask (Michal/Badal)
- Improve kernel-doc (Michal)
- Use unsigned int instead of abusing enum (Michal)
v5
- Use unsigned int for return (MattB/Badal/Rodrigo)
- use xe_gt_WARN for domain awake ack failure (Badal/Rodrigo)
v6
- Change XE_FORCEWAKE_ALL to single bit, this helps accommodate
actually refcounted domains in return. (Michal)
- Modify commit message and warn message (Badal)
- Remove unnecessary information in kernel-doc (Michal)
v7
- Add assert condition for valid input domains (Badal)
v9
- Update kernel-doc and simplify conditions (Michal)
Shuicheng Lin [Tue, 15 Oct 2024 16:12:07 +0000 (16:12 +0000)]
drm/xe: Enlarge the invalidation timeout from 150 to 500
There are error messages like below that are occurring during stress
testing: "[ 31.004009] xe 0000:03:00.0: [drm] ERROR GT0: Global
invalidation timeout". Previously it was hitting this 3 out of 1000
executions of warm reboot. After raising it to 500, 1000 warm reboot
executions passed and it didn't fail.
Due to the way xe_mmio_wait32() is implemented, the timeout is able to
expire early when the register matches the expected value due to the
wait increments starting small. So, the larger timeout value should have
no effect during normal use cases.
v2 (Jonathan):
- rework the commit message
v3 (Lucas):
- add conclusive message for the fail rate and test case
v4:
- add suggested-by
drm/xe: add interface to request physical alignment for buffer objects
Add xe_bo_create_pin_map_at_aligned() which augment
xe_bo_create_pin_map_at() with alignment parameter allowing to pass
required alignemnt if it differ from default.
Matthew Auld [Fri, 11 Oct 2024 13:36:34 +0000 (14:36 +0100)]
drm/xe/xe_sync: initialise ufence.signalled
We can incorrectly think that the fence has signalled, if we get a
non-zero value here from the kmalloc, which is quite plausible. Just use
kzalloc to prevent stuff like this.
Nirmoy Das [Fri, 11 Oct 2024 15:10:29 +0000 (17:10 +0200)]
drm/xe/ufence: ufence can be signaled right after wait_woken
do_comapre() can return success after a timedout wait_woken() which was
treated as -ETIME. The loop calling wait_woken() sets correct err so
there is no need to re-evaluate err.
v2: Remove entire check that reevaluate err at the end(Matt)
Lucas De Marchi [Fri, 11 Oct 2024 03:56:17 +0000 (20:56 -0700)]
drm/xe/query: Move timestamp reg to hwe_read_timestamp()
__read_timestamps() is actually reading the timestamp from a certain
hwe. Use it as parameter, move register declarations to be inside that
function and rename it.
Lucas De Marchi [Fri, 11 Oct 2024 03:56:16 +0000 (20:56 -0700)]
drm/xe/query: Increase timestamp width
Starting with Xe2 the timestamp is a full 64 bit counter, contrary to
the 36 bit that was available before. Although 36 should be sufficient
for any reasonable delta calculation (for Xe2, of about 30min), it's
surprising to userspace to get something truncated. Also if the
timestamp being compared to is coming from the GPU and the application
is not careful enough to apply the width there, a delta calculation
would be wrong.
Extend it to full 64-bits starting with Xe2.
v2: Expand width=64 to media gt, as it's just a wrong tagging in the
spec - empirical tests show it goes beyond 36 bits and match the engines
for the main gt
Matthew Brost [Wed, 11 Sep 2024 15:26:22 +0000 (08:26 -0700)]
drm/xe: Use bookkeep slots for external BO's in exec IOCTL
Fix external BO's dma-resv usage in exec IOCTL using bookkeep slots
rather than write slots. This leaves syncing to user space rather than
the KMD blindly enforcing write semantics on every external BO.
Matthew Brost [Thu, 3 Oct 2024 00:16:57 +0000 (17:16 -0700)]
drm/xe: Don't free job in TDR
Freeing job in TDR is not safe as TDR can pass the run_job thread
resulting in UAF. It is only safe for free job to naturally be called by
the scheduler. Rather free job in TDR, add to pending list.
Matthew Brost [Thu, 3 Oct 2024 00:16:56 +0000 (17:16 -0700)]
drm/xe: Take job list lock in xe_sched_add_pending_job
A fragile micro optimization in xe_sched_add_pending_job relied on both
the GPU scheduler being stopped and fence signaling stopped to safely
add a job to the pending list without the job list lock in
xe_sched_add_pending_job. Remove this optimization and just take the job
list lock.
Imre Deak [Wed, 9 Oct 2024 19:43:58 +0000 (22:43 +0300)]
drm/xe/display: Add missing HPD interrupt enabling during non-d3cold RPM resume
Atm the display HPD interrupts that got disabled during runtime
suspend, are re-enabled only if d3cold is enabled. Fix things by
also re-enabling the interrupts if d3cold is disabled.
Imre Deak [Wed, 9 Oct 2024 19:43:57 +0000 (22:43 +0300)]
drm/xe/display: Separate the d3cold and non-d3cold runtime PM handling
For clarity separate the d3cold and non-d3cold runtime PM handling. The
only change in behavior is disabling polling later during runtime
resume. This shouldn't make a difference, since the poll disabling is
handled from a work, which could run at any point wrt. the runtime
resume handler. The work will also require a runtime PM reference,
syncing it with the resume handler.
Colin Ian King [Wed, 9 Oct 2024 16:05:10 +0000 (17:05 +0100)]
drm/xe/guc: Fix inverted logic on snapshot->copy check
Currently the check to see if snapshot->copy has been allocated is
inverted and ends up dereferencing snapshot->copy when free'ing
objects in the array when it is null or not free'ing the objects
when snapshot->copy is allocated. Fix this by using the correct
non-null pointer check logic.
Matthew Auld [Wed, 9 Oct 2024 08:48:10 +0000 (09:48 +0100)]
drm/xe: fix unbalanced rpm put() with declare_wedged()
Technically the or_reset() means we call the action on failure, however
that would lead to unbalanced rpm put(). Move the get() earlier to fix
this. It should be extremely unlikely to ever trigger this in practice.
Matthew Auld [Wed, 9 Oct 2024 08:48:09 +0000 (09:48 +0100)]
drm/xe: fix unbalanced rpm put() with fence_fini()
Currently we can call fence_fini() twice if something goes wrong when
sending the GuC CT for the tlb request, since we signal the fence and
return an error, leading to the caller also calling fini() on the error
path in the case of stack version of the flow, which leads to an extra
rpm put() which might later cause device to enter suspend when it
shouldn't. It looks like we can just drop the fini() call since the
fence signaller side will already call this for us.
There are known mysterious splats with device going to sleep even with
an rpm ref, and this could be one candidate.
v2 (Matt B):
- Prefer warning if we detect double fini()
Matthew Auld [Mon, 7 Oct 2024 07:45:42 +0000 (08:45 +0100)]
drm/xe/bmg: improve cache flushing behaviour
The BSpec says that EN_L3_RW_CCS_CACHE_FLUSH must be toggled
on for manual global invalidation to take effect and actually flush
device cache, however this also turns on flushing for things like
pipecontrol, which occurs between submissions for compute/render. This
sounds like massive overkill for our needs, where we already have the
manual flushing on the display side with the global invalidation. Some
observations on BMG:
1. Disabling l2 caching for host writes and stubbing out the driver
global invalidation but keeping EN_L3_RW_CCS_CACHE_FLUSH enabled, has
no impact on wb-transient-vs-display IGT, which makes sense since the
pipecontrol is now flushing the device cache after the render copy.
Without EN_L3_RW_CCS_CACHE_FLUSH the test then fails, which is also
expected since device cache is now dirty and display engine can't see
the writes.
2. Disabling EN_L3_RW_CCS_CACHE_FLUSH, but keeping the driver global
invalidation also has no impact on wb-transient-vs-display. This
suggests that the global invalidation still works as expected and is
flushing the device cache without EN_L3_RW_CCS_CACHE_FLUSH turned on.
With that drop EN_L3_RW_CCS_CACHE_FLUSH. This helps some workloads since
we no longer flush the device cache between submissions as part of
pipecontrol.
Edit: We now also have clarification from HW side that BSpec was indeed
wrong here.
Zhanjun Dong [Fri, 4 Oct 2024 19:34:27 +0000 (12:34 -0700)]
drm/xe/guc: Plumb GuC-capture into dev coredump
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.
Zhanjun Dong [Fri, 4 Oct 2024 19:34:26 +0000 (12:34 -0700)]
drm/xe/guc: Extract GuC error capture lists
Upon the G2H Notify-Err-Capture event, parse through the
GuC Log Buffer (error-capture-subregion) and generate one or
more capture-nodes. A single node represents a single "engine-
instance-capture-dump" and contains at least 3 register lists:
global, engine-class and engine-instance. An internal link
list is maintained to store one or more nodes.
Because the link-list node generation happen before the call
to devcoredump, duplicate global and engine-class register
lists for each engine-instance register dump if we find
dependent-engine resets in a engine-capture-group.
To avoid dynamically allocate the output nodes during gt reset,
pre-allocate a fixed number of empty nodes up front (at the
time of ADS registration) that we can consume from or return to
an internal cached list of nodes.
Zhanjun Dong [Fri, 4 Oct 2024 19:34:25 +0000 (12:34 -0700)]
drm/xe/guc: Add capture size check in GuC log buffer
Capture-nodes generated by GuC are placed in the GuC capture ring
buffer which is a sub-region of the larger Guc-Log-buffer.
Add capture output size check before allocating the shared buffer.
Zhanjun Dong [Fri, 4 Oct 2024 19:34:23 +0000 (12:34 -0700)]
drm/xe/guc: Prepare GuC register list and update ADS size for error capture
Add referenced registers defines and list of registers.
Update GuC ADS size allocation to include space for
the lists of error state capture register descriptors.
Then, populate GuC ADS with the lists of registers we want
GuC to report back to host on engine reset events. This list
should include global, engine-class and engine-instance
registers for every engine-class type on the current hardware.
Ensure we allocate a persistent storage for the register lists
that are populated into ADS so that we don't need to allocate
memory during GT resets when GuC is reloaded and ADS population
happens again.
Matt Roper [Tue, 8 Oct 2024 01:35:09 +0000 (18:35 -0700)]
drm/xe/xe3lpm: Add new "instance0" steering table
MCR steering on Xe3 media IP is almost the same as it was on Xe2, except
for one new range (0x38D0D0 - 0x38D0FF) which has changed to an MCR
"MEDIAINF" range on Xe3. Since we can always steer to grpid /
instanceid 0 for MEDIAINF ranges, define a new "INSTANCE0" steering
table for Xe3 media. Xe3 can continue to use the same OADDRM/GPMXMT
table as Xe2.
Shekhar Chauhan [Mon, 7 Oct 2024 15:41:44 +0000 (08:41 -0700)]
drm/xe/ptl: L3bank mask is not available on the media GT
On PTL platforms with media version 30.00, the fuse registers for
reporting L3 bank availability to the GT just read out as ~0 and do not
provide proper values. Xe does not use the L3 bank mask for anything
internally; it only passes the mask through to userspace via the GT
topology query.
Since we don't have any way to get the real L3 bank mask, we don't want
to pass garbage to userspace. Passing a zeroed mask or a copy of the
primary GT's L3 bank mask would also be inaccurate and likely to cause
confusion for userspace. The best approach is to simply not include L3
in the list of masks returned by the topology query in cases where we
aren't able to provide a meaningful value. This won't change the
behavior for any existing platforms (where we can always obtain L3 masks
successfully for all GTs), it will only prevent us from mis-reporting
bad information on upcoming platform(s).
There's a good chance this will become a formal workaround in the
future, but for now we don't have a lineage number so "no_media_l3" is
used in place of a lineage as the OOB workaround descriptor.
v2:
- Re-calculate query size to properly match data returned. (Gustavo)
- Update kerneldoc to clarify that the L3bank mask may not be included
in the query results if the hardware doesn't make it available.
(Gustavo)
John Harrison [Thu, 3 Oct 2024 00:46:11 +0000 (17:46 -0700)]
drm/xe/guc: Add a helper function for dumping GuC log to dmesg
Create a helper function that can be used to dump the GuC log to dmesg
in a manner that is reliable for extraction and decode. The intention
is that calls to this can be added by developers when debugging
specific issues that require a GuC log but do not allow easy capture
of the log - e.g. failures in selftests and failues that lead to
kernel hangs.
Also note that this is really a temporary stop-gap. The aim is to
allow on demand creation and dumping of devcoredump captures (which
includes the GuC log and much more). Currently this is not possible as
much of the devcoredump code requires a 'struct xe_sched_job' and
those are not available at many places that might want to do the dump.
v2: Add kerneldoc - review feedback from Michal W.
John Harrison [Thu, 3 Oct 2024 00:46:10 +0000 (17:46 -0700)]
drm/xe/guc: Add GuC log to devcoredump captures
Include the GuC log in devcoredump captures because they can be useful
with debugging certain types of bug.
v2: Fix kerneldoc
v3: Drop module parameter as now using more compact ascii85 encoding
rather than hexdump (although still not compressed) (review feedback
from Matthew B). Rebase onto recent refactoring of devcoredump code.
v4: Don't move the submission snapshot inside the GuC internals
structure 'cos it really doesn't belong there.
John Harrison [Thu, 3 Oct 2024 00:46:09 +0000 (17:46 -0700)]
drm/xe/guc: Dump entire CTB on errors
The dump of the CT buffers was only showing the unprocessed data which
is not generally useful for saying why a hang occurred - because it
was probably caused by the commands that were just processed. So save
and dump the entire buffer but in a more compact dump format. Also
zero fill it on allocation to avoid confusion over uninitialised data
in the dump.
v2: Add kerneldoc - review feedback from Michal W.
v3: Fix kerneldoc.
v4: Use ascii85 instead of hexdump (review feedback from Matthew B).
v5: Dump the entire CTB object rather than separately dumping just the
H2G and G2H sections. That way it includes the full header info.
John Harrison [Thu, 3 Oct 2024 00:46:08 +0000 (17:46 -0700)]
drm/xe/guc: Dead CT helper
Add a worker function helper for asynchronously dumping state when an
internal/fatal error is detected in CT processing. Being asynchronous
is required to avoid deadlocks and scheduling-while-atomic or
process-stalled-for-too-long issues. Also check for a bunch more error
conditions and improve the handling of some existing checks.
v2: Use compile time CONFIG check for new (but not directly CT_DEAD
related) checks and use unsigned int for a bitmask, rename
CT_DEAD_RESET to CT_DEAD_REARM and add some explaining comments,
rename 'hxg' macro parameter to 'ctb' - review feedback from Michal W.
Drop CT_DEAD_ALIVE as no need for a bitfield define to just set the
entire mask to zero.
v3: Fix kerneldoc
v4: Nullify some floating pointers after free.
v5: Add section headings and device info to make the state dump look
more like a devcoredump to allow parsing by the same tools (eventual
aim is to just call the devcoredump code itself, but that currently
requires an xe_sched_job, which is not available in the CT code).
v6: Fix potential for leaking snapshots with concurrent error
conditions (review feedback from Julia F).
v7: Don't complain about unexpected G2H messages yet because there is
a known issue causing them. Fix bit shift bug with v6 change. Add GT
id to fake coredump headers and use puts instead of printf.
v8: Disable the head mis-match check in g2h_read because it is failing
on various discrete platforms due to unknown reasons.
This drm printer wrapper can be used to increase the robustness of
the captured output generated by any other drm_printer to make sure
we didn't lost any intermediate lines of the output by adding line
numbers to each output line. Helpful for capturing some crash data.
v2: Extended short int counters to full int (JohnH)
John Harrison [Thu, 3 Oct 2024 00:46:06 +0000 (17:46 -0700)]
drm/xe/guc: Use a two stage dump for GuC logs and add more info
Split the GuC log dump into a two stage snapshot and print mechanism.
This allows the log to be captured at the point of an error (which may
be in a restricted context) and then dump it out later (from a regular
context such as a worker function or a sysfs file handler).
Also add a bunch of other useful pieces of information that can help
(or are fundamentally required!) to decode and parse the log.
v2: Add kerneldoc and fix a couple of comment typos - review feedback
from Michal W.
v3: Move chunking code to this patch as it makes the deltas simpler.
Fix a bunch of kerneldoc issues.
v4: Move the CS frequency out of the coredump snapshot function into
the debugfs only code (as that info is already part of the main
devcoredump). Add a header to the debugfs log to match the one in the
devcoredump to aid processing by a unified tool. Add forcewake to the
GuC timestamp read so it actually works.
v6: Add colon to GuC version string (review feedback by Julia F).
John Harrison [Thu, 3 Oct 2024 00:46:05 +0000 (17:46 -0700)]
drm/xe/guc: Copy GuC log prior to dumping
Add an extra stage to the GuC log print to copy the log buffer into
regular host memory first, rather than printing the live GPU buffer
object directly. Doing so helps prevent inconsistencies due to the log
being updated as it is being dumped. It also allows the use of the
ASCII85 helper function for printing the log in a more compact form
than a straight hex dump.
v2: Use %zx instead of %lx for size_t prints.
v3: Replace hexdump code with ascii85 call (review feedback from
Matthew B). Move chunking code into next patch as that reduces the
deltas of both.
v4: Add a prefix to the ASCII85 output to aid tool parsing.
John Harrison [Thu, 3 Oct 2024 00:46:04 +0000 (17:46 -0700)]
drm/xe/devcoredump: Add ASCII85 dump helper function
There is a need to include the GuC log and other large binary objects
in core dumps and via dmesg. So add a helper for dumping to a printer
function via conversion to ASCII85 encoding.
Another issue with dumping such a large buffer is that it can be slow,
especially if dumping to dmesg over a serial port. So add a yield to
prevent the 'task has been stuck for 120s' kernel hang check feature
from firing.
v2: Add a prefix to the output string. Fix memory allocation bug.
v3: Correct a string size calculation and clean up a define (review
feedback from Julia F).
John Harrison [Thu, 3 Oct 2024 00:46:03 +0000 (17:46 -0700)]
drm/xe/devcoredump: Improve section headings and add tile info
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.
John Harrison [Thu, 3 Oct 2024 00:46:02 +0000 (17:46 -0700)]
drm/xe/devcoredump: Use drm_puts and already cached local variables
There are a bunch of calls to drm_printf with static strings. Switch
them to drm_puts instead.
There are also a bunch of 'coredump->snapshot.XXX' references when
'coredump->snapshot' has alread been cached locally as 'ss'. So use
'ss->XXX' instead.
John Harrison [Thu, 3 Oct 2024 00:46:01 +0000 (17:46 -0700)]
drm/xe/guc: Remove spurious line feed in debug print
Including line feeds at the start of a debug print messes up the
output when sent to dmesg. The break appears between all the useful
prefix information and the actual string being printed. In this case,
each block of data has a very clear start line and an extra delimeter
is really not necessary. So don't do it.
v2: Fix typo in commit message (review feedback from Michal W.)
Michal Wajdeczko [Thu, 19 Sep 2024 17:15:28 +0000 (19:15 +0200)]
drm/xe/pf: Allow to save and restore VF config blob from debugfs
For feature enabling and testing purposes, allow to capture and
replace full VF configuration using debugfs blob file, but only
under strict debug config.
Michal Wajdeczko [Thu, 19 Sep 2024 17:15:27 +0000 (19:15 +0200)]
drm/xe/pf: Add functions to save and restore VF configuration blob
We already have support to save and restore GuC VF state, but that
will only work when the target VF configuration (provisioning) will
be exactly the same as the source VF configuration. To help with
assuring that both configurations match, allow to encode whole VF
configuration that can be saved as blob and restored later.
In the future we may want to use such captured configuration blobs
as templates to make sure we provision VFs with exactly the same
configuration that was previously tested or recommended, or when
debugfs knobs are not be available.
Michal Wajdeczko [Thu, 19 Sep 2024 17:15:26 +0000 (19:15 +0200)]
drm/xe/pf: Allow to encode subset of VF configuration KLVs
We want to reuse format of the GuC KLVs while saving and restoring
VF configuration by the PF driver, but some of those KLVs (like
doorbell begin index or GGTT starting offset) are not strictly
needed to correctly restore VF configuration. Modify functions to
omit encoding of some of the KLVs with GuC only details.
Michal Wajdeczko [Thu, 19 Sep 2024 17:15:25 +0000 (19:15 +0200)]
drm/xe/pf: Update success code of pf_validate_vf_config()
This function may return negative error codes on invalid or
incomplete VF configuration, but unlike other int functions,
it was returning 1 instead of 0 on success, which might be
little inconvinient if we would like to use it directly in
other functions.
Michal Wajdeczko [Thu, 19 Sep 2024 17:15:24 +0000 (19:15 +0200)]
drm/xe/guc: Add yet another helper macro for threshold
We already have and use MAKE_GUC_KLV_VF_CFG_THRESHOLD_KEY, but
in upcoming patch we want to generate the key for the length.
Add MAKE_GUC_KLV_VF_CFG_THRESHOLD_LEN for this purpose.
Michal Wajdeczko [Wed, 25 Sep 2024 08:44:45 +0000 (10:44 +0200)]
drm/xe: Add memirq report page address helpers
Both xe_memirq_{source,status}_ptr functions are now strictly for
obtaining an address of the memory based interrupt report pages
used by the HW engines. When initializing the GuC that does not
require special per instance page preparation, we don't need to
abuse these public functions and pass a NULL instead of valid hwe
pointer. Also, without further fixes, this actually may lead to
NPD crash once the hw_reports_to_instance_zero() will be true.
Add internal helpers that will provide report page addresses based
solely on the instance number, which will be always 0 for both GuCs.
Matt Roper [Wed, 2 Oct 2024 23:06:21 +0000 (16:06 -0700)]
drm/xe: Make wedged_mode debugfs writable
The intent of this debugfs entry is to allow modification of wedging
behavior, either from IGT tests or during manual debug; it should be
marked as writable to properly reflect this. In practice this hasn't
caused a problem because we always access wedged_mode as root, which
ignores file permissions, but it's still misleading to have the entry
incorrectly marked as RO.
drm/xe: Use fault injection infrastructure to find issues at probe time
The kernel fault injection infrastructure is used to test proper error
handling during probe. The return code of the functions using
ALLOW_ERROR_INJECTION() can be conditionnally modified at runtime by
tuning some debugfs entries. This requires CONFIG_FUNCTION_ERROR_INJECTION
(among others).
One way to use fault injection at probe time by making each of those
functions fail one at a time is:
FAILTYPE=fail_function
DEVICE="0000:00:08.0" # depends on the system
ERRNO=-12 # -ENOMEM, can depend on the function
Matthew Auld [Tue, 1 Oct 2024 08:43:50 +0000 (09:43 +0100)]
drm/xe/ct: drop irq usage of xa_erase()
Unclear why disabling interrupts is needed here. Nothing seems to be
touching fence_lookup and its corresponding lock from an irq so there
should be no risk of deadlock.
Matthew Auld [Tue, 1 Oct 2024 08:43:47 +0000 (09:43 +0100)]
drm/xe/ct: prevent UAF in send_recv()
Ensure we serialize with completion side to prevent UAF with fence going
out of scope on the stack, since we have no clue if it will fire after
the timeout before we can erase from the xa. Also we have some dependent
loads and stores for which we need the correct ordering, and we lack the
needed barriers. Fix this by grabbing the ct->lock after the wait, which
is also held by the completion side.
v2 (Badal):
- Also print done after acquiring the lock and seeing timeout.
Matthew Brost [Fri, 27 Sep 2024 23:22:28 +0000 (16:22 -0700)]
drm/xe: Fix memory leak when aborting binds
Make sure to call xe_pt_update_ops_fini in xe_pt_update_ops_abort to
free any memory the bind allocated.
Caught by kmemleak when running Vulkan CTS tests on LNL. The leak
seems to happen only when there's some kind of failure happening, like
the lack of memory. Example output:
drm/xe: Prevent null pointer access in xe_migrate_copy
xe_migrate_copy designed to copy content of TTM resources. When source
resource is null, it will trigger a NULL pointer dereference in
xe_migrate_copy. To avoid this situation, update lacks source flag to
true for this case, the flag will trigger xe_migrate_clear rather than
xe_migrate_copy.
projects / linux.git / log