[Intel-gfx] [PATCH v13 00/13] Add GuC Error Capture Support

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Tue Apr 5 07:56:37 UTC 2022


On 23/03/2022 13:19, Tvrtko Ursulin wrote:
> 
> Hi,
> 
> On 21/03/2022 16:45, Alan Previn wrote:
>> This series:
>>    1. Enables support of GuC to report error-state-capture
>>       using a list of MMIO registers the driver registers
>>       and GuC will dump, log and notify right before a GuC
>>       triggered engine-reset event.
>>    2. Updates the ADS blob creation to register said lists
>>       of global, engine class and engine instance registers
>>       with GuC.
>>    3. Defines tables of register lists that are global or
>>       engine class or engine instance in scope.
>>    4. Updates usage and buffer-state data for the regions
>>       of the shared GuC log-buffer to accomdate both
>>       the existing relay logging of general debug logs
>>       along with the new error state capture usage.
>>    5. Using a pool of preallocated memory, provide ability
>>       to extract and format the GuC reported register-capture
>>       data into chunks consistent with existing i915 error-
>>       state collection flows and structures.
>>    6. Connects the i915_gpu_coredump reporting function
>>       to the GuC error capture module to print all GuC
>>       error state capture dumps that is reported.
> 
> Story is finished with this series and we have everything on feature 
> parity? Intel_error_decode handles it fine?
> 
> Would you have an example error capture at hand, I am curious how it looks?

Ping? I need to write a meaningful pull request text in ~2 weeks time 
when submitting the feature to drm-next so looking for a high level 
statement as described - does it complete the work and is error capture 
now at feature parity and all usual userspace tools to access it handle it?

Regards,

Tvrtko




> Regards,
> 
> Tvrtko
> 
>>
>> This is the 13th rev of this series where the first 3 revs
>> are RFC
>>
>> Prior receipts of rvb's:
>>    - Patch #2, #3, #4, #5, #10, #11, #12, #13 have received
>>      R-v-b's from Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
>>    - Patch #1, #6, #7, #8, #9 has received an R-v-b from Matthew Brost
>>      <matthew.brost at intel.com>. NOTE: some of these came in on the
>>      trybot series. https://patchwork.freedesktop.org/series/100831/
>>
>> Changes from prior revs:
>>    v13:- Fixing register list definition styling as per Jani's request.
>>    v12:- Re-sending it because previous revs only got to intel-gfx,
>>          and only cover letter was in dri-devel. Also rebased again.
>>    v11:- Rebase again on latest drm-tip to fix merge error.
>>    v10:- Rebase on latest drm-tip again. Fix a number of checkpatch
>>          warnings and an error Reported-by: kernel test robot 
>> <lkp at intel.com>.
>>    v9: - Rebase on latest drm-tip to solve CI merge-build error.
>>    v8: - Fix a bug found by CI in rev7: Create a cached ADS
>>          capture list for null-header like the other lists.
>>        - Fixed a bug on the ggtt offset calculation in the
>>          ADS population loop. Thanks to Matt Brost.
>>        - Change the storage uses for initial allocation and
>>          caching of the ADS register lists so we only store
>>          a regular pointer instead of file handle.
>>        - Multiple improvements on code styling, variable names,
>>          comments and code reduction from Umesh suggestions
>>          across multiple patches.
>>
>>    v7: - Rebased on lastest drm_tip that has the ADS now using
>>          shmem based ads_blob_write utilities. Stress test
>>          was performed with this patch included to fix a
>>          legacy bug:
>>          https://patchwork.freedesktop.org/series/100768/
>>
>>    v6: - In patch #1, ADS reg-list population, we now alloc
>>          regular memory to create the lists and cache them for
>>          simpler and faster use by GuC ADS module at init,
>>          suspend-resume and reset cycles. This was in response
>>          to review comments from Lucas De Marchi that also
>>          wanted to ensure the GuC ADS module owns the final
>>          copying into the ADS phyical memory.
>>        - Thanks to Jani Nikula for pointing out that patch #2
>>          and #3 should ensure static tables as constant and
>>          dynamic lists should be allocated and cached but
>>          attached to the GT level for the case of multiple
>>          cards with different fusings for steered registers.
>>          These are addressed now along with multiple code
>>          style fixups (thanks to review comment from Umesh)
>>          and splitting the steered register list generation
>>          as a seperate patch.
>>        - The extraction functionality, Patch #10 and #11 (was
>>          patch #7), has fixed all of Umesh's review comments
>>          related to the code styling. Additionally, it was
>>          discovered during stress tests that the extraction
>>          function could be called by the ct processing thread
>>          at the same time as the start of a GT reset event.
>>          Thus, a redesign was done whereby the linked list of
>>          processed capture-output-nodes are allocated up
>>          front and reused throughout the driver's life to
>>          ensure no memory locks are taken during extraction.
>>        - For patch #6 (now 7, 8 and 9), updates to
>>          intel_guc_log was split into smaller chunks and the
>>          log_state structure was returned back to inside of
>>          the intel_guc_log struct as opposed to the
>>          intel_guc struct in prior rev. This is in response
>>          to review comments by Matt Brost.
>>        - #Patch 13 (previously #10) is mostly identical but
>>          addresses all of the code styling comments reviews
>>          from Umesh.
>>    v5: - Added Gen9->Gen11 register list for CI coverage that
>>          included Gen9 with GuC submission.
>>        - Redesigned the extraction of the GuC error-capture
>>          dumps by grouping them into complete per-engine-reset
>>          nodes. Complete here means each node includes the
>>          global, engine-class and engine-instance register
>>          lists in a single structure.
>>        - Extraction is decoupled from the print-out. We now
>>          do the extraction immediately when receiving the
>>          G2H for error-capture notification. A link list of
>>          nodes is maintained with a FIFO based threshold
>>          while awaiting retrieval from i915_gpu_coredump's
>>          capture_engine function.
>>        - Added new plumbing through the i915_gpu_coredump
>>          allocation and capture functions to include a flag
>>          that is used indicate that GuC had triggered the
>>          reset. This new plumbing guarantees an exact match
>>          from i915_gpu_coredump's per-engine vma recording
>>          and node-retrieval from the guc-error-capture.
>>        - Broke the coredump gt_global capture and recording
>>          functions into smaller subsets so we can reuse as
>>          much of the existing legacy register reading + printing
>>          functions and only rely on GuC error-capture for
>>          the smaller subset of registers that are tied to
>>          engine workload execution.
>>        - Updated the register list to follow the legacy execlist
>>          format of printout.
>>    v4:
>>        - Rebased on latest drm-tip that has been merged with the
>>          support of GuC firmware version 69.0.3 that is required
>>          for GuC error-state-catpure to work.
>>        - Added register list for DG2 which is the same as XE_LP
>>          except an additional steering register set.
>>        - Fixed a bug in the end of capture parsing loop in
>>          intel_guc_capture_out_print_next_group that was not
>>          properly comparing the engine-instance and engine-
>>          class being parsed against the one that triggered
>>          the i915_gpu_coredump.
>>    v3:
>>        - Fixed all review comments from rev2 except the following:
>>            - Michal Wajdeczko proposed adding a seperate function
>>              to lookup register string nameslookup (based on offset)
>>              but decided against it because of offset conflicts
>>              and the current table layout is easier to maintain.
>>            - Last set of checkpatch errors pertaining to "COMPLEX
>>              MACROS" should be fixed on next rev.
>>        - Abstracted internal-to-guc-capture information into a new
>>          __guc_state_capture_priv structure that allows the exclusion
>>          of intel_guc.h and intel_guc_fwif.h from intel_guc_capture.h.
>>          Now, only the first 2 patches have a wider build time
>>          impact because of the changes to intel_guc_fwif.h but
>>          subsequent changes to guc-capture internal structures
>>          or firmware interfaces used solely by guc-capture module
>>          shoudn't impact the rest of the driver build.
>>        - Added missing Gen12LP registers and added slice+subslice
>>          indices when reporting extended steered registers.
>>        - Add additional checks to ensure that the GuC reported
>>          error capture information matches the i915_gpu_coredump
>>          that is being printed before we print out the corresponding
>>          VMA dumps such as the batch buffer.
>>     v2:
>>        - Ignore - failed CI retest.
>>
>> Alan Previn (13):
>>    drm/i915/guc: Update GuC ADS size for error capture lists
>>    drm/i915/guc: Add XE_LP static registers for GuC error capture.
>>    drm/i915/guc: Add XE_LP steered register lists support
>>    drm/i915/guc: Add DG2 registers for GuC error state capture.
>>    drm/i915/guc: Add Gen9 registers for GuC error state capture.
>>    drm/i915/guc: Add GuC's error state capture output structures.
>>    drm/i915/guc: Update GuC-log relay function names
>>    drm/i915/guc: Add capture region into intel_guc_log
>>    drm/i915/guc: Check sizing of guc_capture output
>>    drm/i915/guc: Extract GuC error capture lists on G2H notification.
>>    drm/i915/guc: Pre-allocate output nodes for extraction
>>    drm/i915/guc: Plumb GuC-capture into gpu_coredump
>>    drm/i915/guc: Print the GuC error capture output register list.
>>
>>   drivers/gpu/drm/i915/Makefile                 |    1 +
>>   drivers/gpu/drm/i915/gt/intel_engine_cs.c     |    4 +-
>>   .../drm/i915/gt/intel_execlists_submission.c  |    4 +-
>>   drivers/gpu/drm/i915/gt/intel_reset.c         |    2 +-
>>   .../gpu/drm/i915/gt/uc/abi/guc_actions_abi.h  |    7 +
>>   drivers/gpu/drm/i915/gt/uc/guc_capture_fwif.h |  218 +++
>>   drivers/gpu/drm/i915/gt/uc/intel_guc.c        |   13 +-
>>   drivers/gpu/drm/i915/gt/uc/intel_guc.h        |   12 +-
>>   drivers/gpu/drm/i915/gt/uc/intel_guc_ads.c    |  127 +-
>>   .../gpu/drm/i915/gt/uc/intel_guc_capture.c    | 1655 +++++++++++++++++
>>   .../gpu/drm/i915/gt/uc/intel_guc_capture.h    |   33 +
>>   drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h   |   14 +-
>>   drivers/gpu/drm/i915/gt/uc/intel_guc_log.c    |  127 +-
>>   drivers/gpu/drm/i915/gt/uc/intel_guc_log.h    |    7 +-
>>   .../gpu/drm/i915/gt/uc/intel_guc_submission.c |   18 +-
>>   drivers/gpu/drm/i915/i915_debugfs.c           |    3 +-
>>   drivers/gpu/drm/i915/i915_gpu_error.c         |  282 ++-
>>   drivers/gpu/drm/i915/i915_gpu_error.h         |   35 +-
>>   18 files changed, 2379 insertions(+), 183 deletions(-)
>>   create mode 100644 drivers/gpu/drm/i915/gt/uc/guc_capture_fwif.h
>>   create mode 100644 drivers/gpu/drm/i915/gt/uc/intel_guc_capture.c
>>   create mode 100644 drivers/gpu/drm/i915/gt/uc/intel_guc_capture.h
>>


More information about the Intel-gfx mailing list