[Deepin-Kernel-SIG] [linux 6.6-y] [Upstream] [Intel] sync intel i915 drm driver to upstream v6.7

[opsiff]

Summary by Sourcery

Sync the Intel i915 driver to kernel v6.7 by adding GuC-assisted TLB invalidation paths and new ABI actions, integrating GSC firmware support, standardizing GT logging, refining engine list handling, updating reset timing, and applying various bug fixes and selftest adjustments.

New Features:

• Add GuC-based TLB invalidation path with H2G commands, G2H responses, waiters, and xarray tracking
• Introduce GSC firmware support definitions and loading paths using compatibility-only major versions
• Add asynchronous dead GuC worker to handle crash and exception notifications outside G2H context

Bug Fixes:

• Wake up all pending TLB invalidation waiters on GuC or GT reset/suspend to avoid hung waits
• Fix out-of-date OA buffer head/tail handling by removing GTT offset miscalculations

Enhancements:

• Standardize logging with gt_print macros (gt_err, gt_dbg, etc.) to include GT identifier
• Refactor user engine registration lists to use a unified llist/list/rbtree union in engine structures
• Adjust GPU reset timings and atomic waits to new microsecond-based wait helpers
• Add early resume hook to sanitize steer semaphores during driver resume

Tests:

• Tweak selftest spinner and migration timings (longer msleep for OTHER_CLASS engines)
• Use memset32 to pad spinner batches and replace strncpy with strscpy in mock_context

[sourcery-ai]

Reviewer's Guide

Sync Intel i915 driver to v6.7: implements GuC-based TLB invalidation using XArray waitqueues, introduces GT‐aware logging macros, extends GPU reset timeouts, enhances GSC firmware handling, refactors user‐engine data structures, cleans up performance buffer pointer arithmetic, and updates the GEM shrinker to iterate per GT.

Sequence Diagram: GuC-based TLB Invalidation Process

sequenceDiagram
    participant D as Driver
    participant GUC as Intel GuC
    participant TLB_HW as GT TLB Hardware

    D->>D: intel_gt_invalidate_tlb_full() called
    alt GuC TLB Invalidation Path (HAS_GUC_TLB_INVALIDATION and GuC ready)
        D->>D: Chooses intel_guc_invalidate_tlb_engines() or _guc()
        Note right of D: Internally calls guc_send_invalidate_tlb()
        D->>D: Prepares intel_guc_tlb_wait (entry in tlb_lookup or serial_slot)
        D->>GUC: H2G Command: INTEL_GUC_ACTION_TLB_INVALIDATION (with seqno)
        D->>D: Waits on tlb_wait.wq using must_wait_woken()
        GUC->>TLB_HW: Performs TLB invalidation on hardware
        GUC->>D: G2H Message: INTEL_GUC_ACTION_TLB_INVALIDATION_DONE (with seqno)
        D->>D: G2H Handler (ct_process_request -> intel_guc_tlb_invalidation_done)
        Note right of D: Wakes up waiting thread via wait_wake_outstanding_tlb_g2h()
    else Legacy MMIO Invalidation Path
        D->>TLB_HW: Direct MMIO based TLB Invalidation (mmio_invalidate_full)
    end
    D->>D: TLB Invalidation Process Complete

Loading

Sequence Diagram: GuC Crash and Exception Handling via Worker

sequenceDiagram
    participant GUC_HW as Intel GuC (Hardware/Firmware)
    participant DRV_CT as Driver (CT Layer / G2H Handler)
    participant GUC_MOD as Driver (intel_guc module)
    participant WORKER as Driver (dead_guc_worker)
    participant GT_MOD as Driver (Intel GT module)

    GUC_HW->>DRV_CT: G2H Msg (e.g., NOTIFY_CRASH_DUMP_POSTED / NOTIFY_EXCEPTION)
    DRV_CT->>GUC_MOD: ct_process_request() calls intel_guc_crash_process_msg(action)
    GUC_MOD->>WORKER: queue_work(system_unbound_wq, &guc->dead_guc_worker)
    activate WORKER
    WORKER->>WORKER: guc_dead_worker_func() executes
    WORKER->>GUC_MOD: Reads guc.last_dead_guc_jiffies
    alt Short interval since last failure (delta < 500ms)
        WORKER->>GT_MOD: intel_gt_set_wedged(gt)
    else (First failure or longer interval)
        WORKER->>GT_MOD: intel_gt_handle_error(gt, ALL_ENGINES, ..., "dead GuC")
        WORKER->>GUC_MOD: Updates guc.last_dead_guc_jiffies = jiffies
    end
    deactivate WORKER

Loading

Updated Class Diagram for intel_guc and new intel_guc_tlb_wait

classDiagram
    class intel_guc {
        +xarray tlb_lookup
        +u32 serial_slot
        +u32 next_seqno
        +work_struct dead_guc_worker
        +unsigned_long last_dead_guc_jiffies
        +wake_up_all_tlb_invalidate()
        +intel_guc_tlb_invalidation_is_available() bool
        +init_tlb_lookup() int
        +fini_tlb_lookup()
        +wait_wake_outstanding_tlb_g2h(u32 seqno)
        +intel_guc_tlb_invalidation_done(const u32* payload, u32 len) int
        +guc_send_invalidate_tlb(intel_guc_tlb_invalidation_type type) int
        +intel_guc_invalidate_tlb_engines() int
        +intel_guc_invalidate_tlb_guc() int
        +intel_guc_crash_process_msg(u32 action) int
    }

    class intel_guc_tlb_wait {
        +wait_queue_head wq
        +bool busy
    }

    intel_guc "1" -- "*" intel_guc_tlb_wait : Uses via tlb_lookup

Loading

Class Diagram: UABI Engine List Structure Refactoring

classDiagram
    class drm_i915_private {
      +llist_head uabi_engines_llist
      +list_head uabi_engines_list
      +rb_root uabi_engines
      note "Fields are part of a union for UABI engines"
    }

    class intel_engine_cs {
      +llist_node uabi_llist
      +list_head uabi_list
      +rb_node uabi_node
      note "Fields are part of a union for UABI linkage"
    }

Loading

File-Level Changes

Change	Details	Files
GuC TLB invalidation framework	Add XArray "tlb_lookup" with init/fini and cyclic allocation for waiter slots Implement wake_up_all_tlb_invalidate to release blocked waiters on reset or cancellation Introduce guc_send_invalidate_tlb with waitqueues, sequence numbers, retry under memory pressure Handle H2G/G2H actions INTEL_GUC_ACTION_TLB_INVALIDATION and DONE in CT and host message paths	drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c drivers/gpu/drm/i915/gt/uc/intel_guc.h drivers/gpu/drm/i915/gt/uc/intel_guc_ct.c drivers/gpu/drm/i915/gt/uc/intel_guc.c drivers/gpu/drm/i915/gt/uc/abi/guc_actions_abi.h drivers/gpu/drm/i915/gt/intel_tlb.c drivers/gpu/drm/i915/gt/selftest_tlb.c drivers/gpu/drm/i915/gt/uc/intel_guc_ct.h
GT‐aware logging macros	Define gt_dbg/gt_err/gt_notice and once variants in intel_gt_print.h Replace drm_dbg/drm_err/drm_notice calls with GT‐context macros across reset, GSC, workarounds modules	drivers/gpu/drm/i915/gt/intel_gt_print.h drivers/gpu/drm/i915/gt/intel_reset.c drivers/gpu/drm/i915/gt/uc/intel_gsc.c
Extended GPU reset timeouts	Increase PCI reset assertion delays to 50 µs and use _wait_for_atomic with 50 ms timeouts Unify reset polling calls to use _wait_for_atomic for media and render resets	drivers/gpu/drm/i915/gt/intel_reset.c
GSC firmware support enhancements	Add INTEL_GSC_FIRMWARE_DEFS and MAKE_GSC_FW_PATH macros Declare GSC blobs in firmware tables and auto‐select lists Use gt_err/gt_dbg in GSC init and IRQ handling	drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c drivers/gpu/drm/i915/gt/uc/intel_gsc_uc.c drivers/gpu/drm/i915/gt/uc/abi/guc_actions_abi.h
Engine UABI data structure refactor	Change uabi_nodes to a union of llist, list, and rb_node for multi‐stage init Use llist_head for user engine registration and convert to sorted list Update iterators in intel_engine_user.c and i915_drv.h accordingly	drivers/gpu/drm/i915/gt/intel_engine_types.h drivers/gpu/drm/i915/gt/intel_engine_user.c drivers/gpu/drm/i915/i915_drv.h
OA buffer pointer arithmetic cleanup	Remove manual GTT offset adjustments when reading tail/head Simplify tail comparison logic to use stream->oa_buffer.tail directly Reset head/tail to zero on buffer init	drivers/gpu/drm/i915/i915_perf.c
GEM shrinker multi-GT iteration	Loop over all GTs when retiring active contexts under I915_SHRINK_ACTIVE Iterate per‐GT ggtt vm.bound_list to unbind stale IO maps Aggregate freed pages across multiple GTs	drivers/gpu/drm/i915/gem/i915_gem_shrinker.c

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Hey @opsiff - I've reviewed your changes and they look great!

Here's what I looked at during the review

• 🟢 General issues: all looks good
• 🟢 Security: all looks good
• 🟢 Testing: all looks good
• 🟡 Complexity: 1 issue found
• 🟢 Documentation: all looks good

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[sourcery-ai]

issue (complexity): Consider replacing the custom jiffies-based throttling and manual work management with the kernel's delayed_work API for dead GuC handling.

Suggested change

	static void guc_dead_worker_func(struct work_struct *w)
	Rather than roll your own jiffies‐throttling + plain work_struct + ad hoc flushing everywhere, you can leverage delayed_work to get both throttling and a single “flushable” queue point. This drops the extra `last_dead_guc_jiffies` field, the manual `jiffies_to_msecs()` dance, and scattered `flush_work()` calls.
	1. In your struct, switch to delayed_work:
	```c
	- struct work_struct dead_guc_worker;
	+ struct delayed_work dead_guc_dw;

2. Init it once in init_early:

-   INIT_WORK(&guc->dead_guc_worker, guc_dead_worker_func);
+   INIT_DELAYED_WORK(&guc->dead_guc_dw, guc_dead_worker_fn);

3. Simplify the worker fn – no timestamps needed:

static void guc_dead_guc_worker_fn(struct work_struct *w)
{
    struct intel_guc *guc =
        container_of(to_delayed_work(w), struct intel_guc, dead_guc_dw);
    struct intel_gt *gt = guc_to_gt(guc);

    /* always capture/reset via the standard path */
    intel_gt_handle_error(gt, ALL_ENGINES, I915_ERROR_CAPTURE, "dead GuC");
}

4. Centralize scheduling + throttle via delay:

static void guc_schedule_dead_guc(struct intel_guc *guc)
{
    /* schedule only once per 500ms window */
    schedule_delayed_work(&guc->dead_guc_dw,
                          msecs_to_jiffies(500));
}

5. Replace all queue_work(...) + flush_work(...) or queue_work(system_unbound_wq, ...) calls with:

    guc_schedule_dead_guc(guc);

and remove all last_dead_guc_jiffies and flush_work() invocations.

6. In fini/_suspend/reset paths, do one final flush:

    cancel_delayed_work_sync(&guc->dead_guc_dw);

This collapses multiple reset paths onto the existing delayed_work API, preserves the 500 ms throttle, and gives you a single point to cancel/flush.

[Copilot]

Pull Request Overview

This PR syncs the Intel i915 driver to kernel v6.7 and incorporates several enhancements including GuC-assisted TLB invalidation support, updated firmware ABI actions, refined logging, and adjusted reset timings.

• Introduces new GuC TLB invalidation actions and associated macros.
• Updates logging calls to use gt_dbg/gt_err and adjusts GPU reset timing values.
• Refactors engine user list handling and updates selftest timing and string handling.

Reviewed Changes

Copilot reviewed 37 out of 37 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
drivers/gpu/drm/i915/gt/uc/intel_gsc_uc.c	Replaces drm_* logging calls with gt_* equivalents for consistency.
drivers/gpu/drm/i915/gt/uc/abi/guc_actions_abi.h	Adds new GuC action codes and macros for TLB invalidation with a comment typo.
drivers/gpu/drm/i915/gt/selftest_tlb.c	Adjusts sleep delay for OTHER_CLASS engines with an added FIXME comment.
drivers/gpu/drm/i915/gt/selftest_migrate.c	Updates msleep argument to account for the number of CPUs.
drivers/gpu/drm/i915/gt/intel_workarounds.c	Replaces some logging calls and refines workaround implementations.
drivers/gpu/drm/i915/gt/intel_tlb.c	Adds GuC TLB invalidation path to the full TLB invalidation function.
drivers/gpu/drm/i915/gt/intel_reset.c	Increases reset assertion delays and uses _wait_for_atomic instead of wait_for_atomic.
drivers/gpu/drm/i915/gt/intel_engine_user.c, intel_engine_types.h	Updates list handling for engine registration via union of list types.
drivers/gpu/drm/i915/gt/intel_engine_cs.c	Minor logging updates and safety checks in context initialization.
drivers/gpu/drm/i915/gem/selftests/mock_context.c	Uses strscpy in place of strncpy for safe string copy.
drivers/gpu/drm/i915/gem/i915_gem_shrinker.c, i915_gem_execbuffer.c	Removes inline qualifiers from several helper routines for clarity.

Comments suppressed due to low confidence (2)

drivers/gpu/drm/i915/gt/uc/abi/guc_actions_abi.h:210

• There is a typo in the comment: 'GOd' should be corrected to 'gone'.

 * in-flight transactions are GOd.

drivers/gpu/drm/i915/gt/intel_reset.c:168

• Verify that the increased timeout value (50000 microseconds) and use of _wait_for_atomic align with hardware requirements and expected reset behavior.

err = _wait_for_atomic(i915_in_reset(pdev), 50000, 0);

[Avenger-285714]

/lgtm

[Avenger-285714]

/approve

[deepin-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: Avenger-285714

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• deepin/OWNERS [Avenger-285714]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment