[PATCH v2 1/2] drm/xe: introduce xe_gt_reset and xe_gt_reset_wait

[Lucas De Marchi]

On Tue, Dec 10, 2024 at 03:37:47PM +0100, Maciej Patelczyk wrote:
>Adding synchronous version gt reset as there are few places where it
>is expected.
>Adding also a wait helper to wait until gt reset is done.

s/Adding/Add/g

>
>Signed-off-by: Maciej Patelczyk <maciej.patelczyk at intel.com>
>---
> drivers/gpu/drm/xe/tests/xe_bo.c   |  7 +++----
> drivers/gpu/drm/xe/tests/xe_mocs.c |  3 +--
> drivers/gpu/drm/xe/xe_gt.h         | 25 +++++++++++++++++++++++++
> drivers/gpu/drm/xe/xe_gt_debugfs.c |  2 +-
> 4 files changed, 30 insertions(+), 7 deletions(-)
>
>diff --git a/drivers/gpu/drm/xe/tests/xe_bo.c b/drivers/gpu/drm/xe/tests/xe_bo.c
>index c9ec7a313c6b..a1af45e06189 100644
>--- a/drivers/gpu/drm/xe/tests/xe_bo.c
>+++ b/drivers/gpu/drm/xe/tests/xe_bo.c
>@@ -264,10 +264,9 @@ static int evict_test_run_tile(struct xe_device *xe, struct xe_tile *tile, struc
> 		 * however seems quite fragile not to also restart the GT. Try
> 		 * to do that here by triggering a GT reset.
> 		 */
>-		for_each_gt(__gt, xe, id) {
>-			xe_gt_reset_async(__gt);
>-			flush_work(&__gt->reset.worker);
>-		}
>+		for_each_gt(__gt, xe, id)
>+			xe_gt_reset(__gt);
>+
> 		if (err) {
> 			KUNIT_FAIL(test, "restore kernel err=%pe\n",
> 				   ERR_PTR(err));
>diff --git a/drivers/gpu/drm/xe/tests/xe_mocs.c b/drivers/gpu/drm/xe/tests/xe_mocs.c
>index 6f9b7a266b41..2698e433c6fa 100644
>--- a/drivers/gpu/drm/xe/tests/xe_mocs.c
>+++ b/drivers/gpu/drm/xe/tests/xe_mocs.c
>@@ -162,8 +162,7 @@ static int mocs_reset_test_run_device(struct xe_device *xe)
> 		if (flags & HAS_LNCF_MOCS)
> 			read_l3cc_table(gt, &mocs.table);
>
>-		xe_gt_reset_async(gt);
>-		flush_work(&gt->reset.worker);
>+		xe_gt_reset(gt);
>
> 		kunit_info(test, "mocs_reset_test after reset\n");
> 		if (flags & HAS_GLOBAL_MOCS)
>diff --git a/drivers/gpu/drm/xe/xe_gt.h b/drivers/gpu/drm/xe/xe_gt.h
>index 82b9b7f82fca..8db64e289892 100644
>--- a/drivers/gpu/drm/xe/xe_gt.h
>+++ b/drivers/gpu/drm/xe/xe_gt.h
>@@ -56,6 +56,31 @@ void xe_gt_sanitize(struct xe_gt *gt);
> int xe_gt_sanitize_freq(struct xe_gt *gt);
> void xe_gt_remove(struct xe_gt *gt);
>
>+/**
>+ * xe_gt_reset_wait - wait for gt's async reset to finalize.

IMO it would read better as xe_gt_wait_for_reset(),
otherwise it's not clear from name if this function calls the reset
*and* wait or if it just waits a previously posted reset request.

>+ * @gt: GT structure
>+ * Return:
>+ * %true if xe_gt_reset_wait() waited for the work to finish execution,

%true if it waited...

>+ * %false if there was no scheduled reset or it was done.
>+ */
>+static inline bool xe_gt_reset_wait(struct xe_gt *gt)
>+{
>+	return flush_work(&gt->reset.worker);
>+}
>+
>+/**
>+ * xe_gt_reset - perform synchronous reset
>+ * @gt: GT structure
>+ * Return:
>+ * %true if waited for the reset to finish,
>+ * %false if there was no scheduled reset.
>+ */
>+static inline bool xe_gt_reset(struct xe_gt *gt)
>+{
>+	xe_gt_reset_async(gt);
>+	return xe_gt_reset_wait(gt);
>+}
>+
> /**
>  * xe_gt_any_hw_engine_by_reset_domain - scan the list of engines and return the
>  * first that matches the same reset domain as @class
>diff --git a/drivers/gpu/drm/xe/xe_gt_debugfs.c b/drivers/gpu/drm/xe/xe_gt_debugfs.c
>index 3e8c351a0eab..79f61532fd00 100644
>--- a/drivers/gpu/drm/xe/xe_gt_debugfs.c
>+++ b/drivers/gpu/drm/xe/xe_gt_debugfs.c
>@@ -135,7 +135,7 @@ static int force_reset_sync(struct xe_gt *gt, struct drm_printer *p)
> 	xe_gt_reset_async(gt);
> 	xe_pm_runtime_put(gt_to_xe(gt));
>
>-	flush_work(&gt->reset.worker);
>+	xe_gt_reset_wait(gt);

this can become a xe_gt_reset() since the only thing we are doing
between the calls is to release the runtime pm.... doesn't really
matter for this debugfs call.

other than the nitpicks,  Reviewed-by: Lucas De Marchi <lucas.demarchi at intel.com>

thanks
Lucas De Marchi

>
> 	return 0;
> }
>-- 
>2.43.0
>