drm/xe/guc: Fix handling of GUC_HXG_TYPE_NO_RESPONSE_BUSY
authorMichal Wajdeczko <michal.wajdeczko@intel.com>
Thu, 16 Nov 2023 15:12:41 +0000 (16:12 +0100)
committerRodrigo Vivi <rodrigo.vivi@intel.com>
Thu, 21 Dec 2023 16:44:38 +0000 (11:44 -0500)
If GuC responds with the NO_RESPONSE_BUSY message, we extend
our timeout while waiting for the actual response, but we wrongly
assumed that the next message will be RESPONSE_SUCCESS, missing
that we still can get RESPONSE_FAILURE.

Change the condition for the expected message type, using only
common bits from RESPONSE_SUCCESS and RESPONSE_FAILURE (as they
differ, by ABI design, only by the last bit).

v2: add comment/checks to the code (Matt)

Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
drivers/gpu/drm/xe/xe_guc.c

index aa4af536c1e98fdb890ddb0a218a28b8784b0612..56edcb2b0e45234a92182ab9b218a8878bf48e7b 100644 (file)
@@ -659,9 +659,20 @@ timeout:
        header = xe_mmio_read32(gt, reply_reg);
        if (FIELD_GET(GUC_HXG_MSG_0_TYPE, header) ==
            GUC_HXG_TYPE_NO_RESPONSE_BUSY) {
-
-               ret = xe_mmio_wait32(gt, reply_reg, GUC_HXG_MSG_0_TYPE,
-                                    FIELD_PREP(GUC_HXG_MSG_0_TYPE, GUC_HXG_TYPE_RESPONSE_SUCCESS),
+               /*
+                * Once we got a BUSY reply we must wait again for the final
+                * response but this time we can't use ORIGIN mask anymore.
+                * To spot a right change in the reply, we take advantage that
+                * response SUCCESS and FAILURE differ only by the single bit
+                * and all other bits are set and can be used as a new mask.
+                */
+               u32 resp_bits = GUC_HXG_TYPE_RESPONSE_SUCCESS & GUC_HXG_TYPE_RESPONSE_FAILURE;
+               u32 resp_mask = FIELD_PREP(GUC_HXG_MSG_0_TYPE, resp_bits);
+
+               BUILD_BUG_ON(FIELD_MAX(GUC_HXG_MSG_0_TYPE) != GUC_HXG_TYPE_RESPONSE_SUCCESS);
+               BUILD_BUG_ON((GUC_HXG_TYPE_RESPONSE_SUCCESS ^ GUC_HXG_TYPE_RESPONSE_FAILURE) != 1);
+
+               ret = xe_mmio_wait32(gt, reply_reg,  resp_mask, resp_mask,
                                     1000000, &header, false);
 
                if (unlikely(FIELD_GET(GUC_HXG_MSG_0_ORIGIN, header) !=