Merge v6.9.4

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

5 files changed, 115 insertions, 75 deletions

diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
index d32ff3857e65..b3b37ed832ca 100644
--- a/drivers/gpu/drm/xe/xe_device.c
+++ b/drivers/gpu/drm/xe/xe_device.c
@@ -389,8 +389,14 @@ mask_err:
 	return err;
 }
 
-/*
- * Initialize MMIO resources that don't require any knowledge about tile count.
+/**
+ * xe_device_probe_early: Device early probe
+ * @xe: xe device instance
+ *
+ * Initialize MMIO resources that don't require any
+ * knowledge about tile count. Also initialize pcode
+ *
+ * Return: 0 on success, error code on failure
  */
 int xe_device_probe_early(struct xe_device *xe)
 {
@@ -404,6 +410,10 @@ int xe_device_probe_early(struct xe_device *xe)
 	if (err)
 		return err;
 
+	err = xe_pcode_probe_early(xe);
+	if (err)
+		return err;
+
 	return 0;
 }
 
@@ -482,11 +492,8 @@ int xe_device_probe(struct xe_device *xe)
 	if (err)
 		return err;
 
-	for_each_gt(gt, xe, id) {
-		err = xe_pcode_probe(gt);
-		if (err)
-			return err;
-	}
+	for_each_gt(gt, xe, id)
+		xe_pcode_init(gt);
 
 	err = xe_display_init_noirq(xe);
 	if (err)
diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c
index 2ba4fb9511f6..aca519f5b85d 100644
--- a/drivers/gpu/drm/xe/xe_migrate.c
+++ b/drivers/gpu/drm/xe/xe_migrate.c
@@ -33,7 +33,6 @@
 #include "xe_sync.h"
 #include "xe_trace.h"
 #include "xe_vm.h"
-#include "xe_wa.h"
 
 /**
  * struct xe_migrate - migrate context.
@@ -299,10 +298,6 @@ static int xe_migrate_prepare_vm(struct xe_tile *tile, struct xe_migrate *m,
 }
 
 /*
- * Due to workaround 16017236439, odd instance hardware copy engines are
- * faster than even instance ones.
- * This function returns the mask involving all fast copy engines and the
- * reserved copy engine to be used as logical mask for migrate engine.
  * Including the reserved copy engine is required to avoid deadlocks due to
  * migrate jobs servicing the faults gets stuck behind the job that faulted.
  */
@@ -316,8 +311,7 @@ static u32 xe_migrate_usm_logical_mask(struct xe_gt *gt)
 		if (hwe->class != XE_ENGINE_CLASS_COPY)
 			continue;
 
-		if (!XE_WA(gt, 16017236439) ||
-		    xe_gt_is_usm_hwe(gt, hwe) || hwe->instance & 1)
+		if (xe_gt_is_usm_hwe(gt, hwe))
 			logical_mask |= BIT(hwe->logical_instance);
 	}
 
@@ -368,6 +362,10 @@ struct xe_migrate *xe_migrate_init(struct xe_tile *tile)
 		if (!hwe || !logical_mask)
 			return ERR_PTR(-EINVAL);
 
+		/*
+		 * XXX: Currently only reserving 1 (likely slow) BCS instance on
+		 * PVC, may want to revisit if performance is needed.
+		 */
 		m->q = xe_exec_queue_create(xe, vm, logical_mask, 1, hwe,
 					    EXEC_QUEUE_FLAG_KERNEL |
 					    EXEC_QUEUE_FLAG_PERMANENT |
diff --git a/drivers/gpu/drm/xe/xe_pcode.c b/drivers/gpu/drm/xe/xe_pcode.c
index b324dc2a5deb..81f4ae2ea08f 100644
--- a/drivers/gpu/drm/xe/xe_pcode.c
+++ b/drivers/gpu/drm/xe/xe_pcode.c
@@ -10,6 +10,7 @@
 
 #include <drm/drm_managed.h>
 
+#include "xe_device.h"
 #include "xe_gt.h"
 #include "xe_mmio.h"
 #include "xe_pcode_api.h"
@@ -43,8 +44,6 @@ static int pcode_mailbox_status(struct xe_gt *gt)
 		[PCODE_ERROR_MASK] = {-EPROTO, "Unknown"},
 	};
 
-	lockdep_assert_held(&gt->pcode.lock);
-
 	err = xe_mmio_read32(gt, PCODE_MAILBOX) & PCODE_ERROR_MASK;
 	if (err) {
 		drm_err(&gt_to_xe(gt)->drm, "PCODE Mailbox failed: %d %s", err,
@@ -55,17 +54,15 @@ static int pcode_mailbox_status(struct xe_gt *gt)
 	return 0;
 }
 
-static int pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1,
-			    unsigned int timeout_ms, bool return_data,
-			    bool atomic)
+static int __pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1,
+			      unsigned int timeout_ms, bool return_data,
+			      bool atomic)
 {
 	int err;
 
 	if (gt_to_xe(gt)->info.skip_pcode)
 		return 0;
 
-	lockdep_assert_held(&gt->pcode.lock);
-
 	if ((xe_mmio_read32(gt, PCODE_MAILBOX) & PCODE_READY) != 0)
 		return -EAGAIN;
 
@@ -87,6 +84,18 @@ static int pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1,
 	return pcode_mailbox_status(gt);
 }
 
+static int pcode_mailbox_rw(struct xe_gt *gt, u32 mbox, u32 *data0, u32 *data1,
+			    unsigned int timeout_ms, bool return_data,
+			    bool atomic)
+{
+	if (gt_to_xe(gt)->info.skip_pcode)
+		return 0;
+
+	lockdep_assert_held(&gt->pcode.lock);
+
+	return __pcode_mailbox_rw(gt, mbox, data0, data1, timeout_ms, return_data, atomic);
+}
+
 int xe_pcode_write_timeout(struct xe_gt *gt, u32 mbox, u32 data, int timeout)
 {
 	int err;
@@ -109,15 +118,19 @@ int xe_pcode_read(struct xe_gt *gt, u32 mbox, u32 *val, u32 *val1)
 	return err;
 }
 
-static int xe_pcode_try_request(struct xe_gt *gt, u32 mbox,
-				u32 request, u32 reply_mask, u32 reply,
-				u32 *status, bool atomic, int timeout_us)
+static int pcode_try_request(struct xe_gt *gt, u32 mbox,
+			     u32 request, u32 reply_mask, u32 reply,
+			     u32 *status, bool atomic, int timeout_us, bool locked)
 {
 	int slept, wait = 10;
 
 	for (slept = 0; slept < timeout_us; slept += wait) {
-		*status = pcode_mailbox_rw(gt, mbox, &request, NULL, 1, true,
-					   atomic);
+		if (locked)
+			*status = pcode_mailbox_rw(gt, mbox, &request, NULL, 1, true,
+						   atomic);
+		else
+			*status = __pcode_mailbox_rw(gt, mbox, &request, NULL, 1, true,
+						     atomic);
 		if ((*status == 0) && ((request & reply_mask) == reply))
 			return 0;
 
@@ -158,8 +171,8 @@ int xe_pcode_request(struct xe_gt *gt, u32 mbox, u32 request,
 
 	mutex_lock(&gt->pcode.lock);
 
-	ret = xe_pcode_try_request(gt, mbox, request, reply_mask, reply, &status,
-				   false, timeout_base_ms * 1000);
+	ret = pcode_try_request(gt, mbox, request, reply_mask, reply, &status,
+				false, timeout_base_ms * 1000, true);
 	if (!ret)
 		goto out;
 
@@ -177,8 +190,8 @@ int xe_pcode_request(struct xe_gt *gt, u32 mbox, u32 request,
 		"PCODE timeout, retrying with preemption disabled\n");
 	drm_WARN_ON_ONCE(&gt_to_xe(gt)->drm, timeout_base_ms > 1);
 	preempt_disable();
-	ret = xe_pcode_try_request(gt, mbox, request, reply_mask, reply, &status,
-				   true, timeout_base_ms * 1000);
+	ret = pcode_try_request(gt, mbox, request, reply_mask, reply, &status,
+				true, 50 * 1000, true);
 	preempt_enable();
 
 out:
@@ -238,59 +251,71 @@ unlock:
 }
 
 /**
- * xe_pcode_init - Ensure PCODE is initialized
- * @gt: gt instance
+ * xe_pcode_ready - Ensure PCODE is initialized
+ * @xe: xe instance
+ * @locked: true if lock held, false otherwise
  *
- * This function ensures that PCODE is properly initialized. To be called during
- * probe and resume paths.
+ * PCODE init mailbox is polled only on root gt of root tile
+ * as the root tile provides the initialization is complete only
+ * after all the tiles have completed the initialization.
+ * Called only on early probe without locks and with locks in
+ * resume path.
  *
- * It returns 0 on success, and -error number on failure.
+ * Returns 0 on success, and -error number on failure.
  */
-int xe_pcode_init(struct xe_gt *gt)
+int xe_pcode_ready(struct xe_device *xe, bool locked)
 {
 	u32 status, request = DGFX_GET_INIT_STATUS;
+	struct xe_gt *gt = xe_root_mmio_gt(xe);
 	int timeout_us = 180000000; /* 3 min */
 	int ret;
 
-	if (gt_to_xe(gt)->info.skip_pcode)
+	if (xe->info.skip_pcode)
 		return 0;
 
-	if (!IS_DGFX(gt_to_xe(gt)))
+	if (!IS_DGFX(xe))
 		return 0;
 
-	mutex_lock(&gt->pcode.lock);
-	ret = xe_pcode_try_request(gt, DGFX_PCODE_STATUS, request,
-				   DGFX_INIT_STATUS_COMPLETE,
-				   DGFX_INIT_STATUS_COMPLETE,
-				   &status, false, timeout_us);
-	mutex_unlock(&gt->pcode.lock);
+	if (locked)
+		mutex_lock(&gt->pcode.lock);
+
+	ret = pcode_try_request(gt, DGFX_PCODE_STATUS, request,
+				DGFX_INIT_STATUS_COMPLETE,
+				DGFX_INIT_STATUS_COMPLETE,
+				&status, false, timeout_us, locked);
+
+	if (locked)
+		mutex_unlock(&gt->pcode.lock);
 
 	if (ret)
-		drm_err(&gt_to_xe(gt)->drm,
+		drm_err(&xe->drm,
 			"PCODE initialization timedout after: 3 min\n");
 
 	return ret;
 }
 
 /**
- * xe_pcode_probe - Prepare xe_pcode and also ensure PCODE is initialized.
+ * xe_pcode_init: initialize components of PCODE
  * @gt: gt instance
  *
- * This function initializes the xe_pcode component, and when needed, it ensures
- * that PCODE has properly performed its initialization and it is really ready
- * to go. To be called once only during probe.
- *
- * It returns 0 on success, and -error number on failure.
+ * This function initializes the xe_pcode component.
+ * To be called once only during probe.
  */
-int xe_pcode_probe(struct xe_gt *gt)
+void xe_pcode_init(struct xe_gt *gt)
 {
 	drmm_mutex_init(&gt_to_xe(gt)->drm, &gt->pcode.lock);
+}
 
-	if (gt_to_xe(gt)->info.skip_pcode)
-		return 0;
-
-	if (!IS_DGFX(gt_to_xe(gt)))
-		return 0;
-
-	return xe_pcode_init(gt);
+/**
+ * xe_pcode_probe_early: initializes PCODE
+ * @xe: xe instance
+ *
+ * This function checks the initialization status of PCODE
+ * To be called once only during early probe without locks.
+ *
+ * Returns 0 on success, error code otherwise
+ */
+int xe_pcode_probe_early(struct xe_device *xe)
+{
+	return xe_pcode_ready(xe, false);
 }
diff --git a/drivers/gpu/drm/xe/xe_pcode.h b/drivers/gpu/drm/xe/xe_pcode.h
index 08cb1d047cba..3f54c6d2a57d 100644
--- a/drivers/gpu/drm/xe/xe_pcode.h
+++ b/drivers/gpu/drm/xe/xe_pcode.h
@@ -8,9 +8,11 @@
 
 #include <linux/types.h>
 struct xe_gt;
+struct xe_device;
 
-int xe_pcode_probe(struct xe_gt *gt);
-int xe_pcode_init(struct xe_gt *gt);
+void xe_pcode_init(struct xe_gt *gt);
+int xe_pcode_probe_early(struct xe_device *xe);
+int xe_pcode_ready(struct xe_device *xe, bool locked);
 int xe_pcode_init_min_freq_table(struct xe_gt *gt, u32 min_gt_freq,
 				 u32 max_gt_freq);
 int xe_pcode_read(struct xe_gt *gt, u32 mbox, u32 *val, u32 *val1);
diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c
index 53b3b0b019ac..944cf4d76099 100644
--- a/drivers/gpu/drm/xe/xe_pm.c
+++ b/drivers/gpu/drm/xe/xe_pm.c
@@ -54,13 +54,15 @@ int xe_pm_suspend(struct xe_device *xe)
 	u8 id;
 	int err;
 
+	drm_dbg(&xe->drm, "Suspending device\n");
+
 	for_each_gt(gt, xe, id)
 		xe_gt_suspend_prepare(gt);
 
 	/* FIXME: Super racey... */
 	err = xe_bo_evict_all(xe);
 	if (err)
-		return err;
+		goto err;
 
 	xe_display_pm_suspend(xe);
 
@@ -68,7 +70,7 @@ int xe_pm_suspend(struct xe_device *xe)
 		err = xe_gt_suspend(gt);
 		if (err) {
 			xe_display_pm_resume(xe);
-			return err;
+			goto err;
 		}
 	}
 
@@ -76,7 +78,11 @@ int xe_pm_suspend(struct xe_device *xe)
 
 	xe_display_pm_suspend_late(xe);
 
+	drm_dbg(&xe->drm, "Device suspended\n");
 	return 0;
+err:
+	drm_dbg(&xe->drm, "Device suspend failed %d\n", err);
+	return err;
 }
 
 /**
@@ -92,14 +98,14 @@ int xe_pm_resume(struct xe_device *xe)
 	u8 id;
 	int err;
 
+	drm_dbg(&xe->drm, "Resuming device\n");
+
 	for_each_tile(tile, xe, id)
 		xe_wa_apply_tile_workarounds(tile);
 
-	for_each_gt(gt, xe, id) {
-		err = xe_pcode_init(gt);
-		if (err)
-			return err;
-	}
+	err = xe_pcode_ready(xe, true);
+	if (err)
+		return err;
 
 	xe_display_pm_resume_early(xe);
 
@@ -109,7 +115,7 @@ int xe_pm_resume(struct xe_device *xe)
 	 */
 	err = xe_bo_restore_kernel(xe);
 	if (err)
-		return err;
+		goto err;
 
 	xe_irq_resume(xe);
 
@@ -120,9 +126,13 @@ int xe_pm_resume(struct xe_device *xe)
 
 	err = xe_bo_restore_user(xe);
 	if (err)
-		return err;
+		goto err;
 
+	drm_dbg(&xe->drm, "Device resumed\n");
 	return 0;
+err:
+	drm_dbg(&xe->drm, "Device resume failed %d\n", err);
+	return err;
 }
 
 static bool xe_pm_pci_d3cold_capable(struct xe_device *xe)
@@ -310,11 +320,9 @@ int xe_pm_runtime_resume(struct xe_device *xe)
 	xe->d3cold.power_lost = xe_guc_in_reset(&gt->uc.guc);
 
 	if (xe->d3cold.allowed && xe->d3cold.power_lost) {
-		for_each_gt(gt, xe, id) {
-			err = xe_pcode_init(gt);
-			if (err)
-				goto out;
-		}
+		err = xe_pcode_ready(xe, true);
+		if (err)
+			goto out;
 
 		/*
 		 * This only restores pinned memory which is the memory