Merge tag 'drm-xe-next-2024-10-31' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-next

UAPI Changes:
- Define and parse OA sync properties (Ashutosh)

Driver Changes:
- Add caller info to xe_gt_reset_async (Nirmoy)
- A large forcewake rework / cleanup (Himal)
- A g2h response timeout fix (Badal)
- A PTL workaround (Vinay)
- Handle unreliable MMIO reads during forcewake (Shuicheng)
- Ufence user-space access fixes (Nirmoy)
- Annotate flexible arrays (Matthew Brost)
- Enable GuC lite restore (Fei)
- Prevent GuC register capture on VF (Zhanjun)
- Show VFs VRAM / LMEM provisioning summary over debugfs (Michal)
- Parallel queues fix on GT reset (Nirmoy)
- Move reference grabbing to a job's dma-fence (Matt Brost)
- Mark a number of local workqueues WQ_MEM_RECLAIM (Matt Brost)
- OA synchronization support (Ashutosh)
- Capture all available bits of GuC timestamp to GuC log (John)
- Increase readability of guc_info debugfs (John)
- Add a mmio barrier before GGTT invalidate (Matt Brost)
- Don't short-circuit TDR on jobs not started (Matt Brost)

Signed-off-by: Dave Airlie <[e-mail]>

From: Thomas Hellstrom <[e-mail]>
Link: https://patchwork.freedesktop.org/patch/msgid/ZyNvA_vZZYR-1eWE@fedora

drivers/gpu/drm/xe/abi/guc_klvs_abi.h

···
 	GUC_WORKAROUND_KLV_ID_DISABLE_MTP_DURING_ASYNC_COMPUTE				= 0x9007,
 	GUC_WA_KLV_NP_RD_WRITE_TO_CLEAR_RCSM_AT_CGP_LATE_RESTORE			= 0x9008,
 	GUC_WORKAROUND_KLV_ID_BACK_TO_BACK_RCS_ENGINE_RESET				= 0x9009,
+	GUC_WA_KLV_WAKE_POWER_DOMAINS_FOR_OUTBOUND_MMIO					= 0x900a,
 };
 
 #endif

drivers/gpu/drm/xe/display/xe_hdcp_gsc.c

···
 	struct xe_gt *gt = tile->media_gt;
 	struct xe_gsc *gsc = &gt->uc.gsc;
 	bool ret = true;
+	unsigned int fw_ref;
 
 	if (!gsc && !xe_uc_fw_is_enabled(&gsc->fw)) {
 		drm_dbg_kms(&xe->drm,
···
 	}
 
 	xe_pm_runtime_get(xe);
-	if (xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC)) {
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
+	if (!fw_ref) {
 		drm_dbg_kms(&xe->drm,
 			    "failed to get forcewake to check proxy status\n");
 		ret = false;
···
 	if (!xe_gsc_proxy_init_done(gsc))
 		ret = false;
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GSC);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 out:
 	xe_pm_runtime_put(xe);
 	return ret;

drivers/gpu/drm/xe/regs/xe_guc_regs.h

···
 #define   HUC_LOADING_AGENT_GUC			REG_BIT(1)
 #define   GUC_WOPCM_OFFSET_VALID		REG_BIT(0)
 #define GUC_MAX_IDLE_COUNT			XE_REG(0xc3e4)
-#define GUC_PMTIMESTAMP				XE_REG(0xc3e8)
+#define GUC_PMTIMESTAMP_LO			XE_REG(0xc3e8)
+#define GUC_PMTIMESTAMP_HI			XE_REG(0xc3ec)
 
 #define GUC_SEND_INTERRUPT			XE_REG(0xc4c8)
 #define   GUC_SEND_TRIGGER			REG_BIT(0)

drivers/gpu/drm/xe/tests/xe_mocs.c

···
 {
 	struct kunit *test = kunit_get_current_test();
 	u32 l3cc, l3cc_expected;
-	unsigned int i;
+	unsigned int fw_ref, i;
 	u32 reg_val;
-	u32 ret;
 
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	KUNIT_ASSERT_EQ_MSG(test, ret, 0, "Forcewake Failed.\n");
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	KUNIT_ASSERT_NE_MSG(test, fw_ref, 0, "Forcewake Failed.\n");
 
 	for (i = 0; i < info->num_mocs_regs; i++) {
 		if (!(i & 1)) {
···
 		KUNIT_EXPECT_EQ_MSG(test, l3cc_expected, l3cc,
 				    "l3cc idx=%u has incorrect val.\n", i);
 	}
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static void read_mocs_table(struct xe_gt *gt,
···
 {
 	struct kunit *test = kunit_get_current_test();
 	u32 mocs, mocs_expected;
-	unsigned int i;
+	unsigned int fw_ref, i;
 	u32 reg_val;
-	u32 ret;
 
 	KUNIT_EXPECT_TRUE_MSG(test, info->unused_entries_index,
 			      "Unused entries index should have been defined\n");
 
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	KUNIT_ASSERT_EQ_MSG(test, ret, 0, "Forcewake Failed.\n");
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	KUNIT_ASSERT_NE_MSG(test, fw_ref, 0, "Forcewake Failed.\n");
 
 	for (i = 0; i < info->num_mocs_regs; i++) {
 		if (regs_are_mcr(gt))
···
 				    "mocs reg 0x%x has incorrect val.\n", i);
 	}
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static int mocs_kernel_test_run_device(struct xe_device *xe)

drivers/gpu/drm/xe/xe_debugfs.c

···
 {
 	struct xe_device *xe = inode->i_private;
 	struct xe_gt *gt;
-	u8 id;
+	u8 id, last_gt;
+	unsigned int fw_ref;
 
 	xe_pm_runtime_get(xe);
-	for_each_gt(gt, xe, id)
-		XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	for_each_gt(gt, xe, id) {
+		last_gt = id;
+
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+		if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
+			goto err_fw_get;
+	}
 
 	return 0;
+
+err_fw_get:
+	for_each_gt(gt, xe, id) {
+		if (id < last_gt)
+			xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+		else if (id == last_gt)
+			xe_force_wake_put(gt_to_fw(gt), fw_ref);
+		else
+			break;
+	}
+
+	xe_pm_runtime_put(xe);
+	return -ETIMEDOUT;
 }
 
 static int forcewake_release(struct inode *inode, struct file *file)
···
 	u8 id;
 
 	for_each_gt(gt, xe, id)
-		XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+		xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
 	xe_pm_runtime_put(xe);
 
 	return 0;

drivers/gpu/drm/xe/xe_devcoredump.c

···
 {
 	struct xe_devcoredump_snapshot *ss = container_of(work, typeof(*ss), work);
 	struct xe_devcoredump *coredump = container_of(ss, typeof(*coredump), snapshot);
+	unsigned int fw_ref;
 
 	/* keep going if fw fails as we still want to save the memory and SW data */
-	if (xe_force_wake_get(gt_to_fw(ss->gt), XE_FORCEWAKE_ALL))
+	fw_ref = xe_force_wake_get(gt_to_fw(ss->gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 		xe_gt_info(ss->gt, "failed to get forcewake for coredump capture\n");
 	xe_vm_snapshot_capture_delayed(ss->vm);
 	xe_guc_exec_queue_snapshot_capture_delayed(ss->ge);
-	xe_force_wake_put(gt_to_fw(ss->gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(ss->gt), fw_ref);
 
 	/* Calculate devcoredump size */
 	ss->read.size = __xe_devcoredump_read(NULL, INT_MAX, coredump);
···
 	u32 width_mask = (0x1 << q->width) - 1;
 	const char *process_name = "no process";
 
-	int i;
+	unsigned int fw_ref;
 	bool cookie;
+	int i;
 
 	ss->snapshot_time = ktime_get_real();
 	ss->boot_time = ktime_get_boottime();
···
 	}
 
 	/* keep going if fw fails as we still want to save the memory and SW data */
-	if (xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL))
-		xe_gt_info(ss->gt, "failed to get forcewake for coredump capture\n");
+	fw_ref = xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL);
 
 	ss->guc.log = xe_guc_log_snapshot_capture(&guc->log, true);
-	ss->guc.ct = xe_guc_ct_snapshot_capture(&guc->ct, true);
+	ss->guc.ct = xe_guc_ct_snapshot_capture(&guc->ct);
 	ss->ge = xe_guc_exec_queue_snapshot_capture(q);
 	ss->job = xe_sched_job_snapshot_capture(job);
 	ss->vm = xe_vm_snapshot_capture(q->vm);
···
 
 	queue_work(system_unbound_wq, &ss->work);
 
-	xe_force_wake_put(gt_to_fw(q->gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(q->gt), fw_ref);
 	dma_fence_end_signalling(cookie);
 }
 

drivers/gpu/drm/xe/xe_device.c

···
 static int probe_has_flat_ccs(struct xe_device *xe)
 {
 	struct xe_gt *gt;
+	unsigned int fw_ref;
 	u32 reg;
-	int err;
 
 	/* Always enabled/disabled, no runtime check to do */
 	if (GRAPHICS_VER(xe) < 20 || !xe->info.has_flat_ccs)
···
 
 	gt = xe_root_mmio_gt(xe);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		return err;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return -ETIMEDOUT;
 
 	reg = xe_gt_mcr_unicast_read_any(gt, XE2_FLAT_CCS_BASE_RANGE_LOWER);
 	xe->info.has_flat_ccs = (reg & XE2_FLAT_CCS_ENABLE);
···
 		drm_dbg(&xe->drm,
 			"Flat CCS has been disabled in bios, May lead to performance impact");
 
-	return xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
+	return 0;
 }
 
 int xe_device_probe(struct xe_device *xe)
···
 void xe_device_td_flush(struct xe_device *xe)
 {
 	struct xe_gt *gt;
+	unsigned int fw_ref;
 	u8 id;
 
 	if (!IS_DGFX(xe) || GRAPHICS_VER(xe) < 20)
···
 		if (xe_gt_is_media_type(gt))
 			continue;
 
-		if (xe_force_wake_get(gt_to_fw(gt), XE_FW_GT))
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+		if (!fw_ref)
 			return;
 
 		xe_mmio_write32(&gt->mmio, XE2_TDF_CTRL, TRANSIENT_FLUSH_REQUEST);
···
 				   150, NULL, false))
 			xe_gt_err_once(gt, "TD flush timeout\n");
 
-		xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	}
 }
 
 void xe_device_l2_flush(struct xe_device *xe)
 {
 	struct xe_gt *gt;
-	int err;
+	unsigned int fw_ref;
 
 	gt = xe_root_mmio_gt(xe);
 
 	if (!XE_WA(gt, 16023588340))
 		return;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		return;
 
 	spin_lock(&gt->global_invl_lock);
···
 		xe_gt_err_once(gt, "Global invalidation timeout\n");
 	spin_unlock(&gt->global_invl_lock);
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 u32 xe_device_ccs_bytes(struct xe_device *xe, u64 size)

drivers/gpu/drm/xe/xe_drm_client.c

···
 	struct xe_hw_engine *hwe;
 	struct xe_exec_queue *q;
 	u64 gpu_timestamp;
+	unsigned int fw_ref;
 
 	xe_pm_runtime_get(xe);
 
···
 			continue;
 
 		fw = xe_hw_engine_to_fw_domain(hwe);
-		if (xe_force_wake_get(gt_to_fw(gt), fw)) {
+
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), fw);
+		if (!xe_force_wake_ref_has_domain(fw_ref, fw)) {
 			hwe = NULL;
+			xe_force_wake_put(gt_to_fw(gt), fw_ref);
 			break;
 		}
 
 		gpu_timestamp = xe_hw_engine_read_timestamp(hwe);
-		XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), fw));
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 		break;
 	}
 

drivers/gpu/drm/xe/xe_exec_queue_types.h

···
 	/** @hw_engine_group_link: link into exec queues in the same hw engine group */
 	struct list_head hw_engine_group_link;
 	/** @lrc: logical ring context for this exec queue */
-	struct xe_lrc *lrc[];
+	struct xe_lrc *lrc[] __counted_by(width);
 };
 
 /**

drivers/gpu/drm/xe/xe_execlist.c

···
 	q->ring_ops->emit_job(job);
 	xe_execlist_make_active(exl);
 
-	return dma_fence_get(job->fence);
+	return job->fence;
 }
 
 static void execlist_job_free(struct drm_sched_job *drm_job)

drivers/gpu/drm/xe/xe_force_wake.c

···
 	return wake ? "wake" : "sleep";
 }
 
-static void domain_init(struct xe_force_wake_domain *domain,
+static void mark_domain_initialized(struct xe_force_wake *fw,
+				    enum xe_force_wake_domain_id id)
+{
+	fw->initialized_domains |= BIT(id);
+}
+
+static void init_domain(struct xe_force_wake *fw,
 			enum xe_force_wake_domain_id id,
 			struct xe_reg reg, struct xe_reg ack)
 {
+	struct xe_force_wake_domain *domain = &fw->domains[id];
+
 	domain->id = id;
 	domain->reg_ctl = reg;
 	domain->reg_ack = ack;
 	domain->val = FORCEWAKE_MT(FORCEWAKE_KERNEL);
 	domain->mask = FORCEWAKE_MT_MASK(FORCEWAKE_KERNEL);
+
+	mark_domain_initialized(fw, id);
 }
 
 void xe_force_wake_init_gt(struct xe_gt *gt, struct xe_force_wake *fw)
···
 	xe_gt_assert(gt, GRAPHICS_VER(gt_to_xe(gt)) >= 11);
 
 	if (xe->info.graphics_verx100 >= 1270) {
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_GT],
-			    XE_FW_DOMAIN_ID_GT,
+		init_domain(fw, XE_FW_DOMAIN_ID_GT,
 			    FORCEWAKE_GT,
 			    FORCEWAKE_ACK_GT_MTL);
 	} else {
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_GT],
-			    XE_FW_DOMAIN_ID_GT,
+		init_domain(fw, XE_FW_DOMAIN_ID_GT,
 			    FORCEWAKE_GT,
 			    FORCEWAKE_ACK_GT);
 	}
···
 	xe_gt_assert(gt, GRAPHICS_VER(gt_to_xe(gt)) >= 11);
 
 	if (!xe_gt_is_media_type(gt))
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_RENDER],
-			    XE_FW_DOMAIN_ID_RENDER,
+		init_domain(fw, XE_FW_DOMAIN_ID_RENDER,
 			    FORCEWAKE_RENDER,
 			    FORCEWAKE_ACK_RENDER);
 
···
 		if (!(gt->info.engine_mask & BIT(i)))
 			continue;
 
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_MEDIA_VDBOX0 + j],
-			    XE_FW_DOMAIN_ID_MEDIA_VDBOX0 + j,
+		init_domain(fw, XE_FW_DOMAIN_ID_MEDIA_VDBOX0 + j,
 			    FORCEWAKE_MEDIA_VDBOX(j),
 			    FORCEWAKE_ACK_MEDIA_VDBOX(j));
 	}
···
 		if (!(gt->info.engine_mask & BIT(i)))
 			continue;
 
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_MEDIA_VEBOX0 + j],
-			    XE_FW_DOMAIN_ID_MEDIA_VEBOX0 + j,
+		init_domain(fw, XE_FW_DOMAIN_ID_MEDIA_VEBOX0 + j,
 			    FORCEWAKE_MEDIA_VEBOX(j),
 			    FORCEWAKE_ACK_MEDIA_VEBOX(j));
 	}
 
 	if (gt->info.engine_mask & BIT(XE_HW_ENGINE_GSCCS0))
-		domain_init(&fw->domains[XE_FW_DOMAIN_ID_GSC],
-			    XE_FW_DOMAIN_ID_GSC,
+		init_domain(fw, XE_FW_DOMAIN_ID_GSC,
 			    FORCEWAKE_GSC,
 			    FORCEWAKE_ACK_GSC);
 }
···
 					 (ffs(tmp__) - 1))) && \
 					 domain__->reg_ctl.addr)
 
-int xe_force_wake_get(struct xe_force_wake *fw,
-		      enum xe_force_wake_domains domains)
+/**
+ * xe_force_wake_get() : Increase the domain refcount
+ * @fw: struct xe_force_wake
+ * @domains: forcewake domains to get refcount on
+ *
+ * This function wakes up @domains if they are asleep and takes references.
+ * If requested domain is XE_FORCEWAKE_ALL then only applicable/initialized
+ * domains will be considered for refcount and it is a caller responsibility
+ * to check returned ref if it includes any specific domain by using
+ * xe_force_wake_ref_has_domain() function. Caller must call
+ * xe_force_wake_put() function to decrease incremented refcounts.
+ *
+ * Return: opaque reference to woken domains or zero if none of requested
+ * domains were awake.
+ */
+unsigned int __must_check xe_force_wake_get(struct xe_force_wake *fw,
+					    enum xe_force_wake_domains domains)
 {
 	struct xe_gt *gt = fw->gt;
 	struct xe_force_wake_domain *domain;
-	enum xe_force_wake_domains tmp, woken = 0;
+	unsigned int ref_incr = 0, awake_rqst = 0, awake_failed = 0;
+	unsigned int tmp, ref_rqst;
 	unsigned long flags;
-	int ret = 0;
 
+	xe_gt_assert(gt, is_power_of_2(domains));
+	xe_gt_assert(gt, domains <= XE_FORCEWAKE_ALL);
+	xe_gt_assert(gt, domains == XE_FORCEWAKE_ALL || fw->initialized_domains & domains);
+
+	ref_rqst = (domains == XE_FORCEWAKE_ALL) ? fw->initialized_domains : domains;
 	spin_lock_irqsave(&fw->lock, flags);
-	for_each_fw_domain_masked(domain, domains, fw, tmp) {
+	for_each_fw_domain_masked(domain, ref_rqst, fw, tmp) {
 		if (!domain->ref++) {
-			woken |= BIT(domain->id);
+			awake_rqst |= BIT(domain->id);
 			domain_wake(gt, domain);
 		}
+		ref_incr |= BIT(domain->id);
 	}
-	for_each_fw_domain_masked(domain, woken, fw, tmp) {
-		ret |= domain_wake_wait(gt, domain);
+	for_each_fw_domain_masked(domain, awake_rqst, fw, tmp) {
+		if (domain_wake_wait(gt, domain) == 0) {
+			fw->awake_domains |= BIT(domain->id);
+		} else {
+			awake_failed |= BIT(domain->id);
+			--domain->ref;
+		}
 	}
-	fw->awake_domains |= woken;
+	ref_incr &= ~awake_failed;
 	spin_unlock_irqrestore(&fw->lock, flags);
 
-	return ret;
+	xe_gt_WARN(gt, awake_failed, "Forcewake domain%s %#x failed to acknowledge awake request\n",
+		   str_plural(hweight_long(awake_failed)), awake_failed);
+
+	if (domains == XE_FORCEWAKE_ALL && ref_incr == fw->initialized_domains)
+		ref_incr |= XE_FORCEWAKE_ALL;
+
+	return ref_incr;
 }
 
-int xe_force_wake_put(struct xe_force_wake *fw,
-		      enum xe_force_wake_domains domains)
+/**
+ * xe_force_wake_put - Decrement the refcount and put domain to sleep if refcount becomes 0
+ * @fw: Pointer to the force wake structure
+ * @fw_ref: return of xe_force_wake_get()
+ *
+ * This function reduces the reference counts for domains in fw_ref. If
+ * refcount for any of the specified domain reaches 0, it puts the domain to sleep
+ * and waits for acknowledgment for domain to sleep within 50 milisec timeout.
+ * Warns in case of timeout of ack from domain.
+ */
+void xe_force_wake_put(struct xe_force_wake *fw, unsigned int fw_ref)
 {
 	struct xe_gt *gt = fw->gt;
 	struct xe_force_wake_domain *domain;
-	enum xe_force_wake_domains tmp, sleep = 0;
+	unsigned int tmp, sleep = 0;
 	unsigned long flags;
-	int ret = 0;
+	int ack_fail = 0;
+
+	/*
+	 * Avoid unnecessary lock and unlock when the function is called
+	 * in error path of individual domains.
+	 */
+	if (!fw_ref)
+		return;
+
+	if (xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
+		fw_ref = fw->initialized_domains;
 
 	spin_lock_irqsave(&fw->lock, flags);
-	for_each_fw_domain_masked(domain, domains, fw, tmp) {
+	for_each_fw_domain_masked(domain, fw_ref, fw, tmp) {
+		xe_gt_assert(gt, domain->ref);
+
 		if (!--domain->ref) {
 			sleep |= BIT(domain->id);
 			domain_sleep(gt, domain);
 		}
 	}
 	for_each_fw_domain_masked(domain, sleep, fw, tmp) {
-		ret |= domain_sleep_wait(gt, domain);
+		if (domain_sleep_wait(gt, domain) == 0)
+			fw->awake_domains &= ~BIT(domain->id);
+		else
+			ack_fail |= BIT(domain->id);
 	}
-	fw->awake_domains &= ~sleep;
 	spin_unlock_irqrestore(&fw->lock, flags);
 
-	return ret;
+	xe_gt_WARN(gt, ack_fail, "Forcewake domain%s %#x failed to acknowledge sleep request\n",
+		   str_plural(hweight_long(ack_fail)), ack_fail);
 }

drivers/gpu/drm/xe/xe_force_wake.h

···
 			   struct xe_force_wake *fw);
 void xe_force_wake_init_engines(struct xe_gt *gt,
 				struct xe_force_wake *fw);
-int xe_force_wake_get(struct xe_force_wake *fw,
-		      enum xe_force_wake_domains domains);
-int xe_force_wake_put(struct xe_force_wake *fw,
-		      enum xe_force_wake_domains domains);
+unsigned int __must_check xe_force_wake_get(struct xe_force_wake *fw,
+					    enum xe_force_wake_domains domains);
+void xe_force_wake_put(struct xe_force_wake *fw, unsigned int fw_ref);
 
 static inline int
 xe_force_wake_ref(struct xe_force_wake *fw,
···
 {
 	xe_gt_assert(fw->gt, domain != XE_FORCEWAKE_ALL);
 	xe_gt_assert(fw->gt, fw->awake_domains & domain);
+}
+
+/**
+ * xe_force_wake_ref_has_domain - verifies if the domains are in fw_ref
+ * @fw_ref : the force_wake reference
+ * @domain : forcewake domain to verify
+ *
+ * This function confirms whether the @fw_ref includes a reference to the
+ * specified @domain.
+ *
+ * Return: true if domain is refcounted.
+ */
+static inline bool
+xe_force_wake_ref_has_domain(unsigned int fw_ref, enum xe_force_wake_domains domain)
+{
+	return fw_ref & domain;
 }
 
 #endif

drivers/gpu/drm/xe/xe_force_wake_types.h

···
 	XE_FW_MEDIA_VEBOX2	= BIT(XE_FW_DOMAIN_ID_MEDIA_VEBOX2),
 	XE_FW_MEDIA_VEBOX3	= BIT(XE_FW_DOMAIN_ID_MEDIA_VEBOX3),
 	XE_FW_GSC		= BIT(XE_FW_DOMAIN_ID_GSC),
-	XE_FORCEWAKE_ALL	= BIT(XE_FW_DOMAIN_ID_COUNT) - 1
+	XE_FORCEWAKE_ALL	= BIT(XE_FW_DOMAIN_ID_COUNT)
 };
 
 /**
···
 	/** @lock: protects everything force wake struct */
 	spinlock_t lock;
 	/** @awake_domains: mask of all domains awake */
-	enum xe_force_wake_domains awake_domains;
+	unsigned int awake_domains;
+	/** @initialized_domains: mask of all initialized domains */
+	unsigned int initialized_domains;
 	/** @domains: force wake domains */
 	struct xe_force_wake_domain domains[XE_FW_DOMAIN_ID_COUNT];
 };

drivers/gpu/drm/xe/xe_ggtt.c

···
 	else
 		ggtt->pt_ops = &xelp_pt_ops;
 
-	ggtt->wq = alloc_workqueue("xe-ggtt-wq", 0, 0);
+	ggtt->wq = alloc_workqueue("xe-ggtt-wq", 0, WQ_MEM_RECLAIM);
 
 	drm_mm_init(&ggtt->mm, xe_wopcm_size(xe),
 		    ggtt->size - xe_wopcm_size(xe));
···
 	 * vs. correct GGTT page. Not particularly a hot code path so blindly
 	 * do a mmio read here which results in GuC reading correct GGTT page.
 	 */
-	xe_mmio_read32(&xe_root_mmio_gt(xe)->mmio, VF_CAP_REG);
+	xe_mmio_read32(xe_root_tile_mmio(xe), VF_CAP_REG);
 
 	/* Each GT in a tile has its own TLB to cache GGTT lookups */
 	ggtt_invalidate_gt_tlb(ggtt->tile->primary_gt);

drivers/gpu/drm/xe/xe_gsc.c

···
 {
 	struct xe_gt *gt = gsc_to_gt(gsc);
 	struct xe_tile *tile = gt_to_tile(gt);
+	unsigned int fw_ref;
 	int ret;
 
 	if (XE_WA(tile->primary_gt, 14018094691)) {
-		ret = xe_force_wake_get(gt_to_fw(tile->primary_gt), XE_FORCEWAKE_ALL);
+		fw_ref = xe_force_wake_get(gt_to_fw(tile->primary_gt), XE_FORCEWAKE_ALL);
 
 		/*
 		 * If the forcewake fails we want to keep going, because the worst
 		 * case outcome in failing to apply the WA is that PXP won't work,
-		 * which is not fatal. We still throw a warning so the issue is
-		 * seen if it happens.
+		 * which is not fatal. Forcewake get warns implicitly in case of failure
 		 */
-		xe_gt_WARN_ON(tile->primary_gt, ret);
-
 		xe_gt_mcr_multicast_write(tile->primary_gt,
 					  EU_SYSTOLIC_LIC_THROTTLE_CTL_WITH_LOCK,
 					  EU_SYSTOLIC_LIC_THROTTLE_CTL_LOCK_BIT);
···
 	ret = gsc_upload(gsc);
 
 	if (XE_WA(tile->primary_gt, 14018094691))
-		xe_force_wake_put(gt_to_fw(tile->primary_gt), XE_FORCEWAKE_ALL);
+		xe_force_wake_put(gt_to_fw(tile->primary_gt), fw_ref);
 
 	if (ret)
 		return ret;
···
 	struct xe_gsc *gsc = container_of(work, typeof(*gsc), work);
 	struct xe_gt *gt = gsc_to_gt(gsc);
 	struct xe_device *xe = gt_to_xe(gt);
+	unsigned int fw_ref;
 	u32 actions;
 	int ret;
 
···
 	spin_unlock_irq(&gsc->lock);
 
 	xe_pm_runtime_get(xe);
-	xe_gt_WARN_ON(gt, xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC));
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
 
 	if (actions & GSC_ACTION_ER_COMPLETE) {
 		ret = gsc_er_complete(gt);
···
 		xe_gsc_proxy_request_handler(gsc);
 
 out:
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GSC);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_pm_runtime_put(xe);
 }
 
···
 {
 	struct xe_gt *gt = gsc_to_gt(gsc);
 	struct xe_mmio *mmio = &gt->mmio;
-	int err;
+	unsigned int fw_ref;
 
 	xe_uc_fw_print(&gsc->fw, p);
 
···
 	if (!xe_uc_fw_is_enabled(&gsc->fw))
 		return;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
+	if (!fw_ref)
 		return;
 
 	drm_printf(p, "\nHECI1 FWSTS: 0x%08x 0x%08x 0x%08x 0x%08x 0x%08x 0x%08x\n",
···
 			xe_mmio_read32(mmio, HECI_FWSTS5(MTL_GSC_HECI1_BASE)),
 			xe_mmio_read32(mmio, HECI_FWSTS6(MTL_GSC_HECI1_BASE)));
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GSC);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }

drivers/gpu/drm/xe/xe_gsc_proxy.c

···
 {
 	struct xe_gt *gt = gsc_to_gt(gsc);
 	struct xe_device *xe = gt_to_xe(gt);
-	int err = 0;
+	unsigned int fw_ref = 0;
 
 	if (!gsc->proxy.component_added)
 		return;
 
 	/* disable HECI2 IRQs */
 	xe_pm_runtime_get(xe);
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GSC);
+	if (!fw_ref)
 		xe_gt_err(gt, "failed to get forcewake to disable GSC interrupts\n");
 
 	/* try do disable irq even if forcewake failed */
 	gsc_proxy_irq_toggle(gsc, false);
 
-	if (!err)
-		xe_force_wake_put(gt_to_fw(gt), XE_FW_GSC);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_pm_runtime_put(xe);
 
 	xe_gsc_wait_for_worker_completion(gsc);

drivers/gpu/drm/xe/xe_gt.c

···
 		return ERR_PTR(-ENOMEM);
 
 	gt->tile = tile;
-	gt->ordered_wq = alloc_ordered_workqueue("gt-ordered-wq", 0);
+	gt->ordered_wq = alloc_ordered_workqueue("gt-ordered-wq",
+						 WQ_MEM_RECLAIM);
 
 	err = drmm_add_action_or_reset(&gt_to_xe(gt)->drm, gt_fini, gt);
 	if (err)
···
 
 static void xe_gt_enable_host_l2_vram(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	u32 reg;
-	int err;
 
 	if (!XE_WA(gt, 16023588340))
 		return;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (WARN_ON(err))
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		return;
 
 	if (!xe_gt_is_media_type(gt)) {
···
 	}
 
 	xe_gt_mcr_multicast_write(gt, XEHPC_L3CLOS_MASK(3), 0x3);
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static void xe_gt_disable_host_l2_vram(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	u32 reg;
-	int err;
 
 	if (!XE_WA(gt, 16023588340))
 		return;
···
 	if (xe_gt_is_media_type(gt))
 		return;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (WARN_ON(err))
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		return;
 
 	reg = xe_gt_mcr_unicast_read_any(gt, XE2_GAMREQSTRM_CTRL);
 	reg &= ~CG_DIS_CNTLBUS;
 	xe_gt_mcr_multicast_write(gt, XE2_GAMREQSTRM_CTRL, reg);
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 /**
···
 
 static int gt_fw_domain_init(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err, i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref) {
+		err = -ETIMEDOUT;
 		goto err_hw_fence_irq;
+	}
 
 	if (!xe_gt_is_media_type(gt)) {
 		err = xe_ggtt_init(gt_to_tile(gt)->mem.ggtt);
···
 	 */
 	gt->info.gmdid = xe_mmio_read32(&gt->mmio, GMD_ID);
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-	XE_WARN_ON(err);
-
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	return 0;
 
 err_force_wake:
 	dump_pat_on_error(gt);
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 err_hw_fence_irq:
 	for (i = 0; i < XE_ENGINE_CLASS_MAX; ++i)
 		xe_hw_fence_irq_finish(&gt->fence_irq[i]);
···
 
 static int all_fw_domain_init(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err, i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
-		goto err_hw_fence_irq;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		err = -ETIMEDOUT;
+		goto err_force_wake;
+	}
 
 	xe_gt_mcr_set_implicit_defaults(gt);
 	xe_reg_sr_apply_mmio(&gt->reg_sr, gt);
···
 	if (IS_SRIOV_PF(gt_to_xe(gt)))
 		xe_gt_sriov_pf_init_hw(gt);
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	XE_WARN_ON(err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
 	return 0;
 
 err_force_wake:
-	xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-err_hw_fence_irq:
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	for (i = 0; i < XE_ENGINE_CLASS_MAX; ++i)
 		xe_hw_fence_irq_finish(&gt->fence_irq[i]);
 
···
  */
 int xe_gt_init_hwconfig(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto out;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return -ETIMEDOUT;
 
 	xe_gt_mcr_init_early(gt);
 	xe_pat_init(gt);
···
 	xe_gt_enable_host_l2_vram(gt);
 
 out_fw:
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-out:
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	return err;
 }
 
···
 
 static int gt_reset(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err;
 
 	if (xe_device_wedged(gt_to_xe(gt)))
···
 
 	xe_gt_sanitize(gt);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
-		goto err_msg;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		err = -ETIMEDOUT;
+		goto err_out;
+	}
 
 	xe_uc_gucrc_disable(&gt->uc);
 	xe_uc_stop_prepare(&gt->uc);
···
 	if (err)
 		goto err_out;
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	XE_WARN_ON(err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_pm_runtime_put(gt_to_xe(gt));
 
 	xe_gt_info(gt, "reset done\n");
···
 	return 0;
 
 err_out:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
-err_msg:
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	XE_WARN_ON(xe_uc_start(&gt->uc));
 err_fail:
 	xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err));
···
 
 void xe_gt_reset_async(struct xe_gt *gt)
 {
-	xe_gt_info(gt, "trying reset\n");
+	xe_gt_info(gt, "trying reset from %ps\n", __builtin_return_address(0));
 
 	/* Don't do a reset while one is already in flight */
 	if (!xe_fault_inject_gt_reset() && xe_uc_reset_prepare(&gt->uc))
···
 
 void xe_gt_suspend_prepare(struct xe_gt *gt)
 {
-	XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	unsigned int fw_ref;
+
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
 
 	xe_uc_stop_prepare(&gt->uc);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 int xe_gt_suspend(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err;
 
 	xe_gt_dbg(gt, "suspending\n");
 	xe_gt_sanitize(gt);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 		goto err_msg;
 
 	err = xe_uc_suspend(&gt->uc);
···
 
 	xe_gt_disable_host_l2_vram(gt);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_gt_dbg(gt, "suspended\n");
 
 	return 0;
 
-err_force_wake:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
 err_msg:
+	err = -ETIMEDOUT;
+err_force_wake:
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_gt_err(gt, "suspend failed (%pe)\n", ERR_PTR(err));
 
 	return err;
···
 
 void xe_gt_shutdown(struct xe_gt *gt)
 {
-	xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	unsigned int fw_ref;
+
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
 	do_gt_reset(gt);
-	xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 /**
···
 
 int xe_gt_resume(struct xe_gt *gt)
 {
+	unsigned int fw_ref;
 	int err;
 
 	xe_gt_dbg(gt, "resuming\n");
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 		goto err_msg;
 
 	err = do_gt_restart(gt);
···
 
 	xe_gt_idle_enable_pg(gt);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_gt_dbg(gt, "resumed\n");
 
 	return 0;
 
-err_force_wake:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
 err_msg:
+	err = -ETIMEDOUT;
+err_force_wake:
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_gt_err(gt, "resume failed (%pe)\n", ERR_PTR(err));
 
 	return err;

drivers/gpu/drm/xe/xe_gt_debugfs.c

···
 	struct xe_device *xe = gt_to_xe(gt);
 	struct xe_hw_engine *hwe;
 	enum xe_hw_engine_id id;
-	int err;
+	unsigned int fw_ref;
 
 	xe_pm_runtime_get(xe);
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err) {
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
 		xe_pm_runtime_put(xe);
-		return err;
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
+		return -ETIMEDOUT;
 	}
 
 	for_each_hw_engine(hwe, gt, id)
 		xe_hw_engine_print(hwe, p);
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_pm_runtime_put(xe);
-	if (err)
-		return err;
 
 	return 0;
 }

drivers/gpu/drm/xe/xe_gt_idle.c

···
 	struct xe_gt_idle *gtidle = &gt->gtidle;
 	struct xe_mmio *mmio = &gt->mmio;
 	u32 vcs_mask, vecs_mask;
+	unsigned int fw_ref;
 	int i, j;
 
 	if (IS_SRIOV_VF(xe))
···
 						     VDN_MFXVDENC_POWERGATE_ENABLE(j));
 	}
 
-	XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FW_GT));
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
 	if (xe->info.skip_guc_pc) {
 		/*
 		 * GuC sets the hysteresis value when GuC PC is enabled
···
 	}
 
 	xe_mmio_write32(mmio, POWERGATE_ENABLE, gtidle->powergate_enable);
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FW_GT));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 void xe_gt_idle_disable_pg(struct xe_gt *gt)
 {
 	struct xe_gt_idle *gtidle = &gt->gtidle;
+	unsigned int fw_ref;
 
 	if (IS_SRIOV_VF(gt_to_xe(gt)))
 		return;
···
 	xe_device_assert_mem_access(gt_to_xe(gt));
 	gtidle->powergate_enable = 0;
 
-	XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FW_GT));
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
 	xe_mmio_write32(&gt->mmio, POWERGATE_ENABLE, gtidle->powergate_enable);
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FW_GT));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 /**
···
 	enum xe_gt_idle_state state;
 	u32 pg_enabled, pg_status = 0;
 	u32 vcs_mask, vecs_mask;
-	int err, n;
+	unsigned int fw_ref;
+	int n;
 	/*
 	 * Media Slices
 	 *
···
 
 	/* Do not wake the GT to read powergating status */
 	if (state != GT_IDLE_C6) {
-		err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-		if (err)
-			return err;
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+		if (!fw_ref)
+			return -ETIMEDOUT;
 
 		pg_enabled = xe_mmio_read32(&gt->mmio, POWERGATE_ENABLE);
 		pg_status = xe_mmio_read32(&gt->mmio, POWERGATE_DOMAIN_STATUS);
 
-		XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FW_GT));
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	}
 
 	if (gt->info.engine_mask & XE_HW_ENGINE_RCS_MASK) {
···
 {
 	struct kobject *kobj = arg;
 	struct xe_gt *gt = kobj_to_gt(kobj->parent);
+	unsigned int fw_ref;
 
 	xe_gt_idle_disable_pg(gt);
 
 	if (gt_to_xe(gt)->info.skip_guc_pc) {
-		XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FW_GT));
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
 		xe_gt_idle_disable_c6(gt);
-		xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	}
 
 	sysfs_remove_files(kobj, gt_idle_attrs);

drivers/gpu/drm/xe/xe_gt_sriov_pf_config.c

···
 }
 
 /**
+ * xe_gt_sriov_pf_config_print_lmem - Print LMEM configurations.
+ * @gt: the &xe_gt
+ * @p: the &drm_printer
+ *
+ * Print LMEM allocations across all VFs.
+ * VFs without LMEM allocation are skipped.
+ *
+ * This function can only be called on PF.
+ * Return: 0 on success or a negative error code on failure.
+ */
+int xe_gt_sriov_pf_config_print_lmem(struct xe_gt *gt, struct drm_printer *p)
+{
+	unsigned int n, total_vfs = xe_sriov_pf_get_totalvfs(gt_to_xe(gt));
+	const struct xe_gt_sriov_config *config;
+	char buf[10];
+
+	xe_gt_assert(gt, IS_SRIOV_PF(gt_to_xe(gt)));
+	mutex_lock(xe_gt_sriov_pf_master_mutex(gt));
+
+	for (n = 1; n <= total_vfs; n++) {
+		config = &gt->sriov.pf.vfs[n].config;
+		if (!config->lmem_obj)
+			continue;
+
+		string_get_size(config->lmem_obj->size, 1, STRING_UNITS_2,
+				buf, sizeof(buf));
+		drm_printf(p, "VF%u:\t%zu\t(%s)\n",
+			   n, config->lmem_obj->size, buf);
+	}
+
+	mutex_unlock(xe_gt_sriov_pf_master_mutex(gt));
+	return 0;
+}
+
+/**
  * xe_gt_sriov_pf_config_print_available_ggtt - Print available GGTT ranges.
  * @gt: the &xe_gt
  * @p: the &drm_printer

drivers/gpu/drm/xe/xe_gt_sriov_pf_config.h

···
 int xe_gt_sriov_pf_config_print_ggtt(struct xe_gt *gt, struct drm_printer *p);
 int xe_gt_sriov_pf_config_print_ctxs(struct xe_gt *gt, struct drm_printer *p);
 int xe_gt_sriov_pf_config_print_dbs(struct xe_gt *gt, struct drm_printer *p);
+int xe_gt_sriov_pf_config_print_lmem(struct xe_gt *gt, struct drm_printer *p);
 
 int xe_gt_sriov_pf_config_print_available_ggtt(struct xe_gt *gt, struct drm_printer *p);
 

drivers/gpu/drm/xe/xe_gt_sriov_pf_debugfs.c

···
 		.data = xe_gt_sriov_pf_config_print_dbs,
 	},
 	{
+		"lmem_provisioned",
+		.show = xe_gt_debugfs_simple_show,
+		.data = xe_gt_sriov_pf_config_print_lmem,
+	},
+	{
 		"runtime_registers",
 		.show = xe_gt_debugfs_simple_show,
 		.data = xe_gt_sriov_pf_service_print_runtime,

drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c

···
 int xe_gt_tlb_invalidation_ggtt(struct xe_gt *gt)
 {
 	struct xe_device *xe = gt_to_xe(gt);
+	unsigned int fw_ref;
 
 	if (xe_guc_ct_enabled(&gt->uc.guc.ct) &&
 	    gt->uc.guc.submission_state.enabled) {
···
 		if (IS_SRIOV_VF(xe))
 			return 0;
 
-		xe_gt_WARN_ON(gt, xe_force_wake_get(gt_to_fw(gt), XE_FW_GT));
+		fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
 		if (xe->info.platform == XE_PVC || GRAPHICS_VER(xe) >= 20) {
 			xe_mmio_write32(mmio, PVC_GUC_TLB_INV_DESC1,
 					PVC_GUC_TLB_INV_DESC1_INVALIDATE);
···
 			xe_mmio_write32(mmio, GUC_TLB_INV_CR,
 					GUC_TLB_INV_CR_INVALIDATE);
 		}
-		xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	}
 
 	return 0;

drivers/gpu/drm/xe/xe_guc.c

···
 
 static u32 guc_ctl_feature_flags(struct xe_guc *guc)
 {
-	u32 flags = 0;
+	u32 flags = GUC_CTL_ENABLE_LITE_RESTORE;
 
 	if (!guc_to_xe(guc)->info.skip_guc_pc)
 		flags |= GUC_CTL_ENABLE_SLPC;
···
 {
 	struct xe_guc *guc = arg;
 	struct xe_gt *gt = guc_to_gt(guc);
+	unsigned int fw_ref;
 
-	xe_gt_WARN_ON(gt, xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
 	xe_uc_fini_hw(&guc_to_gt(guc)->uc);
-	xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 /**
···
 void xe_guc_print_info(struct xe_guc *guc, struct drm_printer *p)
 {
 	struct xe_gt *gt = guc_to_gt(guc);
+	unsigned int fw_ref;
 	u32 status;
-	int err;
 	int i;
 
 	xe_uc_fw_print(&guc->fw, p);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		return;
 
 	status = xe_mmio_read32(&gt->mmio, GUC_STATUS);
···
 			   i, xe_mmio_read32(&gt->mmio, SOFT_SCRATCH(i)));
 	}
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
-	xe_guc_ct_print(&guc->ct, p);
+	drm_puts(p, "\n");
+	xe_guc_ct_print(&guc->ct, p, false);
+
+	drm_puts(p, "\n");
 	xe_guc_submit_print(guc, p);
 }
 

drivers/gpu/drm/xe/xe_guc_ads.c

···
 					GUC_WORKAROUND_KLV_ID_DISABLE_MTP_DURING_ASYNC_COMPUTE,
 					&offset, &remain);
 
+	if (XE_WA(gt, 14022866841))
+		guc_waklv_enable_simple(ads,
+					GUC_WA_KLV_WAKE_POWER_DOMAINS_FOR_OUTBOUND_MMIO,
+					&offset, &remain);
+
 	/*
 	 * On RC6 exit, GuC will write register 0xB04 with the default value provided. As of now,
 	 * the default value for this register is determined to be 0xC40. This could change in the

drivers/gpu/drm/xe/xe_guc_capture.c

···
 	u16 guc_id = 0;
 	u32 lrca = 0;
 
+	if (IS_SRIOV_VF(xe))
+		return;
+
 	new = guc_capture_get_prealloc_node(guc);
 	if (!new)
 		return;
···
 		return NULL;
 
 	xe = gt_to_xe(q->gt);
-	if (xe->wedged.mode >= 2 || !xe_device_uc_enabled(xe))
+	if (xe->wedged.mode >= 2 || !xe_device_uc_enabled(xe) || IS_SRIOV_VF(xe))
 		return NULL;
 
 	ss = &xe->devcoredump.snapshot;
···
 	struct xe_hw_engine *hwe;
 	enum xe_hw_engine_id id;
 	u32 adj_logical_mask = q->logical_mask;
+
+	if (IS_SRIOV_VF(xe))
+		return;
 
 	for_each_hw_engine(hwe, q->gt, id) {
 		if (hwe->class != q->hwe->class ||

drivers/gpu/drm/xe/xe_guc_ct.c

···
 
 	xe_gt_assert(gt, !(guc_ct_size() % PAGE_SIZE));
 
-	ct->g2h_wq = alloc_ordered_workqueue("xe-g2h-wq", 0);
+	ct->g2h_wq = alloc_ordered_workqueue("xe-g2h-wq", WQ_MEM_RECLAIM);
 	if (!ct->g2h_wq)
 		return -ENOMEM;
 
···
 	receive_g2h(ct);
 }
 
-struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_alloc(struct xe_guc_ct *ct, bool atomic)
+static struct xe_guc_ct_snapshot *guc_ct_snapshot_alloc(struct xe_guc_ct *ct, bool atomic,
+							bool want_ctb)
 {
 	struct xe_guc_ct_snapshot *snapshot;
 
···
 	if (!snapshot)
 		return NULL;
 
-	if (ct->bo) {
+	if (ct->bo && want_ctb) {
 		snapshot->ctb_size = ct->bo->size;
 		snapshot->ctb = kmalloc(snapshot->ctb_size, atomic ? GFP_ATOMIC : GFP_KERNEL);
 	}
···
 	drm_printf(p, "\tstatus (memory): 0x%x\n", snapshot->desc.status);
 }
 
-/**
- * xe_guc_ct_snapshot_capture - Take a quick snapshot of the CT state.
- * @ct: GuC CT object.
- * @atomic: Boolean to indicate if this is called from atomic context like
- * reset or CTB handler or from some regular path like debugfs.
- *
- * This can be printed out in a later stage like during dev_coredump
- * analysis.
- *
- * Returns: a GuC CT snapshot object that must be freed by the caller
- * by using `xe_guc_ct_snapshot_free`.
- */
-struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct,
-						      bool atomic)
+static struct xe_guc_ct_snapshot *guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic,
+							  bool want_ctb)
 {
 	struct xe_device *xe = ct_to_xe(ct);
 	struct xe_guc_ct_snapshot *snapshot;
 
-	snapshot = xe_guc_ct_snapshot_alloc(ct, atomic);
+	snapshot = guc_ct_snapshot_alloc(ct, atomic, want_ctb);
 	if (!snapshot) {
 		xe_gt_err(ct_to_gt(ct), "Skipping CTB snapshot entirely.\n");
 		return NULL;
···
 		xe_map_memcpy_from(xe, snapshot->ctb, &ct->bo->vmap, 0, snapshot->ctb_size);
 
 	return snapshot;
+}
+
+/**
+ * xe_guc_ct_snapshot_capture - Take a quick snapshot of the CT state.
+ * @ct: GuC CT object.
+ *
+ * This can be printed out in a later stage like during dev_coredump
+ * analysis. This is safe to be called during atomic context.
+ *
+ * Returns: a GuC CT snapshot object that must be freed by the caller
+ * by using `xe_guc_ct_snapshot_free`.
+ */
+struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct)
+{
+	return guc_ct_snapshot_capture(ct, true, true);
 }
 
 /**
···
 		drm_printf(p, "\tg2h outstanding: %d\n",
 			   snapshot->g2h_outstanding);
 
-		if (snapshot->ctb) {
+		if (snapshot->ctb)
 			xe_print_blob_ascii85(p, "CTB data", snapshot->ctb, 0, snapshot->ctb_size);
-		} else {
-			drm_printf(p, "CTB snapshot missing!\n");
-			return;
-		}
 	} else {
 		drm_puts(p, "CT disabled\n");
 	}
···
  * xe_guc_ct_print - GuC CT Print.
  * @ct: GuC CT.
  * @p: drm_printer where it will be printed out.
+ * @want_ctb: Should the full CTB content be dumped (vs just the headers)
  *
- * This function quickly capture a snapshot and immediately print it out.
+ * This function will quickly capture a snapshot of the CT state
+ * and immediately print it out.
  */
-void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p)
+void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p, bool want_ctb)
 {
 	struct xe_guc_ct_snapshot *snapshot;
 
-	snapshot = xe_guc_ct_snapshot_capture(ct, false);
+	snapshot = guc_ct_snapshot_capture(ct, false, want_ctb);
 	xe_guc_ct_snapshot_print(snapshot, p);
 	xe_guc_ct_snapshot_free(snapshot);
 }
···
 		return;
 
 	snapshot_log = xe_guc_log_snapshot_capture(&guc->log, true);
-	snapshot_ct = xe_guc_ct_snapshot_capture((ct), true);
+	snapshot_ct = xe_guc_ct_snapshot_capture((ct));
 
 	spin_lock_irqsave(&ct->dead.lock, flags);
 

drivers/gpu/drm/xe/xe_guc_ct.h

···
 void xe_guc_ct_stop(struct xe_guc_ct *ct);
 void xe_guc_ct_fast_path(struct xe_guc_ct *ct);
 
-struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_alloc(struct xe_guc_ct *ct, bool atomic);
-struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct, bool atomic);
+struct xe_guc_ct_snapshot *xe_guc_ct_snapshot_capture(struct xe_guc_ct *ct);
 void xe_guc_ct_snapshot_print(struct xe_guc_ct_snapshot *snapshot, struct drm_printer *p);
 void xe_guc_ct_snapshot_free(struct xe_guc_ct_snapshot *snapshot);
-void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p);
+void xe_guc_ct_print(struct xe_guc_ct *ct, struct drm_printer *p, bool want_ctb);
 
 static inline bool xe_guc_ct_enabled(struct xe_guc_ct *ct)
 {

drivers/gpu/drm/xe/xe_guc_debugfs.c

···
 	return 0;
 }
 
+static int guc_ctb(struct seq_file *m, void *data)
+{
+	struct xe_guc *guc = node_to_guc(m->private);
+	struct xe_device *xe = guc_to_xe(guc);
+	struct drm_printer p = drm_seq_file_printer(m);
+
+	xe_pm_runtime_get(xe);
+	xe_guc_ct_print(&guc->ct, &p, true);
+	xe_pm_runtime_put(xe);
+
+	return 0;
+}
+
 static const struct drm_info_list debugfs_list[] = {
 	{"guc_info", guc_info, 0},
 	{"guc_log", guc_log, 0},
+	{"guc_ctb", guc_ctb, 0},
 };
 
 void xe_guc_debugfs_register(struct xe_guc *guc, struct dentry *parent)

drivers/gpu/drm/xe/xe_guc_fwif.h

···
 
 #define GUC_CTL_FEATURE			2
 #define   GUC_CTL_ENABLE_SLPC		BIT(2)
+#define   GUC_CTL_ENABLE_LITE_RESTORE	BIT(4)
 #define   GUC_CTL_DISABLE_SCHEDULER	BIT(14)
 
 #define GUC_CTL_DEBUG			3

drivers/gpu/drm/xe/xe_guc_log.c

···
 	struct xe_device *xe = log_to_xe(log);
 	struct xe_guc *guc = log_to_guc(log);
 	struct xe_gt *gt = log_to_gt(log);
+	unsigned int fw_ref;
 	size_t remain;
-	int i, err;
+	int i;
 
 	if (!log->bo) {
 		xe_gt_err(gt, "GuC log buffer not allocated\n");
···
 		remain -= size;
 	}
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err) {
-		snapshot->stamp = ~0;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref) {
+		snapshot->stamp = ~0ULL;
 	} else {
-		snapshot->stamp = xe_mmio_read32(&gt->mmio, GUC_PMTIMESTAMP);
-		xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+		snapshot->stamp = xe_mmio_read64_2x32(&gt->mmio, GUC_PMTIMESTAMP_LO);
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	}
 	snapshot->ktime = ktime_get_boottime_ns();
 	snapshot->level = log->level;
···
 		   snapshot->ver_found.major, snapshot->ver_found.minor, snapshot->ver_found.patch,
 		   snapshot->ver_want.major, snapshot->ver_want.minor, snapshot->ver_want.patch);
 	drm_printf(p, "Kernel timestamp: 0x%08llX [%llu]\n", snapshot->ktime, snapshot->ktime);
-	drm_printf(p, "GuC timestamp: 0x%08X [%u]\n", snapshot->stamp, snapshot->stamp);
+	drm_printf(p, "GuC timestamp: 0x%08llX [%llu]\n", snapshot->stamp, snapshot->stamp);
 	drm_printf(p, "Log level: %u\n", snapshot->level);
 
 	remain = snapshot->size;

drivers/gpu/drm/xe/xe_guc_log_types.h

···
 	/** @ktime: Kernel time the snapshot was taken */
 	u64 ktime;
 	/** @stamp: GuC timestamp at which the snapshot was taken */
-	u32 stamp;
+	u64 stamp;
 	/** @level: GuC log verbosity level */
 	u32 level;
 	/** @ver_found: GuC firmware version */

drivers/gpu/drm/xe/xe_guc_pc.c

···
 int xe_guc_pc_get_cur_freq(struct xe_guc_pc *pc, u32 *freq)
 {
 	struct xe_gt *gt = pc_to_gt(pc);
-	int ret;
+	unsigned int fw_ref;
 
 	/*
 	 * GuC SLPC plays with cur freq request when GuCRC is enabled
 	 * Block RC6 for a more reliable read.
 	 */
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (ret)
-		return ret;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
+		return -ETIMEDOUT;
+	}
 
 	*freq = xe_mmio_read32(&gt->mmio, RPNSWREQ);
 
 	*freq = REG_FIELD_GET(REQ_RATIO_MASK, *freq);
 	*freq = decode_freq(*freq);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	return 0;
 }
 
···
 int xe_guc_pc_get_min_freq(struct xe_guc_pc *pc, u32 *freq)
 {
 	struct xe_gt *gt = pc_to_gt(pc);
+	unsigned int fw_ref;
 	int ret;
 
 	mutex_lock(&pc->freq_lock);
···
 	 * GuC SLPC plays with min freq request when GuCRC is enabled
 	 * Block RC6 for a more reliable read.
 	 */
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (ret)
-		goto out;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		ret = -ETIMEDOUT;
+		goto fw;
+	}
 
 	ret = pc_action_query_task_state(pc);
 	if (ret)
···
 	*freq = pc_get_min_freq(pc);
 
 fw:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 out:
 	mutex_unlock(&pc->freq_lock);
 	return ret;
···
 {
 	struct xe_device *xe = pc_to_xe(pc);
 	struct xe_gt *gt = pc_to_gt(pc);
+	unsigned int fw_ref;
 	int ret = 0;
 
 	if (xe->info.skip_guc_pc)
···
 	if (ret)
 		return ret;
 
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (ret)
-		return ret;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
+		return -ETIMEDOUT;
+	}
 
 	xe_gt_idle_disable_c6(gt);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
 	return 0;
 }
···
 	struct xe_device *xe = pc_to_xe(pc);
 	struct xe_gt *gt = pc_to_gt(pc);
 	u32 size = PAGE_ALIGN(sizeof(struct slpc_shared_data));
+	unsigned int fw_ref;
 	int ret;
 
 	xe_gt_assert(gt, xe_device_uc_enabled(xe));
 
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (ret)
-		return ret;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
+		return -ETIMEDOUT;
+	}
 
 	if (xe->info.skip_guc_pc) {
 		if (xe->info.platform != XE_PVC)
···
 	ret = pc_action_setup_gucrc(pc, GUCRC_FIRMWARE_CONTROL);
 
 out:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	return ret;
 }
 
···
 {
 	struct xe_guc_pc *pc = arg;
 	struct xe_device *xe = pc_to_xe(pc);
+	unsigned int fw_ref;
 
 	if (xe_device_wedged(xe))
 		return;
 
-	XE_WARN_ON(xe_force_wake_get(gt_to_fw(pc_to_gt(pc)), XE_FORCEWAKE_ALL));
+	fw_ref = xe_force_wake_get(gt_to_fw(pc_to_gt(pc)), XE_FORCEWAKE_ALL);
 	xe_guc_pc_gucrc_disable(pc);
 	XE_WARN_ON(xe_guc_pc_stop(pc));
 
 	/* Bind requested freq to mert_freq_cap before unload */
 	pc_set_cur_freq(pc, min(pc_max_freq_cap(pc), pc->rpe_freq));
 
-	xe_force_wake_put(gt_to_fw(pc_to_gt(pc)), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(pc_to_gt(pc)), fw_ref);
 }
 
 /**

drivers/gpu/drm/xe/xe_guc_submit.c

···
 	struct xe_exec_queue *q = job->q;
 	struct xe_guc *guc = exec_queue_to_guc(q);
 	struct xe_device *xe = guc_to_xe(guc);
+	struct dma_fence *fence = NULL;
 	bool lr = xe_exec_queue_is_lr(q);
 
 	xe_assert(xe, !(exec_queue_destroyed(q) || exec_queue_pending_disable(q)) ||
···
 
 	if (lr) {
 		xe_sched_job_set_error(job, -EOPNOTSUPP);
-		return NULL;
-	} else if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags)) {
-		return job->fence;
+		dma_fence_put(job->fence);	/* Drop ref from xe_sched_job_arm */
 	} else {
-		return dma_fence_get(job->fence);
+		fence = job->fence;
 	}
+
+	return fence;
 }
 
 static void guc_exec_queue_free_job(struct drm_sched_job *drm_job)
···
 	struct xe_guc *guc = exec_queue_to_guc(q);
 	const char *process_name = "no process";
 	struct xe_device *xe = guc_to_xe(guc);
+	unsigned int fw_ref;
 	int err = -ETIME;
 	pid_t pid = -1;
 	int i = 0;
···
 	if (!exec_queue_killed(q) && !xe->devcoredump.captured &&
 	    !xe_guc_capture_get_matching_and_lock(job)) {
 		/* take force wake before engine register manual capture */
-		if (xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL))
+		fw_ref = xe_force_wake_get(gt_to_fw(q->gt), XE_FORCEWAKE_ALL);
+		if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 			xe_gt_info(q->gt, "failed to get forcewake for coredump capture\n");
 
 		xe_engine_snapshot_capture_for_job(job);
 
-		xe_force_wake_put(gt_to_fw(q->gt), XE_FORCEWAKE_ALL);
+		xe_force_wake_put(gt_to_fw(q->gt), fw_ref);
 	}
 
 	/*

drivers/gpu/drm/xe/xe_huc.c

···
 void xe_huc_print_info(struct xe_huc *huc, struct drm_printer *p)
 {
 	struct xe_gt *gt = huc_to_gt(huc);
-	int err;
+	unsigned int fw_ref;
 
 	xe_uc_fw_print(&huc->fw, p);
 
 	if (!xe_uc_fw_is_enabled(&huc->fw))
 		return;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		return;
 
 	drm_printf(p, "\nHuC status: 0x%08x\n",
 		   xe_mmio_read32(&gt->mmio, HUC_KERNEL_LOAD_INFO));
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }

drivers/gpu/drm/xe/xe_mocs.c

···
 
 void xe_mocs_dump(struct xe_gt *gt, struct drm_printer *p)
 {
-	struct xe_mocs_info table;
-	unsigned int flags;
-	u32 ret;
 	struct xe_device *xe = gt_to_xe(gt);
+	struct xe_mocs_info table;
+	unsigned int fw_ref, flags;
 
 	flags = get_mocs_settings(xe, &table);
 
 	xe_pm_runtime_get_noresume(xe);
-	ret = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-
-	if (ret)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
 		goto err_fw;
 
 	table.ops->dump(&table, flags, gt, p);
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 err_fw:
-	xe_assert(xe, !ret);
 	xe_pm_runtime_put(xe);
 }
 

drivers/gpu/drm/xe/xe_oa.c

···
 #include "xe_pm.h"
 #include "xe_sched_job.h"
 #include "xe_sriov.h"
+#include "xe_sync.h"
 
 #define DEFAULT_POLL_FREQUENCY_HZ 200
 #define DEFAULT_POLL_PERIOD_NS (NSEC_PER_SEC / DEFAULT_POLL_FREQUENCY_HZ)
 #define XE_OA_UNIT_INVALID U32_MAX
+
+enum xe_oa_submit_deps {
+	XE_OA_SUBMIT_NO_DEPS,
+	XE_OA_SUBMIT_ADD_DEPS,
+};
+
+enum xe_oa_user_extn_from {
+	XE_OA_USER_EXTN_FROM_OPEN,
+	XE_OA_USER_EXTN_FROM_CONFIG,
+};
 
 struct xe_oa_reg {
 	struct xe_reg addr;
···
 };
 
 struct xe_oa_open_param {
+	struct xe_file *xef;
 	u32 oa_unit_id;
 	bool sample;
 	u32 metric_set;
···
 	struct xe_exec_queue *exec_q;
 	struct xe_hw_engine *hwe;
 	bool no_preempt;
+	struct drm_xe_sync __user *syncs_user;
+	int num_syncs;
+	struct xe_sync_entry *syncs;
 };
 
 struct xe_oa_config_bo {
···
 
 	struct xe_oa_config *oa_config;
 	struct xe_bb *bb;
+};
+
+struct xe_oa_fence {
+	/* @base: dma fence base */
+	struct dma_fence base;
+	/* @lock: lock for the fence */
+	spinlock_t lock;
+	/* @work: work to signal @base */
+	struct delayed_work work;
+	/* @cb: callback to schedule @work */
+	struct dma_fence_cb cb;
 };
 
 #define DRM_FMT(x) DRM_XE_OA_FMT_TYPE_##x
···
 	return oa_config;
 }
 
-static void free_oa_config_bo(struct xe_oa_config_bo *oa_bo)
+static void free_oa_config_bo(struct xe_oa_config_bo *oa_bo, struct dma_fence *last_fence)
 {
 	xe_oa_config_put(oa_bo->oa_config);
-	xe_bb_free(oa_bo->bb, NULL);
+	xe_bb_free(oa_bo->bb, last_fence);
 	kfree(oa_bo);
 }
 
···
 	return ret;
 }
 
-static int xe_oa_submit_bb(struct xe_oa_stream *stream, struct xe_bb *bb)
+static struct dma_fence *xe_oa_submit_bb(struct xe_oa_stream *stream, enum xe_oa_submit_deps deps,
+					 struct xe_bb *bb)
 {
 	struct xe_sched_job *job;
 	struct dma_fence *fence;
-	long timeout;
 	int err = 0;
 
 	/* Kernel configuration is issued on stream->k_exec_q, not stream->exec_q */
···
 		goto exit;
 	}
 
+	if (deps == XE_OA_SUBMIT_ADD_DEPS) {
+		for (int i = 0; i < stream->num_syncs && !err; i++)
+			err = xe_sync_entry_add_deps(&stream->syncs[i], job);
+		if (err) {
+			drm_dbg(&stream->oa->xe->drm, "xe_sync_entry_add_deps err %d\n", err);
+			goto err_put_job;
+		}
+	}
+
 	xe_sched_job_arm(job);
 	fence = dma_fence_get(&job->drm.s_fence->finished);
 	xe_sched_job_push(job);
 
-	timeout = dma_fence_wait_timeout(fence, false, HZ);
-	dma_fence_put(fence);
-	if (timeout < 0)
-		err = timeout;
-	else if (!timeout)
-		err = -ETIME;
+	return fence;
+err_put_job:
+	xe_sched_job_put(job);
 exit:
-	return err;
+	return ERR_PTR(err);
 }
 
 static void write_cs_mi_lri(struct xe_bb *bb, const struct xe_oa_reg *reg_data, u32 n_regs)
···
 
 	xe_oa_config_put(stream->oa_config);
 	llist_for_each_entry_safe(oa_bo, tmp, stream->oa_config_bos.first, node)
-		free_oa_config_bo(oa_bo);
+		free_oa_config_bo(oa_bo, stream->last_fence);
+	dma_fence_put(stream->last_fence);
 }
 
 static void xe_oa_store_flex(struct xe_oa_stream *stream, struct xe_lrc *lrc,
···
 static int xe_oa_modify_ctx_image(struct xe_oa_stream *stream, struct xe_lrc *lrc,
 				  const struct flex *flex, u32 count)
 {
+	struct dma_fence *fence;
 	struct xe_bb *bb;
 	int err;
 
···
 
 	xe_oa_store_flex(stream, lrc, bb, flex, count);
 
-	err = xe_oa_submit_bb(stream, bb);
+	fence = xe_oa_submit_bb(stream, XE_OA_SUBMIT_NO_DEPS, bb);
+	if (IS_ERR(fence)) {
+		err = PTR_ERR(fence);
+		goto free_bb;
+	}
+	xe_bb_free(bb, fence);
+	dma_fence_put(fence);
+
+	return 0;
+free_bb:
 	xe_bb_free(bb, NULL);
 exit:
 	return err;
···
 
 static int xe_oa_load_with_lri(struct xe_oa_stream *stream, struct xe_oa_reg *reg_lri)
 {
+	struct dma_fence *fence;
 	struct xe_bb *bb;
 	int err;
 
···
 
 	write_cs_mi_lri(bb, reg_lri, 1);
 
-	err = xe_oa_submit_bb(stream, bb);
+	fence = xe_oa_submit_bb(stream, XE_OA_SUBMIT_NO_DEPS, bb);
+	if (IS_ERR(fence)) {
+		err = PTR_ERR(fence);
+		goto free_bb;
+	}
+	xe_bb_free(bb, fence);
+	dma_fence_put(fence);
+
+	return 0;
+free_bb:
 	xe_bb_free(bb, NULL);
 exit:
 	return err;
···
 
 	xe_oa_free_oa_buffer(stream);
 
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
 	xe_pm_runtime_put(stream->oa->xe);
 
 	/* Wa_1509372804:pvc: Unset the override of GUCRC mode to enable rc6 */
···
 		xe_gt_WARN_ON(gt, xe_guc_pc_unset_gucrc_mode(&gt->uc.guc.pc));
 
 	xe_oa_free_configs(stream);
+	xe_file_put(stream->xef);
 }
 
 static int xe_oa_alloc_oa_buffer(struct xe_oa_stream *stream)
···
 	return oa_bo;
 }
 
+static void xe_oa_update_last_fence(struct xe_oa_stream *stream, struct dma_fence *fence)
+{
+	dma_fence_put(stream->last_fence);
+	stream->last_fence = dma_fence_get(fence);
+}
+
+static void xe_oa_fence_work_fn(struct work_struct *w)
+{
+	struct xe_oa_fence *ofence = container_of(w, typeof(*ofence), work.work);
+
+	/* Signal fence to indicate new OA configuration is active */
+	dma_fence_signal(&ofence->base);
+	dma_fence_put(&ofence->base);
+}
+
+static void xe_oa_config_cb(struct dma_fence *fence, struct dma_fence_cb *cb)
+{
+	/* Additional empirical delay needed for NOA programming after registers are written */
+#define NOA_PROGRAM_ADDITIONAL_DELAY_US 500
+
+	struct xe_oa_fence *ofence = container_of(cb, typeof(*ofence), cb);
+
+	INIT_DELAYED_WORK(&ofence->work, xe_oa_fence_work_fn);
+	queue_delayed_work(system_unbound_wq, &ofence->work,
+			   usecs_to_jiffies(NOA_PROGRAM_ADDITIONAL_DELAY_US));
+	dma_fence_put(fence);
+}
+
+static const char *xe_oa_get_driver_name(struct dma_fence *fence)
+{
+	return "xe_oa";
+}
+
+static const char *xe_oa_get_timeline_name(struct dma_fence *fence)
+{
+	return "unbound";
+}
+
+static const struct dma_fence_ops xe_oa_fence_ops = {
+	.get_driver_name = xe_oa_get_driver_name,
+	.get_timeline_name = xe_oa_get_timeline_name,
+};
+
 static int xe_oa_emit_oa_config(struct xe_oa_stream *stream, struct xe_oa_config *config)
 {
 #define NOA_PROGRAM_ADDITIONAL_DELAY_US 500
 	struct xe_oa_config_bo *oa_bo;
-	int err, us = NOA_PROGRAM_ADDITIONAL_DELAY_US;
+	struct xe_oa_fence *ofence;
+	int i, err, num_signal = 0;
+	struct dma_fence *fence;
+
+	ofence = kzalloc(sizeof(*ofence), GFP_KERNEL);
+	if (!ofence) {
+		err = -ENOMEM;
+		goto exit;
+	}
 
 	oa_bo = xe_oa_alloc_config_buffer(stream, config);
 	if (IS_ERR(oa_bo)) {
···
 		goto exit;
 	}
 
-	err = xe_oa_submit_bb(stream, oa_bo->bb);
+	/* Emit OA configuration batch */
+	fence = xe_oa_submit_bb(stream, XE_OA_SUBMIT_ADD_DEPS, oa_bo->bb);
+	if (IS_ERR(fence)) {
+		err = PTR_ERR(fence);
+		goto exit;
+	}
 
-	/* Additional empirical delay needed for NOA programming after registers are written */
-	usleep_range(us, 2 * us);
+	/* Point of no return: initialize and set fence to signal */
+	spin_lock_init(&ofence->lock);
+	dma_fence_init(&ofence->base, &xe_oa_fence_ops, &ofence->lock, 0, 0);
+
+	for (i = 0; i < stream->num_syncs; i++) {
+		if (stream->syncs[i].flags & DRM_XE_SYNC_FLAG_SIGNAL)
+			num_signal++;
+		xe_sync_entry_signal(&stream->syncs[i], &ofence->base);
+	}
+
+	/* Additional dma_fence_get in case we dma_fence_wait */
+	if (!num_signal)
+		dma_fence_get(&ofence->base);
+
+	/* Update last fence too before adding callback */
+	xe_oa_update_last_fence(stream, fence);
+
+	/* Add job fence callback to schedule work to signal ofence->base */
+	err = dma_fence_add_callback(fence, &ofence->cb, xe_oa_config_cb);
+	xe_gt_assert(stream->gt, !err || err == -ENOENT);
+	if (err == -ENOENT)
+		xe_oa_config_cb(fence, &ofence->cb);
+
+	/* If nothing needs to be signaled we wait synchronously */
+	if (!num_signal) {
+		dma_fence_wait(&ofence->base, false);
+		dma_fence_put(&ofence->base);
+	}
+
+	/* Done with syncs */
+	for (i = 0; i < stream->num_syncs; i++)
+		xe_sync_entry_cleanup(&stream->syncs[i]);
+	kfree(stream->syncs);
+
+	return 0;
 exit:
+	kfree(ofence);
 	return err;
 }
 
···
 	}
 
 	return xe_oa_emit_oa_config(stream, stream->oa_config);
+}
+
+static int decode_oa_format(struct xe_oa *oa, u64 fmt, enum xe_oa_format_name *name)
+{
+	u32 counter_size = FIELD_GET(DRM_XE_OA_FORMAT_MASK_COUNTER_SIZE, fmt);
+	u32 counter_sel = FIELD_GET(DRM_XE_OA_FORMAT_MASK_COUNTER_SEL, fmt);
+	u32 bc_report = FIELD_GET(DRM_XE_OA_FORMAT_MASK_BC_REPORT, fmt);
+	u32 type = FIELD_GET(DRM_XE_OA_FORMAT_MASK_FMT_TYPE, fmt);
+	int idx;
+
+	for_each_set_bit(idx, oa->format_mask, __XE_OA_FORMAT_MAX) {
+		const struct xe_oa_format *f = &oa->oa_formats[idx];
+
+		if (counter_size == f->counter_size && bc_report == f->bc_report &&
+		    type == f->type && counter_sel == f->counter_select) {
+			*name = idx;
+			return 0;
+		}
+	}
+
+	return -EINVAL;
+}
+
+static int xe_oa_set_prop_oa_unit_id(struct xe_oa *oa, u64 value,
+				     struct xe_oa_open_param *param)
+{
+	if (value >= oa->oa_unit_ids) {
+		drm_dbg(&oa->xe->drm, "OA unit ID out of range %lld\n", value);
+		return -EINVAL;
+	}
+	param->oa_unit_id = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_sample_oa(struct xe_oa *oa, u64 value,
+				    struct xe_oa_open_param *param)
+{
+	param->sample = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_metric_set(struct xe_oa *oa, u64 value,
+				     struct xe_oa_open_param *param)
+{
+	param->metric_set = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_oa_format(struct xe_oa *oa, u64 value,
+				    struct xe_oa_open_param *param)
+{
+	int ret = decode_oa_format(oa, value, &param->oa_format);
+
+	if (ret) {
+		drm_dbg(&oa->xe->drm, "Unsupported OA report format %#llx\n", value);
+		return ret;
+	}
+	return 0;
+}
+
+static int xe_oa_set_prop_oa_exponent(struct xe_oa *oa, u64 value,
+				      struct xe_oa_open_param *param)
+{
+#define OA_EXPONENT_MAX 31
+
+	if (value > OA_EXPONENT_MAX) {
+		drm_dbg(&oa->xe->drm, "OA timer exponent too high (> %u)\n", OA_EXPONENT_MAX);
+		return -EINVAL;
+	}
+	param->period_exponent = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_disabled(struct xe_oa *oa, u64 value,
+				   struct xe_oa_open_param *param)
+{
+	param->disabled = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_exec_queue_id(struct xe_oa *oa, u64 value,
+					struct xe_oa_open_param *param)
+{
+	param->exec_queue_id = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_engine_instance(struct xe_oa *oa, u64 value,
+					  struct xe_oa_open_param *param)
+{
+	param->engine_instance = value;
+	return 0;
+}
+
+static int xe_oa_set_no_preempt(struct xe_oa *oa, u64 value,
+				struct xe_oa_open_param *param)
+{
+	param->no_preempt = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_num_syncs(struct xe_oa *oa, u64 value,
+				    struct xe_oa_open_param *param)
+{
+	param->num_syncs = value;
+	return 0;
+}
+
+static int xe_oa_set_prop_syncs_user(struct xe_oa *oa, u64 value,
+				     struct xe_oa_open_param *param)
+{
+	param->syncs_user = u64_to_user_ptr(value);
+	return 0;
+}
+
+static int xe_oa_set_prop_ret_inval(struct xe_oa *oa, u64 value,
+				    struct xe_oa_open_param *param)
+{
+	return -EINVAL;
+}
+
+typedef int (*xe_oa_set_property_fn)(struct xe_oa *oa, u64 value,
+				     struct xe_oa_open_param *param);
+static const xe_oa_set_property_fn xe_oa_set_property_funcs_open[] = {
+	[DRM_XE_OA_PROPERTY_OA_UNIT_ID] = xe_oa_set_prop_oa_unit_id,
+	[DRM_XE_OA_PROPERTY_SAMPLE_OA] = xe_oa_set_prop_sample_oa,
+	[DRM_XE_OA_PROPERTY_OA_METRIC_SET] = xe_oa_set_prop_metric_set,
+	[DRM_XE_OA_PROPERTY_OA_FORMAT] = xe_oa_set_prop_oa_format,
+	[DRM_XE_OA_PROPERTY_OA_PERIOD_EXPONENT] = xe_oa_set_prop_oa_exponent,
+	[DRM_XE_OA_PROPERTY_OA_DISABLED] = xe_oa_set_prop_disabled,
+	[DRM_XE_OA_PROPERTY_EXEC_QUEUE_ID] = xe_oa_set_prop_exec_queue_id,
+	[DRM_XE_OA_PROPERTY_OA_ENGINE_INSTANCE] = xe_oa_set_prop_engine_instance,
+	[DRM_XE_OA_PROPERTY_NO_PREEMPT] = xe_oa_set_no_preempt,
+	[DRM_XE_OA_PROPERTY_NUM_SYNCS] = xe_oa_set_prop_num_syncs,
+	[DRM_XE_OA_PROPERTY_SYNCS] = xe_oa_set_prop_syncs_user,
+};
+
+static const xe_oa_set_property_fn xe_oa_set_property_funcs_config[] = {
+	[DRM_XE_OA_PROPERTY_OA_UNIT_ID] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_SAMPLE_OA] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_OA_METRIC_SET] = xe_oa_set_prop_metric_set,
+	[DRM_XE_OA_PROPERTY_OA_FORMAT] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_OA_PERIOD_EXPONENT] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_OA_DISABLED] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_EXEC_QUEUE_ID] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_OA_ENGINE_INSTANCE] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_NO_PREEMPT] = xe_oa_set_prop_ret_inval,
+	[DRM_XE_OA_PROPERTY_NUM_SYNCS] = xe_oa_set_prop_num_syncs,
+	[DRM_XE_OA_PROPERTY_SYNCS] = xe_oa_set_prop_syncs_user,
+};
+
+static int xe_oa_user_ext_set_property(struct xe_oa *oa, enum xe_oa_user_extn_from from,
+				       u64 extension, struct xe_oa_open_param *param)
+{
+	u64 __user *address = u64_to_user_ptr(extension);
+	struct drm_xe_ext_set_property ext;
+	int err;
+	u32 idx;
+
+	err = __copy_from_user(&ext, address, sizeof(ext));
+	if (XE_IOCTL_DBG(oa->xe, err))
+		return -EFAULT;
+
+	BUILD_BUG_ON(ARRAY_SIZE(xe_oa_set_property_funcs_open) !=
+		     ARRAY_SIZE(xe_oa_set_property_funcs_config));
+
+	if (XE_IOCTL_DBG(oa->xe, ext.property >= ARRAY_SIZE(xe_oa_set_property_funcs_open)) ||
+	    XE_IOCTL_DBG(oa->xe, ext.pad))
+		return -EINVAL;
+
+	idx = array_index_nospec(ext.property, ARRAY_SIZE(xe_oa_set_property_funcs_open));
+
+	if (from == XE_OA_USER_EXTN_FROM_CONFIG)
+		return xe_oa_set_property_funcs_config[idx](oa, ext.value, param);
+	else
+		return xe_oa_set_property_funcs_open[idx](oa, ext.value, param);
+}
+
+typedef int (*xe_oa_user_extension_fn)(struct xe_oa *oa,  enum xe_oa_user_extn_from from,
+				       u64 extension, struct xe_oa_open_param *param);
+static const xe_oa_user_extension_fn xe_oa_user_extension_funcs[] = {
+	[DRM_XE_OA_EXTENSION_SET_PROPERTY] = xe_oa_user_ext_set_property,
+};
+
+#define MAX_USER_EXTENSIONS	16
+static int xe_oa_user_extensions(struct xe_oa *oa, enum xe_oa_user_extn_from from, u64 extension,
+				 int ext_number, struct xe_oa_open_param *param)
+{
+	u64 __user *address = u64_to_user_ptr(extension);
+	struct drm_xe_user_extension ext;
+	int err;
+	u32 idx;
+
+	if (XE_IOCTL_DBG(oa->xe, ext_number >= MAX_USER_EXTENSIONS))
+		return -E2BIG;
+
+	err = __copy_from_user(&ext, address, sizeof(ext));
+	if (XE_IOCTL_DBG(oa->xe, err))
+		return -EFAULT;
+
+	if (XE_IOCTL_DBG(oa->xe, ext.pad) ||
+	    XE_IOCTL_DBG(oa->xe, ext.name >= ARRAY_SIZE(xe_oa_user_extension_funcs)))
+		return -EINVAL;
+
+	idx = array_index_nospec(ext.name, ARRAY_SIZE(xe_oa_user_extension_funcs));
+	err = xe_oa_user_extension_funcs[idx](oa, from, extension, param);
+	if (XE_IOCTL_DBG(oa->xe, err))
+		return err;
+
+	if (ext.next_extension)
+		return xe_oa_user_extensions(oa, from, ext.next_extension, ++ext_number, param);
+
+	return 0;
+}
+
+static int xe_oa_parse_syncs(struct xe_oa *oa, struct xe_oa_open_param *param)
+{
+	int ret, num_syncs, num_ufence = 0;
+
+	if (param->num_syncs && !param->syncs_user) {
+		drm_dbg(&oa->xe->drm, "num_syncs specified without sync array\n");
+		ret = -EINVAL;
+		goto exit;
+	}
+
+	if (param->num_syncs) {
+		param->syncs = kcalloc(param->num_syncs, sizeof(*param->syncs), GFP_KERNEL);
+		if (!param->syncs) {
+			ret = -ENOMEM;
+			goto exit;
+		}
+	}
+
+	for (num_syncs = 0; num_syncs < param->num_syncs; num_syncs++) {
+		ret = xe_sync_entry_parse(oa->xe, param->xef, &param->syncs[num_syncs],
+					  &param->syncs_user[num_syncs], 0);
+		if (ret)
+			goto err_syncs;
+
+		if (xe_sync_is_ufence(&param->syncs[num_syncs]))
+			num_ufence++;
+	}
+
+	if (XE_IOCTL_DBG(oa->xe, num_ufence > 1)) {
+		ret = -EINVAL;
+		goto err_syncs;
+	}
+
+	return 0;
+
+err_syncs:
+	while (num_syncs--)
+		xe_sync_entry_cleanup(&param->syncs[num_syncs]);
+	kfree(param->syncs);
+exit:
+	return ret;
 }
 
 static void xe_oa_stream_enable(struct xe_oa_stream *stream)
···
 
 static long xe_oa_config_locked(struct xe_oa_stream *stream, u64 arg)
 {
-	struct drm_xe_ext_set_property ext;
+	struct xe_oa_open_param param = {};
 	long ret = stream->oa_config->id;
 	struct xe_oa_config *config;
 	int err;
 
-	err = __copy_from_user(&ext, u64_to_user_ptr(arg), sizeof(ext));
-	if (XE_IOCTL_DBG(stream->oa->xe, err))
-		return -EFAULT;
+	err = xe_oa_user_extensions(stream->oa, XE_OA_USER_EXTN_FROM_CONFIG, arg, 0, &param);
+	if (err)
+		return err;
 
-	if (XE_IOCTL_DBG(stream->oa->xe, ext.pad) ||
-	    XE_IOCTL_DBG(stream->oa->xe, ext.base.name != DRM_XE_OA_EXTENSION_SET_PROPERTY) ||
-	    XE_IOCTL_DBG(stream->oa->xe, ext.base.next_extension) ||
-	    XE_IOCTL_DBG(stream->oa->xe, ext.property != DRM_XE_OA_PROPERTY_OA_METRIC_SET))
-		return -EINVAL;
-
-	config = xe_oa_get_oa_config(stream->oa, ext.value);
+	config = xe_oa_get_oa_config(stream->oa, param.metric_set);
 	if (!config)
 		return -ENODEV;
 
-	if (config != stream->oa_config) {
-		err = xe_oa_emit_oa_config(stream, config);
-		if (!err)
-			config = xchg(&stream->oa_config, config);
-		else
-			ret = err;
+	param.xef = stream->xef;
+	err = xe_oa_parse_syncs(stream->oa, &param);
+	if (err)
+		goto err_config_put;
+
+	stream->num_syncs = param.num_syncs;
+	stream->syncs = param.syncs;
+
+	err = xe_oa_emit_oa_config(stream, config);
+	if (!err) {
+		config = xchg(&stream->oa_config, config);
+		drm_dbg(&stream->oa->xe->drm, "changed to oa config uuid=%s\n",
+			stream->oa_config->uuid);
 	}
 
+err_config_put:
 	xe_oa_config_put(config);
 
-	return ret;
+	return err ?: ret;
 }
 
 static long xe_oa_status_locked(struct xe_oa_stream *stream, unsigned long arg)
···
 {
 	struct xe_oa_unit *u = param->hwe->oa_unit;
 	struct xe_gt *gt = param->hwe->gt;
+	unsigned int fw_ref;
 	int ret;
 
 	stream->exec_q = param->exec_q;
···
 	stream->periodic = param->period_exponent > 0;
 	stream->period_exponent = param->period_exponent;
 	stream->no_preempt = param->no_preempt;
+
+	stream->xef = xe_file_get(param->xef);
+	stream->num_syncs = param->num_syncs;
+	stream->syncs = param->syncs;
 
 	/*
 	 * For Xe2+, when overrun mode is enabled, there are no partial reports at the end
···
 
 	/* Take runtime pm ref and forcewake to disable RC6 */
 	xe_pm_runtime_get(stream->oa->xe);
-	XE_WARN_ON(xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL)) {
+		ret = -ETIMEDOUT;
+		goto err_fw_put;
+	}
 
 	ret = xe_oa_alloc_oa_buffer(stream);
 	if (ret)
···
 err_free_oa_buf:
 	xe_oa_free_oa_buffer(stream);
 err_fw_put:
-	XE_WARN_ON(xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL));
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 	xe_pm_runtime_put(stream->oa->xe);
 	if (stream->override_gucrc)
 		xe_gt_WARN_ON(gt, xe_guc_pc_unset_gucrc_mode(&gt->uc.guc.pc));
 err_free_configs:
 	xe_oa_free_configs(stream);
 exit:
+	xe_file_put(stream->xef);
 	return ret;
 }
 
···
 	}
 }
 
-static int decode_oa_format(struct xe_oa *oa, u64 fmt, enum xe_oa_format_name *name)
-{
-	u32 counter_size = FIELD_GET(DRM_XE_OA_FORMAT_MASK_COUNTER_SIZE, fmt);
-	u32 counter_sel = FIELD_GET(DRM_XE_OA_FORMAT_MASK_COUNTER_SEL, fmt);
-	u32 bc_report = FIELD_GET(DRM_XE_OA_FORMAT_MASK_BC_REPORT, fmt);
-	u32 type = FIELD_GET(DRM_XE_OA_FORMAT_MASK_FMT_TYPE, fmt);
-	int idx;
-
-	for_each_set_bit(idx, oa->format_mask, __XE_OA_FORMAT_MAX) {
-		const struct xe_oa_format *f = &oa->oa_formats[idx];
-
-		if (counter_size == f->counter_size && bc_report == f->bc_report &&
-		    type == f->type && counter_sel == f->counter_select) {
-			*name = idx;
-			return 0;
-		}
-	}
-
-	return -EINVAL;
-}
-
 /**
  * xe_oa_unit_id - Return OA unit ID for a hardware engine
  * @hwe: @xe_hw_engine
···
 	return ret;
 }
 
-static int xe_oa_set_prop_oa_unit_id(struct xe_oa *oa, u64 value,
-				     struct xe_oa_open_param *param)
-{
-	if (value >= oa->oa_unit_ids) {
-		drm_dbg(&oa->xe->drm, "OA unit ID out of range %lld\n", value);
-		return -EINVAL;
-	}
-	param->oa_unit_id = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_sample_oa(struct xe_oa *oa, u64 value,
-				    struct xe_oa_open_param *param)
-{
-	param->sample = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_metric_set(struct xe_oa *oa, u64 value,
-				     struct xe_oa_open_param *param)
-{
-	param->metric_set = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_oa_format(struct xe_oa *oa, u64 value,
-				    struct xe_oa_open_param *param)
-{
-	int ret = decode_oa_format(oa, value, &param->oa_format);
-
-	if (ret) {
-		drm_dbg(&oa->xe->drm, "Unsupported OA report format %#llx\n", value);
-		return ret;
-	}
-	return 0;
-}
-
-static int xe_oa_set_prop_oa_exponent(struct xe_oa *oa, u64 value,
-				      struct xe_oa_open_param *param)
-{
-#define OA_EXPONENT_MAX 31
-
-	if (value > OA_EXPONENT_MAX) {
-		drm_dbg(&oa->xe->drm, "OA timer exponent too high (> %u)\n", OA_EXPONENT_MAX);
-		return -EINVAL;
-	}
-	param->period_exponent = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_disabled(struct xe_oa *oa, u64 value,
-				   struct xe_oa_open_param *param)
-{
-	param->disabled = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_exec_queue_id(struct xe_oa *oa, u64 value,
-					struct xe_oa_open_param *param)
-{
-	param->exec_queue_id = value;
-	return 0;
-}
-
-static int xe_oa_set_prop_engine_instance(struct xe_oa *oa, u64 value,
-					  struct xe_oa_open_param *param)
-{
-	param->engine_instance = value;
-	return 0;
-}
-
-static int xe_oa_set_no_preempt(struct xe_oa *oa, u64 value,
-				struct xe_oa_open_param *param)
-{
-	param->no_preempt = value;
-	return 0;
-}
-
-typedef int (*xe_oa_set_property_fn)(struct xe_oa *oa, u64 value,
-				     struct xe_oa_open_param *param);
-static const xe_oa_set_property_fn xe_oa_set_property_funcs[] = {
-	[DRM_XE_OA_PROPERTY_OA_UNIT_ID] = xe_oa_set_prop_oa_unit_id,
-	[DRM_XE_OA_PROPERTY_SAMPLE_OA] = xe_oa_set_prop_sample_oa,
-	[DRM_XE_OA_PROPERTY_OA_METRIC_SET] = xe_oa_set_prop_metric_set,
-	[DRM_XE_OA_PROPERTY_OA_FORMAT] = xe_oa_set_prop_oa_format,
-	[DRM_XE_OA_PROPERTY_OA_PERIOD_EXPONENT] = xe_oa_set_prop_oa_exponent,
-	[DRM_XE_OA_PROPERTY_OA_DISABLED] = xe_oa_set_prop_disabled,
-	[DRM_XE_OA_PROPERTY_EXEC_QUEUE_ID] = xe_oa_set_prop_exec_queue_id,
-	[DRM_XE_OA_PROPERTY_OA_ENGINE_INSTANCE] = xe_oa_set_prop_engine_instance,
-	[DRM_XE_OA_PROPERTY_NO_PREEMPT] = xe_oa_set_no_preempt,
-};
-
-static int xe_oa_user_ext_set_property(struct xe_oa *oa, u64 extension,
-				       struct xe_oa_open_param *param)
-{
-	u64 __user *address = u64_to_user_ptr(extension);
-	struct drm_xe_ext_set_property ext;
-	int err;
-	u32 idx;
-
-	err = __copy_from_user(&ext, address, sizeof(ext));
-	if (XE_IOCTL_DBG(oa->xe, err))
-		return -EFAULT;
-
-	if (XE_IOCTL_DBG(oa->xe, ext.property >= ARRAY_SIZE(xe_oa_set_property_funcs)) ||
-	    XE_IOCTL_DBG(oa->xe, ext.pad))
-		return -EINVAL;
-
-	idx = array_index_nospec(ext.property, ARRAY_SIZE(xe_oa_set_property_funcs));
-	return xe_oa_set_property_funcs[idx](oa, ext.value, param);
-}
-
-typedef int (*xe_oa_user_extension_fn)(struct xe_oa *oa, u64 extension,
-				       struct xe_oa_open_param *param);
-static const xe_oa_user_extension_fn xe_oa_user_extension_funcs[] = {
-	[DRM_XE_OA_EXTENSION_SET_PROPERTY] = xe_oa_user_ext_set_property,
-};
-
-#define MAX_USER_EXTENSIONS	16
-static int xe_oa_user_extensions(struct xe_oa *oa, u64 extension, int ext_number,
-				 struct xe_oa_open_param *param)
-{
-	u64 __user *address = u64_to_user_ptr(extension);
-	struct drm_xe_user_extension ext;
-	int err;
-	u32 idx;
-
-	if (XE_IOCTL_DBG(oa->xe, ext_number >= MAX_USER_EXTENSIONS))
-		return -E2BIG;
-
-	err = __copy_from_user(&ext, address, sizeof(ext));
-	if (XE_IOCTL_DBG(oa->xe, err))
-		return -EFAULT;
-
-	if (XE_IOCTL_DBG(oa->xe, ext.pad) ||
-	    XE_IOCTL_DBG(oa->xe, ext.name >= ARRAY_SIZE(xe_oa_user_extension_funcs)))
-		return -EINVAL;
-
-	idx = array_index_nospec(ext.name, ARRAY_SIZE(xe_oa_user_extension_funcs));
-	err = xe_oa_user_extension_funcs[idx](oa, extension, param);
-	if (XE_IOCTL_DBG(oa->xe, err))
-		return err;
-
-	if (ext.next_extension)
-		return xe_oa_user_extensions(oa, ext.next_extension, ++ext_number, param);
-
-	return 0;
-}
-
 /**
  * xe_oa_stream_open_ioctl - Opens an OA stream
  * @dev: @drm_device
···
 		return -ENODEV;
 	}
 
-	ret = xe_oa_user_extensions(oa, data, 0, &param);
+	param.xef = xef;
+	ret = xe_oa_user_extensions(oa, XE_OA_USER_EXTN_FROM_OPEN, data, 0, &param);
 	if (ret)
 		return ret;
 
···
 		drm_dbg(&oa->xe->drm, "Using periodic sampling freq %lld Hz\n", oa_freq_hz);
 	}
 
+	ret = xe_oa_parse_syncs(oa, &param);
+	if (ret)
+		goto err_exec_q;
+
 	mutex_lock(&param.hwe->gt->oa.gt_lock);
 	ret = xe_oa_stream_open_ioctl_locked(oa, &param);
 	mutex_unlock(&param.hwe->gt->oa.gt_lock);
+	if (ret < 0)
+		goto err_sync_cleanup;
+
+	return ret;
+
+err_sync_cleanup:
+	while (param.num_syncs--)
+		xe_sync_entry_cleanup(&param.syncs[param.num_syncs]);
+	kfree(param.syncs);
 err_exec_q:
-	if (ret < 0 && param.exec_q)
+	if (param.exec_q)
 		xe_exec_queue_put(param.exec_q);
 	return ret;
 }

drivers/gpu/drm/xe/xe_oa_types.h

···
 
 	/** @no_preempt: Whether preemption and timeslicing is disabled for stream exec_q */
 	u32 no_preempt;
+
+	/** @xef: xe_file with which the stream was opened */
+	struct xe_file *xef;
+
+	/** @last_fence: fence to use in stream destroy when needed */
+	struct dma_fence *last_fence;
+
+	/** @num_syncs: size of @syncs array */
+	u32 num_syncs;
+
+	/** @syncs: syncs to wait on and to signal */
+	struct xe_sync_entry *syncs;
 };
 #endif

drivers/gpu/drm/xe/xe_pat.c

···
 static void xelp_dump(struct xe_gt *gt, struct drm_printer *p)
 {
 	struct xe_device *xe = gt_to_xe(gt);
-	int i, err;
+	unsigned int fw_ref;
+	int i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto err_fw;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return;
 
 	drm_printf(p, "PAT table:\n");
 
···
 			   XELP_MEM_TYPE_STR_MAP[mem_type], pat);
 	}
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-err_fw:
-	xe_assert(xe, !err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static const struct xe_pat_ops xelp_pat_ops = {
···
 static void xehp_dump(struct xe_gt *gt, struct drm_printer *p)
 {
 	struct xe_device *xe = gt_to_xe(gt);
-	int i, err;
+	unsigned int fw_ref;
+	int i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto err_fw;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return;
 
 	drm_printf(p, "PAT table:\n");
 
···
 			   XELP_MEM_TYPE_STR_MAP[mem_type], pat);
 	}
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-err_fw:
-	xe_assert(xe, !err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static const struct xe_pat_ops xehp_pat_ops = {
···
 static void xehpc_dump(struct xe_gt *gt, struct drm_printer *p)
 {
 	struct xe_device *xe = gt_to_xe(gt);
-	int i, err;
+	unsigned int fw_ref;
+	int i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto err_fw;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return;
 
 	drm_printf(p, "PAT table:\n");
 
···
 			   REG_FIELD_GET(XEHPC_CLOS_LEVEL_MASK, pat), pat);
 	}
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-err_fw:
-	xe_assert(xe, !err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static const struct xe_pat_ops xehpc_pat_ops = {
···
 static void xelpg_dump(struct xe_gt *gt, struct drm_printer *p)
 {
 	struct xe_device *xe = gt_to_xe(gt);
-	int i, err;
+	unsigned int fw_ref;
+	int i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto err_fw;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return;
 
 	drm_printf(p, "PAT table:\n");
 
···
 			   REG_FIELD_GET(XELPG_INDEX_COH_MODE_MASK, pat), pat);
 	}
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-err_fw:
-	xe_assert(xe, !err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 /*
···
 static void xe2_dump(struct xe_gt *gt, struct drm_printer *p)
 {
 	struct xe_device *xe = gt_to_xe(gt);
-	int i, err;
+	unsigned int fw_ref;
 	u32 pat;
+	int i;
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		goto err_fw;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return;
 
 	drm_printf(p, "PAT table:\n");
 
···
 		   REG_FIELD_GET(XE2_COH_MODE, pat),
 		   pat);
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
-err_fw:
-	xe_assert(xe, !err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 }
 
 static const struct xe_pat_ops xe2_pat_ops = {

drivers/gpu/drm/xe/xe_query.c

···
 	__ktime_func_t cpu_clock;
 	struct xe_hw_engine *hwe;
 	struct xe_gt *gt;
+	unsigned int fw_ref;
 
 	if (query->size == 0) {
 		query->size = size;
···
 	if (!hwe)
 		return -EINVAL;
 
-	if (xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL))
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))  {
+		xe_force_wake_put(gt_to_fw(gt), fw_ref);
 		return -EIO;
+	}
 
 	hwe_read_timestamp(hwe, &resp.engine_cycles, &resp.cpu_timestamp,
 			   &resp.cpu_delta, cpu_clock);
 
-	xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
 	if (GRAPHICS_VER(xe) >= 20)
 		resp.width = 64;
···
 			du->oa_unit_id = u->oa_unit_id;
 			du->oa_unit_type = u->type;
 			du->oa_timestamp_freq = xe_oa_timestamp_frequency(gt);
-			du->capabilities = DRM_XE_OA_CAPS_BASE;
+			du->capabilities = DRM_XE_OA_CAPS_BASE | DRM_XE_OA_CAPS_SYNCS;
 
 			j = 0;
 			for_each_hw_engine(hwe, gt, hwe_id) {

drivers/gpu/drm/xe/xe_reg_sr.c

···
 {
 	struct xe_reg_sr_entry *entry;
 	unsigned long reg;
-	int err;
+	unsigned int fw_ref;
 
 	if (xa_empty(&sr->xa))
 		return;
 
 	xe_gt_dbg(gt, "Applying %s save-restore MMIOs\n", sr->name);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 		goto err_force_wake;
 
 	xa_for_each(&sr->xa, reg, entry)
 		apply_one_mmio(gt, entry);
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	XE_WARN_ON(err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
 	return;
 
 err_force_wake:
-	xe_gt_err(gt, "Failed to apply, err=%d\n", err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
+	xe_gt_err(gt, "Failed to apply, err=-ETIMEDOUT\n");
 }
 
 void xe_reg_sr_apply_whitelist(struct xe_hw_engine *hwe)
···
 	u32 mmio_base = hwe->mmio_base;
 	unsigned long reg;
 	unsigned int slot = 0;
-	int err;
+	unsigned int fw_ref;
 
 	if (xa_empty(&sr->xa))
 		return;
 
 	drm_dbg(&xe->drm, "Whitelisting %s registers\n", sr->name);
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	if (err)
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FORCEWAKE_ALL);
+	if (!xe_force_wake_ref_has_domain(fw_ref, XE_FORCEWAKE_ALL))
 		goto err_force_wake;
 
 	p = drm_dbg_printer(&xe->drm, DRM_UT_DRIVER, NULL);
···
 		xe_mmio_write32(&gt->mmio, RING_FORCE_TO_NONPRIV(mmio_base, slot), addr);
 	}
 
-	err = xe_force_wake_put(gt_to_fw(gt), XE_FORCEWAKE_ALL);
-	XE_WARN_ON(err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
 
 	return;
 
 err_force_wake:
-	drm_err(&xe->drm, "Failed to apply, err=%d\n", err);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
+	drm_err(&xe->drm, "Failed to apply, err=-ETIMEDOUT\n");
 }
 
 /**

drivers/gpu/drm/xe/xe_sched_job.c

···
 		fence = &chain->base;
 	}
 
-	job->fence = fence;
+	job->fence = dma_fence_get(fence);	/* Pairs with put in scheduler */
 	drm_sched_job_arm(&job->drm);
 }
 

drivers/gpu/drm/xe/xe_sched_job_types.h

···
 	 * @fence: dma fence to indicate completion. 1 way relationship - job
 	 * can safely reference fence, fence cannot safely reference job.
 	 */
-#define JOB_FLAG_SUBMIT		DMA_FENCE_FLAG_USER_BITS
 	struct dma_fence *fence;
 	/** @user_fence: write back value when BB is complete */
 	struct {
···
 
 struct xe_sched_job_snapshot {
 	u16 batch_addr_len;
-	u64 batch_addr[];
+	u64 batch_addr[] __counted_by(batch_addr_len);
 };
 
 #endif

drivers/gpu/drm/xe/xe_sync.c

···
 			XE_WARN_ON("Copy to user failed");
 		kthread_unuse_mm(ufence->mm);
 		mmput(ufence->mm);
+	} else {
+		drm_dbg(&ufence->xe->drm, "mmget_not_zero() failed, ufence wasn't signaled\n");
 	}
 
 	wake_up_all(&ufence->xe->ufence_wq);

drivers/gpu/drm/xe/xe_vram.c

···
 {
 	struct xe_device *xe = tile_to_xe(tile);
 	struct xe_gt *gt = tile->primary_gt;
+	unsigned int fw_ref;
 	u64 offset;
-	int err;
 	u32 reg;
 
 	if (IS_SRIOV_VF(xe)) {
···
 		return 0;
 	}
 
-	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
-	if (err)
-		return err;
+	fw_ref = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
+	if (!fw_ref)
+		return -ETIMEDOUT;
 
 	/* actual size */
 	if (unlikely(xe->info.platform == XE_DG1)) {
···
 	/* remove the tile offset so we have just the available size */
 	*vram_size = offset - *tile_offset;
 
-	return xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
+	xe_force_wake_put(gt_to_fw(gt), fw_ref);
+
+	return 0;
 }
 
 static void vram_fini(void *arg)

drivers/gpu/drm/xe/xe_wa_oob.rules

···
 14019789679	GRAPHICS_VERSION(1255)
 		GRAPHICS_VERSION_RANGE(1270, 2004)
 no_media_l3	MEDIA_VERSION(3000)
[phone]	GRAPHICS_VERSION(3000), GRAPHICS_STEP(A0, B0)
+		MEDIA_VERSION(3000), MEDIA_STEP(A0, B0)

include/uapi/drm/xe_drm.h

···
 	/** @capabilities: OA capabilities bit-mask */
 	__u64 capabilities;
 #define DRM_XE_OA_CAPS_BASE		(1 << 0)
+#define DRM_XE_OA_CAPS_SYNCS		(1 << 1)
 
 	/** @oa_timestamp_freq: OA timestamp freq */
 	__u64 oa_timestamp_freq;
···
 	 * to be disabled for the stream exec queue.
 	 */
 	DRM_XE_OA_PROPERTY_NO_PREEMPT,
+
+	/**
+	 * @DRM_XE_OA_PROPERTY_NUM_SYNCS: Number of syncs in the sync array
+	 * specified in @DRM_XE_OA_PROPERTY_SYNCS
+	 */
+	DRM_XE_OA_PROPERTY_NUM_SYNCS,
+
+	/**
+	 * @DRM_XE_OA_PROPERTY_SYNCS: Pointer to struct @drm_xe_sync array
+	 * with array size specified via @DRM_XE_OA_PROPERTY_NUM_SYNCS. OA
+	 * configuration will wait till input fences signal. Output fences
+	 * will signal after the new OA configuration takes effect. For
+	 * @DRM_XE_SYNC_TYPE_USER_FENCE, @addr is a user pointer, similar
+	 * to the VM bind case.
+	 */
+	DRM_XE_OA_PROPERTY_SYNCS,
 };
 
 /**