Merge: RHEL 9.8: DRM Main Backport (6.17)

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/7329

## Overview

The DRM backport's goal is to backport all the changes in the DRM subsystem to the kernel target version, with the biggest value being that we get the upstream hardware enablement (and bug fixes) into RHEL.

The first step of the DRM backport is identifying dependencies of the DRM backport outside of `drivers/gpu`. Done in https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/7327 for kernel 6.17.

This MR is the second step: Backporting the changes in the DRM subsystem.

## Dependencies:

This MR depends on the 6.16 dependencies and main backports and the 6.17 dependencies backport:

Depends: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/7321
Depends: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/7322
Depends: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/7327

## Omitted fixes:

#### Non applicable changes:

MSM, omapdrm and Mediatek drivers are not supported: \
Omitted-fix: 08c5b42 \
Omitted-fix: 42464c5 \
Omitted-fix: de651b6 \
Omitted-fix: f4ca529 \
Omitted-fix: f8f6e72 \
Omitted-fix: f5b1819 \
Omitted-fix: 86404a9 ("drm/msm: make sure last_fence is always updated") \
Omitted-fix: 00d5f09 ("drm/msm: Ensure vm is created in VM_BIND ioctl") \
Omitted-fix: c34e08b ("drm/msm: Fix GEM free for imported dma-bufs") \
Omitted-fix: c34e08b ("drm/msm: Fix GEM free for imported dma-bufs")

Non Applicable: \
Omitted-fix: 7463755 \
Omitted-fix: e1bf212 \
Omitted-fix: e1bf212

To be included in the 6.18 backport: \
Omitted-fix: 476a4e1 \
Omitted-fix: 92e34a5 \
Omitted-fix: 2650bc4 \
Omitted-fix: 92e34a5 \
Omitted-fix: 278f890 \
Omitted-fix: 3a3bef6 \
Omitted-fix: 1cf1205 \
Omitted-fix: 6ec4b94 \
Omitted-fix: 0a51bf3 \
Omitted-fix: 4d22db6 \
Omitted-fix: 0a6e9e0 \
Omitted-fix: bbfe987 \
Omitted-fix: bba9aa4 \
Omitted-fix: 33f8f32 \
Omitted-fix: 4aa8961 \
Omitted-fix: 20a8e04 \
Omitted-fix: 7413e9f \
Omitted-fix: 1af59cd \
Omitted-fix: ff780f4 \
Omitted-fix: 1f22fcb \
Omitted-fix: 357d90b \
Omitted-fix: 0335594 ("dma-fence: Fix safe access wrapper to call timeline name method") \
Omitted-fix: dc8aa0c ("drm/i915/dmc: Clear HRR EVT_CTL/HTP to zero on ADL-S") \
Omitted-fix: 3257bd1 ("drm/imx: parallel-display: convert to devm_drm_bridge_alloc() API") \
Omitted-fix: 3257bd1 ("drm/imx: parallel-display: convert to devm_drm_bridge_alloc() API")

## Other:

JIRA: https://issues.redhat.com/browse/RHEL-113577
Signed-off-by: José Expósito <jexposit@redhat.com>

Approved-by: Jan Stancek <jstancek@redhat.com>
Approved-by: Jocelyn Falempe <jfalempe@redhat.com>
Approved-by: Anusha Srivatsa <asrivats@redhat.com>
Approved-by: Eder Zulian <ezulian@redhat.com>
Approved-by: David Airlie <airlied@redhat.com>
Approved-by: Enric Balletbo i Serra <eballetbo@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: CKI GitLab Kmaint Pipeline Bot <26919896-cki-kmaint-pipeline-bot@users.noreply.gitlab.com>

Author: CKI KWF Bot

Makefile

@@ -8,7 +8,7 @@ NAME = Opossums on Parade
 # DRM backport version
 #
 RHEL_DRM_VERSION = 6
-RHEL_DRM_PATCHLEVEL = 16
+RHEL_DRM_PATCHLEVEL = 17
 RHEL_DRM_SUBLEVEL = 
 
 # *DOCUMENTATION*

drivers/accel/drm_accel.c

@@ -20,8 +20,6 @@
 
 DEFINE_XARRAY_ALLOC(accel_minors_xa);
 
-static struct dentry *accel_debugfs_root;
-
 static const struct device_type accel_sysfs_device_minor = {
 	.name = "accel_minor"
 };
@@ -73,17 +71,6 @@ static const struct drm_info_list accel_debugfs_list[] = {
 };
 #define ACCEL_DEBUGFS_ENTRIES ARRAY_SIZE(accel_debugfs_list)
 
-/**
- * accel_debugfs_init() - Initialize debugfs for device
- * @dev: Pointer to the device instance.
- *
- * This function creates a root directory for the device in debugfs.
- */
-void accel_debugfs_init(struct drm_device *dev)
-{
-	drm_debugfs_dev_init(dev, accel_debugfs_root);
-}
-
 /**
  * accel_debugfs_register() - Register debugfs for device
  * @dev: Pointer to the device instance.
@@ -194,7 +181,6 @@ static const struct file_operations accel_stub_fops = {
 void accel_core_exit(void)
 {
 	unregister_chrdev(ACCEL_MAJOR, "accel");
-	debugfs_remove(accel_debugfs_root);
 	accel_sysfs_destroy();
 	WARN_ON(!xa_empty(&accel_minors_xa));
 }
@@ -209,8 +195,6 @@ int __init accel_core_init(void)
 		goto error;
 	}
 
-	accel_debugfs_root = debugfs_create_dir("accel", NULL);
-
 	ret = register_chrdev(ACCEL_MAJOR, "accel", &accel_stub_fops);
 	if (ret < 0)
 		DRM_ERROR("Cannot register ACCEL major: %d\n", ret);

drivers/accel/ivpu/ivpu_drv.c

@@ -677,7 +677,7 @@ static void ivpu_bo_unbind_all_user_contexts(struct ivpu_device *vdev)
 static void ivpu_dev_fini(struct ivpu_device *vdev)
 {
 	ivpu_jobs_abort_all(vdev);
-	ivpu_pm_cancel_recovery(vdev);
+	ivpu_pm_disable_recovery(vdev);
 	ivpu_pm_disable(vdev);
 	ivpu_prepare_for_reset(vdev);
 	ivpu_shutdown(vdev);
@@ -704,6 +704,7 @@ static struct pci_device_id ivpu_pci_ids[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_ARL) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_LNL) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_PTL_P) },
+	{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_WCL) },
 	{ }
 };
 MODULE_DEVICE_TABLE(pci, ivpu_pci_ids);

drivers/accel/ivpu/ivpu_drv.h

@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0-only */
 /*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
  */
 
 #ifndef __IVPU_DRV_H__
@@ -26,6 +26,7 @@
 #define PCI_DEVICE_ID_ARL	0xad1d
 #define PCI_DEVICE_ID_LNL	0x643e
 #define PCI_DEVICE_ID_PTL_P	0xb03e
+#define PCI_DEVICE_ID_WCL	0xfd3e
 
 #define IVPU_HW_IP_37XX 37
 #define IVPU_HW_IP_40XX 40
@@ -165,6 +166,7 @@ struct ivpu_device {
 		int boot;
 		int jsm;
 		int tdr;
+		int inference;
 		int autosuspend;
 		int d0i3_entry_msg;
 		int state_dump_msg;
@@ -207,10 +209,11 @@ extern bool ivpu_force_snoop;
 #define IVPU_TEST_MODE_D0I3_MSG_ENABLE    BIT(5)
 #define IVPU_TEST_MODE_MIP_DISABLE        BIT(6)
 #define IVPU_TEST_MODE_DISABLE_TIMEOUTS   BIT(8)
-#define IVPU_TEST_MODE_TURBO		  BIT(9)
-#define IVPU_TEST_MODE_CLK_RELINQ_DISABLE BIT(10)
-#define IVPU_TEST_MODE_CLK_RELINQ_ENABLE  BIT(11)
-#define IVPU_TEST_MODE_D0I2_DISABLE       BIT(12)
+#define IVPU_TEST_MODE_TURBO_ENABLE       BIT(9)
+#define IVPU_TEST_MODE_TURBO_DISABLE      BIT(10)
+#define IVPU_TEST_MODE_CLK_RELINQ_DISABLE BIT(11)
+#define IVPU_TEST_MODE_CLK_RELINQ_ENABLE  BIT(12)
+#define IVPU_TEST_MODE_D0I2_DISABLE       BIT(13)
 extern int ivpu_test_mode;
 
 struct ivpu_file_priv *ivpu_file_priv_get(struct ivpu_file_priv *file_priv);
@@ -240,6 +243,7 @@ static inline int ivpu_hw_ip_gen(struct ivpu_device *vdev)
 	case PCI_DEVICE_ID_LNL:
 		return IVPU_HW_IP_40XX;
 	case PCI_DEVICE_ID_PTL_P:
+	case PCI_DEVICE_ID_WCL:
 		return IVPU_HW_IP_50XX;
 	default:
 		dump_stack();
@@ -256,6 +260,7 @@ static inline int ivpu_hw_btrs_gen(struct ivpu_device *vdev)
 		return IVPU_HW_BTRS_MTL;
 	case PCI_DEVICE_ID_LNL:
 	case PCI_DEVICE_ID_PTL_P:
+	case PCI_DEVICE_ID_WCL:
 		return IVPU_HW_BTRS_LNL;
 	default:
 		dump_stack();

drivers/accel/ivpu/ivpu_hw.c

@@ -94,26 +94,30 @@ static void timeouts_init(struct ivpu_device *vdev)
 		vdev->timeout.boot = -1;
 		vdev->timeout.jsm = -1;
 		vdev->timeout.tdr = -1;
+		vdev->timeout.inference = -1;
 		vdev->timeout.autosuspend = -1;
 		vdev->timeout.d0i3_entry_msg = -1;
 	} else if (ivpu_is_fpga(vdev)) {
 		vdev->timeout.boot = 50;
 		vdev->timeout.jsm = 15000;
 		vdev->timeout.tdr = 30000;
+		vdev->timeout.inference = 900000;
 		vdev->timeout.autosuspend = -1;
 		vdev->timeout.d0i3_entry_msg = 500;
 		vdev->timeout.state_dump_msg = 10000;
 	} else if (ivpu_is_simics(vdev)) {
 		vdev->timeout.boot = 50;
 		vdev->timeout.jsm = 500;
 		vdev->timeout.tdr = 10000;
+		vdev->timeout.inference = 300000;
 		vdev->timeout.autosuspend = 100;
 		vdev->timeout.d0i3_entry_msg = 100;
 		vdev->timeout.state_dump_msg = 10;
 	} else {
 		vdev->timeout.boot = 1000;
 		vdev->timeout.jsm = 500;
 		vdev->timeout.tdr = 2000;
+		vdev->timeout.inference = 60000;
 		if (ivpu_hw_ip_gen(vdev) == IVPU_HW_IP_37XX)
 			vdev->timeout.autosuspend = 10;
 		else

drivers/accel/ivpu/ivpu_hw_ip.c

@@ -683,6 +683,7 @@ static void pwr_island_delay_set(struct ivpu_device *vdev)
 		return;
 
 	switch (ivpu_device_id(vdev)) {
+	case PCI_DEVICE_ID_WCL:
 	case PCI_DEVICE_ID_PTL_P:
 		post = high ? 18 : 0;
 		post1 = 0;

drivers/accel/ivpu/ivpu_job.c

@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 /*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
  */
 
 #include <drm/drm_file.h>
@@ -100,15 +100,51 @@ static struct ivpu_cmdq *ivpu_cmdq_alloc(struct ivpu_file_priv *file_priv)
 	return NULL;
 }
 
+/**
+ * ivpu_cmdq_get_entry_count - Calculate the number of entries in the command queue.
+ * @cmdq: Pointer to the command queue structure.
+ *
+ * Returns the number of entries that can fit in the command queue memory.
+ */
+static inline u32 ivpu_cmdq_get_entry_count(struct ivpu_cmdq *cmdq)
+{
+	size_t size = ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header);
+
+	return size / sizeof(struct vpu_job_queue_entry);
+}
+
+/**
+ * ivpu_cmdq_get_flags - Get command queue flags based on input flags and test mode.
+ * @vdev: Pointer to the ivpu device structure.
+ * @flags: Input flags to determine the command queue flags.
+ *
+ * Returns the calculated command queue flags, considering both the input flags
+ * and the current test mode settings.
+ */
+static u32 ivpu_cmdq_get_flags(struct ivpu_device *vdev, u32 flags)
+{
+	u32 cmdq_flags = 0;
+
+	if ((flags & DRM_IVPU_CMDQ_FLAG_TURBO) && (ivpu_hw_ip_gen(vdev) >= IVPU_HW_IP_40XX))
+		cmdq_flags |= VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+
+	/* Test mode can override the TURBO flag coming from the application */
+	if (ivpu_test_mode & IVPU_TEST_MODE_TURBO_ENABLE)
+		cmdq_flags |= VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+	if (ivpu_test_mode & IVPU_TEST_MODE_TURBO_DISABLE)
+		cmdq_flags &= ~VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+
+	return cmdq_flags;
+}
+
 static void ivpu_cmdq_free(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
 {
 	ivpu_preemption_buffers_free(file_priv->vdev, file_priv, cmdq);
 	ivpu_bo_free(cmdq->mem);
 	kfree(cmdq);
 }
 
-static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 priority,
-					  bool is_legacy)
+static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 priority, u32 flags)
 {
 	struct ivpu_device *vdev = file_priv->vdev;
 	struct ivpu_cmdq *cmdq = NULL;
@@ -121,18 +157,22 @@ static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 p
 		ivpu_err(vdev, "Failed to allocate command queue\n");
 		return NULL;
 	}
-
-	cmdq->priority = priority;
-	cmdq->is_legacy = is_legacy;
-
 	ret = xa_alloc_cyclic(&file_priv->cmdq_xa, &cmdq->id, cmdq, file_priv->cmdq_limit,
 			      &file_priv->cmdq_id_next, GFP_KERNEL);
 	if (ret < 0) {
 		ivpu_err(vdev, "Failed to allocate command queue ID: %d\n", ret);
 		goto err_free_cmdq;
 	}
 
-	ivpu_dbg(vdev, JOB, "Command queue %d created, ctx %d\n", cmdq->id, file_priv->ctx.id);
+	cmdq->entry_count = ivpu_cmdq_get_entry_count(cmdq);
+	cmdq->priority = priority;
+
+	cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
+	cmdq->jobq->header.engine_idx = VPU_ENGINE_COMPUTE;
+	cmdq->jobq->header.flags = ivpu_cmdq_get_flags(vdev, flags);
+
+	ivpu_dbg(vdev, JOB, "Command queue %d created, ctx %d, flags 0x%08x\n",
+		 cmdq->id, file_priv->ctx.id, cmdq->jobq->header.flags);
 	return cmdq;
 
 err_free_cmdq:
@@ -188,27 +228,14 @@ static int ivpu_register_db(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *
 	return ret;
 }
 
-static void ivpu_cmdq_jobq_init(struct ivpu_device *vdev, struct vpu_job_queue *jobq)
+static void ivpu_cmdq_jobq_reset(struct ivpu_device *vdev, struct vpu_job_queue *jobq)
 {
-	jobq->header.engine_idx = VPU_ENGINE_COMPUTE;
 	jobq->header.head = 0;
 	jobq->header.tail = 0;
 
-	if (ivpu_test_mode & IVPU_TEST_MODE_TURBO) {
-		ivpu_dbg(vdev, JOB, "Turbo mode enabled");
-		jobq->header.flags = VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
-	}
-
 	wmb(); /* Flush WC buffer for jobq->header */
 }
 
-static inline u32 ivpu_cmdq_get_entry_count(struct ivpu_cmdq *cmdq)
-{
-	size_t size = ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header);
-
-	return size / sizeof(struct vpu_job_queue_entry);
-}
-
 static int ivpu_cmdq_register(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
 {
 	struct ivpu_device *vdev = file_priv->vdev;
@@ -219,10 +246,7 @@ static int ivpu_cmdq_register(struct ivpu_file_priv *file_priv, struct ivpu_cmdq
 	if (cmdq->db_id)
 		return 0;
 
-	cmdq->entry_count = ivpu_cmdq_get_entry_count(cmdq);
-	cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
-
-	ivpu_cmdq_jobq_init(vdev, cmdq->jobq);
+	ivpu_cmdq_jobq_reset(vdev, cmdq->jobq);
 
 	if (vdev->fw->sched_mode == VPU_SCHEDULING_MODE_HW) {
 		ret = ivpu_hws_cmdq_init(file_priv, cmdq, VPU_ENGINE_COMPUTE, cmdq->priority);
@@ -291,9 +315,10 @@ static struct ivpu_cmdq *ivpu_cmdq_acquire_legacy(struct ivpu_file_priv *file_pr
 			break;
 
 	if (!cmdq) {
-		cmdq = ivpu_cmdq_create(file_priv, priority, true);
+		cmdq = ivpu_cmdq_create(file_priv, priority, 0);
 		if (!cmdq)
 			return NULL;
+		cmdq->is_legacy = true;
 	}
 
 	return cmdq;
@@ -891,7 +916,7 @@ int ivpu_cmdq_create_ioctl(struct drm_device *dev, void *data, struct drm_file *
 
 	mutex_lock(&file_priv->lock);
 
-	cmdq = ivpu_cmdq_create(file_priv, ivpu_job_to_jsm_priority(args->priority), false);
+	cmdq = ivpu_cmdq_create(file_priv, ivpu_job_to_jsm_priority(args->priority), args->flags);
 	if (cmdq)
 		args->cmdq_id = cmdq->id;
 

drivers/accel/ivpu/ivpu_pm.c

@@ -33,8 +33,11 @@ static unsigned long ivpu_tdr_timeout_ms;
 module_param_named(tdr_timeout_ms, ivpu_tdr_timeout_ms, ulong, 0644);
 MODULE_PARM_DESC(tdr_timeout_ms, "Timeout for device hang detection, in milliseconds, 0 - default");
 
+static unsigned long ivpu_inference_timeout_ms;
+module_param_named(inference_timeout_ms, ivpu_inference_timeout_ms, ulong, 0644);
+MODULE_PARM_DESC(inference_timeout_ms, "Inference maximum duration, in milliseconds, 0 - default");
+
 #define PM_RESCHEDULE_LIMIT     5
-#define PM_TDR_HEARTBEAT_LIMIT  30
 
 static void ivpu_pm_prepare_cold_boot(struct ivpu_device *vdev)
 {
@@ -191,15 +194,21 @@ static void ivpu_job_timeout_work(struct work_struct *work)
 {
 	struct ivpu_pm_info *pm = container_of(work, struct ivpu_pm_info, job_timeout_work.work);
 	struct ivpu_device *vdev = pm->vdev;
+	unsigned long timeout_ms = ivpu_tdr_timeout_ms ? ivpu_tdr_timeout_ms : vdev->timeout.tdr;
+	unsigned long inference_timeout_ms = ivpu_inference_timeout_ms ? ivpu_inference_timeout_ms :
+					     vdev->timeout.inference;
+	u64 inference_max_retries;
 	u64 heartbeat;
 
 	if (ivpu_jsm_get_heartbeat(vdev, 0, &heartbeat) || heartbeat <= vdev->fw->last_heartbeat) {
 		ivpu_err(vdev, "Job timeout detected, heartbeat not progressed\n");
 		goto recovery;
 	}
 
-	if (atomic_fetch_inc(&vdev->job_timeout_counter) > PM_TDR_HEARTBEAT_LIMIT) {
-		ivpu_err(vdev, "Job timeout detected, heartbeat limit exceeded\n");
+	inference_max_retries = DIV_ROUND_UP(inference_timeout_ms, timeout_ms);
+	if (atomic_fetch_inc(&vdev->job_timeout_counter) >= inference_max_retries) {
+		ivpu_err(vdev, "Job timeout detected, heartbeat limit (%lld) exceeded\n",
+			 inference_max_retries);
 		goto recovery;
 	}
 
@@ -408,10 +417,10 @@ void ivpu_pm_init(struct ivpu_device *vdev)
 	ivpu_dbg(vdev, PM, "Autosuspend delay = %d\n", delay);
 }
 
-void ivpu_pm_cancel_recovery(struct ivpu_device *vdev)
+void ivpu_pm_disable_recovery(struct ivpu_device *vdev)
 {
 	drm_WARN_ON(&vdev->drm, delayed_work_pending(&vdev->pm->job_timeout_work));
-	cancel_work_sync(&vdev->pm->recovery_work);
+	disable_work_sync(&vdev->pm->recovery_work);
 }
 
 void ivpu_pm_enable(struct ivpu_device *vdev)

drivers/accel/ivpu/ivpu_pm.h

@@ -25,7 +25,7 @@ struct ivpu_pm_info {
 void ivpu_pm_init(struct ivpu_device *vdev);
 void ivpu_pm_enable(struct ivpu_device *vdev);
 void ivpu_pm_disable(struct ivpu_device *vdev);
-void ivpu_pm_cancel_recovery(struct ivpu_device *vdev);
+void ivpu_pm_disable_recovery(struct ivpu_device *vdev);
 
 int ivpu_pm_suspend_cb(struct device *dev);
 int ivpu_pm_resume_cb(struct device *dev);

drivers/dma-buf/Kconfig

@@ -36,7 +36,6 @@ config UDMABUF
 	depends on DMA_SHARED_BUFFER
 	depends on MEMFD_CREATE || COMPILE_TEST
 	depends on MMU
-	select VMAP_PFN
 	help
 	  A driver to let userspace turn memfd regions into dma-bufs.
 	  Qemu can use this to create host dmabufs for guest framebuffers.