drm merge window fixes part 1

cgroup:
 - fix Koncfig fallout from new dmem controller
 
 Driver Changes:
 - v3d NULL pointer regression fix in fence signalling race
 - virtio: uaf in dma_buf free path
 - xlnx: fix kerneldoc
 - bochs: fix double-free on driver removal
 - zynqmp: add missing locking to DP bridge driver
 
 - amdgpu fixes all over: documentation, display, sriov, various hw block
   drivers
 - amdgpu: use drm/sched helper
 - amdgpu: mark some debug module options as unsafe
 - amdkfd: mark some debug module options as unsafe, trap handler
   updates, fix partial migration handling
 
 DRM core:
 - client: fix fbdev Kconfig select rules, improve tiled-based display
   support
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEEciDa6W7nX7VwIxa1EpWk/0GBDXwFAmeXhyMACgkQEpWk/0GB
 DXwdqA/+Nd/V0WeZingyt4MxhwF068xG9RJLl6EkcdvdJI08MPh0FbqLjGbMiBsa
 o8dcmQWsxecAFxODwV5CtLtQ0/igHalTav0n28lJ9hr1QxQWLdeDX4x0RN6JwX2l
 xp7GT6e+L5wgc0TSBh6zh7mt4Q0nRIY89GCk1nUI2vY12LUep/upG413HfTvAw56
 62qS2i6UZt1IMv8Syv0RiIm5jLJvdpVuDG43nlpK/z6PpzRVtaPOWCI120kTRPKc
 Xp/dUohf6cqIipQBaSEPQ5f45GlMPWbY7PCoX/mKXSTpbNsXnscbH9BELQIVkR3g
 APh/GHyG2qq7dmhdEs8oxpOwkPJv9LLRlyh9p4M4kDdZxr0VI+6cbgb75Tm6n/Xe
 rRCMuY13APeXMg2WQfi/uQjauCzdC/4fhP77h6PhjXkgb8g7VTYPncExlI+QEJGX
 8Ey+nZPvp5LmohXtfmFNvwoDy8O2C0JfymvLN94HxOCe6i3aDPQm1pnlrYnxbp8N
 hgFV8dnn0w2qoEO0CFVqr3k8q/r4S1wR4GXNA8wGKI5rUJSn8vSzQshvs5i9KD2h
 1T3fqq+i/JiBHNB+MxuSBMggHb0zY9CQ2zjbWo+dz8Rx3jAff2whilWGmZltA5/u
 fesjOgMO/fpTzcc4kdwFQDiD0YMlGLWtoakvoYp4FFSKlqd4N94=
 =sfwb
 -----END PGP SIGNATURE-----

Merge tag 'drm-next-2025-01-27' of https://gitlab.freedesktop.org/drm/kernel

Pull drm fixes from Simona Vetter:
 "cgroup:
   - fix Koncfig fallout from new dmem controller

  Driver Changes:
   - v3d NULL pointer regression fix in fence signalling race
   - virtio: uaf in dma_buf free path
   - xlnx: fix kerneldoc
   - bochs: fix double-free on driver removal
   - zynqmp: add missing locking to DP bridge driver
   - amdgpu fixes all over:
       - documentation, display, sriov, various hw block drivers
       - use drm/sched helper
       - mark some debug module options as unsafe
       - amdkfd: mark some debug module options as unsafe, trap handler
         updates, fix partial migration handling

  DRM core:
   - fix fbdev Kconfig select rules, improve tiled-based display
     support"

* tag 'drm-next-2025-01-27' of https://gitlab.freedesktop.org/drm/kernel: (40 commits)
  drm/amd/display: Optimize cursor position updates
  drm/amd/display: Add hubp cache reset when powergating
  drm/amd/amdgpu: Enable scratch data dump for mes 12
  drm/amd: Clarify kdoc for amdgpu.gttsize
  drm/amd/amdgpu: Prevent null pointer dereference in GPU bandwidth calculation
  drm/amd/display: Fix error pointers in amdgpu_dm_crtc_mem_type_changed
  drm/amdgpu: fix ring timeout issue in gfx10 sr-iov environment
  drm/amd/pm: Fix smu v13.0.6 caps initialization
  drm/amd/pm: Refactor SMU 13.0.6 SDMA reset firmware version checks
  revert "drm/amdgpu/pm: add definition PPSMC_MSG_ResetSDMA2"
  revert "drm/amdgpu/pm: Implement SDMA queue reset for different asic"
  drm/amd/pm: Add capability flags for SMU v13.0.6
  drm/amd/display: fix SUBVP DC_DEBUG_MASK documentation
  drm/amd/display: fix CEC DC_DEBUG_MASK documentation
  drm/amdgpu: fix the PCIe lanes reporting in the INFO IOCTL
  drm/amdgpu: cache gpu pcie link width
  drm/amd/display: mark static functions noinline_for_stack
  drm/amdkfd: Clear MODE.VSKIP in gfx9 trap handler
  drm/amdgpu: Refine ip detection log message
  drm/amdgpu: Add handler for SDMA context empty
  ...

Documentation/gpu/zynqmp.rst

@@ -144,6 +144,4 @@ Internals
 
 .. kernel-doc:: drivers/gpu/drm/xlnx/zynqmp_dp.c
 
-.. kernel-doc:: drivers/gpu/drm/xlnx/zynqmp_dpsub.c
-
 .. kernel-doc:: drivers/gpu/drm/xlnx/zynqmp_kms.c

drivers/gpu/drm/Kconfig

@@ -294,6 +294,7 @@ config DRM_TTM_HELPER
 	tristate
 	depends on DRM
 	select DRM_TTM
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
 	select FB_CORE if DRM_FBDEV_EMULATION
 	select FB_SYSMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help
@@ -302,6 +303,7 @@ config DRM_TTM_HELPER
 config DRM_GEM_DMA_HELPER
 	tristate
 	depends on DRM
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
 	select FB_CORE if DRM_FBDEV_EMULATION
 	select FB_DMAMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help
@@ -310,6 +312,7 @@ config DRM_GEM_DMA_HELPER
 config DRM_GEM_SHMEM_HELPER
 	tristate
 	depends on DRM && MMU
+	select DRM_KMS_HELPER if DRM_FBDEV_EMULATION
 	select FB_CORE if DRM_FBDEV_EMULATION
 	select FB_SYSMEM_HELPERS_DEFERRED if DRM_FBDEV_EMULATION
 	help

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -2365,8 +2365,8 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
 		break;
 	}
 
-	DRM_INFO("add ip block number %d <%s>\n", adev->num_ip_blocks,
+	dev_info(adev->dev, "detected ip block number %d <%s>\n",
-		  ip_block_version->funcs->name);
+		 adev->num_ip_blocks, ip_block_version->funcs->name);
 
 	adev->ip_blocks[adev->num_ip_blocks].adev = adev;
 
@@ -6157,6 +6157,44 @@ static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev,
 	}
 }
 
+/**
+ * amdgpu_device_gpu_bandwidth - find the bandwidth of the GPU
+ *
+ * @adev: amdgpu_device pointer
+ * @speed: pointer to the speed of the link
+ * @width: pointer to the width of the link
+ *
+ * Evaluate the hierarchy to find the speed and bandwidth capabilities of the
+ * AMD dGPU which may be a virtual upstream bridge.
+ */
+static void amdgpu_device_gpu_bandwidth(struct amdgpu_device *adev,
+					enum pci_bus_speed *speed,
+					enum pcie_link_width *width)
+{
+	struct pci_dev *parent = adev->pdev;
+
+	if (!speed || !width)
+		return;
+
+	parent = pci_upstream_bridge(parent);
+	if (parent && parent->vendor == PCI_VENDOR_ID_ATI) {
+		/* use the upstream/downstream switches internal to dGPU */
+		*speed = pcie_get_speed_cap(parent);
+		*width = pcie_get_width_cap(parent);
+		while ((parent = pci_upstream_bridge(parent))) {
+			if (parent->vendor == PCI_VENDOR_ID_ATI) {
+				/* use the upstream/downstream switches internal to dGPU */
+				*speed = pcie_get_speed_cap(parent);
+				*width = pcie_get_width_cap(parent);
+			}
+		}
+	} else {
+		/* use the device itself */
+		*speed = pcie_get_speed_cap(adev->pdev);
+		*width = pcie_get_width_cap(adev->pdev);
+	}
+}
+
 /**
  * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot
  *
@@ -6168,9 +6206,8 @@ static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev,
  */
 static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 {
-	struct pci_dev *pdev;
 	enum pci_bus_speed speed_cap, platform_speed_cap;
-	enum pcie_link_width platform_link_width;
+	enum pcie_link_width platform_link_width, link_width;
 
 	if (amdgpu_pcie_gen_cap)
 		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
@@ -6192,11 +6229,10 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 
 	amdgpu_device_partner_bandwidth(adev, &platform_speed_cap,
 					&platform_link_width);
+	amdgpu_device_gpu_bandwidth(adev, &speed_cap, &link_width);
 
 	if (adev->pm.pcie_gen_mask == 0) {
 		/* asic caps */
-		pdev = adev->pdev;
-		speed_cap = pcie_get_speed_cap(pdev);
 		if (speed_cap == PCI_SPEED_UNKNOWN) {
 			adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 |
 						  CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 |
@@ -6252,51 +6288,103 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 		}
 	}
 	if (adev->pm.pcie_mlw_mask == 0) {
+		/* asic caps */
+		if (link_width == PCIE_LNK_WIDTH_UNKNOWN) {
+			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_ASIC_PCIE_MLW_MASK;
+		} else {
+			switch (link_width) {
+			case PCIE_LNK_X32:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X16:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X12:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X8:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X4:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X2:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X1:
+				adev->pm.pcie_mlw_mask |= CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1;
+				break;
+			default:
+				break;
+			}
+		}
+		/* platform caps */
 		if (platform_link_width == PCIE_LNK_WIDTH_UNKNOWN) {
 			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_PCIE_MLW_MASK;
 		} else {
 			switch (platform_link_width) {
 			case PCIE_LNK_X32:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X16:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X12:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X8:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X4:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X2:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X1:
-				adev->pm.pcie_mlw_mask = CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
+				adev->pm.pcie_mlw_mask |= CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
 				break;
 			default:
 				break;

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -280,7 +280,7 @@ module_param_named(gartsize, amdgpu_gart_size, uint, 0600);
 /**
  * DOC: gttsize (int)
  * Restrict the size of GTT domain (for userspace use) in MiB for testing.
- * The default is -1 (Use 1/2 RAM, minimum value is 3GB).
+ * The default is -1 (Use value specified by TTM).
  */
 MODULE_PARM_DESC(gttsize, "Size of the GTT userspace domain in megabytes (-1 = auto)");
 module_param_named(gttsize, amdgpu_gtt_size, int, 0600);
@@ -399,7 +399,7 @@ module_param_named(runpm, amdgpu_runtime_pm, int, 0444);
  * the kernel log for the list of IPs on the asic. The default is 0xffffffff (enable all blocks on a device).
  */
 MODULE_PARM_DESC(ip_block_mask, "IP Block Mask (all blocks enabled (default))");
-module_param_named(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
+module_param_named_unsafe(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
 
 /**
  * DOC: bapm (int)
@@ -457,7 +457,7 @@ module_param_named(vm_update_mode, amdgpu_vm_update_mode, int, 0444);
  * Enable experimental hw support (1 = enable). The default is 0 (disabled).
  */
 MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))");
-module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
+module_param_named_unsafe(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
 
 /**
  * DOC: dc (int)
@@ -568,14 +568,14 @@ module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444);
  * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV).
  */
 MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (1 = enable, 0 = disable, -1 = auto)");
-module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
+module_param_named_unsafe(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
 
 /**
  * DOC: emu_mode (int)
  * Set value 1 to enable emulation mode. This is only needed when running on an emulator. The default is 0 (disabled).
  */
 MODULE_PARM_DESC(emu_mode, "Emulation mode, (1 = enable, 0 = disable)");
-module_param_named(emu_mode, amdgpu_emu_mode, int, 0444);
+module_param_named_unsafe(emu_mode, amdgpu_emu_mode, int, 0444);
 
 /**
  * DOC: ras_enable (int)
@@ -730,7 +730,7 @@ module_param_named(noretry, amdgpu_noretry, int, 0644);
  */
 MODULE_PARM_DESC(force_asic_type,
 	"A non negative value used to specify the asic type for all supported GPUs");
-module_param_named(force_asic_type, amdgpu_force_asic_type, int, 0444);
+module_param_named_unsafe(force_asic_type, amdgpu_force_asic_type, int, 0444);
 
 /**
  * DOC: use_xgmi_p2p (int)
@@ -749,7 +749,7 @@ module_param_named(use_xgmi_p2p, amdgpu_use_xgmi_p2p, int, 0444);
  * assigns queues to HQDs.
  */
 int sched_policy = KFD_SCHED_POLICY_HWS;
-module_param(sched_policy, int, 0444);
+module_param_unsafe(sched_policy, int, 0444);
 MODULE_PARM_DESC(sched_policy,
 	"Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)");
 
@@ -799,7 +799,7 @@ MODULE_PARM_DESC(send_sigterm,
  * Setting 1 enables halt on hang.
  */
 int halt_if_hws_hang;
-module_param(halt_if_hws_hang, int, 0644);
+module_param_unsafe(halt_if_hws_hang, int, 0644);
 MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (default), 1 = on)");
 
 /**
@@ -808,7 +808,7 @@ MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (defau
  * check says. Default value: false (rely on MEC2 firmware version check).
  */
 bool hws_gws_support;
-module_param(hws_gws_support, bool, 0444);
+module_param_unsafe(hws_gws_support, bool, 0444);
 MODULE_PARM_DESC(hws_gws_support, "Assume MEC2 FW supports GWS barriers (false = rely on FW version check (Default), true = force supported)");
 
 /**
@@ -841,7 +841,7 @@ MODULE_PARM_DESC(no_system_mem_limit, "disable system memory limit (false = defa
  */
 int amdgpu_no_queue_eviction_on_vm_fault;
 MODULE_PARM_DESC(no_queue_eviction_on_vm_fault, "No queue eviction on VM fault (0 = queue eviction, 1 = no queue eviction)");
-module_param_named(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
+module_param_named_unsafe(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
 #endif
 
 /**
@@ -849,7 +849,7 @@ module_param_named(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm
  */
 int amdgpu_mtype_local;
 MODULE_PARM_DESC(mtype_local, "MTYPE for local memory (0 = MTYPE_RW (default), 1 = MTYPE_NC, 2 = MTYPE_CC)");
-module_param_named(mtype_local, amdgpu_mtype_local, int, 0444);
+module_param_named_unsafe(mtype_local, amdgpu_mtype_local, int, 0444);
 
 /**
  * DOC: pcie_p2p (bool)
@@ -953,7 +953,7 @@ module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444);
  * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)
  */
 MODULE_PARM_DESC(reset_method, "GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco/bamaco)");
-module_param_named(reset_method, amdgpu_reset_method, int, 0644);
+module_param_named_unsafe(reset_method, amdgpu_reset_method, int, 0644);
 
 /**
  * DOC: bad_page_threshold (int) Bad page threshold is specifies the
@@ -1049,7 +1049,7 @@ module_param_named(seamless, amdgpu_seamless, int, 0444);
  * - 0x4: Disable GPU soft recovery, always do a full reset
  */
 MODULE_PARM_DESC(debug_mask, "debug options for amdgpu, disabled by default");
-module_param_named(debug_mask, amdgpu_debug_mask, uint, 0444);
+module_param_named_unsafe(debug_mask, amdgpu_debug_mask, uint, 0444);
 
 /**
  * DOC: agp (int)

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

@@ -297,7 +297,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
-	if (vm && ring->funcs->emit_switch_buffer)
+	if (job && ring->funcs->emit_switch_buffer)
 		amdgpu_ring_emit_switch_buffer(ring);
 
 	if (ring->funcs->emit_wave_limit &&

drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c

@@ -846,7 +846,7 @@ out:
 	case AMDGPU_INFO_DEV_INFO: {
 		struct drm_amdgpu_info_device *dev_info;
 		uint64_t vm_size;
-		uint32_t pcie_gen_mask;
+		uint32_t pcie_gen_mask, pcie_width_mask;
 
 		dev_info = kzalloc(sizeof(*dev_info), GFP_KERNEL);
 		if (!dev_info)
@@ -934,15 +934,18 @@ out:
 		dev_info->tcc_disabled_mask = adev->gfx.config.tcc_disabled_mask;
 
 		/* Combine the chip gen mask with the platform (CPU/mobo) mask. */
-		pcie_gen_mask = adev->pm.pcie_gen_mask & (adev->pm.pcie_gen_mask >> 16);
+		pcie_gen_mask = adev->pm.pcie_gen_mask &
+			(adev->pm.pcie_gen_mask >> CAIL_PCIE_LINK_SPEED_SUPPORT_SHIFT);
+		pcie_width_mask = adev->pm.pcie_mlw_mask &
+			(adev->pm.pcie_mlw_mask >> CAIL_PCIE_LINK_WIDTH_SUPPORT_SHIFT);
 		dev_info->pcie_gen = fls(pcie_gen_mask);
 		dev_info->pcie_num_lanes =
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
 
 		dev_info->tcp_cache_size = adev->gfx.config.gc_tcp_l1_size;
 		dev_info->num_sqc_per_wgp = adev->gfx.config.gc_num_sqc_per_wgp;

drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h

@@ -40,7 +40,7 @@
 #define AMDGPU_MES_VERSION_MASK		0x00000fff
 #define AMDGPU_MES_API_VERSION_MASK	0x00fff000
 #define AMDGPU_MES_FEAT_VERSION_MASK	0xff000000
-#define AMDGPU_MES_MSCRATCH_SIZE	0x8000
+#define AMDGPU_MES_MSCRATCH_SIZE	0x40000
 
 enum amdgpu_mes_priority_level {
 	AMDGPU_MES_PRIORITY_LEVEL_LOW       = 0,

drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h

@@ -107,6 +107,7 @@ struct amdgpu_sdma {
 	struct amdgpu_irq_src	doorbell_invalid_irq;
 	struct amdgpu_irq_src	pool_timeout_irq;
 	struct amdgpu_irq_src	srbm_write_irq;
+	struct amdgpu_irq_src	ctxt_empty_irq;
 
 	int			num_instances;
 	uint32_t 		sdma_mask;

drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c

@@ -427,7 +427,7 @@ void amdgpu_xcp_release_sched(struct amdgpu_device *adev,
 		return;
 
 	sched = entity->entity.rq->sched;
-	if (sched->ready) {
+	if (drm_sched_wqueue_ready(sched)) {
 		ring = to_amdgpu_ring(entity->entity.rq->sched);
 		atomic_dec(&adev->xcp_mgr->xcp[ring->xcp_id].ref_cnt);
 	}

drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c

@@ -1352,6 +1352,14 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
 	}
 
 	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (adev->gfx.me_fw_version  >= 2480 &&
+		    adev->gfx.pfp_fw_version >= 2530 &&
+		    adev->gfx.mec_fw_version >= 2680 &&
+		    adev->mes.fw_version[0] >= 100)
+			adev->gfx.enable_cleaner_shader = true;
+		break;
 	default:
 		adev->gfx.enable_cleaner_shader = false;
 		break;

drivers/gpu/drm/amd/amdgpu/mes_v12_0.c

@@ -756,7 +756,8 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe)
 
 	if (amdgpu_mes_log_enable) {
 		mes_set_hw_res_pkt.enable_mes_event_int_logging = 1;
-		mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr = mes->event_log_gpu_addr + pipe * AMDGPU_MES_LOG_BUFFER_SIZE;
+		mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr = mes->event_log_gpu_addr +
+				pipe * (AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE);
 	}
 
 	if (enforce_isolation)
@@ -983,29 +984,50 @@ static void mes_v12_0_enable(struct amdgpu_device *adev, bool enable)
 	uint32_t pipe, data = 0;
 
 	if (enable) {
-		data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1);
-		WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
-
 		mutex_lock(&adev->srbm_mutex);
 		for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) {
 			soc21_grbm_select(adev, 3, pipe, 0, 0);
+			if (amdgpu_mes_log_enable) {
+				u32 log_size = AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE;
+				/* In case uni mes is not enabled, only program for pipe 0 */
+				if (adev->mes.event_log_size >= (pipe + 1) * log_size) {
+					WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO,
+						     lower_32_bits(adev->mes.event_log_gpu_addr +
+						     pipe * log_size + AMDGPU_MES_LOG_BUFFER_SIZE));
+					WREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI,
+						     upper_32_bits(adev->mes.event_log_gpu_addr +
+						     pipe * log_size + AMDGPU_MES_LOG_BUFFER_SIZE));
+					dev_info(adev->dev, "Setup CP MES MSCRATCH address : 0x%x. 0x%x\n",
+						 RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_HI),
+						 RREG32_SOC15(GC, 0, regCP_MES_MSCRATCH_LO));
+				}
+			}
+
+			data = RREG32_SOC15(GC, 0, regCP_MES_CNTL);
+			if (pipe == 0)
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1);
+			else
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1);
+			WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
 
 			ucode_addr = adev->mes.uc_start_addr[pipe] >> 2;
 			WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START,
 				     lower_32_bits(ucode_addr));
 			WREG32_SOC15(GC, 0, regCP_MES_PRGRM_CNTR_START_HI,
 				     upper_32_bits(ucode_addr));
+
+			/* unhalt MES and activate one pipe each loop */
+			data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1);
+			if (pipe)
+				data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 1);
+			dev_info(adev->dev, "program CP_MES_CNTL : 0x%x\n", data);
+
+			WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
+
 		}
 		soc21_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		/* unhalt MES and activate pipe0 */
-		data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1);
-		data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 1);
-		WREG32_SOC15(GC, 0, regCP_MES_CNTL, data);
-
 		if (amdgpu_emu_mode)
 			msleep(100);
 		else if (adev->enable_uni_mes)
@@ -1479,8 +1501,9 @@ static int mes_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
 	adev->mes.kiq_hw_fini = &mes_v12_0_kiq_hw_fini;
 	adev->mes.enable_legacy_queue_map = true;
 
-	adev->mes.event_log_size = adev->enable_uni_mes ? (AMDGPU_MAX_MES_PIPES * AMDGPU_MES_LOG_BUFFER_SIZE) : AMDGPU_MES_LOG_BUFFER_SIZE;
+	adev->mes.event_log_size = adev->enable_uni_mes ?
-
+		(AMDGPU_MAX_MES_PIPES * (AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE)) :
+		(AMDGPU_MES_LOG_BUFFER_SIZE + AMDGPU_MES_MSCRATCH_SIZE);
 	r = amdgpu_mes_init(adev);
 	if (r)
 		return r;

drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c

@@ -1406,6 +1406,12 @@ static int sdma_v4_4_2_sw_init(struct amdgpu_ip_block *ip_block)
 				      &adev->sdma.srbm_write_irq);
 		if (r)
 			return r;
+
+		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
+				      SDMA0_4_0__SRCID__SDMA_CTXEMPTY,
+				      &adev->sdma.ctxt_empty_irq);
+		if (r)
+			return r;
 	}
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
@@ -1814,6 +1820,16 @@ static int sdma_v4_4_2_process_srbm_write_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int sdma_v4_4_2_process_ctxt_empty_irq(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      struct amdgpu_iv_entry *entry)
+{
+	/* There is nothing useful to be done here, only kept for debug */
+	dev_dbg_ratelimited(adev->dev, "SDMA context empty interrupt");
+	sdma_v4_4_2_print_iv_entry(adev, entry);
+	return 0;
+}
+
 static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
 	struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
 {
@@ -2096,6 +2112,10 @@ static const struct amdgpu_irq_src_funcs sdma_v4_4_2_srbm_write_irq_funcs = {
 	.process = sdma_v4_4_2_process_srbm_write_irq,
 };
 
+static const struct amdgpu_irq_src_funcs sdma_v4_4_2_ctxt_empty_irq_funcs = {
+	.process = sdma_v4_4_2_process_ctxt_empty_irq,
+};
+
 static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 {
 	adev->sdma.trap_irq.num_types = adev->sdma.num_instances;
@@ -2104,6 +2124,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 	adev->sdma.doorbell_invalid_irq.num_types = adev->sdma.num_instances;
 	adev->sdma.pool_timeout_irq.num_types = adev->sdma.num_instances;
 	adev->sdma.srbm_write_irq.num_types = adev->sdma.num_instances;
+	adev->sdma.ctxt_empty_irq.num_types = adev->sdma.num_instances;
 
 	adev->sdma.trap_irq.funcs = &sdma_v4_4_2_trap_irq_funcs;
 	adev->sdma.illegal_inst_irq.funcs = &sdma_v4_4_2_illegal_inst_irq_funcs;
@@ -2112,6 +2133,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev)
 	adev->sdma.doorbell_invalid_irq.funcs = &sdma_v4_4_2_doorbell_invalid_irq_funcs;
 	adev->sdma.pool_timeout_irq.funcs = &sdma_v4_4_2_pool_timeout_irq_funcs;
 	adev->sdma.srbm_write_irq.funcs = &sdma_v4_4_2_srbm_write_irq_funcs;
+	adev->sdma.ctxt_empty_irq.funcs = &sdma_v4_4_2_ctxt_empty_irq_funcs;
 }
 
 /**

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm

@@ -447,7 +447,9 @@ L_SAVE:
     s_getreg_b32    s_save_m0, hwreg(HW_REG_MODE)						    //MODE
     write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset)
 
-
+    // Clear VSKIP state now that MODE.VSKIP has been saved.
+    // If user shader set it then vector instructions would be skipped.
+    s_setvskip	0,0
 
     /*	    the first wave in the threadgroup	 */
     s_and_b32	    s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK	// extract fisrt wave bit

drivers/gpu/drm/amd/amdkfd/kfd_migrate.c

@@ -278,10 +278,11 @@ svm_migrate_copy_to_vram(struct kfd_node *node, struct svm_range *prange,
 			 struct migrate_vma *migrate, struct dma_fence **mfence,
 			 dma_addr_t *scratch, uint64_t ttm_res_offset)
 {
-	uint64_t npages = migrate->cpages;
+	uint64_t npages = migrate->npages;
 	struct amdgpu_device *adev = node->adev;
 	struct device *dev = adev->dev;
 	struct amdgpu_res_cursor cursor;
+	uint64_t mpages = 0;
 	dma_addr_t *src;
 	uint64_t *dst;
 	uint64_t i, j;
@@ -295,14 +296,16 @@ svm_migrate_copy_to_vram(struct kfd_node *node, struct svm_range *prange,
 
 	amdgpu_res_first(prange->ttm_res, ttm_res_offset,
 			 npages << PAGE_SHIFT, &cursor);
-	for (i = j = 0; i < npages; i++) {
+	for (i = j = 0; (i < npages) && (mpages < migrate->cpages); i++) {
 		struct page *spage;
 
-		dst[i] = cursor.start + (j << PAGE_SHIFT);
+		if (migrate->src[i] & MIGRATE_PFN_MIGRATE) {
-		migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
+			dst[i] = cursor.start + (j << PAGE_SHIFT);
-		svm_migrate_get_vram_page(prange, migrate->dst[i]);
+			migrate->dst[i] = svm_migrate_addr_to_pfn(adev, dst[i]);
-		migrate->dst[i] = migrate_pfn(migrate->dst[i]);
+			svm_migrate_get_vram_page(prange, migrate->dst[i]);
-
+			migrate->dst[i] = migrate_pfn(migrate->dst[i]);
+			mpages++;
+		}
 		spage = migrate_pfn_to_page(migrate->src[i]);
 		if (spage && !is_zone_device_page(spage)) {
 			src[i] = dma_map_page(dev, spage, 0, PAGE_SIZE,
@@ -353,9 +356,12 @@ svm_migrate_copy_to_vram(struct kfd_node *node, struct svm_range *prange,
 out_free_vram_pages:
 	if (r) {
 		pr_debug("failed %d to copy memory to vram\n", r);
-		while (i--) {
+		for (i = 0; i < npages && mpages; i++) {
+			if (!dst[i])
+				continue;
 			svm_migrate_put_vram_page(adev, dst[i]);
 			migrate->dst[i] = 0;
+			mpages--;
 		}
 	}
 

drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

@@ -2034,6 +2034,9 @@ static int amdgpu_dm_init(struct amdgpu_device *adev)
 	if (amdgpu_dc_debug_mask & DC_FORCE_SUBVP_MCLK_SWITCH)
 		adev->dm.dc->debug.force_subvp_mclk_switch = true;
 
+	if (amdgpu_dc_debug_mask & DC_DISABLE_SUBVP)
+		adev->dm.dc->debug.force_disable_subvp = true;
+
 	if (amdgpu_dc_debug_mask & DC_ENABLE_DML2) {
 		adev->dm.dc->debug.using_dml2 = true;
 		adev->dm.dc->debug.using_dml21 = true;
@@ -11483,6 +11486,11 @@ static bool amdgpu_dm_crtc_mem_type_changed(struct drm_device *dev,
 		new_plane_state = drm_atomic_get_plane_state(state, plane);
 		old_plane_state = drm_atomic_get_plane_state(state, plane);
 
+		if (IS_ERR(new_plane_state) || IS_ERR(old_plane_state)) {
+			DRM_ERROR("Failed to get plane state for plane %s\n", plane->name);
+			return false;
+		}
+
 		if (old_plane_state->fb && new_plane_state->fb &&
 		    get_mem_type(old_plane_state->fb) != get_mem_type(new_plane_state->fb))
 			return true;

drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c

@@ -1736,7 +1736,7 @@ static void CalculateBytePerPixelAndBlockSizes(
 #endif
 } // CalculateBytePerPixelAndBlockSizes
 
-static dml_float_t CalculateTWait(
+static noinline_for_stack dml_float_t CalculateTWait(
 		dml_uint_t PrefetchMode,
 		enum dml_use_mall_for_pstate_change_mode UseMALLForPStateChange,
 		dml_bool_t SynchronizeDRRDisplaysForUCLKPStateChangeFinal,
@@ -4458,7 +4458,7 @@ static void CalculateSwathWidth(
 	}
 } // CalculateSwathWidth
 
-static  dml_float_t CalculateExtraLatency(
+static noinline_for_stack dml_float_t CalculateExtraLatency(
 		dml_uint_t RoundTripPingLatencyCycles,
 		dml_uint_t ReorderingBytes,
 		dml_float_t DCFCLK,
@@ -5915,7 +5915,7 @@ static dml_uint_t DSCDelayRequirement(
 	return DSCDelayRequirement_val;
 }
 
-static dml_bool_t CalculateVActiveBandwithSupport(dml_uint_t NumberOfActiveSurfaces,
+static noinline_for_stack dml_bool_t CalculateVActiveBandwithSupport(dml_uint_t NumberOfActiveSurfaces,
 										dml_float_t ReturnBW,
 										dml_bool_t NotUrgentLatencyHiding[],
 										dml_float_t ReadBandwidthLuma[],
@@ -6019,7 +6019,7 @@ static void CalculatePrefetchBandwithSupport(
 #endif
 }
 
-static dml_float_t CalculateBandwidthAvailableForImmediateFlip(
+static noinline_for_stack dml_float_t CalculateBandwidthAvailableForImmediateFlip(
 													dml_uint_t NumberOfActiveSurfaces,
 													dml_float_t ReturnBW,
 													dml_float_t ReadBandwidthLuma[],
@@ -6213,7 +6213,7 @@ static dml_uint_t CalculateMaxVStartup(
 	return max_vstartup_lines;
 }
 
-static void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *mode_lib,
+static noinline_for_stack void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *mode_lib,
 						   struct CalculatePrefetchSchedule_params_st *CalculatePrefetchSchedule_params,
 						   dml_uint_t j,
 						   dml_uint_t k)
@@ -6265,7 +6265,7 @@ static void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *m
 				CalculatePrefetchSchedule_params->Tno_bw = &mode_lib->ms.Tno_bw[k];
 }
 
-static void dml_prefetch_check(struct display_mode_lib_st *mode_lib)
+static noinline_for_stack void dml_prefetch_check(struct display_mode_lib_st *mode_lib)
 {
 	struct dml_core_mode_support_locals_st *s = &mode_lib->scratch.dml_core_mode_support_locals;
 	struct CalculatePrefetchSchedule_params_st *CalculatePrefetchSchedule_params = &mode_lib->scratch.CalculatePrefetchSchedule_params;

drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c

@@ -2778,7 +2778,7 @@ static double dml_get_return_bandwidth_available(
 	return return_bw_mbps;
 }
 
-static void calculate_bandwidth_available(
+static noinline_for_stack void calculate_bandwidth_available(
 	double avg_bandwidth_available_min[dml2_core_internal_soc_state_max],
 	double avg_bandwidth_available[dml2_core_internal_soc_state_max][dml2_core_internal_bw_max],
 	double urg_bandwidth_available_min[dml2_core_internal_soc_state_max], // min between SDP and DRAM
@@ -3625,7 +3625,7 @@ static void CalculateDCFCLKDeepSleepTdlut(
 	dml2_printf("DML::%s: DCFClkDeepSleep = %f (final)\n", __func__, *DCFClkDeepSleep);
 }
 
-static void CalculateDCFCLKDeepSleep(
+static noinline_for_stack void CalculateDCFCLKDeepSleep(
 	const struct dml2_display_cfg *display_cfg,
 	unsigned int NumberOfActiveSurfaces,
 	unsigned int BytePerPixelY[],
@@ -4142,7 +4142,7 @@ static bool ValidateODMMode(enum dml2_odm_mode ODMMode,
 	return true;
 }
 
-static void CalculateODMMode(
+static noinline_for_stack void CalculateODMMode(
 	unsigned int MaximumPixelsPerLinePerDSCUnit,
 	unsigned int HActive,
 	enum dml2_output_format_class OutFormat,
@@ -4239,7 +4239,7 @@ static void CalculateODMMode(
 #endif
 }
 
-static void CalculateOutputLink(
+static noinline_for_stack void CalculateOutputLink(
 	struct dml2_core_internal_scratch *s,
 	double PHYCLK,
 	double PHYCLKD18,
@@ -5999,7 +5999,7 @@ static double calculate_impacted_Tsw(unsigned int exclude_plane_idx, unsigned in
 }
 
 // a global check against the aggregate effect of the per plane prefetch schedule
-static bool CheckGlobalPrefetchAdmissibility(struct dml2_core_internal_scratch *scratch,
+static noinline_for_stack bool CheckGlobalPrefetchAdmissibility(struct dml2_core_internal_scratch *scratch,
 											 struct dml2_core_calcs_CheckGlobalPrefetchAdmissibility_params *p)
 {
 	struct dml2_core_calcs_CheckGlobalPrefetchAdmissibility_locals *s = &scratch->CheckGlobalPrefetchAdmissibility_locals;
@@ -7012,7 +7012,7 @@ static void calculate_bytes_to_fetch_required_to_hide_latency(
 	}
 }
 
-static void calculate_vactive_det_fill_latency(
+static noinline_for_stack void calculate_vactive_det_fill_latency(
 		const struct dml2_display_cfg *display_cfg,
 		unsigned int num_active_planes,
 		unsigned int bytes_required_l[],

drivers/gpu/drm/amd/display/dc/dpp/dcn10/dcn10_dpp.c

@@ -194,6 +194,9 @@ void dpp_reset(struct dpp *dpp_base)
 	dpp->filter_h = NULL;
 	dpp->filter_v = NULL;
 
+	memset(&dpp_base->pos, 0, sizeof(dpp_base->pos));
+	memset(&dpp_base->att, 0, sizeof(dpp_base->att));
+
 	memset(&dpp->scl_data, 0, sizeof(dpp->scl_data));
 	memset(&dpp->pwl_data, 0, sizeof(dpp->pwl_data));
 }
@@ -480,10 +483,11 @@ void dpp1_set_cursor_position(
 	if (src_y_offset + cursor_height <= 0)
 		cur_en = 0;  /* not visible beyond top edge*/
 
-	REG_UPDATE(CURSOR0_CONTROL,
+	if (dpp_base->pos.cur0_ctl.bits.cur0_enable != cur_en) {
-			CUR0_ENABLE, cur_en);
+		REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en);
 
-	dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en;
+		dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en;
+	}
 }
 
 void dpp1_cnv_set_optional_cursor_attributes(

drivers/gpu/drm/amd/display/dc/dpp/dcn401/dcn401_dpp_cm.c

@@ -154,9 +154,11 @@ void dpp401_set_cursor_position(
 	struct dcn401_dpp *dpp = TO_DCN401_DPP(dpp_base);
 	uint32_t cur_en = pos->enable ? 1 : 0;
 
-	REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en);
+	if (dpp_base->pos.cur0_ctl.bits.cur0_enable != cur_en) {
+		REG_UPDATE(CURSOR0_CONTROL, CUR0_ENABLE, cur_en);
 
-	dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en;
+		dpp_base->pos.cur0_ctl.bits.cur0_enable = cur_en;
+	}
 }
 
 void dpp401_set_optional_cursor_attributes(

drivers/gpu/drm/amd/display/dc/hubp/dcn10/dcn10_hubp.c

@@ -546,6 +546,12 @@ void hubp1_dcc_control(struct hubp *hubp, bool enable,
 			SECONDARY_SURFACE_DCC_IND_64B_BLK, dcc_ind_64b_blk);
 }
 
+void hubp_reset(struct hubp *hubp)
+{
+	memset(&hubp->pos, 0, sizeof(hubp->pos));
+	memset(&hubp->att, 0, sizeof(hubp->att));
+}
+
 void hubp1_program_surface_config(
 	struct hubp *hubp,
 	enum surface_pixel_format format,
@@ -1351,8 +1357,9 @@ static void hubp1_wait_pipe_read_start(struct hubp *hubp)
 
 void hubp1_init(struct hubp *hubp)
 {
-	//do nothing
+	hubp_reset(hubp);
 }
+
 static const struct hubp_funcs dcn10_hubp_funcs = {
 	.hubp_program_surface_flip_and_addr =
 			hubp1_program_surface_flip_and_addr,
@@ -1365,6 +1372,7 @@ static const struct hubp_funcs dcn10_hubp_funcs = {
 	.hubp_set_vm_context0_settings = hubp1_set_vm_context0_settings,
 	.set_blank = hubp1_set_blank,
 	.dcc_control = hubp1_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_hubp_blank_en = hubp1_set_hubp_blank_en,
 	.set_cursor_attributes	= hubp1_cursor_set_attributes,

drivers/gpu/drm/amd/display/dc/hubp/dcn10/dcn10_hubp.h

@@ -746,6 +746,8 @@ void hubp1_dcc_control(struct hubp *hubp,
 		bool enable,
 		enum hubp_ind_block_size independent_64b_blks);
 
+void hubp_reset(struct hubp *hubp);
+
 bool hubp1_program_surface_flip_and_addr(
 	struct hubp *hubp,
 	const struct dc_plane_address *address,

drivers/gpu/drm/amd/display/dc/hubp/dcn20/dcn20_hubp.c

@@ -1058,11 +1058,13 @@ void hubp2_cursor_set_position(
 	if (src_y_offset + cursor_height <= 0)
 		cur_en = 0;  /* not visible beyond top edge*/
 
-	if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0)
+	if (hubp->pos.cur_ctl.bits.cur_enable != cur_en) {
-		hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr);
+		if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0)
+			hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr);
 
-	REG_UPDATE(CURSOR_CONTROL,
+		REG_UPDATE(CURSOR_CONTROL,
 			CURSOR_ENABLE, cur_en);
+	}
 
 	REG_SET_2(CURSOR_POSITION, 0,
 			CURSOR_X_POSITION, pos->x,
@@ -1674,6 +1676,7 @@ static struct hubp_funcs dcn20_hubp_funcs = {
 	.set_blank = hubp2_set_blank,
 	.set_blank_regs = hubp2_set_blank_regs,
 	.dcc_control = hubp2_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_cursor_attributes	= hubp2_cursor_set_attributes,
 	.set_cursor_position	= hubp2_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn201/dcn201_hubp.c

@@ -121,6 +121,7 @@ static struct hubp_funcs dcn201_hubp_funcs = {
 	.set_cursor_position	= hubp1_cursor_set_position,
 	.set_blank = hubp1_set_blank,
 	.dcc_control = hubp1_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.hubp_clk_cntl = hubp1_clk_cntl,
 	.hubp_vtg_sel = hubp1_vtg_sel,

drivers/gpu/drm/amd/display/dc/hubp/dcn21/dcn21_hubp.c

@@ -811,6 +811,8 @@ static void hubp21_init(struct hubp *hubp)
 	struct dcn21_hubp *hubp21 = TO_DCN21_HUBP(hubp);
 	//hubp[i].HUBPREQ_DEBUG.HUBPREQ_DEBUG[26] = 1;
 	REG_WRITE(HUBPREQ_DEBUG, 1 << 26);
+
+	hubp_reset(hubp);
 }
 static struct hubp_funcs dcn21_hubp_funcs = {
 	.hubp_enable_tripleBuffer = hubp2_enable_triplebuffer,
@@ -823,6 +825,7 @@ static struct hubp_funcs dcn21_hubp_funcs = {
 	.hubp_set_vm_system_aperture_settings = hubp21_set_vm_system_aperture_settings,
 	.set_blank = hubp1_set_blank,
 	.dcc_control = hubp1_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = hubp21_set_viewport,
 	.set_cursor_attributes	= hubp2_cursor_set_attributes,
 	.set_cursor_position	= hubp1_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn30/dcn30_hubp.c

@@ -499,6 +499,8 @@ void hubp3_init(struct hubp *hubp)
 	struct dcn20_hubp *hubp2 = TO_DCN20_HUBP(hubp);
 	//hubp[i].HUBPREQ_DEBUG.HUBPREQ_DEBUG[26] = 1;
 	REG_WRITE(HUBPREQ_DEBUG, 1 << 26);
+
+	hubp_reset(hubp);
 }
 
 static struct hubp_funcs dcn30_hubp_funcs = {
@@ -513,6 +515,7 @@ static struct hubp_funcs dcn30_hubp_funcs = {
 	.set_blank = hubp2_set_blank,
 	.set_blank_regs = hubp2_set_blank_regs,
 	.dcc_control = hubp3_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_cursor_attributes	= hubp2_cursor_set_attributes,
 	.set_cursor_position	= hubp2_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn31/dcn31_hubp.c

@@ -79,6 +79,7 @@ static struct hubp_funcs dcn31_hubp_funcs = {
 	.hubp_set_vm_system_aperture_settings = hubp3_set_vm_system_aperture_settings,
 	.set_blank = hubp2_set_blank,
 	.dcc_control = hubp3_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_cursor_attributes	= hubp2_cursor_set_attributes,
 	.set_cursor_position	= hubp2_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn32/dcn32_hubp.c

@@ -181,6 +181,7 @@ static struct hubp_funcs dcn32_hubp_funcs = {
 	.set_blank = hubp2_set_blank,
 	.set_blank_regs = hubp2_set_blank_regs,
 	.dcc_control = hubp3_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_cursor_attributes	= hubp32_cursor_set_attributes,
 	.set_cursor_position	= hubp2_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn35/dcn35_hubp.c

@@ -199,6 +199,7 @@ static struct hubp_funcs dcn35_hubp_funcs = {
 	.hubp_set_vm_system_aperture_settings = hubp3_set_vm_system_aperture_settings,
 	.set_blank = hubp2_set_blank,
 	.dcc_control = hubp3_dcc_control,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = min_set_viewport,
 	.set_cursor_attributes	= hubp2_cursor_set_attributes,
 	.set_cursor_position	= hubp2_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hubp/dcn401/dcn401_hubp.c

@@ -141,7 +141,7 @@ void hubp401_update_mall_sel(struct hubp *hubp, uint32_t mall_sel, bool c_cursor
 
 void hubp401_init(struct hubp *hubp)
 {
-	//For now nothing to do, HUBPREQ_DEBUG_DB register is removed on DCN4x.
+	hubp_reset(hubp);
 }
 
 void hubp401_vready_at_or_After_vsync(struct hubp *hubp,
@@ -742,11 +742,13 @@ void hubp401_cursor_set_position(
 			dc_fixpt_from_int(dst_x_offset),
 			param->h_scale_ratio));
 
-	if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0)
+	if (hubp->pos.cur_ctl.bits.cur_enable != cur_en) {
-		hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr);
+		if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0)
+			hubp->funcs->set_cursor_attributes(hubp, &hubp->curs_attr);
 
-	REG_UPDATE(CURSOR_CONTROL,
+		REG_UPDATE(CURSOR_CONTROL,
-		CURSOR_ENABLE, cur_en);
+			CURSOR_ENABLE, cur_en);
+	}
 
 	REG_SET_2(CURSOR_POSITION, 0,
 		CURSOR_X_POSITION, x_pos,
@@ -998,6 +1000,7 @@ static struct hubp_funcs dcn401_hubp_funcs = {
 	.hubp_set_vm_system_aperture_settings = hubp3_set_vm_system_aperture_settings,
 	.set_blank = hubp2_set_blank,
 	.set_blank_regs = hubp2_set_blank_regs,
+	.hubp_reset = hubp_reset,
 	.mem_program_viewport = hubp401_set_viewport,
 	.set_cursor_attributes	= hubp32_cursor_set_attributes,
 	.set_cursor_position	= hubp401_cursor_set_position,

drivers/gpu/drm/amd/display/dc/hwss/dcn10/dcn10_hwseq.c

@@ -1286,6 +1286,7 @@ void dcn10_plane_atomic_power_down(struct dc *dc,
 		if (hws->funcs.hubp_pg_control)
 			hws->funcs.hubp_pg_control(hws, hubp->inst, false);
 
+		hubp->funcs->hubp_reset(hubp);
 		dpp->funcs->dpp_reset(dpp);
 
 		REG_SET(DC_IP_REQUEST_CNTL, 0,
@@ -1447,6 +1448,7 @@ void dcn10_init_pipes(struct dc *dc, struct dc_state *context)
 		/* Disable on the current state so the new one isn't cleared. */
 		pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i];
 
+		hubp->funcs->hubp_reset(hubp);
 		dpp->funcs->dpp_reset(dpp);
 
 		pipe_ctx->stream_res.tg = tg;

drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_hwseq.c

@@ -800,6 +800,7 @@ void dcn35_init_pipes(struct dc *dc, struct dc_state *context)
 		/* Disable on the current state so the new one isn't cleared. */
 		pipe_ctx = &dc->current_state->res_ctx.pipe_ctx[i];
 
+		hubp->funcs->hubp_reset(hubp);
 		dpp->funcs->dpp_reset(dpp);
 
 		pipe_ctx->stream_res.tg = tg;
@@ -956,6 +957,7 @@ void dcn35_plane_atomic_disable(struct dc *dc, struct pipe_ctx *pipe_ctx)
 /*to do, need to support both case*/
 	hubp->power_gated = true;
 
+	hubp->funcs->hubp_reset(hubp);
 	dpp->funcs->dpp_reset(dpp);
 
 	pipe_ctx->stream = NULL;

drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h

@@ -163,6 +163,8 @@ struct hubp_funcs {
 	void (*dcc_control)(struct hubp *hubp, bool enable,
 			enum hubp_ind_block_size blk_size);
 
+	void (*hubp_reset)(struct hubp *hubp);
+
 	void (*mem_program_viewport)(
 			struct hubp *hubp,
 			const struct rect *viewport,

drivers/gpu/drm/amd/include/amd_pcie.h

@@ -49,6 +49,17 @@
 				      | CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN3)
 
 /* Following flags shows PCIe lane width switch supported in driver which are decided by chipset and ASIC */
+
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1          0x00000001
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2          0x00000002
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4          0x00000004
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8          0x00000008
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12         0x00000010
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16         0x00000020
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32         0x00000040
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_MASK        0x0000FFFF
+#define CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_SHIFT       0
+
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X1          0x00010000
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X2          0x00020000
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X4          0x00040000
@@ -56,6 +67,7 @@
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X12         0x00100000
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X16         0x00200000
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_X32         0x00400000
+#define CAIL_PCIE_LINK_WIDTH_SUPPORT_MASK        0xFFFF0000
 #define CAIL_PCIE_LINK_WIDTH_SUPPORT_SHIFT       16
 
 /* 1/2/4/8/16 lanes */
@@ -65,4 +77,10 @@
 				      | CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 \
 				      | CAIL_PCIE_LINK_WIDTH_SUPPORT_X16)
 
+#define AMDGPU_DEFAULT_ASIC_PCIE_MLW_MASK (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1 \
+					   | CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 \
+					   | CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 \
+					   | CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 \
+					   | CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16)
+
 #endif

drivers/gpu/drm/amd/include/amd_shared.h

@@ -345,10 +345,15 @@ enum DC_DEBUG_MASK {
 	 */
 	DC_DISABLE_ACPI_EDID = 0x8000,
 
-	/*
+	/**
 	 * @DC_DISABLE_HDMI_CEC: If set, disable HDMI-CEC feature in amdgpu driver.
 	 */
 	DC_DISABLE_HDMI_CEC = 0x10000,
+
+	/**
+	 * @DC_DISABLE_SUBVP: If set, disable DCN Sub-Viewport feature in amdgpu driver.
+	 */
+	DC_DISABLE_SUBVP = 0x20000,
 };
 
 enum amd_dpm_forced_level;

drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_ppsmc.h

@@ -93,7 +93,6 @@
 #define PPSMC_MSG_SelectPLPDMode                    0x40
 #define PPSMC_MSG_RmaDueToBadPageThreshold          0x43
 #define PPSMC_MSG_SelectPstatePolicy                0x44
-#define PPSMC_MSG_ResetSDMA2                        0x45
 #define PPSMC_MSG_ResetSDMA                         0x4D
 #define PPSMC_Message_Count                         0x4E
 

drivers/gpu/drm/amd/pm/swsmu/inc/smu_types.h

@@ -276,8 +276,7 @@
 	__SMU_DUMMY_MAP(SelectPstatePolicy), \
 	__SMU_DUMMY_MAP(MALLPowerController), \
 	__SMU_DUMMY_MAP(MALLPowerState), \
-	__SMU_DUMMY_MAP(ResetSDMA), \
+	__SMU_DUMMY_MAP(ResetSDMA),
-	__SMU_DUMMY_MAP(ResetSDMA2),
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type

drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h

@@ -107,6 +107,7 @@ struct smu_13_0_dpm_context {
 	struct smu_13_0_dpm_tables  dpm_tables;
 	uint32_t                    workload_policy_mask;
 	uint32_t                    dcef_min_ds_clk;
+	uint64_t                    caps;
 };
 
 enum smu_13_0_power_state {

drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c

@@ -101,38 +101,24 @@ MODULE_FIRMWARE("amdgpu/smu_13_0_14.bin");
 #define MCA_BANK_IPID(_ip, _hwid, _type) \
 	[AMDGPU_MCA_IP_##_ip] = { .hwid = _hwid, .mcatype = _type, }
 
-static inline bool smu_v13_0_6_is_unified_metrics(struct smu_context *smu)
+#define SMU_CAP(x) SMU_13_0_6_CAPS_##x
-{
-	return (smu->adev->flags & AMD_IS_APU) &&
-		smu->smc_fw_version <= 0x4556900;
-}
 
-static inline bool smu_v13_0_6_is_other_end_count_available(struct smu_context *smu)
+enum smu_v13_0_6_caps {
-{
+	SMU_CAP(DPM),
-	switch (amdgpu_ip_version(smu->adev, MP1_HWIP, 0)) {
+	SMU_CAP(UNI_METRICS),
-	case IP_VERSION(13, 0, 6):
+	SMU_CAP(DPM_POLICY),
-		return smu->smc_fw_version >= 0x557600;
+	SMU_CAP(OTHER_END_METRICS),
-	case IP_VERSION(13, 0, 14):
+	SMU_CAP(SET_UCLK_MAX),
-		return smu->smc_fw_version >= 0x05550E00;
+	SMU_CAP(PCIE_METRICS),
-	default:
+	SMU_CAP(HST_LIMIT_METRICS),
-		return false;
+	SMU_CAP(MCA_DEBUG_MODE),
-	}
+	SMU_CAP(PER_INST_METRICS),
-}
+	SMU_CAP(CTF_LIMIT),
-
+	SMU_CAP(RMA_MSG),
-static inline bool smu_v13_0_6_is_blw_host_limit_available(struct smu_context *smu)
+	SMU_CAP(ACA_SYND),
-{
+	SMU_CAP(SDMA_RESET),
-	if (smu->adev->flags & AMD_IS_APU)
+	SMU_CAP(ALL),
-		return smu->smc_fw_version >= 0x04556F00;
+};
-
-	switch (amdgpu_ip_version(smu->adev, MP1_HWIP, 0)) {
-	case IP_VERSION(13, 0, 6):
-		return smu->smc_fw_version >= 0x557900;
-	case IP_VERSION(13, 0, 14):
-		return smu->smc_fw_version >= 0x05551000;
-	default:
-		return false;
-	}
-}
 
 struct mca_bank_ipid {
 	enum amdgpu_mca_ip ip;
@@ -209,7 +195,6 @@ static const struct cmn2asic_msg_mapping smu_v13_0_6_message_map[SMU_MSG_MAX_COU
 	MSG_MAP(RmaDueToBadPageThreshold,            PPSMC_MSG_RmaDueToBadPageThreshold,        0),
 	MSG_MAP(SelectPstatePolicy,                  PPSMC_MSG_SelectPstatePolicy,              0),
 	MSG_MAP(ResetSDMA,                           PPSMC_MSG_ResetSDMA,                       0),
-	MSG_MAP(ResetSDMA2,                          PPSMC_MSG_ResetSDMA2,                      0),
 };
 
 // clang-format on
@@ -297,6 +282,162 @@ struct smu_v13_0_6_dpm_map {
 	uint32_t *freq_table;
 };
 
+static inline void smu_v13_0_6_cap_set(struct smu_context *smu,
+				       enum smu_v13_0_6_caps cap)
+{
+	struct smu_13_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context;
+
+	dpm_context->caps |= BIT_ULL(cap);
+}
+
+static inline void smu_v13_0_6_cap_clear(struct smu_context *smu,
+					 enum smu_v13_0_6_caps cap)
+{
+	struct smu_13_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context;
+
+	dpm_context->caps &= ~BIT_ULL(cap);
+}
+
+static inline bool smu_v13_0_6_cap_supported(struct smu_context *smu,
+					     enum smu_v13_0_6_caps cap)
+{
+	struct smu_13_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context;
+
+	return !!(dpm_context->caps & BIT_ULL(cap));
+}
+
+static void smu_v13_0_14_init_caps(struct smu_context *smu)
+{
+	enum smu_v13_0_6_caps default_cap_list[] = { SMU_CAP(DPM),
+						     SMU_CAP(UNI_METRICS),
+						     SMU_CAP(SET_UCLK_MAX),
+						     SMU_CAP(DPM_POLICY),
+						     SMU_CAP(PCIE_METRICS),
+						     SMU_CAP(CTF_LIMIT),
+						     SMU_CAP(MCA_DEBUG_MODE),
+						     SMU_CAP(RMA_MSG),
+						     SMU_CAP(ACA_SYND) };
+	uint32_t fw_ver = smu->smc_fw_version;
+
+	for (int i = 0; i < ARRAY_SIZE(default_cap_list); i++)
+		smu_v13_0_6_cap_set(smu, default_cap_list[i]);
+
+	if (fw_ver >= 0x05550E00)
+		smu_v13_0_6_cap_set(smu, SMU_CAP(OTHER_END_METRICS));
+	if (fw_ver >= 0x05551000)
+		smu_v13_0_6_cap_set(smu, SMU_CAP(HST_LIMIT_METRICS));
+	if (fw_ver >= 0x05550B00)
+		smu_v13_0_6_cap_set(smu, SMU_CAP(PER_INST_METRICS));
+	if (fw_ver >= 0x5551200)
+		smu_v13_0_6_cap_set(smu, SMU_CAP(SDMA_RESET));
+}
+
+static void smu_v13_0_12_init_caps(struct smu_context *smu)
+{
+	enum smu_v13_0_6_caps default_cap_list[] = { SMU_CAP(DPM),
+						     SMU_CAP(UNI_METRICS),
+						     SMU_CAP(PCIE_METRICS),
+						     SMU_CAP(CTF_LIMIT),
+						     SMU_CAP(MCA_DEBUG_MODE),
+						     SMU_CAP(RMA_MSG),
+						     SMU_CAP(ACA_SYND) };
+	uint32_t fw_ver = smu->smc_fw_version;
+
+	for (int i = 0; i < ARRAY_SIZE(default_cap_list); i++)
+		smu_v13_0_6_cap_set(smu, default_cap_list[i]);
+
+	if (fw_ver < 0x00561900)
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(DPM));
+
+	if (fw_ver >= 0x00561700)
+		smu_v13_0_6_cap_set(smu, SMU_CAP(SDMA_RESET));
+}
+
+static void smu_v13_0_6_init_caps(struct smu_context *smu)
+{
+	enum smu_v13_0_6_caps default_cap_list[] = { SMU_CAP(DPM),
+						     SMU_CAP(UNI_METRICS),
+						     SMU_CAP(SET_UCLK_MAX),
+						     SMU_CAP(DPM_POLICY),
+						     SMU_CAP(PCIE_METRICS),
+						     SMU_CAP(CTF_LIMIT),
+						     SMU_CAP(MCA_DEBUG_MODE),
+						     SMU_CAP(RMA_MSG),
+						     SMU_CAP(ACA_SYND) };
+	struct amdgpu_device *adev = smu->adev;
+	uint32_t fw_ver = smu->smc_fw_version;
+	uint32_t pgm = (fw_ver >> 24) & 0xFF;
+
+	for (int i = 0; i < ARRAY_SIZE(default_cap_list); i++)
+		smu_v13_0_6_cap_set(smu, default_cap_list[i]);
+
+	if (fw_ver < 0x552F00)
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(DPM));
+	if (fw_ver < 0x554500)
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(CTF_LIMIT));
+
+	if (adev->flags & AMD_IS_APU) {
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(PCIE_METRICS));
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(DPM_POLICY));
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(RMA_MSG));
+		smu_v13_0_6_cap_clear(smu, SMU_CAP(ACA_SYND));
+
+		if (fw_ver <= 0x4556900)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(UNI_METRICS));
+		if (fw_ver >= 0x04556F00)
+			smu_v13_0_6_cap_set(smu, SMU_CAP(HST_LIMIT_METRICS));
+		if (fw_ver >= 0x04556A00)
+			smu_v13_0_6_cap_set(smu, SMU_CAP(PER_INST_METRICS));
+	} else {
+		if (fw_ver >= 0x557600)
+			smu_v13_0_6_cap_set(smu, SMU_CAP(OTHER_END_METRICS));
+		if (fw_ver < 0x00556000)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(DPM_POLICY));
+		if (amdgpu_sriov_vf(adev) && (fw_ver < 0x556600))
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(SET_UCLK_MAX));
+		if (fw_ver < 0x556300)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(PCIE_METRICS));
+		if (fw_ver < 0x554800)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(MCA_DEBUG_MODE));
+		if (fw_ver >= 0x556F00)
+			smu_v13_0_6_cap_set(smu, SMU_CAP(PER_INST_METRICS));
+		if (fw_ver < 0x00555a00)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(RMA_MSG));
+		if (fw_ver < 0x00555600)
+			smu_v13_0_6_cap_clear(smu, SMU_CAP(ACA_SYND));
+		if (pgm == 0 && fw_ver >= 0x557900)
+			smu_v13_0_6_cap_set(smu, SMU_CAP(HST_LIMIT_METRICS));
+	}
+	if (((pgm == 7) && (fw_ver >= 0x7550700)) ||
+	    ((pgm == 0) && (fw_ver >= 0x00557900)) ||
+	    ((pgm == 4) && (fw_ver >= 0x4557000)))
+		smu_v13_0_6_cap_set(smu, SMU_CAP(SDMA_RESET));
+}
+
+static void smu_v13_0_x_init_caps(struct smu_context *smu)
+{
+	switch (amdgpu_ip_version(smu->adev, MP1_HWIP, 0)) {
+	case IP_VERSION(13, 0, 12):
+		return smu_v13_0_12_init_caps(smu);
+	case IP_VERSION(13, 0, 14):
+		return smu_v13_0_14_init_caps(smu);
+	default:
+		return smu_v13_0_6_init_caps(smu);
+	}
+}
+
+static int smu_v13_0_6_check_fw_version(struct smu_context *smu)
+{
+	int r;
+
+	r = smu_v13_0_check_fw_version(smu);
+	/* Initialize caps flags once fw version is fetched */
+	if (!r)
+		smu_v13_0_x_init_caps(smu);
+
+	return r;
+}
+
 static int smu_v13_0_6_init_microcode(struct smu_context *smu)
 {
 	const struct smc_firmware_header_v2_1 *v2_1;
@@ -618,7 +759,7 @@ static int smu_v13_0_6_setup_driver_pptable(struct smu_context *smu)
 	MetricsTableA_t *metrics_a = (MetricsTableA_t *)smu_table->metrics_table;
 	struct PPTable_t *pptable =
 		(struct PPTable_t *)smu_table->driver_pptable;
-	bool flag = smu_v13_0_6_is_unified_metrics(smu);
+	bool flag = !smu_v13_0_6_cap_supported(smu, SMU_CAP(UNI_METRICS));
 	int ret, i, retry = 100;
 	uint32_t table_version;
 
@@ -814,8 +955,7 @@ static int smu_v13_0_6_set_default_dpm_table(struct smu_context *smu)
 	smu_v13_0_6_setup_driver_pptable(smu);
 
 	/* DPM policy not supported in older firmwares */
-	if (!(smu->adev->flags & AMD_IS_APU) &&
+	if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(DPM_POLICY))) {
-	    (smu->smc_fw_version < 0x00556000)) {
 		struct smu_dpm_context *smu_dpm = &smu->smu_dpm;
 
 		smu_dpm->dpm_policies->policy_mask &=
@@ -992,7 +1132,7 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu,
 	struct smu_table_context *smu_table = &smu->smu_table;
 	MetricsTableX_t *metrics_x = (MetricsTableX_t *)smu_table->metrics_table;
 	MetricsTableA_t *metrics_a = (MetricsTableA_t *)smu_table->metrics_table;
-	bool flag = smu_v13_0_6_is_unified_metrics(smu);
+	bool flag = !smu_v13_0_6_cap_supported(smu, SMU_CAP(UNI_METRICS));
 	struct amdgpu_device *adev = smu->adev;
 	int ret = 0;
 	int xcc_id;
@@ -1005,7 +1145,7 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu,
 	switch (member) {
 	case METRICS_CURR_GFXCLK:
 	case METRICS_AVERAGE_GFXCLK:
-		if (smu->smc_fw_version >= 0x552F00) {
+		if (smu_v13_0_6_cap_supported(smu, SMU_CAP(DPM))) {
 			xcc_id = GET_INST(GC, 0);
 			*value = SMUQ10_ROUND(GET_METRIC_FIELD(GfxclkFrequency, flag)[xcc_id]);
 		} else {
@@ -1692,7 +1832,7 @@ static int smu_v13_0_6_notify_unload(struct smu_context *smu)
 static int smu_v13_0_6_mca_set_debug_mode(struct smu_context *smu, bool enable)
 {
 	/* NOTE: this ClearMcaOnRead message is only supported for smu version 85.72.0 or higher */
-	if (smu->smc_fw_version < 0x554800)
+	if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(MCA_DEBUG_MODE)))
 		return 0;
 
 	return smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ClearMcaOnRead,
@@ -1837,9 +1977,8 @@ static int smu_v13_0_6_set_soft_freq_limited_range(struct smu_context *smu,
 			if (max == pstate_table->uclk_pstate.curr.max)
 				return 0;
 			/* For VF, only allowed in FW versions 85.102 or greater */
-			if (amdgpu_sriov_vf(adev) &&
+			if (!smu_v13_0_6_cap_supported(smu,
-			    ((smu->smc_fw_version < 0x556600) ||
+						       SMU_CAP(SET_UCLK_MAX)))
-			     (adev->flags & AMD_IS_APU)))
 				return -EOPNOTSUPP;
 			/* Only max clock limiting is allowed for UCLK */
 			ret = smu_v13_0_set_soft_freq_limited_range(
@@ -2043,7 +2182,7 @@ static int smu_v13_0_6_get_enabled_mask(struct smu_context *smu,
 
 	ret = smu_cmn_get_enabled_mask(smu, feature_mask);
 
-	if (ret == -EIO && smu->smc_fw_version < 0x552F00) {
+	if (ret == -EIO && !smu_v13_0_6_cap_supported(smu, SMU_CAP(DPM))) {
 		*feature_mask = 0;
 		ret = 0;
 	}
@@ -2336,11 +2475,10 @@ static int smu_v13_0_6_get_current_pcie_link_speed(struct smu_context *smu)
 
 static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table)
 {
-	bool per_inst, smu_13_0_6_per_inst, smu_13_0_14_per_inst, apu_per_inst;
 	struct smu_table_context *smu_table = &smu->smu_table;
 	struct gpu_metrics_v1_7 *gpu_metrics =
 		(struct gpu_metrics_v1_7 *)smu_table->gpu_metrics_table;
-	bool flag = smu_v13_0_6_is_unified_metrics(smu);
+	bool flag = !smu_v13_0_6_cap_supported(smu, SMU_CAP(UNI_METRICS));
 	int ret = 0, xcc_id, inst, i, j, k, idx;
 	struct amdgpu_device *adev = smu->adev;
 	MetricsTableX_t *metrics_x;
@@ -2348,6 +2486,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 	struct amdgpu_xcp *xcp;
 	u16 link_width_level;
 	u32 inst_mask;
+	bool per_inst;
 
 	metrics_x = kzalloc(max(sizeof(MetricsTableX_t), sizeof(MetricsTableA_t)), GFP_KERNEL);
 	ret = smu_v13_0_6_get_metrics_table(smu, metrics_x, true);
@@ -2421,7 +2560,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 		 * table for both pf & one vf for smu version 85.99.0 or higher else report only
 		 * for pf from registers
 		 */
-		if (smu->smc_fw_version >= 0x556300) {
+		if (smu_v13_0_6_cap_supported(smu, SMU_CAP(PCIE_METRICS))) {
 			gpu_metrics->pcie_link_width = metrics_x->PCIeLinkWidth;
 			gpu_metrics->pcie_link_speed =
 				pcie_gen_to_speed(metrics_x->PCIeLinkSpeed);
@@ -2450,7 +2589,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 				metrics_x->PCIeNAKSentCountAcc;
 		gpu_metrics->pcie_nak_rcvd_count_acc =
 				metrics_x->PCIeNAKReceivedCountAcc;
-		if (smu_v13_0_6_is_other_end_count_available(smu))
+		if (smu_v13_0_6_cap_supported(smu, SMU_CAP(OTHER_END_METRICS)))
 			gpu_metrics->pcie_lc_perf_other_end_recovery =
 				metrics_x->PCIeOtherEndRecoveryAcc;
 
@@ -2475,17 +2614,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 
 	gpu_metrics->num_partition = adev->xcp_mgr->num_xcps;
 
-	apu_per_inst = (adev->flags & AMD_IS_APU) && (smu->smc_fw_version >= 0x04556A00);
+	per_inst = smu_v13_0_6_cap_supported(smu, SMU_CAP(PER_INST_METRICS));
-	smu_13_0_6_per_inst = !(adev->flags & AMD_IS_APU) &&
-				(amdgpu_ip_version(smu->adev, MP1_HWIP, 0)
-				 == IP_VERSION(13, 0, 6)) &&
-				(smu->smc_fw_version >= 0x556F00);
-	smu_13_0_14_per_inst = !(adev->flags & AMD_IS_APU) &&
-				(amdgpu_ip_version(smu->adev, MP1_HWIP, 0)
-				 == IP_VERSION(13, 0, 14)) &&
-				(smu->smc_fw_version >= 0x05550B00);
-
-	per_inst = apu_per_inst || smu_13_0_6_per_inst || smu_13_0_14_per_inst;
 
 	for_each_xcp(adev->xcp_mgr, xcp, i) {
 		amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
@@ -2516,7 +2645,8 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
 				gpu_metrics->xcp_stats[i].gfx_busy_acc[idx] =
 					SMUQ10_ROUND(metrics_x->GfxBusyAcc[inst]);
 
-				if (smu_v13_0_6_is_blw_host_limit_available(smu))
+				if (smu_v13_0_6_cap_supported(
+					    smu, SMU_CAP(HST_LIMIT_METRICS)))
 					gpu_metrics->xcp_stats[i].gfx_below_host_limit_acc[idx] =
 						SMUQ10_ROUND(metrics_x->GfxclkBelowHostLimitAcc
 								[inst]);
@@ -2624,7 +2754,7 @@ static int smu_v13_0_6_get_thermal_temperature_range(struct smu_context *smu,
 		return -EINVAL;
 
 	/*Check smu version, GetCtfLimit message only supported for smu version 85.69 or higher */
-	if (smu->smc_fw_version < 0x554500)
+	if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(CTF_LIMIT)))
 		return 0;
 
 	/* Get SOC Max operating temperature */
@@ -2726,11 +2856,10 @@ static int smu_v13_0_6_smu_send_hbm_bad_page_num(struct smu_context *smu,
 
 static int smu_v13_0_6_send_rma_reason(struct smu_context *smu)
 {
-	struct amdgpu_device *adev = smu->adev;
 	int ret;
 
 	/* NOTE: the message is only valid on dGPU with pmfw 85.90.0 and above */
-	if ((adev->flags & AMD_IS_APU) || smu->smc_fw_version < 0x00555a00)
+	if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(RMA_MSG)))
 		return 0;
 
 	ret = smu_cmn_send_smc_msg(smu, SMU_MSG_RmaDueToBadPageThreshold, NULL);
@@ -2744,31 +2873,13 @@ static int smu_v13_0_6_send_rma_reason(struct smu_context *smu)
 
 static int smu_v13_0_6_reset_sdma(struct smu_context *smu, uint32_t inst_mask)
 {
-	uint32_t smu_program;
 	int ret = 0;
 
-	smu_program = (smu->smc_fw_version >> 24) & 0xff;
+	if (!smu_v13_0_6_cap_supported(smu, SMU_CAP(SDMA_RESET)))
-	switch (amdgpu_ip_version(smu->adev, MP1_HWIP, 0)) {
+		return -EOPNOTSUPP;
-	case IP_VERSION(13, 0, 6):
-		if (((smu_program == 7) && (smu->smc_fw_version > 0x07550700)) ||
-			((smu_program == 0) && (smu->smc_fw_version > 0x00557700)))
-			ret = smu_cmn_send_smc_msg_with_param(smu,
-				SMU_MSG_ResetSDMA, inst_mask, NULL);
-		else if ((smu_program == 4) &&
-			(smu->smc_fw_version > 0x4556e6c))
-			ret = smu_cmn_send_smc_msg_with_param(smu,
-				      SMU_MSG_ResetSDMA2, inst_mask, NULL);
-		break;
-	case IP_VERSION(13, 0, 14):
-		if ((smu_program == 5) &&
-			(smu->smc_fw_version > 0x05550f00))
-			ret = smu_cmn_send_smc_msg_with_param(smu,
-				      SMU_MSG_ResetSDMA2, inst_mask, NULL);
-		break;
-	default:
-		break;
-	}
 
+	ret = smu_cmn_send_smc_msg_with_param(smu,
+						SMU_MSG_ResetSDMA, inst_mask, NULL);
 	if (ret)
 		dev_err(smu->adev->dev,
 			"failed to send ResetSDMA event with mask 0x%x\n",
@@ -3087,7 +3198,7 @@ static bool mca_smu_bank_is_valid(const struct mca_ras_info *mca_ras, struct amd
 	if (instlo != 0x03b30400)
 		return false;
 
-	if (!(adev->flags & AMD_IS_APU) && smu->smc_fw_version >= 0x00555600) {
+	if (smu_v13_0_6_cap_supported(smu, SMU_CAP(ACA_SYND))) {
 		errcode = MCA_REG__SYND__ERRORINFORMATION(entry->regs[MCA_REG_IDX_SYND]);
 		errcode &= 0xff;
 	} else {
@@ -3373,9 +3484,10 @@ static int aca_smu_get_valid_aca_bank(struct amdgpu_device *adev,
 
 static int aca_smu_parse_error_code(struct amdgpu_device *adev, struct aca_bank *bank)
 {
+	struct smu_context *smu = adev->powerplay.pp_handle;
 	int error_code;
 
-	if (!(adev->flags & AMD_IS_APU) && adev->pm.fw_version >= 0x00555600)
+	if (smu_v13_0_6_cap_supported(smu, SMU_CAP(ACA_SYND)))
 		error_code = ACA_REG__SYND__ERRORINFORMATION(bank->regs[ACA_REG_IDX_SYND]);
 	else
 		error_code = ACA_REG__STATUS__ERRORCODE(bank->regs[ACA_REG_IDX_STATUS]);
@@ -3413,7 +3525,7 @@ static const struct pptable_funcs smu_v13_0_6_ppt_funcs = {
 	.fini_power = smu_v13_0_fini_power,
 	.check_fw_status = smu_v13_0_6_check_fw_status,
 	/* pptable related */
-	.check_fw_version = smu_v13_0_check_fw_version,
+	.check_fw_version = smu_v13_0_6_check_fw_version,
 	.set_driver_table_location = smu_v13_0_set_driver_table_location,
 	.set_tool_table_location = smu_v13_0_set_tool_table_location,
 	.notify_memory_pool_location = smu_v13_0_notify_memory_pool_location,

drivers/gpu/drm/drm_client_modeset.c

@@ -743,6 +743,15 @@ retry:
 	if ((conn_configured & mask) != mask && conn_configured != conn_seq)
 		goto retry;
 
+	for (i = 0; i < count; i++) {
+		struct drm_connector *connector = connectors[i];
+
+		if (connector->has_tile)
+			drm_client_get_tile_offsets(dev, connectors, connector_count,
+						    modes, offsets, i,
+						    connector->tile_h_loc, connector->tile_v_loc);
+	}
+
 	/*
 	 * If the BIOS didn't enable everything it could, fall back to have the
 	 * same user experiencing of lighting up as much as possible like the

drivers/gpu/drm/drm_fb_helper.c

@@ -1354,14 +1354,14 @@ int drm_fb_helper_set_par(struct fb_info *info)
 }
 EXPORT_SYMBOL(drm_fb_helper_set_par);
 
-static void pan_set(struct drm_fb_helper *fb_helper, int x, int y)
+static void pan_set(struct drm_fb_helper *fb_helper, int dx, int dy)
 {
 	struct drm_mode_set *mode_set;
 
 	mutex_lock(&fb_helper->client.modeset_mutex);
 	drm_client_for_each_modeset(mode_set, &fb_helper->client) {
-		mode_set->x = x;
+		mode_set->x += dx;
-		mode_set->y = y;
+		mode_set->y += dy;
 	}
 	mutex_unlock(&fb_helper->client.modeset_mutex);
 }
@@ -1370,16 +1370,18 @@ static int pan_display_atomic(struct fb_var_screeninfo *var,
 			      struct fb_info *info)
 {
 	struct drm_fb_helper *fb_helper = info->par;
-	int ret;
+	int ret, dx, dy;
 
-	pan_set(fb_helper, var->xoffset, var->yoffset);
+	dx = var->xoffset - info->var.xoffset;
+	dy = var->yoffset - info->var.yoffset;
+	pan_set(fb_helper, dx, dy);
 
 	ret = drm_client_modeset_commit_locked(&fb_helper->client);
 	if (!ret) {
 		info->var.xoffset = var->xoffset;
 		info->var.yoffset = var->yoffset;
 	} else
-		pan_set(fb_helper, info->var.xoffset, info->var.yoffset);
+		pan_set(fb_helper, -dx, -dy);
 
 	return ret;
 }

drivers/gpu/drm/tiny/bochs.c

@@ -757,7 +757,6 @@ static void bochs_pci_remove(struct pci_dev *pdev)
 
 	drm_dev_unplug(dev);
 	drm_atomic_helper_shutdown(dev);
-	drm_dev_put(dev);
 }
 
 static void bochs_pci_shutdown(struct pci_dev *pdev)

drivers/gpu/drm/v3d/v3d_irq.c

@@ -107,8 +107,10 @@ v3d_irq(int irq, void *arg)
 
 		v3d_job_update_stats(&v3d->bin_job->base, V3D_BIN);
 		trace_v3d_bcl_irq(&v3d->drm, fence->seqno);
-		dma_fence_signal(&fence->base);
+
 		v3d->bin_job = NULL;
+		dma_fence_signal(&fence->base);
+
 		status = IRQ_HANDLED;
 	}
 
@@ -118,8 +120,10 @@ v3d_irq(int irq, void *arg)
 
 		v3d_job_update_stats(&v3d->render_job->base, V3D_RENDER);
 		trace_v3d_rcl_irq(&v3d->drm, fence->seqno);
-		dma_fence_signal(&fence->base);
+
 		v3d->render_job = NULL;
+		dma_fence_signal(&fence->base);
+
 		status = IRQ_HANDLED;
 	}
 
@@ -129,8 +133,10 @@ v3d_irq(int irq, void *arg)
 
 		v3d_job_update_stats(&v3d->csd_job->base, V3D_CSD);
 		trace_v3d_csd_irq(&v3d->drm, fence->seqno);
-		dma_fence_signal(&fence->base);
+
 		v3d->csd_job = NULL;
+		dma_fence_signal(&fence->base);
+
 		status = IRQ_HANDLED;
 	}
 
@@ -167,8 +173,10 @@ v3d_hub_irq(int irq, void *arg)
 
 		v3d_job_update_stats(&v3d->tfu_job->base, V3D_TFU);
 		trace_v3d_tfu_irq(&v3d->drm, fence->seqno);
-		dma_fence_signal(&fence->base);
+
 		v3d->tfu_job = NULL;
+		dma_fence_signal(&fence->base);
+
 		status = IRQ_HANDLED;
 	}
 

drivers/gpu/drm/virtio/virtgpu_prime.c

@@ -189,10 +189,11 @@ static void virtgpu_dma_buf_free_obj(struct drm_gem_object *obj)
 	struct virtio_gpu_object *bo = gem_to_virtio_gpu_obj(obj);
 	struct virtio_gpu_device *vgdev = obj->dev->dev_private;
 	struct dma_buf_attachment *attach = obj->import_attach;
-	struct dma_resv *resv = attach->dmabuf->resv;
 
 	if (attach) {
-		dma_resv_lock(resv, NULL);
+		struct dma_buf *dmabuf = attach->dmabuf;
+
+		dma_resv_lock(dmabuf->resv, NULL);
 
 		virtio_gpu_detach_object_fenced(bo);
 
@@ -200,10 +201,10 @@ static void virtgpu_dma_buf_free_obj(struct drm_gem_object *obj)
 			dma_buf_unmap_attachment(attach, bo->sgt,
 						 DMA_BIDIRECTIONAL);
 
-		dma_resv_unlock(resv);
+		dma_resv_unlock(dmabuf->resv);
 
-		dma_buf_detach(attach->dmabuf, attach);
+		dma_buf_detach(dmabuf, attach);
-		dma_buf_put(attach->dmabuf);
+		dma_buf_put(dmabuf);
 	}
 
 	if (bo->created) {

drivers/gpu/drm/xlnx/zynqmp_dp.c

@@ -1564,7 +1564,7 @@ static void zynqmp_dp_bridge_atomic_enable(struct drm_bridge *bridge,
 
 	pm_runtime_get_sync(dp->dev);
 
-	mutex_lock(&dp->lock);
+	guard(mutex)(&dp->lock);
 	zynqmp_dp_disp_enable(dp, old_bridge_state);
 
 	/*
@@ -1624,7 +1624,6 @@ static void zynqmp_dp_bridge_atomic_enable(struct drm_bridge *bridge,
 	zynqmp_dp_write(dp, ZYNQMP_DP_SOFTWARE_RESET,
 			ZYNQMP_DP_SOFTWARE_RESET_ALL);
 	zynqmp_dp_write(dp, ZYNQMP_DP_MAIN_STREAM_ENABLE, 1);
-	mutex_unlock(&dp->lock);
 }
 
 static void zynqmp_dp_bridge_atomic_disable(struct drm_bridge *bridge,

drivers/gpu/drm/xlnx/zynqmp_dpsub.h

@@ -60,6 +60,7 @@ struct zynqmp_dpsub_audio;
  * @layers: Video and graphics layers
  * @dp: The DisplayPort controller
  * @dma_align: DMA alignment constraint (must be a power of 2)
+ * @audio: DP audio data
  */
 struct zynqmp_dpsub {
 	struct device *dev;

init/Kconfig

@@ -1131,7 +1131,6 @@ config CGROUP_PIDS
 
 config CGROUP_RDMA
 	bool "RDMA controller"
-	select PAGE_COUNTER
 	help
 	  Provides enforcement of RDMA resources defined by IB stack.
 	  It is fairly easy for consumers to exhaust RDMA resources, which