path: root/common/recipes-kernel/linux/linux-yocto-4.19.8/2825-drm-amdkfd-enable-CONFIG_HSA_AMD-and-rebased-some-ch.patch

From a08e41a45d528e77adbed0f751a1ea3d9d5b8aac Mon Sep 17 00:00:00 2001
From: Chaudhary Amit Kumar <Chaudharyamit.Kumar@amd.com>
Date: Tue, 30 Jul 2019 18:16:41 +0530
Subject: [PATCH 2825/2940] drm/amdkfd: enable CONFIG_HSA_AMD and rebased some
 change in kfd

---
 drivers/gpu/drm/amd/amdgpu/Kconfig            |   1 +
 drivers/gpu/drm/amd/amdgpu/Makefile           |  11 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    |  23 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h    |   9 +-
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c  |   2 +
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c  | 121 ++++--------------
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c       |  85 ++++++------
 drivers/gpu/drm/amd/amdkfd/kfd_device.c       |   8 ++
 .../drm/amd/amdkfd/kfd_device_queue_manager.c |   7 +-
 drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c     |  26 +++-
 drivers/gpu/drm/amd/amdkfd/kfd_module.c       |   2 +
 drivers/gpu/drm/amd/amdkfd/kfd_process.c      |   2 +-
 kernel/fork.c                                 |   1 +
 security/device_cgroup.c                      |   1 +
 14 files changed, 146 insertions(+), 153 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig
index 3cdee30995a1..bdb7b74f3879 100644
--- a/drivers/gpu/drm/amd/amdgpu/Kconfig
+++ b/drivers/gpu/drm/amd/amdgpu/Kconfig
@@ -43,3 +43,4 @@ config DRM_AMDGPU_GART_DEBUGFS
 
 source "drivers/gpu/drm/amd/acp/Kconfig"
 source "drivers/gpu/drm/amd/display/Kconfig"
+source "drivers/gpu/drm/amd/amdkfd/Kconfig"
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index 8ea52c00fc8d..3c0a4675d36c 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/drivers/gpu/drm/amd/amdgpu/Makefile
@@ -35,7 +35,8 @@ ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \
 	-I$(FULL_AMD_DISPLAY_PATH) \
 	-I$(FULL_AMD_DISPLAY_PATH)/include \
 	-I$(FULL_AMD_DISPLAY_PATH)/dc \
-	-I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm
+	-I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm \
+	-I$(FULL_AMD_PATH)/amdkfd
 
 amdgpu-y := amdgpu_drv.o
 
@@ -150,7 +151,15 @@ amdgpu-y += \
 # add amdkfd interfaces
 amdgpu-y += amdgpu_amdkfd.o
 
+# linux-headers-4.15.0-20-generic/.config defined
+# CONFIG_HSA_AMD=m by default,
+# here we disable HSA just for rebase before enable kfd.
+# override CONFIG_HSA_AMD =
+
 ifneq ($(CONFIG_HSA_AMD),)
+AMDKFD_PATH := ../amdkfd
+include $(FULL_AMD_PATH)/amdkfd/Makefile
+amdgpu-y += $(AMDKFD_FILES)
 amdgpu-y += \
 	 amdgpu_amdkfd_fence.o \
 	 amdgpu_amdkfd_gpuvm.o \
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 1a699e47777d..ea1273734008 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -29,8 +29,10 @@
 #include <drm/drmP.h>
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
+#include "amdgpu_dma_buf.h"
 #include <linux/module.h>
 #include <linux/dma-buf.h>
+#include "amdgpu_xgmi.h"
 
 static unsigned int compute_vmid_bitmap = 0xFF00;
 
@@ -566,6 +568,27 @@ uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd)
 	return amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]);
 }
 
+uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
+
+	return adev->gmc.xgmi.hive_id;
+}
+uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src)
+{
+	struct amdgpu_device *peer_adev = (struct amdgpu_device *)src;
+	struct amdgpu_device *adev = (struct amdgpu_device *)dst;
+	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
+
+	if (ret < 0) {
+		DRM_ERROR("amdgpu: failed to get  xgmi hops count between node %d and %d. ret = %d\n",
+			adev->gmc.xgmi.physical_node_id,
+			peer_adev->gmc.xgmi.physical_node_id, ret);
+		ret = 0;
+	}
+	return  (uint8_t)ret;
+}
+
 uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)kgd;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 5331d9c8c9f6..bdc550dcc2ca 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -177,9 +177,10 @@ int amdgpu_amdkfd_get_dmabuf_info(struct kgd_dev *kgd, int dma_buf_fd,
 				  size_t buffer_size, uint32_t *metadata_size,
 				  uint32_t *flags);
 uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd);
-
+uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd);
 uint64_t amdgpu_amdkfd_get_mmio_remap_phys_addr(struct kgd_dev *kgd);
 uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd);
+uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct kgd_dev *dst, struct kgd_dev *src);
 
 #define read_user_wptr(mmptr, wptr, dst)				\
 	({								\
@@ -197,11 +198,11 @@ uint32_t amdgpu_amdkfd_get_num_gws(struct kgd_dev *kgd);
 	})
 
 /* GPUVM API */
-int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm,
-					void **process_info,
+int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, unsigned int pasid,
+					void **vm, void **process_info,
 					struct dma_fence **ef);
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
-					struct file *filp,
+					struct file *filp, unsigned int pasid,
 					void **vm, void **process_info,
 					struct dma_fence **ef);
 void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
index 3107b9575929..24d2a623d9f0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
@@ -80,6 +80,7 @@ struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
 	return fence;
 }
 
+#ifdef CONFIG_HSA_AMD
 struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 {
 	struct amdgpu_amdkfd_fence *fence;
@@ -93,6 +94,7 @@ struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 
 	return NULL;
 }
+#endif
 
 static const char *amdkfd_fence_get_driver_name(struct dma_fence *f)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 97bd545045f2..2e7a32fc0dbe 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -31,6 +31,7 @@
 #include "amdgpu_object.h"
 #include "amdgpu_vm.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_dma_buf.h"
 
 /* Special VM and GART address alignment needed for VI pre-Fiji due to
  * a HW bug.
@@ -204,39 +205,25 @@ void amdgpu_amdkfd_unreserve_memory_limit(struct amdgpu_bo *bo)
 	unreserve_mem_limit(adev, amdgpu_bo_size(bo), domain, sg);
 }
 
-
-/* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence(s) from BO's
+/* amdgpu_amdkfd_remove_eviction_fence - Removes eviction fence from BO's
  *  reservation object.
  *
  * @bo: [IN] Remove eviction fence(s) from this BO
- * @ef: [IN] If ef is specified, then this eviction fence is removed if it
+ * @ef: [IN] This eviction fence is removed if it
  *  is present in the shared list.
- * @ef_list: [OUT] Returns list of eviction fences. These fences are removed
- *  from BO's reservation object shared list.
- * @ef_count: [OUT] Number of fences in ef_list.
  *
- * NOTE: If called with ef_list, then amdgpu_amdkfd_add_eviction_fence must be
- *  called to restore the eviction fences and to avoid memory leak. This is
- *  useful for shared BOs.
  * NOTE: Must be called with BO reserved i.e. bo->tbo.resv->lock held.
  */
 static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
-					struct amdgpu_amdkfd_fence *ef,
-					struct amdgpu_amdkfd_fence ***ef_list,
-					unsigned int *ef_count)
+					struct amdgpu_amdkfd_fence *ef)
 {
 	struct reservation_object *resv = bo->tbo.resv;
 	struct reservation_object_list *old, *new;
 	unsigned int i, j, k;
 
-	if (!ef && !ef_list)
+	if (!ef)
 		return -EINVAL;
 
-	if (ef_list) {
-		*ef_list = NULL;
-		*ef_count = 0;
-	}
-
 	old = reservation_object_get_list(resv);
 	if (!old)
 		return 0;
@@ -255,8 +242,7 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
 		f = rcu_dereference_protected(old->shared[i],
 					      reservation_object_held(resv));
 
-		if ((ef && f->context == ef->base.context) ||
-		    (!ef && to_amdgpu_amdkfd_fence(f)))
+		if (f->context == ef->base.context)
 			RCU_INIT_POINTER(new->shared[--j], f);
 		else
 			RCU_INIT_POINTER(new->shared[k++], f);
@@ -264,21 +250,6 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
 	new->shared_max = old->shared_max;
 	new->shared_count = k;
 
-	if (!ef) {
-		unsigned int count = old->shared_count - j;
-
-		/* Alloc memory for count number of eviction fence pointers.
-		 * Fill the ef_list array and ef_count
-		 */
-		*ef_list = kcalloc(count, sizeof(**ef_list), GFP_KERNEL);
-		*ef_count = count;
-
-		if (!*ef_list) {
-			kfree(new);
-			return -ENOMEM;
-		}
-	}
-
 	/* Install the new fence list, seqcount provides the barriers */
 	preempt_disable();
 	write_seqcount_begin(&resv->seq);
@@ -292,10 +263,7 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
 
 		f = rcu_dereference_protected(new->shared[i],
 					      reservation_object_held(resv));
-		if (!ef)
-			(*ef_list)[k++] = to_amdgpu_amdkfd_fence(f);
-		else
-			dma_fence_put(f);
+		dma_fence_put(f);
 	}
 	kfree_rcu(old, rcu);
 
@@ -347,19 +315,8 @@ static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain,
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (ret)
 		goto validate_fail;
-	if (wait) {
-		struct amdgpu_amdkfd_fence **ef_list;
-		unsigned int ef_count;
-
-		ret = amdgpu_amdkfd_remove_eviction_fence(bo, NULL, &ef_list,
-							  &ef_count);
-		if (ret)
-			goto validate_fail;
-
-		ttm_bo_wait(&bo->tbo, false, false);
-		amdgpu_amdkfd_add_eviction_fence(bo, ef_list, ef_count);
-	}
-
+	if (wait) 
+		amdgpu_bo_sync_wait(bo, AMDGPU_FENCE_OWNER_KFD, false);
 validate_fail:
 	return ret;
 }
@@ -448,7 +405,6 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem,
 {
 	int ret;
 	struct kfd_bo_va_list *bo_va_entry;
-	struct amdgpu_bo *pd = vm->root.base.bo;
 	struct amdgpu_bo *bo = mem->bo;
 	uint64_t va = mem->va;
 	struct list_head *list_bo_va = &mem->bo_va_list;
@@ -487,28 +443,16 @@ static int add_bo_to_vm(struct amdgpu_device *adev, struct kgd_mem *mem,
 	if (p_bo_va_entry)
 		*p_bo_va_entry = bo_va_entry;
 
-	/* Allocate new page tables if needed and validate
-	 * them. Clearing of new page tables and validate need to wait
-	 * on move fences. We don't want that to trigger the eviction
-	 * fence, so remove it temporarily.
-	 */
-	amdgpu_amdkfd_remove_eviction_fence(pd,
-					vm->process_info->eviction_fence,
-					NULL, NULL);
-
+	/* Allocate validate page tables if needed */
 	ret = vm_validate_pt_pd_bos(vm);
 	if (ret) {
 		pr_err("validate_pt_pd_bos() failed\n");
 		goto err_alloc_pts;
 	}
 
-	/* Add the eviction fence back */
-	amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true);
-
 	return 0;
 
 err_alloc_pts:
-	amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true);
 	amdgpu_vm_bo_rmv(adev, bo_va_entry->bo_va);
 	list_del(&bo_va_entry->bo_list);
 err_vmadd:
@@ -818,24 +762,11 @@ static int unmap_bo_from_gpuvm(struct amdgpu_device *adev,
 {
 	struct amdgpu_bo_va *bo_va = entry->bo_va;
 	struct amdgpu_vm *vm = bo_va->base.vm;
-	struct amdgpu_bo *pd = vm->root.base.bo;
 
-	/* Remove eviction fence from PD (and thereby from PTs too as
-	 * they share the resv. object). Otherwise during PT update
-	 * job (see amdgpu_vm_bo_update_mapping), eviction fence would
-	 * get added to job->sync object and job execution would
-	 * trigger the eviction fence.
-	 */
-	amdgpu_amdkfd_remove_eviction_fence(pd,
-					    vm->process_info->eviction_fence,
-					    NULL, NULL);
 	amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
 
 	amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
 
-	/* Add the eviction fence back */
-	amdgpu_bo_fence(pd, &vm->process_info->eviction_fence->base, true);
-
 	amdgpu_sync_fence(NULL, sync, bo_va->last_pt_update, false);
 
 	return 0;
@@ -1050,9 +981,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	}
 	return ret;
 }
-
-int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm,
-					  void **process_info,
+int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, unsigned int pasid,
+					  void **vm, void **process_info,
 					  struct dma_fence **ef)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);
@@ -1064,7 +994,7 @@ int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm,
 		return -ENOMEM;
 
 	/* Initialize AMDGPU part of the VM */
-	ret = amdgpu_vm_init(adev, new_vm, AMDGPU_VM_CONTEXT_COMPUTE, 0);
+	ret = amdgpu_vm_init(adev, new_vm, AMDGPU_VM_CONTEXT_COMPUTE, pasid);
 	if (ret) {
 		pr_err("Failed init vm ret %d\n", ret);
 		goto amdgpu_vm_init_fail;
@@ -1087,7 +1017,7 @@ int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm,
 }
 
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
-					   struct file *filp,
+					   struct file *filp, unsigned int pasid,
 					   void **vm, void **process_info,
 					   struct dma_fence **ef)
 {
@@ -1116,6 +1046,7 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd,
 	return 0;
 }
 
+#ifdef CONFIG_HSA_AMD
 void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 				    struct amdgpu_vm *vm)
 {
@@ -1149,6 +1080,7 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 		kfree(process_info);
 	}
 }
+#endif
 
 void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm)
 {
@@ -1410,8 +1342,8 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	 * attached
 	 */
 	amdgpu_amdkfd_remove_eviction_fence(mem->bo,
-					process_info->eviction_fence,
-					NULL, NULL);
+					process_info->eviction_fence);
+
 	pr_debug("Release VA 0x%llx - 0x%llx\n", mem->va,
 		mem->va + bo_size * (1 + mem->aql_queue));
 
@@ -1638,8 +1570,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 	if (mem->mapped_to_gpu_memory == 0 &&
 	    !amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && !mem->bo->pin_count)
 		amdgpu_amdkfd_remove_eviction_fence(mem->bo,
-						process_info->eviction_fence,
-						    NULL, NULL);
+						process_info->eviction_fence);
 
 unreserve_out:
 	unreserve_bo_and_vms(&ctx, false, false);
@@ -1700,7 +1631,7 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd,
 	}
 
 	amdgpu_amdkfd_remove_eviction_fence(
-		bo, mem->process_info->eviction_fence, NULL, NULL);
+		bo, mem->process_info->eviction_fence);
 	list_del_init(&mem->validate_list.head);
 
 	if (size)
@@ -1965,6 +1896,7 @@ int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_dev *kgd, void *vm,
 	return 0;
 }
 
+#ifdef CONFIG_HSA_AMD
 /* Evict a userptr BO by stopping the queues if necessary
  *
  * Runs in MMU notifier, may be in RECLAIM_FS context. This means it
@@ -1996,6 +1928,7 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem,
 
 	return r;
 }
+#endif
 
 /* Update invalid userptr BOs
  *
@@ -2142,16 +2075,6 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 
 	amdgpu_sync_create(&sync);
 
-	/* Avoid triggering eviction fences when unmapping invalid
-	 * userptr BOs (waits for all fences, doesn't use
-	 * FENCE_OWNER_VM)
-	 */
-	list_for_each_entry(peer_vm, &process_info->vm_list_head,
-			    vm_list_node)
-		amdgpu_amdkfd_remove_eviction_fence(peer_vm->root.base.bo,
-						process_info->eviction_fence,
-						NULL, NULL);
-
 	ret = process_validate_vms(process_info);
 	if (ret)
 		goto unreserve_out;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index 604571935072..d571d482948f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -604,48 +604,6 @@ MODULE_PARM_DESC(mes,
        "Enable Micro Engine Scheduler (0 = disabled (default), 1 = enabled)");
 module_param_named(mes, amdgpu_mes, int, 0444);
 
-/**
- * DOC: pcie_p2p (bool)
- * Enable PCIe P2P (requires large-BAR). Default value: true (on)
- */
-bool pcie_p2p = true;
-module_param(pcie_p2p, bool, 0444);
-MODULE_PARM_DESC(pcie_p2p, "Enable PCIe P2P (requires large-BAR). (N = off, Y = on(default))");
-
-/**
- * DOC: dcfeaturemask (uint)
- * Override display features enabled. See enum DC_FEATURE_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
- * The default is the current set of stable display features.
- */
-MODULE_PARM_DESC(dcfeaturemask, "all stable DC features enabled (default))");
-module_param_named(dcfeaturemask, amdgpu_dc_feature_mask, uint, 0444);
-
-/*
- * DOC: hws_gws_support(bool)
- * Whether HWS support gws barriers. Default value: false (not supported)
- * This will be replaced with a MEC firmware version check once firmware
- * is ready
- */
-bool hws_gws_support;
-module_param(hws_gws_support, bool, 0444);
-MODULE_PARM_DESC(hws_gws_support, "MEC FW support gws barriers (false = not supported (Default), true = supported)");
-
-/**
- * DOC: abmlevel (uint)
- * Override the default ABM (Adaptive Backlight Management) level used for DC
- * enabled hardware. Requires DMCU to be supported and loaded.
- * Valid levels are 0-4. A value of 0 indicates that ABM should be disabled by
- * default. Values 1-4 control the maximum allowable brightness reduction via
- * the ABM algorithm, with 1 being the least reduction and 4 being the most
- * reduction.
- *
- * Defaults to 0, or disabled. Userspace can still override this level later
- * after boot.
- */
-uint amdgpu_dm_abm_level = 0;
-MODULE_PARM_DESC(abmlevel, "ABM level (0 = off (default), 1-4 = backlight reduction level) ");
-module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
-
 #ifdef CONFIG_HSA_AMD
 /**
  * DOC: sched_policy (int)
@@ -723,14 +681,14 @@ MODULE_PARM_DESC(ignore_crat,
 
 /**
  * DOC: noretry (int)
- * This parameter sets sh_mem_config.retry_disable. Default value, 0, enables retry.
- * Setting 1 disables retry.
+ * This parameter sets sh_mem_config.retry_disable. Default value, 1, disable retry.
+ * Setting 0 enables retry.
  * Retry is needed for recoverable page faults.
  */
-int noretry;
+int noretry = 1;
 module_param(noretry, int, 0644);
 MODULE_PARM_DESC(noretry,
-	"Set sh_mem_config.retry_disable on Vega10 (0 = retry enabled (default), 1 = retry disabled)");
+	"Set sh_mem_config.retry_disable on Vega10 (1 = retry disabled (default), 0 = retry enabled)");
 
 /**
  * DOC: halt_if_hws_hang (int)
@@ -776,6 +734,38 @@ module_param(queue_preemption_timeout_ms, int, 0644);
 MODULE_PARM_DESC(queue_preemption_timeout_ms, "queue preemption timeout in ms (1 = Minimum, 9000 = default)");
 #endif
 
+/**
+ * DOC: pcie_p2p (bool)
+ * Enable PCIe P2P (requires large-BAR). Default value: true (on)
+ */
+bool pcie_p2p = true;
+module_param(pcie_p2p, bool, 0444);
+MODULE_PARM_DESC(pcie_p2p, "Enable PCIe P2P (requires large-BAR). (N = off, Y = on(default))");
+
+/**
+ * DOC: dcfeaturemask (uint)
+ * Override display features enabled. See enum DC_FEATURE_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
+ * The default is the current set of stable display features.
+ */
+MODULE_PARM_DESC(dcfeaturemask, "all stable DC features enabled (default))");
+module_param_named(dcfeaturemask, amdgpu_dc_feature_mask, uint, 0444);
+
+/**
+ * DOC: abmlevel (uint)
+ * Override the default ABM (Adaptive Backlight Management) level used for DC
+ * enabled hardware. Requires DMCU to be supported and loaded.
+ * Valid levels are 0-4. A value of 0 indicates that ABM should be disabled by
+ * default. Values 1-4 control the maximum allowable brightness reduction via
+ * the ABM algorithm, with 1 being the least reduction and 4 being the most
+ * reduction.
+ *
+ * Defaults to 0, or disabled. Userspace can still override this level later
+ * after boot.
+ */
+uint amdgpu_dm_abm_level = 0;
+MODULE_PARM_DESC(abmlevel, "ABM level (0 = off (default), 1-4 = backlight reduction level) ");
+module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
+
 static const struct pci_device_id pciidlist[] = {
 #ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
@@ -1482,6 +1472,9 @@ static int __init amdgpu_init(void)
 #endif
 	kms_driver.num_ioctls = amdgpu_max_kms_ioctl;
 	amdgpu_register_atpx_handler();
+
+	/* Ignore KFD init failures. Normal when CONFIG_HSA_AMD is not set. */
+	amdgpu_amdkfd_init();
 	/* let modprobe override vga console setting */
 	return pci_register_driver(&amdgpu_kms_pci_driver);
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index bbab3c740dbe..da6826343180 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -483,6 +483,7 @@ static const struct kfd_device_info *lookup_device_info(unsigned short did)
 	return NULL;
 }
 
+#ifdef CONFIG_HSA_AMD
 struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 	struct pci_dev *pdev, const struct kfd2kgd_calls *f2g)
 {
@@ -528,6 +529,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 
 	return kfd;
 }
+#endif
 
 static void kfd_cwsr_init(struct kfd_dev *kfd)
 {
@@ -550,6 +552,7 @@ static void kfd_cwsr_init(struct kfd_dev *kfd)
 	}
 }
 
+#ifdef CONFIG_HSA_AMD
 bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 struct drm_device *ddev,
 			 const struct kgd2kfd_shared_resources *gpu_resources)
@@ -742,12 +745,14 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd)
 
 	return 0;
 }
+#endif
 
 bool kfd_is_locked(void)
 {
 	return  (atomic_read(&kfd_locked) > 0);
 }
 
+#ifdef CONFIG_HSA_AMD
 void kgd2kfd_suspend(struct kfd_dev *kfd)
 {
 	if (!kfd->init_complete)
@@ -780,6 +785,7 @@ int kgd2kfd_resume(struct kfd_dev *kfd)
 
 	return ret;
 }
+#endif
 
 static int kfd_resume(struct kfd_dev *kfd)
 {
@@ -808,6 +814,7 @@ static int kfd_resume(struct kfd_dev *kfd)
 	return err;
 }
 
+#ifdef CONFIG_HSA_AMD
 /* This is called directly from KGD at ISR. */
 void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 {
@@ -834,6 +841,7 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 
 	spin_unlock_irqrestore(&kfd->interrupt_lock, flags);
 }
+#endif
 
 int kgd2kfd_quiesce_mm(struct mm_struct *mm)
 {
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index a0344d7165b4..07caeda8eed4 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1222,7 +1222,9 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA ||
 		q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {
+		dqm_lock(dqm);
 		retval = allocate_sdma_queue(dqm, q);
+		dqm_unlock(dqm);
 		if (retval)
 			goto out;
 	}
@@ -1282,8 +1284,11 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 	deallocate_doorbell(qpd, q);
 out_deallocate_sdma_queue:
 	if (q->properties.type == KFD_QUEUE_TYPE_SDMA ||
-		q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI)
+		q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI) {
+		dqm_lock(dqm);
 		deallocate_sdma_queue(dqm, q);
+		dqm_unlock(dqm);
+	}
 out:
 	return retval;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index 278bc22edf49..9f5024bce095 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -126,10 +126,34 @@ void kfd_doorbell_fini(struct kfd_dev *kfd)
 		iounmap(kfd->doorbell_kernel_ptr);
 }
 
+static void kfd_doorbell_open(struct vm_area_struct *vma)
+{
+	/* Don't track the parent's PDD in a child process. We do set
+	 * VM_DONTCOPY, but that can be overridden from user mode.
+	 */
+	vma->vm_private_data = NULL;
+}
+
+static void kfd_doorbell_close(struct vm_area_struct *vma)
+{
+	struct kfd_process_device *pdd = vma->vm_private_data;
+
+	if (!pdd)
+		return;
+
+	mutex_lock(&pdd->qpd.doorbell_lock);
+	pdd->qpd.doorbell_vma = NULL;
+	/* Remember if the process was evicted without doorbells
+	 * mapped to user mode.
+	 */
+	if (pdd->qpd.doorbell_mapped == 0)
+		pdd->qpd.doorbell_mapped = -1;
+	mutex_unlock(&pdd->qpd.doorbell_lock);
+}
+
 static int kfd_doorbell_vm_fault(struct vm_fault *vmf)
 {
 	struct kfd_process_device *pdd = vmf->vma->vm_private_data;
-	pr_debug("Process %d doorbell vm page fault\n", process->pasid);
         if (!pdd)
                 return VM_FAULT_SIGBUS;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index 9c6dfb856797..d0a05024d38c 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
@@ -95,7 +95,9 @@ int kgd2kfd_init()
 	return kfd_init();
 }
 
+#ifdef CONFIG_HSA_AMD
 void kgd2kfd_exit(void)
 {
 	kfd_exit();
 }
+#endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index edaad275b869..bcdc1f915bc8 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -864,7 +864,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
 
 	if (drm_file)
 		ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(
-			dev->kgd, drm_file,
+			dev->kgd, drm_file, p->pasid,
 			&pdd->vm, &p->kgd_process_info, &p->ef);
 	else
 		ret = amdgpu_amdkfd_gpuvm_create_process_vm(dev->kgd, p->pasid,
diff --git a/kernel/fork.c b/kernel/fork.c
index f0b58479534f..a455cca4fc10 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1157,6 +1157,7 @@ struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
 
 	return mm;
 }
+EXPORT_SYMBOL_GPL(mm_access);
 
 static void complete_vfork_done(struct task_struct *tsk)
 {
diff --git a/security/device_cgroup.c b/security/device_cgroup.c
index cd97929fac66..e3a9ad5db5a0 100644
--- a/security/device_cgroup.c
+++ b/security/device_cgroup.c
@@ -824,3 +824,4 @@ int __devcgroup_check_permission(short type, u32 major, u32 minor,
 
 	return 0;
 }
+EXPORT_SYMBOL(__devcgroup_check_permission);
-- 
2.17.1