Merge tag 'amd-drm-next-6.2-2022-11-18' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.2-2022-11-18:

amdgpu:
- SR-IOV fixes
- Clean up DC checks
- DCN 3.2.x fixes
- DCN 3.1.x fixes
- Don't enable degamma on asics which don't support it
- IP discovery fixes
- BACO fixes
- Fix vbios allocation handling when vkms is enabled
- Drop buggy tdr advanced mode GPU reset handling
- Fix the build when DCN is not set in kconfig
- MST DSC fixes
- Userptr fixes
- FRU and RAS EEPROM fixes
- VCN 4.x RAS support
- Aldrebaran CU occupancy reporting fix
- PSP ring cleanup

amdkfd:
- Memory limit fix
- Enable cooperative launch on gfx 10.3

amd-drm-next-6.2-2022-11-11:

amdgpu:
- SMU 13.x updates
- GPUVM TLB race fix
- DCN 3.1.4 updates
- DCN 3.2.x updates
- PSR fixes
- Kerneldoc fix
- Vega10 fan fix
- GPUVM locking fixes in error pathes
- BACO fix for Beige Goby
- EEPROM I2C address cleanup
- GFXOFF fix
- Fix DC memory leak in error pathes
- Flexible array updates
- Mtype fix for GPUVM PTEs
- Move Kconfig into amdgpu directory
- SR-IOV updates
- Fix possible memory leak in CS IOCTL error path

amdkfd:
- Fix possible memory overrun
- CRIU fixes

radeon:
- ACPI ref count fix
- HDA audio notifier support
- Move Kconfig into radeon directory

UAPI:
- Add new GEM_CREATE flags to help to transition more KFD functionality to the DRM UAPI.
  These are used internally in the driver to align location based memory coherency
  requirements from memory allocated in the KFD with how we manage GPUVM PTEs.  They
  are currently blocked in the GEM_CREATE IOCTL as we don't have a user right now.
  They are just used internally in the kernel driver for now for existing KFD memory
  allocations. So a change to the UAPI header, but no functional change in the UAPI.

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221118170807.6505-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>

1 files changed, 71 insertions, 42 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 0ba87ca6f318..7b5074e776f4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -58,6 +58,7 @@
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_sdma.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_hmm.h"
 #include "amdgpu_atomfirmware.h"
 #include "amdgpu_res_cursor.h"
 #include "bif/bif_4_1_d.h"
@@ -634,9 +635,6 @@ struct amdgpu_ttm_tt {
 	struct task_struct	*usertask;
 	uint32_t		userflags;
 	bool			bound;
-#if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
-	struct hmm_range	*range;
-#endif
 };
 
 #define ttm_to_amdgpu_ttm_tt(ptr)	container_of(ptr, struct amdgpu_ttm_tt, ttm)
@@ -649,7 +647,8 @@ struct amdgpu_ttm_tt {
  * Calling function must call amdgpu_ttm_tt_userptr_range_done() once and only
  * once afterwards to stop HMM tracking
  */
-int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages)
+int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
+				 struct hmm_range **range)
 {
 	struct ttm_tt *ttm = bo->tbo.ttm;
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
@@ -659,16 +658,15 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages)
 	bool readonly;
 	int r = 0;
 
+	/* Make sure get_user_pages_done() can cleanup gracefully */
+	*range = NULL;
+
 	mm = bo->notifier.mm;
 	if (unlikely(!mm)) {
 		DRM_DEBUG_DRIVER("BO is not registered?\n");
 		return -EFAULT;
 	}
 
-	/* Another get_user_pages is running at the same time?? */
-	if (WARN_ON(gtt->range))
-		return -EFAULT;
-
 	if (!mmget_not_zero(mm)) /* Happens during process shutdown */
 		return -ESRCH;
 
@@ -685,9 +683,8 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages)
 	}
 
 	readonly = amdgpu_ttm_tt_is_readonly(ttm);
-	r = amdgpu_hmm_range_get_pages(&bo->notifier, mm, pages, start,
-				       ttm->num_pages, &gtt->range, readonly,
-				       true, NULL);
+	r = amdgpu_hmm_range_get_pages(&bo->notifier, start, ttm->num_pages,
+				       readonly, NULL, pages, range);
 out_unlock:
 	mmap_read_unlock(mm);
 	if (r)
@@ -704,30 +701,24 @@ out_unlock:
  *
  * Returns: true if pages are still valid
  */
-bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm)
+bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
+				       struct hmm_range *range)
 {
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
-	bool r = false;
 
-	if (!gtt || !gtt->userptr)
+	if (!gtt || !gtt->userptr || !range)
 		return false;
 
 	DRM_DEBUG_DRIVER("user_pages_done 0x%llx pages 0x%x\n",
 		gtt->userptr, ttm->num_pages);
 
-	WARN_ONCE(!gtt->range || !gtt->range->hmm_pfns,
-		"No user pages to check\n");
-
-	if (gtt->range) {
-		/*
-		 * FIXME: Must always hold notifier_lock for this, and must
-		 * not ignore the return code.
-		 */
-		r = amdgpu_hmm_range_get_pages_done(gtt->range);
-		gtt->range = NULL;
-	}
+	WARN_ONCE(!range->hmm_pfns, "No user pages to check\n");
 
-	return !r;
+	/*
+	 * FIXME: Must always hold notifier_lock for this, and must
+	 * not ignore the return code.
+	 */
+	return !amdgpu_hmm_range_get_pages_done(range);
 }
 #endif
 
@@ -804,20 +795,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_device *bdev,
 	/* unmap the pages mapped to the device */
 	dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0);
 	sg_free_table(ttm->sg);
-
-#if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
-	if (gtt->range) {
-		unsigned long i;
-
-		for (i = 0; i < ttm->num_pages; i++) {
-			if (ttm->pages[i] !=
-			    hmm_pfn_to_page(gtt->range->hmm_pfns[i]))
-				break;
-		}
-
-		WARN((i == ttm->num_pages), "Missing get_user_page_done\n");
-	}
-#endif
 }
 
 static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
@@ -1168,8 +1145,9 @@ int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo,
  * @addr:  The address in the current tasks VM space to use
  * @flags: Requirements of userptr object.
  *
- * Called by amdgpu_gem_userptr_ioctl() to bind userptr pages
- * to current task
+ * Called by amdgpu_gem_userptr_ioctl() and kfd_ioctl_alloc_memory_of_gpu() to
+ * bind userptr pages to current task and by kfd_ioctl_acquire_vm() to
+ * initialize GPU VM for a KFD process.
  */
 int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo,
 			      uint64_t addr, uint32_t flags)
@@ -1553,6 +1531,23 @@ static void amdgpu_ttm_fw_reserve_vram_fini(struct amdgpu_device *adev)
 		NULL, &adev->mman.fw_vram_usage_va);
 }
 
+/*
+ * Driver Reservation functions
+ */
+/**
+ * amdgpu_ttm_drv_reserve_vram_fini - free drv reserved vram
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * free drv reserved vram if it has been reserved.
+ */
+static void amdgpu_ttm_drv_reserve_vram_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->mman.drv_vram_usage_reserved_bo,
+						  NULL,
+						  NULL);
+}
+
 /**
  * amdgpu_ttm_fw_reserve_vram_init - create bo vram reservation from fw
  *
@@ -1579,6 +1574,31 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev)
 					  &adev->mman.fw_vram_usage_va);
 }
 
+/**
+ * amdgpu_ttm_drv_reserve_vram_init - create bo vram reservation from driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * create bo vram reservation from drv.
+ */
+static int amdgpu_ttm_drv_reserve_vram_init(struct amdgpu_device *adev)
+{
+	uint64_t vram_size = adev->gmc.visible_vram_size;
+
+	adev->mman.drv_vram_usage_reserved_bo = NULL;
+
+	if (adev->mman.drv_vram_usage_size == 0 ||
+	    adev->mman.drv_vram_usage_size > vram_size)
+		return 0;
+
+	return amdgpu_bo_create_kernel_at(adev,
+					  adev->mman.drv_vram_usage_start_offset,
+					  adev->mman.drv_vram_usage_size,
+					  AMDGPU_GEM_DOMAIN_VRAM,
+					  &adev->mman.drv_vram_usage_reserved_bo,
+					  NULL);
+}
+
 /*
  * Memoy training reservation functions
  */
@@ -1747,6 +1767,14 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	}
 
 	/*
+	 *The reserved vram for driver must be pinned to the specified
+	 *place on the VRAM, so reserve it early.
+	 */
+	r = amdgpu_ttm_drv_reserve_vram_init(adev);
+	if (r)
+		return r;
+
+	/*
 	 * only NAVI10 and onwards ASIC support for IP discovery.
 	 * If IP discovery enabled, a block of memory should be
 	 * reserved for IP discovey.
@@ -1871,6 +1899,7 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
 	amdgpu_bo_free_kernel(&adev->mman.sdma_access_bo, NULL,
 					&adev->mman.sdma_access_ptr);
 	amdgpu_ttm_fw_reserve_vram_fini(adev);
+	amdgpu_ttm_drv_reserve_vram_fini(adev);
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {