drm/amdkfd: Fix memory reporting on GFX 9.4.3

This patch fixes memory reporting on the GFX 9.4.3 APU and dGPU
by reporting available memory on a per partition basis. If its an
APU, available and used memory calculations take into account
system and TTM memory.

v2: squash in fix ("drm/amdkfd: Fix array out of bound warning")
    squash in fix ("drm/amdgpu: Update memory reporting for GFX9.4.3")

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

5 files changed, 84 insertions, 31 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 4bf6f5659568..948d362adabb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -35,6 +35,7 @@
 #include <drm/ttm/ttm_execbuf_util.h>
 #include "amdgpu_sync.h"
 #include "amdgpu_vm.h"
+#include "amdgpu_xcp.h"
 
 extern uint64_t amdgpu_amdkfd_total_mem_size;
 
@@ -98,8 +99,8 @@ struct amdgpu_amdkfd_fence {
 
 struct amdgpu_kfd_dev {
 	struct kfd_dev *dev;
-	int64_t vram_used;
-	uint64_t vram_used_aligned;
+	int64_t vram_used[MAX_XCP];
+	uint64_t vram_used_aligned[MAX_XCP];
 	bool init_complete;
 	struct work_struct reset_work;
 
@@ -287,7 +288,8 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
 					void *drm_priv);
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev);
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					uint8_t xcp_id);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		struct amdgpu_device *adev, uint64_t va, uint64_t size,
 		void *drm_priv, struct kgd_mem **mem,
@@ -327,9 +329,9 @@ void amdgpu_amdkfd_block_mmu_notifications(void *p);
 int amdgpu_amdkfd_criu_resume(void *p);
 bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev);
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
 
 #define KFD_XCP_MEM_ID(adev, xcp_id) \
 		((adev)->xcp_mgr && (xcp_id) >= 0 ?\
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 8724a0be31b8..cc37f04651e2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -157,12 +157,13 @@ void amdgpu_amdkfd_reserve_system_mem(uint64_t size)
  * Return: returns -ENOMEM in case of error, ZERO otherwise
  */
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
 	size_t system_mem_needed, ttm_mem_needed, vram_needed;
 	int ret = 0;
+	uint64_t vram_size = 0;
 
 	system_mem_needed = 0;
 	ttm_mem_needed = 0;
@@ -177,6 +178,17 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 		 * 2M BO chunk.
 		 */
 		vram_needed = size;
+		/*
+		 * For GFX 9.4.3, get the VRAM size from XCP structs
+		 */
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			return -EINVAL;
+
+		vram_size = KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+		if (adev->gmc.is_app_apu) {
+			system_mem_needed = size;
+			ttm_mem_needed = size;
+		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		system_mem_needed = size;
 	} else if (!(alloc_flag &
@@ -196,8 +208,8 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	     kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) ||
 	    (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
 	     kfd_mem_limit.max_ttm_mem_limit) ||
-	    (adev && adev->kfd.vram_used + vram_needed >
-	     adev->gmc.real_vram_size - reserved_for_pt)) {
+	    (adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] + vram_needed >
+	     vram_size - reserved_for_pt)) {
 		ret = -ENOMEM;
 		goto release;
 	}
@@ -207,9 +219,11 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	 */
 	WARN_ONCE(vram_needed && !adev,
 		  "adev reference can't be null when vram is used");
-	if (adev) {
-		adev->kfd.vram_used += vram_needed;
-		adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
+	if (adev && xcp_id >= 0) {
+		adev->kfd.vram_used[xcp_id] += vram_needed;
+		adev->kfd.vram_used_aligned[xcp_id] += adev->gmc.is_app_apu ?
+				vram_needed :
+				ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
 	}
 	kfd_mem_limit.system_mem_used += system_mem_needed;
 	kfd_mem_limit.ttm_mem_used += ttm_mem_needed;
@@ -220,7 +234,7 @@ release:
 }
 
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
@@ -230,9 +244,19 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		WARN_ONCE(!adev,
 			  "adev reference can't be null when alloc mem flags vram is set");
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			goto release;
+
 		if (adev) {
-			adev->kfd.vram_used -= size;
-			adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			adev->kfd.vram_used[xcp_id] -= size;
+			if (adev->gmc.is_app_apu) {
+				adev->kfd.vram_used_aligned[xcp_id] -= size;
+				kfd_mem_limit.system_mem_used -= size;
+				kfd_mem_limit.ttm_mem_used -= size;
+			} else {
+				adev->kfd.vram_used_aligned[xcp_id] -=
+					ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			}
 		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		kfd_mem_limit.system_mem_used -= size;
@@ -242,8 +266,8 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 		pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag);
 		goto release;
 	}
-	WARN_ONCE(adev && adev->kfd.vram_used < 0,
-		  "KFD VRAM memory accounting unbalanced");
+	WARN_ONCE(adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] < 0,
+		  "KFD VRAM memory accounting unbalanced for xcp: %d", xcp_id);
 	WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0,
 		  "KFD TTM memory accounting unbalanced");
 	WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
@@ -259,7 +283,8 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
 	u32 alloc_flags = bo->kfd_bo->alloc_flags;
 	u64 size = amdgpu_bo_size(bo);
 
-	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags,
+					  bo->xcp_id);
 
 	kfree(bo->kfd_bo);
 }
@@ -1609,23 +1634,42 @@ out_unlock:
 	return ret;
 }
 
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev)
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					  uint8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
 	ssize_t available;
+	uint64_t vram_available, system_mem_available, ttm_mem_available;
 
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
-	available = adev->gmc.real_vram_size
-		- adev->kfd.vram_used_aligned
+	vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+		- adev->kfd.vram_used_aligned[xcp_id]
 		- atomic64_read(&adev->vram_pin_size)
 		- reserved_for_pt;
+
+	if (adev->gmc.is_app_apu) {
+		system_mem_available = no_system_mem_limit ?
+					kfd_mem_limit.max_system_mem_limit :
+					kfd_mem_limit.max_system_mem_limit -
+					kfd_mem_limit.system_mem_used;
+
+		ttm_mem_available = kfd_mem_limit.max_ttm_mem_limit -
+				kfd_mem_limit.ttm_mem_used;
+
+		available = min3(system_mem_available, ttm_mem_available,
+				 vram_available);
+		available = ALIGN_DOWN(available, PAGE_SIZE);
+	} else {
+		available = ALIGN_DOWN(vram_available, VRAM_AVAILABLITY_ALIGN);
+	}
+
 	spin_unlock(&kfd_mem_limit.mem_limit_lock);
 
 	if (available < 0)
 		available = 0;
 
-	return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN);
+	return available;
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
@@ -1713,7 +1757,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 	amdgpu_sync_create(&(*mem)->sync);
 
-	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags);
+	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags,
+					      xcp_id);
 	if (ret) {
 		pr_debug("Insufficient memory\n");
 		goto err_reserve_limit;
@@ -1781,7 +1826,7 @@ err_node_allow:
 	/* Don't unreserve system mem limit twice */
 	goto err_reserve_limit;
 err_bo_create:
-	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
 err_reserve_limit:
 	mutex_destroy(&(*mem)->lock);
 	if (gobj)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
index 68b63b970ce8..9c5912b9d8bd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
@@ -24,8 +24,11 @@
 #ifndef AMDGPU_XCP_H
 #define AMDGPU_XCP_H
 
+#include <linux/pci.h>
 #include <linux/xarray.h>
 
+#include "amdgpu_ctx.h"
+
 #define MAX_XCP 8
 
 #define AMDGPU_XCP_MODE_NONE -1
@@ -34,6 +37,8 @@
 #define AMDGPU_XCP_FL_NONE 0
 #define AMDGPU_XCP_FL_LOCKED (1 << 0)
 
+struct amdgpu_fpriv;
+
 enum AMDGPU_XCP_IP_BLOCK {
 	AMDGPU_XCP_GFXHUB,
 	AMDGPU_XCP_GFX,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 1ae867482bc7..a9efff94390b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1044,7 +1044,8 @@ static int kfd_ioctl_get_available_memory(struct file *filep,
 
 	if (!pdd)
 		return -EINVAL;
-	args->available = amdgpu_amdkfd_get_available_memory(pdd->dev->adev);
+	args->available = amdgpu_amdkfd_get_available_memory(pdd->dev->adev,
+							pdd->dev->node_id);
 	kfd_unlock_pdd(pdd);
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 45959892bc0f..c1ab70faf36e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -280,7 +280,7 @@ static void svm_range_free(struct svm_range *prange, bool update_mem_usage)
 	if (update_mem_usage && !p->xnack_enabled) {
 		pr_debug("unreserve prange 0x%p size: 0x%llx\n", prange, size);
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 	}
 	mutex_destroy(&prange->lock);
 	mutex_destroy(&prange->migrate_mutex);
@@ -313,7 +313,7 @@ svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
 	p = container_of(svms, struct kfd_process, svms);
 	if (!p->xnack_enabled && update_mem_usage &&
 	    amdgpu_amdkfd_reserve_mem_limit(NULL, size << PAGE_SHIFT,
-					    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR)) {
+				    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0)) {
 		pr_info("SVM mapping failed, exceeds resident system memory limit\n");
 		kfree(prange);
 		return NULL;
@@ -3037,10 +3037,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 			size = (pchild->last - pchild->start + 1) << PAGE_SHIFT;
 			if (xnack_enabled) {
 				amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			} else {
 				r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 				if (r)
 					goto out_unlock;
 				reserved_size += size;
@@ -3050,10 +3050,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 		size = (prange->last - prange->start + 1) << PAGE_SHIFT;
 		if (xnack_enabled) {
 			amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 		} else {
 			r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			if (r)
 				goto out_unlock;
 			reserved_size += size;
@@ -3066,7 +3066,7 @@ out_unlock:
 
 	if (r)
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, reserved_size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 	else
 		/* Change xnack mode must be inside svms lock, to avoid race with
 		 * svm_range_deferred_list_work unreserve memory in parallel.