From 82c9ff3fa8819e03e30677943abc355d544293f1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
Date: Thu, 17 Mar 2022 14:08:36 +0100
Subject: drm/amdkfd: use tlb_seq from the VM subsystem for SVM as well v2
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Git-commit: 4d30a83c740e9904c7f54f071ec121f9e6932f63
Patch-mainline: v5.19-rc1
References: jsc#PED-1166 jsc#PED-1168 jsc#PED-1170 jsc#PED-1218 jsc#PED-1220 jsc#PED-1222 jsc#PED-1223 jsc#PED-1225
Instead of hand rolling the table_freed parameter.
v2: add some changes suggested by Philip
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Philip Yang<Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Patrik Jakobsson <pjakobsson@suse.de>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 5 ++---
.../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 18 ++++++++----------
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 19 ++++++++-----------
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 2 +-
drivers/gpu/drm/amd/amdkfd/kfd_svm.c | 7 +++----
5 files changed, 22 insertions(+), 29 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 0838926a8ef0..f8b9f27adcf5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -273,9 +273,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
uint64_t *size);
-int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
- struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv,
- bool *table_freed);
+int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev,
+ struct kgd_mem *mem, void *drm_priv);
int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv);
int amdgpu_amdkfd_gpuvm_sync_memory(
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 3dc5ab2764ff..1cccf2e7e0ed 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -1058,8 +1058,7 @@ static void unmap_bo_from_gpuvm(struct kgd_mem *mem,
static int update_gpuvm_pte(struct kgd_mem *mem,
struct kfd_mem_attachment *entry,
- struct amdgpu_sync *sync,
- bool *table_freed)
+ struct amdgpu_sync *sync)
{
struct amdgpu_bo_va *bo_va = entry->bo_va;
struct amdgpu_device *adev = entry->adev;
@@ -1070,7 +1069,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem,
return ret;
/* Update the page tables */
- ret = amdgpu_vm_bo_update(adev, bo_va, false, table_freed);
+ ret = amdgpu_vm_bo_update(adev, bo_va, false, NULL);
if (ret) {
pr_err("amdgpu_vm_bo_update failed\n");
return ret;
@@ -1082,8 +1081,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem,
static int map_bo_to_gpuvm(struct kgd_mem *mem,
struct kfd_mem_attachment *entry,
struct amdgpu_sync *sync,
- bool no_update_pte,
- bool *table_freed)
+ bool no_update_pte)
{
int ret;
@@ -1100,7 +1098,7 @@ static int map_bo_to_gpuvm(struct kgd_mem *mem,
if (no_update_pte)
return 0;
- ret = update_gpuvm_pte(mem, entry, sync, table_freed);
+ ret = update_gpuvm_pte(mem, entry, sync);
if (ret) {
pr_err("update_gpuvm_pte() failed\n");
goto update_gpuvm_pte_failed;
@@ -1710,7 +1708,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
struct amdgpu_device *adev, struct kgd_mem *mem,
- void *drm_priv, bool *table_freed)
+ void *drm_priv)
{
struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
int ret;
@@ -1797,7 +1795,7 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
entry->va, entry->va + bo_size, entry);
ret = map_bo_to_gpuvm(mem, entry, ctx.sync,
- is_invalid_userptr, table_freed);
+ is_invalid_userptr);
if (ret) {
pr_err("Failed to map bo to gpuvm\n");
goto out_unreserve;
@@ -2265,7 +2263,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
continue;
kfd_mem_dmaunmap_attachment(mem, attachment);
- ret = update_gpuvm_pte(mem, attachment, &sync, NULL);
+ ret = update_gpuvm_pte(mem, attachment, &sync);
if (ret) {
pr_err("%s: update PTE failed\n", __func__);
/* make sure this gets validated again */
@@ -2476,7 +2474,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
continue;
kfd_mem_dmaunmap_attachment(mem, attachment);
- ret = update_gpuvm_pte(mem, attachment, &sync_obj, NULL);
+ ret = update_gpuvm_pte(mem, attachment, &sync_obj);
if (ret) {
pr_debug("Memory eviction: update PTE failed. Try again\n");
goto validate_map_fail;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 607f65ab39ac..60438193c0c1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1146,7 +1146,6 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
long err = 0;
int i;
uint32_t *devices_arr = NULL;
- bool table_freed = false;
if (!args->n_devices) {
pr_debug("Device IDs array empty\n");
@@ -1208,7 +1207,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
peer_pdd->dev->adev, (struct kgd_mem *)mem,
- peer_pdd->drm_priv, &table_freed);
+ peer_pdd->drm_priv);
if (err) {
struct pci_dev *pdev = peer_pdd->dev->adev->pdev;
@@ -1233,13 +1232,11 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
}
/* Flush TLBs after waiting for the page table updates to complete */
- if (table_freed || !kfd_flush_tlb_after_unmap(dev)) {
- for (i = 0; i < args->n_devices; i++) {
- peer_pdd = kfd_process_device_data_by_id(p, devices_arr[i]);
- if (WARN_ON_ONCE(!peer_pdd))
- continue;
- kfd_flush_tlb(peer_pdd, TLB_FLUSH_LEGACY);
- }
+ for (i = 0; i < args->n_devices; i++) {
+ peer_pdd = kfd_process_device_data_by_id(p, devices_arr[i]);
+ if (WARN_ON_ONCE(!peer_pdd))
+ continue;
+ kfd_flush_tlb(peer_pdd, TLB_FLUSH_LEGACY);
}
kfree(devices_arr);
@@ -2206,8 +2203,8 @@ static int criu_restore_bo(struct kfd_process *p,
if (IS_ERR(peer_pdd))
return PTR_ERR(peer_pdd);
- ret = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(peer->adev, kgd_mem, peer_pdd->drm_priv,
- NULL);
+ ret = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(peer->adev, kgd_mem,
+ peer_pdd->drm_priv);
if (ret) {
pr_err("Failed to map to gpu %d/%d\n", j, p->n_pdds);
return ret;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 4a8a047b7593..ac8123c1ee8f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -722,7 +722,7 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
goto err_alloc_mem;
err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->adev, *mem,
- pdd->drm_priv, NULL);
+ pdd->drm_priv);
if (err)
goto err_map_mem;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index b3fc3e958227..3929377ce8ca 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -1243,7 +1243,6 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
{
struct amdgpu_device *adev = pdd->dev->adev;
struct amdgpu_vm *vm = drm_priv_to_vm(pdd->drm_priv);
- bool table_freed = false;
uint64_t pte_flags;
unsigned long last_start;
int last_domain;
@@ -1284,7 +1283,7 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
last_start - prange->start,
NULL, dma_addr,
&vm->last_update,
- &table_freed);
+ NULL);
for (j = last_start - prange->start; j <= i; j++)
dma_addr[j] |= last_domain;
@@ -1306,8 +1305,6 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
if (fence)
*fence = dma_fence_get(vm->last_update);
- if (table_freed)
- kfd_flush_tlb(pdd, TLB_FLUSH_LEGACY);
out:
return r;
}
@@ -1363,6 +1360,8 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
break;
}
}
+
+ kfd_flush_tlb(pdd, TLB_FLUSH_LEGACY);
}
return r;
--
2.38.1