aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdkfd/kfd_process.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_process.c')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_process.c66
1 files changed, 31 insertions, 35 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index b993011cfa64..f1930ff2c74a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -251,14 +251,13 @@ cleanup:
}
/**
- * @kfd_get_cu_occupancy - Collect number of waves in-flight on this device
+ * kfd_get_cu_occupancy - Collect number of waves in-flight on this device
* by current process. Translates acquired wave count into number of compute units
* that are occupied.
*
- * @atr: Handle of attribute that allows reporting of wave count. The attribute
+ * @attr: Handle of attribute that allows reporting of wave count. The attribute
* handle encapsulates GPU device it is associated with, thereby allowing collection
* of waves in flight, etc
- *
* @buffer: Handle of user provided buffer updated with wave count
*
* Return: Number of bytes written to user buffer or an error value
@@ -288,7 +287,7 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
/* Collect wave count from device if it supports */
wave_cnt = 0;
max_waves_per_cu = 0;
- dev->kfd2kgd->get_cu_occupancy(dev->kgd, proc->pasid, &wave_cnt,
+ dev->kfd2kgd->get_cu_occupancy(dev->adev, proc->pasid, &wave_cnt,
&max_waves_per_cu);
/* Translate wave count to number of compute units */
@@ -692,12 +691,12 @@ static void kfd_process_free_gpuvm(struct kgd_mem *mem,
struct kfd_dev *dev = pdd->dev;
if (kptr) {
- amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(dev->kgd, mem);
+ amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(dev->adev, mem);
kptr = NULL;
}
- amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->kgd, mem, pdd->drm_priv);
- amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->kgd, mem, pdd->drm_priv,
+ amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(dev->adev, mem, pdd->drm_priv);
+ amdgpu_amdkfd_gpuvm_free_memory_of_gpu(dev->adev, mem, pdd->drm_priv,
NULL);
}
@@ -714,24 +713,24 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
struct kfd_dev *kdev = pdd->dev;
int err;
- err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->kgd, gpu_va, size,
+ err = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(kdev->adev, gpu_va, size,
pdd->drm_priv, mem, NULL, flags);
if (err)
goto err_alloc_mem;
- err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, *mem,
+ err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->adev, *mem,
pdd->drm_priv, NULL);
if (err)
goto err_map_mem;
- err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->kgd, *mem, true);
+ err = amdgpu_amdkfd_gpuvm_sync_memory(kdev->adev, *mem, true);
if (err) {
pr_debug("Sync memory failed, wait interrupted by user signal\n");
goto sync_memory_failed;
}
if (kptr) {
- err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->kgd,
+ err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kdev->adev,
(struct kgd_mem *)*mem, kptr, NULL);
if (err) {
pr_debug("Map GTT BO to kernel failed\n");
@@ -742,10 +741,10 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
return err;
sync_memory_failed:
- amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(kdev->kgd, *mem, pdd->drm_priv);
+ amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(kdev->adev, *mem, pdd->drm_priv);
err_map_mem:
- amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->kgd, *mem, pdd->drm_priv,
+ amdgpu_amdkfd_gpuvm_free_memory_of_gpu(kdev->adev, *mem, pdd->drm_priv,
NULL);
err_alloc_mem:
*mem = NULL;
@@ -940,10 +939,10 @@ static void kfd_process_device_free_bos(struct kfd_process_device *pdd)
if (!peer_pdd->drm_priv)
continue;
amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
- peer_pdd->dev->kgd, mem, peer_pdd->drm_priv);
+ peer_pdd->dev->adev, mem, peer_pdd->drm_priv);
}
- amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->kgd, mem,
+ amdgpu_amdkfd_gpuvm_free_memory_of_gpu(pdd->dev->adev, mem,
pdd->drm_priv, NULL);
kfd_process_device_remove_obj_handle(pdd, id);
}
@@ -974,7 +973,7 @@ static void kfd_process_kunmap_signal_bo(struct kfd_process *p)
if (!mem)
goto out;
- amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kdev->kgd, mem);
+ amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(kdev->adev, mem);
out:
mutex_unlock(&p->mutex);
@@ -1003,7 +1002,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
if (pdd->drm_file) {
amdgpu_amdkfd_gpuvm_release_process_vm(
- pdd->dev->kgd, pdd->drm_priv);
+ pdd->dev->adev, pdd->drm_priv);
fput(pdd->drm_file);
}
@@ -1011,7 +1010,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
free_pages((unsigned long)pdd->qpd.cwsr_kaddr,
get_order(KFD_CWSR_TBA_TMA_SIZE));
- kfree(pdd->qpd.doorbell_bitmap);
+ bitmap_free(pdd->qpd.doorbell_bitmap);
idr_destroy(&pdd->alloc_idr);
kfd_free_process_doorbells(pdd->dev, pdd->doorbell_index);
@@ -1317,14 +1316,13 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported)
* support the SVM APIs and don't need to be considered
* for the XNACK mode selection.
*/
- if (dev->device_info->asic_family < CHIP_VEGA10)
+ if (!KFD_IS_SOC15(dev))
continue;
/* Aldebaran can always support XNACK because it can support
* per-process XNACK mode selection. But let the dev->noretry
* setting still influence the default XNACK mode.
*/
- if (supported &&
- dev->device_info->asic_family == CHIP_ALDEBARAN)
+ if (supported && KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 2))
continue;
/* GFXv10 and later GPUs do not support shader preemption
@@ -1332,7 +1330,7 @@ bool kfd_process_xnack_mode(struct kfd_process *p, bool supported)
* management and memory-manager-related preemptions or
* even deadlocks.
*/
- if (dev->device_info->asic_family >= CHIP_NAVI10)
+ if (KFD_GC_VERSION(dev) >= IP_VERSION(10, 1, 1))
return false;
if (dev->noretry)
@@ -1431,12 +1429,11 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
int range_start = dev->shared_resources.non_cp_doorbells_start;
int range_end = dev->shared_resources.non_cp_doorbells_end;
- if (!KFD_IS_SOC15(dev->device_info->asic_family))
+ if (!KFD_IS_SOC15(dev))
return 0;
- qpd->doorbell_bitmap =
- kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
- BITS_PER_BYTE), GFP_KERNEL);
+ qpd->doorbell_bitmap = bitmap_zalloc(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
+ GFP_KERNEL);
if (!qpd->doorbell_bitmap)
return -ENOMEM;
@@ -1448,9 +1445,9 @@ static int init_doorbell_bitmap(struct qcm_process_device *qpd,
for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS / 2; i++) {
if (i >= range_start && i <= range_end) {
- set_bit(i, qpd->doorbell_bitmap);
- set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
- qpd->doorbell_bitmap);
+ __set_bit(i, qpd->doorbell_bitmap);
+ __set_bit(i + KFD_QUEUE_DOORBELL_MIRROR_OFFSET,
+ qpd->doorbell_bitmap);
}
}
@@ -1547,7 +1544,7 @@ int kfd_process_device_init_vm(struct kfd_process_device *pdd,
dev = pdd->dev;
ret = amdgpu_amdkfd_gpuvm_acquire_process_vm(
- dev->kgd, drm_file, p->pasid,
+ dev->adev, drm_file, p->pasid,
&p->kgd_process_info, &p->ef);
if (ret) {
pr_err("Failed to create process VM object\n");
@@ -1779,14 +1776,13 @@ int kfd_process_gpuidx_from_gpuid(struct kfd_process *p, uint32_t gpu_id)
}
int
-kfd_process_gpuid_from_kgd(struct kfd_process *p, struct amdgpu_device *adev,
+kfd_process_gpuid_from_adev(struct kfd_process *p, struct amdgpu_device *adev,
uint32_t *gpuid, uint32_t *gpuidx)
{
- struct kgd_dev *kgd = (struct kgd_dev *)adev;
int i;
for (i = 0; i < p->n_pdds; i++)
- if (p->pdds[i] && p->pdds[i]->dev->kgd == kgd) {
+ if (p->pdds[i] && p->pdds[i]->dev->adev == adev) {
*gpuid = p->pdds[i]->dev->id;
*gpuidx = i;
return 0;
@@ -1951,10 +1947,10 @@ void kfd_flush_tlb(struct kfd_process_device *pdd, enum TLB_FLUSH_TYPE type)
* only happens when the first queue is created.
*/
if (pdd->qpd.vmid)
- amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->kgd,
+ amdgpu_amdkfd_flush_gpu_tlb_vmid(dev->adev,
pdd->qpd.vmid);
} else {
- amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->kgd,
+ amdgpu_amdkfd_flush_gpu_tlb_pasid(dev->adev,
pdd->process->pasid, type);
}
}
Copyright © 1996 – 2023 Jason A. Donenfeld. All Rights Reverse Engineered.