summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd/kfd_device.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_device.c')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_device.c123
1 files changed, 60 insertions, 63 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 3f95f7cb4019..61fff25b4ce7 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -26,7 +26,7 @@
#include <linux/slab.h>
#include "kfd_priv.h"
#include "kfd_device_queue_manager.h"
-#include "kfd_pm4_headers.h"
+#include "kfd_pm4_headers_vi.h"
#define MQD_SIZE_ALIGNED 768
@@ -98,11 +98,14 @@ static const struct kfd_device_info *lookup_device_info(unsigned short did)
for (i = 0; i < ARRAY_SIZE(supported_devices); i++) {
if (supported_devices[i].did == did) {
- BUG_ON(supported_devices[i].device_info == NULL);
+ WARN_ON(!supported_devices[i].device_info);
return supported_devices[i].device_info;
}
}
+ dev_warn(kfd_device, "DID %04x is missing in supported_devices\n",
+ did);
+
return NULL;
}
@@ -114,8 +117,10 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
const struct kfd_device_info *device_info =
lookup_device_info(pdev->device);
- if (!device_info)
+ if (!device_info) {
+ dev_err(kfd_device, "kgd2kfd_probe failed\n");
return NULL;
+ }
kfd = kzalloc(sizeof(*kfd), GFP_KERNEL);
if (!kfd)
@@ -152,15 +157,16 @@ static bool device_iommu_pasid_init(struct kfd_dev *kfd)
}
if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) {
- dev_err(kfd_device, "error required iommu flags ats(%i), pri(%i), pasid(%i)\n",
+ dev_err(kfd_device, "error required iommu flags ats %i, pri %i, pasid %i\n",
(iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
(iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
- (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP) != 0);
+ (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP)
+ != 0);
return false;
}
pasid_limit = min_t(unsigned int,
- (unsigned int)1 << kfd->device_info->max_pasid_bits,
+ (unsigned int)(1 << kfd->device_info->max_pasid_bits),
iommu_info.max_pasids);
/*
* last pasid is used for kernel queues doorbells
@@ -211,9 +217,8 @@ static int iommu_invalid_ppr_cb(struct pci_dev *pdev, int pasid,
flags);
dev = kfd_device_by_pci_dev(pdev);
- BUG_ON(dev == NULL);
-
- kfd_signal_iommu_event(dev, pasid, address,
+ if (!WARN_ON(!dev))
+ kfd_signal_iommu_event(dev, pasid, address,
flags & PPR_FAULT_WRITE, flags & PPR_FAULT_EXEC);
return AMD_IOMMU_INV_PRI_RSP_INVALID;
@@ -234,9 +239,9 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
* calculate max size of runlist packet.
* There can be only 2 packets at once
*/
- size += (KFD_MAX_NUM_OF_PROCESSES * sizeof(struct pm4_map_process) +
- max_num_of_queues_per_device *
- sizeof(struct pm4_map_queues) + sizeof(struct pm4_runlist)) * 2;
+ size += (KFD_MAX_NUM_OF_PROCESSES * sizeof(struct pm4_mes_map_process) +
+ max_num_of_queues_per_device * sizeof(struct pm4_mes_map_queues)
+ + sizeof(struct pm4_mes_runlist)) * 2;
/* Add size of HIQ & DIQ */
size += KFD_KERNEL_QUEUE_SIZE * 2;
@@ -247,42 +252,37 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
if (kfd->kfd2kgd->init_gtt_mem_allocation(
kfd->kgd, size, &kfd->gtt_mem,
&kfd->gtt_start_gpu_addr, &kfd->gtt_start_cpu_ptr)){
- dev_err(kfd_device,
- "Could not allocate %d bytes for device (%x:%x)\n",
- size, kfd->pdev->vendor, kfd->pdev->device);
+ dev_err(kfd_device, "Could not allocate %d bytes\n", size);
goto out;
}
- dev_info(kfd_device,
- "Allocated %d bytes on gart for device(%x:%x)\n",
- size, kfd->pdev->vendor, kfd->pdev->device);
+ dev_info(kfd_device, "Allocated %d bytes on gart\n", size);
/* Initialize GTT sa with 512 byte chunk size */
if (kfd_gtt_sa_init(kfd, size, 512) != 0) {
- dev_err(kfd_device,
- "Error initializing gtt sub-allocator\n");
+ dev_err(kfd_device, "Error initializing gtt sub-allocator\n");
goto kfd_gtt_sa_init_error;
}
- kfd_doorbell_init(kfd);
-
- if (kfd_topology_add_device(kfd) != 0) {
+ if (kfd_doorbell_init(kfd)) {
dev_err(kfd_device,
- "Error adding device (%x:%x) to topology\n",
- kfd->pdev->vendor, kfd->pdev->device);
+ "Error initializing doorbell aperture\n");
+ goto kfd_doorbell_error;
+ }
+
+ if (kfd_topology_add_device(kfd)) {
+ dev_err(kfd_device, "Error adding device to topology\n");
goto kfd_topology_add_device_error;
}
if (kfd_interrupt_init(kfd)) {
- dev_err(kfd_device,
- "Error initializing interrupts for device (%x:%x)\n",
- kfd->pdev->vendor, kfd->pdev->device);
+ dev_err(kfd_device, "Error initializing interrupts\n");
goto kfd_interrupt_error;
}
if (!device_iommu_pasid_init(kfd)) {
dev_err(kfd_device,
- "Error initializing iommuv2 for device (%x:%x)\n",
+ "Error initializing iommuv2 for device %x:%x\n",
kfd->pdev->vendor, kfd->pdev->device);
goto device_iommu_pasid_error;
}
@@ -292,15 +292,13 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
kfd->dqm = device_queue_manager_init(kfd);
if (!kfd->dqm) {
- dev_err(kfd_device,
- "Error initializing queue manager for device (%x:%x)\n",
- kfd->pdev->vendor, kfd->pdev->device);
+ dev_err(kfd_device, "Error initializing queue manager\n");
goto device_queue_manager_error;
}
- if (kfd->dqm->ops.start(kfd->dqm) != 0) {
+ if (kfd->dqm->ops.start(kfd->dqm)) {
dev_err(kfd_device,
- "Error starting queuen manager for device (%x:%x)\n",
+ "Error starting queue manager for device %x:%x\n",
kfd->pdev->vendor, kfd->pdev->device);
goto dqm_start_error;
}
@@ -308,10 +306,10 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
kfd->dbgmgr = NULL;
kfd->init_complete = true;
- dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor,
+ dev_info(kfd_device, "added device %x:%x\n", kfd->pdev->vendor,
kfd->pdev->device);
- pr_debug("kfd: Starting kfd with the following scheduling policy %d\n",
+ pr_debug("Starting kfd with the following scheduling policy %d\n",
sched_policy);
goto out;
@@ -325,11 +323,13 @@ device_iommu_pasid_error:
kfd_interrupt_error:
kfd_topology_remove_device(kfd);
kfd_topology_add_device_error:
+ kfd_doorbell_fini(kfd);
+kfd_doorbell_error:
kfd_gtt_sa_fini(kfd);
kfd_gtt_sa_init_error:
kfd->kfd2kgd->free_gtt_mem(kfd->kgd, kfd->gtt_mem);
dev_err(kfd_device,
- "device (%x:%x) NOT added due to errors\n",
+ "device %x:%x NOT added due to errors\n",
kfd->pdev->vendor, kfd->pdev->device);
out:
return kfd->init_complete;
@@ -342,6 +342,7 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd)
amd_iommu_free_device(kfd->pdev);
kfd_interrupt_exit(kfd);
kfd_topology_remove_device(kfd);
+ kfd_doorbell_fini(kfd);
kfd_gtt_sa_fini(kfd);
kfd->kfd2kgd->free_gtt_mem(kfd->kgd, kfd->gtt_mem);
}
@@ -351,8 +352,6 @@ void kgd2kfd_device_exit(struct kfd_dev *kfd)
void kgd2kfd_suspend(struct kfd_dev *kfd)
{
- BUG_ON(kfd == NULL);
-
if (kfd->init_complete) {
kfd->dqm->ops.stop(kfd->dqm);
amd_iommu_set_invalidate_ctx_cb(kfd->pdev, NULL);
@@ -366,14 +365,15 @@ int kgd2kfd_resume(struct kfd_dev *kfd)
unsigned int pasid_limit;
int err;
- BUG_ON(kfd == NULL);
-
pasid_limit = kfd_get_pasid_limit();
if (kfd->init_complete) {
err = amd_iommu_init_device(kfd->pdev, pasid_limit);
- if (err < 0)
+ if (err < 0) {
+ dev_err(kfd_device, "failed to initialize iommu\n");
return -ENXIO;
+ }
+
amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
iommu_pasid_shutdown_callback);
amd_iommu_set_invalid_ppr_cb(kfd->pdev, iommu_invalid_ppr_cb);
@@ -402,26 +402,27 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size,
unsigned int chunk_size)
{
- unsigned int num_of_bits;
+ unsigned int num_of_longs;
- BUG_ON(!kfd);
- BUG_ON(!kfd->gtt_mem);
- BUG_ON(buf_size < chunk_size);
- BUG_ON(buf_size == 0);
- BUG_ON(chunk_size == 0);
+ if (WARN_ON(buf_size < chunk_size))
+ return -EINVAL;
+ if (WARN_ON(buf_size == 0))
+ return -EINVAL;
+ if (WARN_ON(chunk_size == 0))
+ return -EINVAL;
kfd->gtt_sa_chunk_size = chunk_size;
kfd->gtt_sa_num_of_chunks = buf_size / chunk_size;
- num_of_bits = kfd->gtt_sa_num_of_chunks / BITS_PER_BYTE;
- BUG_ON(num_of_bits == 0);
+ num_of_longs = (kfd->gtt_sa_num_of_chunks + BITS_PER_LONG - 1) /
+ BITS_PER_LONG;
- kfd->gtt_sa_bitmap = kzalloc(num_of_bits, GFP_KERNEL);
+ kfd->gtt_sa_bitmap = kcalloc(num_of_longs, sizeof(long), GFP_KERNEL);
if (!kfd->gtt_sa_bitmap)
return -ENOMEM;
- pr_debug("kfd: gtt_sa_num_of_chunks = %d, gtt_sa_bitmap = %p\n",
+ pr_debug("gtt_sa_num_of_chunks = %d, gtt_sa_bitmap = %p\n",
kfd->gtt_sa_num_of_chunks, kfd->gtt_sa_bitmap);
mutex_init(&kfd->gtt_sa_lock);
@@ -455,8 +456,6 @@ int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size,
{
unsigned int found, start_search, cur_size;
- BUG_ON(!kfd);
-
if (size == 0)
return -EINVAL;
@@ -467,7 +466,7 @@ int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size,
if ((*mem_obj) == NULL)
return -ENOMEM;
- pr_debug("kfd: allocated mem_obj = %p for size = %d\n", *mem_obj, size);
+ pr_debug("Allocated mem_obj = %p for size = %d\n", *mem_obj, size);
start_search = 0;
@@ -479,7 +478,7 @@ kfd_gtt_restart_search:
kfd->gtt_sa_num_of_chunks,
start_search);
- pr_debug("kfd: found = %d\n", found);
+ pr_debug("Found = %d\n", found);
/* If there wasn't any free chunk, bail out */
if (found == kfd->gtt_sa_num_of_chunks)
@@ -497,12 +496,12 @@ kfd_gtt_restart_search:
found,
kfd->gtt_sa_chunk_size);
- pr_debug("kfd: gpu_addr = %p, cpu_addr = %p\n",
+ pr_debug("gpu_addr = %p, cpu_addr = %p\n",
(uint64_t *) (*mem_obj)->gpu_addr, (*mem_obj)->cpu_ptr);
/* If we need only one chunk, mark it as allocated and get out */
if (size <= kfd->gtt_sa_chunk_size) {
- pr_debug("kfd: single bit\n");
+ pr_debug("Single bit\n");
set_bit(found, kfd->gtt_sa_bitmap);
goto kfd_gtt_out;
}
@@ -537,7 +536,7 @@ kfd_gtt_restart_search:
} while (cur_size > 0);
- pr_debug("kfd: range_start = %d, range_end = %d\n",
+ pr_debug("range_start = %d, range_end = %d\n",
(*mem_obj)->range_start, (*mem_obj)->range_end);
/* Mark the chunks as allocated */
@@ -551,7 +550,7 @@ kfd_gtt_out:
return 0;
kfd_gtt_no_free_chunk:
- pr_debug("kfd: allocation failed with mem_obj = %p\n", mem_obj);
+ pr_debug("Allocation failed with mem_obj = %p\n", mem_obj);
mutex_unlock(&kfd->gtt_sa_lock);
kfree(mem_obj);
return -ENOMEM;
@@ -561,13 +560,11 @@ int kfd_gtt_sa_free(struct kfd_dev *kfd, struct kfd_mem_obj *mem_obj)
{
unsigned int bit;
- BUG_ON(!kfd);
-
/* Act like kfree when trying to free a NULL object */
if (!mem_obj)
return 0;
- pr_debug("kfd: free mem_obj = %p, range_start = %d, range_end = %d\n",
+ pr_debug("Free mem_obj = %p, range_start = %d, range_end = %d\n",
mem_obj, mem_obj->range_start, mem_obj->range_end);
mutex_lock(&kfd->gtt_sa_lock);