summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_chardev.c2
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_crat.c3
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_device.c28
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c48
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_migrate.c14
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_priv.h1
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_process.c28
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c10
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c11
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h3
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_svm.c7
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_topology.c2
12 files changed, 113 insertions, 44 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 3e6b4736a7fe..065d87841459 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -365,7 +365,7 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
p->pasid,
dev->id);
- err = pqm_create_queue(&p->pqm, dev, filep, &q_properties, &queue_id,
+ err = pqm_create_queue(&p->pqm, dev, &q_properties, &queue_id,
NULL, NULL, NULL, &doorbell_offset_in_process);
if (err != 0)
goto err_create_queue;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 48caecf7e72e..723f1220e1cc 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -28,6 +28,7 @@
#include "kfd_topology.h"
#include "amdgpu.h"
#include "amdgpu_amdkfd.h"
+#include "amdgpu_xgmi.h"
/* GPU Processor ID base for dGPUs for which VCRAT needs to be created.
* GPU processor ID are expressed with Bit[31]=1.
@@ -2329,6 +2330,8 @@ static int kfd_create_vcrat_image_gpu(void *pcrat_image,
continue;
if (peer_dev->gpu->kfd->hive_id != kdev->kfd->hive_id)
continue;
+ if (!amdgpu_xgmi_get_is_sharing_enabled(kdev->adev, peer_dev->gpu->adev))
+ continue;
sub_type_hdr = (typeof(sub_type_hdr))(
(char *)sub_type_hdr +
sizeof(struct crat_subtype_iolink));
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index fad1c8f2bc83..956198da7859 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -534,7 +534,8 @@ static void kfd_cwsr_init(struct kfd_dev *kfd)
kfd->cwsr_isa = cwsr_trap_gfx11_hex;
kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx11_hex);
} else {
- BUILD_BUG_ON(sizeof(cwsr_trap_gfx12_hex) > PAGE_SIZE);
+ BUILD_BUG_ON(sizeof(cwsr_trap_gfx12_hex)
+ > KFD_CWSR_TMA_OFFSET);
kfd->cwsr_isa = cwsr_trap_gfx12_hex;
kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx12_hex);
}
@@ -1392,6 +1393,13 @@ void kfd_dec_compute_active(struct kfd_node *node)
WARN_ONCE(count < 0, "Compute profile ref. count error");
}
+static bool kfd_compute_active(struct kfd_node *node)
+{
+ if (atomic_read(&node->kfd->compute_profile))
+ return true;
+ return false;
+}
+
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
{
/*
@@ -1485,6 +1493,24 @@ int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id)
return node->dqm->ops.halt(node->dqm);
}
+bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id)
+{
+ struct kfd_node *node;
+
+ if (!kfd->init_complete)
+ return false;
+
+ if (node_id >= kfd->num_nodes) {
+ dev_warn(kfd->adev->dev, "Invalid node ID: %u exceeds %u\n",
+ node_id, kfd->num_nodes - 1);
+ return false;
+ }
+
+ node = kfd->nodes[node_id];
+
+ return kfd_compute_active(node);
+}
+
#if defined(CONFIG_DEBUG_FS)
/* This function will send a package to HIQ to hang the HWS
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 648f40091aa3..c79fe9069e22 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -202,6 +202,8 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q,
int r, queue_type;
uint64_t wptr_addr_off;
+ if (!dqm->sched_running || dqm->sched_halt)
+ return 0;
if (!down_read_trylock(&adev->reset_domain->sem))
return -EIO;
@@ -270,6 +272,8 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q,
int r;
struct mes_remove_queue_input queue_input;
+ if (!dqm->sched_running || dqm->sched_halt)
+ return 0;
if (!down_read_trylock(&adev->reset_domain->sem))
return -EIO;
@@ -292,7 +296,7 @@ static int remove_queue_mes(struct device_queue_manager *dqm, struct queue *q,
return r;
}
-static int remove_all_queues_mes(struct device_queue_manager *dqm)
+static int remove_all_kfd_queues_mes(struct device_queue_manager *dqm)
{
struct device_process_node *cur;
struct device *dev = dqm->dev->adev->dev;
@@ -319,6 +323,33 @@ static int remove_all_queues_mes(struct device_queue_manager *dqm)
return retval;
}
+static int add_all_kfd_queues_mes(struct device_queue_manager *dqm)
+{
+ struct device_process_node *cur;
+ struct device *dev = dqm->dev->adev->dev;
+ struct qcm_process_device *qpd;
+ struct queue *q;
+ int retval = 0;
+
+ list_for_each_entry(cur, &dqm->queues, list) {
+ qpd = cur->qpd;
+ list_for_each_entry(q, &qpd->queues_list, list) {
+ if (!q->properties.is_active)
+ continue;
+ retval = add_queue_mes(dqm, q, qpd);
+ if (retval) {
+ dev_err(dev, "%s: Failed to add queue %d for dev %d",
+ __func__,
+ q->properties.queue_id,
+ dqm->dev->id);
+ return retval;
+ }
+ }
+ }
+
+ return retval;
+}
+
static int suspend_all_queues_mes(struct device_queue_manager *dqm)
{
struct amdgpu_device *adev = (struct amdgpu_device *)dqm->dev->adev;
@@ -1742,7 +1773,7 @@ static int halt_cpsch(struct device_queue_manager *dqm)
KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0,
USE_DEFAULT_GRACE_PERIOD, false);
else
- ret = remove_all_queues_mes(dqm);
+ ret = remove_all_kfd_queues_mes(dqm);
}
dqm->sched_halt = true;
dqm_unlock(dqm);
@@ -1768,6 +1799,9 @@ static int unhalt_cpsch(struct device_queue_manager *dqm)
ret = execute_queues_cpsch(dqm,
KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES,
0, USE_DEFAULT_GRACE_PERIOD);
+ else
+ ret = add_all_kfd_queues_mes(dqm);
+
dqm_unlock(dqm);
return ret;
@@ -1867,7 +1901,7 @@ static int stop_cpsch(struct device_queue_manager *dqm)
if (!dqm->dev->kfd->shared_resources.enable_mes)
unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0, USE_DEFAULT_GRACE_PERIOD, false);
else
- remove_all_queues_mes(dqm);
+ remove_all_kfd_queues_mes(dqm);
dqm->sched_running = false;
@@ -2048,7 +2082,7 @@ int amdkfd_fence_wait_timeout(struct device_queue_manager *dqm,
{
unsigned long end_jiffies = msecs_to_jiffies(timeout_ms) + jiffies;
struct device *dev = dqm->dev->adev->dev;
- uint64_t *fence_addr = dqm->fence_addr;
+ uint64_t *fence_addr = dqm->fence_addr;
while (*fence_addr != fence_value) {
/* Fatal err detected, this response won't come */
@@ -2254,6 +2288,7 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
goto out;
*dqm->fence_addr = KFD_FENCE_INIT;
+ mb();
pm_send_query_status(&dqm->packet_mgr, dqm->fence_gpu_addr,
KFD_FENCE_COMPLETED);
/* should be timed out */
@@ -3173,7 +3208,7 @@ struct copy_context_work_handler_workarea {
struct kfd_process *p;
};
-static void copy_context_work_handler (struct work_struct *work)
+static void copy_context_work_handler(struct work_struct *work)
{
struct copy_context_work_handler_workarea *workarea;
struct mqd_manager *mqd_mgr;
@@ -3200,6 +3235,9 @@ static void copy_context_work_handler (struct work_struct *work)
struct qcm_process_device *qpd = &pdd->qpd;
list_for_each_entry(q, &qpd->queues_list, list) {
+ if (q->properties.type != KFD_QUEUE_TYPE_COMPUTE)
+ continue;
+
mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_CP];
/* We ignore the return value from get_wave_state
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
index 8ee3d07ffbdf..eacfeb32f35d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
@@ -445,14 +445,13 @@ svm_migrate_vma_to_vram(struct kfd_node *node, struct svm_range *prange,
pr_debug("successful/cpages/npages 0x%lx/0x%lx/0x%lx\n",
mpages, cpages, migrate.npages);
- kfd_smi_event_migration_end(node, p->lead_thread->pid,
- start >> PAGE_SHIFT, end >> PAGE_SHIFT,
- 0, node->id, trigger);
-
svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
out_free:
kvfree(buf);
+ kfd_smi_event_migration_end(node, p->lead_thread->pid,
+ start >> PAGE_SHIFT, end >> PAGE_SHIFT,
+ 0, node->id, trigger, r);
out:
if (!r && mpages) {
pdd = svm_range_get_pdd_by_node(prange, node);
@@ -751,14 +750,13 @@ svm_migrate_vma_to_ram(struct kfd_node *node, struct svm_range *prange,
svm_migrate_copy_done(adev, mfence);
migrate_vma_finalize(&migrate);
- kfd_smi_event_migration_end(node, p->lead_thread->pid,
- start >> PAGE_SHIFT, end >> PAGE_SHIFT,
- node->id, 0, trigger);
-
svm_range_dma_unmap_dev(adev->dev, scratch, 0, npages);
out_free:
kvfree(buf);
+ kfd_smi_event_migration_end(node, p->lead_thread->pid,
+ start >> PAGE_SHIFT, end >> PAGE_SHIFT,
+ node->id, 0, trigger, r);
out:
if (!r && cpages) {
mpages = cpages - upages;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 26e48fdc8728..9e5ca0b93b2a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1347,7 +1347,6 @@ int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p);
void pqm_uninit(struct process_queue_manager *pqm);
int pqm_create_queue(struct process_queue_manager *pqm,
struct kfd_node *dev,
- struct file *f,
struct queue_properties *properties,
unsigned int *qid,
const struct kfd_criu_queue_priv_data *q_data,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index d4aa843aacfd..87cd52cf4ee9 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -271,11 +271,9 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
struct kfd_process *proc = NULL;
struct kfd_process_device *pdd = NULL;
int i;
- struct kfd_cu_occupancy cu_occupancy[AMDGPU_MAX_QUEUES];
+ struct kfd_cu_occupancy *cu_occupancy;
u32 queue_format;
- memset(cu_occupancy, 0x0, sizeof(cu_occupancy));
-
pdd = container_of(attr, struct kfd_process_device, attr_cu_occupancy);
dev = pdd->dev;
if (dev->kfd2kgd->get_cu_occupancy == NULL)
@@ -293,6 +291,10 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
wave_cnt = 0;
max_waves_per_cu = 0;
+ cu_occupancy = kcalloc(AMDGPU_MAX_QUEUES, sizeof(*cu_occupancy), GFP_KERNEL);
+ if (!cu_occupancy)
+ return -ENOMEM;
+
/*
* For GFX 9.4.3, fetch the CU occupancy from the first XCC in the partition.
* For AQL queues, because of cooperative dispatch we multiply the wave count
@@ -318,6 +320,7 @@ static int kfd_get_cu_occupancy(struct attribute *attr, char *buffer)
/* Translate wave count to number of compute units */
cu_cnt = (wave_cnt + (max_waves_per_cu - 1)) / max_waves_per_cu;
+ kfree(cu_occupancy);
return snprintf(buffer, PAGE_SIZE, "%d\n", cu_cnt);
}
@@ -338,8 +341,8 @@ static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
attr_sdma);
struct kfd_sdma_activity_handler_workarea sdma_activity_work_handler;
- INIT_WORK(&sdma_activity_work_handler.sdma_activity_work,
- kfd_sdma_activity_worker);
+ INIT_WORK_ONSTACK(&sdma_activity_work_handler.sdma_activity_work,
+ kfd_sdma_activity_worker);
sdma_activity_work_handler.pdd = pdd;
sdma_activity_work_handler.sdma_activity_counter = 0;
@@ -347,6 +350,7 @@ static ssize_t kfd_procfs_show(struct kobject *kobj, struct attribute *attr,
schedule_work(&sdma_activity_work_handler.sdma_activity_work);
flush_work(&sdma_activity_work_handler.sdma_activity_work);
+ destroy_work_on_stack(&sdma_activity_work_handler.sdma_activity_work);
return snprintf(buffer, PAGE_SIZE, "%llu\n",
(sdma_activity_work_handler.sdma_activity_counter)/
@@ -850,8 +854,10 @@ struct kfd_process *kfd_create_process(struct task_struct *thread)
goto out;
}
- /* A prior open of /dev/kfd could have already created the process. */
- process = find_process(thread, false);
+ /* A prior open of /dev/kfd could have already created the process.
+ * find_process will increase process kref in this case
+ */
+ process = find_process(thread, true);
if (process) {
pr_debug("Process already found\n");
} else {
@@ -899,8 +905,6 @@ struct kfd_process *kfd_create_process(struct task_struct *thread)
init_waitqueue_head(&process->wait_irq_drain);
}
out:
- if (!IS_ERR(process))
- kref_get(&process->ref);
mutex_unlock(&kfd_processes_mutex);
mmput(thread->mm);
@@ -1186,10 +1190,8 @@ static void kfd_process_ref_release(struct kref *ref)
static struct mmu_notifier *kfd_process_alloc_notifier(struct mm_struct *mm)
{
- int idx = srcu_read_lock(&kfd_processes_srcu);
- struct kfd_process *p = find_process_by_mm(mm);
-
- srcu_read_unlock(&kfd_processes_srcu, idx);
+ /* This increments p->ref counter if kfd process p exists */
+ struct kfd_process *p = kfd_lookup_process_by_mm(mm);
return p ? &p->mmu_notifier : ERR_PTR(-ESRCH);
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 01b960b15274..c76db22a1000 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -235,7 +235,7 @@ void pqm_uninit(struct process_queue_manager *pqm)
static int init_user_queue(struct process_queue_manager *pqm,
struct kfd_node *dev, struct queue **q,
struct queue_properties *q_properties,
- struct file *f, unsigned int qid)
+ unsigned int qid)
{
int retval;
@@ -300,7 +300,6 @@ cleanup:
int pqm_create_queue(struct process_queue_manager *pqm,
struct kfd_node *dev,
- struct file *f,
struct queue_properties *properties,
unsigned int *qid,
const struct kfd_criu_queue_priv_data *q_data,
@@ -374,7 +373,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
* allocate_sdma_queue() in create_queue() has the
* corresponding check logic.
*/
- retval = init_user_queue(pqm, dev, &q, properties, f, *qid);
+ retval = init_user_queue(pqm, dev, &q, properties, *qid);
if (retval != 0)
goto err_create_queue;
pqn->q = q;
@@ -395,7 +394,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
goto err_create_queue;
}
- retval = init_user_queue(pqm, dev, &q, properties, f, *qid);
+ retval = init_user_queue(pqm, dev, &q, properties, *qid);
if (retval != 0)
goto err_create_queue;
pqn->q = q;
@@ -1029,8 +1028,7 @@ int kfd_criu_restore_queue(struct kfd_process *p,
print_queue_properties(&qp);
- ret = pqm_create_queue(&p->pqm, pdd->dev, NULL, &qp, &queue_id, q_data, mqd, ctl_stack,
- NULL);
+ ret = pqm_create_queue(&p->pqm, pdd->dev, &qp, &queue_id, q_data, mqd, ctl_stack, NULL);
if (ret) {
pr_err("Failed to create new queue err:%d\n", ret);
goto exit;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index de8b9abf7afc..9b8169761ec5 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -44,7 +44,7 @@ struct kfd_smi_client {
bool suser;
};
-#define MAX_KFIFO_SIZE 1024
+#define KFD_MAX_KFIFO_SIZE 8192
static __poll_t kfd_smi_ev_poll(struct file *, struct poll_table_struct *);
static ssize_t kfd_smi_ev_read(struct file *, char __user *, size_t, loff_t *);
@@ -86,7 +86,7 @@ static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user,
struct kfd_smi_client *client = filep->private_data;
unsigned char *buf;
- size = min_t(size_t, size, MAX_KFIFO_SIZE);
+ size = min_t(size_t, size, KFD_MAX_KFIFO_SIZE);
buf = kmalloc(size, GFP_KERNEL);
if (!buf)
return -ENOMEM;
@@ -292,12 +292,13 @@ void kfd_smi_event_migration_start(struct kfd_node *node, pid_t pid,
void kfd_smi_event_migration_end(struct kfd_node *node, pid_t pid,
unsigned long start, unsigned long end,
- uint32_t from, uint32_t to, uint32_t trigger)
+ uint32_t from, uint32_t to, uint32_t trigger,
+ int error_code)
{
kfd_smi_event_add(pid, node, KFD_SMI_EVENT_MIGRATE_END,
KFD_EVENT_FMT_MIGRATE_END(
ktime_get_boottime_ns(), pid, start, end - start,
- from, to, trigger));
+ from, to, trigger, error_code));
}
void kfd_smi_event_queue_eviction(struct kfd_node *node, pid_t pid,
@@ -354,7 +355,7 @@ int kfd_smi_event_open(struct kfd_node *dev, uint32_t *fd)
return -ENOMEM;
INIT_LIST_HEAD(&client->list);
- ret = kfifo_alloc(&client->fifo, MAX_KFIFO_SIZE, GFP_KERNEL);
+ ret = kfifo_alloc(&client->fifo, KFD_MAX_KFIFO_SIZE, GFP_KERNEL);
if (ret) {
kfree(client);
return ret;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
index 85010b8307f8..503bff13d815 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
@@ -44,7 +44,8 @@ void kfd_smi_event_migration_start(struct kfd_node *node, pid_t pid,
uint32_t trigger);
void kfd_smi_event_migration_end(struct kfd_node *node, pid_t pid,
unsigned long start, unsigned long end,
- uint32_t from, uint32_t to, uint32_t trigger);
+ uint32_t from, uint32_t to, uint32_t trigger,
+ int error_code);
void kfd_smi_event_queue_eviction(struct kfd_node *node, pid_t pid,
uint32_t trigger);
void kfd_smi_event_queue_restore(struct kfd_node *node, pid_t pid);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 1893c27746a5..3e2911895c74 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -3111,8 +3111,6 @@ retry_write_locked:
start = max_t(unsigned long, ALIGN_DOWN(addr, size), prange->start);
last = min_t(unsigned long, ALIGN(addr + 1, size) - 1, prange->last);
if (prange->actual_loc != 0 || best_loc != 0) {
- migration = true;
-
if (best_loc) {
r = svm_migrate_to_vram(prange, best_loc, start, last,
mm, KFD_MIGRATE_TRIGGER_PAGEFAULT_GPU);
@@ -3135,7 +3133,9 @@ retry_write_locked:
if (r) {
pr_debug("failed %d to migrate svms %p [0x%lx 0x%lx]\n",
r, svms, start, last);
- goto out_unlock_range;
+ goto out_migrate_fail;
+ } else {
+ migration = true;
}
}
@@ -3145,6 +3145,7 @@ retry_write_locked:
pr_debug("failed %d to map svms 0x%p [0x%lx 0x%lx] to gpus\n",
r, svms, start, last);
+out_migrate_fail:
kfd_smi_event_page_fault_end(node, p->lead_thread->pid, addr,
migration);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index 3871591c9aec..9476e30d6baa 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -1998,6 +1998,8 @@ static void kfd_topology_set_capabilities(struct kfd_topology_device *dev)
if (KFD_GC_VERSION(dev->gpu) >= IP_VERSION(9, 4, 2))
dev->node_props.capability |=
HSA_CAP_TRAP_DEBUG_PRECISE_MEMORY_OPERATIONS_SUPPORTED;
+
+ dev->node_props.capability |= HSA_CAP_PER_QUEUE_RESET_SUPPORTED;
} else {
dev->node_props.debug_prop |= HSA_DBG_WATCH_ADDR_MASK_LO_BIT_GFX10 |
HSA_DBG_WATCH_ADDR_MASK_HI_BIT;
generated by cgit v1.2.3 (git 2.39.0) at 2025-01-19 03:36:22 +0300