summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c')
-rw-r--r--drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c146
1 files changed, 89 insertions, 57 deletions
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index 0472b56de245..a499449fcb06 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -29,6 +29,7 @@
#include "amdgpu_vm.h"
#include "kfd_priv.h"
#include "kfd_smi_events.h"
+#include "amdgpu_reset.h"
struct kfd_smi_client {
struct list_head list;
@@ -36,14 +37,14 @@ struct kfd_smi_client {
wait_queue_head_t wait_queue;
/* events enabled */
uint64_t events;
- struct kfd_dev *dev;
+ struct kfd_node *dev;
spinlock_t lock;
struct rcu_head rcu;
pid_t pid;
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 *);
@@ -85,7 +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;
@@ -149,7 +150,7 @@ static void kfd_smi_ev_client_free(struct rcu_head *p)
static int kfd_smi_ev_release(struct inode *inode, struct file *filep)
{
struct kfd_smi_client *client = filep->private_data;
- struct kfd_dev *dev = client->dev;
+ struct kfd_node *dev = client->dev;
spin_lock(&dev->smi_lock);
list_del_rcu(&client->list);
@@ -162,16 +163,15 @@ static int kfd_smi_ev_release(struct inode *inode, struct file *filep)
static bool kfd_smi_ev_enabled(pid_t pid, struct kfd_smi_client *client,
unsigned int event)
{
- uint64_t all = KFD_SMI_EVENT_MASK_FROM_INDEX(KFD_SMI_EVENT_ALL_PROCESS);
uint64_t events = READ_ONCE(client->events);
- if (pid && client->pid != pid && !(client->suser && (events & all)))
+ if (pid && client->pid != pid && !client->suser)
return false;
return events & KFD_SMI_EVENT_MASK_FROM_INDEX(event);
}
-static void add_event_to_kfifo(pid_t pid, struct kfd_dev *dev,
+static void add_event_to_kfifo(pid_t pid, struct kfd_node *dev,
unsigned int smi_event, char *event_msg, int len)
{
struct kfd_smi_client *client;
@@ -196,7 +196,7 @@ static void add_event_to_kfifo(pid_t pid, struct kfd_dev *dev,
}
__printf(4, 5)
-static void kfd_smi_event_add(pid_t pid, struct kfd_dev *dev,
+static void kfd_smi_event_add(pid_t pid, struct kfd_node *dev,
unsigned int event, char *fmt, ...)
{
char fifo_in[KFD_SMI_EVENT_MSG_SIZE];
@@ -215,9 +215,11 @@ static void kfd_smi_event_add(pid_t pid, struct kfd_dev *dev,
add_event_to_kfifo(pid, dev, event, fifo_in, len);
}
-void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
+void kfd_smi_event_update_gpu_reset(struct kfd_node *dev, bool post_reset,
+ struct amdgpu_reset_context *reset_context)
{
unsigned int event;
+ char reset_cause[64];
if (post_reset) {
event = KFD_SMI_EVENT_GPU_POST_RESET;
@@ -225,83 +227,92 @@ void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
event = KFD_SMI_EVENT_GPU_PRE_RESET;
++(dev->reset_seq_num);
}
- kfd_smi_event_add(0, dev, event, "%x\n", dev->reset_seq_num);
+
+ memset(reset_cause, 0, sizeof(reset_cause));
+
+ if (reset_context)
+ amdgpu_reset_get_desc(reset_context, reset_cause,
+ sizeof(reset_cause));
+
+ kfd_smi_event_add(0, dev, event, KFD_EVENT_FMT_UPDATE_GPU_RESET(
+ dev->reset_seq_num, reset_cause));
}
-void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
+void kfd_smi_event_update_thermal_throttling(struct kfd_node *dev,
uint64_t throttle_bitmask)
{
- kfd_smi_event_add(0, dev, KFD_SMI_EVENT_THERMAL_THROTTLE, "%llx:%llx\n",
+ kfd_smi_event_add(0, dev, KFD_SMI_EVENT_THERMAL_THROTTLE, KFD_EVENT_FMT_THERMAL_THROTTLING(
throttle_bitmask,
- amdgpu_dpm_get_thermal_throttling_counter(dev->adev));
+ amdgpu_dpm_get_thermal_throttling_counter(dev->adev)));
}
-void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
+void kfd_smi_event_update_vmfault(struct kfd_node *dev, uint16_t pasid)
{
- struct amdgpu_task_info task_info;
-
- memset(&task_info, 0, sizeof(struct amdgpu_task_info));
- amdgpu_vm_get_task_info(dev->adev, pasid, &task_info);
- /* Report VM faults from user applications, not retry from kernel */
- if (!task_info.pid)
- return;
-
- kfd_smi_event_add(0, dev, KFD_SMI_EVENT_VMFAULT, "%x:%s\n",
- task_info.pid, task_info.task_name);
+ struct amdgpu_task_info *task_info;
+
+ task_info = amdgpu_vm_get_task_info_pasid(dev->adev, pasid);
+ if (task_info) {
+ /* Report VM faults from user applications, not retry from kernel */
+ if (task_info->task.pid)
+ kfd_smi_event_add(0, dev, KFD_SMI_EVENT_VMFAULT, KFD_EVENT_FMT_VMFAULT(
+ task_info->task.pid, task_info->task.comm));
+ amdgpu_vm_put_task_info(task_info);
+ }
}
-void kfd_smi_event_page_fault_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_page_fault_start(struct kfd_node *node, pid_t pid,
unsigned long address, bool write_fault,
ktime_t ts)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_PAGE_FAULT_START,
- "%lld -%d @%lx(%x) %c\n", ktime_to_ns(ts), pid,
- address, dev->id, write_fault ? 'W' : 'R');
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_PAGE_FAULT_START,
+ KFD_EVENT_FMT_PAGEFAULT_START(ktime_to_ns(ts), pid,
+ address, node->id, write_fault ? 'W' : 'R'));
}
-void kfd_smi_event_page_fault_end(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_page_fault_end(struct kfd_node *node, pid_t pid,
unsigned long address, bool migration)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_PAGE_FAULT_END,
- "%lld -%d @%lx(%x) %c\n", ktime_get_boottime_ns(),
- pid, address, dev->id, migration ? 'M' : 'U');
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_PAGE_FAULT_END,
+ KFD_EVENT_FMT_PAGEFAULT_END(ktime_get_boottime_ns(),
+ pid, address, node->id, migration ? 'M' : 'U'));
}
-void kfd_smi_event_migration_start(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_migration_start(struct kfd_node *node, pid_t pid,
unsigned long start, unsigned long end,
uint32_t from, uint32_t to,
uint32_t prefetch_loc, uint32_t preferred_loc,
uint32_t trigger)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_MIGRATE_START,
- "%lld -%d @%lx(%lx) %x->%x %x:%x %d\n",
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_MIGRATE_START,
+ KFD_EVENT_FMT_MIGRATE_START(
ktime_get_boottime_ns(), pid, start, end - start,
- from, to, prefetch_loc, preferred_loc, trigger);
+ from, to, prefetch_loc, preferred_loc, trigger));
}
-void kfd_smi_event_migration_end(struct kfd_dev *dev, 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, dev, KFD_SMI_EVENT_MIGRATE_END,
- "%lld -%d @%lx(%lx) %x->%x %d\n",
+ 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_dev *dev, pid_t pid,
+void kfd_smi_event_queue_eviction(struct kfd_node *node, pid_t pid,
uint32_t trigger)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_QUEUE_EVICTION,
- "%lld -%d %x %d\n", ktime_get_boottime_ns(), pid,
- dev->id, trigger);
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_QUEUE_EVICTION,
+ KFD_EVENT_FMT_QUEUE_EVICTION(ktime_get_boottime_ns(), pid,
+ node->id, trigger));
}
-void kfd_smi_event_queue_restore(struct kfd_dev *dev, pid_t pid)
+void kfd_smi_event_queue_restore(struct kfd_node *node, pid_t pid)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_QUEUE_RESTORE,
- "%lld -%d %x\n", ktime_get_boottime_ns(), pid,
- dev->id);
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_QUEUE_RESTORE,
+ KFD_EVENT_FMT_QUEUE_RESTORE(ktime_get_boottime_ns(), pid,
+ node->id, 0));
}
void kfd_smi_event_queue_restore_rescheduled(struct mm_struct *mm)
@@ -318,22 +329,43 @@ void kfd_smi_event_queue_restore_rescheduled(struct mm_struct *mm)
kfd_smi_event_add(p->lead_thread->pid, pdd->dev,
KFD_SMI_EVENT_QUEUE_RESTORE,
- "%lld -%d %x %c\n", ktime_get_boottime_ns(),
- p->lead_thread->pid, pdd->dev->id, 'R');
+ KFD_EVENT_FMT_QUEUE_RESTORE(ktime_get_boottime_ns(),
+ p->lead_thread->pid, pdd->dev->id, 'R'));
}
kfd_unref_process(p);
}
-void kfd_smi_event_unmap_from_gpu(struct kfd_dev *dev, pid_t pid,
+void kfd_smi_event_unmap_from_gpu(struct kfd_node *node, pid_t pid,
unsigned long address, unsigned long last,
uint32_t trigger)
{
- kfd_smi_event_add(pid, dev, KFD_SMI_EVENT_UNMAP_FROM_GPU,
- "%lld -%d @%lx(%lx) %x %d\n", ktime_get_boottime_ns(),
- pid, address, last - address + 1, dev->id, trigger);
+ kfd_smi_event_add(pid, node, KFD_SMI_EVENT_UNMAP_FROM_GPU,
+ KFD_EVENT_FMT_UNMAP_FROM_GPU(ktime_get_boottime_ns(),
+ pid, address, last - address + 1, node->id, trigger));
+}
+
+void kfd_smi_event_process(struct kfd_process_device *pdd, bool start)
+{
+ struct amdgpu_task_info *task_info;
+ struct amdgpu_vm *avm;
+
+ if (!pdd->drm_priv)
+ return;
+
+ avm = drm_priv_to_vm(pdd->drm_priv);
+ task_info = amdgpu_vm_get_task_info_vm(avm);
+
+ if (task_info) {
+ kfd_smi_event_add(0, pdd->dev,
+ start ? KFD_SMI_EVENT_PROCESS_START :
+ KFD_SMI_EVENT_PROCESS_END,
+ KFD_EVENT_FMT_PROCESS(task_info->task.pid,
+ task_info->task.comm));
+ amdgpu_vm_put_task_info(task_info);
+ }
}
-int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)
+int kfd_smi_event_open(struct kfd_node *dev, uint32_t *fd)
{
struct kfd_smi_client *client;
int ret;
@@ -343,7 +375,7 @@ int kfd_smi_event_open(struct kfd_dev *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;
generated by cgit (git 2.34.1) at 2025-12-26 21:10:34 +0000