[RFC PATCH 15/17] drm/amdkfd: CRIU dump and restore events
Felix Kuehling
Felix.Kuehling at amd.com
Sat May 1 01:57:50 UTC 2021
From: David Yat Sin <david.yatsin at amd.com>
Add support to existing CRIU ioctl's to save and restore events during
criu checkpoint and restore.
Signed-off-by: David Yat Sin <david.yatsin at amd.com>
Change-Id: I1635b1fa91a81abcbd19290cb88c8ca142c390e0
---
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 185 ++++++++++++++++-------
drivers/gpu/drm/amd/amdkfd/kfd_events.c | 163 ++++++++++++++++----
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 11 +-
3 files changed, 279 insertions(+), 80 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 71f734eae071..ce511b246beb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -997,6 +997,55 @@ static int kfd_ioctl_get_process_apertures_new(struct file *filp,
return 0;
}
+static int kmap_event_page(struct kfd_process *p, uint64_t event_page_offset)
+{
+ struct kfd_dev *kfd;
+ struct kfd_process_device *pdd;
+ void *mem, *kern_addr;
+ uint64_t size;
+ int err = 0;
+
+ if (p->signal_page) {
+ pr_err("Event page is already set\n");
+ return -EINVAL;
+ }
+
+ kfd = kfd_device_by_id(GET_GPU_ID(event_page_offset));
+ if (!kfd) {
+ pr_err("Getting device by id failed in %s\n", __func__);
+ return -EINVAL;
+ }
+
+ pdd = kfd_bind_process_to_device(kfd, p);
+ if (IS_ERR(pdd)) {
+ mutex_unlock(&p->mutex);
+ return PTR_ERR(pdd);
+ }
+
+ mem = kfd_process_device_translate_handle(pdd,
+ GET_IDR_HANDLE(event_page_offset));
+ if (!mem) {
+ pr_err("Can't find BO, offset is 0x%llx\n", event_page_offset);
+
+ mutex_unlock(&p->mutex);
+ return -EINVAL;
+ }
+
+ err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd,
+ mem, &kern_addr, &size);
+ if (err) {
+ pr_err("Failed to map event page to kernel\n");
+ return err;
+ }
+
+ err = kfd_event_page_set(p, kern_addr, size, event_page_offset);
+ if (err) {
+ pr_err("Failed to set event page\n");
+ return err;
+ }
+ return err;
+}
+
static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
void *data)
{
@@ -1008,51 +1057,11 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
* through the event_page_offset field.
*/
if (args->event_page_offset) {
- struct kfd_dev *kfd;
- struct kfd_process_device *pdd;
- void *mem, *kern_addr;
- uint64_t size;
-
- if (p->signal_page) {
- pr_err("Event page is already set\n");
- return -EINVAL;
- }
-
- kfd = kfd_device_by_id(GET_GPU_ID(args->event_page_offset));
- if (!kfd) {
- pr_err("Getting device by id failed in %s\n", __func__);
- return -EINVAL;
- }
-
mutex_lock(&p->mutex);
- pdd = kfd_bind_process_to_device(kfd, p);
- if (IS_ERR(pdd)) {
- err = PTR_ERR(pdd);
- goto out_unlock;
- }
-
- mem = kfd_process_device_translate_handle(pdd,
- GET_IDR_HANDLE(args->event_page_offset));
- if (!mem) {
- pr_err("Can't find BO, offset is 0x%llx\n",
- args->event_page_offset);
- err = -EINVAL;
- goto out_unlock;
- }
+ err = kmap_event_page(p, args->event_page_offset);
mutex_unlock(&p->mutex);
-
- err = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(kfd->kgd,
- mem, &kern_addr, &size);
- if (err) {
- pr_err("Failed to map event page to kernel\n");
- return err;
- }
-
- err = kfd_event_page_set(p, kern_addr, size);
- if (err) {
- pr_err("Failed to set event page\n");
+ if (err)
return err;
- }
}
err = kfd_event_create(filp, p, args->event_type,
@@ -1061,10 +1070,7 @@ static int kfd_ioctl_create_event(struct file *filp, struct kfd_process *p,
&args->event_page_offset,
&args->event_slot_index);
- return err;
-
-out_unlock:
- mutex_unlock(&p->mutex);
+ pr_debug("Created event (id:0x%08x) (%s)\n", args->event_id, __func__);
return err;
}
@@ -2062,6 +2068,7 @@ static int kfd_ioctl_criu_dumper(struct file *filep,
struct kfd_process *p, void *data)
{
struct kfd_ioctl_criu_dumper_args *args = data;
+ struct kfd_criu_ev_bucket *ev_buckets = NULL;
struct kfd_criu_bo_buckets *bo_bucket;
struct amdgpu_bo *dumper_bo;
int ret, id, index, i = 0;
@@ -2088,14 +2095,22 @@ static int kfd_ioctl_criu_dumper(struct file *filep,
return -EINVAL;
}
- pr_debug("num of bos = %llu\n", args->num_of_bos);
+ pr_debug("num of bos = %llu queues = %u events = %u\n", args->num_of_bos, args->num_of_queues, args->num_of_events);
bo_bucket = kvzalloc((sizeof(struct kfd_criu_bo_buckets) *
args->num_of_bos), GFP_KERNEL);
if (!bo_bucket)
return -ENOMEM;
- pr_debug("num of queues = %u\n", args->num_of_queues);
+ if (args->num_of_events) {
+ ev_buckets = kvzalloc((sizeof(struct kfd_criu_ev_bucket) *
+ args->num_of_events), GFP_KERNEL);
+
+ if (!ev_buckets) {
+ ret = -ENOMEM;
+ goto clean;
+ }
+ }
mutex_lock(&p->mutex);
@@ -2182,6 +2197,23 @@ static int kfd_ioctl_criu_dumper(struct file *filep,
goto err_unlock;
}
+ /* Dump events */
+ ret = kfd_event_dump(p, &args->event_page_offset, ev_buckets,
+ args->num_of_events);
+ if (ret) {
+ pr_err("failed to dump events, ret=%d\n", ret);
+ goto err_unlock;
+ }
+ ret = copy_to_user((void __user *)args->kfd_criu_ev_buckets_ptr,
+ ev_buckets,
+ (args->num_of_events *
+ sizeof(struct kfd_criu_ev_bucket)));
+ kvfree(ev_buckets);
+ if (ret) {
+ ret = -EFAULT;
+ goto err_unlock;
+ }
+
ret = copy_to_user((void __user *)args->kfd_criu_bo_buckets_ptr,
bo_bucket,
(args->num_of_bos *
@@ -2195,6 +2227,8 @@ static int kfd_ioctl_criu_dumper(struct file *filep,
err_unlock:
kfd_process_restore_queues(p);
mutex_unlock(&p->mutex);
+clean:
+ kvfree(ev_buckets);
pr_err("Dumper ioctl failed err:%d\n", ret);
return ret;
}
@@ -2388,6 +2422,47 @@ static int criu_restore_queues(struct kfd_process *p,
return ret;
}
+/* criu_restore_queues_events runs with the process mutex locked */
+static int criu_restore_events(struct file *filp, struct kfd_process *p,
+ struct kfd_ioctl_criu_restorer_args *args)
+{
+ int i;
+ struct kfd_criu_ev_bucket *events;
+ int ret = 0;
+
+ if (args->event_page_offset) {
+ ret = kmap_event_page(p, args->event_page_offset);
+ if (ret)
+ return ret;
+ }
+
+ if (!args->num_of_events)
+ return 0;
+
+ events = kvmalloc_array(args->num_of_events,
+ sizeof(struct kfd_criu_ev_bucket),
+ GFP_KERNEL);
+ if (!events)
+ return -ENOMEM;
+
+ ret = copy_from_user(events, (void __user *) args->kfd_criu_ev_buckets_ptr,
+ args->num_of_events * sizeof(struct kfd_criu_ev_bucket));
+
+ if (ret) {
+ ret = -EFAULT;
+ goto exit;
+ }
+
+ for (i = 0; i < args->num_of_events; i++) {
+ ret = kfd_event_restore(filp, p, &events[i]);
+ if (ret)
+ pr_err("Failed to restore event with id (%d)\n", ret);
+ }
+exit:
+ kvfree(events);
+ return ret;
+}
+
static int kfd_ioctl_criu_restorer(struct file *filep,
struct kfd_process *p, void *data)
{
@@ -2618,13 +2693,18 @@ static int kfd_ioctl_criu_restorer(struct file *filep,
goto err_unlock;
}
+ ret = criu_restore_events(filep, p, args);
+ if (ret) {
+ pr_err("Failed to restore events (%d)", ret);
+ err = ret;
+ goto err_unlock;
+ }
+
ret = copy_to_user((void __user *)args->restored_bo_array_ptr,
restored_bo_offsets_arr,
(args->num_of_bos * sizeof(*restored_bo_offsets_arr)));
- if (ret) {
+ if (ret)
err = -EFAULT;
- goto err_unlock;
- }
err_unlock:
mutex_unlock(&p->mutex);
@@ -2739,8 +2819,9 @@ static int kfd_ioctl_criu_helper(struct file *filep,
args->num_of_bos = num_of_bos;
args->num_of_queues = q_index;
args->queues_data_size = queues_data_size;
- dev_dbg(kfd_device, "Num of bos = %llu\n", num_of_bos);
+ args->num_of_events = kfd_get_num_events(p);
+ dev_dbg(kfd_device, "Num of bos = %llu queues:%u events:%u\n", args->num_of_bos, args->num_of_queues, args->num_of_events);
err_unlock:
mutex_unlock(&p->mutex);
return ret;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
index ba2c2ce0c55a..c8d5458ba2c0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c
@@ -53,9 +53,9 @@ struct kfd_signal_page {
uint64_t *kernel_address;
uint64_t __user *user_address;
bool need_to_free_pages;
+ uint64_t user_handle; /* Needed for CRIU dumped and restore */
};
-
static uint64_t *page_slots(struct kfd_signal_page *page)
{
return page->kernel_address;
@@ -92,7 +92,8 @@ static struct kfd_signal_page *allocate_signal_page(struct kfd_process *p)
}
static int allocate_event_notification_slot(struct kfd_process *p,
- struct kfd_event *ev)
+ struct kfd_event *ev,
+ const int *restore_id)
{
int id;
@@ -104,14 +105,19 @@ static int allocate_event_notification_slot(struct kfd_process *p,
p->signal_mapped_size = 256*8;
}
- /*
- * Compatibility with old user mode: Only use signal slots
- * user mode has mapped, may be less than
- * KFD_SIGNAL_EVENT_LIMIT. This also allows future increase
- * of the event limit without breaking user mode.
- */
- id = idr_alloc(&p->event_idr, ev, 0, p->signal_mapped_size / 8,
- GFP_KERNEL);
+ if (restore_id) {
+ id = idr_alloc(&p->event_idr, ev, *restore_id, *restore_id + 1,
+ GFP_KERNEL);
+ } else {
+ /*
+ * Compatibility with old user mode: Only use signal slots
+ * user mode has mapped, may be less than
+ * KFD_SIGNAL_EVENT_LIMIT. This also allows future increase
+ * of the event limit without breaking user mode.
+ */
+ id = idr_alloc(&p->event_idr, ev, 0, p->signal_mapped_size / 8,
+ GFP_KERNEL);
+ }
if (id < 0)
return id;
@@ -178,9 +184,8 @@ static struct kfd_event *lookup_signaled_event_by_partial_id(
return ev;
}
-static int create_signal_event(struct file *devkfd,
- struct kfd_process *p,
- struct kfd_event *ev)
+static int create_signal_event(struct file *devkfd, struct kfd_process *p,
+ struct kfd_event *ev, const int *restore_id)
{
int ret;
@@ -193,7 +198,7 @@ static int create_signal_event(struct file *devkfd,
return -ENOSPC;
}
- ret = allocate_event_notification_slot(p, ev);
+ ret = allocate_event_notification_slot(p, ev, restore_id);
if (ret) {
pr_warn("Signal event wasn't created because out of kernel memory\n");
return ret;
@@ -209,16 +214,21 @@ static int create_signal_event(struct file *devkfd,
return 0;
}
-static int create_other_event(struct kfd_process *p, struct kfd_event *ev)
+static int create_other_event(struct kfd_process *p, struct kfd_event *ev, const int *restore_id)
{
- /* Cast KFD_LAST_NONSIGNAL_EVENT to uint32_t. This allows an
- * intentional integer overflow to -1 without a compiler
- * warning. idr_alloc treats a negative value as "maximum
- * signed integer".
- */
- int id = idr_alloc(&p->event_idr, ev, KFD_FIRST_NONSIGNAL_EVENT_ID,
- (uint32_t)KFD_LAST_NONSIGNAL_EVENT_ID + 1,
- GFP_KERNEL);
+ int id;
+ if (restore_id)
+ id = idr_alloc(&p->event_idr, ev, *restore_id, *restore_id + 1,
+ GFP_KERNEL);
+ else
+ /* Cast KFD_LAST_NONSIGNAL_EVENT to uint32_t. This allows an
+ * intentional integer overflow to -1 without a compiler
+ * warning. idr_alloc treats a negative value as "maximum
+ * signed integer".
+ */
+ id = idr_alloc(&p->event_idr, ev, KFD_FIRST_NONSIGNAL_EVENT_ID,
+ (uint32_t)KFD_LAST_NONSIGNAL_EVENT_ID + 1,
+ GFP_KERNEL);
if (id < 0)
return id;
@@ -296,7 +306,7 @@ static bool event_can_be_cpu_signaled(const struct kfd_event *ev)
}
int kfd_event_page_set(struct kfd_process *p, void *kernel_address,
- uint64_t size)
+ uint64_t size, uint64_t user_handle)
{
struct kfd_signal_page *page;
@@ -315,7 +325,7 @@ int kfd_event_page_set(struct kfd_process *p, void *kernel_address,
p->signal_page = page;
p->signal_mapped_size = size;
-
+ p->signal_page->user_handle = user_handle;
return 0;
}
@@ -343,14 +353,14 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p,
switch (event_type) {
case KFD_EVENT_TYPE_SIGNAL:
case KFD_EVENT_TYPE_DEBUG:
- ret = create_signal_event(devkfd, p, ev);
+ ret = create_signal_event(devkfd, p, ev, NULL);
if (!ret) {
*event_page_offset = KFD_MMAP_TYPE_EVENTS;
*event_slot_index = ev->event_id;
}
break;
default:
- ret = create_other_event(p, ev);
+ ret = create_other_event(p, ev, NULL);
break;
}
@@ -366,6 +376,105 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p,
return ret;
}
+int kfd_event_restore(struct file *devkfd, struct kfd_process *p,
+ struct kfd_criu_ev_bucket *restore_ev)
+{
+ int ret = 0;
+ struct kfd_event *ev = kzalloc(sizeof(*ev), GFP_KERNEL);
+
+ if (!ev)
+ return -ENOMEM;
+
+ ev->type = restore_ev->type;
+ ev->auto_reset = restore_ev->auto_reset;
+ ev->signaled = restore_ev->signaled;
+
+ init_waitqueue_head(&ev->wq);
+
+ mutex_lock(&p->event_mutex);
+ switch (ev->type) {
+ case KFD_EVENT_TYPE_SIGNAL:
+ case KFD_EVENT_TYPE_DEBUG:
+ ret = create_signal_event(devkfd, p, ev, &restore_ev->event_id);
+ break;
+ case KFD_EVENT_TYPE_MEMORY:
+ memcpy(&ev->memory_exception_data,
+ &restore_ev->memory_exception_data,
+ sizeof(struct kfd_hsa_memory_exception_data));
+
+ ret = create_other_event(p, ev, &restore_ev->event_id);
+ break;
+ case KFD_EVENT_TYPE_HW_EXCEPTION:
+ memcpy(&ev->hw_exception_data,
+ &restore_ev->hw_exception_data,
+ sizeof(struct kfd_hsa_hw_exception_data));
+
+ ret = create_other_event(p, ev, &restore_ev->event_id);
+ break;
+ }
+
+ if (ret)
+ kfree(ev);
+
+ mutex_unlock(&p->event_mutex);
+
+ return ret;
+}
+
+int kfd_event_dump(struct kfd_process *p, uint64_t *user_handle,
+ struct kfd_criu_ev_bucket *ev_buckets,
+ uint32_t num_events)
+{
+ struct kfd_event *ev;
+ uint32_t ev_id;
+ int i = 0;
+
+ *user_handle = 0;
+
+ if (p->signal_page)
+ *user_handle = p->signal_page->user_handle;
+
+ idr_for_each_entry(&p->event_idr, ev, ev_id) {
+ if (i >= num_events) {
+ pr_err("Number of events exceeds number allocated\n");
+ return -ENOMEM;
+ }
+
+ ev_buckets[i].event_id = ev->event_id;
+ ev_buckets[i].auto_reset = ev->auto_reset;
+ ev_buckets[i].type = ev->type;
+ ev_buckets[i].signaled = ev->signaled;
+
+ if (ev_buckets[i].type == KFD_EVENT_TYPE_MEMORY) {
+ memcpy(&ev_buckets[i].memory_exception_data,
+ &ev->memory_exception_data,
+ sizeof(struct kfd_hsa_memory_exception_data));
+
+ } else if (ev_buckets[i].type == KFD_EVENT_TYPE_HW_EXCEPTION) {
+ memcpy(&ev_buckets[i].hw_exception_data,
+ &ev->hw_exception_data,
+ sizeof(struct kfd_hsa_hw_exception_data));
+ }
+ pr_debug("Dumped event[%d] id = 0x%08x auto_reset = %x type = %x signaled = %x\n",
+ i, ev_buckets[i].event_id, ev_buckets[i].auto_reset,
+ ev_buckets[i].type, ev_buckets[i].signaled);
+ i++;
+ }
+ return 0;
+}
+
+int kfd_get_num_events(struct kfd_process *p)
+{
+ struct kfd_event *ev;
+ uint32_t id;
+ u32 num_events = 0;
+
+ idr_for_each_entry(&p->event_idr, ev, id)
+ num_events++;
+
+ return num_events++;
+}
+
/* Assumes that p is current. */
int kfd_event_destroy(struct kfd_process *p, uint32_t event_id)
{
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 3ebca0a4e046..d3265860e78b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -1190,11 +1190,20 @@ void kfd_signal_hw_exception_event(u32 pasid);
int kfd_set_event(struct kfd_process *p, uint32_t event_id);
int kfd_reset_event(struct kfd_process *p, uint32_t event_id);
int kfd_event_page_set(struct kfd_process *p, void *kernel_address,
- uint64_t size);
+ uint64_t size, uint64_t user_handle);
+
int kfd_event_create(struct file *devkfd, struct kfd_process *p,
uint32_t event_type, bool auto_reset, uint32_t node_id,
uint32_t *event_id, uint32_t *event_trigger_data,
uint64_t *event_page_offset, uint32_t *event_slot_index);
+
+int kfd_event_restore(struct file *devkfd, struct kfd_process *p,
+ struct kfd_criu_ev_bucket *restore_ev);
+
+int kfd_event_dump(struct kfd_process *p, uint64_t *user_handle,
+ struct kfd_criu_ev_bucket *ev_buckets, uint32_t num_events);
+
+int kfd_get_num_events(struct kfd_process *p);
int kfd_event_destroy(struct kfd_process *p, uint32_t event_id);
void kfd_signal_vm_fault_event(struct kfd_dev *dev, u32 pasid,
--
2.17.1
More information about the amd-gfx
mailing list