[PATCH] drm/amdkfd: allow compute partition mode switch with cgroup exclusions
Jonathan Kim
jonathan.kim at amd.com
Thu May 29 19:09:45 UTC 2025
The KFD currently bars a compute partition mode switch while a KFD
process exists.
Since cgroup excluded devices remain excluded for the lifetime of a KFD
process and user space is able to mode switch single devices, allow
users to mode switch a device with any running process that has been
cgroup excluded from this device.
Signed-off-by: Jonathan Kim <jonathan.kim at amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 4 +-
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 8 +--
drivers/gpu/drm/amd/amdkfd/kfd_device.c | 69 ++++++++++++++++++----
drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 5 +-
drivers/gpu/drm/amd/amdkfd/kfd_process.c | 2 +-
5 files changed, 70 insertions(+), 18 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 4cec3a873995..7cfe4677bf56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -746,12 +746,12 @@ int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev)
{
- return kgd2kfd_check_and_lock_kfd();
+ return kgd2kfd_check_and_lock_kfd(adev->kfd.dev);
}
void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
{
- kgd2kfd_unlock_kfd();
+ kgd2kfd_unlock_kfd(adev->kfd.dev);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index b6ca41859b53..3d5812269ea0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -419,8 +419,8 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd);
void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry);
void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
-int kgd2kfd_check_and_lock_kfd(void);
-void kgd2kfd_unlock_kfd(void);
+int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd);
+void kgd2kfd_unlock_kfd(struct kfd_dev *kfd);
int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id);
int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id);
bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id);
@@ -489,12 +489,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
{
}
-static inline int kgd2kfd_check_and_lock_kfd(void)
+static inline int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd)
{
return 0;
}
-static inline void kgd2kfd_unlock_kfd(void)
+static inline void kgd2kfd_unlock_kfd(struct kfd_dev *kfd)
{
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index bf0854bd5555..a12e1433943d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -1013,10 +1013,30 @@ int kgd2kfd_post_reset(struct kfd_dev *kfd)
return 0;
}
-bool kfd_is_locked(void)
+bool kfd_is_locked(struct kfd_dev *kfd)
{
+ uint8_t id = 0;
+ struct kfd_node *dev;
+
lockdep_assert_held(&kfd_processes_mutex);
- return (kfd_locked > 0);
+
+ /* check reset/suspend lock */
+ if (kfd_locked > 0)
+ return true;
+
+ if (kfd)
+ return kfd->kfd_dev_lock > 0;
+
+ /* check lock on all cgroup accessible devices */
+ while (kfd_topology_enum_kfd_devices(id++, &dev) == 0) {
+ if (!dev || kfd_devcgroup_check_permission(dev))
+ continue;
+
+ if (dev->kfd->kfd_dev_lock > 0)
+ return true;
+ }
+
+ return false;
}
void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
@@ -1442,24 +1462,53 @@ unsigned int kfd_get_num_xgmi_sdma_engines(struct kfd_node *node)
kfd_get_num_sdma_engines(node);
}
-int kgd2kfd_check_and_lock_kfd(void)
+int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd)
{
+ struct kfd_process *p;
+ int r = 0, temp, idx;
+
mutex_lock(&kfd_processes_mutex);
- if (!hash_empty(kfd_processes_table) || kfd_is_locked()) {
- mutex_unlock(&kfd_processes_mutex);
- return -EBUSY;
+
+ if (hash_empty(kfd_processes_table) && !kfd_is_locked(kfd))
+ goto out;
+
+ /* fail under system reset/resume or kfd device is partition switching. */
+ if (kfd_is_locked(kfd)) {
+ r = -EBUSY;
+ goto out;
+ }
+
+ /*
+ * ensure all running processes are cgroup excluded from device before mode switch.
+ * i.e. no pdd was created on the process socket.
+ */
+ idx = srcu_read_lock(&kfd_processes_srcu);
+ hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) {
+ int i;
+
+ for (i = 0; i < p->n_pdds; i++) {
+ if (p->pdds[i]->dev->kfd != kfd)
+ continue;
+
+ r = -EBUSY;
+ goto proc_check_unlock;
+ }
}
- ++kfd_locked;
+proc_check_unlock:
+ srcu_read_unlock(&kfd_processes_srcu, idx);
+out:
+ if (!r)
+ ++kfd->kfd_dev_lock;
mutex_unlock(&kfd_processes_mutex);
- return 0;
+ return r;
}
-void kgd2kfd_unlock_kfd(void)
+void kgd2kfd_unlock_kfd(struct kfd_dev *kfd)
{
mutex_lock(&kfd_processes_mutex);
- --kfd_locked;
+ --kfd->kfd_dev_lock;
mutex_unlock(&kfd_processes_mutex);
}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index d221c58dccc3..67694bcd9464 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -372,6 +372,9 @@ struct kfd_dev {
/* bitmap for dynamic doorbell allocation from doorbell object */
unsigned long *doorbell_bitmap;
+
+ /* for dynamic partitioning */
+ int kfd_dev_lock;
};
enum kfd_mempool {
@@ -1536,7 +1539,7 @@ static inline bool kfd_flush_tlb_after_unmap(struct kfd_dev *dev)
int kfd_send_exception_to_runtime(struct kfd_process *p,
unsigned int queue_id,
uint64_t error_reason);
-bool kfd_is_locked(void);
+bool kfd_is_locked(struct kfd_dev *kfd);
/* Compute profile */
void kfd_inc_compute_active(struct kfd_node *dev);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 722ac1662bdc..5be28c6c4f6a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -854,7 +854,7 @@ struct kfd_process *kfd_create_process(struct task_struct *thread)
*/
mutex_lock(&kfd_processes_mutex);
- if (kfd_is_locked()) {
+ if (kfd_is_locked(NULL)) {
pr_debug("KFD is locked! Cannot create process");
process = ERR_PTR(-EINVAL);
goto out;
--
2.34.1
More information about the amd-gfx
mailing list