+
+ int (*unmap_queues_cpsch_poison)(struct device_queue_manager *dqm,
+ uint16_t pasid);
};
Might be better call it reset_queue directly (match with update_queue, create_queue, .etc.,)
Others look good to me
The series (4 patches) is
Reviewed-by: Hawking Zhang <Hawking.Zhang@xxxxxxx>
Regards,
Hawking
-----Original Message-----
From: Zhou1, Tao <Tao.Zhou1@xxxxxxx>
Sent: Thursday, December 16, 2021 19:36
To: amd-gfx@xxxxxxxxxxxxxxxxxxxxx; Zhang, Hawking <Hawking.Zhang@xxxxxxx>; Yang, Stanley <Stanley.Yang@xxxxxxx>; Chai, Thomas <YiPeng.Chai@xxxxxxx>; Kuehling, Felix <Felix.Kuehling@xxxxxxx>
Cc: Zhou1, Tao <Tao.Zhou1@xxxxxxx>
Subject: [PATCH 3/4] drm/amdkfd: add reset queue function for RAS poison
The new interface unmaps queues with reset mode for the process consumes RAS poison, it's only for compute queue.
Signed-off-by: Tao Zhou <tao.zhou1@xxxxxxx>
---
.../drm/amd/amdkfd/kfd_device_queue_manager.c | 16 ++++++++++++++++
.../drm/amd/amdkfd/kfd_device_queue_manager.h | 5 +++++
2 files changed, 21 insertions(+)
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 01a2cc3928ac..b4b0495c7024 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1476,6 +1476,21 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
return retval;
}
+/* only for compute queue */
+static int unmap_queues_cpsch_poison(struct device_queue_manager *dqm,
+ uint16_t pasid)
+{
+ int retval;
+
+ dqm_lock(dqm);
+
+ retval = unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_BY_PASID,
+ pasid, true);
+
+ dqm_unlock(dqm);
+ return retval;
+}
+
/* dqm->lock mutex has to be locked before calling this function */ static int execute_queues_cpsch(struct device_queue_manager *dqm,
enum kfd_unmap_queues_filter filter, @@ -1896,6 +1911,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
dqm->ops.evict_process_queues = evict_process_queues_cpsch;
dqm->ops.restore_process_queues = restore_process_queues_cpsch;
dqm->ops.get_wave_state = get_wave_state;
+ dqm->ops.unmap_queues_cpsch_poison = unmap_queues_cpsch_poison;
break;
case KFD_SCHED_POLICY_NO_HWS:
/* initialize dqm for no cp scheduling */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index 499fc0ea387f..19ec3e8859e8 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -81,6 +81,8 @@ struct device_process_node {
*
* @get_wave_state: Retrieves context save state and optionally copies the
* control stack, if kept in the MQD, to the given userspace address.
+ *
+ * @unmap_queues_cpsch_poison: reset queue which consumes RAS poison
*/
struct device_queue_manager_ops {
@@ -134,6 +136,9 @@ struct device_queue_manager_ops {
void __user *ctl_stack,
u32 *ctl_stack_used_size,
u32 *save_area_used_size);
+
+ int (*unmap_queues_cpsch_poison)(struct device_queue_manager *dqm,
+ uint16_t pasid);
};
struct device_queue_manager_asic_ops {
--
2.17.1
