On 2022-09-07 15:43, Rajneesh Bhardwaj wrote:
Recently introduced change to allocate doorbells only when the first
queue is created or mapped for CPU / GPU access, did not consider
Checkpoint Restore scenario completely. This fix allows the CRIU restore
operation by extending the doorbell optimization to CRIU restore
scenario.
Fixes: 'commit 15bcfbc55b57 ("drm/amdkfd: Allocate doorbells only when needed")'
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@xxxxxxx>
Reviewed-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>
Thanks!
---
Changes in v2:
* Addressed review feedback from Felix
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 6 ++++++
drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c | 3 +++
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 7 +++++++
3 files changed, 16 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index 84da1a9ce37c..56f7307c21d2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -2153,6 +2153,12 @@ static int criu_restore_devices(struct kfd_process *p,
ret = PTR_ERR(pdd);
goto exit;
}
+
+ if (!pdd->doorbell_index &&
+ kfd_alloc_process_doorbells(pdd->dev, &pdd->doorbell_index) < 0) {
+ ret = -ENOMEM;
+ goto exit;
+ }
}
/*
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
index b33798f89ef0..cd4e61bf0493 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c
@@ -303,6 +303,9 @@ int kfd_alloc_process_doorbells(struct kfd_dev *kfd, unsigned int *doorbell_inde
if (r > 0)
*doorbell_index = r;
+ if (r < 0)
+ pr_err("Failed to allocate process doorbells\n");
+
return r;
}
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 6e3e7f54381b..5137476ec18e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -857,6 +857,13 @@ int kfd_criu_restore_queue(struct kfd_process *p,
ret = -EINVAL;
goto exit;
}
+
+ if (!pdd->doorbell_index &&
+ kfd_alloc_process_doorbells(pdd->dev, &pdd->doorbell_index) < 0) {
+ ret = -ENOMEM;
+ goto exit;
+ }
+
/* data stored in this order: mqd, ctl_stack */
mqd = q_extra_data;
ctl_stack = mqd + q_data->mqd_size;
