drm/amdgpu: use doorbell mgr for kfd process doorbells

This patch:
- adds a doorbell object in kfd pdd structure.
- allocates doorbells for a process while creating its queue.
- frees the doorbells with pdd destroy.
- moves doorbell bitmap init function to kfd_doorbell.c

PS: This patch ensures that we don't break the existing KFD
    functionality, but now KFD userspace library should also
    create doorbell pages as AMDGPU GEM objects using libdrm
    functions in userspace. The reference code for the same
    is available with AMDGPU Usermode queue libdrm MR. Once
    this is done, we will not need to create process doorbells
    in kernel.

V2: - Do not use doorbell wrapper API, use amdgpu_bo_create_kernel
      instead (Alex).
    - Do not use custom doorbell structure, instead use separate
      variables for bo and doorbell_bitmap (Alex)
V3:
   - Do not allocate doorbell page with PDD, delay doorbell process
     page allocation until really needed (Felix)

Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felilx.Kuehling@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 10 insertions, 13 deletions

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 60e6b37b43ba..adb5e4bdc0b2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -368,17 +368,20 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 		goto err_create_queue;
 	}
 
-	if (q && p_doorbell_offset_in_process)
+	if (q && p_doorbell_offset_in_process) {
 		/* Return the doorbell offset within the doorbell page
 		 * to the caller so it can be passed up to user mode
 		 * (in bytes).
-		 * There are always 1024 doorbells per process, so in case
-		 * of 8-byte doorbells, there are two doorbell pages per
-		 * process.
+		 * relative doorbell index = Absolute doorbell index -
+		 * absolute index of first doorbell in the page.
 		 */
-		*p_doorbell_offset_in_process =
-			(q->properties.doorbell_off * sizeof(uint32_t)) &
-			(kfd_doorbell_process_slice(dev->kfd) - 1);
+		uint32_t first_db_index = amdgpu_doorbell_index_on_bar(pdd->dev->adev,
+								       pdd->qpd.proc_doorbells,
+								       0);
+
+		*p_doorbell_offset_in_process = (q->properties.doorbell_off
+						- first_db_index) * sizeof(uint32_t);
+	}
 
 	pr_debug("PQM After DQM create queue\n");
 
@@ -933,12 +936,6 @@ int kfd_criu_restore_queue(struct kfd_process *p,
 		goto exit;
 	}
 
-	if (!pdd->doorbell_index &&
-	    kfd_alloc_process_doorbells(pdd->dev->kfd, &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;