drm/amdkfd: Simplify eviction state logic

Always mark evicted queues with q->properties.is_evicted = true, even
queues that are inactive for other reason. This simplifies maintaining
the eviction state as it doesn't require updating is_evicted when other
queue activation conditions change.

On the other hand, we now need to check those other queue activation
conditions whenever an evicted queues is restored. To minimize code
duplication, move the queue activation check into a macro so it can be
maintained in one central place.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Philip Cox <Philip.Cox@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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 ece35c7a77b5ec9126d83c8eb3115a46784621fb..c18355d4cb956208b6b250e308383b07e99c232d 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -289,13 +289,11 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm,
        }
        q->properties.vmid = qpd->vmid;
        /*
-        * Eviction state logic: we only mark active queues as evicted
-        * to avoid the overhead of restoring inactive queues later
+        * Eviction state logic: mark all queues as evicted, even ones
+        * not currently active. Restoring inactive queues later only
+        * updates the is_evicted flag but is a no-op otherwise.
         */
-       if (qpd->evicted)
-               q->properties.is_evicted = (q->properties.queue_size > 0 &&
-                                           q->properties.queue_percent > 0 &&
-                                           q->properties.queue_address != 0);
+       q->properties.is_evicted = !!qpd->evicted;
 
        q->properties.tba_addr = qpd->tba_addr;
        q->properties.tma_addr = qpd->tma_addr;
@@ -518,14 +516,6 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q)
        }
        mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
                        q->properties.type)];
-       /*
-        * Eviction state logic: we only mark active queues as evicted
-        * to avoid the overhead of restoring inactive queues later
-        */
-       if (pdd->qpd.evicted)
-               q->properties.is_evicted = (q->properties.queue_size > 0 &&
-                                           q->properties.queue_percent > 0 &&
-                                           q->properties.queue_address != 0);
 
        /* Save previous activity state for counters */
        prev_active = q->properties.is_active;
@@ -590,7 +580,7 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
        struct queue *q;
        struct mqd_manager *mqd_mgr;
        struct kfd_process_device *pdd;
-       int retval = 0;
+       int retval, ret = 0;
 
        dqm_lock(dqm);
        if (qpd->evicted++ > 0) /* already evicted, do nothing */
@@ -600,25 +590,31 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
        pr_info_ratelimited("Evicting PASID %u queues\n",
                            pdd->process->pasid);
 
-       /* unactivate all active queues on the qpd */
+       /* Mark all queues as evicted. Deactivate all active queues on
+        * the qpd.
+        */
        list_for_each_entry(q, &qpd->queues_list, list) {
+               q->properties.is_evicted = true;
                if (!q->properties.is_active)
                        continue;
+
                mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
                                q->properties.type)];
-               q->properties.is_evicted = true;
                q->properties.is_active = false;
                retval = mqd_mgr->destroy_mqd(mqd_mgr, q->mqd,
                                KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN,
                                KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);
-               if (retval)
-                       goto out;
+               if (retval && !ret)
+                       /* Return the first error, but keep going to
+                        * maintain a consistent eviction state
+                        */
+                       ret = retval;
                dqm->queue_count--;
        }
 
 out:
        dqm_unlock(dqm);
-       return retval;
+       return ret;
 }
 
 static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
@@ -636,11 +632,14 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
        pr_info_ratelimited("Evicting PASID %u queues\n",
                            pdd->process->pasid);
 
-       /* unactivate all active queues on the qpd */
+       /* Mark all queues as evicted. Deactivate all active queues on
+        * the qpd.
+        */
        list_for_each_entry(q, &qpd->queues_list, list) {
+               q->properties.is_evicted = true;
                if (!q->properties.is_active)
                        continue;
-               q->properties.is_evicted = true;
+
                q->properties.is_active = false;
                dqm->queue_count--;
        }
@@ -662,7 +661,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
        struct mqd_manager *mqd_mgr;
        struct kfd_process_device *pdd;
        uint64_t pd_base;
-       int retval = 0;
+       int retval, ret = 0;
 
        pdd = qpd_to_pdd(qpd);
        /* Retrieve PD base */
@@ -696,22 +695,28 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
         */
        mm = get_task_mm(pdd->process->lead_thread);
        if (!mm) {
-               retval = -EFAULT;
+               ret = -EFAULT;
                goto out;
        }
 
-       /* activate all active queues on the qpd */
+       /* Remove the eviction flags. Activate queues that are not
+        * inactive for other reasons.
+        */
        list_for_each_entry(q, &qpd->queues_list, list) {
-               if (!q->properties.is_evicted)
+               q->properties.is_evicted = false;
+               if (!QUEUE_IS_ACTIVE(q->properties))
                        continue;
+
                mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
                                q->properties.type)];
-               q->properties.is_evicted = false;
                q->properties.is_active = true;
                retval = mqd_mgr->load_mqd(mqd_mgr, q->mqd, q->pipe,
                                       q->queue, &q->properties, mm);
-               if (retval)
-                       goto out;
+               if (retval && !ret)
+                       /* Return the first error, but keep going to
+                        * maintain a consistent eviction state
+                        */
+                       ret = retval;
                dqm->queue_count++;
        }
        qpd->evicted = 0;
@@ -719,7 +724,7 @@ out:
        if (mm)
                mmput(mm);
        dqm_unlock(dqm);
-       return retval;
+       return ret;
 }
 
 static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
@@ -751,16 +756,16 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
 
        /* activate all active queues on the qpd */
        list_for_each_entry(q, &qpd->queues_list, list) {
-               if (!q->properties.is_evicted)
-                       continue;
                q->properties.is_evicted = false;
+               if (!QUEUE_IS_ACTIVE(q->properties))
+                       continue;
+
                q->properties.is_active = true;
                dqm->queue_count++;
        }
        retval = execute_queues_cpsch(dqm,
                                KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
-       if (!retval)
-               qpd->evicted = 0;
+       qpd->evicted = 0;
 out:
        dqm_unlock(dqm);
        return retval;
@@ -1199,13 +1204,12 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
        mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
                        q->properties.type)];
        /*
-        * Eviction state logic: we only mark active queues as evicted
-        * to avoid the overhead of restoring inactive queues later
+        * Eviction state logic: mark all queues as evicted, even ones
+        * not currently active. Restoring inactive queues later only
+        * updates the is_evicted flag but is a no-op otherwise.
         */
-       if (qpd->evicted)
-               q->properties.is_evicted = (q->properties.queue_size > 0 &&
-                                           q->properties.queue_percent > 0 &&
-                                           q->properties.queue_address != 0);
+       q->properties.is_evicted = !!qpd->evicted;
+
        dqm->asic_ops.init_sdma_vm(dqm, q, qpd);
        q->properties.tba_addr = qpd->tba_addr;
        q->properties.tma_addr = qpd->tma_addr;

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
index 6e8509ec29d9e0ebae9af491842768abafa390ec..5370a897526ad03c9770697248f6895547a4b6b1 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c
@@ -231,10 +231,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd,
 
        update_cu_mask(mm, mqd, q);
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }
@@ -275,10 +272,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd,
        m->sdma_engine_id = q->sdma_engine_id;
        m->sdma_queue_id = q->sdma_queue_id;
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }
@@ -358,10 +352,7 @@ static int update_mqd_hiq(struct mqd_manager *mm, void *mqd,
 
        m->cp_hqd_vmid = q->vmid;
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index 4750338199b6d8911284bc5bba8930fd866582d1..a65efca362177ef3b4f69a61744ed1eb7ba19c4a 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -247,10 +247,7 @@ static int update_mqd(struct mqd_manager *mm, void *mqd,
 
        update_cu_mask(mm, mqd, q);
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }
@@ -402,10 +399,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd,
        m->sdma_queue_id = q->sdma_queue_id;
        m->sdmax_rlcx_dummy_reg = SDMA_RLC_DUMMY_DEFAULT;
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
index b550dea9b10a93964a87a48e53287b8a2f79aa7e..335aded8855d1495659f00116d389a1d719b2734 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c
@@ -238,10 +238,7 @@ static int __update_mqd(struct mqd_manager *mm, void *mqd,
 
        update_cu_mask(mm, mqd, q);
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }
@@ -396,10 +393,7 @@ static int update_mqd_sdma(struct mqd_manager *mm, void *mqd,
        m->sdma_engine_id = q->sdma_engine_id;
        m->sdma_queue_id = q->sdma_queue_id;
 
-       q->is_active = (q->queue_size > 0 &&
-                       q->queue_address != 0 &&
-                       q->queue_percent > 0 &&
-                       !q->is_evicted);
+       q->is_active = QUEUE_IS_ACTIVE(*q);
 
        return 0;
 }

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index b61dc53f42d2475aa04be2155f4113d5aaa25284..6d29d589a4a5cf9c905f51e90c6f7d6eb14de18b 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -430,6 +430,11 @@ struct queue_properties {
        uint32_t *cu_mask;
 };
 
+#define QUEUE_IS_ACTIVE(q) ((q).queue_size > 0 &&      \
+                           (q).queue_address != 0 &&   \
+                           (q).queue_percent > 0 &&    \
+                           !(q).is_evicted)
+
 /**
  * struct queue
  *