Skip to content

Commit 7c6b6e1

Browse files
dayatsin-amdalexdeucher
dayatsin-amd
authored and
alexdeucher
committed
drm/amdkfd: Fix GWS queue count
dqm->gws_queue_count and pdd->qpd.mapped_gws_queue need to be updated each time the queue gets evicted. Fixes: b8020b0 ("drm/amdkfd: Enable over-subscription with >1 GWS queue") Signed-off-by: David Yat Sin <david.yatsin@amd.com> Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
1 parent b2d229d commit 7c6b6e1

1 file changed

Lines changed: 37 additions & 46 deletions

File tree

‎drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c‎

Lines changed: 37 additions & 46 deletions
Original file line numberDiff line numberDiff line change
@@ -130,19 +130,33 @@ void program_sh_mem_settings(struct device_queue_manager *dqm,
130130
}
131131

132132
static void increment_queue_count(struct device_queue_manager *dqm,
133-
enum kfd_queue_type type)
133+
struct qcm_process_device *qpd,
134+
struct queue *q)
134135
{
135136
dqm->active_queue_count++;
136-
if (type == KFD_QUEUE_TYPE_COMPUTE || type == KFD_QUEUE_TYPE_DIQ)
137+
if (q->properties.type == KFD_QUEUE_TYPE_COMPUTE ||
138+
q->properties.type == KFD_QUEUE_TYPE_DIQ)
137139
dqm->active_cp_queue_count++;
140+
141+
if (q->properties.is_gws) {
142+
dqm->gws_queue_count++;
143+
qpd->mapped_gws_queue = true;
144+
}
138145
}
139146

140147
static void decrement_queue_count(struct device_queue_manager *dqm,
141-
enum kfd_queue_type type)
148+
struct qcm_process_device *qpd,
149+
struct queue *q)
142150
{
143151
dqm->active_queue_count--;
144-
if (type == KFD_QUEUE_TYPE_COMPUTE || type == KFD_QUEUE_TYPE_DIQ)
152+
if (q->properties.type == KFD_QUEUE_TYPE_COMPUTE ||
153+
q->properties.type == KFD_QUEUE_TYPE_DIQ)
145154
dqm->active_cp_queue_count--;
155+
156+
if (q->properties.is_gws) {
157+
dqm->gws_queue_count--;
158+
qpd->mapped_gws_queue = false;
159+
}
146160
}
147161

148162
/*
@@ -412,7 +426,7 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm,
412426
list_add(&q->list, &qpd->queues_list);
413427
qpd->queue_count++;
414428
if (q->properties.is_active)
415-
increment_queue_count(dqm, q->properties.type);
429+
increment_queue_count(dqm, qpd, q);
416430

417431
/*
418432
* Unconditionally increment this counter, regardless of the queue's
@@ -601,13 +615,8 @@ static int destroy_queue_nocpsch_locked(struct device_queue_manager *dqm,
601615
deallocate_vmid(dqm, qpd, q);
602616
}
603617
qpd->queue_count--;
604-
if (q->properties.is_active) {
605-
decrement_queue_count(dqm, q->properties.type);
606-
if (q->properties.is_gws) {
607-
dqm->gws_queue_count--;
608-
qpd->mapped_gws_queue = false;
609-
}
610-
}
618+
if (q->properties.is_active)
619+
decrement_queue_count(dqm, qpd, q);
611620

612621
return retval;
613622
}
@@ -700,12 +709,11 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q,
700709
* dqm->active_queue_count to determine whether a new runlist must be
701710
* uploaded.
702711
*/
703-
if (q->properties.is_active && !prev_active)
704-
increment_queue_count(dqm, q->properties.type);
705-
else if (!q->properties.is_active && prev_active)
706-
decrement_queue_count(dqm, q->properties.type);
707-
708-
if (q->gws && !q->properties.is_gws) {
712+
if (q->properties.is_active && !prev_active) {
713+
increment_queue_count(dqm, &pdd->qpd, q);
714+
} else if (!q->properties.is_active && prev_active) {
715+
decrement_queue_count(dqm, &pdd->qpd, q);
716+
} else if (q->gws && !q->properties.is_gws) {
709717
if (q->properties.is_active) {
710718
dqm->gws_queue_count++;
711719
pdd->qpd.mapped_gws_queue = true;
@@ -767,11 +775,7 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
767775
mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
768776
q->properties.type)];
769777
q->properties.is_active = false;
770-
decrement_queue_count(dqm, q->properties.type);
771-
if (q->properties.is_gws) {
772-
dqm->gws_queue_count--;
773-
qpd->mapped_gws_queue = false;
774-
}
778+
decrement_queue_count(dqm, qpd, q);
775779

776780
if (WARN_ONCE(!dqm->sched_running, "Evict when stopped\n"))
777781
continue;
@@ -817,7 +821,7 @@ static int evict_process_queues_cpsch(struct device_queue_manager *dqm,
817821
continue;
818822

819823
q->properties.is_active = false;
820-
decrement_queue_count(dqm, q->properties.type);
824+
decrement_queue_count(dqm, qpd, q);
821825
}
822826
pdd->last_evict_timestamp = get_jiffies_64();
823827
retval = execute_queues_cpsch(dqm,
@@ -888,11 +892,7 @@ static int restore_process_queues_nocpsch(struct device_queue_manager *dqm,
888892
mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type(
889893
q->properties.type)];
890894
q->properties.is_active = true;
891-
increment_queue_count(dqm, q->properties.type);
892-
if (q->properties.is_gws) {
893-
dqm->gws_queue_count++;
894-
qpd->mapped_gws_queue = true;
895-
}
895+
increment_queue_count(dqm, qpd, q);
896896

897897
if (WARN_ONCE(!dqm->sched_running, "Restore when stopped\n"))
898898
continue;
@@ -950,7 +950,7 @@ static int restore_process_queues_cpsch(struct device_queue_manager *dqm,
950950
continue;
951951

952952
q->properties.is_active = true;
953-
increment_queue_count(dqm, q->properties.type);
953+
increment_queue_count(dqm, &pdd->qpd, q);
954954
}
955955
retval = execute_queues_cpsch(dqm,
956956
KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
@@ -1378,7 +1378,7 @@ static int create_kernel_queue_cpsch(struct device_queue_manager *dqm,
13781378
dqm->total_queue_count);
13791379

13801380
list_add(&kq->list, &qpd->priv_queue_list);
1381-
increment_queue_count(dqm, kq->queue->properties.type);
1381+
increment_queue_count(dqm, qpd, kq->queue);
13821382
qpd->is_debug = true;
13831383
execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
13841384
dqm_unlock(dqm);
@@ -1392,7 +1392,7 @@ static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm,
13921392
{
13931393
dqm_lock(dqm);
13941394
list_del(&kq->list);
1395-
decrement_queue_count(dqm, kq->queue->properties.type);
1395+
decrement_queue_count(dqm, qpd, kq->queue);
13961396
qpd->is_debug = false;
13971397
execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
13981398
/*
@@ -1467,7 +1467,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
14671467
qpd->queue_count++;
14681468

14691469
if (q->properties.is_active) {
1470-
increment_queue_count(dqm, q->properties.type);
1470+
increment_queue_count(dqm, qpd, q);
14711471

14721472
execute_queues_cpsch(dqm,
14731473
KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
@@ -1683,15 +1683,11 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
16831683
list_del(&q->list);
16841684
qpd->queue_count--;
16851685
if (q->properties.is_active) {
1686-
decrement_queue_count(dqm, q->properties.type);
1686+
decrement_queue_count(dqm, qpd, q);
16871687
retval = execute_queues_cpsch(dqm,
16881688
KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);
16891689
if (retval == -ETIME)
16901690
qpd->reset_wavefronts = true;
1691-
if (q->properties.is_gws) {
1692-
dqm->gws_queue_count--;
1693-
qpd->mapped_gws_queue = false;
1694-
}
16951691
}
16961692

16971693
/*
@@ -1932,7 +1928,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
19321928
/* Clean all kernel queues */
19331929
list_for_each_entry_safe(kq, kq_next, &qpd->priv_queue_list, list) {
19341930
list_del(&kq->list);
1935-
decrement_queue_count(dqm, kq->queue->properties.type);
1931+
decrement_queue_count(dqm, qpd, kq->queue);
19361932
qpd->is_debug = false;
19371933
dqm->total_queue_count--;
19381934
filter = KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES;
@@ -1945,13 +1941,8 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,
19451941
else if (q->properties.type == KFD_QUEUE_TYPE_SDMA_XGMI)
19461942
deallocate_sdma_queue(dqm, q);
19471943

1948-
if (q->properties.is_active) {
1949-
decrement_queue_count(dqm, q->properties.type);
1950-
if (q->properties.is_gws) {
1951-
dqm->gws_queue_count--;
1952-
qpd->mapped_gws_queue = false;
1953-
}
1954-
}
1944+
if (q->properties.is_active)
1945+
decrement_queue_count(dqm, qpd, q);
19551946

19561947
dqm->total_queue_count--;
19571948
}

0 commit comments

Comments
 (0)