From: Amber Lin <Amber.Lin@xxxxxxx>
When amdgpu enable enforce_isolation, KFD enables single-process mode in
HWS and sets exec_cleaner_shader bit in MAP_PROCESS.
Signed-off-by: Amber Lin <Amber.Lin@xxxxxxx>
Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
---
drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c | 14 +++++++++++++-
drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h | 5 +++--
.../gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h | 2 +-
3 files changed, 17 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
index 00776f08351c..1f9f5bfeaf86 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager_v9.c
@@ -37,11 +37,14 @@ static int pm_map_process_v9(struct packet_manager *pm,
struct kfd_node *kfd = pm->dqm->dev;
struct kfd_process_device *pdd =
container_of(qpd, struct kfd_process_device, qpd);
+ struct amdgpu_device *adev = kfd->adev;
packet = (struct pm4_mes_map_process *)buffer;
memset(buffer, 0, sizeof(struct pm4_mes_map_process));
packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS,
sizeof(struct pm4_mes_map_process));
+ if (adev->enforce_isolation[kfd->node_id])
+ packet->bitfields2.exec_cleaner_shader = 1;
packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0;
packet->bitfields2.process_quantum = 10;
packet->bitfields2.pasid = qpd->pqm->process->pasid;
@@ -89,14 +92,18 @@ static int pm_map_process_aldebaran(struct packet_manager *pm,
struct pm4_mes_map_process_aldebaran *packet;
uint64_t vm_page_table_base_addr = qpd->page_table_base;
struct kfd_dev *kfd = pm->dqm->dev->kfd;
+ struct kfd_node *knode = pm->dqm->dev;
struct kfd_process_device *pdd =
container_of(qpd, struct kfd_process_device, qpd);
int i;
+ struct amdgpu_device *adev = kfd->adev;
packet = (struct pm4_mes_map_process_aldebaran *)buffer;
memset(buffer, 0, sizeof(struct pm4_mes_map_process_aldebaran));
packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS,
sizeof(struct pm4_mes_map_process_aldebaran));
+ if (adev->enforce_isolation[knode->node_id])
+ packet->bitfields2.exec_cleaner_shader = 1;
packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0;
packet->bitfields2.process_quantum = 10;
packet->bitfields2.pasid = qpd->pqm->process->pasid;
@@ -144,17 +151,22 @@ static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer,
int concurrent_proc_cnt = 0;
struct kfd_node *kfd = pm->dqm->dev;
+ struct amdgpu_device *adev = kfd->adev;
/* Determine the number of processes to map together to HW:
* it can not exceed the number of VMIDs available to the
* scheduler, and it is determined by the smaller of the number
* of processes in the runlist and kfd module parameter
* hws_max_conc_proc.
+ * However, if enforce_isolation is set (toggle LDS/VGPRs/SGPRs
+ * cleaner between process switch), enable single-process mode
+ * in HWS.
* Note: the arbitration between the number of VMIDs and
* hws_max_conc_proc has been done in
* kgd2kfd_device_init().
*/
- concurrent_proc_cnt = min(pm->dqm->processes_count,
+ concurrent_proc_cnt = adev->enforce_isolation[kfd->node_id] ?
+ 1 : min(pm->dqm->processes_count,
kfd->max_proc_per_quantum);
packet = (struct pm4_mes_runlist *)buffer;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
index 8b6b2bd5c148..cd8611401a66 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h
@@ -145,8 +145,9 @@ struct pm4_mes_map_process {
union {
struct {
- uint32_t pasid:16;
- uint32_t reserved1:2;
+ uint32_t pasid:16; /* 0 - 15 */
+ uint32_t reserved1:1; /* 16 */
+ uint32_t exec_cleaner_shader:1; /* 17 */
uint32_t debug_vmid:4;
uint32_t new_debug:1;
uint32_t reserved2:1;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
index 38f5cb6a222a..e0ed62c4ade0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_aldebaran.h
@@ -37,7 +37,7 @@ struct pm4_mes_map_process_aldebaran {
struct {
uint32_t pasid:16; /* 0 - 15 */
uint32_t single_memops:1; /* 16 */
- uint32_t reserved1:1; /* 17 */
+ uint32_t exec_cleaner_shader:1; /* 17 */
uint32_t debug_vmid:4; /* 18 - 21 */
uint32_t new_debug:1; /* 22 */
uint32_t tmz:1; /* 23 */
--
2.46.0
