Skip to contentSkip to conversation

drm/amdgpu: extend halt_if_hws_hang to MES

PATCH Superseded Graham Sider 2 participants 2 messages Oct 25, 2022 – Oct 31, 2022

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: extend halt_if_hws_hang to MES

Hang on MES timeout if halt_if_hws_hang is set to 1.

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h    | 2 ++
 drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 4 ++++
 2 files changed, 6 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 0e6ddf05c23c..9999c18e7d8e 100644
@@ -219,10 +219,12 @@ extern int amdgpu_use_xgmi_p2p;
 extern int sched_policy;
 extern bool debug_evictions;
 extern bool no_system_mem_limit;
+extern int halt_if_hws_hang;
 #else
 static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS;
 static const bool __maybe_unused debug_evictions; /* = false */
 static const bool __maybe_unused no_system_mem_limit;
+static const int __maybe_unused halt_if_hws_hang;
 #endif
 #ifdef CONFIG_HSA_AMD_P2P
 extern bool pcie_p2p;
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +4 -0
index e14f314902b1..d09f92d38ad1 100644
@@ -129,6 +129,10 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
 	if (r < 1) {
 		DRM_ERROR("MES failed to response msg=%d\n",
 			  x_pkt->header.opcode);
+
+		while (halt_if_hws_hang)
+			schedule();
+
 		return -ETIMEDOUT;
 	}
 
Signature · 1 line
2.25.1
#1 lore raw

Am 2022-10-25 um 09:42 schrieb Graham Sider:

Quoted 30 lines

Hang on MES timeout if halt_if_hws_hang is set to 1.

---
  drivers/gpu/drm/amd/amdgpu/amdgpu.h    | 2 ++
  drivers/gpu/drm/amd/amdgpu/mes_v11_0.c | 4 ++++
  2 files changed, 6 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +2 -0
index 0e6ddf05c23c..9999c18e7d8e 100644
@@ -219,10 +219,12 @@ extern int amdgpu_use_xgmi_p2p;
  extern int sched_policy;
  extern bool debug_evictions;
  extern bool no_system_mem_limit;
+extern int halt_if_hws_hang;
  #else
  static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS;
  static const bool __maybe_unused debug_evictions; /* = false */
  static const bool __maybe_unused no_system_mem_limit;
+static const int __maybe_unused halt_if_hws_hang;
  #endif
  #ifdef CONFIG_HSA_AMD_P2P
  extern bool pcie_p2p;
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +0 -0
index e14f314902b1..d09f92d38ad1 100644

drivers/gpu/drm/amd/amdgpu/mes_v11_0.c static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, · line 135

Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +4 -0
@@ -129,6 +129,10 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes,
  	if (r < 1) {
  		DRM_ERROR("MES failed to response msg=%d\n",
  			  x_pkt->header.opcode);
+
+		while (halt_if_hws_hang)
+			schedule();
+

I think we should add the same in mes_v10_1.c to keep it consistent.

Regards,
  Felix

Quoted 3 lines
drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +0 -0
  		return -ETIMEDOUT;
  	}