Skip to contentSkip to conversation

drm/amd/powerplay: add SMU mode1 reset

PATCH In review Wenhui Sheng 4 participants 21 messages Jul 10, 2020 – Jul 13, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/3 drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +
 drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +20 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev)
+{
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev)
+{
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev);
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +34 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu)
+{
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu)
+{
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)
 {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +4 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max);
@@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);
 int smu_baco_enter(struct smu_context *smu);
 int smu_baco_exit(struct smu_context *smu);
 
+bool smu_mode1_reset_is_support(struct smu_context *smu);
+int smu_mode1_reset(struct smu_context *smu);
 int smu_mode2_reset(struct smu_context *smu);
 
 extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table,
drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
index dff2295705be..7b585e205a5a 100644
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)
 int smu_v11_0_baco_enter(struct smu_context *smu);
 int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +28 -3
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = {
@@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
 	return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false;
 }
 
+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context *smu)
+{
+	struct amdgpu_device *adev = smu->adev;
+	uint32_t val;
+	u32 smu_version;
+
+	/**
+	 * SRIOV env will not support SMU mode1 reset
+	 * PM FW support mode1 reset from 58.26
+	 */
+	smu_get_smc_version(smu, NULL, &smu_version);
+	if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+		return false;
+
+	/**
+	 * mode1 reset relies on PSP, so we should check if
+	 * PSP is alive.
+	 */
+	val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+	return val != 0x0;
+}
+
 static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
 				       struct smu_temperature_range range)
 {
@@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
 	.baco_set_state = smu_v11_0_baco_set_state,
 	.baco_enter = smu_v11_0_baco_enter,
 	.baco_exit = smu_v11_0_baco_exit,
+	.mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+	.mode1_reset = smu_v11_0_mode1_reset,
 	.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
 	.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
 	.override_pcie_parameters = smu_v11_0_override_pcie_parameters,
drivers/gpu/drm/amd/powerplay/smu_v11_0.c +13 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu)
+{
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {
Signature · 6 lines
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#1 lore raw

PATCH 2/3 drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +12 -11
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	struct amdgpu_hive_info *hive = NULL;
 	struct amdgpu_device *tmp_adev = NULL;
 	int i, r = 0;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
-	bool use_baco =
-		(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-		true : false;
+	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 
+	/**
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 	/*
 	 * Flush RAM to disk so that after reboot
 	 * the user can read log and see why the system rebooted.
 	 */
-	if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 		DRM_WARN("Emergency reboot.");
 
 		ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	}
 
 	dev_info(adev->dev, "GPU %s begin!\n",
-		(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+		need_emergency_restart ? "jobs stop":"reset");
 
 	/*
 	 * Here we trylock to avoid chain of resets executing from
@@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		amdgpu_fbdev_set_suspend(tmp_adev, 1);
 
 		/* disable ras on ALL IPs */
-		if (!(in_ras_intr && !use_baco) &&
+		if (!need_emergency_restart &&
 		      amdgpu_device_ip_need_full_reset(tmp_adev))
 			amdgpu_ras_suspend(tmp_adev);
 
@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 
 			drm_sched_stop(&ring->sched, job ? &job->base : NULL);
 
-			if (in_ras_intr && !use_baco)
+			if (need_emergency_restart)
 				amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
 		}
 	}
 
-	if (in_ras_intr && !use_baco)
+	if (need_emergency_restart)
 		goto skip_sched_resume;
 
 	/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 skip_sched_resume:
 	list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) {
 		/*unlock kfd: SRIOV would do it separately */
-		if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+		if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
 	                amdgpu_amdkfd_post_reset(tmp_adev);
 		if (audio_suspended)
 			amdgpu_device_resume_display_audio(tmp_adev);
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +11 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 		amdgpu_ras_reset_gpu(adev);
 	}
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev)
+{
+	if (adev->asic_type == CHIP_VEGA20 &&
+	    adev->pm.fw_version <= 0x283400) {
+		return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+				amdgpu_ras_intr_triggered();
+	}
+
+	return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +12 -7
index 356849136d1d..9f1240bd0310 100644
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-	dev_info(adev->dev, "GPU mode1 reset\n");
-
 	/* disable BM */
 	pci_clear_master(adev->pdev);
 
 	pci_save_state(adev->pdev);
 
-	ret = psp_gpu_reset(adev);
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
+
 	if (ret)
 		dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */
@@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)
 {
 	struct smu_context *smu = &adev->smu;
 
-	if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+	if (smu_baco_is_support(smu))
 		return AMD_RESET_METHOD_BACO;
 	else
 		return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
 	struct smu_context *smu = &adev->smu;
 
 	if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+		dev_info(adev->dev, "GPU BACO reset\n");
+
 		ret = smu_baco_enter(smu);
 		if (ret)
 			return ret;
 		ret = smu_baco_exit(smu);
 		if (ret)
 			return ret;
-	} else {
+	} else
 		ret = nv_asic_mode1_reset(adev);
-	}
 
 	return ret;
 }
drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -1
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
Signature · 6 lines
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#2 lore raw

RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Official Use Only - Internal Distribution Only]

It seems this patch contains different changes. Better to split into several patches:
One for amdgpu_ras_need_emergency_restart() changes.
One for the following changes:
+if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+dev_info(adev->dev, "GPU smu mode1 reset\n");
+ret = amdgpu_dpm_mode1_reset(adev);
+} else {
+dev_info(adev->dev, "GPU psp mode1 reset\n");
+ret = psp_gpu_reset(adev);
+}
+
One for the changes below
-if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+if (smu_baco_is_support(smu))
And maybe the final one for coding style fixes.

BR
Evan
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +8 -7
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 struct amdgpu_hive_info *hive = NULL;
 struct amdgpu_device *tmp_adev = NULL;
 int i, r = 0;
-bool in_ras_intr = amdgpu_ras_intr_triggered();
-bool use_baco =
-(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-true : false;
+bool need_emergency_restart = false;
 bool audio_suspended = false;
 
+/**
+ * Special case: RAS triggered and full reset isn't supported
+ */
+need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 /*
  * Flush RAM to disk so that after reboot
  * the user can read log and see why the system rebooted.
  */
-if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 DRM_WARN("Emergency reboot.");
 
 ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 }
 
 dev_info(adev->dev, "GPU %s begin!\n",
-(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+need_emergency_restart ? "jobs stop":"reset");
 
 /*
  * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

amdgpu_fbdev_set_suspend(tmp_adev, 1);

/* disable ras on ALL IPs */
-if (!(in_ras_intr && !use_baco) &&
+if (!need_emergency_restart &&
amdgpu_device_ip_need_full_reset(tmp_adev))
amdgpu_ras_suspend(tmp_adev);

@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

drm_sched_stop(&ring->sched, job ? &job->base : NULL);

-if (in_ras_intr && !use_baco)
+if (need_emergency_restart)
amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
}
}

-if (in_ras_intr && !use_baco)
+if (need_emergency_restart)
goto skip_sched_resume;

/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
skip_sched_resume:
list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { /*unlock kfd: SRIOV would do it separately */
-if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev)) +if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))

                 amdgpu_amdkfd_post_reset(tmp_adev);

if (audio_suspended)
amdgpu_device_resume_display_audio(tmp_adev);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +10 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 amdgpu_ras_reset_gpu(adev);
 }
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) {
+if (adev->asic_type == CHIP_VEGA20 &&
+    adev->pm.fw_version <= 0x283400) {
+return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+amdgpu_ras_intr_triggered();
+}
+
+return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +8 -4
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-dev_info(adev->dev, "GPU mode1 reset\n");
-
 /* disable BM */
 pci_clear_master(adev->pdev);
 
 pci_save_state(adev->pdev);
 
-ret = psp_gpu_reset(adev);
+if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+dev_info(adev->dev, "GPU smu mode1 reset\n");
+ret = amdgpu_dpm_mode1_reset(adev);
+} else {
+dev_info(adev->dev, "GPU psp mode1 reset\n");
+ret = psp_gpu_reset(adev);
+}
+
 if (ret)
 dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 pci_restore_state(adev->pdev);
 
 /* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {

struct smu_context *smu = &adev->smu;

-if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else
return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
struct smu_context *smu = &adev->smu;

if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+dev_info(adev->dev, "GPU BACO reset\n");
+
ret = smu_baco_enter(smu);
if (ret)
return ret;
ret = smu_baco_exit(smu);
if (ret)
return ret;
-} else {
+} else
ret = nv_asic_mode1_reset(adev);
-}

return ret;
}

drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -1
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 struct amdgpu_device *adev = smu->adev;
 uint32_t val;
 
-if (!smu_v11_0_baco_is_support(smu))
+if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 return false;
 
 val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C5f3fbaad4309421b1e7508d824dbed1e%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874318950168&amp;sdata=iVrmi9ovJz57IzZfR3wFAJDEhVmwXmTGOlzvMIxAZq4%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#3 lore raw

RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Official Use Only - Internal Distribution Only]

[AMD Official Use Only - Internal Distribution Only]

Ok will refine it

Brs
Wenhui

-----Original Message-----
From: Quan, Evan <[email protected]>
Sent: Monday, July 13, 2020 11:26 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Official Use Only - Internal Distribution Only]

It seems this patch contains different changes. Better to split into several patches:
One for amdgpu_ras_need_emergency_restart() changes.
One for the following changes:
+if (amdgpu_dpm_is_mode1_reset_supported(adev)) { dev_info(adev->dev, +"GPU smu mode1 reset\n"); ret = amdgpu_dpm_mode1_reset(adev); } else { +dev_info(adev->dev, "GPU psp mode1 reset\n"); ret = +psp_gpu_reset(adev); }
+
One for the changes below
-if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+if (smu_baco_is_support(smu))
And maybe the final one for coding style fixes.

BR
Evan
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +7 -5
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,  struct amdgpu_hive_info *hive = NULL;  struct amdgpu_device *tmp_adev = NULL;  int i, r = 0; -bool in_ras_intr = amdgpu_ras_intr_triggered(); -bool use_baco =
-(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-true : false;
+bool need_emergency_restart = false;
 bool audio_suspended = false;
 
+/**
+ * Special case: RAS triggered and full reset isn't supported  */
+need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 /*
  * Flush RAM to disk so that after reboot
  * the user can read log and see why the system rebooted.
  */
-if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 DRM_WARN("Emergency reboot.");
 
 ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,  }
 
 dev_info(adev->dev, "GPU %s begin!\n",
-(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+need_emergency_restart ? "jobs stop":"reset");
 
 /*
  * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,  amdgpu_fbdev_set_suspend(tmp_adev, 1);
 

/* disable ras on ALL IPs */
-if (!(in_ras_intr && !use_baco) &&
+if (!need_emergency_restart &&
amdgpu_device_ip_need_full_reset(tmp_adev))
amdgpu_ras_suspend(tmp_adev);

@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

drm_sched_stop(&ring->sched, job ? &job->base : NULL);

-if (in_ras_intr && !use_baco)
+if (need_emergency_restart)
amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
}
}

-if (in_ras_intr && !use_baco)
+if (need_emergency_restart)
goto skip_sched_resume;

/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
skip_sched_resume:
list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) { /*unlock kfd: SRIOV would do it separately */ -if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))

                 amdgpu_amdkfd_post_reset(tmp_adev);

if (audio_suspended)
amdgpu_device_resume_display_audio(tmp_adev);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +9 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)  amdgpu_ras_reset_gpu(adev);  }  }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) { if
+(adev->asic_type == CHIP_VEGA20 &&
+    adev->pm.fw_version <= 0x283400) {
+return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+amdgpu_ras_intr_triggered(); }
+
+return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +5 -4
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-dev_info(adev->dev, "GPU mode1 reset\n");
-
 /* disable BM */
 pci_clear_master(adev->pdev);
 
 pci_save_state(adev->pdev);
 
-ret = psp_gpu_reset(adev);
+if (amdgpu_dpm_is_mode1_reset_supported(adev)) { dev_info(adev->dev,
+"GPU smu mode1 reset\n"); ret = amdgpu_dpm_mode1_reset(adev); } else {
+dev_info(adev->dev, "GPU psp mode1 reset\n"); ret =
+psp_gpu_reset(adev); }
+
 if (ret)
 dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 pci_restore_state(adev->pdev);
 
 /* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {  struct smu_context *smu = &adev->smu;

-if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else
return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev) struct smu_context *smu = &adev->smu;

if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+dev_info(adev->dev, "GPU BACO reset\n");
+
ret = smu_baco_enter(smu);
if (ret)
return ret;
ret = smu_baco_exit(smu);
if (ret)
return ret;
-} else {
+} else
ret = nv_asic_mode1_reset(adev);
-}

return ret;
}

drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -2
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)  struct amdgpu_device *adev = smu->adev;  uint32_t val;
 
-if (!smu_v11_0_baco_is_support(smu))
+if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 return false;
 
 val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
--

2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C5f3fbaad4309421b1e7508d824dbed1e%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874318950168&amp;sdata=iVrmi9ovJz57IzZfR3wFAJDEhVmwXmTGOlzvMIxAZq4%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#4 lore raw

RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

In reply to Wenhui Sheng

[AMD Public Use]

V2 looks good to me expect for the following that can be improved

+ if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {

The check of reboot could be centralized to the new function amdgpu_ras_need_emergency_restart you added in amdgpu_ras.c. And it would be better you can make amdgpu_ras_need_emergency_restart related as a separated patch so that it helps to isolate ras recovery related issue from general gpu reset issue.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +8 -7
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	struct amdgpu_hive_info *hive = NULL;
 	struct amdgpu_device *tmp_adev = NULL;
 	int i, r = 0;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
-	bool use_baco =
-		(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-		true : false;
+	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 
+	/**
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 	/*
 	 * Flush RAM to disk so that after reboot
 	 * the user can read log and see why the system rebooted.
 	 */
-	if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 		DRM_WARN("Emergency reboot.");
 
 		ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	}
 
 	dev_info(adev->dev, "GPU %s begin!\n",
-		(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+		need_emergency_restart ? "jobs stop":"reset");
 
 	/*
 	 * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

amdgpu_fbdev_set_suspend(tmp_adev, 1);

/* disable ras on ALL IPs */
- if (!(in_ras_intr && !use_baco) &&
+ if (!need_emergency_restart &&
amdgpu_device_ip_need_full_reset(tmp_adev))
amdgpu_ras_suspend(tmp_adev);

@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

drm_sched_stop(&ring->sched, job ? &job->base : NULL);

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
}
}

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
goto skip_sched_resume;

/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
skip_sched_resume:
list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) {
/*unlock kfd: SRIOV would do it separately */
- if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+ if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
amdgpu_amdkfd_post_reset(tmp_adev);
if (audio_suspended)
amdgpu_device_resume_display_audio(tmp_adev);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +10 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 		amdgpu_ras_reset_gpu(adev);
 	}
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) {
+	if (adev->asic_type == CHIP_VEGA20 &&
+	    adev->pm.fw_version <= 0x283400) {
+		return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+				amdgpu_ras_intr_triggered();
+	}
+
+	return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +8 -4
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-	dev_info(adev->dev, "GPU mode1 reset\n");
-
 	/* disable BM */
 	pci_clear_master(adev->pdev);
 
 	pci_save_state(adev->pdev);
 
-	ret = psp_gpu_reset(adev);
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
+
 	if (ret)
 		dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {

struct smu_context *smu = &adev->smu;

- if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+ if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else
return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
struct smu_context *smu = &adev->smu;

if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+ dev_info(adev->dev, "GPU BACO reset\n");
+
ret = smu_baco_enter(smu);
if (ret)
return ret;
ret = smu_baco_exit(smu);
if (ret)
return ret;
- } else {
+ } else
ret = nv_asic_mode1_reset(adev);
- }

return ret;
}

drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -1
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);

--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#5 lore raw

RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Public Use]

So first make a patch to refine need_emergency_resetart logic in amdgpu_device_gpu_recover function. Then apply you next series to add new mode1 reset support

Regards,
Hawking

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Zhang, Hawking
Sent: Monday, July 13, 2020 13:34
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Public Use]

V2 looks good to me expect for the following that can be improved

+ if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {

The check of reboot could be centralized to the new function amdgpu_ras_need_emergency_restart you added in amdgpu_ras.c. And it would be better you can make amdgpu_ras_need_emergency_restart related as a separated patch so that it helps to isolate ras recovery related issue from general gpu reset issue.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +8 -7
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	struct amdgpu_hive_info *hive = NULL;
 	struct amdgpu_device *tmp_adev = NULL;
 	int i, r = 0;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
-	bool use_baco =
-		(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-		true : false;
+	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 
+	/**
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 	/*
 	 * Flush RAM to disk so that after reboot
 	 * the user can read log and see why the system rebooted.
 	 */
-	if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 		DRM_WARN("Emergency reboot.");
 
 		ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	}
 
 	dev_info(adev->dev, "GPU %s begin!\n",
-		(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+		need_emergency_restart ? "jobs stop":"reset");
 
 	/*
 	 * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

amdgpu_fbdev_set_suspend(tmp_adev, 1);

/* disable ras on ALL IPs */
- if (!(in_ras_intr && !use_baco) &&
+ if (!need_emergency_restart &&
amdgpu_device_ip_need_full_reset(tmp_adev))
amdgpu_ras_suspend(tmp_adev);

@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

drm_sched_stop(&ring->sched, job ? &job->base : NULL);

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
}
}

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
goto skip_sched_resume;

/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
skip_sched_resume:
list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) {
/*unlock kfd: SRIOV would do it separately */
- if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+ if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
amdgpu_amdkfd_post_reset(tmp_adev);
if (audio_suspended)
amdgpu_device_resume_display_audio(tmp_adev);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +10 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 		amdgpu_ras_reset_gpu(adev);
 	}
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) {
+	if (adev->asic_type == CHIP_VEGA20 &&
+	    adev->pm.fw_version <= 0x283400) {
+		return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+				amdgpu_ras_intr_triggered();
+	}
+
+	return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +8 -4
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-	dev_info(adev->dev, "GPU mode1 reset\n");
-
 	/* disable BM */
 	pci_clear_master(adev->pdev);
 
 	pci_save_state(adev->pdev);
 
-	ret = psp_gpu_reset(adev);
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
+
 	if (ret)
 		dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {

struct smu_context *smu = &adev->smu;

- if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+ if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else
return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
struct smu_context *smu = &adev->smu;

if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+ dev_info(adev->dev, "GPU BACO reset\n");
+
ret = smu_baco_enter(smu);
if (ret)
return ret;
ret = smu_baco_exit(smu);
if (ret)
return ret;
- } else {
+ } else
ret = nv_asic_mode1_reset(adev);
- }

return ret;
}

drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -1
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);

--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Chawking.zhang%40amd.com%7Cba02ce03c500432b668108d826ee5cd6%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637302152493471927&amp;sdata=1Kn2Q6DqkrGj2hgmT6jBHibeDPCmc%2BytqGz3Bma0fv0%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#6 lore raw

RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Public Use]

HI Hawking

I have merged in_ras_intr & use_baco to the function amdgpu_ras_need_emergency_restart, If I also move the check of reboot logic code to the new added function, variable need_emergency_restart Couldn't be used for the following if check in function amdgpu_device_gpu_recover.

Brs
Wenhui

-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Monday, July 13, 2020 1:40 PM
To: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Public Use]

So first make a patch to refine need_emergency_resetart logic in amdgpu_device_gpu_recover function. Then apply you next series to add new mode1 reset support

Regards,
Hawking

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Zhang, Hawking
Sent: Monday, July 13, 2020 13:34
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 2/3] drm/amdgpu: enable mode1 reset

[AMD Public Use]

V2 looks good to me expect for the following that can be improved

+ if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {

The check of reboot could be centralized to the new function amdgpu_ras_need_emergency_restart you added in amdgpu_ras.c. And it would be better you can make amdgpu_ras_need_emergency_restart related as a separated patch so that it helps to isolate ras recovery related issue from general gpu reset issue.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 2/3] drm/amdgpu: enable mode1 reset

For sienna cichlid, add mode1 reset path for sGPU. And fix some corner cases about mode1 mode reuse.

v2: hiding MP0/MP1 mode1 reset under AMD_RESET_METHOD_MODE1

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 +++++++++++-----------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c    | 11 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h    |  1 +
 drivers/gpu/drm/amd/amdgpu/nv.c            | 19 +++++++++++-------
 drivers/gpu/drm/amd/powerplay/navi10_ppt.c |  2 +-
 5 files changed, 37 insertions(+), 19 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +8 -7
index fec6cf3f0c8a..6ce6e6bb8b50 100644
@@ -4274,18 +4274,19 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	struct amdgpu_hive_info *hive = NULL;
 	struct amdgpu_device *tmp_adev = NULL;
 	int i, r = 0;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
-	bool use_baco =
-		(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) ?
-		true : false;
+	bool need_emergency_restart = false;
 	bool audio_suspended = false;
 
+	/**
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
 	/*
 	 * Flush RAM to disk so that after reboot
 	 * the user can read log and see why the system rebooted.
 	 */
-	if (in_ras_intr && !use_baco && amdgpu_ras_get_context(adev)->reboot) {
-
+	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
 		DRM_WARN("Emergency reboot.");
 
 		ksys_sync_helper();
@@ -4293,7 +4294,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 	}
 
 	dev_info(adev->dev, "GPU %s begin!\n",
-		(in_ras_intr && !use_baco) ? "jobs stop":"reset");
+		need_emergency_restart ? "jobs stop":"reset");
 
 	/*
 	 * Here we trylock to avoid chain of resets executing from @@ -4365,7 +4366,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

amdgpu_fbdev_set_suspend(tmp_adev, 1);

/* disable ras on ALL IPs */
- if (!(in_ras_intr && !use_baco) &&
+ if (!need_emergency_restart &&
amdgpu_device_ip_need_full_reset(tmp_adev))
amdgpu_ras_suspend(tmp_adev);

@@ -4377,12 +4378,12 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,

drm_sched_stop(&ring->sched, job ? &job->base : NULL);

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
amdgpu_job_stop_all_jobs_on_sched(&ring->sched);
}
}

- if (in_ras_intr && !use_baco)
+ if (need_emergency_restart)
goto skip_sched_resume;

/*
@@ -4459,7 +4460,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
skip_sched_resume:
list_for_each_entry(tmp_adev, device_list_handle, gmc.xgmi.head) {
/*unlock kfd: SRIOV would do it separately */
- if (!(in_ras_intr && !use_baco) && !amdgpu_sriov_vf(tmp_adev))
+ if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
amdgpu_amdkfd_post_reset(tmp_adev);
if (audio_suspended)
amdgpu_device_resume_display_audio(tmp_adev);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +10 -0
index 3a3fa8567c94..6f06e1214622 100644
@@ -2135,3 +2135,14 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
 		amdgpu_ras_reset_gpu(adev);
 	}
 }
+
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev) {
+	if (adev->asic_type == CHIP_VEGA20 &&
+	    adev->pm.fw_version <= 0x283400) {
+		return !(amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) &&
+				amdgpu_ras_intr_triggered();
+	}
+
+	return false;
+}
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -0
index e7df5d8429f8..b2667342cf67 100644
@@ -633,4 +633,5 @@ void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
+bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev);
 #endif
drivers/gpu/drm/amd/amdgpu/nv.c +8 -4
@@ -265,17 +265,21 @@ static int nv_asic_mode1_reset(struct amdgpu_device *adev)
 
 	amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-	dev_info(adev->dev, "GPU mode1 reset\n");
-
 	/* disable BM */
 	pci_clear_master(adev->pdev);
 
 	pci_save_state(adev->pdev);
 
-	ret = psp_gpu_reset(adev);
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
+
 	if (ret)
 		dev_err(adev->dev, "GPU mode1 reset failed\n");
-
 	pci_restore_state(adev->pdev);
 
 	/* wait for asic to come out of reset */ @@ -307,7 +311,7 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {

struct smu_context *smu = &adev->smu;

- if (!amdgpu_sriov_vf(adev) && smu_baco_is_support(smu))
+ if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else
return AMD_RESET_METHOD_MODE1;
@@ -319,15 +323,16 @@ static int nv_asic_reset(struct amdgpu_device *adev)
struct smu_context *smu = &adev->smu;

if (nv_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
+ dev_info(adev->dev, "GPU BACO reset\n");
+
ret = smu_baco_enter(smu);
if (ret)
return ret;
ret = smu_baco_exit(smu);
if (ret)
return ret;
- } else {
+ } else
ret = nv_asic_mode1_reset(adev);
- }

return ret;
}

drivers/gpu/drm/amd/powerplay/navi10_ppt.c +1 -1
index 4f7d064e16e4..014815bcae93 100644
@@ -2039,7 +2039,7 @@ static bool navi10_is_baco_supported(struct smu_context *smu)
 	struct amdgpu_device *adev = smu->adev;
 	uint32_t val;
 
-	if (!smu_v11_0_baco_is_support(smu))
+	if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
 		return false;
 
 	val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);

--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Chawking.zhang%40amd.com%7Cba02ce03c500432b668108d826ee5cd6%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637302152493471927&amp;sdata=1Kn2Q6DqkrGj2hgmT6jBHibeDPCmc%2BytqGz3Bma0fv0%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#7 lore raw

PATCH 3/3 drm/amdgpu: add module parameter choose reset mode

In reply to Wenhui Sheng

Default value is auto, doesn't change
original reset method logic.

v2: change to use parameter reset_method

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 8 ++++++++
 drivers/gpu/drm/amd/amdgpu/cik.c        | 4 ++++
 drivers/gpu/drm/amd/amdgpu/nv.c         | 4 ++++
 drivers/gpu/drm/amd/amdgpu/soc15.c      | 5 +++++
 drivers/gpu/drm/amd/amdgpu/vi.c         | 4 ++++
 6 files changed, 26 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index 4de93cef79b9..06bfb8658dec 100644
@@ -196,6 +196,7 @@ static const bool debug_evictions; /* = false */
 #endif
 
 extern int amdgpu_tmz;
+extern int amdgpu_reset_method;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +8 -0
index 94c83a9d4987..ed9f32b89f36 100644
@@ -154,6 +154,7 @@ int amdgpu_mes = 0;
 int amdgpu_noretry = 1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
+int amdgpu_reset_method = -1; /* auto */
 
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
@@ -793,6 +794,13 @@ module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
 MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto, 0 = off (default), 1 = on)");
 module_param_named(tmz, amdgpu_tmz, int, 0444);
 
+/**
+ * DOC: reset_method (int)
+ * SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)
+ */
+MODULE_PARM_DESC(reset_method, "SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)");
+module_param_named(reset_method, amdgpu_reset_method, int, 0444);
+
 static const struct pci_device_id pciidlist[] = {
 #ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
drivers/gpu/drm/amd/amdgpu/cik.c +4 -0
index fe306d0f73f7..a99418cd2f5e 100644
@@ -1326,6 +1326,10 @@ cik_asic_reset_method(struct amdgpu_device *adev)
 {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
drivers/gpu/drm/amd/amdgpu/nv.c +4 -0
index 9f1240bd0310..309ff402d7be 100644
@@ -311,6 +311,10 @@ nv_asic_reset_method(struct amdgpu_device *adev)
 {
 	struct smu_context *smu = &adev->smu;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	if (smu_baco_is_support(smu))
 		return AMD_RESET_METHOD_BACO;
 	else
drivers/gpu/drm/amd/amdgpu/soc15.c +5 -0
index 8c739b285915..a8a134f7bb61 100644
@@ -532,6 +532,11 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 	bool baco_reset = false;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+		amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_RAVEN:
 	case CHIP_RENOIR:
drivers/gpu/drm/amd/amdgpu/vi.c +4 -0
index 4e5e91888d87..2cc8ad19ebed 100644
@@ -710,6 +710,10 @@ vi_asic_reset_method(struct amdgpu_device *adev)
 {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_FIJI:
 	case CHIP_TONGA:
Signature · 6 lines
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
#8 lore raw

RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

Default value is auto, doesn't change
original reset method logic.

v2: change to use parameter reset_method

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 8 ++++++++
 drivers/gpu/drm/amd/amdgpu/cik.c        | 4 ++++
 drivers/gpu/drm/amd/amdgpu/nv.c         | 4 ++++
 drivers/gpu/drm/amd/amdgpu/soc15.c      | 5 +++++
 drivers/gpu/drm/amd/amdgpu/vi.c         | 4 ++++
 6 files changed, 26 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index 4de93cef79b9..06bfb8658dec 100644
@@ -196,6 +196,7 @@ static const bool debug_evictions; /* = false */
 #endif
 
 extern int amdgpu_tmz;
+extern int amdgpu_reset_method;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +8 -0
index 94c83a9d4987..ed9f32b89f36 100644
@@ -154,6 +154,7 @@ int amdgpu_mes = 0;
 int amdgpu_noretry = 1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
+int amdgpu_reset_method = -1; /* auto */
 
 struct amdgpu_mgpu_info mgpu_info = {
 .mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
@@ -793,6 +794,13 @@ module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
 MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto, 0 = off (default), 1 = on)");
 module_param_named(tmz, amdgpu_tmz, int, 0444);
 
+/**
+ * DOC: reset_method (int)
+ * SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)
+ */
+MODULE_PARM_DESC(reset_method, "SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)");
+module_param_named(reset_method, amdgpu_reset_method, int, 0444);
+
 static const struct pci_device_id pciidlist[] = {
 #ifdef  CONFIG_DRM_AMDGPU_SI
 {0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
drivers/gpu/drm/amd/amdgpu/cik.c +4 -0
index fe306d0f73f7..a99418cd2f5e 100644
@@ -1326,6 +1326,10 @@ cik_asic_reset_method(struct amdgpu_device *adev)
 {
 bool baco_reset;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+return amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_BONAIRE:
 case CHIP_HAWAII:
drivers/gpu/drm/amd/amdgpu/nv.c +4 -0
index 9f1240bd0310..309ff402d7be 100644
@@ -311,6 +311,10 @@ nv_asic_reset_method(struct amdgpu_device *adev)
 {
 struct smu_context *smu = &adev->smu;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+return amdgpu_reset_method;
+

[Quan, Evan] I kind of wonder what's the expected behavior if user specifies amdgpu_reset_mthod as "0 = legacy" here? User's settings seems ignored silently?
if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else

drivers/gpu/drm/amd/amdgpu/soc15.c +5 -0
index 8c739b285915..a8a134f7bb61 100644
@@ -532,6 +532,11 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 bool baco_reset = false;
 struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+return amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_RAVEN:
 case CHIP_RENOIR:
drivers/gpu/drm/amd/amdgpu/vi.c +4 -0
index 4e5e91888d87..2cc8ad19ebed 100644
@@ -710,6 +710,10 @@ vi_asic_reset_method(struct amdgpu_device *adev)
 {
 bool baco_reset;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+return amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_FIJI:
 case CHIP_TONGA:

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C259262f416404c23d64108d824dbf044%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874389349093&amp;sdata=uyCH2W4ox82ta%2BZRO4PXeYhHf2o7RH3fMoawcsFy3M4%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#9 lore raw

RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

[AMD Official Use Only - Internal Distribution Only]

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: Quan, Evan <[email protected]>
Sent: Monday, July 13, 2020 11:30 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

Default value is auto, doesn't change
original reset method logic.

v2: change to use parameter reset_method

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 8 ++++++++
 drivers/gpu/drm/amd/amdgpu/cik.c        | 4 ++++
 drivers/gpu/drm/amd/amdgpu/nv.c         | 4 ++++
 drivers/gpu/drm/amd/amdgpu/soc15.c      | 5 +++++
 drivers/gpu/drm/amd/amdgpu/vi.c         | 4 ++++
 6 files changed, 26 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index 4de93cef79b9..06bfb8658dec 100644
@@ -196,6 +196,7 @@ static const bool debug_evictions; /* = false */  #endif
 
 extern int amdgpu_tmz;
+extern int amdgpu_reset_method;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +9 -0
index 94c83a9d4987..ed9f32b89f36 100644
@@ -154,6 +154,7 @@ int amdgpu_mes = 0;
 int amdgpu_noretry = 1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
+int amdgpu_reset_method = -1; /* auto */
 
 struct amdgpu_mgpu_info mgpu_info = {
 .mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
@@ -793,6 +794,13 @@ module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);  MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto, 0 = off (default), 1 = on)");  module_param_named(tmz, amdgpu_tmz, int, 0444);
 
+/**
+ * DOC: reset_method (int)
+ * SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 =
+mode1, 3 = mode2, 4 = baco)  */ MODULE_PARM_DESC(reset_method, "SMU
+reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3
+= mode2, 4 = baco)"); module_param_named(reset_method,
+amdgpu_reset_method, int, 0444);
+
 static const struct pci_device_id pciidlist[] = {  #ifdef  CONFIG_DRM_AMDGPU_SI  {0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI}, diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c

index fe306d0f73f7..a99418cd2f5e 100644

drivers/gpu/drm/amd/amdgpu/cik.c +4 -0
@@ -1326,6 +1326,10 @@ cik_asic_reset_method(struct amdgpu_device *adev)  {  bool baco_reset;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO) return
+amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_BONAIRE:
 case CHIP_HAWAII:
drivers/gpu/drm/amd/amdgpu/nv.c +4 -0
@@ -311,6 +311,10 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {  struct smu_context *smu = &adev->smu;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO) return
+amdgpu_reset_method;
+

[Quan, Evan] I kind of wonder what's the expected behavior if user specifies amdgpu_reset_mthod as "0 = legacy" here? User's settings seems ignored silently? [Wenhui, Sheng] yes, if this asic doesn't support legacy, should choose one by auto, do you think this is make sense? Do we need add more debug message here?

if (smu_baco_is_support(smu))
return AMD_RESET_METHOD_BACO;
else

drivers/gpu/drm/amd/amdgpu/soc15.c +5 -0
index 8c739b285915..a8a134f7bb61 100644
@@ -532,6 +532,11 @@ soc15_asic_reset_method(struct amdgpu_device *adev)  bool baco_reset = false;  struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+amdgpu_reset_method == AMD_RESET_METHOD_BACO) return
+amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_RAVEN:
 case CHIP_RENOIR:
drivers/gpu/drm/amd/amdgpu/vi.c +4 -1
@@ -710,6 +710,10 @@ vi_asic_reset_method(struct amdgpu_device *adev)  {  bool baco_reset;
 
+if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+    amdgpu_reset_method == AMD_RESET_METHOD_BACO) return
+amdgpu_reset_method;
+
 switch (adev->asic_type) {
 case CHIP_FIJI:
 case CHIP_TONGA:
--

2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C259262f416404c23d64108d824dbf044%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874389349093&amp;sdata=uyCH2W4ox82ta%2BZRO4PXeYhHf2o7RH3fMoawcsFy3M4%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#10 lore raw

RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

In reply to Wenhui Sheng

[AMD Public Use]

Hi Wenhui,

I think we shall deal with the case where user specify an unexpected/unsupported reset method via the module parameter.

So in each asic_reset_method function, rule out this case, force to auto, and print warning message to inform user.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

Default value is auto, doesn't change
original reset method logic.

v2: change to use parameter reset_method

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 8 ++++++++
 drivers/gpu/drm/amd/amdgpu/cik.c        | 4 ++++
 drivers/gpu/drm/amd/amdgpu/nv.c         | 4 ++++
 drivers/gpu/drm/amd/amdgpu/soc15.c      | 5 +++++
 drivers/gpu/drm/amd/amdgpu/vi.c         | 4 ++++
 6 files changed, 26 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index 4de93cef79b9..06bfb8658dec 100644
@@ -196,6 +196,7 @@ static const bool debug_evictions; /* = false */  #endif
 
 extern int amdgpu_tmz;
+extern int amdgpu_reset_method;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +9 -0
index 94c83a9d4987..ed9f32b89f36 100644
@@ -154,6 +154,7 @@ int amdgpu_mes = 0;
 int amdgpu_noretry = 1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
+int amdgpu_reset_method = -1; /* auto */
 
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
@@ -793,6 +794,13 @@ module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);  MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto, 0 = off (default), 1 = on)");  module_param_named(tmz, amdgpu_tmz, int, 0444);
 
+/**
+ * DOC: reset_method (int)
+ * SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = 
+mode1, 3 = mode2, 4 = baco)  */ MODULE_PARM_DESC(reset_method, "SMU 
+reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 
+= mode2, 4 = baco)"); module_param_named(reset_method, 
+amdgpu_reset_method, int, 0444);
+
 static const struct pci_device_id pciidlist[] = {  #ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI}, diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c

index fe306d0f73f7..a99418cd2f5e 100644

drivers/gpu/drm/amd/amdgpu/cik.c +4 -0
@@ -1326,6 +1326,10 @@ cik_asic_reset_method(struct amdgpu_device *adev)  {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
drivers/gpu/drm/amd/amdgpu/nv.c +4 -0
@@ -311,6 +311,10 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {
 	struct smu_context *smu = &adev->smu;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	if (smu_baco_is_support(smu))
 		return AMD_RESET_METHOD_BACO;
 	else
drivers/gpu/drm/amd/amdgpu/soc15.c +5 -0
index 8c739b285915..a8a134f7bb61 100644
@@ -532,6 +532,11 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 	bool baco_reset = false;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+		amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_RAVEN:
 	case CHIP_RENOIR:
drivers/gpu/drm/amd/amdgpu/vi.c +4 -1
@@ -710,6 +710,10 @@ vi_asic_reset_method(struct amdgpu_device *adev)  {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_FIJI:
 	case CHIP_TONGA:
--

2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#11 lore raw

RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

[AMD Public Use]

Ok got it.

Brs
Wenhui
-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Monday, July 13, 2020 1:52 PM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

[AMD Public Use]

Hi Wenhui,

I think we shall deal with the case where user specify an unexpected/unsupported reset method via the module parameter.

So in each asic_reset_method function, rule out this case, force to auto, and print warning message to inform user.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 3/3] drm/amdgpu: add module parameter choose reset mode

Default value is auto, doesn't change
original reset method logic.

v2: change to use parameter reset_method

---
 drivers/gpu/drm/amd/amdgpu/amdgpu.h     | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 8 ++++++++
 drivers/gpu/drm/amd/amdgpu/cik.c        | 4 ++++
 drivers/gpu/drm/amd/amdgpu/nv.c         | 4 ++++
 drivers/gpu/drm/amd/amdgpu/soc15.c      | 5 +++++
 drivers/gpu/drm/amd/amdgpu/vi.c         | 4 ++++
 6 files changed, 26 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
index 4de93cef79b9..06bfb8658dec 100644
@@ -196,6 +196,7 @@ static const bool debug_evictions; /* = false */  #endif
 
 extern int amdgpu_tmz;
+extern int amdgpu_reset_method;
 
 #ifdef CONFIG_DRM_AMDGPU_SI
 extern int amdgpu_si_support;
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +9 -0
index 94c83a9d4987..ed9f32b89f36 100644
@@ -154,6 +154,7 @@ int amdgpu_mes = 0;
 int amdgpu_noretry = 1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
+int amdgpu_reset_method = -1; /* auto */
 
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
@@ -793,6 +794,13 @@ module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);  MODULE_PARM_DESC(tmz, "Enable TMZ feature (-1 = auto, 0 = off (default), 1 = on)");  module_param_named(tmz, amdgpu_tmz, int, 0444);
 
+/**
+ * DOC: reset_method (int)
+ * SMU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = 
+mode1, 3 = mode2, 4 = baco)  */ MODULE_PARM_DESC(reset_method, "SMU 
+reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 
+= mode2, 4 = baco)"); module_param_named(reset_method, 
+amdgpu_reset_method, int, 0444);
+
 static const struct pci_device_id pciidlist[] = {  #ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI}, diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c

index fe306d0f73f7..a99418cd2f5e 100644

drivers/gpu/drm/amd/amdgpu/cik.c +4 -0
@@ -1326,6 +1326,10 @@ cik_asic_reset_method(struct amdgpu_device *adev)  {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
drivers/gpu/drm/amd/amdgpu/nv.c +4 -0
@@ -311,6 +311,10 @@ nv_asic_reset_method(struct amdgpu_device *adev)  {
 	struct smu_context *smu = &adev->smu;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	if (smu_baco_is_support(smu))
 		return AMD_RESET_METHOD_BACO;
 	else
drivers/gpu/drm/amd/amdgpu/soc15.c +5 -0
index 8c739b285915..a8a134f7bb61 100644
@@ -532,6 +532,11 @@ soc15_asic_reset_method(struct amdgpu_device *adev)
 	bool baco_reset = false;
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_MODE1 ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_MODE2 ||
+		amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_RAVEN:
 	case CHIP_RENOIR:
drivers/gpu/drm/amd/amdgpu/vi.c +4 -1
@@ -710,6 +710,10 @@ vi_asic_reset_method(struct amdgpu_device *adev)  {
 	bool baco_reset;
 
+	if (amdgpu_reset_method == AMD_RESET_METHOD_LEGACY ||
+	    amdgpu_reset_method == AMD_RESET_METHOD_BACO)
+		return amdgpu_reset_method;
+
 	switch (adev->asic_type) {
 	case CHIP_FIJI:
 	case CHIP_TONGA:
--

2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#12 lore raw
In reply to Wenhui Sheng

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;

[Dennis Li] It is better to changed to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register.

+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7CDennis.Li%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874279933866&amp;sdata=DDzKv9mwDCDUyGQUYfKDk4nM0kqt07Qt45Iyr6RQQQU%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#13 lore raw

[AMD Official Use Only - Internal Distribution Only]

Ok, will refine it.

Brs
Wenhui

-----Original Message-----
From: Li, Dennis <[email protected]>
Sent: Monday, July 13, 2020 11:10 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;

[Dennis Li] It is better to changed to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register.

+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7CDennis.Li%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874279933866&amp;sdata=DDzKv9mwDCDUyGQUYfKDk4nM0kqt07Qt45Iyr6RQQQU%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#14 lore raw

[AMD Public Use]

RE - [Dennis Li] It is better change to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register

RREG32_SOC15_NO_KIQ should have no difference from RREG32_SOC15 for this use scenario. This is the feature only supported in bare-metal environment and never run from guest environment. But this do remind us to exclude the feature from guest run-time environment by checking amdgpu_vf_sriov()

Regards,
Hawking

-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Monday, July 13, 2020 11:43
To: Li, Dennis <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

Ok, will refine it.

Brs
Wenhui

-----Original Message-----
From: Li, Dennis <[email protected]>
Sent: Monday, July 13, 2020 11:10 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;

[Dennis Li] It is better to changed to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register.

+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7CDennis.Li%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874279933866&amp;sdata=DDzKv9mwDCDUyGQUYfKDk4nM0kqt07Qt45Iyr6RQQQU%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Reviewed-by
#15 lore raw

[AMD Public Use]

Hi, Hawking,
Got it. Thanks for your expiation. It looks good to me now.

Best Regards
Dennis Li
-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Monday, July 13, 2020 1:44 PM
To: Sheng, Wenhui <[email protected]>; Li, Dennis <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Public Use]

RE - [Dennis Li] It is better change to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register

RREG32_SOC15_NO_KIQ should have no difference from RREG32_SOC15 for this use scenario. This is the feature only supported in bare-metal environment and never run from guest environment. But this do remind us to exclude the feature from guest run-time environment by checking amdgpu_vf_sriov()

Regards,
Hawking

-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Monday, July 13, 2020 11:43
To: Li, Dennis <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

Ok, will refine it.

Brs
Wenhui

-----Original Message-----
From: Li, Dennis <[email protected]>
Sent: Monday, July 13, 2020 11:10 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;

[Dennis Li] It is better to changed to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register.

+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7CDennis.Li%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874279933866&amp;sdata=DDzKv9mwDCDUyGQUYfKDk4nM0kqt07Qt45Iyr6RQQQU%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#16 lore raw
In reply to Zhang, Hawking

[AMD Public Use]

Hi Hawking,

Not sure If there is a situation that we pass through the physical device to a vm, the condition check " if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) " in function amdgpu_mm_rreg should be true when we call RREG32_SOC15, is it correct?

Brs
Wenhui

-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Monday, July 13, 2020 1:44 PM
To: Sheng, Wenhui <[email protected]>; Li, Dennis <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Public Use]

RE - [Dennis Li] It is better change to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register

RREG32_SOC15_NO_KIQ should have no difference from RREG32_SOC15 for this use scenario. This is the feature only supported in bare-metal environment and never run from guest environment. But this do remind us to exclude the feature from guest run-time environment by checking amdgpu_vf_sriov()

Regards,
Hawking

-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Monday, July 13, 2020 11:43
To: Li, Dennis <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

Ok, will refine it.

Brs
Wenhui

-----Original Message-----
From: Li, Dennis <[email protected]>
Sent: Monday, July 13, 2020 11:10 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;

[Dennis Li] It is better to changed to use RREG32_SOC15_NO_KIQ, because when GPU hang, RREG32_SOC15 will fail if it use RREG32_KIQ to read register.

+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7CDennis.Li%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874279933866&amp;sdata=DDzKv9mwDCDUyGQUYfKDk4nM0kqt07Qt45Iyr6RQQQU%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#17 lore raw
In reply to Wenhui Sheng

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+struct smu_context *smu = &adev->smu;
+
+if (is_support_sw_smu(adev))
+return smu_mode1_reset_is_support(smu);
+
+return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+struct smu_context *smu = &adev->smu;
+
+if (is_support_sw_smu(adev))
+return smu_mode1_reset(smu);
+
+return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
     enum PP_SMC_POWER_PROFILE type,
     bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev);
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
      enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+bool ret = false;
+
+if (!smu->pm_enabled)
+return false;
+
+mutex_lock(&smu->mutex);
+
+if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+mutex_unlock(&smu->mutex);
+
+return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+int ret = 0;
+
+if (!smu->pm_enabled)
+return -EOPNOTSUPP;
+
+mutex_lock(&smu->mutex);
+
+if (smu->ppt_funcs->mode1_reset)
+ret = smu->ppt_funcs->mode1_reset(smu);
+
+mutex_unlock(&smu->mutex);
+
+return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 int (*baco_enter)(struct smu_context *smu);
 int (*baco_exit)(struct smu_context *smu);
+bool (*mode1_reset_is_support)(struct smu_context *smu);
+int (*mode1_reset)(struct smu_context *smu);
 int (*mode2_reset)(struct smu_context *smu);
 int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 __SMU_DUMMY_MAP(GmiPwrDnControl), \
 __SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 __SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
  uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 MSG_MAP(PowerDownJpeg,PPSMC_MSG_PowerDownJpeg),
 MSG_MAP(BacoAudioD3PME,PPSMC_MSG_BacoAudioD3PME),
 MSG_MAP(ArmD3,PPSMC_MSG_ArmD3),
+MSG_MAP(Mode1Reset,PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

-if (!smu_v11_0_baco_is_support(smu))
+if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;
[Quan, Evan] This seems irrelevant. Better to create a separate patch for this.

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+struct amdgpu_device *adev = smu->adev;
+uint32_t val;
+u32 smu_version;
+
+/**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+smu_get_smc_version(smu, NULL, &smu_version);
+if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00)) +return false;
+
+/**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+return val != 0x0;
+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,

        struct smu_temperature_range range)  { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {

.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+.mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported, +.mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq, .set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range, .override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
       uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+int ret = 0;
+
+ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+if (!ret)
+msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
  uint32_t *min, uint32_t *max)
 {

--
2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874326587249&amp;sdata=2pLBrCIvPkTh9tepSjuI2XU3VHhLtJjCmgm4MrzLl%2Fo%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#18 lore raw

[AMD Official Use Only - Internal Distribution Only]

[AMD Official Use Only - Internal Distribution Only]

Ok will refine it.

Brs
Wenhui

-----Original Message-----
From: Quan, Evan <[email protected]>
Sent: Monday, July 13, 2020 11:19 AM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Official Use Only - Internal Distribution Only]

-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Wenhui Sheng
Sent: Friday, July 10, 2020 10:17 PM
To: [email protected]
Cc: Gao, Likun <[email protected]>; Sheng, Wenhui <[email protected]>; Zhang, Hawking <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)  return 0;  }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+struct smu_context *smu = &adev->smu;
+
+if (is_support_sw_smu(adev))
+return smu_mode1_reset_is_support(smu);
+
+return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) { struct
+smu_context *smu = &adev->smu;
+
+if (is_support_sw_smu(adev))
+return smu_mode1_reset(smu);
+
+return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
     enum PP_SMC_POWER_PROFILE type,
     bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev);
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
      enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +31 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)  return ret;  }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) { bool ret =
+false;
+
+if (!smu->pm_enabled)
+return false;
+
+mutex_lock(&smu->mutex);
+
+if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+mutex_unlock(&smu->mutex);
+
+return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) { int ret = 0;
+
+if (!smu->pm_enabled)
+return -EOPNOTSUPP;
+
+mutex_lock(&smu->mutex);
+
+if (smu->ppt_funcs->mode1_reset)
+ret = smu->ppt_funcs->mode1_reset(smu);
+
+mutex_unlock(&smu->mutex);
+
+return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {  int ret = 0; diff --git a/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h

index 7b349e038972..ba59620950d7 100644

drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +4 -0
@@ -561,6 +561,8 @@ struct pptable_funcs {  int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);  int (*baco_enter)(struct smu_context *smu);  int (*baco_exit)(struct smu_context *smu);
+bool (*mode1_reset_is_support)(struct smu_context *smu); int
+(*mode1_reset)(struct smu_context *smu);
 int (*mode2_reset)(struct smu_context *smu);  int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);  int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);
 
+bool smu_mode1_reset_is_support(struct smu_context *smu); int
+smu_mode1_reset(struct smu_context *smu);
 int smu_mode2_reset(struct smu_context *smu);
 

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 __SMU_DUMMY_MAP(GmiPwrDnControl), \
 __SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \  __SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
  uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 MSG_MAP(PowerDownJpeg,PPSMC_MSG_PowerDownJpeg),
 MSG_MAP(BacoAudioD3PME,PPSMC_MSG_BacoAudioD3PME),
 MSG_MAP(ArmD3,PPSMC_MSG_ArmD3),
+MSG_MAP(Mode1Reset,PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)  struct amdgpu_device *adev = smu->adev;  uint32_t val;

-if (!smu_v11_0_baco_is_support(smu))
+if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;
[Quan, Evan] This seems irrelevant. Better to create a separate patch for this.

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0); return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context +*smu) {
+struct amdgpu_device *adev = smu->adev; uint32_t val;
+u32 smu_version;
+
+/**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26 */ smu_get_smc_version(smu, +NULL, &smu_version); if (amdgpu_sriov_vf(adev) || (smu_version < +0x003a1a00)) return false;
+
+/**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); return val != 0x0; } +
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,

        struct smu_temperature_range range)  { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {  .baco_set_state = smu_v11_0_baco_set_state,  .baco_enter = smu_v11_0_baco_enter,  .baco_exit = smu_v11_0_baco_exit,

+.mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported, +.mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq, .set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range, .override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +10 -1
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
       uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)  return ret;  }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) { int ret = 0;
+
+ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL); if (!ret)
+msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
  uint32_t *min, uint32_t *max)
 {
--

2.17.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&amp;data=02%7C01%7Cevan.quan%40amd.com%7C5125fa5f71e245a34f4608d824dbec23%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637299874326587249&amp;sdata=2pLBrCIvPkTh9tepSjuI2XU3VHhLtJjCmgm4MrzLl%2Fo%3D&amp;reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#19 lore raw
In reply to Wenhui Sheng

[AMD Public Use]

V2 looks good to me except for the following nitpicks

+#define SMU11_MODE1_RESET_WAIT_TIME 500 //500ms
It would be better explicitly call ms units out in the macro name, say SMU11_MODE1_RESET_WAIT_TIMEOUT_IN_MS.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context
+*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;
+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

#20 lore raw

[AMD Public Use]

Good suggestion, will refine the name.

Brs
Wenhui

-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Monday, July 13, 2020 1:55 PM
To: Sheng, Wenhui <[email protected]>; [email protected]
Cc: Gao, Likun <[email protected]>
Subject: RE: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

[AMD Public Use]

V2 looks good to me except for the following nitpicks

+#define SMU11_MODE1_RESET_WAIT_TIME 500 //500ms
It would be better explicitly call ms units out in the macro name, say SMU11_MODE1_RESET_WAIT_TIMEOUT_IN_MS.

Regards,
Hawking
-----Original Message-----
From: Sheng, Wenhui <[email protected]>
Sent: Friday, July 10, 2020 22:17
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Sheng, Wenhui <[email protected]>; Gao, Likun <[email protected]>
Subject: [PATCH 1/3] drm/amd/powerplay: add SMU mode1 reset

From PM FW 58.26.0 for sienna cichlid, SMU mode1 reset is support, driver sends PPSMC_MSG_Mode1Reset message to PM FW could trigger this reset.

v2: add mode1 reset dpm interface

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c       | 20 +++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h       |  3 ++
 drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 34 +++++++++++++++++++
 .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  4 +++
 drivers/gpu/drm/amd/powerplay/inc/smu_types.h |  1 +  drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h |  2 ++
 .../drm/amd/powerplay/sienna_cichlid_ppt.c    | 31 +++++++++++++++--
 drivers/gpu/drm/amd/powerplay/smu_v11_0.c     | 13 +++++++
 8 files changed, 105 insertions(+), 3 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +18 -0
index 65472b3dd815..16668fc52d0d 100644
@@ -1141,6 +1141,26 @@ int amdgpu_dpm_baco_reset(struct amdgpu_device *adev)
 	return 0;
 }
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset_is_support(smu);
+
+	return false;
+}
+
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev) {
+	struct smu_context *smu = &adev->smu;
+
+	if (is_support_sw_smu(adev))
+		return smu_mode1_reset(smu);
+
+	return -EOPNOTSUPP;
+}
+
 int amdgpu_dpm_switch_power_profile(struct amdgpu_device *adev,
 				    enum PP_SMC_POWER_PROFILE type,
 				    bool en)
drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +3 -0
index 6a8aae70a0e6..7f3cd7185650 100644
@@ -529,6 +529,9 @@ int amdgpu_dpm_mode2_reset(struct amdgpu_device *adev);
 
 bool amdgpu_dpm_is_baco_supported(struct amdgpu_device *adev);
 
+bool amdgpu_dpm_is_mode1_reset_supported(struct amdgpu_device *adev); 
+int amdgpu_dpm_mode1_reset(struct amdgpu_device *adev);
+
 int amdgpu_dpm_set_mp1_state(struct amdgpu_device *adev,
 			     enum pp_mp1_state mp1_state);
 
drivers/gpu/drm/amd/powerplay/amdgpu_smu.c +32 -0
index fe4948aa662f..b5a7422d9548 100644
@@ -2737,6 +2737,40 @@ int smu_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+bool smu_mode1_reset_is_support(struct smu_context *smu) {
+	bool ret = false;
+
+	if (!smu->pm_enabled)
+		return false;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs && smu->ppt_funcs->mode1_reset_is_support)
+		ret = smu->ppt_funcs->mode1_reset_is_support(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
+int smu_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	if (!smu->pm_enabled)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&smu->mutex);
+
+	if (smu->ppt_funcs->mode1_reset)
+		ret = smu->ppt_funcs->mode1_reset(smu);
+
+	mutex_unlock(&smu->mutex);
+
+	return ret;
+}
+
 int smu_mode2_reset(struct smu_context *smu)  {
 	int ret = 0;
drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h +2 -0
index 7b349e038972..ba59620950d7 100644
@@ -561,6 +561,8 @@ struct pptable_funcs {
 	int (*baco_set_state)(struct smu_context *smu, enum smu_baco_state state);
 	int (*baco_enter)(struct smu_context *smu);
 	int (*baco_exit)(struct smu_context *smu);
+	bool (*mode1_reset_is_support)(struct smu_context *smu);
+	int (*mode1_reset)(struct smu_context *smu);
 	int (*mode2_reset)(struct smu_context *smu);
 	int (*get_dpm_ultimate_freq)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t *min, uint32_t *max);
 	int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); @@ -672,6 +674,8 @@ int smu_baco_get_state(struct smu_context *smu, enum smu_baco_state *state);  int smu_baco_enter(struct smu_context *smu);  int smu_baco_exit(struct smu_context *smu);

+bool smu_mode1_reset_is_support(struct smu_context *smu); int +smu_mode1_reset(struct smu_context *smu);
int smu_mode2_reset(struct smu_context *smu);

extern int smu_get_atom_data_table(struct smu_context *smu, uint32_t table, diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu_types.h b/drivers/gpu/drm/amd/powerplay/inc/smu_types.h
index dff2295705be..7b585e205a5a 100644

drivers/gpu/drm/amd/powerplay/inc/smu_types.h +1 -0
@@ -173,6 +173,7 @@
 	__SMU_DUMMY_MAP(GmiPwrDnControl), \
 	__SMU_DUMMY_MAP(DAL_DISABLE_DUMMY_PSTATE_CHANGE), \
 	__SMU_DUMMY_MAP(DAL_ENABLE_DUMMY_PSTATE_CHANGE), \
+	__SMU_DUMMY_MAP(Mode1Reset), \
 
 #undef __SMU_DUMMY_MAP
 #define __SMU_DUMMY_MAP(type)	SMU_MSG_##type
drivers/gpu/drm/amd/powerplay/inc/smu_v11_0.h +2 -0
index d07bf4fe6e4a..38599112ae59 100644
@@ -252,6 +252,8 @@ int smu_v11_0_baco_set_state(struct smu_context *smu, enum smu_baco_state state)  int smu_v11_0_baco_enter(struct smu_context *smu);  int smu_v11_0_baco_exit(struct smu_context *smu);
 
+int smu_v11_0_mode1_reset(struct smu_context *smu);
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max);
 
drivers/gpu/drm/amd/powerplay/sienna_cichlid_ppt.c +3 -2
index dc5ca9121db5..319480550bb7 100644
@@ -39,8 +39,8 @@
 #include "nbio/nbio_2_3_sh_mask.h"
 #include "thm/thm_11_0_2_offset.h"
 #include "thm/thm_11_0_2_sh_mask.h"
-
-#include "asic_reg/mp/mp_11_0_sh_mask.h"
+#include "mp/mp_11_0_offset.h"
+#include "mp/mp_11_0_sh_mask.h"
 
 /*
  * DO NOT use these for err/warn/info/debug messages.
@@ -116,6 +116,7 @@ static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_message_map[SMU_MSG_MAX_C
 	MSG_MAP(PowerDownJpeg,			PPSMC_MSG_PowerDownJpeg),
 	MSG_MAP(BacoAudioD3PME,			PPSMC_MSG_BacoAudioD3PME),
 	MSG_MAP(ArmD3,				PPSMC_MSG_ArmD3),
+	MSG_MAP(Mode1Reset,			PPSMC_MSG_Mode1Reset),
 };
 
 static struct smu_11_0_cmn2aisc_mapping sienna_cichlid_clk_map[SMU_CLK_COUNT] = { @@ -1760,13 +1761,35 @@ static bool sienna_cichlid_is_baco_supported(struct smu_context *smu)

struct amdgpu_device *adev = smu->adev;
uint32_t val;

- if (!smu_v11_0_baco_is_support(smu))
+ if (amdgpu_sriov_vf(adev) || (!smu_v11_0_baco_is_support(smu)))
return false;

val = RREG32_SOC15(NBIO, 0, mmRCC_BIF_STRAP0);
return (val & RCC_BIF_STRAP0__STRAP_PX_CAPABLE_MASK) ? true : false; }

+static bool sienna_cichlid_is_mode1_reset_supported(struct smu_context
+*smu) {
+ struct amdgpu_device *adev = smu->adev;
+ uint32_t val;
+ u32 smu_version;
+
+ /**
+ * SRIOV env will not support SMU mode1 reset
+ * PM FW support mode1 reset from 58.26
+ */
+ smu_get_smc_version(smu, NULL, &smu_version);
+ if (amdgpu_sriov_vf(adev) || (smu_version < 0x003a1a00))
+ return false;
+
+ /**
+ * mode1 reset relies on PSP, so we should check if
+ * PSP is alive.
+ */
+ val = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81);
+ return val != 0x0;
+}
+
static int sienna_cichlid_set_thermal_range(struct smu_context *smu,
struct smu_temperature_range range) { @@ -2538,6 +2561,8 @@ static const struct pptable_funcs sienna_cichlid_ppt_funcs = {
.baco_set_state = smu_v11_0_baco_set_state,
.baco_enter = smu_v11_0_baco_enter,
.baco_exit = smu_v11_0_baco_exit,
+ .mode1_reset_is_support = sienna_cichlid_is_mode1_reset_supported,
+ .mode1_reset = smu_v11_0_mode1_reset,
.get_dpm_ultimate_freq = sienna_cichlid_get_dpm_ultimate_freq,
.set_soft_freq_limited_range = sienna_cichlid_set_soft_freq_limited_range,
.override_pcie_parameters = smu_v11_0_override_pcie_parameters,

drivers/gpu/drm/amd/powerplay/smu_v11_0.c +12 -0
index 48e15885e9c3..c620dccb82e5 100644
@@ -63,6 +63,8 @@ MODULE_FIRMWARE("amdgpu/sienna_cichlid_smc.bin");
 
 #define SMU11_VOLTAGE_SCALE 4
 
+#define SMU11_MODE1_RESET_WAIT_TIME 500  //500ms
+
 static int smu_v11_0_send_msg_without_waiting(struct smu_context *smu,
 					      uint16_t msg)
 {
@@ -1741,6 +1743,17 @@ int smu_v11_0_baco_exit(struct smu_context *smu)
 	return ret;
 }
 
+int smu_v11_0_mode1_reset(struct smu_context *smu) {
+	int ret = 0;
+
+	ret = smu_send_smc_msg(smu, SMU_MSG_Mode1Reset, NULL);
+	if (!ret)
+		msleep(SMU11_MODE1_RESET_WAIT_TIME);
+
+	return ret;
+}
+
 int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type clk_type,
 						 uint32_t *min, uint32_t *max)
 {

--
2.17.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx