Commit e528ccf9 authored by Evan Quan's avatar Evan Quan Committed by Alex Deucher
Browse files

drm/amd/powerplay: shutdown on HW CTF 



To prevent further damage.

Signed-off-by: default avatarEvan Quan <evan.quan@amd.com>
Reviewed-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: default avatarAlex Deucher <alexander.deucher@amd.com>
parent 94952205

drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.c

+14 −2
Original line number Original line Diff line number Diff line
@@ -612,11 +612,17 @@ int phm_irq_process(struct amdgpu_device *adev, 
					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_FUNC(adev->pdev->devfn));

					PCI_FUNC(adev->pdev->devfn));

		else if (src_id == VISLANDS30_IV_SRCID_GPIO_19)

		else if (src_id == VISLANDS30_IV_SRCID_GPIO_19) {

			pr_warn("GPU Critical Temperature Fault detected on PCIe %d:%d.%d!\n",

			pr_warn("GPU Critical Temperature Fault detected on PCIe %d:%d.%d!\n",

					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_FUNC(adev->pdev->devfn));

					PCI_FUNC(adev->pdev->devfn));

			/*

			 * HW CTF just occurred. Shutdown to prevent further damage.

			 */

			dev_emerg(adev->dev, "System is going to shutdown due to HW CTF!\n");

			orderly_poweroff(true);

		}

	} else if (client_id == SOC15_IH_CLIENTID_THM) {

	} else if (client_id == SOC15_IH_CLIENTID_THM) {

		if (src_id == 0) {

		if (src_id == 0) {

			pr_warn("GPU over temperature range detected on PCIe %d:%d.%d!\n",

			pr_warn("GPU over temperature range detected on PCIe %d:%d.%d!\n",
@@ -634,11 +640,17 @@ int phm_irq_process(struct amdgpu_device *adev, 
					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_BUS_NUM(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_SLOT(adev->pdev->devfn),

					PCI_FUNC(adev->pdev->devfn));

					PCI_FUNC(adev->pdev->devfn));

	} else if (client_id == SOC15_IH_CLIENTID_ROM_SMUIO)

	} else if (client_id == SOC15_IH_CLIENTID_ROM_SMUIO) {

		pr_warn("GPU Critical Temperature Fault detected on PCIe %d:%d.%d!\n",

		pr_warn("GPU Critical Temperature Fault detected on PCIe %d:%d.%d!\n",

				PCI_BUS_NUM(adev->pdev->devfn),

				PCI_BUS_NUM(adev->pdev->devfn),

				PCI_SLOT(adev->pdev->devfn),

				PCI_SLOT(adev->pdev->devfn),

				PCI_FUNC(adev->pdev->devfn));

				PCI_FUNC(adev->pdev->devfn));

		/*

		 * HW CTF just occurred. Shutdown to prevent further damage.

		 */

		dev_emerg(adev->dev, "System is going to shutdown due to HW CTF!\n");

		orderly_poweroff(true);

	}





	return 0;

	return 0;

}

}

drivers/gpu/drm/amd/powerplay/smu_v11_0.c

+19 −1
Original line number Original line Diff line number Diff line
@@ -1548,6 +1548,8 @@ static int smu_v11_0_ack_ac_dc_interrupt(struct smu_context *smu) 
#define THM_11_0__SRCID__THM_DIG_THERM_L2H		0		/* ASIC_TEMP > CG_THERMAL_INT.DIG_THERM_INTH  */

#define THM_11_0__SRCID__THM_DIG_THERM_L2H		0		/* ASIC_TEMP > CG_THERMAL_INT.DIG_THERM_INTH  */

#define THM_11_0__SRCID__THM_DIG_THERM_H2L		1		/* ASIC_TEMP < CG_THERMAL_INT.DIG_THERM_INTL  */

#define THM_11_0__SRCID__THM_DIG_THERM_H2L		1		/* ASIC_TEMP < CG_THERMAL_INT.DIG_THERM_INTL  */





#define SMUIO_11_0__SRCID__SMUIO_GPIO19			83



static int smu_v11_0_irq_process(struct amdgpu_device *adev,

static int smu_v11_0_irq_process(struct amdgpu_device *adev,

				 struct amdgpu_irq_src *source,

				 struct amdgpu_irq_src *source,

				 struct amdgpu_iv_entry *entry)

				 struct amdgpu_iv_entry *entry)
@@ -1582,8 +1584,17 @@ static int smu_v11_0_irq_process(struct amdgpu_device *adev, 
				PCI_SLOT(adev->pdev->devfn),

				PCI_SLOT(adev->pdev->devfn),

				PCI_FUNC(adev->pdev->devfn));

				PCI_FUNC(adev->pdev->devfn));

		break;

		break;



		}

		}

	} else if (client_id == SOC15_IH_CLIENTID_ROM_SMUIO) {

		pr_warn("GPU Critical Temperature Fault detected on PCIe %d:%d.%d!\n",

				PCI_BUS_NUM(adev->pdev->devfn),

				PCI_SLOT(adev->pdev->devfn),

				PCI_FUNC(adev->pdev->devfn));

		/*

		 * HW CTF just occurred. Shutdown to prevent further damage.

		 */

		dev_emerg(adev->dev, "System is going to shutdown due to HW CTF!\n");

		orderly_poweroff(true);

	} else if (client_id == SOC15_IH_CLIENTID_MP1) {

	} else if (client_id == SOC15_IH_CLIENTID_MP1) {

		if (src_id == 0xfe)

		if (src_id == 0xfe)

			smu_v11_0_ack_ac_dc_interrupt(&adev->smu);

			smu_v11_0_ack_ac_dc_interrupt(&adev->smu);
@@ -1626,6 +1637,13 @@ int smu_v11_0_register_irq_handler(struct smu_context *smu) 
	if (ret)

	if (ret)

		return ret;

		return ret;





	/* Register CTF(GPIO_19) interrupt */

	ret = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_ROM_SMUIO,

				SMUIO_11_0__SRCID__SMUIO_GPIO19,

				irq_src);

	if (ret)

		return ret;



	ret = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_MP1,

	ret = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_MP1,

				0xfe,

				0xfe,

				irq_src);

				irq_src);