Re: [PATCH v4 05/13] drm/xe/xe_ras: Initialize Uncorrectable AER Registers

[Raag Jadav <raag.jadav@intel.com>]

On Fri, Apr 17, 2026 at 02:28:17PM +0530, Riana Tauro wrote:
> Uncorrectable errors from different endpoints in the device are steered to
> the USP(Upstream Switch Port) which is a PCI Advanced Error Reporting (AER)
> Compliant device. Downgrade all the errors to non-fatal to prevent PCIe
> bus driver from triggering a Secondary Bus Reset (SBR). This allows error
> detection, containment and recovery in the driver.
> 
> The Uncorrectable Error Severity Register has the 'Uncorrectable
> Internal Error Severity' set to fatal by default. Set this to
> non-fatal and unmask the error.
> 
> Signed-off-by: Riana Tauro <riana.tauro@intel.com>
> ---
> v2: clear stale uncorrectable internal status in status register
> (Aravind)
> 
> v3: Abbrevate TLA's (Raag)
>     Add a info message if USP does not support AER
> ---
>  drivers/gpu/drm/xe/Makefile    |  1 +
>  drivers/gpu/drm/xe/xe_device.c |  3 ++
>  drivers/gpu/drm/xe/xe_ras.c    | 84 ++++++++++++++++++++++++++++++++++
>  drivers/gpu/drm/xe/xe_ras.h    | 13 ++++++
>  4 files changed, 101 insertions(+)
>  create mode 100644 drivers/gpu/drm/xe/xe_ras.c
>  create mode 100644 drivers/gpu/drm/xe/xe_ras.h
> 
> diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile
> index 69c233d9a488..e29a4ae99ac6 100644
> --- a/drivers/gpu/drm/xe/Makefile
> +++ b/drivers/gpu/drm/xe/Makefile
> @@ -113,6 +113,7 @@ xe-y += xe_bb.o \
>  	xe_pxp_debugfs.o \
>  	xe_pxp_submit.o \
>  	xe_query.o \
> +	xe_ras.o \
>  	xe_range_fence.o \
>  	xe_reg_sr.o \
>  	xe_reg_whitelist.o \
> diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> index cbdf7426e09c..c1c54836ac73 100644
> --- a/drivers/gpu/drm/xe/xe_device.c
> +++ b/drivers/gpu/drm/xe/xe_device.c
> @@ -62,6 +62,7 @@
>  #include "xe_psmi.h"
>  #include "xe_pxp.h"
>  #include "xe_query.h"
> +#include "xe_ras.h"
>  #include "xe_shrinker.h"
>  #include "xe_soc_remapper.h"
>  #include "xe_survivability_mode.h"
> @@ -1074,6 +1075,8 @@ int xe_device_probe(struct xe_device *xe)
>  
>  	xe_vsec_init(xe);
>  
> +	xe_ras_init(xe);
> +
>  	err = xe_sriov_init_late(xe);
>  	if (err)
>  		goto err_unregister_display;
> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
> new file mode 100644
> index 000000000000..4f705deaeefa
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_ras.c
> @@ -0,0 +1,84 @@
> +// SPDX-License-Identifier: MIT
> +/*
> + * Copyright © 2026 Intel Corporation
> + */
> +
> +#include "xe_device_types.h"
> +#include "xe_ras.h"
> +
> +#ifdef CONFIG_PCIEAER
> +static void aer_unmask_and_downgrade_internal_error(struct xe_device *xe)
> +{
> +	struct pci_dev *pdev = to_pci_dev(xe->drm.dev);
> +	struct pci_dev *vsp, *usp;
> +	u32 aer_uncorr_mask, aer_uncorr_sev, aer_uncorr_status;
> +	u16 aer_cap;
> +
> +	/*
> +	 * Device Hierarchy:
> +	 *
> +	 * Upstream Switch Port (USP)--> Virtual Switch Port (VSP)--> SGunit (GPU endpoint)
> +	 */
> +	vsp = pci_upstream_bridge(pdev);
> +	if (!vsp)
> +		return;
> +
> +	usp = pci_upstream_bridge(vsp);
> +	if (!usp)
> +		return;
> +
> +	aer_cap = usp->aer_cap;
> +
> +	if (!aer_cap) {
> +		dev_info(&usp->dev, "USP doesn't support AER capability\n");
> +		return;
> +	}
> +
> +	/*
> +	 * Clear any stale Uncorrectable Internal Error Status event in Uncorrectable Error
> +	 * Status Register.
> +	 */
> +	pci_read_config_dword(usp, aer_cap + PCI_ERR_UNCOR_STATUS, &aer_uncorr_status);
> +	if (aer_uncorr_status & PCI_ERR_UNC_INTN)
> +		pci_write_config_dword(usp, aer_cap + PCI_ERR_UNCOR_STATUS, PCI_ERR_UNC_INTN);
> +
> +	/*
> +	 * All errors are steered to USP which is a PCIe AER Compliant device.
> +	 * Downgrade all the errors to non-fatal to prevent PCIe bus driver
> +	 * from triggering a Secondary Bus Reset (SBR). This allows error
> +	 * detection, containment and recovery in the driver.
> +	 *
> +	 * The Uncorrectable Error Severity Register has the 'Uncorrectable
> +	 * Internal Error Severity' set to fatal by default. Set this to
> +	 * non-fatal and unmask the error.
> +	 */
> +
> +	/* Initialize Uncorrectable Error Severity Register */
> +	pci_read_config_dword(usp, aer_cap + PCI_ERR_UNCOR_SEVER, &aer_uncorr_sev);
> +	aer_uncorr_sev &= ~PCI_ERR_UNC_INTN;
> +	pci_write_config_dword(usp, aer_cap + PCI_ERR_UNCOR_SEVER, aer_uncorr_sev);
> +
> +	/* Initialize Uncorrectable Error Mask Register */
> +	pci_read_config_dword(usp, aer_cap + PCI_ERR_UNCOR_MASK, &aer_uncorr_mask);
> +	aer_uncorr_mask &= ~PCI_ERR_UNC_INTN;
> +	pci_write_config_dword(usp, aer_cap + PCI_ERR_UNCOR_MASK, aer_uncorr_mask);
> +
> +	pci_save_state(usp);
> +}
> +#endif
> +
> +/**
> + * xe_ras_init - Initialize Xe RAS
> + * @xe: xe device instance
> + *
> + * Initialize Xe RAS
> + */
> +void xe_ras_init(struct xe_device *xe)
> +{
> +	if (!xe->info.has_sysctrl)
> +		return;
> +
> +#ifdef CONFIG_PCIEAER
> +	aer_unmask_and_downgrade_internal_error(xe);

If we fail silently we'd most likely be clueless why RAS isn't working.
So either add error log here or have an explicit success log inside
downgrade function.

Raag

> +#endif
> +}
> diff --git a/drivers/gpu/drm/xe/xe_ras.h b/drivers/gpu/drm/xe/xe_ras.h
> new file mode 100644
> index 000000000000..14cb973603e7
> --- /dev/null
> +++ b/drivers/gpu/drm/xe/xe_ras.h
> @@ -0,0 +1,13 @@
> +/* SPDX-License-Identifier: MIT */
> +/*
> + * Copyright © 2026 Intel Corporation
> + */
> +
> +#ifndef _XE_RAS_H_
> +#define _XE_RAS_H_
> +
> +struct xe_device;
> +
> +void xe_ras_init(struct xe_device *xe);
> +
> +#endif
> -- 
> 2.47.1
>