Re: [PATCH v2 2/2] drm/xe/xe_ras: Add RAS support for GPU health indicator

[Andi Shyti <andi.shyti@kernel.org>]

Hi Soham,

...

> diff --git a/Documentation/ABI/testing/sysfs-driver-intel-xe-ras b/Documentation/ABI/testing/sysfs-driver-intel-xe-ras
> new file mode 100644
> index 000000000000..085cb79a6e00
> --- /dev/null
> +++ b/Documentation/ABI/testing/sysfs-driver-intel-xe-ras

Thanks for adding the documentation!

> @@ -0,0 +1,33 @@
> +What:		/sys/bus/pci/drivers/.../gpu_health
> +Date:		April 2026
> +KernelVersion:	7.0
> +Contact:	intel-xe@lists.freedesktop.org
> +Description:
> +		This file exposes the current GPU health state and, for Physical
> +		Functions (PFs), allows GPU health state to be updated.
> +
> +		This sysfs file is only accessible to administrative users and is
> +		present only on Intel Xe platforms that support the GPU health
> +		indicator interface for RAS.
> +
> +		For Physical Functions (PFs), the file is read-write, while for
> +		Virtual Functions (VFs), it is read-only and does not support GPU
> +		health state updates.
> +
> +		Read return a single line containing one of the valid values for

/Read/Reads/ or /Read return/A read returns/

> +		the current device health state. Only for PFs, writing one of the
> +		valid values updates the current device health state.

...

> +static const char * const gpu_health_states[] = {
> +	[XE_RAS_HEALTH_STATUS_OK]		= "ok",
> +	[XE_RAS_HEALTH_STATUS_WARNING]		= "warning",
> +	[XE_RAS_HEALTH_STATUS_CRITICAL]		= "critical"
> +};

Thanks for making it one word, it makes much more sense to me.

...

> +static ssize_t gpu_health_show(struct device *dev, struct device_attribute *attr, char *buf)
> +{
> +	struct xe_device *xe = kdev_to_xe_device(dev);
> +	struct xe_sysctrl_mailbox_command command = {0};
> +	struct xe_ras_health_get_response response = {0};
> +	struct xe_ras_health_get_input request = {0};
> +	enum xe_sysctrl_mailbox_command_id cmd = XE_SYSCTRL_CMD_GET_HEALTH;

do we need 'cmd' here?

> +	enum xe_ras_health_status health;
> +	int ret;
> +	size_t rlen = 0;
> +
> +	prepare_sysctrl_command(&command, cmd, &request,
> +				sizeof(request), &response, sizeof(response));
> +	guard(xe_pm_runtime)(xe);
> +	ret = xe_sysctrl_send_command(&xe->sc, &command, &rlen);
> +	if (ret)
> +		return ret;
> +
> +	if (rlen != sizeof(response)) {
> +		xe_err(xe,
> +		       "[RAS][GET_HEALTH]: invalid Sysctrl response length %zu (expected %zu)\n",
> +		       rlen, sizeof(response));
> +		return -EPROTO;
> +	}
> +	if (response.current_health > XE_RAS_HEALTH_STATUS_CRITICAL) {
> +		xe_err(xe, "[RAS][GET_HEALTH]: invalid health state %u from Sysctrl\n",
> +		       response.current_health);
> +		return -EPROTO;
> +	}
> +
> +	health = (enum xe_ras_health_status)response.current_health;
> +
> +	xe_dbg(xe, "[RAS][GET_HEALTH]: current GPU health state = %d (%s)\n",
> +	       health, gpu_health_states[health]);
> +
> +	return sysfs_emit(buf, "%s\n", gpu_health_states[health]);
> +}
> +
> +static ssize_t gpu_health_store(struct device *dev, struct device_attribute *attr,
> +				const char *buf, size_t count)
> +{
> +	struct xe_device *xe = kdev_to_xe_device(dev);
> +	struct xe_sysctrl_mailbox_command command = {0};
> +	struct xe_ras_health_set_input request = {0};
> +	struct xe_ras_health_set_response response = {0};
> +	enum xe_sysctrl_mailbox_command_id cmd = XE_SYSCTRL_CMD_SET_HEALTH;

do we need 'cmd' here?

Andi

> +	enum xe_ras_health_status health;
> +	int ret;
> +	size_t rlen = 0;
> +	int state;
> +	int ras_status;
> +
> +	state = sysfs_match_string(gpu_health_states,
> +				   buf);
> +	if (state < 0)
> +		return -EINVAL;
> +
> +	request.new_health = (u8)state;
> +
> +	prepare_sysctrl_command(&command, cmd, &request,
> +				sizeof(request), &response, sizeof(response));