On 06/08/24 10:02, Raag Jadav wrote:
This change was originally sent by Himal, so may be you should keep his authorship.
> From: Lucas De Marchi <lucas.demarchi@xxxxxxxxx>
>
> Bring back uevent for gt reset failure with better uapi naming.
> With this in place we can receive failure event using udev.
>
> $ udevadm monitor --property --kernel
> monitor will print the received events for:
> KERNEL - the kernel uevent
>
> KERNEL[871.188570] change /devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:01.0/0000:03:00.0 (pci)
> ACTION=change
> DEVPATH=/devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:01.0/0000:03:00.0
> SUBSYSTEM=pci
> DEVICE_STATUS=NEEDS_RESET
> REASON=GT_RESET_FAILED
> TILE_ID=0
> GT_ID=0
> DRIVER=xe
> PCI_CLASS=30000
> PCI_ID=8086:56B1
> PCI_SUBSYS_ID=8086:1210
> PCI_SLOT_NAME=0000:03:00.0
> MODALIAS=pci:v00008086d000056B1sv00008086sd00001210bc03sc00i00
> SEQNUM=6104
>
> Signed-off-by: Lucas De Marchi <lucas.demarchi@xxxxxxxxx>
> Signed-off-by: Raag Jadav <raag.jadav@xxxxxxxxx>
> ---
> drivers/gpu/drm/xe/xe_gt.c | 27 +++++++++++++++++++++++++--
> include/uapi/drm/xe_drm.h | 17 +++++++++++++++++
> 2 files changed, 42 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c
> index b04e47186f5b..5ceef0059861 100644
> --- a/drivers/gpu/drm/xe/xe_gt.c
> +++ b/drivers/gpu/drm/xe/xe_gt.c
> @@ -740,6 +740,30 @@ static int do_gt_restart(struct xe_gt *gt)
> return 0;
> }
>
> +static void xe_uevent_gt_reset_failure(struct pci_dev *pdev, u8 tile_id, u8 gt_id)
> +{
> + char *reset_event[5];
> +
> + reset_event[0] = DRM_XE_RESET_REQUIRED_UEVENT;
> + reset_event[1] = DRM_XE_RESET_REQUIRED_UEVENT_REASON_GT;
> + reset_event[2] = kasprintf(GFP_KERNEL, "TILE_ID=%d", tile_id);
> + reset_event[3] = kasprintf(GFP_KERNEL, "GT_ID=%d", gt_id);
> + reset_event[4] = NULL;
> + kobject_uevent_env(&pdev->dev.kobj, KOBJ_CHANGE, reset_event);
> +
> + kfree(reset_event[2]);
> + kfree(reset_event[3]);
> +}
> +
> +static void gt_reset_failed(struct xe_gt *gt, int err)
> +{
> + xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err));
> +
> + /* Notify userspace about gt reset failure */
> + xe_uevent_gt_reset_failure(to_pci_dev(gt_to_xe(gt)->drm.dev),
> + gt_to_tile(gt)->id, gt->info.id);
> +}
> +
> static int gt_reset(struct xe_gt *gt)
> {
> int err;
> @@ -795,8 +819,7 @@ static int gt_reset(struct xe_gt *gt)
> XE_WARN_ON(xe_uc_start(>->uc));
> xe_pm_runtime_put(gt_to_xe(gt));
> err_fail:
> - xe_gt_err(gt, "reset failed (%pe)\n", ERR_PTR(err));
> -
> + gt_reset_failed(gt, err);
> xe_device_declare_wedged(gt_to_xe(gt));
Also, we might want to have a RESET_REQUIRED event whenever
device is wedged.
Thanks,
Aravind.
>
> return err;
> diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h
> index 19619d4952a8..9ea3be97535e 100644
> --- a/include/uapi/drm/xe_drm.h
> +++ b/include/uapi/drm/xe_drm.h
> @@ -20,6 +20,7 @@ extern "C" {
> * 2. Extension definition and helper structs
> * 3. IOCTL's Query structs in the order of the Query's entries.
> * 4. The rest of IOCTL structs in the order of IOCTL declaration.
> + * 5. uEvents
> */
>
> /**
> @@ -1686,6 +1687,22 @@ struct drm_xe_oa_stream_info {
> __u64 reserved[3];
> };
>
> +/**
> + * DOC: uevent generated by xe on it's pci node.
> + *
> + * DRM_XE_RESET_REQUIRED_UEVENT - Event is generated when device needs reset.
> + * The REASON is provided along with the event for which reset is required.
> + * On the basis of REASONS, additional information might be supplied.
> + */
> +#define DRM_XE_RESET_REQUIRED_UEVENT "DEVICE_STATUS=NEEDS_RESET"
> +
> +/**
> + * DRM_XE_RESET_REQUIRED_UEVENT_REASON_GT - Reason provided to DRM_XE_RESET_REQUIRED_UEVENT
> + * incase of gt reset failure. The additional information supplied is tile id and
> + * gt id of the gt unit for which reset has failed.
> + */
> +#define DRM_XE_RESET_REQUIRED_UEVENT_REASON_GT "REASON=GT_RESET_FAILED"
> +
> #if defined(__cplusplus)
> }
> #endif
