[PATCH v2] drm/xe/guc: Escalate GuC load failure immediately
John Harrison
john.c.harrison at intel.com
Thu Feb 13 22:43:54 UTC 2025
On 2/13/2025 13:39, Daniele Ceraolo Spurio wrote:
> When the Xe was first introduced, we intentionally avoided escalating
> GuC load failures, to not abort mid-probe. Xe is now mature enough and
> we gracefully handle probe failures, so we can start escalating
> immediately.
>
> Note that even without this patch the probe is still aborted because
> the attempt to enable CTs after GuC load will fail and that failure is
> already escalated.
>
> v2: only return success if the load_done variable is 1 (Tejas)
>
> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
> Cc: Matthew Brost <matthew.brost at intel.com>
> Cc: John Harrison <John.C.Harrison at Intel.com>
> Reviewed-by: Tejas Upadhyay <tejas.upadhyay at intel.com>
> ---
> drivers/gpu/drm/xe/xe_guc.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_guc.c b/drivers/gpu/drm/xe/xe_guc.c
> index 1619c0a52db9..407399cc4764 100644
> --- a/drivers/gpu/drm/xe/xe_guc.c
> +++ b/drivers/gpu/drm/xe/xe_guc.c
> @@ -938,7 +938,7 @@ static s32 guc_pc_get_cur_freq(struct xe_guc_pc *guc_pc)
> #endif
> #define GUC_LOAD_TIME_WARN_MS 200
>
> -static void guc_wait_ucode(struct xe_guc *guc)
> +static int guc_wait_ucode(struct xe_guc *guc)
Should we add an 'ALLOW_ERROR_INJECTION' on this? Make sure CI checks
that GuC load failures are handled gracefully?
Either way, this change as is looks good now.
Reviewed-by: John Harrison <John.C.Harrison at Intel.com>
John.
> {
> struct xe_gt *gt = guc_to_gt(guc);
> struct xe_mmio *mmio = >->mmio;
> @@ -1045,6 +1045,8 @@ static void guc_wait_ucode(struct xe_guc *guc)
> delta_ms, xe_guc_pc_get_act_freq(guc_pc), guc_pc_get_cur_freq(guc_pc),
> before_freq, status, count);
> }
> +
> + return load_done == 1 ? 0 : -EIO;
> }
>
> static int __xe_guc_upload(struct xe_guc *guc)
> @@ -1077,14 +1079,16 @@ static int __xe_guc_upload(struct xe_guc *guc)
> goto out;
>
> /* Wait for authentication */
> - guc_wait_ucode(guc);
> + ret = guc_wait_ucode(guc);
> + if (ret)
> + goto out;
>
> xe_uc_fw_change_status(&guc->fw, XE_UC_FIRMWARE_RUNNING);
> return 0;
>
> out:
> xe_uc_fw_change_status(&guc->fw, XE_UC_FIRMWARE_LOAD_FAIL);
> - return 0 /* FIXME: ret, don't want to stop load currently */;
> + return ret;
> }
>
> static int vf_guc_min_load_for_hwconfig(struct xe_guc *guc)
More information about the Intel-xe
mailing list