Hi,
> Currently only "interrupt-based" errors have their own history,
> however there are some "non-interrupt-based" errors or events
> which need history to improve debugging or help know the health
> status of UFS devices.
>
> Example of fatal errors,
> - Link startup error
> - Suspend error
> - Resume error
>
> Example of abnormal events,
> - Task or request abort
> - Device reset (now equals to Logical Unit Reset)
> - Host reset
>
> This patch tries to track above errors and events by existed UFS error
> history mechanism.
>
> Signed-off-by: Stanley Chu <stanley.chu@xxxxxxxxxxxx>
Reviewed-by: Avri Altman <avri.altman@xxxxxxx>
> /**
> * ufshcd_link_startup - Initialize unipro link startup
> * @hba: per adapter instance
> @@ -4356,6 +4372,8 @@ static int ufshcd_link_startup(struct ufs_hba
> *hba)
>
> /* check if device is detected by inter-connect layer */
> if (!ret && !ufshcd_is_device_present(hba)) {
> + ufshcd_update_reg_hist(&hba-
> >ufs_stats.link_startup_err,
> + 0);
So no device = no error, instead of ENXIO.
Well I guess this is a fair choice,
although less informative if the device stop responding.
