About ufshcd_err_handling_unprepare

About ufshcd_err_handling_unprepare

[Kiwoong Kim]

Dear all

I have one question about ufshcd_clear_ua_wluns in ufshcd_err_handling_unprepare.
You probably know a scsi command (request sense) is issued in there to clear UAC for W-LUs.

Let's think about a situation that a read command is timed-out.
And then scmd_eh_abort_handler is called, shost's state is transitioned to SHOST_RECOVERY and scsi_sh is waken up.
If this is the case that the scsi_eh goes up to eh_host_reset_handler,
ufshcd_eh_host_reset_handler queues ufshcd_err_handler and waits for its completion.
And this function can call ufshcd_err_handling_unprepare at the end.

But I think, at this time, the scsi command, i.e. request sense, could not be dispatched because of the shost's state.
Is it needed to be fixed or did I miss something?

Thanks.
Kiwoong Kim

Re: About ufshcd_err_handling_unprepare

[Bart Van Assche]

On 9/27/21 12:39 AM, Kiwoong Kim wrote:
> I have one question about ufshcd_clear_ua_wluns in ufshcd_err_handling_unprepare.
> You probably know a scsi command (request sense) is issued in there to clear UAC for W-LUs.
> 
> Let's think about a situation that a read command is timed-out.
> And then scmd_eh_abort_handler is called, shost's state is transitioned to SHOST_RECOVERY and scsi_sh is waken up.
> If this is the case that the scsi_eh goes up to eh_host_reset_handler,
> ufshcd_eh_host_reset_handler queues ufshcd_err_handler and waits for its completion.
> And this function can call ufshcd_err_handling_unprepare at the end.
> 
> But I think, at this time, the scsi command, i.e. request sense, could not be dispatched because of the shost's state.
> Is it needed to be fixed or did I miss something?

Hi Kiwoong,

Please help with reviewing this patch series since this series should resolve the
issue described above:

https://lore.kernel.org/linux-scsi/20210922093842.18025-1-adrian.hunter@intel.com/

Thanks,

Bart.

RE: About ufshcd_err_handling_unprepare

[Kiwoong Kim]

> Hi Kiwoong,
> 
> Please help with reviewing this patch series since this series should
> resolve the issue described above:
> 
> https://lore.kernel.org/linux-scsi/20210922093842.18025-1-
> adrian.hunter@intel.com/

That's what I'm finding. Thanks.

> 
> Thanks,
> 
> Bart.