RE: [PATCH v3] scsi: ufs: core: handle PM SSU timeout before SCSI EH

["Fang Hongjie(方洪杰)" <hongjiefang@asrmicro.com>]

> From: Bart Van Assche [mailto:bvanassche@acm.org]
> Sent: Tuesday, June 2, 2026 1:42 AM
> To: Fang Hongjie(方洪杰) <hongjiefang@asrmicro.com>;
> alim.akhtar@samsung.com; avri.altman@wdc.com;
> James.Bottomley@HansenPartnership.com; martin.petersen@oracle.com;
> peter.wang@mediatek.com; beanhuo@micron.com
> Cc: linux-scsi@vger.kernel.org; linux-kernel@vger.kernel.org
> Subject: Re: [PATCH v3] scsi: ufs: core: handle PM SSU timeout before SCSI
> EH
> 
> On 5/31/26 9:56 PM, Hongjie Fang wrote:
> > @@ -9466,7 +9499,7 @@ static enum scsi_timeout_action
> ufshcd_eh_timed_out(struct scsi_cmnd *scmd)
> >   {
> >   	struct ufs_hba *hba = shost_priv(scmd->device->host);
> >
> > -	if (!hba->system_suspending) {
> > +	if (!hba->pm_op_in_progress || !ufshcd_is_scsi_cmd(scmd)) {
> >   		/* Activate the error handler in the SCSI core. */
> >   		return SCSI_EH_NOT_HANDLED;
> >   	}
> 
> I think we want to avoid SCSI error handler activation for all commands
> that time out while a power management operation is in progress,
> including device management commands.
> 
> Please consider completing commands from inside the timeout handler
> instead of adding force_compl support for the legacy single doorbell
> mode. The very lightly tested patch below should realize this.
> 

Thanks, this makes sense.
I will rework the patch in that direction for v4. It will use
hba->pm_op_in_progress for all PM timeout commands and no longer
relay on the extra legacy single-doorbell force-completion path.

For regular SCSI commands it releases the SCSI command resources before
scsi_done(). For internal device-management commands it only
completes the request, since those commands do not use the
ufshcd_release_scsi_cmd() lifetime rules.



Best.