Re: [PATCH v4 10/10] scsi: ufs: Apply more limitations to user access

[Can Guo]

On 2021-06-25 06:25, Bart Van Assche wrote:
> On 6/23/21 7:23 PM, Can Guo wrote:
> > On 2021-06-24 05:51, Bart Van Assche wrote:
> > > On 6/23/21 12:35 AM, Can Guo wrote:
> > > - During system suspend, user space software is paused before the 
> > > device
> > >   driver freeze callbacks are invoked. Hence, the 
> > > hba->is_sys_suspended
> > >   check can be left out.
> >
> > is_sys_suspended indicates that system resume failed (power/clk is 
> > OFF).
> >
> > > - If a LUN is runtime suspended, it should be resumed if accessed 
> > > from
> > >   user space instead of failing user space accesses. In other words, 
> > > the
> > >   hba->is_wlu_sys_suspended check seems inappropriate to me.
> >
> > hba->is_wlu_sys_suspended indicates that wl system resume failed, 
> > device
> > is not operational.
>
> Hi Can,
>
> Thanks for the clarification. How about converting the above two 
> answers
> into comments inside ufshcd_is_user_access_allowed()?

Sure.

> Should ufshcd_is_user_access_allowed() perhaps be called after
> ufshcd_rpm_get_sync() instead of before to prevent that the value of
> hba->is_sys_suspended or hba->is_wlu_sys_suspended changes after having
> been checked and before the UFS device is accessed?

is_sys_suspended and is_wlu_sys_suspended only represent system PM 
status,
not runtime PM status.

My understanding is that user threads are frozen before system PM 
starts,
so it does not matter we call  ufshcd_is_user_access_allowed() before or
after ufshcd_rpm_get_sync().

If my understanding is wrong, then we need to either call 
lock_system_sleep()
in get_user_access() or wrap 
ufshcd_suspend_prepare/ufshcd_resume_complete()
with host_sem.

Thanks,

Can Guo.

> Thanks,
>
> Bart.