RE: [PATCH v6 05/10] scsi: ufs: add quirk to fix abnormal ocs fatal error

[Avri Altman <Avri.Altman@wdc.com>]

> 
> > -----Original Message-----
> > From: Avri Altman <Avri.Altman@wdc.com>
> > Sent: Monday, April 20, 2020 5:56 PM
> > To: Alim Akhtar <alim.akhtar@samsung.com>; robh@kernel.org
> > Cc: devicetree@vger.kernel.org; linux-scsi@vger.kernel.org;
> > krzk@kernel.org; martin.petersen@oracle.com;
> kwmad.kim@samsung.com;
> > stanley.chu@mediatek.com; cang@codeaurora.org; linux-samsung-
> > soc@vger.kernel.org; linux-arm-kernel@lists.infradead.org; linux-
> > kernel@vger.kernel.org
> > Subject: RE: [PATCH v6 05/10] scsi: ufs: add quirk to fix abnormal ocs
> > fatal error
> >
> > >
> > > From: Kiwoong Kim <kwmad.kim@samsung.com>
> > >
> > > Some architectures determines if fatal error for OCS occurrs to check
> > > status in response upiu. This patch
> > Typo - occurs
> >
> > > is to prevent from reporting command results with that.
> > >
> > > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > > Signed-off-by: Alim Akhtar <alim.akhtar@samsung.com>
> > > ---
> > >  drivers/scsi/ufs/ufshcd.c | 6 ++++++
> > >  drivers/scsi/ufs/ufshcd.h | 6 ++++++
> > >  2 files changed, 12 insertions(+)
> > >
> > > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > > index b32fcedcdcb9..8c07caff0a5c 100644
> > > --- a/drivers/scsi/ufs/ufshcd.c
> > > +++ b/drivers/scsi/ufs/ufshcd.c
> > > @@ -4794,6 +4794,12 @@ ufshcd_transfer_rsp_status(struct ufs_hba
> *hba,
> > > struct ufshcd_lrb *lrbp)
> > >         /* overall command status of utrd */
> > >         ocs = ufshcd_get_tr_ocs(lrbp);
> > >
> > > +       if (hba->quirks & UFSHCD_QUIRK_BROKEN_OCS_FATAL_ERROR) {
> > > +               if (be32_to_cpu(lrbp->ucd_rsp_ptr->header.dword_1) &
> > > +                                       MASK_RSP_UPIU_RESULT)
> > > +                       ocs = OCS_SUCCESS;
> > > +       }
> > > +
> > Not sure that I follow what this quirk is all about.
> > Your code overrides ocs by open coding ufshcd_get_rsp_upiu_result.
> >
> > Normally OCS is in utp transfer req descriptor, dword 2, bits 0..7.
> > My understanding from your description, is that some fatal error might
> > occur, But the host controller does not report it, and it still needs to
> > be checked in the response upiu.
> > Evidently you are not doing so.
> > Please elaborate your description.
> >
> > P.S.
> > The ocs is being evaluated in device management commands as well, Isn't
> > this something you need to attend?
> >
> > Thanks,
> > Avri
> >
> > >         switch (ocs) {
> > >         case OCS_SUCCESS:
> > >                 result = ufshcd_get_req_rsp(lrbp->ucd_rsp_ptr);
> > > diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h
> > > index a9b9ace9fc72..e1d09c2c4302 100644
> > > --- a/drivers/scsi/ufs/ufshcd.h
> > > +++ b/drivers/scsi/ufs/ufshcd.h
> > > @@ -541,6 +541,12 @@ enum ufshcd_quirks {
> > >          * resolution of the values of PRDTO and PRDTL in UTRD as byte.
> > >          */
> > >         UFSHCD_QUIRK_PRDT_BYTE_GRAN                     = 1 << 9,
> > > +
> > > +       /*
> > > +        * This quirk needs to be enabled if the host controller reports
> > > +        * OCS FATAL ERROR with device error through sense data
> > > +        */
> > > +       UFSHCD_QUIRK_BROKEN_OCS_FATAL_ERROR             = 1 << 10,
> > >  };
> > >
> > >  enum ufshcd_caps {
> > > --
> > > 2.17.1
> Avri
> 
> As specified in the spec, OCS isn't supposed to refer to the contents of
> RESPONSE UPIU.
> But, Exynos host behaves like that in some cases, e.g. a value of 'state' in is
> isn't GOOD(00h).
OK.
I still think that you might consider rewording your commit, explaining this quirk better.
Specifically you might not want to say "if fatal..." because fatal code (0x7) is just one error code out of many.
Also you might want to use ufshcd_get_rsp_upiu_result() in the quirk body instead of open coding it.

> 
> For QUERY RESPONSE, its offset, i.e. " dword_1" is reserved, so currently no
> impact, I think.
> But if you feel another condition is necessary to identify if this request is
> QUERY REQEUST or not, we can add more.
No need, as long as you are ok with whatever ufshcd_get_tr_ocs() returns in ufshcd_wait_for_dev_cmd().

Thanks,
Avri

> 
> Thanks