Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   "Kiwoong Kim" <kwmad.kim@samsung.com>
To:     "'Avri Altman'" <Avri.Altman@wdc.com>,
        "'Alim Akhtar'" <alim.akhtar@samsung.com>, <robh@kernel.org>,
        <cpgs@samsung.com>
Cc:     <devicetree@vger.kernel.org>, <linux-scsi@vger.kernel.org>,
        <krzk@kernel.org>, <martin.petersen@oracle.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>
In-Reply-To: <SN6PR04MB46408CF4DD05DB9B48DFE412FCD40@SN6PR04MB4640.namprd04.prod.outlook.com>
Subject: RE: [PATCH v6 05/10] scsi: ufs: add quirk to fix abnormal ocs fatal
 error
Date:   Tue, 21 Apr 2020 16:46:16 +0900
Message-ID: <1288235781.41587455282171.JavaMail.epsvc@epcpadp2>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Thread-Index: AQJQDPill8kAmUENoj8JBCZ49rLenQIFk/WPAcfxkaoBmLD4p6dkFifQ
Content-Language: ko
Content-Type: text/plain; charset="utf-8"
References: <20200417175944.47189-1-alim.akhtar@samsung.com>
        <CGME20200417181016epcas5p2ee7ac86d743ceee9076690dc5b1e2f08@epcas5p2.samsung.com>
        <20200417175944.47189-6-alim.akhtar@samsung.com>
        <SN6PR04MB46408CF4DD05DB9B48DFE412FCD40@SN6PR04MB4640.namprd04.prod.outlook.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

> -----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
>=20
> >
> > 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
>=20
> > 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 =3D 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 =3D 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.
>=20
> 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.
>=20
> P.S.
> The ocs is being evaluated in device management commands as well, Isn't
> this something you need to attend?
>=20
> Thanks,
> Avri
>=20
> >         switch (ocs) {
> >         case OCS_SUCCESS:
> >                 result =3D 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                     =3D 1 << 9,
> > +
> > +       /*
> > +        * This quirk needs to be enabled if the host controller report=
s
> > +        * OCS FATAL ERROR with device error through sense data
> > +        */
> > +       UFSHCD_QUIRK_BROKEN_OCS_FATAL_ERROR             =3D 1 << 10,
> >  };
> >
> >  enum ufshcd_caps {
> > --
> > 2.17.1
Avri

As specified in the spec, OCS isn't supposed to refer to the contents of RE=
SPONSE UPIU.
But, Exynos host behaves like that in some cases, e.g. a value of 'state' i=
n is isn't GOOD(00h).

For QUERY RESPONSE, its offset, i.e. " dword_1" is reserved, so currently n=
o 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.

Thanks