Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 12 Apr 2022 10:28:32 +0200
From:   Miquel Raynal <miquel.raynal@bootlin.com>
To:     Miaoqian Lin <linmq006@gmail.com>
Cc:     Richard Weinberger <richard@nod.at>,
        Vignesh Raghavendra <vigneshr@ti.com>,
        Pratyush Yadav <p.yadav@ti.com>,
        Paul Cercueil <paul@crapouillou.net>,
        Alexandre Belloni <alexandre.belloni@bootlin.com>,
        Bastian Hecht <hechtb@googlemail.com>,
        Artem Bityutskiy <artem.bityutskiy@linux.intel.com>,
        Guennadi Liakhovetski <g.liakhovetski@gmx.de>,
        linux-mtd@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] mtd: rawnand: Fix return value check of
 wait_for_completion_timeout
Message-ID: <20220412102832.09957c35@xps13>
In-Reply-To: <bf603638-cc34-b185-3001-9caeba1884d7@gmail.com>
References: <202204121253.NcZifMQi-lkp@intel.com>
        <20220412063703.8537-1-linmq006@gmail.com>
        <20220412090649.33bb3f8b@xps13>
        <61783400-4df2-47af-78a7-7acb715c3a71@gmail.com>
        <20220412094822.3818ebc2@xps13>
        <bf603638-cc34-b185-3001-9caeba1884d7@gmail.com>
Organization: Bootlin
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi Miaoqian,

linmq006@gmail.com wrote on Tue, 12 Apr 2022 16:23:24 +0800:

> Hi Miquel,
>=20
> On 2022/4/12 15:48, Miquel Raynal wrote:
> > =20
> >>> Hi Miaoqian,
> >>>
> >>> linmq006@gmail.com wrote on Tue, 12 Apr 2022 06:36:52 +0000:
> >>>   =20
> >>>> wait_for_completion_timeout() returns unsigned long not int.
> >>>> It returns 0 if timed out, and positive if completed.
> >>>> The check for <=3D 0 is ambiguous and should be =3D=3D 0 here
> >>>> indicating timeout which is the only error case.
> >>>>
> >>>> Fixes: 83738d87e3a0 ("mtd: sh_flctl: Add DMA capabilty")
> >>>> Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
> >>>> ---
> >>>> change in v2:
> >>>> - initialize ret to 1.
> >>>> ---
> >>>>  drivers/mtd/nand/raw/sh_flctl.c | 8 +++++---
> >>>>  1 file changed, 5 insertions(+), 3 deletions(-)
> >>>>
> >>>> diff --git a/drivers/mtd/nand/raw/sh_flctl.c b/drivers/mtd/nand/raw/=
sh_flctl.c
> >>>> index b85b9c6fcc42..2373251f585b 100644
> >>>> --- a/drivers/mtd/nand/raw/sh_flctl.c
> >>>> +++ b/drivers/mtd/nand/raw/sh_flctl.c
> >>>> @@ -384,7 +384,8 @@ static int flctl_dma_fifo0_transfer(struct sh_fl=
ctl *flctl, unsigned long *buf,
> >>>>  	dma_addr_t dma_addr;
> >>>>  	dma_cookie_t cookie;
> >>>>  	uint32_t reg;
> >>>> -	int ret;
> >>>> +	int ret =3D 1;   =20
> >>> Does not look right. I know this function returns > 0 on positive
> >>> outcomes but this does not make any sense in the first place.   =20
> >> Yes, I made a mistake, Now I realize that in v2, it will return 1 in e=
rror path
> >>
> >> when DMA submit failed. =20
> > Not 1, but a proper error code please (-ETIMEDOUT, -EINVAL, whatever)
> > =20
> >> And for patch v1, it will return 0=C2=A0 if calls wait_for_completion_=
timeout succeeds.
> >> =20
> >>> This function is static and only called twice, please turn it into
> >>> something like:
> >>>
> >>> if (dma_fifo_transfer())
> >>> 	error
> >>> else
> >>> 	ok   =20
> >> So I want to keep ret>0 means success.
> >>
> >> Or could I set ret > 0 after in wait_for_completion_timeout() success =
path?
> >>
> >> like:
> >>
> >> =C2=A0=C2=A0=C2=A0 if(time_left =3D=3D 0)
> >>
> >> =C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0 ret =3D -ETIM=
EDOUT;
> >>
> >> =C2=A0=C2=A0=C2=A0 else
> >>
> >> =C2=A0=C2=A0=C2=A0 =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 =C2=A0 ret =3D 1; =20
> > You can initialize ret to zero at to top. So that anything !=3D 0 is an
> > error (like a lot of functions in the kernel).  =20
>=20
> Thanks for your advice, I will do this.
> > And use:
> >
> > 	if (dma_fifo_transfer())
> > 		error(); =20
> I think keeping the original condition structure is better,
> something like:
>=20
> if (dma_fifo_transfer()=3D=3D0)

  if (cond && cond && !dma_fifo_transfer())

> 	succeed();
>=20
> In this way, only minor changes is needed=E2=80=94=E2=80=94only need to u=
pdate the symbol in condition.
> Otherwise It needs to restructure the code and be more complicated.
>=20
>=20
> Thanks,
>=20
> >> What do you think?
> >>
> >>
> >> Thanks,
> >> =20
> >>>> +	unsigned long time_left;
> >>>> =20
> >>>>  	if (dir =3D=3D DMA_FROM_DEVICE) {
> >>>>  		chan =3D flctl->chan_fifo0_rx;
> >>>> @@ -425,13 +426,14 @@ static int flctl_dma_fifo0_transfer(struct sh_=
flctl *flctl, unsigned long *buf,
> >>>>  		goto out;
> >>>>  	}
> >>>> =20
> >>>> -	ret =3D
> >>>> +	time_left =3D
> >>>>  	wait_for_completion_timeout(&flctl->dma_complete,
> >>>>  				msecs_to_jiffies(3000));
> >>>> =20
> >>>> -	if (ret <=3D 0) {
> >>>> +	if (time_left =3D=3D 0) {
> >>>>  		dmaengine_terminate_all(chan);
> >>>>  		dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
> >>>> +		ret =3D -ETIMEDOUT;
> >>>>  	}
> >>>> =20
> >>>>  out:   =20
> >>> Thanks,
> >>> Miqu=C3=A8l   =20
> >
> > Thanks,
> > Miqu=C3=A8l =20


Thanks,
Miqu=C3=A8l