From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from down.free-electrons.com ([37.187.137.238] helo=mail.free-electrons.com) by bombadil.infradead.org with esmtp (Exim 4.80.1 #2 (Red Hat Linux)) id 1auxZQ-00083u-HD for linux-mtd@lists.infradead.org; Tue, 26 Apr 2016 07:38:09 +0000 Date: Tue, 26 Apr 2016 09:37:36 +0200 From: Boris Brezillon To: Brian Norris , =?UTF-8?B?UmFmYcWCIE1pxYJl?= =?UTF-8?B?Y2tp?= Cc: linux-mtd@lists.infradead.org, Kamal Dasu , Richard Weinberger , David Woodhouse , "open list:BROADCOM STB NAND FLASH DRIVER" , open list Subject: Re: [PATCH 3/3] mtd: brcmnand: respect ECC algorithm set by NAND subsystem Message-ID: <20160426093736.3978a718@bbrezillon> In-Reply-To: <20160426055355.GA25981@localhost> References: <1461324197-1333-1-git-send-email-zajec5@gmail.com> <1461324197-1333-3-git-send-email-zajec5@gmail.com> <20160426055355.GA25981@localhost> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Hi Brian, On Mon, 25 Apr 2016 22:53:55 -0700 Brian Norris wrote: > On Fri, Apr 22, 2016 at 01:23:15PM +0200, Rafa=C5=82 Mi=C5=82ecki wrote: > > It's more reliable than guessing based on ECC strength. It allows using > > NAND on devices with BCH-1 (e.g. D-Link DIR-885L). > >=20 > > Signed-off-by: Rafa=C5=82 Mi=C5=82ecki > > --- > > drivers/mtd/nand/brcmnand/brcmnand.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > >=20 > > diff --git a/drivers/mtd/nand/brcmnand/brcmnand.c b/drivers/mtd/nand/br= cmnand/brcmnand.c > > index c3331ff..dcb22dc 100644 > > --- a/drivers/mtd/nand/brcmnand/brcmnand.c > > +++ b/drivers/mtd/nand/brcmnand/brcmnand.c > > @@ -1927,7 +1927,7 @@ static int brcmnand_setup_dev(struct brcmnand_hos= t *host) > > =20 > > switch (chip->ecc.size) { > > case 512: > > - if (chip->ecc.strength =3D=3D 1) /* Hamming */ > > + if (chip->ecc.algo =3D=3D NAND_ECC_HAMMING) =20 >=20 > This doesn't handle most of the problems I noted on the early version of > this series. (But thank you for following through on the algorithm > selection refactoring!) >=20 > Particularly, this change will > (a) break any existing DTs which used to have 'nand-ecc-size =3D <1>', and > would assume this gets Hamming ECC; and > (b) allows DTs to specify Hamming ECC for unsupported modes, like 1024B > sectors, or ecc_level !=3D 1. None of these are supported in HW. Indeed. >=20 > > cfg->ecc_level =3D 15; > > else > > cfg->ecc_level =3D chip->ecc.strength; =20 >=20 > Something like the following probably works better (not tested): >=20 > ---8<--- >=20 > From: Brian Norris > Date: Mon, 25 Apr 2016 20:48:02 -0700 > Subject: [PATCH] mtd: brcmnand: respect ECC algorithm set by the NAND > subsystem >=20 > This is more obvious than guessing based on ECC strength. It allows > using NAND on devices with BCH-1 (e.g. D-Link DIR-885L). >=20 > This maintains DT backward compatibility by defaulting to Hamming if a > 1-bit ECC algorithm is specified without a corresponding algorithm > selection. i.e., to use BCH-1, you must specify: >=20 > nand-ecc-strength =3D <1>; > nand-ecc-step-size =3D <512>; > nand-ecc-algo =3D "bch"; >=20 > Also adds a check to ensure we haven't allowed someone to get by with SW > ECC. If we want to support SW ECC, we need to refactor some other pieces > of this driver. I'm waiting a bit before applying this patch (I'd like to have a Tested-by first). >=20 > Signed-off-by: Brian Norris > --- > drivers/mtd/nand/brcmnand/brcmnand.c | 24 +++++++++++++++++++++++- > 1 file changed, 23 insertions(+), 1 deletion(-) >=20 > diff --git a/drivers/mtd/nand/brcmnand/brcmnand.c b/drivers/mtd/nand/brcm= nand/brcmnand.c > index c3331ffcaffd..b76ad7c0144f 100644 > --- a/drivers/mtd/nand/brcmnand/brcmnand.c > +++ b/drivers/mtd/nand/brcmnand/brcmnand.c > @@ -1925,9 +1925,31 @@ static int brcmnand_setup_dev(struct brcmnand_host= *host) > cfg->col_adr_bytes =3D 2; > cfg->blk_adr_bytes =3D get_blk_adr_bytes(mtd->size, mtd->writesize); > =20 > + if (chip->ecc.mode !=3D NAND_ECC_HW) { > + dev_err(ctrl->dev, "only HW ECC supported; selected: %d\n", > + chip->ecc.mode); > + return -EINVAL; > + } > + > + if (chip->ecc.algo =3D=3D NAND_ECC_UNKNOWN) { > + if (chip->ecc.strength =3D=3D 1 && chip->ecc.size =3D=3D 512) > + /* Default to Hamming for 1-bit ECC, if unspecified */ > + chip->ecc.algo =3D NAND_ECC_HAMMING; > + else > + /* Otherwise, BCH */ > + chip->ecc.algo =3D NAND_ECC_BCH; > + } > + > + if (chip->ecc.algo =3D=3D NAND_ECC_HAMMING && (chip->ecc.strength !=3D = 1 || > + chip->ecc.size !=3D 512)) { > + dev_err(ctrl->dev, "invalid Hamming params: %d bits per %d bytes\n", > + chip->ecc.strength, chip->ecc.size); > + return -EINVAL; > + } > + > switch (chip->ecc.size) { > case 512: > - if (chip->ecc.strength =3D=3D 1) /* Hamming */ > + if (chip->ecc.algo =3D=3D NAND_ECC_HAMMING) > cfg->ecc_level =3D 15; > else > cfg->ecc_level =3D chip->ecc.strength; --=20 Boris Brezillon, Free Electrons Embedded Linux and Kernel engineering http://free-electrons.com