From: Scott Wood Date: Mon, 16 Sep 2013 16:14:44 -0500 Subject: [U-Boot] [PATCH v4 1/5] mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform In-Reply-To: <20980858CB6D3A4BAE95CA194937D5E73EA0F496@DBDE04.ent.ti.com> References: <20980858CB6D3A4BAE95CA194937D5E73EA0F496@DBDE04.ent.ti.com> Message-ID: <1379366084.2536.173.camel@snotra.buserror.net> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Fri, 2013-09-13 at 07:02 +0000, Gupta, Pekon wrote: > (resending to u-boot maillist as previous mail was discarded by mailman > due to mail-encoding issue.) > > > On Tue, 2013-09-03 at 11:26 +0530, Pekon Gupta wrote: > > > diff --git a/doc/README.nand b/doc/README.nand > > > index 913e9b5..f72f618 100644 > > > --- a/doc/README.nand > > > +++ b/doc/README.nand > > > @@ -169,6 +169,19 @@ Configuration Options: > > > Please convert your driver even if you don't need the extra > > > flexibility, so that one day we can eliminate the old mechanism. > > > > > > + CONFIG_SYS_NAND_ONFI_DETECTION > > > + Enables detection of ONFI compliant devices during probe. > > > + And fetching device parameters flashed on device, by parsing > > > + ONFI parameter page. > > > > I don't see this used anywhere in the patch (defined in config headers, > > yes, but not ifdeffed). > > > > What does "DETECTION" add here, versus just CONFIG_NAND_ONFI? > > > > It should be CONFIG rather than CONFIG_SYS because it just controls > > whether the support is built, not making an assertion about the > > hardware, right? > > > This is not the new define, it was already there in generic code (nand_base.c) > But not documented, so I documented it and enabled it for am33xx platform > Please refer: $UBOOT/drivers/mtd/nand/nand_base.c > It was added as part of following commit > commit 0272c718ba69c60a9d719db6806971d98db98090 > Author: Florian Fainelli > AuthorDate: 2011-02-25 Oops... I looked for it and didn't see it, but maybe I was looking at the Linux version or something similarly silly. > > > + CONFIG_BCH > > > + Enables software based BCH ECC algorithm present in lib/bch.c > > > + This is used by SoC platforms which do not have in-build hardware > > > + engine to calculate and correct BCH ECC. > > > > s/in-build/a built-in/ > > > > > + CONFIG_SYS_NAND_ECCSCHEME > > > + specifies which ECC scheme to use. > > > > Specifies it how? What are the possible values? > > > > If the answer involves "enum omap_ecc", then OMAP should be > > somewhere in the name. > > > This is generic define which can be used by all NAND drivers to specify > ECC scheme for SPL boot. Therefore CONFIG_SYS_ > How each device/driver will use it depends should be mentioned in > its board level document. For AM335x its updated in Patch: > [PATCH v5 5/5] board/ti/am335x/README: update for NAND boot > Do you want it to be made omap specific ? Yes, it should be omap specific. If the semantics are not generic I don't see the value in making the name be generic and splitting the documentation in two places. > > > +/** > > > + * omap_select_ecc_scheme - configures driver for particular ecc-scheme > > > + * @nand: NAND chip device structure > > > + * @ecc_scheme: ecc scheme to configure > > > + * @pagesize: number of main-area bytes per page of NAND device > > > + * @oobsize: number of OOB/spare bytes per page of NAND device > > > + */ > > > +static int omap_select_ecc_scheme(struct nand_chip *nand, int > > ecc_scheme, > > > + unsigned int pagesize, unsigned int oobsize) { > > > + struct nand_bch_priv *bch = nand->priv; > > > + struct nand_ecclayout *ecclayout = nand->ecc.layout; > > > + int i; > > > + > > > + /* Reset ecc interface */ > > > + nand->ecc.mode = NAND_ECC_NONE; > > > + nand->ecc.read_page = NULL; > > > + nand->ecc.write_page = NULL; > > > + nand->ecc.read_oob = NULL; > > > + nand->ecc.write_oob = NULL; > > > + nand->ecc.hwctl = NULL; > > > + nand->ecc.correct = NULL; > > > + nand->ecc.calculate = NULL; > > > + > > > + switch (ecc_scheme) { > > > + case OMAP_ECC_HAM1_CODE_SW: > > > + debug("nand: selected OMAP_ECC_HAM1_CODE_SW\n"); > > > + nand->ecc.mode = NAND_ECC_SOFT; > > > + nand->ecc.layout = NULL; > > > + nand->ecc.size = 0; > > > + nand->ecc.strength = 1; > > > + bch->ecc_scheme = > > OMAP_ECC_HAM1_CODE_SW; > > > + break; > > > + case OMAP_ECC_HAM1_CODE_HW_ROMCODE: > > > + debug("nand: selected > > OMAP_ECC_HAM1_CODE_HW_ROMCODE\n"); > > > + nand->ecc.mode = NAND_ECC_HW; > > > + nand->ecc.strength = 1; > > > + nand->ecc.size = 512; > > > + nand->ecc.bytes = 3; > > > + nand->ecc.hwctl = omap_enable_hwecc; > > > + nand->ecc.correct = omap_correct_data; > > > + nand->ecc.calculate = omap_calculate_ecc; > > > + /* define ecc-layout */ > > > + ecclayout->eccbytes = nand->ecc.bytes * > > > + (pagesize / SECTOR_BYTES); > > > + for (i = 0; i < ecclayout->eccbytes; i++) > > > + ecclayout->eccpos[i] = i + > > > + > > BADBLOCK_MARKER_LENGTH; > > > + ecclayout->oobfree[0].offset = i + > > > + > > BADBLOCK_MARKER_LENGTH; > > > + ecclayout->oobfree[0].length = oobsize - nand->ecc.bytes - > > > + > > BADBLOCK_MARKER_LENGTH; > > > + bch->ecc_scheme = > > OMAP_ECC_HAM1_CODE_HW_ROMCODE; > > > + break; > > > +#ifdef CONFIG_BCH > > > + case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW: > > > + debug("nand: selected > > OMAP_ECC_BCH8_CODE_HW_DETECTION_SW\n"); > > > + nand->ecc.mode = NAND_ECC_HW; > > > + nand->ecc.strength = 8; > > > + nand->ecc.size = 512; > > > + nand->ecc.bytes = 13; > > > + nand->ecc.hwctl = omap_enable_ecc_bch; > > > + nand->ecc.correct = omap_correct_data_bch_sw; > > > + nand->ecc.calculate = omap_calculate_ecc_bch_sw; > > > + /* BCH SW library is used for error detection */ > > > + bch_priv.control = init_bch(13, 8, 0x201b); > > > + if (!bch_priv.control) { > > > + printf("nand: error: could not init_bch()\n"); > > > + return -ENODEV; > > > + } > > > + /* define ecc-layout */ > > > + ecclayout->eccbytes = nand->ecc.bytes * > > > + (pagesize / SECTOR_BYTES); > > > + for (i = 0; i < ecclayout->eccbytes; i++) > > > + ecclayout->eccpos[i] = i + (oobsize - > > > + ecclayout->eccbytes); > > > + ecclayout->oobfree[0].offset = > > BADBLOCK_MARKER_LENGTH; > > > + ecclayout->oobfree[0].length = oobsize - nand->ecc.bytes - > > > + > > BADBLOCK_MARKER_LENGTH; > > > + omap_hwecc_init_bch(nand, NAND_ECC_READ); > > > + bch->ecc_scheme = > > OMAP_ECC_BCH8_CODE_HW_DETECTION_SW; > > > + break; > > > +#endif > > > + case OMAP_ECC_BCH8_CODE_HW: > > > + debug("nand: selected OMAP_ECC_BCH8_CODE_HW\n"); > > > + nand->ecc.mode = NAND_ECC_HW; > > > + nand->ecc.strength = 8; > > > + nand->ecc.size = 512; > > > + nand->ecc.bytes = 14; > > > + nand->ecc.hwctl = omap_enable_ecc_bch; > > > + nand->ecc.correct = omap_correct_data_bch; > > > + nand->ecc.calculate = omap_calculate_ecc_bch; > > > + /* ELM is used for ECC error detection */ > > > + elm_init(); > > > + /* define ecc-layout */ > > > + ecclayout->eccbytes = nand->ecc.bytes * > > > + (pagesize / SECTOR_BYTES); > > > + for (i = 0; i < ecclayout->eccbytes; i++) > > > + ecclayout->eccpos[i] = i + > > > + > > BADBLOCK_MARKER_LENGTH; > > > + ecclayout->oobfree[0].offset = i + > > > + > > BADBLOCK_MARKER_LENGTH; > > > + ecclayout->oobfree[0].length = oobsize - nand->ecc.bytes - > > > + > > BADBLOCK_MARKER_LENGTH; > > > + bch->ecc_scheme = > > OMAP_ECC_BCH8_CODE_HW; > > > + break; > > > + default: > > > + debug("nand: error: ecc scheme not enabled or > > supported\n"); > > > + return -EINVAL; > > > + } > > > > This will result in NAND_ECC_NONE and likely an eventual NULL pointer > > dereference if a bad value is passed and the -EINVAL return path is > > taken. > > > > OK, I now see that you've got the caller patching this up in the error > > case, but that's awkward and error prone. > > > Sorry, din't understand your feedback here. > omap_select_ecc_scheme() is common function which is used for > - changing ECC scheme if 'nandecc' like utility is needed later > - selecting ECC scheme during probe. > This function populates all necessary fields in struct *nand_chip > for nand driver to work. During probe (both SPL and u-boot), > 'CONFIG_SYS_NAND_ECCSCHEME' is used to determine the ecc-scheme. > > 'default' statement protects the driver from any invalid or un-supported > ECC scheme selected by user in his board file. So that driver probe cleanly > exits probe instead of causing an exception. > Can you please elaborate, what I'm missing here ? Before you get into the switch statement you set ecc.mode to NAND_ECC_NONE and wipe out all of the function pointers. Then, if the switch hits default:, you'll return leaving the ecc in a bad state, relying on the caller to clean it up. > > > @@ -767,67 +811,41 @@ void omap_nand_switch_ecc(uint32_t hardware, > > uint32_t eccstrength) > > > { > > > struct nand_chip *nand; > > > struct mtd_info *mtd; > > > + struct nand_bch_priv *bch; > > > + int err = 0; > > > > > > if (nand_curr_device < 0 || > > > nand_curr_device >= CONFIG_SYS_MAX_NAND_DEVICE || > > > !nand_info[nand_curr_device].name) { > > > - printf("Error: Can't switch ecc, no devices available\n"); > > > + printf("nand: error: no devices available\n"); > > > return; > > > } > > > > Why are you making the error message less specific? > > > There is already another error message in board_nand_init() which will be > flagged when there is problem in selecting ECC. > printf("nand: error: cannot select ecc scheme\n"); > But thanks, you pointed out a bug, I should return -ENODEV here, > Otherwise caller would not know that ecc-scheme could not be changed. > > > Better would be something like: > > printf("%s: no devices available\n", __func__); Still, it's best for error messages to be unique and include the function name. > > > #define CONFIG_SYS_NAND_ECCSIZE 512 > > > #define CONFIG_SYS_NAND_ECCBYTES 13 > > > +#define CONFIG_SYS_NAND_ECCSCHEME 3 > > > +#define CONFIG_SYS_NAND_ONFI_DETECTION > > > > What does 3 mean? Please use a symbolic name. > > > For AM335x this is updated in board/ti/am335x/README as part of > following patch: > [PATCH v5 5/5] board/ti/am335x/README: update for NAND boot Patches are generally reviewed in order. Don't introduce something bad in 1/5 and then fix it in 5/5, even if it's just a style issue that doesn't affect bisect. -Scott