From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <1496200531.492.0.camel@mhfsdcap03> Subject: Re: [PATCH v3 3/4] mtd: nand: mediatek: add support for different MTK NAND FLASH Controller IP From: xiaolei li To: Boris Brezillon CC: , , , , , , , , Date: Wed, 31 May 2017 11:15:31 +0800 In-Reply-To: <20170529184842.06ec8275@bbrezillon> References: <1494851979-14416-1-git-send-email-xiaolei.li@mediatek.com> <1494851979-14416-4-git-send-email-xiaolei.li@mediatek.com> <20170529184842.06ec8275@bbrezillon> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit MIME-Version: 1.0 List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Hi Boris, On Mon, 2017-05-29 at 18:48 +0200, Boris Brezillon wrote: > Hi Xiaolei, > > On Mon, 15 May 2017 20:39:38 +0800 > Xiaolei Li wrote: > > > ECC strength and spare size supported may be different among MTK NAND > > FLASH Controller IPs. > > > > This patch contains changes as following: > > (1) add new struct mtk_nfc_caps to support different spare size. > > (2) add new struct mtk_ecc_caps to support different ecc strength. > > (3) remove ECC_CNFG_xBIT define, use a for loop to do ecc strength config. > > (4) remove PAGEFMT_SPARE_ define, use a for loop to do spare format config. > > (5) malloc ecc->eccdata buffer according to max ecc strength of this IP. > > > > Signed-off-by: Xiaolei Li > > --- > > drivers/mtd/nand/mtk_ecc.c | 181 ++++++++++++++++++-------------------------- > > drivers/mtd/nand/mtk_ecc.h | 2 +- > > drivers/mtd/nand/mtk_nand.c | 175 +++++++++++++++++++++--------------------- > > 3 files changed, 166 insertions(+), 192 deletions(-) > > > > diff --git a/drivers/mtd/nand/mtk_ecc.c b/drivers/mtd/nand/mtk_ecc.c > > index dbf2562..ae51303 100644 > > --- a/drivers/mtd/nand/mtk_ecc.c > > +++ b/drivers/mtd/nand/mtk_ecc.c > > @@ -33,26 +33,6 @@ > > [...] > > > -static void mtk_nfc_set_spare_per_sector(u32 *sps, struct mtd_info *mtd) > > +static int cmp_map(const void *m0, const void *m1) > > +{ > > + return *(u8 *)m0 > *(u8 *)m1 ? 1 : -1; > > +} > > + > > +static int mtk_nfc_set_spare_per_sector(u32 *sps, struct mtd_info *mtd) > > { > > struct nand_chip *nand = mtd_to_nand(mtd); > > - u32 spare[] = {16, 26, 27, 28, 32, 36, 40, 44, > > - 48, 49, 50, 51, 52, 62, 63, 64}; > > + struct mtk_nfc *nfc = nand_get_controller_data(nand); > > + u8 *spare; > > u32 eccsteps, i; > > > > + spare = vmalloc(nfc->caps->num_spare_size); > > + if (!spare) > > + return -ENOMEM; > > + > > + memcpy(spare, nfc->caps->spare_size, nfc->caps->num_spare_size); > > + > > + /* > > + * Spare size array of some IP may be not sorted, because the spare > > + * size does not always increase as the spare size bitfiled value of > > + * register NFI_PAGEFMT. So, sort it here. > > + */ > > + sort(spare, nfc->caps->num_spare_size, sizeof(u8), cmp_map, NULL); > > + > > Let's avoid this dup+sort dance. If it's not sorted, we'd better > iterate over all entries of the table and pick the best one (the array > is rather small, so it shouldn't be a problem). > OK. will remove sort next version. > > eccsteps = mtd->writesize / nand->ecc.size; > > *sps = mtd->oobsize / eccsteps; > > > > if (nand->ecc.size == 1024) > > *sps >>= 1; > > > > - for (i = 0; i < ARRAY_SIZE(spare); i++) { > > + for (i = 0; i < nfc->caps->num_spare_size; i++) { > > if (*sps <= spare[i]) { > > if (!i) > > *sps = spare[i]; >