lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Fri, 15 Oct 2021 11:35:15 +0200
From:   Miquel Raynal <miquel.raynal@...tlin.com>
To:     Paul Cercueil <paul@...pouillou.net>
Cc:     Richard Weinberger <richard@....at>,
        Vignesh Raghavendra <vigneshr@...com>,
        Harvey Hunt <harveyhuntnexus@...il.com>, list@...ndingux.net,
        linux-mtd@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-mips@...r.kernel.org, stable@...r.kernel.org
Subject: Re: [PATCH 2/3] mtd: rawnand: Export
 nand_read_page_hwecc_oob_first()

Hi Paul,

> >>  */  
> >>  >> >>   /* An ECC layout for using 4-bit ECC with small-page flash, >> storing  
> >>  >>  @@ -648,7 +580,7 @@ static int davinci_nand_attach_chip(struct >> >> nand_chip *chip)
> >>  >>   			} else if (chunks == 4 || chunks == 8) {
> >>  >>   				mtd_set_ooblayout(mtd,
> >>  >>   						  nand_get_large_page_ooblayout());
> >>  >>  -				chip->ecc.read_page = >> nand_davinci_read_page_hwecc_oob_first;
> >>  >>  +				chip->ecc.read_page = nand_read_page_hwecc_oob_first;
> >>  >>   			} else {
> >>  >>   				return -EIO;
> >>  >>   			}
> >>  >>  diff --git a/drivers/mtd/nand/raw/nand_base.c >> >> b/drivers/mtd/nand/raw/nand_base.c
> >>  >>  index 3d6c6e880520..cb5f343b9fa2 100644
> >>  >>  --- a/drivers/mtd/nand/raw/nand_base.c
> >>  >>  +++ b/drivers/mtd/nand/raw/nand_base.c
> >>  >>  @@ -3160,6 +3160,75 @@ static int nand_read_page_hwecc(struct >> >> nand_chip *chip, uint8_t *buf,
> >>  >>   	return max_bitflips;
> >>  >>   }  
> >>  >> >>  +/**  
> >>  >>  + * nand_read_page_hwecc_oob_first - Hardware ECC page read >> with ECC
> >>  >>  + *                                  data read from OOB area
> >>  >>  + * @chip: nand chip info structure
> >>  >>  + * @buf: buffer to store read data
> >>  >>  + * @oob_required: caller requires OOB data read to >> chip->oob_poi
> >>  >>  + * @page: page number to read
> >>  >>  + *
> >>  >>  + * Hardware ECC for large page chips, require OOB to be read >> >> first. For this  
> >>  >
> >>  > requires
> >>  >
> >>  > With this ECC configuration?
> >>  >  
> >>  >>  + * ECC mode, the write_page method is re-used from ECC_HW. >> These >> methods  
> >>  >
> >>  > I do not understand this sentence nor the next one about >> syndrome. I
> >>  > believe it is related to your engine and should not leak into the >> > core.
> >>  >  
> >>  >>  + * read/write ECC from the OOB area, unlike the >> ECC_HW_SYNDROME >> support with
> >>  >>  + * multiple ECC steps, follows the "infix ECC" scheme and >> >> reads/writes ECC from
> >>  >>  + * the data area, by overwriting the NAND manufacturer bad >> block >> markings.  
> >>  >
> >>  > That's a sentence I don't like. What do you mean exactly?
> >>  >
> >>  > What "Infix ECC" scheme is?
> >>  >
> >>  > Do you mean that unlike the syndrome  mode it *does not* >> overwrite the
> >>  > BBM ?  
> >> >>  I don't mean anything. I did not write that comment. I just moved >> the function verbatim with no changes. If something needs to be >> fixed, then it needs to be fixed before/after this patch.  
> > 
> > Well, this comment should be adapted because as-is I don't think it's
> > wise to move it around.  
> 
> OK.
> 
> I think it says that BBM can be overwritten with this configuration, but that would be if the OOB layout covers the BBM area.

If the ooblayout prevents the BBM to be smatched I'm fine and this
sentence should disappear because it's misleading.

> >> >>  >>  + */  
> >>  >>  +int nand_read_page_hwecc_oob_first(struct nand_chip *chip, >> uint8_t >> *buf,
> >>  >>  +				   int oob_required, int page)
> >>  >>  +{
> >>  >>  +	struct mtd_info *mtd = nand_to_mtd(chip);
> >>  >>  +	int i, eccsize = chip->ecc.size, ret;
> >>  >>  +	int eccbytes = chip->ecc.bytes;
> >>  >>  +	int eccsteps = chip->ecc.steps;
> >>  >>  +	uint8_t *p = buf;
> >>  >>  +	uint8_t *ecc_code = chip->ecc.code_buf;
> >>  >>  +	unsigned int max_bitflips = 0;
> >>  >>  +
> >>  >>  +	/* Read the OOB area first */
> >>  >>  +	ret = nand_read_oob_op(chip, page, 0, chip->oob_poi, >> >> mtd->oobsize);
> >>  >>  +	if (ret)
> >>  >>  +		return ret;
> >>  >>  +
> >>  >>  +	ret = nand_read_page_op(chip, page, 0, NULL, 0);  
> >>  >
> >>  > Definitely not, your are requesting the chip to do the read_page
> >>  > operation twice. You only need a nand_change_read_column I >> believe.  
> >> >>  Again, this code is just being moved around - don't shoot the >> messenger :)  
> > 
> > haha
> > 
> > Well, now you touch the core, so I need to be more careful, and the
> > code is definitely wrong, so even if we don't move that code off, you
> > definitely want to fix it in order to improve your performances.  
> 
> I don't see the read_page being done twice?
> 
> There's one read_oob, one read_page, then read_data in the loop.

read_oob and read_page both end up sending READ0 and READSTART so
they do request the chip to perform an internal read twice. You
need this only once. The call to nand_read_page_op() should be a
nand_change_read_column() with no data requested. 

> >>  >>   /**
> >>  >>    * nand_read_page_syndrome - [REPLACEABLE] hardware ECC >> syndrome >> based page read
> >>  >>    * @chip: nand chip info structure
> >>  >>  diff --git a/include/linux/mtd/rawnand.h >> >> b/include/linux/mtd/rawnand.h
> >>  >>  index b2f9dd3cbd69..5b88cd51fadb 100644
> >>  >>  --- a/include/linux/mtd/rawnand.h
> >>  >>  +++ b/include/linux/mtd/rawnand.h
> >>  >>  @@ -1539,6 +1539,8 @@ int nand_read_data_op(struct nand_chip >> *chip, >> void *buf, unsigned int len,
> >>  >>   		      bool force_8bit, bool check_only);
> >>  >>   int nand_write_data_op(struct nand_chip *chip, const void *buf,
> >>  >>   		       unsigned int len, bool force_8bit);
> >>  >>  +int nand_read_page_hwecc_oob_first(struct nand_chip *chip, >> uint8_t >> *buf,
> >>  >>  +				   int oob_required, int page);  
> >>  >
> >>  > You certainly want to add this symbol closer to the other >> read/write
> >>  > page helpers?  
> >> >>  Where would that be? The other read/write page helpers are all >> "static" so they don't appear in any header.  
> > 
> > I believe we should keep this header local as long as there are no
> > other users.  
> 
> I'll move it to internal.h then.

Why do you want to put it there is there is only one user?

Thanks,
Miquèl

Powered by blists - more mailing lists