| Message ID | 20231215123208.516590-3-miquel.raynal@bootlin.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | mtd: rawnand: Fix sequential reads | expand |
On Fri, 2023-12-15 at 12:32:06 UTC, Miquel Raynal wrote: > A couple of reports pointed at some strange failures happening a bit > randomly since the introduction of sequential page reads support. After > investigation it turned out the most likely reason for these issues was > the fact that sometimes a (longer) read might happen, starting at the > same page that was read previously. This is optimized by the raw NAND > core, by not sending the READ_PAGE command to the NAND device and just > reading out the data in a local cache. When this page is also flagged as > being the starting point for a sequential read, it means the page right > next will be accessed without the right instructions. The NAND chip will > be confused and will not output correct data. In order to avoid such > situation from happening anymore, we can however handle this case with a > bit of additional logic, to postpone the initialization of the read > sequence by one page. > > Reported-by: Alexander Shiyan <eagle.alexander923@gmail.com> > Closes: https://lore.kernel.org/linux-mtd/CAP1tNvS=NVAm-vfvYWbc3k9Cx9YxMc2uZZkmXk8h1NhGX877Zg@mail.gmail.com/ > Reported-by: Måns Rullgård <mans@mansr.com> > Closes: https://lore.kernel.org/linux-mtd/yw1xfs6j4k6q.fsf@mansr.com/ > Reported-by: Martin Hundebøll <martin@geanix.com> > Closes: https://lore.kernel.org/linux-mtd/9d0c42fcde79bfedfe5b05d6a4e9fdef71d3dd52.camel@geanix.com/ > Fixes: 003fe4b9545b ("mtd: rawnand: Support for sequential cache reads") > Cc: stable@vger.kernel.org > Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com> Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next. Miquel
diff --git a/drivers/mtd/nand/raw/nand_base.c b/drivers/mtd/nand/raw/nand_base.c index 04e80ace4182..1b0a984d181d 100644 --- a/drivers/mtd/nand/raw/nand_base.c +++ b/drivers/mtd/nand/raw/nand_base.c @@ -3478,6 +3478,18 @@ static void rawnand_enable_cont_reads(struct nand_chip *chip, unsigned int page, rawnand_cap_cont_reads(chip); } +static void rawnand_cont_read_skip_first_page(struct nand_chip *chip, unsigned int page) +{ + if (!chip->cont_read.ongoing || page != chip->cont_read.first_page) + return; + + chip->cont_read.first_page++; + if (chip->cont_read.first_page == chip->cont_read.pause_page) + chip->cont_read.first_page++; + if (chip->cont_read.first_page >= chip->cont_read.last_page) + chip->cont_read.ongoing = false; +} + /** * nand_setup_read_retry - [INTERN] Set the READ RETRY mode * @chip: NAND chip object @@ -3652,6 +3664,8 @@ static int nand_do_read_ops(struct nand_chip *chip, loff_t from, buf += bytes; max_bitflips = max_t(unsigned int, max_bitflips, chip->pagecache.bitflips); + + rawnand_cont_read_skip_first_page(chip, page); } readlen -= bytes;
A couple of reports pointed at some strange failures happening a bit randomly since the introduction of sequential page reads support. After investigation it turned out the most likely reason for these issues was the fact that sometimes a (longer) read might happen, starting at the same page that was read previously. This is optimized by the raw NAND core, by not sending the READ_PAGE command to the NAND device and just reading out the data in a local cache. When this page is also flagged as being the starting point for a sequential read, it means the page right next will be accessed without the right instructions. The NAND chip will be confused and will not output correct data. In order to avoid such situation from happening anymore, we can however handle this case with a bit of additional logic, to postpone the initialization of the read sequence by one page. Reported-by: Alexander Shiyan <eagle.alexander923@gmail.com> Closes: https://lore.kernel.org/linux-mtd/CAP1tNvS=NVAm-vfvYWbc3k9Cx9YxMc2uZZkmXk8h1NhGX877Zg@mail.gmail.com/ Reported-by: Måns Rullgård <mans@mansr.com> Closes: https://lore.kernel.org/linux-mtd/yw1xfs6j4k6q.fsf@mansr.com/ Reported-by: Martin Hundebøll <martin@geanix.com> Closes: https://lore.kernel.org/linux-mtd/9d0c42fcde79bfedfe5b05d6a4e9fdef71d3dd52.camel@geanix.com/ Fixes: 003fe4b9545b ("mtd: rawnand: Support for sequential cache reads") Cc: stable@vger.kernel.org Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com> --- drivers/mtd/nand/raw/nand_base.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+)