2023-03-23 14:12:02

by Bang Li

Subject: [PATCH] mtdblock: tolerate corrected bit-flips

mtd_read() may return -EUCLEAN in case of corrected bit-flips.This
particular condition should not be treated like an error.

Signed-off-by: Bang Li <[email protected]>
---
drivers/mtd/mtdblock.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 1e94e7d10b8b..a0a1194dc1d9 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -153,7 +153,7 @@ static int do_cached_write (struct mtdblk_dev *mtdblk, unsigned long pos,
mtdblk->cache_state = STATE_EMPTY;
ret = mtd_read(mtd, sect_start, sect_size,
&retlen, mtdblk->cache_data);
- if (ret)
+ if (ret && !mtd_is_bitflip(ret))
return ret;
if (retlen != sect_size)
return -EIO;
@@ -188,8 +188,12 @@ static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
pr_debug("mtdblock: read on \"%s\" at 0x%lx, size 0x%x\n",
mtd->name, pos, len);

- if (!sect_size)
- return mtd_read(mtd, pos, len, &retlen, buf);
+ if (!sect_size) {
+ ret = mtd_read(mtd, pos, len, &retlen, buf);
+ if (ret && !mtd_is_bitflip(ret))
+ return ret;
+ return 0;
+ }

while (len > 0) {
unsigned long sect_start = (pos/sect_size)*sect_size;
@@ -209,7 +213,7 @@ static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
memcpy (buf, mtdblk->cache_data + offset, size);
} else {
ret = mtd_read(mtd, pos, size, &retlen, buf);
- if (ret)
+ if (ret && !mtd_is_bitflip(ret))
return ret;
if (retlen != size)
return -EIO;
--
2.25.1

2023-03-28 13:08:15

by Miquel Raynal

Subject: Re: [PATCH] mtdblock: tolerate corrected bit-flips

Hello,

[email protected] wrote on Thu, 23 Mar 2023 22:03:50 +0800:

> mtd_read() may return -EUCLEAN in case of corrected bit-flips.This
> particular condition should not be treated like an error.

I'm fine with the patch, Richard, are you okay as well?

> Signed-off-by: Bang Li <[email protected]>
> ---
> drivers/mtd/mtdblock.c | 12 ++++++++----
> 1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
> index 1e94e7d10b8b..a0a1194dc1d9 100644
> --- a/drivers/mtd/mtdblock.c
> +++ b/drivers/mtd/mtdblock.c
> @@ -153,7 +153,7 @@ static int do_cached_write (struct mtdblk_dev *mtdblk, unsigned long pos,
> mtdblk->cache_state = STATE_EMPTY;
> ret = mtd_read(mtd, sect_start, sect_size,
> &retlen, mtdblk->cache_data);
> - if (ret)
> + if (ret && !mtd_is_bitflip(ret))
> return ret;
> if (retlen != sect_size)
> return -EIO;
> @@ -188,8 +188,12 @@ static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
> pr_debug("mtdblock: read on \"%s\" at 0x%lx, size 0x%x\n",
> mtd->name, pos, len);
>
> - if (!sect_size)
> - return mtd_read(mtd, pos, len, &retlen, buf);
> + if (!sect_size) {
> + ret = mtd_read(mtd, pos, len, &retlen, buf);
> + if (ret && !mtd_is_bitflip(ret))
> + return ret;
> + return 0;
> + }
>
> while (len > 0) {
> unsigned long sect_start = (pos/sect_size)*sect_size;
> @@ -209,7 +213,7 @@ static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
> memcpy (buf, mtdblk->cache_data + offset, size);
> } else {
> ret = mtd_read(mtd, pos, size, &retlen, buf);
> - if (ret)
> + if (ret && !mtd_is_bitflip(ret))
> return ret;
> if (retlen != size)
> return -EIO;


Thanks,
Miquèl

2023-03-28 13:28:08

by Richard Weinberger

Subject: Re: [PATCH] mtdblock: tolerate corrected bit-flips

Hi!

----- Ursprüngliche Mail -----
> [email protected] wrote on Thu, 23 Mar 2023 22:03:50 +0800:
>
>> mtd_read() may return -EUCLEAN in case of corrected bit-flips.This
>> particular condition should not be treated like an error.
>
> I'm fine with the patch, Richard, are you okay as well?

Hm yes. I see the bug. I think it broke about 10 years ago
by e47f68587b82 ("mtd: check for max_bitflips in mtd_read_oob()").

In the early day I'd expect someone to hit this bug but I guess
on the last 10 years mtdblock had no real users.

Thanks,
//richard

2023-03-28 14:06:42

by Miquel Raynal

Subject: Re: [PATCH] mtdblock: tolerate corrected bit-flips

Hi Richard,

[email protected] wrote on Tue, 28 Mar 2023 15:21:50 +0200 (CEST):

> Hi!
>
> ----- Ursprüngliche Mail -----
> > [email protected] wrote on Thu, 23 Mar 2023 22:03:50 +0800:
> >
> >> mtd_read() may return -EUCLEAN in case of corrected bit-flips.This
> >> particular condition should not be treated like an error.
> >
> > I'm fine with the patch, Richard, are you okay as well?
>
> Hm yes. I see the bug. I think it broke about 10 years ago
> by e47f68587b82 ("mtd: check for max_bitflips in mtd_read_oob()").
>
> In the early day I'd expect someone to hit this bug but I guess
> on the last 10 years mtdblock had no real users.

Right.

Well, anyhow I guess this patch deserves a Fixes/Cc: stable # v3.7+

Li, can you please re-send with the proper tags? Thanks.

Thanks,
Miquèl