diff mbox

[v2] MTD: modify mtd api to return bitflip info on read operations

Message ID 1322943640-11728-1-git-send-email-mikedunn@newsguy.com
State New, archived
Headers show

Commit Message

Mike Dunn Dec. 3, 2011, 8:20 p.m. UTC 
Hi,

This patch proposes a change to the mtd API for the purpose of returning to
the caller information on the number of bit errors corrected by the ecc
facilities of the device during read operations.  The affected functions are
read() and read_oob().

Currently, the -EUCLEAN value returned by read() and read_oob() is the only
information available to the caller regarding bit error corrections.  This
return value indicates simply that one or more bit errors were corrected.  To
make matters worse, this applies to the entire read operation, which can
potentially span the entire device.  Some NAND flash chips are error prone, and
compensate for that by using strong ecc algorithms capable of correcting
multiple errors in a single page.  In order for higher level code (e.g. UBI) to
effectively detect degradation of erase blocks on these devices, more detailed
information is needed.

For the read() method, an unsigned int * argument is added, which the driver
uses to return to the caller the maximum number of bitflips that were corrected
on any single page.  If NULL is passed in this argument, the driver ignores it.
For read_oob(), an element is added to the mtd_oob_ops structure for the same
purpose.  Devices without ecc capabilities (NOR flash, etc) would simply set the
value to 0.

This v2 includes the review comments from Artem Bityutskiy and Thomas Petazzoni,
mainly adding the option to pass max_bitflips = NULL (yes, of course! <smacks
forehead>), as well as some minor changes requested by Artem.  It has been
tested on mtdram, nandsim, onenandsim, and the diskonchip G4 flash (nand driver
currently out-of-tree), on partitioned and unpartitioned devices.  Drivers for
other devices have been compile-tested only, but the changes are trivial in most
cases.  Comments, criticisms, objections, gratefully received.

Thanks,
Mike


Signed-off-by: Mike Dunn <mikedunn@newsguy.com>
---
 drivers/mtd/afs.c                   |    4 ++--
 drivers/mtd/ar7part.c               |    8 ++++----
 drivers/mtd/chips/cfi_cmdset_0001.c |   10 ++++++++--
 drivers/mtd/chips/cfi_cmdset_0002.c |   10 ++++++++--
 drivers/mtd/chips/cfi_cmdset_0020.c |   10 ++++++++--
 drivers/mtd/chips/map_absent.c      |    7 +++++--
 drivers/mtd/chips/map_ram.c         |    8 ++++++--
 drivers/mtd/chips/map_rom.c         |    8 ++++++--
 drivers/mtd/devices/block2mtd.c     |    6 +++++-
 drivers/mtd/devices/doc2000.c       |   11 +++++++++--
 drivers/mtd/devices/doc2001.c       |    7 +++++--
 drivers/mtd/devices/doc2001plus.c   |    6 ++++--
 drivers/mtd/devices/docg3.c         |    6 +++++-
 drivers/mtd/devices/lart.c          |    5 ++++-
 drivers/mtd/devices/m25p80.c        |    5 ++++-
 drivers/mtd/devices/ms02-nv.c       |    6 ++++--
 drivers/mtd/devices/mtd_dataflash.c |    5 ++++-
 drivers/mtd/devices/mtdram.c        |    4 +++-
 drivers/mtd/devices/phram.c         |    5 ++++-
 drivers/mtd/devices/pmc551.c        |    6 +++++-
 drivers/mtd/devices/slram.c         |    8 ++++++--
 drivers/mtd/devices/sst25l.c        |    6 +++++-
 drivers/mtd/ftl.c                   |   24 +++++++++++++-----------
 drivers/mtd/inftlcore.c             |    7 ++++---
 drivers/mtd/inftlmount.c            |    6 +++---
 drivers/mtd/lpddr/lpddr_cmds.c      |    8 ++++++--
 drivers/mtd/maps/bcm963xx-flash.c   |    4 ++--
 drivers/mtd/mtdblock.c              |    7 ++++---
 drivers/mtd/mtdblock_ro.c           |    2 +-
 drivers/mtd/mtdchar.c               |    2 +-
 drivers/mtd/mtdconcat.c             |   16 ++++++++++++----
 drivers/mtd/mtdoops.c               |    3 ++-
 drivers/mtd/mtdpart.c               |    4 ++--
 drivers/mtd/mtdswap.c               |    4 ++--
 drivers/mtd/nand/diskonchip.c       |    4 ++--
 drivers/mtd/nand/nand_base.c        |   21 ++++++++++++++++++---
 drivers/mtd/nand/nand_bbt.c         |    7 ++++---
 drivers/mtd/nftlcore.c              |    6 +++---
 drivers/mtd/nftlmount.c             |    6 +++---
 drivers/mtd/onenand/onenand_base.c  |    6 +++++-
 drivers/mtd/redboot.c               |    2 +-
 drivers/mtd/rfd_ftl.c               |   12 +++++++-----
 drivers/mtd/ssfdc.c                 |    4 ++--
 drivers/mtd/tests/mtd_pagetest.c    |   30 ++++++++++++++++--------------
 drivers/mtd/tests/mtd_readtest.c    |    2 +-
 drivers/mtd/tests/mtd_speedtest.c   |    8 ++++----
 drivers/mtd/tests/mtd_stresstest.c  |    2 +-
 drivers/mtd/tests/mtd_subpagetest.c |    9 +++++----
 drivers/mtd/tests/mtd_torturetest.c |    2 +-
 drivers/mtd/ubi/io.c                |    6 +++---
 include/linux/mtd/mtd.h             |   11 ++++++++++-
 include/linux/mtd/pmc551.h          |    3 ++-
 52 files changed, 254 insertions(+), 125 deletions(-)

Comments

Peter Horton Dec. 4, 2011, 8:43 a.m. UTC | #1 
On 03/12/2011 20:20, Mike Dunn wrote:
>
> This patch proposes a change to the mtd API for the purpose of returning to
> the caller information on the number of bit errors corrected by the ecc
> facilities of the device during read operations.  The affected functions are
> read() and read_oob().
>

Do the number of bit-flips mean anything to the higher layers like UBI? 
As the ECC strength / error rate are a chip dependent thing how do the 
higher layers know what is good/normal/bad?

P.
Mike Dunn Dec. 4, 2011, 1:52 p.m. UTC | #2 
On 12/04/2011 12:43 AM, Peter Horton wrote:
> On 03/12/2011 20:20, Mike Dunn wrote:
>>
>> This patch proposes a change to the mtd API for the purpose of returning to
>> the caller information on the number of bit errors corrected by the ecc
>> facilities of the device during read operations.  The affected functions are
>> read() and read_oob().
>>
>
> Do the number of bit-flips mean anything to the higher layers like UBI? 


The change was motivated primarily by the desire to get ubifs working well on
nand flash.  Currently it works well only on onenand devices where single
bitflips are rare and random.  On nand with frequent and consistent bitflips,
ubi marks a large portion of the blocks as "bad".


> As the ECC strength / error rate are a chip dependent thing how do the higher
> layers know what is good/normal/bad?


Good point.  To be determined.  Probably just another element in the mtd_info
struct named ecc_strength or some such.  UBI e.g., can determine a suitable
"scrublevel" based on that.

Thanks,
Mike
Artem Bityutskiy Dec. 4, 2011, 2:33 p.m. UTC | #3 
On Sat, 2011-12-03 at 12:20 -0800, Mike Dunn wrote:
> +                       bitflips = max_t(unsigned int, bitflips,
> +                                        (unsigned)nb_errors); 

You do not  need the (unsigned) cast because you use max_t().
Artem Bityutskiy Dec. 4, 2011, 2:43 p.m. UTC | #4 
On Sun, 2011-12-04 at 05:52 -0800, Mike Dunn wrote:
> On 12/04/2011 12:43 AM, Peter Horton wrote:
> > On 03/12/2011 20:20, Mike Dunn wrote:
> >>
> >> This patch proposes a change to the mtd API for the purpose of returning to
> >> the caller information on the number of bit errors corrected by the ecc
> >> facilities of the device during read operations.  The affected functions are
> >> read() and read_oob().
> >>
> >
> > Do the number of bit-flips mean anything to the higher layers like UBI? 
> 
> 
> The change was motivated primarily by the desire to get ubifs working well on
> nand flash.  Currently it works well only on onenand devices where single
> bitflips are rare and random.  On nand with frequent and consistent bitflips,
> ubi marks a large portion of the blocks as "bad".

Well, I think non-onenands are also supported. The very modern NAND
support has issues because of new problems which did not exist or were
not visible in the past.

> 
> > As the ECC strength / error rate are a chip dependent thing how do the higher
> > layers know what is good/normal/bad?
> 
> 
> Good point.  To be determined.  Probably just another element in the mtd_info
> struct named ecc_strength or some such.  UBI e.g., can determine a suitable
> "scrublevel" based on that.

I think UBI should scrub only if max_bitflips == ecc_strength by
default. If the driver supplies scrublevel scrub if max_bitflips ==
scrublevel.
Mike Dunn Dec. 4, 2011, 2:55 p.m. UTC | #5 
On 12/04/2011 06:43 AM, Artem Bityutskiy wrote:
> On Sun, 2011-12-04 at 05:52 -0800, Mike Dunn wrote:
>> On 12/04/2011 12:43 AM, Peter Horton wrote:
>>> On 03/12/2011 20:20, Mike Dunn wrote:
>>>> This patch proposes a change to the mtd API for the purpose of returning to
>>>> the caller information on the number of bit errors corrected by the ecc
>>>> facilities of the device during read operations.  The affected functions are
>>>> read() and read_oob().
>>>>
>>> Do the number of bit-flips mean anything to the higher layers like UBI? 
>>
>> The change was motivated primarily by the desire to get ubifs working well on
>> nand flash.  Currently it works well only on onenand devices where single
>> bitflips are rare and random.  On nand with frequent and consistent bitflips,
>> ubi marks a large portion of the blocks as "bad".
> Well, I think non-onenands are also supported. The very modern NAND
> support has issues because of new problems which did not exist or were
> not visible in the past.


I stand corrected.  Not intended as a knock on UBI :)


>>> As the ECC strength / error rate are a chip dependent thing how do the higher
>>> layers know what is good/normal/bad?
>>
>> Good point.  To be determined.  Probably just another element in the mtd_info
>> struct named ecc_strength or some such.  UBI e.g., can determine a suitable
>> "scrublevel" based on that.
> I think UBI should scrub only if max_bitflips == ecc_strength by
> default. If the driver supplies scrublevel scrub if max_bitflips ==
> scrublevel.


So you're thinking that the driver would supply both ecc_strength and
"scrublevel" (or maybe bb_threshold)?  Would these go into struct mtd_info?

Thanks,
Mike
Artem Bityutskiy Dec. 5, 2011, 6:07 a.m. UTC | #6 
On Sun, 2011-12-04 at 06:55 -0800, Mike Dunn wrote:
> So you're thinking that the driver would supply both ecc_strength and
> "scrublevel" (or maybe bb_threshold)?  Would these go into struct mtd_info?

Probably yes. After all, UBI has no idea about what kind of flash is
that and what kind of ECC it uses and what bit-flip level needs
scrubbing. So I think this kind of information should come from the
driver or from the user via mtd sysfs files. What do you think?

Artem.
Artem Bityutskiy Dec. 5, 2011, 6:23 a.m. UTC | #7 
On Sat, 2011-12-03 at 12:20 -0800, Mike Dunn wrote:
> Hi,
> 
> This patch proposes a change to the mtd API for the purpose of returning to
> the caller information on the number of bit errors corrected by the ecc
> facilities of the device during read operations.  The affected functions are
> read() and read_oob().

Mike,

this is quite big patch, and I've realized that it is difficult to
review it because of the size. I know I suggested one big patch, but now
I think we should try to split it, if we can. This way it will also be
easier to pass through dwmw2 because at the end he is the MTD
maintainer.

I can see the following parts in your patch:

1. Mechanical part - no much brains needed, just change prototypes, add
few comments, add NULL arguments everywhere. This is the biggest part. 

2. Implementation part - should be much smaller - implements
max_bitflips support in MTD.

Part 2 is interesting to reveiw, and currently part 1 adds so much noise
that the review becomes difficult. Can you split your patches like that?

I apologize for not suggesting this from the very beginning.

Artem.


P.S. As a side note, I am thinking that with your patch the -EUCLEAN
return code may go away. It has always been a bit ugly interface anyway.
What do you think? My thinking is that we can do this separately later.
But you need to add assertions like:

	WARN_ON(err == -EUCLEAN && max_bitflips == 0);

in interesting places.

It would be easier to do if MTD interface had a single entry point,
though.
Mike Dunn Dec. 5, 2011, 4:58 p.m. UTC | #8 
On 12/04/2011 10:07 PM, Artem Bityutskiy wrote:
> On Sun, 2011-12-04 at 06:55 -0800, Mike Dunn wrote:
>> So you're thinking that the driver would supply both ecc_strength and
>> "scrublevel" (or maybe bb_threshold)?  Would these go into struct mtd_info?
> Probably yes. After all, UBI has no idea about what kind of flash is
> that and what kind of ECC it uses and what bit-flip level needs
> scrubbing. So I think this kind of information should come from the
> driver or from the user via mtd sysfs files. What do you think?


I'm not a flash expert, but that sounds reasonable, especially if the scrublevel
parameter is optional, with UBI using ecc_strength as the default value.  As for
how to pass the parameter, sysfs might be a good idea for scrublevel, allowing
it to be tunable without having to modify the driver if experience show that the
driver author's original assumptions about how a block degrades were incorrect.

Thanks,
Mike
Peter Horton Dec. 5, 2011, 5:09 p.m. UTC | #9 
On 05/12/2011 16:58, Mike Dunn wrote:
> On 12/04/2011 10:07 PM, Artem Bityutskiy wrote:
>> On Sun, 2011-12-04 at 06:55 -0800, Mike Dunn wrote:
>>> So you're thinking that the driver would supply both ecc_strength and
>>> "scrublevel" (or maybe bb_threshold)?  Would these go into struct mtd_info?
>> Probably yes. After all, UBI has no idea about what kind of flash is
>> that and what kind of ECC it uses and what bit-flip level needs
>> scrubbing. So I think this kind of information should come from the
>> driver or from the user via mtd sysfs files. What do you think?
>
>
> I'm not a flash expert, but that sounds reasonable, especially if the scrublevel
> parameter is optional, with UBI using ecc_strength as the default value.  As for
> how to pass the parameter, sysfs might be a good idea for scrublevel, allowing
> it to be tunable without having to modify the driver if experience show that the
> driver author's original assumptions about how a block degrades were incorrect.
>

Surely the check for "do we need to scrub ?" should be done lower down 
otherwise all users of the mtd NAND interface (UBI / JFFS2 etc) are 
going to have to re-implement those sysfs files and the corresponding 
checks.

P.
Mike Dunn Dec. 5, 2011, 6:13 p.m. UTC | #10 
On 12/04/2011 10:23 PM, Artem Bityutskiy wrote:
> On Sat, 2011-12-03 at 12:20 -0800, Mike Dunn wrote:
>> Hi,
>>
>> This patch proposes a change to the mtd API for the purpose of returning to
>> the caller information on the number of bit errors corrected by the ecc
>> facilities of the device during read operations.  The affected functions are
>> read() and read_oob().
> Mike,
>
> this is quite big patch, and I've realized that it is difficult to
> review it because of the size. I know I suggested one big patch, but now
> I think we should try to split it, if we can. This way it will also be
> easier to pass through dwmw2 because at the end he is the MTD
> maintainer.
>
> I can see the following parts in your patch:
>
> 1. Mechanical part - no much brains needed, just change prototypes, add
> few comments, add NULL arguments everywhere. This is the biggest part. 
>
> 2. Implementation part - should be much smaller - implements
> max_bitflips support in MTD.
>
> Part 2 is interesting to reveiw, and currently part 1 adds so much noise
> that the review becomes difficult. Can you split your patches like that?


OK.  But you may still be underestimating the quantity of code that warrants
review.  Since we're going back to multiple patches anyway, how about four patches:

1. api change to mtd.h (a few lines just to put the rest in context)
2. nand, onenand
3. mtd infrastructructure: mtdblock, mtdchar, mtdpart, mtdconcat, ...
4. rote changes

Patch 4 will contain nothing beyond function definition changes, and the addition of

if(max_bitflips != NULL)
    *max_bitflips = 0;

to the top of the read function.  Anything beyond that will go into patch 3. 
Patch 4 will include all drivers that are not nand or onenand, since these don't
have ecc (two exceptions being devices/docg3.c and nand/alauda.c, which are are
nand drivers that do not use the nand interface).

Please let me know if that sounds reasonable.


> I apologize for not suggesting this from the very beginning.


No problem.  I expected this would be an iterative process and resigned myself
to a lot of tedium :)


>
> P.S. As a side note, I am thinking that with your patch the -EUCLEAN
> return code may go away. It has always been a bit ugly interface anyway.
> What do you think? 


I agree.  And with the max_bitflips change, it become cruft.  The patch required
to accomplish this would not be very large, either.



> My thinking is that we can do this separately later.
> But you need to add assertions like:
>
> 	WARN_ON(err == -EUCLEAN && max_bitflips == 0);
>
> in interesting places.


OK, I'll include this with the next patch set.

Thanks,
Mike
Mike Dunn Dec. 5, 2011, 6:57 p.m. UTC | #11 
On 12/05/2011 09:09 AM, Peter Horton wrote:
> On 05/12/2011 16:58, Mike Dunn wrote:
>> On 12/04/2011 10:07 PM, Artem Bityutskiy wrote:
>>> On Sun, 2011-12-04 at 06:55 -0800, Mike Dunn wrote:
>>>> So you're thinking that the driver would supply both ecc_strength and
>>>> "scrublevel" (or maybe bb_threshold)?  Would these go into struct mtd_info?
>>> Probably yes. After all, UBI has no idea about what kind of flash is
>>> that and what kind of ECC it uses and what bit-flip level needs
>>> scrubbing. So I think this kind of information should come from the
>>> driver or from the user via mtd sysfs files. What do you think?
>>
>>
>> I'm not a flash expert, but that sounds reasonable, especially if the scrublevel
>> parameter is optional, with UBI using ecc_strength as the default value.  As for
>> how to pass the parameter, sysfs might be a good idea for scrublevel, allowing
>> it to be tunable without having to modify the driver if experience show that the
>> driver author's original assumptions about how a block degrades were incorrect.
>>
>
> Surely the check for "do we need to scrub ?" should be done lower down
> otherwise all users of the mtd NAND interface (UBI / JFFS2 etc) are going to
> have to re-implement those sysfs files and the corresponding checks.


Well, anything higher up that wants to avail itself of this api change will need
some rework regardless.  Currently the only info passed up from the driver is
that at least one bitflip occurred somewhere during the read.  The plan is to
eventually make some changes to UBI so that the decision to scrub is made more
intelligently.

Also, in addition to being too restricting, hard-coding the decision whether to
scrub into the driver would be tantamount to the driver setting policy, I think.

Thanks,
Mike
Artem Bityutskiy Dec. 5, 2011, 9:16 p.m. UTC | #12 
On Mon, 2011-12-05 at 10:13 -0800, Mike Dunn wrote:
> OK.  But you may still be underestimating the quantity of code that warrants
> review.  Since we're going back to multiple patches anyway, how about four patches:
> 
> 1. api change to mtd.h (a few lines just to put the rest in context)
> 2. nand, onenand
> 3. mtd infrastructructure: mtdblock, mtdchar, mtdpart, mtdconcat, ...
> 4. rote changes

But you need to maintain bisectability. The kernel has to be compilable
and working between all these steps. The only way to do this is to add 2
new interfaces without removing the old ones, the do all the changes in
N stpes, then remove old interfaces. When I start thinking about that,
it looks too much, so I am not sure.

Artem.
Robert Jarzmik Dec. 6, 2011, 9:52 p.m. UTC | #13 
Mike Dunn <mikedunn@newsguy.com> writes:
> On 12/05/2011 09:09 AM, Peter Horton wrote:
>> Surely the check for "do we need to scrub ?" should be done lower down
>> otherwise all users of the mtd NAND interface (UBI / JFFS2 etc) are going to
>> have to re-implement those sysfs files and the corresponding checks.
> Well, anything higher up that wants to avail itself of this api change will need
> some rework regardless.  Currently the only info passed up from the driver is
> that at least one bitflip occurred somewhere during the read.  The plan is to
> eventually make some changes to UBI so that the decision to scrub is made more
> intelligently.

I'd like to second Mike's proposition here.

Policy should not be put into drivers code whenever possible. The question "do
we need to scrub" is to be answered in upper layers IMHO. Moreover, upper layer
would then have the choice to trigger scrubbing on their own policy. This policy
could rely on the information provided by the driver :
 - how much bitflips the ECC can fix

Even better, but I don't know how, the policy should be a userspace matter and
not a kernel one, as I was told many times before :)

Therefore, a upperlayer which strives at very high security level could scrub at
1 bitflip even if the ECC can fix 8. A "normal" upperlayer could go up to 8
before scrubbing.

Now for the "re-implement sysfs and checks", that's true. What could be done is
in a first time, copycat the current behaviour, ie. UBI scrubs if nb_bitflips >
0, and the same for JFFS I suppose.

Then in a second pass, add sysfs values for the trigger (whether in upperlayers
or MTD). This needs a consensus from the list.

Cheers.
Mike Dunn Dec. 7, 2011, 2:16 a.m. UTC | #14 
On 12/06/2011 01:52 PM, Robert Jarzmik wrote:
> Mike Dunn <mikedunn@newsguy.com> writes:
>> On 12/05/2011 09:09 AM, Peter Horton wrote:
>>> Surely the check for "do we need to scrub ?" should be done lower down
>>> otherwise all users of the mtd NAND interface (UBI / JFFS2 etc) are going to
>>> have to re-implement those sysfs files and the corresponding checks.
>> Well, anything higher up that wants to avail itself of this api change will need
>> some rework regardless.  Currently the only info passed up from the driver is
>> that at least one bitflip occurred somewhere during the read.  The plan is to
>> eventually make some changes to UBI so that the decision to scrub is made more
>> intelligently.
> I'd like to second Mike's proposition here.
>
> Policy should not be put into drivers code whenever possible. The question "do
> we need to scrub" is to be answered in upper layers IMHO. Moreover, upper layer
> would then have the choice to trigger scrubbing on their own policy. This policy
> could rely on the information provided by the driver :
>  - how much bitflips the ECC can fix
>
> Even better, but I don't know how, the policy should be a userspace matter and
> not a kernel one, as I was told many times before :)
>
> Therefore, a upperlayer which strives at very high security level could scrub at
> 1 bitflip even if the ECC can fix 8. A "normal" upperlayer could go up to 8
> before scrubbing.


Yes, this flexibility is good, and why policy should not be set from below. 
Good point.  I just mentioned the "don't set policy" doctrine because I blindly
follow dogma :)


> Now for the "re-implement sysfs and checks", that's true. What could be done is
> in a first time, copycat the current behaviour, ie. UBI scrubs if nb_bitflips >
> 0, and the same for JFFS I suppose.
> Then in a second pass, add sysfs values for the trigger (whether in upperlayers
> or MTD). This needs a consensus from the list.


For UBI, the direction of the discussion was that the driver would indicate the
ecc strength of the device (via a new element in struct mtd_info), and this
would be the default UBI scrublevel.  A non-zero sysfs parameter would override
the default.

BTW, it is currently possible (and easy) to do what Peter suggests; i.e., to
avoid any UBI rework and make the decision to scrub in the driver.  All the
driver has to do is not return -EUCLEAN unless it thinks the block should be
scrubbed.  Of course this would be a lie if there were in fact corrected
bitflips.  For drivers using the mtd nand interface, the lie would be in the ecc
stats, since the nand infrastructure code returns -EUCLEAN (or 0) based on
that.  This is what we willl have to do in the diskonchip drivers in order to
get ubi/ubifs to work on them if we don't change the mtd api (and afterwards
make the changes to ubi).

Thanks,
Mike
Ricard Wanderlof Dec. 7, 2011, 8:01 a.m. UTC | #15 
On Wed, 7 Dec 2011, Mike Dunn wrote:

> For UBI, the direction of the discussion was that the driver would 
> indicate the ecc strength of the device (via a new element in struct 
> mtd_info), and this would be the default UBI scrublevel.  A non-zero 
> sysfs parameter would override the default.


>
> BTW, it is currently possible (and easy) to do what Peter suggests; i.e., to
> avoid any UBI rework and make the decision to scrub in the driver.  All the
> driver has to do is not return -EUCLEAN unless it thinks the block should be
> scrubbed.  Of course this would be a lie if there were in fact corrected
> bitflips.  For drivers using the mtd nand interface, the lie would be in the ecc
> stats, since the nand infrastructure code returns -EUCLEAN (or 0) based on
> that.  This is what we willl have to do in the diskonchip drivers in order to
> get ubi/ubifs to work on them if we don't change the mtd api (and afterwards
> make the changes to ubi).

The idea of using -EUCLEAN to indicate that the lower levels think it's 
time for scrubbing appeals to me. One reason is that an upper layer such 
as UBI should not really be concerned with details such as ECC strength 
and number of flipped bits, it seems to me that such things should be left 
to a lower level. Also, one could imagine different variants of flash 
technology requiring different consideration to bit flips and other 
phenomena which again should not be something that for instance UBI needs 
to know about. It makes sense for mtd to have a sweeping knowledge of the 
general types of errors that can occur so that part of the decision can be 
made there.

That said, the level at which the lowest level reports -EUCLEAN could be 
something that is configurable from user space (sysfs), so that it's 
userspace that sets the actual policy and no one else.

So, something like this:
- The driver reports number of bit flips and current ECC strength etc to
   the mtd layer.
- Based on some userspace knob, the mtd framework reports -EUCLEAN if
   scrubbing is needed.
- Upper layers perform scrubbing if they want to (i.e. UBI) or not (i.e.
   JFFS2).

I don't really see this as lying, more of a redefinition of what -EUCLEAN 
really means. The current behavior was invented when single-bit errors 
could be correct and nothing else, and furthermore such bitflips were 
rare; now that multiple-bit errors are commonplace it's time to put more 
detail into what -EUCLEAN implies.  It's not really breaking anything old 
either, if sane defaults are used.

/Ricard
Peter Horton Dec. 7, 2011, 9:42 a.m. UTC | #16 
On 06/12/2011 21:52, Robert Jarzmik wrote:
> Mike Dunn<mikedunn@newsguy.com>  writes:
>> On 12/05/2011 09:09 AM, Peter Horton wrote:
>>> Surely the check for "do we need to scrub ?" should be done lower down
>>> otherwise all users of the mtd NAND interface (UBI / JFFS2 etc) are going to
>>> have to re-implement those sysfs files and the corresponding checks.
>> Well, anything higher up that wants to avail itself of this api change will need
>> some rework regardless.  Currently the only info passed up from the driver is
>> that at least one bitflip occurred somewhere during the read.  The plan is to
>> eventually make some changes to UBI so that the decision to scrub is made more
>> intelligently.
>
> I'd like to second Mike's proposition here.
>
> Policy should not be put into drivers code whenever possible. The question "do
> we need to scrub" is to be answered in upper layers IMHO. Moreover, upper layer
> would then have the choice to trigger scrubbing on their own policy. This policy
> could rely on the information provided by the driver :
>   - how much bitflips the ECC can fix
>

It's not policy, it's a property of the NAND device.
Mike Dunn Dec. 7, 2011, 6:32 p.m. UTC | #17 
Hi Ricard, thanks for chiming in.

On 12/07/2011 12:01 AM, Ricard Wanderlof wrote:
>
>
> So, something like this:
> - The driver reports number of bit flips and current ECC strength etc to
>   the mtd layer.
> - Based on some userspace knob, the mtd framework reports -EUCLEAN if
>   scrubbing is needed.
> - Upper layers perform scrubbing if they want to (i.e. UBI) or not (i.e.
>   JFFS2).


Sounds like a nice compromise between user configurability and keeping the
decision in the mtd layer.  The problem is that the read method currently goes
directly to the driver, so all drivers would have to be patched.  The last patch
I submitted was rejected as being too large.  And it can't be broken into
smaller patches because they are interdependent and not bisectable. 
Implementing this scheme would be an even larger patch, and the changes made to
every driver would bear scrutiny.


>
> I don't really see this as lying, more of a redefinition of what -EUCLEAN
> really means. The current behavior was invented when single-bit errors could
> be correct and nothing else, 


That is clear.  In mtdconcat.c, ecc_stats.corrected is incremented by one when
mtd->read() returns -EUCLEAN.


> and furthermore such bitflips were rare; now that multiple-bit errors are
> commonplace it's time to put more detail into what -EUCLEAN implies.  It's not
> really breaking anything old either, if sane defaults are used.
>


I'm willing to implement whatever consensus emerges, but I need buy-in from the
maintainers before I spend more time on it.

Thanks,
Mike
Mike Dunn Dec. 7, 2011, 6:33 p.m. UTC | #18 
On 12/07/2011 01:42 AM, Peter Horton wrote:
>
> It's not policy, it's a property of the NAND device.
>


I respectfully disagree.  Device property is how error prone it is and its error
correction capability.  Policy is how cautious you want to be with regard to
data integrity vis-a-vis erase block management.  But it *is* a fine line, and
one affects the other.  Plus Ricard makes a good point about devices with other
properties besides ecc strength that could potentially influence the policy
decision.

Thanks,
Mike
Artem Bityutskiy Dec. 12, 2011, 12:48 p.m. UTC | #19 
On Wed, 2011-12-07 at 10:32 -0800, Mike Dunn wrote:
> Hi Ricard, thanks for chiming in.
> 
> On 12/07/2011 12:01 AM, Ricard Wanderlof wrote:
> >
> >
> > So, something like this:
> > - The driver reports number of bit flips and current ECC strength etc to
> >   the mtd layer.
> > - Based on some userspace knob, the mtd framework reports -EUCLEAN if
> >   scrubbing is needed.
> > - Upper layers perform scrubbing if they want to (i.e. UBI) or not (i.e.
> >   JFFS2).
> 
> 
> Sounds like a nice compromise between user configurability and keeping the
> decision in the mtd layer.  The problem is that the read method currently goes
> directly to the driver, so all drivers would have to be patched.  The last patch
> I submitted was rejected as being too large.  And it can't be broken into
> smaller patches because they are interdependent and not bisectable. 
> Implementing this scheme would be an even larger patch, and the changes made to
> every driver would bear scrutiny.

Yeah, I think it is OK.

1. Sanitize MTD interfaces by turning mtd->kuku(buh, buh) to
mtd_kuku(mtd, buh, buh). This should be easy to do - introduce wrappers
which just call mtd->kuku(buh, buh).

Then you can add your ECC stuff to mtd_*() functions.

Careful with partitions - we want ECC levels to be per-partition, so
that one could set different scrublevels for different partitions. So I
guess your ECC stuff should be in partition functions.

I do not remember off-the top of my head how it is handled nowadays, but
in the past, when we had mtdpart as a separate module, it was possible
to bypass partition wrappers. You should make it so that even if there
are no partitions - everything goes via mtpart anyways. Nowadays mtdpart
is an integral part of mtd, so this should be easy to do.
Mike Dunn Dec. 14, 2011, 8:46 p.m. UTC | #20 
On 12/12/2011 04:48 AM, Artem Bityutskiy wrote:
> On Wed, 2011-12-07 at 10:32 -0800, Mike Dunn wrote:
>> Hi Ricard, thanks for chiming in.
>>
>> On 12/07/2011 12:01 AM, Ricard Wanderlof wrote:
>>>
>>> So, something like this:
>>> - The driver reports number of bit flips and current ECC strength etc to
>>>   the mtd layer.
>>> - Based on some userspace knob, the mtd framework reports -EUCLEAN if
>>>   scrubbing is needed.
>>> - Upper layers perform scrubbing if they want to (i.e. UBI) or not (i.e.
>>>   JFFS2).
>>
>> Sounds like a nice compromise between user configurability and keeping the
>> decision in the mtd layer.  The problem is that the read method currently goes
>> directly to the driver, so all drivers would have to be patched.  The last patch
>> I submitted was rejected as being too large.  And it can't be broken into
>> smaller patches because they are interdependent and not bisectable. 
>> Implementing this scheme would be an even larger patch, and the changes made to
>> every driver would bear scrutiny.
> Yeah, I think it is OK.

...<snip>...


> You should make it so that even if there
> are no partitions - everything goes via mtpart anyways. Nowadays mtdpart
> is an integral part of mtd, so this should be easy to do.


OK, this makes sense.  Yes, going through the partitioning code and its wrapper
functions for unpartitioned devices does look fairly straightforward, and It
allows the decision of whether or not to return -EUCLEAN to be made in the mtd
code, not the drivers.  Plus it can be its own separate patch.  But it does not
obviate the need to change the api for read() and read_oob() in order to return
the bitflip info.  Currently only the drivers can determine this.  The read
operation can span multiple pages, so e.g., we can't just use ecc_stats.  So the
"one large patch" problem remains.

As an aside, the nand infrastructure code only reads one page at a time, so if
we restricted this to devices that use the nand interface, ecc_stats could be
used, the driivers would not have to be touched, and the patch would be very
reasonable.  Unfortunately, there are two nand devices that use ecc but do not
use the nand interface.

Here's a proposed patch set:

1. mtd api changed so that read() and read_oob() return bitflip info (LARGE,
essentially the previous patch)
2. Unpartitioned devices go through the partitioning code (small patch)
3. Add ecc_strength to struct mtd_info, backport to drivers (moderate patch)
4. Implement functionality to intelligently decide whether to return -EUCLEAN or
0; add sysfs controls (moderate patch)

Please give this some more thought.  Just as a reminder of the size of item 1,
the diff stats of the previous patch are pasted below.

Thanks,
Mike


 drivers/mtd/afs.c                   |    4 ++--
 drivers/mtd/ar7part.c               |    8 ++++----
 drivers/mtd/chips/cfi_cmdset_0001.c |   10 ++++++++--
 drivers/mtd/chips/cfi_cmdset_0002.c |   10 ++++++++--
 drivers/mtd/chips/cfi_cmdset_0020.c |   10 ++++++++--
 drivers/mtd/chips/map_absent.c      |    7 +++++--
 drivers/mtd/chips/map_ram.c         |    8 ++++++--
 drivers/mtd/chips/map_rom.c         |    8 ++++++--
 drivers/mtd/devices/block2mtd.c     |    6 +++++-
 drivers/mtd/devices/doc2000.c       |   11 +++++++++--
 drivers/mtd/devices/doc2001.c       |    7 +++++--
 drivers/mtd/devices/doc2001plus.c   |    6 ++++--
 drivers/mtd/devices/docg3.c         |    6 +++++-
 drivers/mtd/devices/lart.c          |    5 ++++-
 drivers/mtd/devices/m25p80.c        |    5 ++++-
 drivers/mtd/devices/ms02-nv.c       |    6 ++++--
 drivers/mtd/devices/mtd_dataflash.c |    5 ++++-
 drivers/mtd/devices/mtdram.c        |    4 +++-
 drivers/mtd/devices/phram.c         |    5 ++++-
 drivers/mtd/devices/pmc551.c        |    6 +++++-
 drivers/mtd/devices/slram.c         |    8 ++++++--
 drivers/mtd/devices/sst25l.c        |    6 +++++-
 drivers/mtd/ftl.c                   |   24 +++++++++++++-----------
 drivers/mtd/inftlcore.c             |    7 ++++---
 drivers/mtd/inftlmount.c            |    6 +++---
 drivers/mtd/lpddr/lpddr_cmds.c      |    8 ++++++--
 drivers/mtd/maps/bcm963xx-flash.c   |    4 ++--
 drivers/mtd/mtdblock.c              |    7 ++++---
 drivers/mtd/mtdblock_ro.c           |    2 +-
 drivers/mtd/mtdchar.c               |    2 +-
 drivers/mtd/mtdconcat.c             |   16 ++++++++++++----
 drivers/mtd/mtdoops.c               |    3 ++-
 drivers/mtd/mtdpart.c               |    4 ++--
 drivers/mtd/mtdswap.c               |    4 ++--
 drivers/mtd/nand/diskonchip.c       |    4 ++--
 drivers/mtd/nand/nand_base.c        |   21 ++++++++++++++++++---
 drivers/mtd/nand/nand_bbt.c         |    7 ++++---
 drivers/mtd/nftlcore.c              |    6 +++---
 drivers/mtd/nftlmount.c             |    6 +++---
 drivers/mtd/onenand/onenand_base.c  |    6 +++++-
 drivers/mtd/redboot.c               |    2 +-
 drivers/mtd/rfd_ftl.c               |   12 +++++++-----
 drivers/mtd/ssfdc.c                 |    4 ++--
 drivers/mtd/tests/mtd_pagetest.c    |   30 ++++++++++++++++--------------
 drivers/mtd/tests/mtd_readtest.c    |    2 +-
 drivers/mtd/tests/mtd_speedtest.c   |    8 ++++----
 drivers/mtd/tests/mtd_stresstest.c  |    2 +-
 drivers/mtd/tests/mtd_subpagetest.c |    9 +++++----
 drivers/mtd/tests/mtd_torturetest.c |    2 +-
 drivers/mtd/ubi/io.c                |    6 +++---
 include/linux/mtd/mtd.h             |   11 ++++++++++-
 include/linux/mtd/pmc551.h          |    3 ++-
 52 files changed, 254 insertions(+), 125 deletions(-)
Artem Bityutskiy Dec. 16, 2011, 10:09 a.m. UTC | #21 
On Wed, 2011-12-14 at 12:46 -0800, Mike Dunn wrote:
> Here's a proposed patch set:
> 
> 1. mtd api changed so that read() and read_oob() return bitflip info (LARGE,
> essentially the previous patch)
> 2. Unpartitioned devices go through the partitioning code (small patch)
> 3. Add ecc_strength to struct mtd_info, backport to drivers (moderate patch)
> 4. Implement functionality to intelligently decide whether to return -EUCLEAN or
> 0; add sysfs controls (moderate patch)

Sounds good. Send patches :) You could already have done step 1 at
least :-) Remember, you have less and less time to get into 3.3, let's
make sure at least the big step 1 gets there and gets tested.
diff mbox

Patch

diff --git a/drivers/mtd/afs.c b/drivers/mtd/afs.c
index 89a02f6..a85c7b1 100644
--- a/drivers/mtd/afs.c
+++ b/drivers/mtd/afs.c
@@ -75,7 +75,7 @@  afs_read_footer(struct mtd_info *mtd, u_int *img_start, u_int *iis_start,
 	size_t sz;
 	int ret;
 
-	ret = mtd->read(mtd, ptr, sizeof(fs), &sz, (u_char *) &fs);
+	ret = mtd->read(mtd, ptr, sizeof(fs), &sz, (u_char *) &fs, NULL);
 	if (ret >= 0 && sz != sizeof(fs))
 		ret = -EINVAL;
 
@@ -132,7 +132,7 @@  afs_read_iis(struct mtd_info *mtd, struct image_info_struct *iis, u_int ptr)
 	int ret, i;
 
 	memset(iis, 0, sizeof(*iis));
-	ret = mtd->read(mtd, ptr, sizeof(*iis), &sz, (u_char *) iis);
+	ret = mtd->read(mtd, ptr, sizeof(*iis), &sz, (u_char *) iis, NULL);
 	if (ret < 0)
 		goto failed;
 
diff --git a/drivers/mtd/ar7part.c b/drivers/mtd/ar7part.c
index f40ea45..9c8e10a 100644
--- a/drivers/mtd/ar7part.c
+++ b/drivers/mtd/ar7part.c
@@ -74,7 +74,7 @@  static int create_mtd_partitions(struct mtd_info *master,
 	do { /* Try 10 blocks starting from master->erasesize */
 		offset = pre_size;
 		master->read(master, offset,
-			     sizeof(header), &len, (uint8_t *)&header);
+			     sizeof(header), &len, (uint8_t *)&header, NULL);
 		if (!strncmp((char *)&header, "TIENV0.8", 8))
 			ar7_parts[1].offset = pre_size;
 		if (header.checksum == LOADER_MAGIC1)
@@ -96,7 +96,7 @@  static int create_mtd_partitions(struct mtd_info *master,
 		while (header.length) {
 			offset += sizeof(header) + header.length;
 			master->read(master, offset, sizeof(header),
-				     &len, (uint8_t *)&header);
+				     &len, (uint8_t *)&header, NULL);
 		}
 		root_offset = offset + sizeof(header) + 4;
 		break;
@@ -104,7 +104,7 @@  static int create_mtd_partitions(struct mtd_info *master,
 		while (header.length) {
 			offset += sizeof(header) + header.length;
 			master->read(master, offset, sizeof(header),
-				     &len, (uint8_t *)&header);
+				     &len, (uint8_t *)&header, NULL);
 		}
 		root_offset = offset + sizeof(header) + 4 + 0xff;
 		root_offset &= ~(uint32_t)0xff;
@@ -115,7 +115,7 @@  static int create_mtd_partitions(struct mtd_info *master,
 	}
 
 	master->read(master, root_offset,
-		sizeof(header), &len, (u8 *)&header);
+		     sizeof(header), &len, (u8 *)&header, NULL);
 	if (header.checksum != SQUASHFS_MAGIC) {
 		root_offset += master->erasesize - 1;
 		root_offset &= ~(master->erasesize - 1);
diff --git a/drivers/mtd/chips/cfi_cmdset_0001.c b/drivers/mtd/chips/cfi_cmdset_0001.c
index e1e122f..128ad70 100644
--- a/drivers/mtd/chips/cfi_cmdset_0001.c
+++ b/drivers/mtd/chips/cfi_cmdset_0001.c
@@ -54,7 +54,8 @@ 
 #define AT49BV640D	0x02de
 #define AT49BV640DT	0x02db
 
-static int cfi_intelext_read (struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int cfi_intelext_read(struct mtd_info *, loff_t, size_t, size_t *,
+			      u_char *, unsigned int *);
 static int cfi_intelext_write_words(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int cfi_intelext_write_buffers(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int cfi_intelext_writev(struct mtd_info *, const struct kvec *, unsigned long, loff_t, size_t *);
@@ -1444,7 +1445,9 @@  static inline int do_read_onechip(struct map_info *map, struct flchip *chip, lof
 	return 0;
 }
 
-static int cfi_intelext_read (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int cfi_intelext_read(struct mtd_info *mtd, loff_t from, size_t len,
+			     size_t *retlen, u_char *buf,
+			     unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 	struct cfi_private *cfi = map->fldrv_priv;
@@ -1458,6 +1461,9 @@  static int cfi_intelext_read (struct mtd_info *mtd, loff_t from, size_t len, siz
 
 	*retlen = 0;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	while (len) {
 		unsigned long thislen;
 
diff --git a/drivers/mtd/chips/cfi_cmdset_0002.c b/drivers/mtd/chips/cfi_cmdset_0002.c
index 8d70895..1c7ac7a 100644
--- a/drivers/mtd/chips/cfi_cmdset_0002.c
+++ b/drivers/mtd/chips/cfi_cmdset_0002.c
@@ -48,7 +48,8 @@ 
 #define SST49LF008A		0x005a
 #define AT49BV6416		0x00d6
 
-static int cfi_amdstd_read (struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int cfi_amdstd_read(struct mtd_info *, loff_t, size_t, size_t *,
+			   u_char *, unsigned int *);
 static int cfi_amdstd_write_words(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int cfi_amdstd_write_buffers(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int cfi_amdstd_erase_chip(struct mtd_info *, struct erase_info *);
@@ -1004,7 +1005,9 @@  static inline int do_read_onechip(struct map_info *map, struct flchip *chip, lof
 }
 
 
-static int cfi_amdstd_read (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int cfi_amdstd_read(struct mtd_info *mtd, loff_t from, size_t len,
+			   size_t *retlen, u_char *buf,
+			   unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 	struct cfi_private *cfi = map->fldrv_priv;
@@ -1020,6 +1023,9 @@  static int cfi_amdstd_read (struct mtd_info *mtd, loff_t from, size_t len, size_
 
 	*retlen = 0;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	while (len) {
 		unsigned long thislen;
 
diff --git a/drivers/mtd/chips/cfi_cmdset_0020.c b/drivers/mtd/chips/cfi_cmdset_0020.c
index 666c52f..f91af48 100644
--- a/drivers/mtd/chips/cfi_cmdset_0020.c
+++ b/drivers/mtd/chips/cfi_cmdset_0020.c
@@ -35,7 +35,8 @@ 
 #include <linux/mtd/mtd.h>
 
 
-static int cfi_staa_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int cfi_staa_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *,
+			 unsigned int *);
 static int cfi_staa_write_buffers(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int cfi_staa_writev(struct mtd_info *mtd, const struct kvec *vecs,
 		unsigned long count, loff_t to, size_t *retlen);
@@ -382,7 +383,9 @@  static inline int do_read_onechip(struct map_info *map, struct flchip *chip, lof
 	return 0;
 }
 
-static int cfi_staa_read (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int cfi_staa_read(struct mtd_info *mtd, loff_t from, size_t len,
+			 size_t *retlen, u_char *buf,
+			 unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 	struct cfi_private *cfi = map->fldrv_priv;
@@ -396,6 +399,9 @@  static int cfi_staa_read (struct mtd_info *mtd, loff_t from, size_t len, size_t
 
 	*retlen = 0;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	while (len) {
 		unsigned long thislen;
 
diff --git a/drivers/mtd/chips/map_absent.c b/drivers/mtd/chips/map_absent.c
index f2b8729..3649fbf 100644
--- a/drivers/mtd/chips/map_absent.c
+++ b/drivers/mtd/chips/map_absent.c
@@ -26,7 +26,8 @@ 
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/map.h>
 
-static int map_absent_read (struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int map_absent_read(struct mtd_info *, loff_t, size_t, size_t *,
+			   u_char *, unsigned int *);
 static int map_absent_write (struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int map_absent_erase (struct mtd_info *, struct erase_info *);
 static void map_absent_sync (struct mtd_info *);
@@ -68,7 +69,9 @@  static struct mtd_info *map_absent_probe(struct map_info *map)
 }
 
 
-static int map_absent_read(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int map_absent_read(struct mtd_info *mtd, loff_t from, size_t len,
+			   size_t *retlen, u_char *buf,
+			   unsigned int *max_bitflips)
 {
 	*retlen = 0;
 	return -ENODEV;
diff --git a/drivers/mtd/chips/map_ram.c b/drivers/mtd/chips/map_ram.c
index 67640cc..2dd7c63 100644
--- a/drivers/mtd/chips/map_ram.c
+++ b/drivers/mtd/chips/map_ram.c
@@ -15,7 +15,8 @@ 
 #include <linux/mtd/map.h>
 
 
-static int mapram_read (struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int mapram_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *,
+		       unsigned int *);
 static int mapram_write (struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static int mapram_erase (struct mtd_info *, struct erase_info *);
 static void mapram_nop (struct mtd_info *);
@@ -95,12 +96,15 @@  static unsigned long mapram_unmapped_area(struct mtd_info *mtd,
 	return (unsigned long) map->virt + offset;
 }
 
-static int mapram_read (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int mapram_read(struct mtd_info *mtd, loff_t from, size_t len,
+		       size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 
 	map_copy_from(map, buf, from, len);
 	*retlen = len;
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
 	return 0;
 }
 
diff --git a/drivers/mtd/chips/map_rom.c b/drivers/mtd/chips/map_rom.c
index 593f73d..b802c13 100644
--- a/drivers/mtd/chips/map_rom.c
+++ b/drivers/mtd/chips/map_rom.c
@@ -14,7 +14,8 @@ 
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/map.h>
 
-static int maprom_read (struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int maprom_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *,
+		       unsigned int *);
 static int maprom_write (struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 static void maprom_nop (struct mtd_info *);
 static struct mtd_info *map_rom_probe(struct map_info *map);
@@ -69,12 +70,15 @@  static unsigned long maprom_unmapped_area(struct mtd_info *mtd,
 	return (unsigned long) map->virt + offset;
 }
 
-static int maprom_read (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf)
+static int maprom_read(struct mtd_info *mtd, loff_t from, size_t len,
+		       size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 
 	map_copy_from(map, buf, from, len);
 	*retlen = len;
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
 	return 0;
 }
 
diff --git a/drivers/mtd/devices/block2mtd.c b/drivers/mtd/devices/block2mtd.c
index b78f231..8e989c6 100644
--- a/drivers/mtd/devices/block2mtd.c
+++ b/drivers/mtd/devices/block2mtd.c
@@ -97,7 +97,8 @@  static int block2mtd_erase(struct mtd_info *mtd, struct erase_info *instr)
 
 
 static int block2mtd_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf)
+			  size_t *retlen, u_char *buf,
+			  unsigned int *max_bitflips)
 {
 	struct block2mtd_dev *dev = mtd->priv;
 	struct page *page;
@@ -105,6 +106,9 @@  static int block2mtd_read(struct mtd_info *mtd, loff_t from, size_t len,
 	int offset = from & (PAGE_SIZE-1);
 	int cpylen;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	if (from > mtd->size)
 		return -EINVAL;
 	if (from + len > mtd->size)
diff --git a/drivers/mtd/devices/doc2000.c b/drivers/mtd/devices/doc2000.c
index e9fad91..c3413f4 100644
--- a/drivers/mtd/devices/doc2000.c
+++ b/drivers/mtd/devices/doc2000.c
@@ -48,7 +48,7 @@ 
 */
 
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-		    size_t *retlen, u_char *buf);
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips);
 static int doc_write(struct mtd_info *mtd, loff_t to, size_t len,
 		     size_t *retlen, const u_char *buf);
 static int doc_read_oob(struct mtd_info *mtd, loff_t ofs,
@@ -601,7 +601,7 @@  void DoC2k_init(struct mtd_info *mtd)
 EXPORT_SYMBOL_GPL(DoC2k_init);
 
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-		    size_t * retlen, u_char * buf)
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct DiskOnChip *this = mtd->priv;
 	void __iomem *docptr = this->virtadr;
@@ -610,6 +610,7 @@  static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
 	volatile char dummy;
 	int i, len256 = 0, ret=0;
 	size_t left = len;
+	unsigned int bitflips = 0;
 
 	/* Don't allow read past end of device */
 	if (from >= this->totlen)
@@ -716,6 +717,8 @@  static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
 				   checking *retlen */
 				ret = -EIO;
 			}
+			bitflips = max_t(unsigned int, bitflips,
+					 (unsigned)nb_errors);
 		}
 
 #ifdef PSYCHO_DEBUG
@@ -739,6 +742,9 @@  static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
 		buf += len;
 	}
 
+	if (max_bitflips != NULL)
+		*max_bitflips = bitflips;
+
 	mutex_unlock(&this->lock);
 
 	return ret;
@@ -970,6 +976,7 @@  static int doc_read_oob(struct mtd_info *mtd, loff_t ofs,
 
 	DoC_ReadBuf(this, &buf[len256], len - len256);
 
+	ops->max_bitflips = 0;
 	ops->retlen = len;
 	/* Reading the full OOB data drops us off of the end of the page,
          * causing the flash device to go into busy mode, so we need
diff --git a/drivers/mtd/devices/doc2001.c b/drivers/mtd/devices/doc2001.c
index a3f7a27..58fa69f 100644
--- a/drivers/mtd/devices/doc2001.c
+++ b/drivers/mtd/devices/doc2001.c
@@ -29,7 +29,7 @@ 
 #undef USE_MEMCPY
 
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-		    size_t *retlen, u_char *buf);
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips);
 static int doc_write(struct mtd_info *mtd, loff_t to, size_t len,
 		     size_t *retlen, const u_char *buf);
 static int doc_read_oob(struct mtd_info *mtd, loff_t ofs,
@@ -382,7 +382,7 @@  void DoCMil_init(struct mtd_info *mtd)
 EXPORT_SYMBOL_GPL(DoCMil_init);
 
 static int doc_read (struct mtd_info *mtd, loff_t from, size_t len,
-		     size_t *retlen, u_char *buf)
+		     size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	int i, ret;
 	volatile char dummy;
@@ -478,6 +478,8 @@  static int doc_read (struct mtd_info *mtd, loff_t from, size_t len,
 			   MTD-aware stuff can know about it by checking *retlen */
 			ret = -EIO;
 		}
+		if (max_bitflips != NULL)
+			*max_bitflips = (unsigned)nb_errors;
 	}
 
 #ifdef PSYCHO_DEBUG
@@ -671,6 +673,7 @@  static int doc_read_oob(struct mtd_info *mtd, loff_t ofs,
 	buf[len - 1] = ReadDOC(docptr, LastDataRead);
 
 	ops->retlen = len;
+	ops->max_bitflips = 0;
 
 	return 0;
 }
diff --git a/drivers/mtd/devices/doc2001plus.c b/drivers/mtd/devices/doc2001plus.c
index 99351bc..df62fce 100644
--- a/drivers/mtd/devices/doc2001plus.c
+++ b/drivers/mtd/devices/doc2001plus.c
@@ -33,7 +33,7 @@ 
 #undef USE_MEMCPY
 
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf);
+		size_t *retlen, u_char *buf, unsigned int *max_bitflips);
 static int doc_write(struct mtd_info *mtd, loff_t to, size_t len,
 		size_t *retlen, const u_char *buf);
 static int doc_read_oob(struct mtd_info *mtd, loff_t ofs,
@@ -580,7 +580,7 @@  static int doc_dumpblk(struct mtd_info *mtd, loff_t from)
 #endif
 
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-		    size_t *retlen, u_char *buf)
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	int ret, i;
 	volatile char dummy;
@@ -682,6 +682,8 @@  static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
 #endif
 				ret = -EIO;
 		}
+		if (max_bitflips != NULL)
+			*max_bitflips = (unsigned)nb_errors;
 	}
 
 #ifdef PSYCHO_DEBUG
diff --git a/drivers/mtd/devices/docg3.c b/drivers/mtd/devices/docg3.c
index f7490a014..266a893 100644
--- a/drivers/mtd/devices/docg3.c
+++ b/drivers/mtd/devices/docg3.c
@@ -938,6 +938,8 @@  static int doc_read_oob(struct mtd_info *mtd, loff_t from,
 			}
 			if (ret > 0) {
 				mtd->ecc_stats.corrected += ret;
+				ops->max_bitflips = max(ops->max_bitflips,
+							(unsigned int)ret);
 				ret = -EUCLEAN;
 			}
 		}
@@ -974,7 +976,7 @@  err:
  * Returns 0 if read successfull, of -EIO, -EINVAL if an error occured
  */
 static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
-	     size_t *retlen, u_char *buf)
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct mtd_oob_ops ops;
 	size_t ret;
@@ -986,6 +988,8 @@  static int doc_read(struct mtd_info *mtd, loff_t from, size_t len,
 
 	ret = doc_read_oob(mtd, from, &ops);
 	*retlen = ops.retlen;
+	if (max_bitflips != NULL)
+		*max_bitflips = ops.max_bitflips;
 	return ret;
 }
 
diff --git a/drivers/mtd/devices/lart.c b/drivers/mtd/devices/lart.c
index 3a11ea6..09f37b0 100644
--- a/drivers/mtd/devices/lart.c
+++ b/drivers/mtd/devices/lart.c
@@ -434,7 +434,8 @@  static int flash_erase (struct mtd_info *mtd,struct erase_info *instr)
    return (0);
 }
 
-static int flash_read (struct mtd_info *mtd,loff_t from,size_t len,size_t *retlen,u_char *buf)
+static int flash_read(struct mtd_info *mtd, loff_t from, size_t len,
+		      size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 #ifdef LART_DEBUG
    printk (KERN_DEBUG "%s(from = 0x%.8x, len = %d)\n", __func__, (__u32)from, len);
@@ -469,6 +470,8 @@  static int flash_read (struct mtd_info *mtd,loff_t from,size_t len,size_t *retle
    if (len & (BUSWIDTH - 1))
 	 while (len--) *buf++ = read8 (from++);
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
    return (0);
 }
 
diff --git a/drivers/mtd/devices/m25p80.c b/drivers/mtd/devices/m25p80.c
index 7c60ddd..c42f5dd 100644
--- a/drivers/mtd/devices/m25p80.c
+++ b/drivers/mtd/devices/m25p80.c
@@ -340,7 +340,7 @@  static int m25p80_erase(struct mtd_info *mtd, struct erase_info *instr)
  * may be any size provided it is within the physical boundaries.
  */
 static int m25p80_read(struct mtd_info *mtd, loff_t from, size_t len,
-	size_t *retlen, u_char *buf)
+	size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct m25p *flash = mtd_to_m25p(mtd);
 	struct spi_transfer t[2];
@@ -356,6 +356,9 @@  static int m25p80_read(struct mtd_info *mtd, loff_t from, size_t len,
 	if (from + len > flash->mtd.size)
 		return -EINVAL;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	spi_message_init(&m);
 	memset(t, 0, (sizeof t));
 
diff --git a/drivers/mtd/devices/ms02-nv.c b/drivers/mtd/devices/ms02-nv.c
index 8423fb6..0a478fd 100644
--- a/drivers/mtd/devices/ms02-nv.c
+++ b/drivers/mtd/devices/ms02-nv.c
@@ -55,7 +55,8 @@  static struct mtd_info *root_ms02nv_mtd;
 
 
 static int ms02nv_read(struct mtd_info *mtd, loff_t from,
-			size_t len, size_t *retlen, u_char *buf)
+		       size_t len, size_t *retlen, u_char *buf,
+		       unsigned int *max_bitflips)
 {
 	struct ms02nv_private *mp = mtd->priv;
 
@@ -64,7 +65,8 @@  static int ms02nv_read(struct mtd_info *mtd, loff_t from,
 
 	memcpy(buf, mp->uaddr + from, len);
 	*retlen = len;
-
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
 	return 0;
 }
 
diff --git a/drivers/mtd/devices/mtd_dataflash.c b/drivers/mtd/devices/mtd_dataflash.c
index 236057e..825a5f6 100644
--- a/drivers/mtd/devices/mtd_dataflash.c
+++ b/drivers/mtd/devices/mtd_dataflash.c
@@ -240,7 +240,8 @@  static int dataflash_erase(struct mtd_info *mtd, struct erase_info *instr)
  *   buf    : Buffer containing the data
  */
 static int dataflash_read(struct mtd_info *mtd, loff_t from, size_t len,
-			       size_t *retlen, u_char *buf)
+			  size_t *retlen, u_char *buf,
+			  unsigned int *max_bitflips)
 {
 	struct dataflash	*priv = mtd->priv;
 	struct spi_transfer	x[2] = { { .tx_dma = 0, }, };
@@ -253,6 +254,8 @@  static int dataflash_read(struct mtd_info *mtd, loff_t from, size_t len,
 			(unsigned)from, (unsigned)(from + len));
 
 	*retlen = 0;
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
 
 	/* Sanity checks */
 	if (!len)
diff --git a/drivers/mtd/devices/mtdram.c b/drivers/mtd/devices/mtdram.c
index 2562689..7e36b65 100644
--- a/drivers/mtd/devices/mtdram.c
+++ b/drivers/mtd/devices/mtdram.c
@@ -78,13 +78,15 @@  static unsigned long ram_get_unmapped_area(struct mtd_info *mtd,
 }
 
 static int ram_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf)
+		    size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	if (from + len > mtd->size)
 		return -EINVAL;
 
 	memcpy(buf, mtd->priv + from, len);
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
 	*retlen = len;
 	return 0;
 }
diff --git a/drivers/mtd/devices/phram.c b/drivers/mtd/devices/phram.c
index 23423bd..f1175d2 100644
--- a/drivers/mtd/devices/phram.c
+++ b/drivers/mtd/devices/phram.c
@@ -75,13 +75,16 @@  static void phram_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
 }
 
 static int phram_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf)
+		size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	u_char *start = mtd->priv;
 
 	if (from >= mtd->size)
 		return -EINVAL;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	if (len > mtd->size - from)
 		len = mtd->size - from;
 
diff --git a/drivers/mtd/devices/pmc551.c b/drivers/mtd/devices/pmc551.c
index ecff765..72d1b47 100644
--- a/drivers/mtd/devices/pmc551.c
+++ b/drivers/mtd/devices/pmc551.c
@@ -214,7 +214,8 @@  static void pmc551_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
 }
 
 static int pmc551_read(struct mtd_info *mtd, loff_t from, size_t len,
-			size_t * retlen, u_char * buf)
+		       size_t *retlen, u_char *buf,
+		       unsigned int *max_bitflips)
 {
 	struct mypriv *priv = mtd->priv;
 	u32 soff_hi, soff_lo;	/* start address offset hi/lo */
@@ -239,6 +240,9 @@  static int pmc551_read(struct mtd_info *mtd, loff_t from, size_t len,
 		return -EINVAL;
 	}
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	soff_hi = from & ~(priv->asize - 1);
 	eoff_hi = end & ~(priv->asize - 1);
 	soff_lo = from & (priv->asize - 1);
diff --git a/drivers/mtd/devices/slram.c b/drivers/mtd/devices/slram.c
index e585263..4b56578 100644
--- a/drivers/mtd/devices/slram.c
+++ b/drivers/mtd/devices/slram.c
@@ -77,7 +77,8 @@  static int slram_erase(struct mtd_info *, struct erase_info *);
 static int slram_point(struct mtd_info *, loff_t, size_t, size_t *, void **,
 		resource_size_t *);
 static void slram_unpoint(struct mtd_info *, loff_t, size_t);
-static int slram_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int slram_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *,
+		      unsigned int *);
 static int slram_write(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);
 
 static int slram_erase(struct mtd_info *mtd, struct erase_info *instr)
@@ -124,13 +125,16 @@  static void slram_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
 }
 
 static int slram_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf)
+		size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	slram_priv_t *priv = mtd->priv;
 
 	if (from > mtd->size)
 		return -EINVAL;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	if (from + len > mtd->size)
 		len = mtd->size - from;
 
diff --git a/drivers/mtd/devices/sst25l.c b/drivers/mtd/devices/sst25l.c
index e45f62e..95bb989 100644
--- a/drivers/mtd/devices/sst25l.c
+++ b/drivers/mtd/devices/sst25l.c
@@ -215,7 +215,8 @@  static int sst25l_erase(struct mtd_info *mtd, struct erase_info *instr)
 }
 
 static int sst25l_read(struct mtd_info *mtd, loff_t from, size_t len,
-		       size_t *retlen, unsigned char *buf)
+		       size_t *retlen, unsigned char *buf,
+		       unsigned int *max_bitflips)
 {
 	struct sst25l_flash *flash = to_sst25l_flash(mtd);
 	struct spi_transfer transfer[2];
@@ -223,6 +224,9 @@  static int sst25l_read(struct mtd_info *mtd, loff_t from, size_t len,
 	unsigned char command[4];
 	int ret;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	/* Sanity checking */
 	if (len == 0)
 		return 0;
diff --git a/drivers/mtd/ftl.c b/drivers/mtd/ftl.c
index c7382bb..c2d3b7e 100644
--- a/drivers/mtd/ftl.c
+++ b/drivers/mtd/ftl.c
@@ -161,6 +161,7 @@  static int scan_header(partition_t *part)
     loff_t offset, max_offset;
     size_t ret;
     int err;
+
     part->header.FormattedSize = 0;
     max_offset = (0x100000<part->mbd.mtd->size)?0x100000:part->mbd.mtd->size;
     /* Search first megabyte for a valid FTL header */
@@ -169,7 +170,7 @@  static int scan_header(partition_t *part)
 	 offset += part->mbd.mtd->erasesize ? : 0x2000) {
 
 	err = part->mbd.mtd->read(part->mbd.mtd, offset, sizeof(header), &ret,
-			      (unsigned char *)&header);
+				  (unsigned char *)&header, NULL);
 
 	if (err)
 	    return err;
@@ -225,7 +226,7 @@  static int build_maps(partition_t *part)
 	offset = ((i + le16_to_cpu(part->header.FirstPhysicalEUN))
 		      << part->header.EraseUnitSize);
 	ret = part->mbd.mtd->read(part->mbd.mtd, offset, sizeof(header), &retval,
-			      (unsigned char *)&header);
+				  (unsigned char *)&header, NULL);
 
 	if (ret)
 	    goto out_XferInfo;
@@ -291,7 +292,7 @@  static int build_maps(partition_t *part)
 
 	ret = part->mbd.mtd->read(part->mbd.mtd, offset,
 			      part->BlocksPerUnit * sizeof(uint32_t), &retval,
-			      (unsigned char *)part->bam_cache);
+			      (unsigned char *)part->bam_cache, NULL);
 
 	if (ret)
 		goto out_bam_cache;
@@ -486,8 +487,8 @@  static int copy_erase_unit(partition_t *part, uint16_t srcunit,
 	offset = eun->Offset + le32_to_cpu(part->header.BAMOffset);
 
 	ret = part->mbd.mtd->read(part->mbd.mtd, offset,
-			      part->BlocksPerUnit * sizeof(uint32_t),
-			      &retlen, (u_char *) (part->bam_cache));
+				  part->BlocksPerUnit * sizeof(uint32_t),
+				  &retlen, (u_char *) (part->bam_cache), NULL);
 
 	/* mark the cache bad, in case we get an error later */
 	part->bam_index = 0xffff;
@@ -523,8 +524,8 @@  static int copy_erase_unit(partition_t *part, uint16_t srcunit,
 	    break;
 	case BLOCK_DATA:
 	case BLOCK_REPLACEMENT:
-	    ret = part->mbd.mtd->read(part->mbd.mtd, src, SECTOR_SIZE,
-                        &retlen, (u_char *) buf);
+		ret = part->mbd.mtd->read(part->mbd.mtd, src, SECTOR_SIZE,
+					  &retlen, (u_char *) buf, NULL);
 	    if (ret) {
 		printk(KERN_WARNING "ftl: Error reading old xfer unit in copy_erase_unit\n");
 		return ret;
@@ -744,13 +745,14 @@  static uint32_t find_free(partition_t *part)
 
     /* Is this unit's BAM cached? */
     if (eun != part->bam_index) {
+
 	/* Invalidate cache */
 	part->bam_index = 0xffff;
 
 	ret = part->mbd.mtd->read(part->mbd.mtd,
 		       part->EUNInfo[eun].Offset + le32_to_cpu(part->header.BAMOffset),
 		       part->BlocksPerUnit * sizeof(uint32_t),
-		       &retlen, (u_char *) (part->bam_cache));
+		       &retlen, (u_char *) (part->bam_cache), NULL);
 
 	if (ret) {
 	    printk(KERN_WARNING"ftl: Error reading BAM in find_free\n");
@@ -811,7 +813,7 @@  static int ftl_read(partition_t *part, caddr_t buffer,
 	    offset = (part->EUNInfo[log_addr / bsize].Offset
 			  + (log_addr % bsize));
 	    ret = part->mbd.mtd->read(part->mbd.mtd, offset, SECTOR_SIZE,
-			   &retlen, (u_char *) buffer);
+				      &retlen, (u_char *) buffer, NULL);
 
 	    if (ret) {
 		printk(KERN_WARNING "Error reading MTD device in ftl_read()\n");
@@ -849,8 +851,8 @@  static int set_bam_entry(partition_t *part, uint32_t log_addr,
 		  le32_to_cpu(part->header.BAMOffset));
 
 #ifdef PSYCHO_DEBUG
-    ret = part->mbd.mtd->read(part->mbd.mtd, offset, sizeof(uint32_t),
-                        &retlen, (u_char *)&old_addr);
+	ret = part->mbd.mtd->read(part->mbd.mtd, offset, sizeof(uint32_t),
+				  &retlen, (u_char *)&old_addr, NULL);
     if (ret) {
 	printk(KERN_WARNING"ftl: Error reading old_addr in set_bam_entry: %d\n",ret);
 	return ret;
diff --git a/drivers/mtd/inftlcore.c b/drivers/mtd/inftlcore.c
index dd034ef..4f5e301 100644
--- a/drivers/mtd/inftlcore.c
+++ b/drivers/mtd/inftlcore.c
@@ -345,12 +345,12 @@  static u16 INFTL_foldchain(struct INFTLrecord *inftl, unsigned thisVUC, unsigned
 
 		ret = mtd->read(mtd, (inftl->EraseSize * BlockMap[block]) +
 				(block * SECTORSIZE), SECTORSIZE, &retlen,
-				movebuf);
+				movebuf, NULL);
 		if (ret < 0 && !mtd_is_bitflip(ret)) {
 			ret = mtd->read(mtd,
 					(inftl->EraseSize * BlockMap[block]) +
 					(block * SECTORSIZE), SECTORSIZE,
-					&retlen, movebuf);
+					&retlen, movebuf, NULL);
 			if (ret != -EIO)
 				pr_debug("INFTL: error went away on retry?\n");
 		}
@@ -914,7 +914,8 @@  foundit:
 	} else {
 		size_t retlen;
 		loff_t ptr = (thisEUN * inftl->EraseSize) + blockofs;
-		int ret = mtd->read(mtd, ptr, SECTORSIZE, &retlen, buffer);
+		int ret = mtd->read(mtd, ptr, SECTORSIZE, &retlen, buffer,
+				    NULL);
 
 		/* Handle corrected bit flips gracefully */
 		if (ret < 0 && !mtd_is_bitflip(ret))
diff --git a/drivers/mtd/inftlmount.c b/drivers/mtd/inftlmount.c
index 2ff601f..5eff38b 100644
--- a/drivers/mtd/inftlmount.c
+++ b/drivers/mtd/inftlmount.c
@@ -74,7 +74,7 @@  static int find_boot_record(struct INFTLrecord *inftl)
 		 * but later checks fail.
 		 */
 		ret = mtd->read(mtd, block * inftl->EraseSize,
-				SECTORSIZE, &retlen, buf);
+				SECTORSIZE, &retlen, buf, NULL);
 		/* We ignore ret in case the ECC of the MediaHeader is invalid
 		   (which is apparently acceptable) */
 		if (retlen != SECTORSIZE) {
@@ -119,7 +119,7 @@  static int find_boot_record(struct INFTLrecord *inftl)
 
 		/* Read the spare media header at offset 4096 */
 		mtd->read(mtd, block * inftl->EraseSize + 4096,
-			  SECTORSIZE, &retlen, buf);
+			  SECTORSIZE, &retlen, buf, NULL);
 		if (retlen != SECTORSIZE) {
 			printk(KERN_WARNING "INFTL: Unable to read spare "
 			       "Media Header\n");
@@ -342,7 +342,7 @@  static int check_free_sectors(struct INFTLrecord *inftl, unsigned int address,
 	int i;
 
 	for (i = 0; i < len; i += SECTORSIZE) {
-		if (mtd->read(mtd, address, SECTORSIZE, &retlen, buf))
+		if (mtd->read(mtd, address, SECTORSIZE, &retlen, buf, NULL))
 			return -1;
 		if (memcmpb(buf, 0xff, SECTORSIZE) != 0)
 			return -1;
diff --git a/drivers/mtd/lpddr/lpddr_cmds.c b/drivers/mtd/lpddr/lpddr_cmds.c
index 1dca31d..7960cbf 100644
--- a/drivers/mtd/lpddr/lpddr_cmds.c
+++ b/drivers/mtd/lpddr/lpddr_cmds.c
@@ -30,7 +30,7 @@ 
 #include <linux/module.h>
 
 static int lpddr_read(struct mtd_info *mtd, loff_t adr, size_t len,
-					size_t *retlen, u_char *buf);
+		      size_t *retlen, u_char *buf, unsigned int *max_bitflips);
 static int lpddr_write_buffers(struct mtd_info *mtd, loff_t to,
 				size_t len, size_t *retlen, const u_char *buf);
 static int lpddr_writev(struct mtd_info *mtd, const struct kvec *vecs,
@@ -504,7 +504,7 @@  int do_erase_oneblock(struct mtd_info *mtd, loff_t adr)
 }
 
 static int lpddr_read(struct mtd_info *mtd, loff_t adr, size_t len,
-			size_t *retlen, u_char *buf)
+		      size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct map_info *map = mtd->priv;
 	struct lpddr_private *lpddr = map->fldrv_priv;
@@ -513,6 +513,7 @@  static int lpddr_read(struct mtd_info *mtd, loff_t adr, size_t len,
 	int ret = 0;
 
 	mutex_lock(&chip->mutex);
+
 	ret = get_chip(map, chip, FL_READY);
 	if (ret) {
 		mutex_unlock(&chip->mutex);
@@ -522,6 +523,9 @@  static int lpddr_read(struct mtd_info *mtd, loff_t adr, size_t len,
 	map_copy_from(map, buf, adr, len);
 	*retlen = len;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = 0;
+
 	put_chip(map, chip);
 	mutex_unlock(&chip->mutex);
 	return ret;
diff --git a/drivers/mtd/maps/bcm963xx-flash.c b/drivers/mtd/maps/bcm963xx-flash.c
index c7d3949..03645bd 100644
--- a/drivers/mtd/maps/bcm963xx-flash.c
+++ b/drivers/mtd/maps/bcm963xx-flash.c
@@ -67,7 +67,7 @@  static int parse_cfe_partitions(struct mtd_info *master,
 
 	/* Get the tag */
 	ret = master->read(master, master->erasesize, sizeof(struct bcm_tag),
-							&retlen, (void *)buf);
+			   &retlen, (void *)buf, NULL);
 	if (retlen != sizeof(struct bcm_tag)) {
 		vfree(buf);
 		return -EIO;
@@ -160,7 +160,7 @@  static int bcm963xx_detect_cfe(struct mtd_info *master)
 	int ret;
 	size_t retlen;
 
-	ret = master->read(master, idoffset, 8, &retlen, (void *)buf);
+	ret = master->read(master, idoffset, 8, &retlen, (void *)buf, NULL);
 	buf[retlen] = 0;
 	printk(KERN_INFO PFX "Read Signature value of %s\n", buf);
 
diff --git a/drivers/mtd/mtdblock.c b/drivers/mtd/mtdblock.c
index 7c1dc90..f45a46a 100644
--- a/drivers/mtd/mtdblock.c
+++ b/drivers/mtd/mtdblock.c
@@ -185,7 +185,8 @@  static int do_cached_write (struct mtdblk_dev *mtdblk, unsigned long pos,
 				/* fill the cache with the current sector */
 				mtdblk->cache_state = STATE_EMPTY;
 				ret = mtd->read(mtd, sect_start, sect_size,
-						&retlen, mtdblk->cache_data);
+						&retlen, mtdblk->cache_data,
+						NULL);
 				if (ret)
 					return ret;
 				if (retlen != sect_size)
@@ -222,7 +223,7 @@  static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
 			mtd->name, pos, len);
 
 	if (!sect_size)
-		return mtd->read(mtd, pos, len, &retlen, buf);
+		return mtd->read(mtd, pos, len, &retlen, buf, NULL);
 
 	while (len > 0) {
 		unsigned long sect_start = (pos/sect_size)*sect_size;
@@ -241,7 +242,7 @@  static int do_cached_read (struct mtdblk_dev *mtdblk, unsigned long pos,
 		    mtdblk->cache_offset == sect_start) {
 			memcpy (buf, mtdblk->cache_data + offset, size);
 		} else {
-			ret = mtd->read(mtd, pos, size, &retlen, buf);
+			ret = mtd->read(mtd, pos, size, &retlen, buf, NULL);
 			if (ret)
 				return ret;
 			if (retlen != size)
diff --git a/drivers/mtd/mtdblock_ro.c b/drivers/mtd/mtdblock_ro.c
index 0470a6e..33b4e22 100644
--- a/drivers/mtd/mtdblock_ro.c
+++ b/drivers/mtd/mtdblock_ro.c
@@ -30,7 +30,7 @@  static int mtdblock_readsect(struct mtd_blktrans_dev *dev,
 {
 	size_t retlen;
 
-	if (dev->mtd->read(dev->mtd, (block * 512), 512, &retlen, buf))
+	if (dev->mtd->read(dev->mtd, (block * 512), 512, &retlen, buf, NULL))
 		return 1;
 	return 0;
 }
diff --git a/drivers/mtd/mtdchar.c b/drivers/mtd/mtdchar.c
index e7dc732..bb70d80 100644
--- a/drivers/mtd/mtdchar.c
+++ b/drivers/mtd/mtdchar.c
@@ -231,7 +231,7 @@  static ssize_t mtd_read(struct file *file, char __user *buf, size_t count,loff_t
 			break;
 		}
 		default:
-			ret = mtd->read(mtd, *ppos, len, &retlen, kbuf);
+			ret = mtd->read(mtd, *ppos, len, &retlen, kbuf, NULL);
 		}
 		/* Nand returns -EBADMSG on ECC errors, but it returns
 		 * the data. For our userspace tools it is important
diff --git a/drivers/mtd/mtdconcat.c b/drivers/mtd/mtdconcat.c
index 6df4d4d..5102ba4 100644
--- a/drivers/mtd/mtdconcat.c
+++ b/drivers/mtd/mtdconcat.c
@@ -66,17 +66,19 @@  struct mtd_concat {
 
 static int
 concat_read(struct mtd_info *mtd, loff_t from, size_t len,
-	    size_t * retlen, u_char * buf)
+	    size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct mtd_concat *concat = CONCAT(mtd);
 	int ret = 0, err;
 	int i;
+	unsigned int bitflips = 0;
 
 	*retlen = 0;
 
 	for (i = 0; i < concat->num_subdev; i++) {
 		struct mtd_info *subdev = concat->subdev[i];
 		size_t size, retsize;
+		unsigned int dev_bitflips;
 
 		if (from >= subdev->size) {
 			/* Not destined for this subdev */
@@ -91,7 +93,9 @@  concat_read(struct mtd_info *mtd, loff_t from, size_t len,
 			/* Entire transaction goes into this subdev */
 			size = len;
 
-		err = subdev->read(subdev, from, size, &retsize, buf);
+		err = subdev->read(subdev, from, size, &retsize, buf,
+				   &dev_bitflips);
+		bitflips = max(bitflips, dev_bitflips);
 
 		/* Save information about bitflips! */
 		if (unlikely(err)) {
@@ -109,8 +113,11 @@  concat_read(struct mtd_info *mtd, loff_t from, size_t len,
 
 		*retlen += retsize;
 		len -= size;
-		if (len == 0)
+		if (len == 0) {
+			if (max_bitflips != NULL)
+				*max_bitflips = bitflips;
 			return ret;
+		}
 
 		buf += size;
 		from = 0;
@@ -259,7 +266,7 @@  concat_read_oob(struct mtd_info *mtd, loff_t from, struct mtd_oob_ops *ops)
 	struct mtd_oob_ops devops = *ops;
 	int i, err, ret = 0;
 
-	ops->retlen = ops->oobretlen = 0;
+	ops->retlen = ops->oobretlen = ops->max_bitflips = 0;
 
 	for (i = 0; i < concat->num_subdev; i++) {
 		struct mtd_info *subdev = concat->subdev[i];
@@ -276,6 +283,7 @@  concat_read_oob(struct mtd_info *mtd, loff_t from, struct mtd_oob_ops *ops)
 		err = subdev->read_oob(subdev, from, &devops);
 		ops->retlen += devops.retlen;
 		ops->oobretlen += devops.oobretlen;
+		ops->max_bitflips = max(ops->max_bitflips, devops.max_bitflips);
 
 		/* Save information about bitflips! */
 		if (unlikely(err)) {
diff --git a/drivers/mtd/mtdoops.c b/drivers/mtd/mtdoops.c
index 1e2fa62..a29cb34 100644
--- a/drivers/mtd/mtdoops.c
+++ b/drivers/mtd/mtdoops.c
@@ -253,10 +253,11 @@  static void find_next_position(struct mtdoops_context *cxt)
 	size_t retlen;
 
 	for (page = 0; page < cxt->oops_pages; page++) {
+
 		/* Assume the page is used */
 		mark_page_used(cxt, page);
 		ret = mtd->read(mtd, page * record_size, MTDOOPS_HEADER_SIZE,
-				&retlen, (u_char *) &count[0]);
+				&retlen, (u_char *) &count[0], NULL);
 		if (retlen != MTDOOPS_HEADER_SIZE ||
 				(ret < 0 && !mtd_is_bitflip(ret))) {
 			printk(KERN_ERR "mtdoops: read failure at %ld (%td of %d read), err %d\n",
diff --git a/drivers/mtd/mtdpart.c b/drivers/mtd/mtdpart.c
index a0bd2de..68eac92 100644
--- a/drivers/mtd/mtdpart.c
+++ b/drivers/mtd/mtdpart.c
@@ -58,7 +58,7 @@  struct mtd_part {
  */
 
 static int part_read(struct mtd_info *mtd, loff_t from, size_t len,
-		size_t *retlen, u_char *buf)
+		     size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct mtd_part *part = PART(mtd);
 	struct mtd_ecc_stats stats;
@@ -71,7 +71,7 @@  static int part_read(struct mtd_info *mtd, loff_t from, size_t len,
 	else if (from + len > mtd->size)
 		len = mtd->size - from;
 	res = part->master->read(part->master, from + part->offset,
-				   len, retlen, buf);
+				 len, retlen, buf, max_bitflips);
 	if (unlikely(res)) {
 		if (mtd_is_bitflip(res))
 			mtd->ecc_stats.corrected += part->master->ecc_stats.corrected - stats.corrected;
diff --git a/drivers/mtd/mtdswap.c b/drivers/mtd/mtdswap.c
index bd9590c..1249fe5 100644
--- a/drivers/mtd/mtdswap.c
+++ b/drivers/mtd/mtdswap.c
@@ -736,7 +736,7 @@  static int mtdswap_move_block(struct mtdswap_dev *d, unsigned int oldblock,
 	retries = 0;
 
 retry:
-	ret = mtd->read(mtd, readpos, PAGE_SIZE, &retlen, d->page_buf);
+	ret = mtd->read(mtd, readpos, PAGE_SIZE, &retlen, d->page_buf, NULL);
 
 	if (ret < 0 && !mtd_is_bitflip(ret)) {
 		oldeb = d->eb_data + oldblock / d->pages_per_eblk;
@@ -1161,7 +1161,7 @@  static int mtdswap_readsect(struct mtd_blktrans_dev *dev,
 	retries = 0;
 
 retry:
-	ret = mtd->read(mtd, readpos, PAGE_SIZE, &retlen, buf);
+	ret = mtd->read(mtd, readpos, PAGE_SIZE, &retlen, buf, NULL);
 
 	d->mtd_read_count++;
 	if (mtd_is_bitflip(ret)) {
diff --git a/drivers/mtd/nand/diskonchip.c b/drivers/mtd/nand/diskonchip.c
index 5780dba..5929e56 100644
--- a/drivers/mtd/nand/diskonchip.c
+++ b/drivers/mtd/nand/diskonchip.c
@@ -1072,7 +1072,7 @@  static int __init find_media_headers(struct mtd_info *mtd, u_char *buf, const ch
 	size_t retlen;
 
 	for (offs = 0; offs < mtd->size; offs += mtd->erasesize) {
-		ret = mtd->read(mtd, offs, mtd->writesize, &retlen, buf);
+		ret = mtd->read(mtd, offs, mtd->writesize, &retlen, buf, NULL);
 		if (retlen != mtd->writesize)
 			continue;
 		if (ret) {
@@ -1097,7 +1097,7 @@  static int __init find_media_headers(struct mtd_info *mtd, u_char *buf, const ch
 	/* Only one mediaheader was found.  We want buf to contain a
 	   mediaheader on return, so we'll have to re-read the one we found. */
 	offs = doc->mh0_page << this->page_shift;
-	ret = mtd->read(mtd, offs, mtd->writesize, &retlen, buf);
+	ret = mtd->read(mtd, offs, mtd->writesize, &retlen, buf, NULL);
 	if (retlen != mtd->writesize) {
 		/* Insanity.  Give up. */
 		printk(KERN_ERR "Read DiskOnChip Media Header once, but can't reread it???\n");
diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c
index 35b4565..b146a4d 100644
--- a/drivers/mtd/nand/nand_base.c
+++ b/drivers/mtd/nand/nand_base.c
@@ -1458,6 +1458,8 @@  static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 	oob = ops->oobbuf;
 
 	while (1) {
+		__u32 prior_corrected = mtd->ecc_stats.corrected;
+
 		bytes = min(mtd->writesize - col, readlen);
 		aligned = (bytes == mtd->writesize);
 
@@ -1530,8 +1532,11 @@  static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 			buf += bytes;
 		}
 
-		readlen -= bytes;
+		ops->max_bitflips =
+			max(ops->max_bitflips,
+			    mtd->ecc_stats.corrected - prior_corrected);
 
+		readlen -= bytes;
 		if (!readlen)
 			break;
 
@@ -1580,7 +1585,7 @@  static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
  * Get hold of the chip and call nand_do_read.
  */
 static int nand_read(struct mtd_info *mtd, loff_t from, size_t len,
-		     size_t *retlen, uint8_t *buf)
+		     size_t *retlen, uint8_t *buf, unsigned int *max_bitflips)
 {
 	struct nand_chip *chip = mtd->priv;
 	struct mtd_oob_ops ops;
@@ -1598,11 +1603,15 @@  static int nand_read(struct mtd_info *mtd, loff_t from, size_t len,
 	ops.datbuf = buf;
 	ops.oobbuf = NULL;
 	ops.mode = 0;
+	ops.max_bitflips = 0;
 
 	ret = nand_do_read_ops(mtd, from, &ops);
 
 	*retlen = ops.retlen;
 
+	if (max_bitflips != NULL)
+		*max_bitflips = ops.max_bitflips;
+
 	nand_release_device(mtd);
 
 	return ret;
@@ -1799,6 +1808,8 @@  static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
 	page = realpage & chip->pagemask;
 
 	while (1) {
+		__u32 prior_corrected = mtd->ecc_stats.corrected;
+
 		if (ops->mode == MTD_OPS_RAW)
 			sndcmd = chip->ecc.read_oob_raw(mtd, chip, page, sndcmd);
 		else
@@ -1820,6 +1831,10 @@  static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
 				nand_wait_ready(mtd);
 		}
 
+		ops->max_bitflips =
+			max(ops->max_bitflips,
+			    mtd->ecc_stats.corrected - prior_corrected);
+
 		readlen -= len;
 		if (!readlen)
 			break;
@@ -1865,7 +1880,7 @@  static int nand_read_oob(struct mtd_info *mtd, loff_t from,
 	struct nand_chip *chip = mtd->priv;
 	int ret = -ENOTSUPP;
 
-	ops->retlen = 0;
+	ops->retlen = ops->max_bitflips = 0;
 
 	/* Do not allow reads past end of device */
 	if (ops->datbuf && (from + ops->len) > mtd->size) {
diff --git a/drivers/mtd/nand/nand_bbt.c b/drivers/mtd/nand/nand_bbt.c
index 69148ae..6730690 100644
--- a/drivers/mtd/nand/nand_bbt.c
+++ b/drivers/mtd/nand/nand_bbt.c
@@ -201,7 +201,7 @@  static int read_bbt(struct mtd_info *mtd, uint8_t *buf, int page, int num,
 			from += marker_len;
 			marker_len = 0;
 		}
-		res = mtd->read(mtd, from, len, &retlen, buf);
+		res = mtd->read(mtd, from, len, &retlen, buf, NULL);
 		if (res < 0) {
 			if (mtd_is_eccerr(res)) {
 				pr_info("nand_bbt: ECC error in BBT at "
@@ -298,7 +298,7 @@  static int scan_read_raw_data(struct mtd_info *mtd, uint8_t *buf, loff_t offs,
 	if (td->options & NAND_BBT_VERSION)
 		len++;
 
-	return mtd->read(mtd, offs, len, &retlen, buf);
+	return mtd->read(mtd, offs, len, &retlen, buf, NULL);
 }
 
 /* Scan read raw data from flash */
@@ -753,10 +753,11 @@  static int write_bbt(struct mtd_info *mtd, uint8_t *buf,
 
 		/* Must we save the block contents? */
 		if (td->options & NAND_BBT_SAVECONTENT) {
+
 			/* Make it block aligned */
 			to &= ~((loff_t)((1 << this->bbt_erase_shift) - 1));
 			len = 1 << this->bbt_erase_shift;
-			res = mtd->read(mtd, to, len, &retlen, buf);
+			res = mtd->read(mtd, to, len, &retlen, buf, NULL);
 			if (res < 0) {
 				if (retlen != len) {
 					pr_info("nand_bbt: error reading block "
diff --git a/drivers/mtd/nftlcore.c b/drivers/mtd/nftlcore.c
index cda77b5..afc4292 100644
--- a/drivers/mtd/nftlcore.c
+++ b/drivers/mtd/nftlcore.c
@@ -424,11 +424,11 @@  static u16 NFTL_foldchain (struct NFTLrecord *nftl, unsigned thisVUC, unsigned p
 			continue;
 
 		ret = mtd->read(mtd, (nftl->EraseSize * BlockMap[block]) + (block * 512),
-				512, &retlen, movebuf);
+				512, &retlen, movebuf, NULL);
 		if (ret < 0 && !mtd_is_bitflip(ret)) {
 			ret = mtd->read(mtd, (nftl->EraseSize * BlockMap[block])
 					+ (block * 512), 512, &retlen,
-					movebuf);
+					movebuf, NULL);
 			if (ret != -EIO)
 				printk("Error went away on retry.\n");
 		}
@@ -771,7 +771,7 @@  static int nftl_readblock(struct mtd_blktrans_dev *mbd, unsigned long block,
 	} else {
 		loff_t ptr = (lastgoodEUN * nftl->EraseSize) + blockofs;
 		size_t retlen;
-		int res = mtd->read(mtd, ptr, 512, &retlen, buffer);
+		int res = mtd->read(mtd, ptr, 512, &retlen, buffer, NULL);
 
 		if (res < 0 && !mtd_is_bitflip(res))
 			return -EIO;
diff --git a/drivers/mtd/nftlmount.c b/drivers/mtd/nftlmount.c
index ac40925..b1e6663 100644
--- a/drivers/mtd/nftlmount.c
+++ b/drivers/mtd/nftlmount.c
@@ -64,7 +64,7 @@  static int find_boot_record(struct NFTLrecord *nftl)
 		/* Check for ANAND header first. Then can whinge if it's found but later
 		   checks fail */
 		ret = mtd->read(mtd, block * nftl->EraseSize, SECTORSIZE,
-				&retlen, buf);
+				&retlen, buf, NULL);
 		/* We ignore ret in case the ECC of the MediaHeader is invalid
 		   (which is apparently acceptable) */
 		if (retlen != SECTORSIZE) {
@@ -110,7 +110,7 @@  static int find_boot_record(struct NFTLrecord *nftl)
 
 		/* Finally reread to check ECC */
 		if ((ret = mtd->read(mtd, block * nftl->EraseSize, SECTORSIZE,
-				     &retlen, buf) < 0)) {
+				     &retlen, buf, NULL) < 0)) {
 			printk(KERN_NOTICE "ANAND header found at 0x%x in mtd%d, but ECC read failed (err %d)\n",
 			       block * nftl->EraseSize, nftl->mbd.mtd->index, ret);
 			continue;
@@ -274,7 +274,7 @@  static int check_free_sectors(struct NFTLrecord *nftl, unsigned int address, int
 	int i;
 
 	for (i = 0; i < len; i += SECTORSIZE) {
-		if (mtd->read(mtd, address, SECTORSIZE, &retlen, buf))
+		if (mtd->read(mtd, address, SECTORSIZE, &retlen, buf, NULL))
 			return -1;
 		if (memcmpb(buf, 0xff, SECTORSIZE) != 0)
 			return -1;
diff --git a/drivers/mtd/onenand/onenand_base.c b/drivers/mtd/onenand/onenand_base.c
index a839473..d2c5ff7 100644
--- a/drivers/mtd/onenand/onenand_base.c
+++ b/drivers/mtd/onenand/onenand_base.c
@@ -1451,7 +1451,7 @@  static int onenand_read_oob_nolock(struct mtd_info *mtd, loff_t from,
  * Read with ecc
 */
 static int onenand_read(struct mtd_info *mtd, loff_t from, size_t len,
-	size_t *retlen, u_char *buf)
+			size_t *retlen, u_char *buf, unsigned int *max_bitflips)
 {
 	struct onenand_chip *this = mtd->priv;
 	struct mtd_oob_ops ops = {
@@ -1469,6 +1469,8 @@  static int onenand_read(struct mtd_info *mtd, loff_t from, size_t len,
 	onenand_release_device(mtd);
 
 	*retlen = ops.retlen;
+	if (max_bitflips != NULL)
+		*max_bitflips = mtd_is_bitflip(ret) ? 1 : 0;
 	return ret;
 }
 
@@ -1505,6 +1507,8 @@  static int onenand_read_oob(struct mtd_info *mtd, loff_t from,
 		ret = onenand_read_oob_nolock(mtd, from, ops);
 	onenand_release_device(mtd);
 
+	ops->max_bitflips = mtd_is_bitflip(ret) ? 1 : 0;
+
 	return ret;
 }
 
diff --git a/drivers/mtd/redboot.c b/drivers/mtd/redboot.c
index e366b1d..63e3167 100644
--- a/drivers/mtd/redboot.c
+++ b/drivers/mtd/redboot.c
@@ -105,7 +105,7 @@  static int parse_redboot_partitions(struct mtd_info *master,
 	       master->name, offset);
 
 	ret = master->read(master, offset,
-			   master->erasesize, &retlen, (void *)buf);
+			   master->erasesize, &retlen, (void *)buf, NULL);
 
 	if (ret)
 		goto out;
diff --git a/drivers/mtd/rfd_ftl.c b/drivers/mtd/rfd_ftl.c
index 73ae217..ab797c7 100644
--- a/drivers/mtd/rfd_ftl.c
+++ b/drivers/mtd/rfd_ftl.c
@@ -202,7 +202,8 @@  static int scan_header(struct partition *part)
 	for (i=0, blocks_found=0; i<part->total_blocks; i++) {
 		rc = part->mbd.mtd->read(part->mbd.mtd,
 				i * part->block_size, part->header_size,
-				&retlen, (u_char*)part->header_cache);
+					 &retlen, (u_char *)part->header_cache,
+					 NULL);
 
 		if (!rc && retlen != part->header_size)
 			rc = -EIO;
@@ -251,7 +252,7 @@  static int rfd_ftl_readsect(struct mtd_blktrans_dev *dev, u_long sector, char *b
 	addr = part->sector_map[sector];
 	if (addr != -1) {
 		rc = part->mbd.mtd->read(part->mbd.mtd, addr, SECTOR_SIZE,
-						&retlen, (u_char*)buf);
+					 &retlen, (u_char *)buf, NULL);
 		if (!rc && retlen != SECTOR_SIZE)
 			rc = -EIO;
 
@@ -374,7 +375,7 @@  static int move_block_contents(struct partition *part, int block_no, u_long *old
 
 	rc = part->mbd.mtd->read(part->mbd.mtd,
 		part->blocks[block_no].offset, part->header_size,
-		&retlen, (u_char*)map);
+		&retlen, (u_char *)map, NULL);
 
 	if (!rc && retlen != part->header_size)
 		rc = -EIO;
@@ -414,7 +415,7 @@  static int move_block_contents(struct partition *part, int block_no, u_long *old
 			continue;
 		}
 		rc = part->mbd.mtd->read(part->mbd.mtd, addr,
-			SECTOR_SIZE, &retlen, sector_data);
+			SECTOR_SIZE, &retlen, sector_data, NULL);
 
 		if (!rc && retlen != SECTOR_SIZE)
 			rc = -EIO;
@@ -564,7 +565,8 @@  static int find_writable_block(struct partition *part, u_long *old_sector)
 	}
 
 	rc = part->mbd.mtd->read(part->mbd.mtd, part->blocks[block].offset,
-		part->header_size, &retlen, (u_char*)part->header_cache);
+				 part->header_size, &retlen,
+				 (u_char *)part->header_cache, NULL);
 
 	if (!rc && retlen != part->header_size)
 		rc = -EIO;
diff --git a/drivers/mtd/ssfdc.c b/drivers/mtd/ssfdc.c
index 976e3d2..1d0d8d9 100644
--- a/drivers/mtd/ssfdc.c
+++ b/drivers/mtd/ssfdc.c
@@ -124,7 +124,7 @@  static int get_valid_cis_sector(struct mtd_info *mtd)
 	for (k = 0, offset = 0; k < 4; k++, offset += mtd->erasesize) {
 		if (!mtd->block_isbad(mtd, offset)) {
 			ret = mtd->read(mtd, offset, SECTOR_SIZE, &retlen,
-				sect_buf);
+					sect_buf, NULL);
 
 			/* CIS pattern match on the sector buffer */
 			if (ret < 0 || retlen != SECTOR_SIZE) {
@@ -156,7 +156,7 @@  static int read_physical_sector(struct mtd_info *mtd, uint8_t *sect_buf,
 	size_t retlen;
 	loff_t offset = (loff_t)sect_no << SECTOR_SHIFT;
 
-	ret = mtd->read(mtd, offset, SECTOR_SIZE, &retlen, sect_buf);
+	ret = mtd->read(mtd, offset, SECTOR_SIZE, &retlen, sect_buf, NULL);
 	if (ret < 0 || retlen != SECTOR_SIZE)
 		return -1;
 
diff --git a/drivers/mtd/tests/mtd_pagetest.c b/drivers/mtd/tests/mtd_pagetest.c
index afafb69..3677ec7 100644
--- a/drivers/mtd/tests/mtd_pagetest.c
+++ b/drivers/mtd/tests/mtd_pagetest.c
@@ -127,7 +127,7 @@  static int verify_eraseblock(int ebnum)
 	set_random_data(writebuf, mtd->erasesize);
 	for (j = 0; j < pgcnt - 1; ++j, addr += pgsize) {
 		/* Do a read to set the internal dataRAMs to different data */
-		err = mtd->read(mtd, addr0, bufsize, &read, twopages);
+		err = mtd->read(mtd, addr0, bufsize, &read, twopages, NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -135,7 +135,8 @@  static int verify_eraseblock(int ebnum)
 			       (long long)addr0);
 			return err;
 		}
-		err = mtd->read(mtd, addrn - bufsize, bufsize, &read, twopages);
+		err = mtd->read(mtd, addrn - bufsize, bufsize, &read, twopages,
+				NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -145,7 +146,7 @@  static int verify_eraseblock(int ebnum)
 		}
 		memset(twopages, 0, bufsize);
 		read = 0;
-		err = mtd->read(mtd, addr, bufsize, &read, twopages);
+		err = mtd->read(mtd, addr, bufsize, &read, twopages, NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -163,7 +164,7 @@  static int verify_eraseblock(int ebnum)
 	if (addr <= addrn - pgsize - pgsize && !bbt[ebnum + 1]) {
 		unsigned long oldnext = next;
 		/* Do a read to set the internal dataRAMs to different data */
-		err = mtd->read(mtd, addr0, bufsize, &read, twopages);
+		err = mtd->read(mtd, addr0, bufsize, &read, twopages, NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -171,7 +172,8 @@  static int verify_eraseblock(int ebnum)
 			       (long long)addr0);
 			return err;
 		}
-		err = mtd->read(mtd, addrn - bufsize, bufsize, &read, twopages);
+		err = mtd->read(mtd, addrn - bufsize, bufsize, &read, twopages,
+				NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -181,7 +183,7 @@  static int verify_eraseblock(int ebnum)
 		}
 		memset(twopages, 0, bufsize);
 		read = 0;
-		err = mtd->read(mtd, addr, bufsize, &read, twopages);
+		err = mtd->read(mtd, addr, bufsize, &read, twopages, NULL);
 		if (mtd_is_bitflip(err))
 			err = 0;
 		if (err || read != bufsize) {
@@ -230,7 +232,7 @@  static int crosstest(void)
 	/* Read 2nd-to-last page to pp1 */
 	read = 0;
 	addr = addrn - pgsize - pgsize;
-	err = mtd->read(mtd, addr, pgsize, &read, pp1);
+	err = mtd->read(mtd, addr, pgsize, &read, pp1, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -243,7 +245,7 @@  static int crosstest(void)
 	/* Read 3rd-to-last page to pp1 */
 	read = 0;
 	addr = addrn - pgsize - pgsize - pgsize;
-	err = mtd->read(mtd, addr, pgsize, &read, pp1);
+	err = mtd->read(mtd, addr, pgsize, &read, pp1, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -257,7 +259,7 @@  static int crosstest(void)
 	read = 0;
 	addr = addr0;
 	printk(PRINT_PREF "reading page at %#llx\n", (long long)addr);
-	err = mtd->read(mtd, addr, pgsize, &read, pp2);
+	err = mtd->read(mtd, addr, pgsize, &read, pp2, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -271,7 +273,7 @@  static int crosstest(void)
 	read = 0;
 	addr = addrn - pgsize;
 	printk(PRINT_PREF "reading page at %#llx\n", (long long)addr);
-	err = mtd->read(mtd, addr, pgsize, &read, pp3);
+	err = mtd->read(mtd, addr, pgsize, &read, pp3, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -285,7 +287,7 @@  static int crosstest(void)
 	read = 0;
 	addr = addr0;
 	printk(PRINT_PREF "reading page at %#llx\n", (long long)addr);
-	err = mtd->read(mtd, addr, pgsize, &read, pp4);
+	err = mtd->read(mtd, addr, pgsize, &read, pp4, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -344,7 +346,7 @@  static int erasecrosstest(void)
 
 	printk(PRINT_PREF "reading 1st page of block %d\n", ebnum);
 	memset(readbuf, 0, pgsize);
-	err = mtd->read(mtd, addr0, pgsize, &read, readbuf);
+	err = mtd->read(mtd, addr0, pgsize, &read, readbuf, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -382,7 +384,7 @@  static int erasecrosstest(void)
 
 	printk(PRINT_PREF "reading 1st page of block %d\n", ebnum);
 	memset(readbuf, 0, pgsize);
-	err = mtd->read(mtd, addr0, pgsize, &read, readbuf);
+	err = mtd->read(mtd, addr0, pgsize, &read, readbuf, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
@@ -438,7 +440,7 @@  static int erasetest(void)
 		return err;
 
 	printk(PRINT_PREF "reading 1st page of block %d\n", ebnum);
-	err = mtd->read(mtd, addr0, pgsize, &read, twopages);
+	err = mtd->read(mtd, addr0, pgsize, &read, twopages, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (err || read != pgsize) {
diff --git a/drivers/mtd/tests/mtd_readtest.c b/drivers/mtd/tests/mtd_readtest.c
index 550fe51..e11fdfd 100644
--- a/drivers/mtd/tests/mtd_readtest.c
+++ b/drivers/mtd/tests/mtd_readtest.c
@@ -52,7 +52,7 @@  static int read_eraseblock_by_page(int ebnum)
 
 	for (i = 0; i < pgcnt; i++) {
 		memset(buf, 0 , pgcnt);
-		ret = mtd->read(mtd, addr, pgsize, &read, buf);
+		ret = mtd->read(mtd, addr, pgsize, &read, buf, NULL);
 		if (ret == -EUCLEAN)
 			ret = 0;
 		if (ret || read != pgsize) {
diff --git a/drivers/mtd/tests/mtd_speedtest.c b/drivers/mtd/tests/mtd_speedtest.c
index 493b367..951c80a 100644
--- a/drivers/mtd/tests/mtd_speedtest.c
+++ b/drivers/mtd/tests/mtd_speedtest.c
@@ -214,7 +214,7 @@  static int read_eraseblock(int ebnum)
 	int err = 0;
 	loff_t addr = ebnum * mtd->erasesize;
 
-	err = mtd->read(mtd, addr, mtd->erasesize, &read, iobuf);
+	err = mtd->read(mtd, addr, mtd->erasesize, &read, iobuf, NULL);
 	/* Ignore corrected ECC errors */
 	if (mtd_is_bitflip(err))
 		err = 0;
@@ -235,7 +235,7 @@  static int read_eraseblock_by_page(int ebnum)
 	void *buf = iobuf;
 
 	for (i = 0; i < pgcnt; i++) {
-		err = mtd->read(mtd, addr, pgsize, &read, buf);
+		err = mtd->read(mtd, addr, pgsize, &read, buf, NULL);
 		/* Ignore corrected ECC errors */
 		if (mtd_is_bitflip(err))
 			err = 0;
@@ -261,7 +261,7 @@  static int read_eraseblock_by_2pages(int ebnum)
 	void *buf = iobuf;
 
 	for (i = 0; i < n; i++) {
-		err = mtd->read(mtd, addr, sz, &read, buf);
+		err = mtd->read(mtd, addr, sz, &read, buf, NULL);
 		/* Ignore corrected ECC errors */
 		if (mtd_is_bitflip(err))
 			err = 0;
@@ -276,7 +276,7 @@  static int read_eraseblock_by_2pages(int ebnum)
 		buf += sz;
 	}
 	if (pgcnt % 2) {
-		err = mtd->read(mtd, addr, pgsize, &read, buf);
+		err = mtd->read(mtd, addr, pgsize, &read, buf, NULL);
 		/* Ignore corrected ECC errors */
 		if (mtd_is_bitflip(err))
 			err = 0;
diff --git a/drivers/mtd/tests/mtd_stresstest.c b/drivers/mtd/tests/mtd_stresstest.c
index 52ffd91..9dadf4a 100644
--- a/drivers/mtd/tests/mtd_stresstest.c
+++ b/drivers/mtd/tests/mtd_stresstest.c
@@ -153,7 +153,7 @@  static int do_read(void)
 			len = mtd->erasesize - offs;
 	}
 	addr = eb * mtd->erasesize + offs;
-	err = mtd->read(mtd, addr, len, &read, readbuf);
+	err = mtd->read(mtd, addr, len, &read, readbuf, NULL);
 	if (mtd_is_bitflip(err))
 		err = 0;
 	if (unlikely(err || read != len)) {
diff --git a/drivers/mtd/tests/mtd_subpagetest.c b/drivers/mtd/tests/mtd_subpagetest.c
index 1a05bfa..cc782d4 100644
--- a/drivers/mtd/tests/mtd_subpagetest.c
+++ b/drivers/mtd/tests/mtd_subpagetest.c
@@ -196,7 +196,7 @@  static int verify_eraseblock(int ebnum)
 	set_random_data(writebuf, subpgsize);
 	clear_data(readbuf, subpgsize);
 	read = 0;
-	err = mtd->read(mtd, addr, subpgsize, &read, readbuf);
+	err = mtd->read(mtd, addr, subpgsize, &read, readbuf, NULL);
 	if (unlikely(err || read != subpgsize)) {
 		if (mtd_is_bitflip(err) && read == subpgsize) {
 			printk(PRINT_PREF "ECC correction at %#llx\n",
@@ -224,7 +224,7 @@  static int verify_eraseblock(int ebnum)
 	set_random_data(writebuf, subpgsize);
 	clear_data(readbuf, subpgsize);
 	read = 0;
-	err = mtd->read(mtd, addr, subpgsize, &read, readbuf);
+	err = mtd->read(mtd, addr, subpgsize, &read, readbuf, NULL);
 	if (unlikely(err || read != subpgsize)) {
 		if (mtd_is_bitflip(err) && read == subpgsize) {
 			printk(PRINT_PREF "ECC correction at %#llx\n",
@@ -262,7 +262,8 @@  static int verify_eraseblock2(int ebnum)
 		set_random_data(writebuf, subpgsize * k);
 		clear_data(readbuf, subpgsize * k);
 		read = 0;
-		err = mtd->read(mtd, addr, subpgsize * k, &read, readbuf);
+		err = mtd->read(mtd, addr, subpgsize * k, &read, readbuf,
+				NULL);
 		if (unlikely(err || read != subpgsize * k)) {
 			if (mtd_is_bitflip(err) && read == subpgsize * k) {
 				printk(PRINT_PREF "ECC correction at %#llx\n",
@@ -296,7 +297,7 @@  static int verify_eraseblock_ff(int ebnum)
 	for (j = 0; j < mtd->erasesize / subpgsize; ++j) {
 		clear_data(readbuf, subpgsize);
 		read = 0;
-		err = mtd->read(mtd, addr, subpgsize, &read, readbuf);
+		err = mtd->read(mtd, addr, subpgsize, &read, readbuf, NULL);
 		if (unlikely(err || read != subpgsize)) {
 			if (mtd_is_bitflip(err) && read == subpgsize) {
 				printk(PRINT_PREF "ECC correction at %#llx\n",
diff --git a/drivers/mtd/tests/mtd_torturetest.c b/drivers/mtd/tests/mtd_torturetest.c
index 03ab649..36e0ae9 100644
--- a/drivers/mtd/tests/mtd_torturetest.c
+++ b/drivers/mtd/tests/mtd_torturetest.c
@@ -137,7 +137,7 @@  static inline int check_eraseblock(int ebnum, unsigned char *buf)
 	}
 
 retry:
-	err = mtd->read(mtd, addr, len, &read, check_buf);
+	err = mtd->read(mtd, addr, len, &read, check_buf, NULL);
 	if (mtd_is_bitflip(err))
 		printk(PRINT_PREF "single bit flip occurred at EB %d "
 		       "MTD reported that it was fixed.\n", ebnum);
diff --git a/drivers/mtd/ubi/io.c b/drivers/mtd/ubi/io.c
index f20b6f2..13f6b18 100644
--- a/drivers/mtd/ubi/io.c
+++ b/drivers/mtd/ubi/io.c
@@ -170,7 +170,7 @@  int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
 
 	addr = (loff_t)pnum * ubi->peb_size + offset;
 retry:
-	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf);
+	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf, NULL);
 	if (err) {
 		const char *errstr = mtd_is_eccerr(err) ? " (ECC error)" : "";
 
@@ -1357,7 +1357,7 @@  int ubi_dbg_check_write(struct ubi_device *ubi, const void *buf, int pnum,
 		return 0;
 	}
 
-	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf1);
+	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf1, NULL);
 	if (err && !mtd_is_bitflip(err))
 		goto out_free;
 
@@ -1421,7 +1421,7 @@  int ubi_dbg_check_all_ff(struct ubi_device *ubi, int pnum, int offset, int len)
 		return 0;
 	}
 
-	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf);
+	err = ubi->mtd->read(ubi->mtd, addr, len, &read, buf, NULL);
 	if (err && !mtd_is_bitflip(err)) {
 		ubi_err("error %d while reading %d bytes from PEB %d:%d, "
 			"read %zd bytes", err, len, pnum, offset, read);
diff --git a/include/linux/mtd/mtd.h b/include/linux/mtd/mtd.h
index 9f5b312..a1be9a2 100644
--- a/include/linux/mtd/mtd.h
+++ b/include/linux/mtd/mtd.h
@@ -81,6 +81,9 @@  struct mtd_erase_region_info {
  *		mode = MTD_OPS_PLACE_OOB or MTD_OPS_RAW)
  * @datbuf:	data buffer - if NULL only oob data are read/written
  * @oobbuf:	oob data buffer
+ * @max_bitflips: for read operations, maximum number of bit errors corrected
+ *                on any one minimum i/o unit (e.g., nand page)
+ *                (value returned to caller by the driver)
  *
  * Note, it is allowed to read more than one OOB area at one go, but not write.
  * The interface assumes that the OOB write requests program only one page's
@@ -95,6 +98,7 @@  struct mtd_oob_ops {
 	uint32_t	ooboffs;
 	uint8_t		*datbuf;
 	uint8_t		*oobbuf;
+	unsigned int	max_bitflips;
 };
 
 #define MTD_MAX_OOBFREE_ENTRIES_LARGE	32
@@ -201,8 +205,13 @@  struct mtd_info {
 	 */
 	struct backing_dev_info *backing_dev_info;
 
+	/*
+	 * If not NULL, max_bitflips returns to caller the maximum number of bit
+	 * errors corrected on any one minimum i/o unit (e.g., nand page).
+	 */
+	int (*read) (struct mtd_info *mtd, loff_t from, size_t len,
+		     size_t *retlen, u_char *buf, unsigned int *max_bitflips);
 
-	int (*read) (struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen, u_char *buf);
 	int (*write) (struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen, const u_char *buf);
 
 	/* In blackbox flight recorder like scenarios we want to make successful
diff --git a/include/linux/mtd/pmc551.h b/include/linux/mtd/pmc551.h
index 27ad40a..55a3e06 100644
--- a/include/linux/mtd/pmc551.h
+++ b/include/linux/mtd/pmc551.h
@@ -37,7 +37,8 @@  static int pmc551_erase(struct mtd_info *, struct erase_info *);
 static void pmc551_unpoint(struct mtd_info *, loff_t, size_t);
 static int pmc551_point(struct mtd_info *mtd, loff_t from, size_t len,
 		size_t *retlen, void **virt, resource_size_t *phys);
-static int pmc551_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *);
+static int pmc551_read(struct mtd_info *, loff_t, size_t, size_t *, u_char *,
+		       unsigned int *);
 static int pmc551_write(struct mtd_info *, loff_t, size_t, size_t *, const u_char *);