UBI - disable subpage verify and use CONFIG_MTD_NAND_VERIFY_WRITE possible?

UBI - disable subpage verify and use CONFIG_MTD_NAND_VERIFY_WRITE possible?

[Darwin Rambo]

Hello Artem,

Can you please clarify something for me?

http://www.linux-mtd.infradead.org/faq/ubi.html#L_subpage_verify_fail 

I read this note and still, CONFIG_MTD_NAND_VERIFY_WRITE is very useful for me. The question is, if I don't mind wasting a bit of flash, and configure UBI to not do subpage writes and put the EC and VID headers on separate pages (for example ubi.mtd=6,4096), does UBI then always do page writes and would CONFIG_MTD_NAND_VERIFY_WRITE then work reliably for me? Or are there other places in the code that might do subpage writes? So far it seems to work well for me with the verify on and no subpages for the headers. 

By the way, the standard nand_verify_buf() function in nand_base.c uses readb(chip->IO_ADDR_R) and doesn't use something like chip->ecc.read_page() to correct the data so this sounds like a bug to me. But I override this to do corrected reads anyways, which works better for SLC/MLC flash which can show one or more bit errors in a page read.

I apologize in advance if this is already documented somewhere.

Thanks.

Best regards,
Darwin


Disclaimer - Any views or opinions presented in this e-mail are solely those of the author 
and do not necessarily represent those of the company. 

Re: UBI - disable subpage verify and use CONFIG_MTD_NAND_VERIFY_WRITE possible?

[Artem Bityutskiy]

Hi,

On Mon, 2010-04-19 at 17:52 -0700, Darwin Rambo wrote:
> Hello Artem,
> 
> Can you please clarify something for me?
> 
> http://www.linux-mtd.infradead.org/faq/ubi.html#L_subpage_verify_fail 
> 
> I read this note and still, CONFIG_MTD_NAND_VERIFY_WRITE is very
> useful for me.

Ok.

>  The question is, if I don't mind wasting a bit of flash, and
> configure UBI to not do subpage writes and put the EC and VID headers
> on separate pages (for example ubi.mtd=6,4096), does UBI then always
> do page writes and would CONFIG_MTD_NAND_VERIFY_WRITE then work
> reliably for me?

Yes, in this case UBI will never use sub-pages, all writes will be 4KiB
aligned, and the feature should work.

>  Or are there other places in the code that might do subpage writes?

No.

>  So far it seems to work well for me with the verify on and no
> subpages for the headers. 

Right.

> By the way, the standard nand_verify_buf() function in nand_base.c
> uses readb(chip->IO_ADDR_R) and doesn't use something like
> chip->ecc.read_page() to correct the data so this sounds like a bug to
> me.

Yup. I think this feature was not used by anyone for long time.

>  But I override this to do corrected reads anyways, which works better
> for SLC/MLC flash which can show one or more bit errors in a page
> read.

Just send a patch then :-)

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)