Handle eraseblocks containing the bad-block table in UBI

Handle eraseblocks containing the bad-block table in UBI

[Bogdan Harjoc]

With some Macronix NANDs that present the bad-blocks table at a fixed
eraseblock, UBI tries to use the PEB as data and this results in the
"Bad node, ff ff ff ..." symptoms.

Should UBI receive this special status of the bbt PEBs via the
mtd_isbad() result values (in the Macronix case, mtd_block_isbad()
returns false for the PEB containing the bbt), or are 'bbt PEBs' not
supported in UBI ?

Thanks,
Bogdan Harjoc

Re: Handle eraseblocks containing the bad-block table in UBI

[Richard Weinberger]

Bogdan,

On Fri, Jan 12, 2018 at 4:48 PM, Bogdan Harjoc <harjoc@gmail.com> wrote:
> With some Macronix NANDs that present the bad-blocks table at a fixed
> eraseblock, UBI tries to use the PEB as data and this results in the
> "Bad node, ff ff ff ..." symptoms.
>
> Should UBI receive this special status of the bbt PEBs via the
> mtd_isbad() result values (in the Macronix case, mtd_block_isbad()
> returns false for the PEB containing the bbt), or are 'bbt PEBs' not
> supported in UBI ?

Usually blocks containing the BBT are also marked as bad to avoid the
problem you're facing.
So, yes UBI should not see these PEBs because they should look like bad blocks.

Is this a kernel with vendor hackery?

-- 
Thanks,
//richard

Re: Handle eraseblocks containing the bad-block table in UBI

[Bogdan Harjoc]

On Fri, Jan 12, 2018 at 10:52 PM, Richard Weinberger
<richard.weinberger@gmail.com> wrote:
> Usually blocks containing the BBT are also marked as bad to avoid the
> problem you're facing.
> So, yes UBI should not see these PEBs because they should look like bad blocks.
>
> Is this a kernel with vendor hackery?

It is, at least the nand probing code adds its own nand_bbt_descr
structs. Since the issue goes away by adding an

if (ec_hdr->magic == 'Bbt0' or '1bbT')
   treat_it_as_bad_block()

in ubi_io_read_ec_hdr(), I will see where the bad-block status for the
bbt PEBs gets lost.

Thank you,
Bogdan

Re: Handle eraseblocks containing the bad-block table in UBI

[Richard Weinberger]

Am Samstag, 13. Januar 2018, 13:44:54 CET schrieb Bogdan Harjoc:
> On Fri, Jan 12, 2018 at 10:52 PM, Richard Weinberger
> 
> <richard.weinberger@gmail.com> wrote:
> > Usually blocks containing the BBT are also marked as bad to avoid the
> > problem you're facing.
> > So, yes UBI should not see these PEBs because they should look like bad
> > blocks.
> > 
> > Is this a kernel with vendor hackery?
> 
> It is, at least the nand probing code adds its own nand_bbt_descr
> structs. Since the issue goes away by adding an
> 
> if (ec_hdr->magic == 'Bbt0' or '1bbT')
>    treat_it_as_bad_block()
> 
> in ubi_io_read_ec_hdr(), I will see where the bad-block status for the
> bbt PEBs gets lost.

UBI is the wrong layer to address this.
Please make sure that these blocks are handled in the NAND core.
BBT_BLOCK_RESERVED is the flag you need for them.

Thanks,
//richard

-- 
sigma star gmbh - Eduard-Bodem-Gasse 6 - 6020 Innsbruck - Austria
ATU66964118 - FN 374287y