Detection of pre-CFI Sharp NOR chips

[Andrea Adami]

On Wed, Apr 23, 2014 at 10:42 AM, Andrea Adami <andrea.adami at gmail.com> wrote:
> Hello,
>
> I'm trying to debug the detection of some hostic chips that exactly
> ten years ago could be detected by cfi_probe...
>
> http://lists.infradead.org/pipermail/linux-mtd/2004-April/009647.html
>
> My findings are contradicting that possibility...I am testing on a
> real consumer device, not a prototype,  and the setup is 2x16 on 32
> bit bus (sa1100).
>
> The readings are as follow:
>
> Status Register is:80808080
> JEDEC map_read IDentify 1:b000b0
> JEDEC map_read IDentify 2:b000b0
> CFI map_read at offset 0: b000b0
> CFI map_read at offset 1: b000b0
> CFI map_read at offset 2: 10001
> CFI map_read at offset 3: 0
> CFI map_read at offset 4: 0
> CFI map_read at offset 5: ffffffff
> CFI map_read at offset 6: 4000400
> CFI map_read at offset 7: 0
> CFI map_read at offset 8: b000b0
> CFI map_read at offset 9: b000b0
> CFI map_read at offset a: 0
> CFI map_read at offset b: 0
> CFI map_read at offset c: 0
> CFI map_read at offset d: ffffffff
> CFI map_read at offset e: 4000400
> CFI map_read at offset f: 0
> CFI map_read at offset 10: ffff0051
> CFI map_read at offset 11: ffff0052
> CFI map_read at offset 12: ffff0059
> CFI map_read at offset 13: ffff0003
> CFI map_read at offset 14: ffff0000
> CFI map_read at offset 15: ffff0039
> CFI map_read at offset 16: ffff0000
> ...
>
> map_read() does correctly return the values of both chips for the
> first offsets then the answer of one chip is 0xffff as if that part of
> the table were programmed on a single chip of the pair.
>
> So I can read the cfi->mfr and cfi->id in both jedec and cfi mode.
>
> CFI map_read at offset 0: b000b0
> CFI map_read at offset 1: b000b0
>
> For those chips then I have to discard the readings of one chip to
> match the values in cfi_query_present().
>
> So, before sending to the ML my set of patches for the support of this
> chip family  [1] I'd like to refine the CFI detection [2].
>
> How would you best add the fixup for all the four ID's in case of
> interleave? Are more checks in cfi_qry_present() acceptable?
>
> Thanks for your suggestions
>
> Andrea
>
>
> [1] http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto-3.14/collie/collie-LH28F640BF.patch
> [2] http://cgit.openembedded.org/meta-handheld/tree/recipes-kernel/linux/linux-yocto-3.14/collie/collie-match-cfi-qry.patch

Hi all,

mystery is solved: one flash chip is defective on my device!
The "CFI ROM" is missing on one chip, maybe it was badly manufactured
or perhaps is rotting after 10 years.

I have now received bootlogs from two owners of the same device and
they both get the expected sequence:

JEDEC map_read offs 0x0: b000b0
JEDEC map_read offs 0x1: b000b0
CFI map_read at 4*offset 0: b000b0
CFI map_read at 4*offset 1: b000b0
CFI map_read at 4*offset 2: 10001
CFI map_read at 4*offset 3: 0
CFI map_read at 4*offset 4: 0
CFI map_read at 4*offset 5: ffffffff
CFI map_read at 4*offset 6: 4000400
CFI map_read at 4*offset 7: 0
CFI map_read at 4*offset 8: b000b0
CFI map_read at 4*offset 9: b000b0
CFI map_read at 4*offset a: 0
CFI map_read at 4*offset b: 0
CFI map_read at 4*offset c: 0
CFI map_read at 4*offset d: ffffffff
CFI map_read at 4*offset e: 4000400
CFI map_read at 4*offset f: 0
CFI map_read at 4*offset 10: 510051
CFI map_read at 4*offset 11: 520052
CFI map_read at 4*offset 12: 590059
CFI map_read at 4*offset 13: 30003
CFI map_read at 4*offset 14: 0
CFI map_read at 4*offset 15: 390039
....

So no need for detection hacks, I'll send only the patch for
LH28F640BF family support.

Thanks and sorry for the noise....

Regards

Andrea