Strange faults using Intel P33 flash

Strange faults using Intel P33 flash

[jkimble]

Yes it's an old 2.6.10 kernel but it's CFI NOR flash. Shouldn't it work
with mtd circa 2.6.10?

The flash memory I am using is the Intel P33 (256P33B) NOR flash. I
have two chips with the chip selects tied to the same logic line. The
chips are on the Flexbus of a Coldfire processor (MCF5484). I'm able
to work with the flash fine in the boot loader (Colilo) after porting
the driver template from Intel into it. My question is how do I
configure MTD to work with this. I've set the beginning address and
the size. This is CFI complient NOR flash. I've set a chip interleave of 2
with a 32 bit bus. As the kernel is booting I keep getting the
message:

================================================
$Id: ftl.c,v 1.54 2004/11/16 18:33:15 dwmw2 Exp $
Unlock all sector for Intel on Fire Engine...
physmap flash device: 4000000 at f0000000
phys_mapped_flash: Found 2 x16 devices at 0x0 in 32-bit bank
 Intel/Sharp Extended Query Table at 0x010A
  Unknown Intel/Sharp Extended Query version 1.4.
gen_probe: No supported Vendor Command Set found
cmdlinepart partition parsing not available
RedBoot partition parsing not available
Using physmap partition definition
Creating 2 MTD partitions on "phys_mapped_flash":
0x00000000-0x00040000 : "Colilo (256K)"
mtd: Giving out device 0 to Colilo (256K)
ftl_cs: FTL header not found.
0x00040000-0x00200000 : "user (other Flash memory) 2Mb-256K"
mtd: Giving out device 1 to user (other Flash memory) 2Mb-256K
ftl_cs: FTL header not found.
mice: PS/2 mouse device common for all mice
                 :
                 :NET: Registered protocol family 15
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000000:
0x1f8b in
stead
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000004:
0x40c7 in
stead
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000008:
0x0003 in
stead
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x0000000c:
0x6c69 in
stead
========================================================================

Then of course the kernel panics since this is where the root file system
lives. The  "gen_probe: No supported Vendor Command Set found" message is
obviously critical but with this being a CFI device how can that be? Is it
because my chip selects are tied together? Is there a way to configure
this to correctly handle this like one device (CS's tied together) but
with the commands concatenated into one 32 bit message?

Kinda stuck here. Any help/suggestions much appreciated!!

Re: Strange faults using Intel P33 flash

[Chuck Meade]

jkimble@one.net wrote:
> 
> Yes it's an old 2.6.10 kernel but it's CFI NOR flash. Shouldn't it work
> with mtd circa 2.6.10?
> 
> The flash memory I am using is the Intel P33 (256P33B) NOR flash. I
> have two chips with the chip selects tied to the same logic line. The
> chips are on the Flexbus of a Coldfire processor (MCF5484). I'm able
> to work with the flash fine in the boot loader (Colilo) after porting
> the driver template from Intel into it. My question is how do I
> configure MTD to work with this. I've set the beginning address and
> the size. This is CFI complient NOR flash. I've set a chip interleave of 2
> with a 32 bit bus. As the kernel is booting I keep getting the
> message:
> 
> ================================================
> $Id: ftl.c,v 1.54 2004/11/16 18:33:15 dwmw2 Exp $
> Unlock all sector for Intel on Fire Engine...
> physmap flash device: 4000000 at f0000000
> phys_mapped_flash: Found 2 x16 devices at 0x0 in 32-bit bank
>  Intel/Sharp Extended Query Table at 0x010A
>   Unknown Intel/Sharp Extended Query version 1.4.

I ran into something similar on a 2.6.11 variant.  The MTD of that era does
not support the version of CFI Extended Query found in the P33 flash.
What you can do is either a) use a newer kernel version which does support
this Extended Query version, or b) backport the bits of MTD logic that support
this Extended Query version from a newer kernel to your kernel.

If it is possible (for numerous reasons) it is preferable to use a newer
kernel version that has built-in support for the Extended Query version in
the P33 flash.

HTH,
Chuck