DiskOnChip 2000 128Mb problem

[David Woodhouse]

On Wed, 2003-05-07 at 21:37, Matthew Dharm wrote:
> I'm porting Linux to our embedded PPC platform.  We have a DoC 2000
> (apparently known as "Millenium") on the board, 

The DiskOnChip 2000 and DiskOnChip Millennium are separate beasts.
Remember the new Millennium didn't start till 2001 :)

You had a DiskOnChip 2000 on the Ocelots, and it's been working fine --
looks like you really do have a Millennium on the new board though.

Incidentally, why are you using these expensive parts and not just
putting raw flash on there?

> and I'm having problems getting the driver working.  I'm hoping
> someone here can help... I think the issue is that the driver doesn't
> support this part, tho it should (are we really the first to try
> this?).
> 
> We're using kernel 2.4.17, upgraded with MTD from CVS as of today.
> Our DoC is 128MB in size and operates at 3.3 V.
> 
> Here's our story:  The main DoC driver detects the device, but is
> unable to identify any flash devices.  At boot time, it shows:
> 
> Using configured DiskOnChip probe address 0x70400000
> DiskOnChip Millennium found at address 0x70400000
> No flash chips recognised.

Odd. There shouldn't really be much difference in behaviour between the
2000 and the 2001 drivers. They diverged for a while, then we merged
them. Some tracing and debugging ought to show what the merged driver is
doing differently (and, I assume, wrong).

 <...>

> Which is interesting, because it is exactly half the correct size.  A
> quick check through the source code shows that MAX_CHIPS_MIL is set to
> 1, and there is a FIXME in doc2001.c "to deal with multi-flash on
> multi-Millenium case more carefully".  If I change the definition of
> MAX_CHIPS_MIL to 4, we get:
> 
> Using configured DiskOnChip probe address 0x70400000
> DiskOnChip Millennium found at address 0x70400000
> Flash chip found: Manufacturer ID: EC, Chip ID: 76 (Samsung:NAND 64MB 3,3V)
> Flash chip found: Manufacturer ID: EC, Chip ID: 76 (Samsung:NAND 64MB 3,3V)
> 2 flash chips found. Total DiskOnChip size: 128 MiB
> mtd: Giving out device 0 to DiskOnChip Millennium

That looks basically correct now.

> NFTL driver: nftlcore.c $Revision: 1.88 $, nftlmount.c $Revision: 1.31$
> NFTL_notify_add for DiskOnChip Millennium
> mtd->read = c00cd450, size = 134217728, erasesize = 8192
> NFTL_setup
> Could not find valid boot record
> Could not mount NFTL device

At this point, it failed to find the NFTL format on the DiskOnChip. If
the kernel code fails, then 'nftldump' will too. Try 'nanddump' which
should just dump the contents of the raw flash. Look through it for a
block starting 'ANAND' and/or send me a copy.

We used to have endianness bugs in the NFTL code, but they were fixed
when I was playing with the Ocelot. I don't think there have been
changes since then, so it should still be fine.

-- 
dwmw2