DOC2000 issues...

DOC2000 issues...

[Kevin Fenzi]

Hi folks. 

Some background:

we have a device that has a doc2000 (48MB version) on it. 
Currently we have kernel 2.2.14 with a very old m-systems driver
running on it. It boots fine from the doc2000 (thats the only bootable
device on there):

debug: entering NFDC 2148 identification routine.
Debug: identified NFDC 2148.
Debug: starting NFTL mount.
Debug: finished NFTL mount.
Trueffs: Mount status 0h
Flash disk driver for DiskOnChip2000
Copyright (C) 1998 IGEL GmbH
Copyright (C) 1998 M-Systems Flash Disk Pioneers Ltd.
Partition check:
 fla: fla1
VFS: Mounted root (ext2 filesystem) readonly.
Freeing unused kernel memory: 40k freed

It's BIOS reports:

TrueFFS-BIOS -- Version 3.3.7 for DiskOnChip 2000 (V1.23)
Copyright (C) M-Systems, 1992-2000

Ok, so I attempted to get a 2.2.18 + mtdpatch working. 
I ran into problems getting the mtd stuff to compile into the kernel. 
I built the mtdpatch stuff as modules, and tried using a initrd. No
boot. (just hung from the bios)

So, I went to a 2.4.0 kernel, and compiled all the mtd stuff in
(compiled just fine):

...
M-Systems DiskOnChip driver. (C) 1999 Machine Vision Holdings, Inc.
M-Systems NAND Flash Translation Layer driver. (C) 1999 MVHI       
$Id: nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $          
...
NFTL_open                                          
ENODEV: thisNFTL = 0, minor = 23809, ip = c1173cf0, fp = c111bef8
VFS: Cannot open root device "5d01" or 5d:01                     
Please append a correct "root=" boot option 
Kernel panic: VFS: Unable to mount root fs on 5d:01

Then, I tried the newest m-systems driver with 2.2.18:

...
Copyright (C) 1998,2000 M-Systems Flash Disk Pioneers Ltd.
Copyright (C) 2000 Lineo                                  
DOC device(s) found: 4
...
fl_geninit: registered device at major: 100                                     
partition: 0: start_sect: 0, nr_sects: 177a0 Fl_blk_size[]: bbd0kb
partition: 1: start_sect: 0, nr_sects: 0 Fl_blk_size[]: 0kb       
partition: 64: start_sect: 0, nr_sects: 177a0 Fl_blk_size[]: bbd0kb
partition: 65: start_sect: 0, nr_sects: 0 Fl_blk_size[]: 0kb       
partition: 128: start_sect: 0, nr_sects: 177a0 Fl_blk_size[]: bbd0kb
partition: 129: start_sect: 0, nr_sects: 0 Fl_blk_size[]: 0kb       
partition: 192: start_sect: 0, nr_sects: 177a0 Fl_blk_size[]: bbd0kb
partition: 193: start_sect: 0, nr_sects: 0 Fl_blk_size[]: 0kb       
Partition check:                                             
 fla: fla1      
 flb: flb1
 flc: flc1
 fld: fld1
VFS: Cannot open root device 3e:01
Kernel panic: VFS: Unable to mount root fs on 3e:01

So, I know I am doing something wrong...can anyone point me in the
right direction? what is the best kernel to use for this? 

I would be happy to write up a HOWTO or the like if someone can help
me get it going. ;) 

kevin


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> 
> ...
> M-Systems DiskOnChip driver. (C) 1999 Machine Vision Holdings, Inc.
> M-Systems NAND Flash Translation Layer driver. (C) 1999 MVHI       
> $Id: nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $          
> ...


Didn't find hardware. Did you hardcode the address or leave it at zero to 
auto-probe? Common mistake is to give the segment address not the real 
address (0xd800 not 0xd8000). You do need the extra zero.

Note to self: Hide that option and the 55AA one inside if(LINUXBIOS).

> Then, I tried the newest m-systems driver with 2.2.18:

> VFS: Cannot open root device 3e:01
> Kernel panic: VFS: Unable to mount root fs on 3e:01

They changed the major number from the temporary one they were using. But
rather than using the one I'd already registered, they started using an
entirely new one. Major 100 I think.

Geometry info is useful though - can you compare it with what the free 
driver reports, when you get that working?

> I would be happy to write up a HOWTO or the like if someone can help
> me get it going. ;) 

I love you.

-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> On Fri, 2 Feb 2001, Kevin Fenzi wrote:
>>  ...  M-Systems DiskOnChip driver. (C) 1999 Machine Vision
>> Holdings, Inc.  M-Systems NAND Flash Translation Layer driver. (C)
>> 1999 MVHI $Id: nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $ ...

David> Didn't find hardware. Did you hardcode the address or leave it
David> at zero to auto-probe? Common mistake is to give the segment
David> address not the real address (0xd800 not 0xd8000). You do need
David> the extra zero.

I left it at 0. How can I tell what address the DOC is at? from the
bios? 

So it would print something here if it found the DOC? 

David> Note to self: Hide that option and the 55AA one inside
David> if(LINUXBIOS).

yeah, also help for those kernel options would be nice. ;) 
I had to look at the code to see that 0 meant to auto-probe. ;) 

>> Then, I tried the newest m-systems driver with 2.2.18:

>> VFS: Cannot open root device 3e:01 Kernel panic: VFS: Unable to
>> mount root fs on 3e:01

David> They changed the major number from the temporary one they were
David> using. But rather than using the one I'd already registered,
David> they started using an entirely new one. Major 100 I think.

yeah, I tried that too...no dice. That just should need to be the
'root=' option for the kernel in lilo.conf, right? 

David> Geometry info is useful though - can you compare it with what
David> the free driver reports, when you get that working?

sure. 

>> I would be happy to write up a HOWTO or the like if someone can
>> help me get it going. ;)

David> I love you.

ha. I have to get it working first tho. ;) 

David> -- dwmw2

kevin




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> I left it at 0. How can I tell what address the DOC is at? from the
> bios? 
> 
> So it would print something here if it found the DOC? 

Yep, should do. Make it print stuff anyway - litter the 'doccheck' routine
in docprobe.c with printks. You do have CONFIG_MTD_DOCPROBE_HIGH turned
_off_ too, don't you? Another option which wants hiding.


-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> On Fri, 2 Feb 2001, Kevin Fenzi wrote:
>> I left it at 0. How can I tell what address the DOC is at? from the
>> bios?
>> 
>> So it would print something here if it found the DOC?

David> Yep, should do. Make it print stuff anyway - litter the
David> 'doccheck' routine in docprobe.c with printks. You do have

I can do that...

David> CONFIG_MTD_DOCPROBE_HIGH turned _off_ too, don't you? Another
David> option which wants hiding.

nope. I think I have it on. Would this be causing it to not detect it?

Let me add some printk's and turn that option off...

David> -- dwmw2

kevin




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> I can do that...

Don't bother, if DOCPROBE_HIGH was on - just turn it off.

> David> CONFIG_MTD_DOCPROBE_HIGH turned _off_ too, don't you? Another
> David> option which wants hiding.
> 
> nope. I think I have it on. Would this be causing it to not detect it?

Yep. It probes at different addresses (0xfffc8000-0xfffee000 instead of 
0xc8000-0xee000). Another LinuxBIOS thing.

-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> On Fri, 2 Feb 2001, Kevin Fenzi wrote:
>> I can do that...

David> Don't bother, if DOCPROBE_HIGH was on - just turn it off.

ok. Doing that now. ;) 

David> CONFIG_MTD_DOCPROBE_HIGH turned _off_ too, don't you? Another
David> option which wants hiding.
>>  nope. I think I have it on. Would this be causing it to not detect
>> it?

David> Yep. It probes at different addresses (0xfffc8000-0xfffee000
David> instead of 0xc8000-0xee000). Another LinuxBIOS thing.

Is there a way to eliminate this option, just probe the normal
addresses first and then the High addresses? or does that cause a
problem? 

David> -- dwmw2

kevin



To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> David> CONFIG_MTD_DOCPROBE_HIGH turned _off_ too, don't you? Another
> David> option which wants hiding.
> >>  nope. I think I have it on. Would this be causing it to not detect
> >> it?
> 
> David> Yep. It probes at different addresses (0xfffc8000-0xfffee000
> David> instead of 0xc8000-0xee000). Another LinuxBIOS thing.
> 
> Is there a way to eliminate this option, just probe the normal
> addresses first and then the High addresses? or does that cause a
> problem? 

That should be OK. But LinuxBIOS needs all the other options too. I think 
it's best just to hide all three inside CONFIG_LINUXBIOS or something.

-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> Don't bother, if DOCPROBE_HIGH was on - just turn it off.

David> CONFIG_MTD_DOCPROBE_HIGH turned _off_ too, don't you? Another
David> option which wants hiding.
>>  nope. I think I have it on. Would this be causing it to not detect
>> it?

David> Yep. It probes at different addresses (0xfffc8000-0xfffee000
David> instead of 0xc8000-0xee000). Another LinuxBIOS thing.

ok, here's what the 2.4.0 kernel gives me now: (making progress!)

M-Systems DiskOnChip driver. (C) 1999 Machine Vision Holdings, Inc.
DiskOnChip 2000 found at address 0xE0000                           
Flash chip found: Manufacturer ID: 98, Chip ID: 73 (Toshiba TH58V128DC)
3 flash chips found. Total DiskOnChip size: 48 Mb                      
mtd: Giving out device 0 to DiskOnChip 2000      
Ignoring DiskOnChip 2000 at 0xE2000 - already configured
Ignoring DiskOnChip 2000 at 0xE4000 - already configured
Ignoring DiskOnChip 2000 at 0xE6000 - already configured
Ignoring DiskOnChip 2000 at 0xE8000 - already configured
Ignoring DiskOnChip 2000 at 0xEA000 - already configured
Ignoring DiskOnChip 2000 at 0xEC000 - already configured
Ignoring DiskOnChip 2000 at 0xEE000 - already configured
ftl_cs: FTL header not found.                           
M-Systems NAND Flash Translation Layer driver. (C) 1999 MVHI
$Id: nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $          
NFTL_notify_add for DiskOnChip 2000               
mtd->read = c01c66f0, size = 50331648, erasesize = 16384
NFTL_setup                                              
Could not find valid boot record
Could not mount NFTL device     
...

NFTL_open                                          
ENODEV: thisNFTL = 0, minor = 23809, ip = c1173cf0, fp = c111bef8
VFS: Cannot open root device "5d01" or 5d:01                     
Please append a correct "root=" boot option 
Kernel panic: VFS: Unable to mount root fs on 5d:01

I do have the  Probe for 0x55 0xAA BIOS Extension Signature on, should
that also be off? 

David> -- dwmw2

kevin


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> M-Systems DiskOnChip driver. (C) 1999 Machine Vision Holdings, Inc.
> DiskOnChip 2000 found at address 0xE0000                           
> Flash chip found: Manufacturer ID: 98, Chip ID: 73 (Toshiba TH58V128DC)
> 3 flash chips found. Total DiskOnChip size: 48 Mb                      
> mtd: Giving out device 0 to DiskOnChip 2000      
> Ignoring DiskOnChip 2000 at 0xE2000 - already configured
> Ignoring DiskOnChip 2000 at 0xE4000 - already configured
> Ignoring DiskOnChip 2000 at 0xE6000 - already configured
> Ignoring DiskOnChip 2000 at 0xE8000 - already configured
> Ignoring DiskOnChip 2000 at 0xEA000 - already configured
> Ignoring DiskOnChip 2000 at 0xEC000 - already configured
> Ignoring DiskOnChip 2000 at 0xEE000 - already configured

That's amusing. you said the latest binary-only driver actually detected 
four units, didn't you? Looks like our code is doing better alias 
detection :)

> M-Systems NAND Flash Translation Layer driver. (C) 1999 MVHI
> $Id: nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $          
> Could not find valid boot record
> Could not mount NFTL device     

Hmmm. Failing to recognise the Media Header. #define ECC_DEBUG in
doc2000.c. It may well be the same problem that's causing 72MiB units to
fail.

If it complains about ECC errors, go to nftlmount.c. Find the bit which 
does this:

     if ((i & (SECTORSIZE - 1)) == 0) {
	/* read one sector for every SECTORSIZE of blocks */
	if (MTD_READECC(nftl->mtd, block * nftl->EraseSize +
		i + SECTORSIZE, SECTORSIZE,
		&retlen, buf, (char *)&oob) < 0)
		goto ReplUnitTable;
     }


Replace MTD_READECC with MTD_READ and omit the final (&oob) argument.

> I do have the  Probe for 0x55 0xAA BIOS Extension Signature on, should
> that also be off? 

No, on is good.


-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> That's amusing. you said the latest binary-only driver actually
David> detected four units, didn't you? Looks like our code is doing
David> better alias detection :)

yeah. ;) 

>> M-Systems NAND Flash Translation Layer driver. (C) 1999 MVHI $Id:
>> nftl.c,v 1.57 2000/12/01 17:51:54 dwmw2 Exp $ Could not find valid
>> boot record Could not mount NFTL device

David> Hmmm. Failing to recognise the Media Header. #define ECC_DEBUG
David> in doc2000.c. It may well be the same problem that's causing
David> 72MiB units to fail.

yeah, looks like:

...
DiskOnChip ECC Error: Read at 728000
Errors corrected: ffffffff
...

for all kinds of sectors. ;) 

David> If it complains about ECC errors, go to nftlmount.c. Find the
David> bit which does this:

David>      if ((i & (SECTORSIZE - 1)) == 0) { /* read one sector for
David> every SECTORSIZE of blocks */ if (MTD_READECC(nftl->mtd, block
David> * nftl->EraseSize + i + SECTORSIZE, SECTORSIZE, &retlen, buf,
David> (char *)&oob) < 0) goto ReplUnitTable; }

David> Replace MTD_READECC with MTD_READ and omit the final (&oob)
David> argument.

ok. 

So I have:

                                                /* read one sector for every SECTORSIZE of blocks */
                                                if (MTD_READ(nftl->mtd, block * nftl->EraseSize +
                                                                i + SECTORSIZE, SECTORSIZE,
                                                                &retlen, buf) < 0)

I still get:

tons of these...

DiskOnChip ECC Error: Read at 2fec000
Errors corrected: ffffffff           
DiskOnChip ECC Error: Read at 2ff0000
Errors corrected: ffffffff           
DiskOnChip ECC Error: Read at 2ff4000
Errors corrected: ffffffff           
DiskOnChip ECC Error: Read at 2ff8000
Errors corrected: ffffffff           
Could not find valid boot record
Could not mount NFTL device     
...
NFTL_open                                          
ENODEV: thisNFTL = 0, minor = 23809, ip = c1173cf0, fp = c111bef8
VFS: Cannot open root device "5d01" or 5d:01                     
Please append a correct "root=" boot option 
Kernel panic: VFS: Unable to mount root fs on 5d:01

Should I have to have the DEBUG_ECC off? 

Looks like in find_boot_record() at the top of that file is another
MTD_READECC...should I change that to MTD_READ? 

David> -- dwmw2

kevin


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[David Woodhouse]

On Fri, 2 Feb 2001, Kevin Fenzi wrote:

> Should I have to have the DEBUG_ECC off? 

Doesn't matter.

> Looks like in find_boot_record() at the top of that file is another
> MTD_READECC...should I change that to MTD_READ? 

Try it.

-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org

Re: DOC2000 issues...

[Kevin Fenzi]

>>>>> "David" == David Woodhouse <dwmw2@infradead.org> writes:

David> On Fri, 2 Feb 2001, Kevin Fenzi wrote:
>> Should I have to have the DEBUG_ECC off?

David> Doesn't matter.

>> Looks like in find_boot_record() at the top of that file is another
>> MTD_READECC...should I change that to MTD_READ?

David> Try it.

no dice. The ecc errors stopped, but it still doesn't seem to want to
read the DOC. ;( 

anything else I can try? 

David> -- dwmw2

kevin


To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org