Skip to contentSkip to conversation

AT91 Kernel oops when loading kernel from dataflash.

U-Boot-Users Raúl Sánchez Siles 2 participants 4 messages Nov 8, 2006 – Nov 15, 2006

lore
Share & tools
lore
#0 lore raw

Hello all:

This is the first time I write to the list, and I appreciate the big help it
gives us users.

We're using an AT91RM9200 based board called Portux920T. We have now a quite
stable kernel and u-boot configuration which I attach. We manage to include a dataflash inside the portux board and get it to work. At least almost, please read on.

When doing big transfers in memory (10M), we have some kernel oopses(see
panic.log.zip attached). The oops comes up in the function __wake_up_common in the file kernel/sched.c

The steps to reproduce this are the following:

1- Start the first bootloader (used the binary provided by atmel). 2- Make the first bootloader start u-boot(1.1.6).
3- U-boot downloads kernel(2.6.18) from _dataflash_ into RAM. 4- Rest of booting till shell prompt.
5- Execute for example: dd if=/dev/zero of=/root/borrar bs=1k count=10k 6- Oops!

If we substitute step 3 for U-boot downloads kernel from _parallel flash_
into RAM, the Oops won't happen.

The kernel has been patched with the latest maxim(2.6.18) patchset for the
AT91RM9200 microcontroller. The u-boot configuration is also attached (portux920T.h).

We have also tried using different first stage bootloaders we could find.
Even we compile it ourselves using the RAM initialisation code taken from the u-boot. We also have tested several toolchains, from emdebian to the one provided by portux.

We have 64MB Ram and we have tried using 64MB 32bit wide and 32MB 16bit
wide. Flash and Dataflash are both 4MB.
We will much appreciated whatever info or test that could take out from this
works but... situation. Thank you very much.

Regards,

Signature · 29 lines
Ra?l S?nchez Siles
-------------- next part --------------
A non-text attachment was scrubbed...
Name: config-2.6.18.2.gz
Type: application/x-gzip
Size: 7586 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061108/b6633d0a/attachment.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: portux920T.h.gz
Type: application/x-gzip
Size: 3064 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061108/b6633d0a/attachment-0001.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: panic.log.gz
Type: application/x-gzip
Size: 6518 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061108/b6633d0a/attachment-0002.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061108/b6633d0a/attachment.pgp 
#1 lore raw
Quoted 14 lines · Ra?l S?nchez Siles

Hello all:

This is the first time I write to the list, and I appreciate the
big help it gives us users.

We're using an AT91RM9200 based board called Portux920T. We have
now a quite stable kernel and u-boot configuration which I attach. We manage to include a dataflash inside the portux board and get it to work. At least almost, please read on.

When doing big transfers in memory (10M), we have some kernel
oopses(see panic.log.zip attached). The oops comes up in the function __wake_up_common in the file kernel/sched.c

Quoted 8 lines

The steps to reproduce this are the following:

1- Start the first bootloader (used the binary provided by atmel). 2- Make the first bootloader start u-boot(1.1.6).
3- U-boot downloads kernel(2.6.18) from _dataflash_ into RAM. 4- Rest of booting till shell prompt.
5- Execute for example: dd if=/dev/zero of=/root/borrar bs=1k count=10k 6- Oops!

You do not say that you are loading a ramdisk.
Do you have the file system in dataflash?
If not, I do not see how this can be influenced by the dd command... If it is, then the 4 MB flash seems small for a 10 MB copy!

I can see two scenarios where the dataflash can cause problems.

1) The dataflash blocksize is not 1024 bytes, it is 1056 bytes
2) You are running the dataflash > 5 Mbps
The PDC of the SPI must not be starved of bus cycles,
or you are in trouble unless the H/W manages chipselect through GP I/O.

I would try

$ dd if=/dev/zero of=/root/borrar bs=1056 count=10k

to start with, abnd if this works, start digging.

Quoted 13 lines

If we substitute step 3 for U-boot downloads kernel from _parallel
flash_ into RAM, the Oops won't happen.

The kernel has been patched with the latest maxim(2.6.18) patchset
for the AT91RM9200 microcontroller. The u-boot configuration is also attached (portux920T.h).

We have also tried using different first stage bootloaders we could
find. Even we compile it ourselves using the RAM initialisation code taken from the u-boot. We also have tested several toolchains, from emdebian to the one provided by portux.

Quoted 22 lines

We have 64MB Ram and we have tried using 64MB 32bit wide and 32MB
16bit wide. Flash and Dataflash are both 4MB.
We will much appreciated whatever info or test that could take out
from this works but... situation. Thank you very much.

Regards,

Quoted 13 lines

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642

_______________________________________________
U-Boot-Users mailing list
U-Boot-Users at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/u-boot-users

Best Regards,
Ulf Samuelsson
ulf at atmel.com

GSM:  +46 (706) 22 44 57
Tel:     +46  (8) 441 54 22
Fax:     +46 (8) 441 54 29
Mail: Box 2033  174 02 Sundbyberg
Visit: Kavalleriv?gen 24
          174 58 Sundbyberg'

Sweden

#2 lore raw

El Jueves, 9 de Noviembre de 2006 08:02, Ulf Samuelsson escribi?:

Quoted 11 lines
Quoted 7 lines · Ra?l S?nchez Siles

Hello all:

We're using an AT91RM9200 based board called Portux920T. We have
now a quite stable kernel and u-boot configuration which I attach. We manage to include a dataflash inside the portux board and get it to work. At least almost, please read on.

You do not say that you are loading a ramdisk.
Do you have the file system in dataflash?

You are right, sorry. On this test case we are not using initrd, since the
rootfs is on an MMC card formated with ext3.

Quoted 10 lines

If not, I do not see how this can be influenced by the dd command... If it is, then the 4 MB flash seems small for a 10 MB copy!

I can see two scenarios where the dataflash can cause problems.

1) The dataflash blocksize is not 1024 bytes, it is 1056 bytes
2) You are running the dataflash > 5 Mbps
The PDC of the SPI must not be starved of bus cycles,
or you are in trouble unless the H/W manages chipselect through GP I/O.

Quoted 5 lines

I would try

$ dd if=/dev/zero of=/root/borrar bs=1056 count=10k

to start with, abnd if this works, start digging.

I am afraid that in this situation where dataflash is not used once the board is started, the blocksize wouldn't need to make a difference.

BTW we tried that dd command and crashed as well.

Quoted 9 lines
Quoted 9 lines

The kernel has been patched with the latest maxim(2.6.18) patchset
for the AT91RM9200 microcontroller. The u-boot configuration is also attached (portux920T.h).

We have also tried using different first stage bootloaders we could
find. Even we compile it ourselves using the RAM initialisation code taken from the u-boot. We also have tested several toolchains, from emdebian to the one provided by portux.

Quoted 6 lines
Quoted 6 lines

We have 64MB Ram and we have tried using 64MB 32bit wide and 32MB
16bit wide. Flash and Dataflash are both 4MB.
We will much appreciated whatever info or test that could take out
from this works but... situation. Thank you very much.

Regards,

Reviewing my last e-mail I notice a mistake. I tried to send a graphic, but
according the list rules, it is not allowed. So instead I try to explain it with some ascii art ;).

This is the first scenario, where several configurations are tested. This
scenario does not work.

Stored in: Scenario 1
(or) (Problems)

Dataflash 1st Bootloader

Dataflash U-boot
Serial port

Dataflash
Parallel flash Kernel->RAM
Serial port
NFS

This is the second scenario, all these test works:

Scenario 2 Stored in:
(Works)

U-boot Parallel flash(1st position)

Kernel->Ram Dataflash
Parallel flash
Serial port
NFS

Now common to both scenarios the Rootfs can be arranged as explained below:

Scenario 1 | Scenario 2
----------------------ROOTFS----------------------------------
Flash->initrd(ext2) | MMC(ext3 or reiserfs) | NFS *

*(didn't fail in any case AFAIK)

Every scenario has its variables on the left for the 1st scenario and on the
right for the 2nd. The 1st doesn't work, the 2nd works always. The variables means that we have tried that option, i.e.: Copying kernel to ram downloading it from dataflash, parallel flash, etc..

The lower part means the root filesystem layout. In the case I exposed, the
rootfs was on an MMC card formatted with ext3. But we also tried other possibilities, but these are common to both scenarios, on 1st didn't work, on 2nd it worked. Take into account that in this case, once the system is running, AFAIK, we are not accessing the kernel or u-boot downloading source (i.e.: dataflash, flash).

When I say work, I mean dd is successful. We have also tried several dd
cases.

Looking at this scheme, one could think that the problem is on the 1st stage
bootloader. We have tried the binary provided by atmel, several others binaries seen on the internet, we have compiled it ourselves using u-boot initialisation routines, or without them.

I hope this explanation is clarifying enough, so anyone could give us
further info.

Regards,

Signature · 8 lines
Ra?l S?nchez Siles
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061113/7b74f68e/attachment.pgp 
#3 lore raw

El Lunes, 13 de Noviembre de 2006 11:30, escribi?:

Quoted 15 lines

El Jueves, 9 de Noviembre de 2006 08:02, Ulf Samuelsson escribi?:

Quoted 14 lines
Quoted 13 lines · Ra?l S?nchez Siles

Hello all:

We're using an AT91RM9200 based board called Portux920T. We have
now a quite stable kernel and u-boot configuration which I attach. We manage to include a dataflash inside the portux board and get it to work. At least almost, please read on.

We have 64MB Ram and we have tried using 64MB 32bit wide and 32MB
16bit wide. Flash and Dataflash are both 4MB.
We will much appreciated whatever info or test that could take out
from this works but... situation. Thank you very much.

Quoted 6 lines
Quoted 1 line
Quoted 1 line

Regards,

Looking at this scheme, one could think that the problem is on the 1st
stage bootloader. We have tried the binary provided by atmel, several others binaries seen on the internet, we have compiled it ourselves using u-boot initialisation routines, or without them.

...And indeed it was on the first stage bootloader. There was a problem with the SDRAMC Configuration Register. The problem was that the bootloader we took were for boards using 16MB of RAM, 16bit wide, while the Portux board has 2 32MB chips, 16bit wide each one.

We modified the 1st bootloader source, so that the SDRAMC_CR programming
changes the NR (Number of Row bits) from 12 to 13. That was the problem.

If anyone needs further clarification, just ask.

Thank you for the advises.

Regards,

Signature · 8 lines
Ra?l S?nchez Siles
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.denx.de/pipermail/u-boot/attachments/20061115/590ff1d0/attachment.pgp