Hello Jurgen,

Did you ever find out what is going wrong with your BSP? I have the same problem running the Logic Linux demo - haven't quite gotten to the point of building my own one yet. I'd be very grateful if you or anyone else could shed some light on it.

Cheers,

james

Hi James,

no, I haven't had any response from logic. I am trying some other channels but haven't found a solution yet.

regards,

Jurgen

Thanks Jurgen. I've tried tinkering with a lot of things but none of them make any difference. If even the provided Linux demo doesn't work then what hope do I have for building my own? I'd be interested to hear if anyone has actually been able to build and deploy a Linux BSP as the small amount of traffic here suggests that we might be the only people trying to do this

Please post here if you have any joy with Logic.

james

I built my own image today and got the same result when I tried to run it on the board. Here's my output:

Linux version 2.4.26 (jpinakis@qrssrv01) (gcc version 3.4.3) #5 Mon Jan 16 15:27
:04 WST 2006
starting up linux rev 0.2: startmem 0xc022e000, size 61MB
Flex CS0: ff800000 007f0001 00100d80
Flex CS1: e0000000 07ff0001 00100d80
Flex CS4: 40000000 0fff0001 00103580
Flex CS5: 50000000 0fff0001 00101580
On node 0 totalpages: 8192
zone(0): 1024 pages.
zone(1): 7168 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/mtdblock/2 rootfstype=yaffs rw
Have no hardware time clock, returning ToD as 00:00:00 on 1 Jan 1970
*** ILLEGAL INSTRUCTION *** FORMAT=4
Current process id is 0
BAD KERNEL TRAP: 00000000
PC: [<c0047128>]
SR: 2704 SP: c022bf80 a2: c022a000
d0: 00000001 d1: 00000000 d2: c01eb5b0 d3: 00000001
d4: 00000000 d5: 00000000 a0: c01b2f10 a1: c01b1d40
Process swapper (pid: 0, stackpage=c022c000)
Fault Status=0x0:Normal
Stack from c022bfb4:
c00470d4 c00dd3ae c021f322 00000000 00000084 00000000 c022bfd8 c01b2ecc
00000000 c022bffc 00000001 00000000 c01eb5b0 c01eb5b0 c022bffc c021cc34
00067564 c01e24bc 00000000
Call Trace:
[<c00dd3ae>]
Code: 00a3 46e2 eafa 2a2f 9151 214a d051 19de 902d a2a0
Kernel panic: Attempted to kill the idle task!
In idle task - not syncing

Anyone out there got any ideas? We'd really like to start looking into using this processor for development and are being held up by this.

james

> Anyone out there got any ideas? We'd really like to start looking into using this
> processor for development and are being held up by this.

1) Which SOM are you using?
2) Which config are you using?
3) Have you customized the config?
4) Which version of LoLo is on the board?

The error looks valid(i.e.0x00a3 is an invalid opcode for ColdFire, but the traceback doesn't look right). Could you also respond with the output from the following:

[LinuxBSP-22012005]$ m68k-linux-addr2line -e linux/vmlinux 0xc0047128 0xc00dd3ae

Posted By peterb on 17 Jan 2006 0:7 PM
> Anyone out there got any ideas? We''d really like to start looking into using this
> processor for development and are being held up by this.

1) Which SOM are you using?

We have the M5485EVB.

Quote:
2) Which config are you using?

config-m5485cfe-evb-yaffs

Quote:
3) Have you customized the config?

No, we're just using the default one.

Quote:
4) Which version of LoLo is on the board?

Version 2.0.5-MCF5485_10 0001

Quote:
The error looks valid(i.e.0x00a3 is an invalid opcode for ColdFire, but the traceback doesn't look right). Could you also respond with the output from the following:

[LinuxBSP-22012005]$ m68k-linux-addr2line -e linux/vmlinux 0xc0047128 0xc00dd3ae

LinuxBSP-22012005$ /opt/logicpd/ColdFire/m68k-linux/gcc-3.4
.3-glibc-2.3.5/bin/m68k-linux-addr2line -e linux-2.4.26/vmlinux 0xc0047128 0xc00dd3ae
??:0
/home/NT1/jpinakis/LinuxBSP-22012005/linux-2.4.26/mm/slab.c:391

LinuxBSP-22012005$

After I read the post from Logic, I wondered if the version of LoLo didn't have something to do with it, so I downgraded to version 2.0.2. The linux image then booted properly! No idea why, but perhaps that's something Logic could look at.

james

Hi,

that sounds like great news to me. Where can I pick up that Logicloader version? Can you send it to me ?

kind regards,

Jurgen
Email: jurgen@metacom.co.za

You can download this on the product downloads page for your kit under the "LogicLoader Bootloader" category. You're going to want to make sure that you're registered for this product first.

Thanks,

The issue turns out that the download buffer used by LoLo to hold the elf file was moved up in memory between version 2.0.2 and 2.0.5, and this was not an issue for the romfs image(since in that case is just one *giant* segment that has relocatable code at the beginning), but the vmlinux in the yaffs rootfs has a hole in the middle(where the .bss segment is located), and that hole causes LoLo to incorrectly choose to copy down in memory the 2nd segment of the ELF file, and accidentally overwrite a portion of the first segment. Hence the failure.

I've updated the LoLo development tree, and this fix will show up in the next version of LoLo.

I upgraded to the latest version of logicloader (2.0.6), but the bug is still there. When will we get a fixed version?

André Lämmer

Here's a workaround you can use to load the YAFFS kernel from the root filesystem and boot it.

1) build/install the root filesystem.

2) On your linux development host, create a raw binary file of the linux kernel(note its name is vmlinux-absolute), adjusting the load addresses of each segment so the start of the raw file lines up with the start of the kernel (at address 0xc0000) (note the backslashes to break up the command line into something readable):

[peter@fast LinuxBSP-22012005]$ /opt/logicpd/ColdFire/m68k-linux/gcc-3.4.3-glibc-2.3.5/bin/m68k-linux-objcopy -O binary \
--change-section-lma .text-0xc0000 \
--change-section-lma .text.exit-0xc0000 \
--change-section-lma .rodata-0xc0000 \
--change-section-lma .kstrtab-0xc0000 \
--change-section-lma __ex_table-0xc0000 \
--change-section-lma __ksymtab-0xc0000 \
--change-section-lma .data-0xc0000 \
--change-section-lma .exitcall.exit-0xc0000 \
--change-section-lma .data.cacheline_aligned-0xc0000 \
--change-section-lma .text.init-0xc0000 \
--change-section-lma .data.init-0xc0000 \
--change-section-lma .setup.init-0xc0000 \
--change-section-lma .initcall.init-0xc0000 \
--change-section-lma init_task-0xc0000 \
linux-2.4.26/vmlinux-absolute linux-2.4.26/vmlinux.bin

3) Copy this to your tftpserver root (mines on a different machine called 'pogo'):

[peter@fast LinuxBSP-22012005]$ scp linux-2.4.26/vmlinux.bin peter@pogo:/tftpboot/

4) On the board, mount the root filesystem:

losh> add-yaffs flash nor 0xe0000000 0x01000000
losh> mount yaffs /flash

5) bring up the ethernet port (I use DHCP):

losh> ifconfig sm0 dhcp

6) Copy the binary kernel file from the tftpserver over to the root filesystem in /boot:

losh> cp /tftp/192.168.3.3:vmlinux.bin /flash/boot/vmlinux.bin

7) Find out the size of /boot/vmlinux.bin (since its required for the 'load raw' command):

losh> ls /flash/boot/vmlinux.bin
: vmlinux.bin 1490944

8) using the size returned (in my case 1490944), load the binary kernel to address 0xc0000:

losh> load raw 0xc0000 1490944 /flash/boot/vmlinux.bin
loading from /flash/boot/vmlinux.bin:
loading raw binary to 0x000c0000 (ram) len 0016c000:
......................................................................................................................................................................................md5sum of _loaded_ portion of the file:
c64877998e8e6fe84ef9216f67b7d68a - addr: 000c0000 len: 0016c000
file loaded

9) execute it:

losh> exec
Linux version 2.4.26 (peter@fast) (gcc version 3.4.3) #1 Mon Apr 3 09:36:22 EDT
2006
starting up linux rev 0.2: startmem 0xc022e000, size 61MB
Flex CS0: ff800000 007f0001 00100d80
Flex CS1: e0000000 07ff0001 00100d80
Flex CS4: 40000000 0fff0001 00103580
Flex CS5: 50000000 0fff0001 00101580
On node 0 totalpages: 8192
zone(0): 1024 pages.
zone(1): 7168 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/mtdblock/2 rootfstype=yaffs rw
Have no hardware time clock, returning ToD as 00:00:00 on 1 Jan 1970
Calibrating delay loop... 199.88 BogoMIPS
Memory: 62880k/65536k available (968k kernel code, 1632k data, 56k init)
...

I have a fix for the corrupted elf file, and it will be in the next release of LoLo.

Posted By peterb on 5 Apr 2006 10:17 AM
Here''s a workaround you can use to load the YAFFS kernel from the root filesystem and boot it.

1) build/install the root filesystem.

Is the bug now fixed in the latest lolo?

André

Yes, LoLo 2.0.8 has the fix to allow you to exec the kernel from the yaffs image:

*****************************************************************

LogicLoader

(c) Copyright 2002-2005, Logic Product Development, Inc.
All Rights Reserved.
Version 2.0.8-MCF5475_10 0001

*****************************************************************

losh> erase 0xe0000000 0x01000000
unlocking flash blocks
erasing flash: 100%
locking flash blocks
erased '/dev/flash1' start=0x0: len=0x1000000 bytes
losh> ifconfig sm0 dhcp
WARNING -- valid config block not found - using default MAC address
Starting DHCP on sm0
losh> load raw 0xe0000000 9830400 /tftp/192.168.3.3:yaffs-rootfs-m5475cfe-evb.img
loading from /tftp/192.168.3.3:yaffs-rootfs-m5475cfe-evb.img:
loading raw binary to 0xe0000000 (flash) len 00960000:
....ignoring rest of file...done
file loaded to ram buffer, destined for flash addr: 0xe0000000
losh> burn
burning 9830400 bytes from 00200000 to /dev/flash1
unlocking flash blocks
erasing flash: 100%
burning flash: 100%
locking flash blocks
burn good (9830400 bytes)
losh> add-yaffs flash nor 0xe0000000 0x01000000
add-yaffs: inserted: start 0xe0000000 - len 0x1000000
losh> mount yaffs /flash
Mounting a YAFFS partition.

---------------------------------------------------
Starting YAFFS file system:
Current YAFFS partition table:
0: flash 00000000 - 0000007f ( 0)
1: <NULL>
2: <NULL>
3: <NULL>
YAFFS initialized.
---------------------------------------------------

losh> load elf /flash/boot/vmlinux
loading from /flash/boot/vmlinux:
........................................................................................................................................................................
ELF section 1: buffer address: 0x00202000 dest: 0x000c0000
ELF section 2: buffer address: 0x00339000 dest: 0x00237000
warning: guessing combined program header size as 1270540
loaded 1356556 @ 0x000c0000 Ram
file loaded
losh> exec
rec_check: 0xc024c000
Linux version 2.4.26 (peter@thunk) (gcc version 3.4.3) #11 Thu Jul 13 18:50:50 EDT 2006
starting up linux rev 0.2: startmem 0xc024e000, size 61MB
Flex CS0: ff800000 007f0001 00101180
Flex CS1: e0000000 07ff0001 00101180
Flex CS4: 40000000 0fff0001 00104580
Flex CS5: 50000000 0fff0001 00101d80
On node 0 totalpages: 8192
zone(0): 1024 pages.
zone(1): 7168 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/mtdblock/2 rootfstype=yaffs rw
Have no hardware time clock, returning ToD as 00:00:00 on 1 Jan 1970