Re: [RFH] Octeon support for debian.org machines
Hi,
On Thursday 04 March 2010 16:10:42 Martin Zobel-Helas wrote:
> Hi,
>
> [please CC me, i am not subscribed to the list]
>
> i am currently busy setting Debian on MOVIDIS Revolution x16 (Octeon
> 3860NSP) machines. I compiled Linus 2.6.33 with .config [1], including
> the following patches:
>
> * http://www.linux-mips.org/archives/linux-mips/2006-01/msg00181.html
> * in arch/mips/cavium-octeon/csrc-octeon.c (so we always use the ugly
> way):
> -#if (__GNUC__ < 4) || ((__GNUC__ == 4) && (__GNUC_MINOR__ <= 3))
> +#ifndef __NNNNNNNNNNNNNNNNNNNNNNNNNNN__
>
> * set a symlink from linux-2.6.33/include/asm-mips to asm
>
> The kernel compiles fine, but i don't get it running:
>
> Movidis X16 U-Boot version 1.6.1 (Build time: Nov 23 2007 - 18:27:52)
>
> | CUST_WSX16 board revision major:1, minor:1, serial #: XXXXXXXXX
> | OCTEON CN38XX-EXP revision: 3, Core clock: 500 MHz, DDR clock: 333 MHz
> | (666 Mhz data rate)
> | DRAM: 2048 MB
> | Flash: 2 MB
> | IPD backpressure workaround verified, took 31 loops
> | Clearing DRAM........ done
> | BIST check passed.
> | Starting PCI
> | PCI Status: PCI-X 64-bit
> | PCI BAR 0: 0x00000000, PCI BAR 1: Memory 0x00000000 PCI 0xf8000000
> | Net: octeth0, octeth1, octeth2, octeth3, octeth4, octeth5, octeth6,
> | octeth7
> | 0
> | Movidis x16> printenv
> | bootdelay=0
> | baudrate=115200
> | download_baudrate=115200
> | bootloader_flash_update=protect off 0xbf430000 0xbf47ffff;erase
> | 0xbf430000 0xbf47ffff;cp.b 0x100000 0xbf430000 0x50000
> | flash_update=protect off 0xbec30000 0xbec7ffff;erase 0xbec30000
> | 0xbec7ffff;cp.b 0x100000 0xbec30000 0x50000 gatewayip=192.168.1.1
> | netmask=255.255.255.0
> | dnsip=192.168.1.1
> | ipaddr=192.168.1.77
> | serverip=192.168.1.151
> | autostart=yes
> | bootcmd=fsload vmlinux.64; bootoctlinux 100000 numcores=16 mem=0
> | root=/dev/sda1 bootsas=fsload vmlinux.64; bootoctlinux 100000 numcores=16
> | mem=0 root=/dev/sda1 bootfile=vmlinux.64
> | ethact=octeth1
> | loadaddr=0x100000
> | coremask_override=0xffff
> | numcores=16
> | stdin=serial
> | stdout=serial
> | stderr=serial
> | Movidis x16> setenv ipaddr 192.168.2.17
> | Movidis x16> setenv serverip 192.168.2.10
> | Movidis x16> setenv ethact octeth2
> | Movidis x16> tftpboot 0x0 vmlinux-2.6.33-dsa-octeon
> | octeth2: Up 1Gbs Full duplex, (port 2)
> | Using octeth2 device
> | TFTP from server 192.168.2.10; our IP address is 192.168.2.17
> | Filename 'vmlinux-2.6.33-dsa-octeon'.
> | Load address: 0x100000
> | Loading: ##########################################
> | done
> | Bytes transferred = 6000378 (5b8efa hex), 13564 Kbytes/sec
> | Movidis x16> bootoctlinux 0x100000 numcores=16 mem=0 root=/dev/sda1
> | panic=30 argv[2]: numcores=16
> | argv[3]: mem=0
> | argv[4]: root=/dev/sda1
> | argv[5]: panic=30
> | ELF file is 64 bit
> | Allocated memory for ELF segment: addr: 0x1100000, size 0x485ad0
> | Error allocating memory for elf image!
> | ## Loading Linux kernel with entry point: 0x00100000 ...
> | Bootloader: Done loading app on coremask: 0xffff
>
> After that, the machine seems dead.
>
> Any help how to proceed appreciated.
Make sure that your kernel does not include a .notes section. I ran into the
same problem on my evaluation board a while ago. To strip down this section
use the following command:
# Workaround pre-SDK-1.9.0 u-boot versions not handling the .notes
section
$(TARGET_CROSS)strip -R .notes vmlinux.elf -o fixed-vmlinux.elf
>
> Greetings
> Martin
>
> [1] http://people.debian.org:~zobel/octeon.config
>
Reply to: