[OLPC-devel] Weekly work summary [6/12]

[Ronald G Minnich]

weekly status, eh? I will join the fun :-)

Well ...

1. linuxbios continues to work for me on the board. I last flashed it a 
few weeks ago, and have not needed to reflash since.

2. kexec continues not to work, for certain cases, in particular the 
kernels distributed as part of the OLPC images. As kexec is THE key 
piece for our boot strategy using linuxbios, I want to figure this out.

Latest on this is that if the initrd+kernel size is > 1M, then the kexec 
will fail, but the error is BIZARRE:
Probing CS553x NAND controller CS#0 at MMIO 0x20000000
BUG: unable to handle kernel paging request at virtual address c7800000
  printing eip:
c01c1126
*pde = 00000000
Oops: 0000 [#1]
Modules linked in:
CPU:    0
EIP:    0060:[<c01c1126>]    Not tainted VLI
EFLAGS: 00010296   (2.6.17jhc #4)
eax: c77c1f78   ebx: 00000000   ecx: 00000001   edx: 00000098
esi: c76a4214   edi: c7800000   ebp: c76a4160   esp: c77c1f68
ds: 007b   es: 007b   ss: 0068
Process swapper (pid: 1, threadinfo=c77c0000 task=c111a480)
Stack: c0217f28 00000000 c0217f28 000000dc 000000ec c76a4214 c76a4160 
c76a4c0c
        20000000 c022c41e c76a4160 00000001 fffffffa 00000000 00000000 
c0234bf0
        00000000 00000000 c01002dd c0101f86 c020b5a0 c010027a 00000000 
00000000
Call Trace:
  <c022c41e>   <c01002dd>
  <c0101f86>   <c010027a>
  <c010027a>   <c01007f5>
Code: 8b 14 24 8b 42 10 89 45 0c 8b 42 08 89 45 10 c1 e8 05 89 45 14 8b 
5a 14 83 e3 02 8b 15 80 7a 21 c0 85 d2 8d 44 24
EIP: [<c01c1126>]  SS:ESP 0068:c77c1f68
  <0>Kernel panic - not syncing: Attempted to kill init!

This makes ZERO sense to me at present. This appears to be the kernel 
kmalloc'ing data above the top of memory, where the frame buffer lives. EH?

next step is to boot with CS5535x as a module.

If I can get past this, then we've licked this problem.

3. For flash burn, doing this will be fairly easy, but I am concerned 
about some issues w.r.t. the EC, so am waiting on some feedback from 
Quanta on their comments in the sample 'flash enable' code.

4. Steve Goodrich is going to get me the final go-round of register 
settings this week and we will apply to linuxbios. This last change
should bring to a close our register setting issues.

5. I tested using initramfs instead of initrd, and it works, but 
consensus is that we should stick with initrd. Initramfs has some 
baggage. The one good thing about initramfs is that it is one file 
(bzImage) instead of two (bzImage + inird). But general feeling is that
a seperate initrd file is easier.

6. We need a DSDT for linuxbios. AMD will be delivering that this week I 
hope. ACPI is the next big hurdle.

thanks

ron