[lkml]   [2004]   [Aug]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
SubjectRe: How to debug 2.6 PReP boot hang?
> > It's the one that reorganizes boot code:
> > PPC32: Kill off arch/ppc/boot/prep and rearrange some files.
> Sadly, that is what I expected. Try narrowing down the differences
> between prep/head.S and simple/head.S (or rather head.s via make
> arch/ppc/boot/prep/head.s and simple/head.s to strip out comments, etc).

I replaced the whole head.S with old prep one. Works, so head.S is
probably the culprit.

1. L1 cache disabling was moved. I moved it back to before cheking OF
address (but after recording link register into r3 since I left the
disabling a subroutine), it still hangs.

2. The address that the code is relocated to is computed differently.
Old code relocates itself to hardcoded 8M (start), new code to
max(load+size, end).

3. Relocation loop is different (see ms to be memmove, not memcpy, at
the first glance).

I replaced the calculation and relocate with old unconditional relocate,
it still hangs.

Replaced head.S with first half of old head.S (relocate from old,
start_ldr from new), still hangs.

Put a couple of debugging traps around. Trap before load_kernel is
executed, trap after load_kernel is never reached. So head.S changes
cause load_kernel to hang.

Some more changes and the only difference is load_kernel vs
decompress_kernel. Changing load_kernel call in relocate.S to
decompress_kernel makes the kernel boot. Oh, finally.

BTW, the address of OF residual data bounces around in
r3->r29->r4->r11->r6. The old code only did r3->r11->r6 but the new code
uses r11 for cache disabling.

This is the patch I'm using currently. I understand that this is
probably not the right patch for inclusion :)

===== arch/ppc/boot/simple/relocate.S 1.11 vs edited =====
--- 1.11/arch/ppc/boot/simple/relocate.S 2004-04-03 06:13:47 +03:00
+++ edited/arch/ppc/boot/simple/relocate.S 2004-08-18 23:27:39 +03:00
@@ -183,7 +183,7 @@
mr r5,r6 /* Checksum */
mr r6,r11 /* Residual data */
mr r7,r25 /* Validated OFW interface */
- bl load_kernel
+ bl decompress_kernel

* Make sure the kernel knows we don't have things set in
Meelis Roos (

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:05    [W:0.043 / U:25.148 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site