From: Daniel Drake <dsd(a)laptop.org>
Subject: Re: New F11 for the XO-1 Build 6
To: "Yioryos Asprobounitis" <mavrothal(a)yahoo.com>
Cc: dirakx(a)gmail.org, "OLPC Testing" <testing(a)lists.laptop.org>,
"OLPC Development" <devel(a)lists.laptop.org>, "sugar-devel"
<sugar-devel(a)lists.sugarlabs.org>, "Fedora OLPC List"
<fedora-olpc-list(a)redhat.com>, "Mitch Bradley" <wmb(a)laptop.org>
Date: Wednesday, August 26, 2009, 11:16 PM
2009/8/27 Yioryos Asprobounitis
<mavrothal(a)yahoo.com>:
> I had this issue with os5 (see "...-swap" posts above
in the fedora-olpc-list). It is something not related to
swap but is happening often enough. Usually after 1-3
attempts removing all peripherals (and overclocking :)
recovers and boots fine. Didn't have it with os6 yet.
> Of course the old firmware does not help either :)
If it's freezing after the "loading ramdisk" line, then
it's not
related to swap. That's way too early for swap to be having
an effect.
http://dev.laptop.org/ticket/9100
I also see this bug very often and work around it simply by
turning
off the machine for 5 minutes and trying again. Here are my
findings
so far:
- nothing is output from the kernel on the serial console,
so it is a
very early freeze from the kernel standpoint
- the serial console does show a newline character *after*
the
"loading ramdisk" line, which seems to indicate that the
ramdisk does
load correctly and the crash is triggered by something that
happens
after
- the OFW frame key does not cause the debugger to start
- waiting a few minutes does seem to increase success rate
- I have
had very little success with on,off,on,off,on,off type
cycles, but if
you insert a few minutes wait inbetween then it does feel
to be more
likely to boot
- once booted, you can reboot indefinitely and it won't
hang
- I discount reports that this is related to external
devices, swap
partitions, etc. I run my XOs with the standard
configuration and no
external devices and I frequently see this bug
- I've seen this ever since we upgraded kernel after 8.2
- the 8.2 kernels continue to work fine
Daniel
I subscribe 100% to your description. Actually when I say "1-3 attempts" I imply
the old crash practice: Power off-"count 30 elephants"-power on :)
I guess the olpc-aware kernels may not be fully aware of olpc...
How do you debug these *sooo* early things???...