BlaisorBlade wrote:
>>previously I use uml-2.4.20 with a redhat9 rootfs
perfectly. now I'm
>>trying vallina-2.6.3, with uml-patch-2.6.3-rc2-1
donwloaded from sf.
>>rootfs is the same.
>>
>>The symptom is: it hang on boot right after "mount
devfs", and consuming
>>100% CPU all the time until was manually killed.
remove the 'devfs' option
>>dosen't help: it hang again right after "mount ext3
root filesystem
>>readonly". break into gdb only got trace begin from
something like
>>"sigint_handler".
>
>
> There is a problem that happens differently on 2.6
UML kernels. This has been
> reported as "2.6 stops at NET: Registered protocol
family" and so on. Or
> better, we have some evidence about what that hang
relates to; and so we can
> guess you have the same problem. The bug does not
hit devfs, ext3, NET or
> anything such; it hits the core scheduler - or
better wait queues and
> notifications (i.e. when code inside the kernel
decides to wait for
> something, then this bug makes things not work).
>
> Would you post the backtrace you get, so we can be
sure about what is
> happening? Also, you could probably try
vanilla-2.6.0 linux kernel and
> patch-2.6.0-test9 (and disable module support). A
lot of people reported that
> patch makes the problem go away.
Thanks you very much for the reply!
for some reason I must use 2.6.3. following is the
backtrace:
(gdb) bt
#0 0xa011f4a1 in sigprocmask () at seq_file.h:49
#1 0xa00173c2 in change_sig (signal=10, on=1)
at arch/um/kernel/signal_user.c:56
#2 0xa001cb70 in sig_handler_common_tt (sig=11,
sc_ptr=0x58)
at arch/um/kernel/tt/trap_user.c:33
#3 0xa001955f in sig_handler (sig=-1577124956, sc=
{gs = 7, __gsh = 0, fs = 0, __fsh = 0, es = 123,
__esh = 0, ds = 123, __dsh = 0, edi = 134534120, esi =
1073828704, ebp = 2684354264, esp = 2684354216, ebx =
18, edx = 2684354522, ecx = 791621423, eax = 60,
trapno = 14, err = 6, eip = 2686245736, cs = 115,
__csh = 0, eflags = 2163270, esp_at_signal =
2684354216, ss = 123, __ssh = 0, fpstate = 0xa1fefd80,
oldmask = 134283264, cr2 = 60})
at arch/um/kernel/trap_user.c:102
#4 <signal handler called>
#5 0xa01cdb68 in __kernel_vsyscall ()
Previous frame inner to this frame (corrupt stack?)
(gdb)
At the hang time, the two processes which occupy all
CPU power is:
- 1090 pts/0 S 0:12 ./linux [(tracing thread)]
- 1122 pts/0 R 0:17 ./linux [/sbin/init]
AFAIK, tracing thread only get control when process
within UML issue syscalls. So it looks like /sbin/init
is trying to do something via syscalls, but always
failed, so it repeatedly issue out syscalls?
Best Regards
MeiJia
_________________________________________________________
Do You Yahoo!?
完全免费的雅虎电邮，马上注册获赠额外60兆网络存储空间
http://cn.rd.yahoo.com/mail_cn/tag/?http://cn.mail.yahoo.com
-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click
|
