2.5.38 semaphore.c calls sleeping function in illegal context
ksymoops 2.4.6 on i686 2.5.38-2. Options used
-v /mnt/b/2.5.38/vmlinux-2.5.38-2 (specified)
-K (specified)
-L (specified)
-O (specified)
-m /boot/System.map-2.5.38-2 (specified)Reading Oops report from the terminalc02c9f6c c02c9f84 c01175f7 c0260f80 c0261500 0000007e c02c9f98 c011a4a1
c0261500 0000007e 00000000 c02c9fac c011a78a c0312480 c0278213 c0383244
c02c9fb8 c02d8316 c02ba640 c02c9fc0 c02d8164 c02c9fd0 c02d8a4b 00000000
Call Trace: [<c01175f7>] [<c011a4a1>] [<c011a78a>] [<c0105000>] Warning (Oops_read): Code line not seen, dumping what data is available
Trace; c01175f7 <__might_sleep+27/2b>
Trace; c011a4a1 <acquire_console_sem+2d/50>
Trace; c011a78a <register_console+122/1cc>
Trace; c0105000 <_stext+0/0>
1 warning issued. Results may not be reliable.
Quoted 13 lines · William Lee Irwin III
ksymoops 2.4.6 on i686 2.5.38-2. Options used
-v /mnt/b/2.5.38/vmlinux-2.5.38-2 (specified) -K (specified) -L (specified) -O (specified) -m /boot/System.map-2.5.38-2 (specified)Reading Oops report from the terminal
c02c9f6cc02c9f84c01175f7c0260f80c02615000000007ec02c9f98c011a4a1c0261500 0000007e 00000000 c02c9fac c011a78a c0312480 c0278213 c0383244 c02c9fb8 c02d8316 c02ba640 c02c9fc0 c02d8164 c02c9fd0 c02d8a4b 00000000 Call Trace: [<c01175f7>] [<c011a4a1>] [<c011a78a>] [<c0105000>]
Quoted 7 lines
Warning (
Oops_read): Code line not seen, dumping what data is availableTrace;
c01175f7<__might_sleep+27/2b>
Trace;c011a4a1<acquire_console_sem+2d/50>
Trace;c011a78a<register_console+122/1cc>
Trace;c0105000<_stext+0/0>
Don't know. Who called register_console()?
But I suspect in_atomic() is returning incorrect or misleading answers early in boot.
Quoted 4 lines · William Lee Irwin III
Quoted 4 lines
Trace;
c01175f7<__might_sleep+27/2b>
Trace;c011a4a1<acquire_console_sem+2d/50>
Trace;c011a78a<register_console+122/1cc>
Trace;c0105000<_stext+0/0>
Quoted 3 lines · Andrew Morton
Don't know. Who called
register_console()?
But I suspectin_atomic()is returning incorrect or misleading answers early in boot.
I would suspect console_init(). I believe some kind of change was done here for preempt bootstrap ordering issues (vm86_info: BAD) as it's a bit too early to schedule here. Things have gotten interesting down here in other contexts where sleeping and/or waitqueue fiddling is illegal so early on. Getting a better stack dump might be helpful. I'll see if I can do that soon.
Cheers,
Bill