On Wed, 28 Jun 2006 15:09:43 +0200 Paolo Ornati wrote: > On Wed, 28 Jun 2006 14:53:49 +0200 > Paolo Ornati wrote: > > > > [ 430.083347] localedef R running task 0 8577 8558 8578 (NOTLB) > > > [ 430.083352] gzip X ffff81001e612ee0 0 8578 8577 (L-TLB) > > > [ 430.083358] ffff81001395bef8 ffff81001fd1a310 0000000000000246 ffff81001e612ee0 > > > [ 430.083362] ffff81001e4c0080 ffff81001e612ee0 ffff81001e4c0258 0000000000000001 > > > [ 430.083366] 0000000000000046 0000000000000046 ffff81001395bf18 0000000000000010 > > > [ 430.083370] Call Trace: {do_exit+2378} {vfs_write+288} > > > [ 430.083379] {sys_exit_group+0} {system_call+126} > > > > do_exit() -- kernel/exit.c > > > > 0xffffffff80227f66 : mov %rax,0x18(%rbp) > > 0xffffffff80227f6a : callq 0xffffffff8048b850 > > 0xffffffff80227f6f : ud2a > > 0xffffffff80227f71 : pushq $0xffffffff804b7821 > > 0xffffffff80227f76 : retq $0x3ba > > 0xffffffff80227f79 : jmp 0xffffffff80227f79 > > > that is the end of the function: > > ... > schedule(); > BUG(); > /* Avoid "noreturn function does return". */ > for (;;) ; > } > > > So the questions are two: > > 1) why schedule() didn't work? > > 2) why the process is looping around "ud2a" (placed by BUG()) and > presumibly throwing a lot of "invalid opcode" exceptions? It seems that the first mail didn't hit LKML (maybe it was too big), here it is (stripped): ------------------------------------------------------------------------ Running "localedef" triggers an infinite loop in kernel mode (or something) --> localdef becomes unkillable. This is dmesg after pressing ALT+SysRQ+T three times: [CUT] KERNEL: -- Paolo Ornati Linux 2.6.17-ga39727f2-dirty on x86_64 ^^^^^^^ The -dirty is because of a fix to PS2 Keyboard auto-repeat, nothing else. ------------------------------------------------------------------------ Full config attached.