2.6.0-test7: Preempt enabled -> kernel panic
When compiling 2.6.0-test7 with preempt I get a kernel panic when running the tst-eintr1 test program from the nptl 0.60 package. It does not happen every time but running it repeatedly will lead to a panic pretty quickly.
With preempt disabled it's rock solid.
Stack trace is not always the same but there always seems to be infinite recursion. Also sometimes interrupts are disabled (no SysRq) and sometimes not.
This is on an Athlon XP, kernel compiled with gcc 3.3.1.
Example stack trace:
...
do_page_fault+0x12c/0x454 poke_blanked_console+0x5c/0x70 try_to_wake_up+0xa7/0x160 default_wake_function+0x2a/0x30 __wake_up_common+0x31/0x60 do_page_fault+0x0/0x454 error_code+0x2d/0x38 do_exit+0x1d2/0x350 do_page_fault+0x0/0x454 die+0xe1/0xf0 do_page_fault+0x12c/0x454 sys_exit+0x13/0x20 syscall_call+0x7/0xb Code: 8b 5d 68 c7 44 24 20 01 00 03 00 8b 50 14 8b 00 81 e2 ff ff
<6>note: ld-linux.so.2[2069] exited with preempt_count 1
--
Markus
In article <[email protected]>, Markus Schoder <[email protected]> wrote:
| When compiling 2.6.0-test7 with preempt I get a kernel panic | when running the tst-eintr1 test program from the nptl 0.60 package. | It does not happen every time but running it repeatedly will lead | to a panic pretty quickly. | | With preempt disabled it's rock solid. | | Stack trace is not always the same but there always seems to | be infinite recursion. Also sometimes interrupts are disabled | (no SysRq) and sometimes not.
I usually run 2.6 kernels with both softdog and NMI watchdog, have no idea if that would help other than to possibly force a reboot.
| This is on an Athlon XP, kernel compiled with gcc 3.3.1.
Check the list, there was something about gcc version, don't remember exactly what.
Signature · 3 lines
bill davidsen <[email protected]> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979.