[2.6.35-rc3] reboot mutex 'bug'...

[2.6.35-rc3] reboot mutex 'bug'...

[Daniel J Blueman]

When halting 2.6.35-rc3 compiled on ARM with debugging, the kernel
warns [1] about the reboot_mutex being taken at kernel/sys.c:392 [2]
and held across do_exit() since kernel_power_off() doesn't necessarily
stop execution.

Does it make sense to annotate the reboot_mutex lock or is another
approach suggested?

Daniel

--- [1]

=====================================
[ BUG: lock held at task exit time! ]
-------------------------------------
halt/3874 is exiting with locks still held!
1 lock held by halt/3874:
 #0:  (reboot_mutex){+.+...}, at: [<c0070424>] sys_reboot+0x98/0x1c4

stack backtrace:
[<c003ab58>] (unwind_backtrace+0x0/0xe8) from [<c04ba3f0>]
(dump_stack+0x18/0x1c)
[<c04ba3f0>] (dump_stack+0x18/0x1c) from [<c008aea8>]
(debug_check_no_locks_held+0x74/)
[<c008aea8>] (debug_check_no_locks_held+0x74/0x90) from [<c0063dfc>]
(do_exit+0x570/0x)
[<c0063dfc>] (do_exit+0x570/0x5ec) from [<c00704a8>] (sys_reboot+0x11c/0x1c4)
[<c00704a8>] (sys_reboot+0x11c/0x1c4) from [<c0036040>]
(ret_fast_syscall+0x0/0x3c)

--- [2] kernel/sys.c

SYSCALL_DEFINE4(reboot, int, magic1, int, magic2, unsigned int, cmd,
                void __user *, arg)
{
...
        mutex_lock(&reboot_mutex);
...
        case LINUX_REBOOT_CMD_POWER_OFF:
                kernel_power_off();
                do_exit(0);
                break;
...
        mutex_unlock(&reboot_mutex);
...
}
-- 
Daniel J Blueman