On Sat, 15 Mar 2003, Willem Riede wrote:
> Indeed. If you get there, the command in progress is hung.
> To be able to restart the device, the old command needs to be
> aborted. But that is an inherently racy undertaking.
>
> Nominally, I just want to set HWGROUP(drive)->handler = NULL.
> But there is a small chance, that there is actually (interrupt)
> activity going on for the command, which would result in a new
> entry in HWGROUP(drive)->handler popping up after it is cleared.
>
> The loop as programmed significantly increases the odds that
> the old command is really aborted.
>
> It may not be elegant to schedule(1) with the lock taken, but it
> does work.
>
> However, my latest patch doesn't seem to be applied, since in my
> version I have a set_current_state(TASK_INTERRUPTIBLE); before
> the schedule.
Yeah but what happens when a task tries to acquire ide_lock? Incidentally
this one crept in via timer softirq so if ide_timer_expiry gets there
before your scheduled timeout timer?
NMI Watchdog detected LOCKUP on CPU0, eip c02be73d, registers:
CPU: 0
EIP: 0060:[<c02be73d>] Tainted: PF
EFLAGS: 00000086
EIP is at .text.lock.ide_io+0x40/0x93
Call Trace:
[<c03557e1>] i8042_timer_func+0x21/0x30
[<c02bdcc0>] ide_timer_expiry+0x0/0x310
-- function.linuxpower.ca - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Mar 15 2003 - 22:00:43 EST