Jiri Kosina wrote:
> On Tue, 3 Jul 2007, Felipe Balbi wrote:
>
>> So looks like we don't have (yet) a way to make it work nicely in both
>> cases... does anyone has a clue about how to implement this one??
>
> You of course still can
>
> local_irq_save(flags);
> spin_lock(&some_lock);
> ...
> spin_unlock(&some_lock);
> usb_hcd_giveback_urb(hcd, urb);
> local_irq_restore(flags);
>
> which is safe even against future removal of preempt inc/dec from
> spin_lock_irq{save,restore}() functions.
>
Unfortunately that would not work in the RT case either. Because in RT,
that spin_lock can schedule, and we are not allowed to schedule with
interrupts disabled.
What we could do that would work in all cases (but is butt-ugly) is the
following:
{
DEFINE_SPINLOCK(local);
unsigned long flags;
spin_lock_irqsave(&local, flags);
spin_lock(&some_lock);
...
spin_unlock(&some_lock);
usb_hcb_giveback_urb(hcd, urb);
spin_unlock_irqrestore(&local, flags);
}
Because the "local" lock is on the stack, it will never be in
contention. Which means that it basically acts like a local_irq_save.
The benefit is that it also works in RT, because that case the
interrupts would *not* be disabled in that entire path. For those that
don't know, RT has interrupts running as threads, so the need here to
disable interrupts is gone.
I can send a note to Ingo, and see if he has better ideas. Perhaps
there's a way that we can do a local_irq_restore that matches spinlocks.
Maybe we can add a primitive of
local_irqs_restore_from_spinlock(flags);
This would document that the local_irq_save came from a spinlock as it
will allow things like this to seamlessly change for RT.
(/me decides to just add Ingo on this thread)
-- Steve
-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
[email protected]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
