Re: test6-lowlatency-D1 results: 50msec + hangs ....

[Philipp Rumpf]

On Sat, Aug 05, 2000 at 05:52:26PM -0700, David S. Miller wrote:
>    Date: 	Sun, 6 Aug 2000 01:43:04 +0100
>    From: Philipp Rumpf <[e-mail]>
> 
>    That's bogus on UP (spin_is_locked always returns 0).
> 
>    I think it would be reasonably nice to have something like
> 
>    spin_lock_held(spinlock_t *lock);
> 
>    that returns 1 for UP and spin_is_locked(lock) for SMP - basically 
> 
> Really, whatever you name it, the question of a lock being held or not
> on UP has zero meaning.  There is no correct return value, period.

Uhm, I disagree.  "Am I holding this lock" is always true on UP.  "Is
someone else holding this lock" is always false.  "Am I holding this lock"
also isn't the same as spin_is_locked on SMP since you want to look at
who's holding it when debugging is enabled.  Same goes for "Is someone
else holding this lock".

> No matter what constant answer you decide to give on UP it will always
> be incorrect.  I know, because there were old bits of assertions in

You're right.  But no-one actually cares about whether a lock is held -
they care about whether they're holding it or someone else is holding it
or no-one's holding it.

	Philipp