User: Password:
|
|
Subscribe / Log in / New account

The obvious name ...

The obvious name ...

Posted Aug 5, 2009 20:56 UTC (Wed) by felixfix (subscriber, #242)
Parent article: The realtime preemption endgame

Not going to create an account there just for one suggestion. If someone submits this and wins a free beer, enjoy!

The obvious name is raw_spinlock, as opposed to cooked_spinlock. There ie great UNIX tradition in this.


(Log in to post comments)

The obvious name ...

Posted Aug 6, 2009 9:28 UTC (Thu) by eliezert (subscriber, #35757) [Link]

I would call the busywait spinlocks.
This is what they do, and this will also make it obvious they should be used rarely.
I like your cooked / raw proposal, but I think a name the would discourage unnecessary use would be better.

The obvious name ...

Posted Aug 6, 2009 18:03 UTC (Thu) by cpeterso (guest, #305) [Link]

busywait_spinlock is a great name! It accurately describes its behavior and discourages people from using it (without deep thought).

The obvious name ...

Posted Aug 10, 2009 14:44 UTC (Mon) by kjp (subscriber, #39639) [Link]

I thought atomic_spinlock wasn't too bad, but I like busywait_spinlock.

Of course, I do think it would be better to rename the spinlock itself if it no longer does that....

I guess to facilitate the changeover you could have atomic_lock or busywait_lock (spin is a bit redundant) and mutex, and only very old code would need a typdef for the now non existant 'spinlock' type.

The obvious name ...

Posted Aug 7, 2009 0:40 UTC (Fri) by Kamilion (subscriber, #42576) [Link]

Eh, I signed up just to post up "snooze_lock".
Seemed pretty self-explanatory to me. :)

And there's always the alarm-clock connotation, someone always manages to mash the snooze button.

The obvious name ...

Posted Aug 7, 2009 1:08 UTC (Fri) by martinfick (subscriber, #4455) [Link]

But, it's not a snooze lock, it's a spin lock! A snooze lock would imply that a process goes to sleep and yields the processor. A spin lock does not yield, it's in fact the opposite of a "sleep" lock.

If I understand the naming dilemma, the problem is that the old spin_locks which will not be renamed are in fact becoming sleep locks. If your naming convention is adopted, you would have spin_locks which sleep, and snooze_locks which spin!

So, raw_spinlock or real_spinlock seem more appropriate to me.

The obvious name ...

Posted Aug 10, 2009 18:08 UTC (Mon) by kjp (subscriber, #39639) [Link]

It's incredibly sad if the spinlocks wouldn't get renamed. Linux already has incredibly badly named things (struct class anyone?) but now it's also getting incapable of making a mass rename due to what - proprietary or out of tree code?


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds