lockmgr_sleep() (was Re: ssleep() (was Re: mention msleep() in porting_drivers.txt))

Peter Avalos pavalos at theshell.com
Sun Feb 10 09:27:30 PST 2008


Simon 'corecode' Schubert wrote:
Aggelos Economopoulos wrote:
+ * Atomically drop a lockmgr lock and go to sleep. The lock is 
reacquired
+ * before returning from this function. Passes on the value returned by
+ * tsleep().
+ */
+int
+lock_sleep(void *ident, int flags, const char *wmesg, int timo,
+       struct lock *lk)
I like it, but I'd put lock_sleep in kern_lock.c and wouldn't make 
lockstatus_owned public.
I like it too. Having the lock as the first or last arg makes much more 
sense than having it in the middle.  It won't be confusing, especially 
when we document it in a manual page and in porting_drivers.txt.

+10, should go in now.  However then we'd ship a release with a lot of 
deprecation warnings.  Maybe we should follow up with a 
msleep->spin_sleep commit.

+1

--Peter





More information about the Submit mailing list