Eric, Thanks for the explanation. What I don't understand is this statement:
> The clock thread > can block (for example, in trying to grab a lock that's already held), > but that's different from being preempted If the clock thread is blocked, it will go to sleep and relinquish the CPU -- and that necessarily implies preemption. What am I missing? Thanks. -- Sasha On Tue, 5 Aug 2008, Eric Saxe wrote: > Alexandra (Sasha) Fedorova wrote: >> I am writing some code that will be called from the function ts_tick(), >> which is called by clock_tick(). Comments in ts_tick() say that when one >> releases a thread lock s/he should use "thread_unlock_nopreempt()" since >> "clock thread cannot be pre-empted". (See >> http://cvs.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/common/disp/ts.c, >> >> line 1781). >> My question is, if my code is called from ts_tick(), can I not acquire any >> adaptive mutexes? I thought that I cannot use MUTEX_ADAPTIVE >> (MUTEX_DEFAULT) mutex, because it might block, and, according to those >> comments, the thread executing clock_tick cannot be blocked. > Adaptive mutexes are aquired in the clock thread's context. The clock thread > can block (for example, in trying to grab a lock that's already held), but > that's different from being preempted (which could happen when dropping a > thread lock (see the implementation of disp_lock_exit()). > >> So I was going to use MUTEX_SPIN. And then, I went through >> clock_tick_process(), a function that calls clock_tick() (see >> http://cvs.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/common/os/clock_tick.c) >> >> and saw that it acquires an adaptive mutex: >> >> 367 plockp = t->t_plockp; >> 368 mutex_enter(plockp); >> > Right. > >> I haven't found any place where plockp (or the p_lock mutex to which it >> points) was initialized, so I assume it's adaptive. > The thread is referencing the p_lock in the process structure (proc_t). I > don't see any explicit initialization of that lock (other than being > bzero()ed)....but yes, it's adaptive. > >> Could you please shed some light on this issue? Is it safe to use an >> adaptive mutex in the code called from clock_tick? > Yes. It's safe...you probably just want to keep in mind the possibility that > clock might back up, if the mutexes it needs to get it's job done are > contended....which in general can be "painful" for the system. :) > > Thanks, > -Eric > _______________________________________________ opensolaris-code mailing list opensolaris-code@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
