On 2010-08-18 at 19:05:55 [+0200], Stephan Assmus <superstippi@xxxxxx> wrote: > Am 18.08.2010 18:26, schrieb Ingo Weinhold: > > On 2010-08-18 at 08:52:28 [+0200], Stephan Assmus<superstippi@xxxxxx> > > wrote: > >> MultiLocker supports nested write-locks, I don't know > >> what happens when the current write-lock-owner acquires a read-lock. The > >> (read-lock) assert that Axel mentions may indeed be the best solution. > > > > BTW, since there's rw_lock functionality in libroot > > (headers/private/shared/locks.h), MultiLocker could be discarded. > > Thanks for the reminder, it could be more efficient, too, perhaps? Probably not. Synchronizing with the current kernel version could give it an edge, since read locking/unlocking could be inlined, but other than that the differences are probably marginal. It wouldn't use any semaphores anymore, though. CU, Ingo