Il 13/08/2012 21:13, Harsh Bora ha scritto: >> > > Hi Paolo, > > The rcu_read_[un]lock() macros were added as no-ops (based on your > inputs on #qemu) to replace reader-writer locks with RCU based locking > as suggested while proposing QemuRWLock API for RW locks (See > http://lists.gnu.org/archive/html/qemu-devel/2011-10/msg00192.html).
The rwlock was also premature optimization IMHO. > v9fs_synth_mutex was actually a pthread_rwlock_t earlier. > I am not sure if reader lock would be better than having a plain mutex > for readers as well. > > Aneesh, inputs ? > > Also, if we are going forward with this change, we may also want to > remove definition of QLIST_INSERT_HEAD_RCU, since the code being remove > below is the only consumer of that macro. It's in a library file, so I don't mind leaving it in. I think liu ping fan's patches were using it. Paolo
