On Wed, Oct 28, 2009 at 05:15:05PM +0100, Daniel Veillard wrote:
On Fri, Oct 23, 2009 at 02:05:31PM +0100, Daniel P. Berrange wrote:
This implements a thin wrapper around the pthread_rwlock primitives. No impl is provided for Win32 at this time since it is rather hard, and none of our code yet requires it on Win32
* src/util/threads.h: Add virRWLockInit, virRWLockDestroy, virRWLockRead, virRWLockWrite, virRWLockUnlock APIs * src/util/threads-pthread.h: define virRWLock struct * src/util/threads-pthread.c: Implement RWLock APIs --- src/libvirt_private.syms | 6 ++++++ src/util/threads-pthread.c | 30 ++++++++++++++++++++++++++++++ src/util/threads-pthread.h | 4 ++++ src/util/threads.h | 10 ++++++++++ 4 files changed, 50 insertions(+), 0 deletions(-)
diff --git a/src/libvirt_private.syms b/src/libvirt_private.syms index bd9d84a..6ed562d 100644 --- a/src/libvirt_private.syms +++ b/src/libvirt_private.syms @@ -426,6 +426,12 @@ virCondWait; virCondSignal; virCondBroadcast;
+virRWLockInit; +virRWLockDestroy; +virRWLockRead; +virRWLockWrite; +virRWLockUnlock; + # util.h virFileReadAll; virFileWriteStr; diff --git a/src/util/threads-pthread.c b/src/util/threads-pthread.c index 4e00bc5..2052c0a 100644 --- a/src/util/threads-pthread.c +++ b/src/util/threads-pthread.c @@ -57,6 +57,36 @@ void virMutexUnlock(virMutexPtr m) }
+int virRWLockInit(virRWLockPtr m) +{ + if (pthread_rwlock_init(&m->lock, NULL) != 0) { + errno = EINVAL; + return -1; + } + return 0; +} + +void virRWLockDestroy(virRWLockPtr m) +{ + pthread_rwlock_destroy(&m->lock); +} + +void virRWLockRead(virRWLockPtr m) +{ + pthread_rwlock_rdlock(&m->lock); +} + +void virRWLockWrite(virRWLockPtr m) +{ + pthread_rwlock_wrlock(&m->lock); +} + +void virRWLockUnlock(virRWLockPtr m) +{ + pthread_rwlock_unlock(&m->lock); +} + +
Hum it's a small patch, but I would rather fix those function to not be void when the underlying pthread_ counterpart can actually fail. IMHO we should report errors coming from the thread library (whichever is used on a given platform).
I mis-understood the manpage for this - I read it to mean these calls would actually deadlock upon error (as a mutex would), but in fact they return error EDEADLOCK instead which is rather unhelpful. I'm going to withdraw this entire patch. Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|