On 07/19/2011 07:22 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange"<berrange@xxxxxxxxxx>
Remove the need for a virNetSocket object to be protected by
locks from the object using it, by introducing its own native
locking and reference counting
* src/rpc/virnetsocket.c: Add locking& reference counting
---
src/rpc/virnetsocket.c | 147 +++++++++++++++++++++++++++++++++++++++---------
1 files changed, 120 insertions(+), 27 deletions(-)
diff --git a/src/rpc/virnetsocket.c b/src/rpc/virnetsocket.c
index 7ea1ab7..8dd4d3a 100644
--- a/src/rpc/virnetsocket.c
+++ b/src/rpc/virnetsocket.c
@@ -40,6 +40,7 @@
#include "logging.h"
#include "files.h"
#include "event.h"
+#include "threads.h"
#define VIR_FROM_THIS VIR_FROM_RPC
@@ -49,6 +50,9 @@
struct _virNetSocket {
+ virMutex lock;
+ int refs;
Someday we should revive the virObject patches, for lighter-weight
atomic-op reference counting. But that shouldn't hold up this patch.
int virNetSocketAddIOCallback(virNetSocketPtr sock,
int events,
virNetSocketIOFunc func,
void *opaque)
{
+ int ret = -1;
+
+ virMutexLock(&sock->lock);
if (sock->watch> 0) {
VIR_DEBUG("Watch already registered on socket %p", sock);
- return -1;
+ goto cleanup;
}
+ sock->refs++;
This increases the ref-count on registration, but...
void virNetSocketRemoveIOCallback(virNetSocketPtr sock)
{
+ virMutexLock(&sock->lock);
+
if (sock->watch<= 0) {
VIR_DEBUG("Watch not registered on socket %p", sock);
+ virMutexUnlock(&sock->lock);
return;
}
virEventRemoveHandle(sock->watch);
- sock->watch = 0;
+
+ virMutexUnlock(&sock->lock);
}
...this doesn't decrease it. Am I missing something?
Once that question is answered, then ACK to the rest of the patch.
--
Eric Blake eblake@xxxxxxxxxx +1-801-349-2682
Libvirt virtualization library http://libvirt.org
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list