On Thu, Jun 23, 2011 at 05:50:27PM -0600, Eric Blake wrote:
On 06/22/2011 09:33 AM, Daniel P. Berrange wrote:
+++ b/src/Makefile.am @@ -1187,7 +1187,7 @@ else
+libvirt_net_rpc_server_la_SOURCES = \ + rpc/virnetserverprogram.h rpc/virnetserverprogram.c \ + rpc/virnetserverservice.h rpc/virnetserverservice.c \ + rpc/virnetserverclient.h rpc/virnetserverclient.c \ + rpc/virnetserver.h rpc/virnetserver.c +libvirt_net_rpc_server_la_CFLAGS = \ + $(AM_CFLAGS) +libvirt_net_rpc_server_la_LDFLAGS = \ + $(AM_LDFLAGS) \ + $(CYGWIN_EXTRA_LDFLAGS) \ + $(MINGW_EXTRA_LDFLAGS)l
Umm, wonder why that spurious l isn't causing us grief?
I expect libtool discards it for some reason
+ +virNetServerPtr virNetServerNew(size_t min_workers, + size_t max_workers, + size_t max_clients, + virNetServerClientInitHook clientInitHook) +{ + virNetServerPtr srv; + struct sigaction sig_action; + + if (VIR_ALLOC(srv) < 0) { + virReportOOMError(); + return NULL; + } + + srv->refs = 1; + + if (!(srv->workers = virThreadPoolNew(min_workers, max_workers, + virNetServerHandleJob, + srv))) + goto error; + + srv->nclients_max = max_clients; + srv->sigwrite = srv->sigread = -1; + srv->clientInitHook = clientInitHook; + srv->privileged = geteuid() == 0 ? true : false;
I'd have gone with the shorter:
src->privileged = !geteuid();
I find the explicit conditional a little bit more immediately obvious to read.
+ +bool virNetServerIsPrivileged(virNetServerPtr srv) +{ + bool priv; + virNetServerLock(srv); + priv = srv->privileged; + virNetServerUnlock(srv); + return priv;
Does this really need to obtain a lock, since srv->privileged is never changed after construction?
I guess not, but I guess I didn't want to special case this method
+ +static int virNetServerSignalSetup(virNetServerPtr srv) +{ + int fds[2]; + + if (srv->sigwrite != -1) + return 0; + + if (pipe(fds) < 0) {
If you use pipe2(fds, O_CLOEXEC|O_NONBLOCK),
Ah good idea.
+ + /* Count of messages in the 'tx' queue, + * and the server worker pool queue + * ie RPC calls in progress. Does not count + * async events which are not used for + * throttling calculations */ + size_t nrequests; + size_t nrequests_max; + /* Zero or one messages being received. Zero if + * nrequests >= max_clients and throttling */ + virNetMessagePtr rx; + /* Zero or many messages waiting for transmit + * back to client, including async events */ + virNetMessagePtr tx; + + /* Filters to capture messages that would otherwise + * end up on the 'dx' queue */ + virNetServerClientFilterPtr filters;
'dx' queue? Did you mean 'rx'?
No, messages are on the 'rx' queue while data is being read. When a complete message has been read the message is moved from 'rx' to 'dx' queue for processing by a worker. The filter causes the message to go elsewhere, instead of the 'dx' queue. Regards, Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|