On Mon, Nov 01, 2010 at 11:50:01AM -0600, Eric Blake wrote:
On 11/01/2010 10:11 AM, Daniel P. Berrange wrote:
The current remote driver code for streams only supports blocking I/O mode. This is fine for the usage with migration but is a problem for more general use cases, in particular bi-directional streams.
This adds supported for the stream callbacks and non-blocking I/O. with the minor caveat is that it doesn't actually do non-blocking I/O for sending stream data, only receiving it. A future patch will try to do non-blocking sends, but this is quite tricky to get right.
* src/remote/remote_driver.c: Allow non-blocking I/O for streams and support callbacks
+ +static void +remoteStreamEventTimer(int timer ATTRIBUTE_UNUSED, void *opaque) +{ + virStreamPtr st = opaque; + struct private_data *priv = st->conn->privateData; + struct private_stream_data *privst = st->privateData; + + remoteDriverLock(priv); + if (privst->cb && + (privst->cbEvents & VIR_STREAM_EVENT_READABLE) && + privst->incomingOffset) { + virStreamEventCallback cb = privst->cb; + void *cbOpaque = privst->cbOpaque; + virFreeCallback cbFree = privst->cbFree; + + privst->cbDispatch = 1; + remoteDriverUnlock(priv); + (cb)(st, VIR_STREAM_EVENT_READABLE, cbOpaque);
Any reason you aren't using the simpler style?
cp(st, ...);
I prefer the '(cb)(st...)' style, because it makes it clearer that 'cb' is not a function name, but a function pointer. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://deltacloud.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|