On Thu, Nov 29, 2012 at 4:47 PM, Paolo Bonzini pbonz...@redhat.com wrote:
Stefan, perhaps you could replace the stop_notifier mechanism of
event-poll.c with something similar to aio_notify/qemu_notify_event,
and even remove event_poll_run in favor of event_poll (aka aio_wait...).
And also
On Thu, Nov 22, 2012 at 04:16:51PM +0100, Stefan Hajnoczi wrote:
virtio-blk-data-plane is a subset implementation of virtio-blk. It only
handles read, write, and flush requests. It does this using a dedicated
thread that executes an epoll(2)-based event loop and processes I/O
using Linux
On Thu, Nov 22, 2012 at 04:16:51PM +0100, Stefan Hajnoczi wrote:
virtio-blk-data-plane is a subset implementation of virtio-blk. It only
handles read, write, and flush requests. It does this using a dedicated
thread that executes an epoll(2)-based event loop and processes I/O
using Linux
+unsigned int num_reqs;
+QemuMutex num_reqs_lock;
OK the only reason this lock is needed is because
you want to drain outside the thread.
Won't it be better to queue process the drain request through
the thread?
You won't need any locks then.
Draining is processed in the
On Thu, Nov 29, 2012 at 10:21:31AM -0500, Paolo Bonzini wrote:
+unsigned int num_reqs;
+QemuMutex num_reqs_lock;
OK the only reason this lock is needed is because
you want to drain outside the thread.
Won't it be better to queue process the drain request through
the
- Messaggio originale -
Da: Michael S. Tsirkin m...@redhat.com
A: Paolo Bonzini pbonz...@redhat.com
Cc: qemu-devel@nongnu.org, Kevin Wolf kw...@redhat.com, Anthony Liguori
aligu...@us.ibm.com, Blue Swirl
blauwir...@gmail.com, k...@us.ibm.com, Asias He as...@redhat.com,
Stefan
virtio-blk-data-plane is a subset implementation of virtio-blk. It only
handles read, write, and flush requests. It does this using a dedicated
thread that executes an epoll(2)-based event loop and processes I/O
using Linux AIO.
This approach performs very well but can be used for raw image
