On Tue, 3 Jan 2012 22:37:48 +0530, Amit Shah [off-list ref] wrote:

On (Sun) 01 Jan 2012 [13:47:27], Michael S. Tsirkin wrote:

quoted

On Fri, Dec 30, 2011 at 03:08:19PM +0530, Amit Shah wrote:

quoted

quoted

quoted

+	if (!try_fill_recv(vi, GFP_KERNEL))
+		schedule_delayed_work(&vi->refill, 0);

This needs to be switched to non reentrant wq too?

Subsequent invocations could create problems?  Note that this will be
the first item to be queued in the workqueue for the refill work.

		Amit

See recent patches on net-next.
We switched other calls to a non reentrant wq so I think same
reasoning applies there.

There's a patch ordering issue; so this series should land after the
-net series is merged.

Rusty, do you want me to re-spin this patch on top of yours, or is
that something that you'll handle?

Indeed, this can't go via -net, since it needs the virtio-pci patches
which change power management.

I've changed that one line, and will post those patches for DaveM's ack.

Thanks,
Rusty.