On Tue, Apr 17, 2012 at 02:30:27PM +0800, Jason Wang wrote:No, I mean patch the guest driver like this:> On 04/17/2012 02:07 PM, Michael S. Tsirkin wrote:Ah so you patched qemu to do this? Cool. Can you post the patch for testing pls?> >On Tue, Apr 17, 2012 at 01:54:55PM +0800, Jason Wang wrote:> > Looks like we can do this by supplying an invalid hdr_len in vnet> >>On 04/17/2012 12:57 PM, Michael S. Tsirkin wrote:> >Yes but do you have a test that makes guest hit EFAULT or EINVAL?> >>>On Tue, Apr 17, 2012 at 11:27:01AM +0800, Jason Wang wrote:> >>Not very hard, w/o this patch, we can see almost 100% cpu> >>>>On 04/16/2012 09:39 PM, Michael S. Tsirkin wrote:> >>>Sounds OK. BTW how do you test this?> >>>>>On Mon, Apr 16, 2012 at 04:28:10PM +0800, Jason Wang wrote:> >>>>Right, so do you think it makes sense that we only restart polling> >>>>>Does same thing happen if we get an error from copy from user?> >>>>>>> On 04/16/2012 03:16 PM, Michael S. Tsirkin wrote:> >>>>>>-EFAULT, vhost> >>>>>>>> >On Mon, Apr 16, 2012 at 02:08:33PM +0800, Jason Wang wrote:> >>>>>>> > This issue is when guest driver is able to hit the> >>>>>>>>> >>Currently, we restart tx polling unconditionally when sendmsg()> >>>>>>>> >Why is this a problem?
> >>>>>>>>> >>fails. This would cause unnecessary wakeups of vhost wokers as it's
> >>>>>>>>> >>only needed when the socket send buffer were exceeded.
> >>>>>>> discard the the descriptor and restart the polling. This would wake
> >>>>>>> vhost thread and repeat the loop again which waste cpu.
> >>>>>
> >>>>on -EAGAIN or -ENOBUFS?
> >>>
> >>utilization for vhost thread if guest hit EFAULT or EINVAL. With
> >>this patch, the cpu utilization should be very low I think.
> header as tap does the check for this.