On Tue, Jun 04, 2013 at 03:42:19PM +0800, Amos Kong wrote: > On Tue, Jun 04, 2013 at 02:43:11PM +0800, Amos Kong wrote: > > > > > > I did more tests in clear environment, and found that the guest hang/slow > > (no response from monitor) is caused by flooding events. I could not > > reproduce it with upstream qemu [1] > > > > If I set event_throttle to 1 ~ 1000, the problem doesn't occur. > > > > It's easier to reproduce this problem by changing vlan config, > > not because it passes more data with VQ cmd, but it will cause more > > events. > > > > > > In this case, we can set event_throttle to 1 for _RX_FILTER_CHANGED > > event to avoid it slows guest. The 1 ms delay should be acceptabled? > > Just discussed with mst in IRC. > > Here we have two problem: > (1) huge number of events will flood monitor client (management), > (2) emitting huge number of events will slow guest itself. > > Both the flag (nc->rxfilter_notify_enabled) and event_throttle API > can be used to avoid problem (1). > > Event_throttle API can clearly avoid problem (2). > > In real testing, I found it's difficult to reproduce problem (2) if we > already use the flag. It seems response time is larger enough, some > events will be dropped, guest could not be slowed. > > Michael told me that we have many ways to slow guest itself, so it's > not a big issue here. > > We care about the delay of responsing event, so we should only use > control flag (As my patch v4). > > What's your opinion?
Sounds reasonable. Pls send v3 and we'll discuss :)
