Date: Sat, 16 Apr 2011 15:49:10 -0500 From: Alex Anagnos <woodshop2300@gmail.com> To: Julian Elischer <julian@freebsd.org> Cc: freebsd-emulation@freebsd.org Subject: Re: Networking frame loss in transmission Message-ID: <608E001D-14AF-4025-80D0-58369BE95773@gmail.com> In-Reply-To: <4DA9FBCE.2060106@freebsd.org> References: <20110416120030.457A510656DF@hub.freebsd.org> <4DA9FBCE.2060106@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
I sometime see arps int the beginning of the captures. They are followed by the the frames, sometimes the error happens = sometime not. Other times i see arps in the middle of the stream, with nothing wrong. I'm running this test with short 20-30 second gaps between them (to = reset captures).. So the arp table is not even expiring between runs. So on shorter run, i = don't see arps at all, and the problem still sometimes occurs.. I have not tried a FreeBSD guest, but i don't think the guest is the = issue here. I'v taken the same linux Guest, and run it temporally on a Windows host, = and the problem never manifests. On Apr 16, 2011, at 3:27 PM, Julian Elischer wrote: >=20 >> Message: 2 >> Date: Fri, 15 Apr 2011 18:57:19 -0500 >> From: Alex Anagnos<woodshop2300@gmail.com> >> Subject: Networking frame loss in transmission >> To: freebsd-emulation@freebsd.org >> Message-ID:<BANLkTik92Eg0hrHsaK0J8yh3vibuyRDZew@mail.gmail.com> >> Content-Type: text/plain; charset=3DISO-8859-1 >>=20 >> FreeBSD 8.0-RELEASE w/ VirtualBox 3.2.12. >>=20 >> The setup is a small application in a Guest OS that is sending UDP >> Ethernet frames. >> I Am capturing on FreeBSD on the device with both Wireshark, and a >> custom written program that uses BPF zero copy. >>=20 >> What i'm seeing is that when the application send data that is beyond >> the MTU the data is fragmented. >> However, sometimes, the first few fragments never show up in the >> capture, and indeed the receiving app never gets the message >> (reassembly fails). >=20 > maybe it was waiting for arp? >> So when I call send with an 8192 payload, i see, frames 1 and 2 go >> missing, but 3, 4, and , 5 get out fine. (1500 MTU) > do you see an arp reply wbetween 2 and 3? >=20 >> When calling send 2 times in rapid succession, all fragments of the >> first send go missing as well as 1, and 2 of the 2nd send. 3, 4, and = 5 >> get out fine. >> When calling send 3 times in rapid succession, all fragments of send = 1 >> & 2 get lost as well as fragment 1& 2 of the 3rd send. Frags 3, 4, >> and 5 are fine. >=20 > makes arp less likely but not impossible. >> When sending this at like 1 Message per second, things occasionally = don't work. >> When going at 3000FPS, it has so far always been the first few = frames, >> after that the rest of the stream gets through ok almost like the >> driver has to warm up or something.. >>=20 >> The issue happens sending from both a Linux guest and from a Windows = Guest. >> The issue does not care about the speed of the traffic, it happens at >> 1FPS and at 3000FPS. > have you tried a freebsd guest? at least we can debug that.. :-) >=20 >> The issue does seam to occur less often the smaller the message sent. >> When the message does not require fragmentation I have not seen the >> issue happen but i'm not going to guarantee it does not happen. >>=20 >> The problem happens using Bridged, VDE, and Host-Only. >> The problem happens using the FastPCI driver, intel e1000, and = para-virtualized. >> The problem happens with any number of the above 3 combos. >>=20 >> The problem happens in Version 3.2.6 and inversion 3.2.12. >>=20 >> The problem looks to be somewhere in the VirtualBox box code, or = maybe >> in the VirtualBox kernel module integration with FreeBSD >=20 > _______________________________________________ > freebsd-emulation@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-emulation > To unsubscribe, send any mail to = "freebsd-emulation-unsubscribe@freebsd.org"
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?608E001D-14AF-4025-80D0-58369BE95773>