Date: Mon, 22 Sep 2014 12:57:14 +0000 From: bugzilla-noreply@freebsd.org To: freebsd-bugs@FreeBSD.org Subject: [Bug 193834] New: Starting up Vbox instance slows down existing tcp flows on igb nic (tso related?) Message-ID: <bug-193834-8@https.bugs.freebsd.org/bugzilla/>
next in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=193834 Bug ID: 193834 Summary: Starting up Vbox instance slows down existing tcp flows on igb nic (tso related?) Product: Base System Version: 10.1-BETA1 Hardware: amd64 OS: Any Status: Needs Triage Severity: Affects Many People Priority: --- Component: kern Assignee: freebsd-bugs@FreeBSD.org Reporter: mike@sentex.net I noticed on the weekend our offsite rsync virtually stopped. Upon investigation I found that I had started a vbox guest on the machine and this caused the rsync ssh flow to virtually stall. Stopping and restart the rsync process fixed the issue. Its easy to re-create. start a large tcp stream operation eg. pipe an ssh stream to another machine fire up a FreeBSD guest in vbox where the guest bridges the nic. The ssh stream will virtually stall. Another way to fix the issue is to shutdown the vbox guest. The traffic speeds will resume to normal. 1 37 0xffffffff80200000 1141750 kernel 2 1 0xffffffff81342000 4a10 coretemp.ko 3 1 0xffffffff81411000 169e64 zfs.ko 4 1 0xffffffff8157b000 397f opensolaris.ko 5 1 0xffffffff8157f000 3582 ums.ko 6 1 0xffffffff81583000 2b5c uhid.ko 7 1 0xffffffff81586000 176f accf_http.ko 8 1 0xffffffff81588000 cbf accf_data.ko 9 1 0xffffffff81589000 2e57 aesni.ko 10 1 0xffffffff8158c000 1f56c crypto.ko 11 3 0xffffffff815ac000 32710 vboxdrv.ko 12 1 0xffffffff815df000 3ec0 vboxnetadp.ko 13 1 0xffffffff815e3000 28c0 vboxnetflt.ko 14 2 0xffffffff815e6000 b998 netgraph.ko 15 1 0xffffffff815f2000 40b2 ng_ether.ko Another way to do it is to start the ssh stream, pause and then unpause the guest. When the guest restarts, the ssh stream will drop to almost zero. The bug seems to have something to do with tso. If I disable tso on the NIC, this behaviour does not happen. Intel MB. dmesg, dmidecode etc attached # VBoxManage --version 4.3.12_OSEr93733 -- You are receiving this mail because: You are the assignee for the bug.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-193834-8>