kern/104056: VMware ESX 3.0: lnc0: Missed packet -- no receive buffer

[Astrodog]

The following reply was made to PR kern/104056; it has been noted by GNATS.

From: Astrodog <astrodog at gmail.com>
To: "Jonas Nagel" <fireball at zerouptime.ch>
Cc: bug-followup at freebsd.org
Subject: Re: kern/104056: VMware ESX 3.0: lnc0: Missed packet -- no receive buffer
Date: Mon, 9 Oct 2006 00:44:07 -0500

 On 10/8/06, Jonas Nagel <fireball at zerouptime.ch> wrote:
 > Never mind my last mail; I was able to install the tools through the
 > ports. I used the vmware-guestd5 port, since I don't run X on that VM.
 >
 > It wouldn't install the vmxnet.ko though by default.
 >
 > There is also only one precompiled for FreeBSD 5.3 on the ISO; I copied
 > it over and it loaded fine.
 >
 > dns1# ll vmxnet.ko
 > -r-xr-xr-x  1 root  wheel  12837 Oct  8 23:59 vmxnet.ko
 > dns1# kldstat
 > Id Refs Address    Size     Name
 >  1    5 0xc0400000 6f0e44   kernel
 >  2    1 0xc0af1000 59f00    acpi.ko
 >  3    1 0xc5f0e000 3000     vmmemctl.ko
 >  4    1 0xc5f11000 4000     vmxnet.ko
 >
 > dns1# ps ax |grep vmware
 > 85086  ??  Ss     0:01.93 /usr/local/sbin/vmware-guestd --background /var/run/vmware_guestd.pid --halt-command
 >
 > I even put the said vmxnet.ko into /boot/kernel and loaded it using
 > 'vmxnet_load="YES"'. But this all didn't have any influence over network
 > performance.
 >
 > Specially the connection to other virtual machines is less than
 > acceptable. Also note the interesting pattern:
 >
 > dns1# ping -s 16000 hirtdco01
 > PING hirtdco01.hirtdom.local (10.0.0.90): 16000 data bytes
 > 16008 bytes from 10.0.0.90: icmp_seq=0 ttl=128 time=2.148 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=1 ttl=128 time=1.854 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=2 ttl=128 time=878.200 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=3 ttl=128 time=1.963 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=4 ttl=128 time=1007.225 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=5 ttl=128 time=2.003 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=6 ttl=128 time=1007.256 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=7 ttl=128 time=167.444 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=8 ttl=128 time=2.328 ms
 > 16008 bytes from 10.0.0.90: icmp_seq=9 ttl=128 time=1.985 ms
 > (...)
 > --- hirtdco01.hirtdom.local ping statistics ---
 > 88 packets transmitted, 81 packets received, 7% packet loss
 > round-trip min/avg/max/stddev = 1.581/298.857/1010.492/417.300 ms
 >
 > It goes from 1-2 ms up to 1 second and down again or, well, results into
 > timeouts.
 >
 > Any other idea (besides blaming it to VMware)?
 >
 
 What kind of system load is there on that? I've seen FreeBSD under
 VMWare get kinda messed up if the other VMs are heavily loaded.