TCP checksum doesnot work

[Werner Frerichs]

Transfering large data using scp or ssh breaks with the message "Corrupted MAC on input." . This can be a bug in Virtualbox, Hardware, ... or something else. Transfering data using http for examble new packages, updates, ... never breaks because http doesnot check the transport twice (tcp should do this), so the packages are corrupt (wrong md5sum or gunzip cannot open).

I don't know, if the corruption of the transport belongs to a bug in my hardware or in the device driver for the virtual network interface, but there must be a chance to enable TCP checksum to avoid corrupted data.

I found an article/bug report on xensource which describ this better. ​http://lists.xensource.com/archives/html/xen-users/2006-01/msg00741.html

Thanks Werner Frerichs

[Frank Mehnert]

Werner, which host is that, Linux?

[Werner Frerichs]

SLES10 is installed on this host

imapcl1:~ # uname -a Linux imapcl1 2.6.16.60-0.37_f594963d-smp #1 SMP Mon Mar 23 13:39:48 UTC 2009 x86_64 x86_64 x86_64 GNU/Linux

[olivier friard]

Hi,

I noticed the same problem with:

host: Linux Ubuntu 9.06 AMD64 guest: Windows XP pro 32 bits SSH clients: cygwin ssh (openSSH 4.7p1 OpenSSL 0.9.8g)

SSH Secure Shell 3.2.0

Olivier

[Christopher Klein]

Same problem on my machine.

• host is Linux ckl-srv-linux 2.6.28-13-generic #45-Ubuntu SMP Tue Jun 30 19:49:51 UTC 2009 i686

GNU/Linux

• host running OpenSSH 5.1p1
• host network card is nVidia Corporation MCP79 Ethernet (rev b1) (Zotac Ion)
• guest is Windows Server 2003 Std.
• guest network is bridged

I can only access VirtualBox GUI over X11 Forwarding, the message "Corrupted MAC on input" on client occurs randomly. For debugging purposes I opened a second SSH connection to my host. This session did *not* receive the CMoi-message if first connection fails.

It may be pure coincidence but if I use Intel PRO/1000 MT Desktop as network card emulation instead of the network card options the error did *not* occur.

[Christopher Klein]

I have to revert my last sentence. Changing to Intel PRO/1000 MT Desktop did not solve my problem.

While the connection crashes I get the following message in /var/log/messages:

device eth0 left promiscuous mode

[Frank Mehnert]

ckl, which VBox version are you using?

[Christopher Klein]

virtualbox-3.0_3.0.0-49315_Ubuntu_jaunty_i386.deb -> latest

[hajma]

Seeing this in a Solaris 11Express guest on a Solaris11Express host

VirtualBox 4.1.2

[hajma]

Seeing this in a Solaris 10u10 guest. The host is Solaris 11, amd64, VBox 4.1.6. NAT networking.

[vasily Levchenko]

Replying to hajma:

Seeing this in a Solaris 10u10 guest. The host is Solaris 11, amd64, VBox 4.1.6. NAT networking.

Could you please attach the log file? Are there any steps to reproduce with NAT attachment?

[hajma]

log

[hajma]

the config file

[hajma]

to reproduce: try to upload something big using ssh. Is not 100% reproducible. It's likely influenced by system load.

[vasily Levchenko]

Replying to hajma:

to reproduce: try to upload something big using ssh. Is not 100% reproducible. It's likely influenced by system load.

Is it specific for Solaris host/guest or you've seen this issue with other combinations?

[Christopher Klein]

Replying to Hachiman:

Replying to hajma:

to reproduce: try to upload something big using ssh. Is not 100% reproducible. It's likely influenced by system load.

Is it specific for Solaris host/guest or you've seen this issue with other combinations?

Not specific for Solaris host/guest - I noticed this behavior a few days after I had written my ckl.

[hajma]

Replying to Hachiman:

Replying to hajma:

to reproduce: try to upload something big using ssh. Is not 100% reproducible. It's likely influenced by system load.

Is it specific for Solaris host/guest or you've seen this issue with other combinations?

I've seen it on my laptop as well as on a Sun Fire X4170 M2 with a variety of Solaris 10 and Solaris 11 guests

[Frank Mehnert]

Please reopen if still relevant with VBox 4.2.10.