Packet corruption in re0

[Christoph Schug]

On Thu, Feb 21, 2008, Pyun YongHyeon wrote:

> Copy sys/dev/re/if_re.c, sys/pci/if_rlreg.h in HEAD to your box.
> Due to lack of m_defrag(9) in 7-PRERELEASE/RC, you also have to add
> that function to if_re.c(Copy m_defrag() in sys/kern/uipc_mbuf.c on
> HEAD/RELENG_7 to if_re.c). That would make it build on your box.

I am having the very same problem on a rented Hetzner DS8000 root
server offering [1]. According to sysutils/dmidecode from the ports the
main board is a MSI K9AG Neo2-Digital (MS-7368) [2] with an embedded
RTL8168/8111 PCI-E Gigabit Ethernet NIC (re0 at pci0:2:0:0: class=0x020000
card=0x368c1462 chip=0x816810ec rev=0x01 hdr=0x00).

[1] http://www.hetzner.de/rootserver.html
[2] http://www.msicomputer.com/product/p_spec.asp?model=K9AG_Neo2-Digital&class=mb

>  > Yes, I've been down the heat path, but the problem takes ~36 hours to recur
>  > both if the system is turned on from cold, or simply rebooted. Mind you,
>  > North Queensland could do with some cooling! I will take another look at the
>  > circuitry if moving to HEAD's re doesn't fix it.

Interestingly enough, in my case it takes a similar amount of time but
I can speed up the problem to show up when doing many disk (!) I/O
operations on the 3ware RAID controller. Furthermore calling programs
like 'rpm -Uvh' which draw a progress bar by sending many hash ('#')
characters in a short period of time are very "helpful" to freeze
the session while other SSH sessions during the same time still work
flawlessly. 'netstat -an' shows pretty high numbers in the send queue of
the freezed session.

Will give re(4) of HEAD a try despite the fact that I would be very
happy to see a fix in RELENG_7 as I have to bring to machine into
production very soon. Thanks for the hint!

-cs