IP header checksum missing with Realtek 8168, jumbo frames and offloading.

[Arnaud Houdelette]

Pyun YongHyeon a écrit :
> On Sun, Mar 02, 2008 at 11:44:03PM +0100, Arnaud Houdelette wrote:
>  > I encountered connectivity issues with an integrated Realtek 8168 on my 
>  > MSI motherboard after enabling jumbo frames on my other box. 
>  > Investigating the issue, I found that the packets with an ethernet frame 
>  > of length > 2048 get an IP header of 0x0000.
>  > ping -s 3000 192.168.0.11  ==> fail (ethereal on the other box show the 
>  > 0x0000 checksum on IP header)
>  > ping -s 2008 192.168.0.11  ==> fail
>  > ping -s 2006 192.168.0.11  ==> succeed
>  > 
>  > 
>  > re0: <RealTek 8168/8111B PCIe Gigabit Ethernet> port 0xd800-0xd8ff mem 
>  > 0xfeaff000-0xfeafffff irq 19 at device 0.0 on pci2
>  > re0: Using 2 MSI messages
>  > miibus0: <MII bus> on re0
>  > rgephy0: <RTL8169S/8110S/8211B media interface> PHY 1 on miibus0
>  > rgephy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 
>  > 1000baseT-FDX, auto
>  > 
>  > 
>  > The interface re0 is configured with :
>  > ifconfig inet 192.168.0.1/24 media auto mtu 7422 polling
>  > ifconfig re0 -txcsum  solves the issue.
>  > 
>  > I tried to reproduce the issue with a Realtek 8169 (using re(4) too). I 
>  > couln't : checksum offloading works ok on this card.
>  > Is this a known issue (or maybe a bug in the 8168) ?
>  > 
>
> There had been several re(4) instability issues on PCIe based
> controllers. Would you try the following patch and let me know the
> result?
> http://people.freebsd.org/~yongari/re/6.3R/re.busdma.patch
>
> If you use 7.0-RELEASE use the following one.
> http://people.freebsd.org/~yongari/re/7.0R/if_re.c
> http://people.freebsd.org/~yongari/re/7.0R/if_rlreg.h
>   
Sorry, I forgot to precise that I run the amd64 7.0-RELEASE .
I got the patch and recompiled the kernel fine. I'll test this evening.
>  > I can provide some network capture if needed. In the meantime I swapped 
>  > the two cards as I don't need jumbo on one of them.
>  > 
>  > Thanks
>  > 
>  > Arnaud
>