NetBSD-Bugs archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

kern/37762: bge hw assist problems



>Number:         37762
>Category:       kern
>Synopsis:       bge hw assist problems
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jan 15 12:20:00 +0000 2008
>Originator:     arto.huusko%pp2.inet.fi@localhost
>Release:        NetBSD 4.99.49
>Organization:
>Environment:
System: NetBSD vekotin.local 4.99.49 NetBSD 4.99.49 (VEKOTIN) #1: Sun Jan 13 
00:31:02 EET 2008 
root%mutteri.local@localhost:/local/netbsd/build/netbsd-current/amd64/obj/sys/arch/amd64/compile/VEKOTIN
 amd64
Architecture: x86_64
Machine: amd64
>Description:
        There are some problems with hardware assisted TCP segmentation and
        checksumming in the bge driver. The NIC is attached like so:

bge0 at pci2 dev 0 function 0: Broadcom BCM5787M Gigabit Ethernet
bge0: interrupting at ioapic0 pin 18 (irq 11)
bge0: ASIC BCM5754/5787 A2 (0xb002), Ethernet address 00:1b:38:2b:fc:30
bge0: setting short Tx thresholds
brgphy0 at bge0 phy 1: BCM5754/5787 1000BASE-T media interface, rev. 0
bge0: block failed to stop: reg 0x1400, bit 0x00000002

        ifconfig bge0 says:

bge0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
        
capabilities=3f80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx,TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx>
        enabled=0
        address: 00:1b:38:2b:fc:30
        media: Ethernet autoselect (100baseTX 
full-duplex,flowcontrol,rxpause,txpause)
        status: active
        inet 10.0.0.11 netmask 0xffffff00 broadcast 10.0.0.255

        I don't have any specifics what exactly is going wrong, but the symptoms
        are

        - With TSO4, NFS (client) is useless. With even a little traffic the
          client seems to think the server is gone (server not responding
          messages).
          tcpdumping the traffic shows some NFS traffic happening constantly,
          but tcpdump shows for example invalid rpc response codes etc.

        - With TCP checksum options turned on, bad checksums occur.
          One interesting effect was that I couldn't access www.facebook.com,
          but other sites worked mostly. When I turned TCP checksum offload
          off, everything seems to work (except skype).

          netstat -s on this machine shows 208 received packets discarded due
          to bad checksums. The LAN router shows 480 received packets
          discarded due to checksums.

        - IP4 csum offload possibly works, netstat -s shows no IP packets
          discarded due to checksum. Same for UDP.

        - Even with all offload options turned off I still can't connect skype,
          but that could be due to linux32 emulation problems and not bge. But
          I wouldn't rule bge out either.
>How-To-Repeat:
        Run bge on Acer Aspire 5310 with various hw assist options turned on.
>Fix:
        Not known.




Home | Main Index | Thread Index | Old Index