Subject: kern/30651: sk0 borkes big frames (i.e. NFS traffic)
To: None <,,>
From: None <>
List: netbsd-bugs
Date: 07/01/2005 15:58:00
>Number:         30651
>Category:       kern
>Synopsis:       sk0 borkes big frames (i.e. NFS traffic)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Jul 01 15:58:00 +0000 2005
>Originator:     Martin Husemann
>Release:        NetBSD 3.0_BETA
System: NetBSD 3.0_BETA NetBSD 3.0_BETA (EMMAS) #0: Mon Jun 27 08:40:02 CEST 2005 i386
Architecture: i386
Machine: i386
I have this network interface:
skc0 at pci2 dev 5 function 0: irq 10
skc0: 3Com Gigabit LOM (3C940) rev. (0x1)
sk0 at skc0 port A: Ethernet address 00:0e:a6:28:e3:ec
makphy0 at sk0 phy 0: Marvell 88E1011 Gigabit PHY, rev. 3
makphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FDX, auto

in my NFS server. It works fine as long as I connect it to a 100Mbit switch:

	address: 00:0e:a6:28:e3:ec
	media: Ethernet autoselect (100baseTX full-duplex)
	status: active
	inet ....

But as soon as I connect it to a GigE switch (1000baseTX full-duplex) the
NFS traffic stalls. Small packets still make it through, but clients report
"NFS server ... not responding" - even clients (like my U2) that only have a
100baseTX connection themselfs.

On a side note: once connected to a GigE switch it is not possible to force
the media selection to something slower via "ifconfig sk0 media ..." - I am
not sure that is considered a bug though.