DragonFlyBSD

Committed.

Sascha

This breaks bge(4) for NICs I use, see below:
- on nfs client using bge(4) for nfs:
'cat /nfs/file' stalls without printing anything from file.
tcpdump on nfs client shows that data is received;
and resent with doubling interval.
'ls /nfs' normally works before stall.

- using tcp for nfs fixes problem
 - 'ifconfig bge0 -rxcsum' fixes problem
 - backing out commit fixes problem

icmp & tcp doesn't seem to be affected: ping & ssh works without problems.
Trying old kernels from chlamydia pointed to data of this commit.

thomas
- host1
- dmesg
bge0: <Altima AC9100 Gigabit Ethernet, ASIC rev. 0x105> mem 0xec040000-0xec04ffff irq 10 at device 12.0 on pci0
miibus1: <MII bus> on bge0
brgphy0: <BCM5701 10/100/1000baseTX PHY> on miibus1
brgphy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FDX, auto
bge0: MAC address: 00:09:5b:22:d9:f2

- pciconf -lv
bge0@pci0:12:0: class=0x020000 card=0x302a1385 chip=0x03ea173b rev=0x15 hdr=0x00
vendor = 'Broadcom (Was: Altima Communications Inc)'
device = 'AC1002 Gigabit Ethernet Adapter'
class = network
subclass = ethernet

-- host2
-dmesg
bge0: <Broadcom BCM5751 Gigabit Ethernet, ASIC rev. 0x4001> mem 0xdfcf0000-0xdfcfffff irq 11 at device 0.0 on pci2
miibus0: <MII bus> on bge0
brgphy0: <BCM5750 10/100/1000baseT PHY> on miibus0
brgphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FDX, auto
bge0: MAC address: 00:12:3f:5f:b3:3d

- pciconf -lv
bge0@pci2:0:0: class=0x020000 card=0x01791028 chip=0x167714e4 rev=0x01 hdr=0x00
vendor = 'Broadcom Corporation'
device = 'BCM5750A1 NetXtreme Gigabit Ethernet PCI Express'
class = network
subclass = ethernet

Should be fixed on HEAD, please test whether it works for you.

I suggest using polling(4) on this card; interrupt moderation is
completely broken on it.

Best Regards,
sephe

It works for me. Thank you.

-thomas