Well, I'm not sure.  Does the netstat -s output count packets which
are discarded for tcp checksum errors even when they are being transmitted
from the local box?  I think it might be transmission, since my test tcp
traces show us not answering to obvious retransmissions.  Howevr, the the
counter for tcp checksum discards in netstat -s output increments very
steadily as well, leading to my confusion.
	Are you suggesting that I might need the tx short work around your
output lists?
-Brian

On Dec 18,  4:24pm, Ignatios Souvatzis wrote:
} Subject: Re: bge(4) tcp4csum problem -- driver problem or chip problem???
} On Thu, Dec 07, 2006 at 11:40:10AM -0800, Brian Buhrow wrote:
} > 	Hello.  I have a series of Sun V20Z servers with the following
} > Broadcom chips on them, as reported by the bge(4) driver.
} > 
} > bge0 at pci2 dev 2 function 0: Broadcom BCM5703X Gigabit Ethernet
} > bge0: interrupting at irq 5
} > bge0: ASIC BCM5703 A2 (0x1002), Ethernet address 00:09:3d:00:18:26
} > brgphy0 at bge0 phy 1: BCM5703 1000BASE-T media interface, rev. 2
} 
} bge0 at pci1 dev 0 function 0: Broadcom BCM5751M Gigabit Ethernet
} bge0: interrupting at irq 11
} bge0: pcie mode=0x105000
} bge0: ASIC BCM5751 A1 (0x4101), Ethernet address 00:16:d3:29:d0:41
} bge0: setting short Tx thresholds
} brgphy0 at bge0 phy 1: BCM5750 1000BASE-T media interface, rev. 0
} 
} > 	With hardware tcp4 checksumming enabled, I see a large number of tcp
} > packets discarded for checksum errors.  
} 
} NetBSD 4 BETA2, I don't see any  errors, I think. Is this on transmission
} (on the peer) or on reception?
} 
} 	-is
>-- End of excerpt from Ignatios Souvatzis