em driver input errors

[alexpalias-bsdnet at yahoo.com]

--- On Thu, 8/20/09, H.Fazaeli <fazaeli at sepehrs.com> wrote:

> From: H.Fazaeli <fazaeli at sepehrs.com>
> Subject: Re: em driver input errors
> To: alexpalias-bsdnet at yahoo.com
> Cc: freebsd-net at freebsd.org
> Date: Thursday, August 20, 2009, 12:22 PM
> 
>  
> can you provide sysctl dev.em.0.debug=1 output?

em0: Adapter hardware address = 0xfffffffe80230320 
em0: CTRL = 0x40f00241 RCTL = 0x8002 
em0: Packet buffer = Tx=16k Rx=48k 
em0: Flow control watermarks high = 47104 low = 45604
em0: tx_int_delay = 0, tx_abs_int_delay = 0
em0: rx_int_delay = 0, rx_abs_int_delay = 0
em0: fifo workaround = 0, fifo_reset_count = 0
em0: hw tdh = 259, hw tdt = 259
em0: hw rdh = 603, hw rdt = 602
em0: Num Tx descriptors avail = 4095
em0: Tx Descriptors not avail1 = 0
em0: Tx Descriptors not avail2 = 0
em0: Std mbuf failed = 0
em0: Std mbuf cluster failed = 0
em0: Driver dropped packets = 0
em0: Driver tx dma failure in encap = 0


> There are posts that show the following workarounds have 
> fixed problems similar to yours on other interface types:
> 
> - disable tso (ifconfig em0 -tso)

There's no TSO flag active currently, but I did try it... I guess it's a no-op.  Also, the machine is a router, meaning most of the packets are just being forwarded.  As far as I understand, TSO is concerned with TCP connections with one local endpoint.

Anyway, I did this (after looking in the README for the new intel driver):
sysctl net.inet.tcp.tso=0

> - disable msi/msix (hw.em.enable_msi=0 in
> /boot/loader.conf. There are also hw.pci.enable_msix and 
>   hw.pci.enable_msi but I don't know their exact
> effect).

This, and the next one (new driver), will be much harder to try in a production environment... I'll try to find some test machines to torture.

> You may also try upgrading to the latest intel em driver
> (6.9.12).
> http://downloadcenter.intel.com/Filter_Results.aspx?strOSs=52&strTypes=all&ProductID=1067&OSFullName=FreeBSD*&lang=eng
> 


Thanks
    Alex