Subject: Re: Intel Dual Gigabit Adapter i82546GB not supported...
To: NetBSD Networking Technical Discussion List <tech-net@NetBSD.ORG>
From: Greg A. Woods <woods@weird.com>
List: tech-net
Date: 12/03/2004 15:17:04
[ On Friday, December 3, 2004 at 03:42:20 (-0500), Dheeraj Reddy wrote: ]
> Subject: Re: Intel Dual Gigabit Adapter i82546GB not supported...
>
> > > --------------------------------- 8< ---------------------------------
> > > BUGS
> > >      The Intel i82545GM and i82546GB controllers with  internal SERDES are not
> > >      currently supported.
> > > --------------------------------- 8< ---------------------------------
> > > 
> > > Ugh! Now im lost. Are there plans, to support this anytime 
> > > soon? What is an internal SERDES?
> 
> fwiw,
>   SERDES = serial multiplexer/demultiplexer.

I guess the i82546GB card I have borrowed to test in the alpha dosn't
have that infernal "internal SERDES"....   :-)

(maybe because it is a multi-mode fibre (1000baseSX) card?)

(this is from a very recent -current kernel)

wm0 at pci1 dev 1 function 0: Intel i82546GB 1000BASE-X Ethernet, rev. 3
wm0: interrupting at dec 6600 irq 24
wm0: Ethernet address 00:04:23:a8:7c:02
wm0: 1000baseSX, 1000baseSX-FDX, auto
wm1 at pci1 dev 1 function 1: Intel i82546GB 1000BASE-X Ethernet, rev. 3
wm1: interrupting at dec 6600 irq 25
wm1: Ethernet address 00:04:23:a8:7c:03
wm1: 1000baseSX, 1000baseSX-FDX, auto
 

Now if only I could get the thing to work at anything more than a measly
20-30 MB/s (with TCP, e.g. with ttcp).

[console]<@> # ttcp -l 32768 -b 65536 -v -t -s -n 50000 10.10.10.1
ttcp-t: buflen=32768, nbuf=50000, align=16384/0, port=5001, sockbufsize=65536  tcp  -> 10.10.10.1
ttcp-t: socket
ttcp-t: sndbuf
ttcp-t: connect
ttcp-t: 1638400000 bytes in 46.42 real seconds = 34469.37 KB/sec +++
ttcp-t: 1638400000 bytes in 26.19 CPU seconds = 61095.97 KB/cpu sec
ttcp-t: 50000 I/O calls, msec/call = 0.95, calls/sec = 1077.17
ttcp-t: 0.0user 26.1sys 0:46real 56% 0i+0d 0maxrss 0+200008pf 16+513csw
ttcp-t: buffer address 0x20050000


Even with four simultaneous ttcp runs (on this 4-CPU SMP machine) the
aggregate throughput is about the same.


Rumour has it that the wm(4) driver also suffers from the "too much
memory for fast DMA" syndrome (this machine has 16GB of RAM).

-- 
						Greg A. Woods

+1 416 218-0098                  VE3TCP            RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>          Secrets of the Weird <woods@weird.com>