On Mon, Nov 10, 2008 at 01:58:54PM +0300, Boris Samorodov wrote:
> Pyun YongHyeon writes:
> > On Sat, Nov 08, 2008 at 12:27:12PM +0300, Boris Samorodov wrote:
> > > Pyun YongHyeon writes:

> >
> > [...]
> > > > One user reported non-working NFS over UDP and I disabled Rx
> > > > checksum offload as AR81xx hardware is not able to handle
> > > > fragmented IP datagrams correctly. So it's highly recommended to
> > > > disable Rx checksum offload or use the following updated files.
> > > >
> > > > http://people.freebsd.org/~yongari/a...0081108.tar.gz
> > >
> > > Tested at EeePC-1000. The perfomance dropped (seems to be expected)
> > > twice -- to 5.5 MB/s (fetching a big file to tmpfs). Other than that
> > > works fine. This is for:
> > > -----
> > > ale0@pci0:4:0:0: class=0x020000 card=0x83241043 chip=0x10261969 rev=0xb0 hdr=0x00
> > > vendor = 'Attansic (Now owned by Atheros)'
> > > class = network
> > > subclass = ethernet
> > > -----

> >
> > Fortunately, I've managed to add work-around for Rx checksum
> > offload issue. Would you try latest ale(4) at the fowllowing URL?
> > http://people.freebsd.org/~yongari/a...0081110.tar.gz

>
> Great, thanks! The perfomance is returning back. ;-) It shows
> approx. 10.5 MB/s while ftp'ing a big file to tmpfs. I'd like to show


I didn't even expect such a quantum improvement! :-)

> you some additional info:
> -----
> uname -a ftp://ftp.bsam.ru/pub/tmp/EeePC/uname.2
> netstat -w 1 ftp://ftp.bsam.ru/pub/tmp/EeePC/netstat.ale.2
> sysctl dev.ale.0.stats ftp://ftp.bsam.ru/pub/tmp/EeePC/sysctl.ale.stats.2
> iostat -w 1 ftp://ftp.bsam.ru/pub/tmp/EeePC/iostat.ale.2
> -----
>
> The interesting one is a netstat one. I'm not sure what zeroes for
> packets mean while trafic exists.
>


Hmm, I also have no idea why netstat(1) shows such a non-sense
value while transfer is in progress. I guess you can easily write a
script that extracts interesting MAC statistics of ale(4).
For example,
sysctl dev.ale.0.stats.rx.good_frames
sysctl dev.ale.0.stats.tx.good_frames

Use 'sysctl -d dev.ale.0.stats' to get complete descriptoin of each
node.

> BTW, a flood ping created a 233 packets per second trafic.
>


Since ale(4) requires a lot of CPU cycles to saturate link, it may
completely depend on your CPU power.
--
Regards,
Pyun YongHyeon
_______________________________________________
freebsd-current@freebsd.org mailing list
http://lists.freebsd.org/mailman/lis...reebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"