Re: CFT: major wireless changes

On Fri, Jun 01, 2007 at 10:49:26AM -0700, Sam Leffler wrote:
> This patch against HEAD imports the work that's been sitting in the 
> sam_wifi p4 branch:
> 
> http://people.freebsd.org/~sam/sam_wifi-20070601.patch.gz
>... 
> Note that scanning in net80211 has been broken out into separate modules 
> and you must now add device lines for wlan_scan_ap+wlan_scan_sta (as 
> appropriate) or load the modules as you do the crypto modules.

Hmm... not sure I make use of this:  for determining SSID, encryption,
etc., I usse a perl script I cobbbled up a while back that tries various
known configurations (falling back to a default, if nothing else works).
It doesn't use the "scan" caapabilities of ifconfig, per se.
 
>....
> This is the work of many people including Kip Macy, Max Laier, Sephe 
> Ziehau, Andrew Thompson, and Kevin Lo.  Please report problems to this list.

Right.  OK;  for starters:

localhost(7.0-C)[29] uname -a
FreeBSD g1-1.catwhisker.org. 7.0-CURRENT FreeBSD 7.0-CURRENT #441: Fri Jun  1 16:48:51 PDT 2007     root_at_g1-18.catwhisker.org.:/common/S3/obj/usr/src/sys/LAPTOP_30W  i386
localhost(7.0-C)[30] 

This was a result of a fairly normal buildworld & friends from this morning:

g1-18(7.0-C)[5] uname -a
FreeBSD g1-18.catwhisker.org. 7.0-CURRENT FreeBSD 7.0-CURRENT #440: Fri Jun  1 11:25:54 PDT 2007     root_at_g1-18.catwhisker.org.:/common/S4/obj/usr/src/sys/LAPTOP_30W  i386
g1-18(7.0-C)[6]

to which I applied the patchset Sam cited, getting the single expected
reject for sys/net80211/ieee80211_input.c, line 765 for hunk #24.  I then
re-ran the "make buldworld" & friends, as usual; nothing exciting so far.


On boot, things seemmed normal until I noticed that the machine was
ussing "localhost" as the hostname instead of the PTR value for the
IP address it obtained via DHCP.

I started checking what the script (that I wrote, mentioned above)
actually does, and the first clue I received was that running a
bare "ifconfig" command appeared to work, except that it also
dumped core.  ("echo $?" yielded 139, IIRC.)

As expected, the resulting core file wasn't especially illuminating,
so I rebuilt ifconfig with the -g flag, hoping to get a little more
insight.

The resulting executable does not dump core.

Oh.  OK; carrying on, then....

I then tried running my script by hand (vs. the /usr/local/etc/rc.d
automagic), and discovered that although it now was able to see the
wi0 device, the script didn't persuade the wi0 NIC to associate with
either of the APs (which are configured similarly, taking into
account that they are from different vendors -- Apple & Linksys).

I then tried the built-in xl0 NIC; ifconfig showed "active" when I
plugged a cable in.  I  tried a PCcard de(4) device; that also
showed "active" when I plugged a cable in.

I  then tried an PCcard an(4) device (a venerable Cisco/Aironet 350
card that had seen a fair amount of prior service before I obtained
my current laptop, which has the wi0 NIC built in as a miniPCI card),
and it appears to work -- it's what I'm using to access the machine
where I'm writing this message.  (That said, it appears subject to
pauses that last for about 10 seconds or so; I believe that these
messages are associated with the pauses:

Jun  1 17:58:24 localhost kernel: an0: RID access failed
Jun  1 17:58:24 localhost kernel: an0: xmit failed
Jun  1 18:10:13 g1-1 kernel: an0: RID access failed
Jun  1 18:10:13 g1-1 kernel: an0: xmit failed

but I'm not absolutely certain of it.)


Summmary:

* ifconfig dumped core; not sure why, and debug version doesn't.
* wi(4) device didn't associate; will poke around and see if I
  can  figure out what's broken -- it may just be something I
  need to tweak in my script.
* an(4) associates, but appears to "hang" for several seconds
  every once in a while.
* Wired devices I used for reality checks (xl(4) and de(4))
  seem OK,but I didn't actually try using them for traffic yet.

The machine in question is a single-CPU i386 machine; nothing
exotic.  (It's a Dell Inspiron 8200 with 1 GB RAM.)

As I mentioned to Sam privately, I have a local private mirror
of the FreeBSD CVS repository handy, and the laptop is configured
to facilitate testing of this nature (to a certain extent, at
least), so I'm willing to try experiments.  Within (what I
consider) reason, anyway.  :-}

Peace,
david
-- 
David H. Wolfskill				david_at_catwhisker.org
Anything and everything is a (potential) cat toy.

See http://www.catwhisker.org/~david/publickey.gpg for my public key.