I think - something is wrong with atheros driver
-
In latest pfsense releases i observed such nasty thing - when working with soekris net4801 and CM9 Atheros minipci card - i got system sluggish, specially when starteed bittorrent client from my wifi notebook. I can kill my pfsense board in 15-20 minutes, in those conditions System state table grow up to ~1000-2000 states, but this is less than defined maximum states (10000). Running "top" from serial console shows app. 75MB free memory - so I think - this is no problem. Before system reboot console always show "ath0: unable to start recv logic", sometimes system can not reboot in normal way, so i have to do fsck manually.
Any ideas appreciable.
Thanks, Arnis.
p.s. SNAPSHOT-09-18-06 and later, today I downloaded SNAPSHOT-09-27-06 upgrade and embedded kernel - all the same, no more ideas. -
Sorry but nothing has changed in regards to ATH in many many many many many many many months.
-
Thanks, Scott, thats quieten me, so so I will investigate conditions, may be someone will give some more directions.
-
What is the last pfSense version that works well for you?
-
Thats the problem, that don't know exactly. To be completely true, i'm not sure, that it works well at all. Older versions (before RC2) i lost. I found this problem when moved my soekris box from my workplace to my home place - so conditions changed. In my workplace i tested this box mainly with hughe ftp and smb (windows) file transfers - everything works perfectly - I got ~20Mbit troughput, no hangups, no slowdowns. Now in my home place conditions is little bit changed, mostly traffic pattern. And I get such strange things. Bittorrent is only one example - how to kill the box, other traffic, which produces lot of firewall states also will render system completely unusable.
-
Now i can give more information after my investigations. Shortly after system reboots:
#netstat -m
1338/207/1545 mbufs in use (current/cache/total)
1329/181/1510/4800 mbuf clusters in use (current/cache/total/max)
1327/81 mbuf+clusters out of packet secondary zone in use (current/cache)
0/0/0/0 4k (page size) jumbo clusters in use (current/cache/total/max)
0/0/0/0 9k jumbo clusters in use (current/cache/total/max)
0/0/0/0 16k jumbo clusters in use (current/cache/total/max)
2993K/413K/3406K bytes allocated to network (current/cache/total)
0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)
0/0/0 requests for jumbo clusters denied (4k/9k/16k)
0/4/1456 sfbufs in use (current/peak/max)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
0 calls to protocol drain routines…then I started Bitcomet over wireless link (from my notebook), just to kill system faster. Shortly before crash i get:
#netstat -m
4788/207/4995 mbufs in use (current/cache/total)
4779/21/4800/4800 mbuf clusters in use (current/cache/total/max)
4778/21 mbuf+clusters out of packet secondary zone in use (current/cache)
0/0/0/0 4k (page size) jumbo clusters in use (current/cache/total/max)
0/0/0/0 9k jumbo clusters in use (current/cache/total/max)
0/0/0/0 16k jumbo clusters in use (current/cache/total/max)
11082K/93K/11176K bytes allocated to network (current/cache/total)
0/1734/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)
0/0/0 requests for jumbo clusters denied (4k/9k/16k)
0/4/1456 sfbufs in use (current/peak/max)
0 requests for sfbufs denied
0 requests for sfbufs delayed
0 requests for I/O initiated by sendfile
0 calls to protocol drain routinesand after crash (repeated):
ath0: unable to start recv logic
ath0: ath_rx_proc: no mbuf!
ath0: ath_rx_proc: no mbuf!
ath0: ath_rx_proc: no mbuf!
.....(continiously)I think - there is some kernel memory lekage in atheros driver or somewhere else, because i never get such mbuf outage when doing the same things over wired connections.
btw: I bought new R52 (atheros based) card in place of my CM9 - just to be sure that wifi card is ok.
Any ideas - how to solve this problem?p.s. This happens with embedded and regular kernel, configured as "embeded" and "pfSense" platform, i running system on soekris net4801 with microdrive installed, of course - latest version "1.0-RELEASE-embedded", but i observed this problem also in releases candidates and intermediate versions.
p.p.s. some symptoms like this:
http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/78968- but not exactly.
-
You should try to adress this at the appropriate freebsd list as we don't maintain the driver.
-
http://lists.freebsd.org/pipermail/freebsd-mobile/
-
arno, just to let you know…you arn't the only one with this problem. It works fine for me on an atheros 5212 chipset card for browsing but even doing samba file copies over 200meg or so and it crashes to come back a couple minutes later.
-
we have spent many a time looking at the ath driver it is reasonable.
it seems that some of the issues should be fixed with 0.98 hal but
i have yet to work out how to get this into a bsd box.client mode in 11b with -burst on is a good stable setup. also same on ofdm/6 and 12 on 802.11a in client mode
anything else has sperious issues.
-
Try the latest snapshot from http://www.pfsense.com/~sullrich/
If this still has issues then post a message to the FreeBSD-mobile list.
-
Regarding "11b with -burst" and "ofdm/6 and 12 on 802.11a" in client mode… how do you enable/disable these settings in pfSense?
Thanks, -Pete -
i add it into /etc/inc/interfaces under the wireless configuration section.
there are ifconfig settings there just add a -burst in there. maybe someone else has a better place to put this.
but this works for us -
Try the latest snapshot from http://www.pfsense.com/~sullrich/
If this still has issues then post a message to the FreeBSD-mobile list.
have been attempting to clarify the issue well enough to be able to post on freebsd-modile list.
have tryed the 6.2 snapshots with no change. am presently trying to work out how to add the newest ath_hal into a build.
but this is definately proving to be a difficult learning curvewill post outcomes when i have any more information