[ previous ] [ next ] [ threads ]
 
 From:  Manuel Kasper <mk at neon1 dot net>
 To:  m0n0wall dash dev at lists dot m0n0 dot ch
 Subject:  Preliminary lockup survey results
 Date:  Sun, 13 Nov 2005 21:14:29 +0100
17 reports have been submitted through
<http://m0n0.ch/wall/lockup.php> so far (thanks all!). Here are the
statistics gathered by parsing the submitted configs and counting the
number of configs where certain m0n0wall features were enabled:

DNS forwarder:          16
Traffic shaper:         11
ipfw on (CP or shaper): 11
DHCP server on LAN:     10
webGUI in HTTPS mode:    7
PPPoE on WAN:            7
PPTP VPN server:         7
Optional interfaces:     6
SNMP agent:              5
DynDNS updater:          5
IPsec VPN:               4
DHCP on WAN:             4
Advanced outbound NAT:   4
1:1 NAT:                 4
Captive portal:          3
Static routes:           1

Out of these, the most likely thing would be ipfw (loaded for traffic
shaper and/or captive portal), but that's only a wild guess. There
may be several unrelated reasons for the lockups. Two users said that
the lockups would go away if they turned off the captive portal.

m0n0wall platforms:

generic-pc              8
wrap                    5
generic-pc-cdrom        3
net48xx                 1

Except for one machine, all have 64 MB of RAM or more. Reported
chipsets on generic-pc platforms: Intel and VIA; CPUs: Intel
(Pentium, PII, PIII) and AMD. Note that there are no reports of
net45xxs freezing, but with only 17 reports in total, that doesn't
mean much.

NIC drivers reported include sis (9 users), xl (5), rl (2), fxp (2),
dc (2), bge, wi, and ed.

On average, the boxes lock up every 16 days (min 1, max 90), which is
of course very helpful in tracing the cause</sarcasm>.

All users either indicated that the lockups did not occur with 1.11
(10 users), or that they didn't know (7 users).

The last version reported to be working is 1.2b10 (3 users). 1.2b9
has been reported 5 times. I find it odd that 1.2b10 should work
while 1.2 doesn't, since the changes were minimal and did not involve
kernel stuff. 1.2b9 -> 1.2b10 is more likely, but I still don't see
what change could have caused this (only went from 4.11-RELEASE-p8 to
-p11).

The majority (10) of users said that they couldn't ping the LAN
interface when the box had locked up (3 said they could and 4 didn't
know). We should also have asked whether the console works during a
"lockup"...

Is there anybody on this list who experiences the lockups and could
do some more in-depth debugging?

- Manuel