News: This forum is now permanently frozen.
Pages: [1]
Topic: 1.8 on GB-1000 Lockups  (Read 4954 times)
« on: July 20, 2011, 18:50:35 »
Fred Grayson *****
Posts: 994

I thought I'd dive into 1.8. I gave build 473 a try. Naturally, I stared with the generic-pc version that works with my GB-1000 hardware, including the serial console. The 1.33 generic-pc-serial does not work with the console on the GB-1000. I've reported this elsewhere.

Well, generic-pc-1.8.0b475.img does not provide a working serial console, so I next tried generic-pc-serial-1.8.0b475.img which does provide a working serial console!

The problem now is that there is no LAN to WAN connectivity using my working 1.33 config file. Not by name, not by IP, none at all. I can ping and trace by both name and IP from the serial console or the web GUI, just not from apps or shells on LAN machines.

I reset the upgraded 1.8 over 1.33 to Factory Defaults and configured my WAN and LAN interfaces manually. This gets it working, but I really don't want to go thru the many other changes manually to try and find what in my current 1.33 config file breaks 1.8.

I'll have to look at the broken 1.8 with my 1.33 config file closer. Maybe there will be something in the logs that shows what's hosing it.
« Last Edit: August 06, 2011, 16:09:55 by fredg »

--
Google is your friend and Bob's your uncle.
« Reply #1 on: July 20, 2011, 20:22:44 »
Fred Grayson *****
Posts: 994

I started disabling various things within my once working 1.33 configuration file running on generic-pc-serial-1.8.0b475 until LAN to WAN connectivity was restored. It started working once the Traffic Shaper was disabled.

I verified that it was this, and this alone by restoring the full 1.33 config file and then disabling only the Traffic Shaper.

Anyone else seeing anything like this?


--
Google is your friend and Bob's your uncle.
« Reply #2 on: July 22, 2011, 18:45:40 »
Fred Grayson *****
Posts: 994

Running pc-serial-1.8.0b475 with the Traffic Shaper disabled. System locks up within a few minutes and does not respond to anything. Nothing is written to the console, and once locked up the logs are unreadable.

pc-serial-1.8.0b476 does the same thing.

I'll try new releases as they appear, but for now it's back to 1.33.

--
Google is your friend and Bob's your uncle.
« Reply #3 on: July 26, 2011, 22:45:50 »
brushedmoss ****
Posts: 446

Hi Fred,

can you enable syslog to a remote server and see what it reports when the system 'hangs' ?

how much memory is in your system ?

wrt serial console, this was changed between 1.33 and 1.8 to remove some baud rate code, so glad it works on your system now.

wrt traffic shaper, AFAIK some people are using this without issue, with TS enabled you immediately get not packetforwarding ?
« Reply #4 on: July 27, 2011, 00:06:35 »
Fred Grayson *****
Posts: 994

I'll enable remote syslogging, upgrade to pc-serial-1.8.0b476, see what appears in the remote log and get back to you with that when it crashes.

I have 128MB RAM in the system.

As soon as the Traffic Shaper is enabled, packet forwarding out the WAN stops. When I disable the Shaper, forwarding works again. I have whatever rules, pipes, and queues the Magic Shaper Wizard provides.

--
Google is your friend and Bob's your uncle.
« Reply #5 on: July 27, 2011, 00:29:40 »
Fred Grayson *****
Posts: 994

The only things that appear in the remote syslog are the following; I don't think they are related to the lockup/crash. They seem to get in there earlier:

Jul 26 18:08:32 adsl-98-64-126-119.mia.bellsouth.net reboot: rebooted by root
Jul 26 18:08:32 adsl-98-64-126-119.mia.bellsouth.net syslogd: exiting on signal 15

The following is written to the console as soon as the boot is completed:

Enter a number: fxp1: Microcode loaded, int_delay: 1000 usec  bundle_max: 6

Lines like the above for the other configured interfaces (fxp0, fxp2) appear earlier in the console output, well above the console setup menu.

EDIT:

Also, to eliminate the possibility that this is upgrade related, I started over with a bare CF card. After booting, this image locks up almost immediately as well. Whatever is happening takes it down faster than anything can be written to the log, or isn't triggering a log write at all.

Thanks for looking.
« Last Edit: July 27, 2011, 15:54:51 by fredg »

--
Google is your friend and Bob's your uncle.
« Reply #6 on: August 03, 2011, 14:09:50 »
brushedmoss ****
Posts: 446

weird,

I have tested on a 128mb system with traffic shaper on, and didn't get a lock up.  Let me revert back to you  with some suggestions later this week
« Reply #7 on: August 03, 2011, 16:01:25 »
Fred Grayson *****
Posts: 994

Could be a GB-1000 oddity, like that flail we went through with the console. I'm still scratching my head on that one. As I said, it's hard to try things on this setup as it crashes very soon after bootup.

I await your suggestions. Thanks again.

--
Google is your friend and Bob's your uncle.
« Reply #8 on: August 26, 2011, 10:28:45 »
MasterGohan *
Posts: 1

I started disabling various things within my once working 1.33 configuration file running on generic-pc-serial-1.8.0b475 until LAN to WAN connectivity was restored. It started working once the Traffic Shaper was disabled.

I verified that it was this, and this alone by restoring the full 1.33 config file and then disabling only the Traffic Shaper.

Anyone else seeing anything like this?



Registered an account just to comment on this..

Experienced IDENTICAL with all v1.8 beta builds of m0n0wall.  While I don't have a GB-1000, I tested with my Athlon XP system, a Duron, and two Sempron machines.  Basically that being, the moment that Traffic Shaping was enabled routing would completely (and instantly) stop functioning (on systems known working with traffic shaping on 1.33).


I'm finding however, that it seems to be directly tied to the NICs that I'm using.  The AMD motherboards I used all have integrated VIA NIC's, I'd also tried two Realtek PCI NICs to similar outcome.

Now the interesting part:  When using my Zyxel Gigabit (vge0), on the WAN side, outgoing connections function (IE, I can surf the internet) ... but not incoming connections.  When I use it on the LAN end, remote initiated connections work, yet not outgoing.

--I assume that IF I had two of those Zyxel NICs, everything would be peachy and I could enable shaping on 1.8.


So ... I then decided that I'd give it a shot looking at pfsense (since I was starting at this point to wonder if what I was seeing was specific to FreeBSD 8).  Sure enough, the EXACT same thing happens in pfsense 2.0.  Whatever it is, I'm guessing it's not anything in m0n0wall's code directly so much as a problem that came from transitioning to the newer version of FreeBSD.

That said, while it's a shame that my previously good hardware no longer works.  Any suggestions far as confirmed working NICs for m0n0wall 1.8?
« Last Edit: August 26, 2011, 10:32:39 by MasterGohan »
« Reply #9 on: August 26, 2011, 19:05:33 »
Fred Grayson *****
Posts: 994

Just for completeness the GB-1000 has four onboard Intel NICs. From dmesg they report as Intel 82559 Pro/100 Ethernet and use the fxp driver.

I can't blame the NICs for what I am seeing on 1.8, at least not yet. The GB-1000 hardware is undocumented and somewhat mysterious.

The one I am currently using has been flawless with m0n0wall generic-pc-1.33.img on a CF card and 128MB RAM.

--
Google is your friend and Bob's your uncle.
« Reply #10 on: August 30, 2011, 10:34:28 »
brushedmoss ****
Posts: 446

Hi,  was away on vacation and have just changed job, so have and am , very busy, so sorry for silence.

MasterGohan, with you zyxel as LAN, can setup Traffic shaping and give me the output of status.php via a PM ?

Also, it would be worth turning off some of the nic hardware features, from exec.php

ifconfig fxp0 -rxcsum -txcsum -tso -lro -polling

etc.
« Reply #11 on: August 30, 2011, 17:27:23 »
Fred Grayson *****
Posts: 994

I'll be waiting for the next build. Currently, with generic-pc-serial-1.8.0b477.img on GB-1000 there is no WebGUI. Port scanning the LAN IP shows the FTP and DNS proxies open, but nothing on port 80 or 443.

--
Google is your friend and Bob's your uncle.
« Reply #12 on: August 30, 2011, 19:00:59 »
brushedmoss ****
Posts: 446

no webgui in 1.8.0b477 only ?  you had webui in previous 1.8.0b , right ? 

you could put the ifconfig in your config as an shellcmd.
« Reply #13 on: August 30, 2011, 22:02:16 »
Fred Grayson *****
Posts: 994

There was GUI in build 475, but it and build 476 would crash so frequently that I often couldn't get to the GUI to make changes.

--
Google is your friend and Bob's your uncle.
 
Pages: [1]
 
 
Powered by SMF 1.1.20 | SMF © 2013, Simple Machines