OpenWrt Forum Archive

Topic: Build for WNDR3700/WNDR3800

The content of this topic has been archived between 9 Jul 2013 and 6 May 2018. Unfortunately there are posts – most likely complete pages – missing.

Is the 6in4 hotplug handler supposed to restart radvd in trunk now? I don't see anything like that in the handler, and radvd stops working on interface changes.

arokh wrote:

Is the 6in4 hotplug handler supposed to restart radvd in trunk now? I don't see anything like that in the handler, and radvd stops working on interface changes.

No, the 6in4 hotplug handler has no role in that, but Jow made patches to LuCI that should make radvd to restart after interface changes.

However, I have seen no evidence of that restart happening in real life. So, I have still left my own patch '/etc/hotplug.d/iface/92-radvd' in place both for my Trunk and Backfire versions. And my build restarts radvd ok.

Discussion between messages 37-41 in this thread: https://forum.openwrt.org/viewtopic.php … 70#p128970


Jow's LuCI patches: http://luci.subsignal.org/trac/changeset/6923
http://luci.subsignal.org/trac/changeset/6922

(Last edited by hnyman on 6 Apr 2011, 12:14)

Interface changes can happen outside luci, so a hotplug handler is needed it seems. Stealing yours smile

(Last edited by arokh on 6 Apr 2011, 13:10)

hnyman wrote:

Note: as Backfire 10.03.1-rc5 seems to be in the horizon, I will probably switch back to LuCI 0.10 in the Backfire build in the next few builds.

I have switched back to the "old" LuCI 0.10 in the Backfire build for making it compatible with the (hopefully) forthcoming Backfire 10.03.1-rc5.

The LuCI 0.10 looks a bit boring after having used LuCI trunk for a while, so hopefully the developers are still considering using the LuCI trunk for the rc5 build. Especially the old overview/start page is quite unnecessary and the current LuCI trunk version of that summary page would be much better.

Note: my last Backfire build with the LuCI trunk interface is still available: http://koti.welho.com/hnyman1/Openwrt/B … uCI_trunk/

I have added the Wake-on-LAN (WOL) LuCI module in r26941, but otherwise no recent changes.

Regarding the LuCI user interface:
the Backfire build is still using LuCI 0.10, but as the LuCI trunk was copied to 0.10 a few days ago ( http://luci.subsignal.org/trac/changeset/7038) , there is not much difference between LuCI 0.10 and LuCI trunk from user perspective.

Actually, earlier today I happened to be trying your previous 26869 Backfire build, and was having a lot of trouble with a Mac holding an association. It would connect and complete the WPA2 handshake, but within a few minutes the router would log "disassociated due to inactivity". I had to toggle the wireless on the Mac to get a new connection, and it would repeat. I tried your trunk 26869 build too, same result. So, I backed out to the previous 26472 that I was previously running, no issues after the downgrade.

There weren't any older builds on your site to try to bisect where the problem crept in, and I'll try the new build, but I assume you've not seen this?

I am not using the WLAN that extensively, so there is just the default functionality that there is at the time.

I had trouble with LAN connectivity latencies with one build (of April 30th), so I deleted that build from the server. In general I am only keeping the newest builds there, and I am also marking if I haven't tried a build by myself ("untested").

Success so far with  Backfire-r26941-2011-05-18/ , no wifi dropouts from the Mac or other clients.

I did notice the reported signal and signal/noise values are about 20dbmV off from where they were with 26472. I'm assuming this is a LuCI thing, because the clients aren't seeing any signal strength changes. LuCI looks pretty good overall.

Does some package need to be installed to enable the "Realtime {Load,Traffic,Connections}" graphs in the Status tab? The values are all zero and no images. The values under the Statistics tab are ok.

(Last edited by tt on 19 May 2011, 14:58)

Good to hear.

There is some constant tinkering of wireless drivers by the developers, so the signal strength change comes from that. Most recent one were at r26798/9:
https://dev.openwrt.org/changeset/26798  &  https://dev.openwrt.org/changeset/26799

tt wrote:

Does some package need to be installed to enable the "Realtime {Load,Traffic,Connections}" graphs in the Status tab? The values are all zero and no images.

It might be browser dependent: I see pictures there just fine both with Firefox 5.0a2 and IE9. If I remember correctly, those realtime graphs use SVG graphics ( http://luci.subsignal.org/trac/changeset/6574 ), and SVG is not supported by the earlier IE versions. Which browser you are using?

(Last edited by hnyman on 19 May 2011, 15:42)

Whoa - right you are. Moving over to an Ubuntu/Firefox browser I see the graphs and stats no problem. Or, turning off compatibility mode in IE9 for the LuCI pages, I see them too. I had turned IE9 compat mode on because the login page doesn't render well, and a few LuCI pages too. Looks like I will just leave it off. Thanks for the tip.

I have added the VSFTPD FTP server package to the firmware at r27153.
(Otherwise the default config, but connections from both anonymous clients and local clients are prevented in order to provide initial security.)

(Last edited by hnyman on 10 Jun 2011, 17:39)

In latest trunk build I've got issue with my Mac running time machine often get random network interruptions (probably under high load). Anyone with issue with this build? What are the things that are different from this build and trunk?

mikejuni wrote:

In latest trunk build I've got issue with my Mac running time machine often get random network interruptions (probably under high load). Anyone with issue with this build? What are the things that are different from this build and trunk?

There are only config changes plus some scripts added. Otherwise it should be default values from the main trunk, both in Trunk version and in Backfire version. All the differences are available in the diff files in that FTP directory.

Personally, I find that trunk version has sometimes too high latency and higher CPU load, compared to Backfire version with practically identical config. I suspect that for some reason the trunk's default config is not quite optimal for WNDR3700, but I haven't found any clear reasons for that.

Hello hnyman,
can you post an image of your custom firmware for the wndr3700v2...
I have this router and I'd like to try it.
Thanks

Ultimo67 wrote:

can you post an image of your custom firmware for the wndr3700v2...

As I only have v1, I have not been compiling the v2 firmware.
But I might look into it after summer holidays.

Unrelated note:
As I have been building my version with standard kernel options, I believe that the Backfire 10.03.1-rc5 modules (published today) are compatible with my Backfire build, in case somebody wishes to add modules to my build.

(Last edited by hnyman on 15 Jul 2011, 19:47)

hnyman wrote:
Ultimo67 wrote:

can you post an image of your custom firmware for the wndr3700v2...

As I only have v1, I have not been compiling the v2 firmware.

Actually, it looks like v2 gets compiled automatically, but I have not been copying it. Since version r27693-2011-07-19 I will copy also the v2 image, but I will not test the image...

hnyman wrote:
hnyman wrote:

As I only have v1, I have not been compiling the v2 firmware.

Actually, it looks like v2 gets compiled automatically, but I have not been copying it. Since version r27693-2011-07-19 I will copy also the v2 image, but I will not test the image...

The hnyman v2 image has been working fine on my WNDR3700v2; Thank You very much for sharing your work!

Two questions for the thread:

1. Has anyone installed Snort (or other IDS) in a v1 or v2? If so, are you satisfied with running it there?

2. Is anyone logging firewall messages to the wndr syslog (and out port 514)? If so, how? e.g. did you flush/modify/reload iptables via script after start up, or is there some sort of LuCI package?

TIA, newbie

Has anybody seen stability issues regarding recent ar71xx / WNDR3700 trunk builds (r27840 or later)? My router's trunk version built in August seems to randomly reboot after periods of heavy load, and I am wondering if others have seen anything. (Backfire builds seem to work ok.)

I suspect the recent patches for Netfilter (r27840 and later typo fixes for it...) and/or ar71xx memory corruption fixes (r27894-r27897).

The failure goes this way:
after some period of heavy traffic (20-40 Mb/s in, 3-4 Mb/s out to WAN), the so far stable latency first jumps to over 400 ms, and after a while the router (both luci and SSH console) comes inaccessible, but forwarding traffic may still continue to flow through the router for some time. Then the router dies totally and reboots, blinking the leds like in a normal boot process (except that there is no initial phase with the single orange led). Weirdly, the kernel log and system log do not show the complete boot session, but only a few network related items added to the old logs. Uptime gets reseted. I have not been able to reproduce this always, so it seems to be semi-random. After the crash, the traffic returns normal.  (Only wired LAN is use, no WLAN connections although WLAN is active in router.)

I am going to build versions without those patches, and see if I can identify the possible suspect.

(Last edited by hnyman on 10 Aug 2011, 10:51)

hnyman wrote:

I suspect the recent patches for Netfilter (r27840 and later typo fixes for it...) and/or ar71xx memory corruption fixes (r27894-r27897).

I am going to build versions without those patches, and see if I can identify the possible suspect.

FWIW,  WNDR3700v2-trunk-r27826-2011-07-29-1514-squashfs-factory seems rock solid - consistent with your suspicions.

Please post additional suspicions and/or conclusions here.

TIA

ml2517 wrote:

Might or might not be the culprit:
https://dev.openwrt.org/changeset/27923
and/or this:
https://dev.openwrt.org/changeset/27927

Yeah, those two are the "typo fixes" for patch 27840 that I mentioned in my original message.

I have built a trunk version otherwise up-to-date but without 27840, 27923 and 27927, and it seems to be stable so far...
'trunk-r27944-2011-08-09-without-suspect-netfilter-patches'

(Last edited by hnyman on 10 Aug 2011, 21:00)

hnyman wrote:
ml2517 wrote:

Might or might not be the culprit:
https://dev.openwrt.org/changeset/27923
and/or this:
https://dev.openwrt.org/changeset/27927

Yeah, those two are the "typo fixes" for patch 27840 that I mentioned in my original message.

I have built a trunk version otherwise up-to-date but without 27840, 27923 and 27927, and it seems to be stable so far...

Got ya.

The previous build 'trunk-r27944-2011-08-09-without-suspect-netfilter-patches' from August 9th was built without those 601-, 611-, 612- and 613-patches was rock solid for two days, so I am still inclined to believe that one of four those patches might have caused instability in the previous few builds in August.

For the next version 'trunk-r27959-2011-08-11-without-suspect-610-611-netfilter-patch' I included patches 612 and 613, as they look more simple and risk free (no pointers etc. in source code) than the other two. I left out 611-netfilter_match_bypass_default_table.patch and 610-netfilter_match_bypass_default_checks.patch, which are more complicated. So far, after a couple of hours and a few gigabytes of heavy traffic, the build still looks ok. QoS works well and nice throughput in general.

If that build works ok for a while, I will try to further narrow the possible problem in the next build.

Well, all the trunk builds have been stable since August 9th, even with the "suspect" patches included, so I guess that there was something else that has now been (hopefully) fixed.

August ends with r28138:

I have patched the 6in4 tunnel script for Hurricane Electric / tunnelbroker.net to follow their new login format.
Users need to use "User ID" instead of "username" in the 6in4 tunnel login process.
I also added a note about that to my firmware's Luci version.
See discussion at: https://dev.openwrt.org/ticket/10019