**** BEGIN LOGGING AT Mon May 10 02:59:57 2010 May 10 08:30:05 xMff: the patched preinit script did not do any good, unfortunately May 10 08:31:11 Mr_Smoke: okay, need to find some broadcom with serial to actually test it May 10 08:31:40 xMff: I have one handy, can I do anything to help ? May 10 08:32:04 you could flash the jffs2 image May 10 08:32:17 then you can edit preinit stuff without having to reflash for each test May 10 08:32:52 sprinkle it with debug echos and find out why eth0 isn't brought up May 10 08:33:03 Ok that's an idea indeed May 10 08:33:10 echo debugging is fun ;) May 10 08:33:17 Oh also, one thing May 10 08:33:26 Is it totally impossible to soft-reboot from failsafe ? May 10 08:34:20 the normal reboot does not work May 10 08:34:30 Yeah, I noticed that indeed May 10 08:34:31 but echo b > /proc/sysrq-trigger usually does May 10 08:34:40 Ah thanks that's exactly what I couldn't remember May 10 08:34:59 Thanks :) I'll go play with jffs2 image now May 10 09:39:40 Is "alex@" around here by any chance ? May 10 09:45:17 xMff: erm, funny thing. I just flashed the jffs2 image, and /e/c/network is absent, so I don't get any network at all, even in normal boot May 10 09:56:08 build #54 of ep93xx is complete: Failure [failed compile_10] Build details are at http://tksite.gotdns.org:8010/builders/ep93xx/builds/54 May 10 14:33:39 ping Mr_Smoke May 10 14:33:45 ping xMff May 10 14:35:28 sorry I disappeared yesterday....computer trouble May 10 14:38:44 Hey that's quite alright you know à) May 10 14:38:45 :) May 10 14:39:06 I was told to get in touch with you for 2 things: May 10 14:39:19 1) the sdcard preinit thing, which I posted on openwrt-devel@ eventually May 10 14:39:33 2) the failsafe problem, for which I was advised to use a jffs2 image for debugging May 10 14:40:16 But the plot thickens. I've just flashed the jffs2 image, but there is no /etc/config/network May 10 14:40:30 In short : squashfs image = network ok, but no network in failsafe May 10 14:40:39 jffs2 image = no network, ever May 10 14:40:53 Quite strange isn't it ? May 10 14:41:18 Mr_Smoke: failsafe doesn't need /etc/config/network May 10 14:41:28 Mr_Smoke: in fact, it doesn't use it at all May 10 14:41:31 rtz2: that I know May 10 14:41:39 I'm talking about two different issues here May 10 14:41:51 Mr_Smoke: one problem at the time :) May 10 14:41:54 in failsafe, backfire doesn't have any eth0 configured May 10 14:42:03 That's the main issue May 10 14:42:19 (on brcm-2.4/brcm47xx) May 10 14:42:25 brcm-2.4 here yeah May 10 14:42:50 hmmm...ok, Let me look May 10 14:42:56 Sure thing May 10 14:43:48 cshore: preinit for broadcom does some switch setup stuff but it does no vconfig to add the vlan if May 10 14:44:06 cshore: also, since pi_ifname is unset, no ip is configured ever May 10 14:44:19 jow_laptop: I tried applying your patch for this May 10 14:44:37 http://paste.tksite.gotdns.org/d1365d33c <= this one May 10 14:44:41 but no luck. May 10 14:45:29 That's when x.Mff advised me to use the jffs2 image to debug (directly editing the scripts without reflashing, namely), but the jffs2 image doesn't seem to have any /e/c/network, which is the second issue May 10 14:45:31 Mr_Smoke: regarding your /etc/config/network issue, it should be generated by /etc/init.d/netconfig May 10 14:45:41 Hm May 10 14:46:07 Let's try running that manually then May 10 14:46:14 iirc that init scripts checks whether there already is a /etc/config/network file, if yes it does nothing May 10 14:46:28 doesn't 2.4 use eth0 not eth0.0? May 10 14:46:41 or rather ethX May 10 14:46:54 jow_laptop: ok, could be a case of PEBKAC here : "No space left on device" May 10 14:46:56 i.e. not dotted subnames? May 10 14:47:09 cshore: I can try that May 10 14:47:26 cshore: hm not sure May 10 14:47:50 I should use the 64k image, right ? unit is WRT54GL 1.1, so should be 4MB Flash May 10 14:47:59 I know for 2.6 it's the dotted subname May 10 14:48:50 Mr_Smoke: cat /proc/mtd May 10 14:49:10 Mr_Smoke: if erase size is 0x10000 then 64k, if it is 0x20000 then 128k May 10 14:51:26 oh, I remember now....brcm-2.4 original code configured the switch all as one eth device not as vlans May 10 14:51:30 Yeah that's 10000 May 10 14:51:57 I can't understand the "no space left on device" issue though May 10 14:52:02 I'll remove a few packages maybe May 10 14:53:39 jow_laptop: the jffs2 image is quite a bit bigger than the squashfs then I assume May 10 14:53:40 jffs2 is bigger than squashfs...it's easy to run out of space on 4MB May 10 14:53:43 Ok May 10 14:53:56 I'll remove all the extroot stuff I had been using May 10 14:55:31 Maybe we should look at the ar71xx stuff (IIRC). It configures does vlan magic May 10 14:56:16 I just copied what was there before for brcm-2.4 May 10 14:59:19 not ar71xx, atheros May 10 15:00:40 I think for brcm-2.4 the idea was to have no vlans (I'm guessing, that was what it looked like orginally) May 10 15:01:23 having vlans in failsafe is kinda pointless imho May 10 15:01:50 rtz2: the main reason do to vlans is to prevent leakage between vlans May 10 15:03:04 rtz2: it's an issue that needs to be worked on for a few archs May 10 15:03:18 ahhh, right May 10 15:04:28 there was one arch that configured/deconfigured each port as a vlan and did preinit messages to it, but I can't remember which one May 10 15:04:38 to each I mean May 10 15:04:42 [florian]: ping May 10 15:12:35 hey it's actually brcm47xx that has the per vlan code May 10 15:12:48 see 15_set_preinit_interface May 10 15:13:33 The reason there is no vnconfig is that it should be configured in preinit_net_echo May 10 15:15:37 rebooting with a little more space on the jffs2, and I'll read May 10 15:16:34 I think the problem is that the per-vlan stuff is missing vnconfig....or maybe it's the ip deconfig after each net_echo May 10 15:17:29 Darn 32k available and it still can't write a thing May 10 15:18:46 yeah, it really doesn't like less than an erase block (64k on your device) May 10 15:19:01 Should be good now May 10 15:19:25 Ok we're good May 10 15:19:31 now onto that vlan/failsafe thing May 10 15:19:36 craigc * r21415 /packages/net/multiwan/ (Makefile files/usr/bin/multiwan): [packages] multiwan: Tweaked probability for random packet selection May 10 15:19:37 What would you like me to check ? May 10 15:20:00 preinit_net_echo ... let's see May 10 15:20:51 /lib/preinit/15_set_preinit_interface May 10 15:21:23 that's 15_set_preinit_interface_brcm here May 10 15:21:28 yeah May 10 15:22:12 But it provides no alternative in case $pi_ifname is undef, as far as I can see May 10 15:22:46 That's because in that case it does the same as every other arch when there is not pi_ifname....it doesn't use an interface May 10 15:23:05 That can be defined in Image Option | Preinit options in menuconfig May 10 15:23:22 So you're saying that this image option is mandatory ? May 10 15:23:48 If you want network and there is no default for your device, then yes May 10 15:24:11 I see. Then may I suggest either of 2 things: May 10 15:24:20 - create a default for brcm-2.4 May 10 15:24:31 - warn users when using buildroot May 10 15:24:41 Hm on second though, even the default image has that problem May 10 15:24:48 the default is set in 05_init_interfaces_brcm May 10 15:24:52 So in fact the user warning is probably not an option May 10 15:24:53 oh May 10 15:24:56 pi_ifname = ifname if ifname is set May 10 15:24:58 let's vi May 10 15:25:06 And where is ifname set itself ? May 10 15:25:33 Oh, further up May 10 15:25:36 yep May 10 15:25:58 So, erm May 10 15:26:00 I'm kinda lost here May 10 15:26:17 set_preinit_iface sets ifname to eth0 May 10 15:26:21 yes May 10 15:26:33 then later on, since pi_ifname is unset, it should use ifname instead May 10 15:26:35 except for devices where it is eth1 May 10 15:26:39 right May 10 15:26:42 ok May 10 15:26:45 so pi_ifname should = eth0 May 10 15:26:53 But apparently it's not May 10 15:27:00 Lets reboot into failsafe again May 10 15:27:39 Yup, it's not using it May 10 15:28:38 acoul: ping May 10 15:29:22 did you do an echo in the script, or from shell...it's not 'export'ed, only exists withing the preinit scripts May 10 15:29:54 I'll echo in a moment May 10 15:29:57 Just one thing : May 10 15:30:19 the interface is brought up by failsafe_ip() from 05_set-failsafe-switch-brcm, right ? May 10 15:32:10 Hm that's odd May 10 15:32:14 cshore: pi_ifname is notwhere initialized from ifname if unset May 10 15:32:25 cshore: atheros had the same issue May 10 15:32:33 In failsafe_ip, both $if_name and $pi_ifname are set to eth0 May 10 15:33:24 ah ok May 10 15:33:39 So it must be further down May 10 15:33:41 ok, that's as it should be May 10 15:34:28 what about by the time it reaches 15_set_preinit_interface? May 10 15:34:53 I'm adding echoes to that one in a sec May 10 15:35:24 I think the problem is the lack of vnconfig in preinit_net_echo myself May 10 15:36:33 or the preinit_ip_deconfig (if you're trying to use for other than preinit_net_echo) May 10 15:37:10 Darn May 10 15:37:13 It's set everywhere Oo May 10 15:37:34 both are set to eth0 in pretty much every function from step 05 to 15 May 10 15:38:09 Well I can add whatever vnconfig you tell me to :) May 10 15:38:56 Do you know how 2.4 vlans works? I'm not sure, so I'll be testing through you if not May 10 15:39:36 Hm, depends what you mean by "how it works" May 10 15:40:00 well how you're supposed to configure them May 10 15:40:07 In terms of port numbers etc ? May 10 15:40:27 no, I mean echo /proc/switch/blah... and vnconfig May 10 15:40:58 Hm no, I couldn't do that off the top of my head, but I saw a couple of scripts that took care of that May 10 15:41:23 ah, ok, do you remember which ones? May 10 15:42:10 0 1 2 3 5* are vlan0 usually May 10 15:42:22 and 4 5 vlan 1 May 10 15:42:46 right now, from failsafe, I have a different situation obviously : May 10 15:42:54 well, for preinit we do 0 5, 1 5, 2 5, etc each configured as vlan 0 May 10 15:42:58 root@(none):/# cat /proc/switch/eth0/vlan/0/ports May 10 15:42:58 4 5u May 10 15:43:09 I mean which scripts, sorry May 10 15:43:18 Oh May 10 15:43:25 Well I can grep :) May 10 15:43:53 I've got this : May 10 15:43:54 05_set_failsafe_switch_brcm: echo "0 1 2 3 4 ${cpu_port:-5u*}" > /proc/switch/eth0/vlan/0/ports May 10 15:44:24 oh, I know that, I'm not sure if preinit_net_echo is right, is why I ask May 10 15:44:35 And then in 15_set_preinit_interface_brcm, there's a for loop over seq(0,4) May 10 15:44:43 and it does May 10 15:44:44 echo "$port ${cpu_port:-5u*}" > /proc/switch/eth0/vlan/0/ports May 10 15:44:44 port_net_echo $1 May 10 15:45:03 I think it needs vnconfig 0 0 or 0 1 or something May 10 15:45:38 I know. May 10 15:45:44 Oh, yay then :) May 10 15:46:12 Can you echo "0 1 2 3 4 5" >/proc/switch/eth0/vlan/0/ports May 10 15:46:49 Where do you want it ? May 10 15:47:07 actually also, can you cat /proc/switch/eth0/enable from the command line May 10 15:47:17 and /proc/switch/eth0/vlan_enable May 10 15:47:20 oh ok May 10 15:47:38 enable = 1 May 10 15:48:09 No such thing as proc/switch/eth0/vlan_enable though May 10 15:48:18 enable_vlan? May 10 15:48:52 Yeah, it's enabled alright (=1) May 10 15:49:00 ok, so far so good May 10 15:49:15 The ports were really off though, right up until you made me set them again May 10 15:49:25 now can you echo "0 1 2 3 4 5" >/proc/switch/eth0/vlan/0/ports May 10 15:49:29 Done May 10 15:49:30 what do you mean? May 10 15:49:32 Well May 10 15:49:48 Right now, after the command, cat /proc/switch/eth0/vlan/0/ports gives : May 10 15:49:49 0 1 2 3 4 5t* May 10 15:49:59 Before, it was giving me : May 10 15:50:00 4 5u May 10 15:50:14 oh, ok, that is expected May 10 15:50:18 When I say "before", I mean, from plain old failsafe, without touching anything May 10 15:50:19 5t* is not so ideal May 10 15:50:22 ok May 10 15:50:38 should be 5u* May 10 15:50:52 oh, can you redo the echo with 5u? May 10 15:50:56 Sure May 10 15:51:29 Ok, got "5u" now in ports too (cat output gives 5u at the end) May 10 15:51:42 ok, now can you do vnconfig 0 0? May 10 15:51:42 not 5u* though May 10 15:51:44 ok May 10 15:51:51 Done May 10 15:52:01 ifconfig , what ifaces are there? May 10 15:52:04 Nothing May 10 15:52:30 after vncconfig 0 0? May 10 15:52:36 Yup May 10 15:52:38 vnconfig I mean May 10 15:52:42 hmmm..... May 10 15:52:48 vconfig ;) May 10 15:53:08 vconfig 0 0 ? May 10 15:53:20 Hm yeah May 10 15:53:24 vnconfig is not a command here May 10 15:53:30 oh, ok May 10 15:53:34 vconfig eth0 add 0 May 10 15:53:36 (no error message, failsafe inside ;p) May 10 15:53:46 jow_laptop: oh yeah that's from your patch May 10 15:54:06 and no interfaces in ifconfig now? May 10 15:54:08 ifconfig still outputs nothing May 10 15:54:10 :/ May 10 15:54:12 btw, I am not sure whether the brcm-2.4 switch can be reconfigured on the fly May 10 15:54:22 Mr_Smoke: ifconfig -a May 10 15:54:37 Ah my bad May 10 15:54:39 ah, so how does one configure the brcm-2.4 switch? May 10 15:54:40 it's been there all the time May 10 15:54:51 no IP setup though May 10 15:54:56 that's normal May 10 15:55:08 what eth? May 10 15:55:11 usually you have to reboot to apply the switch changes May 10 15:55:11 But ifconfig -a would always show eth0 May 10 15:55:12 eth0 May 10 15:55:20 any others? May 10 15:55:28 lo and that's it May 10 15:55:36 ok, then we can't do that May 10 15:56:06 also May 10 15:56:20 If the switch can't be configured on the fly then do so will kill the config that was there before on the next reboot May 10 15:56:22 I noticed that if I tried to bring up eth0 manually with a proper IP, I still couldn't ping my lan May 10 15:56:27 So it does sound like a vlan config indeed May 10 15:57:24 BUT May 10 15:57:30 right now, I *CAN* May 10 15:57:42 So what you made me do actually worked May 10 15:57:47 oh, ok May 10 15:57:59 I think it's to do with the ports May 10 15:58:00 oh, I know May 10 15:58:04 Yes ? May 10 15:58:10 eth0 gets replaced by the vlan eth0 May 10 15:58:49 so there really only is one interface, but now it's the vlan May 10 15:59:24 try echo "1 5u*" >/proc/switch/eth0/vlan/0/ports May 10 15:59:35 Hm ok, I had to reboot in the meantime May 10 15:59:41 oh, ok May 10 15:59:42 cuz I couldn't break out of ping May 10 15:59:51 right, no job control May 10 15:59:54 so just vconfig eth0 add 0 May 10 16:00:01 and then echo "1 5u*" ? May 10 16:00:02 just checked the old code May 10 16:00:03 after echo May 10 16:00:10 ok May 10 16:00:15 echo then vconfig May 10 16:00:15 they didn't mess with vconfig at all May 10 16:00:25 jow_laptop: I know May 10 16:00:36 jow_laptop: but did it work? May 10 16:00:42 cshore: yes May 10 16:00:50 hmmmm..... May 10 16:01:31 now put your network cable on port 1 and configure the interface and try ping (with a count) May 10 16:02:04 ok May 10 16:02:49 Hm port 1 should be the one furthest from the wan port, right ? May 10 16:02:56 well, it varies May 10 16:03:00 Or the second one maybe, since it starts at 0 May 10 16:03:03 Let's try again May 10 16:03:29 Hm, should I be afraid that it cannot ping itself ? May 10 16:03:34 s/afraid/worried/ ? May 10 16:04:18 you mean ping IP-you-configured doesn't work? May 10 16:04:28 Yes May 10 16:04:47 can you ping the network? May 10 16:04:56 No, nothing May 10 16:05:09 tried all the ports May 10 16:05:35 hmmm....try echo 1 >/proc/switch/eth0/reset, then the port echo again May 10 16:05:36 Hm hangon a sec May 10 16:05:59 I didn't issue vconfig eth0 add 0 May 10 16:06:04 I suppose that's not good :) May 10 16:06:12 oh, ok, try that May 10 16:07:21 No luck May 10 16:07:34 pinging the local IP I configured just drops straight back to the CLI May 10 16:07:37 and I can't ping the network May 10 16:07:40 ok, try the switch reset and port echo, and config May 10 16:07:52 vconfig I mean May 10 16:08:17 k May 10 16:09:32 Wo May 10 16:09:38 There's trafic May 10 16:09:47 success :) May 10 16:09:51 ok, yay May 10 16:10:06 now try again with port 2 May 10 16:10:10 ok May 10 16:10:17 so reset, echo port, vconfig, and re ping May 10 16:10:25 yes May 10 16:10:53 the vconfig may not actually be necessary, so next we'll try without that May 10 16:11:31 Hm that's not working May 10 16:11:35 lemme try again May 10 16:12:20 Ok it's good in fact :) May 10 16:12:27 remember different port for the cable May 10 16:12:30 ok, May 10 16:12:34 Yeah, I just switched to the wrong one May 10 16:12:39 I4ve got the ordering right, now May 10 16:12:40 heh May 10 16:12:56 ok, can you reboot and try without the vconfig? May 10 16:12:59 it's 0,1,2,3 4=Wan, when facing the unit, looking from above. May 10 16:13:00 Ok May 10 16:13:08 I think the magic was the reset May 10 16:13:33 I see what you're getting at ... the default would have been the wan port then May 10 16:13:46 ok let's try port 2 again then May 10 16:14:27 It's good :) May 10 16:14:47 boot => failsafe => echo "2 5u*" > /proc/switch/.... => it pings :) May 10 16:14:48 ok, so vconfig isn't necessary, it's the reset? May 10 16:14:57 No, reset isn't necessary either May 10 16:15:02 CAn I try something ? May 10 16:15:12 betcha it is for the second port your try May 10 16:15:16 Since it was set to "4 5u" without touching anything, I'll try the LAN port May 10 16:15:41 ok May 10 16:16:23 Bingo May 10 16:16:37 It worked. I just had to add the LAN IP to eth0 and bring it up May 10 16:16:43 ant it was pingable on the WAN port (port 4) May 10 16:16:54 root@(none):/# cat /proc/switch/eth0/vlan/0/ports May 10 16:16:55 4 5u May 10 16:16:57 ah, I know what's going on May 10 16:17:05 you're trying to use it for failsafe, right? May 10 16:17:12 Yes May 10 16:17:26 preinit_ip_deconfig is probably the problem then May 10 16:17:36 and the fact it's left on 4 5u* May 10 16:17:55 should switch to 0 5u* and configured May 10 16:18:27 so lan port 0 is the port failsafe is done on May 10 16:18:31 Hm May 10 16:18:41 preinit_ip_deconfig is defined *twice* May 10 16:18:55 yes the later one overrides the first May 10 16:18:57 ok May 10 16:19:04 it's arch specific May 10 16:19:12 whereas the first one is for every arch May 10 16:19:18 ok May 10 16:19:31 So the arch specific is bringing everything down May 10 16:19:33 but I mean preinit_net_echo shouldn't call preinit_ip_deconfig May 10 16:19:52 Oh May 10 16:19:57 yeah, I see that now May 10 16:20:07 besides, I see why it's port 4 May 10 16:20:20 right the seq stops at 4 May 10 16:20:24 yup May 10 16:20:30 we need to do one more that resets to 0 May 10 16:20:33 Wouldn't it be better to use all the LAN ports ? May 10 16:20:37 or seq the other way May 10 16:20:45 then you can vlan leakage May 10 16:20:48 Ah May 10 16:20:49 if vlans are defined May 10 16:21:00 yeah i see the remarks above May 10 16:21:08 wasn't there a reset to 0 1 2 3 4in 15_brcm_something ? May 10 16:21:18 jow_laptop: I think so, lemme find it again May 10 16:21:22 that is before this May 10 16:21:29 that's the initial setup May 10 16:21:32 ah May 10 16:21:51 failsafe_ip() May 10 16:22:09 for reference: https://dev.openwrt.org/browser/trunk/target/linux/brcm-2.4/base-files/etc/preinit.arch?rev=18253 May 10 16:22:30 cshore: so you're saying preinit_net_echo should not contain a call to preinit_ip_deconfig then May 10 16:22:37 and also, it should reset to port 0 maybe ? May 10 16:22:48 right May 10 16:23:14 we need to make sure there is a deconfig in the end of preinit though May 10 16:23:16 Why not use all ports ? After all, it's failsafe, better use whatever's available, right ? May 10 16:23:31 vlan leakage May 10 16:23:39 package from vlan 0 to vlan 1 etc May 10 16:23:55 hm 'k May 10 16:24:21 so, if I remove the call to preinit_ip_deconfig, and then echo "0 5u" > /proc/blah ... it should be good May 10 16:24:27 right May 10 16:24:40 and what about that final call at the end of preinit ? May 10 16:24:49 check there is a preinit_ip_deconfig in /lib/preinit May 10 16:25:29 I think other archs do deconfig in 90 something May 10 16:25:50 grep deconfig /lib/preinit/* May 10 16:26:02 there's one in 99_10_run_init May 10 16:26:09 ok, then we're fine May 10 16:26:18 Ok May 10 16:26:24 Rebooting with these mods May 10 16:26:40 I think in failsafe mode, all ports should be in the vlan May 10 16:26:54 Hooray May 10 16:26:57 on some devices the port order is reversed May 10 16:27:18 jow_laptop: I would think so too, but I must admit I don't understand that vlan leakage thing May 10 16:27:19 jow_laptop: what about vlan leakage? We already have bug reports about that on other archs May 10 16:27:19 so you never know whether to connect on lan1 or lan4 May 10 16:27:57 cshore: well if it only happens due to user request when actually entering failsafe I don't see the issue May 10 16:28:00 But so far, it works : I now have eth0 up, using 192.168.1.1 (nvram was reset so that's quite ok), and pinging on port 0 May 10 16:28:37 hmmm.....maybe we need to do a config in the failsafe hook then May 10 16:28:57 for preinit we keep the vlans separate and for failsafe we combine them May 10 16:29:03 Well, why not set all ports just after/instead-of the for loop ? May 10 16:29:22 because then it happens on every boot May 10 16:29:27 that there is vlan leakage May 10 16:29:54 if you do it in failsafe then it's only when you delibarately enter failsafe May 10 16:30:12 Ok, so failsafe needs its own net_echo/deconfig functions then May 10 16:30:36 not net_echo, but port config, yeah May 10 16:31:03 bbl May 10 16:32:10 Well, that was fun May 10 16:32:38 Just a sec, could I get you to help me with the failsafe....I just am writing a quick script for you to try May 10 16:32:44 Of course May 10 16:32:59 Just in case I fail to answer, I have an errand to run, but I'll be back soon May 10 16:33:16 ok, I need a couple of minutes anyway May 10 16:33:35 Ok then I'm running away for 5 May 10 16:43:53 Mr_Smoke: Could you try this: http://openwrt.pastebin.com/cAyCN9rK as 05_failsafe_config_net_brcm? May 10 16:45:19 cshore: do you want me to undo the changes I made to 15_set_preinit_interface_brcm for that ? May 10 16:45:28 no May 10 16:45:35 in addition to May 10 16:45:41 Ok sure May 10 16:48:34 Hm May 10 16:48:57 I must have done something wrong, ports are still set to 0 5u as per the modification to 15_set_preinit_interface_brcm May 10 16:49:40 hmmm....you said reset wasn't necessary, right? May 10 16:49:59 or was it? May 10 16:50:04 It was not, indeed May 10 16:50:12 I'll throw an "echo" in there, see if it runs May 10 16:51:37 cshore: it's called, but there are calls to preinit_net_echo later ... order problem maybe ? May 10 16:52:29 do they call the 05_failsafe, preinit_echo, or the one in 15_set_preinit? May 10 16:53:02 oh!, I see May 10 16:53:12 I was going to determine that by means of another echo May 10 16:53:55 doh! 05_ will always be overwritten, because it is sourced first May 10 16:53:58 15_set_preinit is called just before I get the console May 10 16:54:02 yeah, looks like it May 10 16:54:21 rename it to 16_ instead ? May 10 16:54:55 craigc * r21416 /packages/net/multiwan/ (Makefile files/usr/bin/multiwan): [packages] multiwan: Fixed last WAN probability to be 1.0. May 10 16:56:58 http://openwrt.pastebin.com/iRtxuYkZ as 20_failsafe_net_echo, and remove preinit_net_echo from 05, please May 10 16:57:44 ok May 10 16:58:17 No hook in there ? just sourced I assume ? May 10 16:58:30 yep May 10 16:58:56 ok May 10 16:59:00 initdate_failsafe is defined elsewhere and hooked, this just overrides the definition May 10 16:59:14 Right May 10 16:59:17 Rebooting May 10 16:59:21 *indicate May 10 17:00:00 Looks better now, let's try pinging any port May 10 17:00:09 cshore: congrats, it's working :) May 10 17:00:20 yay, thank you for helping May 10 17:00:36 Don't mention it, it was a fun thing to do :) May 10 17:00:44 And now I know a little more about OpenWRT's init/failsafe :) May 10 17:00:52 :) May 10 17:01:01 I take it you're going to submit a patch ? :) May 10 17:01:10 commit a change, yep May 10 17:01:17 oh even better May 10 17:01:21 I'll be looking out for that one May 10 17:01:40 Do you think I could pester you some more with a wireless issue, or is it not your "field" ? May 10 17:02:25 Could you look at http://openwrt.pastebin.com/k3W0xCrx and tell me if that is like your changes to 15_set_preinit_interface? May 10 17:02:40 I could try, but I don't know much about wireless May 10 17:03:14 ok May 10 17:04:38 cshore: the only difference is that I added -> echo "0 5u*" > /proc/switch/eth0/vlan/0/ports <- after line 32, but I think we can do without it, since it's overriden either by init or by failsafe now May 10 17:05:01 Oh wait May 10 17:05:07 I see it now May 10 17:06:00 Yeah it's a bit different, but I didn't know exactly where to set port 0, so yours is probably more sensible May 10 17:07:21 no you're right, mine is wrong....it would affect eth1 devices too May 10 17:07:36 Ah May 10 17:07:54 well mine was at line 3 May 10 17:07:56 32* May 10 17:08:11 After the for loop, obviously May 10 17:08:17 yeah, right after the loop May 10 17:08:30 Ok, gotta go again, but I'll be back May 10 17:08:38 ok May 10 17:10:50 cshore * r21417 /trunk/ (4 files in 2 dirs): brcm-2.4: Fixed preinit and failsafe switch configuration May 10 17:52:47 cshore, ping May 10 17:54:04 mirko * r21418 /packages/libs/libdirectfb/Makefile: DirectFB: change package name as upper-case characters in package names result in problems with opkg May 10 17:54:09 mirko * r21419 /packages/ (4 files in 4 dirs): adjust dependencies regarding previous name change "DirectFB" -> "directfb" May 10 17:54:27 cshore, why did you alter trunk/.gitignore in 57ed6332f33d95b4bc0f5afb71d739e548ce2818 ? any special reason? May 10 17:54:44 doh! May 10 17:54:57 no, error May 10 17:55:14 how do I revert that May 10 17:55:50 i'm using git - so i would just checkout this file in a prior version May 10 17:56:14 and commit it once again - partial reverts are not possible in svn afaik May 10 17:56:26 they are May 10 17:56:30 uh, ok May 10 17:56:32 you can just do a reverse merge May 10 17:56:58 xMff: how do I do that? May 10 17:57:29 svn merge -r HEAD:21416 .gitignore May 10 17:58:07 xMff, by the way - i'm looking for a commit a long time ago... - it removed checks for upper-case characters in package-names while creating ipk's... - did you so (I'm not sure any more who committed)? May 10 17:58:32 nope May 10 17:58:37 xMff, scrolled through the log several times, however didn't find it anymore... May 10 17:58:39 ok May 10 17:58:46 nbd, ping May 10 17:58:53 then it was nbd afaik May 10 17:59:07 cshore * r21420 /trunk/.gitignore: .gitignore: Undo accidental commit. May 10 17:59:29 xMff, will you be in berlin this weekend? wireless community weekend? May 10 17:59:33 yes May 10 17:59:49 ok, done May 10 18:02:38 cshore, thanks :) May 10 18:16:04 cshore: the wireless issue I was referring to is what I believe led to this : https://forum.openwrt.org/viewtopic.php?id=23998 May 10 18:16:18 In fact, WDS doesn't work unless there is a matching ap definition in /e/c/wireless May 10 18:16:40 I think we tracked it down one day to a broadcom script that does not consider wds as a vif ... i'll lok again May 10 18:16:44 look* May 10 18:17:48 That's it : /lib/wifi/broadcom.sh, line 44 May 10 18:18:06 I was told (can't remember by whom right now) that it *might* be the cause May 10 18:19:07 jow_laptop: it was you actually :) May 10 18:30:20 ah, I see May 10 18:30:31 IS the website under attack or something ? May 10 18:31:06 It's really slow all of a sudden May 10 18:31:19 cshore: the wireless thing ? you are about to make my day then ;) May 10 18:32:21 can wds have more than one ssid, or if there is one wds is it the only one? May 10 18:33:09 I would think it should have only one SSID, but I cannot say for sure May 10 18:34:58 After line 31, add wds_if="$vif" May 10 18:35:55 and at line 44 add $wds_if to the variable (after $mon_if) May 10 18:36:08 also line 48 May 10 18:37:03 Ok, I'll test that, maybe not tonight though May 10 18:37:13 But that's what jow was pointing at indeed May 10 18:37:37 I don't know if it will work, but that's my best guess May 10 18:37:49 Hm May 10 18:38:09 Oh darn, my buildroot is probably set for target backfire, not trunk May 10 18:38:14 Odd May 10 18:38:16 nbd would be better May 10 18:38:30 I'm getting "At revision 21420" but I don't see your commit from a few hours ago May 10 18:39:09 hmmm...probably backfire then May 10 18:39:29 svn info? May 10 18:39:33 I honestly suck at cvs/svn/git May 10 18:39:37 yeah, backfire May 10 18:39:48 branches/backfire May 10 18:40:01 Mr_Smoke: me too....just enough to get by for now May 10 18:40:20 But, the changes you committed are in trunk ? or backfire ? May 10 18:40:26 trunk May 10 18:40:45 O May 10 18:40:46 Ok. May 10 18:40:51 So I would need to move to trunk then May 10 18:40:56 yes May 10 18:40:59 maybe it should be backported.. May 10 18:41:01 I'll create a separate buildroot dir May 10 18:41:08 Well, maybe yeah May 10 18:41:12 IMHO failsafe is quite important May 10 18:41:22 And failsafe with no network is pretty bad May 10 18:41:24 I'll let the senior devs decide that May 10 18:41:30 Ok :) May 10 18:42:03 checking out trunk for now May 10 18:42:20 and failsafe with network on 2.6 works fine? May 10 18:42:32 hmmm, how to send a patch >100kb? May 10 18:42:44 Hm no idea, I haven't tried 2.6 in ages May 10 18:42:59 I could give it a go if you want May 10 18:43:08 Probably not tonight though May 10 18:43:26 I'm getting sighs from the missus ;) May 10 18:43:34 Iam3r: mailinglist May 10 18:43:42 Iam3r: what's your patch? May 10 18:44:50 Iam3r or ticket: File (size limit 2 MB): May 10 18:45:12 uboot for ar71xx devices (just for tzhe nbg460n as second stage ootloader so far, as i only have that one, and it doesn'T have jtag) May 10 18:46:21 Ok, got to go. May 10 18:46:51 cshore: many thanks for your help, I'll get back to you and/or n.bd on the wds issue and 2.6 failsafe when I can (probably thurdsay/friday) May 10 18:47:34 ok, thanks May 10 18:49:55 crow: ok, but if mailinglist is ok, i use that. I have some more patches depending on that one and i think it is better to post them in relation. I just wondered if it is better to zip it or to send it plain text May 10 18:50:32 100kb is not that big May 10 18:51:33 lam3r: easier to read a patch that's in plaintext (but attachments are better than inline if your MUA mangles patches) May 10 18:52:02 lam3r: especially whitespace May 10 18:52:27 rtz2: ok May 10 18:53:39 cshore: ok, i try to send it plain text (kind if hard work with firefox, but i did it before) May 10 18:59:08 i meant thunderbird... May 10 18:59:36 jow_laptop: ping May 10 19:14:56 cisco uses strange ways of supplying requested source code, using file hosting services May 10 19:17:28 lol May 10 19:17:34 KanjiMonster: what did you request? May 10 19:17:52 rtz2: linksys e2100l sources May 10 19:20:21 I would have thought they put it on their gpl page, but no, a somewhat fishy file hoster (which wants an email address for supplying the actual download link) May 10 19:21:31 which fishy file hoster? May 10 19:21:59 sounds like they told some sw devel to give the darned hippies their sources May 10 19:23:37 noz: pong May 10 19:24:06 maligor: filehosting.org May 10 19:26:28 what was the name of the file to put in ~/.openwrt/ to include the default settings? I cant remember... May 10 19:27:23 I thought it was default.config, but I'm not having any luck May 10 19:28:11 jow * r21421 /branches/backfire/target/linux/brcm-2.4/base-files/lib/preinit/ (3 files): [backfire] merge r21417 May 10 19:31:39 xMff: Thinking about #7030, and the r21112 update you did to branches/backfire, porting across the trunk changes r20679-r20685. Would it be worth bumping up the package revision number so it can be installed as an upgrade? May 10 19:32:08 * xMff looks May 10 19:32:25 ah May 10 19:32:29 surwe May 10 19:33:58 jow * r21422 /trunk/package/hotplug2/Makefile: [package] hotplug2: bump pkg release after hotplug fixes in r20679-r20685 May 10 19:34:39 jow * r21423 /branches/backfire/package/hotplug2/Makefile: [backfire] merge r21422 May 10 20:02:38 xMff: ping May 10 20:02:47 whats the minimum key size for psk2 May 10 20:03:21 8 May 10 20:04:29 byte May 10 20:05:32 xMff: so 846E7A3C3 should be a valid key May 10 20:05:44 yeah May 10 20:05:49 doesnt work May 10 20:05:49 yes May 10 20:06:02 well at least im told it doesnt work May 10 20:06:17 told by who/what? May 10 20:06:28 sounds suspiciously hex encoded May 10 20:06:34 11 bytes May 10 20:06:42 well they say was originall a wep key May 10 20:06:59 too short for that May 10 20:07:08 so im thinking they used some passphrase which generated this May 10 20:07:41 ok ill drive by there and validate what im being told May 10 20:08:13 lastly why would 2 5 ghz radios sudeently stop communicating after 2 weeks is the next problem May 10 20:12:27 nbd: any thoughts, ath2k gear 5ghz Ubiquiti boards, 2 weeks in the field, repeaters stop communication May 10 20:13:03 though they run fine here in the lab for months May 10 20:13:11 i was thinking environmental May 10 20:33:50 xMff: ok heres one for you, the key is 846E7A3C3 but should have been C846E7A3C3 , however even using the C846E7A3C3 itll let you connect and not tell you the key is wrong. gives you an ip and all May 10 20:34:48 so on the AP they was was set for 846E7A3C3, the client was using C846E7A3C3 and cpuld connect, and get an ip, though couldnt surf May 10 20:35:23 and no error messages, i would have though windows and linux clients would have gotten an error stateing key mismatch May 10 20:49:34 ok May 10 20:50:09 I officially delay the gcc 4.5 project for the next two months May 10 20:50:32 .0 is simply too buggy May 11 01:27:28 craigc * r21424 /packages/net/multiwan/ (Makefile files/usr/bin/multiwan): [packages] multiwan: updated load balancer algorithm May 11 02:44:47 build #48 of iop32x is complete: Exception [exception] Build details are at http://tksite.gotdns.org:8010/builders/iop32x/builds/48 May 11 02:57:09 build #48 of adm5120 is complete: Exception [exception] Build details are at http://tksite.gotdns.org:8010/builders/adm5120/builds/48 **** ENDING LOGGING AT Tue May 11 02:59:56 2010