**** BEGIN LOGGING AT Sat Dec 17 02:59:57 2011
Dec 17 04:26:02 <builder_> build #117 of at91 is complete: Failure [failed compile_4]  Build details are at http://buildbot.openwrt.org:8010/builders/at91/builds/117
Dec 17 04:41:43 <builder_> build #115 of ubicom32 is complete: Failure [failed compile_3]  Build details are at http://buildbot.openwrt.org:8010/builders/ubicom32/builds/115
Dec 17 05:02:06 <philipp64|laptop> hmmm.... do I need to do anything to create a vlan interface? I have: http://fpaste.org/HptL/  but this seems to be incorrect.
Dec 17 05:06:52 <roh> philipp64|laptop: nope.. but you need to config the switch accordingly
Dec 17 05:07:31 <philipp64|laptop> switch? no, I just want to have a trunked interface (not an access untagged interface)
Dec 17 05:07:48 <philipp64|laptop> my box doesn't have 'switch' hardware.
Dec 17 05:09:51 <roh> philipp64|laptop: http://fpaste.org/GaJX/
Dec 17 05:09:58 <roh> thats what works for me.
Dec 17 05:11:12 <roh> the t marks the tagged ports.. i guess if you have a real iface which has no switch what you tried should work.
Dec 17 05:11:21 <roh> if the hw can handle the bigger mtu
Dec 17 05:12:15 <roh> what kind of hw do you have on that and the other end?
Dec 17 05:13:14 <philipp64|laptop> it's an Alix 6F2 with via_rhine ethernet ports. it's talking to a Cisco SFP300 switch.
Dec 17 05:13:56 <roh> i see
Dec 17 05:14:37 <roh> can you exchange both devices by a linux notebook with some intel nic and try it 'dry' ?
Dec 17 05:14:55 <roh> maybe you need to set some ios tweak to make it speak vlan to linux.
Dec 17 05:15:48 <roh> ! Little note about VLAN 1. In Cisco systems it is the default VLAN
Dec 17 05:15:48 <roh> so you MUST start using vlans from 2.
Dec 17 05:15:55 <roh> http://www.candelatech.com/~greear/vlan/cisco_howto.html *g*
Dec 17 05:16:43 <philipp64|laptop> I have another Alix box running Astlinux right now with vlans and it works fine.
Dec 17 05:17:22 <roh> well.. just try some vlan above 1  and see if that works
Dec 17 05:25:27 <philipp64|laptop> seems the /lib/network/config.sh script is missing "ip link set vlanN name ethX.N" ...
Dec 17 05:26:12 <roh> uhm. worked for me.. weird
Dec 17 05:26:53 <jow_laptop> philipp64|laptop: using the vlan notation in ifname is enough
Dec 17 05:27:07 <roh> check add_vlan()
Dec 17 05:27:09 <jow_laptop> prepare_interface() will call vconfig then to spawn the vlan
Dec 17 05:27:48 <jow_laptop> the reason why you iface isn't brought up is the missing ip address
Dec 17 05:27:57 <philipp64|laptop> I think the problem is that we're not doing "vconfig set_name_type DEV_PLUS_VID" at the top.
Dec 17 05:28:00 <jow_laptop> not the vlan setup
Dec 17 05:28:14 <jow_laptop> you're barking up the wrong tree
Dec 17 05:28:29 <philipp64|laptop> or that the busybox default is braindead.
Dec 17 05:29:20 <philipp64|laptop> I tried using vconfig to manually create the interface, and "vconfig add eth0 1" results in "vlan1@eth0" as the name according to "ip link show".
Dec 17 05:29:35 <jow_laptop> again, the issue is not ip link or vconfig
Dec 17 05:29:43 <jow_laptop> your config is simply incomplete
Dec 17 05:29:49 <jow_laptop> add an ip address and it will work ;)
Dec 17 05:32:36 <philipp64|laptop> jow_laptop: the line got trimmed, but it was there...
Dec 17 05:33:00 <philipp64|laptop> we definitely need: vconfig set_name_type DEV_PLUS_VID
Dec 17 05:33:02 <jow_laptop> oh and another problem
Dec 17 05:33:12 <jow_laptop> software vlans and bridges do not play well together
Dec 17 05:33:22 <jow_laptop> stuff eth0 into the bridge and create a br-lan.1
Dec 17 05:33:37 <jow_laptop> or simply disable the bridge
Dec 17 05:34:17 <jow_laptop> the vconfig name type call happens in /etc/init.d/boot
Dec 17 05:34:40 <roh> jow_laptop: eh. sure? i use that
Dec 17 05:35:22 <russell--> hmm.  i have a serial console on the airrouter now
Dec 17 05:35:30 <jow_laptop> roh: it works with switch hardware, but not with software vlans
Dec 17 05:35:36 <roh> ah
Dec 17 05:35:43 <jow_laptop> at least not here, on any unit I tried
Dec 17 05:36:41 <russell--> r29289 boots, r29557 doesn't
Dec 17 05:37:28 <philipp64|laptop> this is why I hate busybox... http://fpaste.org/MHOr/
Dec 17 05:37:51 <philipp64|laptop> typing "vconfig set_name_type" with no argument is *supposed* to give the current naming mode.
Dec 17 05:38:07 <philipp64|laptop> even the usage help says so.
Dec 17 05:38:13 <roh> *g* .. well.. feel free to fix it
Dec 17 05:38:19 <jow_laptop> philipp64|laptop: please, you're not the first one who tries vlans on openwrt, we already thought about setting the right naming schema
Dec 17 05:38:41 <jow_laptop> as I said, it happens in /etc/init.d/boot and it is very unlikely to be your problem
Dec 17 05:38:55 <philipp64|laptop> so I can't have a trunked interface in a bridge with wifi?
Dec 17 05:39:02 <jow_laptop> apparently not
Dec 17 05:39:05 <philipp64|laptop> and have the wifi be untrunked?
Dec 17 05:39:15 <philipp64|laptop> hmmm.... that's weird. works fine on Astlinux.
Dec 17 05:39:23 <jow_laptop> then compare the sysctls
Dec 17 05:40:33 <philipp64|laptop> http://fpaste.org/rbbE/
Dec 17 05:41:42 <jow_laptop> kernel version?
Dec 17 05:42:07 <russell--> [    0.290000] VFS: Cannot open root device "(null)" or unknown-block(0,0)
Dec 17 05:42:10 <russell--> [    0.300000] Please append a correct "root=" boot option; here are the available partitions:
Dec 17 05:42:13 <russell--> [    0.310000] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(0,0)
Dec 17 05:43:17 <philipp64|laptop> 2.6.27.7
Dec 17 05:43:43 <jow_laptop> well, try with 32+
Dec 17 05:43:46 <philipp64|laptop> BusyBox v1.12.4
Dec 17 05:45:19 <jow_laptop> did you ensure that on openwrt the eth0 itself is not member of another bridge? This will break vlans as well
Dec 17 05:45:40 <russell--> r29293 nano factory image works too
Dec 17 05:46:48 <philipp64|laptop> it's not.
Dec 17 05:48:01 <jow_laptop> http://marc.info/?l=linux-netdev&m=132007004619404&w=2
Dec 17 05:48:03 <russell--> lol, the current nano image works too, so it's something about the airrouter image
Dec 17 05:48:21 <jow_laptop> russell--: juhosg redid large chunks of the model detection lately
Dec 17 05:50:04 <russell--> [    0.000000] MIPS: no machine found for id 'UBNT-AR', supported machines: [...]
Dec 17 05:51:08 <russell--> the working version says "[    0.080000] MIPS: machine is Ubiquiti Nanostation M
Dec 17 05:51:11 <jow_laptop> russell--: did you compile yourself?
Dec 17 05:51:12 <russell--> "
Dec 17 05:51:14 <russell--> yeah
Dec 17 05:51:21 <jow_laptop> just with svn up && make ?
Dec 17 05:51:40 <russell--> the git equivalents, more or less
Dec 17 05:51:46 <jow_laptop> that will not copy updated mach files into the linux build tree
Dec 17 05:51:53 <russell--> ah
Dec 17 05:52:00 <jow_laptop> run make target/linux/clean world and test that image
Dec 17 05:52:03 <jow_laptop> I bet it'll work
Dec 17 05:52:08 <russell--> okay, thanks!
Dec 17 05:52:09 <philipp64|laptop> jow_laptop: how does busybox retain the state from one call to "vconfig set_name_type ..." to the next "vconfig add eth0 1" ???
Dec 17 05:52:34 <jow_laptop> philipp64|laptop: its kept in the kernel vlan code, that command is basically just a sysctl wrapper
Dec 17 05:53:28 <philipp64|laptop> the only explanation of what I saw was that the "vconfig set_name_type" had been skipped in /etc/init.d/boot the first time, and that the interfaces were being named wrong.
Dec 17 05:53:53 <jow_laptop> no, I believe you're hitting actual kernel bugs or bridge code limitations
Dec 17 05:53:57 <philipp64|laptop> because I did a "vconfig set_name_type DEV_PLUS_VID_NO_PAD" manually and rebooted and everything came up as it should.
Dec 17 05:54:11 <philipp64|laptop> weird persistence.
Dec 17 05:54:38 <jow_laptop> I assume correlation
Dec 17 05:55:46 <philipp64|laptop> well, I could burn a new image and reboot, but without sysupgrade for the x86 pulling CF's from the alix is a real pain. you have to unscrew the board from the base...
Dec 17 05:55:50 <philipp64|laptop> grrrrr...
Dec 17 05:56:47 <philipp64|laptop> well, I'm going to hook it up to my switch and see if it works in the wild...
Dec 17 06:09:45 <philipp64|laptop> jow_laptop: well, bridging wlan0 and eth0.1 seems to work fine....
Dec 17 06:11:06 <philipp64|laptop> btw, when I ssh in, I'm seeing:
Dec 17 06:11:09 <philipp64|laptop> -ash: tty: not found
Dec 17 06:11:10 <philipp64|laptop> ash: /dev/console: unknown operand
Dec 17 06:11:15 <philipp64|laptop> is this a known bug?
Dec 17 06:11:57 <philipp64|laptop> I'm using dropbear, but couldn't figure out how it creates tty sessions...
Dec 17 06:13:22 <philipp64|laptop> so I didn't get very far debugging it.
Dec 17 06:39:11 <russell--> jow_laptop++
Dec 17 06:45:19 <russell--> https://plus.google.com/u/0/114541658134976985870/posts/XA8rywK24gt
Dec 17 06:45:46 <russell--> just to document the serial pinout
Dec 17 07:06:47 <builder_> build #80 of mpc52xx is complete: Failure [failed compile_10]  Build details are at http://buildbot.openwrt.org:8010/builders/mpc52xx/builds/80
Dec 17 12:18:18 <karlp> jow_laptop: reading and removing the uci-defaults scripts in post-inst is just so that you don't need to reboot right?
Dec 17 12:18:55 <karlp> any files in ucidefaults will get read and removed automatically every boot without me adding anything extra to the package makefiles, beyond copying the file to uci-defaults in the install step right?
Dec 17 12:58:58 <karlp> jow_laptop: ok, I've got my postinst and defaults scripts to create a ucitrack entry,
Dec 17 12:59:24 <karlp> but save and apply on the luci page now saves and applys into the config file, but the service isn't restarted.
Dec 17 12:59:36 <karlp> I've double checked the syntax and naming.
Dec 17 12:59:42 <karlp> there's nothing in logread -f
Dec 17 12:59:53 <karlp> (except my own service logs, showing that it doesn't get restarted)
Dec 17 13:04:35 <mbaba> Hi. I need to call a new shell code from net/mac80211/status.c
Dec 17 13:04:59 <mbaba> this shell code do some operations on router
Dec 17 13:15:14 <nbd> wow, that sounds like a *really* ugly hack
Dec 17 13:37:26 <stintel> nbd: sorry for late reply, but your ath9k patch seems to fix my oops
Dec 17 13:37:30 <stintel> thanks again
Dec 17 13:43:26 <stintel> jow_laptop: are you also going to factor the tx power offset into the "iwinfo wlanX info" command ?
Dec 17 14:33:42 <CIA-79> nbd * r29559 /trunk/package/hostapd/patches/760-fix_wds_sta_check.patch: hostapd: add missing check for wds stations
Dec 17 14:33:47 <CIA-79> nbd * r29560 /trunk/package/mac80211/patches/565-ath9k_fix_lockup.patch: ath9k: fix a locking issue in the tx path
Dec 17 15:21:29 <jow_laptop> karlp: see if  luci-reload name  works
Dec 17 15:21:42 <jow_laptop> stintel: will do
Dec 17 15:25:02 <karlp> jow_laptop: I'll give that a go monday morning.
Dec 17 15:25:14 <jow_laptop> karlp: ok
Dec 17 15:25:31 <karlp> that should either work, or tell me more about why it couldn't?
Dec 17 15:25:40 <jow_laptop> it should work
Dec 17 15:26:06 <karlp> when I removed my on_commit hook, luci behaved much better, (I didn't get any raw html back) but the service wasn't restarted.
Dec 17 15:26:16 <karlp> anyway, monday.  I'm about out of here for today
Dec 17 15:27:06 <jow_laptop> if you get raw html back then something is wrong
Dec 17 15:27:32 <karlp> I get that back if I use on_commit... os.excute
Dec 17 15:27:52 <jow_laptop> yeah, you need to move stdout to /dev/null
Dec 17 15:28:02 <karlp> oh, duh.
Dec 17 15:28:07 <karlp> I should have known that.
Dec 17 15:28:13 <karlp> printing text before the content type.
Dec 17 15:28:18 <jow_laptop> luci-reload will call /etc/init.d/foo reload
Dec 17 15:28:31 <karlp> if reload isn't defined, will that call restart?
Dec 17 15:28:41 <jow_laptop> afair yes, need to check
Dec 17 15:28:47 <karlp> the default restart just calls stop, then start,
Dec 17 15:28:57 <jow_laptop> nope
Dec 17 15:29:11 <karlp> oh, then that's probably all it is.
Dec 17 15:29:11 <jow_laptop> reload will just return 1
Dec 17 15:29:31 <jow_laptop> hm
Dec 17 15:29:41 <karlp> so I can either make my app properly understand "reload", or add a reload step to my init.d that just calls restart.
Dec 17 15:30:03 <jow_laptop> yeah
Dec 17 15:30:29 <karlp> ok, well that seems like a pretty 100% explanation.
Dec 17 15:30:35 <karlp> I'll just add a reload step for now,
Dec 17 15:30:47 <karlp> reloading is on the list of things to do, but restarting is good enough for now for me.
Dec 17 15:32:10 <jow_laptop> ah no actually it does  [ "$action" = "reload" ] && action='eval reload "$@" || restart "$@" && :'
Dec 17 15:32:17 <jow_laptop> (rc.common)
Dec 17 15:38:41 <karlp> umm, my brane is full, will that do restart if there is no reload method, but reload and only reload if reload is defined?
Dec 17 15:38:59 <karlp> well, things to try on monday anyway.
Dec 17 15:42:14 <jow_laptop> yes, it will fallback to restart if reload fails
Dec 17 15:42:31 <jow_laptop> reload is implemented as "return 1" by default, so always fails
Dec 17 16:05:13 <Olipro> jow_laptop: how would I prevent a make install procedure of a package from resulting in the files being put into the ipkg-install dir
Dec 17 16:06:04 <Olipro> the new ISC DHCPd relies on a make process for a bunch of export libs which then performs a make install to a dir within the source project folder, but using the OpenWRT environment results in everything ending up inside the ipkg-install dir
Dec 17 16:08:55 <builder_> build #118 of s3c24xx is complete: Failure [failed compile_10]  Build details are at http://buildbot.openwrt.org:8010/builders/s3c24xx/builds/118
Dec 17 16:27:04 <Olipro> ok, you override DESTDIR
Dec 17 16:27:07 <Olipro> lovely
Dec 17 16:41:17 <builder_> build #95 of avr32 is complete: Failure [failed compile_4]  Build details are at http://buildbot.openwrt.org:8010/builders/avr32/builds/95
Dec 17 16:49:13 <mbaba> I need to do some commands on the router such as uci set ..  and I wrote it into .sh file
Dec 17 16:49:30 <mbaba> and I need to call this file from c code
Dec 17 16:49:47 <OutBackDingo> mbaba, bad idea
Dec 17 16:50:14 <mbaba> what can i do
Dec 17 16:50:18 <OutBackDingo> write a patch for the application you want to make the calls from
Dec 17 16:50:22 <OutBackDingo> in C
Dec 17 16:51:44 <mbaba> the problem is that I need to call a function in nl80211.c but I can't prepare the its parameter
Dec 17 16:53:25 <OutBackDingo> mbaba, ok explain what your trying to do
Dec 17 16:53:27 <mbaba> so I think I may write a shell code to do the corresponding function using the uci command
Dec 17 16:53:32 <mbaba> ok
Dec 17 16:56:18 <mbaba> I need to call __nl80211_set_channel which responsible of changing the channel(I think) from another function but I can't get the wdev and rdev to send them to this function
Dec 17 16:57:54 <nbd> mbaba: no, the problem is that apparently the idea what you want to do is flawed and you need to rethink your approach
Dec 17 16:58:25 <mbaba> what do you think
Dec 17 16:59:12 <nbd> calling user space code from the kernel like that is wrong on many levels
Dec 17 16:59:44 <OutBackDingo> especially exec'ing it to a shell
Dec 17 17:01:03 <mbaba> so what I can do instead of that
Dec 17 17:01:45 <nbd> first explain what you want to do without talking about any implementation details
Dec 17 17:02:31 <mbaba> I need to change the channel of the wlan according to some condition
Dec 17 17:02:39 <OutBackDingo> whats the condition
Dec 17 17:03:26 <mbaba> according to if a packet failed to send and dropped
Dec 17 17:04:04 <nbd> what mode are you operating in?
Dec 17 17:04:27 <mbaba> infrastructure
Dec 17 17:04:42 <mbaba> my router is as access point
Dec 17 17:04:54 <nbd> ok, so all channel changes need to come from hostapd (on the ap side) or wpa_supplicant (on the client side)
Dec 17 17:05:30 <nbd> so you should ignore the kernel side for now (aside from ensuring that you can read the relevant counters from user space)
Dec 17 17:05:39 <nbd> and focus on integrating your changes into hostapd
Dec 17 17:06:34 <nbd> with mac80211, it's not possible to have the wireless stack change the channel properly without going through hostapd
Dec 17 17:07:04 <mbaba> you mean hostapd package
Dec 17 17:07:08 <mbaba> or what
Dec 17 17:07:25 <nbd> you cannot run an AP interface without having a running hostapd instance
Dec 17 17:07:32 <OutBackDingo> be interesting to know at what state you detect a packet failed to send / drop then attempt a channel change
Dec 17 17:07:46 <nbd> yeah, i still doubt that this is useful in any way
Dec 17 17:08:30 <nbd> since failed packets can have so many different causes
Dec 17 17:09:08 <OutBackDingo> nbd, like no neighbors / association
Dec 17 17:09:35 <OutBackDingo> sounds like he might be trying to automate channel hunting
Dec 17 17:09:53 <nbd> hostapd already has some code for that
Dec 17 17:10:02 <OutBackDingo> yupp
Dec 17 17:10:04 <nbd> it just need to be changed to run some tests while the interface is up
Dec 17 17:10:24 <mbaba> I found a counter at net/mac80211/status.c which called dot11failedcount
Dec 17 17:10:51 <nbd> mbaba: why do you think it's a good idea to use that as a trigger for channel changes?
Dec 17 17:13:37 <mbaba> yes I need to make an automatic channel switiching
Dec 17 17:13:37 <mbaba> called dynamic channel allocation
Dec 17 17:13:51 <nbd> "why" is not a yes or no question
Dec 17 17:13:57 <OutBackDingo> AKA autochannel
Dec 17 17:14:16 <nbd> i'd like to know why you consider failed transmission useful for dynamic channel selection
Dec 17 17:14:24 <nbd> because i think that as an indicator it's worthless
Dec 17 17:19:47 <OutBackDingo> nbd, which gets me to wonder, if an interface is in monitor mode utlizing packet injection and doesnt have an ssid what would be the smartest way to scan for other nodes and set the chanell to match
Dec 17 17:21:06 <nbd> what kind of nodes?
Dec 17 17:22:18 <OutBackDingo> nbd, mesh based our config is http://pastebin.com/5CeeKU64
Dec 17 17:22:40 <OutBackDingo> notice theres no adhoc/ahdemo
Dec 17 17:22:54 <nbd> scan for APs using an additional client mode interface
Dec 17 17:23:31 <nbd> you could change hostapd to inject a custom IE
Dec 17 17:23:36 <OutBackDingo> nbd, ok... now how?
Dec 17 17:24:30 <nbd> iw phy0 interface add sta0 type managed; ifconfig sta0 hw ether <some_unique_mac_address> up; iw sta0 scan
Dec 17 17:25:04 <nbd> a scan is fast enough that clients will typically not lose their connection to the AP
Dec 17 17:25:16 <nbd> if they're transferring data, they'll get some slight packet loss
Dec 17 17:25:26 <nbd> but TCP easily recovers from that anyway
Dec 17 17:29:36 <OutBackDingo> interesting
Dec 17 17:31:28 <OutBackDingo> then parse the scan results for unique mac, set channel
Dec 17 17:33:12 <OutBackDingo> nbd, so could i simply add a phy0 mode managed with bssid to the default config
Dec 17 17:33:54 <nbd> no, because right now there's no way to specify an interface that should not attempt to connect to anything
Dec 17 17:35:35 <OutBackDingo> and i cant add a bssid to the monitor interface
Dec 17 17:35:45 <nbd> a monitor interface can't scan
Dec 17 17:35:55 <OutBackDingo> so i would still need other interfaces to broadcast something
Dec 17 17:36:05 <nbd> yes
Dec 17 17:36:10 <nbd> unless you come up with something else
Dec 17 17:36:29 <nbd> but i'd suggest doing something simple first
Dec 17 17:36:32 <OutBackDingo> can a monmitor interface broadcast a bssid ?
Dec 17 17:36:39 <nbd> creating an algorithm to allow large networks to converge is not exactly trivial
Dec 17 17:37:06 <nbd> especially if at some point it gets split into two big clusters
Dec 17 17:37:13 <OutBackDingo> that issue is with a channel change from the command center, sometimes its not fully propogated and we orphan a radio or two
Dec 17 17:38:00 <nbd> ah, ok. so if there's some central control, then you can get away with a more simple form of discovery
Dec 17 17:38:06 <OutBackDingo> so i was trying to asses how to get them to scan, find neighbor, change channel, rejoin
Dec 17 17:38:14 <OutBackDingo> without rolling a truck :)
Dec 17 17:38:25 <nbd> does every node bring up an AP mode interface?
Dec 17 17:38:39 <OutBackDingo> yes though not always the same ssid
Dec 17 17:38:49 <nbd> ok, then use a client mode interface for scanning
Dec 17 17:38:59 <nbd> modify hostapd to inject an IE on the AP side
Dec 17 17:39:01 <nbd> for probe responses
Dec 17 17:39:28 <nbd> so that you can write a program for the client side that can find this IE in the scan results and decide which AP is the right one
Dec 17 17:39:43 <nbd> the nl80211 scan API gives you access to IEs
Dec 17 17:40:03 <OutBackDingo> ok ill dig
Dec 17 17:40:09 <OutBackDingo> thanks for your brain
Dec 17 18:36:49 <mazilo_> Does anyone here know what package do I need to enable if I want to compile a package that requires /dev/ptmx (pseudo-terminal master multiplexer)?
Dec 17 18:55:19 <CIA-79> juhosg * r29561 /trunk/target/linux/ar71xx/image/Makefile: ar71xx: use 64kraw template on more TP-Link boards
**** ENDING LOGGING AT Sun Dec 18 02:59:57 2011