[00:00] <nbd> that part looks normal to me
[00:00] <frop> yep
[00:00] <frop> then arrive "install", isn't it?
[00:00] <nbd> install comes after all compile steps
[00:00] <frop> yep
[00:00] <nbd> and only for packages marked as <*>
[00:01] <frop> ok...here it's M
[00:01] <frop> should i try it as * ?
[00:01] <frop> or you don't need it?
[00:01] <nbd> no, i don't need it at the moment
[00:01] <frop> k
[00:02] <nbd> found some more postinst brokenness
[00:05] <frop> $ svn up
[00:05] <frop> At revision 4643.
[00:05] <CIA-17> nbd * r4644 /packages/utils/strace/Makefile: cleanup
[00:05] <frop> but openntpd has the same behaviour
[00:05] <CIA-17> nbd * r4645 /packages/net/ (atftp/Makefile tor/Makefile): fix some more postinst brokenness
[00:05] <nbd> same as before?
[00:05] <frop> yep
[00:05] <nbd> try to rebuild it manually
[00:05] <nbd> maybe it doesn't detect file changes because of the symlink
[00:05] <frop> ishould i "clean" something other?
[00:06] <nbd> make package/openntpd-{clean,compile}
[00:06] <frop> yes, already done
[00:06] <nbd> after svn update in the packages directoryt
[00:06] <frop> ohhh
[00:06] <frop> can't i submit it from openwrt/
[00:06] <frop> ?
[00:06] <nbd> yes
[00:07] <Bartman007> from openwrt/ yes, but not from openwrt/branches/buildroot-ng/openwrt/
[00:07] <frop> ahhh
[00:08] <frop> :)
[00:08] <frop> openwrt/branches/buildroot-ng/packages then
[00:08] <Bartman007> actually those packages are in openwrt/packages/
[00:08] <Bartman007> there is also openwrt/branches/buildroot-ng/openwrt/package
[00:09] <frop> i got openwrt/ and packages/
[00:09] <frop> in openwrt/packages/ there're symlinks to ../../packages/something
[00:10] <nbd> then try svn update in packages/
[00:10] <Bartman007> I was speaking of the locations relative to where they are in the repository
[00:11] <frop> nbd: GRRR you are right
[00:11] <frop> fuck i'm not used still to buildroot-ng and packages
[00:11] <frop> :P
[00:11] <frop> sorry
[00:21] <frop> it get installed now :P
[01:13] <CIA-17> nbd * r4646 /branches/buildroot-ng/openwrt/target/linux/brcm-2.6/config: enable split_iso for ehci on brcm-2.6
[01:17] <CIA-17> nbd * r4647 /branches/whiterussian/openwrt/package/wificonf/wificonf.c: don't convert mw to qdbm for wl0_txpwr, set the value as-is
[05:04] <Bartman007> Kaloz: ping
[05:08] <Bartman007> h3sp4wn: ping
[10:36] <[florian]> nbd: what is the state of the multissid feature with wl ?
[10:36] <nbd> it should work
[10:36] <[florian]> even under win32 ?
[10:37] <nbd> sometimes
[10:37] <[florian]> ah, I have not yet tested under Linux
[10:48] <[florian]> nbd: I don't think I will manage to get assoclist working by myself unfortunately :(, I have tried various codes and they don't seem to work :(
[10:51] <nbd> what kinds of code?
[10:51] <nbd> btw. you should first try to implement it in user space
[10:51] <nbd> makes things easier
[10:52] <nbd> when you're done, move it to the kernel
[10:55] <[florian]> ok
[10:56] <[florian]> which library provides access to wl_ioctl's ?
[10:57] <[florian]> well, I will find it by reading the wificonf makefile
[10:57] <nbd> copy some code from wificonf
[10:57] <nbd> it's easy
[11:05] <[florian]> ok
[11:28] <[florian]> nbd: did you answer the guy that asks for captive portal informations or so ?
[11:28] <nbd> not yet
[11:28] <nbd> did you?
[11:30] <[florian]> neither did I, I don't really understand what he expects from us actually
[11:39] <nbd> groz: ping
[11:39] <groz> pong
[11:39] <nbd> i have another potential fix for the google issue with broadcom-wl
[11:39] <nbd> can you try it?
[11:40] <groz> ok, i'm dealing with another set of gotchas here, it'll be a little while before i can try it
[11:40] <groz> i got a busted network, trying to figure out why dhcp not working on it right now
[11:40] <nbd> when you have time, please install wl and run 'wl vlan_mode 0'
[11:40] <groz> ok, so just check it out as it is now
[11:40] <groz> then ipkg install wl
[11:40] <nbd> yes
[11:40] <groz> and run wl vlan_mode 0
[11:40] <nbd> yes
[11:40] <groz> ok, i'll get to it shortly
[11:40] <nbd> people in the dd-wrt forum report that it works
[11:41] <groz> i'll have to build and flash
[11:41] <groz> it wont take long
[11:41] <nbd> ok
[11:41] <nbd> thanks
[11:41] <groz> but right now, gotta get my dhcp server fixed first, i got a network with things dropping offline
[11:41] <nbd> yeah
[11:41] <nbd> i'll add support for that call to wlc in the mean time
[12:18] <groz> nbd, is there any particular hardware version i should be trying for this first test ?
[12:18] <nbd> no
[12:18] <groz> ok, the one on my desk is gs V2.1
[12:18] <nbd> good
[12:18] <groz> i'll be a few minutes yet
[12:19] <groz> but I found my network problem just validating it's all fixed now
[12:19] <groz> some idiot (that would be me) configured a dhcp server to not acknowledge 'new clients'
[12:19] <nbd> :)
[12:19] <groz> well, i'm experimenting with a slick new setup here
[12:19] <malbon> [florian]: I got my dg834gt today.
[12:26] <CIA-17> nbd * r4648 /branches/buildroot-ng/openwrt/package/broadcom-wl/src/wlc/wlc.c: add vlan_mode to wlc
[12:31] <malbon> groz: sounds secure :)
[12:36] <groz> malbon, actually, it's anything but
[12:36] <groz> im building a monster network of umls
[12:40] <CIA-17> kaloz * r4649 /branches/whiterussian/openwrt/target/linux/linux-2.4/patches/generic/902-hot_fixes-32.7.patch: add new hotfixes patch
[12:42] <malbon> uml, that's Xen on the cheap? ;)
[12:42] <groz> if you only want linux targets, it's far more efficient
[12:42] <h3sp4wn> Kaloz: Did you get round to finishing the image (no probs if you didn't)
[12:42] <groz> there is no emulation layer involved
[12:42] <malbon> in uml or xen?
[12:43] <groz> in uml
[12:43] <groz> it's just 'more threads'
[12:43] <malbon> there is no emulation in xen either.
[12:43] <malbon> (well unless you use hvm)
[12:43] <Kaloz> nope.. but i really should get some time for it today or tomorrow
[12:43] <groz> doesn't xen emulate the hardware ?
[12:43] <malbon> nope not with native guests.
[12:44] <malbon> it's more efficient than uml in my experience, but less flexible.
[12:44] <groz> i've never tried it, but, for what i'm doing, uml works very well
[12:45] <groz> simulating a network with a couple hundred nodes, and, serious propogation delays between them
[12:45] <groz> use teh qos network delay settings for that
[12:45] <malbon> yes, if it works for you, why not. :)
[13:00] <nbd> groz: what kind of software are you running on the nodes? i mean what part specifically are you going to test?
[13:01] <groz> what i'm working up nbd, is a method of simulating long hop radio links
[13:01] <nbd> interesting
[13:01] <groz> long hop in many cases includes satellite round trip times
[13:01] <groz> so for those nodes, i need to introduce 600ms delays
[13:01] <groz> so here's how i'm setting it all up
[13:02] <groz> i have a few vlans, with uml's conntecting to vlans by briding the uml switch with the vlan
[13:02] <groz> then I have a couple of 'real computers' also in the mix
[13:02] <groz> we are setting up a simulation of a rather complex network based on satellite links
[13:02] <groz> to test it for 'useablity' to carry voip
[13:02] <groz> with all the inherent satellite delays
[13:03] <nbd> very interesting
[13:03] <groz> since this company has NO communication now
[13:03] <groz> other than satellite phones at a buck a minute
[13:03] <groz> this could be a very interesting setup when done
[13:03] <nbd> yeah
[13:03] <nbd> definitely
[13:03] <groz> what i'm doing, is basically putting sattellite round trip delays
[13:03] <groz> between a few asterisk boxes
[13:03] <groz> then connecting sip phones to the asterisks on 'local networks'
[13:04] <groz> that's one simulation i'm using
[13:04] <groz> the other, i'm gonna end up simulating a 200+ node wireless mesh
[13:04] <nbd> what kind of routing are you using in the mesh?
[13:05] <groz> proprietary stuff that a client uses
[13:05] <nbd> ah
[13:05] <groz> doing a scale up simulation of thier existing setup
[13:05] <groz> currently the largest deplyed system only has 48 nodes
[13:06] <nbd> the freifunk people are always complaining about people testing mesh protocols only in simulations and not in real deployments
[13:06] <groz> well, we are doing it the other way around
[13:06] <groz> we have deployments, and know what the issues are
[13:06] <nbd> that's good
[13:06] <groz> and now, are testing to see how far we can scale up
[13:07] <nbd> berlin has a working 300+ node mesh
[13:07] <groz> these guys setup is pretty slick, it's self healing, self routing, just take the nodes out, plug em into power and let the magic do it's thing
[13:08] <groz> they use it today mostly to do wifi coverage of places like campgrounds, where strining ethernet to ap's is prohibitively expensive
[13:08] <nbd> yeah
[13:08] <nbd> the main problem of the berlin mesh is that almost nobody bothered to implement the ad-hoc mode properly
[13:09] <nbd> and there are a few issues in the 802.11 standard that also make it very easy to cause trouble
[13:09] <groz> this doesn't use ad-hoc
[13:09] <nbd> yeah, i figured
[13:09] <groz> its using wds connections
[13:09] <groz> so that the aps appear to be 'normal ap' to end users
[13:10] <groz> and windows will just go 'found a wireless network'
[13:10] <groz> when customers are in the locations
[13:10] <nbd> probably with at least two wifi cards on every node, right?
[13:10] <groz> depends on the rollout, but, that's one way of implementing
[13:11] <groz> i do all of my testing of stuff here using one per node, cuz i test on wrt54
[13:11] <groz> now i'm migrating testing into a swack of umls
[13:12] <nbd> because i think if you're using wds and you have clients and mesh neighbors on the same channel, it can get pretty traffic-intensive
[13:12] <groz> well, when you think it' thru
[13:12] <groz> it's being used for internet accesss
[13:13] <groz> and normally fed by a 1.5 mbit dsl or slower sat link in the remote locations
[13:13] <groz> so, the overhead on the wifi for meshing
[13:13] <nbd> yeah, in a setup like that it doesn't matter
[13:13] <groz> is plenty of room for that
[13:13] <groz> 54mbit actually works out to a REAL 10 mbit of thruput in general
[13:14] <groz> where you start to run into issues, is when you get out on the far reaches of the mesh
[13:14] <groz> there's a lot of latency
[13:14] <groz> due to the number of radio hops
[13:14] <groz> a lot of these locations, the only choice for inet access is satellite to begin with
[13:15] <groz> so the question becomes this
[13:15] <groz> how far out can you go from the main downlink, before it's better off to start with another downlink
[13:16] <groz> back in 5, gotta restart a router
[13:28] <nbd> groz: btw. this vlan_mode thing has a very high chance of working
[13:29] <groz> i've got a build just starting, then i gotta flash
[13:29] <nbd> k
[13:40] <groz> nbd, one other issue we were having
[13:40] <groz> was that a centrino laptop was unable to associate with it
[13:40] <groz> any chance the same thing will fix that too ?
[13:41] <groz> i'm guessing this all may be revolving around vlan tags in the wifi packets
[13:41] <groz> from your fix
[13:41] <nbd> i don't know if it'll fix centrino
[13:42] <groz> ok, just getting set up here to tftp on the build
[13:47] <groz> hmm, i seem to have built a broken image
[13:47] <groz> lemme try this again
[13:53] <groz> nbd, dumb question
[13:53] <groz> where in menuconfig do i find the 'propriety broadcom driver'
[13:53] <groz> cuz i seem to have lost it
[13:54] <groz> oh wait
[13:54] <groz> i found it back
[13:54] <groz> explains why this wasn't working
[13:54] <groz> i had it configured out
[14:05] <nbd> groz: seems like i made a mistake in the network scripts. fixing that
[14:13] <[florian]> malbon: great, have you tested yet bcm63xx ?
[14:14] <groz> I'm having some other issues here, not quite sure yet what's going on, but, i'm getting a very broken build, so i just blew away the tree and checked out a fresh copy to try again
[14:15] <nbd> groz: network scripts are broken and don't bring up the interface properly
[14:15] <nbd> groz: i'll have it fixed in a minute
[14:16] <nbd> groz: or is it broke in a different way?
[14:16] <groz> i think my main break was some other hacks i had in from earlier playing
[14:16] <groz> so i just started fresh to be able to do this test
[14:17] <groz> ahh, i see commit messages are not showing up here today
[14:17] <groz> that's why i was wondering 'whats up', you have committed some changes
[14:20] <CIA-17> nbd * r4650 /branches/buildroot-ng/openwrt/package/base-files/ (4 files in 4 dirs): fix remaining *_ifnames references
[14:20] <nbd> fixed now
[14:21] <groz> heh, and i see it here now too
[14:21] <groz> magic
[14:22] <Kaloz> so all the wifi-related stuff is fixed now?
[14:22] <groz> well, lets wait till i got it on here
[14:22] <groz> and i can validate
[14:22] <groz> before we claim it all fixed
[14:22] <nbd> Kaloz: wds is not implemented yet
[14:22] <nbd> Kaloz: and i think this 'fix' won't do anything to help with the centrino stuff
[14:22] <groz> so this build wont do wds ?
[14:22] <Kaloz> groz: btw, it's kinda funny that you were not able to conenct with a centrino card
[14:23] <nbd> groz: it's not implemented in the scripts yet
[14:23] <groz> it wasn't me kaloz, its a friend of mine up north
[14:23] <groz> and if it 'works' here
[14:23] <Kaloz> because maybe the draft-n thingie is working, just my notebook's centrino wasn't able to connect to it :p
[14:23] <groz> i'll have him flash a router later this morning, and try it
[14:23] <nbd> Kaloz: haha
[14:23] <nbd> groz: wlc can set up wds, but the scripts don't set it up from the network config yet
[14:23] <Kaloz> nbd: don't laugh, i can blame you all the time :P
[14:23] <groz> well we had that issue with the ng build, his notebook wouldn't connect
[14:23] <nbd> Kaloz: go ahead
[14:23] <groz> but, all his other stuff would
[14:24] <Kaloz> nbd: i didn't say i want to :P
[14:24] <nbd> hehe
[14:24] <groz> oknbd, if it's just the script sfor wds, wont matter here
[14:24] <Kaloz> nbd: i do make mistakes as well all the time
[14:24] <nbd> i wasn't laughing at you btw. just laughing at the funny situation
[14:25] <Kaloz> well, probably i would have mor eluck with an athros
[14:25] <Kaloz> but HP locks the bios down
[14:25] <nbd> this lockdown thing really sucks
[14:25] <Kaloz> and changing the pci id to pass the bios and then modify madwifi all the time just to be able to pay with that is stupid
[14:25] <Kaloz> and my centrino works quite nicely
[14:26] <Kaloz> the 2915abg that is
[14:26] <nbd> i've always had problems with centrino in a friend's laptop
[14:26] <nbd> 2100 was really crappy
[14:26] <nbd> 2200 worked somewhat better
[14:26] <nbd> but still not that great
[14:26] <groz> as a point of reference, that same laptop has no problems when we have a wr derivative flashed on
[14:27] <nbd> groz: i've heard about these centrino problems before
[14:27] <nbd> groz: from dd-wrt
[14:29] Action: Kaloz never had problems before with centrino stuff
[14:29] Action: groz thinks that's cuz kaloz has never _tried_ centrino stuff
[14:31] <Kaloz> .....
[14:32] <nbd> groz: how's the test going? :)
[14:32] <groz> I'll be reflashing here in a minute
[14:32] Action: groz doing 3 things at once
[14:33] <nbd> you could try this diff then: http://downloads.openwrt.org/people/nbd/tmp/diff.txt
[14:35] <groz> nbd, it just took the tftp, now to see if i built one that'll actually start
[14:35] <nbd> did you remember to rebuild base-files manually?
[14:35] <nbd> because the autorebuild is not fixed there yet
[14:35] <groz> i built 'from scratch'
[14:35] <nbd> ok
[14:46] <[florian]> there are less and less whiterussian tickets, sound good :)
[14:50] <groz> you know, i really hate that default firewall script, when i flash one of these with 'default' stuff, gotta go fuss with cables and such to get connected on the 4 port, then, after I make the changes and reboot it, connected back on the other side
[14:51] <groz> it's just in time to realize it was'nt finished firstboot when i did that
[14:51] <groz> and i get to start over
[15:00] <groz> nbd ping ?
[15:00] <groz> with this new setup, strait out of the box
[15:00] <groz> it should be just acting as a simple masquerading router no ?
[15:00] <nbd> yes
[15:00] <nbd> with dhcp on wan
[15:00] <groz> right now, it doesn't seem to be forwarding anything from my notebook
[15:01] <groz> i can ping 1.1 but nothing beyond, and a quick look
[15:01] <groz> i cant see why
[15:01] <groz> but ssh in, i can ping all over
[15:01] <nbd> does ns lookup work?
[15:01] <groz> yes
[15:01] <groz> i've got dns on the notebook
[15:02] <groz> well, wait, let me verify it's not cached
[15:02] <groz> yup
[15:02] <groz> notebook is talking to router, and getting dns
[15:02] <groz> but it cant ping past the router
[15:02] <groz> and the iptables ruels seem ok
[15:02] <Kaloz> groz: this is a centrino notebook? :P
[15:03] <Kaloz> groz: ip_forward?
[15:03] <groz> no
[15:03] <groz> yes, ip_forward is 1
[15:03] <nbd> did the MASQUERADE rule trigger?
[15:03] <nbd> do iptables -L -v
[15:04] <groz> iptables -t nat -L shows a masqerade rule
[15:04] <groz> iptables -L -v doesn't but i dont expect it there
[15:05] <groz> is there a quick and easy way with all this new network stuff to put it back to using vlan names on the adapters
[15:06] <groz> so i can drop in something known to work well ?
[15:09] <Kaloz> groz: iptables -t nat -L -v
[15:09] <Kaloz> and chekc if the nat rule is triggered imho :)
[15:09] <groz> yes, it is
[15:09] <groz> there is a masquerade rule there
[15:09] <Kaloz> the question is if its trggered
[15:09] <Kaloz> (eg. packet count)
[15:10] <groz> yah, seems to be increasing
[15:11] <Kaloz> i would say something is fishy there with networking
[15:11] <Kaloz> check it with tcpdump
[15:11] <groz> hang on, i'm checking a few more things
[15:11] <groz> ok, i changed policy on FORWARD to ACCEPT
[15:11] <groz> and away i go
[15:12] <Kaloz> groz: can you check it with a centrino? :)
[15:12] <groz> ok, got expected results, yahoo load, google does not
[15:12] <groz> now i try that wl trick
[15:12] <groz> it was wl vlan_mode 0 correct nbd ?
[15:13] <groz> woooohoooooo
[15:13] <groz> google loads
[15:13] <groz> i cant with the centrino right now kaloz, but, in another 4 hours denis will be up
[15:13] <groz> and i can get him to flash a box, and try it
[15:13] <Kaloz> okie
[15:14] <groz> ok, this google fix, for me, it's huge
[15:14] <CIA-17> nbd * r4651 /branches/buildroot-ng/openwrt/package/broadcom-wl/files/lib/wifi/broadcom.sh: add vlan_mode 0 command to fix the weird google issue in broadcom-wl
[15:14] <groz> now, if centrino will associate, it'll be even hugeer
[15:14] <groz> nbd, the vlan_mode read back -1 before i set it
[15:14] <groz> and google did NOT load
[15:14] <groz> i set it
[15:14] <groz> and google came right up on the notebook
[15:15] <nbd> grat
[15:15] <nbd> great
[15:15] <groz> after i changed the forward policy to actually forward :)
[15:15] <nbd> is there a bug in the firewall script?
[15:15] <groz> not sure, i have to go look why, i wasn't really interested in that right now, more interested in google issue :)
[15:15] <groz> all i did was change the polich
[15:16] <groz> iptables -P FORWARD ACCEPT
[15:16] <nbd> yes, there is a bug
[15:16] <groz> first thing i normally do, is toss away all that
[15:16] <groz> and introduce my own
[15:16] <groz> so, i dont normally even see what it does
[15:23] <CIA-17> nbd * r4652 /branches/buildroot-ng/openwrt/package/base-files/default/lib/network/config.sh: default value for *_device in scan_interfaces
[15:27] <groz> nbd, quick dumb question
[15:28] <groz> in this new network setup scripts, where are you setting the name 'br-lan' ?
[15:28] <groz> i cant find it right off
[15:32] <Kaloz> groz: /etc/init.d/S05netconfig
[15:32] <Kaloz> ?
[15:33] <groz> that's the setup, but, what's actually chaning the device name
[15:33] <groz> from br0 to br-lan
[15:33] <groz> that's what i cant find
[15:33] <nbd> /lib/network/config.sh
[15:34] <groz> k thanks
[15:34] <groz> ahah, got it
[15:46] <CIA-17> nbd * r4653 /branches/buildroot-ng/openwrt/package/base-files/default/ (etc/hotplug.d/net/10-net lib/network/config.sh sbin/ifup): revert to using *_ifnames internally, too many problems with firewall, ppp, etc. otherwise
[15:47] <groz> hehe
[15:47] <groz> ok, what did you change now ?
[15:48] <groz> i was just starting to look at some of the side effects of all the adapters having names changed
[15:48] <nbd> after running scan_interfaces, config_get lan ifname will return br-lan instead of the interface list
[15:48] <groz> ahh
[15:48] <nbd> (in the config, ifname specifies the list of interfaces in the bridge)
[15:49] <nbd> so that after scan_interfaces, you can always expect the ifname to contain the interface name that will be used for ip stuff
[15:49] <groz> I like this 'no dependancy on nvram' setup, but, it's goinna take a little thinking to get it all figured out correctly
[15:49] <nbd> yeah
[15:49] <nbd> the new scripts are a bit complex :)
[15:49] <nbd> but they're quite modular and the config files are easy to read and write
[15:49] <murb> the old setup scripts that assumed you had nvram were annoying
[15:49] <groz> yah, and, bash is not my best language, and i'm awful at the ash dialect thereof
[15:50] <groz> but im getting it figured out
[15:50] <nbd> the whole config abstraction takes a bit of getting used to, i think
[15:50] <nbd> especially if you use callbacks on config sections
[15:50] <groz> well, the 'issue' i still have, is the concept of 'wan/lan' and how it still seems to wrap around that
[15:51] <groz> cuz, there's potential for a lot more complex setups
[15:51] <groz> but once i get this config thing figured, i can probably see a way thru that
[15:51] <nbd> you can add other interface groups as well
[15:51] <groz> yah, was just thinking that as i typed it
[15:51] <nbd> the only special treatment of lan and wan is in the firewall stuff
[15:51] <nbd> and it's mostly because of wan
[15:52] <groz> well as an example, the router in front of me right now, it's running -ng built for x86 on a via board
[15:52] <groz> it's got a 'wan' eithernet, and a 'lan' ethernet
[15:52] <groz> but, it's also got a bunch of vlans
[15:52] <groz> so, one vlan is for one specific virtual machine group
[15:52] <groz> another vlan for another
[15:52] <nbd> if you set wan to where your internet uplink is, it's going to be ok
[15:53] <groz> and they could just as easily be separate wires
[15:53] <nbd> if you don't have a central uplink, drop it and do custom stuff
[15:53] <groz> yah, well, the first thing i always do is
[15:53] <groz> del firewall.*
[15:53] <groz> and then put in my own handling for iptables stuff
[15:53] <groz> but that's another totally different story unto itself
[16:14] <[florian]> nbd: there is not a nand chip driver for rb532 ?
[16:14] <nbd> i haven't added it yet
[16:14] <[florian]> ah ok
[16:14] <nbd> but i have a patch
[16:14] <[florian]> but it should work ?
[16:14] <nbd> yeah
[16:14] <nbd> but i heard that the nand chip is not that reliable
[16:14] <[florian]> really great
[16:14] <nbd> so i thought it would be better to start off with booting from cf only
[16:14] <[florian]> ah, I don't know much about the nand technology actually
[16:15] <nbd> me neither
[16:15] <thepeople_work> would it be posible to add the wp54 patches for openwrt kamikaze that compex has made?
[16:15] <[florian]> wp54 ?
[16:16] <thepeople_work> a compex wp54
[16:16] <thepeople_work> http://compex.com.sg/home/OEM/Open_wrt.htm
[16:16] <thepeople_work> the table of hardware is not correct for this device wither
[16:16] <thepeople_work> either*
[16:16] <nbd> thepeople_work: i will get hardware soon and then we'll clean up the patches and merge them
[16:17] <thepeople_work> ok, I have the hardware that is why I was wondering :-)
[16:17] <nbd> well, you'll get a fresh buildroot-ng port for it soon :)
[16:17] <thepeople_work> thanks, nbd
[16:25] <nbd> Hooligan0: hi
[16:48] <[florian]> nbd: we have the sources for libshared right ?
[16:51] <nbd> sure
[16:52] <[florian]> ok, I was digging into some GPL tarballs
[16:55] <[florian]> in the tiwlan.o binary file, I can see that the module licence is given as GPL, but we can't find the sources for this Wi-Fi driver, can we ?
[16:55] <murb> [florian]: oh nice catch!
[16:56] <murb> is that an acx1xx card?
[16:56] <[florian]> yes it is, there are references to Acx111
[16:56] <[florian]> you can find the file in the archive named WAG54GV2-EU_v1.00.19.tgz
[16:56] <[florian]> thanks to linux-mips.org
[16:57] <murb> I saw on lkml/netdev that they are refusing to merge the acx free drivers because of questions about where they came from, if they were legally reverse engineered etc.
[16:58] <murb> [florian]: does that include netfilter code?
[16:58] <[florian]> murb: I don't know yet
[16:58] <murb> i'm just wgetting it.
[16:59] <[florian]> murb: how can it use netfilter code by the way ? if it is a driver ?
[17:00] <murb> [florian]: i mean in that tgz
[17:00] <[florian]> ah you mean maybe the whole tarball :) right ::)
[17:00] <murb> prevent more netfilter work being done by mentioning it to gpl-violations...
[17:01] <murb> (i'd very much like a clean acx driver in the mainline kernel)
[17:01] <nbd> [florian]: the only gpl references in the driver tarball are in the WPA stuff and the config utility
[17:03] <[florian]> ok
[17:03] <[florian]> have to go, see you later !
[17:19] <groz> nbd, i'm almost finished the setup here, once its done, i'll deploy this new driver mod onto a batch of 8 wrts, and run some radio stress tests for a bit
[17:19] <nbd> great
[17:19] <groz> one of them is in client mode
[17:20] <groz> the rest are aps
[17:20] <groz> the aps all talking on wds links
[17:20] <groz> the client hops from one to the next
[17:20] <groz> as a client
[17:20] <groz> whole thing is scripted
[17:20] <nbd> the wlc wds command should be working
[17:20] <groz> at the end, it pops out some stats on overall radio thruput and error rates
[17:20] <groz> ok, this all uses wl tho
[17:20] <groz> it's a set of scripts i've had for a while
[17:21] <nbd> if you want to save space someday, you can replace the stuff with wlc
[17:21] <groz> im just making minor mods to it all to acomodate the fact it's dealing with eth0.x instead of vlanx
[17:21] <nbd> if any particular command is missing, tell me
[17:21] <groz> hehe, ok, but, i think i'm capable of adding one myself too if it's got me stuck
[17:22] <groz> i _do_ want to get switched over
[17:22] <groz> but, it takes time to migrate everything
[17:22] <nbd> yeah
[17:22] <groz> right now, the google thing has been a show stop, and, the centrino thing too
[17:22] <groz> i've tested google fix, will get denis to try centrino when he's up
[17:23] <groz> it's still early in the morning our time here
[17:23] <groz> i'm on a bit of a screwed up schedule, was flying yesterday, and got home around 4pm
[17:23] <groz> then crashed, and slept till midnite
[17:26] <groz> the next time you are playing in those init scripts
[17:26] <groz> i've got one 'request'
[17:26] <groz> i use a dynamic dns setup here, so, int he stanza defining the adapter
[17:26] <groz> would be cool to be able to add a name that it passes as the -H parameter to udhcpc
[17:26] <groz> flash up the box
[17:26] <groz> when it dhcps from the network
[17:26] <groz> ping openwrt will respond
[17:27] <groz> cuz dhcp updates the dns
[17:27] <groz> it belongs in the config stanza tho, because, there is valid times to have it use different names on different adapters
[17:27] <groz> rather than just take the name from hostname
[17:28] <nbd> you can set 'option hostname foobar' in a dhcp interface section
[17:28] <groz> oh, ok, you are already ahead of me
[17:28] <groz> i haven't got enough boxes flashed to play yet, but, rest assured, i will be playing a bunch with it before the day is done
[17:30] <murb> alternatively you could do a nsupdate, but keying requires sychonrised clocks.
[17:31] <groz> that would involve putting nsupdate on the client
[17:32] <nbd> groz: according to some people updating the ipw drivers fixes the issue with the new broadcom stuff
[17:32] <groz> rather the name is just in the dhcp, and do it ther,e for simplicity
[17:32] <groz> I'll have data n that later today nbd, it's still to early in the morning to phone denis
[17:32] <nbd> groz: hmm... for some people it still doesn't work
[17:33] <groz> what, the google, or the ipw ?
[17:33] <nbd> ipw
[17:33] <nbd> not tested with openwrt
[17:34] <nbd> tested with the linksys firmware
[17:34] <nbd> (wrt300n)
[17:47] <Bartman007> Kaloz: if you could post the link to your 7001 binaries with eth drivers I would be greatful. and I may have a armeb uclibc compile error for you in about a hours time.
[17:49] <Kaloz> nbd: so i need to grab some pccard to check it
[17:50] <Kaloz> Bartman007: well, today i don't know how will i have time for it (birtday partie) but i'll try
[17:51] <h3sp4wn> Kaloz: I can test the drivers if you have the link also
[18:27] <Bartman007> no problem, I don't mean to rush you
[18:33] <murb> btw http://www.yuri.org.uk/~murble/tiwlan.txt # they claim their module is GPLed.
[18:34] <nbd> murb: ah, i just noticed, that the source that is included in that tarball is the source for tiap.o
[18:34] <nbd> murb: the one for tiwlan.o is not included
[19:45] <nbd> [florian]: ping
[19:48] <groz> nbd, i just got a call from denis, he's up, but it's gonna be at least an hour or so before he can get to flashing a router
[19:48] <nbd> no problem
[19:48] <nbd> i can wait
[19:49] <groz> hehe, we dont have much choice now do we ?
[19:49] <nbd> i don't really expect the vlan_mode thing to fix the centrino assoc anyway
[19:50] <groz> ok, then i do have a mior request for wlc
[19:50] <groz> wlc fixe_centrino=on
[19:50] <nbd> i don't have a specific call to map it to :)
[19:50] <groz> just a minor request :)
[19:51] <nbd> and without knowing the problem i can't add patchtable entries like for the slot time or the no_ack flag
[19:51] <nbd> :)
[19:51] <groz> what do we have to do to identify the problem ?
[19:51] <nbd> no idea
[19:52] <nbd> have you seen the crazy patchtable stuff?
[19:53] <nbd> :)
[19:53] <groz> no, i haven't really stopped to look at it
[19:53] <groz> i just saw ther eis some really ugly hack on the go there
[19:53] <groz> seems like it's patching a driver in memory on the fly
[19:53] <nbd> yeah
[19:53] <nbd> and the build system generates a patch table by running objdump on wl.o and parsing it with a perl script :)
[19:55] <[florian]> nbd: pong
[19:56] <nbd> [florian]: do you have any major uncommitted bcm63xx changes?
[19:56] <[florian]> let me see
[19:56] <nbd> [florian]: because i was planning on hacking a bit on that port now
[19:56] <[florian]> ah really great
[19:56] <[florian]> do you have any bcm63xx device ?
[19:56] <nbd> yeah
[19:56] <nbd> a 96348
[19:57] <[florian]> ok, CFE or redboot based ?
[19:57] <nbd> cfe
[19:58] <[florian]> ok, really good, so that you can also try to patch the mtd map to use the mapping provided by CFE
[19:58] <[florian]> I have no major commit
[19:58] Action: murb wonders what chipset is in his sinus 1054 dsl thing
[19:58] <[florian]> nbd: there is one thing you can probably do, is fixing the bootloader detection
[19:59] <nbd> yeah, i was planning on doing that
[19:59] <[florian]> nbd: in the inventel livebox patch I found, it was done by parsing the arguments passed to prom_init
[20:01] <[florian]> nbd: here it is : http://pastebin.ca/147823
[20:04] <[florian]> I think there is now, with 2.6 kernels, a better way to get arguments passed by the bootloader
[20:06] <[florian]> this reminds me of getting my huawei hg550 from my parent's house
[20:10] <murb> hmm ok i have a 96345 here.
[20:49] <CIA-17> nbd * r4655 /branches/buildroot-ng/openwrt/include/prereq-build.mk: add prereq check to prevent people from building openwrt as root
[20:55] <Bartman007> nbd: haha, that would have screwed my laptop on saturday if I had run as root
[21:13] <CIA-17> nbd * r4656 /branches/buildroot-ng/openwrt/Makefile: hide rm -rf command in Makefile
[21:14] <nbd> hmm... just realized how ambiguous that commit message is :)
[21:16] <murb> nbd: will that stop me buliding it with fakeroot?
[21:16] <murb> or do you test for that?
[21:17] <nbd> no, but why use fakeroot?
[21:17] <nbd> it should be built as a regular user entirely
[21:40] <groz> nbd, you cannot do the things required to make a bootable image as a regular user
[21:40] <nbd> what kind of a bootable image?
[21:40] <groz> installing lilo onto a loop mounted image
[21:40] <groz> as an example
[21:40] <nbd> then we have to figure out a way to do it without using loopback
[21:41] <nbd> i don't want the build process to require root access at any point
[21:41] <groz> then you cant install boot loaders, doesn't really matter whihc you are trying to install for an x86 bootable image, lilo, grub take your pick
[21:41] <groz> they need to be loop mounted or they wont install
[21:41] <nbd> at the moment, yes
[21:42] <groz> they checkf for being on a block device, and refuse otherwise
[21:42] <nbd> we can fix that
[21:44] <groz> then when all is said and done
[21:45] <groz> you still have to write that image onto the media
[21:45] <nbd> yeah, but that isn't part of the build process
[21:45] <groz> well, i kept the 'make bootable' script out of the main build process
[21:45] <groz> for that reason
[21:45] <groz> it's gotta be run after the fact
[21:48] <nbd> also we should make the images somewhat size independent
[21:48] <groz> i did
[21:48] <nbd> great
[21:48] <groz> you can resize2fs the media
[21:48] <groz> after the fact
[21:48] <groz> but that only works if you dont partition it
[21:49] <groz> which you can do with a usb stick
[21:49] <groz> just write it on raw, treats it like a big floppy instead of like partitioned media
[21:49] <nbd> on rb532 i use block2mtd and a patched partition table parser
[21:49] <groz> it actually works on hd too
[21:49] <nbd> which maps all the free space after the partitions to a separate partition
[21:49] <nbd> and also reduces the size of a partition if it goes beyond the end of the device
[21:50] <groz> the way i make the images for usb keys, there are no paritions
[21:50] <groz> so, you can just resize2fs it, and it'll expand to fill the device
[21:50] <groz> i create an 8 meg image to start with
[21:50] <groz> dont think you can get smaller usb media
[21:50] <nbd> i do the same with rb532, only with jffs2
[21:51] <nbd> or maybe 16m, don't remember
[21:51] <groz> when using a usb key, no real point to jffs2
[21:51] <groz> they do wear levelling internally, and map to a 512 byte sector architecture externally
[21:51] <nbd> is it possible to create an ext2 fs that has exactly the size that it needs?
[21:52] <groz> it is, but the way i'm building it
[21:52] <groz> the same ext2 is used for both uml and for this
[21:52] <groz> and on uml, you want _some_ free space
[21:52] <nbd> well, you can run resize2fs automatically
[21:53] <groz> well, when using it with a uml, the size is fixed once you start it
[21:53] <groz> you cant expand it
[21:53] <groz> on the fly
[21:54] <groz> it's also possible, if you are using initramfs, to have nothing but the kernel on the media
[21:54] <groz> then run from the ramfs once up
[21:54] <groz> but that is only useful for a fixed configuratin where you never want to save config data between reboots
[21:54] <nbd> for the generic x86 port it would probably also be nice to use grub and allow the user to specify rootfs and kernel partition sizes in menuconfig
[21:55] <groz> i had thought of specifying the image size there
[21:55] <groz> i had issues getting grub to properly install
[21:55] <nbd> controlling the kernel partition size could be useful if the user plans on adding more space for different test kernelws
[21:55] <groz> the thing is when you loop mount, it cant read the geometry correctly, and, it wasn't picking up correctly the way lilo was
[21:55] <groz> after the fact
[21:55] <groz> keeping in mind, when you stick a given key into different computers
[21:56] <groz> the bios will give it totally different geometry
[21:56] <groz> with lilo you can force it to go lba, and, it 'just works'
[21:56] <nbd> when i have time, i could check out the grub source code
[21:56] <groz> i tried hard to get extlinux to do that
[21:56] <groz> cuz that would have been the ideal simple choice
[21:56] <groz> but
[21:57] <nbd> or maybe it would be better to use syslinux and fat
[21:57] <groz> it doesn't seem to be capable of doign lba
[21:57] <groz> if you do that, then the fat cannot become your root file system
[21:57] <groz> after the boot
[21:57] <nbd> it doesn't have to
[21:57] <nbd> we can have a separate partition
[21:57] <groz> i WANT my boot device as root
[21:57] <nbd> ok
[21:57] <groz> and i want it without partitions :)
[21:57] <nbd> ok
[21:57] <nbd> :)
[21:58] <groz> and i found a way to make it all 'just work'
[21:58] <groz> nothing stops this setup from working within a partition
[21:58] <groz> ie if you dd the image it currently creates for x86 into a partition
[21:58] <groz> then make the partition bootable
[21:58] <groz> it should just work
[21:59] <groz> and you should be able to resize it up to the size of the partition
[21:59] <groz> but you will have to edit the example lilo conf files i put in there
[21:59] <groz> to account for partiioned media
[22:00] <groz> there is another whole track one could take, and that's to do along the lines of what we have in the wrt
[22:00] <nbd> i think i'll give partitionless grub a shot first
[22:00] <nbd> hopefully without using loopback
[22:00] <groz> and do it with 2 paritions, then jump thru the same hoops
[22:00] <groz> well, i already started a menu section in menuconfig titled 'boot loaders'
[22:00] <groz> :)
[22:00] <nbd> :)
[22:00] <groz> i want to take a peek at uboot as well
[22:01] <groz> in theory,no reason we cant have various ones in there, with appropriate depends
[22:01] <groz> so, you see lilo and grub when building for x86
[22:01] <nbd> yeah
[22:01] <groz> and you see uboot/whathaveyou for other hardware
[22:05] <nbd> [florian]: ever noticed this?
[22:05] <nbd> Calibrating delay loop... 4.97 BogoMIPS (lpj=12448)
[22:05] <nbd> [florian]: on bcm63xx
[22:05] Action: nbd tries again with gcc 3.4.4 instead of 3.4.6
[22:10] <[florian]> nbd: ah, I don't remember seeing that
[22:11] <[florian]> nbd: have you succeeded in booting the current bcm63xx branch on your device ?
[22:11] <nbd> hint: it's the same problem that bcm47xx has with gcc 3.4.6
[22:11] <nbd> nah, it crashes while accessing the flash
[22:11] <nbd> most likely related to this bogomips issue
[22:12] <nbd> what does the bogomips count show on your device?
[22:12] <nbd> and with which compiler did you build it?
[22:13] <[florian]> nbd: I don't exactly remember, and I have not the device set up right now
[22:13] <nbd> k
[22:13] <nbd> would be interesting to know
[22:13] <[florian]> nbd: 3.4.6/2.17
[22:14] <nbd> i'll try 3.4.4/2.16 now
[22:14] <[florian]> ok
[22:14] <nbd> if it works, i'll try 3.4.6/2.17
[22:14] <[florian]> sounds weird to have gcc versions changing the bogomips calculations
[22:14] <nbd> well, it's a known problem on bcm47xx
[22:15] <nbd> it's the only reason why we're still using 3.4.4 for it in kamikaze
[22:15] <[florian]> ah, and do you have a fix ?
[22:15] <nbd> no
[22:27] <[florian]> nbd: confirmed, I had the same bug : Calibrating delay loop... 4.12 BogoMIPS (lpj=10304)
[22:28] <[florian]> using gcc-3.4.6
[22:28] <nbd> and this is probably the reason for the random crashes in various places
[22:29] <[florian]> that's what I was going to ask
[22:29] <nbd> the weird thing is that i've never seen this bug in linux 2.6 on bcm47xx
[22:29] <[florian]> ha such a pity
[22:30] <[florian]> why could it cause random crashes in various places ?
[22:31] <nbd> [florian]: because wrong bogomips messes up lots of parts in the kernel - flash access, hardware access
[22:32] <nbd> [florian]: all the timings are suddenly wrong
[22:32] <[florian]> ok, I did not know it was used like that too
[22:40] <groz> nbd, i've got an 'interesting' one here
[22:40] <groz> ifconfig -a shows wl0.3
[22:40] <groz> and i'm not sure why it's there, it's not up
[22:40] <nbd> weird
[22:40] <groz> but i have no idea what it is, or where it came from
[22:40] <nbd> normally it should only show up if you add 4 virtual interfaces
[22:41] <nbd> are wl0.2, wl0.1 also there?
[22:41] <groz> http://pastebin.ca/148007
[22:41] <groz> that's output from ifconfig -a
[22:41] <nbd> really weird
[22:42] <nbd> shouldn't be a problem, though
[22:42] <groz> well, who knows what kind of crap is old and in nvram on this box
[22:42] <nbd> it doesn't use nvram
[22:42] <groz> but i'm thinking, nvram not actually used anymore, or is it ?
[22:42] <nbd> at least not for setting up interfaces
[22:43] <groz> its probably something my program has done, the ting it, what, i'll have to go look
[22:43] <groz> but it's working
[22:43] <groz> i can associate and load google with notebook and with the nokia
[22:43] <groz> just getting ready to start loading on a couple wds connections
[22:43] <nbd> :)
[22:43] <groz> and see what happens
[22:43] <nbd> run wlc vif 3 ssid
[22:43] <nbd> to see if it returns anything
[22:44] <groz> nope
[22:44] <groz> so that would be a virtual interface then ?
[22:44] <nbd> normally it uses wl0 for the primary one, wl0.1 for the secondary one, etc.
[22:45] <groz> ok, is there an equivalent to wl status for the virtual guys ?
[22:45] <nbd> no
[22:45] <nbd> not yet
[22:45] <groz> ok, just asking, looking for how i 'get info' outa this funny thing :)
[22:46] <nbd> are you using client mode?
[22:46] <groz> hmm, i do cycle thru it at one point
[22:46] <groz> and then back
[22:47] <nbd> because status was only for client mode, right?
[22:47] <groz> i do scans every now and then
[22:47] <nbd> ah
[22:47] <groz> wl status does return valid info in ap mode
[22:47] <nbd> interesting
[22:47] <groz> i could never figure out how to get signal strengths on other aps cept by scanning in client mode
[22:48] <groz> so, i switch to client mode for 2 seconds, do a scan, then switch back
[22:49] <groz> and that part does seem to be working ok, cuz i've got scan results squirreled away
[22:49] <groz> hehe, this is one of my main interests in the virtual adapters
[22:49] <groz> leave one in client mode
[22:49] <groz> for scans
[22:50] <groz> but i'm curious how that'll affect ap operation
[22:50] <groz> if a client mode vif scans
[22:50] <nbd> if you switch to sta, then all other vifs die
[22:50] <nbd> there's a special 'apsta' mode
[22:51] <groz> oh, i thought it was possible to have one vif as an ap, and another in client mode
[22:51] <nbd> only with a special mode
[22:51] <groz> that would elimintate the need for wds connections
[22:51] <nbd> wds is much better than client mode if you intend to do bridging
[22:52] <groz> in some cases ya, bridging
[22:52] <groz> in other cases, it would be routing
[22:52] <nbd> because the implicit layer2-nat done in the 'client bridge' mode sucks
[22:53] <groz> hehe, you will like my 'concept', it's devious
[22:53] <groz> see, i got a few here, and, i also have a couple of 'linksys' visible
[22:53] <groz> so, the concept is
[22:53] <groz> i'll have one of them connect to a 'linksys' , voila, more bandwidth on my net connection
[22:53] <groz> then, i can use filtering to send all p2p traffic out the neighbors ap
[22:53] <nbd> hehe
[22:54] <groz> but i guess i can dedicate a station mode unit to that, would be nice to combine it all on one tho
[22:54] <groz> hehe
[22:55] <nbd> in that case, the apsta mode might work
[22:55] <groz> imagine that, my wrt sitting here, 2 vifs in sta, one in ap mode
[22:55] <nbd> that's not possible
[22:56] <groz> each of the 2 as sta connected to a different neighbors ap
[22:56] <groz> oh, ok, asking to much eh ?
[22:56] <nbd> you can only have one sta
[22:56] Action: groz hands nbd a 'per wrt' and reminds him, i got a dozen of them in the apartment
[22:57] <nbd> hehe
[22:57] <nbd> with sta i meant sta vif
[22:57] <groz> so, they can each have a sta, and then backhaul it all with wds
[22:57] <groz> that'll work
[22:57] <nbd> are your neighbors on the same channel?
[22:57] <groz> lemme look it changes from time to time
[22:58] <nbd> then you should use cable to connect the two sta routers
[22:58] <groz> i can already do that
[22:58] <groz> dont need special new toys for that
[22:58] <groz> but, to get the best connection, i gotta put one wrt out on the balcony, and it's a pain to run cables out there
[22:59] <groz> when folks tell me to use cables, my standard response is 'what part of the word wireless is unclear??'
[23:01] <groz> anyways, what this is doing, under the covers, it'll do this once in a while
[23:01] <groz> wl ap 0; wl scan; sleep 2; wl scanresults; wl ap 1
[23:01] <groz> that's probably the bit that's somehow triggering the wl0.3
[23:01] <groz> i have some work to do to convert this all to using wlc
[23:02] <nbd> [florian]: same bogomips issue with the old compiler
[23:02] <nbd> trying a new one now
[23:11] <[florian]> nbd: ok
[23:15] <nbd> btw. broadcom uses a nonstandard utility to lzma-encode the kernel binaries for cfe
[23:15] <nbd> i've written a weird shell replacement for it that uses the standard lzma: http://downloads.openwrt.org/people/nbd/tmp/cmplzma.sh
[23:16] <nbd> for some reason the standard format has 8x \xff at offset 5
[23:16] <nbd> and the broadcom format doesn't have it there
[23:20] <[florian]> nbd: do you know this useful site : http://skaya.enix.org/wiki/BroadCom96345 ?
[23:20] <nbd> interesting
[23:24] <groz> it doesn't load here
[23:28] <nbd> it also doesn't load here anymore
[23:28] <[florian]> with your custom lzma'd kernel ?
[23:28] <groz> it loaded here now
[23:28] <nbd> the web page
[23:29] <[florian]> ah ok :)
[23:29] <nbd> still rebuilding the whole thing with gcc 4.0.3
[23:29] <groz> why are so many folks re-inventing this wheel ?
[23:29] <[florian]> whoah
[23:29] <[florian]> groz: what do you mean ?
[23:29] <nbd> groz: because they don't know any better
[23:29] <groz> yet another batch going the hard way, starting with a box 'break in' then trying to build from scratch
[23:29] <groz> reading this page about this router
[23:30] <[florian]> yes
[23:30] <groz> mbm and i went down that road a couple years ago, the vast majority of the hard part is encapsulated in openwrt
[23:31] <[florian]> yes
[23:32] <[florian]> nbd: I did not write yet the "addpatern" for bcm63xx, altough it should be quite simple I hope
[23:32] <nbd> what kind of addpattern?
[23:33] <[florian]> the program that produces valid images, description here : http://skaya.enix.org/wiki/FirmwareFormat
[23:34] <[florian]> I go sleeping, so you tomorrow
[23:34] <nbd> see you tomorrow
[00:00] --- Fri Aug 25 2006
[00:00] <frop> yep
[00:00] <frop> then arrive "install", isn't it?
[00:00] <nbd> install comes after all compile steps
[00:00] <frop> yep
[00:00] <nbd> and only for packages marked as <*>
[00:01] <frop> ok...here it's M
[00:01] <frop> should i try it as * ?
[00:01] <frop> or you don't need it?
[00:01] <nbd> no, i don't need it at the moment
[00:01] <frop> k
[00:02] <nbd> found some more postinst brokenness
[00:05] <frop> $ svn up
[00:05] <frop> At revision 4643.
[00:05] <CIA-17> nbd * r4644 /packages/utils/strace/Makefile: cleanup
[00:05] <frop> but openntpd has the same behaviour
[00:05] <CIA-17> nbd * r4645 /packages/net/ (atftp/Makefile tor/Makefile): fix some more postinst brokenness
[00:05] <nbd> same as before?
[00:05] <frop> yep
[00:05] <nbd> try to rebuild it manually
[00:05] <nbd> maybe it doesn't detect file changes because of the symlink
[00:05] <frop> ishould i "clean" something other?
[00:06] <nbd> make package/openntpd-{clean,compile}
[00:06] <frop> yes, already done
[00:06] <nbd> after svn update in the packages directoryt
[00:06] <frop> ohhh
[00:06] <frop> can't i submit it from openwrt/
[00:06] <frop> ?
[00:06] <nbd> yes
[00:07] <Bartman007> from openwrt/ yes, but not from openwrt/branches/buildroot-ng/openwrt/
[00:07] <frop> ahhh
[00:08] <frop> :)
[00:08] <frop> openwrt/branches/buildroot-ng/packages then
[00:08] <Bartman007> actually those packages are in openwrt/packages/
[00:08] <Bartman007> there is also openwrt/branches/buildroot-ng/openwrt/package
[00:09] <frop> i got openwrt/ and packages/
[00:09] <frop> in openwrt/packages/ there're symlinks to ../../packages/something
[00:10] <nbd> then try svn update in packages/
[00:10] <Bartman007> I was speaking of the locations relative to where they are in the repository
[00:11] <frop> nbd: GRRR you are right
[00:11] <frop> fuck i'm not used still to buildroot-ng and packages
[00:11] <frop> :P
[00:11] <frop> sorry
[00:21] <frop> it get installed now :P
[01:13] <CIA-17> nbd * r4646 /branches/buildroot-ng/openwrt/target/linux/brcm-2.6/config: enable split_iso for ehci on brcm-2.6
[01:17] <CIA-17> nbd * r4647 /branches/whiterussian/openwrt/package/wificonf/wificonf.c: don't convert mw to qdbm for wl0_txpwr, set the value as-is
[05:04] <Bartman007> Kaloz: ping
[05:08] <Bartman007> h3sp4wn: ping
[10:36] <[florian]> nbd: what is the state of the multissid feature with wl ?
[10:36] <nbd> it should work
[10:36] <[florian]> even under win32 ?
[10:37] <nbd> sometimes
[10:37] <[florian]> ah, I have not yet tested under Linux
[10:48] <[florian]> nbd: I don't think I will manage to get assoclist working by myself unfortunately :(, I have tried various codes and they don't seem to work :(
[10:51] <nbd> what kinds of code?
[10:51] <nbd> btw. you should first try to implement it in user space
[10:51] <nbd> makes things easier
[10:52] <nbd> when you're done, move it to the kernel
[10:55] <[florian]> ok
[10:56] <[florian]> which library provides access to wl_ioctl's ?
[10:57] <[florian]> well, I will find it by reading the wificonf makefile
[10:57] <nbd> copy some code from wificonf
[10:57] <nbd> it's easy
[11:05] <[florian]> ok
[11:28] <[florian]> nbd: did you answer the guy that asks for captive portal informations or so ?
[11:28] <nbd> not yet
[11:28] <nbd> did you?
[11:30] <[florian]> neither did I, I don't really understand what he expects from us actually
[11:39] <nbd> groz: ping
[11:39] <groz> pong
[11:39] <nbd> i have another potential fix for the google issue with broadcom-wl
[11:39] <nbd> can you try it?
[11:40] <groz> ok, i'm dealing with another set of gotchas here, it'll be a little while before i can try it
[11:40] <groz> i got a busted network, trying to figure out why dhcp not working on it right now
[11:40] <nbd> when you have time, please install wl and run 'wl vlan_mode 0'
[11:40] <groz> ok, so just check it out as it is now
[11:40] <groz> then ipkg install wl
[11:40] <nbd> yes
[11:40] <groz> and run wl vlan_mode 0
[11:40] <nbd> yes
[11:40] <groz> ok, i'll get to it shortly
[11:40] <nbd> people in the dd-wrt forum report that it works
[11:41] <groz> i'll have to build and flash
[11:41] <groz> it wont take long
[11:41] <nbd> ok
[11:41] <nbd> thanks
[11:41] <groz> but right now, gotta get my dhcp server fixed first, i got a network with things dropping offline
[11:41] <nbd> yeah
[11:41] <nbd> i'll add support for that call to wlc in the mean time
[12:18] <groz> nbd, is there any particular hardware version i should be trying for this first test ?
[12:18] <nbd> no
[12:18] <groz> ok, the one on my desk is gs V2.1
[12:18] <nbd> good
[12:18] <groz> i'll be a few minutes yet
[12:19] <groz> but I found my network problem just validating it's all fixed now
[12:19] <groz> some idiot (that would be me) configured a dhcp server to not acknowledge 'new clients'
[12:19] <nbd> :)
[12:19] <groz> well, i'm experimenting with a slick new setup here
[12:19] <malbon> [florian]: I got my dg834gt today.
[12:26] <CIA-17> nbd * r4648 /branches/buildroot-ng/openwrt/package/broadcom-wl/src/wlc/wlc.c: add vlan_mode to wlc
[12:31] <malbon> groz: sounds secure :)
[12:36] <groz> malbon, actually, it's anything but
[12:36] <groz> im building a monster network of umls
[12:40] <CIA-17> kaloz * r4649 /branches/whiterussian/openwrt/target/linux/linux-2.4/patches/generic/902-hot_fixes-32.7.patch: add new hotfixes patch
[12:42] <malbon> uml, that's Xen on the cheap? ;)
[12:42] <groz> if you only want linux targets, it's far more efficient
[12:42] <h3sp4wn> Kaloz: Did you get round to finishing the image (no probs if you didn't)
[12:42] <groz> there is no emulation layer involved
[12:42] <malbon> in uml or xen?
[12:43] <groz> in uml
[12:43] <groz> it's just 'more threads'
[12:43] <malbon> there is no emulation in xen either.
[12:43] <malbon> (well unless you use hvm)
[12:43] <Kaloz> nope.. but i really should get some time for it today or tomorrow
[12:43] <groz> doesn't xen emulate the hardware ?
[12:43] <malbon> nope not with native guests.
[12:44] <malbon> it's more efficient than uml in my experience, but less flexible.
[12:44] <groz> i've never tried it, but, for what i'm doing, uml works very well
[12:45] <groz> simulating a network with a couple hundred nodes, and, serious propogation delays between them
[12:45] <groz> use teh qos network delay settings for that
[12:45] <malbon> yes, if it works for you, why not. :)
[13:00] <nbd> groz: what kind of software are you running on the nodes? i mean what part specifically are you going to test?
[13:01] <groz> what i'm working up nbd, is a method of simulating long hop radio links
[13:01] <nbd> interesting
[13:01] <groz> long hop in many cases includes satellite round trip times
[13:01] <groz> so for those nodes, i need to introduce 600ms delays
[13:01] <groz> so here's how i'm setting it all up
[13:02] <groz> i have a few vlans, with uml's conntecting to vlans by briding the uml switch with the vlan
[13:02] <groz> then I have a couple of 'real computers' also in the mix
[13:02] <groz> we are setting up a simulation of a rather complex network based on satellite links
[13:02] <groz> to test it for 'useablity' to carry voip
[13:02] <groz> with all the inherent satellite delays
[13:03] <nbd> very interesting
[13:03] <groz> since this company has NO communication now
[13:03] <groz> other than satellite phones at a buck a minute
[13:03] <groz> this could be a very interesting setup when done
[13:03] <nbd> yeah
[13:03] <nbd> definitely
[13:03] <groz> what i'm doing, is basically putting sattellite round trip delays
[13:03] <groz> between a few asterisk boxes
[13:03] <groz> then connecting sip phones to the asterisks on 'local networks'
[13:04] <groz> that's one simulation i'm using
[13:04] <groz> the other, i'm gonna end up simulating a 200+ node wireless mesh
[13:04] <nbd> what kind of routing are you using in the mesh?
[13:05] <groz> proprietary stuff that a client uses
[13:05] <nbd> ah
[13:05] <groz> doing a scale up simulation of thier existing setup
[13:05] <groz> currently the largest deplyed system only has 48 nodes
[13:06] <nbd> the freifunk people are always complaining about people testing mesh protocols only in simulations and not in real deployments
[13:06] <groz> well, we are doing it the other way around
[13:06] <groz> we have deployments, and know what the issues are
[13:06] <nbd> that's good
[13:06] <groz> and now, are testing to see how far we can scale up
[13:07] <nbd> berlin has a working 300+ node mesh
[13:07] <groz> these guys setup is pretty slick, it's self healing, self routing, just take the nodes out, plug em into power and let the magic do it's thing
[13:08] <groz> they use it today mostly to do wifi coverage of places like campgrounds, where strining ethernet to ap's is prohibitively expensive
[13:08] <nbd> yeah
[13:08] <nbd> the main problem of the berlin mesh is that almost nobody bothered to implement the ad-hoc mode properly
[13:09] <nbd> and there are a few issues in the 802.11 standard that also make it very easy to cause trouble
[13:09] <groz> this doesn't use ad-hoc
[13:09] <nbd> yeah, i figured
[13:09] <groz> its using wds connections
[13:09] <groz> so that the aps appear to be 'normal ap' to end users
[13:10] <groz> and windows will just go 'found a wireless network'
[13:10] <groz> when customers are in the locations
[13:10] <nbd> probably with at least two wifi cards on every node, right?
[13:10] <groz> depends on the rollout, but, that's one way of implementing
[13:11] <groz> i do all of my testing of stuff here using one per node, cuz i test on wrt54
[13:11] <groz> now i'm migrating testing into a swack of umls
[13:12] <nbd> because i think if you're using wds and you have clients and mesh neighbors on the same channel, it can get pretty traffic-intensive
[13:12] <groz> well, when you think it' thru
[13:12] <groz> it's being used for internet accesss
[13:13] <groz> and normally fed by a 1.5 mbit dsl or slower sat link in the remote locations
[13:13] <groz> so, the overhead on the wifi for meshing
[13:13] <nbd> yeah, in a setup like that it doesn't matter
[13:13] <groz> is plenty of room for that
[13:13] <groz> 54mbit actually works out to a REAL 10 mbit of thruput in general
[13:14] <groz> where you start to run into issues, is when you get out on the far reaches of the mesh
[13:14] <groz> there's a lot of latency
[13:14] <groz> due to the number of radio hops
[13:14] <groz> a lot of these locations, the only choice for inet access is satellite to begin with
[13:15] <groz> so the question becomes this
[13:15] <groz> how far out can you go from the main downlink, before it's better off to start with another downlink
[13:16] <groz> back in 5, gotta restart a router
[13:28] <nbd> groz: btw. this vlan_mode thing has a very high chance of working
[13:29] <groz> i've got a build just starting, then i gotta flash
[13:29] <nbd> k
[13:40] <groz> nbd, one other issue we were having
[13:40] <groz> was that a centrino laptop was unable to associate with it
[13:40] <groz> any chance the same thing will fix that too ?
[13:41] <groz> i'm guessing this all may be revolving around vlan tags in the wifi packets
[13:41] <groz> from your fix
[13:41] <nbd> i don't know if it'll fix centrino
[13:42] <groz> ok, just getting set up here to tftp on the build
[13:47] <groz> hmm, i seem to have built a broken image
[13:47] <groz> lemme try this again
[13:53] <groz> nbd, dumb question
[13:53] <groz> where in menuconfig do i find the 'propriety broadcom driver'
[13:53] <groz> cuz i seem to have lost it
[13:54] <groz> oh wait
[13:54] <groz> i found it back
[13:54] <groz> explains why this wasn't working
[13:54] <groz> i had it configured out
[14:05] <nbd> groz: seems like i made a mistake in the network scripts. fixing that
[14:13] <[florian]> malbon: great, have you tested yet bcm63xx ?
[14:14] <groz> I'm having some other issues here, not quite sure yet what's going on, but, i'm getting a very broken build, so i just blew away the tree and checked out a fresh copy to try again
[14:15] <nbd> groz: network scripts are broken and don't bring up the interface properly
[14:15] <nbd> groz: i'll have it fixed in a minute
[14:16] <nbd> groz: or is it broke in a different way?
[14:16] <groz> i think my main break was some other hacks i had in from earlier playing
[14:16] <groz> so i just started fresh to be able to do this test
[14:17] <groz> ahh, i see commit messages are not showing up here today
[14:17] <groz> that's why i was wondering 'whats up', you have committed some changes
[14:20] <CIA-17> nbd * r4650 /branches/buildroot-ng/openwrt/package/base-files/ (4 files in 4 dirs): fix remaining *_ifnames references
[14:20] <nbd> fixed now
[14:21] <groz> heh, and i see it here now too
[14:21] <groz> magic
[14:22] <Kaloz> so all the wifi-related stuff is fixed now?
[14:22] <groz> well, lets wait till i got it on here
[14:22] <groz> and i can validate
[14:22] <groz> before we claim it all fixed
[14:22] <nbd> Kaloz: wds is not implemented yet
[14:22] <nbd> Kaloz: and i think this 'fix' won't do anything to help with the centrino stuff
[14:22] <groz> so this build wont do wds ?
[14:22] <Kaloz> groz: btw, it's kinda funny that you were not able to conenct with a centrino card
[14:23] <nbd> groz: it's not implemented in the scripts yet
[14:23] <groz> it wasn't me kaloz, its a friend of mine up north
[14:23] <groz> and if it 'works' here
[14:23] <Kaloz> because maybe the draft-n thingie is working, just my notebook's centrino wasn't able to connect to it :p
[14:23] <groz> i'll have him flash a router later this morning, and try it
[14:23] <nbd> Kaloz: haha
[14:23] <nbd> groz: wlc can set up wds, but the scripts don't set it up from the network config yet
[14:23] <Kaloz> nbd: don't laugh, i can blame you all the time :P
[14:23] <groz> well we had that issue with the ng build, his notebook wouldn't connect
[14:23] <nbd> Kaloz: go ahead
[14:23] <groz> but, all his other stuff would
[14:24] <Kaloz> nbd: i didn't say i want to :P
[14:24] <nbd> hehe
[14:24] <groz> oknbd, if it's just the script sfor wds, wont matter here
[14:24] <Kaloz> nbd: i do make mistakes as well all the time
[14:24] <nbd> i wasn't laughing at you btw. just laughing at the funny situation
[14:25] <Kaloz> well, probably i would have mor eluck with an athros
[14:25] <Kaloz> but HP locks the bios down
[14:25] <nbd> this lockdown thing really sucks
[14:25] <Kaloz> and changing the pci id to pass the bios and then modify madwifi all the time just to be able to pay with that is stupid
[14:25] <Kaloz> and my centrino works quite nicely
[14:26] <Kaloz> the 2915abg that is
[14:26] <nbd> i've always had problems with centrino in a friend's laptop
[14:26] <nbd> 2100 was really crappy
[14:26] <nbd> 2200 worked somewhat better
[14:26] <nbd> but still not that great
[14:26] <groz> as a point of reference, that same laptop has no problems when we have a wr derivative flashed on
[14:27] <nbd> groz: i've heard about these centrino problems before
[14:27] <nbd> groz: from dd-wrt
[14:29] Action: Kaloz never had problems before with centrino stuff
[14:29] Action: groz thinks that's cuz kaloz has never _tried_ centrino stuff
[14:31] <Kaloz> .....
[14:32] <nbd> groz: how's the test going? :)
[14:32] <groz> I'll be reflashing here in a minute
[14:32] Action: groz doing 3 things at once
[14:33] <nbd> you could try this diff then: http://downloads.openwrt.org/people/nbd/tmp/diff.txt
[14:35] <groz> nbd, it just took the tftp, now to see if i built one that'll actually start
[14:35] <nbd> did you remember to rebuild base-files manually?
[14:35] <nbd> because the autorebuild is not fixed there yet
[14:35] <groz> i built 'from scratch'
[14:35] <nbd> ok
[14:46] <[florian]> there are less and less whiterussian tickets, sound good :)
[14:50] <groz> you know, i really hate that default firewall script, when i flash one of these with 'default' stuff, gotta go fuss with cables and such to get connected on the 4 port, then, after I make the changes and reboot it, connected back on the other side
[14:51] <groz> it's just in time to realize it was'nt finished firstboot when i did that
[14:51] <groz> and i get to start over
[15:00] <groz> nbd ping ?
[15:00] <groz> with this new setup, strait out of the box
[15:00] <groz> it should be just acting as a simple masquerading router no ?
[15:00] <nbd> yes
[15:00] <nbd> with dhcp on wan
[15:00] <groz> right now, it doesn't seem to be forwarding anything from my notebook
[15:01] <groz> i can ping 1.1 but nothing beyond, and a quick look
[15:01] <groz> i cant see why
[15:01] <groz> but ssh in, i can ping all over
[15:01] <nbd> does ns lookup work?
[15:01] <groz> yes
[15:01] <groz> i've got dns on the notebook
[15:02] <groz> well, wait, let me verify it's not cached
[15:02] <groz> yup
[15:02] <groz> notebook is talking to router, and getting dns
[15:02] <groz> but it cant ping past the router
[15:02] <groz> and the iptables ruels seem ok
[15:02] <Kaloz> groz: this is a centrino notebook? :P
[15:03] <Kaloz> groz: ip_forward?
[15:03] <groz> no
[15:03] <groz> yes, ip_forward is 1
[15:03] <nbd> did the MASQUERADE rule trigger?
[15:03] <nbd> do iptables -L -v
[15:04] <groz> iptables -t nat -L shows a masqerade rule
[15:04] <groz> iptables -L -v doesn't but i dont expect it there
[15:05] <groz> is there a quick and easy way with all this new network stuff to put it back to using vlan names on the adapters
[15:06] <groz> so i can drop in something known to work well ?
[15:09] <Kaloz> groz: iptables -t nat -L -v
[15:09] <Kaloz> and chekc if the nat rule is triggered imho :)
[15:09] <groz> yes, it is
[15:09] <groz> there is a masquerade rule there
[15:09] <Kaloz> the question is if its trggered
[15:09] <Kaloz> (eg. packet count)
[15:10] <groz> yah, seems to be increasing
[15:11] <Kaloz> i would say something is fishy there with networking
[15:11] <Kaloz> check it with tcpdump
[15:11] <groz> hang on, i'm checking a few more things
[15:11] <groz> ok, i changed policy on FORWARD to ACCEPT
[15:11] <groz> and away i go
[15:12] <Kaloz> groz: can you check it with a centrino? :)
[15:12] <groz> ok, got expected results, yahoo load, google does not
[15:12] <groz> now i try that wl trick
[15:12] <groz> it was wl vlan_mode 0 correct nbd ?
[15:13] <groz> woooohoooooo
[15:13] <groz> google loads
[15:13] <groz> i cant with the centrino right now kaloz, but, in another 4 hours denis will be up
[15:13] <groz> and i can get him to flash a box, and try it
[15:13] <Kaloz> okie
[15:14] <groz> ok, this google fix, for me, it's huge
[15:14] <CIA-17> nbd * r4651 /branches/buildroot-ng/openwrt/package/broadcom-wl/files/lib/wifi/broadcom.sh: add vlan_mode 0 command to fix the weird google issue in broadcom-wl
[15:14] <groz> now, if centrino will associate, it'll be even hugeer
[15:14] <groz> nbd, the vlan_mode read back -1 before i set it
[15:14] <groz> and google did NOT load
[15:14] <groz> i set it
[15:14] <groz> and google came right up on the notebook
[15:15] <nbd> grat
[15:15] <nbd> great
[15:15] <groz> after i changed the forward policy to actually forward :)
[15:15] <nbd> is there a bug in the firewall script?
[15:15] <groz> not sure, i have to go look why, i wasn't really interested in that right now, more interested in google issue :)
[15:15] <groz> all i did was change the polich
[15:16] <groz> iptables -P FORWARD ACCEPT
[15:16] <nbd> yes, there is a bug
[15:16] <groz> first thing i normally do, is toss away all that
[15:16] <groz> and introduce my own
[15:16] <groz> so, i dont normally even see what it does
[15:23] <CIA-17> nbd * r4652 /branches/buildroot-ng/openwrt/package/base-files/default/lib/network/config.sh: default value for *_device in scan_interfaces
[15:27] <groz> nbd, quick dumb question
[15:28] <groz> in this new network setup scripts, where are you setting the name 'br-lan' ?
[15:28] <groz> i cant find it right off
[15:32] <Kaloz> groz: /etc/init.d/S05netconfig
[15:32] <Kaloz> ?
[15:33] <groz> that's the setup, but, what's actually chaning the device name
[15:33] <groz> from br0 to br-lan
[15:33] <groz> that's what i cant find
[15:33] <nbd> /lib/network/config.sh
[15:34] <groz> k thanks
[15:34] <groz> ahah, got it
[15:46] <CIA-17> nbd * r4653 /branches/buildroot-ng/openwrt/package/base-files/default/ (etc/hotplug.d/net/10-net lib/network/config.sh sbin/ifup): revert to using *_ifnames internally, too many problems with firewall, ppp, etc. otherwise
[15:47] <groz> hehe
[15:47] <groz> ok, what did you change now ?
[15:48] <groz> i was just starting to look at some of the side effects of all the adapters having names changed
[15:48] <nbd> after running scan_interfaces, config_get lan ifname will return br-lan instead of the interface list
[15:48] <groz> ahh
[15:48] <nbd> (in the config, ifname specifies the list of interfaces in the bridge)
[15:49] <nbd> so that after scan_interfaces, you can always expect the ifname to contain the interface name that will be used for ip stuff
[15:49] <groz> I like this 'no dependancy on nvram' setup, but, it's goinna take a little thinking to get it all figured out correctly
[15:49] <nbd> yeah
[15:49] <nbd> the new scripts are a bit complex :)
[15:49] <nbd> but they're quite modular and the config files are easy to read and write
[15:49] <murb> the old setup scripts that assumed you had nvram were annoying
[15:49] <groz> yah, and, bash is not my best language, and i'm awful at the ash dialect thereof
[15:50] <groz> but im getting it figured out
[15:50] <nbd> the whole config abstraction takes a bit of getting used to, i think
[15:50] <nbd> especially if you use callbacks on config sections
[15:50] <groz> well, the 'issue' i still have, is the concept of 'wan/lan' and how it still seems to wrap around that
[15:51] <groz> cuz, there's potential for a lot more complex setups
[15:51] <groz> but once i get this config thing figured, i can probably see a way thru that
[15:51] <nbd> you can add other interface groups as well
[15:51] <groz> yah, was just thinking that as i typed it
[15:51] <nbd> the only special treatment of lan and wan is in the firewall stuff
[15:51] <nbd> and it's mostly because of wan
[15:52] <groz> well as an example, the router in front of me right now, it's running -ng built for x86 on a via board
[15:52] <groz> it's got a 'wan' eithernet, and a 'lan' ethernet
[15:52] <groz> but, it's also got a bunch of vlans
[15:52] <groz> so, one vlan is for one specific virtual machine group
[15:52] <groz> another vlan for another
[15:52] <nbd> if you set wan to where your internet uplink is, it's going to be ok
[15:53] <groz> and they could just as easily be separate wires
[15:53] <nbd> if you don't have a central uplink, drop it and do custom stuff
[15:53] <groz> yah, well, the first thing i always do is
[15:53] <groz> del firewall.*
[15:53] <groz> and then put in my own handling for iptables stuff
[15:53] <groz> but that's another totally different story unto itself
[16:14] <[florian]> nbd: there is not a nand chip driver for rb532 ?
[16:14] <nbd> i haven't added it yet
[16:14] <[florian]> ah ok
[16:14] <nbd> but i have a patch
[16:14] <[florian]> but it should work ?
[16:14] <nbd> yeah
[16:14] <nbd> but i heard that the nand chip is not that reliable
[16:14] <[florian]> really great
[16:14] <nbd> so i thought it would be better to start off with booting from cf only
[16:14] <[florian]> ah, I don't know much about the nand technology actually
[16:15] <nbd> me neither
[16:15] <thepeople_work> would it be posible to add the wp54 patches for openwrt kamikaze that compex has made?
[16:15] <[florian]> wp54 ?
[16:16] <thepeople_work> a compex wp54
[16:16] <thepeople_work> http://compex.com.sg/home/OEM/Open_wrt.htm
[16:16] <thepeople_work> the table of hardware is not correct for this device wither
[16:16] <thepeople_work> either*
[16:16] <nbd> thepeople_work: i will get hardware soon and then we'll clean up the patches and merge them
[16:17] <thepeople_work> ok, I have the hardware that is why I was wondering :-)
[16:17] <nbd> well, you'll get a fresh buildroot-ng port for it soon :)
[16:17] <thepeople_work> thanks, nbd
[16:25] <nbd> Hooligan0: hi
[16:48] <[florian]> nbd: we have the sources for libshared right ?
[16:51] <nbd> sure
[16:52] <[florian]> ok, I was digging into some GPL tarballs
[16:55] <[florian]> in the tiwlan.o binary file, I can see that the module licence is given as GPL, but we can't find the sources for this Wi-Fi driver, can we ?
[16:55] <murb> [florian]: oh nice catch!
[16:56] <murb> is that an acx1xx card?
[16:56] <[florian]> yes it is, there are references to Acx111
[16:56] <[florian]> you can find the file in the archive named WAG54GV2-EU_v1.00.19.tgz
[16:56] <[florian]> thanks to linux-mips.org
[16:57] <murb> I saw on lkml/netdev that they are refusing to merge the acx free drivers because of questions about where they came from, if they were legally reverse engineered etc.
[16:58] <murb> [florian]: does that include netfilter code?
[16:58] <[florian]> murb: I don't know yet
[16:58] <murb> i'm just wgetting it.
[16:59] <[florian]> murb: how can it use netfilter code by the way ? if it is a driver ?
[17:00] <murb> [florian]: i mean in that tgz
[17:00] <[florian]> ah you mean maybe the whole tarball :) right ::)
[17:00] <murb> prevent more netfilter work being done by mentioning it to gpl-violations...
[17:01] <murb> (i'd very much like a clean acx driver in the mainline kernel)
[17:01] <nbd> [florian]: the only gpl references in the driver tarball are in the WPA stuff and the config utility
[17:03] <[florian]> ok
[17:03] <[florian]> have to go, see you later !
[17:19] <groz> nbd, i'm almost finished the setup here, once its done, i'll deploy this new driver mod onto a batch of 8 wrts, and run some radio stress tests for a bit
[17:19] <nbd> great
[17:19] <groz> one of them is in client mode
[17:20] <groz> the rest are aps
[17:20] <groz> the aps all talking on wds links
[17:20] <groz> the client hops from one to the next
[17:20] <groz> as a client
[17:20] <groz> whole thing is scripted
[17:20] <nbd> the wlc wds command should be working
[17:20] <groz> at the end, it pops out some stats on overall radio thruput and error rates
[17:20] <groz> ok, this all uses wl tho
[17:20] <groz> it's a set of scripts i've had for a while
[17:21] <nbd> if you want to save space someday, you can replace the stuff with wlc
[17:21] <groz> im just making minor mods to it all to acomodate the fact it's dealing with eth0.x instead of vlanx
[17:21] <nbd> if any particular command is missing, tell me
[17:21] <groz> hehe, ok, but, i think i'm capable of adding one myself too if it's got me stuck
[17:22] <groz> i _do_ want to get switched over
[17:22] <groz> but, it takes time to migrate everything
[17:22] <nbd> yeah
[17:22] <groz> right now, the google thing has been a show stop, and, the centrino thing too
[17:22] <groz> i've tested google fix, will get denis to try centrino when he's up
[17:23] <groz> it's still early in the morning our time here
[17:23] <groz> i'm on a bit of a screwed up schedule, was flying yesterday, and got home around 4pm
[17:23] <groz> then crashed, and slept till midnite
[17:26] <groz> the next time you are playing in those init scripts
[17:26] <groz> i've got one 'request'
[17:26] <groz> i use a dynamic dns setup here, so, int he stanza defining the adapter
[17:26] <groz> would be cool to be able to add a name that it passes as the -H parameter to udhcpc
[17:26] <groz> flash up the box
[17:26] <groz> when it dhcps from the network
[17:26] <groz> ping openwrt will respond
[17:27] <groz> cuz dhcp updates the dns
[17:27] <groz> it belongs in the config stanza tho, because, there is valid times to have it use different names on different adapters
[17:27] <groz> rather than just take the name from hostname
[17:28] <nbd> you can set 'option hostname foobar' in a dhcp interface section
[17:28] <groz> oh, ok, you are already ahead of me
[17:28] <groz> i haven't got enough boxes flashed to play yet, but, rest assured, i will be playing a bunch with it before the day is done
[17:30] <murb> alternatively you could do a nsupdate, but keying requires sychonrised clocks.
[17:31] <groz> that would involve putting nsupdate on the client
[17:32] <nbd> groz: according to some people updating the ipw drivers fixes the issue with the new broadcom stuff
[17:32] <groz> rather the name is just in the dhcp, and do it ther,e for simplicity
[17:32] <groz> I'll have data n that later today nbd, it's still to early in the morning to phone denis
[17:32] <nbd> groz: hmm... for some people it still doesn't work
[17:33] <groz> what, the google, or the ipw ?
[17:33] <nbd> ipw
[17:33] <nbd> not tested with openwrt
[17:34] <nbd> tested with the linksys firmware
[17:34] <nbd> (wrt300n)
[17:47] <Bartman007> Kaloz: if you could post the link to your 7001 binaries with eth drivers I would be greatful. and I may have a armeb uclibc compile error for you in about a hours time.
[17:49] <Kaloz> nbd: so i need to grab some pccard to check it
[17:50] <Kaloz> Bartman007: well, today i don't know how will i have time for it (birtday partie) but i'll try
[17:51] <h3sp4wn> Kaloz: I can test the drivers if you have the link also
[18:27] <Bartman007> no problem, I don't mean to rush you
[18:33] <murb> btw http://www.yuri.org.uk/~murble/tiwlan.txt # they claim their module is GPLed.
[18:34] <nbd> murb: ah, i just noticed, that the source that is included in that tarball is the source for tiap.o
[18:34] <nbd> murb: the one for tiwlan.o is not included
[19:45] <nbd> [florian]: ping
[19:48] <groz> nbd, i just got a call from denis, he's up, but it's gonna be at least an hour or so before he can get to flashing a router
[19:48] <nbd> no problem
[19:48] <nbd> i can wait
[19:49] <groz> hehe, we dont have much choice now do we ?
[19:49] <nbd> i don't really expect the vlan_mode thing to fix the centrino assoc anyway
[19:50] <groz> ok, then i do have a mior request for wlc
[19:50] <groz> wlc fixe_centrino=on
[19:50] <nbd> i don't have a specific call to map it to :)
[19:50] <groz> just a minor request :)
[19:51] <nbd> and without knowing the problem i can't add patchtable entries like for the slot time or the no_ack flag
[19:51] <nbd> :)
[19:51] <groz> what do we have to do to identify the problem ?
[19:51] <nbd> no idea
[19:52] <nbd> have you seen the crazy patchtable stuff?
[19:53] <nbd> :)
[19:53] <groz> no, i haven't really stopped to look at it
[19:53] <groz> i just saw ther eis some really ugly hack on the go there
[19:53] <groz> seems like it's patching a driver in memory on the fly
[19:53] <nbd> yeah
[19:53] <nbd> and the build system generates a patch table by running objdump on wl.o and parsing it with a perl script :)
[19:55] <[florian]> nbd: pong
[19:56] <nbd> [florian]: do you have any major uncommitted bcm63xx changes?
[19:56] <[florian]> let me see
[19:56] <nbd> [florian]: because i was planning on hacking a bit on that port now
[19:56] <[florian]> ah really great
[19:56] <[florian]> do you have any bcm63xx device ?
[19:56] <nbd> yeah
[19:56] <nbd> a 96348
[19:57] <[florian]> ok, CFE or redboot based ?
[19:57] <nbd> cfe
[19:58] <[florian]> ok, really good, so that you can also try to patch the mtd map to use the mapping provided by CFE
[19:58] <[florian]> I have no major commit
[19:58] Action: murb wonders what chipset is in his sinus 1054 dsl thing
[19:58] <[florian]> nbd: there is one thing you can probably do, is fixing the bootloader detection
[19:59] <nbd> yeah, i was planning on doing that
[19:59] <[florian]> nbd: in the inventel livebox patch I found, it was done by parsing the arguments passed to prom_init
[20:01] <[florian]> nbd: here it is : http://pastebin.ca/147823
[20:04] <[florian]> I think there is now, with 2.6 kernels, a better way to get arguments passed by the bootloader
[20:06] <[florian]> this reminds me of getting my huawei hg550 from my parent's house
[20:10] <murb> hmm ok i have a 96345 here.
[20:49] <CIA-17> nbd * r4655 /branches/buildroot-ng/openwrt/include/prereq-build.mk: add prereq check to prevent people from building openwrt as root
[20:55] <Bartman007> nbd: haha, that would have screwed my laptop on saturday if I had run as root
[21:13] <CIA-17> nbd * r4656 /branches/buildroot-ng/openwrt/Makefile: hide rm -rf command in Makefile
[21:14] <nbd> hmm... just realized how ambiguous that commit message is :)
[21:16] <murb> nbd: will that stop me buliding it with fakeroot?
[21:16] <murb> or do you test for that?
[21:17] <nbd> no, but why use fakeroot?
[21:17] <nbd> it should be built as a regular user entirely
[21:40] <groz> nbd, you cannot do the things required to make a bootable image as a regular user
[21:40] <nbd> what kind of a bootable image?
[21:40] <groz> installing lilo onto a loop mounted image
[21:40] <groz> as an example
[21:40] <nbd> then we have to figure out a way to do it without using loopback
[21:41] <nbd> i don't want the build process to require root access at any point
[21:41] <groz> then you cant install boot loaders, doesn't really matter whihc you are trying to install for an x86 bootable image, lilo, grub take your pick
[21:41] <groz> they need to be loop mounted or they wont install
[21:41] <nbd> at the moment, yes
[21:42] <groz> they checkf for being on a block device, and refuse otherwise
[21:42] <nbd> we can fix that
[21:44] <groz> then when all is said and done
[21:45] <groz> you still have to write that image onto the media
[21:45] <nbd> yeah, but that isn't part of the build process
[21:45] <groz> well, i kept the 'make bootable' script out of the main build process
[21:45] <groz> for that reason
[21:45] <groz> it's gotta be run after the fact
[21:48] <nbd> also we should make the images somewhat size independent
[21:48] <groz> i did
[21:48] <nbd> great
[21:48] <groz> you can resize2fs the media
[21:48] <groz> after the fact
[21:48] <groz> but that only works if you dont partition it
[21:49] <groz> which you can do with a usb stick
[21:49] <groz> just write it on raw, treats it like a big floppy instead of like partitioned media
[21:49] <nbd> on rb532 i use block2mtd and a patched partition table parser
[21:49] <groz> it actually works on hd too
[21:49] <nbd> which maps all the free space after the partitions to a separate partition
[21:49] <nbd> and also reduces the size of a partition if it goes beyond the end of the device
[21:50] <groz> the way i make the images for usb keys, there are no paritions
[21:50] <groz> so, you can just resize2fs it, and it'll expand to fill the device
[21:50] <groz> i create an 8 meg image to start with
[21:50] <groz> dont think you can get smaller usb media
[21:50] <nbd> i do the same with rb532, only with jffs2
[21:51] <nbd> or maybe 16m, don't remember
[21:51] <groz> when using a usb key, no real point to jffs2
[21:51] <groz> they do wear levelling internally, and map to a 512 byte sector architecture externally
[21:51] <nbd> is it possible to create an ext2 fs that has exactly the size that it needs?
[21:52] <groz> it is, but the way i'm building it
[21:52] <groz> the same ext2 is used for both uml and for this
[21:52] <groz> and on uml, you want _some_ free space
[21:52] <nbd> well, you can run resize2fs automatically
[21:53] <groz> well, when using it with a uml, the size is fixed once you start it
[21:53] <groz> you cant expand it
[21:53] <groz> on the fly
[21:54] <groz> it's also possible, if you are using initramfs, to have nothing but the kernel on the media
[21:54] <groz> then run from the ramfs once up
[21:54] <groz> but that is only useful for a fixed configuratin where you never want to save config data between reboots
[21:54] <nbd> for the generic x86 port it would probably also be nice to use grub and allow the user to specify rootfs and kernel partition sizes in menuconfig
[21:55] <groz> i had thought of specifying the image size there
[21:55] <groz> i had issues getting grub to properly install
[21:55] <nbd> controlling the kernel partition size could be useful if the user plans on adding more space for different test kernelws
[21:55] <groz> the thing is when you loop mount, it cant read the geometry correctly, and, it wasn't picking up correctly the way lilo was
[21:55] <groz> after the fact
[21:55] <groz> keeping in mind, when you stick a given key into different computers
[21:56] <groz> the bios will give it totally different geometry
[21:56] <groz> with lilo you can force it to go lba, and, it 'just works'
[21:56] <nbd> when i have time, i could check out the grub source code
[21:56] <groz> i tried hard to get extlinux to do that
[21:56] <groz> cuz that would have been the ideal simple choice
[21:56] <groz> but
[21:57] <nbd> or maybe it would be better to use syslinux and fat
[21:57] <groz> it doesn't seem to be capable of doign lba
[21:57] <groz> if you do that, then the fat cannot become your root file system
[21:57] <groz> after the boot
[21:57] <nbd> it doesn't have to
[21:57] <nbd> we can have a separate partition
[21:57] <groz> i WANT my boot device as root
[21:57] <nbd> ok
[21:57] <groz> and i want it without partitions :)
[21:57] <nbd> ok
[21:57] <nbd> :)
[21:58] <groz> and i found a way to make it all 'just work'
[21:58] <groz> nothing stops this setup from working within a partition
[21:58] <groz> ie if you dd the image it currently creates for x86 into a partition
[21:58] <groz> then make the partition bootable
[21:58] <groz> it should just work
[21:59] <groz> and you should be able to resize it up to the size of the partition
[21:59] <groz> but you will have to edit the example lilo conf files i put in there
[21:59] <groz> to account for partiioned media
[22:00] <groz> there is another whole track one could take, and that's to do along the lines of what we have in the wrt
[22:00] <nbd> i think i'll give partitionless grub a shot first
[22:00] <nbd> hopefully without using loopback
[22:00] <groz> and do it with 2 paritions, then jump thru the same hoops
[22:00] <groz> well, i already started a menu section in menuconfig titled 'boot loaders'
[22:00] <groz> :)
[22:00] <nbd> :)
[22:00] <groz> i want to take a peek at uboot as well
[22:01] <groz> in theory,no reason we cant have various ones in there, with appropriate depends
[22:01] <groz> so, you see lilo and grub when building for x86
[22:01] <nbd> yeah
[22:01] <groz> and you see uboot/whathaveyou for other hardware
[22:05] <nbd> [florian]: ever noticed this?
[22:05] <nbd> Calibrating delay loop... 4.97 BogoMIPS (lpj=12448)
[22:05] <nbd> [florian]: on bcm63xx
[22:05] Action: nbd tries again with gcc 3.4.4 instead of 3.4.6
[22:10] <[florian]> nbd: ah, I don't remember seeing that
[22:11] <[florian]> nbd: have you succeeded in booting the current bcm63xx branch on your device ?
[22:11] <nbd> hint: it's the same problem that bcm47xx has with gcc 3.4.6
[22:11] <nbd> nah, it crashes while accessing the flash
[22:11] <nbd> most likely related to this bogomips issue
[22:12] <nbd> what does the bogomips count show on your device?
[22:12] <nbd> and with which compiler did you build it?
[22:13] <[florian]> nbd: I don't exactly remember, and I have not the device set up right now
[22:13] <nbd> k
[22:13] <nbd> would be interesting to know
[22:13] <[florian]> nbd: 3.4.6/2.17
[22:14] <nbd> i'll try 3.4.4/2.16 now
[22:14] <[florian]> ok
[22:14] <nbd> if it works, i'll try 3.4.6/2.17
[22:14] <[florian]> sounds weird to have gcc versions changing the bogomips calculations
[22:14] <nbd> well, it's a known problem on bcm47xx
[22:15] <nbd> it's the only reason why we're still using 3.4.4 for it in kamikaze
[22:15] <[florian]> ah, and do you have a fix ?
[22:15] <nbd> no
[22:27] <[florian]> nbd: confirmed, I had the same bug : Calibrating delay loop... 4.12 BogoMIPS (lpj=10304)
[22:28] <[florian]> using gcc-3.4.6
[22:28] <nbd> and this is probably the reason for the random crashes in various places
[22:29] <[florian]> that's what I was going to ask
[22:29] <nbd> the weird thing is that i've never seen this bug in linux 2.6 on bcm47xx
[22:29] <[florian]> ha such a pity
[22:30] <[florian]> why could it cause random crashes in various places ?
[22:31] <nbd> [florian]: because wrong bogomips messes up lots of parts in the kernel - flash access, hardware access
[22:32] <nbd> [florian]: all the timings are suddenly wrong
[22:32] <[florian]> ok, I did not know it was used like that too
[22:40] <groz> nbd, i've got an 'interesting' one here
[22:40] <groz> ifconfig -a shows wl0.3
[22:40] <groz> and i'm not sure why it's there, it's not up
[22:40] <nbd> weird
[22:40] <groz> but i have no idea what it is, or where it came from
[22:40] <nbd> normally it should only show up if you add 4 virtual interfaces
[22:41] <nbd> are wl0.2, wl0.1 also there?
[22:41] <groz> http://pastebin.ca/148007
[22:41] <groz> that's output from ifconfig -a
[22:41] <nbd> really weird
[22:42] <nbd> shouldn't be a problem, though
[22:42] <groz> well, who knows what kind of crap is old and in nvram on this box
[22:42] <nbd> it doesn't use nvram
[22:42] <groz> but i'm thinking, nvram not actually used anymore, or is it ?
[22:42] <nbd> at least not for setting up interfaces
[22:43] <groz> its probably something my program has done, the ting it, what, i'll have to go look
[22:43] <groz> but it's working
[22:43] <groz> i can associate and load google with notebook and with the nokia
[22:43] <groz> just getting ready to start loading on a couple wds connections
[22:43] <nbd> :)
[22:43] <groz> and see what happens
[22:43] <nbd> run wlc vif 3 ssid
[22:43] <nbd> to see if it returns anything
[22:44] <groz> nope
[22:44] <groz> so that would be a virtual interface then ?
[22:44] <nbd> normally it uses wl0 for the primary one, wl0.1 for the secondary one, etc.
[22:45] <groz> ok, is there an equivalent to wl status for the virtual guys ?
[22:45] <nbd> no
[22:45] <nbd> not yet
[22:45] <groz> ok, just asking, looking for how i 'get info' outa this funny thing :)
[22:46] <nbd> are you using client mode?
[22:46] <groz> hmm, i do cycle thru it at one point
[22:46] <groz> and then back
[22:47] <nbd> because status was only for client mode, right?
[22:47] <groz> i do scans every now and then
[22:47] <nbd> ah
[22:47] <groz> wl status does return valid info in ap mode
[22:47] <nbd> interesting
[22:47] <groz> i could never figure out how to get signal strengths on other aps cept by scanning in client mode
[22:48] <groz> so, i switch to client mode for 2 seconds, do a scan, then switch back
[22:49] <groz> and that part does seem to be working ok, cuz i've got scan results squirreled away
[22:49] <groz> hehe, this is one of my main interests in the virtual adapters
[22:49] <groz> leave one in client mode
[22:49] <groz> for scans
[22:50] <groz> but i'm curious how that'll affect ap operation
[22:50] <groz> if a client mode vif scans
[22:50] <nbd> if you switch to sta, then all other vifs die
[22:50] <nbd> there's a special 'apsta' mode
[22:51] <groz> oh, i thought it was possible to have one vif as an ap, and another in client mode
[22:51] <nbd> only with a special mode
[22:51] <groz> that would elimintate the need for wds connections
[22:51] <nbd> wds is much better than client mode if you intend to do bridging
[22:52] <groz> in some cases ya, bridging
[22:52] <groz> in other cases, it would be routing
[22:52] <nbd> because the implicit layer2-nat done in the 'client bridge' mode sucks
[22:53] <groz> hehe, you will like my 'concept', it's devious
[22:53] <groz> see, i got a few here, and, i also have a couple of 'linksys' visible
[22:53] <groz> so, the concept is
[22:53] <groz> i'll have one of them connect to a 'linksys' , voila, more bandwidth on my net connection
[22:53] <groz> then, i can use filtering to send all p2p traffic out the neighbors ap
[22:53] <nbd> hehe
[22:54] <groz> but i guess i can dedicate a station mode unit to that, would be nice to combine it all on one tho
[22:54] <groz> hehe
[22:55] <nbd> in that case, the apsta mode might work
[22:55] <groz> imagine that, my wrt sitting here, 2 vifs in sta, one in ap mode
[22:55] <nbd> that's not possible
[22:56] <groz> each of the 2 as sta connected to a different neighbors ap
[22:56] <groz> oh, ok, asking to much eh ?
[22:56] <nbd> you can only have one sta
[22:56] Action: groz hands nbd a 'per wrt' and reminds him, i got a dozen of them in the apartment
[22:57] <nbd> hehe
[22:57] <nbd> with sta i meant sta vif
[22:57] <groz> so, they can each have a sta, and then backhaul it all with wds
[22:57] <groz> that'll work
[22:57] <nbd> are your neighbors on the same channel?
[22:57] <groz> lemme look it changes from time to time
[22:58] <nbd> then you should use cable to connect the two sta routers
[22:58] <groz> i can already do that
[22:58] <groz> dont need special new toys for that
[22:58] <groz> but, to get the best connection, i gotta put one wrt out on the balcony, and it's a pain to run cables out there
[22:59] <groz> when folks tell me to use cables, my standard response is 'what part of the word wireless is unclear??'
[23:01] <groz> anyways, what this is doing, under the covers, it'll do this once in a while
[23:01] <groz> wl ap 0; wl scan; sleep 2; wl scanresults; wl ap 1
[23:01] <groz> that's probably the bit that's somehow triggering the wl0.3
[23:01] <groz> i have some work to do to convert this all to using wlc
[23:02] <nbd> [florian]: same bogomips issue with the old compiler
[23:02] <nbd> trying a new one now
[23:11] <[florian]> nbd: ok
[23:15] <nbd> btw. broadcom uses a nonstandard utility to lzma-encode the kernel binaries for cfe
[23:15] <nbd> i've written a weird shell replacement for it that uses the standard lzma: http://downloads.openwrt.org/people/nbd/tmp/cmplzma.sh
[23:16] <nbd> for some reason the standard format has 8x \xff at offset 5
[23:16] <nbd> and the broadcom format doesn't have it there
[23:20] <[florian]> nbd: do you know this useful site : http://skaya.enix.org/wiki/BroadCom96345 ?
[23:20] <nbd> interesting
[23:24] <groz> it doesn't load here
[23:28] <nbd> it also doesn't load here anymore
[23:28] <[florian]> with your custom lzma'd kernel ?
[23:28] <groz> it loaded here now
[23:28] <nbd> the web page
[23:29] <[florian]> ah ok :)
[23:29] <nbd> still rebuilding the whole thing with gcc 4.0.3
[23:29] <groz> why are so many folks re-inventing this wheel ?
[23:29] <[florian]> whoah
[23:29] <[florian]> groz: what do you mean ?
[23:29] <nbd> groz: because they don't know any better
[23:29] <groz> yet another batch going the hard way, starting with a box 'break in' then trying to build from scratch
[23:29] <groz> reading this page about this router
[23:30] <[florian]> yes
[23:30] <groz> mbm and i went down that road a couple years ago, the vast majority of the hard part is encapsulated in openwrt
[23:31] <[florian]> yes
[23:32] <[florian]> nbd: I did not write yet the "addpatern" for bcm63xx, altough it should be quite simple I hope
[23:32] <nbd> what kind of addpattern?
[23:33] <[florian]> the program that produces valid images, description here : http://skaya.enix.org/wiki/FirmwareFormat
[23:34] <[florian]> I go sleeping, so you tomorrow
[23:34] <nbd> see you tomorrow
[00:00] --- Fri Aug 25 2006