**** BEGIN LOGGING AT Tue Oct 12 02:59:57 2010
Oct 12 03:09:21 <CIA-39> jow * r23407 /trunk/scripts/remote-gdb: [scripts] remote-gdb: fix parsing of arch and libc from given path as well
Oct 12 06:14:14 <russell--> hmm.  seeing some weirdness from dnsmasq
Oct 12 06:15:02 <russell--> i have an "option local" line in /etc/config/dhcp
Oct 12 06:15:21 <russell--> which is supposed to redirect queries for certain domains to an external dns server
Oct 12 06:15:35 <russell--> (works on old-ish wgt634u build)
Oct 12 06:15:48 <russell--> find it doesn't work on newish atheros (ap51) build
Oct 12 06:17:03 <russell--> it reports in syslog that it is setting it up, e.g.: Oct 11 02:36:34 216.99.214.10 dnsmasq[717]: using nameserver 70.102.34.162#53 for domain ptp
Oct 12 06:17:08 <russell--> but, it doesn't
Oct 12 06:17:12 <russell--> like, work
Oct 12 06:17:18 <russell--> wht?
Oct 12 06:28:30 <xMff> russell--: "which is supposed to redirect queries for certain domains to an external dns server"
Oct 12 06:28:38 <xMff> no, it does the exact opposite
Oct 12 06:28:55 <xMff> everything amtching local is _not_ forwarded
Oct 12 07:01:03 <russell--> xMff: what I mean to say, i don't get resolutions for the domain ptp on atheros-r23351, but I do on brcm47xx-r18781 with the same configuration file.
Oct 12 07:01:32 <xMff> russell--: does it resolve to private ips?
Oct 12 07:02:03 <russell--> i am a client on that the device's network
Oct 12 07:02:07 <russell--> in both cases
Oct 12 07:02:46 <xMff> I mean, is *.ptp resolved to RFC1918 addresses?
Oct 12 07:03:08 <russell--> they are 10.11.x.y
Oct 12 07:03:20 <xMff> then you need to disable the rebind protection
Oct 12 07:03:48 * russell-- prepares a pastebin of the current config
Oct 12 07:03:49 <xMff> or whitelist ptp
Oct 12 07:04:28 <xMff> so either "option rebind_protection 0" (very broad) or "list rebind_domain ptp" (limited whitelisting)
Oct 12 07:05:43 <russell--> http://openwrt.pastebin.org/149857
Oct 12 07:05:51 <russell--> i think i may have already spotted the problem
Oct 12 07:07:33 <russell--> hmm. no, maybe not.
Oct 12 07:07:45 <xMff> again
Oct 12 07:07:52 <xMff> you need to disable rebind protection
Oct 12 07:07:59 <russell--> yeah, /me trying your suggestion
Oct 12 07:08:00 <xMff> it defaults to on if not set
Oct 12 07:09:10 <russell--> fixed, thanks.
Oct 12 07:09:53 <russell--> i can't understand how my ancient config files could possible not work today!!! ;-)
Oct 12 07:10:24 <xMff> there should've been a big warning in logread that RFC1918 responses are dropped
Oct 12 07:12:33 <russell--> Oct 12 00:00:34 test daemon.warn dnsmasq[717]: possible DNS-rebind attack detected: donk.ptp
Oct 12 07:12:51 <russell--> like that?
Oct 12 07:12:58 <xMff> yes
Oct 12 07:13:20 <xMff> its a new security feature
Oct 12 07:13:33 <CIA-39> kerneis * r23408 /packages/net/babel/ (Makefile patches/001-buggy-packed.patch):
Oct 12 07:13:34 <CIA-39> [package] babel: upgrade to version 1.0.2
Oct 12 07:13:34 <CIA-39> This version works around a gcc bug that would cause assertion failures
Oct 12 07:13:34 <CIA-39> on MIPS. Other than that, there are no user-visible changes.
Oct 12 07:13:34 <CIA-39> Signed-off-by: Gabriel Kerneis <kerneis@pps.jussieu.fr>
Oct 12 07:13:48 <russell--> cool
Oct 12 07:13:49 <xMff> actually its not new but it has been enabled by default recently
Oct 12 07:13:56 <russell--> right, makes sense
Oct 12 07:15:08 * russell-- looking at the dnsmasq changelog atm
Oct 12 07:16:38 <russell--> 2.51-1 vs 2.55-5
Oct 12 07:19:26 <acoul> hello
Oct 12 09:56:19 <builder_> build #6 of ramips is complete: Success [build successful]  Build details are at http://tksite.gotdns.org:8010/builders/ramips/builds/6
Oct 12 10:04:26 <russell--> hmm.  looks like atheros has a similar problem to one i solved on brcm47xx 1.5 years ago with perl signals?
Oct 12 10:04:42 * russell-- looks for my old bug report
Oct 12 10:04:53 <russell--> forked children don't get reaped
Oct 12 10:05:01 <russell--> hang around as zombies
Oct 12 10:12:34 <russell--> this was the fix back then:  https://dev.openwrt.org/changeset/16539
Oct 12 10:32:09 * russell-- going to blindly try applying the same change for mips as mipsel
Oct 12 10:55:22 <russell--> hmm. that seemed to work
Oct 12 10:55:47 <russell--> i did have a little trouble with opkg thinking that the rootfs was full
Oct 12 10:56:24 <russell--> had to use --force-space to convince it
Oct 12 10:56:36 <russell--> there was room in /overlay
Oct 12 11:02:05 <russell--> jow_laptop: http:www.personaltelco.net/~russell/mips-perl-signal.patch
Oct 12 11:02:46 <russell--> i'm not sure anyone else uses perl on these things but me!
Oct 12 11:12:45 <jow_laptop> russell--: this patch does the trick?
Oct 12 11:13:05 <russell--> not sure how/where the signal numbers are defined, i got these by autoconfiguring *on* a wgt634u using a cross-built toolchain on an nfs mounted root or something.  maybe it was a usb-mounted filesystem, it was a while ago.
Oct 12 11:13:17 <russell--> it makes the zombies go away
Oct 12 11:13:54 <russell--> it's possible different hell will break lose, but i haven't seen it yet
Oct 12 11:14:23 <dkostousov> Hello. I'm working with openwrt 8.09.1 kamikaze with luci 0.8.7 Which http server does luci use?
Oct 12 11:14:49 <jow_laptop> dkostousov: on 8.09.1 either busybox httpd or lucid
Oct 12 11:16:00 <dkostousov> Where can I find "root" index.html?
Oct 12 11:16:05 <jow_laptop>  /www
Oct 12 11:17:51 <dkostousov> where is it in source tree?
Oct 12 11:18:54 <russell--> dkostousov: if you just want to override something, put your files in files/www/
Oct 12 11:18:55 <jow_laptop> libs/sgi-cgi/htdocs/
Oct 12 11:20:29 <russell--> but i don't really know what dkostousov has in mind, so i'll just shut up now
Oct 12 11:21:13 <jow_laptop> ;)
Oct 12 11:24:58 <russell--> before i wander off to bed, i just want to say for the record that i remain *stunned* that perl (and nocatauth) runs on an 8meg flash/32meg ram device.  thank you.  i feel better.
Oct 12 11:28:58 <dkostousov> i want to disable caching for this page
Oct 12 11:33:36 <loswillios> meh, building with GLIBC has some errors: canPATH_MAX in limits.h http://pastebin.com/KN46S2yU
Oct 12 11:33:57 <loswillios> *can't find PATH_MAX although limits.h is included
Oct 12 11:38:20 <russell--> jow_laptop: i compared my patch against the first 30-odd signal names in linux-2.6's arch/mips/include/asm/signals.h and they match.
Oct 12 11:41:16 <dkostousov> How can I rebuild some packages?
Oct 12 11:42:54 <loswillios> dkostousov: make package/<name>/{clean,compile,install} V=99
Oct 12 11:46:56 <russell--> jow_laptop: as far as i can tell they *all* match, my confidence in the patch is high, there doesn't seem to be any indication that mips and mipsel signal definitions should not be the same, my patch makes them the same.
Oct 12 11:50:10 * russell-- wanders off to bed
Oct 12 12:07:10 <CIA-39> nbd * r23409 /trunk/package/mac80211/patches/020-nl80211_fix.patch: mac80211: fix a crash in the netlink backport
Oct 12 12:07:16 <CIA-39> nbd * r23410 /trunk/package/mac80211/patches/527-ath9k_ps_survey_fix.patch: ath9k: fix channel time updates when the interface is idle
Oct 12 12:08:16 <builder_> build #5 of uml is complete: Exception [exception shell_8 upload]  Build details are at http://tksite.gotdns.org:8010/builders/uml/builds/5
Oct 12 13:06:46 <builder_> build #5 of x86 is complete: Exception [exception shell_8 upload]  Build details are at http://tksite.gotdns.org:8010/builders/x86/builds/5
Oct 12 13:28:46 <CIA-39> acoul * r23411 /packages/net/click/ (3 files in 2 dirs): net/click: update to click-20101002 git (#8072)
Oct 12 14:01:13 <builder_> build #5 of xburst is complete: Exception [exception shell_8 upload]  Build details are at http://tksite.gotdns.org:8010/builders/xburst/builds/5
Oct 12 14:13:43 <CIA-39> nbd * r23412 /trunk/package/mac80211/patches/ (3 files): ath9k: fix a few ani issues
Oct 12 14:23:42 <oldcalvin> Can anybody help me how to debug the b43 wireless driver?
Oct 12 14:24:05 <oldcalvin> thanks a lot !
Oct 12 14:24:16 <builder_> build #7 of pxcab is complete: Success [build successful]  Build details are at http://tksite.gotdns.org:8010/builders/pxcab/builds/7
Oct 12 14:26:55 <CIA-39> acoul * r23413 /packages/net/udpxy/ (. Makefile): net/udpxy: add udpxy package. (closes ticket #8077)
Oct 12 15:00:50 <builder_> build #8 of brcm63xx is complete: Failure [failed compile_4]  Build details are at http://tksite.gotdns.org:8010/builders/brcm63xx/builds/8
Oct 12 15:29:52 <CIA-39> acoul * r23414 /trunk/target/linux/x86/image/Config.in: target/linux: fix Geos baudrate. (thank you Philip Prindeville)
Oct 12 16:01:53 <CIA-39> acoul * r23415 /trunk/target/linux/generic/ (4 files in 4 dirs): target/linux: Add support for cores with slow WDT clock like bcm5354. (thank you Aleksandar Radovanovic)
Oct 12 16:13:54 <builder_> build #7 of rb532 is complete: Success [build successful]  Build details are at http://tksite.gotdns.org:8010/builders/rb532/builds/7
Oct 12 16:15:20 <CIA-39> acoul * r23416 /trunk/target/linux/generic/config-2.6.36: generic/config-2.6.36: Add missing symbols
Oct 12 16:22:01 <builder_> build #9 of ppc40x is complete: Exception [exception shell_8 upload]  Build details are at http://tksite.gotdns.org:8010/builders/ppc40x/builds/9
Oct 12 16:37:49 <kristianpaul> Helo
Oct 12 16:37:52 <xMff> ping KanjiMonster
Oct 12 16:37:52 <kristianpaul> Hello**
Oct 12 16:38:42 <EqUaTe> xMff: based on past form, he'll likely be around in an hour or two.
Oct 12 16:39:22 <kristianpaul> An OpenWRT target is able to manage things, like custom version of packages already included in a repositty?  what about linux patches too? and custom root_fs too?
Oct 12 16:39:58 <xMff> kristianpaul: sounds like you want a subtarget or even a completely new arch
Oct 12 16:40:52 <xMff> however as a rule of thumb you should try to generalize stuff and shift build-time decisions to runtime detections where possible
Oct 12 16:41:49 <kristianpaul> well for pachages it could create in other git the modified versions
Oct 12 16:42:30 <kristianpaul> root_fs is not big deal, is mainly yo modify some defaults in /etc/profile and may be network and inittab
Oct 12 16:42:56 <kristianpaul> but linux seems to be the more dificult to support isnt xMff ?
Oct 12 16:43:03 <xMff> do you want to create an "official" port or something for your personal use?
Oct 12 16:43:23 <KanjiMonster> xMff: pong
Oct 12 16:43:36 <EqUaTe> wow, this is early for you KM.. not gone home yet? :P
Oct 12 16:43:46 <kristianpaul> well not personal if for electronic board wich uses a procesor support in openwrt already
Oct 12 16:44:18 <KanjiMonster> EqUaTe: just came home
Oct 12 16:44:21 <kristianpaul> is the SIE board, produced for soem guy in the qihardware comuinity, same from the Ben Nanonote come from
Oct 12 16:44:40 <xMff> KanjiMonster: I wondered whether you could explain some switch configs to me. I currently bring LuCI back in shape and want to rework the switch config.
Oct 12 16:44:42 <EqUaTe> ahh
Oct 12 16:44:56 <KanjiMonster> xMff: sure
Oct 12 16:45:11 <kristianpaul> just those guys used that make SIE hardware are not good in sofware maintenance so they made some ugly patch for the current openwrt-xbusrt branch
Oct 12 16:46:02 <xMff> kristianpaul: mirko and larsc maintain xburst, they may be able to help you
Oct 12 16:47:06 <kristianpaul> xMff: yeah sure they pointme to do some things in uptream like ie Linux, but besides that seems i dont have clear the meaning of a "Target" in openwrt
Oct 12 16:47:09 <xMff> KanjiMonster: ok, first for general understanding. Am I right that rtl8366s and rtl8366rb don't follow the same principles with regard to vlan config?
Oct 12 16:47:14 <kristianpaul> or subtarget?
Oct 12 16:47:37 <kristianpaul> can you explain it a bit or point me documentation about it?
Oct 12 16:47:54 <kristianpaul> or may be should  i ask in #openwtry if this is not the right place for
Oct 12 16:47:56 <larsc> kristianpaul: right now we have 3 subtrgets for xburst. it should be easy to add another for SIE
Oct 12 16:48:06 <KanjiMonster> xMff: ugh, good question. I didn't work with the realtek code (don't have one), so I can't really tell except by looking at the code.
Oct 12 16:48:11 <kristianpaul> larsc: hey :)
Oct 12 16:48:22 <kristianpaul> larsc: hanvon ebook is one of then?
Oct 12 16:49:03 <larsc> yes
Oct 12 16:49:06 <xMff> KanjiMonster: ok. Right now I try to understand the semantics of the pvid option for the switch_port section
Oct 12 16:49:54 <KanjiMonster> xMff: ah, thats easy. pvid (the "real" pvid ;) is the port vlan; this is (normally) the default vlan id added to untagged frames going into the port
Oct 12 16:50:09 <xMff> ah I see
Oct 12 16:50:31 <kristianpaul> larsc: okay i'll take a look, i just want make thing clear in case somebody want help with that messy hacks became more serios :)
Oct 12 16:50:36 <kristianpaul> Thanks
Oct 12 16:51:37 <xMff> KanjiMonster: so it basically only affects incoming, untagged frames?
Oct 12 16:53:28 <KanjiMonster> xMff: mostly
Oct 12 16:56:39 <KanjiMonster> xMff: some switches support a mixed tagged/untagged mode with all except one VID tagged, and the pvid defines the untagged VID
Oct 12 16:57:09 <xMff> hmm
Oct 12 17:00:37 <xMff> another thing
Oct 12 17:00:59 <xMff> vlan index and vlan id may differ?
Oct 12 17:01:09 <xMff> e.g. vlan 1 with three ports and vid 7
Oct 12 17:02:46 <xMff> I believe to recall something like "driver can only handle 16 vlans but supports ids up to 4096"
Oct 12 17:05:16 <xMff> I think I got it, the vlan number is like a name and an implicit result of the order of switch_vlan sections in the config
Oct 12 17:05:36 <xMff> the vlan id is what's specified by the "vlan" attribute
Oct 12 17:05:42 <KanjiMonster> right
Oct 12 17:06:01 <KanjiMonster> the ar8216 works the same way (also has only 16 entries in its table)
Oct 12 17:06:46 <xMff> ok and the need for empty "padding vlans" was due to bugs in swconfig?
Oct 12 17:06:47 <KanjiMonster> although it looks like the realtek switches do support a full 4096-entries mode (one for each vid)
Oct 12 17:06:57 <KanjiMonster> I think yes
Oct 12 17:07:46 <KanjiMonster> according to the datasheet, it handles frames with the VID 0 as untagged
Oct 12 17:21:50 <xMff> KanjiMonster: do you think such a layoud would represent the configuration methology correctly? http://luci.subsignal.org/~jow/sw.html
Oct 12 17:22:27 <xMff> *layout
Oct 12 17:25:53 <KanjiMonster> xMff: it represents the vlan table quite clearly. I would probably make "tagged/untagged" a port attribute, but this would need changing the current swconfig api or serious workarounds
Oct 12 17:27:05 <xMff> a port attribute?
Oct 12 17:27:25 <xMff> how would you make a port untagged in one vlan and tagged in another then?
Oct 12 17:27:39 <KanjiMonster> you can have at most one untagged vlan at a port
Oct 12 17:27:49 <xMff> a right
Oct 12 17:28:34 <xMff> so in the upper table if one cell has U set all others in the column must be locked to --- or T only
Oct 12 17:29:06 <xMff> this is trivial with a bit of javascript
Oct 12 17:29:51 <xMff> the the pvids in the lower table could be turned into a select which contains only vids from the table above
Oct 12 17:29:54 <KanjiMonster> yes, but I have to admit I don't know whether the atheros switches actually support this (or at least If the code supports this)
Oct 12 17:30:04 <xMff> or is it valid to set a pvid which does not exist as vlan?
Oct 12 17:30:47 <KanjiMonster> let me take a look at the code
Oct 12 17:31:06 <builder_> build #5 of ar7 is complete: Exception [exception shell_8 upload]  Build details are at http://tksite.gotdns.org:8010/builders/ar7/builds/5
Oct 12 17:31:08 <xMff> as you can guess vlans are not really my speciality...
Oct 12 17:31:36 <KanjiMonster> I remember something about the pvid being in index into the vlan table in the ar8216 code
Oct 12 17:31:42 <xMff> at least not the tagged / untagged stuff. I regarded that always as just a way to partition the switch ports
Oct 12 17:32:07 <xMff> the concept of multiple vlans over the same ports was beyound my imagination until recently :P
Oct 12 17:32:38 <xMff> well I think its reasonable to restrict possible pvid values to existing vids
Oct 12 17:32:52 <KanjiMonster> yes
Oct 12 17:33:10 <xMff> but on the other hand one could use use such an extra pvid to catch rogue traffic
Oct 12 17:33:30 <xMff> if the driver would allow that
Oct 12 17:33:48 <KanjiMonster> normally the switch is configured to drop any packets not allowed at this port for the vid
Oct 12 17:33:56 <xMff> ah now, the connection to the cpu would be missing then
Oct 12 17:33:59 <xMff> *no
Oct 12 17:34:21 <xMff> so you always need to refer to an existing vlan if you actually want to see the traffic on the host
Oct 12 17:34:40 <xMff> because the cpu must be linked in somehow
Oct 12 17:35:16 <KanjiMonster> if you want to let the cpu be part of the vlan yes, but nothing is stopping you to create a vlan between two ports
Oct 12 17:35:25 <xMff> yes I know
Oct 12 17:35:44 <EqUaTe> but xMffs point is quite correct
Oct 12 17:35:53 <EqUaTe> if you want the device to see the traffic, the cpu has to be on it
Oct 12 17:35:59 <KanjiMonster> ah, yes
Oct 12 17:37:20 <xMff> ok, great - I think I'll implement it this way
Oct 12 17:39:38 <xMff> one thing I miss in swconfig is an indicator whether an attribute is readonly or not
Oct 12 17:40:03 <xMff> a machine readable indicator I mean
Oct 12 17:40:22 <KanjiMonster> you should probably check that VIDs 0 or 4095 aren't used as pvids (btw, "Port PVID" is like "PIN number" ;), and probably check the max VID (if swconfig exposes that)
Oct 12 17:40:46 <xMff> yes it does, almost all info I need is there
Oct 12 17:40:53 <xMff> just have to screen scrape it a bit
Oct 12 17:41:55 <KanjiMonster> hm, here it doesn't, only the number of vlans (at least swconfig ... help doesn't list that)
Oct 12 17:42:20 <xMff> mine shows VLAN 0 - VLAN 4074 in help
Oct 12 17:42:43 <xMff> erm show
Oct 12 17:43:01 <xMff> root@OpenWrt:~# swconfig dev rtl8366s show | grep VLAN | wc -l
Oct 12 17:43:01 <xMff> 4111
Oct 12 17:43:25 <EqUaTe> vlan 0 is invalid
Oct 12 17:43:31 <xMff> yes I know
Oct 12 17:43:35 <EqUaTe> and 4094+ are almost always reserved
Oct 12 17:43:45 <EqUaTe> sometimes 4091 and up are reserved
Oct 12 17:43:58 <xMff> the loop printing the lines seems to be a bit off here as well
Oct 12 17:44:04 <EqUaTe> ah
Oct 12 17:44:37 <xMff> it looks like it starts at -1
Oct 12 17:44:59 <KanjiMonster> xMff: that would yield 16 on the ar8216, although it supports the full VID-range
Oct 12 17:45:01 <xMff> ah no
Oct 12 17:45:27 <xMff> KanjiMonster: do you have an swconfig help output of that particular chip?
Oct 12 17:45:39 <xMff> and a show output
Oct 12 17:48:24 <KanjiMonster> xMff: http://pastebin.com/R0zmYaH4 (I shortened the vlan list a bit ;)
Oct 12 17:49:41 <xMff> hm
Oct 12 17:49:52 <xMff> that one has an additional pvid attribute per vlan
Oct 12 17:50:19 <KanjiMonster> which sets the VID for the vlan
Oct 12 17:50:27 <KanjiMonster> (which defaults to the vlan index)
Oct 12 17:50:44 <xMff> my rtl8366s has no such attribute
Oct 12 17:51:03 <xMff> which would mean that vid always vlan index
Oct 12 17:51:12 <xMff> +equal
Oct 12 17:51:56 <KanjiMonster> the vlan table of the ar8216 can only take 16 entries, but you can use any VID for these entries, therefore the extra pvid attribute
Oct 12 17:52:32 <xMff> the ar8216 way is reasonable imo
Oct 12 17:53:07 <xMff> but on the rtl here I'd need to add 1000 empty switch_vlan sections to define a vlan with id 1001 ?! This can't be right
Oct 12 17:54:05 <xMff> I should try a newer kernel
Oct 12 17:54:14 <KanjiMonster> the ar8316's table can take 4096 entries, but the driver stores it in an array, which gets bigger than kmallocs max size after ~2.5k entries, limiting it to 128 entries was the easiest way of working around it
Oct 12 17:55:10 <xMff> instrad of pvid I have an "info" attribute here
Oct 12 17:55:16 <xMff> *instead
Oct 12 17:58:13 <KanjiMonster> btw, just creating an switch_section with option vlan 1001 should be enough for the realtek switches
Oct 12 17:58:20 <KanjiMonster> err.. switch_vlan section
Oct 12 17:58:28 <xMff> yes thats what I thought
Oct 12 17:58:42 <xMff> but I wonder why "vlan" isn't exposed as attribute in swconfig then
Oct 12 17:58:58 <xMff> and why its called "vlan" for rtl and "pvid" for ar
Oct 12 18:01:56 <KanjiMonster> perhaps it was designed with the roboswitches in mind, which only support 16 VLANs, and incidentially only VIDs 0-15, so there's a nice 1:1 relation
Oct 12 18:02:22 <xMff> in your /etc/config/network you use "option pvid X" as well?
Oct 12 18:02:39 <KanjiMonster> yes
Oct 12 18:03:16 <xMff> that should be aligned between the drivers imo
Oct 12 18:03:27 <KanjiMonster> (at least in theory, I don't actually use the whole vlan stuff at all, but I once confirmed it works ;)
Oct 12 18:03:43 <xMff> its already three variants right now
Oct 12 18:03:43 <KanjiMonster> yes
Oct 12 18:03:50 <xMff> roboswitch, rtl style, ar style
Oct 12 18:04:00 <xMff> I bet the ip17xx is different yet again
Oct 12 18:05:23 <xMff> heh indeed
Oct 12 18:05:30 <xMff> its called "tag" there
Oct 12 18:06:19 <KanjiMonster> roboswitch will hopefully be gone once the swconfig driver for it gets completed
Oct 12 18:08:26 <xMff> on which board is the ar8216 used?
Oct 12 18:09:32 <KanjiMonster> its on one planex I think, and some atheros targets do have it also
Oct 12 18:09:42 <xMff> ok I think I understand it now
Oct 12 18:09:53 <xMff> option vlan is actually the index in the drivers vlan table
Oct 12 18:10:02 <KanjiMonster> the integrated switch in the ar7240 is also a (slightly modified) ar8216
Oct 12 18:10:12 <xMff> and some drivers allow a differend vid from the index
Oct 12 18:10:18 <KanjiMonster> yes
Oct 12 18:10:26 <xMff> some call it tag, some pvid
Oct 12 18:11:19 <xMff> imo the option vlan should always be the vid and the driver should assign it to a suitable table index by itself, or perform some transparent mapping between table idx and vid
Oct 12 18:12:07 <xMff> maybe just fill the slots sequentially and then emit an error if the table is full
Oct 12 18:12:08 <KanjiMonster> exactly my thought
Oct 12 18:13:37 <xMff> that would give a nice uniform config and userspace would not need to worry about driver internals
Oct 12 18:13:39 <KanjiMonster> this would probably work nicely with the vlan prefix on the roboswitches (you can actually use all 12 bits for VIDs, but the first 8 are shared by all entries in the table)
Oct 12 18:14:01 <xMff> the current system requires me to write one gui backend for each driver
Oct 12 18:14:41 <xMff> I had to do the same for wifi already, I don't feel like doing the same for switches :/
Oct 12 18:16:13 <KanjiMonster> I had started to collect some thoughts for a redesigned swconfig interface, but currently have my hands full with other stuff
Oct 12 18:16:47 <xMff> well the current situation is bad for documentation, cli users and gui users
Oct 12 18:17:00 <KanjiMonster> yeah
Oct 12 18:19:29 <xMff> at the very list all drivers with extended vlan support should agree on a common name for it
Oct 12 18:19:34 <xMff> *least
Oct 12 18:20:08 <xMff> since no default network config reverences any pvids/tags/ids it should have no big impact if its changed
Oct 12 18:20:14 <xMff> references
Oct 12 18:20:22 <xMff> my spelling is bad today...
Oct 12 18:20:53 <KanjiMonster> 'vid' (without the p) would be the most obvious imo
Oct 12 18:21:02 <xMff> yes
Oct 12 18:21:07 <xMff> +1
Oct 12 18:21:18 <xMff> I'll make an RFC patch later
Oct 12 18:22:03 <KanjiMonster> great
Oct 12 18:22:44 <xMff> then swconfig should report the size of its vlan table so that userspace can do the mapping
Oct 12 18:28:22 <KanjiMonster> it already does, at least for atheros switches ("vlans: 128" in my case)
Oct 12 18:33:15 <KanjiMonster> hrm, I misremembered, the switch_dev doesn't have a max vid attribute
Oct 12 18:48:00 <russell--> someone should apply my perl patch for mips
Oct 12 18:48:40 <russell--> http:www.personaltelco.net/~russell/mips-perl-signal.patch
Oct 12 18:58:21 <KanjiMonster> russell--: create a ticket or send it to the devel-ml; makes it a bit more visible
Oct 12 19:51:56 <builder_> build #8 of at91 is complete: Failure [failed compile_4]  Build details are at http://tksite.gotdns.org:8010/builders/at91/builds/8
Oct 12 20:07:41 <xMff> KanjiMonster: http://paste.openwrt.org/d87d6e9e
Oct 12 20:10:38 <CIA-39> nico * r23417 /packages/libs/gd/ (3 files in 2 dirs):
Oct 12 20:10:39 <CIA-39> packages/gd: various fixes
Oct 12 20:10:39 <CIA-39> * fix patch to tweak configure.ac instead of configure since autoreconf is used in Build/Configure
Oct 12 20:10:39 <CIA-39> * pass --force to libtoolize to actually update libtool
Oct 12 20:10:39 <CIA-39> * add a patch to properly deal with LIBICONV
Oct 12 20:10:39 <CIA-39> * install gdlib-config in $(STAGING_DIR) and a symlink in $(STAGING_DIR_HOST)
Oct 12 20:11:34 <CIA-39> nico * r23418 /packages/libs/freetype/ (Makefile patches/): packages/freetype: update to 2.3.12
Oct 12 20:32:05 <KanjiMonster> xMff: looks okay, but why don't you make the max_vlans a global switch_attr in switch.c? Or directly use https://dev.openwrt.org/browser/trunk/target/linux/generic/files/include/linux/switch.h#L156 ?
Oct 12 20:32:27 <xMff> KanjiMonster: yeah that was just stupid, it already there
Oct 12 20:32:36 <xMff> Switch 1: rtl8366s(RTL8366S), ports: 6 (cpu @ 5), vlans: 4096
Oct 12 20:32:59 <xMff> KanjiMonster: however, http://paste.openwrt.org/d77eb3428
Oct 12 20:33:54 <KanjiMonster> xMff: that looks good.
Oct 12 20:51:54 <CIA-39> jow * r23419 /trunk/target/linux/generic/files/drivers/net/phy/ (ar8216.c ip17xx.c psb6970.c rtl8306.c):
Oct 12 20:51:54 <CIA-39> [generic] unify extended vlan id swconfig attributes.
Oct 12 20:51:54 <CIA-39> AR8216 and PSB6970 used "pvid", IP17xx used "tag" and RTL8306 called it "vid".
Oct 12 20:51:54 <CIA-39> Change all to "vid" and annotate the description with the valid ID range.
Oct 12 21:36:23 <builder_> build #8 of adm5120 is complete: Exception [exception failed slave lost shell_10 compile_11]  Build details are at http://tksite.gotdns.org:8010/builders/adm5120/builds/8
Oct 12 21:36:23 <builder_> build #8 of iop32x is complete: Exception [exception failed slave lost shell shell_10 compile_11]  Build details are at http://tksite.gotdns.org:8010/builders/iop32x/builds/8
Oct 12 23:18:09 <xMff> KanjiMonster: thanks for your help. I got the vlan stuff sorted out
Oct 12 23:21:29 <xMff> http://imagebin.ca/view/QAbjD8U.html
Oct 12 23:22:42 <KanjiMonster> xMff: looks nice :)
Oct 12 23:24:22 <KanjiMonster> xMff: I assume the cpu port is dynamically named based on the help output?
Oct 12 23:24:45 <xMff> yes
Oct 12 23:25:21 <xMff> I need to test on ip17xx tomorrow but I'm sure it will work there too
Oct 13 00:35:55 <CIA-39> lars * r23420 /trunk/target/linux/generic/config-2.6.36: [kernel] Add some missing symbols to 2.6.36 config
Oct 13 01:58:13 <CIA-39> lars * r23421 /trunk/target/linux/xburst/ (67 files in 6 dirs): [xburst] Add 2.6.36 patches
**** ENDING LOGGING AT Wed Oct 13 02:59:57 2010