**** BEGIN LOGGING AT Thu May 19 02:59:58 2011
May 19 04:32:45 <CIA-107> cshore * r26945 /trunk/target/linux/x86/alix2/target.mk:
May 19 04:32:45 <CIA-107> Simple typo for kmod- prefix in alix2 target makefile.
May 19 04:32:45 <CIA-107> Signed-off-by: Philip Prindeville <philipp@redfish-solutions.com>
May 19 05:24:03 <RealOpty> cshore, since your fixing typos, fix this one :) https://dev.openwrt.org/ticket/9425
May 19 05:29:10 <CIA-107> cshore * r26946 /packages/utils/restorefactory/Makefile: [packages] utils: restorefactory: Fixed typo in package description define. Thank RealOpty.
May 19 05:40:06 <RealOpty> cshore, ty
May 19 07:53:07 <cshore> bother, changing the ip on an interface with ifconfig results in a non-responsive interface, but changing /e/c/network and rebooting works) (but not rebooting and ifup iface isn't enough either)
May 19 07:56:46 <cshore> on ar71xx (dir 601 a1 to be exact)
May 19 07:57:11 <KanjiMonster> hey cshore
May 19 07:57:25 <cshore> oh, wait, no it's the tplink wr841nd
May 19 07:57:28 <cshore> KanjiMonster: hey
May 19 07:58:25 <KanjiMonster> cshore: I'm currently rewriting the bcm5325 driver to include an spi driver; if I have it ready, it would be great if you could test it :)
May 19 07:58:45 <cshore> ok, you have my email?
May 19 07:58:52 <KanjiMonster> yeah
May 19 07:59:42 <KanjiMonster> but first I have to make sure my rewrite is working (totally rewrote the driver)
May 19 07:59:50 <KanjiMonster> in mdio mode
May 19 08:00:50 <cshore> ok
May 19 08:03:17 <KanjiMonster> spi looks quite straight forward, but since I don't have any device to test ...
May 19 08:08:46 <cshore> be glad to help
May 19 08:09:30 <cshore> it's not like it's hurting me ;-)  (considering I believed I've mention wanting it....)
May 19 08:11:27 <cshore> ok, maybe I'm just doing something wrong with this ip change, but:
May 19 08:11:27 <cshore> I should just be able to do:
May 19 08:11:27 <cshore> ifconfig br-lan 192.168.x.x/24
May 19 08:11:27 <cshore> and have it work, right? (took the firewall down first)
May 19 08:13:22 <KanjiMonster> cshore: works for me
May 19 08:13:46 <cshore> works on brcm63xx for me too...but this ar71xx device doesn't like it
May 19 08:14:10 <KanjiMonster> ah, heh, tested it on brcm63xx, too ;)
May 19 08:15:21 <cshore> heh, that wasn't what I meant, but funny
May 19 08:15:49 <xMff> that ip thing would be a rather serious regression
May 19 08:16:31 <KanjiMonster> cshore: hm, does ifconfig eth0 down && ifconfig eth0 up help? (substitute eth0 with whatever your lan port is ;)
May 19 08:16:37 <cshore> let me update trunk (I was using an week or two ago because I was trying to sync versions with devices)
May 19 08:16:56 <KanjiMonster> ah, wait, this shouldn't have any influence on layer3
May 19 08:19:21 <cshore> ok, I wasn't sure if it was because of ifconfig vs ip (I've been using ip on my devices)
May 19 08:35:33 <cshore> hmmm....wait....wonder if it's related to mklibs
May 19 09:36:51 <CIA-107> nbd * r26947 /trunk/package/mac80211/patches/580-ath9k_tx_fifo_cleanup.patch: ath9k: fix some locking issues in the tx fifo cleanup patch
May 19 09:42:35 <cshore> xMff: ok, if the problem isn't mklibs (my next test) then there is a regression
May 19 10:04:26 <builder_> build #40 of at91 is complete: Failure [failed compile_4]  Build details are at http://buildbot.openwrt.org:8010/builders/at91/builds/40
May 19 10:07:04 <tripolar> jow_laptop: ping
May 19 10:07:06 <tripolar> hi
May 19 10:07:45 <tripolar> jow_laptop: did you follow the mailinglist about wget? here and on a lot of other systems wget doesnt depend on libidn
May 19 10:08:03 <xMff> yeah, will fix it lter
May 19 10:09:13 <tripolar> xMff: or enable idn ...
May 19 10:09:16 <tripolar> for wget
May 19 10:09:28 <tripolar> idn - Enable support for Internationalized Domain Names
May 19 10:11:03 <CIA-107> nbd * r26948 /trunk/package/broadcom-wl/patches/003-compat-2.6.35.patch:
May 19 10:11:03 <CIA-107> broadcom-wl: fix uninitialized variable
May 19 10:11:03 <CIA-107> It was causing an occasional kernel oops.
May 19 10:11:03 <CIA-107> Signed-off-by: Nathan Hintz <nlhintz@hotmail.com>
May 19 10:11:08 <CIA-107> nbd * r26949 /trunk/package/broadcom-wl/patches/006-generic-dma-api.patch: (log message trimmed)
May 19 10:11:08 <CIA-107> broadcom-wl: fix wild ssb_device accesses as pci_dev for legacy pci dma api
May 19 10:11:08 <CIA-107> broadcom-wl driver bound to ssb device with ssb driver probe
May 19 10:11:08 <CIA-107> have osh handle struct pdev pointer value initialized with
May 19 10:11:08 <CIA-107> ssb_device pointer. Later on pdev is used with legacy pci
May 19 10:11:08 <CIA-107> dma api as pci_dev thus causing oops sometimes.
May 19 10:11:09 <CIA-107> The patch replaces legacy pci dma api and pass relevant
May 19 10:23:25 <xMff> tripolar: adding idn support using the iconv stub adds c.a 5k to the wget executable
May 19 10:23:43 <xMff> tripolar: plus 60k for idn
May 19 10:23:47 <xMff> libidn
May 19 10:24:15 <xMff> not sure if it is worth it
May 19 10:24:36 <tripolar> yeah but wget is an optinal package
May 19 10:24:47 <builder_> build #39 of ubicom32 is complete: Failure [failed compile_3]  Build details are at http://buildbot.openwrt.org:8010/builders/ubicom32/builds/39
May 19 10:24:51 <tripolar> and the small version is included in busybox so .....
May 19 10:24:52 <[florian]> tripolar: still, people might want the full feature wget without it's additionnal blaot
May 19 10:25:00 <tripolar> yeah thats true
May 19 10:25:33 <[florian]> as you wish
May 19 10:27:09 <xMff> we could add a wget-idn-nossl, wget-noidn-ssl, wget-idn-ssl, ...
May 19 10:27:17 <xMff> :P
May 19 10:27:31 <[florian]> :)
May 19 10:27:42 <tripolar> yeah but i think 60k isnt that much und sites like http://www.köln.de/ should also work
May 19 10:29:00 <tripolar> but it comes to nothing als openwrt has no utf8 support so ...
May 19 10:29:05 <tripolar> als=as
May 19 10:29:31 <PsyMan> wonder if those issues are resolved yet :\
May 19 10:29:31 <xMff> I understand that
May 19 10:29:44 <xMff> but what annoys me is the usual gnu bloat
May 19 10:29:54 <xMff> libidn, a simple punicode decoder, needs 60k
May 19 10:29:58 <xMff> plzus iconv... wtf
May 19 10:30:03 <PsyMan> last time i ran a torrent owrt would recycle the router
May 19 10:30:05 <PsyMan> :\
May 19 10:30:27 <xMff> one can decode punicode to codepoints and encode those codepoints to utf-8 sequences or vice versa in maybe a 100 lines of code
May 19 10:30:32 <cshore> PsyMan: how much RAM in the router?
May 19 10:30:38 <xMff> one does not nead a 60k crap-blob for that
May 19 10:30:41 <PsyMan> not much
May 19 10:30:45 <PsyMan> only 16mb
May 19 10:31:01 <PsyMan> but no matter the settings, it would recycle
May 19 10:31:10 <tripolar> xMff: yeah but instead of updating and rewriting stuff they use stuff that is allready tested and functional and constantly updated by the lib devs
May 19 10:31:20 <PsyMan> (default fw does not, ever)
May 19 10:31:26 <tripolar> and 60k isnt that much on a normal desktop sys :)
May 19 10:31:27 <cshore> PsyMan: I don't think you can do torrents with that....unless there is some torrent that's more low mem than I know about
May 19 10:31:28 <xMff> they follow the infinite memory theorem
May 19 10:31:42 <xMff> pc = desktop = loads of ram
May 19 10:32:16 <tripolar> xMff: look at kde the versions until 4.3 or something like this depended on mysql ;)
May 19 10:32:25 <xMff> same for the crackheads in comitees designing protocols like tr-069... soap, xml, rest, ...   bullshit orgy
May 19 10:32:35 <xMff>  /rant off
May 19 10:33:01 <tripolar> xMff: they dont write code they just create the papers ... i think most of them never really wrote code ;)
May 19 10:33:22 <nbd> a committee is a creatue with six or more legs and no brain ;)
May 19 10:33:28 <nbd> s/creatue/creature/
May 19 10:34:02 <tripolar> nbd: so its a brain amputated spider ;)
May 19 10:35:52 <cshore> xMff: you could always try to convince everyone that an 10x more expensive x86 is a superior embedded device because it can do more.... (well really it depends on what you're doing, but while I think they have their place, cost accountants won't love you for that choice).
May 19 10:35:59 <xMff> yeah... ugly, hairy, stunning and everybody wants to stomp on it
May 19 10:36:00 <PsyMan> ?
May 19 10:36:25 <PsyMan> sure you can
May 19 10:36:41 <PsyMan> i had 3 clients connected torrenting on default f/w
May 19 10:37:00 <PsyMan> same torrents on just one client kills owrt
May 19 10:37:02 <xMff> PsyMan: well just lower the conntrack maximum
May 19 10:37:11 <PsyMan> but a lot of things do
May 19 10:37:12 <xMff> openwrt allows more than it can handle with 16mb
May 19 10:37:16 <PsyMan> there's the issue
May 19 10:37:24 <PsyMan> i did
May 19 10:37:34 <PsyMan> all settings/ports/limits
May 19 10:37:41 <PsyMan> there's a memory leak somewhere
May 19 10:37:41 <cshore> PsyMan: hmmm....maybe the torrent program you're using then ... there's  few choices
May 19 10:37:45 <PsyMan> or more
May 19 10:37:49 <xMff> ok
May 19 10:37:50 <xMff> bbl
May 19 10:38:03 <PsyMan> cshore: no
May 19 10:38:25 <PsyMan> the program i'm using can't freeze the router with one f/w and work with another
May 19 10:38:35 <PsyMan> with 3x the clients connected
May 19 10:38:37 <PsyMan> :p
May 19 10:39:28 <cshore> PsyMan: ah, ok, so you've installed and compiled the program in both cases?
May 19 10:39:48 <PsyMan> the program?
May 19 10:40:05 <PsyMan> the torrent client is not the issue
May 19 10:40:09 <cshore> PsyMan: torrent client
May 19 10:40:43 <PsyMan> i used the same executable in both cases
May 19 10:40:53 <PsyMan> and a different one just to make sure
May 19 10:40:55 <cshore> PsyMan: right....that's what I'm getting at
May 19 10:41:13 <cshore> asking I mean
May 19 10:41:27 <PsyMan> so yeah it's a leak somewhere
May 19 10:41:40 <tripolar> with which program can i create wordlist? with fixed length and only a subset of caracters?
May 19 10:41:48 <PsyMan> might be wlan, qos or upnp
May 19 10:42:15 <cshore> depends, how much RAM does the other firmware use without the torrent...openwrt is in the 12MB or so not including cache IIRC
May 19 10:42:22 <acoul> 2.6.39 is out
May 19 10:42:59 <PsyMan> the "other" uses 15
May 19 10:43:02 <tripolar> acoul: nice :)
May 19 10:43:09 <PsyMan> owrt uses around 12-13
May 19 10:43:23 <cshore> PsyMan: so you only have 1 MB free on the other?
May 19 10:43:28 <PsyMan> yeah
May 19 10:43:29 <cshore> before the torrent?
May 19 10:43:34 <PsyMan> yeah
May 19 10:43:50 <PsyMan> it has efficient memory handling apparently
May 19 10:43:52 <PsyMan> :p
May 19 10:44:06 <cshore> what torrent program?
May 19 10:44:21 <PsyMan> here we go again
May 19 10:44:29 <cshore> no, I'm curious
May 19 10:44:39 <cshore> I need a low mem client
May 19 10:45:12 <PsyMan> utorrent azureus transimission
May 19 10:45:13 <tripolar> acoul: do you know if this bug is fixed in the final release?
May 19 10:45:14 <tripolar> http://www.phoronix.com/scan.php?page=news_item&px=OTQ1Ng
May 19 10:45:25 <PsyMan> *transmission
May 19 10:46:01 <PsyMan> and the router won't handle it
May 19 10:46:37 <PsyMan> worse part is the issues while just browsing too
May 19 10:46:53 <PsyMan> would end up recycling under that much load too
May 19 10:46:54 <PsyMan> :\
May 19 10:50:00 <tripolar> PsyMan you search for a small torrent programm?
May 19 10:50:53 <cshore> PsyMan: I know owrt doesn't like 16MB, but transmission (for example) with only 1MB sounds impossible to me....I've looked at mem usage of transmission and I can't imagine how it'd be able to work with that little memory.  Are you sure that 15MB is the 'used' memory and not the 'not-totally-free-mem-including-cache'
May 19 10:51:03 <tripolar> PsyMan: ctorrent is the smallest i know ... start it with -C 0  .. here it runs fine on a sys with 32mb ram
May 19 10:51:49 <tripolar> PsyMan: torrent programms normally use a cache for the pieces they just download when the peace is ready its copied to the disk or when the cache is full
May 19 11:03:25 <PsyMan> you don't get it
May 19 11:03:30 <PsyMan> at all
May 19 11:04:05 <PsyMan> the problem is the router being unable to handle the connections
May 19 11:04:23 <PsyMan> (cause of a mem leak elsewhere most likely)
May 19 11:04:55 <cshore> how have you verified this?
May 19 11:04:58 <PsyMan> i'm not running the client on the router
May 19 11:05:13 <cshore> PsyMan: ah, well, that's completely different
May 19 11:05:30 <PsyMan> i'm running the torrent client(s) on the clients connected to the router
May 19 11:05:39 <PsyMan> so yeah, it is unrelated
May 19 11:06:45 <cshore> then you're probably running into the fact openwrt uses default conntrack table size that's for larger routers, and you need to reduce conntrack table size and or reduce timeouts
May 19 11:07:11 <cshore> in /etc/sysctl.conf
May 19 11:08:10 <PsyMan> i told ya i messed with all those settings and that it sometimes recycles while just browsing
May 19 11:08:13 <cshore> unfortunately atm there is no easy way to specify per-model sysctl settings
May 19 11:08:18 <PsyMan> is it so hard to ger? :p
May 19 11:08:23 <PsyMan> *get
May 19 11:09:29 <cshore> what do you have besides stock on the router?
May 19 11:10:09 <PsyMan> ipv6 tables and dhcp, upnp daemon, qos
May 19 11:10:19 <cshore> qos is known to have issues
May 19 11:10:25 <cshore> it's being worked on
May 19 11:11:02 <PsyMan> oh and wireless driver, adsl driver, ethernet driver and basic wireless security
May 19 11:11:06 <PsyMan> those will be all
May 19 11:11:18 <cshore> qos is probably the problem
May 19 11:11:25 <PsyMan> hmm :|
May 19 11:11:34 <cshore> or rather the layer7 stuff
May 19 11:11:42 <PsyMan> layer7?
May 19 11:12:02 <cshore> qos for protocol types instead of ports
May 19 11:13:28 <PsyMan> ahh
May 19 11:13:39 <PsyMan> is that integrated and enabled?
May 19 11:13:48 <cshore> unfortunately yes
May 19 11:13:54 <PsyMan> ouch
May 19 11:14:02 <cshore> though recent trunk disables it I think
May 19 11:14:15 <cshore> I have to ask nbd
May 19 11:14:42 <nbd> layer7 leaks memory like mad
May 19 11:14:46 <nbd> so i removed it from the default qos config
May 19 11:14:52 <PsyMan> when?
May 19 11:14:53 <nbd> (but it's still available)
May 19 11:14:59 <nbd> a while ago
May 19 11:15:01 <nbd> not sure
May 19 11:15:20 <PsyMan> define "a while" pls kk? :>
May 19 11:17:06 <cshore> PsyMan 7 weeks: https://dev.openwrt.org/changeset/26365
May 19 11:17:13 <PsyMan> https://dev.openwrt.org/changeset/26365
May 19 11:17:16 <PsyMan> ahh
May 19 11:17:18 <PsyMan> you go me
May 19 11:17:21 <PsyMan> got
May 19 11:17:50 <PsyMan> ehm, were those still used if qos was disabled?
May 19 11:19:31 <cshore> I'm not really sure how all that works
May 19 11:19:43 <PsyMan> me neither xD
May 19 11:19:50 * PsyMan looks at nbd
May 19 11:23:19 <nbd> ?
May 19 11:23:27 <PsyMan> https://dev.openwrt.org/changeset/25756 <-- maybe this should be updated? (had some issues)
May 19 11:24:00 <PsyMan> nbd: if qos was disabled, whould the leak still occur?
May 19 11:24:13 <PsyMan> the one related to the l7 stuff
May 19 11:24:31 <nbd> no
May 19 11:24:46 <acoul> @tripolar: no clue
May 19 11:24:47 <PsyMan> hmm
May 19 11:24:52 <PsyMan> i see
May 19 11:26:37 <PsyMan> maybe it was the wireless driver too :\
May 19 11:26:40 <PsyMan> oh well
May 19 12:13:59 <cshore> ping ar71xx devs; regression with vlans
May 19 12:38:34 <gzanan> Does anyone build a 2.6.39 images for ar71xx,especially tp-link device?
May 19 14:22:56 <jmccrohan> cshore, KanjiMonster: Trying to port to new brcm63xx device but failing. Anything obvious I'm missing? https://gist.github.com/980837
May 19 14:24:34 <cshore> er, what exactly isn't working?
May 19 14:26:26 <cshore> awk: /proc/cpuinfo: No such file or directory    -   I was pretty sure that solved ages ago....are you doing a clean build?
May 19 14:26:55 <jmccrohan> cshore: latest svn
May 19 14:27:28 <jmccrohan> the roboswitch errors are the main problem i think.
May 19 14:30:18 <cshore> oh, nvm.....that's something that I forgot about (the cpuinfo stuff)
May 19 14:30:38 <cshore> well what's not working?  the switch, or only vlans?
May 19 14:32:25 <cshore> that is can you ping other hosts, or do you have no network at all (without vlans I mean)
May 19 14:32:45 <jmccrohan> no network at all
May 19 14:33:00 <jmccrohan> booting it now, hang on i'll show you output
May 19 14:35:14 <cshore> what SoC and switch?
May 19 14:36:27 <jmccrohan> BCM5325
May 19 14:37:10 <cshore> ok, and what did you do for board definition (it looks like it's looking for the switch on eth0, normally with these boards it's eth1)
May 19 14:38:00 <jmccrohan> yeah. i'm just realising that after looking at /etc/config/network
May 19 14:39:01 <cshore> does it have a separate eth0 and eth1 ? (one which is the switch, the other being the lan?)
May 19 14:39:12 <cshore> sorry wan
May 19 14:41:16 <jmccrohan> na. if config shows only lo and eth0
May 19 14:41:26 <jmccrohan> wan would normally be adsl
May 19 14:41:29 <cshore> no, I mean on the device
May 19 14:41:41 <cshore> (physically)
May 19 14:41:55 <cshore> I have boards that have a wan and switch with 6348
May 19 14:42:15 <cshore> that's why I ask (non-adsl, regular ethernet I mean)
May 19 14:43:52 <cshore> can you paste your board definition
May 19 14:43:54 <cshore> ?
May 19 14:45:38 <jmccrohan> oh right. yea, it has 4 ether connected to the BCM5325, and then ADSL. Board info should be in the ticket I opened a while back. https://dev.openwrt.org/ticket/8810
May 19 14:46:38 <mazilo> Under packages/lang, I don't see a java package. Don't we support java?
May 19 14:47:07 <cshore> mazil: nope
May 19 14:47:31 <mazilo> cshore: I was hoping we have a support for java.
May 19 14:47:50 <cshore> mazilo: not at present....no one has worked on it
May 19 14:48:00 <mazilo> cshore: OK.
May 19 14:48:25 <cshore> jmccrohan: did you do .has_phy, or .force_speed, type enet0 ?
May 19 14:49:01 <cshore> jmccrohan: or did you define both enet0 and enet1 even though there is only one?
May 19 14:51:45 <jmccrohan> cshore: I essentially copypasta'd the RTA1025W_16 definition and editied to match the board name, and to define the UART.
May 19 14:54:16 <cshore> ah, not the best choice.  You should look at CT536_CT5261 (for example instead) because it properly define a switch only (no wan ethernet).
May 19 14:56:06 <jmccrohan> ok so. I thought they were the same because of their board definitions in the GPL'd source.
May 19 14:56:21 <jmccrohan> https://dev.openwrt.org/attachment/ticket/8810/boardparms.c#L1130 vs https://dev.openwrt.org/attachment/ticket/8810/boardparms.c#L1286
May 19 14:56:52 <cshore> the RTA1025W_16 is (AFAIK) an untested...guessed definition
May 19 14:58:02 <jmccrohan> ahhh.. makes more sense now. i'll have a tinker around with this and get back to you
May 19 14:59:11 <jmccrohan> also, is there a handy way for creating diffs from board defs? or just a case of editing, and then redownloading a second copy of file and manually do the diff?
May 19 15:00:06 <cshore> you mean in the kernel tree under build_dir?
May 19 15:00:12 <jmccrohan> yeah
May 19 15:05:27 <cshore> build_dir/linux.../linux       (KERNELROOT)   has a patches dir for quilt
May 19 15:05:27 <cshore> so if you, before editing
May 19 15:05:27 <cshore> quilt series
May 19 15:05:27 <cshore> quilt push <name_of_patch_that_is_to_be_below_the_patch>
May 19 15:05:27 <cshore> quilt new platform/xxx_some_patch_name
May 19 15:05:27 <cshore> qulit add path/to/bcm63xx_board.c
May 19 15:05:28 <cshore> edit
May 19 15:05:29 <cshore> quilt refresh
May 19 15:05:59 <cshore> then patches/platform/xx_some_patch_name will have your diff
May 19 15:07:37 <cshore> if you copy that to target/linux/brcm63xx/patches-2.6.whatever, then when you make target/linux/prepare (or make/target/linux/{clean,compile} which implicitly does prepare, the build_dir will have the patch on a fresh tree
May 19 15:08:27 <cshore> do submit to openwrt you could then svn add / svn diff
May 19 15:18:45 <gzanan> Is DIR-665 kirdwood based supported?
May 19 15:20:07 <_lore_> ping nbd
May 19 15:25:29 <jmccrohan> chshore: thanks. :)
May 19 15:27:20 <nbd> _lore_: pong
May 19 15:27:30 <_lore_> hi
May 19 15:27:33 <nbd> hi
May 19 15:27:34 <_lore_> how are you?
May 19 15:28:01 <nbd> fine, how are you?
May 19 15:28:14 <_lore_> fine tnx
May 19 15:28:28 <_trine> I'm ok too if it matters :P
May 19 15:28:42 <_trine> is everyone else ok
May 19 15:28:57 <_lore_> I carried out some more tests on the reset of ath9k
May 19 15:28:57 <nbd> of course it matters, i don't want you to feel left out or anything
May 19 15:28:58 <nbd> ;)
May 19 15:29:02 <_trine> please forgive my poor sense of humour :)
May 19 15:29:44 <_lore_> I can reproduce the issue turning off a device...
May 19 15:30:22 <_lore_> but more in general the issue is related to frame failures using AMPDU
May 19 15:31:38 <nbd> interesting
May 19 15:31:50 <_trine> would madwifi work on my nanostation if I compiled it in ?
May 19 15:32:03 <nbd> _lore_: which chipset on the AP side?
May 19 15:32:06 <_trine> and removed ath9k
May 19 15:32:14 <nbd> _trine: no
May 19 15:32:28 <_lore_> disabling AMPDU the issue disappears...
May 19 15:32:45 <_lore_> I am injecting frame in monitor mode
May 19 15:32:54 <_lore_> AR9280
May 19 15:33:09 <nbd> take a look at xmit.c
May 19 15:33:11 <nbd> look for needreset = 1
May 19 15:33:24 <_lore_> yes I did
May 19 15:33:24 <nbd> check the comment above
May 19 15:33:59 <_trine> I have some weird issues with ath9k on my nanostation when using airodump
May 19 15:34:08 <_lore_> in the ath_tx_comple_aggr..
May 19 15:34:53 <_lore_> I have no reset in AP mode, just in monitor
May 19 15:35:20 <_lore_> I am not able to reproduce the issue in AP mode
May 19 15:35:24 <nbd> ok
May 19 15:35:26 <_lore_> till now
May 19 15:35:37 <nbd> try forcing the driver to set ah->opmode to AP mode
May 19 15:35:40 <nbd> when in monitor mode
May 19 15:35:48 <_trine> I see multiple instances of the same AP with multiple different clients attached to it and I never see this using my laptop
May 19 15:36:25 <nbd> _trine: doesn't sound like an ath9k bug to me
May 19 15:36:36 <_trine> I dont know what causes it
May 19 15:36:45 <nbd> _trine: when in monitor mode, ath9k and mac80211 will simply deliver frames that they see to user space
May 19 15:36:59 <nbd> the interpretation of how many APs and clients there are, and which ones, is left up to the user space program
May 19 15:37:08 <nbd> so any error in that interpretation can only be in the user space application
May 19 15:37:13 <nbd> => airodump issue, driver independent
May 19 15:38:09 <_trine> I only see it on the nanostation my only thought would be that the  nanostation hears more with its better antenna but there are such a lot of these different clients it makes me doubt the authenticity of them
May 19 15:38:23 <_lore_> the system is really stucked..and I need the reset in ath_tx_complete_poll_work to restart the transmission
May 19 15:38:59 <_lore_> could it be due to the AMPDU burst size?
May 19 15:39:30 <nbd> how are you controlling aggregation?
May 19 15:41:29 <_lore_> I use the sta info of the neighbour
May 19 15:42:33 <_lore_> I enstablish the AGGR connections in mac80211
May 19 15:42:41 <nbd> ok
May 19 15:43:03 <nbd> so when you do this, do you only have an active monitor mode interface, or any other interface types as well?
May 19 15:43:41 <_lore_> I use just one VAP in monitor mode
May 19 15:44:36 <_lore_> hower I set these two parameters to fixed valus: ampdu_factor = 0x3 and ampdu_density=0x6
May 19 15:45:11 <_lore_> I do not know if they are corrects or not
May 19 15:45:22 <nbd> maybe there are some hardware settings that are different for pure monitor mode compared to AP mode
May 19 15:45:28 <nbd> that affect behaviour related to A-MPDU
May 19 15:45:43 <nbd> hence my suggestion of making the driver use NL80211_IFTYPE_AP for ah->opmode
May 19 15:45:58 <nbd> which should be fine for monitor mode as well
May 19 15:46:05 <_lore_> which are these values?
May 19 15:46:16 <nbd> dunno, they're spread over the driver
May 19 15:46:21 <nbd> but tied to ah->opmode
May 19 15:46:26 <nbd> so just change ah->opmode
May 19 15:46:33 <nbd> since the AP iftype will allow monitoring as well
May 19 15:46:37 <nbd> as long as the monitor options are set
May 19 15:46:46 <nbd> just go to the function that calculates the opmode
May 19 15:47:25 <_lore_> ok...tnx...I will have a look to this
May 19 15:47:33 <nbd> right now it defaults to NL80211_IFTYPE_STATION
May 19 15:47:46 <nbd> which may not be a good idea for something meant to do point-to-multipoint communication
May 19 15:48:30 <_lore_> yes
May 19 15:48:46 <_lore_> I have to use monit interface
May 19 15:48:52 <_lore_> *monitor
May 19 15:49:05 <nbd> monitor interfaces are not registered with the driver
May 19 15:49:12 <nbd> they're handled only by the stack internally
May 19 15:49:53 <nbd> so for the case of no registered virtual interfaces it should default to AP
May 19 15:50:02 <nbd> and only if there is an interface leave the default at station
May 19 15:50:21 <nbd> you could probably send that change upstream as well
May 19 15:51:48 <_lore_> I think this issue is related to the frame failures, in fact I observed this behaviour soon  after a packet loss or when I turn off  the other station
May 19 15:52:44 <nbd> yes, but internally the handling of such failures might be different between ap-mode configurations and sta-mode configurations
May 19 15:52:57 <nbd> since you mentioned that this does not happen in normal AP mode
May 19 15:53:13 <_lore_> yes, exactly
May 19 15:53:25 <nbd> so just change the mode and see what happens
May 19 15:53:38 <_lore_> yes
May 19 15:54:10 <_lore_> ok..tnx
May 19 15:54:18 <_lore_> I am going to try
May 19 16:24:16 <_lore_> I tried to set ah->opmode = NL80211_IFTYPE_AP but same issue
May 19 16:34:43 <CIA-107> mazilo * r26950 /packages/net/freeswitch/Makefile: fix MD5SUM on freeswitch-sounds-en-us-callie-8000-1.0.15.tar.gz package
May 19 17:52:50 <philipp64|laptop> nbd and KanjiMonster: can we get the mtd patches committed?  Thanks.
May 19 18:09:14 <netprince> I've been away for a while, is there any info on the resolver problem in trunk?
May 19 18:10:16 <netprince> ping and wget both segfault when doing the name lookup unless I specify -4
May 19 18:15:15 <jmccrohan> cshore: Working perfectly now with a proper device definition. Still needed to manually edit /etc/config/network from console though. Will swapping .has_enet0 and .has_enet1 force the switch to be created on eth1 by default?
May 19 18:54:16 <KanjiMonster> jmccrohan: does this device really use both eths? else you only need to use .has_enet1
May 19 18:57:24 <jmccrohan> KanjiMonster: it doesn't. i was talking about swapping the 0 and 1 definitions.
May 19 18:58:11 <KanjiMonster> jmccrohan: then I would recommend that you just use .has_enet1 = 1, so that there's only eth0
May 19 18:58:39 <KanjiMonster> no need to create non working eth's ;)
May 19 18:59:42 <KanjiMonster> you have to create a default /etc/config/network then though, since the default currently is two eths; this should be eventually swapped (so default is only one eth)
May 19 19:00:38 <jmccrohan> Yeah, that was going to be my next question. where do I create the modified /etc/config/network?
May 19 19:02:12 <KanjiMonster> jmccrohan: target/linux/brcm63xx/base-files/etc/defconfig/<boardname>
May 19 19:03:20 <jmccrohan> KanjiMonster: ta.
May 19 19:04:51 <KanjiMonster> jmccrohan: also don't forget to add it to target/linux/brcm63xx/base-files/lib/brcm63xx.sh ;)
May 19 19:10:18 <jmccrohan> KanjiMonster: thanks. anywhere else I need to patch?
May 19 19:11:36 <KanjiMonster> jmccrohan: I think that's all (assuming you already edited the image/Makefile to build an image for it)
May 19 19:25:07 <jmccrohan> KanjiMonster: those two files were exactly what where needed. basic networking is now working on flash, and LEDs are no longer wonky.
May 19 19:25:21 <jmccrohan> time to assemble the patches
May 19 19:25:39 <KanjiMonster> great
May 19 19:26:17 <jmccrohan> KanjiMonster, cshore: thanks for the help.
**** ENDING LOGGING AT Fri May 20 03:00:00 2011