**** BEGIN LOGGING AT Sat Jan 31 02:59:57 2009 Jan 31 03:07:18 2nd time... is it my imagination... or was Languages->Python->python-pyserial removed? Jan 31 03:08:51 never mind Jan 31 03:23:26 nbd * r14295 /packages/net/chillispot/patches/120-fix_bufferoverflow.patch: chillispot: fix buffer overflow (thx to Sebastian Gottschall) Jan 31 03:46:03 thepeople * r14296 /packages/net/n2n/Makefile: update n2n Jan 31 05:05:47 libxt_layer7.c:25:39: warning: linux/netfilter/xt_layer7.h: No such file or directory Jan 31 05:05:47 libxt_layer7.c:48: warning: 'struct xt_layer7_info' declared inside parameter list Jan 31 05:06:16 that's from iptables... Jan 31 05:21:24 trunk/package/iptables/patches/1.4.1.1/002-layer7_2.17.patch doesn't seem to be applied? Jan 31 05:21:48 does that mean my BR is screwed? Jan 31 05:25:44 agb * r14297 /trunk/target/linux/etrax/ (config-default patches/400-scsi_import.patch): [kernel] etrax: add scsi symbols to kernel config, thanks claudyus. update symbols after enabling scsi Jan 31 06:30:53 ok i fixed it... Jan 31 07:18:21 nbd * r14298 /trunk/package/lua/ (Makefile patches/500-eglibc_config.patch): fix lua compile with eglibc selected and utmp support disabled Jan 31 11:14:31 xMff: howdy Jan 31 12:30:28 mirko * r14299 /packages/utils/oprofile/Makefile: adding missing dependency Jan 31 12:55:19 florian * r14300 /packages/utils/sysstat/Makefile: Jan 31 12:55:19 Creating directory /var/log/sysstat during install is wrong, as it will corrupt the symbolic link /var -> /tmp Jan 31 12:55:19 The directory is already created dynamically from the init script. Jan 31 12:55:19 Signed-off-by: Thomas Langer Jan 31 12:55:57 florian * r14301 /trunk/package/e2fsprogs/ (Makefile files/ files/e2fsck.init): Jan 31 12:55:57 Add e2fsck init script - scans every ext2/ext3 mount from fstab Jan 31 12:55:57 Signed-off-by: Vasilis Tsiligiannis Jan 31 13:03:28 * jb42 ponders Jan 31 13:39:06 florian * r14302 /packages/sound/mpd/Makefile: Jan 31 13:39:06 Update mpd to 0.14. Jan 31 13:39:06 Signed-off-by: Michael Vogt Jan 31 13:51:21 florian * r14303 /packages/net/cups/files/ (cupsd.init etc/cups/cupsd.conf): (log message trimmed) Jan 31 13:51:21 By default OpenWRT's /var is a symlink to /tmp. Jan 31 13:51:21 This is unusual under GNU/Linux but justified by the unusual Jan 31 13:51:21 circumstances of limited flash space and lifetime. But there is Jan 31 13:51:21 no reason for OpenWRT packages to presume that /tmp and /var Jan 31 13:51:25 are always interchangeable. Jan 31 13:51:27 The patch below fixes the cups package so that it behaves better Jan 31 15:37:55 zandbelt * r14304 /packages/net/asterisk-1.6.x/ (7 files in 3 dirs): add basic asterisk 1.6.x package based on 1.6.1-rc1 Jan 31 16:01:22 florian * r14305 /trunk/target/linux/brcm63xx/ (2 files in 2 dirs): [brcm63xx] add support for user9102 boards (#4537) Jan 31 16:08:32 agb * r14306 /trunk/package/broadcom-wl/ (Makefile files/lib/wifi/broadcom.sh): [package] broadcom-wl: move txpower back to wifi-device section, but remain backwards compatible Jan 31 16:09:16 agb * r14307 /trunk/package/madwifi/ (Makefile files/lib/wifi/madwifi.sh): [package] madwifi: mode txpower to wifi-device section, but remain backwards compatible Jan 31 16:09:31 nico * r14308 /trunk/package/kernel/modules/netfilter.mk: [package] kernel: add modules for nfnetlink based packet logging & queuing and conntracking Jan 31 16:09:40 agb * r14309 /trunk/package/mac80211/ (Makefile files/lib/wifi/mac80211.sh): [package] mac80211: move txpower to wifi-device section, but remain backwards compatible Jan 31 16:12:57 florian * r14310 /trunk/package/kernel/modules/netdevices.mk: [package] add libphy module to tg3 driver package - fixes unresolved symbols (#4353) Jan 31 16:15:31 agb * r14311 /trunk/docs/wireless.tex: [docs] wireless.tex: update docs to reflect txpower move Jan 31 16:18:02 agb * r14312 /branches/8.09/ (7 files in 7 dirs): Jan 31 16:18:02 merge r14306, r14307, r14309, and r14311 to 8.09: Jan 31 16:18:02 [package] broadcom,madwifi,mac80211: move txpower from wifi-iface to wifi-device but retain backward compatiblity Jan 31 16:19:01 nico * r14313 /packages/multimedia/ffmpeg/Makefile: [packages] ffmpeg: fix typo Jan 31 16:26:02 nico * r14314 /packages/net/mini_snmpd/ (. Makefile patches/ patches/101-opt_flags.patch): [packages] add mini_snmpd, an SNMP server for embedded devices Jan 31 16:29:41 nico * r14315 /packages/libs/libnfnetlink/Makefile: [packages] libnfnetlink: update to v0.40, install dev pkgconfig file Jan 31 16:31:44 nico * r14316 /packages/libs/libnetfilter-queue/ (Makefile patches/): [packages] libnetfilter-queue: update to v0.0.16, add dependency on kmod-nfnetlink-queue Jan 31 16:32:36 nico * r14317 /packages/libs/libnetfilter-log/ (. Makefile): [packages] add libnetfilter-log Jan 31 16:34:16 nico * r14318 /packages/libs/libnetfilter-conntrack/ (. Makefile): [packages] add libnetfilter-conntrack Jan 31 16:35:31 nico * r14319 /packages/net/conntrack-tools/ (. Makefile): [packages] add conntrack-tools Jan 31 16:38:01 nico * r14320 /packages/net/xtables-addons/ (. Makefile): [packages] add xtables-addons, providing netfilter/xtables modules & extensions that aren't merged in mainline Jan 31 16:39:43 nbd * r14321 /branches/8.09/target/linux/ppc44x/Makefile: ppc44x is broken after all Jan 31 16:43:25 nbd * r14322 /packages/utils/pcmciautils/Makefile: fix pcmciautils build Jan 31 16:51:51 florian * r14323 /trunk/target/linux/x86/ (5 files): [x86] resync kernel configurations, patch from Hauke Mehrtens (#4396) Jan 31 16:52:55 florian * r14324 /trunk/target/linux/x86/Makefile: [x86] bump kernel version to 2.6.28.2 (#4396) Jan 31 17:30:05 nbd * r14325 /branches/8.09/package/mac80211/files/lib/wifi/mac80211.sh: fix version control conflict Jan 31 17:49:07 nbd * r14326 /packages/libs/uclibc++/patches/009-compile_fixes.patch: remove a part of r14231 which breaks packages such as sipp Jan 31 17:57:33 florian * r14327 /packages/lang/ruby/patches/500-openssl_missing.patch: [package] fix ruby compilation failure (#4535) Jan 31 18:08:42 jow * r14328 /branches/8.09/feeds.conf.default: branches/8.09: update to LuCI 0.8.5 Jan 31 18:33:53 juhosg * r14329 /trunk/package/mac80211/patches/210-remove_unused_stuff.patch: [package] mac80211: refresh patches Jan 31 18:36:11 juhosg * r14330 /trunk/package/mac80211/patches/401-ath9k-dont-register-leds-on-ar9100.patch: [package] mac80211/ath9k: disable leds on the ar9100 devices Jan 31 18:46:33 juhosg * r14331 /trunk/package/mac80211/patches/401-ath9k-dont-register-leds-on-ar9100.patch: [package] mac80211/ath9k: replace the accidentaly committed patch Jan 31 18:47:14 nbd * r14332 /branches/packages_8.09/: branch packages for 8.09 Jan 31 18:56:30 was my msmtp patch no good? Jan 31 18:56:49 i guess nobody got around to integrating it yet Jan 31 18:57:24 For now over 3 weeks, the wiki refuses to allow one page to be edited: Permission denied: '/var/www/openwrt.org/wiki/data/pages/AtherosPort/current-locked' I can revise other pages, but not this one. Jan 31 18:58:16 nbd: that's okay okay okay Jan 31 18:58:29 nbd * r14333 /branches/packages_8.09/net/xtables-addons/: xtables-addons does not work on 8.09 Jan 31 19:02:11 I updated the WIki hardware listings to reflect the Athros AP81 devices from Trendnet, Planex, Netgear, D-Link Jan 31 19:11:35 juhosg: So without the LED is the ath9k driver faster? Jan 31 19:12:15 RoundSparrow: i guess it's mostly because on ar9100, leds are handled through platform code Jan 31 19:13:26 there is discussion on forum about madwifi and turning off LED making performance difference. I have no idea cpu cost of doing gpio operation Jan 31 19:17:59 I am also taking another fresh search again for the madwifi source for the ar9100 chips. I see again on WNR2000 the dmesg says: "ath_ahb: 0.9.4.5 (Atheros/multi-bss) initialize tx/rx engine, txbuf: 512 rxbuf: 128 wifi0: Atheros AR9100 WiSoC: mem=0xb80c0000, irq=2 " - the ath9k driver output is so terse it doesn't mention anything about txbuf rxbuf - wonder if the AHB bus versions differ from PCI on this. Jan 31 19:19:45 I am frustrated with simple things I don't know about embedded and OpenWRT. WHy when I try to compile ath9k with debug does the compile fail - not not integrated into tree correctly? Also with busybox, how do I do equal to "modprobe -v ath9k debug=0x00002000 " Jan 31 19:21:10 insmod ath9k debug=0x00002000 Jan 31 19:22:27 RoundSparrow: on the ar913x one LED status change costs ~380000 ns approximately Jan 31 19:22:56 juhosg: is it done per packet? Jan 31 19:23:12 yup Jan 31 19:23:21 juhosg * r14334 /trunk/package/mac80211/patches/402-ath9k-enable-debug.patch: [package] mac80211/ath9k: enable debug stuff Jan 31 19:23:24 .38ms is HUGE! Jan 31 19:23:36 maybe not ? Jan 31 19:23:57 I mean to me this add .38ms to a ping time ;) Jan 31 19:24:21 for packet latency it'd be irrelevant Jan 31 19:24:31 and both the LAN and WAN led flash on same packet, so really .76ms Jan 31 19:24:32 but you have to consider that the cpu is busy during those .38ms Jan 31 19:25:11 I mean both WiFi and LAN LED Jan 31 19:26:54 nbd: but our problem is CPU usage, so perhaps it is a concern Jan 31 19:27:45 not to say is the big issue, as the binary-only madwifi driver that Atheros has for 2.6.15 has no such CPU issue. but would be nice to have options while ath9k performance improves Jan 31 19:28:06 juhosg: does CONFIG_ATH9K_DEBUG=y compile for you in trunk? Jan 31 19:28:30 yes Jan 31 19:28:36 RoundSparrow: i don't think the led stuff is even relevant wrt. cpu usage yet Jan 31 19:28:50 compared to some of the other parts where cpu time is wasted Jan 31 19:29:08 for instance the hostapd change should help a lot Jan 31 19:29:22 juhosg: do I put in trunk/.config or "make CONFIG_ATH9K_DEBUG=y" ? Jan 31 19:30:35 juhosg: do you see the issue I mention about TEW-652BRP and now two of them fault on first wifi use ("wifi up") and funny mac address? I have two very close serial number 652BRP and one does this with identical setup and the other does not. Jan 31 19:31:02 nbd: i saw at best 15% improvement on hostapd change Jan 31 19:31:20 my wireless transfers to a local LAN server max out at 750KB/s with CPU at 100% Jan 31 19:33:44 interesting Jan 31 19:33:57 that's much worse than what i expected Jan 31 19:34:31 juhosg / nbd: I also wish to emphasize thank you for your hard work, I think things are progressing well and don't wish to sound like a pest ;) I cheer at your hard work. Jan 31 19:34:37 nbd, where are you located in this world? Jan 31 19:34:40 germany Jan 31 19:38:22 I can perhaps send you a Trendnet AP81 if you can not source them cheap locally. Jan 31 19:39:18 nbd: i guess the ar5416 gpio code simply breaks something on the ar913x: http://rafb.net/p/uDRzwG99.html Jan 31 19:40:22 RoundSparrow: i did not see this problem yet Jan 31 19:42:37 juhosg: so you think this gpio issue also impact the LAN LED performance too? I haven't really tested LAN to WAN usage of CPU Jan 31 19:44:24 RoundSparrow: i don't know it yet Jan 31 19:45:31 RoundSparrow: the ath9k driver gets the mac address from the eeprom data currently, and the TEW-6x2 boards store it on different location Jan 31 19:46:16 juhosg: sorry without symbols: http://openwrt.pastebin.com/f65c49c7c Jan 31 19:48:53 that is from build of svn yesterday - 12484 http://sites.google.com/site/stephengutknecht/Home/openwrt_14284_k28p2-ar71xx-tew-632brp-squashfs.bin?attredirects=0 - and identical config and binary on one router do this, the other does not. Jan 31 19:49:38 RoundSparrow: weird Jan 31 19:49:38 14284, can't type today Jan 31 19:50:09 and now this is the second one. my first trendnet did the same 3 weeks ago on first build, but I bricked it by building and flashing 4.5mb symbol build ;) Jan 31 19:51:35 juhosg: on this problem router, d-link firmware his wireless shows mac 00 : 14 : d1 : 5d : 27 : fa , but openwrt shows: wlan0 Link encap:Ethernet HWaddr 21:11:11:11:11:11 Jan 31 19:54:42 RoundSparrow: the D-Link/Trendnet firmware reads the mac address from the 'config' partition Jan 31 19:55:33 ok. however, is the 11:11:11 unusual? Jan 31 19:56:05 on my working trendnet, it is much more normal. I wonder if the eprom on some out of the factory has junk data. Jan 31 20:02:09 juhosg: where is this output from? oprofile? Jan 31 20:03:42 RoundSparrow: the 21:11:11:11:11:11 is a multicast address Jan 31 20:03:55 nbd: nope, i have added some debug stuff into the code :p Jan 31 20:04:06 ah Jan 31 20:05:01 what is best way to get symbols for a kernel oops? can't fit a symbols build on my device Jan 31 20:05:14 compile the kernel with KALLSYMS Jan 31 20:05:20 in trunk there is now a top level menuconfig option for that Jan 31 20:05:22 doesn't fit Jan 31 20:05:26 juhosg: can you share a patch for your debug stuff? Jan 31 20:05:29 i don't get the .bin because would be too big Jan 31 20:05:36 kernel oops in a module or in the core kernel? Jan 31 20:05:55 you can see that from the addresses most of the time Jan 31 20:06:14 checking, holdon Jan 31 20:06:58 same as this, though this isn't from my system: http://forum.openwrt.org/viewtopic.php?pid=78928#p78928 Jan 31 20:07:07 RoundSparrow: 20:23:20 < CIA-31> juhosg * r14334 /trunk/package/mac80211/patches/402-ath9k-enable-debug.patch: [package] mac80211/ath9k: enable debug stuff Jan 31 20:07:41 cool Jan 31 20:09:11 thelsdj is the guy with the trendnet, I pasted his crash here: http://openwrt.pastebin.com/f65c49c7c - he is much more experienced Linux hacker than I am Jan 31 20:12:17 yep thats it Jan 31 20:13:21 happens on brctl addif but guessing the problem is with detection of my ath9k device Jan 31 20:13:48 ifconfig wlan0 doesn't show proper mac address is first sign Jan 31 20:15:34 thelsdj: we discussed that, on the factory firmware they read from config partition, OpenWRT reads it from eprom. my current theory is your eprom from factory has bad values - and juhosg said that address you have is multicast - that would probably be the issue Jan 31 20:16:14 thelsdj: if we can figure out where the fault is happening and where the mac is set, we could try substituting a valid one (such as the one you know from the other firmware) and ignore the one read from eprom - and see if that resolves issue Jan 31 20:18:55 so is that ath9k debug patch enabled by default as long as i've got latest svn? Jan 31 20:19:36 anything else i should do for a new build to help debugging? Jan 31 20:20:19 sorry guys, i have to go, but i will check this mac address issue tomorrow Jan 31 20:22:35 ok, thanks juhosg Jan 31 20:24:20 np Jan 31 20:26:44 thelsdj: i say review the patch, I was getting build failure when I did "CONFIG_ATH9K_DEBUG=y" before. You can also set that in trunk/.config Jan 31 20:27:52 the patch sets it in config.mk of mac80211 package i believe Jan 31 20:28:04 thelsdj: you may also have to get the right magic of make world to trigger actual recompile Jan 31 20:29:30 ya i'm doing a fresh build anyways Jan 31 20:33:48 so earlier today juhosg checked in then reverted a patch on disable the led. there was talk on the forum about the older madwifi having poor performance on some routers with LED on. He said: "on the ar913x one LED status change costs ~380000 ns approximately" (.38ms); he also said: "i guess the ar5416 gpio code simply breaks something on the ar913x: http://rafb.net/p/uDRzwG99.html" Jan 31 20:34:03 I'm going to try with and without his LED patch to see if it improves performance Jan 31 20:52:52 very cool. rmmod ath9k; insmod ath9k debug=0x00002000 Jan 31 20:52:57 worked for me Jan 31 21:01:54 thelsdj: I posted what output looks normal on ath9k with debug: http://forum.openwrt.org/viewtopic.php?pid=80825#p80825 Jan 31 21:14:52 blogic: ping Jan 31 21:14:58 yes Jan 31 21:15:00 that is me Jan 31 21:15:15 aroscha: 1 sec, after that mail i am getting lots of /pm Jan 31 21:15:16 :) Jan 31 21:16:39 aroscha: yes Jan 31 21:16:41 i am here now Jan 31 21:16:46 ok... Jan 31 21:16:48 what mail? Jan 31 21:17:00 ah Jan 31 21:17:03 hehe Jan 31 21:17:14 i just announce www.l2gvpn.org on several lists Jan 31 21:17:29 oh! cool. Jan 31 21:17:34 yes, i know Jan 31 21:17:36 should I test it on FONs? Jan 31 21:18:11 haha Jan 31 21:18:13 you can Jan 31 21:18:21 my setup has about 6 fons :) Jan 31 21:18:32 ok, hmm... Jan 31 21:18:47 its really nice Jan 31 21:19:00 i actually just replace my openvpn setup :) Jan 31 21:19:46 nice Jan 31 21:19:58 you should definitely talk more about authentication etc with henning Jan 31 21:20:11 his masters thesis was security in MANETs Jan 31 21:20:20 ok Jan 31 21:20:22 i guess the same principles apply to l2gvpn Jan 31 21:20:23 good Jan 31 21:20:28 yes Jan 31 21:20:29 same problems will appear there Jan 31 21:20:34 exactly Jan 31 21:20:39 brb... Jan 31 21:20:45 can you send a mail to me and him please ? Jan 31 21:25:49 sure, i will introduce you - he was the tall guy next to me Jan 31 21:25:52 at the CCC meeting Jan 31 21:31:45 ah ok Jan 31 21:33:13 ok, mail sent Feb 01 01:16:15 lars * r14335 /trunk/scripts/metadata.pl: Feb 01 01:16:15 Reset condition for each dependeny when generating mconf files. Previously all Feb 01 01:16:15 dependencys listed after a conditional dependency had the same condition. Feb 01 02:09:51 lars * r14336 /packages/Xorg/xorg/xserver/ (10 files in 4 dirs): Seperate xserver-xorg and xserver-kdrive-*. **** ENDING LOGGING AT Sun Feb 01 02:59:57 2009