**** BEGIN LOGGING AT Thu Jun 06 03:00:22 2013 Jun 06 09:53:25 cyrus r36865 trunk/package/ base-files/files/etc/uci-defaults/12_network-generate-ula base-files/Makefile * base-files: use less entropy when generating ULAs Jun 06 09:54:40 cyrus r36866 branches/attitude_adjustment/package/ipv6-support/files/network-enable-ipv6.defaults * AA: backport r36865 Jun 06 11:51:43 mazilo: hi Jun 06 11:53:07 nbd: I already updated to the latest svn trunk and am compiling. Hopefully, that will fix the issue. Jun 06 11:53:17 :) Jun 06 11:53:30 nbd: It will be awhile. :( Jun 06 12:15:38 /win 13 Jun 06 13:07:42 nbd: Now, the gcc-4.8.x with Linaro compilation gets this error message (http://dpaste.com/1213811). Jun 06 13:08:15 was it a clean rebuild? Jun 06 13:08:40 did you run make dirclean or remove build_dir/staging_dir? Jun 06 13:08:57 nbd: I did a 'make dirclean'. Jun 06 13:09:10 which target did you build for? Jun 06 13:09:57 nbd: Marvell Kirkwood. Jun 06 13:12:16 build #258 of rb532 is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/rb532/builds/258 Jun 06 13:12:17 nbd: I manually check the gcc/libgcc/config/arm/t-linux file and it has been patched with your toolchain/gcc/patches/4.8-linaro/810-arm-softfloat-libgcc.patch Jun 06 13:12:45 i'll try a build for kirkwood myself Jun 06 13:12:54 nbd: OK. Jun 06 13:12:54 just tried orion and the toolchain build for that one worked Jun 06 13:16:03 build #258 of ppc44x is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/ppc44x/builds/258 Jun 06 13:16:16 nbd: The orion toolchain is the same for Marvell Kirkwood? Jun 06 13:21:56 _rmk_: that thread has some serious popcorn value Jun 06 14:03:45 jow r36867 trunk/package/network/utils/iptables/Makefile * iptables: install libext*.a into staging dir Jun 06 14:03:49 jow r36868 trunk/package/network/config/firewall/Makefile Jun 06 14:03:50 firewall: copy libext*.a from staging dir and drop kernel.mk includes, solves problem with colliding CONFIG_IPV6 symbols Jun 06 14:06:44 jow r36869 branches/attitude_adjustment/package/iptables/Makefile * AA: iptables: backport r36867 (#13653) Jun 06 14:06:50 jow r36870 branches/attitude_adjustment/package/firewall/Makefile * AA: firewall: backport r36868 (#13653) Jun 06 14:10:56 mazilo: hm, kirkwood built just fine for me Jun 06 14:11:06 mazilo: please show me the output of ./scripts/diffconfig.sh Jun 06 14:12:01 nbd: Hold on a second. Jun 06 14:13:16 nbd: http://dpaste.com/1213949/ Jun 06 14:13:23 jow r36871 trunk/package/network/config/firewall/Makefile * firewall: udpate to git head (#13652, #13654, #13658) Jun 06 14:15:45 nbd: not sure if you read this reply, so posting it again Jun 06 14:15:46 04|18:38:49< stintel> nbd: just tested r36836, streaming stalled after 1 minute, still associated but can't ping the phone Jun 06 14:21:20 stintel: i found the bug yesterday, fixed in r36859 Jun 06 14:21:22 jow r36872 branches/attitude_adjustment/package/firewall/Makefile * AA: firewall: merge r36871 Jun 06 14:27:49 is dev.openwrt.org down or just slow? Jun 06 14:28:16 it usually oscillates between being down or just slow :) Jun 06 14:28:33 thanks! :) Jun 06 14:33:13 nbd: ah, I missed that commit, I'll retry tonigh Jun 06 14:33:15 +t Jun 06 14:33:47 i'm pretty sure it'll work. it's already been confirmed by a user in the ticket Jun 06 14:34:10 alright, if something is still wrong I'll let you know Jun 06 16:00:20 I've got an openwrt x86 image running in virtualbox, and typing "reboot" at the console isn't actually rebooting it. Jun 06 16:00:24 is there a trick for that? Jun 06 16:00:41 the console prints out "Restarting system." and "machine restart" but it just hangs after that. Jun 06 16:07:03 hrmm, working thjis time. weird. Jun 06 19:41:39 nbd, forgot to mention in my mail reply to Guy and you: I'm also on IRC :) Jun 06 19:42:20 hey guerby Jun 06 19:42:25 just saw your email Jun 06 19:43:08 as for implementing AirMax in ath9k - I don't really see the point Jun 06 19:44:13 a few people in italy have run some tests and told me that plain ath9k with no airmax is faster than ubnt with airmax enabled Jun 06 19:45:37 nbd, it probably depend on the situation and number of stations but my reference to airmax was more to ask to what point it's possible to tweak sending and receiving algorithm Jun 06 19:45:56 it's possible, yes Jun 06 19:45:59 but not very high on my todo list Jun 06 19:46:20 i will be working on some big changes to tx queueing first Jun 06 19:46:21 nbd, what are your current priroties on ath9k ? Jun 06 19:46:33 the plan is to fix bufferbloat Jun 06 19:46:47 nbd, ah ok Jun 06 19:46:49 by keeping hw queue depth very short and making software queues properly managed Jun 06 19:47:07 with individual queue length being determined by (among other things) the tx rate Jun 06 19:47:37 nbd: fyi, happily streaming youtube to the htc media link now Jun 06 19:47:44 stintel: cool Jun 06 19:47:54 lol, I hit enter and it started choking for a few seconds Jun 06 19:48:16 ok, so now a dumb question Jun 06 19:48:17 but continued fine Jun 06 19:48:22 nbd, about tx rate and modulation is it possible to break compat with regular wifi on radio frames? (eg not using MCS0 for the header) Jun 06 19:48:25 what is the 'magic' of airmax ? Jun 06 19:48:31 guerby: why? Jun 06 19:48:32 does it really do anything super special ? Jun 06 19:48:39 and what do you mean by 'header'? Jun 06 19:48:40 * groz wonders if it's mostly marketing hype Jun 06 19:49:30 groz: it seems to be adding some polling protocol on top of 802.11 to avoid collisions Jun 06 19:49:55 nbd, IIRC (I'm not a specialist) the beginning of a radio frame is always MCS0 and that's why they added frame aggregation to get closer to the highest rate in practice Jun 06 19:50:30 it's not MCS0 Jun 06 19:50:34 and it's not possible to change it Jun 06 19:50:53 I keep reading about 'tdma' as part of it, but, it's not possible to do tdma unless ALL devices on the frequency participate Jun 06 19:50:59 and even if it were possible to change it, it would make things worse for everybody when sharing the channel with other networks Jun 06 19:51:18 because if other devices can't recognize the header, they will stomp on the transmissions with no regard for collision avoidance Jun 06 19:51:46 hehe, and lots of them do that already nbd Jun 06 19:53:17 nbd, most of our important links (backbone to go to the fiber) are in rural areas, long distance and with high gain antennas, given the space in 5 GHz we're alone on the frequency Jun 06 19:55:15 we also often use non 20/40 MHz channel width Jun 06 19:55:43 what hardware are you using guerby ? Jun 06 19:57:23 guerby: i suggest you try some links with openwrt/ath9k and see how well it performs Jun 06 19:57:43 make sure you use something recent, either current 12.09 or trunk git versions Jun 06 19:57:46 groz, ubnt M5 gear Jun 06 19:57:47 instead of the release binaries Jun 06 19:58:06 nbd, ok Jun 06 19:59:10 I've got a few m2 systems here Jun 06 19:59:26 and some unifi stuff Jun 06 20:00:15 using trunk from a few weeks ago, can consistently get between 4 and 5 megabytes a second thruput between them Jun 06 20:00:32 nbd, do you know in what part of 802.11n-2009.pdf they describe the modulation to choose for various parts of the frames Jun 06 20:00:41 ? Jun 06 20:00:56 build #272 of uml is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/uml/builds/272 Jun 06 20:01:34 guerby: get 802.11-2012 Jun 06 20:01:37 groz, we made some tests of various antennas on a 8.6km perfect LoS link, resultats here: http://lists.tetaneutral.net/pipermail/technique/2012-July/000446.html Jun 06 20:02:27 (it's in french but the report lines should be readable :) Jun 06 20:02:31 I'm not using anythign other than stock stuff here Jun 06 20:02:39 got a handful of picostations Jun 06 20:02:44 and a couple nanostations Jun 06 20:02:57 along with 4 of the unifi smoke detector form factor Jun 06 20:03:32 on the omni to parabola link we see the curve of effective bandwidth go down after 10 MHz channel size Jun 06 20:04:07 I run it all with ht20, found using ht40 works well, as long as there isn't a lot of other ap's in the vicinity Jun 06 20:04:13 nbd, done (ouch 2793 pages :) Jun 06 20:04:24 but, if there's other stuff (not mine), get better performance using ht20 Jun 06 20:04:46 and on 2.4ghz, there's ALWAYS neighbors with access points Jun 06 20:05:35 groz, yep we don't use 2.4 except indoor Jun 06 20:05:58 groz, it used to work outdoor a few years ago but now except in the middle of nowhere it's hard to use Jun 06 20:06:00 all of mine are tasked as access points Jun 06 20:06:08 so, got to have 2.4 Jun 06 20:06:37 using the example I had before we moved last week Jun 06 20:06:41 we have to provide 2.4 outdoor for the 36th European Juggling Convention in Toulouse this july Jun 06 20:06:50 using wds hops back to the house from a ways off Jun 06 20:06:52 but it's a bit out of the city Jun 06 20:07:24 I'm not trying to do long distance back hauls Jun 06 20:08:54 but I'm a bit miffed with ubiquiti right now Jun 06 20:09:14 right after I bought the stuff, and filled out the forms online for their 'sdk', got the note saying Jun 06 20:09:20 licenses revoked Jun 06 20:09:37 * groz tries to understand how they can revoke my license to use openwrt derived stuff Jun 06 20:11:20 groz, I haven't followed what ubnt currently does licensing wise (we used the SDK a few years ago) Jun 06 20:12:13 nbd, I assume it's somewhere in chapter 9 :) Jun 06 20:16:02 guerby, one of the things they are trying to enforce now, is prevent folks from removing the ubnt web interface Jun 06 20:19:41 groz, I'm not sure they're worse than mikrotik on their hardware Jun 06 20:20:01 (I've never seen mikrotik tweaked linux sources, may be I missed something though) Jun 06 20:26:30 hehe, nope, I've never seen them either Jun 06 20:31:47 <_rmk_> blogic: thankfully not roasted popcorn... yet. Jun 06 20:31:57 <_rmk_> or maybe carbonized popcorn Jun 06 20:51:48 ok the MCS are defined in chapter 20 Jun 06 20:53:21 not found yet where the modulation to use for frames/part of frames is defined Jun 06 21:25:31 9.7 Multirate support Jun 06 21:34:05 ok found it 18.3.2 PLCP frame format Jun 06 21:34:41 for G Jun 06 21:36:11 PLCP Header coded BPSK, r = 1/2 so MCS0 Jun 06 21:36:57 now 20 for N Jun 06 21:54:35 for N : 20.3.2 PPDU format Jun 06 21:56:44 The non-HT-STF (L-STF), the non-HT-LTF (L-LTF), and the non-HT SIGNAL field (L-SIG) are defined so they can be decoded by non-HT Clause 18 and Clause 19 STAs. Jun 06 21:58:01 Figure 20-1—PPDU format => 8+8+4 = 20us using BPSK/MCS0 Jun 06 22:00:20 if I'm not mistaken Jun 06 22:00:57 18.3.3 PLCP preamble (SYNC) Jun 06 22:03:20 it's not MCS0 => looks like the first 20 us are always MCS0 or did I get something wrong? Jun 06 22:03:25 obinou, hi :) Jun 06 22:15:21 haven't looked at this in detail yet, but you may be right Jun 06 22:15:42 either way, there's no way to configure the rate for this Jun 06 22:15:50 and with aggregation it's typically not necessary either Jun 06 22:16:36 nbd, indeed (I haven't looked yet if there are restrictions when aggregating frames) Jun 06 22:19:37 nbd, with ath9k do we have access on the rate selection algorithm? (or at least statistics on packet loss per modulation) Jun 06 22:20:03 rate control is handled in mac80211, and there are detailed statistics available via debugfs Jun 06 22:20:14 and if you have questions about the actual algorithm - i wrote that thing Jun 06 22:20:33 nbd, eh :) Jun 06 22:21:27 nbd, if you have slides or video of conferences you made and that I should read/watch let me know Jun 06 22:21:42 not really Jun 06 22:21:53 hmmmm, I think I have to upgrade Jun 06 22:22:05 just found out, new internet packages coming from my provider, available next month Jun 06 22:22:10 250/15 Jun 06 22:22:19 gonna need a faster router me thinks Jun 06 22:23:02 groz: who is your ISP? Jun 06 22:23:07 shaw cable Jun 06 22:23:14 florian r36873 trunk/tools/firmware-utils/ src/bcmalgo.c Makefile src/bcmalgo.h src/hcsmakeimage.c * tools: add Broadcom cable modem firmware image creator Jun 06 22:23:19 nbd, latest release of AirOS from ubiquity added some user visible options on rate selection algorithms (I dont't know much about what RouterOS from mikrotik provides) Jun 06 22:23:19 florian r36874 trunk/target/linux/ brcm63xx/patches-3.9/424-bcm3368_support.patch brcm63xx/patches-3.8/424-bcm3368_support.patch brcm63xx/generic/config-default * brcm63xx: add support for BCM3368 Jun 06 22:23:20 groz: How much you will be paying for that speed? Jun 06 22:23:29 but I've only got 100mbit adapter in my router right now Jun 06 22:23:30 florian r36875 trunk/target/linux/ brcm63xx/patches-3.9/425-bcm933xx_hcs.patch brcm63xx/patches-3.8/425-bcm933xx_hcs.patch * brcm63xx: add Broadcom cable modem firmware format header Jun 06 22:23:36 florian r36876 trunk/target/linux/ (43 files in 3 dirs) * brcm63xx: add kernel space and image support for the Netgear CVG834G Jun 06 22:23:41 florian r36877 trunk/target/linux/ brcm63xx/base-files/lib/brcm63xx.sh brcm63xx/base-files/etc/uci-defaults/02_network * brcm63xx: add user-space support for Netgear CVG834G Jun 06 22:23:43 $115 a month with 1 tb data cap Jun 06 22:23:44 guerby: what kind of options do they offer? Jun 06 22:23:48 195 uncapped Jun 06 22:24:00 groz: Wow. Jun 06 22:24:28 it has auto bump, so, go over 1tb in a month, automatically bump to the higher billing level Jun 06 22:24:36 groz: We are with Comcast $200/month and only get 25/2 Mbps. Jun 06 22:24:41 nbd, it's basic: "Data Rate Module:" Default or Alternative Jun 06 22:24:52 heh, whatever that means Jun 06 22:24:53 [florian]: wtf, what interfaces are supported on this bcm one? do you have something for the cable side as well? Jun 06 22:24:55 the manual says to try both and keep the one that works best :) Jun 06 22:25:02 nbd, yeah :) Jun 06 22:25:02 haha Jun 06 22:25:15 i prefer having one algorithm that works well for all cases Jun 06 22:25:18 ;) Jun 06 22:25:20 nbd, you can also specify a max MCS Jun 06 22:25:33 nbd, this one is quite useful Jun 06 22:25:41 why? Jun 06 22:25:58 nbd, makes some links more stable Jun 06 22:26:20 only with a stupid algorithm or hardware issues Jun 06 22:26:27 :) Jun 06 22:27:42 nbd, all algorithms will have to try higher MCS and loose packet from time to time Jun 06 22:27:55 <[florian]> wigyori: no unfortunately, it did not even run linux, but ecos Jun 06 22:28:01 trying a higher MCS doesn't mean losing packets Jun 06 22:28:13 hardware does retransmission for that sort of thing Jun 06 22:28:39 nbd, yes of course but you make bandwidth fluctuate then Jun 06 22:28:49 more than if you didn't try Jun 06 22:28:51 not if rate probing is implemented properly Jun 06 22:29:15 [florian]: drat, okay. i've been trying to get something for a motorola sb5100 (bcm-based as well), but nothing Jun 06 22:29:20 in minstrel_ht (the mac80211 rate control module) i optimized it to reduce the overhead of rate probing a lot Jun 06 22:29:38 <[florian]> wigyori: well bcm3348 is on my todo, but yet again getting cable to work is quite difficult Jun 06 22:30:17 nbd, http://lwn.net/Articles/376765/ Jun 06 22:30:30 yep, that's the one Jun 06 22:30:42 made some more optimizations on it a few weeks ago Jun 06 22:31:43 nbd, ok good Jun 06 22:31:56 nbd, does ath9k support non 20/40 mhz channel width? Jun 06 22:32:59 5 and 10 mhz is supported Jun 06 22:34:32 nbd, nice :) what are the constraints from the hw? (just doc?) Jun 06 22:36:21 http://wireless.kernel.org/en/developers/Documentation/mac80211/RateControl/minstrel Jun 06 22:44:50 gn Jun 06 22:52:07 http://bobcopeland.com/srcs/copeland-linux-rc.pdf Jun 06 23:39:05 nbd: I tried to compile toolchain gcc-4.8.x with Linaro Enhancement for Marvell Orion and got the same error messages. **** ENDING LOGGING AT Fri Jun 07 02:59:58 2013