**** BEGIN LOGGING AT Sun Nov 15 02:59:56 2009 Nov 15 13:10:47 florian * r18424 /trunk/target/linux/rdc/image/Makefile: [rdc] fix image create after the conversion to subtargets Nov 15 18:43:49 jow * r18425 /trunk/ (2 files in 2 dirs): [brcm-2.4] change switch config to swconfig style format to remain consistent accross platforms Nov 15 20:10:55 the new mac80211 had really bad performance on ath9k, STA-AP has only 1mbps, AP-STA drops connection, rcstat reports 300mbps is used while it's physically impossible for the SR71 card(only 130mbps), things are really wrong Nov 15 20:11:25 last update(mac80211-11-03) is way better than this (mac80211-11-13) package from the trunk Nov 15 20:13:35 hauke: ping Nov 15 20:13:43 pong Nov 15 20:14:03 Hauke:ever tested compat-1113? Nov 15 20:14:20 no I have never tested it Nov 15 20:14:50 sigh, not sure if nbd did it, he made some changes on rc, it seems to me it's broken Nov 15 20:15:30 nbd: ping Nov 15 20:15:39 it's not the rc changes Nov 15 20:15:45 according to reports of other people, things broke upstream Nov 15 20:15:52 i'm currently trying to figure out what's going on Nov 15 20:16:07 ok...this time it broke hard Nov 15 20:16:10 but one guy claims it happens with standard wireless-testing after a big ath9k commit a few days ago Nov 15 20:16:20 yes, i'm seeing the same breakage and i'm working on fixing it Nov 15 20:16:37 i did a diff those changes are indeed dramatic Nov 15 20:16:49 yes, but i didn't spot any obvious breakage Nov 15 20:16:54 must be something subtle Nov 15 20:17:07 anyway, will use 1103 for now, working on HT for ibss/ath9k Nov 15 20:17:35 i'm currently working on making ath9k use the mac80211 rate control api properly Nov 15 20:17:40 so i can write 11n minstrel Nov 15 20:17:50 great Nov 15 20:18:04 i can never understand the rc code Nov 15 20:18:16 the atheros rc code is really limited and oversimplified Nov 15 20:18:25 i think the minstrel approach can improve throughput Nov 15 20:18:48 nobody seems care about HT on ibss/adhoc Nov 15 20:18:49 :9 Nov 15 20:18:51 :( Nov 15 20:18:58 i know a lot of people that do Nov 15 20:19:04 unfortunately most of them aren't into hacking wifi driver stuff Nov 15 20:19:22 and i care about it as well Nov 15 20:19:26 just didn't have the time for it yet Nov 15 20:19:30 yeah i'm hacking it off work, hope to have some nice patch before christmas Nov 15 20:19:35 sounds good Nov 15 20:19:44 i'll help with getting it out to users so it'll be tested Nov 15 20:19:56 and i'll also do some code review once you're done Nov 15 20:20:07 i'd like to get this upstream soon after it's done Nov 15 20:20:16 nice. now back to 1103 to test more on ibss/HT Nov 15 20:25:30 kaloz * r18426 /trunk/tools/ (Makefile lzma/Makefile): make the lzma version kernel version dependent, and only build the one we do need Nov 15 21:04:47 xxiao: i figured out the ath9k bug Nov 15 21:04:49 it's a missing memset Nov 15 21:08:54 nbd: great. any patch that i can have a try Nov 15 21:09:24 i'll commit it in a minute Nov 15 21:09:48 ok. will monitor here and pull Nov 15 21:21:54 nbd * r18427 /trunk/package/mac80211/patches/ (520-driver_flags.patch 530-ath9k_cleanup.patch): mac80211: sync with changes to patches submitted upstream Nov 15 21:21:55 nbd * r18428 /trunk/package/mac80211/patches/550-ath9k_rx_fix.patch: ath9k: fix massive packet loss issue introduced by upstream cleanups Nov 15 21:21:58 nbd * r18429 /trunk/package/mac80211/patches/560-ath9k_rate_control_api.patch: ath9k: work-in-progress patch to adapt ath9k to the mac80211 rate control api Nov 15 21:22:00 xxiao: there you go Nov 15 21:22:32 great! will test it now Nov 15 21:23:05 nbd, are you anxious to have folks testing that ? Nov 15 21:23:20 * groz can check it out and have onto an ath9k fairly quickly if so Nov 15 21:23:40 just tell me specifically what you want done for tests, just plain thruput, or anything special ? Nov 15 21:23:56 groz: https://dev.openwrt.org/ticket/6165 Nov 15 21:26:01 throughput + packageloss Nov 15 21:26:13 ok, just read the ticket Nov 15 21:26:47 i have a couple 160nl boxes here, so, will have it on them shortly Nov 15 21:27:53 i'm pretty sure the issue is fixed Nov 15 21:28:04 before my change i could very easily trigger the issue immediately Nov 15 21:28:08 now i can't anymore ;) Nov 15 21:28:37 ok, i've got it building here, will have it on a couple boxes shortly Nov 15 21:28:49 and i'll see about stress testing them a couple of ways Nov 15 21:30:36 ath9k -- xmit.c:1581: error: too many arguments to function 'ath9k_hw_computetxtime' ---not sure this has nothing to do with the check-ins, build a new trunk now Nov 15 21:30:52 oh Nov 15 21:30:59 hm, must have forgotten a quilt add Nov 15 21:35:37 yup, same error here Nov 15 21:36:15 ath9k/xmit.c:1581: error: too many arguments to function 'ath9k_hw_computetxtime' Nov 15 21:36:42 nbd * r18430 /trunk/package/mac80211/patches/560-ath9k_rate_control_api.patch: mac80211: add some missing patch chunks that were accidentally left out in the previous commit Nov 15 21:36:45 try again now Nov 15 21:37:47 ok, building Nov 15 21:38:10 i've got some trivial patches for package/kernel, adding some more modules for more hardware, lm87 and ath9k Nov 15 21:38:15 err, not 9k Nov 15 21:38:20 i meant atl2 Nov 15 21:38:31 whats the best way to forward em for somebody to add in ? Nov 15 21:41:02 ath9k compiles ok now Nov 15 21:41:20 * nbd is getting very stable 65-70 mbit in HT20 @2.4 GHz now Nov 15 21:41:43 tested on ixp4xx with an ar9220 card Nov 15 21:44:52 nbd: what exactly ar9220 card is that? i'm still using the engineering version of ar9160 from ubnt Nov 15 21:45:09 i don't know if it's on the market Nov 15 21:45:10 let me check Nov 15 21:45:33 yup, it's a senao emp-7605 Nov 15 21:47:26 google finds it's only in europe Nov 15 21:54:59 nbd: the fix works great, both AP/STA and IBSS work well Nov 15 21:55:41 one little thing, the mac80211.sh replaced wlan0 with wifi0, however ifconfig insists only wlan0 can be used Nov 15 21:58:59 not sure why it's needed to change to wifi0, wlan0 worked well Nov 15 21:59:40 the wifi-device name is not what's supposed to be used in ifconfig Nov 15 21:59:41 wifi0 is just a faked master interface iirc Nov 15 21:59:51 wlan0 is a "vap" on top of it Nov 15 21:59:54 i should replace it with something else that doesn't clash with other drivers Nov 15 21:59:56 just an internal identifier Nov 15 22:00:21 maybe i should just reuse those old wmaster* names that mac80211 no longer creates Nov 15 22:00:30 nbd: or phyX ? Nov 15 22:00:31 then it doesn't confuse users with phy* names whose numbers don't match the actual phy numbers Nov 15 22:00:33 etc-config-wireless is now using wifi0, ifconfig always showed wlan0 Nov 15 22:00:34 ;) Nov 15 22:00:47 phy numbers change on every driver reload Nov 15 22:00:51 I see Nov 15 22:00:52 and the order might change when you add cards Nov 15 22:01:08 xxiao: virtual interface names are always autogenerated Nov 15 22:01:09 just call it "this_is_no_wireless_interface_name0" ;) Nov 15 22:01:12 you shouldn't expect them to be in the config Nov 15 22:01:27 ok...make sense Nov 15 22:02:33 with these new changes my throughput improved about 15% Nov 15 22:03:10 nbd: wmasterX sounds good, but waht about "cardX" or "radioX" ? Nov 15 22:03:23 radioX sounds good Nov 15 22:03:27 i think i'll pick that one Nov 15 22:05:06 nbd * r18431 /trunk/package/mac80211/patches/ (520-driver_flags.patch 530-ath9k_cleanup.patch): mac80211: adjust patches based on upstream feedback Nov 15 22:05:24 what's this new patch about? Nov 15 22:05:41 hehe, i'm about to flash cards Nov 15 22:05:48 should i rebuild again first ? Nov 15 22:06:24 i thought my AR9160(SR71A) is 130mbps max, debugfs rcstat showed 300mbps, hmm Nov 15 22:06:34 xxiao: no functional changes Nov 15 22:06:44 xxiao: just cleanup for upstream submission Nov 15 22:07:12 ok no need to rebuild then Nov 15 22:08:15 rcstat is right, ubnt says it's up to 300mbps(using 5Ghz HT40-) Nov 15 22:21:58 howdy y'all, offering myself as a bitch for helping with redoing docs/wiki Nov 15 22:22:28 that sounds great Nov 15 22:22:35 we can always use help with the wiki Nov 15 22:23:26 I hear it is being overhauled? Nov 15 22:23:28 true? Nov 15 22:23:35 the new wiki has been online for a while Nov 15 22:23:47 there's information on the structure in the 'meta' namespace Nov 15 22:24:11 oic, ok... I take it this is the wiki linked by the root http://openwrt.org page? Nov 15 22:24:29 yes Nov 15 22:25:11 ok, I just (wrongly) ASSumed (as it is said this does of you and me) that this was a straight copy over Nov 15 22:25:16 what is left to be done? Nov 15 22:25:37 the oldwiki has been copied over in its own namespace Nov 15 22:25:55 what's missing is extending the articles that are there, adding new docs, etc. Nov 15 22:25:59 ok, so maybe just copy/modifying the missing links? Nov 15 22:26:24 stuff in the new namespace shouldn't point to stuff in the old namespace Nov 15 22:26:36 stuff from the old namespace that's worth keeping should be moved over Nov 15 22:26:39 makes sense Nov 15 22:26:40 and integrated properly Nov 15 22:27:09 please read the stuff described under http://wiki.openwrt.org/meta/start Nov 15 22:27:19 it should explain most of the important questions Nov 15 22:27:25 and if you need further info, let me know ;) Nov 15 22:27:27 ok, sounds good Nov 15 22:27:35 will do, thank you Nov 15 23:27:33 jow * r18432 /trunk/target/linux/brcm-2.4/base-files/etc/init.d/netconfig: [brcm-2.4] add migration code to automatically convert old to new switch config on first boot, this will be removed at some point in the future Nov 16 00:44:41 nbd * r18433 /trunk/package/mac80211/patches/570-ath_use_gfp_dma.patch: ath: use GFP_DMA for rx buffer allocations to fix allocation errors on ixp4xx devices with more than 64 MB RAM **** ENDING LOGGING AT Mon Nov 16 02:59:56 2009