**** BEGIN LOGGING AT Wed Jul 04 02:59:56 2007 Jul 04 03:55:21 nbd * r7869 /trunk/target/linux/atheros-2.6/files/arch/mips/atheros/ (19 files in 3 dirs): refactor atheros system code - also add support for the reset button (sends netlink messages in the same format as broadcom-diag) Jul 04 04:35:41 nbd * r7870 /trunk/ (6 files in 5 dirs): add failsafe support for atheros-2.6 (#1501) - only works with ar5315+ at the moment Jul 04 04:39:03 nbd * r7871 /tags/kamikaze_7.07/ (25 files in 8 dirs): merge r7869 and r7870 to 7.07 Jul 04 04:42:38 :) Jul 04 04:43:41 <[mbm]> spiffy, now if only we could figure out my pesky wifi issue Jul 04 04:44:13 what wifi issue? Jul 04 04:44:40 <[mbm]> the one we were talking about 6 or so hours ago Jul 04 04:44:54 ah, the broadcom one Jul 04 04:45:00 no idea Jul 04 04:45:05 stupid wl driver Jul 04 04:45:08 <[mbm]> yeah, I can give you ssh access if you wanted to poke at it Jul 04 04:45:14 i have a device with broadcom wifi here Jul 04 04:45:33 <[mbm]> but basically I got sick of it and haven't looked back Jul 04 04:46:22 <[mbm]> I can't find ifdown being triggered, yet it's the "ifconfig ... down" in ifdown that's somehow causing the problem Jul 04 04:46:40 ifdown isn't being triggered, trust me Jul 04 04:47:04 <[mbm]> right, was getting to that conclusion, which is what makes it surprising that the issue can be bypassed via ifdown Jul 04 04:47:38 try ifup -a; ifconfig wl0 up; sleep 2; wifi Jul 04 04:47:45 or did you try that one already? Jul 04 04:48:01 <[mbm]> can't remember; let me try it Jul 04 04:50:03 <[mbm]> ifup wifi; ifconfig wl0 up; sleep 20; wifi up .. nope .. first call wifi always results in a subsequent teardown Jul 04 04:50:33 <[mbm]> in other words if I stop it in the /lib/wireless/broadcom.sh, it actually does bring up the interface correctly Jul 04 04:51:08 <[mbm]> s/wireless/wifi/ Jul 04 04:51:46 <[mbm]> think before when you'd hinted at wlc not always working you were actually running into the issue I'm seeing Jul 04 04:51:56 hmm Jul 04 04:51:56 <[mbm]> which iwn't a wlc/wl problem Jul 04 04:52:00 strange Jul 04 04:54:19 <[mbm]> got the end of enable_broadcom doing a 'wl statis; sleep; wl status' and it looks like I can sleep as long as I want there Jul 04 04:54:36 <[mbm]> teardown happens after enable_broadcom returns Jul 04 04:54:55 wtf? Jul 04 04:54:59 let me check Jul 04 04:56:05 [mbm]: please run sh -x Jul 04 04:56:09 sh -x /sbin/wifi up Jul 04 04:57:16 <[mbm]> not seeing anything after my last wl status which returned that wl was up and had the correct ssid Jul 04 04:57:36 <[mbm]> err.. doh.. wrong log Jul 04 04:57:55 <[mbm]> wl status; sleep 10; wl status .. first passed, second failed Jul 04 04:59:30 [mbm]: http://nbd.name/diff.txt Jul 04 04:59:44 i have a theory Jul 04 04:59:51 the device is brought up correctly Jul 04 04:59:56 the network scripts create a bridge for the wifi interface Jul 04 05:00:06 but before ifconfig br-wifi is called, wl0 is added to this bridge Jul 04 05:00:14 and because the bridge is down, wl0 is brought down with it Jul 04 05:00:24 and thus loses its settings Jul 04 05:00:33 <[mbm]> sounds plausable Jul 04 05:00:48 this diff should fix it Jul 04 05:00:54 <[mbm]> but a sleep doesn't seem like the best way to resolve a race condition Jul 04 05:01:03 this will never take long Jul 04 05:01:07 it's probably not necessary Jul 04 05:01:16 i just want to make sure i'm not doing things too fast Jul 04 05:01:38 because of setfd 0, the bridge should come up immediately Jul 04 05:03:44 <[mbm]> hm that was the bug, although it doesn't require the sleep in this condition Jul 04 05:03:49 ok Jul 04 05:03:53 i'll remove it and commit Jul 04 05:05:43 nbd * r7872 / (2 files in 2 dirs): Jul 04 05:05:43 fix a race condition with broadcom wl devices in a bridge Jul 04 05:05:43 if a device is added before the bridge is brought up, the wifi settings are lost Jul 04 05:05:59 * [mbm] happy Jul 04 05:06:16 <[mbm]> although somewhat ashamed I didn't find the solution myself Jul 04 05:06:44 blame it on the complexity of the network scripts :) Jul 04 05:06:57 * [mbm] choses to blame the lack of documentation Jul 04 05:07:04 that's ok as well ;) Jul 04 05:14:15 * nbd kicks off a new prerelease build Jul 04 05:17:12 <[mbm]> hmm give me a few min and I'll probably remember something else from the 7.07 todo list, although I think we've fixed (almost) all of them Jul 04 05:17:26 there are some open issues still Jul 04 05:17:33 i know that Jul 04 05:17:35 <[mbm]> well, I mean from my list Jul 04 05:17:39 ah Jul 04 05:17:48 if things still need fixing, please make tickets for them Jul 04 05:18:06 i'm currently using the 7.07 milestone as todo list for the release Jul 04 05:18:41 <[mbm]> don't think it's fixing so much as testing, we need to have a smoketest proceedure for releases Jul 04 05:18:48 yeah Jul 04 10:49:53 Kaloz: Do you have an issue switching "-mcpu=xscale" with "-march=armv5te -mtune=xscale" (some apps are starting to complain (-mcpu= depreciated and then error out - used to just be a warning) - and those are what you use for madwifi Jul 04 11:15:11 mbm * r7873 / (4 files in 4 dirs): clean up default wifi config Jul 04 11:16:06 <[mbm]> ought to be a little more blatent ;) Jul 04 11:20:43 florian * r7874 /trunk/package/util-linux/Makefile: Package the hardware clock utility and fix english typos (#2015) Jul 04 16:49:06 mbm * r7875 / (4 files in 4 dirs): patch wlc and scripts to control wl's builtin wpa supplicant Jul 04 16:50:57 mbm * r7876 / (2 files in 2 dirs): rearrange diag's platform detection code to look for known platform strings before guessing Jul 04 18:34:09 mbm * r7877 / (2 files in 2 dirs): fix bug with ap+sta modes using wpa Jul 04 20:06:49 olli * r7878 /packages/utils/zile/ (. Makefile): Add zile (#2028) Jul 04 20:19:24 ok if anyone has been frustrated that mii-tool doesn't seem to work with a SIOCGMIIPHY invalid argument error , I'v got it working, with a bit of code hack'n slashing. Jul 04 20:21:05 but I'm afraid I dont know how to make a patch or if it works with other hardware Jul 04 20:29:14 olli * r7879 /packages/utils/zile/Makefile: zile: all is not required Jul 04 21:05:15 <[florian]> Wipster: the ioctl is probably ethernet driver specific Jul 04 21:05:37 <[florian]> Wipster: doing a patch is pretty easy : download the original sources and use diff to produce the differences between the untouched sources and yours Jul 04 21:09:17 [florian]: is it worth me submitting it? its just a slight change on how it access, so it works Jul 04 21:47:57 olli * r7880 /packages/utils/vim/ (3 files in 2 dirs): vim: finally update vim to 7.1 with a small binary again (#2031) Jul 04 22:22:53 hey **** ENDING LOGGING AT Thu Jul 05 02:59:56 2007