**** BEGIN LOGGING AT Thu Jan 21 02:59:59 2016 Jan 21 05:59:58 rmilecki r48423 trunk/target/linux/ bcm53xx/patches-4.4/180-PCI-iproc-Fix-BCMA-PCIe-bus-scanning-regression.patch bcm53xx/patches-4.4/800-Revert-PCI-iproc-Add-PAXC-interface-support.patch bcm53xx/patches-4.1/800-Revert-PCI-iproc-Add-PAXC-interface-support.patch bcm53xx/patches-4.1/175-PCI-iproc-Fix-BCMA-PCIe-bus-scanning-regression.patch * bcm53xx: f Jan 21 05:59:59 OpenWrtix iProc regression with sent patch instead of a revert Jan 21 06:41:26 hm Jan 21 06:41:27 this is really weird Jan 21 06:41:39 writing sysupgrade stuff directly to the partition doesn't always boot past the kernel Jan 21 06:41:59 however when i pull the partition, it matches the rootfs exactly Jan 21 06:58:31 morning guys Jan 21 07:11:23 nbd, ping Jan 21 09:05:09 cyrusff: hey, I was looking into ipsec/strongswan setup and couldn't find the file which generates the strongswan config from /etc/config/ipsec Jan 21 09:05:17 cyrusff: any pointers? Jan 21 09:52:39 nitroshift: pong Jan 21 10:01:25 is there any "better" way of doing dependencies on one of two packages? I'm getting some circular dependencies: http://paste.fedoraproject.org/313119/70444145/ Jan 21 10:18:09 nbd, nevermind, i found my own error Jan 21 10:18:44 thanks anyway :) Jan 21 10:21:41 https://dev.openwrt.org/changeset/45873 What does controlling this independently accomplish? Jan 21 11:47:02 nbd r48424 trunk/package/boot/grub2/Makefile * grub2: switch back to installing to STAGING_DIR_HOST Jan 21 11:49:22 unni2: I have always configured /etc/ipsec.conf manually with strongswn Jan 21 11:53:04 stintel: that's what I'm doing also atm. Trying to connect to aws vpc Jan 21 12:04:15 there is some uci module for strongswan Jan 21 12:04:18 but afaik it doesn't work Jan 21 12:05:51 Jan 18 18:44:19 ar0 : 00[LIB] plugin 'uci' failed to load: Error relocating /usr/lib/ipsec/plugins/libstrongswan-uci.so: uci_lookup: symbol not found Jan 21 12:06:06 I should probably report that at some point ;) Jan 21 12:06:39 is there some trick to using config_set from /lib/functions.sh? I'm trying something similar to the existing "fixup_interface" in /lib/network/config.sh but it doesn't seem to work? Jan 21 12:14:41 https://paste.fedoraproject.org/313168/37844614/ has what I'm trying, which as far as I can tell is the same as the fixup_interface. Jan 21 12:26:58 is it something related to the "cfg" bit being a section, not an entire config? Jan 21 12:27:27 calling "uci set mlifter.$cfg.plugin="user--$plugin"" instead of "config_set $cfg plugin "user--$plugin"" works just fine, but seems a bit borked. Jan 21 13:28:47 nbd r48425 trunk/package/kernel/ mac80211/files/lib/netifd/wireless/mac80211.sh mac80211/files/lib/wifi/mac80211.sh Jan 21 13:28:47 mac80211: make the path phy lookup more robust regarding config upgrades, allow partial path matching Jan 21 13:28:51 nbd r48426 trunk/package/kernel/mac80211/files/lib/wifi/mac80211.sh Jan 21 13:28:51 mac80211: strip platform/ prefix from device path to improve compatibility between kernel versions Jan 21 13:29:00 nbd r48427 trunk/include/download.mk * build: download.mk: bugfix download.pl argument call order Jan 21 13:29:07 nbd r48428 trunk/target/linux/x86/patches-4.4/100-fix_cs5535_clockevt.patch * x86: fix kernel crash on geode devices (#21667) Jan 21 13:43:08 Reverting 0ca8e85462074e713be6468a3ea7259caca1b1ea seems to allow perf to compile again. Jan 21 14:07:27 KevinDB: found the bug Jan 21 14:07:42 elfutils devs apparently assume that you always run a very recent glibc ;) Jan 21 14:08:40 nbd: To assume is to make an ass out of u & me :-) Jan 21 14:09:16 nbd r48429 trunk/package/libs/elfutils/patches/110-elf_compressed_header_support.patch * elfutils: fix compatibility with non-glibc builds Jan 21 14:09:19 exactly Jan 21 14:09:22 this should fix perf Jan 21 14:09:56 nbd: Thank you kind Sir, I'll get compilin' :-) Jan 21 14:18:36 blogic r48430 branches/chaos_calmer/target/linux/ramips/image/Makefile * ramips: HiLink HLK-RM04 image name typo (RM02 -> RM04) Jan 21 15:37:29 nbd r48431 trunk/package/devel/oprofile * oprofile: delete the package, it has become useless now that perf is working Jan 21 15:37:36 nbd r48432 trunk/package/kernel/linux/modules/video.mk * kernel: add backlight kernel module support Jan 21 15:37:44 nbd r48433 trunk/package/kernel/linux/modules/video.mk * kernel: video: add fbcon support Jan 21 15:37:52 nbd r48434 trunk/package/kernel/linux/modules/video.mk * kernel: video: add DRM core and IMX DRM support for HDMI/LVDS Jan 21 15:38:03 nbd r48435 trunk/package/kernel/linux/modules/video.mk * kernel: reduce AddDepends related cargo cult programming in video.mk Jan 21 15:38:13 nbd r48436 trunk/package/kernel/linux/modules/video.mk Jan 21 15:38:13 kernel: mark kmod-drm as hidden so that it only gets selected if a driver for it can be selected Jan 21 15:38:20 nbd r48437 trunk/package/kernel/linux/modules/video.mk * kernel: make backlight support depend on display support Jan 21 15:38:25 nbd r48438 trunk/package/kernel/mac80211/patches/310-ath9k_hw-ignore-eeprom-magic-mismatch-on-flash-based.patch * ath9k: ignore EEPROM mismatch on flash based devices (fixes #21611) Jan 21 15:44:40 nbd: Hi Felix, I'm having trouble with extra radio interfaces being created on reboot, I suspect related to 48426/48425 Jan 21 15:45:38 hm Jan 21 15:45:38 right Jan 21 15:45:40 i see the same Jan 21 15:45:44 will look into it Jan 21 15:46:13 nbd: Thank you Jan 21 15:52:34 nbd r48439 trunk/package/kernel/mac80211/files/lib/wifi/mac80211.sh * mac80211: fix check for existing config section from r48426 Jan 21 15:52:39 KevinDB: fixed now Jan 21 15:56:16 nbd r48440 trunk/package/kernel/mac80211/files/lib/wifi/mac80211.sh * mac80211: only strip the platform/ prefix from path for pci devices Jan 21 15:59:36 nbd: Thank you - will let you know if any further problems found :-) Jan 21 16:00:21 nbd r48441 trunk/package/libs/uclient/Makefile * uclient: update to the latest version, fixes overwrite with wget -O Jan 21 16:05:28 nbd: thanks for the drm cleanup Jan 21 16:06:30 I had to google the term 'cargo cult programming' :) Jan 21 16:07:15 it is true that likely most of us learn from example in openwrt and there are lots of perhaps bad examples and little documentation showing the proper methods Jan 21 16:08:57 nbd r48442 trunk/target/linux/generic/config-4.4 * kernel: switch to SLUB by default in 4.4 (for improved performance) Jan 21 16:09:02 nbd r48443 trunk/target/linux/ (5 files in 5 dirs) * kernel: clean up redundant SLUB/SLAB related defines for 4.4 Jan 21 16:09:06 nbd: I'm still confused why some depends have + and others dont Jan 21 16:09:35 that will probably also require some more cleanup Jan 21 16:09:56 not having + makes sense if the options are grouped together it's natural to enable the subsystem first and then select the drivers Jan 21 16:10:11 but it makes auto-selecting such items more difficult Jan 21 16:10:23 and it's really counter-intuitive if the options are not in the same menu and users don't know where to look for stuff Jan 21 16:10:28 so i tend to prefer using + Jan 21 16:12:28 speaking of +, is there way of having a package c, that depends on a|b, and have package d do something like "+c +a" ? I get circular dependencies problems when I tried it earlier Jan 21 16:13:05 ok, makes sense - and I see the dependency types are well documented in https://wiki.openwrt.org/doc/devel/packages#dependencytypes Jan 21 17:12:58 nbd: Can confirm the duplicate radio thing is un-broken again :-) Thanks for the fix. Jan 21 18:43:11 hm.... mwlwifi is on hold until someone more qualified or someone with more time than the linux-wireless maintainer can review the patches Jan 21 19:01:34 nbd r48444 trunk/toolchain/musl/patches/001-eliminate-protected-visibility-data-in-libc.so-with-.patch Jan 21 19:01:34 musl: backport a fix that caused the wrong relocations to be emitted on some architectures Jan 21 19:11:07 plntyk, partx doesn't list sector size unfortunately.. Jan 21 19:11:35 guess i can divide size by sectors Jan 21 19:11:53 will adjust Jan 21 19:17:10 nyt: i found that kpartx seems to print sector size - dunno what/why each program is used best (havent checked program sizes) Jan 21 19:18:08 yeah im doing that now Jan 21 19:18:29 73k partx, 246k fdisk Jan 21 22:27:00 nbd r48445 trunk/package/network/config/qos-scripts/files/usr/lib/qos/generate.sh * qos-scripts: stop overriding tx queue length Jan 21 23:11:15 nbd: I've got some slow path warnings in mt76x2_mac_process if you're interested, or is my image already out of date? https://paste.fedoraproject.org/313478/34178301/ Jan 21 23:16:28 karlp: how frequent are those? Jan 21 23:17:42 I'm trying to keep a better eye on it, noticed this tonight, last night I got a crash when my ssiter connected a 5gig device to it, but was then stable, there's been some blips, but it's kinda hard to quantify. Jan 21 23:17:58 this warning itself should be harmless Jan 21 23:17:58 need to get some more persistent logging somewhere/somehow. Jan 21 23:18:19 yeah, I've had more visible connection drops and resumptions that have less log entries, Jan 21 23:19:31 oh, looks like my build is _just_ behind the last mt76 update Jan 21 23:20:43 that last one is just for compatibility with current mac80211 Jan 21 23:20:46 no functional changes Jan 21 23:21:41 I had a problem with that build when it first came up as it created an extra radio interface, but thankfully the other one still worked and I could get on and fix it, Jan 21 23:22:20 well, I'll try keeping an eye on it a bit more, is remote syslog the best way of trying to keep logs across crashes and things? Jan 21 23:22:37 I guess I could put a usb stick in and have it keep them for longer too maybe Jan 21 23:22:48 nbd r48446 trunk/package/network/config/qos-scripts/files/usr/lib/qos/generate.sh * qos-scripts: Allow classification by the traffic's source interface Jan 21 23:22:54 nbd r48447 trunk/tools/flock/Makefile * flock: do not use ccache, fixes build error Jan 21 23:36:32 nbd: Sorry about that. I actually ran in to that particular issue while I was testing it, but I thought I had it solved. Jan 22 00:04:04 so whats the best way to do some simple division in ash during sysupgrade Jan 22 00:04:27 expr? Jan 22 00:55:16 nbd r48448 trunk/package/kernel/mac80211/patches/ 543-ath9k_entropy_from_adc.patch 311-ath9k-do-not-limit-the-number-of-DFS-interfaces-to-1.patch 530-ath9k_extra_leds.patch 501-ath9k_ahb_init.patch * ath9k: allow multiple virtual interfaces on DFS channels Jan 22 01:00:00 plntyk, http://pastebin.com/raw/jbbVD5dx changed to partx, gonna test it in a few Jan 22 01:12:23 nbd: When you tested the qos-scripts IPv6 patch, what platform was that on and how were you restarting it? I can't seem to reproduce the problem you found by either multiple invocations of "/etc/init.d/qos restart" or "/etc/init.d/qos start". Jan 22 01:12:52 i tested ar71xx Jan 22 01:13:01 and i used /etc/init.d/qos start Jan 22 01:13:14 I also tried manually running generate.sh with the "firewall stop" argument and verified that the printed commands were correct to remove the rules. Jan 22 01:13:57 i'll see if i can find the time to look into this some more soon Jan 22 01:14:14 what issue were you seeing? Jan 22 01:14:21 because i see qos-start and qos-stop eat shit all the time lol Jan 22 01:14:39 Yeah, they always spit out warnings. That isn't new with my patch. Jan 22 01:15:16 nbd: Could you pastebin the output of "/usr/lib/qos/generate.sh firewall stop"? That would be tremendously helpful for debugging. Jan 22 01:15:28 ok, sec Jan 22 01:15:33 I don't have a spare ar71xx device I can test, sorry. Jan 22 01:15:43 once i submit this stupid partition shit Jan 22 01:15:50 gonna see why throughput to he tunnels is such crap :/ Jan 22 01:17:24 mamarley: no need for pastebin, the output is empty Jan 22 01:17:58 Argh. But the iptables rules are there? Jan 22 01:18:03 yes Jan 22 01:19:16 That is extremely odd. Jan 22 01:24:00 It would appear that either grep or sed is not working as intended, which sound crazy when I say it. Jan 22 01:24:11 i was just reaching the same conclusion Jan 22 01:24:17 iptables -t mangle -S | grep '^-N qos_\|-j qos_' Jan 22 01:24:20 returns nothing Jan 22 01:24:25 so already the first grep command fails Jan 22 01:25:08 My test environment is using glibc; is there any chance that has something to do with it? Jan 22 01:25:15 (I assume you are using musl?) Jan 22 01:25:35 Let me recompile with musl and see if that makes a difference. This will take a bit. Jan 22 01:28:08 yes, i use musl Jan 22 01:28:26 and i found a way to make the grep statement work with musl Jan 22 01:28:36 grep -E '(^-N qos_|-j qos_)' Jan 22 01:28:42 which i think makes more sense than what was there before Jan 22 01:29:07 then there's other issues if i fix that one Jan 22 01:29:13 Another app is currently holding the xtables lock. Perhaps you want to use the -w option? Jan 22 01:29:16 ip6tables: Directory not empty. Jan 22 01:29:18 ip6tables: Too many links. Jan 22 01:29:32 i'll drop the patch from my tree again Jan 22 01:42:40 I can reproduce that. It looks to be timing-related, if one tries to add the rules too soon after removing them or vice-versa. Let me see if I can figure something out. Jan 22 01:45:13 ok, this sysupgrade stuff seems solid.... mailing patch to the list Jan 22 01:54:25 https://patchwork.ozlabs.org/patch/571420/ wee Jan 22 02:45:31 nbd: I have fixed the issue with the warnings by passing "-w" to the iptables commands that remove the rules, which causes them to wait for the lock. This makes it marginally slower, but it doesn't print any warnings anymore. I will reply with an updated patch. **** ENDING LOGGING AT Fri Jan 22 02:59:58 2016