**** BEGIN LOGGING AT Wed Dec 23 02:59:58 2015 Dec 23 03:03:46 build #183 of oxnas is complete: Failure [failed compile_5] Build details are at http://buildbot.openwrt.org:8010/builders/oxnas/builds/183 Dec 23 03:05:44 build #182 of brcm63xx is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/brcm63xx/builds/182 Dec 23 03:08:05 build #138 of mpc83xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/mpc83xx/builds/138 Dec 23 03:11:38 build #137 of ep93xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/ep93xx/builds/137 Dec 23 04:29:06 http://rznt.com/iproute2.diff - iproute 4.3.0 patches updated to build properly, fixes tc connmark issues on 4.x Dec 23 04:30:59 uploaded patch to https://dev.openwrt.org/ticket/21374 Dec 23 04:32:18 nyt: when the version changes, you can reset PKG_RELEASE to 1 Dec 23 04:32:30 ah cool, wasn't sure Dec 23 04:32:45 was just a rough patch for nbd or whoever maintains to go over Dec 23 05:17:10 build #187 of ramips is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/ramips/builds/187 Dec 23 05:25:17 nyt: also you should add a signed-of-by and mail it to openwrt-devel@lists.openwrt.org Dec 23 05:25:51 i just fixed some diffs Dec 23 07:20:18 build #138 of mvebu is complete: Failure [failed compile_5] Build details are at http://buildbot.openwrt.org:8010/builders/mvebu/builds/138 Dec 23 08:19:33 nbd ping Dec 23 08:34:41 nbd, regarding changeset 47960, the same dependency is needed by hwmon-pwmfan Dec 23 08:35:34 so line 292 in hwmon.mk should read $(call AddDepends/hwmon,+kmod-thermal) Dec 23 09:16:11 is this safe to call write_lock in a timer callback function? Dec 23 09:16:45 build #183 of ramips.mt7628 is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/ramips.mt7628/builds/183 Dec 23 10:10:06 rmilecki: i don't think so Dec 23 10:10:29 *facepalm* https://github.com/xelerance/xl2tpd/commit/23e47d52915c70515b8aa828e2ec3b2272a280d7 Dec 23 10:10:42 rmilecki: let me check Dec 23 10:10:52 adding a dependency to libm to compute "pow(2, buf->retries-1)", what a good idea Dec 23 10:11:02 rmilecki: or hm, there are _bh and _irq variants of write_lock, so it might be safe after all Dec 23 10:11:08 rmilecki: what are you trying to do? Dec 23 10:12:01 zorun: heh, i guess somebody didn't know about bit shifting ;) Dec 23 10:12:41 nbd: what bugs me is that it got merged upstream :) Dec 23 10:14:04 i think it's not a big deal though, libm is present on pretty much every system anyway Dec 23 10:14:16 and on a libc like musl it's not even a separate library Dec 23 10:15:41 ah ok, I would have thought it would add bloat on OpenWRT Dec 23 10:17:32 rmilecki r47962 trunk/target/linux/ (5 files in 5 dirs) * ledtrig-netdev: update base driver instead of patching it for every kernel Dec 23 10:20:20 nbd: but it means i need to use _irqsave then Dec 23 10:20:31 not necessarily Dec 23 10:20:31 plain write_lock will lead to a deadlock Dec 23 10:20:37 it depends on the context Dec 23 10:20:46 it works like this: Dec 23 10:21:05 for normal process context you need to use the variant that protects against the most restrictive context Dec 23 10:21:25 so if you have another function that locks from irq context, you need to use _irqsave in process context Dec 23 10:21:34 if you have another tasklet function, you need to use _bh Dec 23 10:21:46 from within the tasklet function, you don't need to use _bh Dec 23 10:21:50 the plain lock is enough Dec 23 10:22:16 because what the _bh lock does is ensure that no tasklet on the same cpu can interrupt the locked region Dec 23 10:22:21 since that would deadlock Dec 23 10:22:32 that makes sense Dec 23 10:22:34 thanks! Dec 23 10:22:43 so where do you intend to use the write_lock Dec 23 10:22:49 i'm asking, because i never needed it for anything myself Dec 23 10:23:27 nbd: i'm fixing deadlock in ledtrig-netdev Dec 23 10:24:03 so far I changed all calls to *_lock_irqsave and *_lock_irqrestore and it fixed deadlock for me Dec 23 10:24:16 i'll try less restrictive way now Dec 23 10:24:49 afaiu, i don't need _irq(save|restore) in netdev_trig_timer Dec 23 10:24:57 i need them only our of timer function Dec 23 10:26:09 i think the locking scheme of this code is crappy and should be replaced entirely Dec 23 10:26:40 how should I improve it? Dec 23 10:27:15 i'm reading the code right now to figure out the best way to deal with it Dec 23 10:28:25 i think if you restructure the code a bit, you can probably even get rid of the locking completely Dec 23 10:28:54 hm, no, that won't work Dec 23 10:29:01 because of device name changes Dec 23 10:30:58 nbd: this is what I'm testing right now: http://pastebin.com/A9X2KbVB Dec 23 10:31:52 i think you should make this work without any irqsave locking at all Dec 23 10:32:04 nbd: i didn't modify netdev_trig_timer (called by a kernel timer API), it still uses plain write_lock and write_unlock Dec 23 10:32:04 step 1: convert the lock to a regular spin lock Dec 23 10:32:11 read/write locking adds useless complexity Dec 23 10:32:23 step 2: remove the lock from the timer function Dec 23 10:32:31 you can do this by calling del_timer() on updates Dec 23 10:32:40 and ensuring that the timer gets armed again after the update is finished Dec 23 10:32:47 that simplifies context handling Dec 23 10:33:20 hm, interesting Dec 23 10:33:25 i didn't think about deletimg timer Dec 23 10:33:42 nbd: but there is still netdev_trig_notify Dec 23 10:33:52 do we want to unregister/register notifier as well? Dec 23 10:33:59 isn't it getting too complex at this point? Dec 23 10:34:14 need to check about the context of the notifier Dec 23 10:34:23 i mean deleting timer + unregistering listener & adding timer + registering listener Dec 23 10:35:04 nbd: ah, but I still should leave locking in the driver Dec 23 10:35:05 ok Dec 23 10:35:08 i mistunderstand that Dec 23 10:35:23 *stood Dec 23 10:36:09 yes, leave locking in the driver Dec 23 10:36:16 use _bh locking Dec 23 10:36:30 the notifier calls have rtnl held Dec 23 10:36:41 so i'm pretty sure it's tasklet context Dec 23 10:36:49 not sure though Dec 23 10:36:56 but _bh locking everywhere should definitely be safe Dec 23 10:37:29 so, about the timer Dec 23 10:38:13 in led_device_name_store and other places that do updates, ensure that set_baseline_state is called unconditionally Dec 23 10:38:58 if you use the pattern of deleting the timer at the beginning of every update, you can remove the del_timer call in set_baseline_state as well Dec 23 10:39:23 you can the rely on the fact that after the call, the timer function will always be armed if necessary Dec 23 10:44:45 does that make sense? Dec 23 10:46:33 yes, thanks Dec 23 10:46:44 i'll work on this soon Dec 23 10:46:52 just taking a short break Dec 23 10:48:23 nbd, have you finished talking to rafal? Dec 23 10:48:44 i didn't want to jump into the discussion Dec 23 10:48:50 yes Dec 23 10:48:56 :) Dec 23 10:49:13 i wrote you earlier regarding the thermal kernel module Dec 23 10:49:26 hwmon-pwmfan relies on it too Dec 23 10:49:41 ok Dec 23 10:49:44 feel free to send a patch Dec 23 10:49:56 not sure if adding +kmod-thermal causes a recursive dependency though Dec 23 10:50:05 it does Dec 23 10:50:07 and maybe you can do the same kind of conditional dependency that i added for gpiofan Dec 23 10:50:24 at least for gpiofan this matched the kernel's kconfig dependencies Dec 23 10:50:35 i.e. if thermal support is selected, gpiofan depends on it Dec 23 10:50:36 otherwise not Dec 23 10:50:57 i see Dec 23 10:51:20 i'll probably get to it after the new year, extremely busy here these days :( Dec 23 10:52:04 ok Dec 23 10:52:07 thanks :) Dec 23 10:54:31 * russell-- curious about progress on rt5350 ethernet, broken since merging the new ethernet driver Dec 23 11:14:13 build #182 of cobalt is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/cobalt/builds/182 Dec 23 11:15:23 nbd r47963 trunk/target/linux/generic/patches-3.18/ (12 files) * kernel: backport all current pppoe kernel fixes to 3.18 Dec 23 11:15:45 nbd r47964 branches/chaos_calmer/ (12 files) * kernel: backport all current pppoe kernel fixes to 3.18 Dec 23 11:17:00 build #182 of cns21xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/cns21xx/builds/182 Dec 23 11:19:52 build #182 of orion is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/orion/builds/182 Dec 23 11:23:40 nbd r47965 trunk/toolchain/gcc/common.mk * gcc: remove v4.6 relicts Dec 23 11:28:46 nbd, https://patchwork.ozlabs.org/patch/560462/ Dec 23 11:29:14 does it look OK? Dec 23 11:30:35 i'm still new to submitting patches... Dec 23 11:30:35 yep Dec 23 11:30:39 thanks :) Dec 23 11:31:16 oh, wait Dec 23 11:31:21 it doesn't work Dec 23 11:31:26 you removed the , Dec 23 11:31:43 true Dec 23 11:31:49 let me resend it Dec 23 11:35:10 done Dec 23 11:37:09 nbd: do you know if there are some OpenWrt ppl doing stuff at the congress / 32C3 ? (lightning talk,presentation) Dec 23 11:41:14 nbd r47966 trunk/target/linux/generic/patches-3.18/081-06-ppp-don-t-set-sk_state-to-PPPOX_ZOMBIE-in-pppoe_disc.patch * kernel: add remaining kernel patch accidentally left out of r47963 Dec 23 11:41:21 plntyk: i'm going to do a talk myself Dec 23 11:41:49 plntyk: https://events.ccc.de/congress/2015/Fahrplan/events/7375.html Dec 23 11:42:03 ah thanks for info :) Dec 23 11:42:13 nbd r47967 branches/chaos_calmer/target/linux/generic/patches-3.18/081-06-ppp-don-t-set-sk_state-to-PPPOX_ZOMBIE-in-pppoe_disc.patch * kernel: add remaining kernel patch accidentally left out of r47963 Dec 23 11:43:39 nitroshift: that one looks better Dec 23 11:43:39 nbd r47968 trunk/package/kernel/linux/modules/hwmon.mk Dec 23 11:43:39 kernel: make kmod-thermal dependency conditional to match upstream kernel module configuration dependency for hwmon-pwmfan Dec 23 11:43:42 nbd, is that conference going to be streamed online? Dec 23 11:43:45 thanks :) Dec 23 11:44:13 yes, all talks are streamed Dec 23 11:44:22 and recordings will be available afterwards Dec 23 11:44:43 great! Dec 23 11:48:27 nitroshift: the CCC conferences are really nice from what i've heard/watched on stream (openwrt related links at : https://wiki.openwrt.org/about/talks ) Dec 23 11:49:32 its a mix between hackers, culture, development, politics, social meet & lots of other thing probably Dec 23 11:49:59 plntyk, thanks, watching now :) Dec 23 11:52:10 nitroshift: for example at Chaos Computer Club Camp they were handing out radiobadges which is potentially openwrt - related (SDR radio stuff) - see https://twitter.com/rad1obadge for some pretty pictures Dec 23 11:53:27 its really difficult to get any "work" or activity done during christmas and the post-christmas congress :) Dec 23 11:55:47 plntyk, true :)) Dec 23 12:00:12 nbd: RFC http://pastebin.com/YgHWUXyG Dec 23 12:00:17 nbd: this is step 1 Dec 23 12:13:06 build #179 of pxa is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/pxa/builds/179 Dec 23 12:58:40 nitroshift: i'm afraid GMail doen't like Yahoo Dec 23 12:58:46 probably because of amount of spam it sends Dec 23 12:58:54 nitroshift: found your patches in spam dir Dec 23 13:16:36 rmilecki, i'll use gmail next time Dec 23 13:28:31 build #185 of ipq806x is complete: Failure [failed compile_4] Build details are at http://buildbot.openwrt.org:8010/builders/ipq806x/builds/185 Dec 23 13:31:59 rmilecki: the patches look good to me Dec 23 14:43:14 blogic r47969 trunk/target/linux/ lantiq/dts/TDW8970.dts lantiq/dts/TDW89X0.dtsi * lantiq: Configure LED polarity for TDW8970 and TDW8980. Dec 23 14:43:20 blogic r47970 trunk/target/linux/rb532/image/Makefile * rb532: increase kernel partition Dec 23 14:43:26 blogic r47971 trunk/target/linux/rb532/Makefile * rb532: build squashfs image by default Dec 23 14:43:32 blogic r47972 trunk/target/linux/rb532/Makefile * rb532: switch to 4.1 Dec 23 14:43:57 blogic r47973 trunk/target/linux/ (14 files in 10 dirs) * ar71xx: Add support for AirTight Networks C-55 Dec 23 14:44:02 blogic r47974 trunk/package/network/services/dnsmasq/files/dnsmasq.init * dnsmasq: Add option --no-ping Dec 23 14:44:11 blogic r47975 trunk/tools/firmware-utils/src/mkfwimage2.c * firmware-utils: allow mkfwimage2 to use - in partition names Dec 23 14:44:26 blogic r47976 trunk/package/network/services/lldpd/Config.in * package/lldpd: Remove extraneous select Dec 23 14:44:34 blogic r47977 trunk/package/utils/util-linux/Makefile * util-linux: remove outdated configure options Dec 23 14:44:39 blogic r47978 trunk/tools/e2fsprogs/Makefile * tools/e2fsprogs: remove outdated configure args Dec 23 14:44:45 blogic r47979 trunk/package/network/services/openvpn/Makefile * openvpn: fix configure options Dec 23 14:44:52 blogic r47980 trunk/package/ base-files/Makefile base-files/files/etc/init.d/system * base-files: allow timezone to be overriden by zonename (proper zoneinfo support) Dec 23 14:45:07 blogic r47981 trunk/target/linux/ ar71xx/patches-4.1/739-MIPS-ath79-add-gpio-func-register-for-QCA955x-SoC.patch ar71xx/patches-4.1/621-MIPS-ath79-add-support-for-QCA956x-SoC.patch * ar71xx: update QCA956x support Dec 23 14:45:18 blogic r47982 trunk/target/linux/ar71xx/patches-4.1/622-MIPS-ath79-add-support-for-QCA956x-ethernet.patch * ar71xx: add support for QCA956x ethernet Dec 23 14:45:27 blogic r47983 trunk/target/linux/ (7 files) * ar71xx: refresh patches Dec 23 14:45:43 blogic r47984 trunk/package/firmware/ath10k-firmware/Makefile * ath10k-firmware: Update QCA988X firmware to 10.2.4.97-1 Dec 23 14:45:51 blogic r47985 trunk/target/linux/ar71xx/base-files/lib/preinit/82_patch_ath10k * ar71xx: Update preinit script for new ath10k firmware Dec 23 15:45:57 rmilecki r47986 trunk/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: switch rwlock to spinlock Dec 23 15:46:04 rmilecki r47987 trunk/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: drop locking from timer callback function Dec 23 15:47:51 rmilecki r47988 branches/chaos_calmer/ target/linux/generic/patches-3.18/831-ledtrig_netdev.patch target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: update base driver instead of patching it for every kernel Dec 23 15:48:03 rmilecki r47989 branches/chaos_calmer/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: switch rwlock to spinlock Dec 23 17:11:19 build #174 of ppc40x is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/ppc40x/builds/174 Dec 23 17:12:10 rmilecki r47990 trunk/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: reset link status & stats after changing device_name Dec 23 17:13:17 rmilecki r47991 branches/chaos_calmer/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: drop locking from timer callback function Dec 23 17:13:25 rmilecki r47992 branches/chaos_calmer/target/linux/generic/files/drivers/leds/ledtrig-netdev.c * ledtrig-netdev: reset link status & stats after changing device_name Dec 23 18:25:04 blogic r47993 trunk/target/linux/lantiq/patches-4.1/0025-NET-MIPS-lantiq-adds-xrx200-net.patch * lantiq: disable SW_PORTMAP usage in the ethernet driver Dec 23 18:25:17 blogic r47994 trunk/target/linux/lantiq/dts/TDW89X0.dtsi * lantiq: TDW89x0 - increase spi frequency Dec 23 18:25:24 blogic r47995 trunk/package/kernel/lantiq/ltq-adsl-mei/patches/100_no-date-time.patch * lantiq: ltq-adsl-mei: fix typo Dec 23 18:25:44 blogic r47996 trunk/target/linux/lantiq/patches-4.1/0001-MIPS-lantiq-add-pcie-driver.patch * lantiq: fix PCI_DEVICE_ID_LANTIQ_PCIE Dec 23 19:24:52 blogic r47997 trunk/target/linux/ generic/files/drivers/net/phy/swconfig.c generic/files/include/linux/switch.h generic/files/include/uapi/linux/switch.h * swconfig: add SWITCH_TYPE_LINK and support sending link info to user space Dec 23 19:25:07 blogic r47998 trunk/package/ network/config/swconfig/src/swlib.h network/config/swconfig/src/cli.c network/config/swconfig/src/swlib.c * swconfig: support receiving SWITCH_TYPE_LINK from kernel Dec 23 19:25:24 blogic r47999 trunk/target/linux/generic/files/drivers/net/phy/swconfig.c * swconfig: switch kernel PORT_LINK support to SWITCH_TYPE_LINK Dec 23 19:25:54 blogic r48000 trunk/package/system/ca-certificates/Makefile * ca-certificates: update to version 20151214 Dec 23 19:26:08 blogic r48001 trunk/package/system/fstools/Makefile * fstools: update to latest git HEAD Dec 23 19:26:21 blogic r48002 trunk/target/linux/lantiq/image/Makefile * lantiq: fix Image Builder Dec 23 19:26:33 blogic r48003 trunk/package/devel/binutils/Makefile * binutils: upgrade to version 2.25 Dec 23 19:26:49 blogic r48004 trunk/package/libs/libnl/ patches patches/001-fix-poll-h-include-warning-on-musl.h * libnl: fix warning with poll.h include on musl Dec 23 19:36:46 build #142 of ramips.rt305x is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/ramips.rt305x/builds/142 Dec 23 20:56:47 build #183 of brcm63xx is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/brcm63xx/builds/183 Dec 23 21:21:38 build #150 of avr32 is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/avr32/builds/150 Dec 23 23:42:40 nbd r48005 trunk/target/linux/ar71xx/base-files/lib/upgrade/platform.sh * ar71xx: fix sysupgrade for wnr2200 Dec 23 23:42:45 build #154 of xburst is complete: Failure [failed compile_5] Build details are at http://buildbot.openwrt.org:8010/builders/xburst/builds/154 Dec 24 02:51:18 build #133 of mpc52xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/mpc52xx/builds/133 **** ENDING LOGGING AT Thu Dec 24 02:59:59 2015