**** BEGIN LOGGING AT Mon Jul 20 02:59:58 2015 Jul 20 06:01:47 nbd r46433 trunk/target/linux/ar71xx/files/drivers/mtd/nand/ar934x_nfc.c Jul 20 06:01:47 ar71xx: remove the use of the obsolete IRQF_DISABLED flag in the NAND flash driver (fixes #20125) Jul 20 09:10:23 build #63 of mpc52xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/mpc52xx/builds/63 Jul 20 09:30:21 jow_laptop, I have a few thoughts on the sysupgrade script, what do you think? https://gist.github.com/hgl/f45c4f40e5beec74fe06 Jul 20 09:33:39 hgl: in principle it looks good to me. In the overlay I suggest to simply rename /upper/lib/modules/* to /upper/lib/modules/*.backup_$OPENWRT_RELEASE Jul 20 09:33:50 or something like that Jul 20 09:34:15 an mv is cheap and allows users to reclaim files if needed Jul 20 09:35:17 as for "ubus call system upgrade" Jul 20 09:35:30 this call instructs procd to replace itself (pid 1) with the passed command Jul 20 09:35:52 which is needed to fully release the rootfs without any pending services holding a lock on it Jul 20 09:36:21 so the pid 1 running from / gets replaced with a sysupgrade stage 2 process running from the /tmp ramdisk Jul 20 09:36:56 jow_laptop, so only mv kernel modules and not wipe out the overlay? that's a good idea, user can just upgrade packages with opkg direactly after the upgrading. Jul 20 09:37:17 yes Jul 20 09:37:38 the jffs2 is implictely wiped Jul 20 09:37:53 with the mtd write command? Jul 20 09:37:55 a "firmware" aprtition usually spans kernel + romfs + overlay Jul 20 09:38:32 mtd partitions may overlap, so for many (most) targets/models theres a "convenience" mtd aprtition which covers the entire firmware space Jul 20 09:39:21 so basically a partitition that covers three other partitions Jul 20 09:39:56 ah, that makes sense. Jul 20 09:40:51 whats currently not easily possible is writing arbritary files / directory structures to jffs2 after it has been formatted Jul 20 09:41:10 the current append code is only able to place one file in the root directory Jul 20 09:41:33 so you need to include the relevant kmods in the sysupgrade .tar.gz Jul 20 09:42:14 and expand that tarball after restarting? Jul 20 09:42:24 then extend the firstboot sysupgrade script to "opkg install" any kmods it finds Jul 20 09:44:57 jow_laptop, it's not currently perfectly clear to me how everything piece together. I think I will have a better idea when improving the script. will consult you then. thanks a lot for the detailed explanations. :) Jul 20 09:45:51 the sysupgrade.tgz extraction happens in preinit: http://git.openwrt.org/?p=openwrt.git;a=blob;f=package/base-files/files/lib/preinit/80_mount_root Jul 20 09:46:34 in line 12 you'd basically do something like: find /tmp/upgrade/ -type f -name '*.ipk' | xargs opkg install Jul 20 09:47:57 what are these ipk packages? the packages needed for usb storage? Jul 20 09:48:23 yes Jul 20 09:49:07 you'd extend do_save_conffiles() in /sbin/sysupgrade to figure out which packages are needed, then run "opkg download" to fetch the required kmods with all dependencies, then move those into /tmp/upgrade/ Jul 20 09:49:49 finally amend the $CONFFILES list file (using something like find /tmp/upgrade -type f -name '*.ipk' >> $CONFFILES) Jul 20 09:50:24 the hard part is figuring out required kmods in advance Jul 20 09:50:24 i don't think they should be added to the tarball with a previous release's script, a new release might use difference packages, I think i should write a new script, packed in the new release, which knows what packages to install for usb storage. Jul 20 09:50:47 it should be figured out by a new release i think. Jul 20 09:50:56 well, on firstboot you cannot rely on having network connectivity Jul 20 09:51:27 required network drivers might even reside on the overlay (think of usb 3g stuff etc.) Jul 20 09:52:20 after upgrading, before restarting, the network connectivity is lost too? Jul 20 09:53:29 or we can extract that script out of trx, and download the packages before doing upgrading. Jul 20 09:54:25 although i'm not sure how to extract files from trx Jul 20 09:54:37 there's more formats than just trx Jul 20 09:55:06 the easiest would be mouting the jsut written romfs Jul 20 09:56:22 another possibility is inferring the repository from the image url Jul 20 09:56:37 then fallback to extracting a version number from ther image filename Jul 20 09:56:43 then fallback to snapshots Jul 20 09:56:52 or something like that Jul 20 09:57:20 can i extract the release name from the image itself? Jul 20 09:58:55 not easily Jul 20 09:59:32 you'd first need to find the offset of the 1st squshfs or ubifs or ext4 partition, then loop-mount it, then look for /etc/relase in it Jul 20 10:00:01 and thats not possible with sysupgrades from http(s) since you often do not have enough space to buffer the image locally Jul 20 10:02:20 i think another way is to get the hash value of the image file, and there is a remote file that maps the hash value to the corresponding packages list... Jul 20 10:04:21 this won't work Jul 20 10:04:31 if users use custom built image. Jul 20 10:06:19 brb Jul 20 10:51:14 nbd r46434 trunk/ tools/Makefile tools/patch/Makefile Jul 20 10:51:14 tools: drop circular dependency between patch and ccache, build patch without ccache (fixes #20132) Jul 20 10:54:05 jow_laptop, i think the only reliable way is to extract the trx file before upgrading. we use the opkg.conf in the image file in case it points to a mirror repo, and use the script in that file to install usb storage packages. I think i'm going to bite the bullet and mount the trx file, is testing for ext4 necessary? is it only used in x86? i doubt people use extroot in that platform. Jul 20 13:03:04 build #63 of ar71xx.mikrotik is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/ar71xx.mikrotik/builds/63 Jul 20 13:43:56 I have got a weird QOS problem and don't know where to start diagnosing... Jul 20 13:44:19 Besides WAN and LAN I have a PUBLIC interface with limited bandwidth for guest users Jul 20 13:44:54 after about a day my internet speed on the LAN network also starts to slow down and seems to have the QOS speed limits for PUBLIC applies Jul 20 13:45:11 the weird thing is that the upload and download speeds seem to be reversed as well Jul 20 13:45:42 I have 100 MBit internet and 1/10 Mbit limit on PUBLIC (up/down from perspective of router) Jul 20 13:45:52 LAN has no limits Jul 20 13:46:05 vanDrunen: This would probably be better for the #openwrt channel. Jul 20 13:46:34 who maintains the QOS package? Jul 20 13:46:54 I wanted to find out how to diagnose and report to maintainer Jul 20 13:47:07 I'll take it to #openwrt Jul 20 16:23:41 Chaps, is there a shell library routine defined already that'll give me the status (enabled/running) of a service? I'd like to check if a particular services is running and if so restart it? Jul 20 16:26:27 kevindb: there's /etc/init.d/ enabled, but I don't know if there is something for started/stoppped Jul 20 16:27:38 note this does not know about any extra enabled uci vars, so goes only by wether * enable was called for the init-script Jul 20 16:28:16 KanjiMonster: Thanks, that's good for seeing if it's supposed to be running, but not whether it actually is. To be fair for my case that might be good enough. Jul 20 16:33:30 kevindb: ubus call service list will give you all procd managed services that are running (with probably more info than you care for ;) Jul 20 16:36:30 KanjiMonster: Yikes, yes it does. Hmmm. Also /lib/functions/service.sh might help. Thinking. Playing Jul 20 16:42:12 luka: https://github.com/openwrt/packages/pull/1596 Jul 20 16:46:15 DonkeyHotei: you should be able to Jul 20 16:46:29 "assign" the pull request to luka Jul 20 16:47:05 i thought only admins could do that Jul 20 16:48:24 hm, don't know, I don't have a non-admit account Jul 20 16:48:48 where is the assign button? Jul 20 16:49:48 on the right hand side I see "Labels" "Milestones" "Assignee" (each with a small gear right next to them) Jul 20 16:50:48 yeah, i don't have the gear Jul 20 16:51:03 ah, okay Jul 20 16:51:37 @mentions might work as well in a comment without admin rights Jul 20 16:53:38 https://github.com/openwrt/packages/pull/1520 is also relevant. it's still open despite identical commits having already been pushed without pulling Jul 20 21:32:23 DonkeyHotei: i commented on your pull Jul 20 22:24:13 luka: i commented on your comments Jul 20 23:17:38 luka: corrections made Jul 21 00:40:23 build #63 of gemini is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/gemini/builds/63 Jul 21 00:44:20 build #65 of cobalt is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/cobalt/builds/65 Jul 21 00:48:22 build #65 of cns21xx is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/cns21xx/builds/65 Jul 21 00:52:17 build #65 of orion is complete: Failure [failed shell_10] Build details are at http://buildbot.openwrt.org:8010/builders/orion/builds/65 Jul 21 01:29:19 someone i'm trying to help wants to change FORCE_MAX_ZONEORDER in the kernel, but not having much luck. he tweaks it in, e.g. target/linux/sunxi/config-$foo, but when the kernel builds, he looks in build_dir and it still has the original value. **** ENDING LOGGING AT Tue Jul 21 02:59:58 2015