**** BEGIN LOGGING AT Thu Sep 26 02:59:57 2019 Sep 26 03:22:26 can somebody help me to get rid of the buildpath in kernel modules? https://rebuild.aparcar.org/SNAPSHOT/ath79/generic/packages/kmod-gpio-button-hotplug_4.19.75-3_mips_24kc.ipk.html#tmpkpg--wcu-content---.-data.tar.gz---data.tar---.-lib-modules--.--.---gpio-button-hotplug.ko---readelf---wide---decompress---string-dump-.rodata.str-.---- Sep 26 03:22:56 the link goes to a diffoscope output showing that the .ko file stores the build path Sep 26 04:08:29 mangix: i don't Sep 26 04:30:53 hrm https://lists.openwrt.org/mailman/listinfo/openwrt-devel is not loading Sep 26 04:31:25 rmilecki: https://gist.github.com/neheb/78c3684cc0c8f6fe35d7eaba9e17387b Sep 26 04:31:41 that's with GCC9 Sep 26 04:32:20 uh, some ugly alignment business Sep 26 04:32:44 i'm not experiented with suchj Sep 26 04:33:48 currently i disable -Werror in CMake to get it to build. Sep 26 08:05:50 aparcar[m]: did you test that xtables lock thimg with user supplied iptabled script? Sep 26 08:17:52 jow: I fired fw3 reload and user iptables commands. Normally it did break but with the patch it would just wait a second Sep 26 08:18:09 jow: does it cause any problem?! Sep 26 08:50:15 aparcar[m]: not sure, some chinese guys is stalking me about some xtables lock thing issue after the latest fw3 bump Sep 26 08:50:35 but I just did a quick test locally here and see no such issues Sep 26 08:50:48 so its probably something hacky he does, or stuff involving hotplug or whatever Sep 26 08:54:22 jow: backported the flow offload changes Sep 26 08:54:30 nbd: thank you, much appreciated Sep 26 08:54:53 btw, do you have any idea how I can programmatically firgure out if hardware offloading is available? SOme module or sysfs path to check? Sep 26 08:55:34 no idea Sep 26 08:55:38 ok Sep 26 08:56:13 i don't think we have anything for that yet Sep 26 09:26:04 jow: could you please give the json image info patch a try and review/merge it? worst thing would be to revert it tomorrow isn't it? Sep 26 09:26:57 > Can I sell openwrt based router Sep 26 09:26:58 Also good mails on openwrt-dev list Sep 26 11:18:24 rmilecki: hi Sep 26 11:18:31 been doing some more debugging Sep 26 11:18:37 ok Sep 26 11:18:38 and the SDK images will report [ 0.079356] Calibrating delay loop... 1594.16 BogoMIPS (lpj=7970816) Sep 26 11:18:41 hey Sep 26 11:18:50 while owrt says 800 bogo mips Sep 26 11:18:54 i was told multiple times that it doesn't matter Sep 26 11:18:59 i'm not sure how true is that Sep 26 11:18:59 oh ok Sep 26 11:19:50 I see that under load the SoC is busy 30% of the time with cache operations Sep 26 11:20:39 yeah, that's quite likely Sep 26 11:21:10 never seen that anywhere Sep 26 11:22:19 I have a feeling that the l310 is not setup correctly Sep 26 11:22:37 it could be, you may ping Hauke about that Sep 26 11:23:32 blogic: you may check that "NAT performance regression caused by vlan GRO support" e-mail thread Sep 26 11:23:39 i posted my perf info there Sep 26 11:23:42 yes Sep 26 11:23:43 see if that matches your results Sep 26 11:23:48 i disabled gro Sep 26 11:23:56 and maybe also "ARM router NAT performance affected by random/unrelated commits" e-mail thread Sep 26 11:26:43 ok Sep 26 11:29:17 blogic: are you maybe woring on bcm53xx hw nat offloading? Sep 26 11:30:26 no Sep 26 11:30:34 well Sep 26 11:30:54 eventually once the unit works i would add support for the af tabes or what they are called Sep 26 11:31:03 but right now i can NAT ~350mbit Sep 26 11:31:13 and 800 if i use sw flow table offloading Sep 26 11:31:31 and in theory this SoC should shift wire speed easily Sep 26 11:31:40 oh, interesting, I think last time sw offloading wasn't not improving anything for me at all Sep 26 11:31:47 i have to get back to that NAT finally Sep 26 11:32:06 I think it is the cache Sep 26 11:32:20 so i disabled the cache driver probe and the NAT drops to 160mit Sep 26 11:32:52 [ 0.263876] L310: cache controller enabled 16 ways, CACHE_ID 0x410000c8, AUX_CTRL 0x7a130000 Sep 26 11:33:07 I wanna compare this between SDK and owrt today Sep 26 11:33:42 let me know about results pls Sep 26 11:33:52 will do Sep 26 11:34:29 rmilecki: sent you a PM Sep 26 11:45:24 i have some ideas on how to tackle GRO performance Sep 26 11:45:57 ok Sep 26 11:46:07 so there's ongoing upstream work that does a simpler form of GRO (mostly for ipsec routing), where skbs are simply chained as a skb list Sep 26 11:46:20 so that when they need to be split again, it doesn't require header mangling, checksum recalc, etc. Sep 26 11:46:46 i think it would be good to use that for TCP as well, but only for routed/bridged traffic Sep 26 11:47:51 that way, packet aggregates can travel through the network stack, which should help a lot with performance Sep 26 11:48:32 the main problem why GRO has such huge overhead on bcm53xx is because for some reason broadcom didn't care to add any checksum offload to the ethernet MAC Sep 26 11:48:57 and with list-GRO we wouldn't need it Sep 26 11:49:54 the subject of the patches i'm talking about on netdev is "[PATCH RFC 0/5] Support fraglist GRO/GSO" Sep 26 12:00:52 nbd: let me see if i can make v3 work ontop of v4.4 Sep 26 12:01:05 *v4.14 Sep 26 14:37:21 Hi all, is there any way to get flash io statistics?? I checked the stats in /proc/iostats and /sys/block/mtdX/* but... Sep 26 14:37:42 they all show no write activity although i am writing to jffs2 and "sync"ing after that Sep 26 14:38:57 also, /sys/class/spi_master/* shows some trnsmission activity, but not all spi transmissions are "write" accesses, right? Sep 26 14:39:26 /sys/block/mtd only shows what's going through mtdblock, and jffs2 I/O is not going through the block layer Sep 26 14:40:09 so, is there any way to get jffs2 write activity? Sep 26 15:50:57 Hi people. Sep 26 15:51:30 My build keeps crapping out at make install packages. Sep 26 15:52:07 I am using Ubuntu onder WSL it is a fresh install after I upgraded my PC to a AMD rizen 5 2600 Sep 26 15:52:11 logs please :) Sep 26 15:53:08 I am shure there is a package I need to install on ubuntu Sep 26 15:53:24 I installed sudo apt install build-essential libncursesw5-dev python unzip Sep 26 15:53:54 I am doint make menuconfig to turn on logging now Sep 26 15:56:31 Tapper: I think magic line is to use "make V=s" in build time Sep 26 15:56:51 thanks Sep 26 15:57:22 (was it v=S or V=s, I can't remember which one was capital) Sep 26 15:57:44 V=s Sep 26 15:58:22 but you Sep 26 15:58:26 I need to pip the log to a text file because I cant coppy and paste the text from the window because of my screen reader Sep 26 15:58:29 but you'd have to use -j1 with V=s Sep 26 15:58:29 brb Sep 26 15:59:15 I use along "make V=s | tee build.log" (and rest of make parameters ofcourse) to also log the whole build process Sep 26 16:08:52 there's a menuconfig option to log things too, which saves them all in sparate little files for each step that can be easier to read sometimes. Sep 26 16:11:21 advanced configuration options -> enable log files during build process Sep 26 16:13:57 to find the faulty package just $grep -r ./logs/ -ie "error" Sep 26 17:17:07 OK people I am back. I had to take my kids to football Sep 26 17:17:32 soca for those in the US Sep 26 17:19:22 popcorn Sep 26 17:19:33 crap, wrong chat Sep 26 17:19:42 that was supposed to be a query to a bot Sep 26 17:21:04 dang... now I miss the info did the bot deliver... Sep 26 17:38:34 OK my log says. Sep 26 17:38:35 find: The relative path 'Files/PuTTY/' is included in the PATH environment variable, which is insecure in combination wi Sep 26 17:38:35 th the -execdir action of find. Please remove that entry from $PATH Sep 26 17:38:35 package/Makefile:65: recipe for target 'package/install' failed Sep 26 17:51:41 Any one know anything about my build? Sep 26 17:52:06 what's going wrong I mien Sep 26 18:33:14 that gro work is interesting. wonder if it means GRO can finally be enabled on ath79 Sep 26 19:18:30 Hi mangix Do you know anything about my build messing up. I posted a pece of my log Sep 26 19:18:53 find: The relative path 'Files/PuTTY/' is included in the PATH environment variable, which is insecure in combination wi Sep 26 19:18:53 Sep 26 19:18:53 th the -execdir action of find. Please remove that entry from $PATH Sep 26 19:25:29 All so in C:\Users\jlanc\AppData\Local\Packages\CanonicalGroupLimited.UbuntuonWindows_79rhkp1fndgsc\LocalState\rootfs\home\tapper\openwrt\logs\package\kernel\linux Sep 26 19:25:44 I have WARNING: kmod-crypto-pcompress is not available in the kernel config - generating empty package Sep 26 21:06:23 tapper, do echo $PATH and see if it has that "Files/PuTTY" relative entry in it, and, presuming it does, remove it. Sep 26 21:06:27 or change it to an absolute path Sep 26 21:07:24 Hi karlp I will try I started my .config from scratch and am trying again now. When it is dun I will look thanks Sep 26 21:08:02 OK it crashed again I will look now Sep 26 21:10:09 I get Sep 26 21:10:11 tapper@DESKTOP-JLSQD45:~/openwrt$ echo $PATH Sep 26 21:10:11 /usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/mnt/c/WINDOWS/system32:/mnt/c/ Sep 26 21:10:11 WINDOWS:/mnt/c/WINDOWS/System32/Wbem:/mnt/c/WINDOWS/System32/WindowsPowerShell/v1.0/:/mnt/c/WINDOWS/System32/OpenSSH/:/m Sep 26 21:10:11 nt/c/Program Files/PuTTY/:/mnt/c/Users/jlanc/AppData/Local/Microsoft/WindowsApps:/snap/bin Sep 26 21:10:31 karlp how I fix that? Sep 26 21:10:38 quote it somehow. Sep 26 21:10:46 it's getting upset about the space in program files. Sep 26 21:11:17 I wonder many man years hgave been lost to that absurd demonstration of "look, we can put spaces in file names" even if it should have never been done in such a critical location Sep 26 21:11:21 Why I am using WSL. Why would that matter? Sep 26 21:11:35 * karlp shrugs Sep 26 21:11:39 lol Sep 26 21:11:46 either remove putty from your path, or work out how to quote it. Sep 26 21:12:10 How would I remove putty from your path, Sep 26 21:12:18 my* Sep 26 21:12:35 putty is installed on windows not in ubuntu Sep 26 21:13:19 I can just uninstall it. Would that work? Sep 26 21:13:22 well, aiui, "ubuntu" on wsl isn't really a whole contained distro. Sep 26 21:13:36 * Tapper nods Sep 26 21:13:41 you still have (as you can see) all the paths to things like program files and windows/system32 and similar Sep 26 21:13:59 I have no idea how to edit paths on current versions of windows sorry. Sep 26 21:14:31 you _might_ be able to do an "export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin" .... Sep 26 21:14:32 I will remove putty and try again. I can install it if i need it again Sep 26 21:14:35 and then run your command Sep 26 21:14:44 and that might do a good enough job for just that session Sep 26 21:15:16 I am sorry karlp I am a noob in linux you would have to tell me what to type Sep 26 21:25:41 you're not as much of a noob as you might think these days :) Sep 26 21:25:59 you're running "make " in your wsl terminal right? Sep 26 21:26:21 you should probably be able to do "export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin" first, Sep 26 21:26:25 and then try your make line again Sep 26 21:26:43 that tosses all the windows stuff out of the path, just leaves what's probably from the wsl ubuntu paths. Sep 26 21:53:34 karlp Thanks for the help mate. I uninstalled putty and the build worked. Sep 26 21:57:50 yw Sep 26 22:32:11 these days MS has ported OpenSSH natively to windows, so putty might not be quite that necessary anymore. but there are better fixes than uninstalling putty ;) Sep 26 22:34:07 uninstalling windows? ;) Sep 26 22:34:49 pkgadd I removed the installer for putty and got a stand alone file that I droped in to program files then sent a shortcut to my desktop Sep 26 22:35:49 I have to go to bed now, but I will rool a build in the morning to see if it still works Sep 26 22:35:58 roll* Sep 26 22:36:19 Sep 26 22:36:33 stintel if it was not for my screen reader I would of bin on linux a long time ago Sep 26 22:37:08 stintel yes? Sep 26 22:38:07 If I ever type a hole sentence with out macking a spelling mastake then ban me from here because it will be a inposter Sep 26 22:38:13 hahaha Sep 26 22:38:38 :D Sep 26 22:38:42 stintel: apt-get remove --purge ^windows* doesn't yield any results... Sep 27 00:13:16 https://pastebin.com/8nn8vCU4 Sep 27 00:13:17 funky Sep 27 01:07:57 aparcar[m]: to become more reproducible builds, we should move to a newer gcc version. I don't know why we're still at 7.4.0 instead of using gcc 8 or 9. Sep 27 01:08:15 aparcar[m]: newer gcc versions have a feature to reduce build path dependencies Sep 27 01:09:13 stintel: can you reproduce it? Sep 27 01:10:42 lynxis: seen it on 2 different machines, on one triggered by uhttpd, on the other by tor Sep 27 01:11:18 probably a bug in cryptodev - either the module or the openssl cryptodev stuff Sep 27 01:12:11 jesus. sourceforge git is cavetime git Sep 27 01:12:28 lynxis: I'd by happy with that, anything I can do for that? Sep 27 01:13:00 lynxis: buildbot slaves too old, no gcc 8 there Sep 27 01:16:44 stintel: i beg to differ Sep 27 01:16:49 stintel: isn't it dockered? Sep 27 01:16:56 lynxis: got pretty close https://paste.debian.net/1102887/ Sep 27 01:17:01 arc* target uses GCC8 Sep 27 01:17:21 many bugs have been fixed that way Sep 27 01:17:46 I guess that's new then, I recall some issues with some gcc or g++ version not being available Sep 27 01:18:02 or are you talking about the host GCC? Sep 27 01:18:12 host GCC on the build bots is GCC6 Sep 27 01:18:51 * mangix had to revert some C++17 stuff as GCC6 does not support Sep 27 01:20:33 mangix: I'm a total noob, can i just somewhere change 7.4 to 9 and sudenly everything is compiled with the new gcc? Sep 27 01:22:44 locally? Sep 27 01:26:04 make menuconfig > advanced > toolchain options > gcc compiler version . I also change the binutils version. Sep 27 01:27:08 interesting.... git pr works even when the main branch is not on github Sep 27 01:27:30 mangix: thanks I'll try that Sep 27 01:28:13 it compiles a new toolchain. hopefully you have the space (I recently ran out. 500GB is not enough for me I guess). Sep 27 01:29:42 right, confused with host gcc version Sep 27 01:29:45 * stintel should go to bed Sep 27 02:04:00 mangix: thanks I'll give it a try **** ENDING LOGGING AT Fri Sep 27 02:59:57 2019