**** BEGIN LOGGING AT Fri Nov 02 02:59:58 2018 Nov 02 03:05:32 yes **** BEGIN LOGGING AT Fri Nov 02 04:09:02 2018 Nov 02 06:48:19 morning Nov 02 06:48:39 morning! Nov 02 06:53:51 ldir: i normally ask the person first then a few devs of list and if i feel that others think the same i write an email Nov 02 06:53:57 morning Nov 02 06:54:10 managed to get mikrotik booting with ath79 last night Nov 02 07:06:19 yay Nov 02 07:19:48 weee Nov 02 07:56:42 are mt7628 and mt7688 register compatible? Nov 02 08:19:47 russell--: yes Nov 02 08:19:55 7628 has pci/e Nov 02 08:20:02 and i think it had more LPC foo Nov 02 08:20:27 and there is a 7688 with on chip flash/ram for ecos Nov 02 08:25:18 blogic: i have the datasheet for the 7688, but not the 7628, so trying to make do Nov 02 08:26:40 yeah, the 7628 has three variants afaict: A with pcie, K with the onboard ram, N without pcie. my board has an N Nov 02 08:27:32 ever so slowly starting to make sense... Nov 02 08:28:56 russell--: if i recall correctly, anything in the 7688 pdf will be valid fro 7628 Nov 02 08:29:02 but 7628 might have extra stuff Nov 02 08:32:47 blogic: roger Nov 02 08:53:22 blogic: do you still have backports-v4.18.5.tar.xz somewhere? Nov 02 08:54:32 stintel: yes Nov 02 08:55:57 blogic: can I get it somewhere, I don't have it anymore and trying to bisect RPi0W wifi breakage Nov 02 08:56:09 already uploading Nov 02 08:56:18 cool Nov 02 08:58:05 stintel: 138.201.250.213:~/backports-v4.18.5.tar.xz Nov 02 09:10:22 blogic: thanks Nov 02 09:11:26 blogic: +1 on your feeling about Hannu Nov 02 09:19:39 hah, brcmfmac doesn't even compile on d9eefa7a7031543571d434693c7f984dfbdc990d :/ Nov 02 09:30:54 stintel: that WARNING seems clear Nov 02 09:31:03 stintel: i suggest looking at sources instead of bisecting Nov 02 09:31:08 it should be simple Nov 02 09:33:13 rmilecki: well it broke with the mac80211 bump to 4.18 Nov 02 09:34:29 master at e9d92bf1e1af71ff19e4cdc753de3f65963c58a5 has working wifi Nov 02 09:35:30 stintel: i suggest analyzing code after the dump Nov 02 09:35:35 look at cfg80211 code and brcmfmac code Nov 02 09:35:47 it's really just a WARNING with a simple check Nov 02 09:37:42 stintel: understand the reason, then use "git log" to see changes, find the commit that has chaned the behavior Nov 02 09:57:53 mkresin: what board is this? https://bugs.openwrt.org/index.php?do=details&task_id=1515 ... it looks like the one i'm working on (from Skylab) Nov 02 10:22:20 russell--: MT7628 Eval board? No idea who manufactures the board/under which brand Nov 02 10:24:22 mkresin: okay, thanks. i'm working on a more specific dts for the board i have. Nov 02 12:18:23 stintel:pushed a commit fixing the missing CVEs in the Curl bump to 7.62.0; completely forgot about that initially Nov 02 12:20:51 jow blogic:what is our stance on partial backported patches ? Nov 02 12:21:14 I'm asking this in the context of PR https://github.com/openwrt/openwrt/pull/915 Nov 02 12:21:44 dedeckeh: ah, you probably missed the short discussion about that afterwards Nov 02 12:21:46 At least one of the patches is a partial backport of an upstream commit Nov 02 12:22:40 dedeckeh: if the package has a CPE ID we maybe don't need to add it anymore Nov 02 12:22:55 stintel: ah ok now I see it Nov 02 12:28:41 but explicitly mentioning it in the Changelog anyway probably doesn't harm Nov 02 12:31:00 indeed this is more clear and doesn't harm Nov 02 12:46:40 Hi! Nov 02 12:52:03 any mikrotik gurus around ? Nov 02 12:52:15 i have a 450g here. mach file says it has a microsd Nov 02 12:52:20 but the pcb does not have that Nov 02 12:57:49 blogic: I'll have a look in the stock area here. Our sister company Citymesh has a ton of Mikrotik devices Nov 02 13:01:20 blogic: There are 2 version of the board. RB450 and RB450G. only the G flavour has an actual slot. It seems the board file is shared for both. Nov 02 13:03:07 blogic: https://i.mt.lv/cdn/rb_files/RB450-both.pdf Nov 02 13:05:19 this is weird... apparently this entry got lost in the great wiki move: https://oldwiki.archive.openwrt.org/toh/hwdata/t-comtelekom/t-comtelekom_speedportw724vtypb Nov 02 13:06:18 xback: i have a G lable box and it has no slot :-) Nov 02 13:07:05 same as the 493g Nov 02 13:07:08 strange Nov 02 13:07:10 blogic Nov 02 13:07:18 the 450G should also have 256MB ram Nov 02 13:07:28 while the non-g is stuck at 46 Nov 02 13:07:30 64* Nov 02 13:07:45 how much does your board have? Nov 02 13:07:47 correct it has 256 Nov 02 13:08:21 smells like another brainfart from Mikrotik in this case .. changing spec along the lifecycle .. Nov 02 13:09:26 * russell-- has a 493g Nov 02 13:09:43 russell--: got one in front of me Nov 02 13:09:53 the 450g boots, testing sysupgrade now and will then move to 493 Nov 02 13:10:08 493g also has microsd in the mach file, yet not on my pcb Nov 02 13:10:09 * russell-- recalls that they didn't provide power on the usb port Nov 02 13:10:29 and none of our mach file make the piezo work Nov 02 13:10:30 * KanjiMonster has a 493g as well, and his has an microsd card slot on the underside of the board Nov 02 13:12:17 * russell-- just spent a couple hours poking at a wink hub Nov 02 13:13:00 ($20 on ebay) Nov 02 13:13:15 KanjiMonster: ok Nov 02 13:13:18 blogic: https://oldwiki.archive.openwrt.org/_detail/media/mikrotik/rb493_pcb-bottom.jpg?id=toh%3Amikrotik%3Arb493g - that matches mine Nov 02 13:14:16 blogic: no 450G in stock here. only 750GL's and higher Nov 02 13:14:44 KanjiMonster: thanks, the 450g will be the same then Nov 02 13:19:36 blogic: confirmed mine have microSD as well Nov 02 13:20:16 rmilecki: think I found the problem. if the modparm roamoff=1, WIPHY_FLAG_SUPPORTS_FW_ROAM is disabled, and we're carrying a patch that sets roamoff=1 package/kernel/mac80211/patches/brcm/864-brcmfmac-do-not-use-internal-roaming-engine-by-default.patch Nov 02 13:22:37 blogic: 450g should look like that https://oldwiki.archive.openwrt.org/_detail/media/mikrotik/rb450g.jpg?id=toh%3Amikrotik%3Arb450g Nov 02 13:24:29 KanjiMonster: ok Nov 02 13:24:38 so now i can flash an image but not boot it Nov 02 13:24:47 just hangs at jumping to kernel Nov 02 13:42:20 rmilecki: no idea how to fix it though. if I understand it correctly, https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7f9a3e150ec7d3596386449c15aefb59904a1266 simply breaks wiphy_register for devices with WIPHY_FLAG_SUPPORTS_FW_ROAM disabled Nov 02 13:42:54 are there any plans to release a OpenWrt 18.06.2 ? Nov 02 13:45:43 jow: ping Nov 02 13:47:39 anyone else has issues pushing to the staging trees ? Nov 02 13:58:42 xback: what kind of issues do you see? Nov 02 14:00:10 jow: i'll mail it now Nov 02 14:00:21 thanks Nov 02 14:00:48 jow: done Nov 02 14:01:23 xback: I have a suspicion, give me five minutes please Nov 02 14:02:11 jow: I'll even give you 6 :) thank you Nov 02 14:04:25 stintel: sent report to the linux-wireless@ Nov 02 14:04:28 *send Nov 02 14:54:47 I keep seeing "zcat: write error: Broken pipe\nzcat: write: Broken pipe" during sysupgrade Nov 02 15:10:08 xback: took a lot longer but should be fixed now Nov 02 15:10:29 xback: if you still encounter issues, please write ted hess Nov 02 15:10:37 will be offline for the next 48 hours or so Nov 02 15:13:27 jow: ill try it now Nov 02 15:14:20 jow: I can confirm it's fixed. thanks for your swift intervention Nov 02 16:02:05 * ldir is having inexperience with compiler trouble Nov 02 16:04:03 I have a package that is barfing with warning: implicit declaration of function warnings - now I can find/or add the relevant header file that defines said missing functions but I still get the warning. How can I get the toolchain to prove to me that it is finding the header file and including it? Nov 02 16:08:29 if anyone wishes to follow along https://pastebin.com/xLFPf1eR :-) Nov 02 16:10:28 ldir: CONFIG_OPENSSL_WITH_EC=y ? Nov 02 16:10:37 ldir: maybe the defines are ifdef'd out because some feature isn't enabled in the library providing these? Nov 02 16:10:49 oh wait, the other way around Nov 02 16:11:09 it's trying DSA which is disabled Nov 02 16:11:35 when running wifi analyzer on my android... I some times see the wifi arcs disappear and the come right back, however in openwrt there's nothing logged about the events Nov 02 16:11:40 or seems to be Nov 02 16:11:41 but if I enable deprecated APIs then ldns builds fine. Nov 02 16:12:05 how would I go about troubleshooting this in openwrt? Nov 02 16:13:03 and I'm naively thinking 'building with old apis is a bad idea, let's try to fix that' Nov 02 16:13:19 ldir: The compiler throws an error if it can't find the #include header file, so it is going to be something else. Nov 02 16:13:32 RSA keys will not work with DSA ciphers... Nov 02 16:13:50 you'd need to use an EC+RSA Nov 02 16:14:37 or create an ECDSA private key for your cert Nov 02 16:14:57 ldir: --enable-ecdsa on ldns ? Nov 02 16:14:57 ldir: If you want to see if the compiler is considering a line of code, you can add a '#error EYECATCHER MESSAGE' line directly before/after that line Nov 02 16:15:32 ldir: ah Nov 02 16:15:35 ldir: --disable-dsa on ldns ;) Nov 02 16:16:14 ./dnssec.c:358:6: warning: implicit declaration of function 'BN_bin2bn'; is the first error Nov 02 16:17:16 which is big numbers - as found in bn.h. and BN_bin2bn doesn't appear to be wrapped in IFDEF hackery. Nov 02 16:17:56 jstefek: that's useful to know Nov 02 16:18:41 ldir: could it be that there is another bn.h somewhere in the include path, and the compiler is finding the wrong one? Nov 02 16:23:00 bwah hahaha - oh dear, what a twit - editing source file for one target....and *building* for another. Let me start again but this time edit the correct source Nov 02 16:27:01 ok that helps - I've got rid of the BN_bin2bn warnings. Nov 02 16:29:02 I did that just this morning myself, happens all the time, especially when switching between things a bunch Nov 02 16:48:13 blogic: http://blackhole.sk/~kabel/tmp/moxis.jpg Nov 02 16:55:00 greearb: ping Nov 02 17:13:12 huaracheguarache, here Nov 02 17:28:33 hi! I saw the email about enabling DBGLOG, and I want to ask you about some things. What flag do you want me to set? Does /sys/kernel/debug/ieee80211/phy0/ath10k/debug_level persist across reboots? Nov 02 17:29:19 it would not magically persist across reboots, could set it with modprobe likely Nov 02 17:29:59 This is suggested: echo 0xc0000032 > /sys/kernel/debug/ieee80211/phy0/ath10k/debug_level Nov 02 17:30:14 your phy might be named differently Nov 02 17:35:52 ok, thanks. Nov 02 17:56:04 greearb_: ath10k-ct 4.19 is still working well. I also tested performance later yesterday and found it to be normal. Nov 02 17:57:14 mamarley, good news! If there are any features there that someone needs, fine with me if someone updates openwrt to use it, but it will likely be a few more weeks before we can do any serious testing or bugfixing on 4.19 in case issues are seen. Nov 02 18:08:57 I noticed on my r7800 that the port labeled '4' on the switch is actually vlan 1. So, I tried changing the /etc/config/network so that vlan != port, like this: Nov 02 18:09:07 config switch_vlan Nov 02 18:09:08 option device 'switch0' Nov 02 18:09:08 option vlan '4' Nov 02 18:09:08 option ports '1 6t' Nov 02 18:09:23 And now, doesn't appear to work at all. Is that expected? Nov 02 18:10:27 did you also update the eth0.1 reference to eth0.4? Nov 02 18:10:57 why are you comparing port numbers to vlan numbers anyway? Nov 02 18:10:58 eth0.5 is the WAN port Nov 02 18:11:28 I figure it will be easier for users to know what is what if the VID matches the labels on the system. Nov 02 18:12:29 I did not change the interface setting, they remain like this: config interface 'lan4' Nov 02 18:12:29 option ifname 'eth1.4' Nov 02 18:12:29 option proto 'static' Nov 02 18:12:47 let me check it is not some other issue....it sort of half way seems to be working Nov 02 18:12:56 please show me the full config Nov 02 18:13:08 just the network file, or more? Nov 02 18:13:22 just the network file Nov 02 18:14:08 sorry, it appears it was some other glitch, I reset the wifi and it is passing traffic now. Nov 02 18:15:03 for reference: https://pastebin.com/tyRyEWbA Nov 02 20:36:52 * russell-- is looking in target/linux/mxs for devicetree files, not finding Nov 02 20:37:27 target/linux/mxs/Makefile:KERNEL_PATCHVER:=4.14 Nov 02 20:37:54 and an arm soc, expected devicetree to be obligatory Nov 02 20:41:46 so I'd like to investigate why poweroff fails for Netgear r7800. Any suggestions on where to start looking? Nov 02 20:44:32 russell--: AFAIK some ar71xx were also migrated to 4.14 despite the lack of DTS, but the idea was to migrate as many of them as possible to DTS Nov 02 20:44:38 I'd start by testing 17.01.0 and then 18.06.0, in the end the easiest approach (assuming that at least 17.01.0 worked in that regard) would be bisecting it. I'm very confident that poweroff was still working after the switch to 4.14 - and probably broke semi-recently, but I don't poweroff my router often enough to have noticed a regression in that regard Nov 02 20:45:25 ar71xx is non-DTS, ath79 is DTS (but the latter is not ready yet) Nov 02 20:45:54 the arm-to-DTS association goes back farther than for mips Nov 02 20:46:00 AFAICS some ath79 targets are working pretty well, some are still in progress Nov 02 20:46:39 ath79 status: https://forum.openwrt.org/t/ath79-target-status/18614/9 Nov 02 20:48:19 oh, maybe the dts are upstream already Nov 02 20:50:47 * russell-- finds some in the linux git Nov 02 20:51:10 (confusion level ebbs slightly) Nov 02 20:57:09 isn't devicetree/OF actualyl _from_ mips? Nov 02 21:14:06 when bisecting, is it expected you have to run 'make clean' each time? Nov 02 21:15:40 karlp: originally sparc and then extended to powerpc, from where it went to arm and mips (and probably more) Nov 02 21:16:06 pkgadd: thanks, must have mixed up ppc in my memory :) Nov 02 21:16:20 each time, no - for the larger steps (changing toolchain versions), yes Nov 02 21:18:32 pkgadd, things like this expected w/out a clean? Nov 02 21:18:35 Applying ./patches/704-rt2x00-use-different-txstatus-timeouts-when-flushing.patch using plaintext: Nov 02 21:18:36 patching file drivers/net/wireless/ralink/rt2x00/rt2800lib.c Nov 02 21:18:36 Hunk #1 FAILED at 1140. Nov 02 21:19:23 I think so, yes Nov 02 21:19:37 that 'should' work... Nov 02 21:29:05 make packages/kernel/mac80211/clean might already be enough for that particular case though Nov 02 21:36:35 it fails even with a full make clean Nov 02 21:37:11 this commit, and one a bit later as well: bc24f0ad2b3cc7ca7b74ecef8051c1bbff9d4277 Nov 02 22:13:42 nbd: FYI - i just ran a master build and that seems a lot more stable (90' so far - had two reboots within the first hour i came home tonight) Nov 02 22:16:44 bleh, build fails even after 'make distclean' Nov 02 22:21:21 kab-el: yes, still pending i simply forgot, sorry Nov 02 22:58:04 ok, clean build of freshly cloned repo, commit HEAD~1000 worked. And poweroff fails to work. Nov 02 23:17:57 when i'm bisecting, it is more work if you aren't *sure* as you try to bracket the problem, so i do a make dirclean between test build Nov 02 23:19:24 if you get a test wrong, you end up searching the wrong half of the search space Nov 02 23:23:50 If you have enough storage and compute power, you can compile several builds in parallel. That's how I found a problem on the zynq target Nov 02 23:25:31 It has the added benefit of having working and non-working builds to test changes Nov 03 00:38:06 ldir: the bn issue is due to missing #include . Removing deprecated APIs does not implicitly include header files. **** ENDING LOGGING AT Sat Nov 03 02:59:58 2018