**** BEGIN LOGGING AT Fri Apr 23 02:59:56 2021 Apr 23 05:08:37 Hi Guys, I am running 21.02 snapshots on a WRT1900ACS and have a wifi client on 5GhZ spuriously loosing connection. I know the iwlwifi driver is a piece of hot garbage. But would like to help figuring out what's going on. I noticed a higher softRQ on the device but don't know how to isolate what the issue is. Apr 23 05:09:04 Anyone here that could help me? Apr 23 05:34:09 the problem is that absolutely no one is maintaining mwlwifi anymore, Marvell sold out their wireless division to NXP - and the first thing NXP did, was EOL'ing these chipsets. the (almost-) fullmac firmware is closed, and the driver has always been problematic and is known buggy Apr 23 05:44:52 Yeah I know it's hot garbage :D. But I didnt experience this with the 19.07.7 so this time it might be something different :D Apr 23 05:45:19 Because the link stays (WiFi is connected) Apr 23 05:45:28 but not IP traffic goes through. Apr 23 06:04:20 wow ouch! whats happened to the trunkn tplink archer v4 builds... no 5ghz now, out of space on the images Apr 23 07:27:57 hi all Apr 23 07:57:41 i see that 21.02 comes with (wolf)ssl by default now Apr 23 07:57:47 * check_data_file_clashes: Package libustream-openssl20201210 wants to install file /lib/libustream-ssl.so Apr 23 07:57:49 But that file is already provided by package * libustream-wolfssl20201210 Apr 23 07:58:03 if I want to use openssl - should I just uninstall wolfssl first? Apr 23 08:02:43 sure Apr 23 08:03:10 remove libustream-wolfssl Apr 23 08:06:08 rmilecki: what exactly is requiring the OpenSSL version of libustream? Normally, I can use other OpenSSL enabled SW without messing with libustream... Apr 23 08:06:47 .. but they just want libopenssl Apr 23 08:10:28 guidosarducci: oh, no, i just wanted openssl in general Apr 23 08:10:39 libustream just as a result of that Apr 23 08:13:07 then drop libustream-openssl as u wish Apr 23 08:18:33 rmilecki: sure, if you just want to replace all things wolfssl with openssl that would be needed. Just be careful not to break anything SSL related that's needed to swap things on a running system (i.e. opkg). Better to do this with an image, or at least use opkg to first download *all* the packages you expect to need, *then* upgrade. Apr 23 08:20:01 guidosarducci: thanks Apr 23 09:03:56 rmilecki: I've just forced build of bcm4908 for 21.02-rc1 as it was missing https://buildbot.openwrt.org/openwrt-21.02/images/#/builders/71/builds/1 Apr 23 09:04:27 perhaps you would like to give it a try once it completes Apr 23 09:06:26 bcm4908 is already present as new target in https://openwrt.org/releases/21.02/notes-21.02.0-rc1#new_hardware_targets Apr 23 09:06:27 ynezz: sure Apr 23 09:06:32 thanks Apr 23 09:41:17 Testing out 21.02 rc1 and I'm getting substantially less internet speed. I have comcast 1 gig down (1200 mbps) and tested with 19.07.7 (908 mbps down), original firmware (915 mbps), now tested with 21.02 and it lowered all the way down to 770 mbps Apr 23 09:41:25 ath79/wzr600dhp still panic'ing on boot with no /dev/root Apr 23 09:41:56 on 5.10 Apr 23 09:42:27 im using mvebu Apr 23 09:43:27 [ 0.738524] spi-nor spi0.0: mx25l12805d (16384 Kbytes) Apr 23 09:43:27 [ 0.748118] spi-nor spi0.1: unrecognized JEDEC id bytes: 20 18 c2 20 18 c2 Apr 23 09:43:30 [ 0.755089] spi-nor: probe of spi0.1 failed with error -2 Apr 23 09:48:12 I wonder if anything touched the soft offload code recently Apr 23 09:51:18 russell--: hmm,i have a device with multiple spi-nor chips running, however it is ar934x Apr 23 09:59:17 * russell-- thinking it might be the david bauer spi-nor commit, try a build with that reverted Apr 23 10:01:50 * russell-- has to go into the basement to recover, because tftp uses a LAN port instead of wan (eth1), have to swap ethernet cables around Apr 23 10:09:05 I'm trying to get apcups work on luci-statistics and im getting this error Apr 23 10:09:08 Uploaded file: https://uploads.kiwiirc.com/files/65c6e8c899d41f042e1e80c8454063d7/pasted.txt Apr 23 10:20:42 reverting 28623cab32 didn't help Apr 23 10:21:06 was working as of r16500-d0adc4865c Apr 23 10:29:16 Namidairo: The obsolete routing cache, maybe? https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=17576b1b2aeacf0a23bb9a911d675bd324754745 Apr 23 10:29:52 think it might be a bit futher back than that Apr 23 10:30:38 kept getting 20s stalls on random http requests until I turned soft offload off Apr 23 10:32:02 Namidairo: what openwrt version are you running, and which kernel are you using? Apr 23 10:33:12 OpenWrt SNAPSHOT r16582-3326b5e75c, 5.4.113 Apr 23 10:33:32 so one commit before that it looks Apr 23 10:36:06 maybe it'll work better after switching to linux 5.10 Apr 23 10:36:18 lots of changes to flow offload in there Apr 23 10:37:05 Sorry, I have been offline for a bit Apr 23 10:37:17 did anyone give me any advice? Apr 23 10:47:04 r16542 is bad Apr 23 10:47:14 trying r16521 Apr 23 10:48:40 OpenWrt 21.02.0-rc1 r16046-59980f7aaf / LuCI openwrt-21.02 branch git-21.106.55967-06dd6b5 Apr 23 10:48:44 running this Apr 23 10:48:50 mvebu Apr 23 10:49:04 Speed test is so slow vs 19.07 Apr 23 10:49:17 or the original firmware Apr 23 10:51:35 thagabe: file a bug report, as I also see this is broken. It's worked the last 4 years for me, since first submitting the feature. Very annoying. Apr 23 10:51:43 nbd: Speaking of flow offload, the CONFIG_NF_FLOW_TABLE_HW symbol is dropped in 5.10, isn't it? Apr 23 10:52:18 Or is it still used? Apr 23 10:55:24 it's dropped Apr 23 10:56:26 no separate module for hw offload anymore Apr 23 11:01:41 On firstboot openwrt copies files to /overlay somehow - not sure exactly how that works. But in the process if you are logged in to SSH, once it finishes you get the error, "getcwd openwrt unknown no such file or directory" and need to logout and login again Apr 23 11:02:09 Is there anyway to know when the copying is done so I make sure I do not try to run any kind of scripts before the overlay copying is done? Apr 23 11:06:29 r16521 is okay, trying r16532 Apr 23 11:08:44 russell--: 2683eeb627aabc7c01fb46fc5e825ece626aeaa5 <-- This probably broke it Apr 23 11:09:18 barhom: on first boot overlay is in ram, and then jffs2 is being prepared in background, and then the files are copied and overlay switches to jffs2. Apr 23 11:10:01 Yes, and I assume that I get this error, "getcwd openwrt unknown no such file or directory" once all of that is done right? Apr 23 11:10:34 russell--: This flag was needed to trigger the CS function of spi-ath79 if the GPIOs provided are valid Apr 23 11:10:36 That happens when the overlay switches to jffs2 Apr 23 11:12:26 nbd: Thanks for the info, I made a kernel_oldconfig and noticed it had gone, I just wanted to make sure. Apr 23 11:14:09 barhom: part of this happens in libfstools/overlay.c switch2jffs() function Apr 23 11:15:31 barhom: you shouldn't need to log out, just "cd" Apr 23 11:16:16 karlp, I think thats true but until I do logout the prompt says "(unknown)" Apr 23 11:17:06 Anyway, the question was more if there is I way to find out when the process copying the jffs is done. I do not want to start cron and other tasks until overlay is finished Apr 23 11:17:08 blocktrron: yeah, that looks likely Apr 23 11:17:48 barhom: afaict it happens after "mount_root done" returns. Apr 23 11:18:24 But I can't tell how to track that. Apr 23 11:19:13 barhom: probably just moving it past 95 is enough? Apr 23 11:19:52 (so taht it runs after S95done) Apr 23 11:19:57 trying master HEAD with 2683eeb6 reverted Apr 23 11:19:58 logread probably Apr 23 11:20:44 logread and follow, pipe into shell with read, check if line corresponds to regex of string describing copying is done, then return from the function Apr 23 11:20:57 PaulFertser, I'll try that - thanks. Cron is S50 right now, Ill try moving it further down Apr 23 11:21:39 procd or sysvinit, barhom? Apr 23 11:21:40 Hmm, yeah, this is steel needed Apr 23 11:21:55 russell--: can you revert it and report if it works again? Apr 23 11:22:10 Thermi, I assume Im using procd - whatever is default in 21.02 Apr 23 11:22:11 Thermi: oh that would be such a dirty trick Apr 23 11:22:15 Most likely it will, it's still needed, as spi-ath79 sets use_gpio_descriptors Apr 23 11:22:44 barhom: I've traced the calls in the source code, it really looks like you just need to run after "done". Apr 23 11:23:41 * PaulFertser read "steel needed" as "steel indeed" and so was wondering when a discussion about materials started :) Apr 23 11:24:17 PaulFertser, in what file? Apr 23 11:24:38 barhom: libfstools/{overlay.c,mount_root} , base-files/files/etc/init.d/done Apr 23 11:26:35 Seems like done is S95, since cron is S50 I should be fine making cron S95+. I cant see that openwrt relies on cron for any other things than my own custom files Apr 23 11:26:52 Thanks for the help Apr 23 11:32:56 blocktrron: yep, that was it, reverting fixes it Apr 23 11:33:20 barhom: procd does not track start completion or service completion. Therefore the problem exists that copying and cron still run into each other Apr 23 11:38:19 What about a way to check if the overlay is mounted? Apr 23 11:38:30 using 'df' for example. What would I look for? Apr 23 11:39:28 barhom: are you sure what I suggest doesn't really work? Apr 23 11:39:56 Putting it after done? Probably it works but that means I need to remove /etc/init.d/cron default file Apr 23 11:40:09 I'd like to avoid removing as much defaults as possible Apr 23 11:40:53 barhom: not removing, just making a symlink with not default name. Apr 23 13:11:21 russell--: okay, I'll revert that with another patch i plan on upstreaming Apr 23 13:11:34 Can i ping you, so you can verify it still works then? Apr 23 18:39:22 Hi openwrt-devs..  I've been working lately to get my comfast ew74 (looks like the 72, but actually rampis, not ath79).  Now that everything is working well enough I wanted to see how I can get some help/feedback on finalizing and cleaning things up enough to contribute the changes back to the community.   I think the two outstanding items Apr 23 18:39:23 would be, making the leds do something more meaningful and a more pretty or abstracted DTS. Apr 23 18:40:46 eric21: usually you can find contributions for similar hardware by just doing git log for the target/linux/ramips directory Apr 23 18:41:03 eric21: for LEDs to be meaningful you assign them appropriate aliases. Apr 23 18:45:35 PaulFertser I do have some leds working the rssi part may not be right, I will work on that if I get time.  I'd really like to contribute what I have since the rest seems to work well,  My dts is pretty good, i leverage the includes as much as possible after starting with on eI extracted from the factory firmware.  I'm more wondering if the Apr 23 18:45:35 community would help polish it? Do i submit a pull as/is? or just leave it only in my world and continue to build images as needed?  I've never written anything to contribute...  I have it "commited" to my local system correctly I think, so I could generate a patch or whatever with a little help. Apr 23 18:48:02 eric21: I think if this hardware is available to the general public it would certainly be beneficial to the community to have it supported in upstream OpenWrt, I can't really understand why you're hesitating. Please ask a specific questions and I hope someone will find a specific answer for you. Apr 23 18:48:42 eric21: if you need a wiki account to document the board please send me your desired wiki nick and e-mail and I'll add you. Apr 23 18:49:13 what do I do next? email a patch or what?  I'm asking now that it builds and works and  I have git add/committed locally, what should I do. Apr 23 18:52:12 eric21: you can either email it to the devel mailing list (better use git send-email for that) or add a "github pull request". Apr 23 18:52:28 eric21: both ways are described on the wiki, I suggest you reread the contribution guidelines there. Apr 23 18:53:23 I will I'll work on getting get send-email to work, I was just trying to understand if it was time to.  Thanks Paul. Apr 23 18:55:39 eric21: git send-email with all default settings (no configuration at all) is actually working fine for the purpose. Apr 23 18:56:27 OK, I'll keep that in mind. Apr 23 18:56:41 (assuming you have a functioning MTA of course) Apr 23 18:59:29 I can make my MTA work, I was practicing sending to my self but I think I need to generate a patch or something, it returns this: Apr 23 19:00:05 fatal: ambiguous argument '001-description.patch': unknown revision or path not in the working tree. Apr 23 19:00:05 Use '--' to separate paths from revisions, like this: Apr 23 19:00:06 'git [...] -- [...]' Apr 23 19:00:06 format-patch -o /tmp/MqJApZQtmf 001-description.patch: command returned error: 128 Apr 23 19:00:28 I appreciate the help BTW.  this is what it takes to motivate me to get over the hump... Apr 23 19:02:30 eric21: if you just need to send the last commit you do "git format-patch HEAD^" Apr 23 19:04:55 I prob have 3 or 4 prior ones of the past week, i'm sure I can find them "git log", I think I merged or switched to master at somepoint.   should I just generate a patchfile for each of my commits? Apr 23 19:06:07 git format-patch origin/master Apr 23 19:06:08 0001-added-support-for-comfast-jw-ew74.patch Apr 23 19:06:08 0002-corrected-packages.patch Apr 23 19:06:09 0003-cleaned-up-comfast-jw-ew74-file-some-more.patch Apr 23 19:06:09 0004-Removed-non-existant-wan-interface.-Fixed-eeprom-and.patch Apr 23 19:06:57 oh, I bet I should remove luci as an included package too shouldn't I? Apr 23 19:07:32 eric21: git format-patch takes a commit range. Implicitely from the current commit. Apr 23 19:07:45 eric21: It creates patch files. Apr 23 19:08:18 And yes, a patch file for every commit Apr 23 19:08:25 so if it generates patch files sufficient for what I want, i just use the same options w/ git-email and off it goes? Apr 23 19:09:04 git send-email? Apr 23 19:09:45 eric21: check man git send-email, scroll down to the bottom. There's an example. Apr 23 19:09:54 great, so I'll clean up a little more, commit one last time, and send it off.  Does that sound right? Apr 23 19:10:07 eric21: you need to squash all your patches as upstream is interested in the final work, not on your development history. Apr 23 19:10:15 eric21: "git rebase -i" is appropriate for squashing. Apr 23 19:10:34 (If you want a nice gui, give sublime merge a try) Apr 23 19:10:46 I think I got it.  will do Apr 23 19:10:54 It visualizes the commit tree quite nicely and helps you understand exactly what happens. Apr 23 19:11:01 eric21: usually adding support for a new board is just a single patch. Apr 23 19:11:09 squashing is what I needed to know Apr 23 21:00:42 Build [#69](https://buildbot.openwrt.org/master/images/#builders/4/builds/69) of `x86/generic` completed successfully. Apr 23 22:36:51 blocktrron: yes, happy to try your patch Apr 23 22:37:43 (fwiw, just figured out the nick/name mapping :) Apr 24 02:58:55 If I want to compile a driver source using the build-system to do it, what's the best way? **** ENDING LOGGING AT Sat Apr 24 03:00:29 2021