**** BEGIN LOGGING AT Sat Oct 24 10:59:57 2020 Oct 24 11:01:49 blogic: sorry, i started collecting info and suddenly realize that info is not from openwrt host. Oct 24 11:02:19 https://gist.github.com/damex/84c09592a931a7e9c61112fdff17c139 full log so no problem there so far Oct 24 11:15:36 iwinfo mesh0 scan ; wifi down ; wifi up Oct 24 11:18:12 Could u tell me how I can check those values here? https://github.com/openwrt/openwrt/blob/master/tools/firmware-utils/src/mkfwimage.c#L132-L140 Oct 24 11:20:58 nick[m]: kern_start would be visible from vendor firmware boot log / partition layout Oct 24 11:21:22 nick[m]: kern_entry from u-boot log when starting the kernel up, and it must match the address used when building the kernel. Oct 24 11:23:40 PaulFertser: okay. I will flash airos with serial again and save all boot logs. Thank u. When I can compare them. Could that be a possible bug from 4.19 to 5.4, or am I wasting my time? ;) Oct 24 11:24:31 nick[m]: probably not needed yet Oct 24 11:24:31 i doubt if you need airos, just look from u-boot Oct 24 11:24:40 nick[m]: can you instead show full bootlog you're currently getting? Oct 24 11:24:55 Bootloader messages should be enough. Oct 24 11:25:02 +1 Oct 24 11:30:34 PaulFertser: russell-- https://gist.github.com/PolynomialDivision/bac6f565a961331ebbe9265265fb0627 Oct 24 11:30:48 that is the log I get with current openwrt trunk Oct 24 11:30:55 should I flash airos again? Oct 24 11:31:15 I backported some jffs2 fix from trunk Oct 24 11:31:16 did not helped :/ Oct 24 11:31:25 what is your problem? Oct 24 11:31:26 so it is not "completely trunk" Oct 24 11:32:18 still trying to fix my nanobeam ac gen2 and nanostation ac loco: https://bugs.openwrt.org/index.php?do=details&task_id=3272&opened=1310&type%5B0%5D=&sev%5B0%5D=&due%5B0%5D=&cat%5B0%5D=&status%5B0%5D=open&percent%5B0%5D=&reported%5B0%5D= Oct 24 11:32:42 " Newly-erased block contained word 0xd31704ef at offset 0x003d0000" and a lot of more messages with no persistent data Oct 24 11:32:56 I found some similar issue on the old bug tracker Oct 24 11:33:08 https://dev.archive.openwrt.org/changeset/48829.html Oct 24 11:33:14 I thought maybe I can just check them Oct 24 11:33:23 I tried the 4.19 spi nor driver with 5.4 kernel Oct 24 11:33:29 did not help Oct 24 11:33:56 and tried to port the 4.19 jffs2 to 5.4 but Iw as not succesfull xD Oct 24 11:34:41 i'm i reading that sysupgrade with -n works? Oct 24 11:34:51 nick[m]: if your kernel boots then those values are correct Oct 24 11:34:53 no. Oct 24 11:35:03 PaulFertser: oh okay. Oct 24 11:35:11 even that firmware_max_length? Oct 24 11:35:34 Since I went from AirOs 8.7? Oct 24 11:36:19 nothing to do with airos Oct 24 11:37:20 svanheule[m]: ping Oct 24 11:37:21 what does the full openwrt bootlog look like? Oct 24 11:38:08 Borromini: pong Oct 24 11:38:22 Borromini: sup :-) Oct 24 11:38:31 o/ Oct 24 11:39:14 russell--: https://gist.github.com/PolynomialDivision/1cc1c98a9977393844e2609d87ee8dee Oct 24 11:39:25 and then a lot of jffs2 errors follow Oct 24 11:40:24 nick[m]: i want to see them too, the offsets will be interesting Oct 24 11:43:48 russell--: https://gist.github.com/PolynomialDivision/dfee22441f5ba8a4ebf8d5a8563f78ba Oct 24 11:44:06 russell--: the device just rebooted Oct 24 11:44:22 russell--: this jffs2 stuff is happening every reboot Oct 24 11:45:37 russell--: here is an issue on the XW plattform: https://dev.archive.openwrt.org/ticket/20982 Oct 24 11:46:08 russell--: they make airos responsible and uboot Oct 24 11:48:22 not likely Oct 24 11:48:55 looks like jffs2 is never created correctly Oct 24 11:49:50 russell--: yep. :/ Oct 24 11:50:30 try getting a dump of /dev/mtdblock2 Oct 24 11:50:30 russell--: should I again try the spi-nor driver from 4.19 and make a bootlog? Oct 24 11:50:35 no Oct 24 11:50:45 Wed Oct 21 14:26:56 2020 kern.notice kernel: [ 0.419103] 0x0000004f0000-0x000000f60000 : "rootfs_data" Oct 24 11:51:03 that's where your kernel thinks jffs2 should live Oct 24 11:53:02 russell--: wifi -> "Command failed: Not found" Oct 24 11:53:28 russell--: it is getting more and more broken ^^ I have to find a way back into this device, then I will make a dump Oct 24 11:54:05 return to vanilla openwrt without your hacks Oct 24 11:57:32 then dd if=/dev/mtdblock2 of=/tmp/mtdblock2.img, scp that to your build box and look at the hexdump. you should see the block headers in the jffs2 section Oct 24 12:01:53 you should only see this: "Wed Oct 21 14:28:32 2020 kern.warn kernel: [ 121.156052] jffs2_scan_eraseblock(): End of filesystem marker found at 0x0" on the first boot Oct 24 12:05:13 russell--: I have 19% percent of that image ^^ the device crashed again. I have to retry that. but the 0x0 address Oct 24 12:05:14 https://gist.github.com/PolynomialDivision/0ea1608e9a8c7c04cde25a229da51827 Oct 24 12:05:45 shouldn't there be 0xdeadc0de? Oct 24 12:06:03 Im stuck at "PCIe: Port 0 aligning TLP counters as workaround to maintain ordering" need help, https://forum.openwrt.org/t/adding-openwrt-support-for-xirrus-xr-2400-target/69764/19?u=n4gi0s Oct 24 12:08:20 firstboot has the same effect or? so I just do firstboot again and dump the flash Oct 24 12:09:26 i'd try to do a new sysupgrade, and dd after it boots the first time Oct 24 12:11:41 if you are running out of ram, you can run the dd from ssh and pipe the output to your machine, something like: ssh root@yourdevice 'dd if=/dev/mtdblock2' > /tmp/mtdblock2.img (or wherever) Oct 24 12:32:33 russell--: can I send u an email? Oct 24 16:34:20 https://github.com/torvalds/linux/blob/master/arch/mips/cavium-octeon/executive/cvmx-helper.c#L252 i wonder if someone could take a look - board finds 3 ports while only 1 is defined. interface 0 and 1 is qsgmii. interface 3 is loop (what is that loop?). we need to disable interface 1 since nothing is connected. not defining it in device tree is not enough Oct 24 16:35:07 so that case switch to port 1. which is not defined in device tree. Oct 24 19:28:42 I'm really not one to ask for help often, but I'm having a hard time expressing dependencies in a package makefile… Oct 24 19:28:49 Any make gurus around? :) Oct 24 19:41:18 Guess not… :P Oct 24 19:44:22 i am not a guru by any stretch but there are quite a few examples in the packages repository Oct 24 19:47:40 Haven't found any example for what I want to do… maybe it's not supported. Oct 24 19:48:28 I want something like DEPENDS:=(pacakge-A || package-B). Oct 24 19:48:51 (Package B being a variant of package A.) Oct 24 19:49:51 dunno but doesnt that indicate that there is some overhead there between package-a and package-b Oct 24 19:50:14 maybe there should be a package-c with the stuff that both package-a and package-b depends on? Oct 24 19:50:36 Wait…! Maybe a PROVIDES directive…? Oct 24 19:50:38 Hmm… Oct 24 19:52:21 rsalvaterra: yes provids does it we have something similar with ip and ip-full Oct 24 19:52:54 Hauke: Thanks, I'll have a look! Oct 24 19:54:14 I also noticed the CONFLICTS directive causes indentation depending on the order of how it's expressed. That's a bit weird. Oct 24 20:00:35 Epic! Exactly what I need, works like a charm! Oct 24 20:01:33 So, package-B provides and conflicts with package-A. :) Oct 24 20:54:39 rsalvaterra: Notice on the ffmpeg suggestion.. isn't ffmpeg like horribly out of date because no one wants to maintain it? Oct 24 21:02:49 How do you walk that line where it's what you need in the package (and why you've sent the PR) vs people who just want to pull down a file without any real work? Legit question, since I'm going to be facing that eventually Oct 24 21:56:55 Grommish: I couldn't care less about ffmpeg in OpenWrt. :P Oct 24 21:57:07 On the desktop, however… Oct 24 21:58:50 Anyway, I'm not sure I understood your question… Oct 24 21:59:10 What do you mean by "pull down a file without any real work"? Oct 24 23:40:30 Wonder if there is gonna be a DMCA takedown request for http://sources.openwrt.org/youtube_dl-* Oct 24 23:44:09 * blocktrron starts packaging libdvdcss Oct 24 23:46:03 Welp, this was a learning experience. Package variants are… "interesting". Oct 24 23:47:53 The youtube-dl repository has been taken down…? :| Oct 24 23:48:24 I've been looking into it a bit for iw/ ip, but didn't really achieve what I wanted (using personal meta packages to select the -full variants on my big flash devices) Oct 24 23:50:44 I *think* I got it working for tor. I spent hours just getting menuconfig not to blow up. Oct 24 23:52:38 for me, the lithmus test was trying to build a multi profile target with large and smaller flash sizes without negative per-device dependencies Oct 24 23:53:50 Sounds quite "fun" too. :) Oct 24 23:54:11 wpad is also difficult to select that way Oct 24 23:54:45 The whole hostapd/wpa_supplicant/wpad configuration is borderline insane. Oct 24 23:55:13 for wpad, I had to declare defeat - and use negative dependencies instead (adding something like "-libustream-wolfssl -wpad-basic -wpad-mini -wpad wpad-openssl -wpad-basic-wolfssl" to each and every device...) Oct 24 23:57:04 Speaking of which, what's the thing with the CONFLICTS directive? It causes strange indentation in the menuconfig, depending on how it's expressed. Oct 24 23:57:17 (patching out the default ssl library/ wpad type would have been easier, but...) Oct 24 23:58:11 And that's one of the reasons why the wireless daemons configuration is crazy… Oct 24 23:59:30 Alright! I have tor-basic and tor-full building correctly! Now to generate a real image… Oct 25 00:00:12 fortunately tor is a bit easier, as there's little danger for one of its variants to become preinstalled by default Oct 25 00:01:07 (in the sense of not having to worry about default selections) Oct 25 00:01:57 Yes, most definitely. I only have to worry about keeping the same functionality with an existing .config. Oct 25 01:24:54 rsalvaterra: Meaning, the users who just want to use imagebuilder or the snapshots without actually building anything Oct 25 01:25:19 rsalvaterra: Do you think bias'ing towards those users is better, since the "power users" can more easily build their own Oct 25 01:25:58 rsalvaterra: or, do you think you put it the way most people (especially you) would need it without stressing over the other cases Oct 25 01:26:39 rsalvaterra: I'm asking regarding the comments on your PR.. I'm asking which way, of two valid opinions, do you think is best Oct 25 01:35:23 Grommish: As long as my particular use case is satisfied, I don't mind at all making things more simple for everybody else. :) Oct 25 01:36:25 In this particular case, I believe the two package variants approach is just fine. Oct 25 01:38:15 And now I'm off to bed. Cheers! **** ENDING LOGGING AT Sun Oct 25 02:59:57 2020