**** BEGIN LOGGING AT Fri Jun 12 02:59:58 2020 Jun 12 03:41:48 build #334 of mpc85xx/p2020 is complete: Failure [failed kmodconfig] Build details are at http://buildbot.openwrt.org/master/images/builders/mpc85xx%2Fp2020/builds/334 blamelist: Stijn Tintel , Kevin Darbyshire-Bryant , INAGAKI Hiroshi Jun 12 06:33:39 anyone else with a Unifi AP AC Pro running ath79 no longer able to sysupgrade? I am stuck at r12464-a017773a92 Jun 12 06:34:02 it appears to be starting the sysupgrade, then reboots with the old version Jun 12 07:16:21 how do I define a partition layout for openwrt to boot from nand? there's 22 mtd partitions on the nand with the last 3 being oem firmware partitions (mtd20 has kernel, rootfs, and overlay; mtd21 is a backup of mtd20; mtd22 is just logging) that add up to around 230M (0xDC00000 in hex). the uboot on this router (RAC2V1K) has ubi support and support for booting directly from nand. Jun 12 07:33:37 stintel, i only remember reading failed sysupgrade on freifunk firmware (base on OpenWrt) - a race condition with olsrd (memory intensive) and crond - oom killer is invoked and kills sysupgrade Jun 12 07:36:07 but only low memory devices: https://github.com/freifunk-berlin/firmware/issues/806 Jun 12 08:33:47 nick[m]: let me know when you want me to update the dawn package. Jun 12 08:34:14 dwmw2_gone: adrian replied to you mt7623 PR Jun 12 08:34:27 yeah, just looking through that. Jun 12 08:34:33 Mostly cosmetic so far AFAICT from my phone. Jun 12 08:34:44 it is just cosmetic Jun 12 08:35:20 I was looking for some way to avoid having to respecify the mt7623n_bpir2 prefix in the image creation, but I can't find a suitable environment variable. Jun 12 08:35:46 That's also available when we're building uboot and the preloader (since those artifacts have the prefix in question) Jun 12 08:37:57 IMAGE_PREFIX ? Jun 12 08:38:17 include/image.mk:IMG_PREFIX_EXTRA:=$(if $(EXTRA_IMAGE_NAME),$(call sanitize,$(EXTRA_IMAGE_NAME))-) Jun 12 08:38:20 include/image.mk: IMAGE_NAME = $$(IMAGE_PREFIX)-$$(1)-$$(2) Jun 12 08:39:51 What makes a package get built for a specific target (and is that the same as for a specific image) instead of generically for that CPU/platform ? Jun 12 08:40:13 Because the mt7623n-preloader package can only put its blob into an image-specific place if it's built for that image, right? Jun 12 08:41:00 target/linux/mediatek/image/mt7623.mk Jun 12 08:44:28 I included image.mk in mt7623n-preloader/Makefile but $(IMAGE_NAME) isn't set. Jun 12 08:44:56 in my Build/InstallDev section. Jun 12 08:45:04 which kind of makes sense, since that isn't image-specific Jun 12 08:45:08 I may just stick with hard-coded names. Jun 12 08:45:17 it just grates a little Jun 12 08:49:24 lmore377: name "ubi" the one with ubi Jun 12 08:49:39 lmore377: OpenWrt's kernel will autoattach "ubi" one and look for volumes in it Jun 12 09:08:47 dwmw2_gone: i pushed a fix but my pc comiles still the toolchain and I have to leave now :S Jun 12 09:09:05 OK, I'll test and update the package, if you're OK with that? Jun 12 09:09:06 so i don't know if the fix is okay Jun 12 09:09:27 no, you needed to cast to size_t not cast to int. Jun 12 09:09:30 It was already an int. Jun 12 09:09:35 - snprintf(hostname, HOST_NAME_MAX, "%.*s", len, ptr.o->v.string); Jun 12 09:09:35 + snprintf(hostname, HOST_NAME_MAX, "%.*s", (int)len, ptr.o->v.string); Jun 12 09:09:39 that cast is redundant Jun 12 09:09:55 er Jun 12 09:09:58 maybe I lie Jun 12 09:10:46 no, you're right. Jun 12 10:06:56 dwmw2_gone: if u have time to test would be awesome. Jun 12 10:06:59 I had to revert some commit that was done to the tcp socket Jun 12 10:08:06 my tree is currently building banana pi but when that's done I'll switch back and test Jun 12 10:35:14 can I please get review of https://patchwork.ozlabs.org/project/openwrt/patch/20200610161040.8903-1-ynezz@true.cz/ ? Jun 12 10:37:07 I need this one badly in order to be able to do automatic testing of snapshot images (tftbooting initramfs) and don't want to wait few more days, thanks! :) Jun 12 10:37:30 it's mvebu related patch Jun 12 10:38:33 that patch is a gem Jun 12 10:38:37 * ynezz hides Jun 12 10:44:57 what's the original purpose of the patch? Jun 12 10:45:36 ynezz: I think we have a bootargs-override somewhere? Jun 12 10:47:22 we do. but it isn't a treewide one: target/linux/mpc85xx/patches-5.4/102-powerpc-add-cmdline-override.patch Jun 12 10:50:27 I don't follow Jun 12 10:51:04 that original patch will clear all commandline arguments passed to kernel if it doesn't contain root= parameter Jun 12 10:51:28 so booting just with `console=` doesn't actually provide console as the kernel commandline is empty Jun 12 10:51:45 bootargs-override would need modified DTB? Jun 12 10:52:23 I just want to tftpboot test pristine images from downloads.openwrt.org, no modding Jun 12 10:55:04 ynezz: sorry. I misread the patch :( I thought it's for some weird u-boot which patch various dts properties like bootargs and mtd partition table. Jun 12 10:58:13 gch981213: no worries, thank you for looking :) Jun 12 11:10:31 Adran: hm Jun 12 11:10:35 why my u-boot has this header Jun 12 11:10:37 00000000 88 16 88 58 6e 98 04 00 55 2d 42 6f 6f 74 00 00 |...Xn...U-Boot..| Jun 12 11:10:37 00000010 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| Jun 12 11:10:37 00000020 00 00 00 00 00 00 00 00 00 00 e0 81 ff ff ff ff |................| Jun 12 11:10:37 00000030 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................| Jun 12 11:10:40 * Jun 12 11:10:40 00000200 b8 00 00 ea 14 f0 9f e5 14 f0 9f e5 14 f0 9f e5 |................| Jun 12 11:10:44 I want what's at 0x200 to be at 0x0 Jun 12 11:13:46 aha, got it Jun 12 11:30:49 mangix: can you please revert the change of the zstd download to zstd? Jun 12 11:31:13 zstd is too new to be universally supported by tar everywhere Jun 12 11:36:45 at least Ubuntu 20.04 LTS does not support it, and I guess neither does CentOS 6.x / 7.x Jun 12 11:36:53 Debian stable (buster) neither Jun 12 11:45:08 jow: why not support zstd? Jun 12 11:46:31 FreeBSD have zstd in base Jun 12 11:47:34 Ivan_83: i think buster is kind of the baseline. you don't want people to go around and install stuff from third party repos to build openwrt. Jun 12 11:47:41 even in kernel: options ZSTDIO #m zstd-compressed kernel and user dumps Jun 12 11:47:55 Ivan_83: it is not about supporting zstd but about silently switching the download of the zstd tarball itself to zstd Jun 12 11:48:08 which is kind of a pointless optimization that just introduces chicken-egg problems Jun 12 11:48:34 ohh, ok, sorry for noise ) Jun 12 11:48:54 given the current LTS distro baseline, you can assume that gzip, bzip2 and (at least to some extend) xz is available Jun 12 11:49:15 zstd is simple too new, iirc .tar.zstd is only supported since tar 1.31 Jun 12 11:49:23 which didn't land in all LTS distros yet Jun 12 11:50:01 IIRC centos 6 is using gcc 4.8 Jun 12 11:50:15 so would be surprised with zstd :p Jun 12 11:50:29 yeah but CentOS 6 is not relevant anymore. there is 7 and 8 Jun 12 11:50:37 if you're still on 6, you need to ask yourself some questions Jun 12 11:52:26 Centos 7 uses tar 1.26 Jun 12 11:52:36 Centos 8 uses tar 1.30 Jun 12 11:52:41 neither support zstd officially Jun 12 11:54:22 Debian Buster uses tar 1.30 but seems to have backported stuff Jun 12 12:04:46 stintel: explain that to the people reporting build issues with gcc 4.8 Jun 12 12:05:43 afaik it's only centos 6 userbase Jun 12 12:09:23 Please install GNU 'find' - I do not want GNU :( Jun 12 12:09:50 and some people don't want zstd :P Jun 12 12:13:45 Ivan_83: welll.... Jun 12 12:13:53 find --version 2>&1 | grep GNU - madness check Jun 12 12:14:07 I want try to build on FreeBSD Jun 12 12:14:08 simpler than trying to check all the options that are used Jun 12 12:14:44 gnulag! :) Jun 12 12:18:55 well, comment the check out if you think your own find will do the right thing. Jun 12 12:18:56 Ivan_83: but you can have gfind on FreeBSD, and gawk and gmake. Jun 12 12:19:06 but thye don't want to... Jun 12 12:21:25 hm, my kernel isn't getting cmdline from uboot Jun 12 12:21:33 bootargs=console=earlyprintk console=ttyS0,115200 console=tty1 fbcon=map:0 root=/dev/mmcblk1p2 rootfstype=ext4 rootwait vmalloc=496M debug=7 initcall_debug=0 video=1920x1080 drm.debug=0x7 Jun 12 12:21:46 [ 0.000000] Kernel command line: console=ttyS2,115200n8 Jun 12 12:23:12 Many OpenWrt targets enable overriding command line because bootloaders are often fully uncooperative. Jun 12 12:24:13 CONFIG_CMDLINE="earlyprintk console=ttyS0,115200 rootfstype=squashfs,jffs2" Jun 12 12:24:13 CONFIG_CMDLINE_FROM_BOOTLOADER=y Jun 12 12:24:13 # CONFIG_CMDLINE_EXTEND is not set Jun 12 12:24:13 # CONFIG_CMDLINE_FORCE is not set Jun 12 12:24:18 This is Banana Pi R2 Jun 12 12:25:40 * dwmw2_gone frowns at target/linux/mediatek/patches-5.4/0006-dts-fix-bpi2-console.patch Jun 12 12:26:14 PaulFertser: I already have find, stat.... Jun 12 12:27:44 Ivan_83: you have to take into account than many of the OpenWrt packages are mostly tested on GNU/Linux systems and so their build scripts might have certain assumptions. Of course it would be nice to fix them to be POSIX-compatible. Jun 12 12:28:26 it builds on macOS as well Jun 12 12:28:35 (using GNU ports) Jun 12 12:29:20 I imagine, that it would be uphill battle otherwise Jun 12 12:29:45 last time I looked at fbsd, I've used the GNU stuff as well Jun 12 12:31:41 IIRC I got stuck at the point where it bombed out on an error with missing host headers for kernel modules or such Jun 12 12:32:43 anyway, it would be nice to have fbsd builder Jun 12 12:36:26 nice for me to - do not like run vbox with ubuntu only for build owrt ) Jun 12 12:46:46 iirc some time ago compiling on a bsd variant worked - just like OS X, main distros or rolling distros Jun 12 12:47:34 plntyk: yes, using g* ports Jun 12 12:47:48 which Ivan_83 do not want on his system Jun 12 12:48:04 there's just too much stuff requiring gawk, gsed, gfind etc. Jun 12 12:48:19 using features which are not present or working differently in non-GNU variants Jun 12 12:49:02 yeah - it like wanting to keep those old versions - but some GCC people decided to do new releases Jun 12 12:49:07 or those evil C++ ppl Jun 12 12:49:34 those GCC people abandoned Kernighand and Ritchie style C ^^ Jun 12 12:52:39 most strange - cp checks Jun 12 13:09:22 missed check: /bin/sh: mkhash: not found Jun 12 13:15:55 mkhash is built by the "make prepare-tmpinfo" target Jun 12 13:16:52 which is a precondition for almost all other make targets (make menuconfig etc.) Jun 12 13:18:43 Hmm... I just noticed, that 19.07.3 appears to miss a hostname binary: Jun 12 13:18:43 /usr/share/openvswitch/scripts/ovs-ctl: line 42: hostname: not found Jun 12 13:18:43 By intuition, I'd have expected busybox to provide some. Jun 12 13:18:59 Do you know what cound be wrong, here? Jun 12 13:47:29 I already patch mkhash.c to build Jun 12 13:48:32 now it try to build one of my router image Jun 12 13:56:27 why owrt need to build cmake? Jun 12 13:57:04 to not require users to install it Jun 12 13:57:25 and to ensure it is present in a minimum supported version Jun 12 13:57:48 but I already have fresh one Jun 12 14:01:50 * ldir It's a long day already - goes for a cycle Jun 12 14:03:35 is there any was to switch to already exist cmake? (cmake from owrd build fails and I prefer to not fix it Jun 12 14:08:33 or at least where I can store patch files for cmake? Jun 12 14:14:11 Ivan_83: patches are in tools/cmake/patches. If you want to use the host one instead, I guess you can drop cmake related lines in tools/Makefile and add a SetupHostCommand in include/prereq-build.mk. (I never tried this though.) Jun 12 14:16:00 I still suggest using a Linux VM Jun 12 14:16:18 I want extactly through all that trouble a few years back (getting base OpenWrt built on FreeBSD) Jun 12 14:16:24 see where it stands now :) Jun 12 14:16:31 its a huge waste of time and effort Jun 12 14:16:50 *Linux VM or g* ports Jun 12 14:17:28 and even if you manage to get base building, then stuff from the extended packages feed is likely a lost cause Jun 12 14:18:29 I already installed g* ports %) Jun 12 15:06:20 Hello everyone. Anybody here with a router that has the RTL8198 chipset? I'm trying to dump the firmware before I try and get OpenWrt on it, but the bootloader has like five commands in total. Jun 12 15:07:42 When the original kernel/firmware combo boots I think I get to a Linux shell, but obviously there's no documentation on the root user/PW combo. So I need to dump the flash somehow and then find the password from the shadow file. Anybody with this chipset have any luck with firmware dumping? Jun 12 15:09:59 Have you checked for a username/password combo for the device? If it's from your ISP, you can sometimes find that info posted online by disgruntled ex-exployees. Jun 12 15:10:45 Or, they actually have it posted legit and just buried.. Spectrum Cable in the states had the info for the Admin panel on the routers they gave out, but never advertised it Jun 12 15:11:46 Somebody reverse-engineered a similar router (but with an Atheros chipset) from the same ISP and posted the root credentials, but they had it taken down because the ISP requested masking :/ other than that there's not much to go on for the router. I can get to the bootloader shell by pressing Esc or to the login screen of the firmware. Jun 12 15:12:50 Does it have TFTP access? Jun 12 15:12:56 I do see a Winbond chip on the router itself, but it's a soldered part with no exposed leads so I can't get any flash clips onto that thing. i can try desoldering it but with my bad skills more likely the chip will die during the attempt Jun 12 15:13:21 Grommish: not sure, I don't see any connection attempts on bootup Jun 12 15:13:35 it wouldn't try any unless you tell it to Jun 12 15:13:48 You'd have to halt the bootloader Jun 12 15:14:31 Yeah, this is what I get over serial when the router is booting up: https://hastebin.com/azakaladug.makefile Jun 12 15:14:32 and look around in there.. If it has tftp client, you can setup a TFTP server and maybe find a firmware to sideload Jun 12 15:15:17 Yeah, but I'd like the original firmware as insurance in case I bork it Jun 12 15:16:04 No doubt.. but that's what I'm saying.. if it has TFTP abilities, it was intended to be used.. so the manu probably has it up somewhere Jun 12 15:16:35 Maybe not, but some routers are just so convoluted sometimes Jun 12 15:17:23 Yeah, but the bootloader menu feels incomplete, like a lot of the commands were stripped out. Jun 12 15:17:52 Here's a printout when I run help: https://hastebin.com/iluyawojaw.xml There's like, six or seven commands and that's it? Jun 12 15:19:12 so does "dw" work ? Jun 12 15:20:19 Not sure what that does, but when I type "dw" it comes up as "Unknown command !" Jun 12 15:20:23 its probably the same bootloader as https://openwrt.org/docs/techref/bootloader/realtek - see number of "-" in COMMAND MODE line Jun 12 15:21:17 ah you listed "db" Jun 12 15:21:29 so you should be able to dump the memory / flash over serial Jun 12 15:21:34 plntyk: thank you! Now I know what the commands are for. Jun 12 15:23:13 hello hello Jun 12 15:23:46 is there a "best practice" to make a package that replaces a specific firmware from linux-firmware package? Jun 12 15:24:05 I can't just make a new package that installs the binary, because it will conflict with linux-firmware Jun 12 15:24:42 and I can't not install linux-firmware Jun 12 15:24:56 matteosilex, look at ath10k-ct firmware that replaces the ath10k "normal" firmware from linux-firmware afaik Jun 12 15:25:23 plntyk: amazing, I'll have a look now, thanks! Jun 12 15:36:15 plntyk: mmm... looks like the ath10k-ct replace the "normal" kernel module so that you can use Candela's special firmware, but the package does not include the firmware itself Jun 12 15:36:25 plntyk: maybe I'm misreading? Jun 12 15:37:32 look under package/firmware Makefiles Jun 12 15:37:56 ath10k-ct-firmware Jun 12 15:40:26 OK, so I've confirmed that I could dump portions of the firmware through serial using the DB command, but how do I know how much to dump? Do I just run "DB 80a00000 1024000000" and let it run until it encounters the end of the flash? Jun 12 15:41:07 I've found that if I try and dump the first bits of the flash ("DB 00000000 32") it will just give me an "Unidentified exception" and freeze the shell Jun 12 15:43:28 well you know the flash size from the info/number thats printed on the chip - BUT the "db" dumps memory and the flash is probably mapped to some adress in memory Jun 12 15:44:07 adress probably more like 0x______ Jun 12 15:44:11 dunno Jun 12 15:45:18 plntyk: oh, I see. what they do works nicely. Unfortunately, the firmware I have to replace is for the brcmfmac driver, which SELECTs the package with the default firmware Jun 12 15:45:21 you should write some short script to convert the lines / dumped chars into binary and then check with strings Jun 12 15:45:45 plntyk: while in the case of ath10k-ct, the driver does not depend on a specific firmware package, so you can install any Jun 12 15:46:20 plntyk: I'm thinking whether I should patch the Makefile that defines the brcmfmac package Jun 12 15:46:48 plntyk: i.e. package/kernel/mac80211/Makefile Jun 12 15:47:12 plntyk: and remove the dependency. it's a bit hacky though, as a solution Jun 12 15:47:29 matteosilex, what about sth like "ALTERNATIVES" ? would that work ? Jun 12 15:47:35 plntyk: thanks! I'll have to find where the flash is mapped in memory...and the flash size Jun 12 15:48:00 Kind of annoying, cos on Atheros platforms you usually have the "bootinfo" command and it spits out all the secrets Jun 12 15:48:50 main problem with dumping via "db" is slow serial speed Jun 12 15:49:24 after finding the right area Jun 12 15:51:05 plntyk: not sure how ALTERNATIVES work (I guess it's a feature of the build system). Wouldn't that still need to be changed in the brcmfmac package (i.e. the one that depends on the firmware package, not the firmware package itself) Jun 12 15:58:11 If you two are bored, and want to help me out, check out https://forum.openwrt.org/t/designing-openwrt-device-from-the-nearly-ground-up/66642 Jun 12 15:58:26 I need community minds :D Jun 12 15:59:54 specs: dual core OcteonIII CN7020 <- lol Jun 12 16:00:45 the other channel had a user wanting do run debian on an octeon platform Jun 12 16:01:17 Its a great box Jun 12 16:01:29 but Itus went phhttp and took everything with them Jun 12 16:01:43 and I don't know enough to ask the right questions.. Jun 12 16:02:32 octeon = mips with extra weirdness (https://lists.nongnu.org/archive/html/qemu-devel/2008-05/msg00805.html) Jun 12 16:02:58 Yeah, but Jun 12 16:03:01 I've already got it working :) Jun 12 16:03:45 My issue is that it isn't working properly from a user perspective. They only made a few hundred, but anything I put out will be their only choice Jun 12 16:04:58 for better tinkering the bootloader should support a way to load and boot different files from mmc partition or via tftp or via usb Jun 12 16:05:17 It does Jun 12 16:05:47 mmc has 4 partitions, 1 fat, 3 ext3 for role support (router, bridge, gateway).. or.. whatever.. selectable via front-panel gpio switch Jun 12 16:06:21 backup the 3 partitions, add a 4th , add a custom compiled kernel and run it with newer userspace on the 4th partition Jun 12 16:06:33 or boot kernel from fat and load root from usb Jun 12 16:07:13 It loads the ELF bin from the fat then pivots the root to the correct extroot Jun 12 16:07:27 but the kernel stays with the bin Jun 12 16:08:05 well you said the bootloader kann load different files - so it should be able to load a different kernel Jun 12 16:08:09 Currently I'm running 4.19.123 built from master branch Jun 12 16:08:39 dunno if you really want to keep the "3" use cases Jun 12 16:08:57 I don't either, but unless I want to build uboot, it's ok Jun 12 16:08:58 it just complicates things Jun 12 16:09:15 Each partition has about 1Gb of space on it Jun 12 16:11:15 I know there is no *vmlin* anywhere on the device that I can find. my assumption is that it is kept in the bin file that's loaded. Jun 12 16:11:46 I jus tnow it works, now I have to make it work cleanly. hense the post hehe so any ideas on how you'd do it if you were in my situation, let me know for sure Jun 12 16:12:36 I'm not against learning to compile a new uboot, but I would need to know how and what info to pull to make the new one. kernel is conjfusing enough, bootloader is a whole new level of huH? Jun 12 16:14:19 hm... maybe you better formulate what you want the device to do - from what you wrote it currently boots mainline OpenWrt fine - am I correct ? Jun 12 16:14:50 Yes. Jun 12 16:15:30 so one problem is the handling of the "switch" and operation mode of "normal" firmware Jun 12 16:16:06 Well.. That's not really an issue.. GPIO switch on the front tells uboot which image to load Jun 12 16:17:08 the switch logic might be just a boot script that is in uboot-environment - meaning you could change it or replace the ".scr" file of that partition if its "normal" boot script Jun 12 16:17:14 I read the GPIO to tell the system which /dev and whatnot Jun 12 16:18:17 Like I said, I'm not committed to anything Itus did, only that I can get what they do to work on a master branch build Jun 12 16:18:42 What would someone who is a networking enthusist do with a reference board, I guess the question is Jun 12 16:18:55 Cavium did release the 4.14 kernel for the Octeon3 Jun 12 16:19:08 and the toolchain, but I can't get it to build out Jun 12 16:20:57 I am not really getting what you are trying to improve / change if you manged to get openwrt running - you can write the kernel to the fat partition and uboot seems to boot from there and the kernel then loads the root depending on gpio hw switch Jun 12 16:21:45 Well, for starters, as it works, i can't update the kernel without replacing the BIN and wiping the extroot Jun 12 16:21:56 you have to add "sysupgrade" to update the right location of kernel (name) and then just copy 3 x the userspace to each partition Jun 12 16:22:16 See, that's just it.. I don't have a kernel location hehe Jun 12 16:22:45 Oh.. Wait.. Each image only handles one mode.. I'm just going to make 3 target defines Jun 12 16:22:55 the kernel should be in the fat partition - at least thats the way on arm boards with uboot Jun 12 16:23:16 The entire bin image is in the fat, yes, but it's the whole ELF bin Jun 12 16:23:50 Again, I know just enough to talk gibberish ;D Jun 12 16:23:54 so ? Jun 12 16:24:38 maybe provide a serial log of uboot and kernel bootup Jun 12 16:25:24 using a non-initramfs kernel image means that openwrt uses a real rootfs and not the one inside its initramfs Jun 12 16:25:33 https://youtu.be/XB-tjcb2Nio Jun 12 16:25:36 so you have to change that Jun 12 16:26:14 openwrt-octeon-itusrouter-initramfs-kernel.bin is the final build file Jun 12 16:26:16 that crappy youtube Jun 12 16:26:27 there is a reason pastebins were invented Jun 12 16:26:36 hahah Yes, but I had that handy Jun 12 16:26:42 one sec Jun 12 16:27:01 select a build option without initramfs Jun 12 16:28:36 unselect "ramdisk" in build options Jun 12 16:29:09 Oopes Jun 12 16:29:42 i am not sure why ramdisk is selected - whats the reasoning there Jun 12 16:30:01 That was not a graceful failover at all haha.. Jun 12 16:30:29 Honestly, because it works and that's how Itus had it.. i think.. between itus and the Ubit ER/ER Lite target, it just works Jun 12 16:30:48 "normally" the initramfs variant of images are only needed if you do not want to change flash / run from ram only Jun 12 16:30:55 which is my whole issue.. great.. but it works for me.. because I can build it.. but doesn't do anything for the other couple of hundred boxes out there Jun 12 16:31:17 See, there isn't a MDT flash Jun 12 16:31:33 It's just /dev/mmcblk1 and the 4 partitions Jun 12 16:31:57 So it's running like a Live CD.. if I comment out the switch_root, it runs just fine based on the image Jun 12 16:32:05 nothing is saved of course, but yeah, runs fine Jun 12 16:32:57 openwrt has build options to change kernel commandline - so after a backup you could try to overwrite one partition with real persistent data Jun 12 16:33:26 plntyk: this is probably just RAM right? Not the flash chip? https://www.mouser.com/ProductDetail/Winbond/W9751G6KB-25 Jun 12 16:33:29 https://hastebin.com/axoloforab.sql Jun 12 16:33:46 Right, it loads it into RAM as a ELF binary Jun 12 16:33:54 It's so weird because I can't find any other chip on the board that resembles flash Jun 12 16:34:19 Maybe it's part of the SoC? Jun 12 16:34:26 I'm surprised it's gotten as far as it has, if I'm honest Jun 12 16:34:40 No, see, I have access to uboot AND the board firmware Jun 12 16:34:47 in the far partition Jun 12 16:34:58 oct2boot.bin and the rhino reference Jun 12 16:35:16 I think the rhino reference is stage 1, calling oct2boot.bin which is uboot Jun 12 16:35:24 so i could change that if I knew how I guess Jun 12 16:37:09 Ideally, i'd be able to build a minimalist image that's ONLY the kernel and libs/deps, which would be what is loaded into RAM at stage 3 Jun 12 16:38:17 but I'm not sure how to deal with installed stuff that might break if the kernel changes, because those are on the mmcblk1p2-4 partitions Jun 12 16:38:33 you have emmc - that might be a more ram like chip size Jun 12 16:38:46 It's for 1GB RAM, 3.8GB eMMC Jun 12 16:38:49 err got Jun 12 16:38:51 i doubt its PoP - package on package - but might be so Jun 12 16:40:10 Here is the working config file.. Jun 12 16:40:11 https://github.com/Grommish/Itus_Shield_v2/blob/master/.config Jun 12 16:40:20 But, keep in mind that my final image was 110MB Jun 12 16:40:36 I've got full python3 support Jun 12 16:40:58 because.. why not.. this is the issue I am facing.. the sandbox seems so large I'm not sure WHAT to do with it Jun 12 16:41:55 Real devs are fighting for every extra byte, and I broke luci stats because my Snort3 process was using 575m RAM and the lua wasn't set for non-numbers Jun 12 16:43:02 Anyway, I've got to run. If you feel up to it, drop in on my thread Jun 12 16:43:05 thanks again! Jun 12 16:49:17 did gromish share any board photos? Jun 12 16:54:07 plntyk: found it! I was so focused on the Winbond logo that I completely missed the (now obvious) Macronix chip Jun 12 16:54:47 Part #: MX25L12845EMI-10G, 128Mb chip I think Jun 12 17:58:13 Signing off for the night, thanks Grommish and plntyk for the help :) Jun 12 18:18:43 when is planned 19.07.4, can more devices be added if they are very simple commits cherrypicked from master? how to propose that? Jun 12 18:19:14 propose means: that I tried on my own on real hardware and it works Jun 12 19:10:40 `char neighbor_report[NEIGHBOR_REPORT_LEN] = "";` can that cause null pointer on some architectures? Jun 12 19:10:41 whups irc Jun 12 19:10:42 char neighbor_report[NEIGHBOR_REPORT_LEN] = ""; Jun 12 19:13:25 nick[m]: if NEIGHBOR_REPORT_LEN != 0 shouldn't be problematic. Jun 12 21:02:29 Hm, is the build system making a uImage with a dtb included for my banana pi r2? Jun 12 21:02:36 Can I make it *not* do that and use the one from u-boot? Jun 12 21:03:12 * dwmw2_gone comments out DEVICE_DTS Jun 12 21:09:07 PaulFertser: thank u. I think u are right! :) Jun 12 21:26:49 what's the consequence of a KCONFIG entry in a package not having =y/=m/=n after it? Jun 12 21:28:32 e.g. KernelPackage/libsas has CONFIG_SCSI_SAS_ATA=y but just CONFIG_SCSI_SAS_ATTRS Jun 12 21:32:54 nick[m]: debugging reproducible faults on targets is relatively easy with gdbserver, there's a nice script to help with running GDB from host for that. Jun 12 21:34:03 PaulFertser: reproducible is the problem xD Jun 12 21:35:59 nick[m]: I hope OpenWrt has working support for core dumps these days. Jun 12 21:39:47 adrianschmutzler: such an entry will inherit the selection state from the package symbol Jun 12 21:42:49 PaulFertser: you mean support will be coming? Jun 12 21:43:16 nick[m]: no, I mean it should be relatively easy to enable core dumps generation. Jun 12 21:43:29 it works ootb since a few years Jun 12 21:43:48 just ulimit -c unlimited; before invokign the crashing process Jun 12 21:44:25 Superb Jun 12 21:44:49 can I put this into init.d? Jun 12 21:46:35 adrianschmutzler: the core logic is in scripts/package-metadata.pl, line 73-85 Jun 12 21:47:29 adrianschmutzler: if a value is set (CONFIG_sym=x) it is used, if just the symbol is specified, it is set to if the parent kmod pacakge is enabled as <*> or Jun 12 21:47:58 unless the target kernel config already enables this symbol as builtin, in this case it is set to <*> and not Jun 12 21:48:28 if the parent kmod package is *not* selected, the value-less kconfig sybols is set to (not set) Jun 12 21:48:47 s/sybols/symbol/ Jun 12 21:52:35 jow: so, CONFIG... without =x is essentially =m if package is selected, =n if pkg not selected, unless setting the symbol explicitly elsewhere will overwrite? Jun 12 21:53:34 right, with elsewhere being the target or subtarget kernel config overrides usually Jun 12 21:53:50 e.g. a subtarget kernel enabling builtin usb support Jun 12 21:54:04 the corresponding kmod-usb-whatever will still be built, but not contain a *.ko Jun 12 21:54:29 and its corresponding KCONFIG:=CONFIG_USB_whatever will not downgrade from =y to =m Jun 12 21:56:16 hmm, so what would be a reason for setting something to =m explicitly then? Jun 12 21:56:36 good question Jun 12 21:56:50 sounds like a bug / cargo-cult / pacakging mistake to me Jun 12 21:58:47 is there some site I can look at cheap routers with ipq4018 ipq4019 ipq4028 ipq4029 nvebu soc? Jun 12 21:58:59 searching for something like fritz!box 4040 Jun 12 22:02:04 jow: okay, thanks. that should be enough understanding for what I want to do. Jun 12 22:02:32 commit 7765e442d04e4c19690f81084a9726776aea8b76 description has some core dump enabling info that may be useful Jun 12 22:15:29 build #335 of mpc85xx/p2020 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/mpc85xx%2Fp2020/builds/335 Jun 13 00:10:20 I finally flashed openwrt and on first boot I get Jun 13 00:10:20 [ 53.336169] jffs2_build_filesystem(): unlocking the mtd device... Jun 13 00:10:20 [ 53.336175] done. Jun 13 00:10:20 [ 53.341265] jffs2_build_filesystem(): erasing all blocks after the end marker... Jun 13 00:10:20 [ 123.607405] done. Jun 13 00:10:22 [ 123.614027] jffs2: notice: (1119) jffs2_build_xattr_subsystem: complete building xattr subsystem, 0 of xdatum (0 unchecked, 0 orphan) and 0 of xref (0 dead, 0 orphan) found. Jun 13 00:10:24 [ 123.788397] overlayfs: upper fs does not support tmpfile. Jun 13 00:11:09 but after a reboot I get Jun 13 00:11:09 [ 2.510218] read error in "firmware" at offset 0xcfc000 Jun 13 00:11:09 [ 71.819199] jffs2: cannot read OOB for EB at 0c960000, requested 8 bytes, read 8 bytes, error -74 Jun 13 00:11:20 and overlay never mounts Jun 13 00:21:51 If my image wasn't a legacy uImage with a dtb compiled into the kernel, but was instead a FIT image that u-boot could see the dtb in, would u-boot be able to set the command line in /chosen properly? Jun 13 02:52:15 build #163 of bcm27xx/bcm2711 is complete: Failure [failed kmodconfig] Build details are at http://buildbot.openwrt.org/master/images/builders/bcm27xx%2Fbcm2711/builds/163 blamelist: Adrian Schmutzler , J. Scott Heppler **** ENDING LOGGING AT Sat Jun 13 02:59:57 2020