**** BEGIN LOGGING AT Mon Jul 08 02:59:58 2019 Jul 08 08:47:37 aloha Jul 08 08:53:31 'lo blogic Jul 08 08:53:51 jow: did you migrate the server btw ? Jul 08 08:54:14 i'd cancle the contract on the old one in that case Jul 08 09:20:30 blogic: mostly. I think you can cancel the contracft now, it should run out at the end of the month, right? Jul 08 09:20:38 yes Jul 08 09:20:43 then cancel it now Jul 08 09:31:48 Hi, where can I find more information regarding SSB and BCMA of the BCM47xx? What do these buses connect? Jul 08 09:59:18 someoner: all SoC parts Jul 08 09:59:38 someoner: CPU, memory, Ethernet, Wireless, PCIE controller, USB controller Jul 08 10:00:46 someoner: on non-SoC devices (PCIe cards, USB cards, SDIO cards), ssb/bcma bus connects all card parts (common core, PCIe core, wireless core) Jul 08 10:13:19 rmilecki: ok this sounds pretty essential. Is there a way to ensure that a chip has SSB or bcma? I have a BCM53312, but no datasheet. GPL sources for ecos are available. Jul 08 10:14:09 someoner: there is a bit in some register to read, I'm not sure if I can find it now Jul 08 10:16:16 This sounds useful. I don't have any version of openwrt running on it yet (probably crashing at bus initialization). Is there an easy way to read a single register from CFE? Jul 08 10:17:19 someoner: CID_TYPE_MASK Jul 08 10:18:16 someoner: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/bcma/bcma_driver_chipcommon.h BCMA_CC_ID_TYPE Jul 08 10:18:32 someoner: i don't CFE allows doing that Jul 08 10:19:51 someoner: according to the https://wikidevi.com/files/Broadcom/product_brochure.pdf BCM53312 is a switch Jul 08 10:19:54 someoner: keep looking for a SoC Jul 08 10:21:15 you are the second guy telling me this. This is the main chip on the pcb and it is labelled BCM53312. THis should be the SoC and contains a MIPS CPU Jul 08 10:22:16 I have a picture Jul 08 10:23:35 maybe SoC chip is hidden? some radiator or sth Jul 08 10:24:50 https://imgur.com/pIdcB4d Jul 08 10:24:56 There is nothing on the backside Jul 08 10:25:11 I had to remove a heatsink from the SoC Jul 08 10:25:53 Direct link: https://i.imgur.com/pIdcB4d.jpg Jul 08 10:27:33 "Integrated Multilayer Switch and CPU" Jul 08 10:27:41 the product brochure you linked actually says: "8-Port GbE + 4-Port GbE Highly Integrated Multilayer Switch SoC" so it actually is a SoC Jul 08 10:31:39 someoner: it seems you're right then Jul 08 10:31:42 interesting Jul 08 10:45:20 someoner: http://hackadayarchive.com/viewtopic.php?f=2&t=3177 Jul 08 11:02:33 DonkeyHotei: Looks similar, but is a different device (I have a GS108Tv2). The post mentions the redboot bootloader, but I have the the CFE bootloader and my serial connection is working Jul 08 11:12:09 someoner: what is your flash size? Jul 08 11:14:08 16MB flash, 64MB RAM Jul 08 11:15:38 someoner: are you sure it's 16 MB? or maybe 16 Mb? Jul 08 11:18:08 I'm sure, I divided the Mb from the datasheet by 8 Jul 08 11:23:04 I think the part number is JS28F128P33TF70A, but I'm not home right now Jul 08 11:23:31 so 128Mb=16MB Jul 08 11:24:50 someoner: ok, it just seems like a weird chocie for eCos Jul 08 11:25:01 I thought eCos was only used for devices with very small flash size Jul 08 11:25:05 interesting device Jul 08 11:30:42 a little bit too intresting for me ;), When I started I thought it shouldn't be too hard to install openwrt. Then the device turned out to be big endian... Jul 08 11:35:29 rmilecki: the FW update is about 4MB compressed so it has a big flash Jul 08 11:36:20 I mean in theory it could run something else than eCos. I only checked the sources. How would I check if the actual software on the device is ecos? Jul 08 11:37:09 basd on the UART output this does not look like Linux Jul 08 11:37:21 and eCos makes sense on such devices Jul 08 11:39:55 Interestingly the ecos sources use the following packages: CYGPKG_HAL_MIPS_BCM47xx and CYGPKG_HAL_MIPS_BCM953710 Jul 08 11:40:41 what is the BCM953710? anyone familiar with the broadcom product line notation? Google isn't finding anything Jul 08 11:44:08 Is it ok to have two boards with the same name, but only one defined a DEVICE_VARANT:=eMMC and the other doe snot define it at all? https://github.com/openwrt/openwrt/pull/2204/files Jul 08 11:52:13 someoner: i think you really have to start working on booting Linux Jul 08 11:52:27 someoner: if you don't get any output, enable earlyprintk Jul 08 11:52:32 that should help Jul 08 12:33:11 is there a 'hook' to run a scrip only the very first time the device boots? Jul 08 12:33:33 /etc/uci-defaults/ Jul 08 12:33:49 scripts in there are sources on boot and deleted when they exit with code 0 Jul 08 12:33:57 *sourced Jul 08 12:34:38 cool, thanks Jul 08 12:37:50 charolastra: (which means they run on every update too...) Jul 08 12:38:10 or after resetting the overlay (in case they're baked into the squashfs) Jul 08 12:38:22 but that is usually what people want when they ask for "first boot" Jul 08 12:39:28 by update you mean new firmware flashing? Jul 08 12:39:35 or the opkg update stuff? Jul 08 12:40:01 new firmware flashing will run uci-defaults again. Jul 08 13:29:15 Hauke, comment out that offending bit of mac80211, it seems broken and useless to me. Jul 08 13:48:05 Hauke: I just replied on PR#2204 Jul 08 13:49:21 aparcar: ^ Jul 08 13:52:03 jow: luci CI https://github.com/openwrt/luci/pull/2835 Jul 08 13:52:26 luaraneda: I have no position on how to handle that, however I think the argument is valid and no VARIANT for the sdcard version is required Jul 08 13:57:44 I'm looking at OpenWRT's image creation logic, and I don't see it using tools like fakeroot or pseudo Jul 08 13:58:00 so how does it set the correct permission/ownership of files for read-only filesystems (like squashfs) ? Jul 08 13:58:12 for example, "ping" needs to be setuid-root to be unusable by non-root users Jul 08 14:06:05 greearb_: was adhoc mode newly added to ath10k-ct? Jul 08 14:06:31 I am just wondering why I get this error message now and not with the older versions Jul 08 14:10:57 odhcpd seems to advertise RDNSS server on link-local addresses, like fe80::20d:b9ff:fe4c:71dd%eth0 Jul 08 14:11:04 and it breaks some clients that don't like it Jul 08 14:11:05 it is that ath10k started advertising min-gcd Jul 08 14:11:12 is it possible to disable RDNSS advertisement ? Jul 08 14:30:42 greearb_: ok Jul 08 15:55:00 Hi I flashed a build with luci in but when I try and log on to it it says bad gate way. Jul 08 15:55:10 Any know how to fix? Jul 08 15:55:15 anyone? Jul 08 15:55:50 clear your browser cache Jul 08 15:59:24 Tapper: Greetings! 2 questions for you. 1) Which linux distro do you use under WSL for building openwrt and 2) where should I send the recovered linksys ? Jul 08 15:59:51 DonkeyHotei Hi mate I have Jul 08 15:59:57 ldir I use ubuntu Jul 08 16:00:09 and I will send you a email with my address Jul 08 16:00:13 thanks Jul 08 16:00:42 excellent. Are you now able to get your builds working again 'cos I know you were having problems? Jul 08 16:00:53 what's the usi command to chang a lan IP? Jul 08 16:00:57 change Jul 08 16:00:59 * Jul 08 16:03:45 ldir sent Jul 08 16:03:57 arrived :-) Jul 08 16:04:31 ldir yeah mate builds are working again. I think it's one of the progams I have open in the back ground all the time that is messing things up. Jul 08 16:05:11 I think it's something to do with memory corruption! Jul 08 16:05:46 https://www.paypal.me/ldir - anyone can threaten me with money if they like :-) Jul 08 16:06:57 If I do a build with my twitter client, thunder bird, VLC and Instantbirunning it just brakis the build on all difrent packages it does not make sence. Jul 08 16:09:18 the command will be something like: uci set network.lan.ipaddr=192.168.1.1 ; uci commit Jul 08 16:13:29 this could all go horribly wrong - tries launching ubuntu within WSL on Windows 10 running in a VM under parallels on macos Jul 08 16:18:02 why......? Jul 08 16:19:06 aparcar[m]: thanks! Jul 08 16:19:10 a machine in a machine in a machine in a..... :-) Jul 08 16:20:14 WSL is not really a machine in a machine though, afaik ? Jul 08 16:21:04 well it's somewhere between magic and witchcraft as far as I'm concerned Jul 08 16:21:40 voodoo and black magic ;) Jul 08 16:24:27 jow: for what? Jul 08 16:25:06 aparcar[m]: ci Jul 08 16:37:27 jow: feel free to merge it Jul 08 16:43:19 OK people I am bac Jul 08 16:43:51 ldir I will sort you out in the morning if that's ok. Jul 08 16:44:21 of course - not sent it yet, will be tomorrow now Jul 08 16:44:26 I now have my dumb AP back on it's rite IP address, but when I go to luci it just says Bad Gateway Jul 08 16:44:26 The process did not produce any response Jul 08 16:44:48 I did a first boot to. so it's not old config files Jul 08 16:45:10 * Tapper nods at ldir Jul 08 16:45:51 I am just going to ruck a new duild and see if that fixes it Jul 08 16:45:59 ruck a new build* Jul 08 16:56:41 jow: can you extend my rights on luci so I can setup the ci? Jul 08 17:03:39 aparcar[m]: allegedly you're already a luci admin member Jul 08 17:46:19 Hi I did a new build but I am still getting the badgateway thing. Could I have a package missing? Jul 08 17:46:29 bad gateway Jul 08 17:47:36 did you set your gateway? Jul 08 17:48:02 Is anyone here actively using gdb inside of OpenWRT? I'm using the latest package in the feeds on the OpenWRT 19.07 trunk, and whenever I try to debug my binary gdb either hangs, tells me theres an illegal instruction, or once threw this error: Jul 08 17:48:03 In make menuconfig I did >luci>collections>luci and did the build Jul 08 17:48:13 fork-child.c:66: internal-error: void prefork_hook(const char*): Assertion `saved_ui == NULL' failed. Jul 08 17:48:20 oops sorry to talk over you Tapper Jul 08 17:49:05 Borromini My mane gateway is 192.168.1.1 and I set same on the dumb I am trying to reach Jul 08 17:50:02 Foxtrot it's OK bud. :) Jul 08 17:55:01 Foxtrot: I'm no expert, but I have used gdb before. On what platform are you trying to run it? Jul 08 18:05:58 Foxtrot, same gdb issue with 18.06 on mvebu Jul 08 18:06:57 *18.06.2 Jul 08 18:10:55 the only openwrt setup with working gdb i got so far is a virtual machine with x86/glibc build. not sure if it is arch or libc related. may be both.. Jul 08 18:13:36 jow: ping Jul 08 18:14:12 interesting, thanks max_g. For what it's worth, I'm not having that issue with 15.05, although that is too ancient to provide a real yardstick Jul 08 18:14:19 I'm targeting ar71xx Jul 08 18:14:37 cotequeiroz: ^ Jul 08 18:15:12 I'm used to using gdb, same binary is able to be debugged fine on my PC Jul 08 18:16:11 WTF I flashed and got rid of my configs and am still getting "Bad Gateway Jul 08 18:16:11 The process did not produce any response" when I try and log on to luci. Jul 08 18:16:31 running master is only for the brave Jul 08 18:17:27 Borromini lol I feel stupid! I bet it's a reel simple thing I have messed up Jul 08 18:18:05 I know I am going to flash 18.xx for now to see if it works Jul 08 18:18:32 why not be a good boy and build off 19.07 for a while huh Jul 08 18:18:39 help test that before it goes gold Jul 08 18:19:23 O shit I cant. There is no ath79 in 18.xx Jul 08 18:19:50 xD Jul 08 18:19:56 there's still ar71xx Jul 08 18:19:59 What is the command to switch to the 19.xx branch? Jul 08 18:20:09 git checkout openwrt-19.07 Jul 08 18:20:21 you are building your own no? Jul 08 18:20:23 "There is no ath79 in 18.xx" ? Jul 08 18:20:28 don't forget to enable luci Jul 08 18:20:36 Foxtrot: there isn't. just ar71xx Jul 08 18:20:37 K and when I run a build it will just build 19.07? Jul 08 18:20:45 Tapper: yes that's how it works mate. Jul 08 18:20:51 just clean out your buildroot Jul 08 18:20:53 sweet!:) Jul 08 18:21:15 i messed up another ath79 device >_> Jul 08 18:21:21 mangix: I was thinking if we really need to have a minimal build of libsodium. For one package (python-pynacl) it is necessary to have the full build of libsodium. I was thinking that I will add libsodium-full. What is your opinion? My colleague advised me to try to do variant as ffmpeg has. Jul 08 18:21:23 through no fault other than my own. Jul 08 18:21:35 The difference between the minimal and full build of libsodium is 20KB. Jul 08 18:25:39 max_g: Is your mvebu target using musl or uclibc? Jul 08 18:25:51 uh, glibc* sorry Jul 08 18:26:07 Foxtrot: I have used in in mips74kc, and I remember having trouble with MIPS16 code. Is your illegal instruction error inside the code being debugged, or gdb itself crashed? Jul 08 18:26:49 It's kind of hard to tell. I don't get Illegal Instruction if I run the binary normally. I only get it in gdb Jul 08 18:27:07 Regardless of whether it's my code or not though, I can't find out because gdb won't run properly Jul 08 18:27:08 It is probably performing cpu detection. Try to continue Jul 08 18:27:15 There is no prompt. Jul 08 18:27:18 It hangs Jul 08 18:28:10 Foxtrot, musl on mvebu Jul 08 18:28:10 Foxtrot: what program is throwing 'illegal instruction'? Jul 08 18:28:18 a custom binary Jul 08 18:29:39 max_g: i'm going to try building a glibc firmware and try again Jul 08 18:31:12 I've run gdb on mvebu with musl, but I compile my own, I haven't tried the standard package. Is the binary you're trying to debug compiled with the -mips16 flag? Jul 08 18:31:42 cotequeiroz: No. Jul 08 18:31:53 Again, I really don't think it's an issue with my program. Jul 08 18:32:18 gdb is throwing other errors itself too, and also explictly telling me not to trust it's output because it is failing asserts Jul 08 18:33:38 Pepe: ack Jul 08 18:34:25 <_abc_> Hello again. I asked yesterday, why tl-wr841n v14 appears supported in snapshot on the device support list, but does not appear in the snapshot under the relevant directory as an image? Is this an ommission? Failed to built? Other? Thanks. Jul 08 18:38:19 how can I enable early printk? shouldn't this option be under Kernel build options? Jul 08 18:39:41 someoner: for "standard" serial drivers you simply pass "earlycon" to the command line Jul 08 18:45:03 I'm trying to boot my image with CFE. How can I pass command line arguments? Normally I would do "boot -tftp -elf 192.168.0.9:/vmlinux-initramfs.elf". Here is the help: https://pastebin.com/PwftA3Ps Jul 08 19:08:06 OK, I'll just try to set CONFIG_EARLY_PRINTK Jul 08 19:20:11 aha! Jul 08 19:20:28 So, I built a firmware for my device using the glibc toolchain instead of the musl toolchain, and gdb works great now Jul 08 19:20:46 No crashes, no hangs, etc Jul 08 19:24:42 https://www.openwall.com/lists/musl/2018/04/06/1 Jul 08 19:24:50 this sounds like a very similar situation Jul 08 19:27:02 mangix: Ok. I will prepare a pull request tomorrow. :) Jul 08 20:10:49 is there some change that was made to the 18.06 branch over the past year that can prevent winxp clients from connecting to wifi altogether? Jul 08 20:15:31 Erm, people still use EOL'd products to get online? <=Vista are unsupported, short of costly backport agreements. Maybe one of the security fixes in the protocol that XP won't get updates to fix? CCMP support, or something else? Jul 08 20:47:04 pekster: ccmp is quite old and xp had it just fine Jul 08 20:48:02 DonkeyHotei: Including the 2017 vulnerability? You've clearly missed the point I was making.. Jul 08 20:48:20 Pretty sure XP didn't offer a patch for a fix published long-after (like, a decade after) it went EOL ;) Jul 08 20:48:22 maybe it has to do with the ath10k-ct-htt firmware Jul 08 20:49:18 * pekster shrugs. Could be. Point is you're trying to support stupidly outdated platforms. If Win7 & above works, tell your users/vendors/partners/staff to upgrade their security-nightmare of a platform, or migrate to one that won't cost to do so Jul 08 20:57:58 which change removed interoperability would be useful information Jul 08 21:27:17 DonkeyHotei: You're most likely to find that out by going through the changes yourself on Git. Jul 08 21:27:57 bisecting for this purpose is silly Jul 08 21:28:12 Lol, okay. Jul 08 21:28:53 i don't believe it's a bug in openwrt at this point Jul 08 21:48:02 jow: should I backport the circleci to the other branches as well? Jul 08 22:00:20 Great I got early printk working: https://pastebin.com/A3M022bx Jul 08 22:00:52 someoner: that looks like fun Jul 08 22:00:55 does this line: "ssb: Found chip with id 0x4710, rev 0x00 and package 0x00" look correct Jul 08 22:01:43 https://patchwork.ozlabs.org/patch/690401/ Jul 08 22:01:52 someone else has the same output and is getting further than you Jul 08 22:02:50 thx! should have googled it :D Jul 08 22:03:53 So most likely this is due to my system beeing a big endian system Jul 08 22:04:58 i don't see what that post has to do with this device Jul 08 22:05:12 The line that he copied and pasted? Jul 08 22:05:38 He asked if it looked right or not, I simply provided someone else with the same output. Jul 08 22:05:40 I was worried, that rev 0x00 and package 0x00 could be an indicator that something is wrong Jul 08 22:06:06 like my device not having an having an ssb bus Jul 08 22:06:25 the id is not 0x0000 Jul 08 22:06:37 DonkeyHotei: what? Jul 08 22:07:02 ssb bus is detected Jul 08 22:07:17 DonkeyHotei: looks like this Jul 08 22:07:42 I'll take a closer look at the ssb driver Jul 08 22:08:32 I'm still rather new to linux kernel development. Can I get some backtrace from the exception? Jul 08 22:10:31 usually yes, but debug symbols are off by default in openwrt because they grow the kernel tremendously Jul 08 22:11:16 OK, I'll try to enable them and hope it still fits Jul 08 22:13:20 oh there are multiple useful sounding options: symbol table information, tracing support, debug information, dynamic printk Jul 08 22:13:28 do I want all of those? Jul 08 22:14:23 someoner: you might want to let scan skip over core 5 as a quick hackfix to see if it is just this one causing issues (or limit the number of cores to 5 or so) Jul 08 22:16:53 KanjiMonster: why core 5? Would you mind to elaborate how you can see this? Jul 08 22:18:18 B8005FFC <- the 5 there. each core has 4k of registers (or 0x1000), so b8000000 ~ b8000fff is core 0, b8001000 ~ b8001fff is core 1 etc Jul 08 22:20:52 OK, I understand. I'll build a quick hack Jul 08 22:21:52 someoner: also 0x4710 / 0x00 / 0x00 is the fallback value when core 0 isn't a chipcommon core (see https://elixir.bootlin.com/linux/latest/source/drivers/ssb/scan.c#L316 ) Jul 08 22:22:43 can you maybe print the idhigh read in https://elixir.bootlin.com/linux/latest/source/drivers/ssb/scan.c#L290 ? Jul 08 22:25:08 sure Jul 08 22:27:12 the expected value for a chipcommon would be 0x4243800? (0x4243 == broadcom, 0x800 = chipcommon, ? = core revision) Jul 08 22:31:33 Hmm, idhi is 0? Jul 08 22:31:48 I used "printk("idhi %d", idhi);" after "idhi = scan_read32(bus, 0, SSB_IDHIGH);" Jul 08 22:32:10 hrm Jul 08 22:32:40 can you print the result of scan_read32(bus, 0, SSB_CHIPCO_CHIPID); ? Jul 08 22:35:41 also if you change the pr_debugs to pr_infos in the loop you should get more output (there is a way to not have touch these lines and get the output regardless, but I always forget how) Jul 08 22:39:01 scan_read32(bus, 0, SSB_CHIPCO_CHIPID); is never completed Jul 08 22:39:08 I guess it crashes there Jul 08 22:39:38 or the if condition never holds true Jul 08 22:39:49 (more likely) Jul 08 22:40:21 yeah, I meant doing it it unconditionally and printing the result ;) sorry for not being clear enough Jul 08 22:40:52 ups Jul 08 22:44:12 this is not looking good: scan_read32(bus, 0, SSB_CHIPCO_CHIPID) returns 1164050, but all ssb cores are unknown Jul 08 22:44:37 https://pastebin.com/kUu7egKt Jul 08 22:45:53 someoner: in general you want to use %x for printing register contents Jul 08 22:46:15 1164050 is 0x0011C312 Jul 08 22:47:01 thx for the hint! I only took a very quick look at the printk documentation. Jul 08 22:48:38 that's more like generic printf stuff ;p Jul 08 22:49:38 true Jul 08 22:49:52 312 at least looks a bit like it might be the right value/byte order, it being a 53*312* Jul 08 22:54:06 I'm sorry, but I cant follow you. Why is 312 good? Jul 08 22:55:05 chipid usually has the well chip id stored, so e.g. a bcm4716 would have xxxx4716 in the chipid register Jul 08 22:55:24 if the byte order is wrong, you would see 1647xxxx Jul 08 22:56:23 ok got it Jul 08 22:56:25 anyway, bed time for me, it almost 1 am here Jul 08 22:56:33 *it's Jul 08 22:56:44 same time zone, I should also go+# Jul 08 22:57:00 then good night Jul 08 22:57:17 KanjiMonster: good night and thanks for the help! Jul 08 22:57:39 (eventually I might need to start looking at that gpl tarball I guess ;P) Jul 08 22:57:56 it actually contains a zip Jul 08 22:58:27 (I already took a very quick look at it) Jul 08 22:59:10 Maybe I should try one build with early print and the other bus. But maybe I'll try this tomorrow. **** ENDING LOGGING AT Tue Jul 09 02:59:57 2019