**** BEGIN LOGGING AT Thu Mar 14 02:59:57 2019 Mar 14 08:14:18 Hauke: BTW I was seeing some new weird behavior in my QEMU x86_64 testing images (processes using openssl blocked for a long time in call to getrandom()) and I've bisected it overnight to this commit be3892284ca77a69615351b106b8dfbadad728c4 Mar 14 08:15:19 Hauke: bisect log http://paste.ubuntu.com/p/GXtP2fXTMh/ Mar 14 08:32:53 morning Mar 14 09:05:46 Hauke: I've added all the details to the PR as well https://github.com/openwrt/openwrt/pull/965#issuecomment-472762625 Mar 14 09:40:11 mangix: Yes? Mar 14 09:55:09 Hauke: this looks like a very good idea! (regarding https://developers.google.com/season-of-docs/ ) Mar 14 10:14:51 Rene__: did you ever get the second GMAC working on mt7621? Mar 14 10:17:43 Yes i did Mar 14 10:19:19 This is what I have running in my UBNT ER-X-SFP: 2nd GMAC connected to P5 of the switch and P5 is connected with P0/P4 Mar 14 10:20:48 Still work in progress but it works https://github.com/vDorst/linux-1/commit/40725f9499932e6d582f6866df7367f7a12f4a39 Mar 14 10:21:37 5.0.1 kernel with new next mtk_eth_soc and mt7530 driver. Both conveted to phylink API. Mar 14 10:23:37 P5 to P0/P4 trick works on every mt7621A device BTW. Mar 14 10:28:28 I also findout that UBNT screwup with the external phy (SFP port). They wired the phy as MAC so TX and RX lines are swap. Mar 14 10:29:16 This means I can't connect the external phy with 2nd GMAC. Mar 14 10:43:13 mangix: what for device do you have? Mar 14 11:35:07 stintel: seen this on Phoronix? https://www.phoronix.com/scan.php?page=article&item=raspberrypi-linux-419&num=1 Mar 14 12:35:37 Rene__: GnuBee PC2 Mar 14 12:58:02 mangix: great you have a external phy. Mar 14 13:06:06 greearb_: Just so you know, the latest beta FW you have for qca988x crashes repeatedly with https://paste.ubuntu.com/p/7xkRWPZcTR/. Mar 14 13:09:35 mangix: you can try this branch https://github.com/vDorst/linux-1/tree/v5.0.1-mt762x-phylink-gmac2-ephy Mar 14 13:20:50 well, I no longer use the device. But I did get an email from Neil Brown, who's trying to get it working. Mar 14 13:21:45 hhaving said that, I am curious what the performance is of the second port as it's not behind a switch Mar 14 13:24:01 mangix: the same Mar 14 13:25:19 I see with iperf3 from/to the device ~930mbit down and ~475mbit up on both GMACs or any PORT Mar 14 13:26:23 interesting... on all my mt7621 devices, iperf speed fluctuates a lot unless I tweak the SMP affinity Mar 14 13:26:29 475mbit is a bug but not sure where it comes from. Mar 14 13:26:40 maybe the upstream driver is better Mar 14 13:27:13 i think an SMP affinity of 4 gives me the best results Mar 14 13:28:27 Rene__: try playing with disabling tso and whatnot with ethtool Mar 14 13:30:14 ok, i'll keep that in mind. For now my goal is convert mtk_eth_soc and mt7530 to phylink API and try to get it upstream Mar 14 13:32:12 phylink help me to support my SFP cage properly. Mar 14 13:33:32 that's great actually. Mar 14 13:35:09 btw, how do you run this kernel on your device without something like openwrt? Mar 14 13:35:53 Kernel is missing NAND and I2C driver at the moment. Mar 14 13:36:06 i2c of openwrt works great so that is fixed. Mar 14 13:37:45 NAND is a issue at the moment. Patches don't apply on the +4.19 kernel. I know that Greg Ungerer is working on that. Mar 14 13:38:48 What I did is make a openwrt cpio image. At kernel build include this in the uimage. Then tftp boot it. Mar 14 13:39:34 Not the best setup but it works. Mar 14 13:39:36 mamarley, thanks, will decode that crash shortly Mar 14 13:40:51 Thanks! Mar 14 13:43:07 Rene__: well, i2c and nand are not present on my platform :) Mar 14 13:50:27 mangix: that makes it easier :) Mar 14 13:58:32 mangix: I sended Neil an email but my work. I hope it is useful. Mar 14 14:13:15 mamarley, please try the firmware I just uploaded Mar 14 15:18:28 greearb_: That one doesn't crash, thanks! Mar 14 15:20:54 no problem, main thing you might care about is that it should enable BT-COEX on 2.4Ghz-only radios....a bug reporter said that feature helped them, but I have no way to test. Mar 14 15:22:05 greearb_: My device only has 5gHz radios using ath10k, the 2.4gHz radio uses ath9k. Mar 14 15:27:22 nod Mar 14 15:47:50 any idea if i need to do more than call ath79_register_usb to get usb on QCA9531 ? Mar 14 15:47:58 it doesn't come up Mar 14 15:48:50 at least /sys/kernel/debug/usb/devices is empty Mar 14 16:03:58 thats the closed source firmware output: https://gist.githubusercontent.com/aep/871978a5a7b671f24885519058723bca/raw/ae3d0f8e911b2bc16a003fc75e7b64b125ce2af0/gistfile1.txt Mar 14 16:04:17 ehci-platform ehci-platform.0: new USB bus registered, assigned bus number 1 Mar 14 16:07:27 aep: ar71xx is dead, almost in maintenance mode only, so you should use ath79 (target/linux/ath79) which is using DTS Mar 14 16:08:51 nah. i cant even get ethernet up with dts Mar 14 16:11:59 I would check GPIO as well, sometimes you just need to toggle USB power via some kind of regulator as well Mar 14 16:12:44 power is up Mar 14 16:13:42 well at least it outputs power. Mar 14 16:15:52 there's a usb hub on this thing. GL850G Mar 14 16:16:26 i was expecting that to just be detected once the usb root is up tho Mar 14 16:20:51 well, just find the datasheet and find out the VCC pin and check if there's power Mar 14 16:21:34 yep, great idea Mar 14 16:22:34 although i'm reasonably sure the board init doesnt work Mar 14 16:22:44 because even if it doesnt detect the switch, it should have a root Mar 14 16:31:11 yeah, switch has power Mar 14 16:50:27 heh. its the wrong chip Mar 14 16:50:56 vendor sheet says QCA9531 but /proc/cpuinfo says QCA9533 Mar 14 16:51:18 on the other hand, does that even matter? Mar 14 17:25:40 * ldir knows this isn't the correct way to develop but iteratively types stuff to stop the compiler complaining anyway Mar 14 17:26:52 LOL, yeah -- some of the wonky #define re-casting I've seen is to prevent compiler warnings, rather than make the code more understandable Mar 14 17:29:17 oh no, this is much worse.... this is me getting basic syntax wrong - the copy & pasted stuff is mostly ok, anything I write is a car crash that I fix up with the aid of the compiler :-) Mar 14 17:30:06 now if every could cross their fingers this might actually go through without any warning after the 27th attempt :-) Mar 14 17:32:30 make that 28 Mar 14 17:34:38 I've been pretty happy with CLion as an IDE. Unfortunately it's a paid product so I only get to use it 90 days at a time when they have their pre-release, testing versions out Mar 14 17:35:42 The auto-completion doesn't work too well without a CMake structure, but its on-the-fly syntax and logic checking has saved me a lot of time Mar 14 17:37:50 Oh this is just me fiddling with the kernel, somewhere I really shouldn't be let loose - enough monkeys.... Mar 14 17:45:20 yikes, at last it's gone through without even a warning! https://github.com/ldir-EDB0/linux-stable/commits/14conndscp Mar 14 17:48:13 :thumbs-up: Mar 14 17:49:53 uuuh, ath79_register_usb isnt even implemented for QCA953x https://elixir.bootlin.com/linux/v4.3/source/arch/mips/ath79/dev-usb.c#L224 Mar 14 17:49:57 what do i do now? Mar 14 17:52:30 it's a BUG() ;) Mar 14 18:13:33 Is there a non-GPL version of U-Boot that OEMs (Linksys, in particular) use on QCA SoCs that would allow them to "decline" to provide Corresponding Source? Mar 14 18:14:10 `U-Boot 2012.07 [Chaos Calmer 15.05.1,r35193] (Nov 02 2017 - 16:33:09)` suggests that it is GPL to me Mar 14 18:17:33 unlikely Mar 14 18:17:47 they just dont care Mar 14 18:21:37 Thanks, lets me make a stronger statement in the request. Mar 14 18:35:15 ffs, i found it. someone intentionally broke QCA953x usb host init for some weird slave mode init hack. Mar 14 18:35:21 WHY Mar 14 18:53:02 aep: if it was within openwrt, the commit logs should tell you ;) Mar 14 18:54:00 nah, file got moved Mar 14 18:54:31 but the code is obvious. someone wanted OTG on a board that doesnt have OTG Mar 14 18:54:39 and broke it for everyone else Mar 14 18:57:00 shouldn't be too hard to trace through file moves Mar 14 18:57:04 don't you just love it when that happens? Mar 14 19:01:27 "Oh, this is cool! Only a couple lines. Sure, commit it to mainline." Mar 14 19:04:22 s/everyone/me/ Mar 14 19:04:29 pretty sure the author wasnt that stupid :P Mar 14 19:05:10 where is it? Mar 14 19:05:47 target/linux/ar71xx/patches-4.14/920-usb-chipidea-AR933x-platform-support.patch Mar 14 19:08:39 is it also in ath79, or just ar71xx? Mar 14 19:11:24 probably squashed Mar 14 19:11:29 since DTS works different Mar 14 19:13:15 blogic: ping Mar 14 19:13:41 aep: https://github.com/openwrt/openwrt/commits/318e19ba6755105bb6cc19937d8fff26cbd2cc6f/target/linux/ar71xx/patches-4.9/920-usb-chipidea-AR933x-platform-support.patch Mar 14 19:17:56 also, https://github.com/openwrt/openwrt/commits/7bbf4117c6fe4b764d9d7c62fb2bcf6dd93bff2c/target/linux/ar71xx/patches-4.4/920-usb-chipidea-AR933x-platform-support.patch Mar 14 19:21:23 thanks Mar 14 19:21:31 apparantly it reads from this bootstrap thing Mar 14 19:21:34 which is zero for me Mar 14 19:21:39 i wonder what's supposed to set it Mar 14 19:24:05 ynezz: regarding your openssl problem, you should probably contact Eneas U de Queiroz he takes good care of openssl Mar 14 19:25:00 zorun: this season of docs looks interesting, but I do not have the time to organize this from OpenWrt side, if someone else would like to take over this would be nice Mar 14 20:08:30 hm is there something like uqmi but for GPS? Mar 14 20:10:32 http://aprs.gids.nl/nmea/ Mar 14 20:11:25 yeah those are obvious, but the modem wants some initial AT commands to start sending them Mar 14 20:11:38 was just wondering if there's an openwrt built in thing Mar 14 20:12:32 hrm, my qca9531 dwvices havw working usb Mar 14 20:14:12 Can confirm that on a GL.iNet AR300M-Lite as well, on the ath79 target Mar 14 20:15:04 aep: I'm using this with my quectel modems https://paste.ubuntu.com/p/nSg6VscGMg/ Mar 14 20:16:05 it's called ugps https://git.openwrt.org/?p=project/ugps.git;a=summary Mar 14 20:16:08 mangix, jeffsf_ : yep, thanks. i suspect its because this device doesnt use uboot so the bootstrap thingy works different Mar 14 20:16:31 ynezz: nice! thanks. Mar 14 20:18:15 ynezz: how do you make sure to hit the right tty? since they're random and no udev rules Mar 14 20:18:57 I know how to cross fingers :) Mar 14 20:19:37 heh Mar 14 20:20:01 I've never had this kind of issue with this modems Mar 14 20:20:20 aye. Mar 14 20:21:49 but you can of course add some sysfs based USB vid/pid handling instead of this hardcoded device path Mar 14 20:22:22 hm, the vlan on my dsl0 interface stopped honouring the 'mtu 1508' config Mar 14 20:22:33 will debug that when I'm actually on the correct side of it... Mar 14 20:22:37 ynezz: btw, which modem do you use? i have a uc20 and its missing /dev/tty* on openwrt Mar 14 20:22:52 quectel ec20 Mar 14 20:23:01 hm Mar 14 20:23:20 could you do me a favor and check dmesg which driver produced /dev/tty? Mar 14 20:23:25 or /dev/acm maybe? Mar 14 20:24:28 drivers/usb/serial/qcserial.c Mar 14 20:24:41 thanks Mar 14 20:26:00 odd, i have that loaded but it doesnt give me any ttys. meh Mar 14 20:26:53 what vid/pid is uc20? Mar 14 20:31:27 05c6:9003 Mar 14 20:32:01 works fine on desktop linux. poking around the driver now Mar 14 20:32:44 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/usb/serial/option.c?id=6e9f44eaaef0df7b846e9316fa9ca72a02025d44 Mar 14 20:33:06 oooh Mar 14 20:33:10 thanks Mar 14 20:42:40 flawless Mar 14 20:44:04 aep: neither does mine Mar 14 20:44:45 aye Mar 14 20:47:38 ynezz:oh btw, could you pastebin your uci qmi config? i cant figure out how to make netifd dial for me, since the docs are only for the web ui thing Mar 14 20:52:03 https://paste.ubuntu.com/p/2BPS72R9zJ/ Mar 14 20:53:22 thanks! Mar 14 20:56:35 https://openwrt.org/docs/guide-user/network/wan/wwan/ltedongle#qmi_protocol_configuration Mar 14 20:58:21 yeah. its not clear where to apply those tho Mar 14 20:58:27 i added it to interface wwan0 Mar 14 20:58:47 you're apparantly creating a new interface **** ENDING LOGGING AT Fri Mar 15 02:59:57 2019