| 04:54 | Coburn | So I figured out what was going on with my Clearfog |
| 04:55 | Coburn | in kernel 4.4.61 there's a bug that seems to cause eth2 which is the port that is on the left of the SFP (?) connector to "disco" |
| 04:56 | Coburn | by that, the connection will go down, the driver will reinitialize the connection to 1gbps full flow control rx/tx, then the link will come up, then the link will come down, then it'll be reconfigured to flow control off. Rinse and repeat. |
| 04:56 | Coburn | So I tried to compile 4.9 from kernel.org. Looked peachy, went to boot the kernel... hangs at "Starting kernel...." |
| 04:57 | Coburn | Do I need to do some "set mainlinelinux=true" in uBoot stuff? Again, the SR wiki has zero documentation on how to boot a non-Marvell kernel. |
| 04:58 | Coburn | the 4.4.61 kernel I'm using is from the LEDE generic build branch, so i assume they use upstream or some SR supplied kernel. Don't know. |
| 07:43 | jnettlet | Coburn, the LEDE kernel is their own mix of patches on top of the 4.4.x LTS branch. We attempted to provide patches but they were refused. |
| 07:44 | jnettlet | if you are seeing a problem then it is best to report it on their forums. |
| 07:51 | Coburn | alright, i will do. |
| 07:51 | Coburn | What is the process of booting a non-marvell kernel on the ClearFog? |
| 07:51 | Coburn | like a mainline 4.9.x on the Armada 388? |
| 08:08 | jnettlet | I think most everything should work except SFP. You may need to check the device-tree name that u-boot is loading. |
| 08:09 | jnettlet | current head of LEDE is running 4.9.30, I just started investigating their latest changes. |
| 08:10 | jnettlet | what Distro are you trying to do bring up on? |
| 08:11 | Coburn | LEDE |
| 08:11 | Coburn | I'm using a kernel compiled on my ODROID-XU4 for the mvebu target |
| 08:11 | Coburn | and selecting the clearfog chipset, armanda 388, as part of the supported one |
| 08:11 | jnettlet | Coburn, I would just grab their HEAD commit and build that then. |
| 08:11 | jnettlet | if you want I have an image you can test. |
| 08:12 | Coburn | I'm using a dtb that was built with the non-working 4.9 kernel, it works on the 4.4.61 kernel that LEDE supplies |
| 08:12 | Coburn | HEAD commit? |
| 08:12 | jnettlet | their latest codebase |
| 08:12 | Coburn | I tried building my own LEDE image but honestly... it's way too complicated |
| 08:12 | Coburn | for me at least |
| 08:12 | Coburn | but if you have an image, sure i can test. |
| 08:12 | Coburn | I just want something that won't cause the lan ports to mess around |
| 08:13 | Coburn | since I use the port to the left of the SFP port as my connection to my Fixed Wireless modem box, it gets frustrating |
| 08:14 | jnettlet | sure. one second. |
| 08:14 | Coburn | ok. Is the image squashfs or ext4? |
| 08:14 | jnettlet | squashfs with overlay |
| 08:14 | jnettlet | I can build an ext4 one if you prefer. easy enough |
| 08:14 | Coburn | ext4 would be great. |
| 08:16 | Coburn | I prefer to have the less overlay stuff the better, I understand LEDE is for routers with limited memory, but it makes trying to figure out where things are being saved difficult - one image I used had actually been writing the overlay stuff to raw sectors on the uSD past the end of the rootfs squashfs partition! |
| 08:17 | jnettlet | yeah that is the LEDE team's idea. It is absolutely terrible since if you upgrade they basically need to put everything in RAM and just hope that you don't lose power |
| 08:20 | Coburn | Yep |
| 08:20 | Coburn | I remember using telnet to watch LEDE upgrade a TP-Link router with only 16MB of RAM |
| 08:21 | Coburn | It loaded the 3.8MB sysupgrade.bin into RAM, killing a few daemons in the process, then wrote it to flash and reboot |
| 08:22 | Coburn | Right now since I had the sh*ts with the clearfog I'm using a 512MB USB on a older Core i3 rig. Not power efficient by any means but hey, works :) |
| 08:24 | jnettlet | Give this a try. https://drive.google.com/open?id=0Bwd16WpGMX4LRmx6RzVKajA1Vk0 |
| 08:25 | Coburn | ok |
| 08:25 | Coburn | what modules are included? |
| 08:25 | Coburn | i need the modules for iwlwifi, the firmware and hostapd essentials |
| 08:27 | Coburn | I will test this image out tomorrow. No chance today |
| 08:52 | jnettlet | it doesn't have iwlwifi built in, but I can push them out. I would just test that it is working and then I will post a repository that you can run imagebuilder against so you can customize your own. |
| 08:54 | Coburn | ok |
| 08:55 | Coburn | Well, how about I test this first and if it works, I'll let you know and go from there. |
| 08:55 | Coburn | One question I have is that is it correct to have a MSI_Enable_Interrupt error on the mPCIe slot? |
| 08:56 | Coburn | like the MSI-X feature returns -7 or something when the kernel tries to enable that interrupt on the LEDE 4.4.61 one. I assume the kernel patches you got rejected fixed those |
| 08:56 | jnettlet | most likely depends on the card |
| 08:57 | Coburn | ah ok. the card i'm using is a Intel wifi... 7925 or something |
| 08:57 | Coburn | It does both N and AC... but I think it can only do one network instead of both |
| 08:57 | Coburn | 7265 rather |
| 08:57 | jnettlet | that should support MSI interrupts as best as I remember. I would test this image and report back. |
| 09:01 | Coburn | sure thing. |
| 09:01 | Coburn | Also, are you the "John" that appears on the SR github? |
| 09:02 | jnettlet | Jon yes |
| 09:03 | Coburn | awesome |
| 09:03 | Coburn | because I was looking through the git there and i noticed commits from you |
| 09:03 | Coburn | and I was like "wait... is this the same guy?" |
| 09:06 | jnettlet | yep that is I :) |
| 09:11 | vpeter | jnettlet: If wan port is fried on clearfog then to act as a router the only way is through sfp transceiver? |
| 09:12 | Coburn | vpeter: how did you fry it? |
| 09:12 | jnettlet | vpeter, you can get an SFP module, or using DSA use one of the switch ports as a WAN port. |
| 09:13 | jnettlet | the latter will reduce performance slightly but would still be fast enough for anything but gigabit uplinks |
| 09:14 | vpeter | Coburn: lightning |
| 09:14 | vpeter | DSA? |
| 09:15 | Coburn | vpeter: ouch. Does the clearfog have any surge protection built-in on the lan ports? |
| 09:15 | vpeter | No idea. But clearfof was connected on belkin protection :( |
| 09:16 | jnettlet | vpeter, that is Marvell's distributed switch architecture. It allows each switch port to be configured individually as its own interface. |
| 09:16 | vpeter | That's good then. |
| 09:16 | jnettlet | you would basically assign lan0 to be the want interface and then add lan1-5 to the lan bridge |
| 09:16 | jnettlet | s/want/wan/ |
| 09:17 | vpeter | Same as I did on old blue linksys :) |
| 09:17 | vpeter | Where can I find documentation for this? Switch datasheet? |
| 09:17 | jnettlet | but that will require a special build of LEDE |
| 09:18 | vpeter | I'm using my own build atm. And this would also mean I could use newer one :-) |
| 09:18 | jnettlet | the drivers are all upstreamed. You just need to revert LEDE's kernel patch that prefers their swconfig switch architecture |
| 09:18 | vpeter | So which lede should I use? official master or the one on sr github? |
| 09:19 | jnettlet | basically if you just revert that patch then all the ports will show up as individual interfaces in the gui and you just need to build your bridges and assign the ports to them. |
| 09:19 | jnettlet | you can use master right now. Artox's patches just got merged end of last week. |
| 09:20 | vpeter | Good. Then I will try this in next week. Thanks for the tip. |
| 09:20 | jnettlet | fyi you can get a SFP 1000Base-T module for around $15-20 |
| 09:20 | jnettlet | ebay is a great place to pick up old SFP modules |
| 09:20 | vpeter | + postage? You have any link at hand maybe? |
| 09:21 | vpeter | Is there any difference between them? |
| 09:21 | jnettlet | but you can find them on Amazon also. https://www.amazon.de/Ubiquiti-kompatibel-1000Base-T-Transceiver-Modul-RJ-45-Kupfer/dp/B01N1PBILH/ref=pd_day0_147_6?_encoding=UTF8&psc=1&refRID=S8AD34QZK121210VP8B0 |
| 09:21 | jnettlet | https://www.amazon.de/Generic-dgs-712-F%C3%BCr-1000Base-T-Ethernet-RJ45-Stecker/dp/B01585WT32/ref=pd_sbs_23_5?_encoding=UTF8&psc=1&refRID=98JBKV8BCA7YJ6PBKSVG |
| 09:22 | vpeter | I will ask a friend for a test. I would say he should have it few if he is running Cisco academy. |
| 09:22 | jnettlet | most likely yes. |
| 09:22 | jnettlet | regardless, any questions just let me know |
| 09:23 | vpeter | I think I will buy one of those anyway. To get one more port if nothing more. |
| 09:23 | vpeter | Is FTP visible in LEDE without any modificications? |
| 09:25 | jnettlet | not sure. I know ssh is. |
| 09:34 | Coburn | SFTP should work but it kicked me out when i tried logging in |
| 09:46 | jnettlet | Coburn, give scp a try |
| 09:49 | Coburn | will do tomorrow |
| 10:06 | Coburn | jnettlet: does your image use the official sr uboot or the lede one? |
| 10:16 | jnettlet | Coburn, this is just a basic LEDE build. I am testing what is upstream vs what we have. |
| 10:32 | Coburn | jnettlet: It booted, that's a start |
| 10:35 | jnettlet | \0/ |
| 10:38 | Coburn | 02:00.0 Network controller: Intel Corporation Wireless 7260 (rev bb) |
| 10:38 | Coburn | It sees the adapter! |
| 10:38 | Coburn | And the ports aren't discoing |
| 10:40 | jnettlet | sounds promising. |
| 10:50 | Coburn | [ 1483.522168] mvneta f1070000.ethernet eth0: Link is Up - 1Gbps/Full - flow control rx/tx |
| 10:50 | Coburn | So if eth0 was eth2 in the 4.4.63 kernel |
| 10:50 | Coburn | that's the discoing port |
| 10:50 | Coburn | seems to behaving itself. I'll have to do further testing tomorrow |
| 10:52 | jnettlet | Coburn, here is the iwlwifi kmod. https://drive.google.com/open?id=0Bwd16WpGMX4LRnhQcmlvUGpfcG8 |
| 10:53 | Coburn | only issue with that is doesn't it need the 80211 stuff as well? |
| 10:58 | jnettlet | that should already be installed |
| 10:58 | jnettlet | I only build the ath10k modules by default on my builds |
| 10:58 | Coburn | ah ok |
| 11:00 | Coburn | [ 6.147494] pci 0000:00:03.0: enabling device (0140 -> 0142) |
| 11:01 | Coburn | [ 6.153183] iwlwifi 0000:02:00.0: enabling device (0140 -> 0142) |
| 11:01 | Coburn | [ 6.185771] iwlmvm: Unknown symbol backport_dev_coredumpsg (err 0) |
| 11:01 | Coburn | [ 6.351483] kmodloader: 1 module could not be probed |
| 11:01 | jnettlet | oh missing iwlmvm...let me see if that was created |
| 11:01 | Coburn | [ 6.356489] kmodloader: - iwlmvm - 0 |
| 11:02 | Coburn | no errors about MSI interrupts |
| 11:02 | Coburn | that's good |
| 11:02 | jnettlet | hmmm it built the module just didn't seem to package it |
| 11:06 | Coburn | manually packaging it? |
| 11:06 | jnettlet | https://drive.google.com/file/d/0Bwd16WpGMX4LTldPY0ZuOC1KS3M/view?usp=sharing |
| 11:07 | jnettlet | just the standalone module |
| 11:07 | Coburn | ok |
| 11:23 | vpeter | jnettlet: seems previously I made a typo - I was asking if sfp module is visible in lede and not ftp service :) |
| 11:24 | jnettlet | ah. as of the patches accepted upstream from Artox, yes it is configured as part of the WAN bridge. |
| 11:25 | vpeter | perfect |
| 11:25 | vpeter | Will do both then (buy adapter and build new lede). |
| 11:29 | vpeter | jnettlet: last think - is there any lede master config at hand? |
| 11:29 | vpeter | Last time Artox gave me one to start with. It was more easy. |
| 11:29 | Coburn | jnettlet: that module already exists |
| 11:29 | Coburn | but it's encountering an error |
| 11:30 | Coburn | [ 1764.435209] iwlmvm: Unknown symbol backport_dev_coredumpsg (err 0) |
| 11:31 | jnettlet | looks like LEDE doesn't have the proper kernel configs to support that. |
| 11:32 | Coburn | bugger. |
| 11:32 | jnettlet | I am looking to see what provides that. |
| 11:38 | jnettlet | oh annoying. This is the same bug I had to patch in yocto with the new gcc6 |
| 11:39 | jnettlet | the module.symvers crc is generated as 0x00000000 which is read as symbol disabled, however it is not. |
| 11:40 | jnettlet | reverted this commit. https://github.com/SolidRun/linux-fslc/commit/f3e761c7e63326e2d7d35c33f6028adecc3504c6 |
| 11:41 | Coburn | so waht do i need to do? |
| 11:45 | Artox | vpeter: there is the dotconfig I used to build images before patches were merged |
| 11:45 | Artox | https://repo.solid-build.xyz/lede/targets/mvebu/generic/config.seed |
| 11:46 | Artox | Or you can use Ledes snapshots configs |
| 11:46 | Artox | https://downloads.lede-project.org/snapshots/targets/mvebu/generic/config.seed |
| 11:47 | Artox | Also, when you build Lede for the clearfog there is a caveat |
| 11:47 | Artox | DON't enable early printk, in its default config it crashes the kernel before printing *any* messages |
| 11:48 | jnettlet | Coburn, just hold on for now I guess. I will need to sort this out for LEDE. |
| 11:48 | jnettlet | this bug seems to have resurfaced due to gcc 6 |
| 11:49 | Coburn | jnettlet: no worries. I am glad that the disco port has stopped having a party. |
| 11:50 | Coburn | I will configure it with my NBN Fixed Wireless box and test the pppoe side of things |
| 11:51 | jnettlet | and I have the SQM packages built into the image by default, so you can tune the QOS to minimize buffer bloat. |
| 11:51 | jnettlet | https://wiki.openwrt.org/doc/howto/sqm |
| 11:53 | Coburn | ou've restored some of my trust in the clearfog that I wasn't just given a platform without any support |
| 11:54 | Coburn | LEDE has been pretty OK, although I keep forgetting that I must hit save and apply after firewall rules. Otherwise changes don't stick. |
| 11:55 | Coburn | Not the best user interface but hey, on this size of $1 million, i'll take it |
| 11:55 | Coburn | side* |
| 11:56 | Coburn | do you have a buildbot or should i come back in a week and ask if you have any updates? |
| 12:25 | Coburn | if i wanted to build a build like yours, jnettlet, what's involved? |
| 13:08 | vpeter | Artox: thanks. |
| 13:48 | Coburn | Guess I'll log off, and rejoin in the morning. Night guys. |
| 14:30 | vpeter | jnettlet: in lede kernel 4.9 is the issue with lan6 port resolved? In 4.4 there were this notice with #if 0 |
| 14:30 | vpeter | /* use lan6 only as 1000Base-FD and no use of ethtool/mii-tool */ |
| 14:30 | vpeter | I assume now all 6 ports will work out of the box? |
| 14:50 | vpeter | Also how to switch pcie to msata with u-boot in lede? |