| 08:41 | topi` | we are nearing October; is it possible to ship the Marvell 8040 board with a 8GB DIMM instead of the 4GB one? since it probably limits my ability to compile some of the nasty stuff, some huge projects won't even compile on 4GB :( |
| 08:43 | topi` | jnettlet: TI used to be good at upstreaming drivers, does it hold true for the Ti WiLink drivers as well? |
| 08:43 | topi` | I, for one, have no nostalgia about brcm4330, it's an old chip and it needs to die |
| 08:44 | jnettlet | topi`, the upstream drivers are good. There is a problem with the bluetooth and device tree since upstream doesn't like how the library that manages it is written. |
| 08:44 | topi` | BT is something we don't use (so far) |
| 08:45 | jnettlet | then the wifi drivers work great |
| 08:45 | topi` | BT is always problematic, I remember the Linaro guys took many months to get BT working on apq8064 even though WiFi was no issue |
| 08:48 | jnettlet | as for the DIMM. You can write to sales. Everything is 4GB by default, but you might be able to order it with none and then buy your own. It is a standard DIMM so you can put 16GB in it if you want. |
| 08:48 | topi` | I need a Linker that would not require >3GB of memory to be able to link even medium-sized projects, like the PyPy |
| 08:48 | topi` | maybe gold would be better |
| 08:48 | jnettlet | gold is better for building chromoium |
| 08:48 | jnettlet | chromium |
| 08:48 | topi` | yeah, that's another memory hog to ocmpile |
| 08:48 | topi` | lately I've tried building Servo for HB quad :) |
| 08:49 | topi` | it's using OpenGLES for almost everything |
| 08:49 | topi` | and even DOM parsing gets parallelized over 4 cores |
| 08:49 | jnettlet | nice. with the imx6 4x4 som you can just about get through Chromium with swap on an SSD |
| 08:49 | topi` | yeah :) unfortunately I have the 2GB HB |
| 08:50 | topi` | I tried getting a baremetal armhf server from scaleway.com but even with 1GB of swap it did not succeed :/ |
| 08:50 | jnettlet | the 8040 boards should be just about ready to go. The production run went through while I was down in Israel. Just need to do spot QA and such |
| 08:50 | topi` | good to hear there's some progress :) |
| 08:51 | jnettlet | there was a backlog getting chips from Marvell for production. |
| 08:51 | topi` | I admire how your company has a very open policy on reporting progress :) when I worked on Nokia, we couldn't say shit about the things we saw in the production line, progress, yields... |
| 08:52 | topi` | jnettlet: well it's also a first for Marvell themselves :) |
| 08:52 | topi` | I guess nobody is really using those 8xxx chips yet |
| 08:52 | jnettlet | since we are a smaller company it is much better to be as transparent as possible. |
| 08:52 | topi` | I totally agree |
| 08:53 | jnettlet | nope we are the first. this is a reference design |
| 08:53 | topi` | some of my ex-Nokia colleagues work now at Intel and I hear it's pretty tight there |
| 08:53 | topi` | they're very reluctant to talk about what they even do there :) |
| 08:53 | jnettlet | yeah my uncle used to work in chip design at Intel and he couldn't talk about work at all |
| 08:53 | jnettlet | I still don't know what chips he actually worked on :) |
| 08:54 | topi` | the chip profession runs in the family, eh? :) |
| 08:54 | topi` | also my uncle is an electical engineer, I wonder if it has something to do with things... |
| 08:54 | jnettlet | I am just a software guy. |
| 08:54 | wbx | i am compiling electron, gold seems to be required ;) |
| 08:54 | topi` | from all SW guys, you work pretty close to metal |
| 08:54 | topi` | most of SW guys just build software that runs under some VM, under something called tomcat, and virtualized to the death |
| 08:55 | topi` | do they know anything about metal? probably not |
| 08:56 | wbx | topi`: but they know how to write non efficient software, more cpu and more ram is no problem in their world ;) |
| 08:56 | topi` | :) |
| 08:56 | topi` | wbx: more cpu and more ram, or then switch to 'gold' :) |
| 08:56 | topi` | there are options |
| 08:57 | wbx | but ninja and gold isn't so great. still hours of compiling just when changing one config parameter... |
| 08:57 | topi` | I feel your pain |
| 08:57 | wbx | hate to get electron running on the embedded intel board. |
| 08:57 | topi` | I sincerely hope something comes out of Servo, it's a great project and much less convoluted than Chromium |
| 08:57 | wbx | i thought openjdk has a fuckuped buildsystem... |
| 08:57 | topi` | what kind of embedded intel board? |
| 08:58 | wbx | some kontron device |
| 08:58 | topi` | probably a Bay Trail :) |
| 08:58 | topi` | I ordered myself a Marvell 8040 board, should be plenty faster than any high end Atom, but we'll see |
| 08:59 | wbx | intel atom with two graphic ports, they want to show different websites with different esolutions. |
| 08:59 | wbx | resolutions.. |
| 08:59 | topi` | why not just bolt together two low-end Hummingboard Edges :) then you have 2x HDMI |
| 09:00 | topi` | the Kontron products are a bit on the expensive side |
| 09:00 | wbx | yeah, but I am only involved lately for the buildsystem, the hardware is already ordered. |
| 09:01 | suihkulokki | is there some "vendor tree" for the 8040 board available already? |
| 09:01 | wbx | jnettlet: you work for solidrun? |
| 09:01 | wbx | jnettlet: i will try the wifi today on my cubox-i |
| 09:02 | suihkulokk | 09:02 * suihkulokki thinks "the 8040" is a bit boring name - we need a nickname for the beast ;) |
| 09:02 | wbx | haven't tried further, the electron stuf keeps be busy. |
| 09:02 | topi` | wbx: how many boards do you ship+configure usually per project? |
| 09:02 | wbx | topi`: this project will be more then 3000 boards. |
| 09:02 | topi` | we're now starting to aim 100 devices for one customer... and lots of problems because we lack infrastructure. So I created a "small" python script that does deployment, install, config based on parameter files |
| 09:03 | topi` | previously, we just did everything by hand , editing a master clone image |
| 09:03 | topi` | which is, of course, stupid |
| 09:03 | topi` | but we never had the time to do things properly :( |
| 09:03 | wbx | topi`: last time we had 70 devices and done the firmware install manually by changing the sdcard. and a 10 port usb hub to write the cards ;) |
| 09:04 | wbx | topi`: a friend of mine has a nfs boot setup and a small dialog script, so that EVERYONE can configure and install devices for him ;) |
| 09:05 | topi` | nfs boot can also be useful |
| 09:05 | wbx | but the most important thing is always a dual-partition setup with the possibility to safely update. (like cisco is doing it.) |
| 09:05 | wbx | last time with u-boot, now with grub2. |
| 09:06 | topi` | but even now, we're happy to be able to ssh to a new box and run the superdeployment.py with proper parameters and it does everything |
| 09:06 | topi` | it's also versioned |
| 09:06 | topi` | an upgrade operation is just a "inspect checksums; uninstall; install" process |
| 09:07 | wbx | so incremental inplace updates of the firmware? |
| 09:07 | topi` | we have plans for a dual-partition setup, to switch between upgraded systems and facilitate failover, but we lack manpower to really do it |
| 09:08 | topi` | we ship just a modified debian/jessie with openvpn and some hardening |
| 09:08 | wbx | topi`: we are only two and managed it ;) with kexec and some sort of verify script after booting the new system |
| 09:08 | topi` | yeah I've looked into kexec, looks simple enough |
| 09:08 | topi` | it's harder to verify that it actually does the things |
| 09:08 | topi` | without a HW watchdog |
| 09:09 | wbx | topi`: debian is to big. the last firmware I created was under 20 megabyte and the pdf docs where more than 50 % ;) |
| 09:10 | wb | 09:10 * wbx is getting some breakfast |
| 09:16 | jnettlet | wbx, yes I work for SR. let me know about the wifi. |
| 09:18 | jnettlet | wbx, for safe updates I am working on integrating gpt partitioning and the swupd client/server that intel has built into the ostroproject |
| 09:18 | jnettlet | of course google also has their reliable update system they built for chrome and are integrating into Android now. |
| 09:26 | topi` | jnettlet: do you think it'd be possible to integrate GPT partitioning for the i.mx6 series as well? |
| 09:26 | topi` | the i.mx6 booting not being wholly compatible with GPT... |
| 09:27 | jnettlet | topi`, yes. |
| 09:27 | jnettlet | it just requires a custom tool set. Once I have everything patched I will try and get it pushed into upstream tools. |
| 09:29 | topi` | cool |
| 09:30 | jnettlet | topi`, here is an existing utility you can use if you want to try GPT partitioning on iMX6. https://github.com/dereulenspiegel/tableknife |
| 09:30 | topi` | it will streamline things for us, since every other piece of HW we use is using GPT |
| 09:30 | jnettlet | I just need to integrate this functionality into gdisk. |
| 09:31 | topi` | tableknife :D |
| 09:31 | topi` | it's like surgery |
| 09:32 | jnettlet | yeah. This lets you stick a binary in between the header and partition table. I want to specify the offset in GPT and then let you write a partition table like normal. Then you can flash any binary you want in between the two |
| 09:33 | jnettlet | for most SDHC cards I recommend starting your partitions at 4MB for performance reasons so there is lots of dead space there. |
| 09:34 | jnettlet | without getting this upstreamed, making sure you have a compatible disk partitioning tool would be a nightmare |
| 09:35 | topi` | the other complication might be the UBOOT |
| 09:35 | topi` | it needs to know where the rootfs is located :) |
| 09:36 | jnettlet | well u-boot just needs to be built with GPT partition support |
| 09:37 | jnettlet | that is just a config option that needs to be enabled |
| 09:42 | topi` | it seems this SR image is using a rather old uboot, 2013.10-rc4 |
| 09:43 | topi` | admittedly, this is installed from a year-old SR .dd |
| 09:43 | topi` | i'm lazy to upgrade ;) |
| 09:44 | topi` | is it possible to swap the HB Gate SoM for the new one with TI WiLink so we can test it? |
| 10:30 | wbx | jnettlet: http://pastebin.com/BKDiEXUU |
| 10:30 | wbx | kernel config: http://debug.openadk.org/cubox/cfg |
| 10:32 | wbx | dmesg: http://debug.openadk.org/cubox/dmesg.txt |
| 10:32 | wbx | do you see anything that might be a problem with this minimal kernel config? |
| 10:34 | jnettlet | so is the wifi module not being detected, or not loading the firmware? |
| 10:37 | wbx | it is not being detected. i have a working 4.1.x kernel and I see brcmfmac dmesg messages. |
| 10:37 | wbx | jnettlet: the 4.1.x was patched. with xbian patches. |
| 10:37 | wbx | jnettlet: the 4.7.3 is plain vanilla |
| 10:37 | wbx | jnettlet: is something in the dts files missing in plain vanilla? |
| 10:38 | jnettlet | for which board? |
| 10:38 | wbx | cubox-i ultra2 |
| 10:38 | wbx | the first cubox-i with two cores. |
| 10:38 | jnettlet | nope, the brcmfmac is loaded through the microsom dtsi file |
| 10:40 | jnettlet | there are no messages at all. also check dmesg for mmc messages |
| 10:40 | wbx | in this one? arch/arm/boot/dts/imx6qdl-microsom.dtsi . i see. |
| 10:40 | jnettlet | yes |
| 10:41 | wbx | any special options in MMC needed for Wifi? is wifi connected to the second mmc controller? |
| 10:41 | jnettlet | it is connected to the first. All the patches needed for wifi support should be upstsream |
| 10:43 | wbx | MMC_SDHCI_ESDHC_IMX is active. |
| 10:46 | wbx | oh, in the working kernel two mmc controllers are found, in the non-working just one |
| 10:47 | jnettlet | hmmm, I wonder if the regulator isn't being enabled |
| 10:49 | wbx | what is the name of the config symbol? |
| 10:52 | jnettlet | everything looks to be there, which I would expect |
| 10:52 | jnettlet | if you modprobe brcmfmac manually nothing happens? |
| 10:55 | wbx | i have no modprobe and no unloading active ;) |
| 10:56 | wbx | i can try, but i think the problem is earlier. only mmc0 in SDCARD mode found. on 4.1 mmc0 is SDIO and mmc1 SDCARD |
| 10:56 | jnettlet | wbx, how are you loading modules? the config showed they were built as modules |
| 10:56 | wbx | jnettlet: via insmod and a shell script |
| 10:56 | wbx | just really simple. |
| 10:57 | wbx | is there a defconfig for 4.7.x I can compare with? |
| 10:58 | jnettlet | arch/arm/configs/imx_v6_v7_defconfig |
| 10:59 | wbx | okay. CONFIG_PWRSEQ_EMMC=m |
| 10:59 | wbx | CONFIG_PWRSEQ_SIMPLE=m |
| 10:59 | wbx | they are missing. i try |
| 11:10 | wbx | jnettlet: GREAT man! it is working |
| 11:11 | jnettlet | excellent |
| 11:13 | wbx | shouldn't such symbols be selected or depends on? |
| 11:13 | wbx | jnettlet: you said yesterday everything gets autoselected ;) |
| 11:14 | jnettlet | wbx, not those options. They are added to the defconfig |
| 11:14 | jnettlet | they are not board or chip specific. |
| 11:14 | wbx | not? |
| 11:15 | jnettlet | so they are only needed for our microsom with wireless enabled. |
| 11:16 | wbx | yeah, so they are not enabled , because they are chip specific. |
| 11:16 | wbx | whatever. thanks a lot. now i can use latest OpenADK on my cubox-i again. |
| 11:16 | jnettlet | great |
| 11:17 | jnettlet | is it working well? |
| 11:20 | wbx | jnettlet: it is my main router for my lab. with 4.1 it worked very well. will see how 4.7.3 performs |
| 11:21 | wbx | jnettlet: i no longer use kodi on it. too much patching involved. |
| 11:21 | jnettlet | interesting. I wonder if we should get OpenADK running on our clearfog platforms |
| 11:22 | wbx | the intel stuff? |
| 11:22 | wbx | jnettlet: at least last time I got a 3000 devices customer convinced to use OpenADK :) |
| 11:23 | jnettlet | wbx, clearfog is our Armada platform. headless very fast router and storage |
| 11:23 | wbx | jnettlet: may be you could get me a device? last time I got a hummingboard1 from rabeeh/malte. and I personal bought a cubox-i |
| 11:23 | wbx | jnettlet: ah, nice. |
| 11:24 | wbx | solidrun got so many devices out, I didn't followed the last two years ;) |
| 11:24 | jnettlet | wbx, if you can port to it and wouldn't mind doing a short write up we can probably send you a clearfog base |
| 11:24 | jnettlet | all this has happened in the last year. it has been hectic |
| 11:24 | wbx | ah. should I send an email to you? |
| 11:25 | wbx | [11/11] LINK electron |
| 11:25 | wbx | make[6]: Nothing to be done for 'post-build'. |
| 11:25 | jnettlet | yeah, send me an email [email protected] |
| 11:25 | wbx | finally after two weeks hacking ;) |
| 11:26 | wbx | jnettlet: i can also prepare a buildroot patch later. I have a good connection to them. |
| 11:26 | jnettlet | wbx, I think we have a buildroot patch, but I will want it updated. I am in the middle of getting upstream u-boot 100% supported. |
| 11:27 | wbx | jnettlet: cool. cubox-i is working with upstream u-boot, too. that is really cool. |
| 11:27 | jnettlet | wbx, it is not working optimally though. there are known bugs and the memory timings are very sub-optimal |
| 11:29 | wbx | jnettlet: for cubox-i or clearfog? |
| 11:29 | jnettlet | cubox-i |
| 11:30 | vpeter | That's good news indeed. |