| 07:30 | jnettlet | This is exactly the reason systemd annoys me. By default SysvInit would set loglevel=3 to reduce logging output. Well we aren't going to do that, we just want to print everything out and it is up to someone else to change the loglevel so it isn't completely unusable. |
| 07:31 | jnettlet | Changing default behaviors for no earthly reason other than to be different |
| 08:13 | jnettlet | rabeeh, interesting. The uhs-1 problems appear to be a timing problem. enabling mmc_debug causes the problems to go away. |
| 08:29 | jnettlet | oh look at that. we are constantly getting card_removed, card_inserted interrupts. |
| 08:29 | jnettlet | sdhci [sdhci_irq()]: *** mmc0 got interrupt: 0x00000082 |
| 08:29 | jnettlet | sdhci [sdhci_irq()]: *** mmc0 got interrupt: 0x00000041 |
| 08:57 | jnettlet | jmontleo, I am slowly narrowing in our sdhc card problems |
| 09:47 | MikeSeth | I am going to pick up my cubox! |
| 12:38 | jnettlet | tomlohave, ping |
| 13:07 | jnettle | 13:07 * jnettlet guesses he will have to build his own geexbox image |
| 13:08 | cbxbiker61_ | jnettlet, i'm prepping an image right at the moment that is ext2-f2fs-swap |
| 13:08 | jnettlet | cbxbiker61_, well there is a problem with the framebuffer detection. It won't work with device-tree enabled kernels |
| 13:08 | jnettlet | it is actually a problem with all of geexbox not just the FSL one. |
| 13:09 | cbxbiker61_ | ic, i'm sticking with 3.0.101 for right now, i'll be happy to go current kernel's once those are working well |
| 13:11 | jnettlet | I just tested the last nightly on the HB1 and it was significantly slower than my build of XBMC running on X11 with the latest kernel |
| 14:11 | jmontleo | jnettlet, cool - if at some point you want another person to test a patch I'm happy to do so. |
| 14:15 | _rmk | 14:15 * _rmk_ is coming to the conclusion that device tree on arm is a massive backwards step... since getting something to boot with dt seems to be an almost impossible task (the only platform I have here which seems capable of DT booting is the HB) |
| 14:16 | _rmk_ | I'm about to put the CT9x4 card (4 CPU Cortex A9) back in the Versatile Express because I can't get the Cortex-A15/big.Little card to boot properly with DT (it spews errors) |
| 14:17 | _rmk_ | and... getting a uboot for my OMAP4 board which can deal with DT over the network is proving to be totally impossible |
| 14:17 | _rmk_ | so basically all my bootable platforms with the exception of HB are all non-DT. |
| 14:45 | jnettlet | _rmk_, the XO-4 can boot device-tree. OFW even creates and exports the device-tree :-) |
| 14:46 | jnettlet | _rmk_, has big.LITTLE made it into mainline yet or is it just booting the big cores? |
| 14:48 | jnettlet | jmontleo, if you want the quickest fix the booting problem add #define DEBUG to the top of drivers/mmc/host/sdhci.c |
| 14:49 | jmontle | 14:49 * jmontleo gives it a try |
| 14:49 | jnettlet | it seems to be the most reliable fix I have found so far. But since it gets rid of the problem it doesn't really help me |
| 14:51 | jnettlet | mostly it is just printing out the interrupts received and slowing down the interrupt processing just a bit fixes the race, or whatever is going bad. |
| 14:52 | jnettlet | I am trying to figure out if it is normal for those card removed, inserted interrupts to constantly happen. I think it may be due to power-saving but haven't verified it yet. |
| 14:54 | _rmk_ | jnettlet: what I object to is not being able to get to a bootable kernel on DT based platforms - with it sometimes taking months to get sorted for whatever reason, while others seemingly are able to boot it |
| 14:56 | jnettlet | +1 |
| 14:56 | _rmk_ | in the case of the TC2 tile, after three weeks, it finally is able to mount its rootfs, but only through finding the right combination of random configuration symbols which don't really mean all that much (who'd have thought that turning off MCPM would let it get past probing the flash when MCPM has nothing to do with flash!) |
| 14:57 | _rmk_ | but... because there's something (god knows what) wrong with the CLCD driver, it causes X to fail, and the boot process stops at that point without progressing to a login prompt. |
| 14:57 | _rmk_ | and... you can't telnet into it because telnetd fails to start because /dev/null appears to be missing (but why is it missing?) |
| 14:58 | jnettlet | wow that does sound like a mess. |
| 14:58 | _rmk_ | but how can an older DT based kernel boot this same filesystem without problem |
| 14:59 | jnettlet | well because it is starting to get used more there is a lot of device-tree churn. |
| 14:59 | jnettlet | I believe there are a few device-tree patches in our OLPC kernel that need to go upstream as well. |
| 15:00 | _rmk_ | then there's my OMAP4430 board which I've ended up having to disable because there seems to be no uboot which can do both network booting and DT. |
| 15:00 | jnettle | 15:00 * jnettlet raises an eyebrow really? |
| 15:01 | _rmk_ | yep - the previous uboot would allow loading multiple files but had no bootz support (it was legacy only) |
| 15:01 | _rmk_ | that was a special version from TI which included network support |
| 15:01 | _rmk_ | mainline uboot still (many years on) has no support for network on this platform |
| 15:02 | jnettlet | do you have a copy of u-boot with network support working? |
| 15:02 | _rmk_ | the latest special version from TI can only do network stuff when you run "dhcp" - if you try to tftp manually the network stuff just breaks |
| 15:03 | _rmk_ | I only have binary uboots for this platform |
| 15:03 | jnettlet | which board is it? |
| 15:03 | _rmk_ | OMAP4430 SDP |
| 15:04 | _rmk_ | it's slightly complicated by uboot needing to be in two parts - "MLO" and uboot itself. MLO is basically a chain-loader for uboot |
| 15:05 | jnettlet | yeah it is using an SPL build which is what cubox-i will use |
| 15:09 | MikeSeth | I oughta ask |
| 15:09 | MikeSeth | d'ya get paid for any of this by anybody? |
| 15:14 | MarcusVinter | I also wondered that. |
| 15:14 | jnettlet | some of us do make an income from hardware development yes. |
| 15:14 | jnettlet | _rmk_, what ethernet driver does that board use in the kernel? |
| 15:14 | jnettlet | KS8851? |
| 15:15 | jnettlet | jmontleo, make any difference on your board? |
| 15:16 | jmontleo | still building |
| 15:20 | _rmk_ | yep, KS8851 |
| 15:20 | _rmk_ | MikeSeth: thankfully for ARM and TI stuff, they're paying :) |
| 15:21 | jnettlet | _rmk_, someone is still paying for TI? |
| 15:22 | MikeSeth | _rmk_: I've been contemplating walking away from enterprisey software for a looooong time |
| 15:24 | _rmk_ | just about - there's not many people left. |
| 16:10 | jmontleo | jnettlet, no difference on my cubox-i; still get error -110 transferring data |
| 16:11 | jnettlet | hmmm |
| 16:11 | jnettlet | jmontleo, okay thanks for testing. |
| 16:11 | jmontleo | you're welcome |
| 16:37 | MikeSeth | what's the ethernet chip on cubox-i? |
| 16:43 | _rmk_ | mikeseth: it's built-in to the imx6, but with an external AR8035 phy |
| 16:44 | MikeSeth | _rmk_: supports vlan tagging right? |
| 16:49 | _rmk_ | I believe it does |
| 17:02 | MikeSeth | wiki specs says i1 doesn't support esata or serial, but I can clearly see an esata port and a microusb one |
| 17:02 | MikeSeth | an I missing something? |
| 17:47 | jnettlet | okay who knows what creates the /dev/video16 17 18 devices? |
| 17:58 | MikeSeth | won't you see in udev logs? |
| 18:01 | jnettlet | well the devices aren't being created, so no |
| 18:03 | MikeSeth | do you mean they're missing or that they're static inodes? IIRC all /dev/video is v4l |
| 18:03 | MikeSeth | ...I am probably less qualified to have this discussion than you are |
| 18:04 | jnettlet | yes v4l, but I couldn't tell what. It is a dma v4l device node I was missing in device-tree |
| 18:49 | jnettlet | hste, I have finally succeeded in getting audio and video decoding under 3.10 beta working |
| 18:53 | hste | jnettlet: nice :) |
| 18:53 | hste | jnettlet: did u change in kernel? |
| 18:53 | MikeSeth | okay, my i1 booted into droid, time to try out debian |
| 18:54 | MarcusVinter | Just writing it now. |
| 18:58 | hste | rabeeh: Just created a installer for jessie on hb http://stende.no-ip.info/imx6/solidrun-hb/scripts/mk_jessie-lxde_hb.sh |
| 19:11 | jnettlet | hste, I needed some device-tree changes/additions |
| 19:16 | MikeSeth | hste: oh, nice, just as I was about to hack uboot into the jessie image manually |
| 19:22 | MikeSeth | hste: any chance you github that? |
| 19:37 | MikeSeth | something is very wrong with israeli internet right now |
| 19:46 | hste | MikeSeth: Its just a simple script so I don't put it on git |
| 19:52 | dv_ | ah, nice. i4pro shipment is progressing. |
| 19:54 | MikeSeth | hste: gist then :) |
| 19:55 | MikeSeth | what's the byte offset of the uboot image on the sd? 1024, right? |
| 19:55 | dv_ | jnettlet: hows the kernel going? |
| 19:56 | jnettlet | dv_, it is going. playing back video but trying to sort out various problems |
| 19:56 | dv_ | VPU problems? |
| 19:56 | dv_ | how are you playing videos? |
| 19:57 | jnettlet | just xbmc right now. |
| 19:57 | MikeSeth | jas-hacks: gonna try your jessi image on i1 right now |
| 19:57 | dv_ | jnettlet: also, for the moment, I should stick to the freescale 3.0.35 one with the 4 solidrun patches on top, right? |
| 19:58 | dv_ | I want to send otavio some hummingboard recipes soon |
| 19:58 | dv_ | (cubox-i will follow when I get these devices) |
| 19:58 | jnettlet | dv_, for now yes. I can try to backport some work |
| 19:59 | jas-hacks | jas-hacks: give it go, is that the 512Mb model? |
| 19:59 | dv_ | jnettlet: nah, dont bother. concentrate on the new kernel instead. |
| 19:59 | dv_ | also, shall I still use your uboot fork from github? |
| 20:00 | jnettlet | good question we should probably ask rabeeh about that. |
| 20:00 | jnettlet | might be time to switch to the "official" repo |
| 20:01 | dv_ | did the official repo accept the humminboard and cubox-i patches? |
| 20:02 | jnettlet | dv_, it is in the works |
| 20:03 | dv_ | otavio, two questions for you. first, kernel: for now, I will use the fsl 3.0.35 with the solidrun patches. this will be added to meta-fsl-arm-extra. eventually though I'd switch to jnettlet's kernel. is that okay? can this be added to meta-fsl-arm-extra like that? |
| 20:03 | otavio | dv_: why not using jnettlet kernel directly? |
| 20:04 | dv_ | second, uboot: essentially the same. right now, I use jnettlet's fork. can I use it in a recipe and add that recipe to meta-fsl-arm-extra? in short, are external, non-fsl repos for uboot and kernel okay? |
| 20:04 | dv_ | otavio: he's not done yet with the kernel |
| 20:05 | otavio | dv_: it is fine; but Fabio sent patches to the mainline U-Boot and seems it is at same status as jnettlet one |
| 20:06 | otavio | dv_: I'd prefer to backport it to 2013.10 and go with it |
| 20:06 | dv_ | hm |
| 20:06 | dv_ | well, jnettlet's fork is based on 2013.10 |
| 20:06 | otavio | dv_: but it is your call |
| 20:06 | otavio | dv_: so for now name it linux-cubox and u-boot-cubox |
| 20:06 | dv_ | linux-cuboxi and u-boot-cuboxi perhaps (linux-cubox already exists in my layer and refers to the kernel for the old cubox) |
| 20:06 | otavio | any progress in 3.10 support? |
| 20:06 | MikeSeth | jas-hacks: yeah |
| 20:06 | otavio | dv_: works |
| 20:07 | MikeSeth | I think I hosed the boot image |
| 20:07 | dv_ | otavio: any links to fabio's patches? |
| 20:07 | MikeSeth | display says unsupported resolution, usb devices aren't initialized |
| 20:07 | dv_ | otavio: better ask jnettlet that :) |
| 20:09 | otavio | dv_: in u-boot mailing list |
| 20:10 | MikeSeth | damn I need a hdmi plexer and a serial cable |
| 20:10 | dv_ | Fabio Estevam ? |
| 20:11 | dv_ | jnettlet: might be interesting: http://www.mail-archive.com/[email protected]/msg128765.html |
| 20:13 | otavio | yes |
| 20:23 | MikeSeth | why are the uboot images different for different cubox models? |
| 20:30 | MikeSeth | ok, I *did* hose the boot image, now I've got it written properly, debian time! |
| 20:33 | MikeSeth | holy god it booted |
| 20:33 | MikeSeth | gentlemen, I am talking to you from my own cubox |
| 20:34 | jnettlet | MikeSeth, because they different versions need different DDR configs. There should be a new version supporting SPL soon that can boot all the devices with the same uboot binary |
| 20:36 | jnettlet | dv_, yep he is submitting the patches for us. We have worked that out behind the scenese |
| 20:37 | dv_ | MikeSeth: yay |
| 20:40 | MikeSeth | jas-hacks: is the default keymap in jessie non-US by design? |
| 20:41 | jas-hacks | jas-hacks: possibly ... might be UK |
| 20:42 | MikeSeth | looks that way, I think there's composition enabled and I got a " instead of the at sign |
| 20:42 | hste | MikeSeth: you can change in /etc/default/keyboard |
| 20:42 | MikeSeth | hste: yeah |
| 20:47 | MikeSeth | 80 mb free ram |
| 20:47 | MikeSeth | okay this box is going to be a headless server |
| 20:48 | dv_ | MikeSeth: quite low. did you reserve a lot of MB for video? |
| 20:48 | sickness | headless++ |
| 20:48 | sickness | :) |
| 20:48 | MikeSeth | dv_: looks like it ate up 256mb |
| 20:48 | dv_ | will you playback video on it? |
| 20:49 | MikeSeth | nah, I'll get a quad one for that |
| 20:49 | dv_ | then set the kernel command argument vram= to something low |
| 20:49 | dv_ | oh wait, this is the cubox-i ? or the old cubox? |
| 20:49 | MikeSeth | -i |
| 20:49 | MikeSeth | i1, 512mb ram |
| 20:49 | dv_ | I dont know if this one supports the vram= option |
| 20:50 | dv_ | if not, you can modify the kernel config |
| 20:50 | hste | gpumem= |
| 20:50 | dv_ | but doesnt the VPU also reserve some mem? |
| 20:50 | MikeSeth | let's try it |
| 20:50 | dv_ | iirc the size is inside the kernel config |
| 20:51 | hste | MikeSeth: you could build the kernel without vivante if you don't plan to use that stuff |
| 20:52 | hste | and not start lightdm |
| 20:52 | MikeSeth | that worked |
| 20:53 | MikeSeth | amazingly good boot time |
| 20:54 | MikeSeth | hste: right now I am going to go through boot logs and see if anything funny is up |
| 20:55 | MikeSeth | EXT3-fs (mmcblk0p1): error: couldn't mount because of unsupported optional features (240) |
| 20:55 | MikeSeth | EXT2-fs (mmcblk0p1): error: couldn't mount because of unsupported optional features (240) |
| 20:55 | MikeSeth | EXT4-fs (mmcblk0p1): mounted filesystem with ordered data mode. Opts: (null) |
| 20:55 | MikeSeth | is this probing for fs type? |
| 20:57 | MikeSeth | okay something very weird |
| 20:57 | MikeSeth | bogomips count quadrupled once I reduced the gpu memory allocation |
| 20:57 | jas-hacks | MikeSeth: you can change to boot args to include 'rootfstype=ext4' |
| 21:02 | MikeSeth | emacs work, and I've just filled my fs |
| 21:05 | MikeSeth | alright, thanks everybody, this is awesome, supper time |
| 21:16 | MikeSeth | jas-hacks: will you be maintaining the debian jessie image? |
| 21:17 | jas-hacks | MikeSeth: will try our best |
| 21:18 | MikeSeth | I'll talk to rabeeh tomorrow about debian packages he wanted maintainers for |
| 21:18 | hste | MikeSeth: If you don't need the vivante etc, you can install a clean debian armhf made with bootstrap |
| 21:18 | MikeSeth | hste: I don't on this box but I already want a quad for multimedia |
| 21:19 | MikeSeth | besides the absolute 100% proof the box works in any case is I can view streaming porn on it, which remains to be proven :P |
| 21:20 | MikeSeth | btw i suspect the android 4.3 image has the same problem with gpu memory allocation, it was very sloppy for me and aps would crash every 5 sec or so |
| 21:21 | MikeSeth | I'm really curious to know why video ram allocation crashes the jiffy counter |
| 21:21 | MikeSeth | may very well be a kernel or driver bug |
| 21:22 | hste | MikeSeth: Hopefully freescale will come with 4.4 sometime. It should run better with only 512MB |
| 21:23 | MikeSeth | hste: linux runs just fine once I set the gpu ram to 64mb, but it had an eerie slow feeling before I did |
| 21:23 | MikeSeth | I also suspect that the i1 boxes arent going to be very popular so I'll run into issues nobody else has |
| 21:23 | hste | I run mine c1 with 32M |
| 21:24 | MikeSeth | jas-hacks: how do you generate the jessie image? |
| 21:27 | jas-hacks | MikeSeth: started with a bootable console rootfs, then lots of sweat & tears to get GPU/VPU stuff working |
| 21:28 | MikeSeth | what does this involve other than kernel patches? |
| 21:29 | MikeSeth | oh I just realized you blogged this |
| 21:29 | MikeSeth | sorry |
| 21:29 | MikeSeth | rtfm time |
| 22:26 | _dab_ | 14:15 * _rmk_ is coming to the conclusion that device tree on arm is a massive backwards step... |
| 22:27 | _dab_ | and _dab_ agrees: I cannot get DT to work reliably with Marvell kirkwood |
| 22:33 | _rmk_ | yea, I now have only one bootable platform on my build/boot test system |
| 22:38 | MikeSeth | r is for regression |
| 22:45 | _dab_ | I had one kernel for 3 different models of dreamplug, with 3.10+ DT none of them work |
| 22:59 | shesselba | _dab_: can you be more specific wrt "none of them work" ? |
| 23:14 | _dab_ | shesselba: "loading kernel" and freeze |
| 23:15 | shesselba | _dab_: just to make sure, you are booting zImage with appended dtb? |
| 23:15 | _dab_ | yes |
| 23:16 | shesselba | what is your u-boot command to boot it? |
| 23:16 | _dab_ | I resurrected the dreamplug setup.c (not sure of the name without looking) and all was fine |
| 23:18 | _dab_ | I can't recall, I need something special on the ext?load? |
| 23:21 | _dab_ | dab remembers I added the DT to the spi flash and pointed at this address (excuse me this was about 3 months ago) |
| 23:23 | shesselba | _dab_: you added the dtb to spi? I thought you said you boot with appended dtb? |
| 23:23 | shesselba | is your u-boot dt capable? |
| 23:28 | _dab_ | I tried both and also tried with and without an updated u-boot |
| 23:30 | _dab_ | The original uboot is not dt aware |
| 23:40 | MikeSeth | serial works on -i1 correct? |