| 11:30 | rabeeh | Artox: notice there is a standard 4 pin header for FAN on the board |
| 11:31 | rabeeh | one of the GPIOs is connected as PWM to control the fan speed |
| 12:03 | Artox | rabeeh: yes, I noticed. Also noticed that my case comes with a 3-pin fan :< |
| 12:06 | jnettlet | Most fans are 2 or 3 pin |
| 15:30 | vpeter | Any idea what could be the cause? http://forum.solid-run.com/software-f38/what-is-the-recommended-linux-os--t3108.html#p22117 |
| 15:54 | topi` | my macchiatoBIN doesn't recognise my 250GB sandisk SSD sata... do I need to connect somethign else than those 7 sata pins? |
| 15:55 | topi` | maybe it needs some power on that 15-pin connector? |
| 15:55 | topi` | but no idea where I could get that power |
| 15:56 | topi` | the SSD seems to require +5V 1.5A |
| 15:56 | topi` | sounds a bit high to me |
| 15:57 | jnettlet | topi`, you probably need a 5V supply |
| 15:58 | jnettlet | you can get a picopsu to handle that. |
| 16:07 | topi` | it seems the only place to get +5V on the board is the usb2.0 header |
| 16:08 | topi` | I have this kind of SATA cable (from my bananapi): http://static3.watterott.com/20110947.jpg |
| 16:08 | topi` | the red/black connector goes to +5V and gnd |
| 16:08 | topi` | but there's no header on the macchiato that I could neatly connect it to :( I need to do some solder/jumper wire to use that usb2.0 header |
| 16:09 | topi` | I probably need to take the board home, since that's where I have proper electronics equipment |
| 16:11 | topi` | I untarred an arm64 rootfs onto the SSD and trying to boot from it now :) |
| 16:11 | topi` | maybe I could get the kernel to see sda, if I use my usb<>sata dongle and connect that to the usb3.0 port |
| 16:20 | topi` | jnettlet: do you happen to remember how to force uboot to try "egiga2" first with tftpboot, otherwise it will try egiga0 and 1 and timeout both... |
| 16:20 | jnettlet | topi`, look at the env variables. There is one like primeth or something to set the first interface to use. |
| 16:21 | topi` | ok |
| 16:22 | topi` | what kind of kernel param is responsible for "waiting up to 110 more seconds for network" messages, counting to 0? |
| 16:22 | topi` | this comes after sd 4:0:0:0: [sda] Attached SCSI disk |
| 16:23 | topi` | I did setenv bootargs console=ttyS0,115200 root=/dev/sda6 rw |
| 16:23 | topi` | and saveenv |
| 16:23 | jnettlet | use env save |
| 16:23 | jnettlet | Marvell broke saveenv |
| 16:23 | topi` | OK |
| 16:24 | topi` | so, you're shipping uboot that was built buy marvell? :) |
| 16:24 | topi` | doing all of the damage control ;) |
| 16:24 | jnettlet | I think with the latest u-boot enough is upstream to finally switch. |
| 16:26 | topi` | woohoo, I was able to set ethprime to egiga2 :) |
| 16:26 | jnettlet | fyi. the mcbin board is a Marvell board made by SolidRun. It is not a SolidRun board. |
| 16:26 | jnettlet | so there is the distinction |
| 16:27 | topi` | ok |
| 16:27 | topi` | I was able to "tftpboot" the kernel image, but only running "tftpboot" only loaded the kerne, not the .dtb |
| 16:27 | topi` | I wonder if there is a script that does both |
| 16:27 | jnettlet | yeah. probably Artox is the one to ask about that. |
| 16:28 | jnettlet | I think he is netbooting his board |
| 16:28 | topi` | I think I know why my kernel params did not change. |
| 16:28 | topi` | there is "set_bootargs" macro that sets the root to nfsroot |
| 16:28 | topi` | I wrote a rootfs on sda, now I just need to load the kernel via tftp |
| 16:29 | jnettlet | okay |
| 16:32 | topi` | ok, found the get_images and that should get me both the image and the dtb |
| 16:32 | topi` | mind you, I'm not a very experienced uboot hacker :) |
| 16:33 | topi` | wtf... A start job is running for dev-ttyS0.device ... why doesn't ttyS0 work? |
| 16:35 | topi` | kernel says "console [ttyS0] enabled". |
| 16:35 | topi` | maybe I'll ask the ubuntu core guys |
| 16:59 | topi` | jnettlet: do you remember if there was some issue with the DDR that I should not run the DIMM at full 2400mhz clocks? |
| 16:59 | topi` | I'll try to see if this particular 8040 can run the cores at 2.0 ghz |
| 17:15 | jnettlet | topi` yes only run the memory at 800mhz for now. Are still waiting for calibration and timing fixes from Marvell |
| 17:15 | jnettlet | Maximum do 1050Mhz |
| 17:16 | jnettlet | I am getting ready to get on a flight so will be sporadic |
| 17:19 | topi` | have a nice flight : |
| 17:19 | topi` | :) |
| 17:19 | topi` | and don't worry if you see your neighbor passenger carry a Samsung Note 7 device :) |
| 17:28 | topi` | I get over 200MB/s over the usb3 connector |
| 17:28 | topi` | not too shabby |
| 17:29 | topi` | oh, and writes at 250+ MB/s :) |
| 17:29 | topi` | this is on a JMS567 usb-sata bridge chip |
| 17:29 | topi` | I'll try out the SATA ports as soon as I figure out how to get the +5V power |
| 17:30 | topi` | the memory perf is slightly disappointing, I get 2.5 GB/s with a dd if=emptyfile of=/dev/null bs=1024k |
| 17:30 | topi` | (do this several times so all the data fits into ram buffers) |
| 17:32 | Ke | topi`: what filesystem is the emptyfile on |
| 17:32 | topi` | just an ext4 |
| 17:32 | Ke | if you want memory perf, why not benchmark userspace memcpy? |
| 17:32 | topi` | I bet xfs would be faster in sequential |
| 17:33 | topi` | reads-from-filesystem-buffers act as a proxy for memory perf |
| 17:33 | topi` | in server workloads, it's rather important |
| 17:33 | topi` | unless, say, mysql uses mmap() to actually bypass the filesys |
| 17:33 | Ke | perhaps, how big was the filesize? |
| 17:33 | topi` | 1 gigabyte |
| 17:34 | Ke | there is a lot more stuff going on that just copy_to_user anyway there |
| 17:35 | Ke | and even copy_to_user has to pay attention |
| 17:36 | Ke | did you try perf to figure out, where the time is spent |
| 17:37 | bencoh | I'd expect mysql to do mmap() at some point |
| 17:37 | bencoh | (I hope it does at least) |
| 17:37 | Ke | btrfs at least sucks at mmap writes though |
| 17:50 | jnettlet | sata should do full 500MB/s |
| 17:50 | jnettle | 17:50 * jnettlet is on Norwegian so in flight wifi :) |
| 17:53 | topi` | it seems the 8040 does not tolerate compiling the kernel with make -j6 |
| 17:53 | topi` | the board reset itself after a few minutes into compiling |
| 17:54 | jnettlet | topi`, what do you have it clocked at? |
| 17:54 | topi` | the default as shipped |
| 17:54 | jnettlet | haven't seen this. nothing on the console? |
| 17:55 | topi` | CPU 1300 DDR 800 FABRIC 800 MSS 200 |
| 17:55 | topi` | that's what the uboot reports |
| 17:55 | topi` | nothing, it just rebooted and back into uboot |
| 17:55 | topi` | reset |
| 17:55 | topi` | makes me believe there was some overheating |
| 17:55 | topi` | and, oh, it started to re-reset itself also at the time of loading uboot |
| 17:56 | jnettlet | yeah sounds like a heat issue. |
| 17:56 | topi` | so it kept printing BootROM 2.13 over and over again |
| 17:56 | jnettlet | I don't actually have the current shipping heatsink |
| 17:56 | topi` | I'll try a fan, if it allows me to actually compile the kernel |
| 17:57 | topi` | this is definitely no mobile SoC :) |
| 17:57 | jnettlet | even a very slow fan is usually enough to really keep the chip cool. |
| 17:58 | Ke | so there is a failsafe that automatically reboots? |
| 17:58 | jnettlet | yes |
| 17:59 | bencoh | building kernel directly on board? well, at least it's a nice test case ... :) |
| 18:00 | topi` | :) |
| 18:00 | topi` | gcc is one of the SPEC sub-tests where the cortex-a72 is really leaps and bounds better than any previous arm cores |
| 18:00 | Ke | heh, even my C201 is self hosting |
| 18:01 | topi` | what is a C201? |
| 18:01 | Ke | Chromebook |
| 18:01 | Ke | rk3288 |
| 18:01 | topi` | ok :) |