IRC #olimex 2016-02-01

[01:19:52] <guest____> egh i cant get my olimexino stm32 running
[15:35:30] <muawija_> hi! i'm looking for power specification for the A20-SOM
[15:35:45] <Tsvetan> consumption?
[15:35:50] <muawija_> NO!
[15:36:08] <Tsvetan> then define power
[15:36:11] <Tsvetan> :)
[15:36:17] <muawija_> i'd like to know how much power I can use on the 3V3
[15:36:37] <muawija_> if I provide 5V/2A to the board
[15:36:55] <muawija_> there is not a single word about this in the manual :-(
[15:37:59] <muawija_> the question is: must I use my own 5V->3V3 converter or whether I can use the 3V3 provided by the board
[15:38:10] <muawija_> ?
[15:39:26] <Tsvetan> depends
[15:40:52] <Tsvetan> the 3.3V dcdc in the SOM can supply 1A
[15:41:12] <Tsvetan> but it's also used by the SOM, so I would not recommend to load it with more than 100mA outside SOM
[15:43:25] <muawija_> ok. but there is no official specification?
[18:10:13] <bgardner> Hello all. I have my shiny new A10-Olinuxo-LIME up and running but the chip seems warmer to the touch than I would like. I installed lm-sensors to get something more tangible than 'seems warm' but with no PCI, lm-sensors can't detect anything. Does this SBC even have temp sensors before I keep wrestling with this?
[18:26:15] <Tenkawa> Is there a way to access the onboard nand on the Lime2 model through the 4.4 or 4.5-rc mainline kernels?
[18:55:52] <Tsvetan> bgardner: look at /sys/devices/platform/a10-tp-hwmon/temp1_input
[18:56:01] <Tsvetan> to read the SOC temperature
[18:56:07] <bgardner> Tsvetan: Looking now
[18:57:46] <bgardner> Tsvetan: 'No such file or directory' - looks like 'a10-tp-hwmon' doesn't exist.
[18:59:33] <Tsvetan> then your image may not be compiled with this option
[19:01:05] <Tsvetan> tomorrow I can ask our people, but right now is 19.00 oclock and I'm leaving office too
[19:01:18] <bgardner> Tsvetan: Hm. Okay, thanks!
[19:06:20] <adj_> Tenkawa, will you ask the same question again and again every day?
[19:28:20] <Tenkawa> adj_: well noone has said yes or no yet or gave any acknowledgement until now so i wasnt sure if anyone was even reading this channel
[19:29:31] <Tenkawa> apparently noone knows so I will just have to do some experimenting. Google results were conflicting which is why I asked here to begin with
[19:29:40] <Tenkawa> cheers all
[19:34:07] <adj_> Tenkawa, thanks for disconnecting every time you ask or say anything, sure you are not going to read the logs
[19:46:53] <lukas2511> i hate him for this
[21:00:08] <nonato> hi
[21:07:13] <nonato> does anyone have any experience with A20 and full disk encryption, including root? how is it performance wise?
[21:24:12] <lukas2511> nonato: i'll try soon, but i think it will be really really bad.
[21:44:25] <nonato> lukas2511: i was wondering about that since it has AES security accelerator
[21:46:30] <adj_> nonato, you could benchmark the ATA bandwidth, and then benchmark the AES encryption/decryption bandwidth
[21:46:59] <adj_> this way you could estimate if you are going to be ATA or CPU bottlenecked
[21:47:41] <nonato> adj_: i could, if i had the hardware
[21:47:47] <adj_> ah, ok
[21:47:58] <adj_> i thought you have it
[21:49:08] <nonato> adj_: i am in the design process
[21:51:36] <nonato> adj_: i am looking for the opportunity to use a lime instead of a low end x86
[21:52:56] <lukas2511> nonato: not sure if aes acceleration is in mainline yet, also i've read that it isn't really that fast and that people had better speeds using just the cpu
[21:53:33] <nonato> adj_: but i can't seem to find one with business projects, i wish it was around when i was dealing with media
[21:53:48] <lukas2511> okay it seems to be supported at 4.3, but no idea about the speed
[21:54:26] <adj_> https://linux-sunxi.org/Benchmarks
[21:55:18] <adj_> aes-128 cbc 23328.09k/s
[21:55:42] <adj_> that's about half the ata bandwidth in a20, right?
[21:56:22] <adj_> so I suppose that you would have half the bandwidth on serial data
[21:58:21] <nonato> adj_: more like speed of storing in ram
[21:58:56] <adj_> I don't understand
[21:59:15] <adj_> do you mean that memory bandwidth would be the bottleneck?
[21:59:43] <nonato> adj_: otherwise 23mb/s is faster that alot of usb2 sticks
[22:00:18] <nonato> adj_: i think "openssl speed" writes in ram
[22:00:42] <adj_> yes
[22:01:25] <adj_> but ata writes/read to/from ram
[22:02:04] <adj_> just trying to know if it will be bounded
[22:02:10] <adj_> just trying to know if it will be processing bounded
[22:02:58] <adj_> the same page says aes-128 cbc 44219.12k/s for A10, maybe one is using aes instructions and other not
[22:02:59] <nonato> adj_: hm... i am running "openssl speed" locally and then i will compare with writing to disk
[22:03:33] <adj_> yes, that's want I'm trying to know
[22:04:09] <adj_> 44MB/s should be enough
[22:04:55] <adj_> i don't have my lime2 at hand, so I can't check it
[22:07:45] <nonato> i suppose uboot will be fine with crypt module loaded from initrd
[22:10:31] <adj_> i think so
[22:18:57] <nonato> ok, done benchmarking
[22:20:34] <nonato> can i past the relative lines of the results here?
[22:27:44] <nonato> anyway
[22:27:44] <nonato> openssl speed, aes-256 cbc, 8192 bytes, 96635.56k.
[22:28:14] <nonato> hdparm -Tt /dev/sda, buffered disk reads, 113.22 MB/sec
[22:28:37] <nonato> hdparm -Tt /dev/mapper/vg-root_crypt, buffered disk reads, 54.92 MB/sec
[22:36:43] <nonato> with dd writing /dev/zero in an unencrypted partition i get an avarege write speed of 83.5
[22:37:21] <nonato> with 79 MB copied
[22:39:08] <nonato> the same on the encrypted results on an average write speed of 63.5 MB/s
[22:40:24] <nonato> all this on my laptop
[22:47:49] <nonato> i am trying to correlate "openssl speed" and disk read/write
[22:50:38] <nonato> a ~50% drop in speed reported from "openssl seed" might be the case
[22:51:04] <nonato> *speed
[23:27:44] <bgardner> Okay, got my A10 configured to stop the screensaver so CPU usage fell to 0.4% average. But, system load is 1.00 1.02 0.97. Root cause here looks to be dead process usb-hardware-sc. System performance is excellent, so does this need to be dealt with or just ignore it?