IRC log of #cubox of Thu 13 Jul 2017. All times are in CEST < Back to index

11:28 topi`_> jnettlet[m]: btw, have you mainlined those patches you sent me to get the wl18xx working on 4.9.21?
11:29 topi`_> I remember there were some missing bits in the .dtb and then some fix that actually changed the timings or something so wl18xx could work
11:30 topi`_> I lost a fair bit of my work when I killed a VPS which I forgot to tar up and those linux kernels were amongst the victims it seems
11:30 jnettlet[m]> topi`_: not yet. we were just talking about this on the arm-linux mailing list
11:31 jnettlet[m]> basically we are going to simplify the upstream device-trees and then provide overlay fragments for the variants
11:31 topi`_> I have those .dts related patches, but I am missing that particular patch that addressed a bit of code somewhere
11:31 jnettlet[m]> upstream u-boot supports this now and it will make things as simple as possible
11:31 topi`_> do you remember what was needed to get wl18xx to work? IIRC it was something you fixed in 2015
11:32 jnettlet[m]> not off the top of my head. Everything is clouded with the new code I am working with.
11:32 topi`_> I'll scan my mailbox
11:33 topi`_> yes, it was all about some oscillator
11:34 topi`_> something about pcf8523_set_12p5_pf
11:35 jnettlet[m]> oh that was to fix the rtc clock
11:35 jnettlet[m]> yes those patches are prepped for upstream as well
11:44 topi`_> sounds good :)
11:45 topi`_> we'd love to be able to upgrade the kernel to whatever newest one there is in mainline
11:45 topi`_> at a future point in time
12:01 topi`_> make -j6 for a clean build of the HB kernel took 9 mins
12:02 topi`_> I cranked up the clocks of the 8040 to 2GHz
12:02 jnettlet[m]> 2Ghz 2133 ddr?
12:02 topi`_> especially nice is that all of the linux kernel .git files etc fit in the RAM cache
12:02 topi`_> I think 2133. let me check
12:03 jnettlet[m]> 2400 is unstable, so avoid it
12:03 topi`_> Clock: CPU 2000 [MHz] DDR 1050 [MHz]
12:05 topi`_> I calculate sha256 sums for all the binaries I produce for deployment.
12:05 topi`_> what is very very nice is the speed of sha256 on the 8040
12:05 topi`_> I built an image that is 500MB compressed, time openssl sha256 image.dd.gz
12:05 topi`_> user 0m0.568s
12:05 topi`_> (it was already in RAM cache)
12:06 Ke> I don't think sha256sum is any faster for trivial data than compressed data
12:06 Ke> well wouldn't
12:06 jnettlet[m]> oh you should definitely bump up the memory speed
12:06 topi`_> no, but when I built the binaries on a quad core HB, that sha256sum used to take *seconds*
12:06 topi`_> like 10 secs
12:06 jnettlet[m]> I used ddr 1050 and 1600Mhz on my pre-binned chip
12:09 topi`_> jnettlet[m]: if you look at the Frequency selection part of the http://wiki.macchiatobin.net/tiki-index.php?page=MACCHIATObin+Interface+list
12:09 topi`_> you'll see that there is no such choice for DDR as 1050mhz
12:09 topi`_> lowest is 1600
12:10 topi`_> I chose the second-to-bottom, which is CPU 2000 DDR 2100
12:14 jnettlet[m]> topi`_: that chart is for the released chip. I have a pre-binned one
12:15 jnettlet[m]> devs always get the pre-release stuff :)
12:20 jnettlet[m]> topi`_: oh when I mean 1050 I am talking Mhz, not MTps
12:35 Ke> how's the upstream support for macchiatobin right now?
12:35 Ke> upstream linux kernel
12:36 jnettlet[m]> will be better after the -rc1 merge window
12:36 jnettlet[m]> although the patch acceptance is slow it is progressing
12:36 Ke> 10g ethernet?
12:39 Artox> topi`_, if u-boot ays script, then it tried to read something liek a boot.scr
12:39 Artox> aka 'source $loadaddr'
14:59 topi`_> jnettlet[m]: would you happen to know how keyctl and kernel's key facilities work? I'm trying to get ecryptfs to work...
14:59 topi`_> but its painful
15:00 topi`_> Artox: OK, I'll try to remember that :) how do you load a "uEnv.txt" file, then? just use "run" I guess?
15:20 Artox> no
15:20 Artox> but I don't actually know
15:20 Artox> hah
15:50 vpeter> topi`_: uEnv.txt is loaded to some address and then used by env import command.