2014-04-23

<cubear> Montjoie: is it in 3.4-sunxi branch yet?
<Montjoie> release delayed due to bad performance
<Montjoie> driver soon to be released for mainline
<Montjoie> cubear, driver availlable for 3.4

2014-04-16

<Montjoie> oh my god, the bug is triggered by commenting/decommenting unused dead code
<Montjoie> ouch my cryptotest program works fine with -O0 and block with -O2, WTF

2014-04-10

<ccaione> Montjoie: platform_get_resource with IORESOURCE_MEM
<Montjoie> ok it is automatic..
<Montjoie> how can I get iomem from devicetree for a device ?
<Montjoie> certainly due to neon improvment
<Montjoie> ssvb, my last bench show identical performance with generic software encryption

2014-04-09

<ssvb> montjoie[home]: how does its performance compare to software encryption now?
<montjoie[home]> I could set the SS clock to the maximum
<montjoie[home]> Turl I think the speed gain is due to better clock support
<Turl> montjoie[home]: why the difference?
<Turl> montjoie[home]: nice speed :)
<montjoie[home]> whhhhoh 50% performance gain for the security system with mainline,,,

2014-04-04

<montjoie[home]> ssvb if mripard could ask AW, go for it
<ssvb> wens: but people were speculating about Montjoie's patches there, so it was a nice thing to finally learn about
<ssvb> Montjoie: but if we can also independently measure something, then we should surely go for it
<ssvb> Montjoie: maybe mripard could add a clarification of the TP thermal sensor calibration to the list of questions for AW?
<ssvb> Montjoie: right, and now imho it's time to get something more credible than just an assumption :)
<Montjoie> the datasheet said nothing for the TP, so i have assumed that the calcul was the same, and for A20 it seems correct
<ssvb> Montjoie: on A20 it seems to somewhat be believable, but still it would be nice to check it with some external temperature measurement tool
<ssvb> Montjoie: yes, for the TP sensor I'm getting strange results on A10, it starts from around 100C on idle
<Montjoie> for the sensor enbdeded in the TS, it is a guess (see my old patch for it)
<Montjoie> ssvb, for AXP the numbers are from the datasheet

2014-03-31

<Montjoie> and need to add the correction to my cfengine
<Montjoie> need to see my history
<Montjoie> labrador, my cubieboard is off, but I forgot if it is touch /etc/adjtime or hwclock --systohc that made it works
<labrador> Montjoie: i have it and current content is: 0.000000 1396287840 0.000000 \n1396287840 \nUTC
<Montjoie> labrador, the no such file or directory is about /etc/adjtime, I touch it for made hwclock works

2014-03-30

<montjoie[home]> I will try sunxi-next also in cas of...
<montjoie[home]> mripard I am sure
<mripard> montjoie[home]: are you sure you're booting that kernel?
<montjoie[home]> mripard the dt is modified with http://pastebin.com/swdUjXuB
<montjoie[home]> latest sunxi-devel
<montjoie[home]> yes
<Turl> montjoie[home]: ahb clock is good and ss one fails?
<montjoie[home]> Turl nothing change:(
<Turl> montjoie[home]: what happens if you put clocks in reverse order?
<montjoie[home]> and i try without it since it is useless
<montjoie[home]> yes
<Turl> montjoie[home]: did you fix interrupts = <...>?
<Turl> montjoie[home]: it's exactly the same as on spi driver, and it works there :)
<Turl> montjoie[home]: looks good, dunno
<Turl> montjoie[home]: can you paste code and dmesg?
<montjoie[home]> Turl nothing changes with or without interupts
<Turl> montjoie[home]: look at other devices for example
<Turl> montjoie[home]: on a20 they look like <N N N>, not, <N>
<Turl> montjoie[home]: maybe it's the interrupt specifier breaking things
<Turl> montjoie[home]: strange hm
<montjoie[home]> Turl http://pastebin.com/utU3LtqG I see the ss clock
<montjoie[home]> (compiling debugfs...)
<Turl> montjoie[home]: can you check if clock tree looks ok and ss is there? it should be on clk/clk_summary in debugfs
<montjoie[home]> -2
<Turl> montjoie[home]: what return code do you get from devm_clk_get?
<Turl> montjoie[home]: and what does your dt look like?
<montjoie[home]> for be precise
<montjoie[home]> ss_ctx->ssclk = devm_clk_get(&pdev->dev, "mod");
<montjoie[home]> devm_clk_get(..., "mod")
<Turl> montjoie[home]: what fails?
<montjoie[home]> so something is wrong with the ss clock
<montjoie[home]> It works for the first clock, but for the ss clock if fail
<montjoie[home]> thanks I will try
<Turl> montjoie[home]: no, right now with that dt you need to use of_clk_get
<montjoie[home]> Turl the name i need to clk_get is ahb right ?
<Turl> montjoie[home]: you can add clock-names to get clock by devm_clk_get if you want
<Turl> montjoie[home]: looks good that way
<montjoie[home]> Turl does I need to modify something for the clocks ?
<Turl> montjoie[home]: don't forget to write a binding document as well :)
<Turl> montjoie[home]: your dt looks sane :)
<montjoie[home]> Wizzup I am writing the driver for it:)
<montjoie[home]> could someone confirme that my Security System DT is correct http://pastebin.com/sXnC6Ygd ?
<montjoie[home]> could someone confirme that my Security System DT is correct http://pastebin.com/sXnC6Ygd ?

2014-03-29

<labrador> montjoie[home]: A13 olinuxino
<montjoie[home]> you need to give more info
<montjoie[home]> for which board ?
<nabblet> montjoie[home]: thank you for the info
<montjoie[home]> nabblet I have just built one, 17m for a kernel
<montjoie[home]> and time to compile crossdev the first time
<montjoie[home]> :)
<nabblet> montjoie[home]: ok. and you were too impatient so you got the arch kernel instead? I think it would take me more time to find a suitable precompiled kernel :P
<montjoie[home]> compiling a kernel is less than a hour, 15min I think
<nabblet> montjoie[home]: fair enough. still how long does it take (roughly)?
<montjoie[home]> a quad core with only one core enabled and undercloked due to hardware problems, not a good example:)
<nabblet> montjoie[home]: what are its specs and how long did you wait? just want to get an idea of how long it might take me
<montjoie[home]> no my main machine
<montjoie[home]> I compile all gentoo in tmpfs, but kernel si too large, and nfs is too slow
<montjoie[home]> now yes
<nabblet> montjoie[home]: ah so you were using crossdev parallely?
<montjoie[home]> it is the first I found, too enthusiast to wait for the compilation to end
<nabblet> montjoie[home]: out of interest: why did you chose to use the arch-kernel instead of crossdev?
<nabblet> montjoie[home]: ok, thank you for your kind answers
<montjoie[home]> no idea
<nabblet> montjoie[home]: oh, and one more question. the olimex a20 can only boot from micro sd-card. now, i have a 4 GB sd-card which will probably suffice for an X-less setup. How much should i take into account for a setup with X (and fluxbox)? do you have some kind of clue?
<nabblet> montjoie[home]: ok. thank you. just wanted to make double-sure i am not starting off in the wrong direction :) I'll go with crossdev on my gentoo vm and see how far this will get me.
<montjoie[home]> It just be slower inside a vm, nothing more as I know
<nabblet> montjoie[home]: i set up a vm with gentoo. i read that (with crossdev) you need to take care about your home arch and your target arch. will that be tricky in a vm?
<montjoie[home]> and all u-boot tricks
<montjoie[home]> and the uImage in /boot
<montjoie[home]> yes just copy stage3
<montjoie[home]> nabblet yes, but if you use gentoo, crossdev is not a pain
<montjoie[home]> for sunxi-devel does someone know if CONFIG_SUN4I_EMAC=y is sufficient to have the network ?
<nabblet> montjoie[home]: ok, so what i need/want is compiled kernel + compiled u-boot + stage 3 ?
<montjoie[home]> nabblet no only cubieboard2
<nabblet> montjoie[home]: do you (personally) have experience with gentoo @ olimex a20 ?
<montjoie[home]> nabblet I forgot the u-boot, I have get an already built one also
<nabblet> montjoie[home]: your trick with the archkernel is nice. didn't think of this option at all!
<nabblet> montjoie[home]: well the wiki says the catch is to emerge/install/setup crossdev
<montjoie[home]> with crossdev it is easy to crosscompile
<nabblet> montjoie[home]: tbh i am willing to wait some time (although i think the a20 will be sufficiently fast) to get around the crosscompiling pain
<montjoie[home]> yes for compiling the real "gentoo" kernel
<nabblet> montjoie[home]: ah, and then you used the arch kernel for compiling?
<montjoie[home]> nabblet the problem is the first kernel creation, I have set a gentoo stage with a arch kernel the first time
<bgal> montjoie[home]: in theory it's already enabled in commandline
<montjoie[home]> compiling in RAM helps
<nabblet> montjoie[home]: hm, but do i need to crosscompile. can't i do the whole thing natively as i do it with normal gentoo installs?
<montjoie[home]> nabblet yes gentoo can be used on it
<montjoie[home]> it's y beginning with uart
<montjoie[home]> ccaione I always use ssh for login before
<montjoie[home]> where can I enable it ?
<bgal> montjoie[home]: it seems that the kernel console is not enabled, you should see something like: console [ttyS0] enabled
<ccaione> montjoie[home]: which console are you using at login time? (inittab?)
<montjoie[home]> it seems that the kernel runs a bit further after but the uart show no nothing, any idea ?
<montjoie[home]> I finaly got booting a sunxi-devel kernel, but now it stop after bootconsole [earlycon0] disabled , http://pastebin.com/W9BhU1jj for boot log
<montjoie[home]> strange my first cubieboard2 have only ext4 and it works with old u-boot
<specing> montjoie[home]: uboot is just for booting, afaik it only reads fat32
<montjoie[home]> hello does uboot can read ext4 fs ? I got some Unrecognized filesystem type with it
<montjoie[home]> and just before fdt_high=ffffffff
<montjoie[home]> I have that in uEnv.txt uenvcmd=fatload mmc 0 0x46000000 uImage; fatload mmc 0 0x49000000 sun7i-a20-cubieboard2.dtb; bootm 0x46000000 - 0x49000000
<mripard> montjoie[home]: please look at http://linux-sunxi.org/Mainline_Kernel_Howto
<mripard> montjoie[home]: no. You're not passing the DT
<montjoie[home]> hello I fail to boot sunxi-devel with http://pastebin.com/GgE0i9sW any idea ? does I need to set machineid as per http://linux-sunxi.org/Manual_build_howto ?

2014-03-28

* montjoie[home] flee with shame
<montjoie[home]> lol it is written 2 lines after
<Turl> montjoie[home]: "-"
<montjoie[home]> if I does no initramfs, what I need to set ?
<montjoie[home]> I ask that because of bootm 0x<kernel-address> 0x<initramfs-address> 0x<dtb-address>
<mripard> montjoie[home]: or nfs
<Turl> montjoie[home]: you can use MMC or SATA with sunxi-devel
<Turl> montjoie[home]: yes it is
<montjoie[home]> hello does it is possible to use a sunxi-devel kernel without initramfs ?

2014-03-27

<Montjoie> Makefile:485: *** "System not configured - see README". Arret
<Montjoie> as said the README
<Montjoie> the solution is to do make board_config
<libv> Montjoie: paste bin the full error
<Montjoie> hello does http://linux-sunxi.org/U-Boot#Compilation is still up to date ? I got System not configured errror when following it

2014-03-11

<Turl> sehraf: no, but montjoie[home] was working on it last I checked. The performance wasn't that great yet, but the offloading of the cpu may still hold some value

2014-03-06

<Montjoie> wens, if you need help for the hwmon:)

2014-02-24

<Montjoie> logs/smart/raid/cpu/sensors
<Montjoie> you have lots of thing to be monitored just with one server
<Montjoie> no a real monitor software like nagios/xymon, I use the latter
<Montjoie> seems you do not monitor enougth:)
<Montjoie> php session must be put away from /tmp, and disk inodes should be monitored and graphed:)

2014-02-19

<mripard> Montjoie: a very common use for PWM are backlights :)
<wens> Montjoie: there's a PWM controller, with PWM outputs muxed on some pins
<Montjoie> I just see a sunxi-pwm driver, it seems to use the lcd output
<CaptHindsight> Montjoie: depends on how the fan is connected to the cubie, the IO pins themselves don't drive much current if that what you mean
<mripard> Montjoie: the PWM are exported on a few pins, I don't know wether they are usable or not on the cubie2
<gzamboni> Montjoie, cant you do pwm using a regular gpio by software ?
<Montjoie> mripard, does cubieboard2 have a PWM, by GPIO I meant the external output pin
<mripard> Montjoie: usually, a PWM is more fit for that
<Montjoie> hello, can I control a fan with GPIO on a cubieboard ?

2014-02-17

<Montjoie> yes, I have backported it to 3.4 for benching against the security module:)
<Turl> Montjoie: the in kernel NEON one?
* Montjoie cry after benchmarking AES with arm assembly

2014-02-13

<oliv3r> montjoie[home]: i prefer puppet over cfengine; but to be fair, i never used either!
<Montjoie> cfengine is for all OS, not only gentoo
<ccaione> Montjoie: I'm a gentoo-noob, so yeah I have a lot to learn. Thank you for the advice :)
<Montjoie> ccaione, now it is time to use a conf engine for gain time, cfengine for example
<Montjoie> dont forget to compile gcc twice:)

2014-02-07

<montjoie[home]> but i could do mistake
<montjoie[home]> my current benchmark said that
<plaes> montjoie[home]: that it's slow?
<montjoie[home]> cubear read the email I have sent on the sunxi mailiing list for crypto hardware acceleration
<oliv3r> if you add support, sure; i think montjoie[home] is adding it via the regular crypto framework
<oliv3r> cubear: for now; yeah; but montjoie[home] is working on the mainline encrpytion driver
<montjoie[home]> mripard didnt buy any:(
<Montjoie> if some french people have some experience with foreign buying here, contact me please:)
<Montjoie> 2x is what I see
* Montjoie is annoyed, trying to buy another cb2, buying it on foreign country is cheaper than mine:(

2014-02-05

<Turl> montjoie[home]: :p you need like a quad socket xeon for that
<montjoie[home]> Turl depend on your RAM/numbers of cpu, on a 48core gentoo is fast...
<ccaione> montjoie[home]: just because I want to take a break from arch :)
<montjoie[home]> another gentoo:)

2014-02-03

<montjoie[home]> ssvb you could check gpg in manifest of all ebuilds
<montjoie[home]> Wizzup no ext2 on RAID0 on nfs host
<Wizzup> montjoie[home]: considered using squashfs for that?
<montjoie[home]> with portage on NFS, gentoo is quite fast on A20
<montjoie[home]> gentoo for all the world!!!

2014-01-25

<montjoie[home]> nieuwbie gentoo here but no x11

2014-01-22

<montjoie[home]> gzamboni see include/linux/printk.h, use dev_dbg instead but according to some ifdefine I think you need to enable DEBUG

2014-01-21

<hramrach> montjoie[home]: sunxi-next probably has still broken clocks so it won't boot. also no drivers
<montjoie[home]> ok I will try that thanks
<Turl> montjoie[home]: try sunxi-devel as hramrach said
<montjoie[home]> Turl cubieboard2
<Turl> montjoie[home]: you have ct? or cb2?
<montjoie[home]> but emac no ?
<Turl> montjoie[home]: that doesn't have gmac support
<montjoie[home]> mripard sunxi-next
<Turl> montjoie[home]: what branch are you using?
<montjoie[home]> Turl thanks for the uEnv.txt but it is not better
<Turl> Montjoie: you may want to add root=... on bootargs too
<Turl> Montjoie: try something like this http://pastebin.com/raw.php?i=hmX3pvRT
<Turl> Montjoie: yeah, that's not how uEnv.txt works :)
<Turl> Montjoie: paste it please
<Montjoie> with the <board> changed by the filename
<Montjoie> the 4 lines needed to type on u-boot
<Turl> Montjoie: what did you add on uEnv?
<Montjoie> mainline kernel KO:(
<Montjoie> hramrach, so if I put all 4 lines in uEnv.txt, it will work ?
<Montjoie> Ok I will update u-boot first
<Montjoie> Cubieboard2
<Montjoie> yes
<Turl> Montjoie: are you on a20?
<Montjoie> does updating uboot is usefull, my uboot is the same since 4/5 month ?
<Montjoie> I interact only with ssh
<Turl> Montjoie: mainline doesn't have video out, how would you interact?
<Montjoie> I want to try sunxi-devel but http://linux-sunxi.org/Mainline_Kernel_Howto said that I mus stop autoboot for doing some hack but I dont have any serial console, so mainline without serial is impossible?
<Montjoie> If he not work on it soon, I will try also for fun, I really need it for the Security System:)
<wens> Montjoie: mripard_ plan is to get SPI working first, then work on DMA with SPI as testbed
<Montjoie> gzamboni, I have read that mripard_ would work on it
<gzamboni> Montjoie, yes, but i'm just playing around, i dont know if will be able to make it work
<Montjoie> gzamboni, you work on the DMA engine ?
<gzamboni> Montjoie, i will dig deeper, shouldnt it break after a -EINVAL ? in almost all EINVAL i have a goto to unload stuff, but i think there are some without
<Montjoie> gzamboni, place a printk before each return -EINVAL ?

2014-01-14

<Montjoie> ok thanks, now I am sure than I have correctly sent it
<ccaione> Montjoie: I guess nobody (but you) is really expert in that field
<Montjoie> i have sent it to linux-sunxi@googlegroups.com
<oliv3r> Montjoie: you forgot linux-sunxi in the CC?
<Montjoie> hello I have sent the security system driver 4 days ago and still no comment on it, does I have done something wrong ?

2014-01-12

<montjoie[home]> 512bytes for dm-crypt
<montjoie[home]> Turl sadly no, I must write all stuff for each request (aka for each 512b)
<Turl> montjoie[home]: don't you need to write key to register once?
<montjoie[home]> and I have a idea that perhaps could boost performance, preload key in security id, perhaps it is more fast than load it from memory
<montjoie[home]> adding DEs/3DES should be simple
<montjoie[home]> I wait for comment on my patch before working more on it:)
<montjoie[home]> no but I will add it
<Turl> montjoie[home]: did they talk to you?
<Turl> montjoie[home]: hi, I got someone asking me about DES/3DES SS linux support on an email the other day
<montjoie[home]> and the security system on A20 could do some key management for boot

2014-01-11

<montjoie[home]> still no comment on my driver:'(

2014-01-10

<montjoie[home]> alea jacta est

2014-01-05

<oliv3r> montjoie[home]: i'll add thjat
<montjoie[home]> and DES 3DES
<montjoie[home]> SS support SHA1
<oliv3r> montjoie[home]: checksumming isn't limitted to md5?
<montjoie[home]> Corentin LABBE
<oliv3r> montjoie[home]: may i steal your real name?
<montjoie[home]> with that everyone could understand
<montjoie[home]> oliv3r please write Security System(crypto engine)
<oliv3r> montjoie[home]: i'll fix it,s ec
<oliv3r> montjoie[home]: the one needs to be removed :p
<montjoie[home]> oliv3r the two need to be merged
<oliv3r> montjoie[home]: wingrime added it under 'to be done' not realizing crypto == SS
<montjoie[home]> hello, could someone modify http://linux-sunxi.org/Linux_mainlining_effort, putting Security System in work in progress by me and not "left to be done"

2014-01-03

<montjoie[home]> if someone can write that I will mainline after
<oliv3r> montjoie[home]: no rush; just documenting
<montjoie[home]> oliv3r I work on the SS driver, I expect to release this month
<wens> oliv3r: Montjoie on security system
<oliv3r> montjoie[home]: I think wasn't it?

2013-12-23

<Montjoie> I look forward any work on DMA

2013-12-21

<montjoie[home]> The actual conversion seems good but...
<montjoie[home]> like as I said in the mail, I have no datasheet of how to convert the raw value
<montjoie[home]> since the two chip are distant of many centimeter both temperature are usefull but note that the temperature of the a20(TP) can be not scaled
<nedko> montjoie[home]: great. thank you!
<montjoie[home]> nedko you get the internal temperature of the AXP chip and the A20 chip
<nedko> Montjoie: i've applied two of your patches for hwsensor support in a20. one gives me axp20_mfd-i2c-0-34 sensor, the other a20_tp-isa-0000. what they actually measure? the temperature of the silicon on both chips?

2013-12-07

<montjoie[home]> thanks ssvb for the link
<ssvb> Montjoie: beaglebone seems to have similar performance issues with hardware "accelerated" dm-crypt - http://thread.gmane.org/gmane.comp.hardware.beagleboard.user/58532 :)

2013-12-06

<mripard> Montjoie: sure, i will
<Montjoie> but I keep hope, each day I found some optimization
<Montjoie> but perhaps my test with dd is bad because iozone give better results for SS
<Montjoie> A20
<ssvb> Montjoie: is this on A10 or A20 hardware?
<Montjoie> for the moment a dd of 250M on a dm-crypt relying on tmpfs is writed in 23s without SS and in 33s with SS (37 with DMA)
<Montjoie> but I need to investigate more
<Montjoie> any number I give to cryptseyup, dm-crypt use 512bytes because my hardrive have 512 bytes sector
<torbenh3> Montjoie: also keep in mind, that having free cpu time, is more important than raw throughput.
<torbenh3> Montjoie: even with 8k blocks ?
<Montjoie> and it made me cry
<Montjoie> for a short teaser, raw peformance of SS seems good, but real life benchmark show that SS is useless for dm-crypt
<Montjoie> ssvb, yes very painful, you will see benchmark in my patch submission soonly
<ssvb> Montjoie: do you have some benchmark numbers in cpu cycles (or nanoseconds) comparing different methods?
<ssvb> Montjoie: ouch, this sounds painful, does it have any advantage over purely software encryption without dma?
<Montjoie> wens, dm-crypt crypt with buffer of 512bytes, too small for having any performance increase with DMA vs SS CPU mode
<wens> Montjoie: dm-crypt doesn't use DMA?
<Montjoie> I need info on DMA only for optimization
<Montjoie> wens, yes it can work without it, and for dm-crypt for example DMA is useless
<wens> Montjoie: can the crypto engine work without DMA?
<Montjoie> mripard, if you start working on DMA, please tell me, I greatly look forward on it for enhancing my crypto driver
<Montjoie> I have asked here several time and nobody seems to know how to get good values for DMA config
<Montjoie> arghhh
<mripard> Montjoie: hmmm, I worked on SPI without DMA
<Montjoie> mripard, If you have work on SPI, could you explain me the DMA magic value 0x07070707 used for DMA config please?