<montjoie>
curlybracket: I dont see this problem in v8
<montjoie>
so dt patch are wrong ?
<curlybracket>
montjoie: I've encountered the same problem with the sun8i-thermal driver you described on Saturday. sun8i_ths_probe was failing with -2 because DT node listed "ahb" clock name instead of "mod" which was expected by the driver.
2019-12-22
<montjoie>
oh very interested! It will permit to remove all my USB dongle on all my H6
<montjoie>
tuxd3v: kernelci have a lime2 so you could always check its status
<tuxd3v>
montjoie, maybe sun4i_ss was ok, before.. I have never succeeded in configuring PRNG correctly... using RNG-tools .. and now out-of-the-box it just work with config CONFIG_CRYPTO_DEV_SUN4I_SS_PRNG=y
<montjoie>
tuxd3v: sun4i_ss was not fine before ?
2019-12-21
<anarsoul>
montjoie: they exist because you have thermal zones in dts
<montjoie>
basicly all kernelci boards
<montjoie>
I hit that on pineh64 model b and sun50i-h5-libretech-all-h3-cc and sun8i-h2-plus-orangepi-r1
<montjoie>
but strange that the /sys files exists without it
<montjoie>
anarsoul: I have probably a problem since i didnt see the driver loaded
<anarsoul>
montjoie: on what platform did you test it?
<anarsoul>
montjoie: I tested it only on A64 (don't have other devices) - works fine for me
<jernej>
montjoie: I just tried thermal patch on top of 5.4 on Tanix TX6 (H6) and it works fine when reading /sys/class/thermal/thermal_zone0/temp
<montjoie>
anarsoul: I test your thermal serie, and I got cat: /sys/devices/virtual/thermal/thermal_zone0/temp: Invalid argument
2019-12-20
<montjoie>
SMB...
<montjoie>
but there are too many users...
<montjoie>
it can pass
<montjoie>
joke, no joke who knows
<montjoie>
1st april is not fat, I could try it on this day
<montjoie>
I try to see if any non-crypto driver use it
<montjoie>
for made the linecount god happy
<montjoie>
time to try a BIG PATCH THAT REMOVE IT (along with DES3)
<montjoie>
I dont know why Linux keep it
<montjoie>
hé hé hé
<willmore>
montjoie, how much to disable DES?
<montjoie>
or more evil, enable AES 1024
<montjoie>
or money. I will create a kickstarter for crypto, larger backer will get extra speed ah ah ahah
2019-12-17
<rellla>
maz, montjoie: it was the commit from linus on top of 5.5-rc2 btw
<rellla>
maz, montjoie: thanks, i'll give master chance then.
<montjoie>
rellla: no problem on my side (kernelci) on 5.5-rc2 gcc-8 sunxi_defconfig
<montjoie>
and when you saw the official advice "put your board vertically for fixing heat problem"...
<montjoie>
wens: I need to add rpi4 on CI, no uboot USB, no uboot net, my only way is to code some kexec fun
<wens>
montjoie: sh*t how? :p
<montjoie>
fALSO: I confirm, rpi4 is so s..t
<montjoie>
mmmh I have THERMAL_HWMON so I should see it
<montjoie>
anarsoul: I have tried your THS serie, but I saw no change, how to test it ? running sensor give no output for example
2019-11-27
<anarsoul>
montjoie: it applies more or less cleanly on linux-next, some dts patches are fuzzy, but it's OK. Anyway, I'll rebase it onto linux-next before sending v7
<montjoie>
I try to finish my boot-all-sunxi-patch-serie-bot
<montjoie>
thanks
<anarsoul>
montjoie: linux master
<montjoie>
I fail to apply it on mripard/for-next and linux-next
<montjoie>
anarsoul|c: on top of which tree your thermal serie is based ?
<jernej>
montjoie: I think there is none. 5.6 related branches will apear once 5.5-rc1 is released.
<montjoie>
mripard_: where is your "next" tree ? the one I find seems obsoletes
<montjoie>
mripard_: you said applied for pine64 model b but I still didnt saw them in linux-next
2019-11-25
<montjoie>
rules for young/old, you are always the young of someone, my grandfather was saying "noob you use transistor based electronics"
<montjoie>
it should
2019-11-22
<montjoie>
ah I just saw that "the ninth gate 2" scenario will be based on Allwinner BSP
<montjoie>
willmore: yes it is worse than any satan ritual or is a part of it
<montjoie>
but the "do RSA in symmetric mode" was the best alarm of the BSP driver usefullness
<montjoie>
I just check for confirming, in BSP the A33 SHA1 problem is not handled...
<willmore>
montjoie, then implementing from scratch makes even more sense. :)
<montjoie>
I am nearly sure it was never tested
<montjoie>
willmore: the BSP driver is more awfull than using old API
<willmore>
wens, I wasn't sure if montjoie started from a clean slate or worked to move the BSP driver over. If it were the latter, I would assume it would have been written in the most out of date API possible. Knowing vendors...
<willmore>
montjoie, oh, good. I got worried when I read that article that you're have another hurdle thrown in your path.
<montjoie>
willmore: wens: yes all sunxi crypto uses skcipher
<montjoie>
swap over NFS in case of missing RAM size
<montjoie>
especially after midnight
<montjoie>
NEVER build on sdcard
2019-11-17
<tuxd3v>
montjoie, thanks for your confirmation :)
<montjoie>
megi: yes I test BE currently, I will enable it on kernelci for all sunxi soon
<megi>
montjoie: you can run H3 in big endian mode?
<montjoie>
fun, cryptsetup crash only on (BigEndian && opi1+)
<montjoie>
tuxd3v: yes sun4i-ss provide ciphers, hash and prng
2019-11-14
<montjoie>
endianess killed me
2019-11-13
<montjoie>
it is a valid compilation target, so it must works
<montjoie>
no, fixes for arm bigendian and for A33
<montjoie>
wow such stack of fix for sun4i-ss comming
2019-11-12
<montjoie>
pfff SS fail on sha1 test on A33 only, probably since the beginning, nobody never reported...
<montjoie>
and listenning some reversed metal
2019-11-09
<montjoie>
uhubctl to the rescue!
<montjoie>
raaah the serial leak power ton the pineh64
2019-11-08
<montjoie>
I saw higher price
<montjoie>
it is cheap
<montjoie>
for our LAB we use cambrionix HUB, but it is 110livre sterling for 8 ports
<montjoie>
for testing you need to automate power cycling
<montjoie>
Power Distribution Unit
<montjoie>
it is an amazon basics hub which have per port power via uhubctl
<montjoie>
someone from NXP give me their cheap power PDU reference
<montjoie>
if you want to automate testing, I recommand to use LAVA for board management
<montjoie>
pineh64 modelA availlable for kernelci
<montjoie>
yes! it works
<montjoie>
and for the moment my ethernet USB dongle is not detected
<montjoie>
it dont
<montjoie>
fALSO: yes I mean for uboot
<montjoie>
smaeul: does your ATF serie is sufficient for having emac in uboot for h6
<montjoie>
I cannot find them
<montjoie>
could someone point me the patch for H6 emac in uboot ?
2019-11-05
<montjoie>
I dont understand why they dont have simple made it x4
<montjoie>
sŧill contention on one register (the start task)
<montjoie>
and I have a patch for boosting even further sun8i-ce
<montjoie>
but the result is still in favor of sun8i-ce
<montjoie>
but sun8i-ce can do only one algo class (rng, sym, asym) at a time and the sun8i-ss could do two
<montjoie>
wens: at least the perf is better
<wens>
montjoie: seems like the newer sun8i-ce is much better, as it uses task descriptors?
2019-11-01
<montjoie>
yes
<montjoie>
ah ah ah 3 crypto driver (with both allwinner ones) in one linux release
2019-10-30
<tuxd3v>
thanks montjoie
<montjoie>
tuxd3v: you could ignore the syscon warning
<montjoie>
thanks megi
<montjoie>
yes adding the "connector:" label did the trick
<megi>
montjoie: grep for hdmi-connector in base dts
<montjoie>
I got Label or path connector not found
<montjoie>
jernej: your "patch" fail to build
2019-10-29
<montjoie>
jernej: thanks
<jernej>
montjoie: you still need that node in pine64 model A DT
<montjoie>
okay so only ethernet remains different
<megi>
montjoie: my ddc-en-gpios patch for the hdmi driver was merged upstream, so dts change should work
<montjoie>
or perhaps it will be easier that you send a patch after mine
<montjoie>
jernej: I will resend the pineH64 model A serie, could you confirm that your patch is still needed and working https://lkml.org/lkml/2019/8/16/309
2019-10-28
<tuxd3v>
montjoie: thanks for the info
<montjoie>
tuxd3v: TRNG is on my TODO list for sun8i-ce
2019-10-25
<smaeul>
montjoie: megi: I posted a u-boot patch for the necessary driver changes a couple of weeks ago
<montjoie>
tomeu: someone is working on H6 ethernet for uboot, but cannot remember who
<tomeu>
montjoie: guess there isn't support for it in u-boot yet?
<montjoie>
tomeu: no the erthernet GMAC is real device
<montjoie>
try all ports
<tomeu>
montjoie: yeah
<montjoie>
tomeu: you mean an "USB ethernet dongle" ?
<tomeu>
montjoie: I'm having trouble getting the eth probed in u-boot on a h64
2019-10-19
<montjoie>
on aliexpress I mean
<montjoie>
on allwinner I found some probable real FTDI but nothing less than 6€
<montjoie>
wow it seems that we can choose to encrypt only some sector
<montjoie>
ah there's more EMCE register directly in sdhc and ndfc controller
<willmore>
montjoie, yep, that's the normal way to do it. Ask at boot.
<montjoie>
a possible way is protect the key by a password and ask it at boot
<montjoie>
it is readable by anyone
<montjoie>
but having to store the key somewhere readable is...
<montjoie>
If I understand well the limited documentation...
<willmore>
montjoie, understood.
<montjoie>
it seems to intercept MMC/NAND "traffic"
<montjoie>
willmore: EMCE bypass dmcrypt/luks
<montjoie>
another solution is to have sdcard only for starting and rootfs on encrypted NAND since EMCE support the NAND
<montjoie>
so I have to add support for it to uboot
<montjoie>
it supports aes-xts
<montjoie>
it seems that the encryption is for the whole sdcard or not
<montjoie>
and where to store the key ?
<montjoie>
but how to set this nicely in linux
<montjoie>
EMCE seems in middle of the mmc "bus" so transparent for the controller
<montjoie>
willmore: the more problem is that the documentation is SMALL
<willmore>
montjoie, LOL, true. I'm sure that it won't work because of layering issues.
<montjoie>
willmore: for mmc the EMCE should be more effective, but who know, it is allwinner stuff...
<montjoie>
willmore: for network, I never tested ipsec (but I should to)
<willmore>
montjoie, that's what I was afraid of. It makes it hard to use it for efficient disk or network encryption.
<montjoie>
it is on my TODO driver queue
<montjoie>
willmore: on H6 it exists a crypto device dedicated for MMC
<montjoie>
willmore: only from memory
<willmore>
montjoie, the encryption engine works like DMA in that it reads data and writes it out somewhere else (or back to the origin), right? Can it read/write from I/O devices like the SATA controller or SD controller? Or does it need to operate only on in-memory buffers?
<montjoie>
wens: yes
<montjoie>
for each request we still need to do lots of cpu action like dmamapping
<montjoie>
the ce already do dma
<montjoie>
the gain is due in fact to: less interupts and less kthread resume
<montjoie>
willmore: not sure to have understood your question
2019-10-02
<montjoie>
I use it also, but not really reliable
<montjoie>
aes-xts-plain64
<montjoie>
xts(ecb)
<montjoie>
200MB/s with dd on LUKS2/4096 not bad
<montjoie>
and I need to add a bench with a 4096 sector size LUKS, I expect it better
<montjoie>
the interest is that the speed is the same when cpu is loaded (last time I benched)
<montjoie>
jernej: not so slow, I need to rebench against arm-crypto
<jernej>
montjoie: so now HW crypto is actually useful?
<willmore>
montjoie, congrats.
<montjoie>
I batch by 8, let's try 32
<montjoie>
on all soce
<montjoie>
so I batch request, but doing that force me to go away the official "linux/cryptoengine" for managing queue
<montjoie>
I use the task list next which I didnt use before
<montjoie>
yeah it seems tham I just found an optimization which increase crypto engine by 20% up to 100%
2019-09-27
<montjoie>
perhaps it is possible, but I dont remember to have saw something
<montjoie>
KotCzarny: no WOL for sunxi
<KotCzarny>
montjoie: do stmmac support wake-on-lan? or in particular, ethernet on sunxi socs?
2019-09-20
<KotCzarny>
montjoie: ^
<montjoie>
send a patch for review as RFC/RFT
<montjoie>
since it is a modified R40, probably it uses the sun8i-emac
<montjoie>
tdebrouw: are you sure that V40 use sun4i-emac ?
2019-09-19
<montjoie>
not yet
<libv>
montjoie: did you try olimexes own uboot tree?
<montjoie>
anyway, I will put it on the testlab with 100M switch
<montjoie>
libv: my lime2 is rev: C, I still tried it and no change