2017-01-14

<montjoie> apritzel if you want, I didnt own H5
<apritzel> montjoie: I can test it on the H5 and A64
<montjoie> I mean test on a83t/a64
<montjoie> yes
<montjoie> I need to test it on non-h3
<montjoie> but stable when running
<montjoie> not clean, no remove functions:)
<MoeIcenowy> montjoie: how is sun8i-stmmac cleaned up?

2017-01-13

<MoeIcenowy> montjoie: sorry... no...
<montjoie> MoeIcenowy: did you test sun4i-ss like we speak one week ago ?

2017-01-10

<montjoie> apritzel one more warning, I think it could "break" other stmmac
<apritzel> montjoie: thanks a lot, looks quite good on a first glance!
<montjoie> warning: no cleaning path
<willmore> montjoie, packets on the wire and everything? Yay!!!! Congratulations.
<montjoie> apritzel: good to know you have also the problem:)
<apritzel> montjoie: didn't have time to debug this yet, have you tried mripard's sunxi/for-next branch, which has more DT fixes?
<apritzel> montjoie: it seems to work fine on the A64/H5
<apritzel> montjoie: btw: I see the same MMC issues on my OPi Zero with 4.10-rc, I believe it's a pinctrl/DT issue
<apritzel> montjoie: no hurry, and thanks a ton for doing this!
<montjoie> apritzel: I work on it, finish to split up patchs
<apritzel> montjoie: if you could push something (even WIP) somewhere, I can give it a try
<montjoie> I will try on my pine64
<montjoie> got some more perf by dont forgetting to compile the realtek phy, but 145Mbit/s is not enought
<MoeIcenowy> montjoie: cannot wait to see sun8i-stmmac! ;-)
<montjoie> nobody here with a bpi m2+ for benching it ?
<montjoie> my opipc have the mmc timeout bug, I need to rollback to 4.9
<montjoie> bpi m2+
<apritzel> montjoie: is that this infamous Pine64 GBit issue?
<montjoie> apritzel KotCzarny, in fact sun8i-emac got the same performance on that board:)
<apritzel> montjoie: if you have something that's stable, please post it
<apritzel> montjoie: I wouldn't care about performance at the moment
<apritzel> montjoie: awesome! great work!
<montjoie> plaes: did you still get the mmc timeout ?
<montjoie> strange
<montjoie> something weird and somewhat related, the mmc work on that board with the same kernel that made mmc timeout on opipc
<montjoie> stmmac is full NAPI, so I need to add some "int stop"
<montjoie> I think I need to check some interrupt handling
<montjoie> on a gigabit link:(
<montjoie> Mbit/s
<montjoie> but at least the link is stable
<montjoie> something is missing
<montjoie> iperf 80M/s and 124M/s
<montjoie> ouch
<montjoie> no perf test for the moment
<montjoie> and just add a new comment "softreset MAC after any syscon change"
<montjoie> I need to made the code beautifull, but yes the hard work is done
<montjoie> just sshed into sun8i-stmmac:)

2017-01-09

<willmore> montjoie, well, the ARM part of the CPU could be off, but the AR part can sill live.
<montjoie> and need to connect hard drive also
<montjoie> My primary idea was to use ATX because I have some old shuttle and so have an embedded DC into
<KotCzarny> montjoie, if you just want dual or single voltage, you can get 65W or 200W dedicated psu
<montjoie> mripard: happy to see that I got the same idea than free-electron with ATX:)
<montjoie> mripard: thanks:D
<mripard> montjoie: you might want to have a look at http://free-electrons.com/blog/hardware-infrastructure-free-electrons-lab/
<montjoie> but power control is the major objective
<montjoie> KotCzarny: uart muxed is also a goal:)
<montjoie> esp8266 is already ordered for test
<deskwizard> montjoie: depending on how remote you want, you could use an esp8266 instead of the arduino, port expander (so you can add another 16 ports later) and a dumb relay board
<montjoie> my first proto is the same (without the usb)
<montjoie> KotCzarny: exactly
<montjoie> deskwizard: all managed board could need to be used at maximum (for benching/testing driver)
<montjoie> so nothing on the board can be used
<montjoie> a state with really no current
<montjoie> willmore: I want also a power up
<deskwizard> montjoie: what kind of load?
<willmore> montjoie, could apritzel's code that runs on the OpenRISC core do something like that?
<montjoie> so I think USB relay will need too much USB HUB:)
<montjoie> 10 at minimum
<montjoie> beeble: thanks but 124¢ is too much:)
<KotCzarny> montjoie: there are relays with usb control
<beeble> montjoie: something like that? http://energenie.com/item.aspx?id=7557
<montjoie> I begin to try with arduino/relay but perhaps it exists a better way
<montjoie> it seems that manageable USB hub doesnt exists
<montjoie> does someone know here a way to do an "IPMI" for ARM cards ? I need remote power control
<willmore> montjoie, details, details. :)
<montjoie> willmore: thanks but wait for the great until packet are sent on the wire...:)
<willmore> montjoie, yay! Thank you for such great work!
<montjoie> normal
<ErwinH> montjoie: The size of the request seem to play a big role. The bigger the request the longer it will take. request 16 does 13180r/s vs 29931r/s
<Nico___> #montjoie for info i had to add NAND definition in nand_ids file
<montjoie> cubieboard2
<Nico___> #montjoie wirh board A20 based do you have ?
<montjoie> will try it on my a20 thanks
<Nico___> #montjoie i tied wihe a Lime_2 board image and it works
<montjoie> for me nand is a total WIP
<montjoie> didnt know it should work on A20
<montjoie> didnt try yet
<montjoie> perhaps tuning the clock is a way for more perf
<montjoie> thanks
<montjoie> could you paste the error in private ?
<montjoie> but good to know it works on H5
<montjoie> like sun4i-ss with DMA
<montjoie> the code is not optimized but yeah, my last test was software/4 or /10
<montjoie> :)
<montjoie> ErwinH: how do you test it ?
<montjoie> only two rename from ahb to stmmaceth
<montjoie> wens: all is done my the same syscon function and I kept the same DT nodes
<montjoie> wens: not a problem
<montjoie> and the good news is that the amount of hack is low (lower than I expected)
<montjoie> good news, sun8i-stmmac is really near to work, just packet are not sent on the wire, (but they leave the MAC)

2017-01-07

<montjoie> plaes: could you try to bisect your problem and confirm it is the same ?
<montjoie> the problem come with "Merge tag 'pinctrl-v4.10-1' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl"
<montjoie> "must not" is the better word
<montjoie> bisecting between 4.9 and 4.10rc1 CANNOT go outside this interval
<montjoie> but clearly git bisect bug
<montjoie> plaes: the cmmit is f233dbca6227703eaae2f67d6d9c79819773f16b pinctrl: sunxi: Rework the pin config building code
<montjoie> all my bisect go to a mripard patch in 4.9 rc, but 4.9 works. It is really strange
<montjoie> plaes: happy to see i am not the only one to have this error:)

2017-01-06

<apritzel> plaes: montjoie is chasing the very same problem
<montjoie> bisecting by hand ...
<montjoie> raaaah second time that git bisect compress time space, and go ousite the good-bad limit

2017-01-05

<montjoie> i bisect between 4.9 and 4.10rc1 and the bad commit is in 4.9rc2. I dont understand at all
<montjoie> apritzel: probably:)
<apritzel> montjoie: btw: I think the "update clk timeout" message is a follow up error, the actual culprit is probably the data error above
<montjoie> no reverting to last kernel work
<apritzel> montjoie: and you are sure that it's not the SD card which is broken?
<montjoie> apritzel: allwinner,sun7i-a20-mmc
<apritzel> montjoie: I meant the compatible string on the MMC node ...
<montjoie> apritzel: opipc, I have begin to bisect
<apritzel> montjoie: what does your MMC node like? which compatible string?
<montjoie> it seems that its a DT change which cause that
<montjoie> dont know what disable them
<NiteHawk> montjoie: disabling the vccs seems fishy - is that specific to the mmc (1.7v card?), or might it bork the peripherals entirely?
<montjoie> so nobody got this http://pastebin.com/GLMqvP94 ?
<montjoie> apritzel: nothing for the moment
<apritzel> montjoie: so do you have something that somehow works?
<apritzel> montjoie: "made" as in: "it's done now"?
<montjoie> stmmac made me crazy
<montjoie> but if someone know the fix:)
<montjoie> typed enter too early, seems that some people already get it
<montjoie> updated my opipc to 4.10rc1 and get "sunxi-mmc 1c0f000.mmc: fatal err update clk timeout

2017-01-04

<montjoie> I read the datasheet you pointed yesterday
<montjoie> V3s I should say
<montjoie> according to datasheet it got one
<montjoie> MoeIcenowy: I see yo have a V3 SoC, could you test(activate in DT) sun4i-ss on it ?

2017-01-03

<montjoie> on A20 without the previous commit, it happend data corruption
<montjoie> and then compare the written file with the origin file
<montjoie> write to it
<montjoie> it create a luks volume
<montjoie> no
<montjoie> no this commit is old
<montjoie> crypto: sun4i-ss - Replace spinlock_bh by spin_lock_irq{save|restore}
<montjoie> it is the commit id of linux
<montjoie> MoeIcenowy: I try to know if the bug solved is specific to A20
<montjoie> MoeIcenowy: could you try https://github.com/montjoie/cryptotest/blob/master/luks/lukstest on them ? and after reverting bdb6cf9f6fe6d9af905ea34b7c4bb78ea601329e
<MoeIcenowy> montjoie: ?
<montjoie> anybody here with A10 A13 A31 A33 who can test something with sun4i-ss ?

2016-12-26

<montjoie> the probe seems good, now I need to populate dma functions
<montjoie> Unable to handle kernel paging request at virtual address f0aad058 exactly
<montjoie> WIP:)
<MoeIcenowy> montjoie: what's the progress of dwmmc modify?
<montjoie> igraltist: the patch could be easily get from git hub
<igraltist> montjoie_: could not found, looks wrong search pattern
<montjoie_> igraltist: why not using directly the patch ?

2016-12-25

<montjoie> allwinner love to complexify naming
<montjoie> raaaah
<montjoie> renamed to CE
<montjoie> the user manual is clear, R8 is a SS
<montjoie> I will try to have access to thoses SoCs for checking
<montjoie> i know that R40 is probably a CE
<montjoie> so if it is really a CE, sun4i-ss will not work
<montjoie> the problem is that CE is the new SS generation
<montjoie> but strange that SS is renamed to CE
<montjoie> BlasterBlade yes crypto engine interrupt

2016-12-21

<montjoie> checkout v2016.11 make the issue disapear
<montjoie> yes
<likewise> montjoie: upstream master, right?
<montjoie> uboot git master yes
<likewise> montjoie: what uboot GIT?
<montjoie> does someone recently compiled uboot ? it failed with "SyntaxError: Missing parentheses in call to 'print'"
<montjoie> yes need more than 4:)

2016-12-19

<montjoie> so . is portable:)
<montjoie> always use only '.', source is a bashisms
<montjoie> MoeIcenowy: you mean you want to decode all address space ? and not only the ID
<MoeIcenowy> montjoie: this sun8i-sid can only read 4 bits
<montjoie> MoeIcenowy: sun8i-sid was already under work, just need to send patch

2016-12-14

<montjoie> wens: any way to debug it ?
<wens> montjoie: unaligned access using aligned access instructions?
<wens> montjoie: unaligned access?
<montjoie> KotCzarny: not fried
<montjoie> does someone know what could cause data abort(http://pastebin.com/uKxhD2X7) with uboot ?
<jski> if I take the latest stable mainline kernel will montjoie patch still fix ethernet?

2016-12-12

<alain__> montjoie: no worries, not a big issue for me, HD works well enough. Was more to report the issue also so that someone else looking in the IRC log sees s/he's not alone.
<montjoie> sorry alain__ didnt have any idea
<montjoie> no more info in dmesg ?
<montjoie> :)

2016-12-01

<montjoie> starting to really modify stmmac, not easy, all dma_ops need to be rewritten, some core file also

2016-11-29

<montjoie> wens: I will add them to V5:)
<wens> montjoie: see the last 2 patches # https://github.com/wens/linux/commits/h3-audio-emac
<montjoie> :)

2016-11-28

<montjoie> but it is strange that anything hit sun4i-ss, since FS encryption use CTS and XTS block mode
<montjoie> all howto I follow fail for ext4 encryption:(
<igraltist> montjoie: the problem with e4crypt is that u have to build a tool that store a hash of password and given dir to avoid mistake
<igraltist> montjoie: no i stop this i keep my cryptsetup
<igraltist> montjoie: i had use this cryptSalt as here
<igraltist> montjoie: http://paste.debian.net/899491/
<montjoie> igraltist: like e4crypt add_key -S s:425341754425341754 && e4crypt set_policy 2cd4da5d322b28ac temp/ but that example does not work
<montjoie> igraltist: does not work, clearly some keyctl command is missing
<igraltist> montjoie: http://paste.debian.net/899487/
<montjoie> igraltist: I still wait for command creating policy for ext4crypt
<montjoie> miasma: you could copy it, but dont forget colors:)
<montjoie> the problem is that it duplicate content
<montjoie> where exactly in the wiki ?
<miasma> montjoie: but it's ok to copy it to the wiki?
<miasma> montjoie: on the todo list :)
<montjoie> miasma: where are my beautilfull colors ?
<miasma> montjoie: if it's ok to include the crypto status table to the wiki, there's a version here https://linux-sunxi.org/User:Miasma/cryptotable

2016-11-26

<apritzel> montjoie: that's great!
<montjoie> :)
<montjoie> yes
<apritzel> montjoie: have you started looking at this already?
<apritzel> montjoie: well, a working DW based driver would be proof enough, I guess
<montjoie> apritzel: official comm to users:)
<apritzel> montjoie: I guess they won't tell you what IP they used
<apritzel> montjoie: why do you need official acknowledgement?
<montjoie> wens: i have already a contact for CE, but past first question, another void
<wens> montjoie: you could ask mripard for an introduction to a contact at allwinner
<montjoie> I need to do an offcial communication about emac mainlining
<tkaiser> teo__: Are you sure montjoie's driver is included in your kernel? I used MoeIcenowy's latest branch with USB and Ethernet working: https://github.com/Icenowy/linux/commits/ice-a64-v6.1 (only tx-delay/rx-delay is missing, I would suggest to use 2/0 on production Pine64+)
<montjoie> igraltist: what is "my_policy" ?
<igraltist> montjoie: i try the sunxi-ss from 4.9 but the same: sun4i-ss 1c15000.crypto-engine: ERROR: Empty IV

2016-11-25

<igraltist> montjoie: would be enough just replace the drivers/crypto/sunxi-ss with 4.9 files
<igraltist> montjoie: i stuck on lt
<montjoie> igraltist: do you have tried with 4.9 ?
<igraltist> montjoie: and just e4encrypt set_policy
<igraltist> montjoie: Linux jaschtschik 4.4.34-rsbac-2+ #2 SMP PREEMPT Thu Nov 24 04:21:51 CET 2016 armv7l Allwinner sun7i (A20) Family GNU/Linux
<montjoie> igraltist: could you share full command list to error
<montjoie> igraltist: which kernel ?
<montjoie> someone reported some month ago, cannot reproduce it for the moment
<KotCzarny> montjoie: ^
<ErwinH> Just these 2 patches are enough to get a thermal readout on the H3 boards: https://github.com/megous/linux/commit/3df4f5100da8d93b54e2f372f8108ea11f88ba87 & https://github.com/megous/linux/commit/6875fd897e344fd54a7bcdfa34ccbff04fcb8d03 (even on montjoie's branch)

2016-11-24

<apritzel> MoeIcenowy: dunno, montjoie started to look into this, AFAIK
<KotCzarny> hmm, make designware work with montjoie's driver then
<apritzel> so unfortunately montjoie's driver will probably go nowhere
<MoeIcenowy> montjoie: what's the mainlining status of sun8i-emac?
<montjoie> for x86
<montjoie> i mean optionnal
<montjoie> extention for x86, but mandatory to have for x86_64

2016-11-23

<apritzel> montjoie: page 232 contains a new diagram and pages 240 contains a whole page of "programming guidelines"
<apritzel> montjoie: sing and rejoice!

2016-11-22

<MoeIcenowy> montjoie seems to have both
<montjoie> tkaiser: there are some rewriting for CE, but not enough
<montjoie> lol like the printing "comment" and "title" on some pcb I have
<montjoie> still not enough info for RSA:(
<montjoie> yeah user manual!

2016-11-21

<ErwinH> I'm not using megi's branch, but the montjoie's sun8i-emac-wip-v5 branch, with just the 2 commits I linked.
<montjoie> now trying TRNG
<montjoie> A83T PRNG work

2016-11-20

<xdeniz> who can i ask for this topic montjoie? i have to install trustzone on this board (optee) and i need help
<montjoie> I.MX6 is clearly out of topic here
<willmore> montjoie, yay!
<montjoie> when I will successfully made TEE work, H3/A64 will have 8 stream
<montjoie> but the dream of having a day a CE with more than 8 stream...
<montjoie> since no PSCI:(
<willmore> montjoie, more than 8?
<montjoie> yeah A83T is back with crypto engine, and more crypto stream than number of cpu:) (since no PSCI)
<montjoie> wens: I lost my a83t source tree, which branch is the best, our a83-emac ?

2016-11-18

<montjoie> the title said sunxi-ng but no answer on his patchset
<wens> montjoie: he has his own version
<montjoie> and it seems that jfmoine work on clk-ng for a83t
<montjoie> in fact it is written in "core stuff"
<montjoie> at least for putting it in "job todo"
<montjoie> wens: do you know the complexity of doing it ?
<montjoie> -it +he
<montjoie> it dont work on it now ?
<wens> montjoie: last attempt was from vishnu, iirc that was the old style clocks
<montjoie> a83t clock is set as WIP, but I cannot see who work on it

2016-11-17

<montjoie> like getmaintainer.pl for linux

2016-11-16

<montjoie> update sun8i-ce with A64 support (A80/A83T is coming)

2016-11-14

<montjoie> my phone got a 3.4.1:)
<montjoie> yeah, it seems A64 TRNG is working
<montjoie> longsleep: with generic/arm linux implementation
<montjoie> longsleep: last bench said perf/10
<longsleep> montjoie: did you make some speed tests of the A64 crypto stuff? I disabled their kernel drivers for my Pine64 tree if i remember correctly
<montjoie> KotCzarny: A64 have HMAC-sha256, so it was not for gaining space
<montjoie> now I need to find why PRNG do not work
<montjoie> dont know why they removed it
<montjoie> just it is the same Crypto Engine
<montjoie> no
<KotCzarny> montjoie: is it advertised anywhere that it is?
<montjoie> lol found why sha384/512 doesnt work on A64, simply it do not support it
<montjoie> strange A64 is getting DMA timeout for PRNG/SHA384/SHA512

2016-11-11

<MoeIcenowy> montjoie: so does H2+ and H5
<montjoie> already sold and still no user manual:(

2016-11-10

<montjoie> The only column in black
<montjoie> I agree:)
<miasma> montjoie: would it be ok to have a crypto engine page in the sunxi wiki with possibly content copied from your page or a link to that page?
<montjoie> I do not remind who request it, but I have updated all my github branch (and so SS/CE with PRNG)

2016-11-09

<montjoie> Keziolio: I wait for an answer on v4 for sending v5
<Keziolio> sorry montjoie i'll pay you a beer
<montjoie> Keziolio: after boot, do you have some packet loss ? ethtool -S eth0 after some time could be usefull
<montjoie> Keziolio: your 3 last mac digit asaid that you have an old uboot
<montjoie> Keziolio: what said dmesg ?
<montjoie> do you have emac support in uboot ?
<montjoie> so you said to have some disconnect ? anything in dmesg ?
<montjoie> v5 ?
<Keziolio> montjoie: it was yours