2019-01-14

<hanetzer> montjoie: CONFIG_DEVTMPFS{,_MOUNT}=y
<montjoie> check also DEVTMPFS CONFIGs
<montjoie> replace init content by a shell script which debug-echo all necessary stuff for runing init
<montjoie> hanetzer: could you paste the console output ?
<montjoie> note that my personnal lab is still not connected to kernelci but it should be soon
<montjoie> plaes: a20 in my lab and seatle, a10 only in our seatle lab
<plaes> montjoie: do you have a10/a20 devices in your lab?
<montjoie> note that there will be an h6 board in our kernelci lab in some days/week

2019-01-11

<montjoie> just in case, I have released my helper script for testing my drivers (and also building and booting linux on lots of targets) https://github.com/montjoie/bbci any comment welcome

2019-01-09

<hanetzer> montjoie: as in, not persuing it? its mostly because I'm a bit embroiled with rockchip stuff atm (u-boot for the rk3288 veyron-speedy chromebook and acquiring a rk3399 samsung chromebook plus for u-boot, and mali work on both), I don't have time for yet another SoC in my life :P
<hanetzer> montjoie: hmm?
<montjoie> hanetzer: it is the better route:D!
<hanetzer> montjoie: very true. I've went ahead and gone gentoo route
<montjoie> hanetzer: there are no good suggestions without some needs/goal

2018-12-21

<MoeIcenowy> montjoie: I tried to let a Banana Pi M2+ to work with fixed link ;-)
<montjoie> MoeIcenowy: I do not have any board with fixedlink, so hard to work on this
<MoeIcenowy> montjoie: I realized a problem that a MDIO mux is not the best way to represent the mux on H3

2018-12-06

<montjoie> crmingle: ask your final question just in case

2018-12-01

<jernej> montjoie: bisect came up by commit a66d972465d15b1d89281258805eb8b47d66bd36, but if I revert it, it doesn't help
<montjoie> but got random memory problem
<montjoie> I ran next 2 or 3 days ago
<jernej> montjoie: Kernel prints "Starting Linux kernel" on uart and nothing more
<montjoie> jernej: symptom ?

2018-11-29

<montjoie> how do you trust kernel from people doing RSA in CBC mode
<montjoie> always use the mainline!

2018-11-28

<montjoie> not sure if it is on a64 or another one
<montjoie> jaganteki: hit it also

2018-11-16

<montjoie> but mripard will confirm if it is know or not for sun4i_drm_hdmi
<montjoie> seems lots of things imbricated
<montjoie> at least on meson drm, its maintainer said "it's complicated to fix it"
<montjoie> like lots of other drm modules

2018-11-15

<karlp> montjoie: depends what you're virting
<montjoie> no enough RAM for virt
<montjoie> wens: got one

2018-11-11

<montjoie> wens: mripard I just found this patch https://lkml.org/lkml/2018/7/11/144 is still not in next, does it is lost ?

2018-11-09

<jernej> montjoie: problem solved by selecting stmmac driver as a built-in
<jernej> montjoie: I enabled all axp related modules, but it still doesn't work
<montjoie> jernej: vcc-phy: disabling perhaps is the cause
<heatfanjohn> montjoie: A64 on PineA64+ board
<jernej> montjoie: would you know from this dmesg https://pastebin.com/raw/SgC3NTZ2 why ethernet fails on pine64+?
<montjoie> heatfanjohn: which SoC ?
<montjoie> mdipolt: does this setup is in wiki (if not it could be worth the cost to add it)

2018-11-08

<montjoie> wens: full messages: https://pastebin.com/adUauUHN
<montjoie> wens: it is a sdcard on bpim3
<montjoie> I got some "sunxi-mmc 1c10000.mmc: fatal err update clk timeout" on A83T, dangerous ?

2018-11-03

<montjoie> ok so I understand the "only one DT"
<MoeIcenowy> montjoie: no
<montjoie> I believed we had final version
<montjoie> MoeIcenowy: ES ?
<montjoie> and how to find difference between A and B in DT ?
<montjoie> MoeIcenowy: you said "dont test on model A" but what are the consequence of doing it ?
<MoeIcenowy> montjoie: sleeping 1s after regulator_enable really helps
<MoeIcenowy> montjoie: I doubt a delay after regulator is enabled is needed
<MoeIcenowy> montjoie: on the new Pine H64 model B the behavior is a little strange

2018-10-17

<montjoie> what did you do for testing audio ? did you have a test plan ?

2018-10-10

<montjoie> a div problem
<montjoie> note that i have also error when playing 32khz audio

2018-10-06

<montjoie> at least I booted arm64 big endian

2018-10-05

<montjoie> perhaps or megous sources:)
<jonkerj> montjoie: this is very strange. my sources (megous' orange-pi-4.19 branch) with your kconfig yield same phy errors
<jonkerj> if that does not help, I'm going to run it with exactly your config, montjoie. If that doesn't work, u-boot could be suspect
<montjoie> it is hungry of CONFIG
<montjoie> so BE die after started /init
<montjoie> at least you have found why
<montjoie> now you can diff to see what was missing
<montjoie> given in pm
<montjoie> do you want my .config ?
<montjoie> 4.19.0-rc6-next-20181004
<montjoie> just booted my pine64, it works
<jonkerj> montjoie: I'm going to try if a different PSU makes any different wrt mdio/phy error on pine64+
<montjoie> and perhaps found some crash in other driver to fix:)
<montjoie> no but I want to be sure to submit a good driver
<montjoie> i need to test my crypto driver on BE
<montjoie> in fact it booted it was just very long:(
<montjoie> for testing purpose
<anarsoul> montjoie: but why?
<montjoie> My BE kernel is stuck at "booting the kernel"
<montjoie> does someone already booted big endian kernel on sunxi ?

2018-10-04

<montjoie> for me you cannot have DWMAC-SUN8I without STMMAC_PLATFORM
<montjoie> yes
<montjoie> tomorow I will have acces to a pine64@work and will try
<montjoie> and i saw lofs of regulator dummy did you enabled all AXP stuff ?
<montjoie> i never seen MDIO device at address 1 is missing.
<montjoie> at least kernelci said ok https://kernelci.org/boot/sun50i-a64-pine64-plus/
<montjoie> jonkerj: could you paste dmesg ?
<montjoie> jonkerj: just DWMAC-SUN8I is enought
<montjoie> jonkerj: perhaps it is related to the syscon problem
<jonkerj> montjoie: I've got similar issues in 4.19-rc6
<montjoie> mine has died
<montjoie> I will try to test on work's one
<montjoie> but I didnt hit anything with next
<montjoie> jonkerj: I will test tomorow

2018-10-03

<montjoie> MoeIcenowy: do you have a recent branch with patchs for H6 dwmac-sun8i ?
<montjoie> ok
<montjoie> wens: why I dont see them in next ?
<wens> montjoie: specifcally, check the tip of regulator/for-4.20
<wens> montjoie: fyi I see additional patches to fix the regulator GPIO descriptor issue, in both the regulator and gpio trees

2018-10-02

<pgreco> agreed, that cable is the first step, but that feels way too similar to montjoie's problem with his bpi-m3
<pgreco> I'm trying to fix https://bugzilla.redhat.com/show_bug.cgi?id=1530582 , which looks a lot like montjoie's bpi-m3 gmac problem
<montjoie> anyone booting sun8i-h3-libretech-all-h3-cc with recent next ?

2018-09-21

<messpl> montjoie: thanks, i will check it
<montjoie> messpl: you have the SS prng

2018-09-19

<montjoie> this patch enable emmc on the xxx. The change was copier from the dt of xxxx
<montjoie> tjader: your commit message is tiny
<montjoie> it will prevent some comment
<montjoie> tjader: also use checkpatch
<montjoie> always use get_maintainer.pl and after you can add more people if needed
<montjoie> since it's DT file you will have additionnal people
<montjoie> tjader: use get_maintainer.pl

2018-09-13

<montjoie> buildroot ?

2018-09-12

<montjoie> timdorohin: I wasnt joking on gentoo, I use it!
<timdorohin> montjoie: h3 can build kernel in under 30 minutes (with optimized config).
<montjoie> bmccarter: so what is your problem ?
<montjoie> bmccarter: gentoo !!!!!!!

2018-09-11

<montjoie> karlp: try ip -6
<montjoie> karlp: no address start with 33 with ip ?
<montjoie> well, seems many multicast address
<montjoie> karlp: could you recompile your kernel with a patch ?
<montjoie> dont know
<montjoie> probably it's busybox ip and not iproute2
<karlp> montjoie: ip maddress? ip address I pasted earlier, just 1 v4 plus 2 v6
<montjoie> I kill my enemies by making them watching old strange movies, this morning a coworker said she like Sean Connery, tonight she will see zardoz as I suggested
<montjoie> ah ah I am a troll hunter, to be sure to remain the only one
<willmore> Okay, I used to like you, montjoie . :P
<montjoie> #likesystemd (troll sorry)
<montjoie> my previous jobs, network guys had a megaphone for "not network fault"
<KotCzarny> montjoie: you still can make some docs, it might be often asked
<montjoie> taking coworkers by surprise
<montjoie> now I can ran in the office yelling "ah ah not a bug"
<megi> montjoie: I had > 8 link entries there :), I see the limit in the driver is 8 addresses
<montjoie> could yo paste ip maddress
<KotCzarny> karlp, there is a possibility driver still needs fixing some bug, talk to montjoie
<megi> montjoie: ip output looks quite regular, and I have default config with one IPv4 address and one IPv6 address
<montjoie> so it is usefull to have an "ip a" output
<montjoie> perhaps you have lots of multicast and or ipv6 spetial address
<montjoie> Too many address mean too many mac address in the filter list
<montjoie> whats 's the ip a output

2018-08-30

<montjoie> last obstable for shipit is ipsec testing
<montjoie> cool no more luks crash with sun8i-ce

2018-08-13

<montjoie> wens: still no answer for a new bpi-m3 on my side:( but at least the problem is fixed now

2018-08-12

<montjoie> input/output error on any read/write on mmc
<montjoie> got the same on R40
<montjoie> did someone hit some mmc error with H3 on linux-next ?

2018-08-09

<rex> hey montjoie, once my new display is here, I will try your branch. atm I won't be able to reboot, once I shut it down, cause I have no display :(
<montjoie> so yes all classic make options exists
<montjoie> rex: its a linux-next + patches
<montjoie> yes
<montjoie> you can try my crypto wip branch on github
<montjoie> yes (shame)
<montjoie> all kernel test pass, but when doing any action after mouting a luks partition...
<montjoie> I have a crash when using cryptsetup
<montjoie> rex: yes
<rex> montjoie: Aren't you working on the crypto part?

2018-07-26

<libv> montjoie: we are looking at just over half a TB
<montjoie> size ?
<montjoie> libv: If volumetry can fit on my server
<montjoie> do you know data size and access volumetry ?
<libv> montjoie: i would then also want to get rid of both the expense and the legal liability of this machine
<montjoie> I have already proposed to do the sysadmin task for it
<montjoie> and also MoeIcenowy
<montjoie> could you set wens and mripard as ops ?

2018-07-23

<montjoie> or wait for H7 ?
<montjoie> MoeIcenowy: I was just trolling... my ":D" seems not enough
<MoeIcenowy> montjoie: because it's broken
<montjoie> removing PCIE ... why ? :D

2018-07-21

<montjoie> this is my regression suite https://github.com/montjoie/lava-tests
<montjoie> anything bad on mmc
<montjoie> I hoped that something different exists (like smartctl for hdd)
<montjoie> but I seek also anything which can make proof that mmc work
<montjoie> I will
<montjoie> system unusable after so I dont know if something is printed, but I least I want to activate it
<montjoie> detect it automaicly
<montjoie> I have uptime+60s problem with mmc on linux-next, and I want to detect it
<MoeIcenowy> montjoie: make a SDIO card ;-)
<montjoie> nothing wrong, I want something more/different
<montjoie> appart from "try to read with dd", any idea on how to test if mmc driver work ? (non destructive way)

2018-07-20

<willmore> montjoie, which ATM? Automated Teller Machines or Asynchronous Transfer Mode? Is the old floppy code beautiful? The controller and hardware sure weren't.
<montjoie> or some old ATM devices
<montjoie> willmore: for beautifullness, did you check some old code from device nobody uses ? (just floppy for example)

2018-07-19

<montjoie> berlin ahci is 0xe90000 0x1000 and sataphy is 0xe900a0 0x200
<montjoie> wens: like berlin with a subaddress space
<montjoie> MoeIcenowy: you are right, since some PHY code is used, it will be easier to create a sata phy driver
<montjoie> MoeIcenowy: I have created a fake PHY and I will ask for advice phy people
<MoeIcenowy> montjoie: I suggest you make a new phy driver

2018-07-18

<montjoie> MoeIcenowy: instead of adding more code to ahci for phy, I have modelled a fake phy and use already existing phy-supply. But since the phy need to load, a fake-phy compatible (and fake driver) is needed. What do you think about it ?

2018-07-16

<montjoie> wens: does this DT patch https://pastebin.com/TDnZeK7m seems correct for you ?
<wens> montjoie: you can't, so you should change the name of one of them

2018-07-15

<montjoie> MoeIcenowy: so i dont understand, how can two clock exists with the same name ?
<montjoie> and nobody respect it
<montjoie> but it fail when respecyin the binding
<montjoie> I have a patch for not crashing and just fail
<montjoie> but if I respect the binding it crash*
<MoeIcenowy> montjoie: the SoC has a internal mux which is the input of CCU
<montjoie> I have problem understanding sun6i-rtc bindings, it create osc32k clock, but osc32k already exists in DT

2018-07-13

<montjoie> atsampson: you miss a fan for hdd
<montjoie> atsampson: nice but my goal is higher than just cluster, I want Ci for all stuff that affect my boards
<atsampson> montjoie: for amusement value, here's my kitchen table build cluster: https://stuff.offog.org/armbox-closed.jpg https://stuff.offog.org/armbox-open.jpg
<[TheBug]> montjoie: e-mail FriendlyElec about it - I will say this - their sales team usually replies pretty quickly, however, I previously wasn't able to resolve the fact that my order had a wrong case included and I asked them to ship the right one and they just ignored me -- however, one would hope that if they shipped the wrong board they would correct that.
<montjoie> I stuck board in the wood (via M3 screw)
<montjoie> and board are not enough stuck to the ground
<montjoie> lego is too soft
<montjoie> for fun this is my work in progress http://kernel.montjoie.ovh/
<montjoie> yeah I distcc more and more
<montjoie> KotCzarny: thanks, but 512M is hard for a board with gentoo:)
<KotCzarny> montjoie: ^
<wens> montjoie: thanks. I don't have my board beside me atm, so I'll compare them next week
<montjoie> wens: you sked for bpim3 photo http://kernel.montjoie.ovh/P1090346.JPG
<montjoie> KotCzarny: no big chip overside http://kernel.montjoie.ovh/P1090350.JPG
<montjoie> hello my friendly elec nanopi neo plus2 only have 512M of RAM but the product is branded as 1G. What can I do ? Someone here asked for RAM photo http://kernel.montjoie.ovh/P1090352.JPG

2018-07-12

<montjoie> let's see if I could do better
<montjoie> MoeIcenowy: fun with reset for AHCI https://www.spinics.net/lists/linux-ide/msg55357.html

2018-07-11

<montjoie> wens: ok
<wens> montjoie: could you take a picture of your bpi-m3 and post it somewhere?

2018-07-08

<ElBarto> montjoie: true
<montjoie> ElBarto: buggy is less worse as untested
<ElBarto> montjoie: you could have ended the sentence at 'buggy' :)
<montjoie> so untested ...
<montjoie> ElBarto: the old bsp was buggy, doing RSA in CBC mode in a kernel which doesnt have RSA API:)
<ElBarto> montjoie: thanks, and good to know that the manual is enough to start
<montjoie> user manual is "enough"
<montjoie> ElBarto: he contact me via mail, I will answer
<ElBarto> montjoie: or did you had to read BSP code ?
<ElBarto> montjoie: someone want to do crypto for a64 on FreeBSD, is the user manual enough ?
<ElBarto> montjoie: hi

2018-07-06

<montjoie> ppfff if I have restarted my cilab, it shoud not have happen
<montjoie> I got the email some hours ago
<montjoie> jernej: patch posted on mailing list

2018-06-30

<montjoie> and the target-supply is for sending power to drives I think
<montjoie> MoeIcenowy: ahci only accept one regulator, so some code modification will be necessary
<montjoie> found them
<montjoie> MoeIcenowy: so you already have sent a patch for it ?
<Pineicenowy> montjoie: I remember someone mentioned that the misc regulator powering should be generic
<montjoie> what can i do for leeping regulators up
<montjoie> MoeIcenowy: for r40 i didnt understand what you mean

2018-06-29

<pgreco> MoeIcenowy, I think montjoie is offline
<MoeIcenowy> montjoie: I have did R40 sata, and only the regulators matter
<montjoie> disabled after sata is up, so disbling bring all down
<montjoie> any idea why https://pastebin.com/w0xcYpHU does not work ? regulator are disabled
<montjoie> is seems that R40 sata is compatible
<MoeIcenowy> montjoie: you can start ;-)

2018-06-28

<montjoie> hello does anyone started working on R40 sata ?

2018-06-27

<vagrantc> montjoie: ah, yes, i tested with an earlier 4.16.x, and it didn't have the issue. now with or without your patch i'm having that issue.
<montjoie> vagrantc: seems unrelated with my patch
<montjoie> vagrantc: could you paste them ?
<vagrantc> montjoie: i tried the sunxi-rsb patch on sun8i-a83t-cubietruck-plus against 4.16.16 ... and it so far has produce two kernel OOPS and hangs

2018-06-26

<montjoie> I think this should apply on anyone
<vagrantc> montjoie: against what tree?
<montjoie> vagrantc: from a cold start
<montjoie> vagrantc: try https://pastebin.com/yWPwiq3Q
<vagrantc> montjoie: missed some of the backscroll ... haven't had a chance to adapt and test the patch you sent yesterday
<montjoie> vagrantc: perhaps you have the same problem, and need to set dldo3 to 3.3
<montjoie> MoeIcenowy: will try with 3.3 and send a patch for it
<montjoie> MoeIcenowy: thanks
<CuteIcenowy> montjoie, wens: mikey confirmed that it's 3.3
<montjoie> MoeIcenowy: it is away i cannot see it, but last time I have the same as wens in my memory
<MoeIcenowy> montjoie: what's your board's version (on silkscreen)?
<montjoie> since my board work well, what is the best solution for me, upgrade vcc-pd to 3.3 ?
<wens> montjoie: the PHYs are dual voltage, and configured by pulling up or down on one of the pins at power on
<MoeIcenowy> montjoie: I think it's decided by board designer
<montjoie> perhaps a batch of 3.3 PHY ?
<MoeIcenowy> montjoie: RGMII use either 2.5 or 3.3
<montjoie> wens: but the fact I am not the only show a more deployed problem
<montjoie> wens: I have done it
<wens> montjoie: I suggest asking for a second board :)
<montjoie> wens: what's your PHY model on your bpim3 ? just in case
<montjoie> so I have a bad PHY which need higher voltage ?
<wens> montjoie: but on the PHY side DVDD33, pins 15 and 21 are correctly wired for 2.5V, while 37 is 3.3V, all following spec
<montjoie> all a83t
<montjoie> and cubietruck+ also
<montjoie> jonkerj: sorry no only bpim3 one
<montjoie> wens: perhaps shematics is wrong ?
<montjoie> wens: and in dt cubietruck+ is also set as 2.5 so it explains why vagrantc hit also that
<wens> montjoie: if you look at the BPI-M3 schematics, on page 3, it does say 2.5V for DLDO3
<montjoie> but I lack knowledge on why vcc-pd need 2500
<montjoie> non 0x12 -> 18 and adding 700 to it give 2500mV as requested
<wens> montjoie: hmm, maybe bad calculation in the driver?
<montjoie> so clearly this voltage drop made PHY unhappy
<montjoie> it try to set DLDO3 from 0x16 to 0x12
<montjoie> it is the only write I skip and gigabit works
<montjoie> wens: MoeIcenowy DLDO3 confirmed!
<montjoie> wens: thanks for the test
<wens> montjoie: 800 / 200 Mbps with iperf and default settings, no packet errors or drops
<MoeIcenowy> wens, montjoie: so maybe dldo3 problem?
<montjoie> letting axp write to those register breaks gigabit
<montjoie> so DLDO1 or DLDO3 voltage is in cause
<montjoie> only 6 writes
<montjoie> I have also a big sleep in atomic to debug with SMP
<montjoie> MoeIcenowy: please try:)
<montjoie> bpim3
<MoeIcenowy> montjoie: what board>
<montjoie> ah ah! disabling all RSB write to AXP made the gigabit working. Now time to find which write is in cause

2018-06-25

<montjoie> and play with ethtool -s eth0 speed 100 duplex full autoneg off/ethtool -s eth0 speed 1000 duplex full autoneg off
<vagrantc> montjoie: have a patch to try, or something else?
<montjoie> vagrantc: could you try it ?