2017-05-02

<qschulz> montjoie: where/how do you read the sensors?
<montjoie> sensors still print one temp (from axp ?)
<montjoie> qschulz: I see the thermal timeout only once
<qschulz> montjoie: and the first message you see (thermal failing to read thermal zone), do you see it multiple times after inserting the module?
<qschulz> montjoie: yes. Anyway, I'm leaning towards the thermal framework not handling the disappearing of a thermal sensor.
<montjoie> rmmod ?:)
<qschulz> montjoie: why would you do such thing?
<montjoie> mripard: will do
<mripard> montjoie: send an email
<montjoie> qschulz: https://pastebin.com/3P8i44Hs after rmmod sun4i_gpadc_iio
<montjoie> and a crash:)
<montjoie> ah finally got temperature on cubiedev

2017-05-01

<montjoie> at gigabit FD
<montjoie> vagrantc: does the link is correctly negotiated ?
<montjoie> wget/curl ?
<montjoie> could you tried iperf ? and which tool give you some kb/s ?
<montjoie> tested iperf this morning an got some 400mb/s
<montjoie> slow ?

2017-04-25

<willmore> montjoie, April first was a long time ago. :)
<montjoie> wens: patch ready:)
<wens> montjoie: yeah, i had meant to send a patch for it, hadn't gotten around to it yet
<montjoie> I will send a patch
<montjoie> mripard: I just found the CCU problem, when no CCU driver is select ccu_common is still compiled and need ccu_gate
<montjoie> mripard: any idea for my clk build error ?
<montjoie> systemd or win95 IOT? still dont know which one scare me the most
<montjoie> ah ah ah a horrible one, it install systemd which run win95 IOT
<montjoie> MoeIcenowy: ?
<MoeIcenowy> montjoie: what the hell
<mripard> montjoie: nice move :)
<montjoie> mripard: inserted backdoor in it
<mripard> montjoie: you should kill it once for good some time :)
<montjoie> I have just created two branch sun8i-emac-v5.1 (yes the old) and dwmac-sun8i-v5
<montjoie> wens: gentoo just set gcc5.4 as default this week

2017-04-24

<montjoie> I need someone to test set drive-strenght of RMII pins to 30
<montjoie> yes
<montjoie> I need a network test
<montjoie> anyone here with a pine64 without gigabit(RMII) ?

2017-04-22

<montjoie> johnemc2: ask your question
<montjoie> uint32_t will trigger checkpatch warning
<montjoie> prefer u32 to uint32_t

2017-04-21

<montjoie> and sanity of people backporting it:)
<montjoie> I could maintain it, but only for fun
<montjoie> sun8i-emac is dead for any mainline kernel
<MoeIcenowy> montjoie: you can publish the new version only on github and in linux-sunxi mailing list, but not to any kernel mainline mailing lists ;-)
<MoeIcenowy> montjoie: ^
<montjoie> I have some fix added
<montjoie> I will publish an updated version
<MoeIcenowy> montjoie: MDIO?
<montjoie> MoeIcenowy: send me the patch:)
<MoeIcenowy> montjoie: yes I fixed it by use the code logic from stmmac_sdio.c . Now I believe that sun8i-emac has many bugs ;-)
<MoeIcenowy> montjoie: I highly doubt that sun8i-emac has bugs on MDIO writing... have you finded any?
<MoeIcenowy> montjoie: a problem observed on sun8i-emac: after the regulator is enabled some delay should be made for RTL8211E to get ready
<montjoie> MoeIcenowy: I think the first step for easy backport is to use the same DT
<montjoie> since I have some idea on how to fix some
<montjoie> so If I do a sun8i-emac v5+, I make some people happy ?
<montjoie> but bugs lying around in it:(
<montjoie> how much
<montjoie> better ?
<montjoie> swiftgeek: old sun8i-emac patch
<vagrantc> montjoie: log if you care to take a look: https://www.aikidev.net/~vagrant/dwmac-sun8i/
<montjoie> at least with W=1 which produce some warnings all over the tree, I get nothing
<montjoie> perhaps some patch needed is not in 4.11-rc7
<vagrantc> montjoie: honestly not sure, just using sbuild with debian and the standard debian toolchains
<montjoie> I compile with W=1
<montjoie> vagrantc: what are your warning flag

2017-04-20

<montjoie> I integrate your yesterday comment and I will publish a branch soon
<montjoie> wens: so you confirm that your problem is just the loopback bit
<montjoie> but yeah the setlink function need to be cleaned
<montjoie> strange that the loopback is set, I clear it
<wens> montjoie: my phy gets detected, but for some reason the loopback bit in the emac is set
<wens> montjoie: i think i made too many changes to figure out why it's not working

2017-04-19

<montjoie> so for me, the syscon reg is read at reset
<montjoie> go below 150Mbit/s
<montjoie> but AC200 cannot
<montjoie> link managed at gigabit
<montjoie> the reason was that all was done with the internal PHY
<montjoie> I have removed this reset in the past on Bpim2+, and strangely I got very bad perf in gigabit
<montjoie> it is with that that I foound that the internal PHY was an AC200 and it accept gigabit
<montjoie> if i didnt reset after syscon, the phy change is ignored
<wens> montjoie: why do you need the reset just after syscon is programmed
<montjoie> wens: syscon values are generated from the default but i think i didnt understand well your question
<wens> montjoie: what values from syscon does it pick up? or needs to pick up?
<wens> montjoie: also, it's custom to have register bit field macros use the register name as its prefix
<montjoie> wens: will fix NULL checks
<montjoie> wens: strange my opipc work
<wens> montjoie: syscon value change?
<wens> montjoie: then that warning is kind of useless :p
<wens> montjoie: h3 orange pi pc
<montjoie> wens: the sun8i_dwmac_reset is needed for let MAC seee the syscon value change
<montjoie> wens: the first message is just a warning, uboot certainly leave emac enabled
<montjoie> wens: you got that on which board?
<wens> montjoie: why do you call sun8i_dwmac_reset() in sun8i_power_phy()?
<wens> montjoie: sun8i-dwmac 1c30000.ethernet: Current syscon value is not the default 148000 (expect 58000)

2017-04-17

<lvrp16> montjoie: i should make a stackable board holder to give out to developers
<montjoie> very old lego dump:) need to force them
<montjoie> but yeah I need to push/force a bit some part
<montjoie> I do not want to do anything definitive so no glue
<KotCzarny> montjoie, i'm afraid about the building blocks structure, it's not looking very stable
<montjoie> KotCzarny: pine64 is the only one with different size, but all cards have different power placement, so nothing can be "straight line"
<montjoie> for the moment the arduino control only one power (need buy more relays)
<montjoie> yeah first proto of cluster test http://sunxi.montjoie.ovh/cluster.jpg

2017-04-12

<montjoie> for my part, their RSAimplementation is fake, so their anwser was the real help
<montjoie> the fun fact will be, in 4.4 do they use our work, or disable it and use their
<montjoie> yes
<montjoie> for me no
<montjoie> tx/rx delay units for emac for example
<montjoie> they answered some questions
<montjoie> I hurry to see their new RSA code
<montjoie> mainlinwhat ?
<wens> montjoie: nope, you need to enable ohci0 and ehci0, the hosts that are tied to the otg port
<montjoie> wens: pine64 in next have at leastohci enabled, will try enable ehci
<wens> montjoie: did you enable ehci0/ohci0 in the device tree? i stumbled on that last week:)

2017-04-11

<montjoie> do you think a "kconfig" section with hints for minimal kernel configs for each subsystem could be a good idea
<montjoie> still missing something since lsusb doesnt show up the dongle plugged
<CuteIcenowy> montjoie: but I tihnk without MUSB one port will still be probed
<montjoie> tkaiser: CuteIcenowy MUSB was the missing part thanks
<CuteIcenowy> montjoie: what did you enabled?
<montjoie> nothing pluged
<montjoie> lsusb return -99
<tkaiser> montjoie: Tried both ports?
<montjoie> Does not having USB working on pine64/4.11rc5 is normal ?
<montjoie> yes
<montjoie> you must wait qschulz patchs for having AXP temp
<montjoie> KotCzarny: dont know, totally outdated stuff anyway.
<KotCzarny> montjoie: The requested URL /patchs_current/0001-mfd-axp20x-Rename-ADC-register-name-IPSOUT-to-APS.patch was not found on this server.

2017-04-09

<montjoie> set it in DT ?

2017-04-08

<montjoie> dont know
<MoeIcenowy> montjoie: how can device tree deal with "PHY-RST" pins?

2017-04-07

<montjoie> I have no DT for the +2
<montjoie> for testing DT changes asked by mripard
<montjoie> anyone with an opi+ ?
<montjoie> but works on mine, wens any output ?
<montjoie> ah
<wens> montjoie: tried your dwmac-sun8i-v3 branch on my opi-pc, just hangs after Starting Kernel...

2017-04-06

<montjoie> anybody here with some tech contact at SiS ?

2017-04-05

<tkaiser> jelle: sun8i-h3-nanopi.dtsi from montjoie's latest Ethernet branch: https://github.com/montjoie/linux/tree/dwmac-sun8i-v3
<tkaiser> montjoie: With this test (combined network/storage test) I got lower numbers with Armbian/mainline compared to Raspbian running on an RPi 3.
<tkaiser> montjoie: Yeah, that's great since it showed the benchmark is the problem and not the driver :)
<montjoie> tkaiser: lower ? perf decrease ?
<tkaiser> montjoie: Good news! With dwmac-sun8i-v3 benchmark numbers are even lower.
<montjoie> tkaiser yes dwmac-sun8i-v3 is the good one for the moment

2017-04-04

<tkaiser> montjoie: I lost track with Ethernet patches. Is this the one to try: https://github.com/montjoie/linux/tree/dwmac-sun8i-v3 ?
<montjoie> seems usefull
<montjoie> lol, testing my netdevice test on stmmac... crash
<MoeIcenowy> montjoie: ;-) thanks
<montjoie> MoeIcenowy: now it is in the documentation:)
<MoeIcenowy> montjoie: ;-) thx I will try to remember it
<montjoie> DELAY unit is in 0.1 ns
<montjoie> MoeIcenowy: mic-e certainly 200ps
<MoeIcenowy> (I remember montjoie mentioned the definition of the TX DELAY number, but I forgot it
<MoeIcenowy> montjoie: ^
<montjoie> Thedemon007: yes sun8i-emac could be compiled as module

2017-04-03

<montjoie> MoeIcenowy: why you need to backport ?:)
<montjoie> so you need to cherry pick also dt
<montjoie> MoeIcenowy: the reset/irq/etc name chanegs between stmmac and sun8i-emac
* montjoie didnt see the warning content
<MoeIcenowy> montjoie: this time I used v2
<montjoie> very few changes between v2/v3
<MoeIcenowy> montjoie: got EMAC reset timeout on opi1 with backported stmmac
<montjoie> and all disapear by using dwmac-sun8i
<montjoie> normal too many bug in sun8i-emac
<montjoie> or backport all stmmac patch
<montjoie> but yeah since I sent lots of patch on stmmac... rsync stmmac dir is the easiest way:D
<montjoie> I miss the april fool "dwmac-sun8i is just a satan project for giving pain do backporters"
<montjoie> I am happy, lots of easy review for v3
<montjoie> yes
<montjoie> MoeIcenowy: just the default "generic PHY"
<MoeIcenowy> montjoie: what's ethernet-phy-ieee802.3-c22 ?

2017-04-02

<montjoie> tkaiser: thanks:)
<tkaiser> montjoie: Congratulations. Jean-Luc did some Ethernet performance tests on NEO 2 and your driver (in pretty outdated version I believe) outperforms Allwinners: http://www.cnx-software.com/2017/04/02/nanopi-neo-2-board-benchmarks-with-ubuntu-16-04-2-using-linux-3-10-and-linux-4-10/

2017-04-01

<montjoie> dirty rsync stmmac dir:)
<MoeIcenowy> montjoie: it's difficult to backport it to a stable version kernel
<montjoie> MoeIcenowy: so what's missing in dwmac-sun8i ?
<KotCzarny> montjoie: i've head you've been bannd from working on stmmac or networking things altogether in kernel
<montjoie> suspens:)

2017-03-31

<montjoie> what's missing in dwmac-sun8i ?
<montjoie> why ?
<MoeIcenowy> montjoie: but currently sun8i-emac is better for daily use ;-)
<montjoie> MoeIcenowy: miasma dwmac-sun8i not sun8i-emac:)

2017-03-30

<montjoie> so that network regression like that will be found
<montjoie> at least, the good news is that khillman agree that I do the network test for kernelci
<montjoie> yes probably
<montjoie> and the breaker is totally silent since...
<montjoie> since he agree to the revert
<montjoie> didnt understand why dmiller do this
<montjoie> so now I need to revert two patch
<montjoie> the patch added patch the reverted patch:)
<montjoie> too late
<montjoie> oh my god, Im doomed, dmiller just add a patch that destroy my revert patch for stmmac

2017-03-27

<montjoie> I could try to contact kevin from baylibre, but baylibre never answered me also:(
<montjoie> I try to contact kernelci also, for adding network test: no answer:(
<montjoie> just for be sure that the regression is not on my fault:)
<montjoie> wens: could you test it ?
<montjoie> he said that he will do, but still try to send fix
<wens> montjoie: hmm, can't you just send a revert?
<montjoie> alsy: yes
<montjoie> the guy who break it delay too long the revert
<alsy> Montjoie: does the dwmac regression still present in linux-next on A20?
<montjoie> I still need someone to test linux-next on sunxi. on any non sun8i hardware

2017-03-26

<MoeIcenowy> montjoie is WIP on Ethernet for OPi, and I'm WIP on USB OTG

2017-03-24

<montjoie> could someone test latest linux-next and confirm me to also have huge network regression ? (if possible dwmac-sunxi not A20)
<montjoie> problem already reported
<montjoie> I wait for the revert
<montjoie> on A20 total break
<montjoie> on H3 TX is now 33Mb/s
<montjoie> upstream
<KotCzarny> montjoie: is it broken because of sunxi or upstream?
<montjoie> so my minor modification cannot be well tested
<montjoie> cannot send v3, stmmac is breaked in next
<montjoie> :(

2017-03-23

<montjoie> net-next does not like sunxi:(
<montjoie_> oups, opipc TX drop to 37Mb/s, seems hitting a regression

2017-03-21

<montjoie> stmmac bug found (2 bugs)
<montjoie> very strange bug on stmmac, TX irq never fire (without reason)
<montjoie> one rack, one board
<montjoie> KotCzarny: exactly what I tryto do with old IDE rack, "ethernet, uart and power thougth it"
<montjoie> ah ah ah my colortable will spread the world
<montjoie> I have no choice, stmmac is a moving target
<montjoie> rx_queues_to_use/tx_queues_to_use is not set, but I still loosing network
<montjoie> I think I found the issue
<montjoie> need to start it to produce result:)
<alex__> montjoie: does your bisect produce any result?
<alex__> montjoie: sun7i-dwmac is also affected in linux-next. It freezes my cubietruck after a random time.
<montjoie> seems lack of testing
<montjoie> lol, breaked also on meson
<montjoie> wens: certainly
<wens> montjoie: maybe the multi queue patches just merged?

2017-03-20

<montjoie> trying to bisect
<montjoie> break even without my patch
<willmore> montjoie, was it your code? ;) If not, then yay!
<montjoie> even for dwmac-sunxi:)
<montjoie> wow linux-next break dwmac

2017-03-19

<montjoie> MoeIcenowy: I will send v3 soon
<MoeIcenowy> montjoie: could you first send out sun8i-stmmac device tree binding (without send out driver itself now)?

2017-03-18

<montjoie> I will try to run it with a big front fan, to enhance uptime
<ElBarto> montjoie: after a bit of resolder everything was ok
<ElBarto> montjoie: fyi on my bpi-m2 (a31s) after a while my uart connection solder was weak
<montjoie> but one of my uart seems dead, coincidence ?
<montjoie> no
<montjoie> no
<montjoie> KotCzarny: nothing, suddenlly more heat than anything
<montjoie> seems alive, but no uart and lots of heat around (LOTS)
<montjoie> RIP bpim2+
<montjoie> not discounted
<montjoie> aliexpress
<montjoie> and 3 different chipset:)
<montjoie> just bought 3 UART, all with RX/TX inverted ... happy
<apritzel> argh, backporting montjoie's new driver to 4.10 is a pain ...

2017-03-17

<montjoie> wens: with which gcc/binutils do you compile your uboot R40 branch ? (got http://pastebin.com/PMyWQh9r)
<montjoie> MoeIcenowy: which uboot do you use for R40 ?
<montjoie> didnt found it:(
<MoeIcenowy> montjoie: ^
<montjoie> just received my BPI ultra
<montjoie> MoeIcenowy: do you have a github branch for your R40 work ?

2017-03-14

<oliv3r> montjoie: lol yes
<montjoie> this stressing moment before pressing enter for gitsendemail...

2017-03-13

<montjoie> Microsoft! I am availlable to code driver:)
<montjoie> lol no A64 network in windows 10

2017-03-10

<montjoie> do someones here get some "DMA-API: cpu touching an active dma mapped cacheline" randomly on H3/A64

2017-03-09

<montjoie> not sure if support both SS and CE will be good
<montjoie> now need to remove spagethi code
<montjoie> it's a good day, a83t SecuritySystem work
<montjoie> wens: mripard just asked for it, and yes it is the same
<wens> montjoie: is it the same for the older (a20) chips?
<beeble> montjoie: thanks
<montjoie> so for bpim3, its 0.7ns
<montjoie> 0.1ns
<montjoie> the last obstacle before sending v2
<montjoie> yes, got units for the delay
<montjoie> hourrah, got answer for tx/rx delay from allwinner
<agin_> montjoie: I'm using the orange pi h2 plus, looks like the DTS isnt updated
<agin_> montjoie: I used modprobe to install it and it appears under lsmod
<agin_> hey montjoie , i have compiled the emac ko

2017-03-08

<montjoie> but half of crash was when idle
<montjoie> tkaiser: like wens said, no psci so no SMP:(
<tkaiser> montjoie: How high are CPU cores clocked and is SMP already working?
<tkaiser> montjoie: I have an eaely BPi M3 which has a Micro USB jack for DC-IN. After soldering a barrel jack to connectors it became stable. Though never used with mainline kernel.
<montjoie> and I try to use my h8homlet without success
<montjoie> bpim3
<tkaiser> montjoie: Which board are you using?
<montjoie> wens: mine crash without any output every 48h
<wens> montjoie: sd card time out
<montjoie> wens: any stability issue with your a83t device ?
<montjoie> agin_: just use my repo
<agin_> montjoie: what is the best method of getting your eth driver?
<montjoie> sorry net-next
<montjoie> I work on top of linux-next
<agin_> montjoie: do you work off the sunxi-next branch?
<montjoie> stmmac needed to be cleaned before adding sunxi stuff
<agin_> montjoie: it's weird that eh, the 'good stable' one is a dead end
<montjoie> yes git cmd line tool, no graphic UI
<montjoie> sun8i-emac will work on nearby any kernel
<montjoie> yes
<montjoie> if you cherry pick
<montjoie> but you need latest kernel source
<montjoie> agin_: use the v5, but it is a dead end, the new driver is dwmac-sun8i
<montjoie> sun8i-emac is total standalone
<montjoie> the dwmac-sun8i need some latest stmmac patch
<montjoie> agin_: work for me:)