ChanServ changed the topic of #linux-sunxi to: Allwinner/sunxi development discussion - Don't ask to ask. Just ask! - See http://linux-sunxi.org | https://github.com/linux-sunxi/ | Logs at http://irclog.whitequark.org/linux-sunxi
<bsdfox> anyone using gpio with a10?
<Epsylon3> i was :p
<bsdfox> do you remember where the register access was?
<bsdfox> is it /sys/devices/platform/gpio-sunxi?
<Epsylon3> in /sys/class/gpio
<Epsylon3> with the pin names
<Epsylon3> like pi18
<bsdfox> I don't have /sys/class/gpio :\
<bsdfox> I've got gpio-sunxi compiled into the kernel
<Epsylon3> and GPIOLIB ?
<bsdfox> perhaps not
<bsdfox> that's in the kernel?
<Epsylon3> which kernel version ?
<bsdfox> 3.4
<Epsylon3> ok, more recent than me then
<Epsylon3> but should be the same, yes gpiolib is the standard linux way
<Epsylon3> like new leds
<Epsylon3> its a config flag
<bsdfox> h36sa@giw ~/Source/sunxi/sunxi-3.4 $ grep GPIOLIB .config
<bsdfox> CONFIG_ARCH_WANT_OPTIONAL_GPIOLIB=y
<bsdfox> CONFIG_GPIOLIB=y
<Epsylon3> SYSFS can help ;)
<Epsylon3> its made for that
<bsdfox> hmm I don't see GPIO_SYSFS in mine
<bsdfox> :)
<Epsylon3> its because GPIOLIB is a parent
<Epsylon3> and was not enabled
<bsdfox> it was enabled already
<Epsylon3> so all sub options are hiddent
<Epsylon3> you need make oldconfig
<Epsylon3> to populate the new ones
<bsdfox> hmm yeah I'm not seeing it even with oldconfig (didn't make any changes)
<Epsylon3> you need CONFIG_GPIOLIB=y
<Epsylon3> then make oldconfig
<bsdfox> I have it
<Epsylon3> hmm
<Epsylon3> config GPIO_SYSFS
<Epsylon3> bool "/sys/class/gpio/... (sysfs interface)"
<Epsylon3> depends on SYSFS && EXPERIMENTAL
<Epsylon3> check the config_experimental to
<bsdfox> that was it
<bsdfox> thanks
christopher has quit [Ping timeout: 272 seconds]
<bsdfox> has anyone gone through and mapped any of the available gpio on mele a1000/2000?
<Turl> bsdfox: well we know it has the uart pins
christopher has joined #linux-sunxi
<bsdfox> yeah I've got those working :)
<Turl> vga has some more I suppose
<bsdfox> there are lots of unpopulated pads on the board
<bsdfox> I was just wondering if anyone had probed them to see what's connected.
<Turl> I haven't looked much inside
<Turl> I don't think anyone did
<bsdfox> I see hno's pio util but I'm trying to figure out proper usage now
<Turl> I used fel-gpio
<bsdfox> I'll post anything I fiigure out
<Turl> but on mele I don't think you can unless you solder an otg cable
<Turl> there's an usb header inside too now that I remember
<Turl> for otg
<Turl> that might contain usable pins :)
<bsdfox> I actually have some OTG cables
<bsdfox> I see at least 3 unpopulated usb headers inside
<ganbold___> where all dts is in linux-sunxi? specially which branch should I check?
<rz2k> check other branches
<ganbold___> yeah, just checking one of wips of mripard
egbert has quit [Disconnected by services]
egbert has joined #linux-sunxi
<bsdfox> anyone used hno
<bsdfox> 's pio tool?
mdfe has quit [Ping timeout: 260 seconds]
Epsylon3 is now known as E3toSleep
<Turl> bsdfox: fel-gpio is just a wrapper for pio isn't it?
<bsdfox> Turl, I didn't get OTG connected
<bsdfox> I'm just using some python to toggle gpio[1-4]
<bsdfox> found the blue and red leds :)
<Turl> what device are you playing with?
<bsdfox> mele a2000
<Turl> where's the red led? o.O
<Turl> and can the front blue one be turned off?
<bsdfox> Turl, yes it can be turned off
<bsdfox> and that same led appears to have red too :)
<bsdfox> GPIO1_PH10 is red led (main led)
<bsdfox> GPIO2_PH20 is blue led (main led)
<bsdfox> I can't find GPIO3_PB3 or GPIO4_PH22 and I just checked all the open headers (lots of unpopulated resistor pads I didn't test yet
<bsdfox> found GPIO4_PH22
<bsdfox> wish I had an entire build on sd.. my sata stuff is in the way so I can't easily test everything
<bsdfox> three pins is better than 0 though :)
slapin_nb has quit [Ping timeout: 245 seconds]
<bsdfox> is there a good place on the website I can put this info with some pictures?
slapin_nb has joined #linux-sunxi
<Turl> bsdfox: I cannot turn PH10 nor 20 off :(
<bsdfox> Turl, did you set direction first?
<bsdfox> also your script.fex/bin needs this section
<Turl> I'm doing sudo ./pio -m "PH20<0><0><0><0>"
<bsdfox> I couldn't figure out pio
<bsdfox> [gpio_para]
<bsdfox> gpio_used = 1
<bsdfox> gpio_num = 4
<bsdfox> gpio_pin_1 = port:PH10<1><default><default><0>
<bsdfox> gpio_pin_2 = port:PH20<1><default><default><1>
<bsdfox> gpio_pin_3 = port:PB03<0><default><default><default>
<bsdfox> gpio_pin_4 = port:PH22<1><default><default><0>
<bsdfox> I'm sure there are other gpio available but those were in my stock script.bin so I figured they're safe
<Turl> nice, sudo ./pio -m "PH10*1" did the trick somehow
<Turl> and PH20
<bsdfox> Turl, try echo "out" > /sys/class/gpio/gpio2_ph20/direction && echo "0" /sys/class/gpio/gpio2_ph20/value
<bsdfox> err add a > on the second
<bsdfox> cool
<Turl> PH20<1><0><0><0> seems to be the right config
<Turl> mode 1, pull,drive and data 0
<bsdfox> 10 minutes writing python and probing with oscope
<Turl> I wonder if the red led on cb is controllable too now
<bsdfox> I don't have one but it's worth a shot
<bsdfox> cubie board has available schematics doesn't it?
<bsdfox> ooh that should be helpful
<bsdfox> I bet we can grab a bunch of other available pins with that
<bsdfox> a10 datasheet isn't available right? last time I looked it wasn't
<bsdfox> I'm suspecting that almost the entire PHnn port is available
cubiefan has joined #linux-sunxi
<cubiefan> hello
<cubiefan> I am using cubieboard
<cubiefan> and... I'm trying to set resoultion to 480x272 through cvbs pin
<cubiefan> but there's no option. what should I do?
<bsdfox> I think you might need to do that in script.bin/fex
jinzo has quit [Ping timeout: 248 seconds]
jinzo has joined #linux-sunxi
<cubiefan> hmm.. there's no option like 480x272
<cubiefan> I've tried using EDID kernel arguments but it didn't work
<bsdfox> does it seem safe to enable all the LCD0/LCD1 GPIO pins on a mele that doesn't (right?) use them? then toggle in output mode to see what I can find?
<cubiefan> bsdfox//maybe not
<bsdfox> yeah probably a bad idea
<bsdfox> I think if I go through this and find any gpio that's not used I might be ok.. http://bob.drinksbeer.org/script.fex although I see PH22 in there for MMC. I'll have to check that pin to see if it's used during boot
<bsdfox> hmm definitely not used when reading my sd card so not sure what's up with that
rz2k has quit []
ZaEarl has quit [Ping timeout: 264 seconds]
wingrime has joined #linux-sunxi
<mripard> Turl: \o/
<wingrime> mriprd: what wrong with IRQ-ctrl
<wingrime> mriprd: why irq pending flag not cleaned ?
<wingrime> mripard: why touchscreen driver should touch IRQ registers to do that
<wingrime> ?
<mripard> hmmmm, they should not.
<mripard> they should only clear their own, and never poke into the interrupt controller
<wingrime> mripard: you can see ts-drivers for sunxi thay all clean IRQ pending flag (thay defenetly shoud not do that)
<mripard> do you have some precise example in mind?
<wingrime> mripard: ft5_ts driver
<wingrime> mripard: see irq handler
<wingrime> mripard: I tryed remove that , but I hung,witch means flag not cleaned by default
<wingrime> mripard: Irq code must do that
<mripard> ah, you're right, it shouldn't do that
<mripard> but I don't think they declared an irqchip driver for the external interrupt
<mripard> so instead of having one single driver that takes care of all of that, like you suggested, and like it should be
<mripard> every driver needs to poke into the PIO IRQ registers
<wingrime> this MAJOR issue
<mripard> it seems to be a pattern at allwinner, they do this for most of the subsystems, like clocks, muxing, etc.
<mripard> I know, but the way the external interrupts work in hardware is... tricky
<mripard> I still haven't figure out exactly how to correctly handle that
<wingrime> mripard:where Irq code for aw placed?
<mripard> for linux-sunxi or mainline kernel?
<wingrime> for linux-sunxi
<mripard> as far as I know, the SoC interrupt controller is in arch/arm/mach-sun3i/core.c
<mripard> and there's no support for the PIO external interrupts, except what you saw
<mripard> it's for the soc interrupt controller
<mripard> it's not related to your problem
<mripard> not directly at least
<wingrime> mripard: me need irq-chip ?
<mripard> yes
<mripard> the PIO hardware handles muxing, gpio, and irq
<mripard> and for each IRQ it gets, it will trigger an IRQ to the soc interrupt controller
<mripard> it's a chain
<mripard> we have the code for the soc interrupt controller
<mripard> we don't for the PIO
<wingrime> look like it not implemented fully
<mripard> yes, it is for the soc interrupt controller.
<wingrime> static struct irq_chip sw_vic_chip
<wingrime> mripard: this is look like top priority to fix
<mripard> then go ahead
<wingrime> docs))
<mnemoc> btw, eint support is in the gpiolib
<mnemoc> driver
<mripard> mnemoc: oh? nice
<wingrime> mnemoc: it must be in irq.c file
<mnemoc> "must" is a relative term in our tree :p
<mripard> wingrime: no
<wingrime> core.c should be divided to irq.c timer.c
<mnemoc> as things go refactored the get into the place where they are implemented with less changes
<mripard> it's a different IP, it needs to be in a different driver
<wingrime> mripard: also why MACHINE_STATE in plat/core.c
<wingrime> *START
<wingrime> it shuld be at least it march folder
<wingrime> *in
<mnemoc> wingrime: plat-sunxi is an step group between mach-sunNi and mach-sunxi
<mnemoc> used to reduce code duplication
<mnemoc> and understand SoC diffs better
<wingrime> mnemoc: MACHINE_START must be in march I think
<wingrime> eachone for sun4i and sun5i
<mnemoc> allwinner has separated trees per so, and mach-sunNi and sunNi-foo everything, while we aim for mach-sunxi and sunxi-foo
<mnemoc> wingrime: we don't want mach-sun[4567]i. those WILL die
<wingrime> mnemoc: agreed
<wingrime> mnemoc: we should move pm to plat
<wingrime> It look same
<wingrime> mnemoc: I want singe kernel for sun5i and sun4i
<wingrime> *single
<mnemoc> mainline is multiplatform, we don't
<mnemoc> but we still have a lot of unify
<mnemoc> unified stuff can then be multiplatform-ized easily
<mnemoc> while mripard is in the reimplement path, sunxi-3.0/sunxi-3.4 are on the refactoring path
<wingrime> mnemoc ^
<mnemoc> aiming at having *tested* sunxi-foo and plat-sunxi code that can be brought into mainline
<mripard> wingrime: I agree, it should be in drivers/gpio/gpio-sunxi
<mripard> and it's already the case.
<wingrime> mriprad: what best way to fix this ?
<mnemoc> of course that doesn't belong there :p
<mripard> remove the code, unbreak the driver?
<mnemoc> move the code somewhere else
<mnemoc> reducing duplication and out of place hackery
<wingrime> mripard: without this system freezesz
<wingrime> dead hung
<mnemoc> don't remove it, move it
<mnemoc> baby steps
<mripard> mnemoc: to me, it's already somewhere else, since you have some irq_chip already for the external interrupts
<mnemoc> in which case the driver must call the external function
<wingrime> mnemoc: irq_chip must do this
<wingrime> mnemoc: see non-sunxi other drivers
<wingrime> mnemoc: it must be done in irq controller code
<mnemoc> fine, refactor it that way and send the patches
<mnemoc> (more frienly to multi-platform/future)
<mnemoc> but a driver is better than mach-/plat-
<wingrime> mnemoc: this problematic
<wingrime> mnemoc: at least it can broke some thig
<wingrime> mnemoc: it can broke some drivers etc
paulk-desktop has joined #linux-sunxi
<mnemoc> wingrime: make it first in a way that doesn't break drivers
<mnemoc> then you go cleaning
<wingrime> mripard: we need irq_eoi ?
<wingrime> mripard: than I can add to core
<wingrime> mripard: strange it have this
<wingrime> mripard: it strange that It hung without clear
christopher has quit [Remote host closed the connection]
n01 has joined #linux-sunxi
n01 has quit [Ping timeout: 248 seconds]
<techn__> Turl: http://pastebin.com/YA8HVFrq , Recovery boots correctly to GUI but OS shows black screen with this log
mdfe has joined #linux-sunxi
<techn__> mali's group/user seem to be root/root..
<oliv3r> argh, i can't get why my kernel won't boot. I know that hexdump shows a completly different header, and 'file' doesn't recognize it as a kernel either (the boot.img kernel) but that does boot, whereas mine doesn't. It can't be that it's compressed, it's double the size, and debug info shouldn't make the kernel twice as big? and make it not being able to identify as kernel :S
<mnemoc> the anger of the dark side...
rellla has joined #linux-sunxi
<oliv3r> how can I liberate my kernel if a liberated kernel refuses to boot :D
<oliv3r> i should check the guys chithub and see how he built the kernel, the .config isn't enough
<oliv3r> still, twice the kernel; and it's not a kernel file says its' data
ol1ver has joined #linux-sunxi
<mnemoc> what device? what kernel? what bootloader?
<ol1ver> inet97f-ii; boot0/1, 3.0.62+
<mnemoc> that's a13, right?
<ol1ver> a10
<mnemoc> weird
<ol1ver> i'm uploading closeup of available pins now; but i probed them with all 8 uarts on; nothing responded
<mnemoc> did you replace nanda's linux/u-boot.bin with lichee-dev's ?
<ol1ver> i did not touch u-boot.bin
<ol1ver> i did replace script.bin
<ol1ver> but i'm running a custom kernel allready!
<ol1ver> that's his repo
<mnemoc> stock linux/u-boot.bin doesn't pass proper machine id and mem info
<ol1ver> he did CM10 for many A10 tablets
<ol1ver> first thing i don't get is why his kernel is twice the size, and doesn't have the kernel 'header' its uncompressed, i can see strings
<mnemoc> why don't you invest your energy in our kernel instead of a random fork?
<ol1ver> that's what I'm trying to do
<ol1ver> i'm trying to boot our kernel
<ol1ver> but uboot appears to not load it, i'm getting stuck at the displaying the u-boot bitmap
<mnemoc> if you are booting 3.0 with "stock" u-boot.bin you need to enable the ignore atag_mem thing
<ol1ver> now your talking
<ol1ver> where do i find info on that?
<mnemoc> but replacing the broken by design stock u-boot.bin is *highly* recomended
<ol1ver> can't find atag_mem on the wiki
<mnemoc> .config
<ol1ver> well i can't brick mine anyway right; i can boot via SD so that's good
<mnemoc> CONFIG_SUNXI_IGNORE_ATAG_MEM
<ol1ver> ok so i'll start replacing u-boot then
<mnemoc> that enables allwinner boot hacks in sunxi-3.0
<mnemoc> but it's not present in 3.4
<ol1ver> well the thing is, i pulled /proc/config.gz from the running kernel
<ol1ver> and loaded that
<mnemoc> an ancient fork
<ol1ver> true
<mnemoc> including allwinner hacks by default
<ol1ver> ok, u-boot first then
<ol1ver> u-boot loads normal kernels then; correct?
<mnemoc> should
<ol1ver> which repo holds the lichee u-boot?
<mnemoc> u-boot-sunxi
<mnemoc> there are 3 important branches
<ol1ver> the wiki mentions two branches, NAND and MMC-only
<mnemoc> lichee-dev (fixed nand drop-in, not improved), sunxi (stable, mmc-only, clean and with nice features), and sunxi-current (development, cleaner, newer, still mmc-only, DTS and tftp capable)
<ol1ver> so which one should I use, if i want to boot android from nand (and regular linux from SD
<ol1ver> i'd assume lichee-dev; since the others are all mmc-only
<mnemoc> yes
<ol1ver> bah; that one doesn't even build :p
<ol1ver> i'll see later why; $work$ first
<ol1ver> oh i can just use that :p
<ol1ver> ah, board setup is quite different, i'll have to redo it
<ol1ver> do'h
<ol1ver> but $work first
<ol1ver> do we want a patch for that tree at all?
<mnemoc> we still need it. so, yes
<mnemoc> fixes only, ... but maybe uEnv.txt might be a good idea
<mnemoc> brb
wingrime has quit [Ping timeout: 276 seconds]
<techn__> whee.. a13 boots to cwm10 gui :)
<techn__> now cleanup and patch bomb
<mnemoc> techn__: are you testing new kernel right now? can you give a shot to a patch I just wrote?
<techn__> mnemoc: I'm now using allwinner-dev-teams kernel with couple cherry-picks from our tree
<mnemoc> ok :(
<techn__> but plan is to test with our kernel at some point
<techn__> I thought it was easier to boot with android "verified" kernel
<mnemoc> yes
n01 has joined #linux-sunxi
rellla has quit [Remote host closed the connection]
eebrah has quit [Ping timeout: 260 seconds]
E3toSleep is now known as Epsylon3
<Epsylon3> mnemoc : do we have infos on the CM9 rom available on cubieboard.org ?
<Epsylon3> its a 3.0.42 kernel
<mnemoc> afaik he just built Turl's
sky770 has joined #linux-sunxi
<Epsylon3> nice :)
<Epsylon3> 20121028
eebrah has joined #linux-sunxi
<n01> uhm, wemac.0: WARNING: no IRQ resource flags set.
<n01> is this normal?
<techn__> Turl: need help
<techn__> Turl: How I can have nande and nandh automaticly formated to ext4? now I need to format them via recovery after flash.. and boot to recovery needs UART with sunxi devices?
<techn__> ok.. no need for UART seems to work with adb
<techn__> but I think it should work without recovery after flash
<cubiefan> hellow is there any way to support 480x272 resolution for cubieboard?
<cubiefan> edid arguments didn't worked
<techn__> cubiefan: which interface?
n01 has quit [Quit: Reconnecting]
n01 has joined #linux-sunxi
eebrah has quit [Ping timeout: 264 seconds]
vinifm has joined #linux-sunxi
ZaEarl has joined #linux-sunxi
rellla has joined #linux-sunxi
brian_ has quit [Ping timeout: 245 seconds]
mdfe has quit [Ping timeout: 252 seconds]
<bsdfox> Epsylon3, I've got GPIO working and have a little python script toggling pins but I've only found 3 that seem to be broken out. There are lots of pads unpopulated.. any advice on figuring out what pins these might go to (need to define parameter for gpio export in the bin/fex)?
<ol1ver> bsdfox: you did look at linux-sunxi.org/A10/PIO
<bsdfox> I did not
<sky770> ..
shineworld has joined #linux-sunxi
<techn__> mnemoc: your 5/7 patch looks pretty dangerous :p
<techn__> did you test that with a13? :)
<mnemoc> tests would be appreciated. I can give you the patchset for 3.4 if you want to test it there
<mnemoc> I have everything packaged (I'm moving to another apartment), so I can't really test it :<
<mnemoc> what's the attribute used to mark functions that always return the same?
sky770 has quit []
<ol1ver> there, that's the best i can do with regards to foto's for my tablet
shineworld has quit [Remote host closed the connection]
<mnemoc> http://linux-sunxi.org/index.php?title=Cubieboard/Cubieboard_Linux-3.9&rcid=4213
ganbold_ has joined #linux-sunxi
<ol1ver> oh god it still fails to build. i'll just use your u-boot mnemoc; lichee-dev u-boot is way to broken
<mnemoc> :(
<mnemoc> ol1ver: techn just submited a patchset for lichee-dev/sun5i
<Turl> it might be pure, it might be constant mnemoc
<mnemoc> I actually have that book. but no idea on which box :p
<Turl> yeah, the book is awesome :)
ganbold___ has quit [Remote host closed the connection]
<mnemoc> so my evil 5/7 needs __attribute__ ((const)) to let sunxi_is_foo() checks efficient
<mnemoc> not that anyone would notice the diff anyway
<Turl> linux has __attribute_const__ you can use
<mnemoc> wtf static struct gpio_eint_data gpio_eint_list[] is in the .h?!
<mnemoc> (w)hy
<Turl> mnemoc: what about detecting the soc version just once early on boot, and then making the macros simple (soc_version == SUNXI_VER_...)
<Turl> ?
<mnemoc> exporting the version symbol?
<mnemoc> isn't that illegal these days?
<mnemoc> anyhow, I still need someone to test this things work before thinking too much in how to make it more efficient :p
<mnemoc> wip/sunxi-3.[04]/soc-detect on my github if anyone has the chance to test on A10S or A13 hw
cubiefan has quit [Ping timeout: 245 seconds]
rellla has quit [Remote host closed the connection]
<techn__> Turl: how you would like patches to allwinner-common?
<Turl> techn__: pull request is okay, otherwise email me (turl at linux sunxi) and tell me what you need added on the repo, preferably with github links to commits
<techn__> ok..I'll send patches for sun5i support
<techn__> and couple renamings to sunxi :p
ganbold_ has quit [Remote host closed the connection]
ganbold_ has joined #linux-sunxi
wingrime has joined #linux-sunxi
ganbold__ has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
<bsdfox> my sata drive is clicking much more often after using PH17 :D
<wingrime> bsdfox: You can set spinup / spindown
<wingrime> bsdfox: or it bug ?
<wingrime> bsdfox: than show dmesg
<Dreadlish> or your power source is not much efficient ;)
<bsdfox> wingrime, no I think PH17 is SMC_VCCEN
torqu3e has quit [Quit: torqu3e]
<bsdfox> the drive spins down as soon as I toggle PH17
<bsdfox> reboot fixed whatever the drive was doing on its own afterward
<wingrime> bsdfox; simply set in to "1" and forgot
<bsdfox> wingrime, I'm probing my mele looking for usable GPIO
<bsdfox> so far I've only found four- PH10, PH20, PH22, PH23
<wingrime> bsdfox: simply find PCB schme
<bsdfox> four is enough to do SPI so I might stop
<bsdfox> wingrime, is it available somewhere? I've looked
mdfe has joined #linux-sunxi
<Turl> bsdfox: you cannot mux random ports to spi though
<bsdfox> I can bit bang it if necessary
<Turl> there's spi pins on PA, PB, PC and PI from what I can see on the docs
<bsdfox> Turl, if I could locate an unused hardware SPI bus that'd be ideal but I haven't yet
<bsdfox> and I don't want to have contention from sd card or whatever else is SPI in the mele
<wingrime> bsdfox: you can use SPI with many slaves
<wingrime> bsdfox: you must use "cs\"
<bsdfox> wingrime, yeah I know but I don't want to get involved with the chip selects for the builtin functions
<bsdfox> it looks like a couple i2c busses are available so maybe I'll just use those
<wingrime> i2c as I remeber can be used without cs , addresses only
<bsdfox> yup
<wingrime> bsdfox: a13 have 3 i2c and 3 spi
<wingrime> bsdfox: you must check docs for wires
<wingrime> bsdfox: a13 have MAJOR advantage it TQFP package
<wingrime> I want that tool for firmare hacking for zet6221 (in datascheet)
<bsdfox> I didn't know it was TQFP
<bsdfox> that's cool. I like sata though
<wingrime> bsdfox: unfortunly lack
<bsdfox> who has access to modify http://rhombus-tech.net/allwinner_a10/hacking_the_mele_a1000/ ? I can add some info
<wingrime> bsdfox: but I think you can use usb bridge
<wingrime> bsdfox: use our wiki
<wingrime> bsdfox: add to it
<wingrime> mele a1000 only 512 Mb ram (((
<wingrime> bsdfox: I can try add IR to my tablet
<bsdfox> wingrime, http://linux-sunxi.org/Mele_A1000 the wiki just links to that other page :P
<Dreadlish> wingrime: you can try to resolder ram chips ;)
<wingrime> DreadLish: no
<wingrime> DreadLish: ddr3 is hard thing as it works at 400 Mhz
<wingrime> DreadLish: any solder error are fatal
<wingrime> DreadLish: and it BGA
<Dreadlish> bga is simple
<Dreadlish> when you have hot air ;)
<Dreadlish> cause there isn't so much balls
<Dreadlish> some people do it in common oven ;D
<wingrime> DreadLish: no need borad healter and and IR solder station
<wingrime> Air works badly
<Dreadlish> not really
<Dreadlish> i have soldered some ddr2 chips
<Dreadlish> with hot air.
<Dreadlish> and they are working
<Dreadlish> ram chips aren't so temperature sensitive, so they don't really need being soldered with ir
<wingrime> Dreadish: don't want remeber when I try solder last time with air
<Dreadlish> my last was hmm...
<Dreadlish> yesterday?
<wingrime> any way ddr have speed that hard with impedance
<wingrime> any flux can be fatal
<Dreadlish> ech
<Dreadlish> it can be done without flux.
<wingrime> I mean It need hi accuracy
<wingrime> and more, it can lacks some address lines for 1G ram
<Dreadlish> it's just matter of experience and good magnifier ;)
torqu3e has joined #linux-sunxi
<Epsylon3> bsdfox: yes its in fex, and some are used for other functions
<Epsylon3> +[gpio_para]
<Epsylon3> +gpio_used = 1
<Epsylon3> +gpio_num = 4
<Epsylon3> +gpio_pin_1 = port:PI13<1><default><default><default>
<Epsylon3> +gpio_pin_2 = port:PI12<1><default><default><default>
<Epsylon3> +gpio_pin_3 = port:PI11<1><default><default><default>
<Epsylon3> +gpio_pin_4 = port:PI10<1><default><default><default>
<Epsylon3> those ones are near the jack
<Epsylon3> the 4 last ones (square)
<ol1ver> is there _any_ way i can see what u-boot booted my tablet, after the kernel is running?
<Epsylon3> tested on leds
<Turl> ol1ver: if you connect uart and check the logs after kernel has booted.. ;)
<ol1ver> if i had uart, i wouldn't be askin' that
<ol1ver> silly turl
<ol1ver> i did order a sd -> uart adapter to go with my mele a1000g; but that order somehow fell through; so posponing it for my A20 based device
<Turl> ol1ver: if you have control over multiple uboots you could add a dummy cmdline parameter to them or something to pass on the kernel
<Turl> and check that
<wingrime> hramrach: are you here ?
<ol1ver> oh
<ol1ver> Turl: well, before i used the stock u-boot; now i'm using mnemoc's one
<ol1ver> so u-boot param's should be different
<ol1ver> unless mnemoc used the exact same u-boot arguments
<ol1ver> wdconsole=ttyS0,115200 rw init=/init loglevel=8
<ol1ver> nothing much to see here :S
<Turl> if they're both nand uboots you could try booting 3.4
<bsdfox> Epsylon3, which jack? my script.fex stock had PH10, PH20, PB3, PH22 in that order
<Turl> 3.4 doesn't boot with AW's stock uboot
<ol1ver> that's the whole problem, i can't boot a custom kernel yet :(
<ol1ver> i can't even get 3.0 to boot
<Turl> get uart then :p
<ol1ver> *sigh*
<ol1ver> i know
<ol1ver> well for starters
<ol1ver> think here with me for a min
<ol1ver> if i do 'file kernel image' before using mkbootimg
<ol1ver> on our 3.0 i get a proper 'kernel' id returned; which makes sense, it's a kernel, duh
<ol1ver> but if i extract the kernel from boot.img; i get 'data', i can grep for strings
<ol1ver> so i know it's not compressed
<ol1ver> the thing is, its twice as big
<ol1ver> yet doesn't have a header at the start
<Turl> they might be using a bImage
<Turl> try file arch/arm/boot/bImage and see if it says data too
<ol1ver> need a few
<wingrime> mnemoc: are you here ?
<Turl> mnemoc: he was moving I think
<Turl> err
<wingrime> Turl: O want merege PM
<Turl> wingrime: ^
<wingrime> Turl: I want merge PM to one singe
<Turl> to what?
<wingrime> plat-sunxi
<Turl> wingrime: have you sent patches to ML?
<wingrime> not yet
<mnemoc> Turl: still packing :)
<Turl> mnemoc: did you get internet for the new place yet?
<wingrime> mnemoc: now I join pm to single one
<wingrime> mnemoc: I want you acept it without check-patch script
<mnemoc> Turl: yes, but no electricity yet :|
<Turl> mnemoc: o.O
<techn__> It would be cool to have automatic build for our stuff.. it takes forever to build and test android on every device :p
<Turl> mnemoc: 0km apartment? :p
<techn__> with automatic build I mean some build farm
<mnemoc> Turl: yes. they couldn't sell them, so they are now renting new flats cheap :p
<mnemoc> wingrime: where is it?
<techn__> Turl: how I can build only bootimg?
<Turl> techn__: make bootimg
<techn__> thanks
<wingrime> mnemoc: Not now
<Turl> wingrime: when unifying stuff we won't complain if you don't use checkpatch
<Turl> in fact it's better if you don't
<Turl> so it's just unification and not mixed with cleanup
<wingrime> mnemoc: I must check it for me
<mnemoc> minimizing the diff is more important than style
<wingrime> mnemoc: and ask some people to test it on a10
<techn__> Turl: can android build system manage build so that you can create common rootimg and device specific bootimg/recoveryimg
<Turl> techn__: hm?
<techn__> I mean that you'll have sunxi-rootimg and sun4i-bootimg/recovery and sun5i-bootimg/recovery
<wingrime> Who can check pm on a10/a13
<wingrime> I will send patch soon
<wingrime> I wan't someone to check this NOW
<wingrime> and I will make DMA patch
* mnemoc has no device available
<Turl> techn__: I'm not sure it would be feasible
<ol1ver> wingrime: can't check, can't boot custom kernels yet :(
<Turl> techn__: system/ has device specific stuff
<Turl> like radio (wifi, bt) configuration
<wingrime> Than who can ?
<Turl> different sensor libs
<techn__> I think kernel modules should handle situation when they are modprobe to wrong device
<Turl> different camera config
<Turl> techn__: yeah but it's not that
<Turl> it's the android<->kernel sensor hal for example
<techn__> camera config could be read from .fex
<Turl> all that on fex, and read from userspace? :p
<wingrime> funny 400 kb patch
<wingrime> mnemoc: pm are same 100%
<ol1ver> that's huge for a patch :)
<wingrime> mnemoc: it must be workable
<mnemoc> wingrime: please review the -dev branches
<mnemoc> or sun5i is *very* early
<mnemoc> our*
<techn__> anyone tried linaro nano images?
<wingrime> mnemoc: diffs http://paste.debian.net/244052/
<wingrime> I fixed pathes only
<wingrime> "s"
<wingrime> mnemoc: you add fresh to -dev ?
<Turl> wingrime: A31s is for phones
<wingrime> Turl: you can see patch
<wingrime> Turl: it simple
<ol1ver> techn__: i tried linaro alip rootfs image, wouldn't display anything
<ol1ver> wingrime: taht's probably why there's that nasty ar100 in the A31*
<ol1ver> wingrime: i think you argue'd a phone really needs it
<wingrime> ollver: ar100 now used only for suspend
<wingrime> ollver: and some not interesting stuff
<wingrime> ollver: but In fututer it will handle (I think) baseband
<ol1ver> i could agree
<wingrime> ollver: gsm baseband some realtime os
<ol1ver> how's your zet6221 coming along?
<ol1ver> did you receive reference driver i emaile you?
<wingrime> yep
<wingrime> but by link on irc
<ol1ver> that was the datasheet?
<wingrime> yes
<ol1ver> i sent you a zip via gmail
<ol1ver> they've sent us a reference driver
<wingrime> whitch theme ?
<wingrime> on email
<ol1ver> er
<ol1ver> don't recall
<ol1ver> from: oliver@linux-sunxi.org
<wingrime> try to find
<wingrime> find
<wingrime> thanks
<wingrime> ollver: it have full sunxi driver
<ol1ver> yeh; 'reference driver'
<ol1ver> it even included a seperated 'firmware' in the .h file
<ol1ver> and GPLed source
<ol1ver> i don't know what this firmware is
<ol1ver> 'init code'
<ol1ver> it init the chip
<ol1ver> or actual firmware
<ol1ver> (which isn't needed, only if oyu have to reflash)
<wingrime> ollver: I wan't more one thing
<wingrime> ollver: Thay have tool for firmware hacking
<ol1ver> lol doubt we can get that
<ol1ver> but what do you mean with 'firmware tool' the source?
<wingrime> try to ask)
<ol1ver> well we have assembly i guess :p
<ol1ver> i don't htink you need 'flash tools'
<ol1ver> as those should be included int he reference driver
<wingrime> you not right
<ol1ver> what's in the .h ?
<wingrime> that tool for fimware config
<ol1ver> d
<wingrime> it can change sensitity
<wingrime> and some interesting parameters
<ol1ver> hmm
<wingrime> see data scheet
<ol1ver> i thought they'd did it via i2c
<ol1ver> i know the ssd253 does it via commands over i2c
<ol1ver> so ideally, you'd need a firmware, which would take those commands from i2c :)
<ol1ver> well once we have an 'awesome' driver, we can show it to them and ask fro it?
<wingrime> ollver: zet6221 have not fully use i2c
<wingrime> ollver: my driver works good but lacks some features now
<wingrime> ollver: 1) flashing
<wingrime> ollver: 2) multitouch (add soon)
<wingrime> ollver: 3) charger mode
<ol1ver> what is charger mode?
<ol1ver> and why does it need flashing?
<wingrime> when you connect charger driver send command to ts
<wingrime> some chargers for tablets give noise to power lines
<bsdfox> does i2cdetect work for anyone? I just get incomplete xfer (0x20) error on all three busses
<wingrime> this commands move ts to charger mode (lowing sensitivy)
<wingrime> bsdfox: with firmware are you using
<bsdfox> Linux linaro-ubuntu-desktop 3.4.29+ #6 PREEMPT Fri Mar 22 09:18:19 PDT 2013 armv7l armv7l armv7l GNU/Linux
<wingrime> bsdfox: offcose you will get xfer error as i2c not connected to anyone
<bsdfox> wingrime, I soldered on an eeprom
<bsdfox> plus the axp209 should be on i2c-0
<bsdfox> I put the pullup resistors in too
<ol1ver> wingrime: does the reference driver do charger mdoe?
<wingrime> yes
<wingrime> it simple
<wingrime> I can do it
<wingrime> but it need axp209 driver patch
<wingrime> driver must know when we in charge mode
<ol1ver> kernel knows
<ol1ver> can't the driver ask the kernel?
<wingrime> ollver: see datascheet for hacking tool
<ol1ver> i'll ask next week
<wingrime> ollver: we must add "export" to some kernel functions to jnow
<wingrime> ollver: how you reciver answer
<wingrime> ollver: I have tryed but no success
<ol1ver> i'll ask for firmware source ;)
<wingrime> ollver: thay ts chip is simple microcontroller seems to be MCU51 based
<wingrime> 8051
<ol1ver> do you think, with source, you could make it more i2c capable?
<wingrime> ollver: nop, but I can ajust
<ol1ver> nope? not enough skills :p jk
<wingrime> ollver: not skills, this simply not required at all
<wingrime> I can use that tool to try add "pressure"
<ol1ver> but would it allow for a more simple and clean driver?
<wingrime> ollver: that all simple now at all))
<ol1ver> well with source, would you be able to write your own 'tool'? or is the tool required to modify the running firmware of the mcu?
<ol1ver> my concern here is, if you need the 'hacking tool' to tune the mcu; it'll be hard to distribute etc
<ol1ver> it will be closed src etc etc
<wingrime> with firmware source I can try add commands to ajust it online
<ol1ver> i'll askt hem next week, but this time, i'll have not very high hopes
<wingrime> than question: how you get this?
<ol1ver> i'll e-mail them :p
<wingrime> at least firmware must be simple disassembler capable
<wingrime> and possible rewrite it
<ol1ver> did you look at the .h from the zip?
<wingrime> yep
<ol1ver> what is that? the mcu firmwarE?
<wingrime> I can cut fw for my own tablet form driver
<wingrime> YES
<ol1ver> or init sequence
<wingrime> firmware
<ol1ver> cool
<ol1ver> should match yours :p
<wingrime> not shure
<wingrime> I can explan you how capicative ts works
<wingrime> if have crossed lines
<wingrime> not connected
<wingrime> 15 for example
<wingrime> as thay close thay have capicacy
<wingrime> between
<wingrime> your finger also conductive
<wingrime> when you move you finger closely it modufy capicacy
<ol1ver> aye
<wingrime> for measure capicaly it sends signal to some lines and recives anver on other
<wingrime> becose capicator can't conduct direct current, thay use ac with some freq
<wingrime> MCU have singe (think) ADC that measue this
<wingrime> signal level are low, witch makes it sensitive to noise
<wingrime> expecly from power
<wingrime> as chargers are impulse down converters thay have dirty spectrum (good chargers have filters,cheap not)
<ol1ver> assume cheap for most devices
<wingrime> china engeners wonder me, how thay did it with so small amount of bipolar transistors
<ol1ver> i heard that apple chargers are really good
<wingrime> I saw good charges baset on TinySWITCH IC
<ol1ver> holy shit that's a huge ripple :D
<ol1ver> Turl: it looks like it may be a bImage indeed; i just compiled my own 3.0 kernel, it's 8megs now and 'data' by 'file'
<ol1ver> hexdump shows a quite similiar file
<techn__> bsdfox: you could try i2cdump with latest stage.. there should be i2c-3 device :p
<techn__> and it should work correctly with i2cdump
<bsdfox> techn__, it's definitely addressing properly
<bsdfox> I can watch the sda/scl lines toggle on a scope
<techn__> i2c-3 device is for hdmi
<bsdfox> I'm actually able to read the eeprom at 0x50 but can't seem to write (no error).. WP pin is low so it shouldn't be write protected
<techn__> bsdfox: yeah.. I didn't implement write support :)
<bsdfox> oh :P
<bsdfox> axp209 doesn't need it?
<techn__> no documentation for hdmi.. so I just moved current logic to its own device
<techn__> oh.. you meant i2c-0.. dunno about that
<bsdfox> i2c-1
<bsdfox> I'm using i2c-1
<techn__> i2c-3 has 0x50.. there is EDID info
<bsdfox> ahh. yeah 0x50 just happens to be the address of 24lc32a w/ A[0:2] low
<bsdfox> 01010000
<bsdfox> so there should be write support for i2c-1?
<techn__> dunno
<techn__> sunxi-i2c device seems to have bug.. https://github.com/linux-sunxi/linux-sunxi/blob/sunxi-3.0/drivers/i2c/busses/i2c-sunxi.c#L723 should return transferred msgs count
<bsdfox> looks like mripard is working on i2c
<techn__> nevermind.. return logic is in different function
<bsdfox> hmm yeah looks like write support is in there
<wingrime> sunxi-i2c have more one bug
<wingrime> it have no timeout in some cases
<wingrime> and hung ((
mdfe has quit [Remote host closed the connection]
<techn__> mnemoc: started to test soc-detect-branch
<techn__> any other I have to check than clocks?
<techn__> btw. clocks seem to be wrong @ begin.. "Mali: mali clock set completed, clock is 180000000 Mhz" with div 2 :)
torqu3e has quit [Quit: torqu3e]
torqu3e has joined #linux-sunxi
<techn__> mnemoc: remotes/amery/wip/sunxi-3.0/soc-detect wont compile.. error: invalid storage class for function ‘sunxi_is_a10s’
<mnemoc> o_O
<techn__> and others
<mnemoc> full paste?
<mnemoc> compiles fine here...
wingrime has quit [Ping timeout: 245 seconds]
<techn__> wot
<techn__> wrong paste
<mnemoc> techn__: did I cause those errors compiling the clocks too???
<techn__> mnemoc: clocks has not been working atleast since "Linux version 3.0.57+"
<mnemoc> what?
<Turl> scan-build: Using 'clang' from path: /usr/bin/clang
<Turl> techn__: ^
<techn__> that compile problem is in soc-detect branch
<Turl> wtf is scan build? and can you build linux with clang?
<techn__> Turl: sparse uses clang
<Turl> since when?
<Turl> I never installed clang yet I've used sparse
<techn__> Turl: it's disabled with C=0
* mnemoc building a fresh a13_olinux hwpack on the vanilla sunxi-bsp
<mnemoc> techn__: everything went well in sunxi-3.0. trying stage/sunxi-3.0 now
<Turl> techn__: I have sparse on this pc but no clang
<Turl> and make C=0 or 2 print nothing about scan-build or clang
* Turl is confused
<Turl> sparse is just a wrapper for gcc isn't it?
<techn__> Turl: I'm using ubuntu 12.10
<Turl> I'm using arch, but I've used debian too and it didn't have clang either
<techn__> mnemoc: that compile problem is in amery/wip/sunxi-3.0/soc-detect
<mnemoc> techn__: you just said the clock error happens since .57
<techn__> mnemoc: that clock problem
<techn__> sorry if I was misleading :/
<ol1ver> feck :S
<ol1ver> while the bImage looks what I need; it still won't boot :(
<ol1ver> mnemoc: not even with your pre-build u-boot
<ol1ver> is it boot1 that displays those ugly bmps?
<Turl> I think so
<ol1ver> ah turl!
<techn__> oliv3r: I think it's boot.axf :/
<ol1ver> your hunch was right, it very much looks like it's bImage
<ol1ver> boot.axf part of boot0/1?
<mnemoc> boot1 calls the .axf, which reads the .ini and then u-boot.bin
<ol1ver> i s hould move that to our wiki
<ol1ver> i did some boot describing
<mnemoc> ssh: connect to host github.com port 22: Connection timed out .... lovely
<ol1ver> mnemoc: so your u-boot should be able to boot any 3.0 kernel?
<mnemoc> I used to use it to run 3.4 from nand
<mnemoc> so yes, 3.0 should work too
<ol1ver> it refuses to boot :(
<ol1ver> so i'll try the nand kernel hack
<ol1ver> what was it again amem something
<techn__> IGNORE_ATAG..
<mnemoc> techn__: http://sprunge.us/IhWg ... github is not letting me commit (timeout!)
<ol1ver> so when using ATAG; use the bImage or the regular uImage
<mnemoc> that's unrelated
<mnemoc> with IGNORE_ATAG thing core.c will read the memory size from the DRAMC register instead of trusting the bootloader
<ol1ver> ok so i can still try both kernels (uImage or bImage
<mnemoc> afaik bImage is for boot.img, and uImage when loaded directly
<Turl> as far as I recall bImage is compatible with zImage
<Turl> uImage is an uboot wrap thing
<Turl> for android boot.imgs you need zImage or bImage if you want it to be slower to boot :p
<Turl> (actually if uboot loads the full partition, it might be faster to use a bImage hm)
<ol1ver> well i haven't figured out what bImage is, but uImage is just the kernel with a u-boot header. the stupid thing is, it's half the size of bImage :)
<ol1ver> atm i dont' care how fast its boot, as long as it bots :)
<Turl> what were you using to make bootimgs?
<ol1ver> mkbootimg from android
<ol1ver> well i hacked it around a bit
<ol1ver> didnt' wanna pull and build the entire android repo yet so
<Turl> yeah but what are you passing to it as kernel? :p
<Turl> zImage, bImage, uImage, Image or what? :)
<ol1ver> i downloaded sha.c sha.h to build libmincrypt.a
<ol1ver> then downloaded mkbootimg.c and mkbootimg.h and cooked all that together to get mkbootimg :)
<Turl> do you want a mkbootimg binary?
<ol1ver> should be legit
<ol1ver> i found a perl script, that splits a bootimg
<ol1ver> i took the boot.img from my tablet, split it using the perl script, and got the bImage and ramdisk, the ramdisk works, the kernel looks allright, so looks like splitimg works
<ol1ver> recombinding them also yields me a bootable boot.img so mkbootimg should be fully functional
<ol1ver> Turl: initially i did all my experimenting with uImage
<ol1ver> which now, i know is stupid :)
<ol1ver> i need bImage
<Turl> :P
<Turl> try with a zImage too
<Turl> should be the size of a uImage but compatible :)
<ol1ver> i have replaced u-boot.bin on nanda/linux/u-boot.bin with a 3.4 compatible one from mnemoc's repo
<ol1ver> i can string the extracted bootimg, and the sizes are nearly the same
n01 has quit [Ping timeout: 248 seconds]
<ol1ver> so that seems to be okay
<ol1ver> i'll now use the atag version
<ol1ver> btw, bImage isn't from arch/arm; but is in the sun4i_crane_defconfig-linux/bImage
<ol1ver> i'll ommit 'pagesize' i have added it cause the original did have it
<ol1ver> but 2048 sounds pretty standard
vinifm has quit [Quit: Ex-Chat]
<Turl> ${OBJCOPY} -R .note.gnu.build-id -S -O binary vmlinux output/bImage
<Turl> that's how it makes the bImage
<techn__> mnemoc: "sunxi: BROM Ver: 1100 1100 1625" and "cf8563 1-0051: retrieved date/time is not valid." are only differences with soc-detect and stage branchs
<techn__> cf8563 print could be normal :/
<ol1ver> so it's nearly the same as vmlinux, not supprising as it matches in size
<Turl> ol1ver: but I'm 99% confident you can use zImage too and it will work
<ol1ver> first get something booting :)
<ol1ver> then expermient
<ol1ver> btw, i can boot 3.0 and 3.4 using the fedora rootfs from sd fine
<ol1ver> still nothing, stuck at the bmp screen :S
<ol1ver> i guess i have to wait for my sd -> uart breakout board
<Turl> ol1ver: did you make disp, lcd and hdmi modules built in? (=y)
<ol1ver> would that help? the kernel doesn't even boot
<ol1ver> it gets stuck at the bmp screen
<Turl> how do you know it doesn't even boot?
<Turl> maybe it booted just fine and it's failing to load the disp modules, so you won't see any change on your screen :)
<ol1ver> you sure? i thought the moment the kernel loads, the bmp goes away
<ol1ver> it 'clears' the screen
<Turl> I think that happens when disp kicks in
<ol1ver> nah
<ol1ver> while i was playing with it
<ol1ver> i kept moving the modules dir back and forth (modules.old etc)
<ol1ver> to keep the original modules (with disp) and make room for the new ones
<ol1ver> i forget doing that a few times
<ol1ver> so i had a booting kernel, which cleared the screen but failed to display the CM10 bootanimation, due to missing screen
<Turl> in any case, you don't lose anything by trying :P
<ol1ver> yeah i did try adb a few times, but nothing :S
<ol1ver> (that's how i knew i forgot to change the modules back :)
<techn__> oliv3r: adb could work when screen is black
<ol1ver> yep it does
<ol1ver> but i'll give it one more shot, with built in dsp stuff
<techn__> ok.. now I understand.. we have API's changed
<techn__> but I'm not sure if they are fatal changes :/
<ol1ver> techn__: ?
<ol1ver> how's the disp called? CONFIG_?
<techn__> oliv3r: different mali than half year ago
<ol1ver> for android?
<ol1ver> regular linux boots and works fine
<techn__> kernel space.. so you'll need same version malilibs to user space
<ol1ver> yeah but i'm not even getting that far
<techn__> I think bootanim needs mali also
<Turl> yeah it does
<ol1ver> what about the bit before that?
<ol1ver> isn't bootanimation an android thing?
<Turl> yeah it is
<ol1ver> well i know from the working kernel, it takes a good 5-10 seconds before the bootanimation comes
<ol1ver> so the kernel clears stuff :)
<Turl> also make sure the nand driver is enabled
<ol1ver> but let me try it with built in disp
<ol1ver> what's the driver?
<Turl> I think it changed names
<ol1ver> cause 'disp' is a little generic :)
<Turl> press / and search for sunxi_nand
<ol1ver> nand is builtin allready
<Turl> FB_SUNXI, FB_SUNXI_LCD/HDMI
<ol1ver> ty
<ol1ver> all built in allready :(
<Turl> try with zImage if you haven't then
<Turl> one last shot :)
<ol1ver> :)
<ol1ver> what frustrates me is
<ol1ver> the kernel IS custom
paulk-desktop has quit [Quit: Ex-Chat]