IRC log for #arm-netbook on 20100319

04:02.42*** join/#arm-netbook ibot (ibot@rikers.org)
04:02.43*** topic/#arm-netbook is ARM netbook GNU/Linux hacking - logs: http://ibot.rikers.org/%23arm-netbook
04:45.49*** join/#arm-netbook TimRiker (~timr@bzflag/projectlead/TimRiker)
05:03.56*** part/#arm-netbook TimRiker (~timr@bzflag/projectlead/TimRiker)
11:22.17*** join/#arm-netbook mnemoc (~amery@shell.opensde.net)
11:26.25*** join/#arm-netbook lkcl (~lkcl@nat66.mia.three.co.uk)
12:12.20lkclhey when do we get to play with our shiny new home on alioth?  can you add me fjp (a/c lkcl) ?
12:32.26lkclfjp: afio and gnu cpio are different.
12:32.43lkclalso take a look at http://mid-linux.org/sdk there may be a different version of cpio there
12:32.53lkclalso fil says that pax does cpio
13:23.34mnemocgnu cpio supports several formats http://www.gnu.org/software/cpio/manual/html_node/Options.html#Options
14:02.51SQlvpapirfjp, the ben nanonote is roughly the size of a gameboy advance SP http://en.wikipedia.org/wiki/Game_Boy_Advance_SP
14:07.09SQlvpapirmy first thought was and 'what the..' and then 'why' until I realized the form factor. choosing qwerty will only trow off a few people. I admit that the 'alignedness' tradeoff for bigger keys is something I would prefer was done different
14:09.21SQlvpapirwas very close to ordering a Dingoo A-320 and the ben nanonote but both was out of stock (luckily). and insurance for my next car has first priority now so it will have to wait
15:15.27*** join/#arm-netbook SQlvpapir (~teis@188.177.95.62)
16:58.21lkclfjp: http://piratepad.net/vnBigRgsug is the poweroff function.
16:59.34mnemoc404
17:03.02fjplkcl: Suggest you use http://paste.debian.net/
17:03.28fjppiratepad does not seem to work; also not in that mail you sent
17:03.48lkclok.
17:04.01lkclRG
17:04.11lkclhttp://piratepad.net/vnBigRGsug
17:05.07lkclpaste.debian.net is editable real-time, is it?
17:05.20fjpI get Loading................................................................................................................................................................
17:05.45fjpAnd no content.
17:07.04fjpNo, but you can use an existing paste as basis for a new one.
17:07.24lkclok... well, poweroff is this:
17:07.29fjpPiratepad still not working.
17:07.31lkclW1_delay_for(1000)
17:07.47lkcls3c_gpio_pullup(S3c_FPK12, S3C_GPK12_INP)
17:08.04lkcls3c_gpio_cfgpin(S3C_GPK12, S3C_GPK12_OUTP)
17:08.05mnemoci see code now in piratepad
17:08.16lkcls3c_gpio_setpin(S3C_GPK12, 0)
17:08.38lkclfjp: try a webkit-based browser or FF 3
17:09.20lkclfjp: but... no worries, those are the four functions: it's basically all S3C_GPK12 that's involved in poweroff
17:09.24fjplkcl: That's in place of 'machine_restart(NULL);' I assume?
17:10.07lkclit's a function ineptly-named poweroff
17:10.39fjpstatic void s3c6410_power_off(void)
17:10.39fjp{
17:10.39fjp<PROTECTED>
17:10.40fjp<PROTECTED>
17:10.41fjp<PROTECTED>
17:10.41fjp<PROTECTED>
17:10.44fjp}
17:10.45fjpCorrect?
17:11.23lkclyep
17:11.26lkclthat's it
17:11.29lkclcapital c there
17:11.29fjpOr should 'machine_restart(NULL);' still be in there?
17:11.35lkcli don't honestly know.
17:12.01mnemocfjp: better ask in #s3cxxxx
17:12.02lkclthe code is _actually_ called from w1_run() which has those "secret" function in it
17:12.21lkclthis code is a complete dog's dinner
17:14.09fjpCommitted. I'll try to work on alioth tomorrow.
17:14.39lkclack.
17:14.47lkclcan you add me to the project?
17:15.39fjpWill do then.
17:15.43lkclok
17:15.47fjpBusy now. Sorry.
17:20.49mnemocwonders if he can (ab)use arm-netbook to make the kernel .deb for the gdium
17:24.46lkclfjp: no prob
17:27.22lkclmnemoc: beh? :)
17:27.39lkclohh, ha ha you want to build .debs?  suuure
17:28.04lkclgcc etc. installed perfectly well.
17:28.19lkclthe chitech kernel actually has NFS rootfs
17:28.22lkclbuilt-in
17:28.43lkclso what i did was used the nfs kernel server DO NOT use the nfs userspace server
17:28.50lkclmounted a random subdirectory
17:29.00lkcldid apt-get build-essential
17:29.05lkcland started building packages :)
17:31.30mnemoclkcl: they (mandriva) finally released a relatively new (.29.something) patchset for the Gdium
17:32.03mnemoclkcl: so it's a good idea to get a .deb for people to be able to install a modern debian :)
17:33.41mnemoci really hate their distro
18:07.24lkclmandriva?  yerrss...
18:10.21lkclfjp: don't freak out too much - chitech (or more specifically seatron, the designers of the SO-DIMM) are reading the DS2431 EEPROM... using GPIO Bank N, pin 4 :)
18:10.43lkclcackles manically with horror and respect
18:12.07SQlvpapirlkcl, do chitech still think the machines they sold us was bought for 'industrial espionage' via. a 'license loophole'?
18:23.24mnemoclkcl: the gdium comes with a fork (maintained by mandriva) of mandriva 2008.1 for mipsel, but they keep it separated from the rest of the world. so only the packages they cared to enable are avaialbe
18:28.41mnemoci hate how mandriva (company) managed to kill the gdium's olph project by their absolute negative to open to a comunity
18:38.38lkclSQlvpapir, that's what they told us that they are _concerned_ about.  i've informed them that they need to be more concerned about simply being overtaken and made irrelevant, and that by giving us the source code and us mainlining it, it actually _strengthens_ their position by increasing brand awareness
18:47.56lkclADC is used to read the battery power
19:18.51lkclfjp: the "sane" function for poweroff, called from kernel_halt and kernel_shutdown_prepare
19:19.00lkclis ineptly named eepc_power_off
19:19.04lkcland does:
19:19.21lkcls3c_gpio_cfgpin(S3C_GPK12, 1);
19:19.28lkcls3c_gpio_pullup(S3C_GPK12, 1);
19:19.40lkcls3c_gpio_setpin(S3C_GPK12, 0);
19:20.38lkclthat's DIFFERENT from the poweroff function which is called when the insane-looking "secrets" reading from the eeprom all go haywire, which does pullup 0 cfgpin 1 setpin 0 - in that order
19:22.23fjpSo this is what should be put in s3c6410_power_off?
19:23.31lkclprobably, yes - if that's the absolute absolute last thing that needs to be done, it sounds like it, yes.
19:23.58lkclin the ctpc89e kernel they've just replaced absolutely everything in kernel_halt with a single call to eepc_power_off
19:24.17lkclthus neatly bypassing all the safety checks performed by the kernel in shutting the machine down
19:25.00fjpSo instead off
19:25.03fjpstatic void s3c6410_power_off(void)
19:25.03fjp{
19:25.04fjp<PROTECTED>
19:25.04fjp<PROTECTED>
19:25.05fjp<PROTECTED>
19:25.05fjp<PROTECTED>
19:25.07fjp<PROTECTED>
19:25.08fjp}
19:25.13fjpwe get:
19:25.15fjpstatic void s3c6410_power_off(void)
19:25.15fjp{
19:25.15fjp<PROTECTED>
19:25.16fjp<PROTECTED>
19:25.17fjp<PROTECTED>
19:25.17fjp<PROTECTED>
19:25.19fjp}
19:25.41fjpAdd a "?" behind "Still needed"
19:25.47lkclyep that's the trick.
19:26.02lkcli have no idea about that machine_restart call
19:26.56lkclmy guess is that this is "pthwappp - power off, bye bye" stuff.
19:27.03fjpI'll leave it in commented for now.
19:28.33mnemochow do you distinguish if it's really powered off or just halted?
19:28.39mnemocby looking at the device
19:28.41fjpmachine_restart() -> arm_pm_restart() -> arm_machine_restart()
19:28.57fjpvoid arm_machine_restart(char mode)
19:28.57lkcli have noooo idea.  the power LED goes off? :)
19:28.57fjp{
19:28.57fjp<PROTECTED>
19:28.58fjp<PROTECTED>
19:28.58fjp<PROTECTED>
19:28.59fjp<PROTECTED>
19:29.00fjp<PROTECTED>
19:29.00fjp<PROTECTED>
19:29.02fjp<PROTECTED>
19:29.03fjp<PROTECTED>
19:29.05fjp<PROTECTED>
19:29.05fjp<PROTECTED>
19:29.07fjp<PROTECTED>
19:29.07fjp<PROTECTED>
19:29.10fjp<PROTECTED>
19:29.10fjp<PROTECTED>
19:29.12fjp<PROTECTED>
19:29.12fjp<PROTECTED>
19:29.15fjp<PROTECTED>
19:29.15fjp<PROTECTED>
19:29.17fjp<PROTECTED>
19:29.17fjp<PROTECTED>
19:29.19fjp<PROTECTED>
19:29.19fjp}
19:29.24lkclcareful with that, fjp - you'll get kicked off for flooding :)
19:29.28mnemoclkcl: it goes off for halting too
19:30.28mnemoc:)
19:31.02lkclarm_machine_restart.... ah ha... they have some GPIO accessing in there, too... 1secc....
19:31.23lkcli made a record of all the functions which access 0xF6800000
19:31.34lkclor was it 0xF4000000 can't remember
19:32.08lkcloh bloody hell
19:32.11lkcl*sigh*
19:32.52lkcldirect s3C_VA_GPIO accessing.  _again_.
19:33.20lkclok they have cpu_proc_fin()
19:33.26lkclthen setup_mm_for_reboot()
19:33.50lkcland then ... arch_reset looks like it's been inlined ...
19:34.17fjpCould be. Could also be that that allows a hook.
19:34.48lkclokaaay.  arch_reset there are tooons of them....
19:35.23lkclS3C_VA_GPIO + 0xE4 ===
19:36.24lkclBank H + 4
19:36.26lkcl===
19:36.53fjpTake a look in include/asm-arm/arch-s3c2410/system.h
19:37.01fjpThat's where arch_reset is defined.
19:38.25lkclnope, it's nothing like that.
19:39.30lkclGPIO Bank H, Pin 4, is set to 0x1
19:40.57fjpThe thing is that I think we may have to implement it there anyway as the preparation that's being done in arm_machine_restart() is probably valid.
19:41.07lkclwait... no... 0xE4
19:41.14lkcl= Bank H + 4
19:42.12lkclS3C_GPH8 cfgpin == 0x1
19:42.37lkclit's s3c_gpio_cfgpin(S3C_GPH8, 0x1)
19:42.41lkclthen...
19:43.46lkcl0xEC = Bank H + 0x8 + 4 which means ....
19:44.16lkclS3C_GPH0..9
19:44.22lkclis the +0x4
19:44.30lkcl+0x08 is pullup
19:44.54lkclBIC 0x30000 means...
19:45.27lkclexcellent yes that's 0x3>>(8*2) so yes excellent that's S3C_GPH8 as well
19:45.45lkclso that means s3c_gpio_pullup(S3C_GPH8, 0x0)
19:46.35lkclthen there's _another_ one argh different pin
19:46.39lkcl0xE8
19:46.42lkcl= Bank H + 8
19:50.18lkclsetpin.
19:50.25lkclit's 0x4 + 0x4 for bank H
19:50.38lkclagain - 0x100>>8 = 1
19:51.04lkclso that's s3c_gpio_setpin(S3C_GPH8, 0x0)
19:51.24lkcland finally A0 =
19:51.26lkcler...
19:51.30lkcl0xA0/32...
19:51.40lkcl5
19:51.59lkclBank F
19:52.25lkcli wonder what the heck all this lot is about, anyway? :)
19:53.18lkcl0x30000000 == 0x3<<(14*2)
19:53.22lkclso that's Bank F, pin 14...
19:54.27lkclcfg.
19:54.47lkclso s3c_gpio_setcfg(S3C_GPF14, 0x0);
19:55.15lkcland then 0xA8 is 0xA0 + 0x8 which is Bank F pullup
19:55.33lkclso s3c_gpio_pullup(S3C_GPF14, 0x0);
19:55.35lkclHOORAY.
19:56.02lkclfjp: ok, got the sequence
19:56.23lkcls3c_gpio_cfgpin(S3C_GPH8, 0x1)
19:56.30lkcls3c_gpio_pullup(S3C_GPH8, 0x0)
19:56.38lkcls3c_gpio_setpin(S3C_GPH8, 0x0)
19:56.45lkcls3c_gpio_setcfg(S3C_GPF14, 0x0);
19:56.49lkcl<PROTECTED>
19:57.07mnemoc\o/
19:57.23fjpCool. Next question is: where to put it :-)
19:57.29lkclfjp: that's in arm_machine_restart, most likely in arch_reset.  actually probably best in s3c24xx_reset_hook
19:58.58lkclachhh... somewhere sensible, pleeease
19:58.59fjpYep. That looks like something we can customize.
19:59.22lkclmnemoc, can i ask you - do you have the kernel that fjp has been booting?  what's your basic-c programming like, e.g. doing an mmap application to read kernel memory at 0xf4000000?
19:59.45lkcli need a memory dump of all the GPIO registers
20:00.34mnemoclkcl: i haven't have time to open the chitech since the day after it arrived. terribly busy with $work$ :(
20:00.47lkclwe may be able to ascertain, by checking if GPF14, GPH8 etc. are set, if they are needed
20:00.50lkcltsk ! :)
20:01.14lkclok, so you're here for moral support - that's good enough :)
20:01.21mnemoc:)
20:04.13lkclok.  end of day for me.
20:06.18fjpHmm. We had the call to power_off commented out until now.
20:07.01lkclfjp: ahhhh - S3C_FPF14 is PWM TIMER0
20:07.10lkcllook in s3c6410_timer_setup
20:07.30lkclcase 0
20:07.34fjpHow does this look:
20:07.35fjphttp://paste.debian.net/64893/
20:07.44lkclfuck me they dick with the PWM TIMER0 directly
20:10.18lkclit's worthwhile establishing why s3c6410_timer_setup hasn't been called - i see no evidence of a printk PWM channel in dmesg on chitech kernel
20:11.16lkclwhat der fuk?  ohh - it's a test signal.
20:11.30lkcl#ifdef PWM_using_sample
20:12.54lkclfind . | xargs grep s3c6410_timer_setup
20:12.54fjp$ git grep s3c6410_timer_setup arch/arm/ :-)
20:12.54lkclokaaay.  the s3cfb_b089aw01.c and also the s3cfb_fimd4x.c :)
20:13.37lkclok fimd4x.c is channel 1 (not the right case value then...)
20:14.00lkcl__set_brightness is also... beh.
20:14.09lkcli'm looking at the code _we're_ creating
20:14.16lkclsod it... need the assembly *sigh*
20:15.40lkclfuckers.
20:15.55lkclthey're using case 0 s3c6410_timer_setup
20:16.10lkcls3c6410_timer_setup(channel=0x0, ....)
20:16.23lkclwhich explains why they "terminate" the timer directly
20:17.29lkclfjp: that's worth making a note of.  the s3cfb_set_brightness function uses a PWM timer, s3c6410_timer_setup channel 0, which uses GPIO Bank F, Pin 14, as a PWM
20:17.42lkclso they modulate the backlight brightness, with that.
20:24.18lkclfjp: worth mentioning then in the http://paste.debian.net/64893/ that GPF14=0 is used for the PWM backlight brightness control.
20:24.23lkclto power it off
20:25.13lkcli wonder if the problems we're seeing with the GL850Gs are because of doing s3c6410_timer_setup with the wrong thing?
20:25.14lkclhmmm
20:27.59fjphttp://paste.debian.net/64898/
20:28.40lkcl1sec..
20:29.30lkclyehhh.... what the heck
20:52.27lkclfjp: hooray!  that was it = change channel=0
20:52.49lkclhave to check ... looks good
20:52.59lkclin s3cfb_b089aw01.c set_brightness
20:53.02lkclchange channel = 1
20:53.07lkclto channel = 0
20:53.10lkcland usb is ok
20:53.12lkcli thiiink
20:54.11lkclyesss, that's it.
20:54.25lkcli'm getting usb devices configured, 4 ports detected x 2
20:54.26lkclHOORAY
20:54.45mnemoc21:05:43 < lkcl> ok.  end of day for me.
20:54.45lkcloh... waitwait...
20:54.46mnemoc:)
20:54.48lkclha ha
20:54.52lkclmnemoc: sssh
20:54.56lkclwait...
20:55.05mnemoc^^
20:55.08lkclawww
20:55.17lkclwossgoing on?
20:55.19lkclhmmm...
20:55.21lkclfalse alarm...
20:55.23lkcldrat
20:56.09fjpCOOL!
20:56.17fjpnot cool?
20:56.23mnemocnope
20:56.28lkclfalse alarm
20:56.30lkcl1sec...
20:56.34lkcltry again...
20:56.57mnemocwonders where is the power adapter of his chitech...
21:01.31lkclargh - was editing smdk6400_machine_init
21:01.36lkclnot smdk6410 whoops
21:03.35lkclsighs
21:03.38lkclnope.
21:33.14fjplkcl: Is that channel change still valid or not?
21:58.39lkclfjp: yes, but it will take some code / testing to confirm it
21:59.01lkclbacklight changing (/proc or /sys whatever)
22:24.20lkclfjp: yesss, fucking got it.
22:24.41lkclOTG_PHY_CLK_VALUE = 0x40
22:24.47lkcli don't know why i didn't get that earlier
22:24.53lkcli realised it was different
22:25.11lkclbut i had accidentally set it to 0x2
22:25.16lkclkeyboard is working.
22:25.27lkclmouse i don't think i compiled in a module for it yet
22:30.04fjpWoot!
22:30.07fjpWhere?
22:30.36lkcldrivers/usb/host/ohci-s3c2410.c
22:30.46lkcl#define
22:31.38fjpIn the if or the else branch?
22:31.46fjpI think the if?
22:32.30lkclelse
22:32.33lkclnot if
22:32.50lkclyou should have #undef USB_HOST_PORT2_EN
22:32.51fjpSo the "UTMI Interface, Oscillator" value
22:32.58lkclyep.
22:33.03lkclchange that to 0x40
22:33.41lkclwhere it's called, in usb_hcd_s3c2410_probe MAKE SRE that it says
22:33.56lkclusb_host_clk_en(S3C_USB_CLKSRC_48M, ....)
22:34.32lkcland then double-check that you haven't pissed about with case 0x1 (S3C_USB_CLKSRC_48M)
22:34.34lkcliiin....
22:34.46lkclwell... in usb_host_clk_en :)
22:34.52fjpHmm. There's no s3c_gpio_setcfg function
22:34.56lkcland that's in smdk6410
22:35.03lkclno. that shouldn't be activated
22:35.12lkcl#undef USB_HOST_PORT2_EN remember?
22:35.25lkcland you should _not_ have CONFIG_PLAT_S3C24xx set either
22:35.33fjpNo talking about your power change.
22:36.12lkclbasically they're _not_ enabling USB-OTG, they're *only* using its clock :)
22:36.25fjp<PROTECTED>
22:36.25fjp<PROTECTED>
22:36.25fjp<PROTECTED>
22:36.29lkclso it's all completely fscked
22:36.31lkcloh right
22:36.32lkclthat
22:36.37fjpThat first function does not exist
22:36.42lkcl1sec...
22:36.47fjpcfgpin?
22:37.13lkclsorry it's....
22:37.14fjpor setpin?
22:37.17lkclehmmm...
22:37.26lkclbollocks
22:38.20lkclcfgpin
22:38.24lkcls3c_gpio_cfgpin
22:44.18bjdooksright, build fixes commited to git head
22:45.20fjpbjdooks: I hope to get a repo for our netbook up tomorrow. Would be great if you could look over our changes sometime.
22:45.39bjdooksright
22:45.54bjdooksi'll state a new next-samsung at the weekend
22:46.01bjdooksi'll start a new next-samsung at the weekend
22:46.28lkclbjdooks: ibot can do s/ for you :)
22:46.39lkcls/ibot/the lovely ibot/
22:46.54bjdooksline-editors rock
22:46.58lkclhehe
22:47.02lkcltim rikers rocks
22:47.15fjplkcl: Have USB too.
22:47.25lkclyou do? fxxxxing fantastic.
22:47.34bjdooksthat reminds me, I need to fix the otg gadget code
22:47.44bjdooksand get moving on finishing the dma support
22:47.46lkclthere's one GPIO for wifi power that'll be needed
22:47.57lkcli'll decode the function tomorrow
22:49.28lkclfjp: did you get keyboard and mouse or just keyboard
22:51.00fjpBoth!
22:51.06lkclyaay
22:51.11lkclbrilliant.
22:51.35fjpBut touchpad with delay. Takes some time.
22:51.44lkcloo.
22:51.50lkclouch
22:51.50lkcloh?
22:52.20lkclanyway.
22:52.23fjpTo get activated. Once it works its fine
22:52.33lkclright. ok.  oops :)
22:52.34fjpHmm. Desktop has disappeared...
22:52.55lkclbacklight just got activated? :)
22:52.59lkclperhaps?
22:53.17fjpNo. I still have the background and the clock, but no menu etc.
22:53.43lkclohh.  you're using the chitech "MOS"?
22:53.44fjpMaybe the rootfs is fckd
22:53.59fjpYes. Our kernel + their rootfs
22:54.02lkclok
22:54.12lkclcould be some interaction with hal...
22:54.19lkcldevices going haywire / non-existent
22:54.56lkclWIFI's non-existent, backlight is fscked, midfun-keys don't exist, and those are hard-coded into hal configs
22:55.06fjptests shutdown from login screen
22:55.12bjdooksright, goodnight all.
22:55.14lkclso to be honest i'm not surprised something went splat, there: the main ...
22:55.21lkclbjdooks: night
22:55.26lkcli should be off too
22:55.29lkclyep
22:55.31fjpNope. Still stays on
22:55.57lkclok: suggest leave it, investigate tomorrow.  i did accidentally kick the power GPIO :)
22:56.09lkclK12 instead of K13 or vice-versa :)
22:56.12fjpYeah.
22:56.20lkclin smdk6410_machine_init
22:56.22lkclha ha
22:56.31lkclswitch on, power off
22:56.35lkclswitch on, power off
22:56.38lkclwhoops
22:57.05lkclS3C_GPK12 is the one that controls power.
22:57.05fjptouchpad sometimes works sometimes not. Same as with their kernel
22:57.12lkcloh dear!
22:57.18lkclthat says hardware fault, then
22:57.27lkclmine's ok
22:58.10lkcli think these touchpads will go the way of the usual "static zapped me and i now activate even if you approach me from a distance of 6 in"
22:58.15fjpStart menu does show initially, but disappears later.
22:58.23lkclwonder why...
22:58.52lkclprobably segfault on one of the applets (wifi, backlight)
22:59.00lkcl_but_
22:59.01lkclok.
22:59.02lkclenough
22:59.05lkclsleep for me
22:59.06lkclreally
22:59.08mnemoclkcl: do you have the "fantastic thoughtful analysis and business case" somewhere i could read it? :)
22:59.14fjpYeah. Progress again :-)
22:59.18lkclmnemoc: tomorrow
22:59.22lkclping me again
22:59.26lkclyehh
22:59.26mnemoclkcl: deal
22:59.38lkclnight folks
22:59.44fjpnight
22:59.58mnemocsleep well lkcl, well deserved rest :)
23:07.21fjp's off too
23:08.03mnemocgood night fjp

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.