mnemoc changed the topic of #arm-netbook to: EOMA: Embedded Open Modular Architecture - Don't ask to ask. Just ask! - http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68 - ML arm-netbook@lists.phcomp.co.uk - Logs http://ibot.rikers.org/%23arm-netbook or http://irclog.whitequark.org/arm-netbook/ - http://rhombus-tech.net/
popolon has quit [Quit: Quitte]
<drachensun> I've been digging through the chat log and ceo16 managed a succesful build but said he had to use g++-4.5 not 4.6
<drachensun> rellla: do you know compiler version you have?
<drachensun> I tried to install 4.5 but from apt only 4.4 and 4.6 were available ( I think)
<drachensun> and 4.4 has problems also
<lundman> now I've got to rememer how to unpack a dep file
techn has quit [Read error: Connection reset by peer]
<lundman> hmm new xbmc does not start in 1080p mode
tuliom has quit [Quit: Konversation terminated!]
gzamboni has quit [Ping timeout: 240 seconds]
gzamboni has joined #arm-netbook
Lupu_ has joined #arm-netbook
Lupul has joined #arm-netbook
alcides has quit [Quit: man woman || $> Segmentation fault (core dumped)]
dyoung has quit [Ping timeout: 246 seconds]
destinal has quit [Ping timeout: 246 seconds]
destinal has joined #arm-netbook
dyoung-away has joined #arm-netbook
dyoung-away is now known as dyoung
dyoung has quit [Changing host]
dyoung has joined #arm-netbook
QingPei has joined #arm-netbook
Lupul has quit [Quit: leaving]
Lupu_ has quit [Quit: Page closed]
<drachensun> lundman: can you put the package up for download?
tekzilla has quit [Ping timeout: 252 seconds]
tekzilla has joined #arm-netbook
<lundman> already is
<drachensun> the build deb package? would you mind sending a link?
<drachensun> built deb package I mean
mSquare has joined #arm-netbook
<drachensun> excellent, thanks
<drachensun> the other guys package file was 87 M
drachensun has quit [Quit: Leaving]
wingrime has joined #arm-netbook
AndChat|128225 has quit [Read error: Connection reset by peer]
rellla has joined #arm-netbook
QingPei has quit [Ping timeout: 246 seconds]
<rellla> drachensun: if I can trust my config.log, then it was 4.6.3-11
QingPei has joined #arm-netbook
QingPei has left #arm-netbook [#arm-netbook]
avernos has quit [Ping timeout: 246 seconds]
avernos has joined #arm-netbook
Quarx has joined #arm-netbook
<rellla> morning all
<lundman> rella
<lundman> I'm guessing your hilight colours of course :)
<rellla> ;-)
<rellla> since yesterday my mele isn't booting anymore
<rellla> full log: http://pastebin.com/ePbxPDU4
<rellla> what does that mean: [sw_hcd0]: <0>Internal error: Oops - undefined instruction: 0 [#1] PREEMPT
<rellla> something corrupted?
<lundman> undefined instruction, that is new
<rm> rellla, did this kernel ever boot for you?
<rellla> yes, i did a 6h compile orgy and xbmc testing yesterday
<rm> well, no idea other than afaik enabling preemption in the kernel config is a bad idea
<rellla> it was the def-config from amerys branch
<lundman> yeah why is everyone turning on PREEMPTIVE.. i think they might believe that means preemptive :)
<lundman> maybe the uImage was corrupted, with 00s
<rellla> i'll try recopy uImage and libs later ...
<rm> defconfig has a number of stupid things :P
<rm> like enabling tickless
<rm> which is not even supposed to work afaik
wingrime has quit [Quit: Bye]
popolon has joined #arm-netbook
<oliv3r> I got a10 hardware for my birthday, :D
<lundman> \o/
<mnemoc> :)
<mnemoc> oliv3r: happy birthday ;-)
<popolon> happy birthday
<valhalla> oliv3r: happy hacking birthday
sspiff has joined #arm-netbook
RaYmAn_ is now known as RaYmAn
phh has quit [Read error: Connection reset by peer]
phh has joined #arm-netbook
avernos has quit [Ping timeout: 246 seconds]
merbzt has joined #arm-netbook
sspiff has quit [Remote host closed the connection]
sspiff has joined #arm-netbook
sspiff has quit [Changing host]
sspiff has joined #arm-netbook
merbzt has quit [Ping timeout: 240 seconds]
<oliv3r> finally! :p
avernos has joined #arm-netbook
<oliv3r> no mele though; a tablet
<oliv3r> yarvik tab264 (momo9) sun4i one
<oliv3r> it has uSD slot, so should be able to boot from it :)
<mnemoc> yes
<mnemoc> and so to play with the registers ;-)
<oliv3r> gonna install cm9/10 first on it though (the default android is horrible) and start playing with uBoot etc via SD
<oliv3r> and! play with registers :D
<oliv3r> and hopefully not break it :p
<oliv3r> but thre's some learning curve to go passing by :p; I really have to refrain myself from starting with registers, so I can finish on the documentation :)
<oliv3r> there's no clockworkmod tree is there
<oliv3r> on the wiki; anyway, if i discover anything i'll write a page on it
<mnemoc> :)
<mnemoc> refactoring existing pages is also welcomed
<mnemoc> every tutorial decided to teach (again) how to install a compiler :|
<oliv3r> while linux documentation overal is good and things are wel explained, i find android/roms to be more like the 90's 'warez' scene, zips, binaries and half/vague instructions
<oliv3r> <- gentoo user
<mnemoc> there are many gentoo users here
<mnemoc> but not me :p
<oliv3r> lol
<oliv3r> its redumentary but there's _something_ :p
<zub> rm: I'd also vote for cleaning up defconfig, as some options seems wrong/useless
<mnemoc> patches welcomed
<zub> rm: btw. I've been playing with my "minimal config" and I keep running into strange issues, crashes on boot. It seems to be related to CONFIG_PREEMPT
<zub> rm: and CONFIG_BLK_DEV_INTEGRITY
<zub> mnemoc: I think I already used up my credit :( and I'm not really sure myself what should be included in the defconfig, jsut some options seems obviously useless, while some just suspect
<mnemoc> the problem with preempt should now be solved with the fix on the regulator detection and the changes in the clock handling
<zub> rm: thx
<mnemoc> rm: i would rather take a patch
<mnemoc> #52 is more of a wishlist
<oliv3r> What are the differences in speed between a reasonable fast SD card, and the onboard nand? how fast IS the nand? will i notice much difference in speed when using an older 512mb SD card for example (for testing)
<mnemoc> oliv3r: I personally totally ignore the NAND
<mnemoc> it's driver is crap and loves to halt the world when there is some writting
<oliv3r> good, since i have an arsenal of old SD cards (512MiB-8GiB)
<oliv3r> one of the 'improvements' needed I assume?/
<mnemoc> totaly rewrite I would say :|
<mnemoc> but we still don't know enough about the nandc to do so
<oliv3r> :S
<mnemoc> using mtd
<zub> mnemoc: I've been getting crashes somewhere in drivers/usb/sun4i_usb/hcd/hcd0/*: http://linux.fjfi.cvut.cz/~zub/crash.log
<zub> but have not had time to investigate deeper
<mnemoc> zub: how old is your kernel?
<zub> it seems to be randomly caused by changing config
<oliv3r> the hardware DOES expoxe the mtd 'layer'?
<mnemoc> nope
<zub> fea836a3876514a9097de4fe6181d0cb1b332a44 is the latest commit, linux-sunxi-3.4
<zub> so I think it's up to date
<mnemoc> usb on 3.4 is known to have troubles, yes
<mnemoc> the forward-port is still incomplete
<mnemoc> specially on the gadget side
<mnemoc> but needs love in general
<zub> mnemoc: the actual line that dies is "status = usb_add_hcd(sw_hcd_to_hcd(sw_hcd), -1, 0);" (line 1480 in sw_hcd0.c), and that is caused by usb_add_hcd (dies in dev_info(hcd->self.controller, "%s\n", hcd->product_desc);
<zub> the pointr is not null... but apart form that I don't know how/why dies it explode
<mnemoc> zub: we still don't have usb maintainer... do you volunteer? ;-)
<zub> funny thing is it's sort-of-random, I didn't figure out one config item that causes it... it seems idfferent ocnfig = different result, maybe affected by build
<zub> mnemoc: I have already failed
<zub> if you need some touchscreen hacks, or freescale framebuffer hacks, I could know something :-P
<zub> too bad there's not freescale FB on the allwinner :)
<mnemoc> failure is a requirement to real success ;-)
<zub> I already asked, but let me try again: CONFIG_BLK_DEV_INTEGRITY - does this make sense on the allwinner? does it even do anything at all? it's on in the defconfig
<mnemoc> silence usually means "no idea"
<zub> it seems to have affected the crashes with usb, but it could be that just the code is laid out differently/soem random effect
<zub> btw. anybody here is using gdb/something to debug the kernel? I guess JTAG is one option. Bit what about things like kgdb? (I don't have JTAG. I guess I could get it via the uSD hack, but don't have it so far.)
<mnemoc> hno and RaYmAn use jtag
<mnemoc> i bought one, but haven't needed to learn to use it yet...
<zub> so you're going with printk? :)
<mnemoc> yup :p
<zub> what irritates me the most is inserting/removing the uSD all the time... :(
<zub> MK802+ doesn't have the eth connected, so I guess I can't boot from net even if uBoot could do it; maybe boot-over-serial (never tried)
<mnemoc> i keep two uImage files, if the development one fails I tell u-boot to use the known-to-work one
<mnemoc> zub: you can try using `fel`
<zub> ah, and then you update the uImage w/o removing the card. good idea
<zub> mnemoc: is it possible to use FEL to receive and start an uImage?
<zub> (or, well, any kernel image)
<mnemoc> spl + atags + uImage, theoretically should work
<zub> A10 is magic! :)
RITRedbeard_ has quit [Ping timeout: 245 seconds]
<zub> I think two uImages will work for me, that sounds easy enough
<mnemoc> err. spl + atags + script.bin + kernel image
<mnemoc> works for me well enough to not need to explore other options yet ;-)
<mnemoc> zub: the A10 is VERY hacker friendly
<lundman> gets a 8/10 in hacking, and 6/10 for developer :)
ZaEarl has joined #arm-netbook
<mnemoc> a 2/10 for user wanting video decoding on linux :<
<RaYmAn> mnemoc: psh, next thing, you'll say tegra is very hacker friendly
<RaYmAn> :P
ZaEarl_ has joined #arm-netbook
ZaEarl has quit [Read error: Connection reset by peer]
<ZaEarl_> ugh, crummy hotel wifi
ZaEarl_ is now known as ZaEarl
<rellla> rm: doing a recopy of uImage and kernel libs did it. it's online again ;-)
<mnemoc> RaYmAn: :p
<mnemoc> RaYmAn: i meant the design of the chip itself.... the lack of documentation and crappy code hurts obviusly
<RaYmAn> ;)
<RaYmAn> Tegra isn't very friendly at all. You look at it wrong and it starts to sleep-of-death
<RaYmAn> (Seriously - ASUS disabled the low-power core in their JB release for the tegra3 tablets because of SOD)
<mnemoc> o_O
<oliv3r> ironically, video decoding in linux is 8/10 for me :)
<ZaEarl> anyone else in Hong Kong this week?
<zub> I wish :)
<oliv3r> next to searching all over the net, what's the best option to identify the harware in my tablet? I'll get busybox with lsusb and lspci to work; but then there's probably some i2c hardware?
<mnemoc> oliv3r: usually people takes them apart :p
<ZaEarl> oliv3r, a lot of devices don't show up in lsusb/lspci
<oliv3r> zaearl why not? I'd expect anything connected to those busses to show up
<mnemoc> ts for example is usually connected via i2c
<oliv3r> oh, gotta look for some teardown pics/stuff; the yarvic tab26[04]/momo9 is quite common
<ZaEarl> because they're connected other ways
<oliv3r> ah, well duh then :)
<zub> wasn't there some tool to list stuff on i2c (to the extent the stuff can be probed...)
<oliv3r> i know lm_sensors probes stuff with sensors detect
<rellla> strange behaviour: boot up mele->ok, 1st reboot->ok, 2nd reboot->kernel hangs, recopy uImage, boot->ok !? kernel version 3.0.42
<mnemoc> rellla: got a serial console?
<rellla> yep
<mnemoc> rellla: where does it hang?
<ZaEarl> I haven't seen hipboi in ages. Anyone know what he's been up to?
<rellla> once it was this sw_hcd PREEMT-thing I mentioned above, i try to reproduce and post log
<rellla> ...waiting 10s for reboot....
<mnemoc> he is usually available on xmpp/gtalk from an android phone
<oliv3r> saw him log on/off yeseterday? day before yesterday for sure
<ZaEarl> nickserv says he hasn't been on for 4 weeks.
<zub> sw_hcd...
<mnemoc> ZaEarl: that only counts users when they authenticate
<mnemoc> ZaEarl: btw, sw_ is short of softwinner... one of the old alter-egos of allwinner's software division
<mnemoc> err
<mnemoc> zub: btw, sw_ is short of softwinner... one of the old alter-egos of allwinner's software division
<mnemoc> ZaEarl: sorry, wrong <tab> completion
<zub> mnemoc: softwinner, hardwinner, allwinner... :)
<oliv3r> I saw on the mailing list, that the A10 rev C has a bug with PLL4 (sata?); /proc/cpuinfo does list some form of rev numbers, but mine is 'rev 2'. I doubt there's been 11 revisions since? the rev C correlate to the /proc/cpuinfo?
<mnemoc> oliv3r: afaik all new A10 chips are rev C
<oliv3r> boxchip may have been better
<oliv3r> :p
<oliv3r> mnemoc: well version 2, could be rev B then?
<mnemoc> i guess so
<mnemoc> unless the code counts from 0
<RaYmAn> i2cdetect is..well.. unstable by definition. There's no standard way to detect if a device is present on the i2c bus (it's device specific to the target device)
<oliv3r> true, well tablets don't use sata ports anyway, so it shouldn't matter, but still
<RaYmAn> but you can probably find some info from sysfs i2c stuff
<oliv3r> well on a pre-configured running stock ROM you should get most/all i2c connected devices i'd say
<RaYmAn> sure, but i2cdetect will not necessarily show them all (is my point)
<oliv3r> yeah, i guess they guestimate on based on known chips
jquip has joined #arm-netbook
<oliv3r> does anybody use any of the windows only tools? Like 'redscorpio's' tools etc? Do they work reliably in wine? I guess any big fuckup can be fixed by booting via uSD? as long as you don't break the BROM
<RaYmAn> If possible, it's a good idea to dump boot0 and boot1 before starting to play
<RaYmAn> but might require JTAG to do properly
<ZaEarl> oliv3r, yeah, a lot of that stuff works nicely in wine
<ZaEarl> you can't break brom
<oliv3r> i'd imagine that with jtag you can still flash BROM, oth, it also makes a lot of sense if it where true rom baked into the asic
<RaYmAn> indeed, but you can end up in a place where your device no longer works if you don't have the boot0/boot1 because they contains ram initialization parameters and similar
<ZaEarl> right, so keep a LiveSuit image around.
<RaYmAn> also, no , BROM is read only
<oliv3r> so boot0/boot1 is still read (by BROM) even when booting from uSD?
<mnemoc> nope
<mnemoc> when booting from uSD both are bypassed
<oliv3r> so in theory, i could fill the entire nand with 0; only have BROM working (duh) and safely boot from uSD and restore boot0/1 (from a backup)
rz2k has quit [Ping timeout: 245 seconds]
<RaYmAn> we don't know how to access the part of the nand that contains boot0/boot1 :/
<RaYmAn> but otherwise, yes
<oliv3r> so you cannot end up in a place where you cannot boot, as long as you have a workable bootable uSD card then
<RaYmAn> correct
<ZaEarl> LiveSuit can install a fresh rom without booting the device
<RaYmAn> ZaEarl: it kind of assumes you have a livesuit image for your device
<RaYmAn> or a compatible one
<oliv3r> livesuit a DFU utility? (direct flash)
<ZaEarl> of course
<mnemoc> he code to do raw access to the nand is available... we only need time and willing to dive into it and export those raw blocks as a /dev
<mnemoc> s/he/the/
<ibot> mnemoc meant: the code to do raw access to the nand is available... we only need time and willing to dive into it and export those raw blocks as a /dev
<RaYmAn> ah
<RaYmAn> mnemoc: where?
<mnemoc> in drivers/block/sunxi_nand/ :)
<ZaEarl> there are tons of generic livesuit images out there, that'll restore basic funcationality to most devices
<RaYmAn> ZaEarl: yes, assuming they have compatible ram.
<oliv3r> once you have restored some functionality via livesuite, you can always go from there I suppose
<oliv3r> I guess a 'livesuite' image, is nothing mmore then a complete nand dump
<RaYmAn> in essense, yeah (as usual, it's a bit more than that, but yeah)
<ZaEarl> it's a bit more complicated than that
<oliv3r> inc. the android stuff? or just boot stuff, u-boot etc
<mnemoc> livesuit seems to do some probing and write whatever it wants in boot1's header. not exactly what is in the provided script.bin
<oliv3r> well since we can't access boot0 and boot1 yet, doinga complete dump is imposslbe atm anyhow
<mnemoc> oliv3r: we can, from nand's uboot
<RaYmAn> without jtag, yeah
<oliv3r> oh, so with jtag we can doa full dump
<RaYmAn> even if jtag can't access nand, it's possible to dump it from memory
<oliv3r> maybe i should talk to hipboi and see if he sells me one of those uSD -> jtag adapters
<zub> what is the relationship between boot0/boot1 and /dev/nand?
<oliv3r> mnemoc: oh; so if i boot into a proper u-boot from uSD, i can use that to dump the full nand?
<mnemoc> oliv3r: no, only if you boot from nand's u-boot
<zub> is boot0/boot1 in an area that is not visible in /dev/nand?
<oliv3r> so based on zub's question, I'd assume, the linux/android kernel doesn't expose ALL of the flash in /dev/nand
ZaEarl_ has joined #arm-netbook
<RaYmAn> zub: yes
<mnemoc> oliv3r: only the logical layer
<zub> :(
<mnemoc> boot0/boot1 live before that
<oliv3r> mnemoc: confusing, but I understand, the original u-boot knows more and thus exposes it properly
<RaYmAn> mnemoc: it must be slightly more complicated than that, otherwise the /dev/nand expose patch would make it appear
<zub> btw. how does the nan partitioning (/dev/nanda, /dev/nandb...) work? can we alter the partitioning?
<zub> nand
<mnemoc> zub: there is a tool in sunxi-tools to repartition /dev/nand
<oliv3r> is this info on the wiki btw? instead of bombing you with the same questions over and over :)
ZaEarl has quit [Read error: Connection reset by peer]
<oliv3r> mnemoc: i assume to partition only the 'known' region
<mnemoc> oliv3r: the logical layer, yes
<oliv3r> so the A10 doesn't need to be told what the partition sizes are! (I know some phones with mtd based partitions, get told the partition sizes at boot time)
<mnemoc> /dev/nand has a custom partition table
<mnemoc> in the first block before nanda
<mnemoc> but "the tool" deals with that
<oliv3r> yeah but the partition table is (exposed) actually written to nand! that's good :) my phone doesn't have a partition table, i think the internal SPL/bootloader tells the kernel the partition dimensions via mtd= boot cmdline
<mnemoc> yuck
<zub> oliv3r: yup, seen that with mtd too
<oliv3r> well it is 'stored' in some way I suppose :)
<oliv3r> you can modify it, but only by overriding the mtd parameter, i think the spl stored value, is no way around it. a lot of phones actually do it that way it seems
<zub> oliv3r: you could hack the kernel to ignore the supplied mtd and use some hardcoded :)
<oliv3r> well yeah
<zub> yay for ugly hacks!
<mnemoc> that's the way of "the industry"
<oliv3r> I still fail to understand why are they so hard trying to make it 'annoying' (not hard, not difficult, just annoying).
<mnemoc> "get it done now! no one cares how. but it has to be done NOW!"
<oliv3r> regular consumers aren't ever gonna mess with this
<oliv3r> i ment their 'protection ''schemes'''
<oliv3r> but yeah, crappy code is deffinatly in that department
<oliv3r> I guess on the hardware side they shouldn't be allowed to work that way
<mnemoc> carriers claim they protect the networks
<oliv3r> it's hard to release a software update for a hardware flaw :)
<mnemoc> because a rooted phone is a liability
<oliv3r> that may be true and well
pawel5870 has joined #arm-netbook
<oliv3r> but you can have hacker friendly hardware, and still sell unrooted phones
<oliv3r> my tablet btw, came pre-rooted
<oliv3r> no official market, and I think it runs some development mode rom
<mnemoc> that's the case for most A10 devices
<oliv3r> Ah ok
<zub> I hate the idea of me buyin a device and not being allowed to do whatever I polease with it. I paid the $. Now give me my device w/o stupid DRM/locks/...
<oliv3r> exactly
<mnemoc> zub: apple fanboy I presume ;-)
<oliv3r> *I* worked for the money, *I* bought the device, *I* own it, yet i cannot do whatever I please with it
<zub> mnemoc: ehm? never had any apple device and I don't plan to buy any
<oliv3r> (cedar/mali even here applies, boot0/1 etc follows)
<zub> mnemoc: I prefer Allwinner to Apple :-P
<oliv3r> I must say, I do find the A10 SoC very very interesting chip (probably due to its hacker friendlyness/price)
<mnemoc> same. no sony, no apple, no amazon. no any device from a company who believes they need to keep control of the products after they are sold
<zub> but I bought an HTC phone. Part of that $ came to M$ (patents). :(
<zub> and it was locked, but I picked one that was easy to root
<ZaEarl_> now, if we could just get the OEMs to respect the GPL
<oliv3r> i got an old (used) iPhone 3gs from work, that i had for a few days and traded it for an old very little used HTC hero
<oliv3r> but it was free, so can't complain. I do own a samsung washingmachine though :)
<zub> :)
<zub> oliv3r: does it have JTAG? :)
<RaYmAn> hacking your washingmachine seems like taking things a bit far :P
<mnemoc> but hacking those cleaning robots seems fun :p
<RaYmAn> yeah
<oliv3r> lol haven't it opened it yet :D
<oliv3r> RaYmAn: now, maybe, but in time, when we have all opensource sofware everywhere, we'll need a new challange i'm sue :p
<oliv3r> also, washing mashines will get more 'powerfull' and will feature a true OS at some point
<oliv3r> think, touch screen, 'options'
<zub> "clone my great washing programme from git:/... :)"
<oliv3r> washes twice as clean, twice as fast, at half the cost
<oliv3r> and no more reboots!
<oliv3r> hmm, the momo9 tablet (twin brother it is said, or rebadged) boots from USB
<oliv3r> oh!
<oliv3r> do'hh I should have looked (but i still have 8 tabs open needing some reading :p)
<mnemoc> nah, it's still empty
<oliv3r> lol
<oliv3r> right, all this talking and having lunchmeans i haven't started on finishing the timers page, which i'll get to in a few :)
<mnemoc> good idea :)
mysteryname has joined #arm-netbook
<mnemoc> hno: around?
<andoma> anyone been using mali together with FB?
<andoma> i get lots of tearing artifacts, etc
<mnemoc> hno: http://sprunge.us/eRRH?diff <--- reworked memory mangling proposal
tzafrir has quit [Ping timeout: 248 seconds]
<oliv3r> one last question before I get to work; all those CWM/Android flashing 'tools/instructions' don't really do anything other then format/mount/rm -rf/extract? If I change the partitioning scheme, those all should remain to work error-free I assume
<mnemoc> CWM/Android flashing are windows-people oriented
<mnemoc> but livesuit images come with encrypted partition images and a .fex file including repartitioning info
<mnemoc> so it will kill anything you do
<oliv3r> yeah, everything is dumbed down too much imo and it's all 'hacker-secret-club' sorta thing
<oliv3r> anyhow, the stock software is quite crap, and some people have build CM9 roms and CWM roms, so installing those should be no prblem after partitioning :)
<mnemoc> Turl has a pretty clean tree for CM10 for A10
rz2k has joined #arm-netbook
<oliv3r> but even things like 'built from allwinner's sources (baseband 1.2) ... what? baseband is a wireless communication thing I would think, I know they use it on radio-firmwares too, but not sure what baseband relates to wrt allwinner
<oliv3r> yeah but I wonder if it'll work on my tablet, can try of course, i'll look it over
<mnemoc> as long as you keep your script.bin and you don't rely in a .ko we don't have in the open tree. it will work
<oliv3r> the wiki mentions for nand-part that you need a kernel patch to expuse the FULL nand block device, i take it that is not entirly true? since there's some hidden bits (where boot0 and boot1 reside)?
<mnemoc> it's the full *logical* nand block
<oliv3r> thought so
<mnemoc> rm: have you tried to enable thumb2 on an A10 kernel?
<rm> no
<rm> you mean "compile kernel as thumb (experimental)"?
<mnemoc> yes
<rm> no, didn't try that
<rm> sounds interesting, would be nice to try and somehow benchmark the effects
<mnemoc> indeed
<rm> but I'm afraid the kernel will flat out not boot
<rm> and I don't have convenient means to debug
<lundman> did you try --force!
<mnemoc> hno: http://sprunge.us/ZaeS?diff <--- improved simmetry and dropped CONFIG_ rename (for the sake of 3.0 "stability")
<jquip> hows linux on nand? i was told to think with ubifs or yaffs... i'm still having an SD card :/ what's the best way to do this? nande apparently has enough space... anybody try nand repartioning? Thinking about a linux only tablet....
<mnemoc> CONFIG_THUMB2_KERNEL=y .... so here we go
<mnemoc> jquip: on linux you don't have raw nand, so raw nand filesystem don't work
<mnemoc> same on SD cards
<jquip> mnemoc: yes yes and no no... ubifs/yaffs works I hear..
<mnemoc> /sys/kernel/debug/kmemleak is always angry :<
<jquip> or am i mistaken?
<mnemoc> jquip: never tried myself
<rm> mnemoc, I was turning it off in rc.local
<rm> and now I think I disabled it in .config
jquip has left #arm-netbook [#arm-netbook]
tuliom has joined #arm-netbook
tzafrir has joined #arm-netbook
<mnemoc> rm: http://dpaste.com/812381/plain/ <--- with thumb2 enabled
<mnemoc> crashed with an alignment error on the usb junk-driver
<rm> turn off preempt!!!!!
<rm> but yeah
<rm> wouldn't bet it will help
<rm> facepalm
<rm> and you also run tickless
<mnemoc> we are supposed to have those problems fixed...
<rm> that's almost like TRYING to build a kernel that won't work
<mnemoc> i don't want to have a kernel that runs only disabling things
<mnemoc> but the point was, it boots
<mnemoc> and that usb driver needs fixing ;-)
<rm> maybe try disabling USB driver!
<rm> :D
<mnemoc> :p
mysteryname has quit [Ping timeout: 252 seconds]
dyoung has quit [Ping timeout: 245 seconds]
zub has quit [Ping timeout: 246 seconds]
destinal has quit [Ping timeout: 246 seconds]
zub has joined #arm-netbook
drachensun has joined #arm-netbook
destinal has joined #arm-netbook
dyoung-away has joined #arm-netbook
dyoung-away is now known as dyoung
dyoung has quit [Changing host]
dyoung has joined #arm-netbook
<mnemoc> rm: anyhow, can you test if you still have inestabilities when using tickless and preempt?
<mnemoc> rm: they should be gone by now
<zub> mnemoc: aaah, my favourite crash (the usb driver)
<mnemoc> :)
<hno> mnemoc, I think you are overdoing it here, but looks fine other than that.
<zub> somebody should fix that :-P
<zub> I know nothing about it... but maybe I can at least try to figure out the cause...
<mnemoc> zub: I only got it on a thumb2 experimental build
<zub> I get very similar crash randomly... http://linux.fjfi.cvut.cz/~zub/crash.log
<zub> very similar ~ in similar parts of the code
<hno> zub, please read code and expand on the register guide so a better driver can be written.
<mnemoc> hno: what do you think it's an excess there?
<hno> Having more than one define.
<zub> hno: I'll give it a try, but knowing myself (I know nothing about USB/kernel, and I don't have that much free time anyway) I can't promise anything :(
<oliv3r> hno: do you know more about the CPUCFG register with regard to the chip version? you write 00 = A, 11 = B and ?? = C. with only 2 bits available, got any more info by any chance now?
<mnemoc> hno: it's one for allwinner boot hacks, and a second to request mali's reserve
<hno> mnemoc, and head.S part is not about mali, it's about booting directly from boot1.
<mnemoc> hno: yes, that change is to kill the machine type hack together with the memsize hacks
<oliv3r> hno: the datasheet says default: 0x03 (I assume that to be rev C, but you claim it to be B?) and says 'reserved to 2'b11 ??
<mnemoc> hno: obviusly the define needs to be renamed
<mnemoc> 'A'+x == 'C' for x == 2
<oliv3r> anywhere in the source I could check that?
<mnemoc> hno: but the idea was to bind both hacks in the same CONFIG_
<hno> And that head.S code should move somewhere else. I.e. zimage header or a custom header.
<oliv3r> mnemoc: i get the simple math behind it :p; but if 0 = A, and 3 = B, C would have to be either 1 or 2, which ... seems odd :p
<hno> oliv3r, rev A & B have the same value.
<oliv3r> so that register is somewhat useless? :S
<hno> Olny RevC needs code to behave slightly different.
<oliv3r> let me rephrase, what shall I put up on the wiki :p
<oliv3r> your notes seem to not match what you say here
<oliv3r> I doubt there's any reference in the source about it? other then reading the value
<oliv3r> hno: I take it you refer to the PLL4/Sata patch?
<mnemoc> oliv3r: in arch/arm/mach-sun4i/core.c of the kernel look for sw_get_ic_ver
<mnemoc> oliv3r: and yes, 0 == MAGIC_VER_A (0), 3 == MAGIC_VER_B (1), anything else == MAGIC_VER_C (2)
<oliv3r> but hno just said that A and B share the same magic number?
<oliv3r> mnemoc: found sw_get_i2_ver; thank you :)
<Turl> rm I run with preempt on all the things A10, no issues
<mnemoc> i don't know... i didn't write those functions
<hno> Hmm.. seems I remember wrong there. A = 0, B = 0b11, C= anything else.
<Turl> $ ssh mele "uname -a; uptime"
<Turl> Linux mele 3.4.12+ #6 PREEMPT Wed Oct 10 19:24:57 UTC 2012 armv7l GNU/Linux
<Turl> 08:55:46 up 18:24, 0 users, load average: 0.00, 0.01, 0.05
<rm> Turl, also with cpufreq enabled?
<Turl> ye
<rm> and on 3.0.x as well?
<Turl> yes
<rm> ok, will try it next time
<oliv3r> hno: you made me all confused! I will change that and then save the timers page; which means its ready for review and done! took me only 2 1/2 days! :(
<zub> oliv3r: looks impressive
<mnemoc> oliv3r: amazing job. thank you!
<oliv3r> i saw the next section though; which scared me
<oliv3r> i think it's about twice as long :S
<oliv3r> anyway, i'll go a few chapters back and start with something smaller :p
<oliv3r> it may need some minor formatting
<oliv3r> I still think a template would have been usefull for those tables :p
<oliv3r> Tech companies should just use wiki's to begin with for their datasheet, much easier to revise :D
<oliv3r> Clock control module will be tomorrow's goal!
<mnemoc> :)
<zub> http://limadriver.org/ ... so there is some hope :)
<mnemoc> that one is critical, after migrating our kernel to pinctrl, common clock framework driver will be next
<mnemoc> zub: that basically means libv
<mnemoc> unfortunatelly he has been way to busy doing $work$ these months :<
<mnemoc> s/to/too/
<ibot> mnemoc meant: unfortunatelly he has been way too busy doing $work$ these months :<
<Turl> mnemoc: have you noticed the lack of spambot registrations? :D
<mnemoc> Turl: yes, thank you! :)
<Turl> I thought it wouldn't work, but it seems it has done the trick :P
<mnemoc> Turl: do you know if we can *delete* the bogus accounts?
<Turl> yeah you can
<Turl> one by one :<
<mnemoc> :<
<mnemoc> mediawiki's admin side sucks badly
<libv> yeah, only tend to spend a few h per weeks on lima, sadly
<libv> s/weeks/week/
<ibot> libv meant: yeah, only tend to spend a few h per week on lima, sadly
<mnemoc> libv: any perspective of change on that?
<mnemoc> libv: hi :)
<libv> not atm, and hi back :)
<oliv3r> pinctrl, is that the first step to mainlining sunxi?
<libv> with any luck, i can get the demo i had planned for XDC going in time for FOSDEM :(
<oliv3r> was* (sounds reasonable as it sounds quite low)
<mnemoc> oliv3r: not the first in general, but it's critical for core mainlining
<oliv3r> I figured CCM and Timers be equally important :)
<oliv3r> well in that case, i'll go with CCM for sure, which would you reccon I should look at after?
<mnemoc> oliv3r: drivers unification and bug fixing is also toward mainlining
<oliv3r> (htough I think ccm will take 2 days easy)
<ManoftheSea> lkcl: Hoy Luke.
<oliv3r> probably system control register; after those two, i'll refactor the ones I did (to match the timer one in style)
<mnemoc> :)
<ManoftheSea> Have people seen ZaReason's got an A10 tablet?
<mnemoc> ManoftheSea: ZaEarl has been here for like a year
<ManoftheSea> Is that it. I figured someone was.
<mnemoc> Turl is also part of that team
<oliv3r> time to go home!
<oliv3r> see you all tomorrow :)
<mnemoc> good night oliv3r
<oliv3r> it's only 16:30 here :p
<oliv3r> but thank you :)
<zub> ha, same TZ :)
* Turl is still in the AM :P
<andoma> CET ftw
<mnemoc> not really when you live in the westernest edge of europe and you are forced to use germany's TZ
<ManoftheSea> I'm thinking on the Laptop shell for an EOMA card... specifically, power.
<ManoftheSea> The laptop shell should be able to charge and turn-on without an EOMA card, right?
<mnemoc> yes, the laptop board needs it's own STM32F controller it
<mnemoc> controlling*
<ManoftheSea> Then, when EOMA card is inserted and user says "go", power goes across the connector.
<mnemoc> right
<ManoftheSea> mnemoc: in the STM32F case, the STM just gets out of the way? Until when? The EOMA signals "hey, turn me off"?
<mnemoc> then the module reads the description of the io board from an eeprom over i2c and learns about it's periferals
<ManoftheSea> mnemoc: right, I understand the idea of the EEPROM, even if I don't totally get device tree yet.
<mnemoc> ManoftheSea: it shouldn't get out of the way... it will still need to take care of battery charging and stuff
<ManoftheSea> well, out of the way of the screen, if it ever was there.
<mnemoc> i don't think it should ever be
<ManoftheSea> But how does the EOMA do power management/backlight control/power-off?
<mnemoc> don't know... isn't that part of the LCD pins set?
<ManoftheSea> Battery info can go across USB, I imagine.
<ManoftheSea> backlight control? I didn't think so.
<mnemoc> afaik the idea was to use openEC to let both talk
<ManoftheSea> openEC refers to x86 ports. I'm not claiming to understand it.
<mnemoc> no clue how is that implemented in reality
<ManoftheSea> Yeah, LCD pins are just pixel data, v- and h-sync, and an enable.
<ManoftheSea> Well, that's what I'm thinking on. How does the card turn its I/O board off?
<ManoftheSea> Or, in the case of a capable laptop, return control to the STM32F
<mnemoc> i think so, the "EC" has to take care of it
<mnemoc> but i'm not familiar with the life at that depth
<ManoftheSea> yeah, that page says an LPC interface.
<ManoftheSea> whatever that is (/me uses the google)
<ZaEarl_> I wouldn't expect the EOMA card to have an EC
<mnemoc> not the card. the device
ZaEarl_ is now known as ZaEarl
<ManoftheSea> the laptop shell side.
<ManoftheSea> Hmm, Low Pin Count, it's 4 wires, and it's like ISA.
<ManoftheSea> I don't think we want it.
<mnemoc> ManoftheSea: send an email to the list about it
<mnemoc> luke rarely looks at here
<ManoftheSea> I will. I'm currently limited only to Google's web interface. Bleh.
* ManoftheSea points finger in mouth.
<mnemoc> you can post from gmane too
<ManoftheSea> mnemoc: regarding luke, I was just hoping to catch him, but nothing important.
<ManoftheSea> Oh hey, good idea.
<ManoftheSea> On the other hand, I'd like to do a little research first, and present what I think are options.
<ManoftheSea> Rather than just asking you (plural) to do everything.
<ManoftheSea> oh hey! GPIOs. Those could bitbang a 4-wire protocol.
<ManoftheSea> Even if it is emulation of something ancient.
<mnemoc> bitbanging is still popular :p
lerc_ has joined #arm-netbook
<drachensun> where can I get the cedar libs for cedarxplayertest
<drachensun> I figured it all out before and I didn't write it down it seems
<mnemoc> the xbmca10 thing has an script to download them ..... from my dropbox :|
lerc has quit [Ping timeout: 260 seconds]
<drachensun> ok cool
<mnemoc> there are tons of .zip and .rar files inside, not sure which he uses
<mnemoc> but the script knows
<drachensun> I'll try and figure it out
<drachensun> maybe put them back up in a github instead
<drachensun> yeah, there were a ton of different pieces in there
<ManoftheSea> do ARM chips "use" ACPI?
<drachensun> I couldn't figure out how they all were different
<mnemoc> drachensun: can you do a pull request to update https://github.com/amery/allwinner-a10-video ?
<mnemoc> once you have figured out what files are worthy :)
<drachensun> heh sure
<mnemoc> thank you :)
<ManoftheSea> mnemoc, I seem to think you've got extensive arm device experience. How does Linux signal an ARM core to turn off the system?
<ManoftheSea> (Very high level is okay! I'll research from there)
<mnemoc> ManoftheSea: i'm a noob. but i know the kernel asks the PMU to do so
<ManoftheSea> So, it goes through to runlevel 0, everything shutdown/unmounted, then sends some kind of signal?
<ManoftheSea> and in x86, that signal seems to be an ACPI or APM signal. Which are told to the chip by the bios.
<mnemoc> [ 5913.230000] Power down.
<mnemoc> [axp] send power-off command!
<mnemoc> ^--- from A10
<ManoftheSea> So, in device tree, it'd have to identify a signal to cut power.
<ManoftheSea> ok.
* ManoftheSea writes that down.
<mnemoc> in EOMA he will tell the EC to do so
<ManoftheSea> yep. I'm just wondering how flexible that is. Given the set of signals that cross the header.... I'm wondering if you can't have a USB device to do that.
<zub> ManoftheSea: w.r.t. the user-space part, see man 2 reboot - an interesting syscall :)
<ManoftheSea> Assuming the STM32F is wonderful and great and all that: It's doing mouse, keyboard, backlight...
<ManoftheSea> zub: I have to tell my current laptop "reboot=e" at the cmdline.
<ManoftheSea> So, certainly there are many ways to cut powre.
<mnemoc> i assume one will be able to write something in a magic address (register)
<ManoftheSea> mnemoc: register of the ARM?
<mnemoc> EC
<ManoftheSea> Er, yeah, I guess I should call it that.
<ManoftheSea> But EC is so tied to x86
<mnemoc> EC = MCU of the i/o board where the module is connected
<ManoftheSea> MCU?
<mnemoc> micro controller
drachensun has quit [Ping timeout: 260 seconds]
<mnemoc> don't know if that "openec" uses a register (memory mapped) or a protocol over i2c
<ManoftheSea> that's the LPC I was talking about.
<mnemoc> first time I see that 3-letter-acronym
<ManoftheSea> The chip does some EC stuff in hardware, from port 62h and 66h.
<ManoftheSea> That's where I said "Low Pin Count". It's a quad pumped 4 wire interface to replace ISA.
<mnemoc> too low for me :)
<ManoftheSea> too low? ISA being Industry Standard Architecture (?)
<mnemoc> for me ISA is the slot that came before PCI
<mnemoc> but nevermind. I got it.
<ManoftheSea> yeah, that's what I'm saying. It's OLD. '85
<mnemoc> we don't need something fancy to talk with the EC
NAiL__ is now known as NAiL
mSquare has left #arm-netbook [#arm-netbook]
NAiL has quit [Changing host]
NAiL has joined #arm-netbook
<ManoftheSea> ok... there's discussion of "gpio-poweroff" driver, through i2c gpios.
<ManoftheSea> Which fits into device tree.
<mnemoc> afaik lkcl goal is to port openec, not to implement something new
<ManoftheSea> Hmm, I'm assembling info on why that's wrong. I'll make a case later.
<ManoftheSea> Though, gpio-poweroff isn't "new"
<mnemoc> :)
<ManoftheSea> It's from the arm discussion list.
<mnemoc> but it's poweroff-specific
<mnemoc> we need full access to whatever is going on there
<ManoftheSea> OpenEC would make the EOMA card have to emulate x86 to communicate.
<ManoftheSea> Which would be great for an x86 EOMA card.
<mnemoc> i suppose the idea is to implement an open standard already suppoered by the kernel
<ManoftheSea> That's what I'm looking for, yes.
<rm> this can fit
<ManoftheSea> I guess it'd be useful to know more about what a BIOS does.
<ManoftheSea> Because, all I can think of is "apply power to EOMA" or "remove power from EOMA"
<mnemoc> ManoftheSea: look at the specs of openec
<ManoftheSea> doing the battery charging is more like an UPS
<mnemoc> it's very detailed on what info is provided
<ZaEarl> would the axp209 do the job, since lots of a10 systems already use it.
<ManoftheSea> I would think so, ZaEarl. Also, the openec list seems to have died in Jan '11
<ManoftheSea> er, sorry... there are two places for power management.
<ManoftheSea> The EOMA card has an axp209, almost certainly.
<ManoftheSea> It does power from 5V to 3.3, 1.?, as well as USB-OTG management
<ManoftheSea> Also, if there's any kind of battery on the EOMA card side.
<ManoftheSea> But the laptop/tablet/monsterboard(maybe) will also have a battery.
<mnemoc> eoma is not a10 specific
<ManoftheSea> And that I/O-board battery is more like an uninterruptable power supply.
<mnemoc> so axp209 has absolutely nothing to do
<ManoftheSea> mnemoc: thank you for correcting my imprecise language.
<ManoftheSea> An A10-based EOMA card will almost certainly have an axp209
<mnemoc> not necesarily
<ManoftheSea> What part of "almost certainly" are you disagreeing with?
<mnemoc> i would say a "maybe" at most
<ManoftheSea> ok.
<ManoftheSea> I believe the current A10-based EOMA card has an AXP209.
<mnemoc> ah, cool then
<ManoftheSea> That's the Wits-tech model, in my understanding
<mnemoc> good point. now i agree with the almost certainly
<ZaEarl> yes, rhombus tech seems to have an axp209
sspiff has quit [Remote host closed the connection]
Maqs has joined #arm-netbook
<ManoftheSea> ZaEarl: does your zatab have an axp209?
<ZaEarl> yup
<mnemoc> very few a1X devices don't have it
<mnemoc> iirc the mk802 and olimex's "micro" board
<mnemoc> but i don't see a reason for the a10 module to need it
<ManoftheSea> alright. Now that we're past that... most of what I can find about openEC is {battery info, temperature info, keyboard controller}
<ManoftheSea> mnemoc: there's not a reason it needs it. It's just part of the reference model and comes from the same company.
<ManoftheSea> right?
<mnemoc> display backlight and that stuff?
<mnemoc> ManoftheSea: yes
<ManoftheSea> actually, I don't see that in what I'm reading.
<Turl> without an axp you cannot use dvfs
<ManoftheSea> Not that I am certain I'm reading the right place.
gzamboni has quit [Ping timeout: 240 seconds]
<ManoftheSea> and then the datasheet at http://wiki.laptop.org/images/a/ab/KB3700-ds-01.pdf
<ManoftheSea> "EC standard commands as described in ACPI 2.0 spec"
<ManoftheSea> So... OpenEC is to implement ACPI?
<mnemoc> seems so...
<ZaEarl> whenever I hear ACPI I cringe.
<ManoftheSea> And that's mostly for old crap (x86).
<ManoftheSea> I'm thinking the EOMA-compliant device wants to use USB as much as possible.
<ManoftheSea> Because it's discoverable.
<ManoftheSea> So, if kbd and mouse are USB devices, you don't need the kbc. Or at least it's not as promenent a device.
alcides has joined #arm-netbook
alcides has joined #arm-netbook
alcides has quit [Changing host]
<mnemoc> most of the stuff will be usb, sure
<mnemoc> but not everything can
<ManoftheSea> mnemoc: no, but whatever's not usb is i2c or gpio, and specified in device tree.
<ManoftheSea> Making it discoverable.
<ZaEarl> I sure wish Allwinner would announce something about sun6i this week.
<ManoftheSea> Hmm, I'll try looking up that dvi-lvds chip that's specified in the EOMA page.
ZaEarl has quit [Quit: Ex-Chat]
ZaEarl_ has joined #arm-netbook
gzamboni has joined #arm-netbook
<mnemoc> ManoftheSea: btw, are you Gordan Bobic?
<ManoftheSea> no. I thought you were.
<mnemoc> Derek LaHousse
<ManoftheSea> yeah
<mnemoc> :)
<ManoftheSea> what? did you /who me?
<mnemoc> yes
<mnemoc> should have done it before asking :p
<ManoftheSea> oh, it's not dvi->lvds, it's RGB->lvds
<ManoftheSea> so yeah, the SN75LVDS83B doesn't do backlight.
phh has quit [Quit: No Ping reply in 180 seconds.]
phh has joined #arm-netbook
<ManoftheSea> Did I sound particularly knowledgeable?
<mnemoc> here? sure, you know far more than I about hw
pawel5870 has quit [Remote host closed the connection]
techn has joined #arm-netbook
vinifr has joined #arm-netbook
pawel5870 has joined #arm-netbook
ZaEarl_ has quit [Ping timeout: 244 seconds]
popolon has quit [Quit: Quitte]
gimli has joined #arm-netbook
rellla has quit [Quit: rellla]
tzafrir has quit [Ping timeout: 240 seconds]
Quarx has quit []
<lkcl> ManoftheSea: yeah OpenEC is far more involved than it seems. it's 10,000 lines of c-code. keyboard matrix. mouse driver. battery charging (PWM and level reading). the works. it's... complex.
<lkcl> ManoftheSea: TFP410 and TFP401. they're DVI-to-RGB/TTL (and vice-versa)
<ManoftheSea> lkcl: I take it you've been reading back?
<lkcl> ManoftheSea: que? :) oh - yes :)
<ManoftheSea> What's the DVI comment about?
<lkcl> "ManoftheSea> Hmm, I'll try looking up that dvi-lvds chip that's specified in the EOMA page."
<lkcl> TFP410 for dvi->rgb/ttl
<ManoftheSea> I was discussing going from EOMA RGB to the panel LVDS. I mistakenly thought the EOMA was DVI
<ManoftheSea> I see.
<lkcl> TFP401 rgb/ttl -> DVI
<ManoftheSea> No, I actually meant the SN75LVDS83B
<lkcl> ah then you want SN75LVDS83B or equivalent.
<ManoftheSea> Which you've specified EOMA must be compatible with.
<lkcl> that's only for panels up to (about) 1440x900
<ManoftheSea> I thought it provided up to 5 LVDS channels.
<lkcl> well, it's the easiest way to put it, that the signals must be 5V tolerant etc. etc.
<lkcl> SN75LVDS83B is only single-channel LVDS
<lkcl> the reason for specifying 24-pin RGB/TTL is because you can do *anything* with it, not costing a fortune in ICs.
<ManoftheSea> oh, is it 5 line drivers in a single LVDS channel?
<lkcl> sounds about right.
<ManoftheSea> So, there's another chip to do higher screens?
<lkcl> a single LVDS channel involves 3 colours and i think one clock line pair.
<lkcl> yes, a different IC is needed - one that does dual LVDS.
<lkcl> i don't know one: if you find one that's popular please do let me know, i need to document it.
<ManoftheSea> hmm... The SN75LVDS83B shows 4 7-bit shift registers, and a clock. Rather than 3 8-bit colors.
<ManoftheSea> If I find something, I'll shout.
<ManoftheSea> But I did want to ask a bit more about OpenEC, and whether it's necessary.
<lkcl> star.
<lkcl> it is.
<lkcl> cost of a battery management IC: $1.50
<lkcl> cost of a 0.5 watt audio driver IC: $1.50
<lkcl> cost of a USB-based keyboard/mouse driver IC: $1.50
<ManoftheSea> No no, not just doing battery management on the I/O board with STM
<ManoftheSea> But specifically the EC protocol
<ManoftheSea> Which seems to pull in ACPI
<lkcl> EC "protocol"?
<ManoftheSea> And ancient x86 cruft
<lkcl> yeah all that would go.
<ManoftheSea> go... out? go in?
<lkcl> out.
rellla has joined #arm-netbook
<ManoftheSea> page 26
<lkcl> it's not like ARM systems support ACPI
<ManoftheSea> EC stuff in firmware, doing ACPI 2.0
<mnemoc> doesn't uefi brings acpi into arm?
<ManoftheSea> ick at that too.
<mnemoc> feel inevitable...
<ManoftheSea> So, from the linked stuff around OpenEC and XO, it looks like EC is a term that means a lot of x86 cruft.
<ManoftheSea> Given that chip is doing it in hardware.
<ManoftheSea> From an "LPC" bus.
<lkcl> ManoftheSea: KB3700-ds-01: not interesting. as in, yes sure the STM32F would do the same job but otherwise ...
<lkcl> ah no, it's...
<ManoftheSea> It may be an overloaded term.
<lkcl> ok, do you know how battery management algorithms work?
<ManoftheSea> can't say I do.
<lkcl> you can't just "apply power"
<lkcl> you have to use PWM
<lkcl> and it depends on what the state of the battery is (how charged / depleted it is)
<lkcl> so you start off if it's very very low doing 100% charge
<lkcl> then as the state rises you reduce it to say... 50%
<ManoftheSea> lkcl, let me ask you this: Would it be more apt to compare the shell to a Uninterruptable Power Supply.
<lkcl> then, when it approaches 80% or above, you start to "pause" for several seconds
<ManoftheSea> Those have batteries, but provide a specific output.
<ManoftheSea> Also, they have USB interfaces, which are discoverable
<lkcl> no it wouldn't - not unless you were looking to do a UPS *using* the STM32F
<lkcl> ManoftheSea: you're really confusing me
<ManoftheSea> yeah, I've been rather confused
<ManoftheSea> I'll claim that as my problem.
<lkcl> i'm endeavouring to explain what tasks a $1.50 STM32F needs to do when used to replace 3 or even 4 other discrete $1 to $1.50 components
<ManoftheSea> And I agree that it seems quite capable of replacing them.
<ManoftheSea> But I don't quite see how it communicates with the EOMA card.
<lkcl> normally those tasks would be covered by e.g. the dedicated periperals of e.g. the A10
<lkcl> over the USB 1.1 bus
<lkcl> an STM32F has USB 1.1
<mnemoc> nice
<ManoftheSea> So, the STM32F has to provide a USB device that emulates the LPC?
<ManoftheSea> Why doesn't it just present itself as an UPS?
<lkcl> so what you would do e.g. in the audio driver case would be to write a linux driver that recognised the STM32
<ManoftheSea> And a keyboard, and a mouse, and a and a and a
<lkcl> because you have to multiplex quite a bit of data onto that... yeeees
<lkcl> exactly
<lkcl> so it will be necessary to write a communications protocol, splitting out the data
<lkcl> quite a challenging and interesting task.
<lkcl> hence the need for a little bit more than libopenstm32, and why something like RTEMS is needed
<ManoftheSea> I agree that far. But I don't believe EC is the right term any more.
<mnemoc> *cough* micro-eng-board-with-stm32f *cough*
<lkcl> yeah all right mnemoc :)
<ManoftheSea> mnemoc: at that point, it's the mini
<lkcl> it's the closest term that the industry will recognise.
<mnemoc> ManoftheSea: we don't need lvds or wifi
<lkcl> yes. do really want someone to come forward and take the leaflab's maple schematics and add the extra connectors
<ManoftheSea> I must bow to your greater experience, but from the research I did, they'll recognize it as "x86 stuff"
<ManoftheSea> I haven't heard from Prof Pierce in 2 weeks either.
<ManoftheSea> er, Pearce.
<mnemoc> uefi on arm is real, and will become the standard regardless who likes it or not :|
raver2046 has joined #arm-netbook
<ManoftheSea> mnemoc: the standard for people who want Windows 8.
<mnemoc> and that means every major manufacturer and most mid/small ones
<mnemoc> then the rest will fall
<mnemoc> as usual
<ManoftheSea> But since Win8 UEFI devices are completely locked out... they're uninteresting.
<lkcl> ManoftheSea: ah hmm... well... that may be an advantage. maybe not. we'll see.
<xxiao> M$ die die die
<mnemoc> uefi doesn't need to lock linux out. that's a contract thing between MS and manufacturers
<lkcl> ManoftheSea: ah? hm. wonder where he's got to?
<ManoftheSea> yes. In the x86 world, it says "consumers must be able to run their own code." In the ARM world, it says "FOAD".
<mnemoc> unlocked devices will come with uefi too
vinifr has quit [Ping timeout: 245 seconds]
<lkcl> xxiao: peace! we want to take M$'s money! :) we would *love* them to make an EOMA-68 CPU Card, because the patent licensing will help fund their replacement :)
<xxiao> hopefully
<xxiao> but that sounds more of a intel thing
<jelly-home> I'd also love a board to plug said card into!
<ManoftheSea> oh hey... there's a USB audio streaming demo for the STL
<ManoftheSea> STM
rellla has quit [Quit: rellla]
<ManoftheSea> It needs an external audio DAC, though...
<xxiao> anyone benchmarked A10's SATA port?
<lkcl> ManoftheSea: yeah. well, it'd work for low-ish quality audio. i don't imagine it'll require that many external discrete components
<xxiao> i was told by a guy that Marvell's kirkwood has issues on sata etc but its price is irresitable
<ManoftheSea> mnemoc: oh cool. The last message on the latter list was Jan '11
<ManoftheSea> brb
<lkcl> mnemoc: great
<mnemoc> the announcement
tzafrir has joined #arm-netbook
raver2046 has quit [Ping timeout: 260 seconds]
<mnemoc> using eluaproject.net would be fun :p
<mnemoc> lkcl: once the stm32 is in the board, how do we flash it? special pins?
<mnemoc> or code lives in the same eeprom as the DT description?
<ManoftheSea> mnemoc: looks like it uses JTAG.
<mnemoc> so that's yet another thing to add to the board
<ManoftheSea> it's a header, is all.
<mnemoc> yes. but has to be considered
<mnemoc> lkcl: ---^
<mnemoc> eoma68-a10 board + expansion header + eng-board + jtag header = eoma68 devkit
<ManoftheSea> what's the "expansion header" here?
<mnemoc> in the card
<ManoftheSea> oh, for no-sleeve cards?
<mnemoc> i think that will be the case until there are no devices
<mnemoc> devkit-ish
<mnemoc> not that we need another A10 dev board after the CB and A10-olinuxino
<mnemoc> the most critical part is having a module and having an stm32 with jtag in the first dev i/o^W^Weng board
<lkcl> mnemoc: ta
<mnemoc> .oO(ta?)o
<lkcl> ok. mnemoc could you update the elinux.org pages to note that a jtag for STM32F is needed?
<lkcl> i gotta run
<lkcl> yeah. 8. gotta go
<mnemoc> lkcl: need to convince you first :p
<ManoftheSea> we all agree, the STM32F needs a JTAG header.
<mnemoc> lkcl: also, there is no micro-eng page
<mnemoc> cu lkcl
<ManoftheSea> mnemoc: the microboard is at the bottom of the miniboard page
<mnemoc> thanks
tuliom has quit [Ping timeout: 245 seconds]
rsalveti has quit [Ping timeout: 245 seconds]
rellla has joined #arm-netbook
tuliom has joined #arm-netbook
eFfeM has joined #arm-netbook
rellla has quit [Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org]
raver2046 has joined #arm-netbook
pawel5870 has quit [Ping timeout: 272 seconds]
xxiao has quit [Read error: Operation timed out]
xxiao has joined #arm-netbook
gimli has quit [Remote host closed the connection]
tuliom has quit [Ping timeout: 245 seconds]
rsalveti has joined #arm-netbook
raver2046 has quit []
<ManoftheSea> So, I've got a problem...
<ManoftheSea> TI's documents say 1 channel of LVDS (18-24 bit color) go up to 1400 x 1050
<ManoftheSea> While it takes two LVDS channels (6-8 LVDS pairs) to reach 1920 x 1200.
<ManoftheSea> So... EOMA-68 is limited to 1400x1050 on the internal connector? BOO!
<ManoftheSea> also, lkcl, DS90C387A is TI's dual-LVDS chip.
specing_ is now known as specing
popolon has joined #arm-netbook
<ManoftheSea> oh, the DS chips are 1.8 V input, while the SN have 3.3
<ManoftheSea> Mah bad.
<ManoftheSea> er... nope. The chip I menioned, ...387, is a 3.3 input and 5V tolerant
<zub> hmmm... if I want to use the OTG USB as a host - ti shoudl work automagically, right? that's the whole point of OTG, is it not?
<ManoftheSea> no, the driver has to do something.
<zub> ehm?
<zub> I don't care what the driver does, I was talking from the user point of view... you plug a device in, and it's recognized, and works
<ManoftheSea> oh, yeah.
<zub> my trouble is: I connect something, lsusb doesn't show it... so whatever needs to happen doesn't happen
<zub> can I force the port to host mode somehow?
<lkcl> ManoftheSea: no! EOMA-68 is *not* limited to 1400x1050! all you do is find (as you did) a dual-channel RGB/TTL-to-LVDS IC!
<lkcl> honestly :)
<ManoftheSea> lkcl: but... there are only 24 parallel color lines.
<Turl> zub: you need an OTG cable usually
<lkcl> ManoftheSea: so you connect all 24 of them to the Dual-channel LVDS IC
<lkcl> the Dual-channel LVDS IC takes 24-pin parallel colour lines as input
<ManoftheSea> the dual-channel lvds IC has 48 inputs...
<ManoftheSea> well, 51.
<lkcl> ah then you've found the wrong IC.
<ManoftheSea> but red1 is 8, and red2 is 8
<lkcl> let me take a look at the datasheet ok?
<Turl> and some devices need 5v feed to them on some pin to switch modes
<zub> Turl: there was some adapter with micro (or whatever) USB on one and, and USB A female on the other, I assumed that is an OTG adapter
<zub> can I just force the port to host? I tried to change the kernel config, but selecting "HOST", it does not compile :(
<ManoftheSea> with 8 "serial data transmitter channels", the DS90C387A looks good...
<ManoftheSea> but I think that inputs is lying.
eFfeM has quit [Quit: Leaving.]
<ManoftheSea> or rather than "it doesn't exist"... I did not find the chip you're looking for.
<ManoftheSea> away
bsdfox has joined #arm-netbook
<lkcl> ManoftheSea: page 2 says that it supports "option single pixel transmitter imputs"
<rm> inexplicably I have 399 MB free instead of 315, and this is on a mali-enabled kernel
<mnemoc> VE disabled?
<mnemoc> VE takes 80
<rm> CedarX?
<mnemoc> yes
<Turl> rm check dmesg
<lkcl> p16 explains that you can set the "DUAL" pin to 1/2 voltage (1.65V) and the chip will go into "single input, dual LVDS output" mode.
<Turl> all the allocations are printed there
<rm> # CONFIG_VIDEO_SUN4I_CEDAR is not set
<rm> right!
<rm> currenlty it does not work anyway in X11, correct?
<rm> tl
<mnemoc> it might work on armel/armv4 systems
<mnemoc> but not on armhf
<rm> I see
<Turl> armv4? o.O
<mnemoc> armel is armv4
<Turl> I thought it was armv5
* mnemoc doubts
<jelly-home> Debian's armel is armv5t
<jelly-home> if you're talking about that ABI
<jelly-home> if you had a v4 you'd run the older "arm", now nonexistent in debian
<mnemoc> yes, libs ABI
<Turl> https://es.wikipedia.org/wiki/Arquitectura_ARM#Familias awesome table for us spanish speakers :)
<jelly-home> mnemoc: dunno, armel chroot completely failed to work last time I tried it on a v4
n6pfk has quit [Remote host closed the connection]
<mnemoc> jelly-home: thumb is required
<mnemoc> jelly-home: that might be the reason
<mnemoc> not all v4 had thumb
<jelly-home> didn't even know there were v4t until I read that wiki page
<mnemoc> :)
Gujs has quit [Quit: Gujs needs some sleep...]
<mnemoc> Turl: armel/armhf are debian platform names, including tons of chosen optimization flags, not exactly bound to an arm family or feature
<mnemoc> armhf for example doesn't only mean hardfloat, also v7, neon and thumb2 iirc
<Turl> mnemoc: I know
<mnemoc> Turl: ah, ok. i got confused by the paste of the link
Gujs has joined #arm-netbook
rsalveti has quit [Ping timeout: 255 seconds]
rsalveti has joined #arm-netbook
<TomNL> does anyone know if cnx hwpack work for mele?
<TomNL> i have one from 19 september which works..but the ones from october dont even get to start u-boot it seems
<mnemoc> problem is no one updates the hwpack generation code when the kernel or u-boot get changed
<TomNL> ah ok, and it's changed?
<mnemoc> yes
<mnemoc> u-boot has changed dramatically during the last weeks
<TomNL> ah ok
<TomNL> hehe
<TomNL> looking for different bin files now?
<mnemoc> it does per-board dram and pmu initialization
<mnemoc> in the case of the mele it means it has to be built using the mele_a1000 target instead of sun4i
<TomNL> ah ok :) i will check the hw-pack create script and see if i can change it
<mnemoc> it's on github
<mnemoc> unfortunatelly it supports boards from which we don't have reliable dram initialization info yet
<mnemoc> but for those which worked with the generic 'sun4i' using the mele's will probably work fine
<mnemoc> and for 'sun5i' the a13_olinuxino's will do too
<mnemoc> but those that required patching stuff need to be rethought
ZaEarl_ has joined #arm-netbook
popolon has quit [Quit: Quitte]
<ManoftheSea> lkcl: but that's just pixel doubling.