#qi-hardware IRC log for Sunday, 2011-06-26

[g2]wpwrak, how did you add the AT86RF231 in kicad ? I can see it on your schematic from git, but not in my own library.00:46
wpwrak[g2]: it's in components/at86rf230.lib00:52
wpwrak(project ben-wpan)00:52
[g2]ah!!! sweet, thx00:54
wpwrakyou probably also want project kicad-libs, for footprints00:55
wpwrakthen, you want fped: http://svn.openmoko.org/trunk/eda/fped/README00:56
wpwraksome distros may have it packaged already, but better build from sources so that you get the latest version00:56
wpwrakfinally, project eda-tools (qi-hw again) may be useful. e.g., when you're in ben-wpan, do "make dsv"00:57
wpwrakthen put eda-tools/dsv/dsv into your PATH00:58
wpwraknow you can invoke the at86rf231 data sheet with  dsv txrx   (while you're under ben-wpan)00:58
[g2]I haven't gotten that far yet.  I've just built the latest kicad from source and I'm running that.01:00
[g2]I now see how the components are added in via the "Component library files"01:01
[g2]I'm warming up to eschema, but it'll take a little time01:01
[g2]I'm going to try and have an ATmega328 TQFN layout done for Monday and then the AT86RF231 daughterboard for following Monday01:03
wolfspraulgood morning everybody01:03
[g2]wolfspraul, morning :)01:03
wpwrak[g2]: the libs are listed in the *.pro file. i usually just add them there. less hassle than the gui :)01:04
[g2]wpwrak, nod.01:05
wpwrak[g2]: in general, it's probably best if you try to imitate the structure of the ben-wpan project. that way, you'll find many useful items that are already in place. including makefiles ;-)01:06
[g2]wpwrak, I'll have to work my way up to that.01:07
wpwraksure. it wasn't built in a day either :)01:07
[g2]right, I'm jumping in after like 4 years01:08
kristianpaulwolfspraul: buenos dias :)01:11
wolfspraulwpwrak: just reading backlog. you definitely 100% don't have to worry about shipping prototypes with leaded solder and running into any rohs issue.01:17
wolfspraulthat takes the entire rohs concept and implementation completely out of proportion01:17
wpwrakwolfspraul: heh ;-) i have a somewhat pessimistic bias with customs from hell right next to me01:20
wolfspraulsure, but I wanted to throw in my experience01:20
wolfspraulso we can focus on real issues not perceived ones01:21
wolfspraulfirst of all I am 99.999% sure that no customs in the world will have lead testing strips and do random checks on anything01:21
wolfspraulthat's because rohs is far more than just 'lead', and it's becoming more all the time (which is good)01:22
wolfspraulrohs is an evolving standard, and big success01:22
wolfspraulthe way it's enforced is mostly via trade or industry groups01:22
wolfspraulthey are 'monitoring' the market01:22
wolfspraulsometimes they act passively, i.e. companies can ask them for advice regarding their products01:22
wolfspraulbut they can also start to become active, going after repeat 'offenders'01:23
wolfspraulmost likely that would be some big Asian manufacturer/importer who just thinks he can make a little extra margin by importing large quantities of non-rohs compliant stuff into the EU (for example)01:23
wpwrakyeah, wouldn't have to worry about repeat offending :)01:23
wolfspraulafaik rohs still doesn't apply in the US01:23
wolfspraulsouth korea, japan, and now even china are following with similar ideas/legislation though01:24
wolfspraulso let's say we have such a large Asian manufacturer/vendor, and they import large quantities. Maybe a competitor notices, and alerts the industry/trade group watching this. They would first try to work with the manufacturer/importer - if that doesn't go anywhere they would start to involve customs.01:24
wolfspraulthe focus of everybody is to _improve_ the quality of products and make them rohs compliant, not to confiscate and discard stuff at customs01:25
wolfspraulin all this, a 1-man operation in Buenos Aires, with a home lab, soldering together a few prototypes, it all really completely will never show up on the radar01:26
wolfspraulit's not what rohs was meant to address, and it's not how it works and why it became successful01:26
wolfspraulhaving said that, of course it's a nice gesture from you that you want to help the world by spreading a few grams of lead less into the environment :-)01:27
wolfspraulthere are still some Chinese manufacturers who are learning what that stuff really is. if they finally get it, that will save tons of lead every month. that's the focus...01:28
wolfspraulif China would start to enforce rohs or similar legislation in their domestic market, that would be a huge win01:29
wolfspraulof course the US could work a little harder too :-)01:30
wpwrakwell, who said that customs behave based on the intent of the rules they're supposed to enforce ? :)01:30
wpwrakat least around here, that's often not the case. instead, the rules give them an opportunity to stop things. and then let you make it worth their effort to get them released again.01:31
wolfspraulin my experience, customs will not want to have anything to do with rohs01:32
wolfspraulit's too complicated01:32
wolfspraulplease keep the big picture in mind. rohs is largely a eu thing.01:32
wolfspraulsouth korea, japan have adopted similar ideas01:32
wpwrakhere, they told me that even something declared as "PCB" could be an issue. because "PCB" also means a prohibited chemical.01:33
wolfspraulsome others are 'debating' or trying a little here and there, like California has some rules that apply to light bulbs now (according to Wikipedia)01:33
wolfspraulah definitely01:33
wolfspraulthat's another story01:33
wolfspraulyou have to be careful with labling01:33
wpwrakyeah ;)01:33
wolfspraulbecause they will lookup acronyms in their books...01:33
wpwrakand on comes the flashing red light ...01:34
wolfspraulyes but how does that relate to you using leaded or lead-free solder when making your 3 prototypes?01:34
wolfspraulI tried to put rohs fear in perspective.01:34
wolfspraulnothing to fear about rohs. rohs is a good thing, and will evolve further.01:34
wolfsprauldoesn't apply to home-made anything01:35
wolfspraulwhen you go into volume production, and make a retail product, and want to sell in the EU or some other countries, then yes, rohs is one thing on a longer list to take care of (like CE, FCC, WEEE, etc)01:36
wolfspraulthat's all about rohs01:36
wpwrakactually, i think it may now apply to pretty much anything. they recently dropped a number of exceptions.01:36
wpwrakwhether they enforce it for tiny quantities is a different story, though01:36
wolfspraulpossible, but trust me, hand-built prototypes that you are sending from Buenos Aires to Berlin are _not_ meant01:36
wpwrakanyway, US, 2 x EU, and TW worked without any trouble, plus 3 x AR outbound. that's reassuring.01:37
wolfspraulI won't even try to sift through thousands of pages to find the exception that would apply to that case, if anybody would even want to take a closer look.01:37
wolfspraulgood :-)01:37
wolfspraulI'm not surprised.01:37
wolfspraulwith labling I totally agree with you - be careful01:37
wolfspraulthat's how customs indeed does work01:38
wolfspraulthey won't turn on your product, or even do chemical tests. but they will go through all documentation, and look around the product for labels or any text that gives them reason for action.01:38
wpwrakyes, exactly01:38
wpwrakso the question in this case is whether they're looking for something that says anything about RoHS compliance or not01:39
wpwrakapparently, they don't01:39
wolfspraulrohs is too difficult/technical01:39
wpwrakit could just be the logo :)01:40
wolfspraulit is implemented by industry groups, working together with government agencies. and starting with high volume.01:40
wolfsprauland it only applies to the EU and some other smaller countries01:40
wolfspraulnot the US, not China, not many other regions and countries01:41
wpwrakwell, china have their own variant01:41
wolfspraulit's amazing what a pull force the EU market can create, I think by now 50% or more of Chinese manufacturers know what rohs is and a lot of them, especially the better ones, have changed all their production to be rohs compliant01:41
wolfspraulnot yet, just being discussed I think01:41
wolfspraulI think after you made the initial investment, into your people to know what rohs is, the different process etc., the production costs must be very very close to non-rohs production costs.01:42
wpwrak(eu pull) yeah. it's too good a market to just give up. plus, running lead-free and leaded in parallel would be crazily expensive01:42
wolfspraulwhat I see a lot with Chinese vendors is that if it's a small vendor, who has never done rohs runs, or barely heard about it, then they will quote you some large extra charge for a rohs run01:43
wolfspraulbasically you are financing his switch :-)01:43
wolfspraulbut then there are many vendors now, especially larger ones but also increasingly smaller ones, they will just tell you "all our runs are rohs compliant now"01:43
wolfsprauleven whatever they ship to US, China, South-East asia, etc.01:43
wolfspraulthat's quite amazing I think01:43
wolfspraulactual rohs countries maybe only 500 million people or so01:44
ds3rohs sucks01:44
wolfspraulbut it's cheaper for the manufacturer to switch everything to rohs compliance and never hear about this anymore :-)01:44
wpwrakpeople with money. good euros, not fragile dollars. well, maybe with the euro weakening, they wouldn't switch now ;-)01:44
wolfspraulds3: why? tell us more :-)01:45
wolfspraulI think it's great. I want to understand more about the chemical composition of the stuff I'm making anyway. talking about rohs is a way to engage vendors in that kind of conversation...01:53
wolfspraulwpwrak: I read the backlog, but still cannot really make sense out of what [g2] is building01:57
wolfspraulhe is building some 802.15.4 product?01:58
wpwrakhe wants to combine a small arduino with the at86rf231, yes01:58
wolfspraulso he's not building an at86rf231 arduino shield, but an integrated board out of arduino+at86rf231?02:02
wpwrakactually, he now seems to want to make two boards. so the final arrangement may be something shield-ish.02:03
kristianpaulwpwrak: http://downloads.qi-hardware.com/people/kristianpaul/tmp/gps_clk_from_mm1.png02:42
kristianpaulwpwrak: is not for overrun i just tried the clk from gps, but this time coming from the fpga02:42
kristianpaulerggg no wait02:44
kristianpaulsmall misconfiguration with scope it seems ;) 02:45
wpwrakso 8 MHz is good ?02:45
kristianpauli just wondered with the first pic, now fixed :-)02:47
wpwrakthe scope indicates quite a bit of jitter. you may want to move the trigger ~125 us to the left, then look at the curve again02:48
wpwrak(first pic) probe not grounded ? ;-)02:48
kristianpaulno no !!02:48
kristianpaulprobe is grounded02:49
wpwrakthen you have a lot of overshoot :)02:49
wpwrakwell, could be just bad ground. (loop too long)02:49
kristianpaulfrom gnd in J21, whatever is grounds to..02:49
wpwrakactually, the jitter is probably okay. that may be just trigger inaccuracy. all the rising edges are narrow, so they're in sync02:51
wpwrakis J21 close to the signal ?02:51
kristianpauldont understand02:51
kristianpaulJ21 is the I/O expasion connector in mm102:51
kristianpaulthis debug I/O i added and the other gps related pins go there02:52
wpwrakand where is the clock signal you're measuring ? also on J21 ?02:52
wpwrakah, okay02:52
wpwrakcan the scope display th trigger frequency ?02:53
wpwraki think the frequency you're showing is a waveform measurement, correct ?02:54
kristianpaul(waveform) yes02:54
kristianpaul(trigger frequency) dunno02:54
wpwrakokay, then it's normal for it to be relatively inaccurate. if you can get the trigger frequency, there may be a much more precise indication.02:55
kristianpaulfor now i'll activa the OUT ping for overrrun signal02:55
qi-bot[commit] Werner Almesberger: tools/Makefile (BEN_DIRS): added "dirtpan", so it's now included by default (master) http://qi-hw.com/p/ben-wpan/417d12c03:06
qi-bot[commit] Werner Almesberger: tools/dirtpan/: added statistics collection, SIGUSR1 to dump, SIGUSR2 to reset (master) http://qi-hw.com/p/ben-wpan/0194dc103:06
kristianpaulwpwrak: does this look sane for you http://paste.debian.net/121012/ ?03:08
wpwrakno :)03:14
kristianpauli know i missed something but not sure wich part..03:15
wpwrak30*2048/2 = ~30 MB03:15
rjeffriesfyi all US manufacturers are observinf ROHS standard. not sure if it is law, butsince any product may ship to EU i is effectiely required. 03:15
wpwrak30*1024 = ~30 MB03:15
wpwrak30*4*256 = ~30 MB03:15
wpwrakso all buffers should be the same03:16
wpwrakor did you mean MSa ? :)03:16
wpwrak(and Mb would be Megabits. careful, if you buy RAM chips, they always give the size in Mb, not in MB ;-)03:17
wpwrakhm .. 4.475 us per bit. 4 us are for the air interface. that's 475 ns for transfer to/from the transceiver, meaning about 237 us per side ... about 4.2 Mbps. well, that may even be correct. having some evilly long inter-byte delays.03:19
wpwrak(inter-byte delays) required by the at86rf230. the ben could go a little faster03:20
rjeffrieswpwrak the over the air data rate is expected to be...maybe 250 Kbps correct? goodput woud be less03:55
kristianpaulwpwrak: i like this of debugging with scope, is more fun than the classical led ;)04:21
kristianpauli wich as usefull verbose as a printf :-)04:22
kristianpaulsecond i'll upload overflow OUT in scope04:23
kristianpaulwpwrak: http://downloads.qi-hardware.com/people/kristianpaul/tmp/debug_overflow_pin_and_counter_clock.png04:25
kristianpaulfor CH1 wich is overflow flag_pin scope is not decied wich freq it have, still chaing from 386 to 43804:26
kristianpaulCH2 is okay 1.024 is half the sampling rate as the shifter output is a byte04:28
kristianpaulahh wait04:30
kristianpaulmy fault04:30
kristianpaulCH1 now is ~200Khz after tune trigger a bit more again04:34
kristianpauloh !05:13
kristianpaulthe overrun clock signal is gone..05:13
kristianpaulwhat i did?05:15
kristianpauljust run some comands to dump some data from the gps acquistion buffer05:15
kristianpauldid i forgot to enable the counter?... could be05:15
kristianpaulhum yes05:16
kristianpaulhe :)05:16
kristianpaulha i foudn him ! http://www.gmat.unsw.edu.au/snap/staff/peter_mumford.htm05:29
kristianpaulwolfspraul: this is namuru  correlator author05:29
wolfspraulnice, looks like he's still active, even bette05:35
kristianpaulhttp://www.gpsworld.com/tech-talk-blog/the-difference-between-software-radio-software-defined-radio-and-software-receivers-1 nice article05:43
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-06252011-0203/05:47
kristianpaulwhere hell is gone the overun signal...06:24
kristianpaulhumm i'll start saving bitstream before synthesize new ones..06:28
kristianpaulwpwrak: http://downloads.qi-hardware.com/people/kristianpaul/tmp/debug_overflow_pin_and_counter_clock_shot3.png overflow pin, wich right value 7ff just before counter start over :)07:09
qi-bot[commit] David Kühling: new package qemu-host: Qemu user-space emulator for host toolchain (master) http://qi-hw.com/p/openwrt-packages/1d560af08:21
wpwrakkristianpaul: so .. the overflow happens at the correct time ? if yes, how do you know ?08:47
wpwrakkristianpaul: something i see in the waveforms is that the falling edges are somewhat slow, about 10-15 ns, while the rising edges are much faster. you may want to check your pin configuration. for debugging signals, that's okay, but in case you're driving any "fast" digital logic, that could be a problem.08:49
wpwrakkristianpaul: maybe check with lekernel. he'll know instantly whether these signals look unusual or not.08:49
wpwrakkristianpaul: also, you may want to disable the bandwidth limiter and use peak detect. that way, you can see digital glitches.08:50
wpwrakkristianpaul: last but not least, you may want to build a resistive probe some day. that'll help you get rid of the reflections :) (the waveforms look very clean. did you use averaging ?)08:51
qi-bot[commit] David Kühling: emacs: enable second-stage "compilation" (i.e. dumping) via qemu (master) http://qi-hw.com/p/openwrt-packages/ed5db4208:53
qi-bot[commit] David Kühling: emacs: make volume keys work like PageUp/PageDown (master) http://qi-hw.com/p/openwrt-packages/bcb7b6408:53
wpwrakdvdk: whee ! another epic struggle comes to an end :)09:11
wpwrakdvdk: the emacs build process must be quite something. actually, does perl still dump/undump itself too ? i vaguely recall that it did, some eons ago09:11
vladkorotnevHello Everyone09:39
dvdkwpwrak: yeah, the emacs build process now involves a few more million LOC building qemu09:40
dvdkwpwrak: i guess the term "nightly build" cannot be really applied if builds take > 1d nowadays :)09:41
wpwrakdvdk: nice :) the should add some code-generation feature that makes each step grow the others a little. that way, you could reach a point where an emacs build outruns moore's law :)09:42
wpwrakdvdk: how well do incremental builds work ?09:42
dvdkwpwrak: that.s easy, just add more layers of qemu emulation09:42
dvdkwpwrak: i'm using 'ccache' here (openwrt has an option to enable that).  this way build is a little incremental, but still 100% reproducable09:43
dvdkother than that, in theory openwrt tracks dependencies and only rebuilds packages that change.  but bad idea to rely on that on a build server09:43
dvdka few (dependency) bugs only crop up if you build from scratch09:43
dvdkbtw about the nightly builds: that's the firmware compilation time, not qemu+emacs (which only contribute about 10 min. i'd guess)09:44
wpwrakhmm, openwrt builds its own toolchain. how do you handle gcc distribution in a ccache environment ? doesn't ccache resent mixing gcc versions ?09:45
wpwrakerr ... sorry, confused09:45
wpwrakno gcc distribution. just mixing gcc versions09:45
wpwrakdistribution would be distcc ;-)09:45
wpwrakdvdk: (nightly) you mean the whole openwrt rootfs+optional packages set ? "firmware" is a bit confusing ...09:46
dvdkwpwrak: yup09:53
qi-bot[commit] Werner Almesberger: atusb/fw/Makefile: introduce target-specific compliation variants (master) http://qi-hw.com/p/ben-wpan/6909fc210:01
qi-bot[commit] Werner Almesberger: atusb/fw/: remove unused items when building the USB driver for the boot loader (master) http://qi-hw.com/p/ben-wpan/8f9498410:01
dvdkwpwrak (about ccache): i think ccache checks command line + date and file size of gcc executable to test whether a file from the cache matches10:02
dvdkin cross-build it's run as a prefix 'ccache mipsel-openwrt-linux-gcc' so it knows you're using a different compiler10:03
wpwrakdvdk: okay, that should be pretty robust. hmm, there was something that still broke it. i once ran into it ... when i reported it, they basically threw up their hands and concluded that they couldn't test for *that* too. hmm, what was it ... it was pretty obscure10:04
wpwrakgrr. don't remember some ELF header that got subtly tweaked, i think10:06
qi-bot[commit] Werner Almesberger: spi_atben: SPI host specialized for atben (ben-wpan) http://qi-hw.com/p/qi-kernel/31ff13d10:25
qi-bot[commit] Werner Almesberger: qi_lb60: changed board setup from atben with spi_gpio to spi_atben (ben-wpan) http://qi-hw.com/p/qi-kernel/36baa2b10:25
qi-bot[commit] Werner Almesberger: at86rf230: assorted fixes (ben-wpan) http://qi-hw.com/p/qi-kernel/f0c38be10:25
qi-bot[commit] Werner Almesberger: Revert "at86rf230: initialize unused buffers in struct spi_transfer to NULL" (ben-wpan) http://qi-hw.com/p/qi-kernel/01c171f10:25
qi-bot[commit] Werner Almesberger: at86rf230: check PHR of inbound packets and make sure frame fits into skb (ben-wpan) http://qi-hw.com/p/qi-kernel/fe7850b10:25
qi-bot[commit] Werner Almesberger: spi_atben: added section titles (ben-wpan) http://qi-hw.com/p/qi-kernel/4926b1010:25
qi-bot[commit] Werner Almesberger: spi_atben: moved atben_reset and at86rf230_platform_data from board to driver (ben-wpan) http://qi-hw.com/p/qi-kernel/9a7d79a10:25
qi-bot[commit] Werner Almesberger: spi_atben: moved qi_lb60_atben platform device from board to spi_atben (ben-wpan) http://qi-hw.com/p/qi-kernel/00fded810:25
qi-bot[commit] Werner Almesberger: spi_atben: moved spi_board_info of atben from board to spi_atben (ben-wpan) http://qi-hw.com/p/qi-kernel/b9def2210:25
qi-bot[commit] Werner Almesberger: spi_atben: allocate SPIP bus_num dynamically (ben-wpan) http://qi-hw.com/p/qi-kernel/e08d15610:25
qi-bot[commit] Werner Almesberger: spi_atben: some cleanup (ben-wpan) http://qi-hw.com/p/qi-kernel/3e7de6010:25
qi-bot[commit] Werner Almesberger: spi_atben: announce atben_reset and don't modify global variables (ben-wpan) http://qi-hw.com/p/qi-kernel/add45cd10:25
qi-bot[commit] Werner Almesberger: spi_atben: we set prv->board_info.platform_data later, don't mis-initialize (ben-wpan) http://qi-hw.com/p/qi-kernel/8a07eed10:25
qi-bot[commit] Werner Almesberger: spi_atben: added optimized unidirectional SPI bitbangers (ben-wpan) http://qi-hw.com/p/qi-kernel/1fe9fab10:25
qi-bot[commit] Werner Almesberger: spi_atben: more minor cleanup (ben-wpan) http://qi-hw.com/p/qi-kernel/db9015210:25
qi-bot[commit] Werner Almesberger: spi_atben: removed classifier (ben-wpan) http://qi-hw.com/p/qi-kernel/ccfe8ad10:25
qi-bot[commit] Werner Almesberger: spi_atben: added detection of FORCE_TX_ON commands for interrupt synchronization (ben-wpan) http://qi-hw.com/p/qi-kernel/46b0b6b10:25
qi-bot[commit] Werner Almesberger: Merge branch 'ben-wpan-atben' into ben-wpan (ben-wpan) http://qi-hw.com/p/qi-kernel/f48f8b910:25
qi-bot[commit] Werner Almesberger: TODO: updated (master) http://qi-hw.com/p/ben-wpan/865d3bb10:25
qi-bot[commit] Werner Almesberger: install/ben-wpan-config-2.6.38: updated for spi_atben driver (master) http://qi-hw.com/p/ben-wpan/0eff15810:25
qi-bot[commit] Werner Almesberger: TODO: update for switch to spi_atben (master) http://qi-hw.com/p/ben-wpan/b3b037e10:25
wpwrakhmm, those merges are indeed a little chatty ...10:25
wpwraktuxbrain_HxxHhzo: heya ! how's lyf ?11:06
lkclhi folks, i've been recommended to ask for some advice in here11:13
lkcli'm looking for people who might be interested to build a free software licensed hardware project, based around the 1ghz Ingenic jz4770.11:14
lkclmore specifically, a GNU FSF "Hardware-endorsed" laptop and/or micro-server11:15
wpwrakah, never heard of the 4770. is this a big brother to the 4760 ?11:15
lkclyeehhhs, it is :)  1ghz, 65nm.  finally it's a CPU where there's no proprietary crap, that is quick enough to take quotes seriously quotes11:16
lkclthe 4760 wasn't quiite there (110nm or thereabouts, 700mhz)11:16
AylaI know there's a GPU on that one, there are open-source drivers for it?11:16
lkclthe GPU is just yet another X-Burst engine, running at 1ghz.11:17
wpwrak(seriously) heh :)11:17
wolfspraullkcl: when will the 4770 ship?11:17
lkclwolfspraul: don't know.11:17
wolfspraulminor issue11:17
wolfspraul"fsf endorsed laptop" has been a recipe for failure for a number of years now11:17
lkcli've been pinging ingenic.11:17
wolfspraulhope you are not adding another one to the pile...11:17
lkclwolfspraul: that's because the CPUs which can qualify are like... um... shit? :)11:18
wpwraklkcl: you may want to reconsider the "FSF-endorsed" requirement. it's not bad as a "nice to have" objective, but you may find that some of the FSF's ideas wrt hardware are at odds with what makes sense on a practical level to optimize your freedoms.11:18
mthlkcl: afaik the VPU is just another X-Burst core, but there is a separate GPU rectangle in block diagram next to CPU and VPU11:18
wolfspraullkcl: initially you said you are "looking for people interested in building..."11:19
lkclthere's a couple of other CPUs that qualify (and are good enough) such as TI's DM37xx and AM series *as long* as you get the versions without the 3D graphics11:19
Aylayes, there's a CPU, a VPU (second Xburst) and a GPU11:19
wolfspraulwhat is your part?11:19
wolfspraulare you able to throw a few hundred thousand USD on the table for work to start?11:19
lkclwolfspraul: get things going.11:19
wolfspraulor you just want to see others doing it while you comment on it from the side?11:19
wpwraklkcl: unless, of course, someone is sponsoring your project specifically with "FSF endorsement" as a requirement ...11:20
wolfspraulor you want to order one for 199 USD?11:20
lkclwolfspraul: hell no!  but i did get rms interested, enough for him to consider authorising at least _some_ funds.11:20
wolfspraullkcl: there are definitely people here who "want to build" something like that :-)11:20
wolfspraulthat sounds like it will go nowhere11:21
lkcland in some other public discussions, i got an offer of $1k out-of-the-blue towards it.11:21
wpwrakalthough we'd rather try something a little smaller. ben-sized would be ideal.11:21
Aylalkcl: nice! 99k more and you can build it!11:21
wpwrakthere's a war coming in the non-intel netbook category. there's little point in getting oneself onto that battlefield.11:22
lkclwpwrak: i've been planning something like this for a while.  how about a 50x70mm CPU card?  i know of a design-house in china that will, if he is provided with a CPU card, make a 14in laptop around it for $2,000.11:22
wpwrakAyla: multply with 150% and you have about the R&D cost11:22
wolfspraulI would agree with Werner. We have to be careful to not waste money in areas where others are doing it already.11:22
wpwrakAyla: no money for initial production, though11:22
wolfspraulI'd rather innovate on copyleft hardware and in areas where others won't.11:22
wolfspraullkcl: oh my :-)11:23
Aylawpwrak: what R&D? You just need to buy a couple of components, and solder them together! voila! :)11:23
wolfspraulplease stay in this community, it will be good for you...11:23
lkclAyla: i'm very much aware of the differences between USA / EU hardware development costs and those of China / Asia :)  it's a factor of 1011:23
wolfspraulright now honestly I wouldn't know where to start...11:23
wolfspraullkcl: all wrong11:23
wolfspraulI'm typing to you from Beijing, China btw11:23
wolfspraulgotta get dinner soon11:23
Aylawolfspraul: lies!11:23
lkclso i have received quotes from China of $10k for doing a CPU board, and $100k from EU / Germany / USA developers11:23
lkclexact same board.11:24
wpwraklkcl: cheap, good, ever sees the light of day. pick two ;-)11:24
wolfspraulthey mean very different things though11:24
Aylawolfspraul: how did you pass the big firewall? :D11:24
lkclwolfspraul: ni how maaaa :)11:24
wpwraklkcl: and that's true in china as much as elsewhere11:24
wolfspraullkcl: you have to be patient. your goals are similar to the goals of many people here.11:24
lkclwolfspraul: wo jiaouu lukehh11:24
lkclahh goood.11:24
wolfspraulI'm not going to talk you into a Ben NanoNote now, just keep lurking here...11:25
lkcli've been working quietly for over a year, banging my head against the wall on this project from a "mass-produced" angle11:26
wolfspraullkcl: say hello to rms next time you see him :-)11:26
lkclhe he11:26
wolfspraulkeep talking. what was your initial thinking, and how did it evolve over the last year?11:26
wolfspraulwhat did you learn along the way?11:26
lkcllong story :)11:27
lkclwe learned that if you contact enough chinese OEMs, _eventually_ you'll break the myth that chinese OEMs will never part with money to a foreigner.11:27
lkclwe've come up with a de-facto industry standard group of interfaces11:28
lkcland are looking to get this so-called "embedded" CPUs into a 50x70mm user-removable hot-swappable card with ethernet, 24-pin RGB/TTL, SATA, I2C, USB-2 and about 10 GPIO pins on the main internal edge connector11:29
lkclthose are pretty generic "lowest common denominator" interfaces (and there's a Genesys Logic IC which does USB-to-SATA if any one SoC happens not to have SATA, it's about $1)11:30
wpwrakUSB 2.0 full-speed ?11:31
wolfspraul"never part with money to a foreigner"? don't understand11:31
AylaUSB-to-SATA? Isn't that like a very bad idea?11:31
lkclon the _other_ end of the 50x70mm card, it's possible to get HDMI, Micro-SD, Audio jacks, USB-OTG and anything else that the CPU does or doesn't have11:31
wolfspraulin hardware, try to take the time to visit the vendors you plan to work with11:31
wpwrakAyla: it think the fun begins with the edge connector ...11:31
lkclAyla: well, it's better than a kick in the teeth, if the CPU just doesn't have SATA at all.11:32
Aylabut then there's no point to use SATA11:32
lkclwolfspraul: been there.  my associates went to HK and PRC last month.11:32
Aylaas you'll be highly limited by the speed of the USB bus11:32
wolfspraullkcl: to give you an idea. If I wanted to know more about the 4770, small details like when it will be available, what the story is behind it, etc. then I would have a meeting with the Ingenic CEO tomorrow. That helps, quite a bit :-)11:33
lkclAyla: like i said, it's better than a kick in the teeth.  if the CPU is so low-cost it just doesn't _have_ SATA, but the de-facto (proposed) standard says SATA has to at least be provided...11:34
wolfsprauldo not make too many assumptions11:34
lkclwolfspraul: you're _really_ in china, and would do that? :)11:34
wolfspraulI will not do that.11:34
Aylalkcl: yes, but then there's no *point* to have SATA11:35
wolfspraullkcl: of course I'm in China, I just said so.11:35
lkclAyla: if you only are picking a standard to support one and only one CPU, i'd agree with you.11:35
lkclAyla: but if you are picking a standard which is designed for modern CPUs, then if someone wants to try to "force" their way into that standard using a lower-cost CPU and conversion ICs (e.g. USB-to-SATA) then... let them.  it's no big deal.11:37
Aylait's big deal; without real SATA there's no way to run a full system on your board; only from USB or SD which is super-slow11:38
Aylathat's the main reason why I did not buy the Panda board11:38
wolfspraullkcl: the product you are making, who's the user?11:38
wolfspraulit competes with what? a notebook? ipad? iphone? arduino?11:38
lkclAyla: then for those systems where high-performance is an issue, you pick CPUs which have real SATA.  the... oo, S5PV310 for example (i _think_ it's got SATA on-board)11:39
lkclwolfspraul: ahh, that's the fun bit.  it's um... almost everything (but i'm aware there's a "walk-before-run" barrier)11:39
wolfspraulpeople buy it because?11:40
lkclwe want to start off with the "popular" band-wagon products (tablets etc.)11:40
lkclok, let me run you through the "story", and it'll answer that question.11:40
lkclyou walk into a supermarket, up to the computer "appliances" shelf.11:40
lkclon the left-hand rack there are CPU cards.  $30, $40, $70 gets you a low-speed, medium-speed and super-dooper-speed CPU.11:41
lkclthe $30 card has say... a Telechips TCC8901 with 256mb RAM, or a WM8650, or... you get the idea11:42
lkclthe sooper-dooper-speed one has a 1.5ghz OMAP4440, or a Samsung S5PV310, or NVidia's "latest" whatever Tegra - again, you get the idea11:42
lkclon the next shelf is "Chassis".11:43
wpwraklkcl: i'd quickly forget about modularity at such a complex level. better to make the design open so that people who want to design for a different cpu can just adapt the design.11:43
wolfspraulI don't get the idea. it sounds scary.11:43
lkclon this shelf there's a MID / Video Player; a PVR / TV Player; a 7in tablet chassis, an 8in tablet chassis, a 10in one, a 10in laptop, a 14in laptop, a 24in "all-in-one" TV11:43
wpwraklkcl: keep are few _simple_ interfaces for DIY and such. but don't put high-speed busses on extension boards and such. that's just cost and trouble.11:44
lkclwolfspraul: many people don't get it.11:44
wolfspraullkcl: I think you have 2 paths forward: one is that your thinking includes manufacturing, then I very much suggest you start to manufacture things yourself asap.11:44
wolfspraulsmall things, you can start with a budget of a few hundred USD and will learn _A LOT_11:44
wolfspraulor you keep your mind entirely out of manufacturing, on the software side only11:44
lkclwpwrak: look at the list of interfaces again.  the high-speed interfaces are all serial lines with balanced pairs, and the rest are maximum 75mhz (RGB/TTL)11:44
wolfspraulthen you need to look at what is really available today, and you can make guesses on what will be available tomorrow, and you can focus your thinking on hacking into those things11:45
wolfspraulI think you have to make that decision first11:45
wolfsprauldinner, bbl11:46
lkclwolfspraul: can't do that :) been planning this for ages.  i can do the software in my sleep (was one of the original HTC linux hackers / reverse-engineers back in 2004-6)11:46
lkclso this is the next challenge for me11:46
wolfspraulyou want to learn about manufacturing?11:46
lkcli'd rather find people who already have that expertise :)11:47
lkclok - i'm going to chase ingenic.  i have a message sooomewhere with the 4760 datasheet... can't bloody find it11:48
wolfsprauloh my11:48
wolfspraulthat's the reason why everybody hates 'support'11:48
wolfspraul"open source will bring expensive support cases"11:48
wolfsprauland what do people like me tell people like Ingenic?11:49
wolfspraul1) just hang up the phone11:49
wolfspraul2) just delete incoming mails11:49
wolfspraulyes, a lot of nonsense will come if you open up, it's unavoidable. But you still have the power to ignore.11:49
wolfspraulseriously, that's what I'm telling them11:49
wolfsprauland it's the only way for them to protect their scarce engineering resources11:50
lekernelmore or less the same with xilinx...11:51
lkclyeahhh, i know that one... that's why we're also working from the mass-volume angle.11:51
lkcli did say it's a long story :)11:51
lkclwolfspraul: go eat!11:52
wolfspraullkcl: what do you mean with "working from the mass-volume angle"?11:54
wpwrakwolfspraul: "promise volume so they don't hang up so quickly" ? :)11:56
lkclwolfspraul: tell you later, ok? got a couple of things to do.  brief answer: i have associates with connections in UK hypermarket retail stores.11:56
lkclhen how!  they've released schematics of the developer board, the RD477011:56
wpwrakwolfspraul: it kinda worked for openmoko :)11:56
wpwraklkcl: why not go for something a bit simpler ? do you know the ben nanonote ? we're looking for ways to make a successor with improved specs and a much more open design and development process.11:58
wpwraklkcl: the nanonote is much simpler than your high-end modular PC.11:58
wpwraklkcl: so, lower R&D cost. simpler internal interfaces. fewer chips to worry about. lower cost of production ramp-up.11:59
lkclwpwrak: apologies, i have an advantage that i've been going over this for some considerable time.12:00
Aylawpwrak: I wouldn't mind a nanonote tablet :)12:00
wpwraklkcl: i would still put the R&D cost around 100-200 kUSD. about 100-300 kUSD to start mass production, depends a bit on your negotiation powers :)12:00
lkclfrom many different perspectives, not just "one product, one design team, one CPU, one sales channel, one end-user customer base"12:00
wpwraklkcl: good that we just started yesterday ;-))12:01
lkclwpwrak: no, it's not.  ok, if you do all the casework (plastics) from scratch, yes.  but that would be crazy12:01
lkclthe casework alone would be $100k (yes i've been following the openpandora...)12:02
lkclwpwrak: :)12:02
wpwraklkcl: the case is indeed one of the tricky bits. but you need to control that too, or you're not really free.12:02
wpwraklkcl: obviously, you'd keep the case radically simple. no point in trying to out-style apple.12:03
Jay7Ya Tablet is good name for russian market :)12:03
lkclwpwrak: that's why we've been talking about "converting" and "adapting" existing OEM casework, starting with systems that already have the holes in the right places.12:03
Jay7Ya Tabletko even12:03
wpwraklkcl: 100 k for the case sounds like they hired a commercial design house12:03
Jay7(kind of slang)12:04
lkclachh, it's a piss-take price, i know.12:04
wpwraklkcl: OEM cases are risky. also, they limit what you can do to the beaten path. little potential for innovation. every change will be a pain. been there (at openmoko), done that. never again ;-)12:04
lkclwpwrak: *snort* :)12:05
wpwrakJay7: so what does it mean ? :)12:05
lkcloo - it looks like the jz4770 has HDMI out!12:05
wpwrakah, lemme check google translate. more fun that way :)12:05
Jay7wpwrak: well.. something like "I'm tablet"12:06
lkclwtf is "EPD"??12:06
Jay7Ya is tralsiteration of letter / which is equal to "I'm" :)12:06
wpwrakJay7: google xlat proposed  90 B01;5B:>  then said  "da pills" ;-)  (not sure if "da" is just a back-transliteration or a slurred "the")12:07
Jay7wpwrak: yes, really good translation12:07
wpwrakJay7: aah ! kewl :)12:07
Jay7da is the, yes12:07
Jay7slangish :)12:07
wpwraklkcl: (hdmi out) that's why you want a single board. no extension connector. particularly if you don't count on owning the case anyway. (but even if you do, you make your life easier if you can freely move things around)12:08
lkclwpwrak: The Plan is: de-facto standard based around that set of interfaces, 50x70mm CPU cards.  the reasons why are numerous and overall compelling.  basically no - no single board.12:10
lkclthere's precedent for "PCBA" design as it's called.12:10
wpwrak"PCBA" = PCB assembled board12:11
lkclthe Seatron / Chitech PC-89E.  Telechips TCC890x tablets. 12:11
lkclwpwrak: ok, ok :) the ODMs we've been talking to understand what we mean.  ok, we _think_ they understand what they mean :)12:12
wpwraklkcl: well, good luck ! i think you'll learn a lot on that first attempt ;-)12:13
wpwraklkcl: at least you seem to have strong backers. that's good. else, you would even have a hard time getting usable chips.12:14
lkclwpwrak: if companies like Seatron didn't already *successfully* have 200-pin SO-DIMMs, and DirectInsight.co.uk didn't have 200-pin SO-DIMMs with a range of CPUs on them, and various other design companies didn't have modular designs which are virtually identical in concept: yes, i wouldn't even bother12:15
wpwraklkcl: and you don't seem to care about hardware openness. that also simplifies the problem.12:15
lkclwpwrak: what do you mean?  i _do_ care - it's a simple, simple standard, so i'm working from all angles, both the proprietary non-open hardware angle *AND* the competely-open hardware angle12:16
wpwraklkcl: oh, modular is possible. but look at the cost. also, if you don't own the case, there's very little real flexibility in your daughterboards.12:16
wpwraklkcl: all you can do choose which holes in the plastic to fill.12:16
lkclok, ok, wpwrak - i have to get on: let it sink in / mull it over a bit.12:17
lkclthis has been a thought experiment / communicating-with-OEMs exercise for well over a year :)12:18
wpwraklkcl: how many products have you designed and had manufactured so far ?12:22
wpwraklkcl: (including specification and sourcing, of course)12:23
wpwrak(me: openmoko very little gta01, gta02, a bit debug board, a bit hxd8, very little gta03; gta02-core design process only (that's a nice bit of work too - your design house in china won't given you open design files; then three boards at qi-hw, all the way from specification, design to fab-ready output with testing)12:25
wpwrak(wolfgang: openmoko a bit gta02, gta03. then qi-hw ben, milkymist one. ben is an OEM design. milkymist one new from scratch.)12:26
AylaI did work on the openmoko for a school project12:27
Aylapretty weird machine12:27
wpwrakAyla: i hope you enjoyed the rugged case design ;-)12:28
wpwrakAyla: and the reliability of the debug board, in case you used that12:28
Aylathe job was to get Android 2.2 running on it, and make test apps for bluetooth, wifi and GPS12:29
wpwrakAyla: pheww. tall order. and, did it work ?12:29
Aylait was quite easy in fact12:29
Aylathere's already a port of Android on it12:29
wpwrakAyla: ah, then you got into it at the right time :)12:30
Aylabut I was amazed by the battery life ;)12:30
wpwrakAyla: how many minutes ? did you dare to suspend ? and if you did, did it resume ? :)12:31
Aylaon suspend, it wouldn't last one night12:31
Aylaafter a full charge :/12:31
Aylamy openmoko was suspending/resuming fine, I never had problems with that12:32
Aylabut my coworker's wouldn't resume12:32
AylaI was also amazed how non-responsive it is12:33
AylaI believe that's because of the slow CPU<->GPU bus12:33
Aylawith a simple calcul, I discovered that the phone can't deliver more than ~10fps12:34
wpwrakAyla: (non-responsive) that must have been thanks to the graphics decelerator. i think that was a novelty. shame we didn't patent it. see also page 31 of http://people.openmoko.org/werner/ols2008.ps12:36
wpwrakAyla: the other novelty was "embedded air". of that it had plenty. i even made a board for it: http://www.almesberger.net/misc/idbg/12:37
qi-bot[commit] Werner Almesberger: atr86rf230: updated from latest linux-zigbee git (ben-wpan) http://qi-hw.com/p/qi-kernel/a7fdd0813:56
qi-bot[commit] Werner Almesberger: re-applied at96rf230.c part of commit 8684b7aecce3963a04d1aeb9ac4154c0c235791e (ben-wpan) http://qi-hw.com/p/qi-kernel/233a11413:56
qi-bot[commit] Werner Almesberger: spi_atben: cleanup error returns in atben_probe (ben-wpan) http://qi-hw.com/p/qi-kernel/48f628f13:56
lekernel"sane semantics"? for SD?14:10
lekernelsane semantics would be one command to read capacity, one command to read a sector (identified by number), one command to write a sector14:10
lekernelinstead of that they have puked out a 600-page specification14:11
wpwrakand part of it secret, to make things more interesting :)14:12
lekernelyeah..I wonder what sucks more, SD or USB?14:13
lekernelat least SD cards don't require you send them ping packets every millisecond14:13
wpwrak(what sucks more) i've asked myself the same question. couldn't quite decide either ;-)14:14
wpwrakIEEE 802.15.4 is also a good competitor. the spec (only 323 pages) cheerfully begins all talk about MAC-layer things with some 100 pages of MIB and intra-stack communication.14:18
wpwrakand there others would say things like "500 us", you're given a lengthy formula making reference to parameters scattered all over the document. naturally, there's no "for example, for <insert most common case>, this would be 500 us"14:19
whitequarkwhat's your best method of soldering wires to lqfp pins?14:58
lekernelask mwalle on #milkymist, he's the only one who managed to do it properly (and get his board to work) when we messed up the video chip pinout on our first M1 prototypes15:03
whitequarkduh, already done15:47
whitequarkmy preferred one is "when it's all fucked up, step out, calm down, and then start from beginning"15:48
kristianpaulwpwrak: (averaging) no17:48
kristianpaulwpwrak: (resistive probe) interesting, is no the same design you used in atben/atusb?17:49
kristianpaulwpwrak: (peak detect) sound interesting and i read how to do it, let me review manual17:49
kristianpaulwpwrak: i think i did some mistakes adding delayed assignment in non clocked block wich the one that fire the overrrun, thats fixed now waiting synthesis to finish so i'll mesure again17:51
kristianpaulwpwrak: overrun signal freq should be just 0.5 hz off 1024 wich is the clock used by the counter, and as it is 11 bits that mean 0,5 hz per count i think17:59
kristianpaulhe, not so hard in the diagram :-) http://www.colorado.edu/geography/gcraft/notes/gps/gif/bitsanim.gif18:05
kristianpaulDocScrutinizer: this a nice link you posted time ago, finally had time to check it :)18:05
DocScrutinizerme posted a link?19:45
DocScrutinizerI'm posting so many links all the time, sorry I can't reacall19:47
stefan_schmidtwpwrak: what is the .slp_tr pin for? I just have seen that it is connected to an GPIO on the atben but I don't see how it will be handled with the atusb19:47
DocScrutinizerkristianpaul: (and who's interested) this is a definitely not so nice link - MSSF aka aegis, on maemo6/meego-harmattan. Straight from hell:  http://www.developer.nokia.com/Community/Wiki/Harmattan:Developer_Library/Developing_for_Harmattan/Harmattan_security/Security_guide19:48
stefan_schmidtwpwrak: and btw, I had to export the irq function to use them in a module under x86. But now I get IRQ 24.19:50
DocScrutinizerI've been bitching and nagging abut it since at least half a year - got ignored. Today the big boom and everybody's upset and OHNOES!!! OMFG19:50
DocScrutinizerstefan_schmidt: still in USB<->OTA ?19:52
stefan_schmidtDocScrutinizer: yup, getting the driver to behave like a sane spi master19:53
DocScrutinizeraja zigbee19:53
stefan_schmidtno, not zigbee19:53
stefan_schmidtthat what zigbee uses as the first two layers19:53
DocScrutinizer:nod: I know. Just tierd19:53
stefan_schmidtheh, ok19:54
Action: stefan_schmidt has to explain it to people often enough :)19:54
kristianpaul:o, thats new for me 19:58
wpwrakstefan_schmidt: (slp_tr) that's a multifunction pin: used to put the transceiver into sleep mode (which means it will stop supplying a clock to the mcu, thereby stopping usb), and also to start a transmission. for the latter, you can also just write a command to TRX_CMD21:16
stefan_schmidtwpwrak: ok, thanks21:18
wpwrakstefan_schmidt: (export) i actually never tried to build or use this as a module :)21:18
stefan_schmidtwpwrak: I'm still fighting various crashes here. I must have some broken logic wrt platform_device and usb_device21:18
stefan_schmidtwpwrak: heh, its easier for me having it as module here :)21:19
wpwrakkristianpaul: (resistive probe) hmm, not sure what you're referring to. i use resistive probes from time to time. they get a lot less noise and artifacts than "normal" probes. (probably also because i optimize their shape a bit)21:21
wpwrakkristianpaul: here's one: http://people.openmoko.org/werner/rigol/r-probe.jpg (not very nicely made, though)21:22
stefan_schmidtwpwrak: I fear I need some help on the atusb driver from you when your devices arrived. But lets see what I get working until then.21:22
stefan_schmidtwpwrak: as I said, it already probes for the at86rf230 and that one gets loaded.21:23
stefan_schmidtwpwrak: and crashes. So somehow broken pdata I suspect.21:23
stefan_schmidtRight now I commented out most of the probe and remove code in at86rf230 and it works. Going to near it down now.21:24
wpwrakstefan_schmidt: can you use the oops ? or is the system too dead for it ?21:24
stefan_schmidtwpwrak: sometimes I get it. NULL deref  during probe or remove. Will save it next time I get it.21:24
stefan_schmidtwpwrak: ah, and one more thing21:25
stefan_schmidtI regulary reload the driver (atusb) and everytime I get a new spi_master and the old one stays.21:25
wpwrakstefan_schmidt: you could printk the pointer as they get generated. that way, you can easily see their values.21:25
stefan_schmidt[ 1970.331211] at86rf230 spi32750.0: registered at86rf23021:25
stefan_schmidt[ 1970.331636] at86rf230 spi32749.0: registered at86rf23021:25
stefan_schmidt[ 1970.331704] at86rf230 spi32748.0: registered at86rf23021:25
stefan_schmidtand more21:25
stefan_schmidtwpwrak: pointer from what? pdata?21:26
wpwrakstefan_schmidt: whenever you do some   foo *p = weird_op(xxx);  then print the p before it's used21:27
stefan_schmidtwpwrak: ok21:28
wpwrakstefan_schmidt: (registration) hmm.21:28
wpwrakspi_master_put should get rid of the master ...21:28
stefan_schmidtwpwrak: I have that here21:29
wpwrak(oops decoding) you can also use ksymoops. it bulk-decodes the oops for you21:30
wpwrak(when it works ;-)21:30
stefan_schmidtthere was also some oops markup script somewhere iirc21:31
wpwrakthat's ksymoops21:35
stefan_schmidtwpwrak: ksymoops has been removed from the kernel.21:39
stefan_schmidtreadme in scripts ksymoops21:39
wpwrakstefan_schmidt: ksymoops should be available as a package in most distributions21:57
wpwrakmarkup_oops.pl looks interesting. haven't used that one yet21:59
stefan_schmidt[ 4266.274468] at86rf230 spi32766.0: PDATA RSTN -212899516022:00
stefan_schmidthmm, that does not look like valid platform data22:00
stefan_schmidt        .rstn   = -1,22:00
stefan_schmidtis what it should be22:00
wpwraklooks like a pointer, 811a1ca822:01
stefan_schmidtlets follow this path first22:02
wpwrakcheck the pdata pointer and compare with the address of the real thing.22:03
stefan_schmidtwill do22:03
stefan_schmidtwpwrak: ffff880133ad37c8 right now and it contains the right value atm22:14
stefan_schmidtand the pointer is the same in both drivers22:15
wpwrakhmm, far off22:15
stefan_schmidtwhere should it be?22:15
wpwrakno, i mean the other value you got before, ffffffff811a1ca822:15
stefan_schmidtwpwrak: its a %p in printk, hope thats right22:16
stefan_schmidtwpwrak: maybe the other value is wrong due to the many spi masters hanging around22:16
stefan_schmidtI need to make sure to only check the output of the latest master22:17
stefan_schmidtthe other may gone wild already :)22:17
wpwrakah .. the change of the master number doesn't mean that the old one wasn't removed. the numbers don't recycle22:18
stefan_schmidt4 reloads in a row and the pointer is always ok and the value right22:18
stefan_schmidtyeah, its seems I had the wrong numbers from the wrong master :)22:18
stefan_schmidtat least we know now that the pdata gets passed correctly. Thats good :)22:19
wpwrakgreat. one step checked off the list :)22:20
stefan_schmidtLetting the 230 now do some more things during probing22:20
wpwrakfor the interrupt, you may want to put printks at _disable, _enable, _irq, and the handler in at86rf230.c22:23
wpwrakthere's also an irq_ack function i didn't provide. you may want to check that one too. but 22:25
wpwraktuxbrain_HxxHhzo: btw, wolfgang was complaining that he hasn't heard from you about the atben/atusb he ordered.22:29
wpwraktuxbrain_HxxHhzo: how are the critters selling anyway ?22:29
qi-bot[commit] Werner Almesberger: ub: updated to handle dynamically allocated SPI bus numbers (master) http://qi-hw.com/p/wernermisc/0ee750322:38
stefan_schmidtwpwrak: ok, one error source I understand now22:38
stefan_schmidtwpwrak: the many spi_master are still coupled with the 230 driver and are _all_ probing the driver22:39
stefan_schmidtwpwrak: and some of them have no longer valid data....22:39
wpwrakoh, nice :)22:40
stefan_schmidtwpwrak: for example the reset test [if (lp->reset)] does on wrong data now thinks it does not have a reset function and tries gpio22:40
stefan_schmidton x86....22:40
stefan_schmidtfixing the spi_master problem is it for now I think :)22:41
wpwrakdoes this help ? http://downloads.qi-hardware.com/people/werner/tmp/patch22:50
wpwrak(the numbers will still count down)22:50
stefan_schmidtwpwrak: will try it22:52
stefan_schmidtwpwrak: your point is that we need to put the spi device as well?22:52
stefan_schmidtto loose the coupling?22:52
wpwrakjust a wild guess22:52
stefan_schmidtok, trying now22:53
wpwrakit's the only thing i don't put, so ...22:53
stefan_schmidtreboot time22:59
stefan_schmidtwpwrak: nope, does not help23:04
stefan_schmidt[  227.217861] at86rf230 spi32766.0: registered at86rf23023:04
stefan_schmidt[  227.217961] at86rf230 spi32765.0: registered at86rf23023:04
stefan_schmidt[  227.218037] at86rf230 spi32764.0: registered at86rf23023:04
wpwrakwho can you tell ?23:04
stefan_schmidtthe decrement is fine but it does register23:05
stefan_schmidtbecause it gets more everytime I reload the driver?23:05
wpwrakso you see more devices pile up ?23:05
stefan_schmidtone more everytime23:06
wpwrakwhere do you see them ?23:06
stefan_schmidtand if I add debug to 230 it will show it probing for every device23:06
stefan_schmidtin a new dmesg after dmesg -c23:06
wpwrak(probe) that's a clear indication ..23:06
stefan_schmidtok, just to check it. moment23:06
wpwrakoh, but if you unload the driver and load again, it should re-register, no ?23:07
wpwrakdo you also see multiple devices with   ls /sys/bus/spi/drivers/at86rf230    ?23:08
stefan_schmidtsudo modprobe -r atusb && sudo modprobe -r at86rf230 &&  sudo modprobe ieee802154 && sudo modprobe mac802154 && sudo insmod drivers/spi/atusb.ko && sudo insmod drivers/ieee802154/at86rf230.ko23:08
stefan_schmidtthats what I do23:08
stefan_schmidt(devices) yes, multiple23:09
stefan_schmidtbind  module  spi32764.0  spi32765.0  spi32766.0  uevent  unbind23:09
wpwrakhmm, looks bad23:10
stefan_schmidtbut from what I can tell I do nothing different from your driver23:11
wpwrakyeah, the bug may also be in my driver. what you do differently is use modules :)23:12
stefan_schmidtwpwrak: may be23:12
wpwrakwithout modules, each "master unload" is a reboot. that cleans out the old masters really well ;-)23:13
stefan_schmidtthe problem is that it makes it very hard to develop this thing when I have to restart for every test to have a sane first try. :D23:13
stefan_schmidtwpwrak: http://pastebin.com/xtsUrHpJ23:14
stefan_schmidtthat is after several loads23:14
stefan_schmidtand you can see that RSTN is correct in the last one but not the ones before23:14
stefan_schmidtdebug is in the fill_data function called from probe23:14
stefan_schmidtwpwrak: you did build a new kernel and booted into it for every try?23:15
wpwrakyeah. for now, there's no alternative to rebooting after one try.23:15
wpwrakthat's my preferred approach23:15
wpwraknice clean kernel. no risk of chasing the mess some earlier bug left behind :)23:16
stefan_schmidtI'm doing the same on embedded devices, but it sucks for your normal workstation23:16
wpwrakit's useful to have > 1 x86 machines ;-)23:16
stefan_schmidtwell, I don't have it :)23:17
stefan_schmidtWe have some more in the lab, of course23:17
stefan_schmidtbut I will not be there for more then one week23:17
wpwrakso try to make each boot count ;-)23:17
wpwrakmeanwhile, i'll see if i can track this down ...23:17
stefan_schmidtwpwrak: your atusb devices are completely broken right now?23:18
wpwrakwell, close to that. the best one doesn't have a working reset line and the RF is severely compromised23:19
stefan_schmidthmm, or maybe something like kvm23:19
wpwrakalso sometimes loses clock, etc.23:19
wpwrakqemu ! ;-)23:19
wpwrakbut i should get a nice parcel from tuxbrain tomorrow :)23:19
stefan_schmidtwould need to figure out how they do usb passthrough23:20
stefan_schmidtwpwrak: ah, tomorrow already.23:20
stefan_schmidtwpwrak: that means we can hack on this thing more effectively tomorrow23:21
wpwrakmaybe. i also have to get things ready for FISL. leaving tuesday morning, volcano willing23:23
zearguys, does the nanonote still have a ~3 delay after you press and hold the power button before it shows anything on the screen?23:27
zeari have some very early firmware (late 2009?) on my unit, and it behaves this way23:27
zearwe have a talk over at #dingoonity about this, since the old versions of the linux port to the dingoo a320 behaved the same way23:27
stefan_schmidtwpwrak: oh, bad timing (for me :))23:28
zearthough a solution for this was found later, which led to an instant boot after powering on23:28
stefan_schmidtwpwrak: It really seems that someone tries to sabotage my diploma thesis with the ieee802154 drivers :)23:28
zear*~3 sec23:29
wpwrakstefan_schmidt: naw, just reboot and you'll have a clean context ;-) you seem to be quite close to having it all work anyway23:31
Action: stefan_schmidt does not have a nanonote, can't tell23:31
stefan_schmidtwpwrak: close is not done :)23:31
wpwrakzear: i still get some delays but i update things sparingly23:31
stefan_schmidtwpwrak: anyway, reboot time...23:31
zearwpwrak, the thing is the delays have returned for us with OpenDingux (nex linux port to the dingoo), and we've kind of.. lost the solution for the delay fix in the legacy linux port23:32
zearso we're currently browsing the old sources and changelogs hoping to rediscover the fix23:32
wpwrakah, maybe larsc or xiangfu can help you there.23:32
zearand i thought once we find it, perhaps it could benefit for the nanonote as well :)23:32
wpwrakregressions are fun ;-)23:33
qi-bot[commit] Werner Almesberger: kernel/irq/: EXPORT_SYMBOL_GPL functions for dynamic interrupt allocation (ben-wpan) http://qi-hw.com/p/qi-kernel/5cdcaf423:37
stefan_schmidtwpwrak: callback into the reset handler inside atusb works. Not reset itself but the callback. Now checking the your tools for the reset23:37
qi-bot[commit] Werner Almesberger: spi_atben: added section titles (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/4926b1023:38
qi-bot[commit] Werner Almesberger: spi_atben: moved atben_reset and at86rf230_platform_data from board to driver (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/9a7d79a23:38
qi-bot[commit] Werner Almesberger: spi_atben: moved qi_lb60_atben platform device from board to spi_atben (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/00fded823:38
qi-bot[commit] Werner Almesberger: spi_atben: moved spi_board_info of atben from board to spi_atben (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/b9def2223:38
qi-bot[commit] Werner Almesberger: spi_atben: allocate SPIP bus_num dynamically (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/e08d15623:38
qi-bot[commit] Werner Almesberger: spi_atben: some cleanup (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/3e7de6023:38
qi-bot[commit] Werner Almesberger: spi_atben: announce atben_reset and don't modify global variables (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/add45cd23:38
qi-bot[commit] Werner Almesberger: spi_atben: we set prv->board_info.platform_data later, don't mis-initialize (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/8a07eed23:38
qi-bot[commit] Stefan Schmidt: Merge remote-tracking branch 'origin/ben-wpan-atben' into ben-wpan-stefan (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/42c875a23:38
qi-bot[commit] Werner Almesberger: spi_atben: added optimized unidirectional SPI bitbangers (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/1fe9fab23:38
qi-bot[commit] Werner Almesberger: spi_atben: more minor cleanup (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/db9015223:38
qi-bot[commit] Stefan Schmidt: Merge remote-tracking branch 'origin/ben-wpan-atben' into ben-wpan-stefan (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/3ce7e2e23:38
qi-bot[commit] Stefan Schmidt: kernel/irq: Export IRQ functions we need for the ATUSB build as module (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/562e0da23:38
qi-bot[commit] Werner Almesberger: spi_atben: removed classifier (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/ccfe8ad23:38
qi-bot[commit] Werner Almesberger: spi_atben: added detection of FORCE_TX_ON commands for interrupt synchronization (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/46b0b6b23:38
qi-bot[commit] Werner Almesberger: Merge branch 'ben-wpan-atben' into ben-wpan (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/f48f8b923:38
qi-bot[commit] Werner Almesberger: atr86rf230: updated from latest linux-zigbee git (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/a7fdd0823:38
qi-bot[commit] Werner Almesberger: re-applied at96rf230.c part of commit 8684b7aecce3963a04d1aeb9ac4154c0c235791e (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/233a11423:38
qi-bot[commit] Werner Almesberger: spi_atben: cleanup error returns in atben_probe (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/48f628f23:38
qi-bot[commit] Stefan Schmidt: Merge remote-tracking branch 'origin/ben-wpan' into ben-wpan-stefan (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/904e54023:38
wpwrakoh .. you're merging ;-)23:39
stefan_schmidtwpwrak: long ago already. I should commit and push more often...23:39
wpwrakwhen i daw the first message, i thought my push has triggered some trouble ;-)23:39
stefan_schmidtreset now and then looking into the SPI transfer functions23:41
wpwrakhmm, rmmod -> dereferences NULL pointer23:44
stefan_schmidtI do modprobe -r here23:46
stefan_schmidtplease check the spi dev for null23:47
stefan_schmidtI somehow remember having issues with this during all the crashes23:47
wpwrakyou mean the one returned by spi_new_device ?23:47
stefan_schmidtwpwrak: in 230 or atben?23:48
stefan_schmidtwpwrak: I have this in 230 remove()23:48
stefan_schmidt      if (spi)23:48
stefan_schmidt                spi_set_drvdata(spi, NULL);23:48
stefan_schmidtnot sure if I still need it23:49
stefan_schmidtusing spi before is fine though...23:49
wpwrakseems to happen in set_irq_chained_handler23:49
Action: stefan_schmidt has opened a can of worms23:50
wpwrakprobably just the wrong order ...23:50
wpwrak(can of worms) yeah. modules just make life complicated and unpleasant :)23:51
wpwrakrmmod now passes23:51
wpwrak... without removing the master23:52
stefan_schmidtthis master thing is really fishy somehow23:52
wpwrakat least it doesn't insmod again ;-)23:52
stefan_schmidtglad that it was not a heisenbug I'm hunting here :)23:53
stefan_schmidtnew tests prepared, reboot time23:53
--- Mon Jun 27 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!