#qi-hardware IRC log for Wednesday, 2012-01-11

wpwrakhehe :)00:05
wpwrak(build from source) i'll have to add some things for boom interaction anyway00:06
wpwrak(upstream) don't you as the author of those patches want to do it yourself ?00:06
wolfspraulI am overloaded00:11
wolfspraulno news in ... months00:11
wolfspraulplus I think my main work nowadays is simply to find money00:12
wolfspraulas much as I like coding00:12
wolfspraulcan't do it00:12
wpwraki think the news concept has failed. at least the news-with-wolfgang-as-editor. pity. but maybe someone else will step up at some point in time.00:13
wpwrak(find money) yes !! :)00:14
wolfspraulwe don't have many readers anyway00:14
wolfspraulso the reason is more for archival purposes, and I do continue to connect those links00:14
wolfspraulof course the intervals are not good00:14
wolfspraulone by one00:14
wpwrak(coding) with a little luck, upstreaming shouldn't be too dreadful. and probably easier to do it yourself than explain to xiangfu why exactly you've done what00:15
wolfspraulpossible, I will see00:15
wolfspraulno worries we will try what works best00:15
wolfspraulno dogma00:15
wolfspraulbut if xiangfu can help me, great00:15
wolfspraulI don't want another up-leveling in 201300:15
wolfspraulI have no idea how 2011 disappeared so fast...00:16
Action: pabs3 wants an open hardware, firmware version of the Crypto Stick02:54
Action: kristianpaul Dormido03:23
wolfspraulpabs3: are you saying the crypto stick is not open?07:59
pabs3wolfspraul: not according to their intro page: "integrated (proprietary) smart card": http://www.crypto-stick.com/en/introduction08:01
wolfspraulhmm, ok08:02
wolfspraulneed to find out more first08:02
wolfspraulbut I doubt it's proprietary because that's their business model08:02
viriclarsc: ah the patch... at the evening. I've that in my home computer08:05
pabs3wolfspraul: if they're assembling off-the-shelf components (like that ATMEL or the uSD card), probably some of those are proprietary in some way: http://www.crypto-stick.com/2011/crypto-stick-2-beta-boards-arrived08:20
wolfspraulsure, unless I see them actively closing up something somewhere, I'm not worried08:22
wolfsprauland I think they are doing a serious and good project, but slowly we will learn more about it08:22
DocScrutinizerstay tuned for c't testing the crypto-stick... ?09:09
wolfspraulDocScrutinizer: thanks a lot for helping out with the J3 audio expansion thing!09:46
wolfspraulwe start to polish the details a bit :-)09:47
DocScrutinizer51I strongly prefer helping during design phase rather than later fixing hw bugs09:47
wolfspraulwell in that case it's just about reducing that big header to something more thoughtful09:48
wolfspraulwhich I'm sure we now have, thanks to your help too09:48
DocScrutinizer51and I'm regularly offering to do so, but alas often it gets ignored. So on GTA04-goldelico09:48
wolfspraulwell here I hope you recognize we appreciate your input a lot. worst case it's being recorded for the moment, and picked up later.09:49
wolfspraulwiki, irc logs, etc.09:49
wolfspraulgta04, don't know. I don't understand the project.09:49
DocScrutinizer51don't worry, me neither :-D09:50
DocScrutinizer51the 'funny' detail: again audio section is in top 3 of messed up circuits09:51
blogicDocScrutinizer51: is audio out hard to layout ?09:53
blogici have a board here with i2c/pcm bus and wanted to make a small breakout with a wolfson codec on it09:53
blogici got the impression it would be trivial09:54
blogicbut reading your comments i might be mistaken09:54
DocScrutinizer51nah, audio isn't particularly hard10:01
DocScrutinizer51wolfspraul: (messed up) in GTA0410:02
DocScrutinizer51wpwrak: I thought about the CD in a bit more10:02
DocScrutinizer51the purpose of 'pulldowns' is obvious. But when you wnt that, you need pulldowns to virtual GND, not to real GND10:03
DocScrutinizer51you may or may not want to tie virtGND to real gnd with a 100kR//1nF then10:05
DocScrutinizer51bbl, work10:05
qi-bot[commit] Wolfgang Spraul: Werner: we can drop the ERC pin exceptions patches. (master) http://qi-hw.com/p/eda-tools/78b197d10:24
qi-bot[commit] Wolfgang Spraul: merged with cmdline patches (master) http://qi-hw.com/p/eda-tools/771599b10:25
qi-bot[commit] Wolfgang Spraul: patch cleanup (master) http://qi-hw.com/p/eda-tools/724ca1e10:50
qi-bot[commit] Wolfgang Spraul: updated patchset to kicad rev 3351 (current head) (master) http://qi-hw.com/p/eda-tools/0bdf90411:21
wolfspraulwpwrak: you can try, I upticked it to 3351 (current latest)11:22
wolfspraulactually there are so many renames going on they managed to break a few things in the last few days alone11:22
wolfspraulactive project, cannot complain. all the more reason to upstream asap.11:22
wolfspraulI left out the pin collision patch for now because it's just a few lines, and of course that area changed as well. it's now called 'SynchronizePins()'11:23
wolfspraulI could uplevel the patch, but that might do more harm than good because I would basically guess the meaning of those variables. better you look at it.11:23
wolfspraulwpwrak: looking at the m1 board a bit more thinking about the 'expansion system'12:38
wolfspraulI think the corner around the audio and infrared makes a lot of sense, basically the entire 'upper floor' of that corner can be where the expansion boards can set12:38
wolfspraulthe other side is crowded with large DMX connectors, video-in, jtag-serial, etc.12:39
wpwrakwolfspra1l: (corner) yes. there's room there. no tall components, plus nothing on the case wall, so one can install a customized wall with openings, if needed13:13
cladamwwpwrak, one quick question: we short L3 for fixing audio problem in rc3. When routing in rc4, just to use 0 ohm or directly short GND and AUDIO_AGND?13:22
cladamwwpwrak, this question will be the same as L19 for video.13:22
cladamwsince there's different copper gnd(digital and analog) area splitted.13:24
cladamwi think that I can just use copper wire to connect/wire two existed ground systems. (i.e. to be the same as rc3 patches now)13:25
cladamwwpwrak, and keep the existing copper plates not to fill them up. or you have another different thinking?13:27
wpwrakcladamw: (l3, l19) i'd short directly. l3 and l19 are clearly wrong and keeping them a "0 R" just adds confusion13:28
cladamwoah~i typed in wrong channel 13:28
cladamwshould go #milkymist13:28
wpwrakwolfspra1l: (pin collision patch) okay, that one can wait a few days. we only need it when editing symbols.13:50
wpwrakwolfspra1l: (things broken in the last few days) that sounds scary. what sort of breakage ?13:51
wolfspra1ljust renaming again13:55
wolfspra1lthey seem to love renaming13:55
wolfspra1lI already mentioned it here, case up and down, underscore in and out13:55
wolfspra1lno consistency either, seems several authors editing each others codes back and forth13:55
wolfspra1lwell then13:55
wolfspra1lrenaming all dialog classes from DIALOG_SOMETHING_BASE to DIALOG_SOMETHING_base begs for another rename :-)13:56
wolfspra1lmixed-case - bah13:56
wpwrakat least they're keeping the indentation/whitespace style, aren't they ? it's one of the ugliest i've ever seen, so it's hard to progress from the status quo13:58
wolfspra1lwpwrak: yes I mostly see battles around underscores and case14:00
wolfspra1lI replied to my 1-yr old thread on the mailing list :-)14:01
wolfspra1lnot very nice, but what's a good mail client for...14:01
wolfspra1lat least it wasn't a 10-yr old thread14:01
wpwraklet's hope people see it. if using a threading MUA, they may not notice14:02
viricI sent the patch18:46
viricthen, why my serial line does not work.grr18:47
viricnow the serial works. weird.18:55
viricbut it fails sometimes at 5760018:56
virichow reliable it should be?18:58
virichm I send 'hello\n' in a loop...19:04
viricand not all reach fine the other side :)19:04
viricmaybe it's me with stty, that I don't know how to use it.19:05
larscmaybe your other side is broken19:05
viricmaybe I have too bad contacts :)19:05
larscsince there is no flow control the jz4740 will happily send out new bytes whether the other side is ready to receive them or not19:06
viricnot at all?19:07
viricisn't xon/xoff working? I used "ixon ixoff" in both sides19:07
larscno hardware flow control19:07
viricI know..19:07
larscsw flow control should work19:07
viric(serial lines, 2012, and still a mistery to get right)19:08
Aylaif you have only 3 wires, then there's no hardware flow control19:08
viricwhy shouldn't xon/xoff work? it's the OS implementing that, or the uart?19:08
larscprobably a bit of both19:08
viricthrough irqs..19:09
larscbut we use the generic 8250 driver19:09
viricuhm loading the driver reprograms the line at 57600 I imagine. Because at 9600, I see until the 'serial8250.0' loading :)19:18
viric[    0.800000] jz4740-nand jz4740-nand.0: No NAND chips found19:27
viricI have MTD_NAND_JZ4740 enabled...19:28
Action: larsc blames mth19:29
mthNAND still works fine on the A320 though, I tested that19:30
viricehem. this kernel is not in its best moment ;)19:30
viricTime for kgdb then! :)19:30
viricwhat makes the nanonote sometimes turn on at connecting power, sometimes not?19:32
viricbtw, why can't I enable CONFIG_FRAME_POINTER?19:32
viricmaybe they are not required.19:33
larscviric: thanks for sending the patch19:50
viricI was only the typing hand19:51
viricI'm going to play with the uart...19:51
viricI want the kernel to disable the gpio function of the pin, and use rx...19:51
DocScrutinizer(xon/xoff) might work suboptimal when TX FIFO isn't handled in a way to place XOFF on top of FIFO, so another 16 or even 64 chars get sent until XOFF gets sent19:52
DocScrutinizerlarsc should know how that's inplemented19:52
viricDocScrutinizer: so the uart has to help in that19:53
DocScrutinizerI do't think UART knows XON/XOFF flow control19:53
viricah. then? 19:53
DocScrutinizerdriver side implementation19:53
DocScrutinizerI guess19:53
viricmaybe the driver can be set some chars to bypass the bufefr19:54
DocScrutinizerthe buffer is HW though19:54
viricI know19:54
DocScrutinizerin UART19:54
viricbut maybe the uart can trigger an irq, if some special char 'enters' the fifo19:54
DocScrutinizerthat had to be checked19:55
viricor maybe xon/xoff can't replace properly hw flow control19:55
DocScrutinizertheoretically it's possible19:55
viricbut I think that proper UARTs should be able to use it19:55
viric(with proper drivers)19:55
virichttp://vicerveza.homeunix.net/~viric/cgi-bin/nanonixos/artifact/3b11d7bf1ea31a579211febf61977d46a8af97e4 my poor start19:56
DocScrutinizerproper driver: RX congestion -> flush TX FIFO, send XOFF, resend flushed bytes19:57
viricyou should handle the possible deadlock :)19:57
DocScrutinizereven ignore TX-XOFF state, as XOFF may be sent even when TX is itself blocked by a XOFF19:58
DocScrutinizerhehe ^^^19:58
viricI don't find that easy to think about :)19:58
DocScrutinizerXOFF/XON alway may get sent19:58
DocScrutinizerIOW, a XOFF may not prevent XOFF/XON to get sent19:59
viricI agree19:59
viricdo you have the jz4740pm.pdf somewhere?20:01
DocScrutinizerme? nope20:01
viricanyone here20:01
viricis it secret? :)20:08
DocScrutinizerhttp://read.pudn.com/downloads160/ebook/723804/Jz4740-spec/Jz4740_ds.pdf hmm, not really good but doesn't mention XOFF IRQ20:24
viricok ok20:30
Action: DocScrutinizer is feeling nostalgic about times when Zilog shipped their Z80 CPU with a proper paperback 150p manual20:32
kristianpaulpaperback manual, nice ! i remenber when i used to print datasheets for PICs ;)20:34
larscnever ever used printed datasheets20:38
kristianpaulpic18f havent chnaged too much since 3 years i remenber.. i hope ;)20:40
kristianpaullarsc: yes sure, for that time i dint have laptop and read on CRT display wasnt very nice..20:41
larscah, ok20:41
larscas DocScrutinizer said, back in the good old days hardware come with printed manuals20:42
kristianpauli always wanted to have a z80... sigh 20:42
larsckristianpaul: implement a z80 on the milkymist ;)20:44
kristianpaulyou serious?20:44
larscsure, why not20:45
larscthe z80 is rather simple20:45
kristianpauli agree with lekernel on this :) waste of time (:20:45
kristianpaulhmm indeed20:46
kristianpaulwhere are the multiplers ! :)20:46
DocScrutinizeror this one: http://www.ebay.de/itm/IC-Z80B-CPU-Zilog-Z8400BPS-/380137224284?pt=Bauteile&hash=item5881f1b85c20:55
DocScrutinizeror this  XP : http://www.ebay.de/itm/Zilog-Z80-CPU-Z0840006PSC-9904-2Q-11070-/190616518598?pt=DE_Elektronik_Computer_CPUs&hash=item2c61a13fc620:56
DocScrutinizer\o/ ASIO: http://www.ebay.de/itm/IC-BAUSTEIN-Z80ASIO-CPU-17488-129-/170714918488?pt=Bauteile&hash=item27bf66ee58 ; CTC: http://www.ebay.de/itm/IC-BAUSTEIN-Z80CTC-CPU-17493-129-/130590019482?pt=Bauteile&hash=item1e67c58f9a ; PIO: http://www.ebay.de/itm/IC-BAUSTEIN-Z80BPIO-CPU-17492-129-/170714918565?pt=Bauteile&hash=item27bf66eea521:02
DocScrutinizerI'd actually consider to buy that one - just because it looks so nice: http://www.ebay.de:80/itm/ws/eBayISAPI.dll?ViewItem&item=250966296618&clk_rvr_id=306143752208&item=250966296618&lgeo=1&vectorid=22948721:05
mthlarsc: there is an open Z80 core for FPGA already21:09
larscmth: yea, though so21:11
mthit's used in this system: http://www.hat.hi-ho.ne.jp/tujikawa/esepld/21:11
viricmth: you introduced lots of changes in the nand driver :)21:17
mthyes, but most of them only affect the behavior after the chips have been detected21:18
virichm ah21:19
DocScrutinizera last goldie oldie one: http://zilog.com/index.php?option=com_product&task=product&businessLine=1&id=139&parent_id=139&Itemid=5921:19
mthyou can ignore the cc_ftl and A320 patches, they don't apply to NanoNote at all21:19
mthe7ca5a665877a030c461959c0853e65a346de2de (multi-bank) does change the detection21:20
mthhmm, "banks = " is missing from board-qi_lb60.c21:21
mthcan you try adding ".banks                  = { 1, 2, 3, 4 }," to qi_lb60_nand_pdata?21:22
mthis there any variation in the NAND banks of different NanoNotes or is it always the same bank?21:23
viricI'll try21:25
viricmth: worked21:26
mthok, then larsc was right to blame me ;)21:26
mthlarsc: should I make it search all 4 banks or should I hardcode the actual bank for the NanoNote?21:27
larschm, i think hardcoding ist ok21:27
mthviric: in which bank was the NAND found? (see dmesg)21:28
viric[    0.830000] jz4740-nand jz4740-nand.0: Found chip 0 on bank 121:29
viricnothing on 2,3,421:30
mthok, thanks21:30
viricthere was in the code: ret = nand_scan_ident(mtd, 1, NULL);21:30
viric'1' is the bank, maybe.21:30
viricno idea.21:31
viricah, 1 is 'maxchips'.21:31
DocScrutinizerA nice reading matter: http://www.zilog.com/docs/z80/um0080.pdf p.42ff >>... When an EI instruction is executed, any pending interrupt request is not accepted until after the instruction following EI is executed. This single instruction delay is necessary when the next instruction is a return instruction. Interrupts are not allowed until a return is completed. The... << - I think I learnt all I know about IRQ from Zilog Z80 UM :-D21:31
viricmth: should I wait for a commit?21:39
qi-bot[commit] Maarten ter Huurne: MIPS: JZ4740: qi_lb60: Look for NAND chip in bank 1. (jz-3.2) http://qi-hw.com/p/qi-kernel/450999b21:43
viricoh ok nice21:43
mththat's a yes :)21:44
Action: wpwrak remembers going to the intel office in zuerich at the age of twelve or so, trying to buy 8080 documentation. they were a bit puzzled but kind enough let me have the manual for free. i still have it. one fun fact is that it mention even the 8088. remarkable long-term planning. needless to say, that was a few years before downloadable pdfs :)21:53
DocScrutinizer:-D http://www.zilog.com/docs/z80/ps0178.pdf  <- scanned from my original ;-)22:01
DocScrutinizerwpwrak: I conclude you are some ~7 years younger than me22:03
DocScrutinizermaybe 522:03
viriclarsc: I could attach a gdb at boot :) but I forgot the debug symbols22:07
wpwrak1967 here22:08
viricbut 'cont' made the kernel boot :)22:08
viriclarsc: I wrote jz_gpio_set_function(JZ_GPIO_UART0_RXD, JZ_GPIO_FUNC_UART0_RXD);   in the 'board_gpio_setup' as a trick to have rx working on ttyS022:08
viricfair? I had no idea what todo22:11
viricnicer would be a kernel cmdline parameter for that22:11
larsclooks good22:11
viricbut who knows how to write that22:11
larscpretty easy22:11
larscthere is one for the avt222:12
larscyou could use that as a skeleton22:12
viricwhat is an avt2?22:12
larscit was a experimental nanonote board with usbhost and 64MB ram22:12
viricnot jz4740?22:13
larscinstead of the jz472522:13
viricthere is only jz4740 and a320, not?22:13
viricI'll recheck22:14
larscin the nanonote board file there is a callback which registers the usb host device22:14
larsccalled avt2_something22:14
larscor something_avt222:14
viricah yes22:14
viricthere are cmdline checks in that driver?22:16
larscno in the nanonote board file22:16
larscjust search for the function name22:17
viricI couldn't see them then22:17
viricI think I did22:17
larsclet me check22:17
larscboard_avt2 is the function22:17
larscand it is hooked up to the avt2 keyword on the commandline via __setup("avt2", board_avt2);22:18
viricI had no idea that was a hook22:18
DocScrutinizer51wpwrak: so my initial guess was absolutely correct :-) tell me which CPU you used and I tell you how old you are22:23
AylaDocScrutinizer51: Pentium 322:26
larscDocScrutinizer51: my first cpu was a pentium I 90 Mhz22:26
Aylahow old am I? :D22:26
Aylalogically, I'm a lot younger than larsc 22:27
DocScrutinizer51too young :-D22:27
larscand my first "compiler" visual studio c++ 6.022:28
larschehe :)22:29
wpwrakglad you managed to become a decent person despite the abuse you suffered in your youth :)22:29
wpwrakor maybe it was that what instilled a strong willingness to escape the misery22:30
virichm I don't manage to debug with kgdb.22:30
Action: nbd started with a cyrix 486 DX and turbo pascal ;)22:30
Jay7turbo pascal is the best!22:31
Action: Jay7 started from P1 100Mhz w/o MMX22:31
Jay716Mb RAM btw..22:31
Action: wpwrak wonders if psychoanalysts ask hackers about their first cpu and compiler. these two are probably a lot more influential on the shape of the blossoming mind than, say, parents22:31
Jay7but before was ZX Spectrum :)22:31
Aylaactually I started with a 6MHz Z80 CPU22:33
Aylaand 32kB of RAM22:33
wpwrakTi 57, 50 program steps, 8 data registers22:35
viricit looks like the vmlinux addresses don't match those of runtime at the nanoonte22:36
larscwpwrak: and my first book on programming book 'c++ in 30 days'22:38
Jay7larsc: I have that book :)22:40
Jay7or may be similar at least22:40
viriclarsc: the kernel is not loaded at the addresses the vmlinux elf says22:40
larscby whom?22:42
viricby my xbboot command22:42
larscviric: the load address or the entry point address is wrong?22:44
virichm I don't know22:44
viricwhen I attach gdb, it sayas:22:44
viric#0  0x81c5274f in ?? ()22:44
viricwhile 'info file', says:22:44
viric        0x80010000 - 0x803210c8 is .text22:44
larscis the kernel halted when you attach gdb?22:46
viricI used the vmlinuz load address calculation, and uploading to  that... it boots fine too, but gives the same gdb symbol strouble22:46
viriclarsc: yes yes. if I write 'cont', it continues22:46
viricBut why is it at 0x81c5274f, I don't know22:46
larscviric: have you tried break some_kernel_symbol? The address you see above is probably just some userspace process22:47
viricI tried22:47
viriclarsc: I use 'kgdbwait'. it's halted just after the serial driver got loaded.22:47
viric(no user space)22:47
larscah, ok22:47
viric'disasm' fails22:48
viricI mean, that 'disasm symbol' does not disassemble the symbol22:48
viricI think the bad boot messages mean: uncompressing kernel at address 8001000022:50
viric(there are some characters 'eaten' there... the early console may be working bad)22:50
viric80010000 is the vmlinux load address.22:50
virichm I see the vmlinuz address is not very important.22:53
viricit has to be valid, and it shouldn't be overwritten by decompression :)22:53
viriclarsc: using kdb, it looks fine.23:12
viricwith gdb, bad.23:12
viricI forgot to start the *mipsel* gdb.23:12
DocScrutinizerwpwrak: EEEEK TI23:14
viricstupid me23:14
viricit works perfect23:14
DocScrutinizerHP-25 here23:14
virickgdb working!23:14
DocScrutinizerwpwrak: and we coded everything from "shortest fastest prime factoring" over "mastermind" and "guess the atom"(send ray in a grid, see where it exits), to auto-adjusting converters for Philips cassette tape counter to real playtime (sub minute accuracy) and even stopwatches (a pity on that "platform", we learnt a lot though about how the CPU works :-D)23:19
DocScrutinizere.g. execution speed for a particular instruction was a sawtooth function vs the location in "RAM" of that instruction23:20
DocScrutinizerand of course the whole thing was highly ambient temperature susceptible23:20
DocScrutinizerand the best time to compete on all those nice efforts was during english and physics lessons ;-D23:22
wpwraki mainly wrote games in my early years :)23:22
viriclarsc: so, enough for today. Victory!23:23
viricmth, Ayla, larsc: http://vicerveza.homeunix.net/~viric/cgi-bin/nanonixos/doc/trunk/doc/kerneldbg/index.wiki23:27
AylaI see that you define the root filesystem as being UBI23:31
Aylaand I remember the nanonote uses UBI too (IIRC)23:32
viricAyla: well, that's what I have in the nand23:32
Aylahow do you guys boot to UBI? I thought u-boot didn't support it23:32
viricthe kernel is in a nand partition directly23:32
viricnot in a file23:32
Aylaah, ok23:33
viricthere are 4MiB for it.23:33
viricalthough uboot loaded only 2MiB (cough)23:33
AylaI started a new bootloader for JZ-based devices,23:33
virichow so?23:34
Aylait's called UBIBoot because it boots to UBI23:34
viricto a file in an ubifs?23:34
Aylano, to a kernel flashed on a UBI volume23:35
viricwhy not implement that on uboot?23:35
Aylawell, I wanted to create a bootloader myself :)23:36
viricah ok23:36
Aylaback on topic - good job with kgbd23:37
viricI like when people write on the documents what they type and also what they see as result :)23:37
viricI remembe as if the qihw wiki had too much of 'type that', without result23:38
viricok, sleep.23:40
DocScrutinizernow uBoot bashing continues *here* ? :-S23:45
--- Thu Jan 12 201200:00

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