IRC log of #maemo-ssu for Wednesday, 2013-09-18

*** M4rtinK has quit IRC00:13
*** dhbiker has joined #maemo-ssu00:45
freemangordonsailus: ping00:46
freemangordonPali: some news - seems there is a bug in clock framework, on 2.6.28 SSI clock chain is:00:47
freemangordon"/sys/kernel/debug/clock/virt_19_2m_ck/osc_sys_ck/sys_ck/dpll3_ck/dpll3_x2_ck/dpll3_m2x2_ck/corex2_fck/ssi_ssr_fck/ssi_sst_fck/"00:47
freemangordonon 3.10 it is:00:48
freemangordonwell, different :)00:48
freemangordonI fixed that, but still no joy :(00:48
freemangordonPali: any idea what notion m2x2 (in clock names) means?00:49
Palino idea00:49
Palifreemangordon: another question: is audio working with 3.10?00:50
freemangordonI am out of ideas, I see cawake interrupts (which IMO means the modem is alive and trying to communicate)00:50
freemangordonPali: no00:50
Paliok00:50
freemangordon... but I see no port interrupts00:50
Paliin 3.5 or 3.8 audio worked00:51
PaliI remember that I heard nokia welcome video00:51
freemangordonI see errors from PA00:51
freemangordonwell, that means DSP was working too00:51
DocScrutinizer05freemangordon: do you have a IRQ-handler set up for that IRQ?00:51
freemangordonyes00:51
freemangordonand I see it in /proc/interrupts00:52
Paliyes, dsp worked before 3.10 (or 3.9) before multiarch for omap3 was introduced00:52
freemangordonbut with count of 000:52
DocScrutinizer05is it ever executed?00:52
freemangordonno00:52
DocScrutinizer05check your IRQ settings00:52
DocScrutinizer05IRQ on OMAP are... weird00:52
Palifreemangordon: for audio you can try to revert my two audio patches in git tree and then manually load audio kernel modules00:53
freemangordonbut cawake interrupt handler is executed00:53
freemangordonport interrupt is not connected to gpio00:53
freemangordonthey are fine, it is irq 6700:53
Palimaybe there is problem00:53
freemangordonPali: I want SSI working :)00:53
DocScrutinizer05e.g the IRQ dispatcher/mux needs *clock* to even notice a IRQ, unless it's a wakeup IRQ, afaik00:53
PaliI tried to migrate audio driver/configuration to new format with hotplug and module autoload support00:53
Paliok00:53
freemangordonDocScrutinizer05: do you have SSI documentation by chance?00:54
freemangordonI hate to work blindly00:54
DocScrutinizer05sorry nope, lost it with my last job00:54
freemangordondo the guys with gta04 have it?00:54
freemangordonI mean, SSI os a part of OMAP00:55
freemangordon*is00:55
*** Martix_ has quit IRC00:55
DocScrutinizer05had several 1000 pages of printed docu about MIPI HSI, IRQ in OMAP4 (or the lack of...), SSI, UART, whatnot else00:55
freemangordonDocScrutinizer05: BTW for sure 3.x sets MUX incorrectly for rx5100:55
DocScrutinizer05I'm not surprised00:56
freemangordonbut this is internal IRQ, MUX config should not matter IMO00:56
freemangordonirq 67, any bell?00:56
DocScrutinizer05right, it's internal interrupt00:56
DocScrutinizer05nope00:56
freemangordonssi_p1_mpu_irq000:57
freemangordonshit00:57
DocScrutinizer05IroN900:~# grep "67:" /proc/interrupts00:57
DocScrutinizer05 67:     204730        INTC  ssi_p1_mpu_irq000:57
freemangordonyep, the same with 3.10, but count is 000:58
freemangordonhonestly, no idea where else to look at00:58
freemangordonmaybe I need to set some consumer?00:59
DocScrutinizer05hmm, 67 not listed in /sys/kernel/debug/omap_gpio00:59
freemangordonit is not gpio00:59
DocScrutinizer05err, THAT rings a bell. Though absolutely nothing more than that00:59
DocScrutinizer05"consumer"01:00
DocScrutinizer05think I heard coleagues bitch about that01:00
freemangordonyep, of some power supply01:00
freemangordonbut that makes no sense either, as driver seems to communicate with SSI core01:00
DocScrutinizer05aaah right01:00
freemangordon?01:00
DocScrutinizer05"power supply consumer"01:01
freemangordonyep01:01
freemangordonfor RAPU/GAZOO01:01
freemangordonbut if cmt is not powered I won;t have cawake interrupts, right?01:02
DocScrutinizer05consider using a different GPIO/IRQ, just for testing purposes. Like GPIO_68 cam-focus01:03
freemangordonhmm, what for?01:03
DocScrutinizer05well, I think THAT is a given fact01:03
freemangordonfor sure this is the correct IRQ, see 2.6.2801:03
DocScrutinizer05sure, I just thought you could press button and see if IRQ-handler jumps01:04
freemangordon67:          0      INTC  ssi_p1_mpu_irq001:04
DocScrutinizer05if you can't debug that pin, just debug another one ;-)01:04
freemangordonI see in init code IRQ handler attached01:05
freemangordonoh, got it01:05
freemangordonbut I think this will be just a waste of time01:05
freemangordonBTW in dsp driver there is a code laying with SSI clocks01:06
freemangordon*playing01:06
DocScrutinizer05I could tell you a lot about "set this bit, clear that bit2 but it's useless since not applicable for the level you're working on01:06
DocScrutinizer05and I have no clue about all those kernel macros used for IRQ and the like01:07
freemangordonin case you are interested, http://pastebin.com/ji3m5He301:08
freemangordongpios01:08
DocScrutinizer05I definitely can debug and hack a kernel driver, but I can't write one, or find the huge missing elephant01:08
freemangordonMUX config http://pastebin.com/8P1e7M1801:09
DocScrutinizer05I wouldn't even notice it's missing01:09
freemangordonthe elephant? :)01:09
freemangordonclocks http://pastebin.com/PR4Vn5Au01:10
DocScrutinizer05yep, the elephant. I don't know where it's supposed to show up01:12
DocScrutinizer05but I will notice the peanut that's laying there despite it should've been eaten. Never heard it's elephant to eat it01:13
DocScrutinizer05yay, MUCH better: http://pastebin.com/raw.php?i=PR4Vn5Au (the format)01:15
*** Martix_ has joined #maemo-ssu01:16
freemangordonyep01:16
DocScrutinizer05felt nausea from seasickness from http://pastebin.com/PR4Vn5Au01:16
DocScrutinizer05wtf is "rate"?01:17
DocScrutinizer05oh nm01:18
freemangordonHz01:18
freemangordonfrequency01:18
DocScrutinizer05sure01:18
DocScrutinizer05no idea01:19
*** dhbiker has quit IRC01:19
DocScrutinizer05except.... use F-Bus to get modem bootlog01:20
DocScrutinizer05;-)01:20
freemangordonwell, the only thing I didn't play with is regulator consumers01:20
DocScrutinizer05there are regulators in linux for BB5?01:21
freemangordonDocScrutinizer05: I am sure modem works (cawake changes)01:21
freemangordonsee schematics, most of power supplies are connected to GAZOO01:21
*** dhbiker has joined #maemo-ssu01:21
DocScrutinizer05sure, but gazoo I'd think is under control of BB5 firmware, not APE01:22
freemangordonyes, but power is controlled by omap (via twl)01:22
DocScrutinizer05main modem power, yeah prolly01:22
DocScrutinizer05some "CMT_EN" or whatever01:23
freemangordonso if linux *thinks* the regulator is not used, it shuts it down01:23
freemangordoncmt_en is a gpio01:23
DocScrutinizer05I think it shuts whole modem down01:23
DocScrutinizer05BB5 is a complete ARM system01:23
DocScrutinizer05quite similar to OMAP301:24
DocScrutinizer05GAZOO == GAIA01:24
freemangordonno, I mean - the hard way, by shutting the regulators down01:24
freemangordonlook at p11 in schematics01:24
DocScrutinizer05the regulators are not even connected to OMAP I'd guess01:24
freemangordonVANA,VIO, V!* etc01:24
freemangordonV1801:24
freemangordonyes, but are controlled by omap01:24
freemangordonand this is the whole idea behind hwmod DT iiuc01:25
DocScrutinizer05how are they controlled when not connected? per RF?01:25
DocScrutinizer05or IrDA?01:25
freemangordonhwmod/DT01:25
freemangordonregulators are controlled by kernel via TWL01:25
freemangordoniiuc if usage count drops to 0 , the regulator is shut down01:26
freemangordonthe same goes for clocks01:26
DocScrutinizer05I can't see a single data line from OMAP to GAZOO01:26
freemangordonthere is no01:27
freemangordonif we dont count gpios as data lines ofc01:27
DocScrutinizer05then you can't control the regulators in GAZOO from OMAP01:27
freemangordonyes, but VANA,V18,VIO, etc are *inputs* to GAZOO iiuc01:28
DocScrutinizer05they are controlled by RAPU like the regulators in GAIA are controlled by OMAP01:28
DocScrutinizer05nope01:28
freemangordonsure?01:28
DocScrutinizer05those are power line *outputs* from GAZOO01:28
freemangordonoh01:28
freemangordonVANA is output from GAZOO?01:29
freemangordonok, if you say so01:29
*** Martix_ has quit IRC01:29
freemangordonhmm, coorect01:29
freemangordon*correct01:29
freemangordonthese are outgoing arrows01:30
DocScrutinizer05and no VANA anywhere outside BB5 domain (except camera)01:30
DocScrutinizer05those are the power supplies fro BB%01:30
DocScrutinizer05501:30
DocScrutinizer05dang01:30
freemangordonok, got it :)01:30
freemangordonhmm, well...01:31
freemangordonNFC where to look then :(01:31
DocScrutinizer05check the single signal lines from OMAP to RAPU01:32
DocScrutinizer05check the (one?) databus from OMAP to RAPU01:32
freemangordonI don't have HF scope here01:33
DocScrutinizer05that's all that connects those two worlds01:33
DocScrutinizer05is'l <10 lines01:33
DocScrutinizer05it's01:33
freemangordonnot to say I don;t want to disassemble my devel n90001:33
DocScrutinizer05Imeant check linux kernel what it does to those lines, compare to 2.6.2801:34
freemangordonthe same driver works with 3.501:34
DocScrutinizer05strace cmt01:34
DocScrutinizer05or libisi01:34
freemangordonDocScrutinizer05: it doesn;t work on lower layer01:35
freemangordonBOOT INFO REQ fails. Or rather receives no answer01:35
freemangordonAnd I think the reason is because IRQ is not triggered01:35
DocScrutinizer05yes, maybe. but anything not working on lower layer should throw an erro on higgher layer. Great diagnose tool01:36
freemangordonwell, I have the errors, something like - "timeout trying to power the modem", etc01:36
DocScrutinizer05strace working system01:36
freemangordonno need, sscd (the one that powers the modem up) traces what it does in syslog01:37
freemangordonwhne started with -d 301:37
DocScrutinizer05timeout trying to power modem most likely means the APE waits for a "Hello, here I am!" from modem after doing cmt_en01:37
freemangordonyep01:37
DocScrutinizer05you need to set up your interface prior tp powering up modem, otherwise "hello, here i am" gets lost.01:38
DocScrutinizer05and maybe you need to send a "^Q" to make the modem start to talk at all01:39
DocScrutinizer05^Q == $random-init-sequence01:39
DocScrutinizer05for coldflash bootloader it's "." afaik01:40
DocScrutinizer05ascii 2."01:40
DocScrutinizer05"." even01:40
DocScrutinizer05a series of dots01:40
DocScrutinizer05which tells bootloader that a "terminal" is attached01:40
DocScrutinizer05for BB5 I guess you need to follow ISI specs on how to establish a communication between APE and BB501:41
DocScrutinizer05wirelssmodemapi stuff01:42
DocScrutinizer05http://www.cncmods.net/files/01:42
dos1wireless modem... lego? :)01:43
DocScrutinizer05http://www.cncmods.net/files/Wireless%20Modem%20API%20G1%20V2%2010w49.zip01:43
DocScrutinizer05initializing that ISI interface is not exactly trivial01:44
freemangordonhmm, wait, wtf?01:44
* freemangordon is going to check something :)01:44
jon_yDocScrutinizer05: why are the wireless docs on a game modding site? :)01:45
DocScrutinizer05it's not like BB5 is supposed to send some plaintext ascii "hello!" via a 192kb UART01:45
DocScrutinizer05jon_y: that's a misconception01:45
jon_yhow so?01:46
DocScrutinizer05in reality they are on wirelessmodemapi.com01:46
jon_yoh ok, vhost01:46
jon_ywell, not vhost01:46
DocScrutinizer05if you want to call it a vhost, I guess we're fine with that :-)01:46
jon_ydns pointint to the same01:46
DocScrutinizer05I'd call it jonwil's mirror01:47
jon_yhe does CnC part time>01:47
jon_y?01:47
DocScrutinizer05*shrug*01:47
DocScrutinizer05~wtf CnC01:47
jon_yjust curious how it ended up sharing01:47
infobotGee...  I don't know what CnC means...01:47
jon_yCommand and Conquer01:48
jon_yin this case CnC3/401:48
DocScrutinizer05sounds OT01:48
jon_ythe site is about game modding01:48
DocScrutinizer05so what?01:48
dos1jon_y: I'd guess that's his page, or he has some access to it, and just used it to store those files, nothing to see here01:48
dos1except some cool lego stuff :)01:48
jon_yjust curious how it manage to share the site together with wireless modems01:49
dos1it's just a personal mirror01:49
jon_yprobably01:49
dos1for sure :P01:49
DocScrutinizer05wirelessmodemapi.com is down since years01:50
DocScrutinizer05actually I guess since decease of meego01:51
*** Martix_ has joined #maemo-ssu02:03
DocScrutinizer05freemangordon: the SSI input buffer may be filled, thus the "ready" signal to BB5 not set, thus BB5 not sending new stuff, thus no IRQ associated with new stuff, thus no handler ever cleaning/flushing the input buffer02:03
*** M4rtinK has joined #maemo-ssu02:16
*** Pali has quit IRC02:20
*** M4rtinK has quit IRC02:25
*** oldtopman has joined #maemo-ssu02:28
*** xes has quit IRC02:47
*** Martix_ has quit IRC03:03
*** nox- has quit IRC03:13
*** jonwil has joined #maemo-ssu03:21
*** DrCode has quit IRC03:22
*** dos1 has quit IRC04:10
*** X-Fade has quit IRC04:15
*** X-Fade has joined #maemo-ssu04:16
*** DrCode has joined #maemo-ssu04:16
*** mkaindl has left #maemo-ssu04:25
*** LauRoman has joined #maemo-ssu04:41
*** freemangordon has quit IRC05:04
*** freemangordon has joined #maemo-ssu05:04
*** LauRoman has quit IRC05:18
*** freemangordon has quit IRC05:31
*** freemangordon has joined #maemo-ssu05:31
*** amiconn_ has joined #maemo-ssu05:57
*** amiconn has quit IRC05:57
*** amiconn_ is now known as amiconn05:57
*** dhbiker has quit IRC06:22
*** n900-dk has joined #maemo-ssu07:21
*** n900-dk_ has quit IRC07:21
*** oldtopman has quit IRC07:42
*** dafox has joined #maemo-ssu07:56
*** dafox has quit IRC08:37
*** dion_ has joined #maemo-ssu08:58
*** Vlad_on_the_road has joined #maemo-ssu09:08
*** mkaindl has joined #maemo-ssu09:14
*** Vlad_on_the_road has quit IRC09:17
*** mkaindl has quit IRC09:45
*** mkaindl has joined #maemo-ssu09:45
*** Pali has joined #maemo-ssu09:54
freemangordonPali: how nice: "new patch, new email" :(10:03
freemangordonWTF is with those guys?!?10:03
Palino idea10:03
freemangordonmaybe you should start writing to linus directly :D10:04
PaliI think I will copy that hunk from previous mail to new and write somethig like guy XYZ was not able to do that10:05
Palifreemangordon: do you remember why mass storage support for g_nokia.ko was rejected?10:05
PaliI think that all usb maintainer do not have everything in head...10:06
freemangordonhmm, no. Wasn't it you that tried to upstream it?10:06
Paliyes I tried to upstream and it10:06
Paliand it was rejected because "only nokia can do any change in his g_nokia.ko driver"10:07
freemangordonoh10:07
freemangordonwhat is that supposed to mean10:07
freemangordon?10:08
Palithis means that they do not accept any change to g_nokia.ko driver10:08
Palimaybe only from somebody with @nokia.com email address10:08
freemangordonBTW even with your patches it doesn't work10:08
freemangordonPali: but how is that possible?10:08
Paliit worked with 3.9 or 3.810:08
Palibut in 3.10 is g_nokia.ko totally broken10:09
freemangordonhmm10:09
Paliand I disabled lot of g_nokia.ko code because it I got always kernel panic10:09
Paliusb networking is still enabled and worked10:09
freemangordonit doesn't work for me10:09
PaliI used only rescueos10:10
Paliand manually loading driver + setting ip address it worked10:10
freemangordontried on both Win XP and Ubuntu 12.04 (my desktop PC and my laptop)10:10
Palibtw, now because usb subsystem (and musb) has bug which cause that charging drivers are non functional, that my patch for usb is required10:10
Paliotherwiese battery charging will not work10:10
freemangordonPali: I really don;t understand, this Tony guy was rejecting smc3 patchset for an year, now everything is fine, he says "because of DT move, bla, bla"10:12
PaliTony did not have time...10:12
Paliand he wait until arm guys accept it10:12
freemangordon"This file will be gone as soon as we're moving to device10:13
freemangordontree based booting. So let's do this in more future proof10:13
freemangordonway.10:13
Paliand only now Dave added: Acked-by10:13
freemangordonthis ^^^10:13
PaliI know10:13
freemangordonI know you know :)10:13
freemangordonsailus: ping10:14
freemangordonno use :)10:15
freemangordonBTW I see no way board code to go. Because of the cameras10:16
Palifreemangordon: can you write some good looking "education" email to that usb patch? Because If I send this which I'm writing now it will not be very good :-)10:17
jonwilhttp://wiki.maemo.org/Porting/Audio :)10:17
Palicopy paste full second hunk of patch to new mail is stupid10:17
jonwilwrote some stuff on audio for Neo90010:17
freemangordonYou means a a reply?10:18
freemangordon*as a10:18
Paliyes10:18
freemangordonby "education" you mean to share my thoughts on the matter?10:18
Paliwrite what do you think10:19
freemangordonok10:19
freemangordonbut later, not now10:19
Paliok10:19
freemangordonhave to run10:19
jonwilPali: Feel free to read http://wiki.maemo.org/Porting and add anything you got :)10:20
Paliok10:20
jonwilAudio in particular :)10:21
jonwilWe need to find someone who has really good pulseaudio knowledge10:21
*** dhbiker has joined #maemo-ssu10:28
*** M4rtinK has joined #maemo-ssu10:53
*** sarha has joined #maemo-ssu10:56
Palifreemangordon: now I'm fed up with that usb guy and going to send not very good looking email10:59
*** dagb has quit IRC11:00
*** dagb_ has joined #maemo-ssu11:00
*** mkaindl has left #maemo-ssu11:03
*** Martix_ has joined #maemo-ssu11:07
*** tg has quit IRC11:27
*** tg` has joined #maemo-ssu11:28
* jonwil is glad he doesn't do kernel stuff11:28
jonwilthat way I dont have to deal with all the mess going on11:28
*** tg` is now known as tg11:29
Palijonwil: I think only usb developers are assholes11:29
jonwilheh11:29
PaliI sent patches to more other subsystems and I never had any problem11:29
Palionly two patches are rejected and both usb11:30
jonwilIts clear now that audio will be the hardest part of the Fremantle porting efforts11:31
*** M4rtinK has quit IRC11:31
Palinow I sent email for that usb guy (also CCed mailinglists)11:34
* Pali waiting for ban on all mailinglists11:34
* jonwil wishes he knew more about PulseAudio and ALSA11:37
* jonwil wishes more people cared about the11:45
* jonwil wishes more people cared about the FPTF :(11:45
PaliFatPhil_: do you know somebody who can still use @nokia.com email address for sending kernel patches?11:47
Palihaving @nokia.com address is enough, nothing more is needed11:47
PaliFatPhil_: usb subsystem maintainers rejected patch for g_nokia.ko usb gadget driver because "only nokia can change this part"11:48
Paliand also they did not understand that nokia is not going to update and use linux g_nokia.ko usb gadget anymore...11:49
keriofreemangordon: apt wants me to reinstall python2.5 from the repo :(11:54
keriowhy didn't you add a +thumb to the version11:54
FatPhil_No results found for "only nokia can change this part".11:56
FatPhil_1 result found for "Guys, WHY ARE YOU SO STUPID AND ARROGANT?"11:56
keriothe fuck is ke-recv-extra11:57
kerioD:11:57
*** freemangordon has quit IRC11:57
keriowtf is hulda11:57
*** freemangordon has joined #maemo-ssu11:57
PaliFatPhil_: https://lkml.org/lkml/2013/1/21/6611:58
PaliFatPhil_: "so unless someone from Nokia says this is how they want to use nokia.c from now on"11:59
Paliand here is patch: https://lkml.org/lkml/2013/1/19/16512:00
FatPhil_I can't simply risk12:01
FatPhil_breaking all other users for your own convenience.12:01
Paliit adding optional/additional support for usb mass storage mode12:01
Paliit is composite usb gadget12:03
Paliand does not break anything if you do not add any device to export12:03
FatPhil_Then if it wasn't for Felipe's final sentence, I'd say it belongs in a new separate file12:04
FatPhil_But if they're moving gadget drivers out of the kernel, it's a step in the wrong direction.12:04
Paliall gadget drivers are still in kernel12:05
Paliand I think they will not be removed, because it break userspace and anything...12:05
jonwilAt least the N900 kernel isn't as bad as the kernel from my old Motorola Z6 linux phone which was a mess12:08
jonwilthis was long before Android12:08
jonwilgotten real quiet in here all of a sudden :P12:40
*** Martix_ has quit IRC12:41
*** iDont has joined #maemo-ssu13:11
*** freemangordon has quit IRC13:16
*** freemangordon has joined #maemo-ssu13:16
freemangordonPali: nice reading :).13:28
*** oooaaaooo has joined #maemo-ssu13:36
*** oooaaaooo has left #maemo-ssu13:36
*** DrCode has quit IRC13:50
*** DrCode has joined #maemo-ssu13:53
freemangordonFatPhil_: IDK if you follow n900 upstreaming efforts, but honestly, it looks to me kernel maintainers don;t wan't to take n900 patches for some political reasons. For sure I do not support "the conspiracy theory" but something is fishy there13:59
jonwilPerhaps they are uninterested in taking (and then having to maintain) patches that the original maintainers/creators have abandoned14:02
jonwilanyone here know of someone who has pulseaudio skills? :P14:03
*** dos1 has joined #maemo-ssu14:03
Palijonwil: ask lennart directly :P14:15
jonwil:P14:18
jonwilin any case I think we need to get more interest in FPTF from anyone out there who can contribute :)14:20
freemangordonjonwil: I give a shit what they are interested in, N900 is (and will be for years to come) the only true linux mobile computer.14:20
*** arcean has joined #maemo-ssu14:21
freemangordonthere is no other consumer device like that14:21
jonwilAnd Neo900 is intended to provide all the good things from the N900 and then a bit more (like potentially better cellular support)14:22
jonwiland like a USB port that wont fail if you look at it wrong14:23
dos1"the only true linux mobile computer"14:24
dos1you made my Neo Freerunner sad :P14:24
dos1sure it's much weaker and without fancy cameras, sensors and what else, but it's still "true linux mobile computer"14:25
*** BCMM has joined #maemo-ssu14:29
keriodos1: your Neo Freerunner is older than the N900!14:30
kerioand the N900 is an obsolete piece of crap!14:30
keriothink about it!14:30
dos1so if it's old, it's not true linux anymore? :P14:31
kerionope14:31
keriotrue linux has systemd!14:31
dos1I have it for more than 5 years already14:31
freemangordondos1: my point was: "consumer device", not geek's gadget :P14:31
kerioand pulseaudio, for speech synth for logins14:31
keriobecause of accessibility14:31
dos1kerio: surprise! my Freerunner has systemd! :D14:31
kerio...dafuq14:31
dos1and it had pulseaudio on Om2007.2, but on SHR we don't use that shit :P14:32
jonwilN9 is a "true linux mobile device"14:32
freemangordonyeah, sure14:32
keriowhy the hell does SHR use systemd? D:14:32
freemangordonjonwil: what do you use as your daily phone?14:33
jonwilN90014:33
jonwilits my only phone14:33
freemangordonwhy you didn't buy n9?14:33
jonwilbecause I like hw kbd14:34
jonwiland because I have no need to buy a new phone when the phone I have works just fine14:34
dos1kerio: well, dunno actually. OE has systemd support, and it makes boot faster for low cost, so maybe that's it14:34
jonwilnot like all those Apple sheeple who have to have the latest and greatest always :)14:34
freemangordon:)14:34
jonwiljust look at the crap that is the iPhone 5S and 5C :P14:35
freemangordonjonwil: and how aegis fits with the "true linux mobile device" statement?14:35
dos1is 5 years already enough to start thinking about helping with creating new device and getting it many months later? :)14:36
dos1I feel kinda hipsterish14:36
dos1heck, I'm probably the last gta02 user in my country14:36
dos1aegis does not fit at all into any "true linux device". N9 is Android-tier for me14:38
freemangordonmy point exactly14:39
freemangordonPali: could you spend some time later on, teaching me how to send patches to the ML. It is too much of a work for you to do it alone and now I have a laptop with ubuntu, I guess it won;t be much of a hassle for me to set it up as it needs to be and help you14:41
Palifreemangordon: no problem14:41
freemangordonok, I'll ping you when I am back home, maybe i 3 hours14:41
freemangordon*in14:41
Paliok14:41
freemangordonmaybe we can convince FatPhil_ to help too :P14:42
Palihm, Sunday, March 24 2013 Sakari Ailus created new rx51-009 branch in omap3camera14:44
PaliI did not know about it14:44
PaliI used camera patches from rx51-007 branch14:45
Palihttps://gitorious.org/omap3camera/rx51/activities14:45
*** sunny_s has joined #maemo-ssu14:47
Palifreemangordon look at rx51-009 branch ^^^^14:47
sailusPali: pong.14:48
Palisailus: hi14:48
sailusI noticed you'd pinged me some time ago. -EBUSY. :-P14:49
Palisailus: freemangordon trying to get n900 cameras to work14:49
Palibut he only see green output14:49
Palifor both back & front14:49
sailusI haven't tried out the camera on the N900 for some time.14:49
Palido you know where can be problem?14:49
sailusDo you know how did he test them and which kernel did he use?14:50
sailusI haven't updated the patches for quite some time, but I recently rebased them.14:50
sailusThey probably don't work.14:50
Palihe used my modified 3.10 tree14:50
Paliand used mediactl for setting connections14:50
Paliand mplayer14:50
sailusWhich entities were involved?14:51
sailusThe resizer can't be used in that setup.14:51
Palisailus: my 3.10 tree contains patches from your branch rx51-007 (+ modified)14:51
sailusAnd... the interfaces that the et8ke8 driver uses are mostly obsolete.14:51
sailusI don't remember if it even uses the pad ops.14:52
Palisailus: he tested both front and back cameras and both green output14:52
sailusProbably it does, but it's definitely not entirely up to date in terms of the V4L2 subdev API.14:52
PaliI'm going to find pastebin of mediactl outputs14:52
sailusI've tested the smiapp driver on N950 and N9.14:53
sailusNot on N900. Although it should work.14:53
Palisailus: http://pastebin.com/st0phiUm14:54
Paliother pastebin exired...14:54
Paliso for more info we need to wait until freemangordon will be back14:54
sailusLooks fine for me.14:55
*** dhbiker has quit IRC14:55
sailusI'd still use yavta just to have less modifiers in the setup compared to a known working one.14:56
*** dhbiker has joined #maemo-ssu14:56
sailusyavta is quite nice and simple.14:56
Paliwhat is yavta?14:57
sailusYet another V4L2 test application.14:58
freemangordonsailus: hi14:58
sailusfreemangordon: Hi!14:58
freemangordonsailus: first of all I had to fix omap3isp code :)14:58
sailusWas there something wrong with it? :-)14:58
freemangordonyes, clk frequency calculation14:59
sailusHas the patch already been submitted to mainline?14:59
sailusI thought there were some uncertainties regarding the 3430 clock rate.14:59
sailusSo those have been resolved now?15:00
freemangordonno, recently we have some, errr... problems with mainline :)15:00
sailus:-P15:00
freemangordonanyway , I'll send the patch soon15:00
sailusCool! :-)15:00
freemangordonbut the main problem is that all I get is green rectangle15:00
sailusHave you tested using yavta?15:01
freemangordonwith no error messages in dmesg15:01
freemangordonno15:01
sailusI'd suggest to do that first to minimise the difference to a known working setup.15:01
freemangordonthis is what I use:15:01
sailusWhich is N9(50) + yavta.15:01
freemangordonsailus: the other thing I had to fix is to restore the xshutdown callback15:02
sailusI'd also use only as much of the pipeline inside the ISP that you actually need to, i.e. capture it straight from the CCP2 receiver.15:02
sailusThat's wrong.15:02
freemangordontried that, no difference15:02
sailusThere should be GPIOs...15:02
sailusAlbeit the hardware is quite strange. It's visible in the schematics.15:02
sailus:-P15:02
sailusPlease try with yavta again.15:03
sailusThe last time I checked mplayer it used V4L1. :-)15:03
freemangordonsailus: well, maybe I don;t know enough, but I dont see how you'll switch the mux gpio without having the callback15:03
sailusBut that was a long time ago.15:03
freemangordonin the correct way that is15:03
sailusI think there would need to be another driver that provides a single gpio that actually translates to two.15:03
sailusAnd that driver would be specific to N900.15:04
freemangordonand who'll call this driver?15:04
freemangordonsome userspace?15:04
sailusIt'd provide a single gpio, I suppose.15:04
sailusBut I don't know the GPIO framework well --- whether it'd even allow doing that.15:05
freemangordondo you have rx51 schematics near you? to see how the cameras are connected15:05
sailusActually toggling the gpio is the responsibility of the sensor driver.15:05
sailusI think they can be easily found using Google.15:05
sailusI don't think I have them around myself.15:05
freemangordonxshutdown gpio of front camera is used to switch between cameras too15:07
freemangordonand I can't see how this can be workarounded having a separate driver15:08
freemangordonunless this driver (or board code) is called from smiapp and et8ek815:09
freemangordonsailus: ^^^15:10
freemangordonthus my "restore xshutdown callback" patch15:10
sailusThe GPIO API is how you'd call it.15:11
Palifreemangordon: if you have some uncommited/unpushed changes in your tree, can you push them to gitorious?15:12
freemangordonsailus: i have a terrible lag as I am connected through gprs, I'll ping you again later when I am back home and try to explain all over why "separate driver" approach won't work15:13
freemangordonPali: I think I pushed all camera patches, lemme check15:13
*** dagb_ is now known as dagb15:13
Paliok15:13
sailusPali: I don't think I have any.15:14
sailusThe GPIO framework seems to be fine btw.15:14
sailusgpiochip_add() and so on...15:14
*** iDont has quit IRC15:15
Palisailus: problem is that n900 has two gpios, one is for switching between cameras and second is for turning camera on/off15:15
Paliwhile omap3isp support only on/off gpio15:15
Paliand cannot interact between more cameras15:16
sailusI don't think it's a problem anymore if you use another driver to toggle them.15:16
sailusThe sensor driver will still see a single one.15:16
Paliso to create new virtual gpio?15:16
Paliand handle changes of this gpio to read two gpios?15:16
Palis/read/real/15:17
infobotPali meant: and handle changes of this gpio to real two gpios?15:17
Paliand how to force omap3isp to allow only one camera to be turned on at same time?15:17
sailusAlthough the GPIO framework seems to assome that the GPIOs are independent. That's not quite the case here.15:18
sailusThey already use the same bus.15:18
sailusThe omap3isp driver handles that already.15:19
sailus(And the mux is switched using on of these gpios.)15:19
Paliok, but in platform data we need to specify gpio which is used for on/off not switch gpio15:20
PaliI still do not understand how to handle it15:20
sailusA new driver showing up as a GPIO chip will request these two GPIOs.15:22
sailusBoth drivers should access only the newly exported GPIO.15:22
sailusHmm.15:23
Paliah, so as I wrote create new virtual GPIO15:23
sailusThat was my though, yes.15:23
Palibut this is overkill15:23
freemangordonwe'd rather need 2 gpios15:23
sailusI actually wonder whether the switch GPIO should be controlled by the ISP driver, not a sensor driver.15:24
sailusIt switches the CSI bus between the sensors.15:24
sailusThe other one is the xshutdown GPIO.15:24
sailusThe solution that's currently in use is somewhat workable but probably will never be accepted to mainline.15:25
freemangordonPali: 1 virtual gpio would not suffice, as smiapp does gpio_get... and does not free it15:25
sailusI think this will need some experimentation.15:25
freemangordonso et8ek8 gpio_get will fail. or vice versa15:25
Paliso two new virtual gpios15:26
*** LauRoman has joined #maemo-ssu15:26
freemangordonPali: sailus: having a driver with 2 virtual gpios will do the job with the current framework imo15:27
PaliI think that omap3isp should be fixed to handle this problem15:27
Paliand not to be too creative and create new virtual gpio driver for omap3isp15:28
freemangordonso instead of having xshutdown callback call, we'll have xshutdown gpio15:28
Palifreemangordon: first use callback and when everyting else start working, then we can start writing new gpio driver....15:28
freemangordonooh, sure :)15:29
PaliI think that higher priority has non working camera, non working modem, non working audio and not upstreamed charging15:29
freemangordon:nod:15:29
Paliand for me new gpio driver for funny omap3isp has lower priority15:30
freemangordongitorious is FUBAR again :(15:30
freemangordonPali: we should move to github ;)15:30
Palifreemangordon: I think this is irrelevant for me :-) I'm using command line git tools15:31
*** iDont has joined #maemo-ssu15:31
Palifreemangordon: or we can ask merlin for hosting git repositories :-)15:31
freemangordonPali: I wanted to show the patches to sailus15:32
Paliah, right15:33
Palifreemangordon: use: https://gitorious.org/linux-n900/linux-n900/commit/<sha1>15:33
Palifreemangordon: or your tree: https://gitorious.org/linux-n900/freemangordons-linux-n900/commit/<sha1>15:34
freemangordonsailus: https://gitorious.org/linux-n900/freemangordons-linux-n900/commit/fd8e4d517284adc7ec1d038e5312142025bf18b715:34
freemangordonomap3isp fix for 343015:34
freemangordonthis chunk somehow disappeared after 3.1 or 3.215:35
freemangordonsailus: btw another problem with framework - you can't have more than one device connected to xclka for example :)15:36
freemangordonhttps://gitorious.org/linux-n900/freemangordons-linux-n900/commit/323664abc42aa0161cfbd750ba24e97b5b546f4f15:37
sailus"omap3isp: Set cam_mclk rate directly" --- that probably broke 3430.15:37
freemangordonyep15:38
freemangordonbut IDK if this missing divisor is the only thing broken by that patch.15:39
sailusLaurent wrote a patch to fix that.15:39
sailusWhat's your e-mail? I'll send it to you.15:39
sailusIt hasn't been sent to a public list since it's completely untested.15:39
freemangordonsailus: freemangordon at abv dot bg15:40
freemangordonsailus: fix what?15:40
sailus(Or I could push it to a public git tree.)15:40
sailusTo fix the clock back-propagation.15:40
freemangordonit is upstreamed15:40
sailusIt is?15:40
freemangordonjust a second15:41
*** oooaaaooo has joined #maemo-ssu15:41
sailusDoesn't look like that, but I could be mistaken.15:41
sailusI have linux-media tree here.15:42
sailusI doubt Laurent would accept it. :-)15:42
oooaaaooohi guys ive installed a couple of apps that dont seem to show up in the app screen; other than command line is there a graphical way to find them?15:42
freemangordonsailus: https://gitorious.org/linux-n900/freemangordons-linux-n900/commit/7b2e1277598e4187c9be3e61fd9b0f0423f9798615:42
oooaaaoooalso is "faster app manager" any good?15:42
freemangordonoooaaaooo: ECHAN, try on #maemo15:43
sailusThat's for 3630 only.15:43
sailusA similar change is required for 3430.15:43
*** oooaaaooo has left #maemo-ssu15:43
freemangordonoh, right, my bad15:44
freemangordonhttps://gitorious.org/linux-n900/freemangordons-linux-n900/commit/92d6ac650784d9e4cdf9e5d219aa1060fae081cd15:44
freemangordonnot upstreamed15:44
freemangordonsailus: see, my tree has all the needed fixes, drivers load without errors, etc15:46
freemangordonso I guess it is either omap3isp or MUX config15:46
freemangordonto blame that is15:46
freemangordonbut I have NFC where and what to look for :)(15:47
freemangordons/:)(/:)/15:47
infobotfreemangordon meant: but I have NFC where and what to look for :)15:47
sailusHere:15:47
sailushttps://gitorious.org/omap3camera/rx51/commit/0878c44382021a8eaf870458c3805f8997356d1215:47
sailusfreemangordon: With that, you shouldn't need to hack anymore.15:48
freemangordonisn't it the same as my patch, just using DEFINE_STRUCT_CLK_FLAGS instead of defining the struct?15:49
freemangordonsailus: ^^^15:51
*** iDont has quit IRC15:52
*** sunny_s has quit IRC16:00
*** arcean has quit IRC16:01
sailusfreemangordon: What matters here is the CLK_SET_RATE_PARENT flag. It was missing.16:01
*** sunny_s has joined #maemo-ssu16:01
Palisailus: another question: et8rk8 driver using some "firmware", but when I looked into source code of firmware it was only some kernel C structure, it is true?16:02
freemangordonsailus: it was not, in my tree :) https://gitorious.org/linux-n900/freemangordons-linux-n900/commit/92d6ac650784d9e4cdf9e5d219aa1060fae081cd16:04
freemangordon.flags = CLK_SET_RATE_PARENT,16:04
sailusPali: true.16:05
*** RST38h has quit IRC16:06
Palisailus: old smia-sensor driver used also some firmware, but new smiapp not using, how it was fixed?16:06
Palistructures was moved to driver?16:07
sailusThe "firmware" is not really firmware but hand-written register values that correspond to some sensor configuration.16:07
sailusIn the smiapp driver the same configuration is passed using a standard IOCTL interface.16:07
sailusThe et8ek8 "firmware" should be moved back to the driver.16:08
Palisailus: so we need some configuration for smiapp driver to work?16:08
sailusIt has a default configuration which is functional.16:10
sailusNot necessarily what you really want, but a working one.16:10
freemangordonPali: once we have that working, we can forward-port resolutions from omap1/KP16:10
sailusSo you have the corresponding patch to Laurent's. Do you still need the other change?16:10
freemangordonsailus: I think I have everything in my tree, it uses clock frameworks, drivers load and init the HW, etc16:11
freemangordonnow I cloned and compiled yavta, will try it when i am back home16:11
sailusAck.16:12
freemangordonsailus: unless you think there is something else needed16:12
freemangordonsailus: please look at the last 10 or so commits here https://gitorious.org/linux-n900/freemangordons-linux-n900/commits/4e272703339f84e65c063c8969e5a69652c0dbda16:13
freemangordondone by "Ivaylo Dimitrov"16:13
freemangordonall those are camera/isp fixes16:14
*** RST38h has joined #maemo-ssu16:18
*** oldtopman has joined #maemo-ssu16:24
*** arcean has joined #maemo-ssu16:27
sailusfreemangordon: Thanks!16:29
freemangordonsailus: hmm? what for?16:32
freemangordonsailus: I don't think those could be upstreamed, I just did a quick hacks to have the drivers working. for example - the correct way to handle multiple devices sharng the same xclk is to use list16:35
freemangordonnot an array16:35
freemangordonI'll make a patch for that when i have the cameras working :)16:35
*** arcean_ has joined #maemo-ssu16:42
*** arcean has quit IRC16:46
*** Martix_ has joined #maemo-ssu16:55
Palifreemangordon: I moved et8ek8 firmware into kernel driver16:59
Paliso driver does not need external "firmware" from userspace16:59
*** Martix_ has quit IRC16:59
freemangordonPali: great :)17:00
freemangordonPali: does it work? :P17:00
Palionly tested compilation17:00
Palibut it must work17:01
*** Martix_ has joined #maemo-ssu17:01
PaliI just replaced pointer from firmware to included structure17:01
freemangordonsure, I was kidding17:01
* freemangordon is on his way home, bbl17:01
*** oldtopman has quit IRC17:02
*** iDont has joined #maemo-ssu17:03
*** dos1 has quit IRC17:05
*** dos1 has joined #maemo-ssu17:07
freemangordonsailus: any recommended cmd line parameters to use for yavta?17:51
*** Martix_ has quit IRC18:05
freemangordonsailus: yavta hangs in ioctl(3, VIDIOC_DQBUF...18:05
*** Martix_ has joined #maemo-ssu18:08
*** NIN101 has joined #maemo-ssu18:18
*** BCMM has quit IRC18:25
*** Martix_ has quit IRC18:32
sailusfreemangordon: That likely mean you don't get any buffers back from the driver.18:46
sailusDo you get interrupts?18:46
freemangordonany idea which gpoi/irq to check?18:47
freemangordon*gpio18:47
freemangordonsailus: BTW:18:47
sailus/proc/interrupts18:47
freemangordon:)18:47
freemangordonI know that, but which particular gpio18:47
sailusomap3isp, I guess.18:47
freemangordonbtw:18:47
freemangordoncam_xclka 1           1            576000018:47
freemangordonis that normal? the frequency I mean18:48
sailusThat looks very low to me.18:48
sailusWhat does it have in the platform data?18:48
freemangordon40:          0      INTC  OMAP3 ISP18:48
sailusNo interrupts then.18:49
sailusSomething definitely is wrong.18:50
freemangordon.ext_clk= 9.6 * 1000 * 1000,18:50
freemangordon.op_sys_clock= (s64 []){ 12000000 * 10 / 2, 0 },18:50
sailusOn both?18:50
freemangordonthis is for smiapp18:50
sailusext_clk is the one you should be concerned about.18:50
sailusAre you using the smia one?18:51
sailusOk.18:51
freemangordonsmia has it hardcoded in the driver iirc18:51
sailusIt doesn't.18:51
sailusIt comes from the platform data.18:52
freemangordon#define ET8EK8_XCLK_HZ960000018:52
freemangordonunsigned int hz = ET8EK8_XCLK_HZ;18:52
freemangordonrval = clk_set_rate(sensor->ext_clk, hz);18:52
freemangordonsailus: ^^^18:53
sailuset8ek8 isn't smia compatible.18:54
freemangordonwell, it is in i2c/smia dir :)18:54
freemangordonanyway18:54
sailusThe problem could be related to discrepancies in how the divisor is chosen.18:55
PaliI can move it from i2c/smia to i2c/18:55
sailusThe information on the parent clock frequency could be wrong.18:55
freemangordonsailus: I saw similar bug on clock framework for SSI clocks18:55
sailusThat frequency must be exactly 9,6 MHz.18:55
freemangordon:nod:18:55
sailus:-P18:55
* freemangordon is going to check how it is on Nokia kernel18:56
sailusThe frequency is the same.18:56
sailusThat has not changed.18:56
freemangordonI meant in /sys/kernel/debug/clk18:56
sailusWhat has changed.... well, is the clock framework.18:56
sailusAck.18:57
freemangordonwhat is reported there and what is the clock tree18:57
sailusExcellent18:57
freemangordonNokia-N900:~# cat /sys/kernel/debug/clock/virt_19_2m_ck/osc_sys_ck/sys_ck/dpll4_ck/dpll4_m5_ck/dpll4_m5x2_ck/cam_mclk/rate19:00
freemangordon17280000019:00
*** dos1 has quit IRC19:00
freemangordonthis is on 2.6.2819:00
*** dos1 has joined #maemo-ssu19:01
freemangordonon 3.10:19:01
freemangordonNokia-N900:~# cat /sys/kernel/debug/clock/virt_19_2m_ck/osc_sys_ck/sys_ck/dpll4_ck/dpll4_m5_ck/dpll4_m5x2_ck/cam_mclk/rate19:01
freemangordon17280000019:01
freemangordonshit19:01
freemangordoncam_mclk     0           0            17280000019:01
freemangordonbut I don;t have xclka child of cam_mclk on 2.6.2819:02
freemangordonhmm, maybe I should revert my divider patch19:03
*** dos1 has quit IRC19:08
*** dos11 has joined #maemo-ssu19:08
*** dos11 has quit IRC19:10
*** dos1 has joined #maemo-ssu19:10
*** dos1 has quit IRC19:16
*** dos1 has joined #maemo-ssu19:16
*** dion_ has left #maemo-ssu19:18
freemangordonsailus: hmm, with that patch reverted I see "unexpected cam_mclk rate:.." from isp driver19:19
Palifreemangordon: is there any other problem except green output?19:19
freemangordonyep, yavta hangs while trying to get video buffer19:20
freemangordonand there are no interrupts from ISP19:20
Palibecause maybe there is one commit which you can try to revert19:20
*** LaoLang_cool has joined #maemo-ssu19:20
Pali450460412ce758f79d02e9f5db720aa09bd8e6ad19:20
freemangordonwhich one?19:20
*** LaoLang_cool has quit IRC19:21
PaliI do not know if this commit it in upstream or not, but sailus send me it months ago via email19:21
freemangordon"smiapp-pll: Take existing divisor into account in minimum divisor check"?19:21
Paliyes, can you check if this commit is in upstream?19:22
Palisailus, can you comment above your patch? ^^^19:22
Paliand there is another commit which changing some color matrix: 3532af7ad45e54e000dfa7f2b7536e7a58ecc61d19:22
Pali"isppreview: new default coeffs for more ambient independent quality"19:23
Palifreemangordon look (and maybe try to revert it) ^^^^19:23
freemangordonPali: ok19:30
freemangordonPali: reverting 450460412ce758f79d02e9f5db720aa09bd8e6ad made it worse :)19:47
Paliok, so that patch is really needed :-)19:48
freemangordonyep19:48
Paliwithout it driver loading failed, right?19:48
freemangordonyep19:48
freemangordoncan't find pll div/mul19:49
Palisailus should comment if that patch is correct ^^^19:49
freemangordonI can try to hack it and feed pll calculations with doupled frequency :)19:50
Palifreemangordon: look also at second commit19:51
freemangordonPali: frequency is incorrect, it should be 9.6Mhz not half of that19:52
sailuscam_mclk isn't in the clock framework in 2.6.28 --- the clock is produced by the ISP itself, not PRCM.19:54
freemangordonsailus: no idea who produces it, but it has the same frequency on 2.6.28 and 3.10, according to /sys/kernel/debug/...19:55
sailusThat's a good sign then.19:55
sailusIs the divisor also the same?19:55
sailusBased on what we discussed a few hours ago the xclk frequency seemed wrong.19:56
freemangordonhow to check the divisor on 2.6.28, any idea?19:56
freemangordonthis is my daily device :)19:56
freemangordon(the one with 2.6.28)19:57
sailus:-)19:57
sailusI'm not sure if the driver prints it.19:57
freemangordonwait, can't we just divide mclk freq with 9600000 to get the correct diviso?19:57
sailusYou might need to use printk() to print it instead.19:57
sailusBut... you can calculate it as well.19:57
freemangordon:nod:19:58
sailus:-)19:58
sailusThis assumes the frequency really is what the clock framework suggests.20:00
sailusNow that the parent clock handling bug has been supposedly fixed, I guess that's ok.20:00
freemangordonhmm, could it be that clock framework reports bogus values for xclk?20:02
freemangordonbecause it is tweaked for 3630, not 343020:02
sailusNot really tweaked, but there was the bug that the divisor between cam_mclk and its parent wasn't accounted for on either one.20:03
sailusAnd only 3430 really worked.20:03
freemangordonwell:20:04
freemangordoncam_mclk     0           0            17280000020:04
freemangordonthis looks fine to me20:04
freemangordonthe same on 2.6.2820:04
freemangordoncam_xclkb 0           0            576000020:04
freemangordonis deffinitely wrong, lets check the divisor20:04
sailusJust by quick check, you should have 18.20:05
sailusAnd... why's cam_xclkb?20:05
sailusBoth cameras use xclka.20:05
freemangordon:nod:20:06
freemangordonsailus: copy/paster error20:06
*** Vlad_on_the_road has joined #maemo-ssu20:06
freemangordonbut xclka is the same20:06
sailusThe divisor would be 30 in that case.20:06
sailusThat's a strange number.20:07
freemangordonsailus: http://pastebin.com/UgdYDUQn20:08
freemangordonNFC at which value to look at :)20:09
DocScrutinizer05hah, clock divisors. real fun20:09
DocScrutinizer05no surprise when new kernel msses it up20:10
*** jonwil has quit IRC20:10
freemangordonsailus: "mul 25 / div 2"?20:11
sailusAh --- now I think I remember.20:12
sailusThe case was that the original configuration from the "firmware" used slightly off-spec frequencies.20:12
sailusAnd that might not be taken into account in the current quirks in smiapp driver.20:12
sailusIt isn't.20:13
freemangordonthere are no quirks for vs655520:13
freemangordon:)20:13
sailusThat's the probable reason for the PLL calculator failure.20:13
*** bsdmaniak has joined #maemo-ssu20:14
freemangordonbut it doesn't fail, just calculates wrong values IMO :)20:14
freemangordonsailus: can I just hardcode the correct values?20:14
freemangordonto prove/not if this is the real problem20:14
sailusIndeed --- if it doesn't fail, then that message can be ignored.20:15
sailusPerhaps it was another mode.20:15
*** dos11 has joined #maemo-ssu20:16
sailusYou're right --- if there was a failure, you'd see this:20:16
sailus        dev_info(dev, "unable to compute pre_pll divisor\n");20:16
freemangordonwhich I get if I revert your "smiapp-pll: Take existing divisor into account in minimum divisor check" patch20:16
*** dos1 has quit IRC20:16
freemangordonsailus: btw, which value in the log I should look at? for divisor that is20:17
sailusThat's the sensor clock tree calculation.20:18
sailusIf it succeeds you should be fine.20:18
sailusWhat matters whether the ISP produces the clock it should to the sensor.20:19
sailusThe cam_xclka frequency suggests it does not.20:19
freemangordon:nod:20:19
sailusBut why, is unknown to me.20:19
freemangordonsailus: which is the value from all those I have to look at?20:19
sailusIt should "just work". :-P20:19
freemangordonheh20:19
freemangordon:)20:19
sailusNone.20:20
sailusext_clk is just input to the PLL calculator.20:20
sailusNothing comes out.20:20
sailusI'd figure out why the divisor is 30 and not 18.20:20
freemangordonwhere you saw that 30?20:20
sailus172800000/576000020:20
freemangordonwhere did you see even :)20:20
freemangordonwell, is it in dmesg log?20:21
sailusI'm quoting you from 20:04. :-)20:21
sailus17 minutes ago.20:21
sailusI need to go; I might be back later today.20:22
freemangordonsure, but do we have that value logged from the pll calculator? or it is the clock framework that calculates it?20:22
freemangordonsailus: ok20:22
sailusSorry.20:22
sailusThat's internal ISP divisor.20:22
sailusThe sensor driver has no part in that.20:23
freemangordonok20:23
sailusxclk = cam_mclk / divisor AFAIR.20:23
freemangordon:nod:20:23
freemangordongoing to check in the clock framework20:23
*** M4rtinK has joined #maemo-ssu20:23
*** Martix_ has joined #maemo-ssu20:25
*** WielkiTost has joined #maemo-ssu20:26
sailusYou should check what the ISP driver does and what frequencies it sees. It calculates the divisor.20:26
freemangordonok, will do20:26
freemangordonthanks!20:26
*** dos11 has quit IRC20:27
*** WielkiTost is now known as dos120:29
freemangordonsailus: I hardcoded 18 in isp_xclk_calc_divider, still the same. And still no ISP interrupts20:40
freemangordonbut at least: cam_xclkb 0           0            960000020:41
freemangordonand cam_xclka 0           0            960000020:41
*** dagb has left #maemo-ssu20:42
*** bsdmaniak has quit IRC20:54
*** bsdmaniak has joined #maemo-ssu20:57
*** bsdmaniak has quit IRC20:58
*** bsdmaniak has joined #maemo-ssu21:13
*** bsdmaniak has quit IRC21:23
*** iDont has quit IRC22:16
*** iDont has joined #maemo-ssu22:18
*** iDont has quit IRC22:23
sailusIt'd be good to know the reason *why* the divisor isn't 18.22:37
DocScrutinizer05HAH! http://projects.goldelico.com/p/openphoenux/page/PaidDeveloper/22:45
DocScrutinizer05Didn't even know of that :-)22:45
*** NIN101 has quit IRC22:51
freemangordonsailus: I bet it is the clock frameworks that screws it23:02
freemangordonAs iiuc dpll4_m5x2 on 3630 runs on a different frequency compared to 343023:03
freemangordonbut there is no difference in cclock3xxx_data.c23:03
*** LauRoman has quit IRC23:03
*** amiconn has quit IRC23:06
*** amiconn has joined #maemo-ssu23:08
DocScrutinizer05dnag. Is that a result of sw dudes writing code to rerpesent hw? ;-)23:10
DocScrutinizer05when you write such definitions, you have to start at the very atoms of a hw design, like crystal frequencies etc23:11
DocScrutinizer05but that knowledge is rarely found in CS-students' heads, while every EE is aware of that23:13
DocScrutinizer05given the EE also develops sw23:14
DocScrutinizer05I heard some CS dudes also develop hardware, and those might actually know, but I guess that's quite exotic a critter to be discovered anywhere23:15
*** Vlad_on_the_road has quit IRC23:17
DocScrutinizer05I guess there's no such thing like "crystal frequency" in DT, eh? ;-)23:26
DocScrutinizer05the fun thing in such stuff: when you have a (simplified) hw like ocillator * multiplier1 / divider1, there may be a range of 10..100 valid for multiplier and a range of 1..16 for divider, but yet some particular combinations of the two factors are explicitly forbidden23:30
DocScrutinizer05and the rules are not as simple as "for multiplier >50, no dividers <10 allowed"23:31
DocScrutinizer05in the end you usually need a table with available output frequencies and the multiplier and divider values you need to use for them. I guess that's a hard thing to implement in DT23:32
DocScrutinizer05to make stuff less boring, the different clocks in system are often entangled to each other in complex ways23:34
DocScrutinizer05like "RAMbus clock *2^N == CPUclock * 2^M" with N,M := int23:35
DocScrutinizer05err23:38
DocScrutinizer05like "RAMbus/4 clock *2^N == CPUclock/3 * 2^M" with N,M := int23:38
*** arcean_ has quit IRC23:39
DocScrutinizer05e.g for the S3C2442 this been a terrible pain in the aft section, for scaling frequency of CPU, since all other clocks based on it and simply changing e.g UART clock on the fly is a really poor idea23:42
*** sunny_s has quit IRC23:42
DocScrutinizer05also we were not able to set RAMbus speed to the max our RAM would allow, without either under- or over-clocking the CPU23:43
DocScrutinizer05wow, TV is really giving insights: 70% of US-americans believe in a physically existing hell23:46
FatPhil_"US-americans"?23:47
DocScrutinizer05I guess at least 50% believe that earth is 5578 years old, and man been made by God, 70 generations ago23:47
FatPhil_yup, those are the US americans23:48
DocScrutinizer05if only those 70% would see the truth, which is: they are _living_ in their hell, every day. they made it themselves23:49
Palifreemangordon: I pushed et8ek8 changes to v3.10-n900 branch on gitorious23:53
*** nox- has joined #maemo-ssu23:54
freemangordonPali: good23:57
PaliI removed code which can oops kernel :-)23:57
freemangordonhmm23:57
* freemangordon is going to check23:57
Paliet8ek8 could oops kernel if you get it working :-)23:58
PaliNULL derefernce23:58
Paliconfigure_interface23:58
Paliit was removed from isp driver long time ago23:58
PaliI removed also platform_data->set_xclk because it is not used23:59

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!