**** BEGIN LOGGING AT Mon Jun 25 02:59:58 2012 Jun 25 07:23:08 heyho Jun 25 08:23:03 Good day :-) Jun 25 08:41:03 Are the embedded drivers for the smartphone hardware reversed, or used as binaries; who is managing the reversing if so? Jun 25 08:43:36 heth: about which kind of smartphone hardware are you talking about? Jun 25 08:43:47 in general Jun 25 08:44:03 that depend Jun 25 08:44:24 I can guess that on android devices you can use the binaries, but not on the others I suppose Jun 25 08:44:50 heth: that really depends on different kind of factors Jun 25 08:45:03 there is no general answer Jun 25 08:46:21 morphis: hi Jun 25 08:46:23 morphis, I understand. Is anyone specialized in reversing and building drivers for the ones that doesn't have other options? Jun 25 08:46:35 morphis: which is the best and latest libsamsung-ipc and android ril source? Jun 25 08:47:16 heth: I guess you can take a look at various drivers from non documented software and see who commit on it ? Jun 25 08:49:04 Alex[sp3dev]: use https://github.com/morphis/libsamsung-ipc for libsamsung-ipc Jun 25 08:49:11 for android ril I don't know Jun 25 08:49:40 heth: what do you have in mind to do? Jun 25 08:51:14 morphis, misc, I'm more looking for some summary on how are the things working, for example if there have been a better solution except reversing every driver; I'm trying to build the big picture on the FOSS smartphone development progress in the last 2-3 years Jun 25 08:51:50 well, that's a quite complex topic Jun 25 08:52:08 you have some documentation on some hardware Jun 25 08:52:23 you have some code for some smartphone thanks to gpl compliance efforts Jun 25 08:53:08 and there is some work to do reverse engineering ( see for example lima , for mali graphical chipset ) Jun 25 08:53:17 As it was some years ago, the biggest problem was there was a lack of specifications and disasembly people as well as tools for ARMs, while the android was still getting started Jun 25 08:53:30 I see. Jun 25 08:53:34 IMHO, the biggest hurdle is the lack of free graphical drivers Jun 25 08:53:41 heth: the short summary is that we have the userland ready but almost no finished device because as soon as one port is finished (which takes like 2 years) the device gets obsolete.. Jun 25 08:53:53 the mosts things are already available thanks to android theses days Jun 25 08:53:58 misc: well, most things, except meego compositor and kwin, work without graphics Jun 25 08:54:02 just somethings like graphics or modem support are missing Jun 25 08:54:37 with samsung and TI opening up drivers and pushing stuff like video cameras upstream, it is getting easier to support all hardware Jun 25 08:54:39 Alex[sp3dev]: yep, maybe hurdle is not the word I should have used :) Jun 25 08:54:50 There was a big issue with power saving, is that a holdback still? Jun 25 08:54:51 a big problem today is that smartphone companys are not following the std. way to do things Jun 25 08:55:02 on some cases Jun 25 08:55:09 Qualcomm is still a big mess Jun 25 08:55:14 morphis: i have an essay on that ;) Jun 25 08:55:16 http://allsoftwaresucks.blogspot.com/2012/06/linux-gone-wrong.html Jun 25 08:55:20 samsung and TI are quite well supported upstream Jun 25 08:55:23 yeah, it didn't change much, I guess :) Jun 25 08:55:43 :) Jun 25 08:55:53 heth: for some vendors , yes Jun 25 08:55:57 look at qualcomm Jun 25 08:56:06 but I'm happy to hear people are openning drivers and pushing for documentation, that is really what is needed Jun 25 08:56:11 but as I said TI and samsung are already quite nice Jun 25 08:56:25 they are doing a lot of upstream devleopment Jun 25 08:56:37 and also an the graphics side things are going forward Jun 25 08:57:13 lets wait one or two years maybe we will have then one or two open source graphics drivers published Jun 25 08:57:43 heth: actually all talks about freedom are a bit hypocritical unless you build the hardware from scratch. modern processors usually contain the non-accessible boot rom in the one-time-programmable memory for securely booting the chain of trust.. and given how small electronics are, you never know what the microcode in a device may do Jun 25 08:58:29 Alex[sp3dev]: does reading your essay require javascript? Jun 25 08:58:46 Alex[sp3dev]: well, i can benefit from freedom without building my own hardware Jun 25 08:58:55 Alex[sp3dev], I think that at some point there will be a longer cycle where the smartphones are useful still from 2y to like 4-5 when the hardware slowers the pace, so the embedded linux will catch up:) Jun 25 08:58:59 PaulFertser: that's a standard blogger.com blog, I dunno. If you have paranoia, I can copy-paste it at pastebin Jun 25 08:59:17 Alex[sp3dev]: i do not have paranoia, i'm just using emacs-w3m usually. Jun 25 08:59:42 PaulFertser: I see. Jun 25 08:59:44 if I take the example of my work phone running android, if the kernel could have been freely recompilable, I could run openvpn to read my mail, I can't recompile ( or I didn't tried hard enough ), so I am not free to read my mail :) Jun 25 09:00:01 PaulFertser: here you are http://pastebin.com/pR5qrtCY Jun 25 09:02:48 Alex[sp3dev], I see. I can guess after a good useful/solid release gets enough momentum, there will be good open hardware for it like the freerunner Jun 25 09:04:31 heth: if we look back to, say, 7 years, there was almost no PDA linux. There were OPIE, GPE and Qtopia, of course, but the only well supported SoC was PXA270 and the only fully supported PDAs were Sharp Zaurus (linux from stock) and iPaq 4700 (reversed by a group of paid programmers by some company) Jun 25 09:05:36 morphis, do you have any rough idea what % of the hardware actually have to be reversed at the moment? Jun 25 09:06:35 puuuh, reverse engineering is not the only problem you have Jun 25 09:06:53 Alex[sp3dev]: and even hx4700 wasn't fully functional, e.g. its power consumption in suspend was several times higher than with the stock fw. Jun 25 09:07:01 Alex[sp3dev]: i think that company was SDG systems. Jun 25 09:07:14 I would say if you have a common device with chips from TI or Samsung there are only the DSP, graphics and modem drivers that needs reverse engineering Jun 25 09:07:46 Alex[sp3dev]: a really nice essay, thanks! Jun 25 09:09:21 morphis: I would say, DSP is not essential, NEON can handle most media just fine, but graphics are a dead end, they need too much time even for simple 2d, not speaking of opengl Jun 25 09:09:23 Alex[sp3dev]: though i'd assume that it's not possible to teach monkey-coders to do anything right. My guess is that's the management at fault because they hire incompetent people to do the work. So one needs to educate managers to hire nice folks, and they'll do the right thing on their own. Jun 25 09:09:49 morphis, sounds really good, compared to what it was. :) Jun 25 09:10:03 heth: yes Jun 25 09:10:06 Alex[sp3dev]: yes Jun 25 09:10:38 Alex[sp3dev]: but need to mention DSP to complete the list (when we're takling about everything and not only essential things) Jun 25 09:14:49 morphis, so the phones are not dying in a day now on battery? Jun 25 09:15:56 my Nexus S with android has a one day usage with a full battery Jun 25 09:22:51 morphis, I mean while not in use with an embedded linux:) Jun 25 10:10:08 moo Jun 25 10:13:46 phone battery depleting during <20h, in idle, usually is caused by any sort of daemon hogging the system, that has not been designed to run on embedded Jun 25 10:14:49 and of course by drivers for e.g. WLAN that do things like constant scanning for APs, instead of playing nice and powering down the wlan hw for 9 out of 10 minutes Jun 25 10:15:31 things like audio amp enabled all the time while there's actually nothing to playback. etc pp Jun 25 10:16:23 or - very simple - a poor freq-governour implementation that never suspends the CPU to zero-clock Jun 25 10:17:31 (the last one not exactly an issue with the [imho faulty by design] suspend-to-ram concept of original OM/SHR) Jun 25 11:02:25 hi there Jun 25 11:03:34 i've cleaned the dust from my freerunner, and i want to flash it to a newer shr release Jun 25 11:04:25 i'm using aurora-image-om-gta02-20120507000408.rootfs.ubi, uImage-2.6.39-r12-om-gta02-20120525082154.bin and qi-ubi-s3c2442-1.0.2-r9+gitr2+aa934abb5e8771100240e344411ddebacca33ece.udfu Jun 25 11:05:37 flashing goes well, but my FR doesn't boot : i get a black screen, only showing a line about "dubgfs dir creation failed" Jun 25 11:05:43 can anyone help? Jun 25 11:15:26 KaZeR_W: that line is normal Jun 25 11:17:54 KaZeR_W: try the shr-image Jun 25 11:18:07 KaZeR_W: try holding in the power button to get a more verbose boot: http://wiki.openmoko.org/wiki/Qi#Enabling_console_messages Jun 25 11:18:15 aurora is not maintained afaik Jun 25 11:23:15 ah, i took the latest image in the repo, and it was aurora :) Jun 25 11:31:40 pabs3, at 9s, i get infos about the headset jack. then at 97,184,275,365s it says it's starting udev version 102 Jun 25 11:31:55 at 275s, it also inserted a module ar6000 Jun 25 11:32:04 i will try with shr-image Jun 25 11:32:12 sounds like it booted fine? Jun 25 11:35:44 yes, but nothing else on the screen :) Jun 25 11:42:58 i'm currently flashing shr-image-om-gta02-20120611074304.rootfs.ubi Jun 25 12:40:56 ok using shr-image it booted, but the usb networking doesn't seem to work Jun 25 12:42:11 KaZeR_W: I think the IP address has changed, not sure what it is now tho Jun 25 12:42:36 does lsusb on laptop/desktop give you a line about the phone? Jun 25 12:42:52 pabs3, http://wiki.openmoko.org/wiki/USB_Networking#Openmoko_Networking_Setup says 192.168.7.1 for the host Jun 25 12:43:05 i do get the usb0 device when i load cdc_ether Jun 25 12:44:11 sounds like udev is missing/broken on your laptop/desktop? Jun 25 12:50:56 i'm not sure it's udev-related Jun 25 12:51:04 interesting, i just noticed that in my logs : Jun 25 12:51:05 Jun 25 14:24:17 gihqdfa1 kernel: [19606.482101] cdc_ether 8-1:3.8: usb0: register 'cdc_ether' at usb-0000:00:1d.2-1, CDC Ethernet De Jun 25 12:51:05 vice, 02:80:37:14:03:00 Jun 25 12:51:05 Jun 25 14:24:17 gihqdfa1 kernel: [19606.488093] cdc_ether 6-1:1.0: eth1: register 'cdc_ether' at usb-0000:00:1d.0-1, CDC Ethernet De Jun 25 12:51:05 vice, 00:1f:11:01:42:e9 Jun 25 12:51:12 i get both usb0 and eth1 :) Jun 25 12:51:59 but, by habit, i configured usb0. i will try again with eth1, but i had to reflash first, i misconfigured the display size in the wizard, icons were way too small Jun 25 15:41:05 KaZeR, on some staging images of shr usb0 on GTA02 side needs to be reset ifup usb0;ifdown usb0 Jun 25 15:50:50 *sigh* someone should really tell Martin Jansa not to do git rebases on public branches as it changes history and makes peoples' lives hard if they follow said branch Jun 25 15:55:50 thanks jluis Jun 25 15:56:21 after a reflash (i broke the display settings ;) ) i got access to the device when configuring eth1 (instead of usb0) on my computer Jun 25 15:57:12 then i rebooted the phone, and network never came back. and i was unable to access the home menu (with all the icons) i only got a empty desktop (so no terminal, i was unable to check if the network device was up at all) Jun 25 18:13:04 KaZeR_W, sorry I had to leave, if the screen has the status bar (home, keyboard,[L/R],square, Wrench) buttons with the time ... Jun 25 18:16:08 pressing the square icon (that cycles one screen,vertical split, horizontal split) will reload the apps icons Jun 25 19:06:40 jake42: ping Jun 25 20:11:23 GNUtoo-hplaptop: hi Jun 25 20:11:30 hi Jun 25 20:11:40 what's up Jun 25 20:12:42 I've been busy with coreboot+flashrom+replicant lately Jun 25 20:13:24 i started hacking galaxy nexus modem. seems similiar to galaxy s2 (also xmm6260) but firmware loading is different, I've only managed to load PSI firmware (there are 4 parts remaining). If I manage to finish it tomorrow, I will get back to integrating with libsamsung-ipc. I've started doing it, but need to make some traffic dumps Jun 25 20:15:01 http://pastebin.com/UZixggzd here's the current log. and it's already 1.5k C code. next time I'll probably use vala or c++. or even haskell, dunno. want to experiment Jun 25 20:15:01 ok wow Jun 25 20:15:34 ok Jun 25 20:19:25 GNUtoo-hplaptop: what do you think about reverse-engineering a GPU driver? seems like a good way to start learning opengl Jun 25 20:19:56 GPU isn't strictly necessary for now, so maybe once the rest would be done it would be great Jun 25 20:20:07 maybe with android 4.0 it is tough Jun 25 20:20:16 which gpu ? Jun 25 20:20:19 what's the GPU? Jun 25 20:20:25 powervr? Jun 25 20:20:30 or mali? Jun 25 20:20:33 GNUtoo-hplaptop: I actually think it might be a good idea to try doing it just to learn how it works Jun 25 20:20:36 which one do you want to reverse? Jun 25 20:20:41 ok Jun 25 20:20:53 yes that may be fun if you're into that kind of things Jun 25 20:20:54 GNUtoo-hplaptop: I have all of them (qcom, tegra, mali, powervr) so I can choose. probably powervr or mali Jun 25 20:21:21 i think both are already being worked on Jun 25 20:21:31 yes but he can still help Jun 25 20:21:36 maybe powervr would be better? Jun 25 20:22:22 but, first I need to finish RIL and integrate S2 to SHR. and then need to improve my electronics skills. but I think I'll be into some opengl programming this summer and will probably start looking at reversing powervr. Jun 25 20:22:25 yep, I mean starting from scratch would be IMHO less efficient that to work with someone :) Jun 25 20:24:11 ok Jun 25 20:28:26 it's a pity there's no decent foss decompiler/disassembler. makes everything complicated Jun 25 20:39:29 seems like powervr has more code in kernel than in userland compared to mali. either way, that's a very large area to explore **** ENDING LOGGING AT Tue Jun 26 02:59:58 2012