irclog2html for #openezx on 20060415

01:04.44TimRikerk
08:56.49*** join/#openezx stefan_schmidt (n=stefan@p54865E8B.dip.t-dialin.net)
09:53.46*** join/#openezx LaF0rge (i=laforge@sunbeam.ipv6.gnumonks.org)
10:15.28stefan_schmidtLaF0rge: If it boots, i can help to test. :)
10:17.28LaF0rgestefan: thanks.  I'll set up my BTUART console later again in order to see how far the booting gets (it could be pxafb related why I don't see anything)
10:17.51LaF0rgebut now I need to buy myself a new wristwatch.  need one before leaving for brasil on tuesday early morning
10:22.19stefan_schmidtLaF0rge: Who needs the exact time? ;)
10:23.08stefan_schmidtHope you have a nice time in brasil.
10:26.39stefan_schmidtbtw: i don't have a wristwatch for the last 6 years now. :)
11:36.51LaF0rgeok, 0,1mm magnet wire wasn't the best idea for connecting the phone UART
11:48.45*** join/#openezx ysakaed (i=intoxica@60.234.51.252)
11:52.14jonwilanyone here?
12:24.11LaF0rgeyes
12:31.43ysakaedhi
14:17.53jonwilhmmm, I have an idea for how to proceed with the bootloader stuff
14:18.18jonwilSomeone should get hold of the documentation for the CPU in the EZX phone and find out where in RAM it starts executing from
14:18.39LaF0rgejonwil: the docs can be freely downloaded ;)
14:18.49LaF0rgeand all the address information should also be present within blob source code
14:18.49jonwiland then read that area (a couple megs from there perhaps) out
14:19.01jonwiland then reverse engineer that
14:19.15jonwilIf there is a layer of code below BLOB (and I think there might be), we could find out that way
14:19.27LaF0rgei have full flahs dumps somewhere here on my disk
14:20.20jonwilif there does exist a lower-layer than blob, we can find that out (and maybe even reverse engineer it if we can find someone that knows ARM assembly)
14:22.27jonwilAlso, all the various variables we are seeing (like that flag) are either set by the lower-level core
14:22.28jonwilcode
14:22.30jonwilif any
14:22.38jonwilor if not, they are set by the other CPU
14:23.07jonwilanyhow, if we reverse engineer the flash stuff, we can at least locate exactly where in memory BLOB is physically located
14:23.15jonwiland confirm once and for all where the phone looks for it
14:23.24jonwilditto kernel
14:24.25jonwilI cant do this since I know zilch about ARM assembler (I know x86 and a bit of stuff on 65816 and a little 68000 but no ARM)
14:27.38*** join/#openezx blkhawk (n=blkhawk@p548A3C79.dip0.t-ipconnect.de)
14:38.44LaF0rgecool. I just flashed a phone while having STUART connected to two testpins.  ramldr prints a lot of debug output there ;)
14:54.33*** join/#openezx ao2 (n=u@2001:1418:117:0:0:0:0:1)
15:18.07stefan_schmidthi
15:23.14stefan_schmidtLaF0rge: News from 2.6 port? Have you got a kernel with serial console working?
15:29.41LaF0rgestefan: unfortunately neither of that
15:29.52LaF0rgethe 2.4.x kernel hangs during boot if i say console=ttyS2
15:34.44stefan_schmidtCONFIG_CMDLINE="root=/dev/ram0 init=/linuxrc initrd=0xa0400000,0x400000 ip=off console=ttyS2, 115200n8"
15:35.07stefan_schmidtThis comes out of arch/arm/def-config/ezx780.initrd
15:35.34stefan_schmidtezx-780.ramdisk.
15:36.01stefan_schmidtDid you set the speed?
15:43.10LaF0rgestefan: i tried with and without speed setting
15:43.19LaF0rgei'm even disabling lots of stuff in the driver now
15:43.21LaF0rgewithout success
15:44.52stefan_schmidtit's a pity
15:45.22LaF0rgeI'm even trying to manually add uart initializaiton to head.S
15:46.47stefan_schmidtah, your fight with arm assembly.
15:49.31stefan_schmidtwithout uart setup i could not help you within this problem. :(
15:50.34LaF0rgethe interesting thing is that even console=tty0 or console=vc/0 breaks booting
15:50.43LaF0rgeso the problem might not even be serial related
15:50.49LaF0rgebut something about the console switch
15:53.51LaF0rgealso, for some strange reason (massive includes?) compiling the 2.4.x kernel takes about 10 times longer than 2.6.x (same compiler)
15:54.14LaF0rgeso most of the time i spend waiting for the compiler to finish
15:58.01stefan_schmidt:)
15:59.11LaF0rgenow it doesn't want to boot at all anymore :(
15:59.24LaF0rgethat motorola kernel never gets the compile dependencies right
16:54.35LaF0rgefsck. now the kernel doesn't boot anymore
16:54.37LaF0rge*sigh*
17:04.50stefan_schmidtbacktracking ;)
17:09.08LaF0rgestrange
17:09.19LaF0rgemy kernels don't always boot
17:09.36LaF0rgesometimes they do, sometimes they just print crap on the framebuffer
17:11.04stefan_schmidtodd
17:11.07LaF0rgeok, the same for your kernel binary
17:11.13stefan_schmidtwow
17:12.02stefan_schmidtsomething goes wrong in blob and or hw initialization here?
17:12.27LaF0rgedon't know. booting from flash always works
17:13.25LaF0rgeok, it works reliably after booting the flash kernel again
17:13.29LaF0rgemaybe some hardware bogosity
17:16.44stefan_schmidtI haven't seen this here so far. But good to know. Perhaps to many kernel loads in short time. :)
17:17.03LaF0rgeno, i think it's something in my code that got the hardware stuck
17:17.12LaF0rgeand blob maybe doesn't initialize everything into a clean state
17:17.31LaF0rgethat combined with too little time without battery for reset
17:18.06LaF0rgeanyway, I recommend you work further on the reblob issue
17:18.27LaF0rgelooking at the start.S code, it _should_ work if you load it to the other address (was it a0100000?)
17:19.58LaF0rgestart.S will copy the second stage to a0000000 (up to a0020000), so there is no overlap/collision
17:20.55stefan_schmidt#define BLOB_RAM_BASE               (0xA0100000)
17:21.28LaF0rgeexactly. try loading to that address with boot_usb
17:21.53stefan_schmidtok, i'll look again into it. Perhaps i did some mistake yesterday.
17:22.06LaF0rgeok, now 'console=vc/0' works, it was my 'broken hardware state', not the kernel
17:22.10LaF0rgedid you already try to load it to that address/
17:22.11LaF0rge?
17:22.18LaF0rgeI thought you only tried a0200000
17:22.24LaF0rgeand what happened?
17:23.43LaF0rgeI mean, how would you notice a working reblob?  in theory, the screen output should stay the same, and it should again react to usb commands, just like before the reblob
17:23.59stefan_schmidti tried both. with no result.
17:24.25stefan_schmidtI changed The version string from 3.0 to 3.1 set ne date. :)
17:24.31LaF0rgeok
17:25.21LaF0rgei must be missing something about uart initialization
17:25.45stefan_schmidtI modified your blob_usb:
17:25.48stefan_schmidt--- boot_usb.c  2006-04-04 01:23:23.000000000 +0200
17:25.48stefan_schmidt+++ boot_usb_blob.c     2006-04-14 22:12:17.000000000 +0200
17:25.48stefan_schmidt@@ -211,7 +211,7 @@
17:25.48stefan_schmidt<PROTECTED>
17:25.48stefan_schmidt<PROTECTED>
17:25.48stefan_schmidt<PROTECTED>
17:25.50stefan_schmidt-#define KERNEL_RAM_BASE        0xA0200000
17:25.52stefan_schmidt+#define BLOB_RAM_BASE  0xA0100000
17:25.54stefan_schmidt<PROTECTED>
17:25.56stefan_schmidt<PROTECTED>
17:26.00stefan_schmidt<PROTECTED>
17:26.02stefan_schmidt@@ -266,9 +266,9 @@
17:26.04stefan_schmidt<PROTECTED>
17:26.06stefan_schmidt<PROTECTED>
17:26.08stefan_schmidt<PROTECTED>
17:26.10stefan_schmidt-       ezx_blob_load_program(KERNEL_RAM_BASE, prog, st.st_size);
17:26.12stefan_schmidt+       ezx_blob_load_program(BLOB_RAM_BASE, prog, st.st_size);
17:26.14stefan_schmidt<PROTECTED>
17:26.16stefan_schmidt-       ezx_blob_cmd_jump(KERNEL_RAM_BASE);
17:26.18stefan_schmidt+       ezx_blob_cmd_jump(BLOB_RAM_BASE);
17:26.20stefan_schmidt<PROTECTED>
17:26.22stefan_schmidt<PROTECTED>
17:26.24stefan_schmidt<PROTECTED>
17:26.26stefan_schmidtWhat is vc/0?
17:26.38LaF0rgethe framebuffer virtual console
17:26.45LaF0rgeYAY
17:26.48LaF0rgeserial c onsole working
17:27.35LaF0rgeLinux version 2.4.20_mvlcee30-mainstone (BJDC@LINUX) (GCC 3.x for XSCALE) #1 Jan 1,2003
17:27.35LaF0rgeCPU: Intel XScale-Bulverde revision 7
17:27.35LaF0rgeMachine: Motorola Ezx Platform
17:27.35LaF0rgeRun Mode clock: 195.00MHz (*15)
17:27.38stefan_schmidtcool
17:27.46LaF0rgeI even have a shell
17:27.53stefan_schmidtreally cool
17:28.04blkhawkO_o
17:28.10blkhawkcongratulations
17:28.12LaF0rgeso half the day I was actually suffering from my bad hardware state, not from anything else
17:28.16stefan_schmidtSo we can debug why it crashes after some seconds.
17:28.23stefan_schmidt:(
17:28.43stefan_schmidtBut you get it now. Congrats.
17:28.45LaF0rgenow the question is why does my 2.6 kernel not run
17:29.14blkhawkwhats needed for a serial cable?
17:30.15LaF0rgeblkhawk: you need to solder thin wires on two testpads of one of the phone PCB's
17:30.26LaF0rgeblkhawk: nothing you want to do to a phone that you still want to carry mobile ;)
17:30.40blkhawki will pass then
17:30.53blkhawkI thought you could somehow switch the usb plug
17:32.23LaF0rgeshould also be possible with some work, but it's more complicated than what I did ;)
17:33.39LaF0rgehttp://svn.gnumonks.org/trunk/a780/doc/stuart-bootlog.txt
17:36.26LaF0rge*rotfl* : XScale Cache/TLB Locking Copyright(c) 2001 MontaVista Software, Inc.
17:37.16LaF0rgeso the phone doesn't actually crash after a couple of seconds
17:37.21LaF0rgeit remains active on the serial port
17:37.38LaF0rgei guess it's jsut the gui apps  that somehow do a blocking read on somme communication to the BP (which doesn't work)
17:40.55stefan_schmidtyes that sound logic.
17:44.48ao2great.
17:46.35ao2So LaF0rge you can boot your own kernel on your dev phone. Was that strange "hadware state" thing that made your kernels to not boot in the past?
17:47.25LaF0rgeyes
17:47.33LaF0rgebut still not my 2.6.x port
17:59.40ao2ah, hoping you can fix that too.
18:09.06*** join/#openezx stefan_ (n=stefan@p548649F8.dip.t-dialin.net)
18:37.48stefan_schmidtNo news on the blob front. *sigh*
18:39.15stefan_schmidtI tried it all again: Load blob, blob-rest-elf32, blob-start-elf32 and blob-start-elf32 catted together with blob-rest-elf32.
18:40.01stefan_schmidtAll of these to both KERNEL_RAM_BASE and BLOB_RAM_BASE.
18:40.55stefan_schmidtThe only interesting is blob and blob-start-elf32 blanks the screen and blob-rest-elf32 and blob-elf32 do nothing.
18:42.37stefan_schmidtThe behaviour is the same for both addresses.
19:27.01webmindpossibly wrong source ?
19:41.21stefan_schmidtwebmind: That would mean that Motorola delivered the wrong source.
19:41.49webmindthat's what I'm saying yes
19:43.23stefan_schmidtOf course we could not exclude this, but prefer to think it's another prblem at the moment.
19:45.23stefan_schmidtWe need more comprehension about blob before we can write to moto again.
19:48.52stefan_schmidtI need contact to blob devs to get answered some questions. But ml is down and the projects seems really dead at all.
19:49.58stefan_schmidtPerhaps TimRiker could help us.
19:51.31stefan_schmidtOf course i can alos not exclude that i make the mistake.
19:56.04webmindok
19:56.21webmindwell we indeed cant make wild acusations
19:58.50stefan_schmidtYes, i don't have a problem talk with moto again, but i like to be sure at this point.
20:09.37*** join/#openezx kokoko (i=intoxica@60.234.51.252)
21:35.45ao2goodnight
22:29.48LaF0rge*sigh*
22:30.59stefan_schmidtLaF0rge: whats up?
22:31.12LaF0rgestefan: I'm still fighting with as/objcopy
22:31.26stefan_schmidtoh
22:31.36LaF0rgewhenever i have code that has relative jumps, those jumps get wrong destination addresses when converting from elf to binary
22:31.55stefan_schmidti stopped working on blob for today. :)
22:33.27stefan_schmidtwhy do you work on the asm uart stuff again?
22:33.50LaF0rgeyes, still
22:35.53stefan_schmidtbtw: I've seen some code to do phone call from cmd. But i don't find it now. Have a link?
22:36.56LaF0rgeseems to work now
22:37.04LaF0rgei don't have such code
22:37.11LaF0rgeit's simple ATD command, though
22:38.05stefan_schmidthmm, ok ATD should be easy.
22:38.38LaF0rgetheres some GSM/3GPP standards on the extended AT command set of mobile phones
22:38.50LaF0rgemotorola seems to have some vendor specific extensions to that, but in general it's a good start
22:44.03LaF0rgeso what I'm currently doing is uploading my own plain binary object code to print hello world and then switch off the backlight
22:44.28LaF0rgei get stuff on the serial port, but not the right characters (though the count is ok)
22:44.30stefan_schmidtlooking forward to free userspace apps Gomunicator seems to be a good basement
22:44.35stefan_schmidthttp://www.linuxdevelopment.org/projects/gomunicator.html
22:45.11stefan_schmidtBut it takes GPE and this means X.
22:45.38stefan_schmidtok, i wonder why you do as.
22:47.18SuNI have a feeling X is going to be faster than Opie/QTe.
22:47.56LaF0rgeok, now the right characters are printed
22:47.59SuNDirectfb means a lot of porting, I'd say :)
22:48.08*** join/#openezx jonwil (n=jonwil@220-244-120-190.static.tpgi.com.au)
22:48.19stefan_schmidtFor LaF0rge porting is fun. ;)
22:48.35LaF0rgei'm just not willing to compromise ;)
22:48.46SuNI guess, but there's only so much he can handle, and sticking with the standards will open a world of software.
22:49.15SuNI mean, you're free to try, just stop sleeping as of now :P
22:50.18SuNAs a viable rootfs, that is.
22:50.42stefan_schmidtYes, this would become nice. :)
22:51.02SuNIt's a bit sad... these machines have SOooo much potential.
22:51.25stefan_schmidthey, we're working on it to get it free.
22:51.40LaF0rgehttp://svn.gnumonks.org/trunk/a780/src/stuart-test/stuart.S
22:52.01SuNstefan_schmidt: yeah, and I'm both patient and grateful.
22:52.44jonwilhave you guys made any progress?
22:53.25stefan_schmidtYou know LaF0rge has a working serial console on startup?
22:53.56stefan_schmidthttp://svn.gnumonks.org/trunk/a780/doc/stuart-bootlog.txt
22:56.47stefan_schmidtSuN: Full feature 2.6, nice userspace (perhaps OPIE) and flashing complete firmware images via USB. That would become great :)
22:57.19stefan_schmidtLaF0rge: I recognize again, why i dislike as. ;)
22:57.24SuNFully agreed.
23:15.16jonwilwhat about my idea of getting the parts of the code that execute first and seeing where blob fits into it?
23:17.37stefan_schmidtjonwil: I've thinking about this idea. But my assembler knowledge is equal to zero.
23:17.55jonwilso is mine :(
23:18.00stefan_schmidtSo i've to pass here.
23:18.23jonwilDoes anyone know what keys you press on ezx to get into "bootloader mode"?
23:18.55stefan_schmidtjogdial and camera hold and power on.
23:18.58stefan_schmidton a780
23:19.21jonwilpower on for how long?
23:19.31stefan_schmidtonly short
23:19.34jonwilok
23:21.00stefan_schmidtjonwil: It seems LaF0rge is fit in arm assembler, but he is working on the 2.6 port so far and i don't think he had time to work on re blob, too.
23:42.28blkhawkn8

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with blootbot logs, split per channel, etc.