| 00:11.28 | landley | Is it jumping to the right spot? |
| 00:16.32 | luffe48 | after getting a headache looking at asm, i went the other way and started hacking the uimage c code |
| 00:16.53 | luffe48 | after changing the address it gunzips the vmlinux to it actually starts doing something |
| 00:17.27 | luffe48 | now i get some sort of pretty printed machine exception |
| 00:19.26 | clever | do you have a memory map of the device? |
| 00:19.36 | clever | maybe it was unzipping to a non-ram location? |
| 00:20.49 | luffe48 | it was just gunzipping to 0x0, start of ram |
| 00:20.56 | luffe48 | i just moved it to 32M |
| 00:53.04 | landley | luffe48: is it pretty-printed from the bootloader, or from the kernel? |
| 00:53.11 | landley | Because if it's from the kernel, you've made progress. |
| 00:53.19 | luffe48 | i don't know :( |
| 00:53.29 | luffe48 | i'm too stupid for this shit |
| 00:53.42 | landley | eh, everybody starts out ignorant. |
| 00:53.49 | landley | Don't confuse ignorance with stupidity. |
| 00:54.59 | luffe48 | First line is "NIP: 0243C3B4 XER: 00000000 LR: 02000024 REGS: 07faebf8 TRAP: 0700 DAR: 00503004", last line is "Call backtrace:", followed by a lack of backtraces |
| 00:55.08 | luffe48 | Does it look familiar? |
| 00:56.34 | luffe48 | At least the 0243C3B4 corresponds to the start of early_init |
| 00:56.42 | luffe48 | the rest i don't understand |
| 01:10.52 | *** join/#edev prpplague (~prpplague@107-206-64-184.lightspeed.rcsntx.sbcglobal.net) |
| 01:24.35 | luffe48 | has a light bulb moment |
| 01:25.23 | luffe48 | before, when vmlinux was unpacked to 0x0, it must have overwritten the exception handler, which seems to be provided by uboot |
| 01:34.04 | clever | grep the source of both u-boot and the kernel for strings in the message |
| 01:34.16 | clever | and figure out who is printing that message |
| 01:34.46 | luffe48 | uboot is, i just found one on the uboot page that looks very similar |
| 01:35.03 | clever | ./arch/powerpc/cpu/mpc5xx/traps.c: printf("NIP: %08lX XER: %08lX LR: %08lX REGS: %p TRAP: %04lx DAR: %08lX\n", |
| 01:35.16 | luffe48 | oh for the love of... |
| 01:35.16 | clever | there are a bunch of different files doing that |
| 01:35.20 | clever | ./arch/powerpc/cpu/ppc4xx/traps.c: printf("NIP: %08lX XER: %08lX LR: %08lX REGS: %p TRAP: %04lx DEAR: %08lX\n", |
| 01:35.23 | clever | one for each cpu |
| 01:35.38 | clever | clever@sakura ~/u-boot-pi $ grep -r NIP .|grep XER|grep DAR |
| 01:35.50 | clever | figure out which cpu you have, then work from there |
| 01:36.30 | clever | it looks like show_regs(struct pt_regs *regs) in the traps.c for your chip |
| 01:36.46 | luffe48 | we agree on uboot? |
| 01:37.16 | clever | yeah, clearly looks like uboot printing the message |
| 01:37.28 | clever | printf("Call backtrace: "); |
| 01:37.34 | clever | also in traps.c |
| 01:38.05 | luffe48 | from looking at some linux exception handling i've deduced TRAP: 700 should mean program check exception |
| 01:38.23 | luffe48 | now i just need to figure out what a program check exception is |
| 01:38.46 | luffe48 | at this speed, i'll have this NAS up and running by easter :) |
| 01:38.54 | clever | <PROTECTED> |
| 01:38.58 | clever | another function in traps.c |
| 01:39.16 | clever | ./arch/powerpc/cpu/ppc4xx/start.S: EXC_XFER_TEMPLATE(ProgramCheck, ProgramCheckException, |
| 01:39.19 | clever | which is referenced by assembly |
| 01:39.42 | clever | - . = 0x700 ProgramCheck: |
| 01:39.43 | clever | - EXCEPTION_PROLOG(SRR0, SRR1) |
| 01:39.43 | clever | | addi r3,r1,STACK_FRAME_OVERHEAD |
| 01:39.43 | clever | | EXC_XFER_TEMPLATE(ProgramCheck, ProgramCheckException, |
| 01:39.43 | clever | |- MSR_KERNEL, COPY_EE) |
| 01:41.49 | clever | also, i'm pasting code from a random ppc chip |
| 01:41.55 | clever | it could be the wrong one for your nas |
| 01:42.37 | luffe48 | i'm fooling around with an mpc8343 |
| 01:42.42 | clever | i would figure out which chip you have, and modify the traps.c on that one, then rebuild u-boot |
| 01:43.20 | clever | <PROTECTED> |
| 01:43.22 | luffe48 | i'm not adventurous enough to mess with uboot - i don't have any way to fix it if i blow that up |
| 01:43.29 | clever | there also appears to be kgdb support in u-boot |
| 01:43.34 | clever | which will catch these exceptions |
| 01:43.41 | clever | enable it, point gdb at the serial port, crash it! |
| 01:44.01 | clever | yeah, you would have to update the u-boot image, and that depends on how the platform is setup |
| 01:44.37 | luffe48 | i can only flash it with uboot |
| 01:45.01 | clever | do you know where u-boot is stored? |
| 01:45.09 | luffe48 | yes |
| 01:45.21 | clever | removable or soldered in flash? |
| 01:45.27 | luffe48 | soldered |
| 01:45.33 | clever | any jtag? |
| 01:45.54 | luffe48 | nope, it's powerpc, so a jtag adapter costs a fortune |
| 01:46.31 | clever | maybe u-boot chain load |
| 01:46.38 | clever | make a 2nd u-boot, which the first one runs |
| 01:46.40 | luffe48 | ah! |
| 01:46.51 | clever | then the 2nd u-boot has kgdb support, and runs the kernel |
| 01:46.55 | prpplague | luffe48: the jtag doesn't cost a fortune, its the software that does |
| 01:47.03 | prpplague | luffe48: jtag is jtag is jtag |
| 01:47.26 | luffe48 | prpplague: ok |
| 01:47.27 | prpplague | luffe48: just need a software package to work with the hardware, there are number of free ones for powerpc |
| 01:47.52 | clever | you can maybe use jtag to reflash the main u-boot image |
| 01:48.57 | luffe48 | chain loading sounds a lot easier |
| 01:49.00 | clever | jtag or a uboot with kgdb would let you debug the problems |
| 01:49.06 | clever | yeah, chain loading would be safer |
| 01:49.16 | clever | less likely to brick it |
| 01:55.43 | luffe48 | i think i'll continue my battles tomorrow |
| 01:55.49 | luffe48 | thanks for the help guys |
| 02:00.59 | *** join/#edev Belliash (~belliash@asiotec/constitutive/belliash) |
| 02:29.14 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 04:29.28 | *** join/#edev [7] (~quassel@rockbox/developer/TheSeven) |
| 04:31.06 | *** join/#edev orhan89 (~orhan89@103.10.171.14) |
| 04:48.59 | *** join/#edev kergoth- (~textual@ip68-109-188-197.ph.ph.cox.net) |
| 04:49.49 | *** join/#edev kergoth- (~textual@ip68-109-188-197.ph.ph.cox.net) |
| 04:57.27 | *** join/#edev kergoth- (~textual@ip68-109-188-197.ph.ph.cox.net) |
| 05:59.48 | *** join/#edev kergoth- (~textual@ip68-109-188-197.ph.ph.cox.net) |
| 06:09.37 | *** join/#edev clio (~andrej@85.159.109.222) |
| 06:16.02 | *** join/#edev mbuf (user@nat/redhat/x-fsfmfznymopupjpi) |
| 08:13.06 | *** join/#edev addams (~addams@87.123.66.0) |
| 08:30.58 | *** join/#edev orbarron (~orbarron@nat/ti/x-ystlmzabmnmfwigr) |
| 08:55.38 | *** join/#edev giusef (~Giuseppe@80.17.44.2) |
| 10:03.17 | *** join/#edev giusef (~Giuseppe@80.17.44.2) |
| 10:18.10 | *** join/#edev addams (~addams@87.123.66.0) |
| 10:27.08 | *** join/#edev luffe48 (~luffe48@0407ds1-esp.0.fullrate.dk) |
| 10:39.22 | *** join/#edev devgiant (~devgiant@unaffiliated/devgiant) |
| 11:25.25 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.222.48) |
| 11:33.07 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.222.48) |
| 11:40.53 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.222.48) |
| 11:48.19 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.222.48) |
| 12:04.38 | *** join/#edev addams (~addams@87.123.66.0) |
| 12:28.45 | *** join/#edev g1powermac (~g1powerma@unaffiliated/g1powermac) |
| 13:17.00 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.222.48) |
| 14:14.35 | *** join/#edev arun_ (~arun@unaffiliated/sindian) |
| 14:29.27 | *** join/#edev Lars_G (~Lars@unaffiliated/lars-g/x-000001) |
| 15:07.29 | *** join/#edev ffio (~fire@unaffiliated/security) |
| 15:18.27 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 15:19.41 | g1powermac | needs a cheap source of inductors |
| 15:23.19 | av500 | hands g1powermac the wire bobbin |
| 15:23.38 | g1powermac | hehe |
| 15:24.00 | av500 | or scavenge old hifi speakers |
| 15:24.13 | av500 | the caps will be bad, but the coils should be fine |
| 15:24.41 | g1powermac | problem is I need very specific inductors though. . .which is making things a bit difficult |
| 15:29.05 | g1powermac | the datasheet for the chip I want to use has a list of recommended inductors, of course most aren't being made anymore |
| 15:29.25 | g1powermac | the one group that is costs $1.30 a piece |
| 15:29.46 | g1powermac | this is the chip I'm using: http://www.maximintegrated.com/datasheet/index.mvp/id/1878/t/al |
| 15:33.09 | g1powermac | wonder if I can get away with this: http://www.digikey.com/product-detail/en/RLB1314-220KL/RLB1314-220KL-ND/2561391 |
| 15:33.31 | g1powermac | its less than a third of the price and has similar ratings |
| 15:35.27 | *** join/#edev luffe48 (~luffe48@0407ds1-esp.0.fullrate.dk) |
| 15:48.03 | *** join/#edev sndcrb (~sndcrb@dsl-hkibrasgw1-58c385-172.dhcp.inet.fi) |
| 15:58.08 | *** join/#edev Lars_G (~Lars@unaffiliated/lars-g/x-000001) |
| 16:51.47 | *** join/#edev prpplague (~danders@rrcs-97-77-26-26.sw.biz.rr.com) |
| 16:55.49 | *** join/#edev [7] (~quassel@rockbox/developer/TheSeven) |
| 16:58.15 | *** join/#edev contempt (contempt@unaffiliated/contempt) |
| 17:04.10 | _av500_ | try and find out |
| 17:04.55 | *** join/#edev [7] (~quassel@rockbox/developer/TheSeven) |
| 17:07.41 | *** join/#edev contempt (contempt@unaffiliated/contempt) |
| 17:08.22 | *** join/#edev TheSeven (~quassel@rockbox/developer/TheSeven) |
| 17:16.55 | *** join/#edev contempt (contempt@unaffiliated/contempt) |
| 17:26.37 | *** join/#edev contempt (contempt@unaffiliated/contempt) |
| 17:36.30 | *** join/#edev contempt (contempt@unaffiliated/contempt) |
| 17:55.43 | *** join/#edev ffio (~fire@unaffiliated/security) |
| 17:57.22 | *** join/#edev NightMonkey (~NightrMon@173-164-139-193-SFBA.hfc.comcastbusiness.net) |
| 17:57.23 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 18:08.01 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 18:16.24 | *** join/#edev luffe48 (~luffe48@0407ds1-esp.0.fullrate.dk) |
| 18:46.23 | *** join/#edev gustavoz_ (~gustavoz@host252.190-31-1.telecom.net.ar) |
| 19:10.21 | *** join/#edev flo_lap (~fuchs@Maemo/community/contributor/florian) |
| 19:29.24 | *** join/#edev gustavoz (~gustavoz@host252.190-31-1.telecom.net.ar) |
| 19:45.21 | luffe48 | FINALLY |
| 19:45.33 | luffe48 | got my kernel to boot on my nas |
| 19:46.34 | luffe48 | turns out, either the cuImage wrapper or the old u-boot cuts everything above 4MiB off the vmlinux at some stage, without ever telling me |
| 20:01.07 | *** join/#edev florian (~fuchs@sign-4db6b0fc.pool.mediaWays.net) |
| 20:01.07 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 20:06.05 | clever | luffe48: aha! |
| 20:06.17 | clever | luffe48: chain loading a new u-boot helps? |
| 20:06.59 | luffe48 | didn't have the patience to figure out how to build it for my board |
| 20:07.17 | clever | how did you fix it? |
| 20:07.19 | clever | smaller kernel image? |
| 20:07.21 | clever | modules? |
| 20:07.22 | luffe48 | yup |
| 20:07.26 | luffe48 | cut everything out |
| 20:07.30 | clever | ah |
| 20:07.42 | clever | so its not really fixed, you just identified the problem and avoided it |
| 20:07.46 | luffe48 | i might not have any filesystem drivers, but at least the kernel boots beautifully :) |
| 20:08.17 | clever | initrd can load with minimal code, and contain the fs drivers in .ko form |
| 20:08.26 | clever | but the question is, can u-boot load a fat initrd? |
| 20:08.43 | clever | brb |
| 20:08.46 | luffe48 | i still have about 700kB up to the 4MiB limit |
| 20:08.52 | luffe48 | :) |
| 20:11.57 | *** join/#edev dalias (dalias@216-12-86-13.cv.mvl.ntelos.net) |
| 20:16.29 | clever | luffe48: 4mb compressed or uncompressed? |
| 20:17.37 | *** join/#edev Lars_G (~Lars@unaffiliated/lars-g/x-000001) |
| 20:22.29 | *** mode/#edev [+o prpplague] by ChanServ |
| 21:18.47 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.217.34) |
| 21:28.54 | luffe48 | clever: i misspoke, "it" seems to cut stuff off the end of the unpacked vmlinux image if it's longer than 4MB |
| 21:29.55 | luffe48 | clever: it didn't actually cut exactly at 4MiB |
| 21:30.17 | luffe48 | i'm not sure what's doing it |
| 21:32.10 | luffe48 | the unpacking routine in the cuImage wrapper seems to work - it's unpacking len bytes, where len is from the vmlinux elf header (i think) |
| 21:33.28 | luffe48 | i think either there's a gotcha in the old u-boot, or a really bad bug the cuimage wrapping shell script |
| 21:33.53 | luffe48 | or possibly i'm just doing something really dumb |
| 21:58.44 | *** join/#edev kuldeepdhaka (~kuldeepdh@117.254.217.34) |
| 22:01.13 | *** join/#edev flo_lap (~fuchs@Maemo/community/contributor/florian) |
| 22:51.42 | *** join/#edev florian (~fuchs@sign-4db6abb5.pool.mediaWays.net) |
| 22:51.42 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 22:52.20 | *** join/#edev NightMonkey (~NightrMon@173-164-139-193-SFBA.hfc.comcastbusiness.net) |
| 22:52.20 | *** join/#edev NightMonkey (~NightrMon@pdpc/supporter/professional/nightmonkey) |
| 23:25.23 | *** join/#edev mozmck (~moses@client-67.210.159.209.dfwtx.partnershipbroadband.com) |