--- Log opened Mon Jun 24 00:00:57 2013 | ||
stekern | olofk, it still doesn't and you have to invalidate the mmus too | 07:15 |
---|---|---|
stekern | oh, and you have to clear r3 too | 07:19 |
stekern | and you don't need uclibc to compile the kernel | 07:25 |
olofk | stekern: Can I just do make CROSS_COMPILE=or32-elf- to build linux? | 19:01 |
olofk | And about r0, and r3 too apparently... is my only option to force them to 0 during RTL simulation? | 19:02 |
olofk | Or can I do it in software somehow? | 19:04 |
olofk | And remind me again... what was the problem with clearing r0 as the first instruction in linux? IIRC there was a patch that was rejected | 19:30 |
hno | olofk, yes. "make CROSS_COMPILE=or32-elf- ARCH=openrisc or1ksim_defconfig" then "make CROSS_COMPILE=or32-elf- ARCH=openrisc -j8", maybe menuconfig inbetween if you want to change the config. | 21:17 |
olofk | hno: I tried it, but with or1k-elf-. It compiled fine after I changed the linker scripts to use elf32-or1k, and it boots up, but I'm getting some crashes | 21:18 |
* hno needs to update his toolchain... | 21:19 | |
* olofk needs to cut down on the amount of openrisc toolchains currently installed | 21:20 | |
hno | Anything later than 3.1 generally crashes on ordb2a for me. But I did get 3.8 running some time ago with some orpsoc/or1200/ordb2a bus & arbiter fixes. | 21:22 |
hno | will try again in some days. Currently stuck on 3.1 in a project with too little time before vacation starts. | 21:22 |
olofk | It's not like I'm doing anything more advanced than cat /proc/cpuinfo after I boot linux, so any old version works fine for me :) | 21:23 |
hno | then try openrisc-3.1 tag. Never had issues with that one.' | 21:24 |
hno | on that level... | 21:24 |
* hno have lost count of how many toolchans are installed for various arches.. | 21:28 | |
olofk | Hmm.. I managed to track my Linux boot in Icarus into early_init_devtree, that is called from or32_early_setup, but now the executed log isn't updating anymore | 21:29 |
olofk | I must have the shareware version of or1200-monitor that only handles 70959 instructions | 21:30 |
olofk | Oh well. Time for bed | 21:30 |
stekern | olofk: that's probably r3 (if you haven't handled that) | 23:16 |
stekern | the kernel expects a pointer to the dev tree in that register, so it's deliberately not cleared | 23:17 |
stekern | and it tries to read the dev tree magic from the pointer location, so if it's x you'll just lock up | 23:18 |
stekern | http://pastie.org/8076820 | 23:21 |
--- Log closed Tue Jun 25 00:00:58 2013 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!