| stekern | blueCmd: mor1kx, the coolest openrisc implementation around (I am biased ;)): https://github.com/openrisc/mor1kx | 05:17 |
|---|---|---|
| stekern | more precisely, I am giving it MMUs | 05:18 |
| jeremybennett | juliusb: Are you around? Which country are you currently in? | 10:48 |
| blueCmd | stekern: ah, cool! | 11:14 |
| LoneTech | .. I thought I had fixed or1ksim disassembly in 64bit systems at some point, but it's broken in svn r797 | 14:40 |
| blueCmd | yeah, I had to compile it in 32 bit :( | 15:24 |
| LoneTech | software breakpoints, incomplete debugger code and delay slots cause me suffering | 15:27 |
| jeremybennett | LoneTech: I thought the Or1ksim disassembly was lifted from binutils? | 15:28 |
| LoneTech | well, it was broken there too | 15:28 |
| jeremybennett | If not, it should be! | 15:28 |
| LoneTech | of course | 15:29 |
| LoneTech | it ought to use libopcodes or something | 15:30 |
| LoneTech | ah, even found the patch I made | 15:32 |
| LoneTech | sure enough, similar code in cpu/or32/or32.c or1ksim_or32_extract | 15:35 |
| stekern | hesienbugs cause me suffering | 15:41 |
| stekern | *heisenbugs | 15:42 |
| jeremybennett | for a moment I was thinking of bugs in woven carpet backing :) | 15:43 |
| jeremybennett | But those would be hessianbugs | 15:43 |
| stekern | heh, well with my snr in spelling mistakes, it could very well have been hessianbugs misspelled ;) | 15:46 |
| blueCmd | stekern: you need to chose words that have enough distance from other words so one can be sure that error correction works then ;) | 15:47 |
| stekern | that's good advice =) | 15:50 |
| blueCmd | what have you guys done with mmap in or1k-linux? :D | 16:34 |
| blueCmd | something is really broke somewhere, not ruling out that it's in my stuff | 16:34 |
| blueCmd | (i'm using mmap2 btw, not sys_mmap which is in sys-or32.c) | 16:35 |
| blueCmd | http://pastebin.com/MRgDGhDC what I'm trying to do (works in qemu, fails in or1ksim with or1k-linux kernel) | 16:37 |
| blueCmd | I will try to run that on a vanilla kernel | 16:38 |
| blueCmd | same result in vanilla :/ | 16:45 |
| blueCmd | d4 is passed to mmap_pgoff but reading from the mmap returns data at 001a8000 | 16:55 |
| blueCmd | bah, PAGE_SIZE missmatch somewhere | 17:41 |
| blueCmd | aand fixed. | 17:53 |
| stekern | blueCmd: how are you running stuff in qemu? | 18:29 |
| blueCmd | stekern: static programs work just by doing qemu-or32 ./a.out | 18:31 |
| blueCmd | dynamic works if you chroot | 18:31 |
| blueCmd | (or should do at least) | 18:32 |
| blueCmd | (I run things staticly) | 18:32 |
| stekern | ah, cool. have to give that a try some day | 18:34 |
| blueCmd | but apparently the mmap is broken in qemu-or32 ;) | 18:34 |
| blueCmd | (it should convert offset which it doesn't do currently) | 18:35 |
| blueCmd | the dream of debian on or1k is approaching! relocations working, need to add some .got things and things should start to work :P | 18:40 |
| stekern | great! | 18:58 |
| stekern | hopefully I've got those heisenbugs ironed out by then so I can try it out on mor1kx =) | 18:59 |
| stekern | but this one seems to be a harder nut to crack... only common denominator between the scenarios seems to be that r2 is getting screwed somewhere | 19:00 |
| stekern | I've got a suspection about arch_local_irq_restore | 19:02 |
| stekern | http://pastie.org/5940856 | 19:02 |
| stekern | a 'l.lwz r2, -4(r1)' in a delayslot right after interrupts gets reenabled | 19:03 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!