| --- Log opened Fri Jul 19 00:00:33 2013 | ||
| stekern | yes, "reproducible", it seems it always crashes in __pagevec_lru_add_fn | 02:17 |
|---|---|---|
| stekern | but on different load/store instructions | 02:17 |
| stekern | and it's not reproducible in that sense that it is some test that triggers it, it's either while ftping or doing chmod +x | 02:19 |
| stekern | the net emulation is flaky too, I get a lot of these: nfs: server 192.168.255.100 not responding, still trying | 02:21 |
| stekern | and "NETDEV WATCHDOG: eth0 (ethoc): transmit queue 0 timed out" | 02:22 |
| stekern | on the bright side, I've got the kernel telling me about the pgd instead of pulling the information out of vmlinux | 02:24 |
| stekern | I'm still puzzled about the MMUCR stuff in spr_defs.h | 03:00 |
| stekern | I've looked through old arch specs, and none of them mentions what's in spr_defs.h | 03:01 |
| stekern | even or1200 manual from 2001 speaks about PTBP | 03:02 |
| stekern | linux doesn't exactly the arch specs definition of the page table entries | 03:35 |
| stekern | *exactly follow | 03:35 |
| stekern | more precisely: http://pastie.org/8154638 | 05:25 |
| poke53281 | Don't know. Ask Jonas. The specification is probably wrong and was inconsistent for some time. Who knows. In principle I would trust the implementation and not the specification. Normally there is a reason. | 06:05 |
| poke53281 | stekern: One idea, you could try the QEMU user mode emulation. | 06:06 |
| poke53281 | http://wiki.qemu.org/download/qemu-doc.html#QEMU-User-space-emulator | 06:06 |
| poke53281 | If this works the tests will be easy and fast. | 06:06 |
| poke53281 | For some reason my qemu still does not work. | 06:09 |
| stekern | on what should I ask Jonas? | 06:11 |
| stekern | the spr_defs? | 06:11 |
| stekern | they make no sense in the header file | 06:12 |
| stekern | as for the Linux vs spec pte differences, I understand those, and other archs (with soft tlb refill) does this kind of thing too | 06:13 |
| stekern | running it in user space qemu has crossed my mind too, right now I was mostly concerned with testing this hw tlb refill thing though | 07:14 |
| stekern | (qemu-user) doesn't seem to work properly with dynamic libraries | 07:22 |
| stekern | static executables run fine | 07:22 |
| stekern | looks like the odd xMMUCR spr defines in spr_defs.h have been there since year 2000, probably some very early draft idea that has just lingered around since no-one have used the xMMUCR registers until now | 11:01 |
| olofk | stekern: You look gorgeous now. :) | 22:34 |
| olofk | Just pushed another patch for or1200-monitor | 23:04 |
| --- Log closed Sat Jul 20 00:00:34 2013 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!