IRC logs for #openrisc Sunday, 2014-08-17

--- Log opened Sun Aug 17 00:00:10 2014
poke53281So, back from hiking.03:03
poke53281stekern: Can you give me a linux which works in or1ksim with a musl-busybox ramfs? This way I can be sure, that my emulator is the problem.03:04
poke53281I have improved the lwa and swa instructions. But it seems, that this doesn't matter. l.swa seems to never fail in my case.05:14
stekernyou'll need to change the or1ksim config to 1GB of mem though05:54
poke53281Of course, your version works in or1ksim06:44
poke53281Mine however,
poke53281doesn't execute "ifup -a", but for example "ifup -av" works (with verbose messages)06:46
poke53281The root option and the virtio devices are ignored. The ram is reduced to 32MB.06:52
poke53281So, doesn't seem to be an emulator problem.  More a kernel problem then. I will try to compile your linux version tomorrow. Would also good if you could provide me with this vmlinux version that doesn't need 1GB :) Firefox is very picky06:55
poke53281I would try your image in jor1k.07:05
stekernI'm out of the house currently, I can build you a version in the evening (EEST)07:56
stekerndoes jor1k support shadowed gpr?07:56
stekernread and write via spr07:57
blueCmd - god, that website is nice10:27
blueCmdit has what I wanted to have, *sigh* :P10:28
poke53281Yes, that looks a little bit more unique than a wiki.17:36
poke53281But who wants to maintain such a website?17:40
poke53281stekern: No, I don't support shadowed gpr. But I get an error message for each unknown access to spr.17:49
poke53281it makes sense for smp systems of course.17:49
stekernyeah, I was mostly wondering if I should turn that of for you if I build you a new kernel17:51
poke53281It works in or1ksim but in jor1k I get. "Error in GetSPR: address 128 unknown"  Will implement it now18:13
stekernoh... right18:15
stekernthat's the coreid spr18:15
stekernI have a local hack that makes the second core just sping if I try to boot a non-smp kernel on the dual-core setup18:16
poke53281what should it return?18:16
stekernwell, the cpu core id ;)18:16
stekernstarting from 018:16
poke53281Ok, then I am right.18:17
stekernI have fixed my hardware so I can boot only one of the cpus and leave the other in reset, so theoretically I could remove that local diff now...18:18
poke53281Ok, the kernel boots. But some reason I doesn't recognize my terminal input.18:18
poke53281But for some reason it doesn't recognize my terminal input.18:19
poke53281I had this problem yesterday too, but ignored it.18:19
poke53281Good, your version works in jor1k too.18:25
poke53281Intressting, I raise two interrupts, one which is known to the kernel and one which is unknown. In this case he seems to ignore both interrupts.18:27
poke53281Both at the same time.18:27
poke53281Ok, but this is not the main problem. The problem is narrowed down to my compiler and the kernel I use.18:28
stekernI'll give you my config18:36
stekernand this is the dts:
poke53281thanks, will try it.18:52
poke53281stekern: You should definitely use the feature register more often. A kernel config which asks for lwa and swa and shadow-registers and so on is not optimal.19:44
poke53281It works!!!20:06
poke53281Finally, the kernel was the problem20:06
poke53281blueCmd: I don't know how and why, but the kernel is broken20:07
poke53281the mexiko branch at least20:13
poke53281The error could produce all kinds of problems. I think it has to do with fork+interrupts+syscalls+scheduling.20:27
poke53281I would suggest to do a diff with
blueCmdpoke53281: use the smp branch in my repo instead20:55
blueCmdpoke53281: mexiko and smp are based on stefans20:55
blueCmdpoke53281: only has 3 local commits from where I synced against stekern's repo20:56
poke53281stefans repo works21:01
poke53281at least your mexiko repo doesn't work for some reason.21:01
poke53281I can try the smp branch from your repo21:02
--- Log closed Mon Aug 18 00:00:12 2014

Generated by 2.15.2 by Marius Gedminas - find it at!