IRC logs for #openrisc Wednesday, 2014-04-16

--- Log opened Wed Apr 16 00:00:07 2014
olofkLimb: There are a few more file that could be interesting to look at05:15
stekernthe X-files?05:22
stekernah, found why the LRU stuff broke mor1kx06:07
mor1kx[mor1kx] skristiansson pushed 1 new commit to master:
mor1kxmor1kx/master 7db7a34 Stefan Kristiansson: dcache: save copy of tag data on refill...06:35
_franck_web_blueCmd: yeah like that but with a 800x480x24bpp LCD ;)06:38
stekernonce again, the atlys board boots linux ;)06:49
stekernlet's see if cache ways > 2 works too06:49
stekernseems like it does:
stekern_franck_web_: what was the issue you had when the kernel got stuck on this line:
_franck_web_couldn't reproduce it :( And because I was working on something else I didn't try hard09:51
stekernok, my atlys board port and mor1kx-generic gets stuck there as well when I enable ethernet09:54
stekerniirc, we left it at suspecting a mtspr ack logic bug09:55
stekernbeing able to reproduc in mor1kx-generic makes things a lot easier to disect09:56
stekernlet me do a quick test with or1200-generic just to make sure09:56
* stekern thinks having those as seperate systems was a wise choice09:57
stekernyeah, this is mor1kx specific10:57
blueCmd_franck_web_: Yeah, we want to build an extension for the bottom header with a display and/or VGA14:22
Limbolofk: I'm more then willing to send any and all files you'd want to look at15:15
Limbolofk: I also implemented a Artix7 compatible clock generator so timing constraints are actually implemented correctly (as far as I can tell)15:17
Limbolofk: Or we don't need to do that15:42
LimbI can't believe I wasted three days on this15:43
LimbSomehow the wb address width was set to 26. And for whatever reason I SWEAR when I kept checking the original code, it was 26. So despite the fact that I knew the width was only 23... I left it as is15:43
LimbI double checked the original source and its of course set to 2315:44
Limbso this entire time, it had been trying to access bits in the address that just _didn't_ exist15:44
stekernLimb: so, it's working now?16:12
Limbstekern: yep16:12
Limbtrying to see if the linux image will work16:13
Limbthough I'm not quite sure how I'm supposed to connect to it lol16:13
LimbI uploaded the image and its sitting at 0x20016:15
Limbdo i have to regnpc 0x100?16:15
Limbused the linxu file from here:
stekernyou should be able to use the or1ksim config as a minimal setup if you comment out the ethernet node and set the cpu clock freq and the serial port clock freq16:16
stekernah, ok, I think that might work as well, if you have gpios16:16
Limbthat I do16:17
stekernbut, it's a bit tricky to load linux with gdb, you have to ensure that r3 is cleared16:17
LimbAlthough I'm more then willing to create a linux image specifically for the nexys4 althogh I just don't know how :P16:17
stekernand you have to reset the bits in SR16:18
stekern(*and* to make things even trickier, mor1kx doesn't allow setting registers over the debug port yet)16:19
Limbstekern: So uploading that image in openocd like it says wasn't enough?16:20
LimbIs there an updated tutorial on how to?16:20
stekern...and at least here in my setup with the atlys board, gdb crashes and starts to leak memory like crazy when I do: set $sr=116:21
stekernwell, what is said there works... for most of the times...16:21
stekernit does reset SR (if you followed my advice and connected the debug reset to the cpu reset)16:22
stekernand if you're lucky, r3 will hold a value that will not cause a bus error (this is less certain)16:22
stekernto handle that, I've been using a small program like this:
stekern...well, not binary, elf16:24
Limbstekern: so upload that elf, then upload the linux image16:30
Limband then telent over tty to the board?16:30
Limber not telnet, but connect to the serial port16:30
stekernwell, upload that elf, run it, upload the linux image16:31
stekernbut I guess that was implied16:31
Limbstekern: connecting the debug reset to the cpu reset gives me crc errors16:42
stekernnow too?16:43
stekernthat's strange16:44
skip_I followed the instructions on for the easy method, then compiled the version of linux the superproject comes with (2.6.37 with some extras), and it works great with or1ksim17:14
skip_I then tried the latest version from git:// and it appears work perfectly, apart from warnings from kernel/rcu/tiny.c:
skip_has anyone seen this happen before?17:17
stekernskip_: by coincident, I just ran that in or1ksim here:
mor1kx[mor1kx] skristiansson pushed 1 new commit to master:
mor1kxmor1kx/master 933f942 Stefan Kristiansson: dcache: invalidate from READ state...17:48
stekern_franck_: ^ we were on the right track back tehn17:49
_franck_stekern: cool another bug squashed18:03
stekern_franck_: yup, but not the last apparently...18:16
stekernstill crashes when I turn on ethernet, but later now18:17
stekernah, this might be SoC related though18:28
stekernyup, I had screwed up the ethmac wb_adr assignment...19:11
stekernit expects [11:2], but I gave it [31:0]19:12
stekernthe ethernet isn't really working though19:16
stekernolofk: have you tested the ethmac in orpsoc-cores?19:16
stekernnm that question for a while, I had commented out the master cyc & stb signal when I was debugging the $clog problem...19:34
stekernah, much better, I've got net!20:11
stekernblueCmd: handhold request!20:11
_franck_stekern: does the ethmac works out of orpsoc-cores ? I'll use it on my neek board soon20:13
stekern_franck_: seems so20:14
stekernlet's see if I can get my old nfs rootfs up and running20:18
blueCmdstekern: crashes seems to be a broken libmpfr21:43
blueCmd(gcc crashes)21:43
--- Log closed Thu Apr 17 00:00:08 2014

Generated by 2.15.2 by Marius Gedminas - find it at!