IRC logs for #openrisc Thursday, 2013-07-04

--- Log opened Thu Jul 04 00:00:11 2013
hnostekern, the mystery tightens.. earlycon=uart,mmio,0x90000000,115200 console=ttyS0,115200  works.. but I don't see how that's different from console=uart...00:57
hnofound it. It is a software error. Last digit of the baud rate is lost somewhere.01:15
hnonow just need to find where it's misplaced.01:16
hnoIt's lost very early.01:16
hnoKernel command line: console=uart,mmio,0x90000000,11520001:16
hnoEarly serial console at MMIO 0x90000000 (options '11520')01:16
hnoAha! Now I feel silly. The change have been staring at me all day and I haven't reacted.01:17
hnoEven logged in the very first lines of kernel output.01:36
hnoKernel command line: console=uart,mmio,0x90000000,11520001:42
hnoEarly serial console at MMIO 0x90000000 (options '11520')01:42
stekernhno, nice work tracking it down!02:18
stekernthat explains why it works in or1ksim, it doesn't care about the baudrate02:20
stekernstill a bit confused regarding verilated model, I thought the baudrate matters there, have to double check02:21
stekernI guess you don't need to run the verilated model, but in case you are still interested:02:22
stekernfirst install systemc libraries and verilator:
stekernwhen that's done, all you need to do is run 'make prepare-vlt', 'ln -sf /path/to/vmlinux.vmem' and '../vlt/Vorpsoc_top' in the sim/run/ directory of orpsocv202:25
stekernthat should have been 'ln -sf /path/to/vmlinux.vmem sram.vmem'02:26
stekernyeah, verilated model cares about the baudrate all right, but the orpsoc that is used for the simulation defines PRESCALER_PRESET_HARD and thus hardcodes the baud rate to 11520002:40
stekernhmm, openocds usb-blaster driver doesn't seem to want to acknowledge sockit08:55
stekernmaybe the low level chip isn't an ftdi anymore, have to take close look when I'm close to the board09:13
stekernbecause it fails when it tries to set the baudrate09:16
stekernolofk: I *know* you just sit on your board waiting for me to get gray hairs over this stuff, so you can jump in when the fun starts ;)09:21
stekernpoke53282: not sure if you heard, but I got native gcc working a while back09:29
-!- Netsplit *.net <-> *.split quits: andresjk, Logxen, hle_, poke5328109:32
stekernyup, no ftdi device, there is an cypress CY7C68013A-56BAXC there09:35
olofkstekern: Generating SoCs won't be part of the ORPSoC scope. It was initially, but I will stick with generating IP-Xact models (and the Altera equivalent if that's useful) and let tools such as Quartus, Kactus2 and Vivado take care of the system level.10:54
olofkIt won't be dependent on those tools, though. It's just that I won't spend time on auto generating top-level connections. Just provide the means for other tools to do that10:55
olofkstekern: And I think I will have plenty of time to play with SoCKit soon. In addition to having a baby, I bought a house two days ago. I heard that those two things really help you free up some spare time to work on side projects :)10:57
olofkDidn't know that you got native GCC working either. That's really cool. Last thing I heard, there were some bison or flex problems10:58
_franck_stekern: it is an embedded USB Blaster *II*11:38
_franck_another good project to come: make an opensources driver for this adapter.....11:39
olofk_franck_: Isn't there already a driver in OpenOCD?11:58
_franck_it for USB Blaster based on FTDI chips12:01
_franck_or my be this new one uses the same protocol12:01
stekern_franck_: yeah, I suspect the "blaster" protocol is the same, so it's probably a matter of getting the data to the chip on a different manner13:03
stekernI'll indulge myself in some usb sniffing excercises13:04
stekernolofk: native gcc is working, self-hosting (i.e. compiling a native gcc with a native gcc) is not, but that's because of uclibc, not gcc14:33
stekern_franck_: weren't you going to the workshop too, or has it already been?14:34
_franck_yes I did attend the workshop14:44
_franck_that wasn't that good, the only question I asked let the guy without voice for a moment :)14:44
stekernso you're now a proud member of the sockittens14:44
_franck_yep !14:45
_franck_I'm waiting for you to do the dirty job ;)14:45
stekernyeah, I'm not overly impressed with the workshop here neither, but I mostly went to get the goodie bag ;)14:46
stekerndirty jobs, just the way I like'em ;)14:48
stekernI didn't know wireshark can parse usb packages, that's handy14:49
_franck_I did see that was on the interfaces choice but I never tried it14:49
poke53282nice stekern. Did you use the gcc on
poke53282jor1k got a nice article on the hungarian unix portal
poke53282Here is the link:
stekern poke53282 yes, but it needs a (genric) patch that I've applied here:
stekernand cool news about the article! (not that I understand what the article says, but that's not so important in this context)19:20
poke53282Ok, thanks. I will try it on my emulator. Just for fun.19:32
poke53282I have also managed to compile and start a X-server for a few seconds before it crashes.19:33
poke53282Wayland is not possible to compile. Too many errors. directfb still crashes.19:34
poke53282I don't like software which are hardware dependent and don't provide a (slow) independent path.19:35
stekernwow, X-server would be nice ;)19:36
poke53282Next I will implement a touch screen controller.19:36
stekernI've tried compiling apache, but it fails with bison19:36
stekernmaybe I should try lighttpd19:37
stekernbut the point with compiling apache was more to see if it would be possible than to get a web server19:38
poke53282I am working on X. But more than xeyes, xclock and twm did not work.19:38
stekernlots to do still19:38
stekernah, but you got that much working still19:38
poke53282What I realized is actually that we need something like "Linux from Scratch" for or1k19:39
poke53282I mainly followed the commands provided by the documentation.19:39
stekernyeah, and completion of eglibc would nice19:40
hnostekern, I use busybox web server.. almost works.19:43
poke53282At the moment I am fine with uClibc. But  eglibc would be nice of course.19:44
poke53282Trying to compile wayland I realized that also libffi is missing support for or1k.19:44
poke53282But the big X11 did compile quiet well.19:45
stekernah, yes, the missing libffi is something we'll need to address too19:45
hnoI do have an automated busybox based root build. Need to patch openrisc uClibc a bit for current busybox.. (the upstream uClibc snapshot it's based on is borked)19:45
poke53282So far I had to patch the .config file. But that is standard.19:46
poke53282of uClibc I mean19:46
stekernhno: don't be shy to share it ;)19:46
poke53282Linux still don't compile with the newest cross compiled gcc.19:47
poke53282The final linking fails.19:47
stekernit does with or1k-elf-gcc19:47
stekernbut that's a bit of a workaround19:48
hnostekern, it's a cherry-pick of 139b8f0c673fed465d27f99c98568e5d5e1b9b72 from upstream.19:49
poke53282Ok, I don't have or1k-elf-gcc in my toolchain. How did you get it?19:50
hnostekern, is also included in the master-next branch.19:50
stekernwe had a discussion about that a while ago:
stekernhno: master-next of what tree?
stekernah, ok, then I think it should be in here too:
stekernI based that on master-next iirc19:53
poke53282Ok, thats it. I will never use Bugzilla again but will write my problems directly into this chat.19:53
stekernpoke53282: you build for the or1k-elf target instead of or1k-linux-uclibc target19:54
poke53282Jonibo wrote in the chat exactly what I realized in my bugzilla entry 6 month? ago.19:54
poke53282@stekern: Seems so. But I have never realized the real difference between or1k-elf- and or1k-linux-.19:57
stekernone is for baremetal (or1k-elf) and one is for linux userspace (or1k-linux-uclibc)19:58
stekernso different libc19:59
poke53282Ok, I understand. But it seems still like a workaround. I have looked up other architectures. They have directly implemented mul and div code in the kernel.20:00
poke53282At the moment we rely on libgcc to link it into the kernel.20:01
stekernyeah, it's a workaround, and a hassle to have different toolchains to compile the kernel and userspace20:01
poke53282Well, better than nothing. I tried to fix it myself but the libgcc code is so terrible for each architecture. Who is writing these. I didn't find an implementation that was similar to one of the others.20:03
hnostekern, yes that also have it.20:04
stekerndunno, but you can't really copy the libgcc code straight into the kernel neither20:04
poke53282Ok, but thanks stekern. I will give it a try.20:06
poke53282When I have a more or less working version of X with mouse support I will send a message in this chat.20:07
stekernthat'll be terrific! =P20:07
poke53282The first X11 running directly in a Web-Browser :)20:07
stekernhaha, yeah, that's awesome20:08
hnostekern, is the github openrisc uClibc maintained by anyone? The page you just linked refers to jonas repository.20:08
hnoAh, developmen version part of the page refers to the github one.20:10
stekernyes, it's part "of that package" so to say20:13
hnoa bit condusing to have multiple "official" repositories for the same thing in the same project.20:15
poke53282.... rebuilding toolchain ....20:16
stekernthere are reasons for the spreadage of repos, but that's a long story20:19
stekernincluded in that story are also the reasons why we have several mailing lists for the same topic, which is even more confusing...20:21
poke53282I can completely understand your confusion. I am confused as well.20:25
poke53282forks, branches, copies....20:26
poke53282But this is how the system works. At least it is documented within git.20:27
hnostekern, it signals a strong sense of fragmentation... branches etc is one thing, but...20:27
poke53282Most of the repositories on are forks.20:29
hnowhat github says is a fork or master do not mean much.20:31
hnowhen done "right" a project repository is normally starting as a fork of some personal repo.20:32
stekernhno: yes, I know, but as I said, there is a story behind it, Most of it can be read in the mailing list archives (take your pick of the two...) if you're interested ;)20:47
stekernbut to make a long story short, some developers wanted to use git instead svn and traditional mailing lists instead of forums... and here we are20:50
--- Log closed Fri Jul 05 00:00:13 2013

Generated by 2.15.2 by Marius Gedminas - find it at!