IRC logs for #openrisc Saturday, 2016-09-10

--- Log opened Sat Sep 10 00:00:05 2016
jia\shorne I can not boot o1k-vmlinux using the github/openrisc/linux repo01:00
jianot sync Attemped to kill init! exitcode=000000701:02
-!- Netsplit *.net <-> *.split quits: simoncook, _franck_, poke53281, blueCmd, dalias, knz, eliask, SMDhome, _florent_, mwfc, (+5 more, use /NETSPLIT to show all of them)01:07
jianot sync: Attempted to kill init! exitcode=0x0000000701:10
stekernjia: how are you building it?01:12
jiaARCH=openrisc CROSS_COMPILE=or1k-linux-uclibc- make or1ksim_defconfig01:12
jiaARCH=openrisc CROSS_COMPILE=or1k-linux-uclibc- make01:13
jianothing changed, clean code01:13
-!- Netsplit over, joins: dalias, blueCmd, jeremybennett, mwfc01:25
jianew vmlinux built by or1k-elf-gcc-5.3, get the same error01:27
stekernjia: what or1ksim are you using?01:28
jiaI'm using qemu01:29
jiaqemu can boot old vmlinux I kept years ago01:29
stekernah, you need to turn on emulation for atomic instructions then01:30
stekernor use the patches blueCmd (iirc) made01:30
jiaOK, I'll try his repo01:31
stekernactually, you should not need to turn on emulation, it should work automatically01:36
stekernso probably something else01:36
jiablueCmd's repo can not be built... 9pfs error01:41
stekernyeah, but as I said, it shouldn't be related to having atomic instruction support or not01:48
stekernunless there's a bug in the emulation, which I can't promise that there isn't01:48
stekernwas a while since I tested it01:49
jiaI'm trying fix it...01:53
stekernI can test with an or1ksim without atomic instruction support to see if that works01:54
jiayes, blueCmd's repo works!01:54
jiaand I find blueCmd made a virtio hda support.01:55
jiaI'll continue in 15 minutes, to eat something01:57
jia./or32-softmmu/qemu-system-or32 -kernel ~/vmlinux -drive "file=/home/jia/or1k.qcow2,media=disk,format=qcow2" --append "rootfstype=ext4 rw root=/dev/hda console=uart,mmio,0x90000000,115200"03:13
jiaVFS:unable to mount root fs on unknown block(0,0)03:13
jiaWhat should I do to make it run?03:14
blueCmdoh man, that was a long time ago. hmmm03:47
blueCmdjia: are you trying to use 9pfs? what's the kernel config you're using?03:48
jiano, I didn't use 9pfs.03:48
jiamay you please tell me how did you build linux and rootfs img?03:49
blueCmdI seem to remember that the IDE support in openrisc was a bit so-so and that 9pfs was the easiest way to get something to work03:49
blueCmdjia: It was a long time ago I played with this, but when I did play with it I used this:
blueCmdthat should be "type make to build everything"03:51
blueCmdbut it was made for me by me, so it's just a collection of scripts so I don't have to type it all the time03:51
blueCmdjia: how I used to run or1k in qemu03:52
jiaIs this the rootfs you build?03:53
blueCmdjia: then you need this kernel I think:
blueCmdjia: it will end up in that directory yes03:55
jiathank you, I'll try.03:56
shorneolofk: wallento: I finally got my work done! book tickets to Italy and put in a Talk Proposal04:14
wallentogreat, looking forward to meet you04:17
mor1kx[mor1kx] spacemonkeydelivers opened pull request #40: Introducing PCU (master...master)
SMDnoteGot my tickets to Italy too.05:04
olofkshorne, SMDnote : Looking forward to meet you over a pizza in Italy :)08:56
olofkshorne: Did wb_intercon_gen work out for you? Had never seen that weird perl script before08:57
olofkjia: Cool to see some work on qemu. Did you get it working?08:57
jianot yet.  I'm too stupid to make linux work:)08:58
olofkjia: Oh, I know all about that feeling :)09:04
olofkWhat's the current status? Have you compiled linux+busybox?09:05
jiaVFS: unable to mount FS on block(0,0)09:05
olofkI'm too stupid to get qemu working, so I might not be too much help :)09:06
olofkIs that running under qemu or or1ksim?09:06
jiarth make qemu update09:06
olofkjia: And have you considered coming to orconf this year?09:07
jiamaybe a virtio device will be good.09:07
jia1, I want, but my boss don't.09:07
jia2, My English is not good, even Italiano...09:08
jia3, I love pizza:)09:09
olofkAh, that sucks :/09:09
jiamay you please suggest all or1k thing commit into the commiunty, on teh ORCONF?09:10
jiabinutils-gdb gcc newlib uclibc glibc09:10
jiait will be a BIG GREAT to use the offical git repos with OR1K support09:11
ZipCPUolofk: Do you have an orpsoc-cores/system for the qemu that jia is working with?09:21
ZipCPUolofk shorne SMDnote: I look forward to joining the three of you.  I like sausage pizza! ;)09:22
olofkjia: newlib and binutils (without gdb) is already upstream09:59
olofkWe've been trying to get gcc upstream, but we are still waiting for the permissions of the final contributor before we can send it in10:00
jiagcc and uclibc to be10:00
olofkI think we can skip uclibc as the musl port is in better shape10:00
olofkZipCPU: qemu is outside of the scope for FuseSoC10:00
olofkAnd I'm also looking forward to the pizza :)10:01
SMDnoteeveryone loves pizza10:01
SMDnoteolofk: Could next(2017) orconf happen somewhere in Northen Europe?10:01
olofkSMDnote: For all the previous years, it has been someone in the community who has suggested hosting it10:02
olofkSo if there is someone somewhere in northen europe who would like to host it, we're all ears10:02
olofkWe've been to Sweden, England, Germany and Switzerland so far10:03
jiayou are all in Europe:)10:04
jiaat last10:04
olofkjia: Where are you located?10:04
olofkAha. Well, we have been talking about going to US and Asia, to make it easier for contributors living there to attend10:05
olofkBut as most of the organizers live in Europe it has just been easier to host it here up until now10:05
olofkBut please let us know if anyone is interested in hosting next year. It's good to have some options10:07
SMDnoteolofk: I'd suggest Moscow, but yu, guys, would have lots of troubles w/ getting visas10:08
olofkSMDnote: Yeah. Moscow would be a nice option, but as you say, the logistics can be a problem. Especially if Sweden decides to join NATO, which we are currently discussing :)10:09
SMDnoteolofk: I'm looking for opportunities to join Sweden ;)10:10
olofkSMDnote: Send me a mail and we'll see what we can do :)10:11
jiaMay I join Sweden too :)10:20
jiavirtio-mmio 93000600.virtio: Wrong magic value 0x76697274!10:21
jiawhy I got this?10:21
olofkjia: Sure! We have room for you too :)10:23
olofkjia: What do you have on base address 0x93000000 ?10:23
jiaa minute10:24
jia    the diff and cmdline10:27
SMDnotejia: Could I ask if you work on openrisc at work?11:35
jianope, only years ago on my spare time11:35
SMDnotejia: it's a pity, I wish some Chineese ppl would make a standalone soc on openrisc :)11:44
jiaI don't know HDL FPGA things...11:45
SMDnoteme neither ;)11:47
olofkHow on earth does this virtio thing work?15:35
shorneolofk: wb_intercon_gen is working fine.  Just when I did a search for wishbone builder that perl script always pops up. We need to link to that in our docs somewhere i.e. a wishbone tutorial.19:59
ZipCPU|Laptopwb2axisp works!! (Although I've only tested it with non-pipelined data so far ...)23:04
ZipCPU|LaptopNow ... if only was up, I might be able to adjust the repository ...23:05
ZipCPU|LaptopAlso ... Xilinx's MIG offers only pitiful memory performance in terms of latency.  Pitiful.23:05
ZipCPU|LaptopIf I send it a address read-valid on clock 2, the data doesn't get returned until clock 25.23:06
ZipCPU|Laptop(Wishbone strobe on clock 1, wishbone ack on clock 27 ...)23:07
ZipCPU|LaptopSad thing is ... it only takes six clocks to do the memory op itself:23:09
ZipCPU|Laptop1. Precharge the row (assumes it was open)23:10
ZipCPU|Laptop2. Activate the row you really want23:10
ZipCPU|Laptop3. Issue the read command.23:10
ZipCPU|Laptop4. Idle23:10
ZipCPU|Laptop5. Data crosses the bus23:10
ZipCPU|Laptop6. Data can be read from bus23:10
ZipCPU|Laptop7. Data should be able to be returned.23:10
ZipCPU|LaptopOkay, suppose you take an extra clock to set up the bus and tear it down, you're still looking at 9 clocks.23:11
ZipCPU|LaptopWhere do Xilinx's other 27 clocks come from??23:11
--- Log closed Sun Sep 11 00:00:06 2016

Generated by 2.15.2 by Marius Gedminas - find it at!