IRC logs for #openrisc Tuesday, 2017-04-04

--- Log opened Tue Apr 04 00:00:14 2017
--- Log closed Tue Apr 04 01:09:44 2017
--- Log opened Tue Apr 04 01:09:52 2017
-!- Irssi: #openrisc: Total of 45 nicks [0 ops, 0 halfops, 0 voices, 45 normal]01:09
-!- Irssi: Join to #openrisc was synced in 29 secs01:10
-!- Netsplit *.net <-> *.split quits: juliusb, shorne, X-Scale, fotis, promach, O01eg01:14
-!- [X-Scale] is now known as X-Scale01:14
-!- Netsplit over, joins: promach01:21
wbxshorne_: hi. qemu-system-or32 does not exit when powering off : Requesting system poweroff -> hangs02:34
wbxshorne_: 2.9.1rc, any idea? this is needed for automatic testing...02:35
wbxshorne_: or is network emulation working? then I can try qmp. to inform qemu via qmp to shutdown the machine.02:36
mithroanyone here understand the start up sequence for Linux on or1k? I'm trying to understand how the MMU gets initialised with the initial page mapping?06:18
mithroLooks like I want the setup_memory function07:47
shorne_wbx: I would have to check what does your software do to halt? it linux?10:16
shorne_mithro: there are a few steps, very initial is here...10:17
shorne_https://github.com/stffrdhrn/linux/blob/openrisc-4.12-smp-qspinlock/arch/openrisc/kernel/head.S#L61410:20
shorne_as soon as l.mtsprr0,r30,SPR_SR is run, it enables mmu, next instruction fetch causes a page fault10:22
shorne_it then sets up that page10:22
shorne_in boot_itlb_miss_handler10:22
shorne_after mmu is setup more there is switch from boot_itlb_miss_handler to itlb_miss_handler10:23
shorne_(thats done with some self writing code updating the jump inscrution at 0x300 Data & 0x400: Insn Page Fault exception10:24
-!- shorne_ is now known as shorne10:25
wbxshorne: yes, linux busybox poweroff13:19
wbxshorne: seems to work with most other qemu system emulation. may be something special in or1k?13:20
shornewbx: if you look at 'arch/openrisc/kernel/process.c' there are some functions like machine_power_off() etc.17:17
shornethats what is called when the power goes down, for openrisc it just does17:18
shornel.nop 117:18
shorneother archs probably do something different, like call some halt instruction, or talk to power hardware17:20
shornethen qemu can emulate that and exit the system17:21
wbxso you think getting network emulation running and qmp will be simpler?17:24
shorneI think getting halt working will be easier17:56
wbxcan you assist me or do you have a concrete idea? i can test patches :)17:58
shorneI am not sure if its the right thing, but there is a power management register17:59
shornemaybe we can set the suspend flag there on powerdown17:59
wbxshorne: could you already see the issue with qemu and linux trying to poweroff?18:00
shornewbx: well, I noticed it didnt actually exit qemu18:01
shornebut I just pretty ctrl+c to exit18:01
wbxyeah, but that is hard to do in non-interactive shell scripts :)18:02
shorneyou can do with an 'expect' script?18:02
wbxurgs. all other systems can either use qmp or poweroff. implementing such a hack for just one test case... hmm18:03
shorneright, it will require patches to qemu and kernel I think18:03
wbxhttps://tests.embedded-test.org/uClibc-ng/1.0.22/18:04
wbxi used or1ksim, but it now fails to start, so i tested qemu again.18:04
shorneit fails to start? oh.. and this works in or1ksim?18:05
shornedid my kernel changes break or1ksim?18:05
wbxhmm. not sure which change breaking it.18:06
shornedoes halt work in or1ksim?18:07
wbxany changes in upstream between 4.9.6 and 4.9.20?18:07
wbxhttps://tests.embedded-test.org/uClibc-ng/1.0.22/REPORT.or1k.libc.uClibc-ng-1.0.2218:07
wbxyes poweroff in or1ksim worked.18:07
wbxafter that console is back.18:07
shorneok, then I can upate qemu to do like or1ksim I guess18:08
shornesorry got to go18:08
wbxno prob18:08
mithroshorne: I'm working on getting Linux booting on our SoC which uses the mor1k but none of your peripherals18:20
--- Log closed Wed Apr 05 00:00:15 2017

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!