Back to Home / #uml / 2007 / 11 / Prev Day | Next Day
#uml IRC Logs for 2007-11-19

---Logopened Mon Nov 19 00:00:32 2007
00:44|-|mgross [~mgross@pool-71-117-236-31.ptldor.fios.verizon.net] has quit [Quit: Leaving]
01:51|-|camgirl29 [~camgirl29@ANantes-257-1-100-96.w90-25.abo.wanadoo.fr] has joined #uml
01:53|-|camgirl29 [~camgirl29@ANantes-257-1-100-96.w90-25.abo.wanadoo.fr] has quit []
02:17|-|Ancalagon [~PtitKosmi@215.208-78-194.adsl-fix.skynet.be] has joined #uml
02:48|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has quit [Remote host closed the connection]
03:09|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has joined #uml
03:52|-|peterz [~peterz@f237116.upc-f.chello.nl] has quit [Quit: Coyote finally caught me]
04:45|-|balbir [~balbir@122.167.179.130] has joined #uml
05:28|-|balbir [~balbir@122.167.179.130] has quit [Ping timeout: 480 seconds]
05:37|-|balbir [~balbir@122.167.179.130] has joined #uml
06:43|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has quit [Remote host closed the connection]
06:47|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has joined #uml
---Logclosed Mon Nov 19 07:32:55 2007
---Logopened Mon Nov 19 07:32:57 2007
07:32|-|mikegrb_ [~michael@mail.thegrebs.com] has joined #uml
07:32|-|Ekipa kanalu #uml: Wszystkich: 31 |-| +op [0] |-| +voice [0] |-| normalnych [31]
07:33|-|caker_ [~caker@ns.theshore.net] has joined #uml
07:34|-|Kanal #uml zsynchronizowany w 92 sekundy
07:34|-|tasaro [~tom@ns.theshore.net] has quit [Ping timeout: 480 seconds]
08:27|-|tasaro_ changed nick to tasaro
08:44|-|dang [~dang@nemesis.fprintf.net] has quit [Quit: Leaving.]
08:50|-|IntuitiveNipple [~TJ@alexandros.tjworld.net] has joined #uml
08:51<IntuitiveNipple-#uml->>Has anyone seen issues with uml getting 'stuck' in a loop executing /sbin/hotplug during run_init_process ?
09:08|-|dang [~dang@aa-redwall.nexthop.com] has joined #uml
09:32<IntuitiveNipple-#uml->>Is it possible to attach gdb to a process running inside uml, when it is stuck in a loop that is driving the CPU hard?
09:40<Magotari-#uml->>IntuitiveNipple: I don't know the answers to your questions, but jdike probably will. He should be here in about an hour or two. It varies.
09:40<IntuitiveNipple-#uml->>ahhh... I tend to always come up with the difficult ones!
09:41<IntuitiveNipple-#uml->>It seems weird, and probably something daft I've done, but it *was* working in an earlier build (creating ubuntu gutsy root-fs images)
09:41<Magotari-#uml->>Hmm...
09:41<Magotari-#uml->>So far I have seen a lot of people have trouble with uml when ubuntu was involved.
09:42<Magotari-#uml->>Why would you want to use ubuntu as a rootfs? Debian should do just as well, but I think it is better suited for virtual machines.
09:42<IntuitiveNipple-#uml->>I'm an ubuntu developer
09:42<Magotari-#uml->>Good answer.
09:43<IntuitiveNipple-#uml->>It worked fine, as I say, but something I've done in fine-tuning the config may have affected it, or it might be there's something up with the debootstrap and apt-get minimal
09:44<IntuitiveNipple-#uml->>gdb was a bit ropey at grabbing the breakpoints as well, but I eventually caught it in kernel_execve()
09:45<Magotari-#uml->>I'm sure jdike will be of more help than I. I try to handle help requests as they come, but very often I have to just pass the problem onto him. I'm new here, you see. A little passtime, tis all.
09:45<IntuitiveNipple-#uml->>even an over-ride with "init=/bin/bash" does the same, which makes me think its a kernel config issue rather than the userland install
09:45<Magotari-#uml->>Does defconfig work?
09:47<IntuitiveNipple-#uml->>I've not seen that! what/where is it?
09:52|-|Electric1lf [~dbharris@bas1-toronto48-1242461494.dsl.bell.ca] has joined #uml
09:59|-|ElectricElf [~dbharris@bas1-toronto48-1279276657.dsl.bell.ca] has quit [Ping timeout: 480 seconds]
10:07<Magotari-#uml->>IntuitiveNipple: Are you compiling your own kernel?
10:07<Magotari-#uml->>There are some settings which are the sane defaults.
10:07<Magotari-#uml->>You can use them by doing "make defconfig" in your kernel directory.
10:07<Magotari-#uml->>You need to have ARCH=um in your environment too.
10:08<IntuitiveNipple-#uml->>Oh, I get it, a make option
10:08<Magotari-#uml->>Yup.
10:08<Magotari-#uml->>Then just edit things with menuconfig or xconfig, whichever you prefer.
10:08<IntuitiveNipple-#uml->>Yes, I'm compiling my own. The stock Ubuntu i386 generic config + the ones required for UML, + the root_fs type
10:09<Magotari-#uml->>Generally, UML's defconfig will work for 80% of what you need.
10:09<Magotari-#uml->>You don't have to change it much at all.
10:09<Magotari-#uml->>Also, a while back there were problems with modules, so sticking to a static kernel might be an idea.
10:10<Magotari-#uml->>Not like you really need modules in a virtual machine, but...
10:10<Magotari-#uml->>I am not sure if those problems ever got fixed.
10:12<IntuitiveNipple-#uml->>Well, this is stuck in uml itself, it isn't get that far! it spins when it does kernel_execve('/sbin/init')
10:12<IntuitiveNipple-#uml->>I've captured an strace log
10:13<Magotari-#uml->>What is your arch? Just curious.
10:13<Magotari-#uml->>64 bit?
10:13<IntuitiveNipple-#uml->>host is x86_64, building 32-bit UML
10:13<Magotari-#uml->>Oh.
10:13<IntuitiveNipple-#uml->>2.6.22.8
10:13<Magotari-#uml->>Well, if you are building a 32bit uml, I think you might want to set SUBARCH or something like that.
10:14<IntuitiveNipple-#uml->>ok, defconfig built... now to try
10:14<Magotari-#uml->>Otherwise you shall build a 64 bit one.
10:14<IntuitiveNipple-#uml->>I do... "i386"
10:14<Magotari-#uml->>Oh... Ok.
10:14<Magotari-#uml->>Is the filesystem 64bit or 32 bit?
10:14<Magotari-#uml->>It would have trouble if the guest kernel and fs are of different number.
10:15<Magotari-#uml->>If the number of bits is the same you can run a 32fs+32uml on a 64 host, or 64fs+64uml on a 64 host.
10:16<Magotari-#uml->>By the bitness of the fs I don't mean the fs's design, but the way the programs inside it were compiled.
10:17<Magotari-#uml->>I helped a person with hangs in init before, and it was the problem, a mismatch.
10:18<IntuitiveNipple-#uml->>nice, defconfig spins too :)
10:18<Magotari-#uml->>Check the bitness of the packages you installed.
10:18<Magotari-#uml->>If 64bits, that explains the problem.
10:19<IntuitiveNipple-#uml->>It's all 32-bit
10:19<Magotari-#uml->>Hmm...
10:20<Magotari-#uml->>Did you patch your host kernel with skas3?
10:23<Magotari-#uml->>Ah well, I don't think I can help you. You need to wait for jdike.
10:25<IntuitiveNipple-#uml->>No, it is using skas0 - standard Ubuntu kernel
10:26<IntuitiveNipple-#uml->>I'm debugging it now, I'll find out why
10:26<Magotari-#uml->>Yeah, I'm sorry but I really don't know why would it fail there.
10:29|-|jdike [~jdike@pool-71-248-190-161.bstnma.fios.verizon.net] has joined #uml
10:29<jdike-#uml->>Hi guys
10:30<Magotari-#uml->>Ah, and there he is.
10:30<Magotari-#uml->>Hey jdike.
10:31<jdike-#uml->>what have I waslked into?
10:31<Magotari-#uml->>jdike: IntuitiveNipple has a problem with a spin in kernel_execve.
10:31<Magotari-#uml->>A 32bit guest and fs on 64bit host.
10:32<IntuitiveNipple-#uml->>Don't worry, I'll get to the bottom of it
10:32<Magotari-#uml->>The host is a stock Ubuntu kernel, the guest is built from source. Even defconfig does not start.
10:32<jdike-#uml->>there are a couple host patches you'll need in order to have 32-bits guests
10:32<IntuitiveNipple-#uml->>uml seems to be spinning in the call to kernel_execve("/sbin/init")
10:32<IntuitiveNipple-#uml->>I had it working fine earlier, it seems something has changed. I'm testing with defconfig atm
10:59<IntuitiveNipple-#uml->>Right. when kernel_execve("/sbin/init") is called, should it return to call do_longjmp() ?
11:01<IntuitiveNipple-#uml->>This is the call trace: http://pastebin.intuitivenipple.net/88
11:06<jdike-#uml->>that looks like a successful exec
11:06<IntuitiveNipple-#uml->>hmmm
11:06<jdike-#uml->>it longjmps back to run_kernel_thread, which returns 1
11:06<IntuitiveNipple-#uml->>ahh, ok
11:06<jdike-#uml->>to new_thread_handler
11:06<jdike-#uml->>which hops into userspace()
11:07<jdike-#uml->>which is the kernel handler for the new process
11:07<IntuitiveNipple-#uml->>is there a way to get gdb to attach to the init process ?
11:08<jdike-#uml->>no
11:08<jdike-#uml->>but what symptoms are you seeing?
11:08<jdike-#uml->>at least you can't attach to the userspace process
11:08<IntuitiveNipple-#uml->>At the point where that trace ends, it is just spinning the CPU forever
11:08<jdike-#uml->>you can attach to the kernel though
11:08<jdike-#uml->>what does strace say?
11:09<IntuitiveNipple-#uml->>42MB of stuff, I checked it from where the printk("/sbin/init) shows up, but I couldn't determine anything, not that I knew what to look for :)
11:10<IntuitiveNipple-#uml->>But there is a lot of this:
11:10<IntuitiveNipple-#uml->>waitpid(12225, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGUSR1}], WSTOPPED) = 12225
11:10<IntuitiveNipple-#uml->>ptrace(PTRACE_SETREGS, 12225, 0, 0x87ca1c0) = 0
11:10<IntuitiveNipple-#uml->>ptrace(PTRACE_SETFPXREGS, 12225, 0, 0x87ca270) = 0
11:10<IntuitiveNipple-#uml->>ptrace(PTRACE_SYSCALL, 12225, 0, SIG_0) = 0
11:10<IntuitiveNipple-#uml->>--- SIGCHLD (Child exited) @ 0 (0) ---
11:10<IntuitiveNipple-#uml->>repeated
11:10<jdike-#uml->>not really
11:10<jdike-#uml->>there is a look, but that's only part of it
11:11<jdike-#uml->>pay attention to the signals coming back from wait
11:11<jdike-#uml->> WSTOPSIG(s) == SIGUSR1
11:11<jdike-#uml->>s/look/loop
11:14<IntuitiveNipple-#uml->>Something like this? http://pastebin.intuitivenipple.net/89
11:14<IntuitiveNipple-#uml->>There's an EAGAIN in there, is that to be expected?
11:17<jdike-#uml->>Are you sure it's hung?
11:17<jdike-#uml->>it's making system calls and doing I/O
11:19<IntuitiveNipple-#uml->>Well, before this started happening, it would proceed with init etc.
11:20<IntuitiveNipple-#uml->>It 'feels' as if it is stuck in a loop someplace, looking at the CPU spinning at about 90%
11:20<jdike-#uml->>the "INIT: version blah" banner doesn't print?
11:20<IntuitiveNipple-#uml->>It won't start with init=/bin/sh etc either
11:21<IntuitiveNipple-#uml->>The last thing I see (without all this debug printing) is "VFS: Mounted root (ext3 filesystem) readonly."
11:21<jdike-#uml->>you pasted stuff from when it seemed truly stuck?
11:22<IntuitiveNipple-#uml->>Yes, when I say stuck... the CPU is being driven hard for *ages* (you can leave it and it is minutes) and there is no further output on the console
11:23<jdike-#uml->>and you pasted stuff from after it had been sitting there for minutes?
11:23<IntuitiveNipple-#uml->>The strace extract? that was from a short time after the /sbin/init call
11:24<IntuitiveNipple-#uml->>This is the section where the call beings: http://pastebin.intuitivenipple.net/90
11:25<jdike-#uml->>at that point, it's happy
11:25<jdike-#uml->>it's faulting in pages and doing I/O
11:25<jdike-#uml->>what about later?
11:25<IntuitiveNipple-#uml->>ok... let me play with it and see what I can find. I bet it is something stupid I've done/not done!
11:26<IntuitiveNipple-#uml->>Now I know what I'm seeing in UML is what is expected, that helps a lot, thanks :)
11:26<jdike-#uml->>just let it sit for 5 minutes
11:26<jdike-#uml->>and then strace it
11:26<jdike-#uml->>and paste a loop
11:27<IntuitiveNipple-#uml->>This is the very last bit of the strace log when I killed the process, you can see a couple iterations of the looping. At this point it had been running about a minute with CPU at 90%
11:27<IntuitiveNipple-#uml->>http://pastebin.intuitivenipple.net/91
11:30<jdike-#uml->>OK, that's a real loop
11:30<jdike-#uml->>what's the host kernel?
11:31<IntuitiveNipple-#uml->>Ubuntu Gutsy x86_64 2.6.22.8
11:31<jdike-#uml->>Feel like patching it?
11:32<IntuitiveNipple-#uml->>Yeah, easy
11:32<IntuitiveNipple-#uml->>Trouble is, I'd rather not since this was working okay all weekend!
11:33<IntuitiveNipple-#uml->>Do you think it could be something as daft as trying a PC restart might clear it?
11:33<jdike-#uml->>same host, same guest?
11:34<IntuitiveNipple-#uml->>guest is i386
11:34<IntuitiveNipple-#uml->>ARCH=um SUBARCH=i386
11:35<jdike-#uml->>same host and guest binaries?
11:35<IntuitiveNipple-#uml->>Yes, 32-bit ELFs
11:37<jdike-#uml->>same host kernel?
11:37<jdike-#uml->>and same UML kernel binary?
11:42<IntuitiveNipple-#uml->>Sorry. Was off getting a cuppa. Yes, regular Ubuntu Gutsy all the way, with my kernel patch to fix UML ftbfs, with SUBARCH=i386
11:46<jdike-#uml->>can you reproduce the working setup?
11:48|-|ram [~ram@pool-72-90-125-50.ptldor.fios.verizon.net] has joined #uml
11:50<IntuitiveNipple-#uml->>lol... that is what I've been trying to do. It stopped working correctly this morning .
11:50<IntuitiveNipple-#uml->>Don't worry too much, I'll solve it and let you know what stupid thing I did :)
11:58<jdike-#uml->>examine your command history :-)
11:58<jdike-#uml->>something changed
12:03<IntuitiveNipple-#uml->>yeah, my brains turned to mush I reckon :p
12:15|-|kowlgi_ [~kowlgi@linux09.ece.utexas.edu] has joined #uml
12:16<kowlgi_-#uml->>I'd like to install some packages on a root_fs for UML. Does someone know of a way to do that?
12:18<jdike-#uml->>then figure out what's happening there
12:18<jdike-#uml->>you're probably going to want that eventually
12:19<jdike-#uml->>yup
12:36|-|tyler29 [~tyler@ARennes-257-1-119-169.w86-210.abo.wanadoo.fr] has joined #uml
12:39<jdike-#uml->>all the popular package managers support installing into an alternate root
12:39<jdike-#uml->>so you can loop-mount the UML filesystem and either do that or chroot to it and update things
12:44|-|the_hydra [~the_hydra@125.164.99.112] has joined #uml
12:44<the_hydra-#uml->>hi all
12:45<the_hydra-#uml->>jdike: hi jeff
12:45<jdike-#uml->>Hi hydra
12:48<the_hydra-#uml->>jdike: how are you lately?
12:49<jdike-#uml->>passable
12:49<the_hydra-#uml->>jdike: days were awful here...got my PSU and proc broken
12:50<jdike-#uml->>yeah, I heard something about a smoking CPU
12:50<the_hydra-#uml->>any plan toward Christmast, jeff?
12:54<jdike-#uml->>not yet
12:57|-|IntuitiveNipple [~TJ@alexandros.tjworld.net] has quit [Quit: The only intuitive interface is the nipple; everything else is learned]
13:13|-|kowlgi_ [~kowlgi@linux09.ece.utexas.edu] has quit [Quit: Leaving]
13:26|-|the_hydra [~the_hydra@125.164.99.112] has left #uml []
13:43|-|IntuitiveNipple [~TJ@alexandros.tjworld.net] has joined #uml
13:47|-|krau [~cktakahas@200.184.118.132] has quit [Quit: Varei!!!]
13:51<IntuitiveNipple-#uml->>I'm debugging the same issue as earlier, on a totally fresh build of uml kernel and userspace. I've been looking at the host 'linux' processes info in /proc/ and noticed that the process that is spinning doesn't report an /exe link. Could this be the issue? http://pastebin.intuitivenipple.net/94
13:51|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has quit [Remote host closed the connection]
13:52<jdike-#uml->>no
13:52<IntuitiveNipple-#uml->>darn!
13:54|-|remus [~remus@76.231.178.131] has joined #uml
14:04|-|krau [~cktakahas@200.184.118.132] has joined #uml
14:08|-|tyler29 [~tyler@ARennes-257-1-119-169.w86-210.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
14:21|-|tyler29 [~tyler@ARennes-257-1-63-214.w81-53.abo.wanadoo.fr] has joined #uml
14:30|-|IntuitiveNipple [~TJ@alexandros.tjworld.net] has quit [Quit: The only intuitive interface is the nipple; everything else is learned]
14:48|-|kos_tom [~thomas@col31-3-82-247-183-72.fbx.proxad.net] has joined #uml
15:02|-|tyler29 [~tyler@ARennes-257-1-63-214.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
15:19|-|tyler29 [~tyler@ARennes-257-1-96-168.w81-48.abo.wanadoo.fr] has joined #uml
15:44<Magotari-#uml->>jdike: I have been working on the patch testing thing, trying to do run_tests()... Right now it is able to do simple host sanity checks, nothing with UML yet. I could get that part done in a few days, but I am quite dissatisfied with the only solution I could come up with to deal with the tests, so I think I shall scrap this project.
15:44<jdike-#uml->>that was quick
15:44<Magotari-#uml->>Right now writing tests is pretty close to bash scripting.
15:45<Magotari-#uml->>I still need to figure out how to execute commands inside UML and in mconsole, but other than that the test part is pretty much ready, result checking included.
15:46<Magotari-#uml->>Anyway, I will keep you posted as significant progress happens.
15:46<Magotari-#uml->>C is fun, btw.
15:46<Magotari-#uml->>Thanks for letting me do those tiny fixup patches, that really encouraged me.
15:51|-|tyler29 [~tyler@ARennes-257-1-96-168.w81-48.abo.wanadoo.fr] has quit [Quit: ++]
15:53<jdike-#uml->>np, your work was useful
16:02<remus-#uml->>jdike: I got a UML crash the other day
16:03<remus-#uml->>jdike: I think it's a UML bug but I'm not sure ... it might be my code though
16:03<remus-#uml->>jdike: the stack trace does not point to my code though
16:03<remus-#uml->>jdike: do you want to take a look ?
16:03<Magotari-#uml->>I'd like to see, just curious.
16:04<Magotari-#uml->>(If it ain't a secret, of course.)
16:05<remus-#uml->>http://rafb.net/p/MbUx1R57.html
16:05<remus-#uml->>Magotari: nope, no secret
16:05<remus-#uml->>Magotari: link above
16:06<Magotari-#uml->>To my eye the thing I see is segv handler twice.
16:06<remus-#uml->>Magotari: it looks like it's blowing in strlen
16:06<Magotari-#uml->>Yeah, it seems to.
16:06<Magotari-#uml->>Well, I'm useless, but thanks for letting me look.
16:07<remus-#uml->>Magotari: you're welcome ;-)
16:08<remus-#uml->>Magotari, jdike: BTW, I don't get a panic when running on a real machine
16:08<Magotari-#uml->>Does it crash deterministically?
16:08<remus-#uml->>on UML I can reproduce this though
16:08<remus-#uml->>yes
16:08|-|hrishi [~hrishi@lawn-143-215-205-55.lawn.gatech.edu] has joined #uml
16:08<Magotari-#uml->>Hmm... Did you try it without that module?
16:09<Magotari-#uml->>I gather that is your code, right?
16:09<remus-#uml->>well, this is where I can crash it ... from the module ;-)
16:09<remus-#uml->>yes, that is my code
16:09<remus-#uml->>that's why I say it could be my code too
16:09<Magotari-#uml->>Ah. I see. Nevermind then.
16:10<remus-#uml->>maybe it's doing something nasty (i.e. memory corruption) and as a result it blows up somewhere else
16:11<remus-#uml->>but maybe not ;-)
16:11<Magotari-#uml->>601f3d18: [<60011842>] segv_handler+0x0/0x6e
16:12<Magotari-#uml->>Interesting to me. I will have to read about the syntax of the traces.
16:12<Magotari-#uml->>I am really curious of what are the things after the +
16:12<Magotari-#uml->>0x0/0x6e
16:17|-|hrishi [~hrishi@lawn-143-215-205-55.lawn.gatech.edu] has quit [Quit: Leaving]
16:19|-|hrishi [~hamur3@legolas.cc.gatech.edu] has joined #uml
16:20<remus-#uml->>Magotari: + is the offset inside the function
16:27<jdike-#uml->>be nice to know who called strlen
16:28<remus-#uml->>jdike: hmm ... I agree ...
16:28<jdike-#uml->>you got a core dump, right?
16:28<remus-#uml->>jdike: that module that I am writing is a filesystem module
16:29<remus-#uml->>no ... how do I turn that on ?
16:31<jdike-#uml->>disable your core dump rlimit
16:32<jdike-#uml->>ulimit -a in bash shows the limit
16:32<jdike-#uml->>s
16:32<Magotari-#uml->>You also need core dump support in the host kernel.
16:33<jdike-#uml->>I've never seen a kernel that didn't have it
16:33<Magotari-#uml->>Mine don't. :)
16:33<jdike-#uml->>didn't even know it was configurable
16:33<jdike-#uml->>why not?
16:33<Magotari-#uml->>No idea, I just never use that feature, so why bother having it.
16:34<jdike-#uml->>remus, anyway get a core dump, gdb it afterwards, and the gdb backtrace should be more informative than the panic
16:34<jdike-#uml->>or just gdb it, put a breakpoint on panic, and do whatever you do
16:34<remus-#uml->>jdike: ok, doing that right now
16:35<remus-#uml->>jdike: how do I put a breakpoint on panic ?
16:35<jdike-#uml->>b panic
16:35<jdike-#uml->>doesn't get much easier than that
16:36<remus-#uml->>jdike: thanks ... I'll post the results
16:46<hrishi-#uml->>Hi, I am trying to profile itimer usage in uml from the host kernel..
16:47<hrishi-#uml->>but things seem to go wrong whenever I use a real timer in a uml user process
16:47<hrishi-#uml->>am i not allowed to do that?
16:47<jdike-#uml->>sure
16:48<hrishi-#uml->>since each new uml user process maps to a new process on the host kernel, i thought this should be possible
16:52<jdike-#uml->>it wouldn't matter if a UML process didn't map to a host process
16:52<jdike-#uml->>that's an implementation detail
16:54<hrishi-#uml->>right, anyway my point was that it is a separate process, so it has its own interval timers
16:55<jdike-#uml->>doesn't matter
16:55<jdike-#uml->>it would have its own interval timers even if it weren't a separate process
16:56<hrishi-#uml->>right, but the uml kernel itself uses its real timer for timekeeping
16:56<jdike-#uml->>only when it's sleeping
16:56<jdike-#uml->>when it's running, it uses its virtual timer
16:57|-|tyler29 [~tyler@ARennes-257-1-155-100.w86-214.abo.wanadoo.fr] has joined #uml
16:58<hrishi-#uml->>oh! i didnt know that. that makes sense then.
16:59<hrishi-#uml->>thanks, that should've struck me!
17:01<jdike-#uml->>but what's wrong with real timers in UML userspace?
17:03|-|dang [~dang@aa-redwall.nexthop.com] has quit [Quit: Leaving.]
17:06|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has joined #uml
17:15<hrishi-#uml->>sorry, what i have is a way in the host kernel to relate a process to which uml kernel it is running on (in case there are more than 1)
17:16<hrishi-#uml->>so this way you can sort of get the itimer usage profile for the uml kernel and all the processes running under it
17:17<jdike-#uml->>Ah
17:18<hrishi-#uml->>but when i run a user process in uml, that simply loops setting itimers, the setitimers from the uml kernel process stopped appearing
17:18<hrishi-#uml->>that was explained..
17:18|-|tyler29 [~tyler@ARennes-257-1-155-100.w86-214.abo.wanadoo.fr] has quit [Quit: ++]
17:18<hrishi-#uml->>i cant see the setitimers from this user process as well, but that may be a bug in the code
17:19<jdike-#uml->>there's only one
17:19<jdike-#uml->>when the process is first created
17:19<hrishi-#uml->>yeah i meant the setitimer() system calls :-)
17:20<jdike-#uml->>you mean the process calling setitimer for itself
17:20<jdike-#uml->>or UML calling setitimer on the host in order to set up a tick?
17:21<hrishi-#uml->>the process calling for itself
17:21<jdike-#uml->>OK
17:21<jdike-#uml->>you won't see that on the host
17:21<hrishi-#uml->>oh i won't? how come?
17:21<jdike-#uml->>because that's not how the kernel does timers
17:22<jdike-#uml->>think of the host as hardware
17:22<jdike-#uml->>the kernel has to manage timers on a system-wide basis
17:23<jdike-#uml->>when a process calls setitimer, UML just drops that info into a list of timers
17:23<jdike-#uml->>and when enough ticks have happened, the next event gets taken off that list, and something done with it
17:23<hrishi-#uml->>ah!
17:25<hrishi-#uml->>ok.. so it doesnt go up to the host at all! evil..
17:25<jdike-#uml->>with a tickless UML, if you're very very careful, you might be able to see a UML setitimer turn into a host setitimer
17:25<jdike-#uml->>but the idle loop does timing by just sleeping until the next event is due
17:26<jdike-#uml->>so you could see a nanosleep of the appropriate duration as well
17:27<hrishi-#uml->>ok cool. i guess i need to check the source to settle that in
17:29<hrishi-#uml->>well thanks for that! id been stuck on this for a while
17:30<jdike-#uml->>np
17:30<jdike-#uml->>UML process system calls rarely translate directly into host system calls
17:37|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has joined #uml
17:44<remus-#uml->>jdike: I am in UML/gdb but I couldn't get any more info
17:44<remus-#uml->>jdike: segv trace is the same
17:44<remus-#uml->>jdike: when I type bt it tells me "No stack."
17:45<remus-#uml->>jdike: creating a core didn't work very well either ...
17:46<remus-#uml->>hold on ... maybe I did the core thing wrong .... trying again ...
17:50<remus-#uml->>jdike: ok, this time was a bit better but it still doesn't tell me who called strlen
17:51<remus-#uml->>http://rafb.net/p/Njsiyw76.html
18:32<Magotari-#uml->>jdike: Any idea if UML can boot from a BSD partitioned disk?
18:32<Magotari-#uml->>Er, Image.
18:33<Magotari-#uml->>I think it should, but I don't want to download a bsd and make an image in qemu to only find out that it does not.
18:36<Magotari-#uml->>I am really damn curious how a crash will look when I try to boot FreeBSD with uml.
18:37<Magotari-#uml->>Probably at the first syscall, as I understand the BSD and Linux calling conventions are different, but...
18:39<Magotari-#uml->>The damn curiosity just won't let up.
18:40|-|kos_tom [~thomas@col31-3-82-247-183-72.fbx.proxad.net] has quit [Quit: I like core dumps]
18:57<jdike-#uml->>odd
18:57<jdike-#uml->>gdb just doesn't want to tell you who called strlen
18:59<jdike-#uml->>I wonder if you have stack corruption on top of that
18:59<jdike-#uml->>I would look at whatever local buffers you have and see if you're overflowing any of them
19:29|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
19:34|-|dang [~dang@nemesis.fprintf.net] has joined #uml
19:38|-|dang [~dang@nemesis.fprintf.net] has left #uml []
20:45|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has quit [Remote host closed the connection]
20:47|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has joined #uml
21:08|-|jdike [~jdike@pool-71-248-190-161.bstnma.fios.verizon.net] has quit [Quit: Leaving]
22:41|-|balbir [~balbir@122.167.179.130] has quit [Ping timeout: 480 seconds]
23:06|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has quit [Remote host closed the connection]
23:17|-|Baltam [~WIKIMOKI@tor-irc.dnsbl.oftc.net] has joined #uml
---Logclosed Tue Nov 20 00:00:09 2007