| |
| |
| |
|
Page: 1 2 3
Comments:
<0> nordle: you tried changing the guest timing method? such as p***ing clock=pit for guest bootloader at boot time?
<0> have you*
<1> dignome: yes
<1> dignome: bootloader = grub, just append clock=pit as mentioned in the help docs, although it doesn't seem to make any difference.
<2> btw my host is linux on amd64 and guest clock is way off wallclock even with clock=pit
<2> I do not use acceleration module
<0> ;/
<1> garrison: same as, it takes 34 seconds to do 1 minutes
<0> garrison: have you tried a printf() on ticks_per_sec once it is set in vl.c cpu_calibrate_ticks() ?
<2> not sure actually
<2> dignome: no, but I suspect the problem is with cpu frequency daemon :)
<0> yeah - always good to disable those before starting qemu ;p
<3> cool'n'quiet must die
<2> or find a way to live with that :)
<1> ahh, is that the problem, the performance governer, as mine is set to userland and fluctuates.
<0> garrison: if hint is doing what I think he's doing it won't be problem in the future.
<3> dignome: and what do you think he's doing?
<2> :)
<0> making the guest be in charge of tsc
<0> well. i mean make qemu in charge of it
<3> that wouldn't solve all the problems
<3> in fact the naive way of doing vtsc is going to solve nothing
<0> ok
<0> so it would not have some dynamic tick counter?
<3> the proper solution is to be notified by the kernel whenever cpu freq governer changes the freq, and even better stable clock source (cheaply)accessible from the user space
<0> no way of answering that... i meant if it had one would it help?
<3> how?
<3> i mean you have a stable vtsc.. then what?
<0> malc_: well. having the vtsc would also require you do something about ticks_per_sec, no?
<3> dignome: not necessarily, in fact hint just wanted to count insns
<0> ok. i misread the conversation as usual ;/
<3> dignome: is the vista beta still downloadable btw?
<0> I think you still get it, but they don't give out keys any more.
<3> dignome: okay. thanks
<4> malc_: Using a virtual TSC forces you to, and to a large extend provides the ifrastructure for solving the host speedstep problems.
<3> pbrook: how?
<4> Because it's almost never directly linked to real time.
<3> you still need a stable host clock source
<4> So you either get it right, or you get it so wrong it's not even close.
<3> in that sense, maybe
<4> The reason qemu uses the host TSC is beacause it's a high-accuracy, low overhead timesource. Except when it isn't :-)
<3> yeah yeah
<4> Using a virtual TSC forces you to decouple virtual time form real time.
<3> only if guest actually uses TSC for timing
<3> which it might as well not do
<3> given much more stable clock sources (such as acpi clock or somesuch)
<4> Well, you could implement a virtual TSC and not use it for anything, but that would be rather pointless.
<0> garrison: http://rafb.net/paste/results/BTBTIi96.html - substitued with uint8_t and uint16_t - compiled.
<4> The point of the virtual TSC is that it's used as the base for all choronological events visible to the guest.
<3> that's what fabrice wants it for
<3> determinisim
<3> -i
<4> Exactly.
<3> he's 3 years late though, and would have saved me a lot of head scratching
<3> oh well
<4> Once you have determinism it's realtively easy to do rate adaptive whatsits to keep in sync with a low-fi realtime.
<2> dignome: strange, compiles fine here...
<2> dignome: though your change is correct - these are left over from slirp.h dependency
<2> dignome: heh, sys/types.h has a typedef here for u_int8_t and u_int16_t (glibc-2.4)
<5> hi
<5> i have some problems to run a mipsel system on x86
<5> if i try qemu-mipsel with -L target target/bin/busybox it tells me that /lib/xx.so could not be found... how to tell that target is the root dir
<5> http://www.onlyfree.de/php/pasteservice/show.php?id=710
<5> if i try to use system-mipsel emulation
<5> any hints?
<4> -L specifies the sysroot.
<4> /lib/xxx.so should be relative to that directory.
<5> it is
<5> ls -l target/lib/
<5> total 332
<5> -rwxr-xr-x 1 crissi crissi 26992 Jul 3 17:57 ld-uClibc.so.0
<5> -rwxr-xr-x 1 crissi crissi 246048 Jul 3 17:57 libc.so.0
<5> -rwxr-xr-x 1 crissi crissi 14208 Jul 3 17:57 libcrypt.so.0
<5> -rwxr-xr-x 1 crissi crissi 9040 Jul 3 17:57 libdl.so.0
<5> -rwxr-xr-x 1 crissi crissi 10260 Jul 3 17:57 libm.so.0
<5> -rwxr-xr-x 1 crissi crissi 1972 Jul 3 17:57 libnsl.so.0
<5> -rwxr-xr-x 1 crissi crissi 5444 Jul 3 17:57 libutil.so.0
<5> drwxr-xr-x 3 crissi crissi 4096 Jul 3 17:57 modules
<5> qemu-mipsel -v
<5> qemu-mips version 0.8.1, Copyright (c) 2003-2005 Fabrice Bellard
<5> qemu-mipsel -L .target/ target/bin/busybox
<5> /lib/ld-uClibc.so.0: No such file or directory
<5> Segmentation fault
<3> with -L the directory name is .target with a leading dot...
<3> wtf?
<5> qemu-mipsel -L ./target target/bin/busybox
<5> /lib/ld-uClibc.so.0: No such file or directory
<5> Segmentation fault
<5> its simply dont do the chroot
<5> any hints?
<0> Crissi: for the system target there is a small hint on how to start it here: http://www.linux-mips.org/wiki/Qemu
<5> http://www.onlyfree.de/php/pasteservice/show.php?id=711
<5> i have already read that
<5> but it doesnt works
<0> "The Linux kernel has a special configuration to support the Qemu platform, no additional patches are needed."
<0> I suppose you built it with that configuration?
<5> http://www.onlyfree.de/php/pasteservice/show.php?id=712
<0> i meant the kernel that you're p***ing to the mips system target
<5> the kernel is fine
<5> qemu: fatal: Trying to execute code outside RAM or ROM at 0xbfc00380
<5> that is the problem
<4> Crissi: You're cofusing teh full system and usermode emulation.
<5> no.
<4> qemu-mipsel runs mips -linux binaries and proxies the syscalls to the host kernel, kinda like wine.
<4> qemu-system-mips emulates acompliete mips-based machine.
<5> usermode dont find the lib
<5> the system emu wont start
<5> i think its a bug
<5> because i have more then one systems for test
<5> for examle the system of the ar7port project
<5> they also wrote:
<5> Run QEMU from a command shell like this:
<5>
<5> ./qemu-system-mipsel -L . -kernel vmlinux -nographic
<5>
<5> You should see the boot messages from the Linux kernel and get
<5> a shell prompt after pressing ENTER.
<5> that starts the kernel in system emu
<5> just i do.
<5> with this system
<5> but i got:
<5> http://www.onlyfree.de/php/pasteservice/show.php?id=711
<5> maybe the cvs of qemu is buggy?
<4> Possible. Maybe you used the wrong kernel.
<0> you mention qemu-system-mipsel - but that paste shows qemu-system-mips (are they not different?)
<5> http://developer.berlios.de/project/showfiles.php?group_id=3721&release_id=10146
<5> http://www.onlyfree.de/php/pasteservice/show.php?id=713
<5> its seems to be a bug in qemu. the patched version of qemu runs: http://prdownload.berlios.de/ar7-firmware/qemu-ar7.tar.bz2
<5> this is the main download page http://developer.berlios.de/project/showfiles.php?group_id=3721&release_id=10146
<0> hm. what is mipsel-ar7-softmmu?
<5> its a patched qemu for ar7 cpu
<5> mips
<5> mipsel
<0> and qemu cvs doesn't support the ar7 cpu?
<5> its dont seems so...
<0> I thought you said there wasn't a problem with the kernel image?
<5> try it: http://prdownload.berlios.de/ar7-firmware/vmlinux.bz2 http://prdownload.berlios.de/ar7-firmware/flashimage.bin.bz2
<0> well. is there a point in trying a kernel image built to run on ar7 cpu when qemu doesn't suppor this officially?
<5> ar7 is working by running the resulting http://prdownload.berlios.de/ar7-firmware/qemu-system-mipsel.bz2
<4> Crissi: Sounds like you're trying to run a different kernel on a patched version of qemu. Ask whoever make that kernel and/or patch.
<5> download the 3 file
<5> and then do: ./qemu-system-mipsel -L . -kernel vmlinux -nographic
<5> after unpacking
<5> if you try to use the normal qemu: http://www.onlyfree.de/php/pasteservice/show.php?id=713
<5> so its interesting to investigate that
<4> Crissi: Use normal qemu with a kenrel compiled for qemu.
<4> Anything else we can't really help you with
<4> Using the wrong kernel (when wrong is any sort of mismatch between what qemu emulates and what the kernel was compiled for) almost certainly won't work.
<4> Same as with real hardware.
<5> pbrook: understand me... i had try normal qemu... it wornt work!
<4> Are you using a kernel compiled for wemu?
Return to
#qemu
or
Go to some related
logs:
cups sucks ps pdf
#gaim
reiserfs data=journal memory leak
#php
acer + linux + eth config
Fatal error: Call to undefined function imageantialias() debian
update-initramfs ubuntu
xf86config-4 damn small linux
accidentally chowned files
nessus-services could not be found
|
|