[Qemu-devel] QEMU regressions ...

[Qemu-devel] QEMU regressions ...

[Herbert Poetzl]

Hi Fabrice!

I'm using QEMU to do kernel development, and because
one of the older redhat kernels, continuously core
dups qemu, I thought, I'll check the cvs version ...

the checkconfig works well, but then the trouble starts ...

  texi2html -monolithic -number qemu-doc.texi
  make: texi2html: Command not found

I don't use texi2html, and I don't want the docu to
be built, so I replace the command with a noop
(actually touch qemu-doc.html)

this seems to do the trick, as the compile runs, and
completes without too many warnings ...

then I wanted to test the vl command, but instead I 
discovered that same command wasn't built ...

I tried make vl, and ideed, something happened:

  gcc -Wall -O2 -g  -g  vl.c   -o vl
  vl.c:47:17: cpu.h: No such file or directory
  In file included from vl.c:49:
  thunk.h:24:20: config.h: No such file or directory
  make: *** [vl] Error 1

did you abandon the vl utility? is there any 'good'
reason for not using it? please advise ...

TIA,
Herbert

[Qemu-devel] Re: QEMU regressions ...

[Herbert Poetzl]

On Sun, Oct 26, 2003 at 11:05:09PM +0100, Fabrice Bellard wrote:
> Hi,
> 
> In the CVS version 'vl' was renamed to 'qemu'. Make install should 
> install it. I will document all that as soon as the release is made (I 
> cannot tell when - I need more time to work on it).

hmm, okay obviously my fault, now 'qemu' binary seems to
work, but the redhat kernel still segfaults/coredumps
this is 2.4.18-27.7.x and I'm not able to get more than

# qemu-cvs-26.10.2003/i386/qemu -nographic -m 128 -snapshot -hda IMGs/TEST_32M.img -hdb IMGs/TEST_256M.img -kernel /usr/src/ALEXEY/kernel-2.4.18-27.7.x-P1/arch/i386/boot/bzImage -append "rw root=/dev/hda1"
warning: could not open /dev/net/tun: no virtual network emulation
Segmentation fault (core dumped)

works with vanilla 2.4.20,21,22,23-pre8
any hints for that one?

TIA,
Herbert


> 
> Fabrice.
> 
> Herbert Poetzl wrote:
> >Hi Fabrice!
> >
> >I'm using QEMU to do kernel development, and because
> >one of the older redhat kernels, continuously core
> >dups qemu, I thought, I'll check the cvs version ...
> >
> >the checkconfig works well, but then the trouble starts ...
> >
> >  texi2html -monolithic -number qemu-doc.texi
> >  make: texi2html: Command not found
> >
> >I don't use texi2html, and I don't want the docu to
> >be built, so I replace the command with a noop
> >(actually touch qemu-doc.html)
> >
> >this seems to do the trick, as the compile runs, and
> >completes without too many warnings ...
> >
> >then I wanted to test the vl command, but instead I 
> >discovered that same command wasn't built ...
> >
> >I tried make vl, and ideed, something happened:
> >
> >  gcc -Wall -O2 -g  -g  vl.c   -o vl
> >  vl.c:47:17: cpu.h: No such file or directory
> >  In file included from vl.c:49:
> >  thunk.h:24:20: config.h: No such file or directory
> >  make: *** [vl] Error 1
> >
> >did you abandon the vl utility? is there any 'good'
> >reason for not using it? please advise ...
> >
> >TIA,
> >Herbert
> >
> >
> >
> 

[Qemu-devel] Re: QEMU regressions ...

[Fabrice Bellard]

It seems that the page table is remapped at an invalid address:

IN:
0x00100026:  movl   $0x30101000,%eax
0x0010002b:  movl   %eax,%cr3

I think more patches are needed to remap this kernel to a right address. 
BUT...

If you take my latest patches, you can use with QEMU any _unpatched_ 
linux kernel now :-)

Fabrice.

Herbert Poetzl wrote:
> On Sun, Oct 26, 2003 at 11:05:09PM +0100, Fabrice Bellard wrote:
> 
>>Hi,
>>
>>In the CVS version 'vl' was renamed to 'qemu'. Make install should 
>>install it. I will document all that as soon as the release is made (I 
>>cannot tell when - I need more time to work on it).
> 
> 
> hmm, okay obviously my fault, now 'qemu' binary seems to
> work, but the redhat kernel still segfaults/coredumps
> this is 2.4.18-27.7.x and I'm not able to get more than
> 
> # qemu-cvs-26.10.2003/i386/qemu -nographic -m 128 -snapshot -hda IMGs/TEST_32M.img -hdb IMGs/TEST_256M.img -kernel /usr/src/ALEXEY/kernel-2.4.18-27.7.x-P1/arch/i386/boot/bzImage -append "rw root=/dev/hda1"
> warning: could not open /dev/net/tun: no virtual network emulation
> Segmentation fault (core dumped)
> 
> works with vanilla 2.4.20,21,22,23-pre8
> any hints for that one?
> 
> TIA,
> Herbert
> 
> 
> 
>>Fabrice.
>>
>>Herbert Poetzl wrote:
>>
>>>Hi Fabrice!
>>>
>>>I'm using QEMU to do kernel development, and because
>>>one of the older redhat kernels, continuously core
>>>dups qemu, I thought, I'll check the cvs version ...
>>>
>>>the checkconfig works well, but then the trouble starts ...
>>>
>>> texi2html -monolithic -number qemu-doc.texi
>>> make: texi2html: Command not found
>>>
>>>I don't use texi2html, and I don't want the docu to
>>>be built, so I replace the command with a noop
>>>(actually touch qemu-doc.html)
>>>
>>>this seems to do the trick, as the compile runs, and
>>>completes without too many warnings ...
>>>
>>>then I wanted to test the vl command, but instead I 
>>>discovered that same command wasn't built ...
>>>
>>>I tried make vl, and ideed, something happened:
>>>
>>> gcc -Wall -O2 -g  -g  vl.c   -o vl
>>> vl.c:47:17: cpu.h: No such file or directory
>>> In file included from vl.c:49:
>>> thunk.h:24:20: config.h: No such file or directory
>>> make: *** [vl] Error 1
>>>
>>>did you abandon the vl utility? is there any 'good'
>>>reason for not using it? please advise ...
>>>
>>>TIA,
>>>Herbert
>>>
>>>
>>>
>>
> 
> 

[Qemu-devel] Re: QEMU regressions ...

[Herbert Poetzl]

On Tue, Oct 28, 2003 at 12:38:05AM +0100, Fabrice Bellard wrote:
> It seems that the page table is remapped at an invalid address:
> 
> IN:
> 0x00100026:  movl   $0x30101000,%eax
> 0x0010002b:  movl   %eax,%cr3
> 
> I think more patches are needed to remap this kernel to a right address. 
> BUT...
> 
> If you take my latest patches, you can use with QEMU any _unpatched_ 
> linux kernel now :-)

hmm, what patches? I used the cvs version ...

but this will use the 'slower' soft-mmu right?
what about gdb support, any improvement on that?

TIA,
Herbert

> Fabrice.
> 
> Herbert Poetzl wrote:
> >On Sun, Oct 26, 2003 at 11:05:09PM +0100, Fabrice Bellard wrote:
> >
> >>Hi,
> >>
> >>In the CVS version 'vl' was renamed to 'qemu'. Make install should 
> >>install it. I will document all that as soon as the release is made (I 
> >>cannot tell when - I need more time to work on it).
> >
> >
> >hmm, okay obviously my fault, now 'qemu' binary seems to
> >work, but the redhat kernel still segfaults/coredumps
> >this is 2.4.18-27.7.x and I'm not able to get more than
> >
> ># qemu-cvs-26.10.2003/i386/qemu -nographic -m 128 -snapshot -hda 
> >IMGs/TEST_32M.img -hdb IMGs/TEST_256M.img -kernel 
> >/usr/src/ALEXEY/kernel-2.4.18-27.7.x-P1/arch/i386/boot/bzImage -append "rw 
> >root=/dev/hda1"
> >warning: could not open /dev/net/tun: no virtual network emulation
> >Segmentation fault (core dumped)
> >
> >works with vanilla 2.4.20,21,22,23-pre8
> >any hints for that one?
> >
> >TIA,
> >Herbert
> >
> >
> >
> >>Fabrice.
> >>
> >>Herbert Poetzl wrote:
> >>
> >>>Hi Fabrice!
> >>>
> >>>I'm using QEMU to do kernel development, and because
> >>>one of the older redhat kernels, continuously core
> >>>dups qemu, I thought, I'll check the cvs version ...
> >>>
> >>>the checkconfig works well, but then the trouble starts ...
> >>>
> >>>texi2html -monolithic -number qemu-doc.texi
> >>>make: texi2html: Command not found
> >>>
> >>>I don't use texi2html, and I don't want the docu to
> >>>be built, so I replace the command with a noop
> >>>(actually touch qemu-doc.html)
> >>>
> >>>this seems to do the trick, as the compile runs, and
> >>>completes without too many warnings ...
> >>>
> >>>then I wanted to test the vl command, but instead I 
> >>>discovered that same command wasn't built ...
> >>>
> >>>I tried make vl, and ideed, something happened:
> >>>
> >>>gcc -Wall -O2 -g  -g  vl.c   -o vl
> >>>vl.c:47:17: cpu.h: No such file or directory
> >>>In file included from vl.c:49:
> >>>thunk.h:24:20: config.h: No such file or directory
> >>>make: *** [vl] Error 1
> >>>
> >>>did you abandon the vl utility? is there any 'good'
> >>>reason for not using it? please advise ...
> >>>
> >>>TIA,
> >>>Herbert
> >>>
> >>>
> >>>
> >>
> >
> >
> 

[Qemu-devel] Re: QEMU regressions ...

[Herbert Poetzl]

On Tue, Oct 28, 2003 at 01:03:54AM +0100, Fabrice Bellard wrote:
> Herbert Poetzl wrote:
> >On Tue, Oct 28, 2003 at 12:38:05AM +0100, Fabrice Bellard wrote:
> >
> >>It seems that the page table is remapped at an invalid address:
> >>
> >>IN:
> >>0x00100026:  movl   $0x30101000,%eax
> >>0x0010002b:  movl   %eax,%cr3
> >>
> >>I think more patches are needed to remap this kernel to a right address. 
> >>BUT...
> >>
> >>If you take my latest patches, you can use with QEMU any _unpatched_ 
> >>linux kernel now :-)
> >
> >
> >hmm, what patches? I used the cvs version ...
> I just commited new patches - I am preparing a new release.
okay, so a 4.4 release is comming up, right?

> >but this will use the 'slower' soft-mmu right?
> Yes.
was there some testing how much slower the soft-mmu is?
or how much overhead it adds?

> >what about gdb support, any improvement on that?
> gdb should work better (single step is implemented but 
> I did not use it a lot).
okay, will test it ...

thanks for doing QEMU,
Herbert

> Fabrice.

[Qemu-devel] os x support & panther

[Satadru Pramanik]

[-- Attachment #1: Type: text/plain, Size: 1443 bytes --]

First, congratulations on the release of qemu 5.0.  The feature set 
looks really nice.

I know that the issue of OS X support has come up recently on this 
list, but had a tiny bit to chip in as I hadn't seen much activity 
regarding it here in the last week.

OS X 10.3 came out this weekend, and Apple's page on OS X 10.3 states:

Panther emulates the System V poll system call on top of BSDís native 
select API, and the dlopen/dlclose dynamic library access routines “ 
used with Executable and Linking Format (ELF) binaries on Linux, BSD, 
and other operating systemsóon top of the native Mach-O format used on 
Mac OS X. With these routines available on Mac OS X, developers of 
cross-platform UNIX applications can port libraries and applications to 
Mac OS X with little or no change to their existing applications.

Does this mean that we should be able to run a linuxppc binary in some 
sort of chroot on an OS X System without any changes?  The whole 
concept of supporting qemu on OS X is a bit of a sharecropper's dream, 
but if we can just run the linuxppc binary on an OS X system, that 
would be sufficient to run things like a linux/i386 version of wine, 
right?

I likely know nothing about what I am talking about, but seeing some 
names on this list from the days ( late 90s) when I ran linuxppc makes 
me very happy, especially with a new powerbook that I am eager to try 
qemu on...

Satadru

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 186 bytes --]