[Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 425 bytes --]

Here are two patches, loosely based on patches in OpenBSD ports system
that enable compiling Qemu. The first patch makes AIO use conditional
on host AIO support. The second one adjusts various include files etc.

I'd like to commit both of these, is that OK?

Background: I got fed up with Sparc glibc mangling global registers
and buggy gdb/gdb64, so I installed OpenBSD. Now debugging Sparc host
support is so much easier.

[-- Attachment #2: config_aio.diff --]
[-- Type: plain/text, Size: 2913 bytes --]

[-- Attachment #3: fix_openbsd_build.diff --]
[-- Type: plain/text, Size: 6973 bytes --]

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Anthony Liguori]

Blue Swirl wrote:
> Here are two patches, loosely based on patches in OpenBSD ports system
> that enable compiling Qemu. The first patch makes AIO use conditional
> on host AIO support. The second one adjusts various include files etc.
>
> I'd like to commit both of these, is that OK?
>   

The first patch is fine but the second patch has some bogus things in 
it.  For instance:
>
> Index: qemu/configure
> ===================================================================
> --- qemu.orig/configure    2008-08-12 17:12:15.000000000 +0000
> +++ qemu/configure    2008-08-12 17:33:36.000000000 +0000
> @@ -150,6 +150,7 @@
> @@ -1076,6 +1081,14 @@
>      echo "#define HAVE_BYTESWAP_H 1" >> $config_h
>    fi
>  fi
> +
> +if [ "$openbsd" = "yes" ] ; then
> +  echo "#define ENOTSUP 4096" >> $config_h
> +  echo "#define qemu_siginfo siginfo_t" >> $config_h
> +else
> +  echo "#define qemu_siginfo struct siginfo" >> $config_h
> +fi

But qemu_siginfo isn't actually used anywhere.  That's just one example, 
the patch seems to have a lot of things like this in it.

Regards,

Anthony Liguori

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Todd T. Fries]

I am responsible for that.  Back when the qemu siginfo stuff was in use
it was necessary.

Sorry for the extra cruft, will look for cleaning it when qemu releases
next..

Looking forward to the next release with OpenBSD/sparc running inside
qemu ;-)

Thanks,
-- 
Todd Fries .. todd@fries.net

 _____________________________________________
|                                             \  1.636.410.0632 (voice)
| Free Daemon Consulting, LLC                 \  1.405.227.9094 (voice)
| http://FreeDaemonConsulting.com             \  1.866.792.3418 (FAX)
| "..in support of free software solutions."  \          250797 (FWD)
|                                             \
 \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
                                                 
              37E7 D3EB 74D0 8D66 A68D  B866 0326 204E 3F42 004A
                        http://todd.fries.net/pgp.txt

Penned by Anthony Liguori on 20080812 14:46.53, we have:
> Blue Swirl wrote:
>> Here are two patches, loosely based on patches in OpenBSD ports system
>> that enable compiling Qemu. The first patch makes AIO use conditional
>> on host AIO support. The second one adjusts various include files etc.
>>
>> I'd like to commit both of these, is that OK?
>>   
>
> The first patch is fine but the second patch has some bogus things in  
> it.  For instance:
>>
>> Index: qemu/configure
>> ===================================================================
>> --- qemu.orig/configure    2008-08-12 17:12:15.000000000 +0000
>> +++ qemu/configure    2008-08-12 17:33:36.000000000 +0000
>> @@ -150,6 +150,7 @@
>> @@ -1076,6 +1081,14 @@
>>      echo "#define HAVE_BYTESWAP_H 1" >> $config_h
>>    fi
>>  fi
>> +
>> +if [ "$openbsd" = "yes" ] ; then
>> +  echo "#define ENOTSUP 4096" >> $config_h
>> +  echo "#define qemu_siginfo siginfo_t" >> $config_h
>> +else
>> +  echo "#define qemu_siginfo struct siginfo" >> $config_h
>> +fi
>
> But qemu_siginfo isn't actually used anywhere.  That's just one example,  
> the patch seems to have a lot of things like this in it.
>
> Regards,
>
> Anthony Liguori
>
>

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Anthony Liguori]

Hi Todd,

Todd T. Fries wrote:
> I am responsible for that.  Back when the qemu siginfo stuff was in use
> it was necessary.
>
> Sorry for the extra cruft, will look for cleaning it when qemu releases
> next..
>   

It would be good to get SVN working consistently on OpenBSD/sparc.  
We're probably a ways off from another release as we still have 
regressed platforms due to the addition of TCG.

Regards,

Anthony Liguori

> Looking forward to the next release with OpenBSD/sparc running inside
> qemu ;-)
>
> Thanks,
>   

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

On 8/12/08, Todd T. Fries <qemu-devel@email.fries.net> wrote:
> I am responsible for that.  Back when the qemu siginfo stuff was in use
>  it was necessary.

Okay, I'll refer to you in the commit message so you'll get the blame. ;-)

>  Sorry for the extra cruft, will look for cleaning it when qemu releases
>  next..
>
>  Looking forward to the next release with OpenBSD/sparc running inside
>  qemu ;-)

Well, don't hold your breath but I'm already making some progress.

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 618 bytes --]

On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
> Blue Swirl wrote:
>
> > Here are two patches, loosely based on patches in OpenBSD ports system
> > that enable compiling Qemu. The first patch makes AIO use conditional
> > on host AIO support. The second one adjusts various include files etc.
> >
> > I'd like to commit both of these, is that OK?
> >
> >
>
>  The first patch is fine but the second patch has some bogus things in it.

Thanks for the review. I removed the unneeded parts, made a separate
patch for fixing OSS selection, and added a configure option for AIO
as requested by Samuel Thibault.

[-- Attachment #2: config_aio.diff --]
[-- Type: plain/text, Size: 3553 bytes --]

[-- Attachment #3: fix_oss.diff --]
[-- Type: plain/text, Size: 408 bytes --]

[-- Attachment #4: fix_openbsd_build.diff --]
[-- Type: plain/text, Size: 6458 bytes --]

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Anthony Liguori]

Blue Swirl wrote:
> On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
>   
>>
>>  The first patch is fine but the second patch has some bogus things in it.
>>     
>
> Thanks for the review. I removed the unneeded parts, made a separate
> patch for fixing OSS selection, and added a configure option for AIO
> as requested by Samuel Thibault.
>   

They look good to me.

Regards,

Anthony Liguori

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[malc]

On Wed, 13 Aug 2008, Blue Swirl wrote:

> On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
>> Blue Swirl wrote:
>>
>>> Here are two patches, loosely based on patches in OpenBSD ports system
>>> that enable compiling Qemu. The first patch makes AIO use conditional
>>> on host AIO support. The second one adjusts various include files etc.
>>>
>>> I'd like to commit both of these, is that OK?
>>>
>>>
>>
>>  The first patch is fine but the second patch has some bogus things in it.
>
> Thanks for the review. I removed the unneeded parts, made a separate
> patch for fixing OSS selection, and added a configure option for AIO
> as requested by Samuel Thibault.

Thanks for the OSS bits, i have committed slightly modified patch that
takes all possible drivers into consideration.

-- 
mailto:av1474@comtv.ru

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

On 8/14/08, malc <av1474@comtv.ru> wrote:
> On Wed, 13 Aug 2008, Blue Swirl wrote:
>
>
> > On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
> >
> > > Blue Swirl wrote:
> > >
> > >
> > > > Here are two patches, loosely based on patches in OpenBSD ports system
> > > > that enable compiling Qemu. The first patch makes AIO use conditional
> > > > on host AIO support. The second one adjusts various include files etc.
> > > >
> > > > I'd like to commit both of these, is that OK?
> > > >
> > > >
> > > >
> > >
> > >  The first patch is fine but the second patch has some bogus things in
> it.
> > >
> >
> > Thanks for the review. I removed the unneeded parts, made a separate
> > patch for fixing OSS selection, and added a configure option for AIO
> > as requested by Samuel Thibault.
> >
>
>  Thanks for the OSS bits, i have committed slightly modified patch that
>  takes all possible drivers into consideration.

Great, but there is a bug in your commit, *) case seems to be checked
before oss|sdl... case gets into play so the error is still present:

Error: Unknown driver 'oss' selected
Possible drivers are: oss sdl esd

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Todd T. Fries]

Actually, the problem is that some driver lists (NetBSD in particular)
were changed to
	list="<a> <b> <c>"
while others (OpenBSD in particular) were left with
	list="a b c"

fun!
-- 
Todd Fries .. todd@fries.net

 _____________________________________________
|                                             \  1.636.410.0632 (voice)
| Free Daemon Consulting, LLC                 \  1.405.227.9094 (voice)
| http://FreeDaemonConsulting.com             \  1.866.792.3418 (FAX)
| "..in support of free software solutions."  \          250797 (FWD)
|                                             \
 \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
                                                 
              37E7 D3EB 74D0 8D66 A68D  B866 0326 204E 3F42 004A
                        http://todd.fries.net/pgp.txt

Penned by Blue Swirl on 20080815 23:05.40, we have:
| On 8/14/08, malc <av1474@comtv.ru> wrote:
| > On Wed, 13 Aug 2008, Blue Swirl wrote:
| >
| >
| > > On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
| > >
| > > > Blue Swirl wrote:
| > > >
| > > >
| > > > > Here are two patches, loosely based on patches in OpenBSD ports system
| > > > > that enable compiling Qemu. The first patch makes AIO use conditional
| > > > > on host AIO support. The second one adjusts various include files etc.
| > > > >
| > > > > I'd like to commit both of these, is that OK?
| > > > >
| > > > >
| > > > >
| > > >
| > > >  The first patch is fine but the second patch has some bogus things in
| > it.
| > > >
| > >
| > > Thanks for the review. I removed the unneeded parts, made a separate
| > > patch for fixing OSS selection, and added a configure option for AIO
| > > as requested by Samuel Thibault.
| > >
| >
| >  Thanks for the OSS bits, i have committed slightly modified patch that
| >  takes all possible drivers into consideration.
| 
| Great, but there is a bug in your commit, *) case seems to be checked
| before oss|sdl... case gets into play so the error is still present:
| 
| Error: Unknown driver 'oss' selected
| Possible drivers are: oss sdl esd
| 

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[malc]

On Fri, 15 Aug 2008, Blue Swirl wrote:

> On 8/14/08, malc <av1474@comtv.ru> wrote:
>> On Wed, 13 Aug 2008, Blue Swirl wrote:
>>
>>
>>> On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:

[..snip..]

>>
>>  Thanks for the OSS bits, i have committed slightly modified patch that
>>  takes all possible drivers into consideration.
>
> Great, but there is a bug in your commit, *) case seems to be checked
> before oss|sdl... case gets into play so the error is still present:

I was ready to apologize for not testing, but then rechecked my scripts
and i in fact did test it properly and it worked (and still does), that
said you are ofcourse right here. Must be some sort of shell difference,
which one did you use?

> Error: Unknown driver 'oss' selected
> Possible drivers are: oss sdl esd
>
>

-- 
mailto:av1474@comtv.ru

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

On 8/16/08, malc <av1474@comtv.ru> wrote:
> On Fri, 15 Aug 2008, Blue Swirl wrote:
>
>
> > On 8/14/08, malc <av1474@comtv.ru> wrote:
> >
> > > On Wed, 13 Aug 2008, Blue Swirl wrote:
> > >
> > >
> > >
> > > > On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
> > > >
> > >
> >
>
>  [..snip..]
>
>
> >
> > >
> > >  Thanks for the OSS bits, i have committed slightly modified patch that
> > >  takes all possible drivers into consideration.
> > >
> >
> > Great, but there is a bug in your commit, *) case seems to be checked
> > before oss|sdl... case gets into play so the error is still present:
> >
>
>  I was ready to apologize for not testing, but then rechecked my scripts
>  and i in fact did test it properly and it worked (and still does), that
>  said you are ofcourse right here. Must be some sort of shell difference,
>  which one did you use?

/bin/sh is hard linked to ksh (Korn shell).

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[malc]

On Sat, 16 Aug 2008, Blue Swirl wrote:

> On 8/16/08, malc <av1474@comtv.ru> wrote:
>> On Fri, 15 Aug 2008, Blue Swirl wrote:
>>
>>
>>> On 8/14/08, malc <av1474@comtv.ru> wrote:
>>>
>>>> On Wed, 13 Aug 2008, Blue Swirl wrote:
>>>>
>>>>
>>>>
>>>>> On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
>>>>>
>>>>
>>>
>>
>>  [..snip..]
>>
>>
>>>
>>>>
>>>>  Thanks for the OSS bits, i have committed slightly modified patch that
>>>>  takes all possible drivers into consideration.
>>>>
>>>
>>> Great, but there is a bug in your commit, *) case seems to be checked
>>> before oss|sdl... case gets into play so the error is still present:
>>>
>>
>>  I was ready to apologize for not testing, but then rechecked my scripts
>>  and i in fact did test it properly and it worked (and still does), that
>>  said you are ofcourse right here. Must be some sort of shell difference,
>>  which one did you use?
>
> /bin/sh is hard linked to ksh (Korn shell).

I see. If my reading of [1] is correct, ksh is right and `GNU bash,
version 3.1.17(2)-release' (which is what i have) is wrong.

[1] http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html

-- 
mailto:av1474@comtv.ru

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 995 bytes --]

On 8/14/08, malc <av1474@comtv.ru> wrote:
> On Wed, 13 Aug 2008, Blue Swirl wrote:
>
>
> > On 8/12/08, Anthony Liguori <anthony@codemonkey.ws> wrote:
> >
> > > Blue Swirl wrote:
> > >
> > >
> > > > Here are two patches, loosely based on patches in OpenBSD ports system
> > > > that enable compiling Qemu. The first patch makes AIO use conditional
> > > > on host AIO support. The second one adjusts various include files etc.
> > > >
> > > > I'd like to commit both of these, is that OK?
> > > >
> > > >
> > > >
> > >
> > >  The first patch is fine but the second patch has some bogus things in
> it.
> > >
> >
> > Thanks for the review. I removed the unneeded parts, made a separate
> > patch for fixing OSS selection, and added a configure option for AIO
> > as requested by Samuel Thibault.
> >
>
>  Thanks for the OSS bits, i have committed slightly modified patch that
>  takes all possible drivers into consideration.

This patch fixes OSS selection and adds the needed libraries to LIBS.

[-- Attachment #2: fix_oss.diff --]
[-- Type: plain/text, Size: 2331 bytes --]

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Samuel Thibault]

Blue Swirl, le Tue 12 Aug 2008 21:45:51 +0300, a écrit :
> The first patch makes AIO use conditional on host AIO support.

Cool, we actually had something very similar in the xen tree.  Could you
add a configure option as well?

Samuel

Re: [Qemu-devel] [RFC, PATCH] Host AIO support detection, OpenBSD host support

[Samuel Thibault]

For BSD, you may need the patch below as well (from the xen repo)

changeset:   15970:04cc0e22a20a 
user:        Keir Fraser <keir@xensource.com> 
date:        Thu Sep 27 16:36:23 2007 +0100 
files:       tools/ioemu/audio/audio.c tools/ioemu/audio/mixeng.c tools/ioemu/audio/ossaudio.c tools/ioemu/block-raw.c tools/ioemu/block-vvfat.c tools/ioemu/bswap.h tools/ioemu/cutils.c tools/ioemu/monitor.c tools/ioemu/osdep.h tools/ioemu/target-i386-dm/exec-dm.c tools/ioemu/usb-linux.c tools/ioemu/vl.c tools/ioemu/vl.h tools/ioemu/vnc.c 
description: 
ioemu: Fixes for BSD. 
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com> 

Samuel


I don't know why this one was done, it looks a bit odd.

diff --git a/audio/mixeng.c b/audio/mixeng.c
index b668c52..f4b8003 100644
--- a/audio/mixeng.c
+++ b/audio/mixeng.c
@@ -146,6 +146,7 @@
 #undef SHIFT
 
 t_sample *mixeng_conv[2][2][2][3] = {
+#ifndef _BSD
     {
         {
             {
@@ -198,9 +199,11 @@ t_sample *mixeng_conv[2][2][2][3] = {
             }
         }
     }
+#endif /* !_BSD */
 };
 
 f_sample *mixeng_clip[2][2][2][3] = {
+#ifndef _BSD
     {
         {
             {
@@ -253,6 +256,7 @@ f_sample *mixeng_clip[2][2][2][3] = {
             }
         }
     }
+#endif /* !_BSD */
 };
 
 /*



Possibly related to the hunk above, something like improper endian
detection?

diff --git a/bswap.h b/bswap.h
index 523d805..fd589da 100644
--- a/bswap.h
+++ b/bswap.h
@@ -5,6 +5,11 @@
 
 #include <inttypes.h>
 
+#ifdef _BSD
+#include <sys/endian.h>
+#include <sys/types.h>
+#else
+
 #ifdef HAVE_BYTESWAP_H
 #include <byteswap.h>
 #else
@@ -73,6 +78,8 @@ static inline void bswap64s(uint64_t *s)
     *s = bswap64(*s);
 }
 
+#endif /* _BSD */
+
 #if defined(WORDS_BIGENDIAN)
 #define be_bswap(v, size) (v)
 #define le_bswap(v, size) bswap ## size(v)