ioemu build failure with GCC > 4.3

ioemu build failure with GCC > 4.3

[He, Qing]

	I got build failure in the qemu-xen-unstable.git tree, and it
seems only to happen with newer GCC versions, namely >4.3.0 which
supports C99 inline.
	The offending part is in tools/ioemu-remote/slirp/misc.c:
138: inline void
139: insque (a, b)

C99 treats inline differently with GNU89, bare `inline' (without linkage
specifier, i.e. `static') in C99 is required not to generate standalone
copy of the function, which is not desired in slirp code. Other
compilation module looks for the function in linking and will not find
them.

Since `-std=gnu99' is specified in the global Config.mk, newer GCC
version uses C99 inline semantics will encounter problems on building
ioemu.

It's most straightforward to remove `inline' in slirp/misc.c, the code
doesn't really seem to benefit from it. But it looks pretty old and not
maintained for sometime. What about patch it locally?

Other options include specify `-fgnu89-inline' (n/a prior to GCC 4.1.3)
and overrides slirp compile line with `-std=gnu89', but may not be
preferred.

Thanks,
Qing

Re: ioemu build failure with GCC > 4.3

[Keir Fraser]

On 10/7/08 10:32, "He, Qing" <qing.he@intel.com> wrote:

> It's most straightforward to remove `inline' in slirp/misc.c, the code
> doesn't really seem to benefit from it. But it looks pretty old and not
> maintained for sometime. What about patch it locally?
> 
> Other options include specify `-fgnu89-inline' (n/a prior to GCC 4.1.3)
> and overrides slirp compile line with `-std=gnu89', but may not be
> preferred.

Could we change the '#define inline' in slirp_config.h to define it to
nothing? None of the usages of 'inline' in the slirp directory look like a
particularly good idea.

 -- Keir

Re: ioemu build failure with GCC > 4.3

[Daniel P. Berrange]

On Thu, Jul 10, 2008 at 10:39:31AM +0100, Keir Fraser wrote:
> On 10/7/08 10:32, "He, Qing" <qing.he@intel.com> wrote:
> 
> > It's most straightforward to remove `inline' in slirp/misc.c, the code
> > doesn't really seem to benefit from it. But it looks pretty old and not
> > maintained for sometime. What about patch it locally?
> > 
> > Other options include specify `-fgnu89-inline' (n/a prior to GCC 4.1.3)
> > and overrides slirp compile line with `-std=gnu89', but may not be
> > preferred.
> 
> Could we change the '#define inline' in slirp_config.h to define it to
> nothing? None of the usages of 'inline' in the slirp directory look like a
> particularly good idea.

None of the entire slirp directory looks like a good idea. This code
isn't even used in Xen is it ? My vote  would be disabling compilation
of entire slirp directory for Xen. 

Daniel
-- 
|: Red Hat, Engineering, London   -o-   http://people.redhat.com/berrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505  -o-  F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

Re: ioemu build failure with GCC > 4.3

[Ian Jackson]

Daniel P. Berrange writes ("Re: [Xen-devel] ioemu build failure with GCC > 4.3"):
> None of the entire slirp directory looks like a good idea. This code
> isn't even used in Xen is it ? My vote  would be disabling compilation
> of entire slirp directory for Xen. 

This is the right answer, yes.

Ian.

Re: ioemu build failure with GCC > 4.3

[Ian Jackson]

iwj writes ("Re: [Xen-devel] ioemu build failure with GCC > 4.3"):
> Daniel P. Berrange writes ("Re: [Xen-devel] ioemu build failure with GCC > 4.3"):
> > None of the entire slirp directory looks like a good idea. This code
> > isn't even used in Xen is it ? My vote  would be disabling compilation
> > of entire slirp directory for Xen. 
> 
> This is the right answer, yes.

I've done this now.

Ian.

commit 29c20f126f298cf2162169b4bbb8206860f11240
Author: Ian Jackson <iwj@mariner.uk.xensource.com>
Date:   Thu Jul 10 10:48:51 2008 +0100

    Disable slirp.
    
    The slirp system is not really a very nice idea, especially not in a
    Xen context.  Also, the code misuses inline and does not compile on
    GCC >4.3ish.
    
    Thanks to Qing He <qing.he@intel.com> for the report.

diff --git a/xen-setup b/xen-setup
index c00122c..755b9ec 100755
--- a/xen-setup
+++ b/xen-setup
@@ -9,7 +9,7 @@ rm -f config-host.mak
 
 if test -f config-host.h; then mv config-host.h config-host.h~; fi
 
-./configure --disable-curses "$@" --prefix=/usr
+./configure --disable-curses --disable-slirp "$@" --prefix=/usr
 
 target=i386-dm

RE: ioemu build failure with GCC > 4.3

[Ross Philipson]

I think the patch also needs "--disable-gcc-check"

Thanks
Ross

-----Original Message-----
From: xen-devel-bounces@lists.xensource.com
[mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Ian Jackson
Sent: Thursday, July 10, 2008 5:50 AM
To: Daniel P. Berrange; Keir Fraser; He, Qing;
xen-devel@lists.xensource.com
Subject: Re: [Xen-devel] ioemu build failure with GCC > 4.3

iwj writes ("Re: [Xen-devel] ioemu build failure with GCC > 4.3"):
> Daniel P. Berrange writes ("Re: [Xen-devel] ioemu build failure with
GCC > 4.3"):
> > None of the entire slirp directory looks like a good idea. This code
> > isn't even used in Xen is it ? My vote  would be disabling
compilation
> > of entire slirp directory for Xen. 
> 
> This is the right answer, yes.

I've done this now.

Ian.

commit 29c20f126f298cf2162169b4bbb8206860f11240
Author: Ian Jackson <iwj@mariner.uk.xensource.com>
Date:   Thu Jul 10 10:48:51 2008 +0100

    Disable slirp.
    
    The slirp system is not really a very nice idea, especially not in a
    Xen context.  Also, the code misuses inline and does not compile on
    GCC >4.3ish.
    
    Thanks to Qing He <qing.he@intel.com> for the report.

diff --git a/xen-setup b/xen-setup
index c00122c..755b9ec 100755
--- a/xen-setup
+++ b/xen-setup
@@ -9,7 +9,7 @@ rm -f config-host.mak
 
 if test -f config-host.h; then mv config-host.h config-host.h~; fi
 
-./configure --disable-curses "$@" --prefix=/usr
+./configure --disable-curses --disable-slirp "$@" --prefix=/usr
 
 target=i386-dm
 

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

RE: ioemu build failure with GCC > 4.3

[Ian Jackson]

Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC > 4.3"):
> I think the patch also needs "--disable-gcc-check"

I don't have such a late gcc4 here to hand so you may well be right.
However --disable-gcc-check seems to be broken right now:

 ./configure: line 1244: test: too many arguments
 ./configure: line 1244: test: too many arguments

I'll look into this.

Ian.

RE: ioemu build failure with GCC > 4.3

[Ross Philipson]

Not sure why you get an error. I have gcc 4.3 and I added it in
ioemu-remote/xen-setup as in:

./configure --disable-gcc-check --disable-slirp --disable-ncurses blah
blah

Xen tools built to completion after I added it.

Thanks
Ross

-----Original Message-----
From: Ian Jackson 
Sent: Thursday, July 10, 2008 10:07 AM
To: Ross Philipson
Cc: Daniel P. Berrange; Keir Fraser; He, Qing;
xen-devel@lists.xensource.com
Subject: RE: [Xen-devel] ioemu build failure with GCC > 4.3

Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC >
4.3"):
> I think the patch also needs "--disable-gcc-check"

I don't have such a late gcc4 here to hand so you may well be right.
However --disable-gcc-check seems to be broken right now:

 ./configure: line 1244: test: too many arguments
 ./configure: line 1244: test: too many arguments

I'll look into this.

Ian.

RE: ioemu build failure with GCC > 4.3

[Ian Jackson]

Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC > 4.3"):
> Not sure why you get an error. I have gcc 4.3 and I added it in
> ioemu-remote/xen-setup as in:
> 
> ./configure --disable-gcc-check --disable-slirp --disable-ncurses blah
> blah
> 
> Xen tools built to completion after I added it.

Can you produce a log of your build ?  I saw what you see below.  It
did exit with status 0 but it appears always to print those error
messages.

This looks like a qemu upstream bug in configure to me.

Ian.

mariner:qemu-iwj.git> ./xen-setup
Install prefix    /usr
BIOS directory    /usr/share/qemu
binary directory  /usr/bin
Manual directory  /usr/share/man
ELF interp prefix /usr/gnemul/qemu-%M
Source path       /u/iwj/work/qemu-iwj.git
C compiler        gcc
Host C compiler   gcc
ARCH_CFLAGS       -m32
make              make
install           install
host CPU          i386
host big endian   no
target list       i386-softmmu x86_64-softmmu arm-softmmu cris-softmmu m68k-softmmu mips-softmmu mipsel-softmmu mips64-softmmu mips64el-softmmu ppc-softmmu ppcemb-softmmu ppc64-softmmu sh4-softmmu sh4eb-softmmu sparc-softmmu i386-linux-user x86_64-linux-user alpha-linux-user arm-linux-user armeb-linux-user cris-linux-user m68k-linux-user mips-linux-user mipsel-linux-user ppc-linux-user ppc64-linux-user ppc64abi32-linux-user sh4-linux-user sh4eb-linux-user sparc-linux-user sparc64-linux-user sparc32plus-linux-user 
gprof enabled     no
profiler          no
static build      no
-Werror enabled   no
SDL support       yes
SDL static link   yes
OpenGL support    yes
curses support    no
mingw32 support   no
Adlib support     no
AC97 support      no
GUS support       no
CoreAudio support no
ALSA support      no
EsounD support    no
DSound support    no
FMOD support      no 
OSS support       yes
VNC TLS support   yes
    TLS CFLAGS     
    TLS LIBS      -lgnutls  
kqemu support     yes
brlapi support    no
Documentation     yes
./configure: line 1244: test: too many arguments
./configure: line 1244: test: too many arguments
emariner:qemu-iwj.git> echo $?
0
mariner:qemu-iwj.git> git-diff | cat
diff --git a/xen-setup b/xen-setup
index 755b9ec..67f038d 100755
--- a/xen-setup
+++ b/xen-setup
@@ -9,7 +9,7 @@ rm -f config-host.mak
 
 if test -f config-host.h; then mv config-host.h config-host.h~; fi
 
-./configure --disable-curses --disable-slirp "$@" --prefix=/usr
+./configure --disable-curses --disable-slirp --disable-gcc-check "$@" --prefix=/usr
 
 target=i386-dm
 
mariner:qemu-iwj.git>

Re: ioemu build failure with GCC > 4.3

[Stefan de Konink]

Ian Jackson schreef:
> Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC > 4.3"):
>> Not sure why you get an error. I have gcc 4.3 and I added it in
>> ioemu-remote/xen-setup as in:
>>
>> ./configure --disable-gcc-check --disable-slirp --disable-ncurses blah
>> blah
>>
>> Xen tools built to completion after I added it.

Could you also tell us/me if you were able to boot a kernel compiled 
with GCC > 4.3 and Xen?


Stefan

Re: ioemu build failure with GCC > 4.3

[prashanth.mundkur]

On Mon, Jul 14, 2008 at 04:12:29PM +0100, Ian Jackson wrote:
> Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC > 4.3"):
> > Not sure why you get an error. I have gcc 4.3 and I added it in
> > ioemu-remote/xen-setup as in:
> > 
> > ./configure --disable-gcc-check --disable-slirp --disable-ncurses blah
> > blah
> > 
> > Xen tools built to completion after I added it.
> 
> Can you produce a log of your build ?  I saw what you see below.  It
> did exit with status 0 but it appears always to print those error
> messages.
> 
> This looks like a qemu upstream bug in configure to me.

I think bug is triggered if gcc3 is not installed on the system. Is it
possible that Ross doesn't see it because he has both gcc3 and gcc4
installed?  Here's how I fixed my build:

diff --git a/configure b/configure
index 2366aa6..0ef2540 100755
--- a/configure
+++ b/configure
@@ -1241,11 +1241,16 @@ case "$target_cpu" in
     fi
     echo "#define CONFIG_DYNGEN_OP 1" >> $config_h
     gcc3minver=`$cc --version 2> /dev/null| fgrep "(GCC) 3." | awk '{ print $3 }' | cut -f2 -d.`
-    if test -n "$gcc3minver" -a $gcc3minver -gt 3
+    if test -n "$gcc3minver"
     then
-      echo "HAVE_GT_GCC_3_3=true" >> $config_mak
+      if test -a $gcc3minver -gt 3;
+      then
+        echo "HAVE_GT_GCC_3_3=true" >> $config_mak
+      else
+        echo "HAVE_GT_GCC_3_3=false" >> $config_mak
+      fi
     else
-      echo "HAVE_GT_GCC_3_3=false" >> $config_mak
+      echo "HAVE_GT_GCC_3_3=true" >> $config_mak
     fi
   ;;
   x86_64)
diff --git a/xen-setup b/xen-setup
index 755b9ec..b0eda89 100755
--- a/xen-setup
+++ b/xen-setup
@@ -9,7 +9,7 @@ rm -f config-host.mak
 
 if test -f config-host.h; then mv config-host.h config-host.h~; fi
 
-./configure --disable-curses --disable-slirp "$@" --prefix=/usr
+./configure --disable-gcc-check --disable-curses --disable-slirp "$@" --prefix=/usr
 
 target=i386-dm
 

--prashanth

RE: ioemu build failure with GCC > 4.3

[Ross Philipson]

[-- Attachment #1: Type: text/plain, Size: 1475 bytes --]

This is in answer to both this question and Ian's earlier request for
the build output on my system. Attached is the build output both with
and without the --disable-gcc-check being used. I see the "too many
arguments" message but earlier in the build output. The build seems to
run to completion. I have gcc 4.3 but no gcc 3.x.

The answer to the question in this email is, no I cannot boot. Xen
crashes early on with an assertion. The serial output from Xen is in the
attached file also.

I am unfortunately engaged in some other priority work and I was just
trying to get the latest Xen going on another machine as an aside task.
I have not spent a great deal of time looking into it.

Thanks
Ross


-----Original Message-----
From: Stefan de Konink [mailto:skinkie@xs4all.nl] 
Sent: Monday, July 14, 2008 11:21 AM
To: Ian Jackson
Cc: Ross Philipson; xen-devel@lists.xensource.com; Daniel P. Berrange;
Keir Fraser; He, Qing
Subject: Re: [Xen-devel] ioemu build failure with GCC > 4.3

Ian Jackson schreef:
> Ross Philipson writes ("RE: [Xen-devel] ioemu build failure with GCC >
4.3"):
>> Not sure why you get an error. I have gcc 4.3 and I added it in
>> ioemu-remote/xen-setup as in:
>>
>> ./configure --disable-gcc-check --disable-slirp --disable-ncurses
blah
>> blah
>>
>> Xen tools built to completion after I added it.

Could you also tell us/me if you were able to boot a kernel compiled 
with GCC > 4.3 and Xen?


Stefan

[-- Attachment #2: tools-build-xen-crash-7-15.tar.gz --]
[-- Type: application/x-gzip, Size: 29871 bytes --]

[-- Attachment #3: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Re: ioemu build failure with GCC > 4.3

[Keir Fraser]

On 15/7/08 14:55, "Ross Philipson" <Ross.Philipson@citrix.com> wrote:

> The answer to the question in this email is, no I cannot boot. Xen
> crashes early on with an assertion. The serial output from Xen is in the
> attached file also.
> 
> I am unfortunately engaged in some other priority work and I was just
> trying to get the latest Xen going on another machine as an aside task.
> I have not spent a great deal of time looking into it.

The crash is during VT-d initialisation. Another argument for disabling VT-d
by default in Xen 3.3.

 -- Keir

RE: ioemu build failure with GCC > 4.3

[Ross Philipson]

Roger that, when I disable VT-d in the BIOS, I can boot up fine.

Ross

-----Original Message-----
From: Keir Fraser 
Sent: Tuesday, July 15, 2008 10:06 AM
To: Ross Philipson; Stefan de Konink; Ian Jackson
Cc: xen-devel@lists.xensource.com; Daniel P. Berrange; He, Qing
Subject: Re: [Xen-devel] ioemu build failure with GCC > 4.3

On 15/7/08 14:55, "Ross Philipson" <Ross.Philipson@citrix.com> wrote:

> The answer to the question in this email is, no I cannot boot. Xen
> crashes early on with an assertion. The serial output from Xen is in
the
> attached file also.
> 
> I am unfortunately engaged in some other priority work and I was just
> trying to get the latest Xen going on another machine as an aside
task.
> I have not spent a great deal of time looking into it.

The crash is during VT-d initialisation. Another argument for disabling
VT-d
by default in Xen 3.3.

 -- Keir

[PATCH] RE: ioemu build failure with GCC > 4.3

[Espen Skoglund]

[Ross Philipson]
> Roger that, when I disable VT-d in the BIOS, I can boot up fine.
> Ross

> -----Original Message-----
> From: Keir Fraser 
> Sent: Tuesday, July 15, 2008 10:06 AM
> To: Ross Philipson; Stefan de Konink; Ian Jackson
> Cc: xen-devel@lists.xensource.com; Daniel P. Berrange; He, Qing
> Subject: Re: [Xen-devel] ioemu build failure with GCC > 4.3

> On 15/7/08 14:55, "Ross Philipson" <Ross.Philipson@citrix.com> wrote:

>> The answer to the question in this email is, no I cannot boot. Xen
>> crashes early on with an assertion. The serial output from Xen is
>> in the attached file also.
>> 
>> I am unfortunately engaged in some other priority work and I was
>> just trying to get the latest Xen going on another machine as an
>> aside task.  I have not spent a great deal of time looking into it.

> The crash is during VT-d initialisation. Another argument for
> disabling VT-d by default in Xen 3.3.

Actually, one should never try to dump the VT-d tables from the DMA
fault interrupt handler since, well, it's in an interrupt handler and
this assertion will always be triggered (on x86_32).

	eSk


---
vt-d: Disable VT-d table output on DMA page faults

Printing VT-d tables requires map_domain_page() which can not be
called from interrupt context (on x86_32).

Signed-off-by: Espen Skoglund <espen.skoglund@netronome.com>

diff -r 1f88a5d4ab95 xen/drivers/passthrough/vtd/iommu.c
--- a/xen/drivers/passthrough/vtd/iommu.c	Tue Jul 15 18:27:59 2008 +0100
+++ b/xen/drivers/passthrough/vtd/iommu.c	Tue Jul 15 19:22:23 2008 +0100
@@ -719,9 +719,11 @@ static int iommu_page_fault_do_one(struc
             PCI_SLOT(source_id & 0xFF), PCI_FUNC(source_id & 0xFF), addr,
             fault_reason, iommu->reg);
 
+#if 0
     if ( fault_reason < 0x20 )
         print_vtd_entries(iommu, (source_id >> 8),
                           (source_id & 0xff), (addr >> PAGE_SHIFT));
+#endif
 
     return 0;
 }

RE: ioemu build failure with GCC > 4.3

[Espen Skoglund]

[Ross Philipson]
> Roger that, when I disable VT-d in the BIOS, I can boot up fine.

According to your logs it's an "interesting" set of ACPI tables you've
got there.  From what I can tell it looks like a USB controller or
something is performing bus transactions to a bogus address in host
memory.  Needless to say, this should (in an ideal world) not be
happening.

To determine if there is a serious problem it would be helpful if you
could try a bootup with the following patch applied.  An `lspci -t`
and `lspci` would also be appreciated.

	eSk


---
vt-d: Some debug output for ACPI DMAR parsing

Signed-off-by: Espen Skoglund <espen.skoglund@netronome.com>

diff -r a9c595d1909c xen/drivers/passthrough/vtd/dmar.c
--- a/xen/drivers/passthrough/vtd/dmar.c	Tue Jul 15 19:22:23 2008 +0100
+++ b/xen/drivers/passthrough/vtd/dmar.c	Tue Jul 15 19:28:54 2008 +0100
@@ -337,6 +337,8 @@ acpi_parse_one_rmrr(struct acpi_dmar_ent
     dev_scope_end   = ((void *)rmrr) + header->length;
     ret = acpi_parse_dev_scope(dev_scope_start, dev_scope_end,
                                rmrru, RMRR_TYPE);
+    dprintk(XENLOG_INFO VTDPREFIX, "RMRR region <%"PRIx64",%"PRIx64">\n",
+            rmrru->base_address, rmrru->end_address);
 
     if ( ret || (rmrru->scope.devices_cnt == 0) )
         xfree(rmrru);