dtrace discuss - Mar 2008 - Bad instruction on x86_64 build on OS X with DTRACE_PROBE

I was looking at mod_trace
(http://prefetch.net/projects/apache_modtrace/index.html
) and playing with getting it to compile on OS X.

When building for x86_64 with -arch x86_64 we get bad instructions  
generated:

gcc -o foo  -arch x86_64 foo.c
/var/folders/rV/rV1x2DafFr0R6tGG+1bbk++++TM/-Tmp-//ccnykQ1o.s:11:bad  
register name `%%esi)''


Using gcc -S I can definitely see we are not generating correct code:

LCFI1:
         __dtrace_probe$6_foo___probe__noargs:
         nop
         leal 0(%%esi), %%esi

Building with -arch i386 works, DTRACE_PROBE1, etc also seem to work  
on OS X based on the building mod_trace.

I attach a simple reproducer, changing the Makefile to build with - 
arch i386 works.  Simply extract onto a Leopard box, cd into the  
extracted dir and run make to see the failure. make foo.s will create  
the generated asm.

$ gcc -v
Using built-in specs.
Target: i686-apple-darwin9
Configured with: /var/tmp/gcc/gcc-5465~16/src/configure --disable- 
checking -enable-werror --prefix=/usr --mandir=/share/man --enable- 
languages=c,objc,c++,obj-c++ --program-transform-name=/^[cg][^.-]*$/s/ 
$/-4.0/ --with-gxx-include-dir=/include/c++/4.0.0 --with-slibdir=/usr/ 
lib --build=i686-apple-darwin9 --with-arch=apple --with-tune=generic -- 
host=i686-apple-darwin9 --target=i686-apple-darwin9
Thread model: posix
gcc version 4.0.1 (Apple Inc. build 5465)
$ dtrace -V
dtrace: Sun D 1.2.2

Paul

-------------- next part --------------
A non-text attachment was scrubbed...
Name: dtrace-osx-bug.tgz
Type: application/octet-stream
Size: 464 bytes
Desc: not available
URL:
<http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20080311/9d464630/attachment.obj>
-------------- next part --------------

On Mar 11, 2008, at 4:00 PM, Paul Nasrat wrote:
> I was looking at mod_trace
(http://prefetch.net/projects/apache_modtrace/index.html
> ) and playing with getting it to compile on OS X.
>
> When building for x86_64 with -arch x86_64 we get bad instructions  
> generated:
>
> gcc -o foo  -arch x86_64 foo.c
> /var/folders/rV/rV1x2DafFr0R6tGG+1bbk++++TM/-Tmp-//ccnykQ1o.s:11:bad  
> register name `%%esi)''
>
>
> Using gcc -S I can definitely see we are not generating correct code:
>
> LCFI1:
>        __dtrace_probe$6_foo___probe__noargs:
>        nop
>        leal 0(%%esi), %%esi
>
> Building with -arch i386 works, DTRACE_PROBE1, etc also seem to work  
> on OS X based on the building mod_trace.
>
> I attach a simple reproducer, changing the Makefile to build with - 
> arch i386 works.  Simply extract onto a Leopard box, cd into the  
> extracted dir and run make to see the failure. make foo.s will  
> create the generated asm.
>
> $ gcc -v
> Using built-in specs.
> Target: i686-apple-darwin9
> Configured with: /var/tmp/gcc/gcc-5465~16/src/configure --disable- 
> checking -enable-werror --prefix=/usr --mandir=/share/man --enable- 
> languages=c,objc,c++,obj-c++ --program-transform-name=/^[cg][^.-]*$/ 
> s/$/-4.0/ --with-gxx-include-dir=/include/c++/4.0.0 --with-slibdir=/ 
> usr/lib --build=i686-apple-darwin9 --with-arch=apple --with- 
> tune=generic --host=i686-apple-darwin9 --target=i686-apple-darwin9
> Thread model: posix
> gcc version 4.0.1 (Apple Inc. build 5465)
> $ dtrace -V
> dtrace: Sun D 1.2.2
Paul,

	The DTRACE_PROBE# macros are not supported on OS X. That is why  
you''re running into issues.

	James M

On 11 Mar 2008, at 23:19, James McIlree wrote:
>
>
>> I was looking at mod_trace
(http://prefetch.net/projects/apache_modtrace/index.html
>> ) and playing with getting it to compile on OS X.
>>
>> When building for x86_64 with -arch x86_64 we get bad instructions  
>> generated:
>>
>> gcc -o foo  -arch x86_64 foo.c
>> /var/folders/rV/rV1x2DafFr0R6tGG+1bbk++++TM/-Tmp-//ccnykQ1o.s: 
>> 11:bad register name `%%esi)''
>>
>
> 	The DTRACE_PROBE# macros are not supported on OS X. That is why  
> you''re running into issues.
If they''re not supported shouldn''t I at least get a warning.
Any
pointers to the right way to do things?

I note that the root cause is actually a bad define in

/usr/include/mach/i386/sdt_isa.h:

#ifdef __x86_64__

#define DTRACE_NOPS                     \
         "nop"                   "\n\t"  \
         "leal 0(%%esi), %%esi"  "\n\t"



Paul

On Mar 12, 2008, at 12:27 AM, Paul Nasrat wrote:
> Any
> pointers to the right way to do things?
% man dtrace
...
"BUILDING CODE CONTAINING USDT PROBES
        The  process of adding USDT probes to code is slightly  
different than documented in the Solaris Dynamic
        Tracing Guide. The steps for adding probes are as follows:"
...

SCP
--
Steve Peters
scp at mac.com

On 12 Mar 2008, at 16:15, Steve Peters wrote:
>
> On Mar 12, 2008, at 12:27 AM, Paul Nasrat wrote:
>
>> Any
>> pointers to the right way to do things?
>
> % man dtrace
> ...
> "BUILDING CODE CONTAINING USDT PROBES
>       The  process of adding USDT probes to code is slightly  
> different than documented in the Solaris Dynamic
>       Tracing Guide. The steps for adding probes are as follows:"
Yeah thanks I''d already figured that out.

Is it worth updating
http://wikis.sun.com/display/DTrace/Statically+Defined+Tracing+for+User+Applications
  ?

Paul

On Wed, Mar 12, 2008 at 04:19:05PM +0000, Paul Nasrat
wrote:
> Yeah thanks I''d already figured that out.
> 
> Is it worth updating
http://wikis.sun.com/display/DTrace/Statically+Defined+Tracing+for+User+Applications
Yes. Here''s another example:

  http://blogs.sun.com/ahl/entry/user_land_tracing_gets_better

Adam

-- 
Adam Leventhal, Fishworks                        http://blogs.sun.com/ahl

On Mar 12, 2008, at 12:27 AM, Paul Nasrat wrote:
>
> On 11 Mar 2008, at 23:19, James McIlree wrote:
>>
>>
>>> I was looking at mod_trace
(http://prefetch.net/projects/apache_modtrace/index.html
>>> ) and playing with getting it to compile on OS X.
>>>
>>> When building for x86_64 with -arch x86_64 we get bad instructions
>>> generated:
>>>
>>> gcc -o foo  -arch x86_64 foo.c
>>> /var/folders/rV/rV1x2DafFr0R6tGG+1bbk++++TM/-Tmp-//ccnykQ1o.s: 
>>> 11:bad register name `%%esi)''
>>>
>>
>> 	The DTRACE_PROBE# macros are not supported on OS X. That is why  
>> you''re running into issues.
>
> If they''re not supported shouldn''t I at least get a
warning. Any
> pointers to the right way to do things?
>
> I note that the root cause is actually a bad define in
>
> /usr/include/mach/i386/sdt_isa.h:
>
> #ifdef __x86_64__
>
> #define DTRACE_NOPS                     \
>        "nop"                   "\n\t"  \
>        "leal 0(%%esi), %%esi"  "\n\t"
	I thought we had #ifdef''d these so they wouldn''t be available
without
some hackery,
but I''m not seeing the guards. Yes, you should have gotten a  
warning :-).

	I filed:

	<rdar://problem/5795255> Disallow use of DTRACE_PROBE macros

	James M