dtrace discuss - Nov 2005 - Multiple returns from a single function

Hello!

Why is it that a call to memcpy has two returns, one from memmove and 
one from memcpy?  Here''s the script and its output:

---------------------------------
The script:
---------------------------------
pid$1:::entry
{
}
pid$1:::return
{
}

----------------------------------
The output:
----------------------------------
...
   0  68204                     memcpy:entry
   0  71091                   memmove:return
   0  71092                    memcpy:return
...

On Thu, Nov 10, 2005 at 10:05:01PM +0100, Josip Gracin
wrote:
> Hello!
> 
> Why is it that a call to memcpy has two returns, one from memmove and 
> one from memcpy?  Here''s the script and its output:
Could you please provide the output of:

	psrinfo -vp
	mount | grep libc
	ldd /bin/lsh

so we can figure out which version of memcpy is being used?  (unfortunately,
there are a multitude of versions)

Thanks,
- jonathan
> ---------------------------------
> The script:
> ---------------------------------
> pid$1:::entry
> {
> }
> pid$1:::return
> {
> }
> 
> ----------------------------------
> The output:
> ----------------------------------
> ...
>   0  68204                     memcpy:entry
>   0  71091                   memmove:return
>   0  71092                    memcpy:return
> ...
> 
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
-- 
Jonathan Adams, Solaris Kernel Development

This is due to the implementation of memcpy() -- it''s rather
intertwined with
memmove():

http://cvs.opensolaris.org/source/xref/on/usr/src/lib/libc/i386_hwcap1/gen/memcpy.s#40

memcpy() is effectively embedded in memmove(). It''s a drag, and it
would be
nice if we could rework the code or the pid provider''s text analysis to
avoid this situation.

Adam

On Thu, Nov 10, 2005 at 10:05:01PM +0100, Josip Gracin wrote:

> Hello!
> 
> Why is it that a call to memcpy has two returns, one from memmove and 
> one from memcpy?  Here''s the script and its output:
> 
> ---------------------------------
> The script:
> ---------------------------------
> pid$1:::entry
> {
> }
> pid$1:::return
> {
> }
> 
> ----------------------------------
> The output:
> ----------------------------------
> ...
>   0  68204                     memcpy:entry
>   0  71091                   memmove:return
>   0  71092                    memcpy:return
> ...
> 
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
-- 
Adam Leventhal, Solaris Kernel Development       http://blogs.sun.com/ahl

On Thu, Nov 10, 2005 at 01:30:33PM -0800, Jonathan Adams
wrote:
> On Thu, Nov 10, 2005 at 10:05:01PM +0100, Josip Gracin wrote:
> > Hello!
> > 
> > Why is it that a call to memcpy has two returns, one from memmove and 
> > one from memcpy?  Here''s the script and its output:
> 
> Could you please provide the output of:
> 
> 	psrinfo -vp
> 	mount | grep libc
> 	ldd /bin/lsh
                 ^^^ should be "ls".  <sigh>

But Adam has already figured this out, and responded, so never mind. =]

- jonathan
> 
> so we can figure out which version of memcpy is being used? 
(unfortunately,
> there are a multitude of versions)
> 
> Thanks,
> - jonathan
> 
> > ---------------------------------
> > The script:
> > ---------------------------------
> > pid$1:::entry
> > {
> > }
> > pid$1:::return
> > {
> > }
> > 
> > ----------------------------------
> > The output:
> > ----------------------------------
> > ...
> >   0  68204                     memcpy:entry
> >   0  71091                   memmove:return
> >   0  71092                    memcpy:return
> > ...
> > 
> > _______________________________________________
> > dtrace-discuss mailing list
> > dtrace-discuss at opensolaris.org
> 
> -- 
> Jonathan Adams, Solaris Kernel Development
-- 
Jonathan Adams, Solaris Kernel Development

Adam Leventhal wrote:
> This is due to the implementation of memcpy() -- it''s rather
intertwined with
> memmove():
I see.  Thanks!