Hi, I am looking for an example dtrace script that is tracing and dereferencing userland data structures. I can not find any examples, and am having a heck of a time getting mine to work as I can not even get my dtrace script to resolve my userland data structures. Peter
Hi Peter,
Here you go:
Here is a C program called sample.c
struct xxx
{
int yyy;
int zzz;
};
main()
{
struct xxx t1 = {20, 30};
sub1 (&t1);
}
sub1 (struct xxx *p)
{
printf ("CProgram: %d %d\n", p->yyy, p->zzz);
}
And here is a D program called sample.d
struct xxx
{
int yyy;
int zzz;
};
pid$target:a.out:sub1:entry
{
sp = (struct xxx *) copyin (arg0, sizeof (struct xxx));
printf ("DTrace: %d %d\n", sp->yyy, sp->zzz);
exit (0);
}
I compiled the C program with:
/usr/sfw/bin/gcc sample.c
and then I ran the D program with:
dtrace -s sample.d -c ./a.out
which produced this output:
dtrace: script ''sample.d'' matched 1 probe
CProgram: 20 30
CPU ID FUNCTION:NAME
0 42282 sub1:entry DTrace: 20 30
Hope this helps! (Note: you probably know this, but just in case, the
reason the C output appears first is that the C program has printed its
output, and run home before the dtrace command has a chance to get the
trace date from the switch buffer.)
Chip
Peter Shoults wrote:> Hi,
>
> I am looking for an example dtrace script that is tracing and
> dereferencing userland data structures.
> I can not find any examples, and am having a heck of a time getting mine
> to work as I can not even
> get my dtrace script to resolve my userland data structures.
>
> Peter
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>
Yeah, I was REALLY hoping to avoid having to define the structures in dtrace. There are many structures I am going to be mucking with, and a few are really large. I know the hex offsets I am going to be working with, so I was hoping there was a way I can avoid having to type in all these structures. peter Chip Bennett wrote:> Hi Peter, > > Here you go: > > Here is a C program called sample.c > > struct xxx > { > int yyy; > int zzz; > }; > > main() > { > struct xxx t1 = {20, 30}; > sub1 (&t1); > } > > sub1 (struct xxx *p) > { > printf ("CProgram: %d %d\n", p->yyy, p->zzz); > } > > And here is a D program called sample.d > > struct xxx > { > int yyy; > int zzz; > }; > > pid$target:a.out:sub1:entry > { > sp = (struct xxx *) copyin (arg0, sizeof (struct xxx)); > printf ("DTrace: %d %d\n", sp->yyy, sp->zzz); > exit (0); > } > > I compiled the C program with: > > /usr/sfw/bin/gcc sample.c > > and then I ran the D program with: > > dtrace -s sample.d -c ./a.out > > which produced this output: > > dtrace: script ''sample.d'' matched 1 probe > CProgram: 20 30 > CPU ID FUNCTION:NAME > 0 42282 sub1:entry DTrace: 20 30 > > Hope this helps! (Note: you probably know this, but just in case, the > reason the C output appears first is that the C program has printed its > output, and run home before the dtrace command has a chance to get the > trace date from the switch buffer.) > > Chip > > > Peter Shoults wrote: > >> Hi, >> >> I am looking for an example dtrace script that is tracing and >> dereferencing userland data structures. >> I can not find any examples, and am having a heck of a time getting mine >> to work as I can not even >> get my dtrace script to resolve my userland data structures. >> >> Peter >> _______________________________________________ >> dtrace-discuss mailing list >> dtrace-discuss at opensolaris.org >> >> > > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss at opensolaris.org >
Well.... why don''t you put all of you structure defs in a header file, and then #include then in both your C programs and your D scripts. You can do that in D using the -C flag. The only things that D doesn''t support syntactically for definitions and declarations is anything that happens to be a D reserved word. (i.e. "string"). Chip Peter Shoults wrote:> Yeah, I was REALLY hoping to avoid having to define the structures in > dtrace. There are many structures I am going to be mucking with, and > a few are really large. I know the hex offsets I am going to be > working with, so I was hoping there was a way I can avoid having to > type in all these structures. > > peter > > > > Chip Bennett wrote: >> Hi Peter, >> >> Here you go: >> >> Here is a C program called sample.c >> >> struct xxx >> { >> int yyy; >> int zzz; >> }; >> >> main() >> { >> struct xxx t1 = {20, 30}; >> sub1 (&t1); >> } >> >> sub1 (struct xxx *p) >> { >> printf ("CProgram: %d %d\n", p->yyy, p->zzz); >> } >> >> And here is a D program called sample.d >> >> struct xxx >> { >> int yyy; >> int zzz; >> }; >> >> pid$target:a.out:sub1:entry >> { >> sp = (struct xxx *) copyin (arg0, sizeof (struct xxx)); >> printf ("DTrace: %d %d\n", sp->yyy, sp->zzz); >> exit (0); >> } >> >> I compiled the C program with: >> >> /usr/sfw/bin/gcc sample.c >> >> and then I ran the D program with: >> >> dtrace -s sample.d -c ./a.out >> >> which produced this output: >> >> dtrace: script ''sample.d'' matched 1 probe >> CProgram: 20 30 >> CPU ID FUNCTION:NAME >> 0 42282 sub1:entry DTrace: 20 30 >> >> Hope this helps! (Note: you probably know this, but just in case, >> the reason the C output appears first is that the C program has >> printed its output, and run home before the dtrace command has a >> chance to get the trace date from the switch buffer.) >> >> Chip >> >> >> Peter Shoults wrote: >> >>> Hi, >>> >>> I am looking for an example dtrace script that is tracing and >>> dereferencing userland data structures. >>> I can not find any examples, and am having a heck of a time getting >>> mine to work as I can not even >>> get my dtrace script to resolve my userland data structures. >>> >>> Peter >>> _______________________________________________ >>> dtrace-discuss mailing list >>> dtrace-discuss at opensolaris.org >>> >> >> _______________________________________________ >> dtrace-discuss mailing list >> dtrace-discuss at opensolaris.org >> >
Hi Peter,> I am looking for an example dtrace script that is tracing and > dereferencing userland data structures. I can not find any examples, > and am having a heck of a time getting mine to work as I can not even > get my dtrace script to resolve my userland data structures.I wrote up my attempts here: <http://blogs.sun.com/peteh/entry/dereferencing_user_space_pointers_in> It took a few attempts to get the complex hops across linked structures in user space working correctly and my resulting D script was not very portable. Jon Haslam has chapter and verse here: <http://blogs.sun.com/jonh/entry/accessing_a_programs_local_variables> -- Peter -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/x-pkcs7-signature Size: 2856 bytes Desc: S/MIME Cryptographic Signature URL: <http://mail.opensolaris.org/pipermail/dtrace-discuss/attachments/20071029/9cc48e0c/attachment.bin>
Adam Leventhal
2007-Oct-31 08:35 UTC
[dtrace-discuss] dtrace example of userland debugging
On Mon, Oct 29, 2007 at 02:23:06PM +0000, Peter Harvey wrote:> It took a few attempts to get the complex hops across linked structures > in user space working correctly and my resulting D script was not very > portable.Yes: we know this, frankly, sucks. We have some ideas for automatic copyins which would allow D scripts to manipulate user-land objects much more naturally. Unfortunately, the implementation is a ways off. Adam -- Adam Leventhal, FishWorks http://blogs.sun.com/ahl