dtrace discuss - Jan 2007 - invalid address (0x0) in action #2 at DIF offset 52

Hi,
  Iam facing this strange problem when i execute the following dtrace script on
on particular thread(tid ==1) where as it works perfectly with another
thread(tid == 4).Can someone help me with this problem?
pid$1::SGetMsg:return
/tid == 1 && self->trace == 1/
{
        printf("SGetMsg Ptr=0x%p\n", arg3);
        printf("Sget Reurn val %x Thread id %d 3
\n",*(int*)copyin(arg3,4),tid);
}

The argument which iam trying to print is pointer to a pointer.
 
 
This message posted from opensolaris.org

Madhusudhan,

	you are using arg3 in "return" probe. If the function has a return 
value, it is stored in argument one. If a function does not
have a return value, argument one is not defined;

what''s arg3?? if it''s fourth input argument of SGetMsg:entry
you must
save it in a thread local var to print it afterwards. In return probe 
you can''t be sure it contains that info.

Cheers,
Gonzalo.


Madhusudhan Reddy wrote:
> Hi,
>   Iam facing this strange problem when i execute the following dtrace
script on on particular thread(tid ==1) where as it works perfectly with another
thread(tid == 4).Can someone help me with this problem?
> pid$1::SGetMsg:return
> /tid == 1 && self->trace == 1/
> {
>         printf("SGetMsg Ptr=0x%p\n", arg3);
>         printf("Sget Reurn val %x Thread id %d 3
\n",*(int*)copyin(arg3,4),tid);
> }
> 
> The argument which iam trying to print is pointer to a pointer.
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org

Gonzalo,
If iam not wrong, input arguments of the function are arg0,arg1, arg2..... and
return arguments are arg1, arg2, arg3.....

3rd Input argument  of SGetMsg(which is pointer to a pointer) is of my interest
so iam trying to print this value using arg3 when the function returns. Anything
wrong ?

Below is the prototype of the function for your reference

PUBLIC S16 SGetMsg
(
REG1 Region region,        
REG2 Pool pool,           
Buffer **mBufPtr        
)

Thanks,
Madhusudhan Reddy
 
 
This message posted from opensolaris.org

Madhusudhan Reddy wrote:
> Gonzalo,
> If iam not wrong, input arguments of the function are arg0,arg1, arg2.....
and return arguments are arg1, arg2, arg3.....
A "C" function has only one (or zero) return arguments; supplied in a
"return" statement.
> 
> 3rd Input argument  of SGetMsg(which is pointer to a pointer) is of my
interest so iam trying to print this value using arg3 when the function returns.
Anything wrong ?
A "C" function can''t change it''s caller''s
view of an input argument either.  What you
might be asking for (I''m not clear, from your question) is the value
pointed to
by the third input argument, at the time of return from the function.

If so, you should do something like capturing that 3rd input argument on entry
to the function (in one D clause) and indirecting through the captured pointer
on exit from the function (in a separate D clause).

- Jeremy

Hi,

You need to grab the value of arg3 on the entry probe to the function 
(pid$1::SGetMsg:entry), save it in a thread-local variable, and then 
reference the thread-local variable in the return probe instead of the 
built-in variable arg3.  arg3 only contains a copy of your function 
argument for the duration of the entry probe clause.

Chip


Madhusudhan Reddy wrote:
> Gonzalo,
> If iam not wrong, input arguments of the function are arg0,arg1, arg2.....
and return arguments are arg1, arg2, arg3.....
>
> 3rd Input argument  of SGetMsg(which is pointer to a pointer) is of my
interest so iam trying to print this value using arg3 when the function returns.
Anything wrong ?
>
> Below is the prototype of the function for your reference
>
> PUBLIC S16 SGetMsg
> (
> REG1 Region region,        
> REG2 Pool pool,           
> Buffer **mBufPtr        
> )
>
> Thanks,
> Madhusudhan Reddy
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>

Thanks a lot chip.Now it works !

Madhusudhan Reddy
 
 
This message posted from opensolaris.org