dtrace discuss - Jul 2005 - speculative tracing on nevada builds ?

Hi,

Has something related to speculative tracing changed between s10 FCS
and the more recent nevada builds ?

I was trying the specopen.d script from the Dtrace guide on a
nevada machine and it failed with :

dtrace: failed to enable ''./spec.d'': DIF program content is
invalid

To try and narrow things down a bit I wrote the following short
script. This works fine on s10, but fails with the above error on
nevada builds 14, 17 and 19 (the only builds I have access to).

I set dtrace_err_verbose=1 but didn''t get any more information.

#!/usr/sbin/dtrace -s

syscall::open:entry
{
        self->spec = speculation();
        speculate(self->spec);
        printf("%s", stringof(copyinstr(arg0)));
}

syscall::open:return
/self->spec && errno != 0/
{
        commit(self->spec);
        self->spec = 0;
}

syscall::open:return
/self->spec && errno == 0/
{
        discard(self->spec);
        self->spec = 0;
}


Anyone got any ideas whats going on ?

Cheers,

Steve

On Thu, Jul 28, 2005 at 12:05:19PM +0100, Stephen Guidon
wrote:
> 
> dtrace: failed to enable ''./spec.d'': DIF program content
is invalid
> 
> To try and narrow things down a bit I wrote the following short
> script. This works fine on s10, but fails with the above error on
> nevada builds 14, 17 and 19 (the only builds I have access to).
> 
> I set dtrace_err_verbose=1 but didn''t get any more information.
> 
> #!/usr/sbin/dtrace -s
> 
> syscall::open:entry
> {
>         self->spec = speculation();
>         speculate(self->spec);
>         printf("%s", stringof(copyinstr(arg0)));
> }
> 
> syscall::open:return
> /self->spec && errno != 0/
> {
>         commit(self->spec);
>         self->spec = 0;
> }
> 
> syscall::open:return
> /self->spec && errno == 0/
> {
>         discard(self->spec);
>         self->spec = 0;
> }
> 
> 
> Anyone got any ideas whats going on ?
This is definitely a bug;  I''ve filed:

6303188 some dtrace scripts with speculations fail to load

Unfortunately, I don''t see an easy workaround at the moment.  It looks
like
some of the validation logic has gotten out of step with itself.

Cheers,
- jonathan

> > #!/usr/sbin/dtrace -s
> > 
> > syscall::open:entry
> > {
> >         self->spec = speculation();
> >         speculate(self->spec);
> >         printf("%s", stringof(copyinstr(arg0)));
> > }
> > 
> > syscall::open:return
> > /self->spec && errno != 0/
> > {
> >         commit(self->spec);
> >         self->spec = 0;
> > }
> > 
> > syscall::open:return
> > /self->spec && errno == 0/
> > {
> >         discard(self->spec);
> >         self->spec = 0;
> > }
> > 
> > 
> > Anyone got any ideas whats going on ?
> 
> This is definitely a bug;  I''ve filed:
> 
> 6303188 some dtrace scripts with speculations fail to load
I actually have a fix for this in my current wad.  (I''m embarrassed to
say
that this was first run into over a month ago -- I''ve got a pretty bad
case
of wad creep at the moment.)

I introduced this with the fix for:

  6253030 adding an action to an ECB takes quadratic time

The fix is a one-liner, and I''m embarrassed that we didn''t
have a test
to catch this in our test suite (suffice it to say that I have added 
such a test).  The workaround would be to rephrase this as:

  syscall::open:entry
  {
          self->spec = speculation();
  }

  syscall::open:entry
  /self->spec/
  {
          speculate(self->spec);
          ...
  }

Apologies for the brokenness -- and for the delay in getting the fix
back in...

	- Bryan

--------------------------------------------------------------------------
Bryan Cantrill, Solaris Kernel Development.       http://blogs.sun.com/bmc

Hi Bryan, Jonathan,

Thanks to you both for the speedy replies. The workaround
is all I needed. Everything is working fine now.

Cheers,

Steve


Bryan Cantrill wrote:
>>>#!/usr/sbin/dtrace -s
>>>
>>>syscall::open:entry
>>>{
>>>        self->spec = speculation();
>>>        speculate(self->spec);
>>>        printf("%s", stringof(copyinstr(arg0)));
>>>}
>>>
>>>syscall::open:return
>>>/self->spec && errno != 0/
>>>{
>>>        commit(self->spec);
>>>        self->spec = 0;
>>>}
>>>
>>>syscall::open:return
>>>/self->spec && errno == 0/
>>>{
>>>        discard(self->spec);
>>>        self->spec = 0;
>>>}
>>>
>>>
>>>Anyone got any ideas whats going on ?
>>
>>This is definitely a bug;  I''ve filed:
>>
>>6303188 some dtrace scripts with speculations fail to load
> 
> 
> I actually have a fix for this in my current wad.  (I''m
embarrassed to say
> that this was first run into over a month ago -- I''ve got a pretty
bad case
> of wad creep at the moment.)
> 
> I introduced this with the fix for:
> 
>   6253030 adding an action to an ECB takes quadratic time
> 
> The fix is a one-liner, and I''m embarrassed that we
didn''t have a test
> to catch this in our test suite (suffice it to say that I have added 
> such a test).  The workaround would be to rephrase this as:
> 
>   syscall::open:entry
>   {
>           self->spec = speculation();
>   }
> 
>   syscall::open:entry
>   /self->spec/
>   {
>           speculate(self->spec);
>           ...
>   }
> 
> Apologies for the brokenness -- and for the delay in getting the fix
> back in...
> 
> 	- Bryan
> 
> --------------------------------------------------------------------------
> Bryan Cantrill, Solaris Kernel Development.       http://blogs.sun.com/bmc
-- 
Cheers,

Steve