dtrace discuss - Apr 2007 - Function shows up in the stack, but can''t probe it

I''m trying to probe a driver function, but am not getting any results. 
I know that ce_drain_fifo is being entered, via below dtrace stack. 
Yet, when I try and probe ce_drain_fifo, I doesn''t fire.

I know that tail calls won''t fire a return probe, but thought that
entry
would.

   ce`ce_drain_fifo+0x79c
   unix`thread_start+0x4

Any clues?

Thanks,
Jim

Hi Jim,

ce_drain_fifo is entered only at CE instance plumb time (or
maybe at up time). Then the thread sits within ce_drain_fifo.
So during the CE normal operation the ce_drain_fifo:entry is 
not fired even if the function shows up in the stack of
active threads.

Could this explain what you see ?

-r

PS tail call does fire the return probe. But it calls it
   does so _before_ the start of the tail called target.


Jim Nissen writes:
 > I''m trying to probe a driver function, but am not getting any
results.
 > I know that ce_drain_fifo is being entered, via below dtrace stack. 
 > Yet, when I try and probe ce_drain_fifo, I doesn''t fire.
 > 
 > I know that tail calls won''t fire a return probe, but thought
that entry
 > would.
 > 
 >    ce`ce_drain_fifo+0x79c
 >    unix`thread_start+0x4
 > 
 > Any clues?
 > 
 > Thanks,
 > Jim
 > _______________________________________________
 > dtrace-discuss mailing list
 > dtrace-discuss at opensolaris.org

All,
Wanted to summarize the findings on this.  First, as Roch pointed out, 
ce_drain_fifo must get called when interface is plumbed up, but that''s 
not the reason I couldn''t get a probe to fire on it.  There were other 
ce functions, that I also could not get probes to fire on.  The reason 
is because the ce driver is compiled with the ''inline'' flag,
which makes
it more efficient, but I''ve been told it offers less
"visibility" for
debugging purposes.

I was offered a ce binary compiled without this flag, and have yet to 
reimplement dtrace probes.

Jim
> Hi Jim,
>
> ce_drain_fifo is entered only at CE instance plumb time (or
> maybe at up time). Then the thread sits within ce_drain_fifo.
> So during the CE normal operation the ce_drain_fifo:entry is 
> not fired even if the function shows up in the stack of
> active threads.
>
> Could this explain what you see ?
>
> -r
>
> PS tail call does fire the return probe. But it calls it
>    does so _before_ the start of the tail called target.
>
>
> Jim Nissen writes:
>  > I''m trying to probe a driver function, but am not getting
any results.
>  > I know that ce_drain_fifo is being entered, via below dtrace stack. 
>  > Yet, when I try and probe ce_drain_fifo, I doesn''t fire.
>  > 
>  > I know that tail calls won''t fire a return probe, but
thought that entry
>  > would.
>  > 
>  >    ce`ce_drain_fifo+0x79c
>  >    unix`thread_start+0x4
>  > 
>  > Any clues?
>  > 
>  > Thanks,
>  > Jim
>  > _______________________________________________
>  > dtrace-discuss mailing list
>  > dtrace-discuss at opensolaris.org
>
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
>