dtrace discuss - Jul 2009 - dtrace behaviour change...?

Lookig at dtrace output from 118, I see the following (as an example):

  6  -> ipnet_bindreq
  6   | ipnet_bindreq:return
  6  <- ipnet_bindreq                         60012e80658
  6  -> dlerrorack
  6    -> mexchange
  6    <- mexchange                           6001094f7a0
  6  <- dlerrorack                            60012e80658
  6  -> qreply
  6  <- qreply                                60012e80560

when doing "/usr/sbin/dtrace -Fs".

What''s wrong here is that dlerrorack is not shown as being executed
from ipnet_bindreq, even though it is.

What this means is that I cannot do the following:

fbt:ipnet:ipnet_bindreq:entry { self->trace = 1; }
fbt:ipnet:ipnet_bindreq:return { self->trace = 0; }
fbt::dlerrorack:entry / self->trace / {}

... because self->trace is always 0 when dlerrorack''entry gets hit.

Is this:
- a build problem
- a dtrace problem
- a dtrace behaviour change
- something else

Darren

Hey Darren,

My guess is that ip_bindreq() is making a tail-call to dlerrorack()  
which
would explain the output below. You can check this by disassembling
ip_bindreq().

Adam

On Jul 24, 2009, at 6:57 PM, Darren Reed wrote:
> Lookig at dtrace output from 118, I see the following (as an example):
>
> 6  -> ipnet_bindreq
> 6   | ipnet_bindreq:return
> 6  <- ipnet_bindreq                         60012e80658
> 6  -> dlerrorack
> 6    -> mexchange
> 6    <- mexchange                           6001094f7a0
> 6  <- dlerrorack                            60012e80658
> 6  -> qreply
> 6  <- qreply                                60012e80560
>
> when doing "/usr/sbin/dtrace -Fs".
>
> What''s wrong here is that dlerrorack is not shown as being
executed
> from ipnet_bindreq, even though it is.
>
> What this means is that I cannot do the following:
>
> fbt:ipnet:ipnet_bindreq:entry { self->trace = 1; }
> fbt:ipnet:ipnet_bindreq:return { self->trace = 0; }
> fbt::dlerrorack:entry / self->trace / {}
>
> ... because self->trace is always 0 when dlerrorack''entry gets
hit.
>
> Is this:
> - a build problem
> - a dtrace problem
> - a dtrace behaviour change
> - something else
>
> Darren
>
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org

--
Adam Leventhal, Fishworks                        http://blogs.sun.com/ahl