dtrace discuss - Mar 2006 - Triggering on close of a written file.

I have just started my first ''serious'' dtrace script and can
use some
advice.

I want to have a probe triggered when close() occurs after a  
succesfull write of /etc/hosts ( I want to regenerate the nameserver  
zone files
from /etc/hosts after it has changed)

At the moment I have the following code:

syscall::open*:entry,
syscall::creat*:entry
{
         self->file=arg0;
}

syscall::open*:return,
syscall::creat*:return
/basename(copyinstr(self->file)) == "hosts"/
{
         self->hostsfd=arg1;
/*        printf("Filename=%s fd=%d",basename(copyinstr(self- 
 >file)),arg1); */
}

syscall::write:entry
/arg0==self->hostsfd/
{
         /* trace("Write hosts"); */
         self->hostswritten=1;
}

syscall::close:entry
/arg0==self->hostsfd && self->hostswritten==1/
{
	system("/usr/local/bin/regen-zonefiles");
         self->hostsfd=-1 ;
}


Is there an easier/better way to do this, I think this script fails  
to detect a failed write, what is the best way to do that...


	Paul

Hi Paul,

You''re correct that this script fails to detect a failed write. You can
examine the value returned from write (arg1 in syscall::write:return) to
check for errors. Another problem is that this script assumes a single-
threaded process or at least that a single thread is opening, writing, and
closing /etc/hosts. If you want to deal with a multi-threaded process you
could use an associative array keyed by the pid rather than thread-local
variables.

Adam

On Fri, Mar 24, 2006 at 03:37:32PM +0100, Paul van der Zwan
wrote:
> I have just started my first ''serious'' dtrace script and
can use some
> advice.
> 
> I want to have a probe triggered when close() occurs after a  
> succesfull write of /etc/hosts ( I want to regenerate the nameserver  
> zone files
> from /etc/hosts after it has changed)
> 
> At the moment I have the following code:
> 
> syscall::open*:entry,
> syscall::creat*:entry
> {
>         self->file=arg0;
> }
> 
> syscall::open*:return,
> syscall::creat*:return
> /basename(copyinstr(self->file)) == "hosts"/
> {
>         self->hostsfd=arg1;
> /*        printf("Filename=%s fd=%d",basename(copyinstr(self- 
> >file)),arg1); */
> }
> 
> syscall::write:entry
> /arg0==self->hostsfd/
> {
>         /* trace("Write hosts"); */
>         self->hostswritten=1;
> }
> 
> syscall::close:entry
> /arg0==self->hostsfd && self->hostswritten==1/
> {
> 	system("/usr/local/bin/regen-zonefiles");
>         self->hostsfd=-1 ;
> }
> 
> 
> Is there an easier/better way to do this, I think this script fails  
> to detect a failed write, what is the best way to do that...
> 
> 
> 	Paul
> 
> 
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
-- 
Adam Leventhal, Solaris Kernel Development       http://blogs.sun.com/ahl

Paul van der Zwan,???

	I don''t get an easier/better way to do this, just add
"write:retuen" to detect a failed write.

syscall::write:return
/self->hostswritten==1 && arg1 < 0 /
{
         /* trace("Write failed"); */
         self->hostswritten=0;
}

======= 2006-03-24 22:37:32 ????????======
>
>I have just started my first ''serious'' dtrace script and
can use some
>advice.
>
>I want to have a probe triggered when close() occurs after a  
>succesfull write of /etc/hosts ( I want to regenerate the nameserver  
>zone files
>from /etc/hosts after it has changed)
>
>At the moment I have the following code:
>
>syscall::open*:entry,
>syscall::creat*:entry
>{
>         self->file=arg0;
>}
>
>syscall::open*:return,
>syscall::creat*:return
>/basename(copyinstr(self->file)) == "hosts"/
>{
>         self->hostsfd=arg1;
>/*        printf("Filename=%s fd=%d",basename(copyinstr(self- 
> >file)),arg1); */
>}
>
>syscall::write:entry
>/arg0==self->hostsfd/
>{
>         /* trace("Write hosts"); */
>         self->hostswritten=1;
>}
>
>syscall::close:entry
>/arg0==self->hostsfd && self->hostswritten==1/
>{
>	system("/usr/local/bin/regen-zonefiles");
>         self->hostsfd=-1 ;
>}
>
>
>Is there an easier/better way to do this, I think this script fails  
>to detect a failed write, what is the best way to do that...
>
>
>	Paul
>
>
>_______________________________________________
>dtrace-discuss mailing list
>dtrace-discuss at opensolaris.org
>
= = = = = = = = = = = = = = = = = = = 			

?????????
??
 
				 
????????zheh
????????zheh at 163.com
??????????2006-03-25

On Sat, 25 Mar 2006, zheh wrote:
> Paul van der Zwan,??????
>
> 	I don''t get an easier/better way to do this, just add
"write:retuen" to detect a failed write.
>
> syscall::write:return
> /self->hostswritten==1 && arg1 < 0 /
> {
>         /* trace("Write failed"); */
>         self->hostswritten=0;
> }
Hi Paul,

There is a good example of analyzing open() system call errors in 
/usr/demo/dtrace/specopen.d. The script uses speculations, which
can be super useful for commiting traced data when an error occurs
(this may not be needed in your case).

Hope this helps,
- Ryan
-- 
UNIX Administrator
http://daemons.net/~matty