efficient mod_perl alternatives to fork()?

[Paul <yd...@yahoo.com>]

Hi all.

I'm writing module code which (for backward compatibility with the CGI
it's replacing) needs to be able to execute commands from a file.
(~urgh~) The files have usually been ksh and/or Perl. Commonly, they
contain a directive to execute a line of shell script.

Ideally, such commands will be replaced with equivelent perl code which
could simply be eval()'d, but I'm concerned that one of my coworkers
might write his perl code with backticks in it (like `grep foo *.bar`)
instead of writing the few extra lines of code, particularly since we
have a lot of legacy code in several languages (such as ksh functions).

Wouldn't that effectively fork an entire server process before exec'ing
the qx//? And is there any simple way to prevent that, or any simple
alternative aside from a big stick? 

(a stoopid question -- Doesn't a Perl eval() with backticks do a full
blown program fork and exec, too?)

Also, I realize the security holes this might present, but the files in
question are never influenced by user input. Security isn't the
question -- just the fork.

Thanks a mil,
Paul

__________________________________________________
Do You Yahoo!?
Yahoo! Shopping - Thousands of Stores. Millions of Products.
http://shopping.yahoo.com/

Re: efficient mod_perl alternatives to fork()?

[Mike Janson <ma...@quios.net>]

You don't mention which OS you're using, but in Richard Stevens'
_Advanced Programming in the UNIX Environment_, on p.189 he states:

"Many current implementations don't perform a complete copy of the 
parent's data, stack, and heap, since a fork is often followed by an 
exec.  Instead, a technique called copy-on-write (COW) is used."

This does exactly what you'd think it would--shared heap is made 
read-only, and additional memory is allocated only when the parent or 
child try to modify data.

So, forking an entire server process isn't as bad as you'd think.

-Mike


At 1:39 PM -0800 11/28/00, Paul wrote:
>Hi all.
>
>I'm writing module code which (for backward compatibility with the CGI
>it's replacing) needs to be able to execute commands from a file.
>(~urgh~) The files have usually been ksh and/or Perl. Commonly, they
>contain a directive to execute a line of shell script.
>
>Ideally, such commands will be replaced with equivelent perl code which
>could simply be eval()'d, but I'm concerned that one of my coworkers
>might write his perl code with backticks in it (like `grep foo *.bar`)
>instead of writing the few extra lines of code, particularly since we
>have a lot of legacy code in several languages (such as ksh functions).
>
>Wouldn't that effectively fork an entire server process before exec'ing
>the qx//? And is there any simple way to prevent that, or any simple
>alternative aside from a big stick?
>
>(a stoopid question -- Doesn't a Perl eval() with backticks do a full
>blown program fork and exec, too?)
>
>Also, I realize the security holes this might present, but the files in
>question are never influenced by user input. Security isn't the
>question -- just the fork.
>
>Thanks a mil,
>Paul
>
>__________________________________________________
>Do You Yahoo!?
>Yahoo! Shopping - Thousands of Stores. Millions of Products.
>http://shopping.yahoo.com/
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: modperl-unsubscribe@apache.org
>For additional commands, e-mail: modperl-help@apache.org

Re: efficient mod_perl alternatives to fork()?

[Ken Williams <ke...@forum.swarthmore.edu>]

ydbxmhc@yahoo.com (Paul) wrote:
>I'm writing module code which (for backward compatibility with the CGI
>it's replacing) needs to be able to execute commands from a file.
>(~urgh~) The files have usually been ksh and/or Perl. Commonly, they
>contain a directive to execute a line of shell script.
>
>Ideally, such commands will be replaced with equivelent perl code which
>could simply be eval()'d, but I'm concerned that one of my coworkers
>might write his perl code with backticks in it (like `grep foo *.bar`)
>instead of writing the few extra lines of code, particularly since we
>have a lot of legacy code in several languages (such as ksh functions).
>
>Wouldn't that effectively fork an entire server process before exec'ing
>the qx//? And is there any simple way to prevent that, or any simple
>alternative aside from a big stick? 
>
>(a stoopid question -- Doesn't a Perl eval() with backticks do a full
>blown program fork and exec, too?)

No, eval() runs the code through the Perl interpreter in the same
process, with full access to any variables that exist in the current
scope.

Any code run by eval() can, of course, fork, use backticks, run system
calls, or do any number of other things that spawn sub-processes.  But
the eval() itself doesn't.

See also
http://perl.apache.org/guide/performance.html#Forking_and_Executing_Subprocess


  -------------------                            -------------------
  Ken Williams                             Last Bastion of Euclidity
  ken@forum.swarthmore.edu                            The Math Forum