httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mathihalli, Madhusudan" <mad...@hp.com>
Subject RE: mod_cgi and apr_setup_signal_thread
Date Tue, 20 Apr 2004 14:36:23 GMT
Any comments ?
 
Here's another perspective : Why should a CGI process block the delivery of all those signals
that are not of interest to httpd ?
 
-Madhu

________________________________

From: Mathihalli, Madhusudan
Sent: Tue 4/13/2004 9:15 AM
To: dev@httpd.apache.org
Subject: RE: mod_cgi and apr_setup_signal_thread




You're partially correct. Here's whatz happening:

worker.c: child_main() invokes apr_setup_signal_thread()
signals.c: apr_setup_signal_thread()
           -> sigfillset(&sig_mask);
              remove_sync_sigs(&sig_mask);
              sigprocmask(SIG_SETMASK, &sig_mask, NULL)
              -> Thus disabling the delivery of async signals to the process

mod_cgi comes along, forks a child process and happily inherits the signal mask from the parent
(the only thing apr_proc_create() does is to change SIGCHLD to default behaviour).


-Madhu

>-----Original Message-----
>From: Jeff Trawick [mailto:trawick@attglobal.net]
>Sent: Tuesday, April 13, 2004 4:55 AM
>To: dev@httpd.apache.org
>Subject: Re: mod_cgi and apr_setup_signal_thread
>
>
>Mathihalli, Madhusudan wrote:
>> I'm using the worker MPM for both mod_cgi and mod_cgid.
>
>I don't see any httpd 2.0/apr 0.9 code which manipulates
>SIGALRM.  Can you give
>me a pointer?
>



Mime
View raw message