RE: cvs commit: httpd-2.0/modules/ssl README mod_ssl.h ssl_engine_init.c ssl_util.c

[MATHIHALLI,MADHUSUDAN (HP-Cupertino,ex1)]

Yes.. APR_LOCKALL can be replaced by APR_INTRAPROCESS.. I've done that
change and also have introduced CRYPTO_dynamic locks.. I shall send that
patch soon..

Thanks
-Madhu

-----Original Message-----
From: Justin Erenkrantz
To: [EMAIL PROTECTED]
Sent: 8/23/01 10:28 PM
Subject: Re: cvs commit: httpd-2.0/modules/ssl README mod_ssl.h
ssl_engine_init.c ssl_util.c

On Fri, Aug 24, 2001 at 04:08:04AM -0000, [EMAIL PROTECTED] wrote:
>   +    *lock_cs = apr_palloc(p, CRYPTO_NUM_LOCKS);
>   +    for (i = 0; i < CRYPTO_NUM_LOCKS; i++)
>   +    {
>   +        lock_count[i]=0;
>   +        apr_lock_create(&(lock_cs[i]), APR_MUTEX, APR_LOCKALL,
>   +                                                mc->szMutexFile,
p);
>   +    }

I meant to review this patch, but I accidentally hit the delete key on 
the original message.  Anyway, that APR_LOCKALL can be APR_INTRAPROCESS.
We're only concerned about safety within our process not across 
multiple processes - there is no need to needlessly lock across
processes here.

IMHO, the call to ssl_util_thread_setup should be moved to 
ssl_init_Child and even potentially surrounded by 
#if APR_HAS_THREADS.  -- justin