Re: [patches] [PATCH] Reload resolv.conf for all threads

[Maxim Kuvyrkov <maxim@xxxxxxxxxxxxxxxx>]

On 9/15/10 4:11 AM, Aurelien Jarno wrote:
> When resolv.conf is modified, the resolver code noticed that and reload
> it. However on a multithread program, the thread which notices the
> change doesn't ask all the other to re-initialize their resolver. The
> patch below fixes the issue.
>
> I submit it here, as this part of the code is EGLIBC specific and not p
> art of GLIBC (though present in most distributions).

Michael (CC'ed) is the author of the code in question.

> 2010-09-15  Aurelien Jarno<aurelien@xxxxxxxxxxx>
>
> 	* resolv/res_libc (__res_maybe_init): Ask other threads to
> 	to re-initialize their resolver states when _PATH_RESCONF
> 	changes.
>
> --- a/resolv/res_libc.c
> +++ b/resolv/res_libc.c
> @@ -100,10 +100,13 @@
>
>   	if (resp->options&  RES_INIT) {
>   		ret = stat (_PATH_RESCONF,&statbuf);
> -		if ((__res_initstamp != resp->_u._ext.initstamp)
> -		    || ((ret == 0)&&  (last_mtime != statbuf.st_mtime))) {
> -			if (ret == 0)
> -				last_mtime = statbuf.st_mtime;
> +		if ((ret == 0)&&  (last_mtime != statbuf.st_mtime)) {
> +			last_mtime = statbuf.st_mtime;
> +			atomicinclock (lock);
> +			atomicinc (__res_initstamp);
> +			atomicincunlock (lock);
> +		}
> +		if (__res_initstamp != resp->_u._ext.initstamp) {
>   			if (resp->nscount>  0)
>   				__res_iclose (resp, true);
>   			return __res_vinit (resp, 1);

This looks correct to me.  Michael, do you have any comments?

--
Maxim Kuvyrkov
CodeSourcery
maxim@xxxxxxxxxxxxxxxx
(650) 331-3385 x724