Re: [PATCH] resolver.3: documents missing options used by _res structure defined in resolv.h and indicates defaults options

["Michael Kerrisk (man-pages)" <mtk.manpages-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On 03/09/2015 08:54 AM, Stéphane Aulery wrote:
> Le lundi 09 mars 2015 à 08:24:38, Michael Kerrisk (man-pages) a écrit :
>> Hello Stéphane,
>>
>> On 03/08/2015 01:03 PM, Stéphane Aulery wrote:
>>> Hi Michael,
>>>
>>> I made another patch with all your remarks.
>>
>> Did you send it already. (No hurry--it's just that I wonder if I missed it.)
> 
> I was waiting your answers to the question bellow before.
> 
>>>>> +.B RES_BLAST
>>>>> +When resolvers want to even out the load on their nameservers, following
>>>>> +them. 
>>>>
>>>> I do not understand that last sentence. Could you clarify please.
>>>
>>> I must admit that it's not clear to me either. I think this means that
>>> server alias are swapped between each request. I have wrong? The word
>>> "blast" leaves me skeptical.
>>>
>>> What I found in code source:
>>>
>>> >From resolv.h:
>>>
>>> #define RES_BLAST       0x00020000      /* blast all recursive servers */
>>>
>>> Form res_send.c
>>>
>>> /*
>>> * Some resolvers want to even out the load on their nameservers.
>>> * Note that RES_BLAST overrides RES_ROTATE.
>>> */
>>> if (__builtin_expect ((statp->options & RES_ROTATE) != 0, 0) &&
>>>    (statp->options & RES_BLAST) == 0) {
>>>       struct sockaddr_in6 *ina;
>>>       unsigned int map;
>>>
>>>       n = 0;
>>>       while (n < MAXNS && EXT(statp).nsmap[n] == MAXNS)
>>>          n++;
>>>       if (n < MAXNS) {
>>>          ina = EXT(statp).nsaddrs[n];
>>>          map = EXT(statp).nsmap[n];
>>>          for (;;) {
>>>             ns = n + 1;
>>>             while (ns < MAXNS
>>>                && EXT(statp).nsmap[ns] == MAXNS)
>>>                ns++;
>>>             if (ns == MAXNS)
>>>                break;
>>>             EXT(statp).nsaddrs[n] = EXT(statp).nsaddrs[ns];
>>>             EXT(statp).nsmap[n] = EXT(statp).nsmap[ns];
>>>             n = ns;
>>>          }
>>>          EXT(statp).nsaddrs[n] = ina;
>>>          EXT(statp).nsmap[n] = map;
>>>       }
>>> }
>>
>> Okay -- I leave it to your discretion to choose what to do here.
> 
> No please, I need your help.

Sorry -- the code is not clear to me, and I don't have the time 
to investigate deeper.

Google turns up this detail from the O'Reilly DNS and BIND
book:

    "Blast" all recursive servers by sending queries to 
    the simultaneously. Not implemented yet.

The first sentence makes sense. I'm not sure whether
the second is true for glibc or not. But, I just grepped
the entire C source code for Fedora 20, and found no users 
of RES_BLAST, so indeed maybe it is not implemented.

Thanks,

Michael



-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html