This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [ping2][PATCH] Fix incorrect getaddrinfo assertion trigger
- From: Allan McRae <allan at archlinux dot org>
- To: libc-alpha at sourceware dot org
- Date: Wed, 16 Oct 2013 12:37:40 +1000
- Subject: Re: [ping2][PATCH] Fix incorrect getaddrinfo assertion trigger
- Authentication-results: sourceware.org; auth=none
- References: <1377864323-14828-1-git-send-email-allan at archlinux dot org> <52395ED0 dot 80908 at archlinux dot org>
Ping!
On 18/09/13 18:05, Allan McRae wrote:
> Ping.
>
> On 30/08/13 22:05, Allan McRae wrote:
>> [BS #9954]
>>
>> With the following /etc/hosts:
>> 127.0.0.1 www.my-domain.es
>> 127.0.1.1 www.my-domain.es
>> 192.168.0.1 www.my-domain.es
>>
>> Using getaddrinfo() on www.my-domain.es, trigger the following assertion:
>> ../sysdeps/posix/getaddrinfo.c:1473: rfc3484_sort: Assertion
>> `src->results[i].native == -1 || src->results[i].native == a1_native' failed.
>>
>> This is due to two different bugs:
>> - In rfc3484_sort() rule 7, src->results[i].native is assigned even if
>> src->results[i].index is -1, meaning that no interface is associated.
>> - In getaddrinfo() the source IP address used with the lo interface needs a
>> special case, as it can be any IP within 127.X.Y.Z.
>>
>> ---
>>
>> This is a patch taken from bugzilla. I can not replicate it myself, even
>> with the attached testcase on bugzilla. However, I have users reporting this
>> issue that also can confirm this patch fixes the assertions for them. I
>> believe this patch is widely used by other distributions (at least Debian
>> and Fedora).
>>
>> There is a negative comment by Ulrich here:
>> https://sourceware.org/bugzilla/show_bug.cgi?id=9954#c2
>>
>> I think that comment is partially correct, in that it is hiding a bug from
>> a faulty DNS provider, in which case this patch is correct in that it
>> improves robustness. That is why this bug does not affect most people.
>> However, I am not fully confident about that assessment...
>>
>> Can anyone replicate or confirm/correct my reasoning why most people can
>> not replicate?
>>
>>
>>
>> 20xx-xx-xx Aurelien Jarno <aurelien@aurel32.net>
>>
>> [BS #9954]
>> * sysdeps/posix/getaddrinfo.c (rfc3484_sort): do not assign native
>> result if the result has no associated interface.
>> * sysdeps/posix/getaddrinfo.c (getaddrinfo): correctly detect
>> interface for all 127.X.Y.Z addresses.
>>
>>
>> sysdeps/posix/getaddrinfo.c | 13 ++++++++++---
>> 1 file changed, 10 insertions(+), 3 deletions(-)
>>
>> diff --git a/sysdeps/posix/getaddrinfo.c b/sysdeps/posix/getaddrinfo.c
>> index 52177e4..c0ded84 100644
>> --- a/sysdeps/posix/getaddrinfo.c
>> +++ b/sysdeps/posix/getaddrinfo.c
>> @@ -1735,13 +1735,13 @@ rfc3484_sort (const void *p1, const void *p2, void *arg)
>>
>> /* Fill in the results in all the records. */
>> for (int i = 0; i < src->nresults; ++i)
>> - if (src->results[i].index == a1_index)
>> + if (a1_index != -1 && src->results[i].index == a1_index)
>> {
>> assert (src->results[i].native == -1
>> || src->results[i].native == a1_native);
>> src->results[i].native = a1_native;
>> }
>> - else if (src->results[i].index == a2_index)
>> + else if (a2_index != -1 && src->results[i].index == a2_index)
>> {
>> assert (src->results[i].native == -1
>> || src->results[i].native == a2_native);
>> @@ -2601,7 +2601,14 @@ getaddrinfo (const char *name, const char *service,
>> tmp.addr[0] = 0;
>> tmp.addr[1] = 0;
>> tmp.addr[2] = htonl (0xffff);
>> - tmp.addr[3] = sinp->sin_addr.s_addr;
>> + /* Special case for lo interface, the source address
>> + being possibly different than the interface
>> + address. */
>> + if ((ntohl(sinp->sin_addr.s_addr) & 0xff000000)
>> + == 0x7f000000)
>> + tmp.addr[3] = htonl(0x7f000001);
>> + else
>> + tmp.addr[3] = sinp->sin_addr.s_addr;
>> }
>> else
>> {
>>
>
>
>