res_query.c:262: __libc_res_nquery: Assertion

[Charles F. Stephens]

Debian report https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=816669
Upstream report https://sourceware.org/bugzilla/show_bug.cgi?id=19791

Whenever, in resolv.conf, there are both IPv6 and IPv4 server entries and either protocol is not routable on a system, resolv throws an assertion:

res_query.c:262: __libc_res_nquery: Assertion `(hp != ((void *)0)) && (hp2 != ((void *)0))' failed.

This problem was introduced as a side-effect of patching CVE-2015-7547.

There is a fix upstream:

commit 10d268070a8aa9a878668e7f060e92ed668de146
Author: Florian Weimer <email address hidden>
Date: Fri Mar 25 11:49:51 2016 +0100

    resolv: Always set *resplen2 out parameter in send_dg [BZ #19791]

    Since commit 44d20bca52ace85850012b0ead37b360e3ecd96e (Implement
    second fallback mode for DNS requests), there is a code path which
    returns early, before *resplen2 is initialized. This happens if the
    name server address is immediately recognized as invalid (because of
    lack of protocol support, or if it is a broadcast address such
    255.255.255.255, or another invalid address).

    If this happens and *resplen2 was non-zero (which is the case if a
    previous query resulted in a failure), __libc_res_nquery would reuse
    an existing second answer buffer. This answer has been previously
    identified as unusable (for example, it could be an NXDOMAIN
    response). Due to the presence of a second answer, no name server
    switching will occur. The result is a name resolution failure,
    although a successful resolution would have been possible if name
    servers have been switched and queries had proceeded along the search
    path.

    The above paragraph still simplifies the situation. Before glibc
    2.23, if the second answer needed malloc, the stub resolver would
    still attempt to reuse the second answer, but this is not possible
    because __libc_res_nsearch has freed it, after the unsuccessful call
    to __libc_res_nquerydomain, and set the buffer pointer to NULL. This
    eventually leads to an assertion failure in __libc_res_nquery:

     /* Make sure both hp and hp2 are defined */
     assert((hp != NULL) && (hp2 != NULL));

    If assertions are disabled, the consequence is a NULL pointer
    dereference on the next line.

    Starting with glibc 2.23, as a result of commit
    e9db92d3acfe1822d56d11abcea5bfc4c41cf6ca (CVE-2015-7547: getaddrinfo()
    stack-based buffer overflow (Bug 18665)), the second answer is always
    allocated with malloc. This means that the assertion failure happens
    with small responses as well because there is no buffer to reuse, as
    soon as there is a name resolution failure which triggers a search for
    an answer along the search path.

    This commit addresses the issue by ensuring that *resplen2 is
    initialized before the send_dg function returns.

    This commit also addresses a bug where an invalid second reply is
    incorrectly returned as a valid to the caller.

    (cherry picked from commit b66d837bb5398795c6b0f651bd5a5d66091d8577)

[Charles F. Stephens (cfs)]

Ping: has anyone looked at this ticket yet? It seems like a trivial fix to incorporate into eglibc. Thanks.