b66d837bb5
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.
The resolver in the GNU C Library ********************************* Starting with version 2.2, the resolver in the GNU C Library comes from BIND 8. Only a subset of the src/lib/resolv part of libbind is included here; basically the parts that are needed to provide the functionality present in the resolver from BIND 4.9.7 that was included in the previous release of the GNU C Library, augmented by the parts needed to provide thread-safety. This means that support for things as dynamic DNS updates and TSIG keys isn't included. If you need those facilities, please take a look at the full BIND distribution. Differences =========== The resolver in the GNU C Library still differs from what's in BIND 8.2.3-T5B: * The resolver in glibc strictly adheres to the recommendations in RFC 1535. BIND 8.2.3-T5B seems to relax those rules a bit (see the code that's wrapped in `#ifndef RFC1535'). * The RES_DEBUG option (`options debug' in /etc/resolv.conf) has been disabled. * The resolver in glibc allows underscores in domain names. * The <resolv.h> header in glibc includes <netinet/in.h> and <arpa/nameser.h> to make it self-contained. * The `res_close' function in glibc only tries to close open files referenced through `_res' if the RES_INIT bit is set in `_res.options'. This fixes a potential security bug with programs that bogusly call `res_close' without initialising the resolver state first. Note that the thread-safe `res_nclose' still doesn't check the RES_INIT bit. By the way, you're not really supposed to call `res_close/res_nclose' directly. * The resolver in glibc can connect to a nameserver over IPv6. Just specify the IPv6 address in /etc/resolv.conf. You cannot change the address of an IPv6 nameserver dynamically in your program though. Using the resolver in multi-threaded code ========================================= The traditional resolver interfaces `res_query', `res_search', `res_mkquery', `res_send' and `res_init', used a static (global) resolver state stored in the `_res' structure. Therefore, these interfaces are not thread-safe. Therefore, BIND 8.2 introduced a set of "new" interfaces `res_nquery', `res_nsearch', `res_nmkquery', `res_nsend' and `res_ninit' that take a `res_state' as their first argument, so you can use a per-thread resolver state. In glibc, when you link with -lpthread, such a per-thread resolver state is already present. It can be accessed using `_res', which has been redefined as a macro, in a similar way to what has been done for the `errno' and `h_errno' variables. This per-thread resolver state is also used for the `gethostby*' family of functions, which means that for example `gethostbyname_r' is now fully thread-safe and re-entrant. The traditional resolver interfaces however, continue to use a single resolver state and are therefore still thread-unsafe. The resolver state is the same resolver state that is used for the initial ("main") thread. This has the following consequences for existing binaries and source code: * Single-threaded programs will continue to work. There should be no user-visible changes when you recompile them. * Multi-threaded programs that use the traditional resolver interfaces in the "main" thread should continue to work, except that they no longer see any changes in the global resolver state caused by calls to, for example, `gethostbyname' in other threads. Again there should be no user-visible changes when you recompile these programs. * Multi-threaded programs that use the traditional resolver interfaces in more than one thread should be just as buggy as before (there are no problems if you use proper locking of course). If you recompile these programs, manipulating the _res structure in threads other than the "main" thread will seem to have no effect though. * In Multi-threaded that manipulate the _res structure, calls to functions like `gethostbyname' in threads other than the "main" thread won't be influenced by the those changes anymore. So if you set RES_USE_INET6, a call to `gethostbyname' won't return any IPv6 hosts anymore. If you recompile such programs, manipulating the _res structure will affect the thread in which you do so instead of the "main" thread. We recommend to use the new thread-safe interfaces in new code, since the traditional interfaces have been deprecated by the BIND folks. For compatibility with other (older) systems you might want to continue to use those interfaces though. Using the resolver in C++ code ============================== There resolver contains some hooks which will allow the user to install some callback functions that make it possible to filter DNS requests and responses. Although we do not encourage you to make use of this facility at all, C++ developers should realise that it isn't safe to throw exceptions from such callback functions. Source code =========== The following files come from the BIND distribution (currently version 8.2.3-T5B): src/include/ arpa/nameser.h arpa/nameser_compat.h resolv.h src/lib/resolv/ herror.c res_comp.c res_data.c res_debug.c res_debug.h res_init.c res_mkquery.c res_query.c res_send.c src/lib/nameser/ ns_name.c ns_netint.c ns_parse.c ns_print.c ns_samedomain.c ns_ttl.c src/lib/inet/ inet_addr.c inet_net_ntop.c inet_net_pton.c inet_neta.c inet_ntop.c inet_pton.c nsap_addr.c src/lib/isc/ base64.c Some of these files have been optimised a bit, and adaptations have been made to make them fit in with the rest of glibc. The more non-obvious changes are wrapped in something like `#ifdef _LIBC'. res_libc.c is home-brewn, although parts of it are taken from res_data.c. res_hconf.c and res_hconf.h were contributed by David Mosberger, and do not come from BIND. The files gethnamaddr.c, mapv4v6addr.h and mapv4v6hostent.h are leftovers from BIND 4.9.7.