[pulseaudio-discuss] [PATCH v2] module-rtp-recv: fix crash on empty UDP packets (CVE-2014-3970)
Alexander E. Patrakov
patrakov at gmail.com
Thu Jun 5 09:29:25 PDT 2014
On FIONREAD returning 0 bytes, we cannot return success, as the caller
(rtpoll_work_cb in module-rtp-recv.c) would then try to
pa_memblock_unref(chunk.memblock) and, because memblock is NULL, trigger
an assertion.
Also we have to read out the possible empty packet from the socket, so
that the kernel doesn't tell us again and again about it.
Signed-off-by: Alexander E. Patrakov <patrakov at gmail.com>
---
src/modules/rtp/rtp.c | 25 +++++++++++++++++++++++--
1 file changed, 23 insertions(+), 2 deletions(-)
diff --git a/src/modules/rtp/rtp.c b/src/modules/rtp/rtp.c
index 570737e..7b75e0e 100644
--- a/src/modules/rtp/rtp.c
+++ b/src/modules/rtp/rtp.c
@@ -182,8 +182,29 @@ int pa_rtp_recv(pa_rtp_context *c, pa_memchunk *chunk, pa_mempool *pool, struct
goto fail;
}
- if (size <= 0)
- return 0;
+ if (size <= 0) {
+ /* size can be 0 due to any of the following reasons:
+ *
+ * 1. Somebody sent us a perfectly valid zero-length UDP packet.
+ * 2. Somebody sent us a UDP packet with a bad CRC.
+ *
+ * It is unknown whether size can actually be less than zero.
+ *
+ * In the first case, the packet has to be read out, otherwise the
+ * kernel will tell us again and again about it, thus preventing
+ * reception of any further packets. So let's just read it out
+ * now and discard it later, when comparing the number of bytes
+ * received (0) with the number of bytes wanted (1, see below).
+ *
+ * In the second case, recvmsg() will fail, thus allowing us to
+ * return the error.
+ *
+ * Just to avoid passing zero-sized memchunks and NULL pointers to
+ * recvmsg(), let's force allocation of at least one byte by setting
+ * size to 1.
+ */
+ size = 1;
+ }
if (c->memchunk.length < (unsigned) size) {
size_t l;
--
1.9.3
More information about the pulseaudio-discuss
mailing list