laforge has submitted this change. ( https://gerrit.osmocom.org/c/osmo-bsc/+/42886?usp=email )
Change subject: ipaccess-proxy: reject oversized IPA frame length ......................................................................
ipaccess-proxy: reject oversized IPA frame length
ipaccess_proxy_read_msg() reads the 16-bit IPA frame length from the wire header and passes it straight as the recv() count into a msgb that was allocated with a fixed PROXY_ALLOC_SIZE (1200) bytes, without ever checking it against the buffer tailroom. A peer that advertises a body length larger than the remaining buffer space makes recv() write past the end of the heap allocation (heap buffer overflow).
Reject frames whose advertised length exceeds the msgb tailroom, the same way the other IPA read paths bound the read to msgb_tailroom().
This issue has been assigned the CVE candidate identifier CAN-2026-2051036.
Change-Id: I05137e114eaa99ff0e85eecccf7645c90945214f --- M src/ipaccess/ipaccess-proxy.c 1 file changed, 6 insertions(+), 0 deletions(-)
Approvals: Jenkins Builder: Verified laforge: Looks good to me, approved fixeria: Looks good to me, but someone else must approve
diff --git a/src/ipaccess/ipaccess-proxy.c b/src/ipaccess/ipaccess-proxy.c index 61c2fc2..71190f6 100644 --- a/src/ipaccess/ipaccess-proxy.c +++ b/src/ipaccess/ipaccess-proxy.c @@ -583,6 +583,12 @@ /* then read the length as specified in header */ msg->l2h = msg->data + sizeof(*hh); len = ntohs(hh->len); + if (len > msgb_tailroom(msg)) { + LOGP(DLINP, LOGL_ERROR, "Oversized IPA frame: len %d > tailroom %d\n", len, msgb_tailroom(msg)); + msgb_free(msg); + *error = -EIO; + return NULL; + } ret = recv(bfd->fd, msg->l2h, len, 0); if (ret < len) { LOGP(DLINP, LOGL_ERROR, "short read!\n");
-
laforge