kernel-2.6.32-696.16.1.el6

The kernel packages contain the Linux kernel, the core of any Linux operating system.

Security Fix(es):

* A race condition issue leading to a use-after-free flaw was found in the way the raw packet sockets are implemented in the Linux kernel networking subsystem handling synchronization. A local user able to open a raw packet socket (requires the CAP_NET_RAW capability) could use this flaw to elevate their privileges on the system. (CVE-2017-1000111, Important)

* An exploitable memory corruption flaw was found in the Linux kernel. The append path can be erroneously switched from UFO to non-UFO in ip_ufo_append_data() when building an UFO packet with MSG_MORE option. If unprivileged user namespaces are available, this flaw can be exploited to gain root privileges. (CVE-2017-1000112, Important)

* A divide-by-zero vulnerability was found in the __tcp_select_window function in the Linux kernel. This can result in a kernel panic causing a local denial of service. (CVE-2017-14106, Moderate)

Asianux would like to thank Willem de Bruijn for reporting CVE-2017-1000111 and Andrey Konovalov for reporting CVE-2017-1000112.

Bug Fix(es):

* When the operating system was booted with Asianux Virtualization, and the eh_deadline sysfs parameter was set to 10s, the Storage Area Network (SAN) issues caused eh_deadline to trigger with no handler. Consequently, a kernel panic occurred. This update fixes the lpfc driver, thus preventing the kernel panic under described circumstances. (BZ#1487220)

* When an NFS server returned the NFS4ERR_BAD_SEQID error to an OPEN request, the open-owner was removed from the state_owners rbtree. Consequently, NFS4 client infinite loop that required a reboot to recover occurred. This update changes NFS4ERR_BAD_SEQID handling to leave the open-owner in the state_owners rbtree by updating the create_time parameter so that it looks like a new open-owner. As a result, an NFS4 client is now able to recover without falling into the infinite recovery loop after receiving NFS4ERR_BAD_SEQID. (BZ#1491123)

* If an NFS client attempted to mount NFSv3 shares from an NFS server exported directly to the client's IP address, and this NFS client had already mounted other shares that originated from the same server but were exported to the subnetwork which this client was part of, the auth.unix.ip cache expiration was not handled correctly. Consequently, the client received the 'stale file handle' errors when trying to mount the share. This update fixes handling of the cache expiration, and the NFSv3 shares now mount as expected without producing the 'stale file handle' errors. (BZ#1497976)

* When running a script that raised the tx ring count to its maximum value supported by the Solarflare Network Interface Controller (NIC) driver, the EF10 family NICs allowed the settings exceeding the hardware's capability. Consequently, the Solarflare hardware became unusable with Asianux Entepripse Linux 6. This update fixes the sfc driver, so that the tx ring can have maximum 2048 entries for all EF10 NICs. As a result, the Solarflare hardware no longer becomes unusable with Asianux Entepripse Linux 6 due to this bug. (BZ#1498019)

CVE-2017-1000111
Linux kernel: heap out-of-bounds in AF_PACKET sockets. This new issue
is analogous to previously disclosed CVE-2016-8655. In both cases, a
socket option that changes socket state may race with safety checks in
packet_set_ring. Previously with PACKET_VERSION. This time with
PACKET_RESERVE. The solution is similar: lock the socket for the
update. This issue may be exploitable, we did not investigate further.
As this issue affects PF_PACKET sockets, it requires CAP_NET_RAW in
the process namespace. But note that with user namespaces enabled, any
process can create a namespace in which it has CAP_NET_RAW.
CVE-2017-1000112
Linux kernel: Exploitable memory corruption due to UFO to non-UFO path
switch. When building a UFO packet with MSG_MORE __ip_append_data()
calls ip_ufo_append_data() to append. However in between two send()
calls, the append path can be switched from UFO to non-UFO one, which
leads to a memory corruption. In case UFO packet lengths exceeds MTU,
copy = maxfraglen - skb->len becomes negative on the non-UFO path and
the branch to allocate new skb is taken. This triggers fragmentation
and computation of fraggap = skb_prev->len - maxfraglen. Fraggap can
exceed MTU, causing copy = datalen - transhdrlen - fraggap to become
negative. Subsequently skb_copy_and_csum_bits() writes out-of-bounds.
A similar issue is present in IPv6 code. The bug was introduced in
e89e9cf539a2 ("[IPv4/IPv6]: UFO Scatter-gather approach") on Oct 18
2005.
CVE-2017-14106
The tcp_disconnect function in net/ipv4/tcp.c in the Linux kernel
before 4.12 allows local users to cause a denial of service
(__tcp_select_window divide-by-zero error and system crash) by
triggering a disconnect within a certain tcp_recvmsg code path.