Whamcloud Community JIRA

[LU-9572] kernel update [RHEL6.9 2.6.32-696.3.1.el6]

Related — Mon, 19 Jun 2017 20:02:24 +0000

Security Fix(es):

A flaw was found in the Linux kernel's handling of packets with the URG flag.
Applications using the splice() and tcp_splice_read() functionality can allow a
remote attacker to force the kernel to enter a condition in which it can loop
indefinitely. (CVE-2017-6214, Moderate)

Bug Fix(es):

When executing certain Hadoop jobs, a kernel panic occasionally occurred on
multiple nodes of a cluster. This update fixes the kernel scheduler, and the
kernel panic no longer occurs under the described circumstances. (BZ#1436241)

Previously, memory leak of the struct cred data structure and related data
structures occasionally occurred. Consequently, system performance was
suboptimal with the symptoms of high I/O operations wait and small amount of
free memory. This update fixes the reference counter of the struct slab cache to
no longer cause imbalance between the calls to the get_cred() function and the
put_cred() function. As a result, the memory leak no longer occurs under the
described circumstances. (BZ#1443234)

Previously, the be2net driver could not detect the link status properly on IBM
Power Systems. Consequently, the link status was always reported as
disconnected. With this update, be2net has been fixed, and the Network Interface
Cards (NICs) now report the link status correctly. (BZ#1442979)

Previously, the RFF_ID and RFT_ID commands in the lpfc driver were issued in
an incorrect order. Consequently, users were not able to access Logical Unit
Numbers (LUNs). With this update, lpfc has been fixed to issue RFT_ID before
RFF_ID, which is the correct order. As a result, users can now access LUNs as
expected. (BZ#1439636)

Previously, the kdump mechanism was trying to get the lock by the
vmalloc_sync_all() function during a kernel panic. Consequently, a deadlock
occurred, and the crashkernel did not boot. This update fixes the
vmalloc_sync_all() function to avoid synchronizing the vmalloc area on the
crashing CPU. As a result, the crashkernel parameter now boots as expected, and
the kernel dump is collected successfully under the described circumstances.
(BZ#1443499)

Bugs fixed (https://bugzilla.redhat.com/):

1426542 - CVE-2017-6214 kernel: ipv4/tcp: Infinite loop in tcp_splice_read()