(RHSA-2017:0501) Important: kernel security and bug fix update

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

Security Fix(es):

• A use-after-free flaw was found in the way the Linux kernel’s Datagram Congestion Control Protocol (DCCP) implementation freed SKB (socket buffer) resources for a DCCP_PKT_REQUEST packet when the IPV6_RECVPKTINFO option is set on the socket. A local, unprivileged user could use this flaw to alter the kernel memory, allowing them to escalate their privileges on the system. (CVE-2017-6074, Important)

Red Hat would like to thank Andrey Konovalov (Google) for reporting this issue.

Bug Fix(es):

• Previously, running the “ethtool -S” command to get the statistics of a Brocade Network Adapter (BNA) sometimes caused a kernel panic. This update applies a set of patches to the bna driver, and the kernel panic no longer occurs in the described scenario. (BZ#1408130)

• Use of a multi-threaded workload with high memory mappings sometimes caused a kernel panic, due to a race condition between the context switch and the pagetable upgrade. This update fixes the switch_mm() by using the complete asce parameter instead of the asce_bits parameter. As a result, the kernel no longer panics in the described scenario. (BZ#1410865)

• Previously, the kernel was sending a Transmission Control Protocol (TCP) window which had a size of zero for a socket with an empty receive queue. Consequently, the TCP session became unresponsive. This update fixes the ibmveth driver to set correct values of the gso_size and gso_type variables and to calculate the value of the gso_segs variable for large packets. As a result, the TCP session no longer hangs in the described scenario. (BZ#1411381)

• Previously, booting a kdump kernel in some cases failed with this error:

  Kernel panic - not syncing: Watchdog detected hard LOCKUP on CPU 0.

This update ensures that the hpet timer software counters, including hpet_default_delta and hpet_t1_cmp, are initialized before an interrupt request is registered, and the kdump kernel now boots without the mentioned error message. (BZ#1404183)

• When one of the drives became unresponsive, all other drives intermittently hung, because the megaraid_sas driver incorrectly sent a reset request to the PowerEdge RAID Controller (PERC). This update fixes megaraid_sas, and thus the hang of one drive no longer leads to intermittent loss of access to all drives on the system. (BZ#1398174)