VPLEX: A directors' firmware may crash due to a bug related to the cache layer

Dell Impacted Hardware:
VPLEX-Series
VPLEX-VS2
VPLEX-VS6
VPLEX-Local
VPLEX-Metro-FC and IP

Dell Impacted Software:
GeoSynchrony 6.0.x

Impact:
Director firmware will restart, however there should not be any impact for hosts configured for HA.  Hosts will lose paths to the director when the firmware crashes and should regain the paths when the firmware automatically restarts.

Issue:
A VPLEX director may experience a firmware crash with the following message logged:

Reason for the crash: /export/local1/jenkins/clone_D50.10/nsfw/snac/cch/cache.c:invalidatePageWithHashImpl/1321: NULL dv on inval page 0x3e3209 flags 0x4 table 15

A multi-page read is outstanding at one director, and for one of the pages (specifically not the last one), we're waiting for access to the page in the cache before we can begin. Now the link goes down, and comes up, and we learn that the other cluster carried on with I/O on the volume in question while the link was down. As part of cleaning up the outstanding I/O on the reading cluster, we make an indexing error, and access the wrong field in an array, leading to the failure of a consistency check.

Resolution:
This issue has been fixed in GeoSynchrony 6.1.x and later

There is no workaround for this issue the only resolution for avoiding encountering this issue again is to upgrade the VPLEX to the latest code release.