When using the Index Replication feature, Apache Solr nodes can pull index
files from a master/leader node using an HTTP API which accepts a file
name. However, Solr before 5.5.4 and 6.x before 6.4.1 did not validate the
file name, hence it was possible to craft a special request involving path
traversal, leaving any file readable to the Solr server process exposed.
Solr servers protected and restricted by firewall rules and/or
authentication would not be at risk since only trusted clients and users
would gain direct HTTP access.

Hrishikesh Gadre discovered that Apache Solr did not validate the file name,
hence an attacker could possible to craft a special request involving path
traversal that could made it crash and provoke a denial of service.