CVE-2018-18281

Desde la versión 3.2 del kernel de Linux, la syscall mremap() realiza vaciados TLB tras soltar bloqueos de tabla de página. Si una syscall como ftruncate() elimina las entradas de las tablas de página de una tarea en medio de mremap(), una entrada TLB obsoleta puede permanecer por poco tiempo, lo que permite el acceso a una página física una vez se ha devuelto al asignador de páginas y se reutiliza. Esto se ha solucionado en las siguientes versiones del kernel: 4.9.135, 4.14.78, 4.18.16 y 4.19.

Since Linux kernel version 3.2, the mremap() syscall performs TLB flushes after dropping pagetable locks. If a syscall such as ftruncate() removes entries from the pagetables of a task that is in the middle of mremap(), a stale TLB entry can remain for a short time that permits access to a physical page after it has been released back to the page allocator and reused.

Jann Horn discovered that the procfs file system implementation in the Linux kernel did not properly restrict the ability to inspect the kernel stack of an arbitrary task. A local attacker could use this to expose sensitive information. Jann Horn discovered that the mremap system call in the Linux kernel did not properly flush the TLB when completing, potentially leaving access to a physical page after it has been released to the page allocator. A local attacker could use this to cause a denial of service , expose sensitive information, or possibly execute arbitrary code. Various other issues were also addressed.