PT-2026-151688
CVE-2026-99477
·
Published
2026-11-19
·
Updated
2026-11-19
Description by
Score
N/A
Not availableNo severity ratings or metrics are available. When they are, we'll update the corresponding info on the page.
In the Linux kernel, the following vulnerability has been resolved:
fs/proc/task mmu: pin mm users in do pagemap scan()
The PAGEMAP SCAN ioctl takes mmap read lock() on an mm obtained from
/proc//pagemap without holding a reference to mm users.
proc mem open() deliberately grabs only mm count, so a consumer that
touches the address space has to call mmget not zero() first. Every other
consumer in this file does -- m start(), do procmap query(),
show smaps rollup() and pagemap read() -- but do pagemap scan() does not.
Without that reference the target can reach exit mmap() concurrently, and
exit mmap() calls unmap vmas() while holding mmap lock only for reading,
which does not exclude the scanner. For ordinary VMAs this is harmless,
as free pgtables() and remove vma() run afterwards under mmap write lock.
A shared hugetlb VMA is different: hugetlb zap end() frees the VMA's
hugetlb vma lock inside the read-locked phase.
walk hugetlb range() can therefore load vma->vm private data, have the
lock freed underneath it, and then down read() a freed rw semaphore. The
load happens before the freeing side clears the pointer, so this is a
use-after-free and not a NULL dereference. Any unprivileged user can
reach it with no huge pages configured, as hugetlb reserve pages()
allocates the lock before the VM NORESERVE early return.
Pin mm users across the scan, the way the other consumers do. mmput()
from the exiting task can then no longer reach exit mmap() while a walk is
in progress.
[akpm@linux-foundation.org: remove unneeded & undesirable test for !mm]
Found an issue in the description? Have something to add? Feel free to write us 👾