PT-2026-151674

CVE-2026-99463

·

Published

2026-11-19

·

Updated

2026-11-19

Description by

Score
N/A
Not available

No 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:
mm/rmap: don't discard lazyfree THPs at PMD level
discard anon folio pmd locked() clears a lazyfree THP's PMD before it determines whether the folio can be discarded, and restores it if the folio is dirty or has extra references. It is done under only the PMD lock and the anon vma read lock inside try to unmap one() and no mmap lock is held. The temporarily none PMD causes UAF.
A parallel munmap() or mremap() is supposed to serialize with it on the PMD lock, but zap pmd range() (called by munmap()) and move page tables() (called by mremap()) do not take the lock when they see the temporarily none PMD and skip the PMD. There is no more page table walk after the none PMD is restored:
  • munmap(): free pgtables() waits in unlink anon vmas() for the anon vma lock, then free pmd range() drops the restored PMD as bad without removing the rmap, so the folio keeps a stale mapcount.
  • whole-VMA MREMAP DONTUNMAP: dontunmap complete() unlinks the source VMA from its anon vma while the restored PMD still maps the folio.
  • plain mremap(): the THP is not moved but unmapped, so the THP's data is lost.
In the first two cases, the folio stays mapped, but no VMA that maps it is linked to its anon vma. When the anon vma is freed, it is still reachable when one rmap walks the folio. Racing lazyfree reclaim against MREMAP DONTUNMAP as an unprivileged user reproduces a KASAN use-after-free in folio lock anon vma read() as reported in the Closes link.
One solution is to use an invalidated PMD instead of a cleared one until the discard is committed, but it needs arch fixes first.
  1. GUP-fast still follows an invalidated PMD on riscv, LoongArch and s390.
  2. sparc64 miscounts thp pte count when an invalidated PMD is set again or cleared.
To keep the fix simple for backport, stop discarding lazyfree THPs at PMD level instead. try to unmap one() splits the PMD and discards the folio at PTE level.
unmap huge pmd locked() and discard anon folio pmd locked() are left in place, unused, to minimize this fix for backport and the invalidated PMD approach is planned as a follow-up series to use them again.
Vendors
Affected products
Related identifiers
Found an issue in the description? Have something to add? Feel free to write us 👾