nday
CVE-2026-97945

x86/mm: Fix user-space data loss with MADV_FREE and THP

Linux / Linux

Disclosed . nday has no fix record for it yet; that does not mean no patch exists. None of the four PoC sources lists it. Not in CISA KEV.

tags productLinux kernel product typeoperating system
why these tags

nday derives these tags by fixed rules from the facts below; NVD, the CNA and other sources do not assign them. Vocabulary and rules: ~/tags.

Linux kernel
affected products in the CVE record: Linux · Linux
operating system
affected products in the CVE record: Linux · Linuxthe product type of Linux kernel in nday's product registry
stage
DIS
disclosed
cvss
no score yet

Lifecycle

x axis UTC; dot = exact minute, wide bar = date only, hollow = unknown or inferred.
time unknown
DISdisclosed
09-25T10:22Z
PATpatch
no record yet
ANAanalysis
no record yet
POCPoC
no record yet
ITWexploited
no record yet
KEVKEV
no record yet

Summary

In the Linux kernel, the following vulnerability has been resolved:

x86/mm: Fix user-space data loss with MADV_FREE and THP

Some of users of Polars (a data analytics library) have lost production data from this bug. They seem to have just the right combination of huge pages, MADV_FREE and heavy reclaim pressure.

pmd_modify() masks the old value with (_HPAGE_CHG_MASK & ~_PAGE_DIRTY), silently discarding the hardware dirty bit. The subsequent pmd_mksaveddirty() call is supposed to transfer _PAGE_DIRTY into _PAGE_SAVED_DIRTY when write-protecting, but the dirty bit was already stripped from the value, so there is nothing left to transfer.

Contrast with pte_modify(), which keeps _PAGE_DIRTY_BITS in its mask, and pud_modify(), which keeps _HPAGE_CHG_MASK untouched: pmd_modify() is the odd one out. Any pmd_modify() on a writable, dirty PMD loses the dirty state.

One visible consequence is data loss with MADV_FREE on PMD-mapped THP:

memset(buf, 0x5A, size);          // PMD-mapped THP, PMD dirty
madvise(buf, size, MADV_FREE);    // PMD cleaned but left writable,
                                  // folio marked lazyfree
memset(buf, 0x5A, size);          // hardware sets _PAGE_DIRTY again
mprotect(buf, size, PROT_READ);   // pmd_modify() drops the dirty bit
mprotect(buf, size, PROT_READ|PROT_WRITE);
// ... memory pressure ...

Reclaim (e.g. under memcg pressure) then finds the lazyfree folio with no dirty bit set anywhere and frees it in __discard_anon_folio_pmd_locked(), even though the data was rewritten after MADV_FREE; subsequent reads fault in fresh zero pages. NUMA hinting alone can trigger the same loss, as do_huge_pmd_numa_page() restores the PMD through pmd_modify() as well.

PMD-mapped file THPs are affected too: mprotect()/NUMA hinting dropping the dirty bit means rewritten data is never written back.

Fix it by keeping _PAGE_DIRTY in the preserved mask, exactly like pte_modify() and pud_modify() do. The existing pmd_mksaveddirty()/pmd_clear_saveddirty() pair then performs the hardware-dirty <-> saved-dirty transition based on the write bit, preserving the shadow-stack encoding rules.

Original text of the CVE record (cvelist), shown as published.

Event log

+DISdisclosed

CVE record published by Linux.

src cvelist · nday , bulk import · raw
{
  "id": 457247,
  "type": "disclosed",
  "source": "cvelist",
  "url": "https://www.cve.org/CVERecord?id=CVE-2026-97945",
  "occurred_at": "2026-09-25T10:22:55.981Z",
  "observed_at": "2026-11-13T05:40:20.493151Z",
  "is_backfill": true,
  "dedupe_key": "",
  "detail": {
    "state": "PUBLISHED",
    "date_published": "2026-09-25T10:22:55.981Z",
    "assigner_short_name": "Linux"
  }
}

Report an error

report an error or omission

Something on this page wrong or missing? Tell the editors; if they accept it, this page's revision log says so publicly. Upstream data is not changed.

Researcher discussion

0 researchers mentioned this in the last 7 days

Posts on X by researchers on nday's list that mention this CVE, shown as posted. nday does not check them.

nday has not collected any post on X by listed researchers that mentions this CVE.

Discussion

0 comments

No comments yet.