CVE-2026-89757

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

mm/mglru: fix and remove redundant unevictable folio handling

sort folio() has a shortcut for moving folios that are no longer evictable but are still sitting on a generation list. However, this shortcut is buggy. It does not follow the PG lru usage convention, and it has a more serious issue.

Unevictable folios are not threaded on lists[LRU UNEVICTABLE], so that folio->lru can be reused to hold folio->mlock count (see the comment in lruvec init()). Hence lruvec add folio() skips the list add() for them, and every other place that turns a folio unevictable initialises mlock count explicitly: lru add() sets it to 0, mlock folio() and mlock new folio() set it to !!folio test mlocked(folio). sort folio() sets nothing, and the lru gen del folio() right above it may have already poisoned folio->lru via list del(), so mlock count ends up aliasing LIST POISON2, which reads as 0x122, i.e. 290. The result is user visible. On munlock, munlock folio() decrements that bogus count, finds it still non-zero and bails out before clearing PG mlocked, so the folio remains unevictable and the Mlocked accounting stays inflated until the folio is freed.

The shortcut also touches the LRU flags in the wrong order. It calls lru gen del folio() while PG lru is still set, so a concurrent folio test clear lru() (e.g. compaction, folio isolate lru()) can succeed on a folio that has already been taken off the generation list, which may lead to unexpected behavior.

So fix it by isolating them as common folios and letting the generic shrink path cull them. This matches the classical LRU behavior, and there should be no visible effect on the generic eviction or isolation behavior.

There is no performance concern either, such a folio goes through this once, and then it is off the generation lists for good.