CVE-2026-99728: mm/huge_memory: fix list corruption when deferred_split_scan() races with deferred_split_folio()
In the Linux kernel, the following vulnerability has been resolved:
mm/hugememory: fix list corruption when deferredsplitscan() races with deferredsplitfolio()
Since the deferred split queue became a listlru, the isolate callback moves the folios it took a reference on to an on-stack list, linked through folio->deferredlist, and deferredsplitscan() takes them off that list with listdelinit() without holding the listlru lock.
deferredsplitfolio() does not know about the on-stack list. It only holds the listlru lock when it looks at the same listhead, and it can run for such a folio: a THP is queued when it is mapped, as a candidate for the underused check, and is queued "again" when a part of it gets unmapped:
CPU0 deferredsplitscan() CPU1 deferredsplitfolio() listdelinit(&folio->deferredlist) next->prev = prev prev->next = next entry->next = entry listlrulockirqsave() foliosetpartiallymapped() listlruadd() listempty(item) -> true listaddtail(item, &l->list) item->next = &l->list item->prev = tail entry->prev = entry
The folio now sits on the listlru with ->prev pointing to itself. The next removal leaves the neighbours pointing to it, and they still do so after the folio has been freed. With CONFIGDEBUGLIST:
listdel corruption. prev->next should be ffd4000000270090, but was ffd4000000278090. (prev=ffd4000000270090)
Fix it by grabbing the lock around listdelinit() so deferredsplitfolio() cannot race with it.
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
linux/Kernelto a version that resolves this vulnerability.Fixed in 7.3.5 - Compensating control
In deferred_split_folio(), hold the list_lru lock around list_del_init(&folio->_deferred_list) so it cannot race with deferred_split_scan(); use list_lru_lock_irqsave().
Frequently Asked Questions
Which systems are exposed to this race?
Systems are exposed when a transparent huge page is placed on the deferred split queue while mapped as an underused-check candidate and is queued again after part of it is unmapped. The issue requires deferred_split_scan() and deferred_split_folio() to operate concurrently on the same folio.
What is the resulting corruption?
The folio can be added to the list_lru with its _deferred_list prev pointer referring to itself. A later removal can then leave neighboring list entries pointing at that folio, corrupting the deferred split list.
Is a fixed kernel version or temporary mitigation provided?
No fixed version, configuration workaround, or temporary mitigation is included in the provided information. The issue is described as resolved in the Linux kernel.