CVE-2026-99705: mm: memory: fix truncation removing mapped folios
In the Linux kernel, the following vulnerability has been resolved:
mm: memory: fix truncation removing mapped folios
Ayush reported an issue that shmem/tmpfs folios are still mapped when truncating them on production: " BUG: Bad page cache in process ... pfn:... page dumped because: still mapped when deleted "
I reproduced the issue with Ayush's reproducer [1] with the latest mm-unstable branch, and I got the following crash info. From the dump message, we can see that truncateinodefolio() is really trying to remove mapped folios, which is incorrect.
" [ 189.171977] page: refcount:3 mapcount:1 mapping:00000000fe168ee1 index:0x3880 pfn:0x191647 [ 189.171995] memcg:ffff0000cc073d40 [ 189.171997] aops:shmemaops ino:3c01 dentry name(?):"memfd:runsc-memory" [ 189.172006] flags: 0x17fffef0002022d(locked|referenced|uptodate|lru| workingset|swapbacked|node=0|zone=2|lastcpupid=0x3ffff) ....... [ 189.172017] page dumped because: VMBUGONFOLIO(foliomapped(folio)) [ 189.172026] ------------[ cut here ]------------ [ 189.172027] kernel BUG at mm/filemap.c:155! [ 189.172057] Internal error: Oops - BUG: 00000000f2000800 [#1] SMP ....... [ 189.177721] CPU: 3 UID: 0 PID: 5521 Comm: repro Kdump: loaded Tainted: 7.3.0-rc4+ #277 PREEMPT(lazy) ....... [ 189.183986] Call trace: [ 189.184124] filemapunaccountfolio+0xf0/0x1e8 (P) [ 189.184391] filemapremovefolio+0x34/0x160 [ 189.184633] filemapremovefolio+0x4c/0xb0 [ 189.184859] truncateinodefolio+0x34/0x58 [ 189.185087] shmemundorange+0x220/0x658 [ 189.185308] shmemfallocate+0x2f0/0x470 [ 189.185527] vfsfallocate+0x128/0x328 ...... [ 189.187078] el0t64sync+0x184/0x188 "
After analysis, I believe the race exists between truncation and MADVDONTNEED, and shmem's ->mappages() merely makes the issue easier to reproduce. Since MADVDONTNEED synchronously releases the pagetable page before calling tlbflushrmaps(), this could cause another thread's truncation to skip zappterange() (due to pmd is none) but still observe the folio's mapcount as non-zero. A possible race scenario is as follows:
CPU 0 CPU 1 shmemfallocate madvisedontneedsinglevma unmapmappingrange ....... ...... (filemapmappages() remap folios) zappterange truncateinodefolio zapemptyptetable(pmd clear) unmapmappingfolio ...... zappmdrange (saw pmd none, skip)
filemapremovefolio BUGON(foliomapped) tlbflushrmaps(remove rmaps)
To fix this issue, we should move the pmd clear operation (via zapemptyptetable()) to after tlbflushrmaps() and use syncwithfoliopmdzap() to sync with concurrent zapping, so that when we find a pmdnone() while unmapping a folio without holding the PTL, the folio's mappings are guaranteed to have been removed via tlbflushrmaps().
Meanwhile move the second pmdnone() check to where it belongs: if splitting zapped the PMD.
Affected Software
Frequently Asked Questions
Which systems are implicated by the reported failure?
The report identifies shmem/tmpfs folios, including a dumped object associated with a memfd named "runsc-memory." The failure occurs when truncation attempts to remove a folio that is still mapped.
What observable symptom indicates this issue is being triggered?
Affected systems may log "Bad page cache" messages stating that a page is "still mapped when deleted," followed by a VM_BUG_ON_FOLIO(folio_mapped(folio)) condition in truncate_inode_folio() and a kernel BUG/Oops.
Is there a documented workaround if the fix cannot be deployed immediately?
No workaround or mitigation is provided in the available information.