[PATCH] hugetlb: Check for anon_vma prior to folio allocation

[Vishal Moola (Oracle)]

Commit 9acad7ba3e25 ("hugetlb: use vmf_anon_prepare() instead of
anon_vma_prepare()") may bailout after allocating a folio if we do not
hold the mmap lock. When this occurs, vmf_anon_prepare() will release the
vma lock. Hugetlb then attempts to call restore_reserve_on_error(),
which depends on the vma lock being held.

We can move vmf_anon_prepare() prior to the folio allocation in order to
avoid calling restore_reserve_on_error() without the vma lock.

Fixes: 9acad7ba3e25 ("hugetlb: use vmf_anon_prepare() instead of anon_vma_prepare()")
Reported-by: syzbot+ad1b592fc4483655438b@xxxxxxxxxxxxxxxxxxxxxxxxx
Signed-off-by: Vishal Moola (Oracle) <vishal.moola@xxxxxxxxx>
---
 mm/hugetlb.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index f826dc681081..fbd278a2e9f6 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -6271,6 +6271,10 @@ static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
 							VM_UFFD_MISSING);
 		}
 
+		ret = vmf_anon_prepare(vmf);
+		if (unlikely(ret))
+			goto out;
+
 		folio = alloc_hugetlb_folio(vma, vmf->address, 0);
 		if (IS_ERR(folio)) {
 			/*
@@ -6310,15 +6314,12 @@ static vm_fault_t hugetlb_no_page(struct mm_struct *mm,
 				restore_reserve_on_error(h, vma, vmf->address,
 							folio);
 				folio_put(folio);
+				ret = VM_FAULT_SIGBUS;
 				goto out;
 			}
 			new_pagecache_folio = true;
 		} else {
 			folio_lock(folio);
-
-			ret = vmf_anon_prepare(vmf);
-			if (unlikely(ret))
-				goto backout_unlocked;
 			anon_rmap = 1;
 		}
 	} else {
-- 
2.43.0


--nyCCuuGrGXDMVrWo--