kernel/fork: only call untrack_pfn_clear() on VMAs duplicated for fork()

Not intuitive, but vm_area_dup() located in kernel/fork.c is not only used
for duplicating VMAs during fork(), but also for duplicating VMAs when
splitting VMAs or when mremap()'ing them.

VM_PFNMAP mappings can at least get ordinarily mremap()'ed (no change in
size) and apparently also shrunk during mremap(), which implies
duplicating the VMA in __split_vma() first.

In case of ordinary mremap() (no change in size), we first duplicate the
VMA in copy_vma_and_data()->copy_vma() to then call untrack_pfn_clear() on
the old VMA: we effectively move the VM_PAT reservation.  So the
untrack_pfn_clear() call on the new VMA duplicating is wrong in that
context.

Splitting of VMAs seems problematic, because we don't duplicate/adjust the
reservation when splitting the VMA.  Instead, in memtype_erase() -- called
during zapping/munmap -- we shrink a reservation in case only the end
address matches: Assume we split a VMA into A and B, both would share a
reservation until B is unmapped.

So when unmapping B, the reservation would be updated to cover only A. 
When unmapping A, we would properly remove the now-shrunk reservation. 
That scenario describes the mremap() shrinking (old_size > new_size),
where we split + unmap B, and the untrack_pfn_clear() on the new VMA when
is wrong.

What if we manage to split a VM_PFNMAP VMA into A and B and unmap A first?
It would be broken because we would never free the reservation.  Likely,
there are ways to trigger such a VMA split outside of mremap().

Affecting other VMA duplication was not intended, vm_area_dup() being used
outside of kernel/fork.c was an oversight.  So let's fix that for; how to
handle VMA splits better should be investigated separately.


With a simple reproducer that uses mprotect() to split such a VMA I can
trigger

x86/PAT: pat_mremap:26448 freeing invalid memtype [mem 0x00000000-0x00000fff]

Link: https://lkml.kernel.org/r/20250422144942.2871395-1-david@redhat.com
Fixes: dc84bc2 ("x86/mm/pat: Fix VM_PAT handling when fork() fails in copy_page_range()")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Rik van Riel <riel@surriel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

kernel/fork.c

@@ -498,10 +498,6 @@ struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
 	vma_numab_state_init(new);
 	dup_anon_vma_name(orig, new);
 
-	/* track_pfn_copy() will later take care of copying internal state. */
-	if (unlikely(new->vm_flags & VM_PFNMAP))
-		untrack_pfn_clear(new);
-
 	return new;
 }
 
@@ -672,6 +668,11 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
 		tmp = vm_area_dup(mpnt);
 		if (!tmp)
 			goto fail_nomem;
+
+		/* track_pfn_copy() will later take care of copying internal state. */
+		if (unlikely(tmp->vm_flags & VM_PFNMAP))
+			untrack_pfn_clear(tmp);
+
 		retval = vma_dup_policy(mpnt, tmp);
 		if (retval)
 			goto fail_nomem_policy;