[PATCH 1/2] x86/mm/pat: Change untrack_pfn() to handle unmapped vma

[Toshi Kani]

mremap() with MREMAP_FIXED, remapping to a new virtual address, on
a VM_PFNMAP range causes the following WARN_ON_ONCE() message in
untrack_pfn().

  WARNING: CPU: 1 PID: 3493 at arch/x86/mm/pat.c:985 untrack_pfn+0xbd/0xd0()
  Call Trace:
  [<ffffffff817729ea>] dump_stack+0x45/0x57
  [<ffffffff8109e4b6>] warn_slowpath_common+0x86/0xc0
  [<ffffffff8109e5ea>] warn_slowpath_null+0x1a/0x20
  [<ffffffff8106a88d>] untrack_pfn+0xbd/0xd0
  [<ffffffff811d2d5e>] unmap_single_vma+0x80e/0x860
  [<ffffffff811d3725>] unmap_vmas+0x55/0xb0
  [<ffffffff811d916c>] unmap_region+0xac/0x120
  [<ffffffff811db86a>] do_munmap+0x28a/0x460
  [<ffffffff811dec33>] move_vma+0x1b3/0x2e0
  [<ffffffff811df113>] SyS_mremap+0x3b3/0x510
  [<ffffffff817793ee>] entry_SYSCALL_64_fastpath+0x12/0x71

MREMAP_FIXED moves a virtual address of VM_PFNMAP, but keeps the pfn
and cache type.  In this case, untrack_pfn() is called with the old
vma after its translation has removed.  Hence, when follow_phys()
fails, track_pfn() is changed to keep the pfn tracked and clears
VM_PAT from the old vma, instead of WARN_ON_ONCE() on the case.

Reference: https://lkml.org/lkml/2015/10/28/865
Reported-by: Stas Sergeev <stsp@xxxxxxx>
Signed-off-by: Toshi Kani <toshi.kani@xxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: H. Peter Anvin <hpa@xxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxx>
---
 arch/x86/mm/pat.c |   17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c
index 188e3e0..f3e391e 100644
--- a/arch/x86/mm/pat.c
+++ b/arch/x86/mm/pat.c
@@ -966,8 +966,14 @@ int track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot,
 
 /*
  * untrack_pfn is called while unmapping a pfnmap for a region.
- * untrack can be called for a specific region indicated by pfn and size or
- * can be for the entire vma (in which case pfn, size are zero).
+ * untrack_pfn can be called for a specific region indicated by pfn and
+ * size or can be for the entire vma (in which case pfn, size are zero).
+ *
+ * NOTE: mremap may move a virtual address of VM_PFNMAP, but keeps the
+ * pfn and cache type.  In this case, untrack_pfn() is called with the
+ * old vma after its translation has removed.  Hence, when follow_phys()
+ * fails, track_pfn() keeps the pfn tracked and clears VM_PAT from the
+ * old vma.
  */
 void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn,
 		 unsigned long size)
@@ -981,14 +987,13 @@ void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn,
 	/* free the chunk starting from pfn or the whole chunk */
 	paddr = (resource_size_t)pfn << PAGE_SHIFT;
 	if (!paddr && !size) {
-		if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr)) {
-			WARN_ON_ONCE(1);
-			return;
-		}
+		if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr))
+			goto out;
 
 		size = vma->vm_end - vma->vm_start;
 	}
 	free_pfn_range(paddr, size);
+out:
 	vma->vm_flags &= ~VM_PAT;
 }
 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/