arm64: mm: always take dirty state from new pte in ptep_set_access_flags

Commit 66dbd6e61a ("arm64: Implement ptep_set_access_flags() for
hardware AF/DBM") ensured that pte flags are updated atomically in the
face of potential concurrent, hardware-assisted updates. However, Alex
reports that:

 | This patch breaks swapping for me.
 | In the broken case, you'll see either systemd cpu time spike (because
 | it's stuck in a page fault loop) or the system hang (because the
 | application owning the screen is stuck in a page fault loop).

It turns out that this is because the 'dirty' argument to
ptep_set_access_flags is always 0 for read faults, and so we can't use
it to set PTE_RDONLY. The failing sequence is:

  1. We put down a PTE_WRITE | PTE_DIRTY | PTE_AF pte
  2. Memory pressure -> pte_mkold(pte) -> clear PTE_AF
  3. A read faults due to the missing access flag
  4. ptep_set_access_flags is called with dirty = 0, due to the read fault
  5. pte is then made PTE_WRITE | PTE_DIRTY | PTE_AF | PTE_RDONLY (!)
  6. A write faults, but pte_write is true so we get stuck

The solution is to check the new page table entry (as would be done by
the generic, non-atomic definition of ptep_set_access_flags that just
calls set_pte_at) to establish the dirty state.

Cc: <stable@vger.kernel.org> # 4.3+
Fixes: 66dbd6e61a ("arm64: Implement ptep_set_access_flags() for hardware AF/DBM")
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reported-by: Alexander Graf <agraf@suse.de>
Tested-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Will Deacon <will.deacon@arm.com>

arch/arm64/mm/fault.c

@@ -109,7 +109,7 @@ int ptep_set_access_flags(struct vm_area_struct *vma,
 	 * PTE_RDONLY is cleared by default in the asm below, so set it in
 	 * back if necessary (read-only or clean PTE).
 	 */
-	if (!pte_write(entry) || !dirty)
+	if (!pte_write(entry) || !pte_sw_dirty(entry))
 		pte_val(entry) |= PTE_RDONLY;
 
 	/*