KVM uses get_user_pages() to resolve its stage2 faults.  KVM sets the
FOLL_HWPOISON flag causing faultin_page() to return -EHWPOISON when it
finds a VM_FAULT_HWPOISON.  KVM handles these hwpoison pages as a
special case.  (check_user_page_hwpoison())
When huge pages are involved, this doesn't work so well.
get_user_pages() calls follow_hugetlb_page(), which stops early if it
receives VM_FAULT_HWPOISON from hugetlb_fault(), eventually returning
-EFAULT to the caller.  The step to map this to -EHWPOISON based on the
FOLL_ flags is missing.  The hwpoison special case is skipped, and
-EFAULT is returned to user-space, causing Qemu or kvmtool to exit.
Instead, move this VM_FAULT_ to errno mapping code into a header file
and use it from faultin_page() and follow_hugetlb_page().
With this, KVM works as expected.
This isn't a problem for arm64 today as we haven't enabled
MEMORY_FAILURE, but I can't see any reason this doesn't happen on x86
too, so I think this should be a fix.  This doesn't apply earlier than
stable's v4.11.1 due to all sorts of cleanup.
[james.morse@arm.com: add vm_fault_to_errno() call to faultin_page()]
suggested.
Link: http://lkml.kernel.org/r/20170525171035.16359-1-james.morse@arm.com
[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20170524160900.28786-1-james.morse@arm.com
Signed-off-by: James Morse <james.morse@arm.com>
Acked-by: Punit Agrawal <punit.agrawal@arm.com>
Acked-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: <stable@vger.kernel.org>	[4.11.1+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
 
 #define FOLL_REMOTE    0x2000  /* we are working on non-current tsk/mm */
 #define FOLL_COW       0x4000  /* internal GUP flag */
 
+static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
+{
+       if (vm_fault & VM_FAULT_OOM)
+               return -ENOMEM;
+       if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
+               return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
+       if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
+               return -EFAULT;
+       return 0;
+}
+
 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
                        void *data);
 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
 
 
        ret = handle_mm_fault(vma, address, fault_flags);
        if (ret & VM_FAULT_ERROR) {
-               if (ret & VM_FAULT_OOM)
-                       return -ENOMEM;
-               if (ret & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
-                       return *flags & FOLL_HWPOISON ? -EHWPOISON : -EFAULT;
-               if (ret & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
-                       return -EFAULT;
+               int err = vm_fault_to_errno(ret, *flags);
+
+               if (err)
+                       return err;
                BUG();
        }
 
        ret = handle_mm_fault(vma, address, fault_flags);
        major |= ret & VM_FAULT_MAJOR;
        if (ret & VM_FAULT_ERROR) {
-               if (ret & VM_FAULT_OOM)
-                       return -ENOMEM;
-               if (ret & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
-                       return -EHWPOISON;
-               if (ret & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
-                       return -EFAULT;
+               int err = vm_fault_to_errno(ret, 0);
+
+               if (err)
+                       return err;
                BUG();
        }
 
 
                        }
                        ret = hugetlb_fault(mm, vma, vaddr, fault_flags);
                        if (ret & VM_FAULT_ERROR) {
+                               int err = vm_fault_to_errno(ret, flags);
+
+                               if (err)
+                                       return err;
+
                                remainder = 0;
                                break;
                        }