exec: allocate PROT_NONE pages on top of RAM

This inserts a read and write protected page between RAM and QEMU
memory, for file-backend RAM.
This makes it harder to exploit QEMU bugs resulting from buffer
overflows in devices using variants of cpu_physical_memory_map,
dma_memory_map etc.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>

exec.c

@@ -84,6 +84,9 @@ static MemoryRegion io_mem_unassigned;
  */
 #define RAM_RESIZEABLE (1 << 2)
 
+/* An extra page is mapped on top of this RAM.
+ */
+#define RAM_EXTRA (1 << 3)
 #endif
 
 struct CPUTailQ cpus = QTAILQ_HEAD_INITIALIZER(cpus);
@@ -1185,10 +1188,13 @@ static void *file_ram_alloc(RAMBlock *block,
     char *filename;
     char *sanitized_name;
     char *c;
+    void *ptr;
     void *area = NULL;
     int fd;
     uint64_t hpagesize;
+    uint64_t total;
     Error *local_err = NULL;
+    size_t offset;
 
     hpagesize = gethugepagesize(path, &local_err);
     if (local_err) {
@@ -1232,6 +1238,7 @@ static void *file_ram_alloc(RAMBlock *block,
     g_free(filename);
 
     memory = ROUND_UP(memory, hpagesize);
+    total = memory + hpagesize;
 
     /*
      * ftruncate is not supported by hugetlbfs in older
@@ -1243,16 +1250,40 @@ static void *file_ram_alloc(RAMBlock *block,
         perror("ftruncate");
     }
 
-    area = mmap(0, memory, PROT_READ | PROT_WRITE,
+    ptr = mmap(0, total, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS,
-                (block->flags & RAM_SHARED ? MAP_SHARED : MAP_PRIVATE),
+                -1, 0);
+    if (ptr == MAP_FAILED) {
+        error_setg_errno(errp, errno,
+                         "unable to allocate memory range for hugepages");
+        close(fd);
+        goto error;
+    }
+
+    offset = QEMU_ALIGN_UP((uintptr_t)ptr, hpagesize) - (uintptr_t)ptr;
+
+    area = mmap(ptr + offset, memory, PROT_READ | PROT_WRITE,
+                (block->flags & RAM_SHARED ? MAP_SHARED : MAP_PRIVATE) |
+                MAP_FIXED,
                 fd, 0);
     if (area == MAP_FAILED) {
         error_setg_errno(errp, errno,
                          "unable to map backing store for hugepages");
+        munmap(ptr, total);
         close(fd);
         goto error;
     }
 
+    if (offset > 0) {
+        munmap(ptr, offset);
+    }
+    ptr += offset;
+    total -= offset;
+
+    if (total > memory + getpagesize()) {
+        munmap(ptr + memory + getpagesize(),
+               total - memory - getpagesize());
+    }
+
     if (mem_prealloc) {
         os_mem_prealloc(fd, area, memory);
     }
@@ -1570,6 +1601,7 @@ ram_addr_t qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
     new_block->used_length = size;
     new_block->max_length = size;
     new_block->flags = share ? RAM_SHARED : 0;
+    new_block->flags |= RAM_EXTRA;
     new_block->host = file_ram_alloc(new_block, size,
                                      mem_path, errp);
     if (!new_block->host) {
@@ -1671,7 +1703,11 @@ static void reclaim_ramblock(RAMBlock *block)
         xen_invalidate_map_cache_entry(block->host);
 #ifndef _WIN32
     } else if (block->fd >= 0) {
-        munmap(block->host, block->max_length);
+        if (block->flags & RAM_EXTRA) {
+            munmap(block->host, block->max_length + getpagesize());
+        } else {
+            munmap(block->host, block->max_length);
+        }
         close(block->fd);
 #endif
     } else {