memory-device: turn alignment assert into check
authorDavid Hildenbrand <david@redhat.com>
Thu, 7 Jun 2018 15:47:04 +0000 (17:47 +0200)
committerPaolo Bonzini <pbonzini@redhat.com>
Thu, 28 Jun 2018 17:05:31 +0000 (19:05 +0200)
The start of the address space indicates which maximum alignment is
supported by our machine (e.g. ppc, x86 1GB). This is helpful to
catch fragmenting guest physical memory in strange fashions.

Right now we can crash QEMU by e.g. (there might be easier examples)

qemu-system-x86_64 -m 256M,maxmem=20G,slots=2 \
 -object memory-backend-file,id=mem0,size=8192M,mem-path=/dev/zero,align=8192M \
 -device pc-dimm,id=dimm1,memdev=mem0

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180607154705.6316-2-david@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
hw/mem/memory-device.c

index 3e04f3954e61b022f95eaeb39c2a804cbfef5881..6de4f70bb4bfc340d779e5e83adb75fc0b634d3f 100644 (file)
@@ -116,9 +116,15 @@ uint64_t memory_device_get_free_addr(MachineState *ms, const uint64_t *hint,
     address_space_start = ms->device_memory->base;
     address_space_end = address_space_start +
                         memory_region_size(&ms->device_memory->mr);
-    g_assert(QEMU_ALIGN_UP(address_space_start, align) == address_space_start);
     g_assert(address_space_end >= address_space_start);
 
+    /* address_space_start indicates the maximum alignment we expect */
+    if (QEMU_ALIGN_UP(address_space_start, align) != address_space_start) {
+        error_setg(errp, "the alignment (0%" PRIx64 ") is not supported",
+                   align);
+        return 0;
+    }
+
     memory_device_check_addable(ms, size, errp);
     if (*errp) {
         return 0;