xen/balloon: Fix crash when ballooning on x86 32 bit PAE

Commit 55b3da9 (xen/balloon: find
non-conflicting regions to place hotplugged memory) caused a
regression in 4.4.

When ballooning on an x86 32 bit PAE system with close to 64 GiB of
memory, the address returned by allocate_resource may be above 64 GiB.
When using CONFIG_SPARSEMEM, this setup is limited to using physical
addresses < 64 GiB.  When adding memory at this address, it runs off
the end of the mem_section array and causes a crash.  Instead, fail
the ballooning request.

Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Cc: <stable@vger.kernel.org> # 4.4+
Signed-off-by: David Vrabel <david.vrabel@citrix.com>

drivers/xen/balloon.c

@@ -151,6 +151,8 @@ static DECLARE_WAIT_QUEUE_HEAD(balloon_wq);
 static void balloon_process(struct work_struct *work);
 static DECLARE_DELAYED_WORK(balloon_worker, balloon_process);
 
+static void release_memory_resource(struct resource *resource);
+
 /* When ballooning out (allocating memory to return to Xen) we don't really
    want the kernel to try too hard since that can trigger the oom killer. */
 #define GFP_BALLOON \
@@ -267,6 +269,20 @@ static struct resource *additional_memory_resource(phys_addr_t size)
 		return NULL;
 	}
 
+#ifdef CONFIG_SPARSEMEM
+	{
+		unsigned long limit = 1UL << (MAX_PHYSMEM_BITS - PAGE_SHIFT);
+		unsigned long pfn = res->start >> PAGE_SHIFT;
+
+		if (pfn > limit) {
+			pr_err("New System RAM resource outside addressable RAM (%lu > %lu)\n",
+			       pfn, limit);
+			release_memory_resource(res);
+			return NULL;
+		}
+	}
+#endif
+
 	return res;
 }