[PATCH] dax: fix potential overflow on 32bit machine

From: "zhangyi (F)" <yi.zhang@huawei.com>
To: <linux-fsdevel@vger.kernel.org>
Cc: <mawilcox@microsoft.com>, <ross.zwisler@linux.intel.com>,
	<viro@zeniv.linux.org.uk>, <yi.zhang@huawei.com>,
	<miaoxie@huawei.com>
Subject: [PATCH] dax: fix potential overflow on 32bit machine
Date: Tue, 5 Dec 2017 11:32:10 +0800	[thread overview]
Message-ID: <20171205033210.38338-1-yi.zhang@huawei.com> (raw)

On 32bit machine, when mmap2 a large enough file with pgoff more than
ULONG_MAX >> PAGE_SHIFT, it will trigger offset overflow and lead to
unmap the wrong page in dax_insert_mapping_entry(). This patch cast
pgoff to 64bit to prevent the overflow.

Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
---
 fs/dax.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/dax.c b/fs/dax.c
index 78b72c4..8e12848 100644
--- a/fs/dax.c
+++ b/fs/dax.c
@@ -539,10 +539,11 @@ static void *dax_insert_mapping_entry(struct address_space *mapping,
 		/* we are replacing a zero page with block mapping */
 		if (dax_is_pmd_entry(entry))
 			unmap_mapping_range(mapping,
-					(vmf->pgoff << PAGE_SHIFT) & PMD_MASK,
+					((loff_t)vmf->pgoff << PAGE_SHIFT) & PMD_MASK,
 					PMD_SIZE, 0);
 		else /* pte entry */
-			unmap_mapping_range(mapping, vmf->pgoff << PAGE_SHIFT,
+			unmap_mapping_range(mapping,
+					(loff_t)vmf->pgoff << PAGE_SHIFT,
 					PAGE_SIZE, 0);
 	}
 
-- 
2.9.5
