nvme: Fix u32 overflow in the number of namespace list calculation
authorJaesoo Lee <jalee@purestorage.com>
Mon, 3 Jun 2019 23:42:28 +0000 (16:42 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 10 Jul 2019 07:55:32 +0000 (09:55 +0200)
[ Upstream commit c8e8c77b3bdbade6e26e8e76595f141ede12b692 ]

The Number of Namespaces (nn) field in the identify controller data structure is
defined as u32 and the maximum allowed value in NVMe specification is
0xFFFFFFFEUL. This change fixes the possible overflow of the DIV_ROUND_UP()
operation used in nvme_scan_ns_list() by casting the nn to u64.

Signed-off-by: Jaesoo Lee <jalee@purestorage.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/nvme/host/core.c

index 979c6ecc6446509876e305213052356457201e5a..8705bfe7bb73162c15a3feedabc79ca68d1102a2 100644 (file)
@@ -1765,7 +1765,8 @@ static int nvme_scan_ns_list(struct nvme_ctrl *ctrl, unsigned nn)
 {
        struct nvme_ns *ns;
        __le32 *ns_list;
-       unsigned i, j, nsid, prev = 0, num_lists = DIV_ROUND_UP(nn, 1024);
+       unsigned i, j, nsid, prev = 0;
+       unsigned num_lists = DIV_ROUND_UP_ULL((u64)nn, 1024);
        int ret = 0;
 
        ns_list = kzalloc(0x1000, GFP_KERNEL);