NVMe: check for integer overflow in nvme_map_user_pages()
authorDan Carpenter <dan.carpenter@oracle.com>
Mon, 13 May 2013 14:59:50 +0000 (17:59 +0300)
committerMatthew Wilcox <matthew.r.wilcox@intel.com>
Fri, 17 May 2013 13:11:03 +0000 (09:11 -0400)
commit5460fc03105fbed01fe27aa572d9f65bb410a61d
tree0e4e62a7fe314ceaa89a7966ef7c2d770686d361
parent5be37bf9c17ffad0590a4044dbb110fe08066923
NVMe: check for integer overflow in nvme_map_user_pages()

You need to have CAP_SYS_ADMIN to trigger this overflow but it makes the
static checkers complain so we should fix it.  The worry is that
"length" comes from copy_from_user() so we need to check that "length +
offset" can't overflow.

I also changed the min_t() cast to be unsigned instead of signed.  Now
that we cap "length" to INT_MAX it doesn't make a difference, but it's a
little easier for reviewers to know that large values aren't cast to
negative.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>
drivers/block/nvme-core.c