block: fix get_max_segment_size() overflow on 32bit arch
authorMing Lei <ming.lei@redhat.com>
Sat, 11 Jan 2020 12:57:43 +0000 (20:57 +0800)
committerJens Axboe <axboe@kernel.dk>
Tue, 14 Jan 2020 20:37:40 +0000 (13:37 -0700)
commit4a2f704eb2d831a2d73d7f4cdd54f45c49c3c353
tree8ff8c6a1e1c0311dba3b81a9482dc5ea87ed456f
parente17016f6dcb047f91a8fc9f46bbf81a21d15ca73
block: fix get_max_segment_size() overflow on 32bit arch

Commit 429120f3df2d starts to take account of segment's start dma address
when computing max segment size, and data type of 'unsigned long'
is used to do that. However, the segment mask may be 0xffffffff, so
the figured out segment size may be overflowed in case of zero physical
address on 32bit arch.

Fix the issue by returning queue_max_segment_size() directly when that
happens.

Fixes: 429120f3df2d ("block: fix splitting segments on boundary masks")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: Christoph Hellwig <hch@lst.de>
Tested-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
block/blk-merge.c