IB/ipath: Convert ipath_user_sdma_pin_pages() to use get_user_pages_fast()
authorJan Kara <jack@suse.cz>
Fri, 4 Oct 2013 13:29:06 +0000 (09:29 -0400)
committerRoland Dreier <roland@purestorage.com>
Fri, 8 Nov 2013 22:43:11 +0000 (14:43 -0800)
commit4adcf7fb6783e354aab38824d803fa8c4f8e8a27
tree6aaa86095a45cf4c845090f025c0a1543d5cbbcb
parent959f58544b7f20c92d5eb43d1232c96c15c01bfb
IB/ipath: Convert ipath_user_sdma_pin_pages() to use get_user_pages_fast()

ipath_user_sdma_queue_pkts() gets called with mmap_sem held for
writing.  Except for get_user_pages() deep down in
ipath_user_sdma_pin_pages() we don't seem to need mmap_sem at all.

Even more interestingly the function ipath_user_sdma_queue_pkts() (and
also ipath_user_sdma_coalesce() called somewhat later) call
copy_from_user() which can hit a page fault and we deadlock on trying
to get mmap_sem when handling that fault.  So just make
ipath_user_sdma_pin_pages() use get_user_pages_fast() and leave
mmap_sem locking for mm.

This deadlock has actually been observed in the wild when the node
is under memory pressure.

Cc: <stable@vger.kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
[ Merged in fix for call to get_user_pages_fast from Tetsuo Handa
  <penguin-kernel@I-love.SAKURA.ne.jp>.  - Roland ]

Signed-off-by: Roland Dreier <roland@purestorage.com>
drivers/infiniband/hw/ipath/ipath_user_sdma.c