NFSv4.1: Remove a bogus BUG_ON() in nfs4_layoutreturn_done
authorTrond Myklebust <Trond.Myklebust@netapp.com>
Wed, 8 Aug 2012 20:03:13 +0000 (16:03 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 14 Sep 2012 17:00:38 +0000 (10:00 -0700)
commit4ad55ffb366e07fdaefc2372e77b98569c9b4619
tree136a867565bd47305149821614572e951b38bcaf
parent9516c03e553f8947ab63592a42e0a6b381a690bf
NFSv4.1: Remove a bogus BUG_ON() in nfs4_layoutreturn_done

commit 47fbf7976e0b7d9dcdd799e2a1baba19064d9631 upstream.

Ever since commit 0a57cdac3f (NFSv4.1 send layoutreturn to fence
disconnected data server) we've been sending layoutreturn calls
while there is potentially still outstanding I/O to the data
servers. The reason we do this is to avoid races between replayed
writes to the MDS and the original writes to the DS.

When this happens, the BUG_ON() in nfs4_layoutreturn_done can
be triggered because it assumes that we would never call
layoutreturn without knowing that all I/O to the DS is
finished. The fix is to remove the BUG_ON() now that the
assumptions behind the test are obsolete.

Reported-by: Boaz Harrosh <bharrosh@panasas.com>
Reported-by: Tigran Mkrtchyan <tigran.mkrtchyan@desy.de>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/nfs/nfs4proc.c