xen/blkback: fix disconnect while I/Os in flight
authorJuergen Gross <jgross@suse.com>
Thu, 18 May 2017 15:28:47 +0000 (17:28 +0200)
committerKonrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Tue, 13 Jun 2017 20:04:18 +0000 (16:04 -0400)
commit46464411307746e6297a034a9983a22c9dfc5a0c
tree10a7411cf52e3cbe830d4a48f44cf1eb93085c31
parent2d4456c73a487abe53863e10641c2f73537edf5c
xen/blkback: fix disconnect while I/Os in flight

Today disconnecting xen-blkback is broken in case there are still
I/Os in flight: xen_blkif_disconnect() will bail out early without
releasing all resources in the hope it will be called again when
the last request has terminated. This, however, won't happen as
xen_blkif_free() won't be called on termination of the last running
request: xen_blkif_put() won't decrement the blkif refcnt to 0 as
xen_blkif_disconnect() didn't finish before thus some xen_blkif_put()
calls in xen_blkif_disconnect() didn't happen.

To solve this deadlock xen_blkif_disconnect() and
xen_blkif_alloc_rings() shouldn't use xen_blkif_put() and
xen_blkif_get() but use some other way to do their accounting of
resources.

This at once fixes another error in xen_blkif_disconnect(): when it
returned early with -EBUSY for another ring than 0 it would call
xen_blkif_put() again for already handled rings on a subsequent call.
This will lead to inconsistencies in the refcnt handling.

Cc: stable@vger.kernel.org
Signed-off-by: Juergen Gross <jgross@suse.com>
Tested-by: Steven Haigh <netwiz@crc.id.au>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
drivers/block/xen-blkback/common.h
drivers/block/xen-blkback/xenbus.c