scsi: fcoe: clear FC_RP_STARTED flags when receiving a LOGO

Johannes Thumshirn · martinkpetersen · commit 1550ec458e0c · 2018-08-02T15:30:03.000-04:00
When receiving a LOGO request we forget to clear the FC_RP_STARTED flag
before starting the rport delete routine.

As the started flag was not cleared, we're not deleting the rport but
waiting for a restart and thus are keeping the reference count of the rdata
object at 1.

This leads to the following kmemleak report:
unreferenced object 0xffff88006542aa00 (size 512):
  comm "kworker/0:2", pid 24, jiffies 4294899222 (age 226.880s)
  hex dump (first 32 bytes):
    68 96 fe 65 00 88 ff ff 00 00 00 00 00 00 00 00  h..e............
    01 00 00 00 08 00 00 00 02 c5 45 24 ac b8 00 10  ..........E$....
  backtrace:
    [&lt;(____ptrval____)&gt;] fcoe_ctlr_vn_add.isra.5+0x7f/0x770 [libfcoe]
    [&lt;(____ptrval____)&gt;] fcoe_ctlr_vn_recv+0x12af/0x27f0 [libfcoe]
    [&lt;(____ptrval____)&gt;] fcoe_ctlr_recv_work+0xd01/0x32f0 [libfcoe]
    [&lt;(____ptrval____)&gt;] process_one_work+0x7ff/0x1420
    [&lt;(____ptrval____)&gt;] worker_thread+0x87/0xef0
    [&lt;(____ptrval____)&gt;] kthread+0x2db/0x390
    [&lt;(____ptrval____)&gt;] ret_from_fork+0x35/0x40
    [&lt;(____ptrval____)&gt;] 0xffffffffffffffff

Signed-off-by: Johannes Thumshirn &lt;jthumshirn@suse.de&gt;
Reported-by: ard &lt;ard@kwaak.net&gt;
Reviewed-by: Hannes Reinecke &lt;hare@suse.com&gt;
Signed-off-by: Martin K. Petersen &lt;martin.petersen@oracle.com&gt;
diff --git a/drivers/scsi/libfc/fc_rport.c b/drivers/scsi/libfc/fc_rport.c
@@ -2164,6 +2164,7 @@ static void fc_rport_recv_logo_req(struct fc_lport *lport, struct fc_frame *fp)
 		FC_RPORT_DBG(rdata, "Received LOGO request while in state %s\n",
 			     fc_rport_state(rdata));
 
+		rdata->flags &= ~FC_RP_STARTED;
 		fc_rport_enter_delete(rdata, RPORT_EV_STOP);
 		mutex_unlock(&rdata->rp_mutex);
 		kref_put(&rdata->kref, fc_rport_destroy);
