ieee1394: sbp2: prevent rare deadlock in shutdown

Scsi_remove_device() may go into uninterruptible sleep if blocked.
Therefore sbp2_remove() unblocks the Scsi_Host before the device is
requested to be removed.  But there could be another 1394 bus reset
after that which would block the host again.  The 1394 subsystem won't
call sbp2_update() concurrently to sbp2_remove(), which is why there is
no chance for sbp2_remove() to be unblocked by sbp2_update().

The fix is to tell sbp2's bus reset handler when a device is to be shut
down so that it skips scsi_block_requests() on that host.  As before,
any new commands after a reset without reconnect will be failed quickly
by sbp2scsi_queuecommand().

In the long term, means to go without scsi_block_requests() should be
found.

Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>

diff --git a/drivers/ieee1394/sbp2.c b/drivers/ieee1394/sbp2.c
index 3cb6b479b2ef451e97969ed9fc3031fbc53368f1..017259cc34f3a692216f2553e82fae0f8a927ec8 100644 (file)
--- a/drivers/ieee1394/sbp2.c
+++ b/drivers/ieee1394/sbp2.c
@@ -478,7 +478,7 @@ static void sbp2util_notify_fetch_agent(struct scsi_id_instance_data *scsi_id,
         * There is a small window after a bus reset within which the node
         * entry's generation is current but the reconnect wasn't completed.
         */
-       if (atomic_read(&scsi_id->unfinished_reset))
+       if (unlikely(atomic_read(&scsi_id->state) == SBP2LU_STATE_IN_RESET))
                return;
 
        if (hpsb_node_write(scsi_id->ne,
@@ -489,7 +489,7 @@ static void sbp2util_notify_fetch_agent(struct scsi_id_instance_data *scsi_id,
         * Now accept new SCSI commands, unless a bus reset happended during
         * hpsb_node_write.
         */
-       if (!atomic_read(&scsi_id->unfinished_reset))
+       if (likely(atomic_read(&scsi_id->state) != SBP2LU_STATE_IN_RESET))
                scsi_unblock_requests(scsi_id->scsi_host);
 }
 
@@ -756,7 +756,7 @@ static int sbp2_remove(struct device *dev)
                        sbp2scsi_complete_all_commands(scsi_id, DID_NO_CONNECT);
                /* scsi_remove_device() will trigger shutdown functions of SCSI
                 * highlevel drivers which would deadlock if blocked. */
-               atomic_set(&scsi_id->unfinished_reset, 0);
+               atomic_set(&scsi_id->state, SBP2LU_STATE_IN_SHUTDOWN);
                scsi_unblock_requests(scsi_id->scsi_host);
        }
        sdev = scsi_id->sdev;
@@ -811,7 +811,7 @@ static int sbp2_update(struct unit_directory *ud)
        /* Accept new commands unless there was another bus reset in the
         * meantime. */
        if (hpsb_node_entry_valid(scsi_id->ne)) {
-               atomic_set(&scsi_id->unfinished_reset, 0);
+               atomic_set(&scsi_id->state, SBP2LU_STATE_RUNNING);
                scsi_unblock_requests(scsi_id->scsi_host);
        }
        return 0;
@@ -842,7 +842,7 @@ static struct scsi_id_instance_data *sbp2_alloc_device(struct unit_directory *ud
        INIT_LIST_HEAD(&scsi_id->sbp2_command_orb_completed);
        INIT_LIST_HEAD(&scsi_id->scsi_list);
        spin_lock_init(&scsi_id->sbp2_command_orb_lock);
-       atomic_set(&scsi_id->unfinished_reset, 0);
+       atomic_set(&scsi_id->state, SBP2LU_STATE_RUNNING);
        INIT_WORK(&scsi_id->protocol_work, NULL, NULL);
 
        ud->device.driver_data = scsi_id;
@@ -926,13 +926,14 @@ static void sbp2_host_reset(struct hpsb_host *host)
        struct scsi_id_instance_data *scsi_id;
 
        hi = hpsb_get_hostinfo(&sbp2_highlevel, host);
-
-       if (hi) {
-               list_for_each_entry(scsi_id, &hi->scsi_ids, scsi_list) {
-                       atomic_set(&scsi_id->unfinished_reset, 1);
+       if (!hi)
+               return;
+       list_for_each_entry(scsi_id, &hi->scsi_ids, scsi_list)
+               if (likely(atomic_read(&scsi_id->state) !=
+                          SBP2LU_STATE_IN_SHUTDOWN)) {
+                       atomic_set(&scsi_id->state, SBP2LU_STATE_IN_RESET);
                        scsi_block_requests(scsi_id->scsi_host);
                }
-       }
 }
 
 /*

diff --git a/drivers/ieee1394/sbp2.h b/drivers/ieee1394/sbp2.h
index 34c52bf4fa34bc40d3a404195818efeac981a1e7..abbe48e646c3a7397d8a355de0e8f0757b426647 100644 (file)
--- a/drivers/ieee1394/sbp2.h
+++ b/drivers/ieee1394/sbp2.h
@@ -347,10 +347,17 @@ struct scsi_id_instance_data {
        /* Device specific workarounds/brokeness */
        unsigned workarounds;
 
-       atomic_t unfinished_reset;
+       atomic_t state;
        struct work_struct protocol_work;
 };
 
+/* For use in scsi_id_instance_data.state */
+enum sbp2lu_state_types {
+       SBP2LU_STATE_RUNNING,           /* all normal */
+       SBP2LU_STATE_IN_RESET,          /* between bus reset and reconnect */
+       SBP2LU_STATE_IN_SHUTDOWN        /* when sbp2_remove was called */
+};
+
 /* Sbp2 host data structure (one per IEEE1394 host) */
 struct sbp2scsi_host_info {
        struct hpsb_host *host;         /* IEEE1394 host */