[SCSI] ibmvfc: Fix deadlock in EH
Fixes the following deadlock scenario shown below. We currently allow
queuecommand to send commands when the ibmvfc workqueue is scanning for
new rports, so we should also allow EH to function at this time as well.
scsi_eh_3 D 0000000000000000 12304 1279 2
Call Trace:
[c0000002f7257730] [c0000002f72577e0] 0xc0000002f72577e0 (unreliable)
[c0000002f7257900] [c0000000000118f4] .__switch_to+0x158/0x1a0
[c0000002f72579a0] [c0000000004f8b40] .schedule+0x8d4/0x9dc
[c0000002f7257b60] [c0000000004f8f08] .schedule_timeout+0xa8/0xe8
[c0000002f7257c50] [d0000000001d23e0] .ibmvfc_wait_while_resetting+0xe4/0x140 [ibmvfc]
[c0000002f7257d20] [d0000000001d3984] .ibmvfc_eh_abort_handler+0x60/0xe4 [ibmvfc]
[c0000002f7257dc0] [d000000000366714] .scsi_error_handler+0x38c/0x674 [scsi_mod]
[c0000002f7257f00] [c0000000000a7470] .kthread+0x78/0xc4
[c0000002f7257f90] [c000000000029b8c] .kernel_thread+0x4c/0x68
ibmvfc_3 D 0000000000000000 12432 1280 2
Call Trace:
[c0000002f7253540] [c0000002f72535f0] 0xc0000002f72535f0 (unreliable)
[c0000002f7253710] [c0000000000118f4] .__switch_to+0x158/0x1a0
[c0000002f72537b0] [c0000000004f8b40] .schedule+0x8d4/0x9dc
[c0000002f7253970] [c0000000004f8e98] .schedule_timeout+0x38/0xe8
[c0000002f7253a60] [c0000000004f80cc] .wait_for_common+0x138/0x220
[c0000002f7253b40] [c0000000000a2784] .flush_cpu_workqueue+0xac/0xcc
[c0000002f7253c10] [c0000000000a2960] .flush_workqueue+0x58/0xa0
[c0000002f7253ca0] [d0000000000827fc] .fc_flush_work+0x4c/0x64 [scsi_transport_fc]
[c0000002f7253d20] [d000000000082db4] .fc_remote_port_add+0x48/0x6c4 [scsi_transport_fc]
[c0000002f7253dd0] [d0000000001d7d04] .ibmvfc_work+0x820/0xa7c [ibmvfc]
[c0000002f7253f00] [c0000000000a7470] .kthread+0x78/0xc4
[c0000002f7253f90] [c000000000029b8c] .kernel_thread+0x4c/0x68
fc_wq_3 D 0000000000000000 10720 1283 2
Call Trace:
[c0000002f559ac30] [c0000002f559ace0] 0xc0000002f559ace0 (unreliable)
[c0000002f559ae00] [c0000000000118f4] .__switch_to+0x158/0x1a0
[c0000002f559aea0] [c0000000004f8b40] .schedule+0x8d4/0x9dc
[c0000002f559b060] [c0000000004f8e98] .schedule_timeout+0x38/0xe8
[c0000002f559b150] [c0000000004f80cc] .wait_for_common+0x138/0x220
[c0000002f559b230] [c0000000002721c4] .blk_execute_rq+0xb4/0x100
[c0000002f559b360] [d00000000036a1f8] .scsi_execute+0x118/0x194 [scsi_mod]
[c0000002f559b420] [d00000000036a32c] .scsi_execute_req+0xb8/0x124 [scsi_mod]
[c0000002f559b500] [d0000000000c1330] .sd_sync_cache+0x8c/0x108 [sd_mod]
[c0000002f559b5e0] [d0000000000c15b4] .sd_shutdown+0x9c/0x158 [sd_mod]
[c0000002f559b660] [d0000000000c16d0] .sd_remove+0x60/0xb4 [sd_mod]
[c0000002f559b700] [c000000000392ecc] .__device_release_driver+0xd0/0x118
[c0000002f559b7a0] [c000000000393080] .device_release_driver+0x30/0x54
[c0000002f559b830] [c000000000392108] .bus_remove_device+0x128/0x16c
[c0000002f559b8d0] [c00000000038f94c] .device_del+0x158/0x234
[c0000002f559b960] [d00000000036f078] .__scsi_remove_device+0x5c/0xd4 [scsi_mod]
[c0000002f559b9f0] [d00000000036f124] .scsi_remove_device+0x34/0x58 [scsi_mod]
[c0000002f559ba80] [d00000000036f204] .__scsi_remove_target+0xb4/0x120 [scsi_mod]
[c0000002f559bb10] [d00000000036f338] .__remove_child+0x2c/0x44 [scsi_mod]
[c0000002f559bb90] [c00000000038f11c] .device_for_each_child+0x54/0xb4
[c0000002f559bc50] [d00000000036f2e0] .scsi_remove_target+0x70/0x9c [scsi_mod]
[c0000002f559bce0] [d000000000083454] .fc_starget_delete+0x24/0x3c [scsi_transport_fc]
[c0000002f559bd70] [c0000000000a2368] .run_workqueue+0x118/0x208
[c0000002f559be30] [c0000000000a2580] .worker_thread+0x128/0x154
[c0000002f559bf00] [c0000000000a7470] .kthread+0x78/0xc4
[c0000002f559bf90] [c000000000029b8c] .kernel_thread+0x4c/0x68
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
diff --git a/drivers/scsi/ibmvscsi/ibmvfc.c b/drivers/scsi/ibmvscsi/ibmvfc.c
index 182c8e7..76ae266 100644
--- a/drivers/scsi/ibmvscsi/ibmvfc.c
+++ b/drivers/scsi/ibmvscsi/ibmvfc.c
@@ -431,6 +431,8 @@
case IBMVFC_TGT_ACTION_DEL_RPORT:
break;
default:
+ if (action == IBMVFC_TGT_ACTION_DEL_RPORT)
+ tgt->add_rport = 0;
tgt->action = action;
break;
}
@@ -483,7 +485,7 @@
switch (vhost->action) {
case IBMVFC_HOST_ACTION_INIT_WAIT:
case IBMVFC_HOST_ACTION_NONE:
- case IBMVFC_HOST_ACTION_TGT_ADD:
+ case IBMVFC_HOST_ACTION_TGT_DEL_FAILED:
vhost->action = action;
break;
default:
@@ -498,7 +500,6 @@
case IBMVFC_HOST_ACTION_TGT_DEL:
case IBMVFC_HOST_ACTION_QUERY_TGTS:
case IBMVFC_HOST_ACTION_TGT_DEL_FAILED:
- case IBMVFC_HOST_ACTION_TGT_ADD:
case IBMVFC_HOST_ACTION_NONE:
default:
vhost->action = action;
@@ -2306,7 +2307,7 @@
done = 1;
}
- if (vhost->state != IBMVFC_NO_CRQ && vhost->action == IBMVFC_HOST_ACTION_NONE)
+ if (vhost->scan_complete)
done = 1;
spin_unlock_irqrestore(shost->host_lock, flags);
return done;
@@ -2820,7 +2821,7 @@
tgt->ids.roles |= FC_PORT_ROLE_FCP_TARGET;
if (parms->service_parms & IBMVFC_PRLI_INITIATOR_FUNC)
tgt->ids.roles |= FC_PORT_ROLE_FCP_INITIATOR;
- ibmvfc_set_tgt_action(tgt, IBMVFC_TGT_ACTION_ADD_RPORT);
+ tgt->add_rport = 1;
} else
ibmvfc_set_tgt_action(tgt, IBMVFC_TGT_ACTION_DEL_RPORT);
} else if (prli_rsp[index].retry)
@@ -3660,7 +3661,6 @@
return 1;
case IBMVFC_HOST_ACTION_INIT:
case IBMVFC_HOST_ACTION_ALLOC_TGTS:
- case IBMVFC_HOST_ACTION_TGT_ADD:
case IBMVFC_HOST_ACTION_TGT_DEL:
case IBMVFC_HOST_ACTION_TGT_DEL_FAILED:
case IBMVFC_HOST_ACTION_QUERY:
@@ -3715,25 +3715,26 @@
static void ibmvfc_tgt_add_rport(struct ibmvfc_target *tgt)
{
struct ibmvfc_host *vhost = tgt->vhost;
- struct fc_rport *rport = tgt->rport;
+ struct fc_rport *rport;
unsigned long flags;
- if (rport) {
- tgt_dbg(tgt, "Setting rport roles\n");
- fc_remote_port_rolechg(rport, tgt->ids.roles);
- spin_lock_irqsave(vhost->host->host_lock, flags);
- ibmvfc_set_tgt_action(tgt, IBMVFC_TGT_ACTION_NONE);
- spin_unlock_irqrestore(vhost->host->host_lock, flags);
- return;
- }
-
tgt_dbg(tgt, "Adding rport\n");
rport = fc_remote_port_add(vhost->host, 0, &tgt->ids);
spin_lock_irqsave(vhost->host->host_lock, flags);
- tgt->rport = rport;
- ibmvfc_set_tgt_action(tgt, IBMVFC_TGT_ACTION_NONE);
+
+ if (rport && tgt->action == IBMVFC_TGT_ACTION_DEL_RPORT) {
+ tgt_dbg(tgt, "Deleting rport\n");
+ list_del(&tgt->queue);
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ fc_remote_port_delete(rport);
+ del_timer_sync(&tgt->timer);
+ kref_put(&tgt->kref, ibmvfc_release_tgt);
+ return;
+ }
+
if (rport) {
tgt_dbg(tgt, "rport add succeeded\n");
+ tgt->rport = rport;
rport->maxframe_size = tgt->service_parms.common.bb_rcv_sz & 0x0fff;
rport->supported_classes = 0;
tgt->target_id = rport->scsi_target_id;
@@ -3811,11 +3812,21 @@
if (vhost->state == IBMVFC_INITIALIZING) {
if (vhost->action == IBMVFC_HOST_ACTION_TGT_DEL_FAILED) {
- ibmvfc_set_host_state(vhost, IBMVFC_ACTIVE);
- ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_TGT_ADD);
- vhost->init_retries = 0;
- spin_unlock_irqrestore(vhost->host->host_lock, flags);
- scsi_unblock_requests(vhost->host);
+ if (vhost->reinit) {
+ vhost->reinit = 0;
+ scsi_block_requests(vhost->host);
+ ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_QUERY);
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ } else {
+ ibmvfc_set_host_state(vhost, IBMVFC_ACTIVE);
+ ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_NONE);
+ wake_up(&vhost->init_wait_q);
+ schedule_work(&vhost->rport_add_work_q);
+ vhost->init_retries = 0;
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ scsi_unblock_requests(vhost->host);
+ }
+
return;
} else {
ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_INIT);
@@ -3846,24 +3857,6 @@
if (!ibmvfc_dev_init_to_do(vhost))
ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_TGT_DEL_FAILED);
break;
- case IBMVFC_HOST_ACTION_TGT_ADD:
- list_for_each_entry(tgt, &vhost->targets, queue) {
- if (tgt->action == IBMVFC_TGT_ACTION_ADD_RPORT) {
- spin_unlock_irqrestore(vhost->host->host_lock, flags);
- ibmvfc_tgt_add_rport(tgt);
- return;
- }
- }
-
- if (vhost->reinit && !ibmvfc_set_host_state(vhost, IBMVFC_INITIALIZING)) {
- vhost->reinit = 0;
- scsi_block_requests(vhost->host);
- ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_QUERY);
- } else {
- ibmvfc_set_host_action(vhost, IBMVFC_HOST_ACTION_NONE);
- wake_up(&vhost->init_wait_q);
- }
- break;
default:
break;
};
@@ -4093,6 +4086,56 @@
}
/**
+ * ibmvfc_rport_add_thread - Worker thread for rport adds
+ * @work: work struct
+ *
+ **/
+static void ibmvfc_rport_add_thread(struct work_struct *work)
+{
+ struct ibmvfc_host *vhost = container_of(work, struct ibmvfc_host,
+ rport_add_work_q);
+ struct ibmvfc_target *tgt;
+ struct fc_rport *rport;
+ unsigned long flags;
+ int did_work;
+
+ ENTER;
+ spin_lock_irqsave(vhost->host->host_lock, flags);
+ do {
+ did_work = 0;
+ if (vhost->state != IBMVFC_ACTIVE)
+ break;
+
+ list_for_each_entry(tgt, &vhost->targets, queue) {
+ if (tgt->add_rport) {
+ did_work = 1;
+ tgt->add_rport = 0;
+ kref_get(&tgt->kref);
+ rport = tgt->rport;
+ if (!rport) {
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ ibmvfc_tgt_add_rport(tgt);
+ } else if (get_device(&rport->dev)) {
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ tgt_dbg(tgt, "Setting rport roles\n");
+ fc_remote_port_rolechg(rport, tgt->ids.roles);
+ put_device(&rport->dev);
+ }
+
+ kref_put(&tgt->kref, ibmvfc_release_tgt);
+ spin_lock_irqsave(vhost->host->host_lock, flags);
+ break;
+ }
+ }
+ } while(did_work);
+
+ if (vhost->state == IBMVFC_ACTIVE)
+ vhost->scan_complete = 1;
+ spin_unlock_irqrestore(vhost->host->host_lock, flags);
+ LEAVE;
+}
+
+/**
* ibmvfc_probe - Adapter hot plug add entry point
* @vdev: vio device struct
* @id: vio device id struct
@@ -4135,6 +4178,7 @@
strcpy(vhost->partition_name, "UNKNOWN");
init_waitqueue_head(&vhost->work_wait_q);
init_waitqueue_head(&vhost->init_wait_q);
+ INIT_WORK(&vhost->rport_add_work_q, ibmvfc_rport_add_thread);
if ((rc = ibmvfc_alloc_mem(vhost)))
goto free_scsi_host;