zfcp: fix missing auto port scan and thus missing target ports
Subsystems (3)
PATCH 1/1 zfcp: fix missing auto port scan and thus missing target ports
Case (1):
The only waiter on wka_port->completion_wq is zfcp_fc_wka_port_get() trying to open a WKA port. As such it should only be woken up by WKA port *open* responses, not by WKA port close responses.
Case (2):
A close WKA port response coming in just after having sent a new open WKA port request and before blocking for the open response with wait_event() in zfcp_fc_wka_port_get() erroneously renders the wait_event a NOP because the close handler overwrites wka_port->status. Hence the wait_event condition is erroneously true and it does not enter blocking state.
With non-negligible probability, the following time space sequence happens depending on timing without this fix:
user process ERP thread zfcp work queue tasklet system work queue
============ ========== =============== ======= =================
$ echo 1 > online
zfcp_ccw_set_online
zfcp_ccw_activate
zfcp_erp_adapter_reopen
msleep scan backoff zfcp_erp_strategy
| ...
| zfcp_erp_action_cleanup
| ...
| queue delayed scan_work
| queue ns_up_work
| ns_up_work:
| zfcp_fc_wka_port_get
| open wka request
| open response
| GSPN FC-GS
| RSPN FC-GS [NPIV-only]
| zfcp_fc_wka_port_put
| (--wka->refcount==0)
| sched delayed wka->work
|
~~~Case (1)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
zfcp_erp_wait
flush scan_work
| wka->work:
| wka->status=CLOSING
| close wka request
| scan_work:
| zfcp_fc_wka_port_get
| (wka->status==CLOSING)
| wka->status=OPENING
| open wka request
| wait_event
| | close response
| | wka->status=OFFLINE
| | wake_up /*WRONG*/
~~~Case (2)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| wka->work:
| wka->status=CLOSING
| close wka request
zfcp_erp_wait
flush scan_work
| scan_work:
| zfcp_fc_wka_port_get
| (wka->status==CLOSING)
| wka->status=OPENING
| open wka request
| close response
| wka->status=OFFLINE
| wake_up /*WRONG&NOP*/
| wait_event /*NOP*/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| (wka->status!=ONLINE)
| return -EIO
| return early
open response
wka->status=ONLINE
wake_up /*NOP*/So we erroneously end up with no automatic port scan. This is a big problem when it happens during boot. The timing is influenced by v3.19 commit 18f87a67e6d6 ("zfcp: auto port scan resiliency").
Fix it by fully mutually excluding zfcp_fc_wka_port_get() and zfcp_fc_wka_port_offline(). For that to work, we make the latter block until we got the response for a close WKA port. In order not to penalize the system workqueue, we move wka_port->work to our own adapter workqueue. Note that before v2.6.30 commit 828bc1212a68 ("[SCSI] zfcp: Set WKA-port to offline on adapter deactivation"), zfcp did block in zfcp_fc_wka_port_offline() as well, but with a different condition.
While at it, make non-functional cleanups to improve code reading in zfcp_fc_wka_port_get(). If we cannot send the WKA port open request, don't rely on the subsequent wait_event condition to immediately let this case pass without blocking. Also don't want to rely on the additional condition handling the refcount to be skipped just to finally return with -EIO.
--- drivers/s390/scsi/zfcp_fc.c | 29 ++++++++++++++++++++--------- drivers/s390/scsi/zfcp_fc.h | 6 ++++-- drivers/s390/scsi/zfcp_fsf.c | 4 ++-- 3 files changed, 26 insertions(+), 13 deletions(-)
drivers/s390/scsi/zfcp_fc.c +20 -9
@@ -145,27 +145,33 @@ void zfcp_fc_enqueue_event(struct zfcp_adapter *adapter,static int zfcp_fc_wka_port_get (struct zfcp_fc_wka_port *wka_port) { +int ret = -EIO ; +if (mutex_lock_interruptible (&wka_port->mutex))return -ERESTARTSYS ;if (wka_port->status ==ZFCP_FC_WKA_PORT_OFFLINE || wka_port->status ==ZFCP_FC_WKA_PORT_CLOSING ) { wka_port->status =ZFCP_FC_WKA_PORT_OPENING ; -if (zfcp_fsf_open_wka_port (wka_port)) +if (zfcp_fsf_open_wka_port (wka_port)) { +/* could not even send request, nothing to wait for */ wka_port->status =ZFCP_FC_WKA_PORT_OFFLINE ; + goto out; + } } -mutex_unlock (&wka_port->mutex); - -wait_event (wka_port->completion_wq, +wait_event (wka_port->opened, wka_port->status ==ZFCP_FC_WKA_PORT_ONLINE || wka_port->status ==ZFCP_FC_WKA_PORT_OFFLINE );if (wka_port->status ==ZFCP_FC_WKA_PORT_ONLINE ) {atomic_inc (&wka_port->refcount); -return 0 ; + ret =0 ; + goto out; } -return -EIO ; +out: +mutex_unlock (&wka_port->mutex); +return ret; }static void zfcp_fc_wka_port_offline (struct work_struct *work)
@@ -181,9 +187,12 @@ static void zfcp_fc_wka_port_offline(struct work_struct *work) wka_port->status =ZFCP_FC_WKA_PORT_CLOSING ;if (zfcp_fsf_close_wka_port (wka_port)) { +/* could not even send request, nothing to wait for */ wka_port->status =ZFCP_FC_WKA_PORT_OFFLINE ; -wake_up (&wka_port->completion_wq); + goto out; } +wait_event (wka_port->closed, + wka_port->status ==ZFCP_FC_WKA_PORT_OFFLINE ); out:mutex_unlock (&wka_port->mutex); }
@@ -193,13 +202,15 @@ static void zfcp_fc_wka_port_put(struct zfcp_fc_wka_port *wka_port)if (atomic_dec_return (&wka_port->refcount) !=0 )return ;/* wait 10 milliseconds, other reqs might pop in */ -schedule_delayed_work (&wka_port->work,HZ /100 ); +queue_delayed_work (wka_port->adapter->work_queue, &wka_port->work, +msecs_to_jiffies (10 )); }static void zfcp_fc_wka_port_init (struct zfcp_fc_wka_port *wka_port,u32 d_id,struct zfcp_adapter *adapter) { -init_waitqueue_head (&wka_port->completion_wq); +init_waitqueue_head (&wka_port->opened); +init_waitqueue_head (&wka_port->closed); wka_port->adapter = adapter; wka_port->d_id = d_id;
drivers/s390/scsi/zfcp_fc.h +4 -2
@@ -185,7 +185,8 @@ enum zfcp_fc_wka_status {/** * struct zfcp_fc_wka_port - representation of well-known-address (WKA) FC port * @adapter: Pointer to adapter structure this WKA port belongs to -* @completion_wq: Wait for completion of open/close command +* @opened: Wait for completion of open command +* @closed: Wait for completion of close command * @status: Current status of WKA port * @refcount: Reference count to keep port open as long as it is in use * @d_id: FC destination id or well-known-address
@@ -195,7 +196,8 @@ enum zfcp_fc_wka_status {*/ struct zfcp_fc_wka_port {struct zfcp_adapter *adapter; -wait_queue_head_t completion_wq; +wait_queue_head_t opened; +wait_queue_head_t closed;enum zfcp_fc_wka_status status;atomic_t refcount;u32 d_id;
drivers/s390/scsi/zfcp_fsf.c +2 -2
@@ -1907,7 +1907,7 @@ static void zfcp_fsf_open_wka_port_handler(struct zfcp_fsf_req *req) wka_port->status =ZFCP_FC_WKA_PORT_ONLINE ; } out: -wake_up (&wka_port->completion_wq); +wake_up (&wka_port->opened); }/**
@@ -1966,7 +1966,7 @@ static void zfcp_fsf_close_wka_port_handler(struct zfcp_fsf_req *req) } wka_port->status =ZFCP_FC_WKA_PORT_OFFLINE ; -wake_up (&wka_port->completion_wq); +wake_up (&wka_port->closed); }/**
base-commit: 71b25693b22ebb9391b27f011d3f4bf9762e24f9
Signature · 1 line
2.34.1
Martin K. Petersen
AppliedSteffen,
Quoted 11 lines
Case (1): The only waiter on
wka_port->completion_wqiszfcp_fc_wka_port_get()trying to open a WKA port. As such it should only be woken up by WKA port *open* responses, not by WKA port close responses.Case (2): A close WKA port response coming in just after having sent a new open WKA port request and before blocking for the open response with
wait_event()inzfcp_fc_wka_port_get()erroneously renders thewait_eventa NOP because the close handler overwriteswka_port->status. Hence thewait_eventcondition is erroneously true and it does not enter blocking state.
Applied to 5.20/scsi-staging, thanks!
Signature · 1 line
Martin K. Petersen Oracle Linux Engineering