#define ZFCP_LOG_AREA ZFCP_LOG_AREA_ERP
/* this drivers version (do not edit !!! generated and updated by cvs) */
-#define ZFCP_ERP_REVISION "$Revision: 1.54 $"
+#define ZFCP_ERP_REVISION "$Revision: 1.65 $"
#include "zfcp_ext.h"
static int zfcp_els(struct zfcp_port *, u8);
-static int zfcp_els_handler(unsigned long);
+static void zfcp_els_handler(unsigned long);
static int zfcp_erp_adapter_reopen_internal(struct zfcp_adapter *, int);
static int zfcp_erp_port_forced_reopen_internal(struct zfcp_port *, int);
static void zfcp_erp_timeout_handler(unsigned long);
static inline void zfcp_erp_timeout_init(struct zfcp_erp_action *);
+/**
+ * zfcp_fsf_request_timeout_handler - called if a request timed out
+ * @data: pointer to adapter for handler function
+ *
+ * This function needs to be called if requests (ELS, Generic Service,
+ * or SCSI commands) exceed a certain time limit. The assumption is
+ * that after the time limit the adapter get stuck. So we trigger a reopen of
+ * the adapter. This should not be used for error recovery, SCSI abort
+ * commands and SCSI requests from SCSI mid-layer.
+ */
+void
+zfcp_fsf_request_timeout_handler(unsigned long data)
+{
+ struct zfcp_adapter *adapter;
+
+ adapter = (struct zfcp_adapter *) data;
+
+ zfcp_erp_adapter_reopen(adapter, 0);
+}
/*
* function: zfcp_fsf_scsi_er_timeout_handler
send_els->completion = NULL;
req = zfcp_sg_to_address(send_els->req);
+ memset(req, 0, PAGE_SIZE);
*(u32*)req = 0;
*(u8*)req = ls_code;
}
-/*
- * function: zfcp_els_handler
- *
- * purpose: Handler for all kind of ELSs
- *
- * returns: 0 - Operation completed successfuly
- * -ENXIO - ELS has been rejected
- * -EPERM - Port forced reopen failed
+/**
+ * zfcp_els_handler - handler for ELS commands
+ * @data: pointer to struct zfcp_send_els
+ * If ELS failed (LS_RJT or timed out) forced reopen of the port is triggered.
*/
-int
+void
zfcp_els_handler(unsigned long data)
{
struct zfcp_send_els *send_els = (struct zfcp_send_els*)data;
struct zfcp_port *port = send_els->port;
- struct zfcp_ls_rjt *rjt;
struct zfcp_ls_rtv_acc *rtv;
struct zfcp_ls_rls_acc *rls;
struct zfcp_ls_pdisc_acc *pdisc;
struct zfcp_ls_adisc_acc *adisc;
void *req, *resp;
- u8 req_code, resp_code;
- int retval = 0;
+ u8 req_code;
- if (send_els->status != 0)
- goto skip_fsfstatus;
+ /* request rejected or timed out */
+ if (send_els->status != 0) {
+ ZFCP_LOG_NORMAL("ELS request timed out, force physical port "
+ "reopen of port 0x%016Lx on adapter %s\n",
+ port->wwpn, zfcp_get_busid_by_port(port));
+ debug_text_event(port->adapter->erp_dbf, 3, "forcreop");
+ if (zfcp_erp_port_forced_reopen(port, 0))
+ ZFCP_LOG_NORMAL("reopen of remote port 0x%016Lx "
+ "on adapter %s failed\n", port->wwpn,
+ zfcp_get_busid_by_port(port));
+ goto out;
+ }
- req = (void*)((page_to_pfn(send_els->req->page) << PAGE_SHIFT) + send_els->req->offset);
- resp = (void*)((page_to_pfn(send_els->resp->page) << PAGE_SHIFT) + send_els->resp->offset);
+ req = zfcp_sg_to_address(send_els->req);
+ resp = zfcp_sg_to_address(send_els->resp);
req_code = *(u8*)req;
- resp_code = *(u8*)resp;
-
- switch (resp_code) {
- case ZFCP_LS_RJT:
- rjt = (struct zfcp_ls_rjt*)resp;
+ switch (req_code) {
- switch (rjt->reason_code) {
-
- case ZFCP_LS_RJT_INVALID_COMMAND_CODE:
- ZFCP_LOG_INFO("invalid LS command code "
- "(wwpn=0x%016Lx, command=0x%02x)\n",
- port->wwpn, req_code);
- break;
-
- case ZFCP_LS_RJT_LOGICAL_ERROR:
- ZFCP_LOG_INFO("logical error (wwpn=0x%016Lx, "
- "reason_expl=0x%02x)\n",
- port->wwpn, rjt->reason_expl);
- break;
-
- case ZFCP_LS_RJT_LOGICAL_BUSY:
- ZFCP_LOG_INFO("logical busy (wwpn=0x%016Lx, "
- "reason_expl=0x%02x)\n",
- port->wwpn, rjt->reason_expl);
- break;
-
- case ZFCP_LS_RJT_PROTOCOL_ERROR:
- ZFCP_LOG_INFO("protocol error (wwpn=0x%016Lx, "
- "reason_expl=0x%02x)\n",
- port->wwpn, rjt->reason_expl);
- break;
-
- case ZFCP_LS_RJT_UNABLE_TO_PERFORM:
- ZFCP_LOG_INFO("unable to perform command requested "
- "(wwpn=0x%016Lx, reason_expl=0x%02x)\n",
- port->wwpn, rjt->reason_expl);
- break;
-
- case ZFCP_LS_RJT_COMMAND_NOT_SUPPORTED:
- ZFCP_LOG_INFO("command not supported (wwpn=0x%016Lx, "
- "command=0x%02x)\n",
- port->wwpn, req_code);
- break;
-
- case ZFCP_LS_RJT_VENDOR_UNIQUE_ERROR:
- ZFCP_LOG_INFO("vendor specific error (wwpn=0x%016Lx, "
- "vendor_unique=0x%02x)\n",
- port->wwpn, rjt->vendor_unique);
- break;
-
- default:
- ZFCP_LOG_NORMAL("ELS rejected by remote port 0x%016Lx "
- "on adapter %s (reason_code=0x%02x)\n",
- port->wwpn,
- zfcp_get_busid_by_port(port),
- rjt->reason_code);
- }
- retval = -ENXIO;
+ case ZFCP_LS_RTV:
+ rtv = (struct zfcp_ls_rtv_acc*)resp;
+ ZFCP_LOG_INFO("RTV response from d_id 0x%08x to s_id "
+ "0x%08x (R_A_TOV=%ds E_D_TOV=%d%cs)\n",
+ port->d_id, port->adapter->s_id,
+ rtv->r_a_tov, rtv->e_d_tov,
+ rtv->qualifier &
+ ZFCP_LS_RTV_E_D_TOV_FLAG ? 'n' : 'm');
break;
- case ZFCP_LS_ACC:
- switch (req_code) {
-
- case ZFCP_LS_RTV:
- rtv = (struct zfcp_ls_rtv_acc*)resp;
- ZFCP_LOG_INFO("RTV response from d_id 0x%08x to s_id "
- "0x%08x (R_A_TOV=%ds E_D_TOV=%d%cs)\n",
- port->d_id, port->adapter->s_id,
- rtv->r_a_tov, rtv->e_d_tov,
- rtv->qualifier &
- ZFCP_LS_RTV_E_D_TOV_FLAG ? 'n' : 'm');
- break;
-
- case ZFCP_LS_RLS:
- rls = (struct zfcp_ls_rls_acc*)resp;
- ZFCP_LOG_INFO("RLS response from d_id 0x%08x to s_id "
- "0x%08x (link_failure_count=%u, "
- "loss_of_sync_count=%u, "
- "loss_of_signal_count=%u, "
- "primitive_sequence_protocol_error=%u, "
- "invalid_transmition_word=%u, "
- "invalid_crc_count=%u)\n",
- port->d_id, port->adapter->s_id,
- rls->link_failure_count,
- rls->loss_of_sync_count,
- rls->loss_of_signal_count,
- rls->prim_seq_prot_error,
- rls->invalid_transmition_word,
- rls->invalid_crc_count);
- break;
-
- case ZFCP_LS_PDISC:
- pdisc = (struct zfcp_ls_pdisc_acc*)resp;
- ZFCP_LOG_INFO("PDISC response from d_id 0x%08x to s_id "
- "0x%08x (wwpn=0x%016Lx, wwnn=0x%016Lx, "
- "vendor='%-16s')\n", port->d_id,
- port->adapter->s_id, pdisc->wwpn,
- pdisc->wwnn, pdisc->vendor_version);
- break;
+ case ZFCP_LS_RLS:
+ rls = (struct zfcp_ls_rls_acc*)resp;
+ ZFCP_LOG_INFO("RLS response from d_id 0x%08x to s_id "
+ "0x%08x (link_failure_count=%u, "
+ "loss_of_sync_count=%u, "
+ "loss_of_signal_count=%u, "
+ "primitive_sequence_protocol_error=%u, "
+ "invalid_transmition_word=%u, "
+ "invalid_crc_count=%u)\n",
+ port->d_id, port->adapter->s_id,
+ rls->link_failure_count,
+ rls->loss_of_sync_count,
+ rls->loss_of_signal_count,
+ rls->prim_seq_prot_error,
+ rls->invalid_transmition_word,
+ rls->invalid_crc_count);
+ break;
- case ZFCP_LS_ADISC:
- adisc = (struct zfcp_ls_adisc_acc*)resp;
- ZFCP_LOG_INFO("ADISC response from d_id 0x%08x to s_id "
- "0x%08x (wwpn=0x%016Lx, wwnn=0x%016Lx, "
- "hard_nport_id=0x%08x, "
- "nport_id=0x%08x)\n", port->d_id,
- port->adapter->s_id, adisc->wwpn,
- adisc->wwnn, adisc->hard_nport_id,
- adisc->nport_id);
- /* FIXME: set wwnn in during open port */
- if (port->wwnn == 0)
- port->wwnn = adisc->wwnn;
- break;
- }
+ case ZFCP_LS_PDISC:
+ pdisc = (struct zfcp_ls_pdisc_acc*)resp;
+ ZFCP_LOG_INFO("PDISC response from d_id 0x%08x to s_id "
+ "0x%08x (wwpn=0x%016Lx, wwnn=0x%016Lx, "
+ "vendor='%-16s')\n", port->d_id,
+ port->adapter->s_id, pdisc->wwpn,
+ pdisc->wwnn, pdisc->vendor_version);
break;
- default:
- ZFCP_LOG_NORMAL("unknown payload code 0x%02x received for "
- "request 0x%02x to d_id 0x%08x, reopen needed "
- "for port 0x%016Lx on adapter %s\n", resp_code,
- req_code, port->d_id, port->wwpn,
- zfcp_get_busid_by_port(port));
- retval = zfcp_erp_port_forced_reopen(port, 0);
- if (retval != 0) {
- ZFCP_LOG_NORMAL("reopen of remote port 0x%016Lx on "
- "adapter %s failed\n", port->wwpn,
- zfcp_get_busid_by_port(port));
- retval = -EPERM;
- }
+ case ZFCP_LS_ADISC:
+ adisc = (struct zfcp_ls_adisc_acc*)resp;
+ ZFCP_LOG_INFO("ADISC response from d_id 0x%08x to s_id "
+ "0x%08x (wwpn=0x%016Lx, wwnn=0x%016Lx, "
+ "hard_nport_id=0x%08x, "
+ "nport_id=0x%08x)\n", port->d_id,
+ port->adapter->s_id, adisc->wwpn,
+ adisc->wwnn, adisc->hard_nport_id,
+ adisc->nport_id);
+ /* FIXME: set wwnn in during open port */
+ if (port->wwnn == 0)
+ port->wwnn = adisc->wwnn;
+ break;
}
-skip_fsfstatus:
+ out:
__free_pages(send_els->req->page, 0);
kfree(send_els->req);
kfree(send_els->resp);
-
- return retval;
+ kfree(send_els);
}
return retval;
}
-/*
- * function:
- *
- * purpose: Wrappper for zfcp_erp_port_reopen_internal
- * used to ensure the correct locking
+/**
+ * zfcp_erp_port_reopen - initiate reopen of a remote port
+ * @port: port to be reopened
+ * @clear_mask: specifies flags in port status to be cleared
+ * Return: 0 on success, < 0 on error
*
- * returns: 0 - initiated action succesfully
- * <0 - failed to initiate action
+ * This is a wrappper function for zfcp_erp_port_reopen_internal. It ensures
+ * correct locking. An error recovery task is initiated to do the reopen.
+ * To wait for the completion of the reopen zfcp_erp_wait should be used.
*/
int
zfcp_erp_port_reopen(struct zfcp_port *port, int clear_mask)
return retval;
}
-/*
- * function:
- *
- * purpose: Wrappper for zfcp_erp_unit_reopen_internal
- * used to ensure the correct locking
+/**
+ * zfcp_erp_unit_reopen - initiate reopen of a unit
+ * @unit: unit to be reopened
+ * @clear_mask: specifies flags in unit status to be cleared
+ * Return: 0 on success, < 0 on error
*
- * returns: 0 - initiated action succesfully
- * <0 - failed to initiate action
+ * This is a wrappper for zfcp_erp_unit_reopen_internal. It ensures correct
+ * locking. An error recovery task is initiated to do the reopen.
+ * To wait for the completion of the reopen zfcp_erp_wait should be used.
*/
int
zfcp_erp_unit_reopen(struct zfcp_unit *unit, int clear_mask)
return retval;
}
-/*
- * function:
- *
- * purpose:
- *
- * returns:
+/**
+ * zfcp_erp_wait - wait for completion of error recovery on an adapter
+ * @adapter: adapter for which to wait for completion of its error recovery
+ * Return: 0
*/
int
zfcp_erp_wait(struct zfcp_adapter *adapter)
struct zfcp_port *port;
list_for_each_entry(port, &adapter->port_list_head, list)
- if (!atomic_test_mask(ZFCP_STATUS_PORT_NAMESERVER, &port->status))
+ if (!atomic_test_mask(ZFCP_STATUS_PORT_WKA, &port->status))
zfcp_erp_port_reopen_internal(port, clear_mask);
return retval;
zfcp_erp_adapter_strategy(struct zfcp_erp_action *erp_action)
{
int retval;
- unsigned long timeout;
struct zfcp_adapter *adapter = erp_action->adapter;
retval = zfcp_erp_adapter_strategy_close(erp_action);
ZFCP_LOG_INFO("Waiting to allow the adapter %s "
"to recover itself\n",
zfcp_get_busid_by_adapter(adapter));
- /*
- * SUGGESTION: substitute by
- * timeout = ZFCP_TYPE2_RECOVERY_TIME;
- * __ZFCP_WAIT_EVENT_TIMEOUT(timeout, 0);
- */
- timeout = ZFCP_TYPE2_RECOVERY_TIME;
- set_current_state(TASK_UNINTERRUPTIBLE);
- schedule_timeout(timeout);
+ msleep(jiffies_to_msecs(ZFCP_TYPE2_RECOVERY_TIME));
}
return retval;
int i;
volatile struct qdio_buffer_element *sbale;
struct zfcp_adapter *adapter = erp_action->adapter;
- int retval_cleanup = 0;
if (atomic_test_mask(ZFCP_STATUS_ADAPTER_QDIOUP, &adapter->status)) {
ZFCP_LOG_NORMAL("bug: second attempt to set up QDIO on "
zfcp_get_busid_by_adapter(adapter));
goto failed_qdio_establish;
}
- ZFCP_LOG_DEBUG("queues established\n");
+ debug_text_event(adapter->erp_dbf, 3, "qdio_est");
if (qdio_activate(adapter->ccw_device, 0) != 0) {
ZFCP_LOG_INFO("error: activation of QDIO queues failed "
zfcp_get_busid_by_adapter(adapter));
goto failed_qdio_activate;
}
- ZFCP_LOG_DEBUG("queues activated\n");
+ debug_text_event(adapter->erp_dbf, 3, "qdio_act");
/*
* put buffers into response queue,
/* NOP */
failed_qdio_activate:
- /* DEBUG */
- //__ZFCP_WAIT_EVENT_TIMEOUT(timeout, 0);
- /* cleanup queues previously established */
- retval_cleanup = qdio_shutdown(adapter->ccw_device,
- QDIO_FLAG_CLEANUP_USING_CLEAR);
- if (retval_cleanup) {
- ZFCP_LOG_NORMAL("bug: shutdown of QDIO queues failed "
- "(retval=%d)\n", retval_cleanup);
- }
+ debug_text_event(adapter->erp_dbf, 3, "qdio_down1a");
+ while (qdio_shutdown(adapter->ccw_device,
+ QDIO_FLAG_CLEANUP_USING_CLEAR) == -EINPROGRESS)
+ msleep(1000);
+ debug_text_event(adapter->erp_dbf, 3, "qdio_down1b");
failed_qdio_establish:
- atomic_clear_mask(ZFCP_STATUS_ADAPTER_QDIOUP, &adapter->status);
-
failed_sanity:
retval = ZFCP_ERP_FAILED;
goto out;
}
- /* cleanup queues previously established */
-
/*
- * MUST NOT LOCK - qdio_cleanup might call schedule
- * FIXME: need another way to make cleanup safe
- */
- /* Note:
- * We need the request_queue lock here, otherwise there exists the
- * following race:
- *
- * queuecommand calls create_fcp_commmand_task...calls req_create,
- * gets sbal x to x+y - meanwhile adapter reopen is called, completes
- * - req_send calls do_QDIO for sbal x to x+y, i.e. wrong indices.
- *
- * with lock:
- * queuecommand calls create_fcp_commmand_task...calls req_create,
- * gets sbal x to x+y - meanwhile adapter reopen is called, waits
- * - req_send calls do_QDIO for sbal x to x+y, i.e. wrong indices
- * but do_QDIO fails as adapter_reopen is still waiting for the lock
- * OR
- * queuecommand calls create_fcp_commmand_task...calls req_create
- * - meanwhile adapter reopen is called...completes,
- * - gets sbal 0 to 0+y, - req_send calls do_QDIO for sbal 0 to 0+y,
- * i.e. correct indices...though an fcp command is called before
- * exchange config data...that should be fine, however
+ * Get queue_lock and clear QDIOUP flag. Thus it's guaranteed that
+ * do_QDIO won't be called while qdio_shutdown is in progress.
*/
- if (qdio_shutdown(adapter->ccw_device, QDIO_FLAG_CLEANUP_USING_CLEAR)) {
- /*
- * FIXME(design):
- * What went wrong? What to do best? Proper retval?
- */
- ZFCP_LOG_NORMAL("bug: shutdown of QDIO queues failed on "
- "adapter %s\n",
- zfcp_get_busid_by_adapter(adapter));
- } else
- ZFCP_LOG_DEBUG("queues cleaned up\n");
+
+ write_lock_irq(&adapter->request_queue.queue_lock);
+ atomic_clear_mask(ZFCP_STATUS_ADAPTER_QDIOUP, &adapter->status);
+ write_unlock_irq(&adapter->request_queue.queue_lock);
+
+ debug_text_event(adapter->erp_dbf, 3, "qdio_down2a");
+ while (qdio_shutdown(adapter->ccw_device,
+ QDIO_FLAG_CLEANUP_USING_CLEAR) == -EINPROGRESS)
+ msleep(1000);
+ debug_text_event(adapter->erp_dbf, 3, "qdio_down2b");
/*
* First we had to stop QDIO operation.
adapter->request_queue.free_index = 0;
atomic_set(&adapter->request_queue.free_count, 0);
adapter->request_queue.distance_from_int = 0;
-
- atomic_clear_mask(ZFCP_STATUS_ADAPTER_QDIOUP, &adapter->status);
out:
return retval;
}
ZFCP_LOG_DEBUG("host connection still initialising... "
"waiting and retrying...\n");
/* sleep a little bit before retry */
- set_current_state(TASK_INTERRUPTIBLE);
- schedule_timeout(ZFCP_EXCHANGE_CONFIG_DATA_SLEEP);
+ msleep(jiffies_to_msecs(ZFCP_EXCHANGE_CONFIG_DATA_SLEEP));
}
} while ((retries--) &&
atomic_test_mask(ZFCP_STATUS_ADAPTER_HOST_CON_INIT,
{
int retval;
- if (atomic_test_mask(ZFCP_STATUS_PORT_NAMESERVER,
+ if (atomic_test_mask(ZFCP_STATUS_PORT_WKA,
&erp_action->port->status))
retval = zfcp_erp_port_strategy_open_nameserver(erp_action);
else
case ZFCP_ERP_STEP_PORT_OPENING:
if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &port->status)) {
- ZFCP_LOG_DEBUG("nameserver port is open\n");
+ ZFCP_LOG_DEBUG("WKA port is open\n");
retval = ZFCP_ERP_SUCCEEDED;
} else {
- ZFCP_LOG_DEBUG("open failed for nameserver port\n");
+ ZFCP_LOG_DEBUG("open failed for WKA port\n");
retval = ZFCP_ERP_FAILED;
}
/* this is needed anyway (dont care for retval of wakeup) */
git://git.onelab.eu / linux-2.6.git / blobdiff