From: James Smart <jsmart2021@gmail.com>
Date: Mon, 4 Jan 2021 10:02:38 -0800
Subject: scsi: lpfc: Implement health checking when aborting I/O
Patch-mainline: Queued in subsystem maintainer repository
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi.git
Git-commit: a22d73b655a8ec6d41f08790e28ee19dc55d0d33
References: bsc#1180891
Several errors have occurred where the adapter stops or fails but does not
raise the register values for the driver to detect failure. Thus driver is
unaware of the failure. The failure typically results in I/O timeouts, the
I/O timeout handler failing (after several seconds), and the error handler
escalating recovery policy and resulting in more errors. Eventually, the
driver is in a position where things have spiraled and it can't do recovery
because other recovery ops are still outstanding and it becomes unusable.
Resolve the situation by having the I/O timeout handler (actually a els,
SCSI I/O, NVMe ls, or NVMe I/O timeout), in addition to aborting the I/O,
perform a mailbox command and look for a response from the hardware. If
the mailbox command fails, it will mark the adapter offline and then invoke
the adapter reset handler to clean up.
The new I/O timeout test will be limited to a test every 5s. If there are
multiple I/O timeouts concurrently, only the 1st I/O timeout will generate
the mailbox command. Further testing will only occur once a timeout occurs
after a 5s delay from the last mailbox command has expired.
Link: https://lore.kernel.org/r/20210104180240.46824-14-jsmart2021@gmail.com
Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-by: Daniel Wagner <dwagner@suse.de>
---
drivers/scsi/lpfc/lpfc.h | 3
drivers/scsi/lpfc/lpfc_attr.c | 2
drivers/scsi/lpfc/lpfc_crtn.h | 2
drivers/scsi/lpfc/lpfc_els.c | 9 +
drivers/scsi/lpfc/lpfc_hbadisc.c | 3
drivers/scsi/lpfc/lpfc_init.c | 177 ++++++++++++++++++++++---------------
drivers/scsi/lpfc/lpfc_nportdisc.c | 2
drivers/scsi/lpfc/lpfc_nvme.c | 8 +
drivers/scsi/lpfc/lpfc_scsi.c | 3
drivers/scsi/lpfc/lpfc_sli.c | 44 ++++++++-
10 files changed, 178 insertions(+), 75 deletions(-)
--- a/drivers/scsi/lpfc/lpfc.h
+++ b/drivers/scsi/lpfc/lpfc.h
@@ -780,6 +780,8 @@ struct lpfc_hba {
#define HBA_FLOGI_ISSUED 0x100000 /* FLOGI was issued */
#define HBA_DEFER_FLOGI 0x800000 /* Defer FLOGI till read_sparm cmpl */
#define HBA_NEEDS_CFG_PORT 0x2000000 /* SLI3 - needs a CONFIG_PORT mbox */
+#define HBA_HBEAT_INP 0x4000000 /* mbox HBEAT is in progress */
+#define HBA_HBEAT_TMO 0x8000000 /* HBEAT initiated after timeout */
uint32_t fcp_ring_in_use; /* When polling test if intr-hndlr active*/
struct lpfc_dmabuf slim2p;
@@ -1135,7 +1137,6 @@ struct lpfc_hba {
unsigned long last_completion_time;
unsigned long skipped_hb;
struct timer_list hb_tmofunc;
- uint8_t hb_outstanding;
struct timer_list rrq_tmr;
enum hba_temp_state over_temp_state;
/*
--- a/drivers/scsi/lpfc/lpfc_attr.c
+++ b/drivers/scsi/lpfc/lpfc_attr.c
@@ -1791,6 +1791,8 @@ lpfc_board_mode_store(struct device *dev
else if (strncmp(buf, "pci_bus_reset", sizeof("pci_bus_reset") - 1)
== 0)
status = lpfc_reset_pci_bus(phba);
+ else if (strncmp(buf, "heartbeat", sizeof("heartbeat") - 1) == 0)
+ lpfc_issue_hb_tmo(phba);
else if (strncmp(buf, "trunk", sizeof("trunk") - 1) == 0)
status = lpfc_set_trunking(phba, (char *)buf + sizeof("trunk"));
else
--- a/drivers/scsi/lpfc/lpfc_crtn.h
+++ b/drivers/scsi/lpfc/lpfc_crtn.h
@@ -360,6 +360,8 @@ lpfc_sli_abort_taskmgmt(struct lpfc_vpor
void lpfc_mbox_timeout(struct timer_list *t);
void lpfc_mbox_timeout_handler(struct lpfc_hba *);
+int lpfc_issue_hb_mbox(struct lpfc_hba *phba);
+void lpfc_issue_hb_tmo(struct lpfc_hba *phba);
struct lpfc_nodelist *lpfc_findnode_did(struct lpfc_vport *, uint32_t);
struct lpfc_nodelist *lpfc_findnode_wwpn(struct lpfc_vport *,
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -1428,6 +1428,9 @@ lpfc_els_abort_flogi(struct lpfc_hba *ph
NULL);
}
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
spin_unlock_irq(&phba->hbalock);
return 0;
@@ -8127,6 +8130,9 @@ lpfc_els_timeout_handler(struct lpfc_vpo
spin_unlock_irq(&phba->hbalock);
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
if (!list_empty(&pring->txcmplq))
if (!(phba->pport->load_flag & FC_UNLOADING))
mod_timer(&vport->els_tmofunc,
@@ -8226,6 +8232,9 @@ lpfc_els_flush_cmd(struct lpfc_vport *vp
lpfc_sli_issue_abort_iotag(phba, pring, piocb, NULL);
spin_unlock_irqrestore(&phba->hbalock, iflags);
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
if (!list_empty(&abort_list))
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"3387 abort list for txq not empty\n");
--- a/drivers/scsi/lpfc/lpfc_hbadisc.c
+++ b/drivers/scsi/lpfc/lpfc_hbadisc.c
@@ -5616,6 +5616,9 @@ lpfc_free_tx(struct lpfc_hba *phba, stru
}
spin_unlock_irq(&phba->hbalock);
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
/* Cancel all the IOCBs from the completions list */
lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
IOERR_SLI_ABORTED);
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -594,7 +594,7 @@ lpfc_config_port_post(struct lpfc_hba *p
/* Set up heart beat (HB) timer */
mod_timer(&phba->hb_tmofunc,
jiffies + msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL));
- phba->hb_outstanding = 0;
+ phba->hba_flag &= ~(HBA_HBEAT_INP | HBA_HBEAT_TMO);
phba->last_completion_time = jiffies;
/* Set up error attention (ERATT) polling timer */
mod_timer(&phba->eratt_poll,
@@ -1207,10 +1207,10 @@ lpfc_hb_mbox_cmpl(struct lpfc_hba * phba
unsigned long drvr_flag;
spin_lock_irqsave(&phba->hbalock, drvr_flag);
- phba->hb_outstanding = 0;
+ phba->hba_flag &= ~(HBA_HBEAT_INP | HBA_HBEAT_TMO);
spin_unlock_irqrestore(&phba->hbalock, drvr_flag);
- /* Check and reset heart-beat timer is necessary */
+ /* Check and reset heart-beat timer if necessary */
mempool_free(pmboxq, phba->mbox_mem_pool);
if (!(phba->pport->fc_flag & FC_OFFLINE_MODE) &&
!(phba->link_state == LPFC_HBA_ERROR) &&
@@ -1384,6 +1384,60 @@ static void lpfc_hb_mxp_handler(struct l
}
/**
+ * lpfc_issue_hb_mbox - Issues heart-beat mailbox command
+ * @phba: pointer to lpfc hba data structure.
+ *
+ * If a HB mbox is not already in progrees, this routine will allocate
+ * a LPFC_MBOXQ_t, populate it with a MBX_HEARTBEAT (0x31) command,
+ * and issue it. The HBA_HBEAT_INP flag means the command is in progress.
+ **/
+int
+lpfc_issue_hb_mbox(struct lpfc_hba *phba)
+{
+ LPFC_MBOXQ_t *pmboxq;
+ int retval;
+
+ /* Is a Heartbeat mbox already in progress */
+ if (phba->hba_flag & HBA_HBEAT_INP)
+ return 0;
+
+ pmboxq = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
+ if (!pmboxq)
+ return -ENOMEM;
+
+ lpfc_heart_beat(phba, pmboxq);
+ pmboxq->mbox_cmpl = lpfc_hb_mbox_cmpl;
+ pmboxq->vport = phba->pport;
+ retval = lpfc_sli_issue_mbox(phba, pmboxq, MBX_NOWAIT);
+
+ if (retval != MBX_BUSY && retval != MBX_SUCCESS) {
+ mempool_free(pmboxq, phba->mbox_mem_pool);
+ return -ENXIO;
+ }
+ phba->hba_flag |= HBA_HBEAT_INP;
+
+ return 0;
+}
+
+/**
+ * lpfc_issue_hb_tmo - Signals heartbeat timer to issue mbox command
+ * @phba: pointer to lpfc hba data structure.
+ *
+ * The heartbeat timer (every 5 sec) will fire. If the HBA_HBEAT_TMO
+ * flag is set, it will force a MBX_HEARTBEAT mbox command, regardless
+ * of the value of lpfc_enable_hba_heartbeat.
+ * If lpfc_enable_hba_heartbeat is set, the timeout routine will always
+ * try to issue a MBX_HEARTBEAT mbox command.
+ **/
+void
+lpfc_issue_hb_tmo(struct lpfc_hba *phba)
+{
+ if (phba->cfg_enable_hba_heartbeat)
+ return;
+ phba->hba_flag |= HBA_HBEAT_TMO;
+}
+
+/**
* lpfc_hb_timeout_handler - The HBA-timer timeout handler
* @phba: pointer to lpfc hba data structure.
*
@@ -1403,9 +1457,9 @@ void
lpfc_hb_timeout_handler(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
- LPFC_MBOXQ_t *pmboxq;
struct lpfc_dmabuf *buf_ptr;
- int retval, i;
+ int retval = 0;
+ int i, tmo;
struct lpfc_sli *psli = &phba->sli;
LIST_HEAD(completions);
@@ -1427,24 +1481,6 @@ lpfc_hb_timeout_handler(struct lpfc_hba
(phba->pport->fc_flag & FC_OFFLINE_MODE))
return;
- spin_lock_irq(&phba->pport->work_port_lock);
-
- if (time_after(phba->last_completion_time +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL),
- jiffies)) {
- spin_unlock_irq(&phba->pport->work_port_lock);
- if (!phba->hb_outstanding)
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL));
- else
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_TIMEOUT));
- return;
- }
- spin_unlock_irq(&phba->pport->work_port_lock);
-
if (phba->elsbuf_cnt &&
(phba->elsbuf_cnt == phba->elsbuf_prev_cnt)) {
spin_lock_irq(&phba->hbalock);
@@ -1464,37 +1500,43 @@ lpfc_hb_timeout_handler(struct lpfc_hba
/* If there is no heart beat outstanding, issue a heartbeat command */
if (phba->cfg_enable_hba_heartbeat) {
- if (!phba->hb_outstanding) {
+ /* If IOs are completing, no need to issue a MBX_HEARTBEAT */
+ spin_lock_irq(&phba->pport->work_port_lock);
+ if (time_after(phba->last_completion_time +
+ msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL),
+ jiffies)) {
+ spin_unlock_irq(&phba->pport->work_port_lock);
+ if (phba->hba_flag & HBA_HBEAT_INP)
+ tmo = (1000 * LPFC_HB_MBOX_TIMEOUT);
+ else
+ tmo = (1000 * LPFC_HB_MBOX_INTERVAL);
+ goto out;
+ }
+ spin_unlock_irq(&phba->pport->work_port_lock);
+
+ /* Check if a MBX_HEARTBEAT is already in progress */
+ if (phba->hba_flag & HBA_HBEAT_INP) {
+ /*
+ * If heart beat timeout called with HBA_HBEAT_INP set
+ * we need to give the hb mailbox cmd a chance to
+ * complete or TMO.
+ */
+ lpfc_printf_log(phba, KERN_WARNING, LOG_INIT,
+ "0459 Adapter heartbeat still outstanding: "
+ "last compl time was %d ms.\n",
+ jiffies_to_msecs(jiffies
+ - phba->last_completion_time));
+ tmo = (1000 * LPFC_HB_MBOX_TIMEOUT);
+ } else {
if ((!(psli->sli_flag & LPFC_SLI_MBOX_ACTIVE)) &&
(list_empty(&psli->mboxq))) {
- pmboxq = mempool_alloc(phba->mbox_mem_pool,
- GFP_KERNEL);
- if (!pmboxq) {
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 *
- LPFC_HB_MBOX_INTERVAL));
- return;
- }
- lpfc_heart_beat(phba, pmboxq);
- pmboxq->mbox_cmpl = lpfc_hb_mbox_cmpl;
- pmboxq->vport = phba->pport;
- retval = lpfc_sli_issue_mbox(phba, pmboxq,
- MBX_NOWAIT);
-
- if (retval != MBX_BUSY &&
- retval != MBX_SUCCESS) {
- mempool_free(pmboxq,
- phba->mbox_mem_pool);
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 *
- LPFC_HB_MBOX_INTERVAL));
- return;
+ retval = lpfc_issue_hb_mbox(phba);
+ if (retval) {
+ tmo = (1000 * LPFC_HB_MBOX_INTERVAL);
+ goto out;
}
phba->skipped_hb = 0;
- phba->hb_outstanding = 1;
} else if (time_before_eq(phba->last_completion_time,
phba->skipped_hb)) {
lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
@@ -1505,30 +1547,23 @@ lpfc_hb_timeout_handler(struct lpfc_hba
} else
phba->skipped_hb = jiffies;
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_TIMEOUT));
- return;
- } else {
- /*
- * If heart beat timeout called with hb_outstanding set
- * we need to give the hb mailbox cmd a chance to
- * complete or TMO.
- */
- lpfc_printf_log(phba, KERN_WARNING, LOG_INIT,
- "0459 Adapter heartbeat still out"
- "standing:last compl time was %d ms.\n",
- jiffies_to_msecs(jiffies
- - phba->last_completion_time));
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_TIMEOUT));
+ tmo = (1000 * LPFC_HB_MBOX_TIMEOUT);
+ goto out;
}
} else {
- mod_timer(&phba->hb_tmofunc,
- jiffies +
- msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL));
+ /* Check to see if we want to force a MBX_HEARTBEAT */
+ if (phba->hba_flag & HBA_HBEAT_TMO) {
+ retval = lpfc_issue_hb_mbox(phba);
+ if (retval)
+ tmo = (1000 * LPFC_HB_MBOX_INTERVAL);
+ else
+ tmo = (1000 * LPFC_HB_MBOX_TIMEOUT);
+ goto out;
+ }
+ tmo = (1000 * LPFC_HB_MBOX_INTERVAL);
}
+out:
+ mod_timer(&phba->hb_tmofunc, jiffies + msecs_to_jiffies(tmo));
}
/**
@@ -2992,7 +3027,7 @@ lpfc_stop_hba_timers(struct lpfc_hba *ph
del_timer_sync(&phba->rrq_tmr);
phba->hba_flag &= ~HBA_RRQ_ACTIVE;
}
- phba->hb_outstanding = 0;
+ phba->hba_flag &= ~(HBA_HBEAT_INP | HBA_HBEAT_TMO);
switch (phba->pci_dev_grp) {
case LPFC_PCI_DEV_LP:
--- a/drivers/scsi/lpfc/lpfc_nportdisc.c
+++ b/drivers/scsi/lpfc/lpfc_nportdisc.c
@@ -252,6 +252,8 @@ lpfc_els_abort(struct lpfc_hba *phba, st
lpfc_sli_issue_abort_iotag(phba, pring, iocb, NULL);
spin_unlock_irq(&phba->hbalock);
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
INIT_LIST_HEAD(&abort_list);
--- a/drivers/scsi/lpfc/lpfc_nvme.c
+++ b/drivers/scsi/lpfc/lpfc_nvme.c
@@ -1699,6 +1699,10 @@ lpfc_nvme_fcp_abort(struct nvme_fc_local
spin_unlock(&lpfc_nbuf->buf_lock);
spin_unlock_irqrestore(&phba->hbalock, flags);
+
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
if (ret_val != WQE_SUCCESS) {
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"6137 Failed abts issue_wqe with status x%x "
@@ -2444,6 +2448,10 @@ lpfc_nvme_wait_for_io_drain(struct lpfc_
}
}
}
+
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
}
void
--- a/drivers/scsi/lpfc/lpfc_scsi.c
+++ b/drivers/scsi/lpfc/lpfc_scsi.c
@@ -5479,6 +5479,9 @@ lpfc_abort_handler(struct scsi_cmnd *cmn
lpfc_sli_abort_fcp_cmpl);
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
if (ret_val != IOCB_SUCCESS) {
/* Indicate the IO is not being aborted by the driver. */
lpfc_cmd->waitq = NULL;
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -4111,6 +4111,8 @@ lpfc_sli_abort_iocb_ring(struct lpfc_hba
lpfc_sli_issue_abort_iotag(phba, pring, iocb, NULL);
spin_unlock_irq(&phba->hbalock);
}
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
/* Cancel all the IOCBs from the completions list */
lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
@@ -7916,7 +7918,7 @@ lpfc_sli4_hba_setup(struct lpfc_hba *phb
/* Start heart beat timer */
mod_timer(&phba->hb_tmofunc,
jiffies + msecs_to_jiffies(1000 * LPFC_HB_MBOX_INTERVAL));
- phba->hb_outstanding = 0;
+ phba->hba_flag &= ~(HBA_HBEAT_INP | HBA_HBEAT_TMO);
phba->last_completion_time = jiffies;
/* start eq_delay heartbeat */
@@ -11090,6 +11092,9 @@ lpfc_sli_host_down(struct lpfc_vport *vp
}
spin_unlock_irqrestore(&phba->hbalock, flags);
+ /* Make sure HBA is alive */
+ lpfc_issue_hb_tmo(phba);
+
/* Cancel all the IOCBs from the completions list */
lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
IOERR_SLI_DOWN);
@@ -12901,7 +12906,21 @@ lpfc_sli_sp_intr_handler(int irq, void *
spin_unlock_irqrestore(
&phba->pport->work_port_lock,
iflag);
- lpfc_mbox_cmpl_put(phba, pmb);
+
+ /* Do NOT queue MBX_HEARTBEAT to the worker
+ * thread for processing.
+ */
+ if (pmbox->mbxCommand == MBX_HEARTBEAT) {
+ /* Process mbox now */
+ phba->sli.mbox_active = NULL;
+ phba->sli.sli_flag &=
+ ~LPFC_SLI_MBOX_ACTIVE;
+ if (pmb->mbox_cmpl)
+ pmb->mbox_cmpl(phba, pmb);
+ } else {
+ /* Queue to worker thread to process */
+ lpfc_mbox_cmpl_put(phba, pmb);
+ }
}
} else
spin_unlock_irqrestore(&phba->hbalock, iflag);
@@ -13497,7 +13516,26 @@ lpfc_sli4_sp_handle_mbox_event(struct lp
phba->pport->work_port_events &= ~WORKER_MBOX_TMO;
spin_unlock_irqrestore(&phba->pport->work_port_lock, iflags);
- /* There is mailbox completion work to do */
+ /* Do NOT queue MBX_HEARTBEAT to the worker thread for processing. */
+ if (pmbox->mbxCommand == MBX_HEARTBEAT) {
+ spin_lock_irqsave(&phba->hbalock, iflags);
+ /* Release the mailbox command posting token */
+ phba->sli.sli_flag &= ~LPFC_SLI_MBOX_ACTIVE;
+ phba->sli.mbox_active = NULL;
+ if (bf_get(lpfc_trailer_consumed, mcqe))
+ lpfc_sli4_mq_release(phba->sli4_hba.mbx_wq);
+ spin_unlock_irqrestore(&phba->hbalock, iflags);
+
+ /* Post the next mbox command, if there is one */
+ lpfc_sli4_post_async_mbox(phba);
+
+ /* Process cmpl now */
+ if (pmb->mbox_cmpl)
+ pmb->mbox_cmpl(phba, pmb);
+ return false;
+ }
+
+ /* There is mailbox completion work to queue to the worker thread */
spin_lock_irqsave(&phba->hbalock, iflags);
__lpfc_mbox_cmpl_put(phba, pmb);
phba->work_ha |= HA_MBATT;