[SCSI] mpt2sas: fix for driver fails EEH, recovery from injected pci bus error

This patch stops the driver to invoke kthread (which remove the dead ioc)
for some time while EEH recovery has started.

[thenzl: add a 'non_operational_loop' reset.]
Signed-off-by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
Signed-off-by: Tomas Henzl <thenzl@redhat.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

diff --git a/drivers/scsi/mpt2sas/mpt2sas_base.c b/drivers/scsi/mpt2sas/mpt2sas_base.c
index ffd85c5..5e24e7e 100644 (file)
--- a/drivers/scsi/mpt2sas/mpt2sas_base.c
+++ b/drivers/scsi/mpt2sas/mpt2sas_base.c
@@ -155,7 +155,7 @@ _base_fault_reset_work(struct work_struct *work)
        struct task_struct *p;
 
        spin_lock_irqsave(&ioc->ioc_reset_in_progress_lock, flags);
-       if (ioc->shost_recovery)
+       if (ioc->shost_recovery || ioc->pci_error_recovery)
                goto rearm_timer;
        spin_unlock_irqrestore(&ioc->ioc_reset_in_progress_lock, flags);
 
@@ -164,6 +164,20 @@ _base_fault_reset_work(struct work_struct *work)
                printk(MPT2SAS_INFO_FMT "%s : SAS host is non-operational !!!!\n",
                        ioc->name, __func__);
 
+               /* It may be possible that EEH recovery can resolve some of
+                * pci bus failure issues rather removing the dead ioc function
+                * by considering controller is in a non-operational state. So
+                * here priority is given to the EEH recovery. If it doesn't
+                * not resolve this issue, mpt2sas driver will consider this
+                * controller to non-operational state and remove the dead ioc
+                * function.
+                */
+               if (ioc->non_operational_loop++ < 5) {
+                       spin_lock_irqsave(&ioc->ioc_reset_in_progress_lock,
+                                                        flags);
+                       goto rearm_timer;
+               }
+
                /*
                 * Call _scsih_flush_pending_cmds callback so that we flush all
                 * pending commands back to OS. This call is required to aovid
@@ -193,6 +207,8 @@ _base_fault_reset_work(struct work_struct *work)
                return; /* don't rearm timer */
        }
 
+       ioc->non_operational_loop = 0;
+
        if ((doorbell & MPI2_IOC_STATE_MASK) == MPI2_IOC_STATE_FAULT) {
                rc = mpt2sas_base_hard_reset_handler(ioc, CAN_SLEEP,
                    FORCE_BIG_HAMMER);
@@ -4386,6 +4402,7 @@ mpt2sas_base_attach(struct MPT2SAS_ADAPTER *ioc)
        if (missing_delay[0] != -1 && missing_delay[1] != -1)
                _base_update_missing_delay(ioc, missing_delay[0],
                    missing_delay[1]);
+       ioc->non_operational_loop = 0;
 
        return 0;
 

diff --git a/drivers/scsi/mpt2sas/mpt2sas_base.h b/drivers/scsi/mpt2sas/mpt2sas_base.h
index 543d8d6..c6ee7aa 100644 (file)
--- a/drivers/scsi/mpt2sas/mpt2sas_base.h
+++ b/drivers/scsi/mpt2sas/mpt2sas_base.h
@@ -835,6 +835,7 @@ struct MPT2SAS_ADAPTER {
        u16             cpu_msix_table_sz;
        u32             ioc_reset_count;
        MPT2SAS_FLUSH_RUNNING_CMDS schedule_dead_ioc_flush_running_cmds;
+       u32             non_operational_loop;
 
        /* internal commands, callback index */
        u8              scsi_io_cb_idx;