[PATCH 1/2 Trusty SRU] lpfc: Fix premature release of rpi bit in bitmask
Tim Gardner
tim.gardner at canonical.com
Wed May 11 13:49:06 UTC 2016
From: James Smart <james.smart at emulex.com>
BugLink: http://bugs.launchpad.net/bugs/1580560
Currently, the driver plays off the fact that older sli4 adapters have a
different rpi access pattern that allowed for the rpi reference to be
released earlier in the teardown sequence, allowing the driver to recycle
the rpi value sooner. Newer sli4 adapters have a different access pattern that
requires us to wait for a later mailbox completion. This changes the put
call location on the newer sli4 adapters.
Symptoms of the error are "0110 ELS" and the "0372 iotag" errors.
Signed-off-by: Dick Kennedy <dick.kennedy at emulex.com>
Signed-off-by: James Smart <james.smart at emulex.com>
Reviewed-by: Hannes Reinecke <hare at suse.de>
Signed-off-by: James Bottomley <JBottomley at Odin.com>
(back ported from commit be6bb94100dc6803a530e20aad05360e6267f56b)
Signed-off-by: Tim Gardner <tim.gardner at canonical.com>
Conflicts:
drivers/scsi/lpfc/lpfc_hbadisc.c
---
drivers/scsi/lpfc/lpfc_crtn.h | 1 +
drivers/scsi/lpfc/lpfc_els.c | 5 ++++
drivers/scsi/lpfc/lpfc_hbadisc.c | 63 ++++++++++++++++++++++++++++++++++------
drivers/scsi/lpfc/lpfc_init.c | 24 +++++++++++++--
drivers/scsi/lpfc/lpfc_sli.c | 49 +++++++++++++++++++++++++++++--
5 files changed, 128 insertions(+), 14 deletions(-)
diff --git a/drivers/scsi/lpfc/lpfc_crtn.h b/drivers/scsi/lpfc/lpfc_crtn.h
index cda076a..d70bd61 100644
--- a/drivers/scsi/lpfc/lpfc_crtn.h
+++ b/drivers/scsi/lpfc/lpfc_crtn.h
@@ -278,6 +278,7 @@ void lpfc_sli_handle_slow_ring_event(struct lpfc_hba *,
struct lpfc_sli_ring *, uint32_t);
void lpfc_sli4_handle_received_buffer(struct lpfc_hba *, struct hbq_dmabuf *);
void lpfc_sli_def_mbox_cmpl(struct lpfc_hba *, LPFC_MBOXQ_t *);
+void lpfc_sli4_unreg_rpi_cmpl_clr(struct lpfc_hba *, LPFC_MBOXQ_t *);
int lpfc_sli_issue_iocb(struct lpfc_hba *, uint32_t,
struct lpfc_iocbq *, uint32_t);
void lpfc_sli_pcimem_bcopy(void *, void *, uint32_t);
diff --git a/drivers/scsi/lpfc/lpfc_els.c b/drivers/scsi/lpfc/lpfc_els.c
index 110445f..24ab05b 100644
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -3668,6 +3668,11 @@ lpfc_mbx_cmpl_dflt_rpi(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
kfree(mp);
mempool_free(pmb, phba->mbox_mem_pool);
if (ndlp) {
+ lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
+ "0006 rpi%x DID:%x flg:%x %d map:%x %p\n",
+ ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
if (NLP_CHK_NODE_ACT(ndlp)) {
lpfc_nlp_put(ndlp);
/* This is the end of the default RPI cleanup logic for
diff --git a/drivers/scsi/lpfc/lpfc_hbadisc.c b/drivers/scsi/lpfc/lpfc_hbadisc.c
index 883ea2d..6a572af 100644
--- a/drivers/scsi/lpfc/lpfc_hbadisc.c
+++ b/drivers/scsi/lpfc/lpfc_hbadisc.c
@@ -3413,6 +3413,11 @@ lpfc_mbx_cmpl_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
pmb->context1 = NULL;
pmb->context2 = NULL;
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+ "0002 rpi:%x DID:%x flg:%x %d map:%x %p\n",
+ ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
if (ndlp->nlp_flag & NLP_REG_LOGIN_SEND)
ndlp->nlp_flag &= ~NLP_REG_LOGIN_SEND;
@@ -3829,6 +3834,11 @@ out:
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+ "0003 rpi:%x DID:%x flg:%x %d map%x %p\n",
+ ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
if (vport->port_state < LPFC_VPORT_READY) {
/* Link up discovery requires Fabric registration. */
@@ -4215,8 +4225,15 @@ lpfc_enable_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
lpfc_initialize_node(vport, ndlp, did);
spin_unlock_irqrestore(&phba->ndlp_lock, flags);
- if (vport->phba->sli_rev == LPFC_SLI_REV4)
+ if (vport->phba->sli_rev == LPFC_SLI_REV4) {
ndlp->nlp_rpi = lpfc_sli4_alloc_rpi(vport->phba);
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
+ "0008 rpi:%x DID:%x flg:%x refcnt:%d "
+ "map:%x %p\n", ndlp->nlp_rpi, ndlp->nlp_DID,
+ ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
+ }
if (state != NLP_STE_UNUSED_NODE)
@@ -4241,9 +4258,12 @@ lpfc_drop_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
if (ndlp->nlp_state == NLP_STE_UNUSED_NODE)
return;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNUSED_NODE);
- if (vport->phba->sli_rev == LPFC_SLI_REV4)
+ if (vport->phba->sli_rev == LPFC_SLI_REV4) {
lpfc_cleanup_vports_rrqs(vport, ndlp);
- lpfc_nlp_put(ndlp);
+ lpfc_unreg_rpi(vport, ndlp);
+ } else {
+ lpfc_nlp_put(ndlp);
+ }
return;
}
@@ -4480,7 +4500,17 @@ lpfc_unreg_rpi(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
mbox->context1 = ndlp;
mbox->mbox_cmpl = lpfc_nlp_logo_unreg;
} else {
- mbox->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
+ if (phba->sli_rev == LPFC_SLI_REV4 &&
+ (!(vport->load_flag & FC_UNLOADING)) &&
+ (bf_get(lpfc_sli_intf_if_type,
+ &phba->sli4_hba.sli_intf) ==
+ LPFC_SLI_INTF_IF_TYPE_2)) {
+ mbox->context1 = lpfc_nlp_get(ndlp);
+ mbox->mbox_cmpl =
+ lpfc_sli4_unreg_rpi_cmpl_clr;
+ } else
+ mbox->mbox_cmpl =
+ lpfc_sli_def_mbox_cmpl;
}
rc = lpfc_sli_issue_mbox(phba, mbox, MBX_NOWAIT);
@@ -4706,6 +4736,11 @@ lpfc_nlp_remove(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
/* For this case we need to cleanup the default rpi
* allocated by the firmware.
*/
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
+ "0005 rpi:%x DID:%x flg:%x %d map:%x %p\n",
+ ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
if ((mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL))
!= NULL) {
rc = lpfc_reg_rpi(phba, vport->vpi, ndlp->nlp_DID,
@@ -5454,7 +5489,11 @@ lpfc_mbx_cmpl_fdmi_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
-
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+ "0004 rpi:%x DID:%x flg:%x %d map:%x %p\n",
+ ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
/*
* Start issuing Fabric-Device Management Interface (FDMI) command to
* 0xfffffa (FDMI well known port) or Delay issuing FDMI command if
@@ -5618,9 +5657,15 @@ lpfc_nlp_init(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
lpfc_initialize_node(vport, ndlp, did);
INIT_LIST_HEAD(&ndlp->nlp_listp);
- if (vport->phba->sli_rev == LPFC_SLI_REV4)
+ if (vport->phba->sli_rev == LPFC_SLI_REV4) {
ndlp->nlp_rpi = lpfc_sli4_alloc_rpi(vport->phba);
-
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
+ "0007 rpi:%x DID:%x flg:%x refcnt:%d "
+ "map:%x %p\n", ndlp->nlp_rpi, ndlp->nlp_DID,
+ ndlp->nlp_flag,
+ atomic_read(&ndlp->kref.refcount),
+ ndlp->nlp_usg_map, ndlp);
+ }
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_NODE,
"node init: did:x%x",
@@ -5646,9 +5691,9 @@ lpfc_nlp_release(struct kref *kref)
lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
"0279 lpfc_nlp_release: ndlp:x%p did %x "
- "usgmap:x%x refcnt:%d\n",
+ "usgmap:x%x refcnt:%d rpi:%x\n",
(void *)ndlp, ndlp->nlp_DID, ndlp->nlp_usg_map,
- atomic_read(&ndlp->kref.refcount));
+ atomic_read(&ndlp->kref.refcount), ndlp->nlp_rpi);
/* remove ndlp from action. */
lpfc_nlp_remove(ndlp->vport, ndlp);
diff --git a/drivers/scsi/lpfc/lpfc_init.c b/drivers/scsi/lpfc/lpfc_init.c
index 39dc1d4..e0f93b4 100644
--- a/drivers/scsi/lpfc/lpfc_init.c
+++ b/drivers/scsi/lpfc/lpfc_init.c
@@ -2634,9 +2634,19 @@ lpfc_sli4_node_prep(struct lpfc_hba *phba)
list_for_each_entry_safe(ndlp, next_ndlp,
&vports[i]->fc_nodes,
nlp_listp) {
- if (NLP_CHK_NODE_ACT(ndlp))
+ if (NLP_CHK_NODE_ACT(ndlp)) {
ndlp->nlp_rpi =
lpfc_sli4_alloc_rpi(phba);
+ lpfc_printf_vlog(ndlp->vport, KERN_INFO,
+ LOG_NODE,
+ "0009 rpi:%x DID:%x "
+ "flg:%x map:%x %p\n",
+ ndlp->nlp_rpi,
+ ndlp->nlp_DID,
+ ndlp->nlp_flag,
+ ndlp->nlp_usg_map,
+ ndlp);
+ }
}
}
}
@@ -2801,8 +2811,18 @@ lpfc_offline_prep(struct lpfc_hba *phba, int mbx_action)
* RPI. Get a new RPI when the adapter port
* comes back online.
*/
- if (phba->sli_rev == LPFC_SLI_REV4)
+ if (phba->sli_rev == LPFC_SLI_REV4) {
+ lpfc_printf_vlog(ndlp->vport,
+ KERN_INFO, LOG_NODE,
+ "0011 lpfc_offline: "
+ "ndlp:x%p did %x "
+ "usgmap:x%x rpi:%x\n",
+ ndlp, ndlp->nlp_DID,
+ ndlp->nlp_usg_map,
+ ndlp->nlp_rpi);
+
lpfc_sli4_free_rpi(phba, ndlp->nlp_rpi);
+ }
lpfc_unreg_rpi(vports[i], ndlp);
}
}
diff --git a/drivers/scsi/lpfc/lpfc_sli.c b/drivers/scsi/lpfc/lpfc_sli.c
index ce21132..796c1b5 100644
--- a/drivers/scsi/lpfc/lpfc_sli.c
+++ b/drivers/scsi/lpfc/lpfc_sli.c
@@ -2240,6 +2240,46 @@ lpfc_sli_def_mbox_cmpl(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
else
mempool_free(pmb, phba->mbox_mem_pool);
}
+ /**
+ * lpfc_sli4_unreg_rpi_cmpl_clr - mailbox completion handler
+ * @phba: Pointer to HBA context object.
+ * @pmb: Pointer to mailbox object.
+ *
+ * This function is the unreg rpi mailbox completion handler. It
+ * frees the memory resources associated with the completed mailbox
+ * command. An additional refrenece is put on the ndlp to prevent
+ * lpfc_nlp_release from freeing the rpi bit in the bitmask before
+ * the unreg mailbox command completes, this routine puts the
+ * reference back.
+ *
+ **/
+void
+lpfc_sli4_unreg_rpi_cmpl_clr(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
+{
+ struct lpfc_vport *vport = pmb->vport;
+ struct lpfc_nodelist *ndlp;
+
+ ndlp = pmb->context1;
+ if (pmb->u.mb.mbxCommand == MBX_UNREG_LOGIN) {
+ if (phba->sli_rev == LPFC_SLI_REV4 &&
+ (bf_get(lpfc_sli_intf_if_type,
+ &phba->sli4_hba.sli_intf) ==
+ LPFC_SLI_INTF_IF_TYPE_2)) {
+ if (ndlp) {
+ lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
+ "0010 UNREG_LOGIN vpi:%x "
+ "rpi:%x DID:%x map:%x %p\n",
+ vport->vpi, ndlp->nlp_rpi,
+ ndlp->nlp_DID,
+ ndlp->nlp_usg_map, ndlp);
+
+ lpfc_nlp_put(ndlp);
+ }
+ }
+ }
+
+ mempool_free(pmb, phba->mbox_mem_pool);
+}
/**
* lpfc_sli_handle_mb_event - Handle mailbox completions from firmware
@@ -15281,14 +15321,14 @@ lpfc_sli4_alloc_rpi(struct lpfc_hba *phba)
struct lpfc_rpi_hdr *rpi_hdr;
unsigned long iflag;
- max_rpi = phba->sli4_hba.max_cfg_param.max_rpi;
- rpi_limit = phba->sli4_hba.next_rpi;
-
/*
* Fetch the next logical rpi. Because this index is logical,
* the driver starts at 0 each time.
*/
spin_lock_irqsave(&phba->hbalock, iflag);
+ max_rpi = phba->sli4_hba.max_cfg_param.max_rpi;
+ rpi_limit = phba->sli4_hba.next_rpi;
+
rpi = find_next_zero_bit(phba->sli4_hba.rpi_bmask, rpi_limit, 0);
if (rpi >= rpi_limit)
rpi = LPFC_RPI_ALLOC_ERROR;
@@ -15297,6 +15337,9 @@ lpfc_sli4_alloc_rpi(struct lpfc_hba *phba)
phba->sli4_hba.max_cfg_param.rpi_used++;
phba->sli4_hba.rpi_count++;
}
+ lpfc_printf_log(phba, KERN_INFO, LOG_SLI,
+ "0001 rpi:%x max:%x lim:%x\n",
+ (int) rpi, max_rpi, rpi_limit);
/*
* Don't try to allocate more rpi header regions if the device limit
--
1.9.1
More information about the kernel-team
mailing list