* [PATCH 1/5] libiscsi: add more informative failure message during iscsi scsi eh
2010-12-31 8:22 iscsi changes for 2.6.38 michaelc
@ 2010-12-31 8:22 ` michaelc
2010-12-31 8:22 ` [PATCH 2/5] be2iscsi: fix gfp use in alloc_pdu michaelc
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: michaelc @ 2010-12-31 8:22 UTC (permalink / raw)
To: linux-scsi; +Cc: Mike Christie
From: Mike Christie <michaelc@cs.wisc.edu>
This adds a more informative error code and message
for the iscsi scsi eh session drop paths. This allows
you to distinguish if the session was dropped due to
a connection failure vs the iscsi layer dropping
the session due to scsi eh failure processing.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
---
drivers/scsi/libiscsi.c | 10 +++++-----
include/scsi/iscsi_if.h | 1 +
2 files changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c
index c15fde8..434d9ae 100644
--- a/drivers/scsi/libiscsi.c
+++ b/drivers/scsi/libiscsi.c
@@ -1795,9 +1795,9 @@ static int iscsi_exec_task_mgmt_fn(struct iscsi_conn *conn,
NULL, 0);
if (!task) {
spin_unlock_bh(&session->lock);
+ iscsi_conn_printk(KERN_ERR, conn, "Could not send TMF.\n");
iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
spin_lock_bh(&session->lock);
- ISCSI_DBG_EH(session, "tmf exec failure\n");
return -EPERM;
}
conn->tmfcmd_pdus_cnt++;
@@ -2202,7 +2202,7 @@ int iscsi_eh_abort(struct scsi_cmnd *sc)
goto success_unlocked;
case TMF_TIMEDOUT:
spin_unlock_bh(&session->lock);
- iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+ iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
goto failed_unlocked;
case TMF_NOT_FOUND:
if (!sc->SCp.ptr) {
@@ -2289,7 +2289,7 @@ int iscsi_eh_device_reset(struct scsi_cmnd *sc)
break;
case TMF_TIMEDOUT:
spin_unlock_bh(&session->lock);
- iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+ iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
goto done;
default:
conn->tmf_state = TMF_INITIAL;
@@ -2370,7 +2370,7 @@ failed:
* we drop the lock here but the leadconn cannot be destoyed while
* we are in the scsi eh
*/
- iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+ iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
ISCSI_DBG_EH(session, "wait for relogin\n");
wait_event_interruptible(conn->ehwait,
@@ -2452,7 +2452,7 @@ int iscsi_eh_target_reset(struct scsi_cmnd *sc)
break;
case TMF_TIMEDOUT:
spin_unlock_bh(&session->lock);
- iscsi_conn_failure(conn, ISCSI_ERR_CONN_FAILED);
+ iscsi_conn_failure(conn, ISCSI_ERR_SCSI_EH_SESSION_RST);
goto done;
default:
conn->tmf_state = TMF_INITIAL;
diff --git a/include/scsi/iscsi_if.h b/include/scsi/iscsi_if.h
index a8631ac..c3e1cbc 100644
--- a/include/scsi/iscsi_if.h
+++ b/include/scsi/iscsi_if.h
@@ -263,6 +263,7 @@ enum iscsi_err {
ISCSI_ERR_INVALID_HOST = ISCSI_ERR_BASE + 18,
ISCSI_ERR_XMIT_FAILED = ISCSI_ERR_BASE + 19,
ISCSI_ERR_TCP_CONN_CLOSE = ISCSI_ERR_BASE + 20,
+ ISCSI_ERR_SCSI_EH_SESSION_RST = ISCSI_ERR_BASE + 21,
};
/*
--
1.7.0.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 2/5] be2iscsi: fix gfp use in alloc_pdu
2010-12-31 8:22 iscsi changes for 2.6.38 michaelc
2010-12-31 8:22 ` [PATCH 1/5] libiscsi: add more informative failure message during iscsi scsi eh michaelc
@ 2010-12-31 8:22 ` michaelc
2010-12-31 8:22 ` [PATCH 3/5] be2iscsi: fix null ptr when accessing task hdr michaelc
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: michaelc @ 2010-12-31 8:22 UTC (permalink / raw)
To: linux-scsi; +Cc: Mike Christie
From: Mike Christie <michaelc@cs.wisc.edu>
The pdu allication callout is called from a spin lock
and in the IO path so we cannot use GFP_KERNEL. This
has the driver use GFP_ATOMIC.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
---
drivers/scsi/be2iscsi/be_main.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index 75a85aa..be07ca0 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -3785,7 +3785,7 @@ static int beiscsi_alloc_pdu(struct iscsi_task *task, uint8_t opcode)
dma_addr_t paddr;
io_task->cmd_bhs = pci_pool_alloc(beiscsi_sess->bhs_pool,
- GFP_KERNEL, &paddr);
+ GFP_ATOMIC, &paddr);
if (!io_task->cmd_bhs)
return -ENOMEM;
io_task->bhs_pa.u.a64.address = paddr;
--
1.7.0.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 3/5] be2iscsi: fix null ptr when accessing task hdr
2010-12-31 8:22 iscsi changes for 2.6.38 michaelc
2010-12-31 8:22 ` [PATCH 1/5] libiscsi: add more informative failure message during iscsi scsi eh michaelc
2010-12-31 8:22 ` [PATCH 2/5] be2iscsi: fix gfp use in alloc_pdu michaelc
@ 2010-12-31 8:22 ` michaelc
2010-12-31 8:22 ` [PATCH 4/5] libiscsi: do not take host lock in queuecommand michaelc
2010-12-31 8:22 ` [PATCH 5/5] libiscsi: use bh locking instead of irq with session lock michaelc
4 siblings, 0 replies; 6+ messages in thread
From: michaelc @ 2010-12-31 8:22 UTC (permalink / raw)
To: linux-scsi; +Cc: Mike Christie
From: Mike Christie <michaelc@cs.wisc.edu>
If alloc_pdu fails then the task->hdr pointer may not be
set. This adds a check for this case in the cleanup callback.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
---
drivers/scsi/be2iscsi/be_main.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)
diff --git a/drivers/scsi/be2iscsi/be_main.c b/drivers/scsi/be2iscsi/be_main.c
index be07ca0..79cefbe 100644
--- a/drivers/scsi/be2iscsi/be_main.c
+++ b/drivers/scsi/be2iscsi/be_main.c
@@ -3914,7 +3914,8 @@ static void beiscsi_cleanup_task(struct iscsi_task *task)
io_task->psgl_handle = NULL;
}
} else {
- if ((task->hdr->opcode & ISCSI_OPCODE_MASK) == ISCSI_OP_LOGIN)
+ if (task->hdr &&
+ ((task->hdr->opcode & ISCSI_OPCODE_MASK) == ISCSI_OP_LOGIN))
return;
if (io_task->psgl_handle) {
spin_lock(&phba->mgmt_sgl_lock);
--
1.7.0.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 4/5] libiscsi: do not take host lock in queuecommand
2010-12-31 8:22 iscsi changes for 2.6.38 michaelc
` (2 preceding siblings ...)
2010-12-31 8:22 ` [PATCH 3/5] be2iscsi: fix null ptr when accessing task hdr michaelc
@ 2010-12-31 8:22 ` michaelc
2010-12-31 8:22 ` [PATCH 5/5] libiscsi: use bh locking instead of irq with session lock michaelc
4 siblings, 0 replies; 6+ messages in thread
From: michaelc @ 2010-12-31 8:22 UTC (permalink / raw)
To: linux-scsi; +Cc: Mike Christie
From: Mike Christie <michaelc@cs.wisc.edu>
iscsi_tcp, ib_iser, cxgb*, be2iscsi and bnx2i do not use
the host lock and do not take the session lock against
a irq, so this patch drops the DEF_SCSI_QCMD use. Instead
we just take the session lock and disable bhs.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
---
drivers/scsi/libiscsi.c | 37 ++++++++++++++-----------------------
include/scsi/libiscsi.h | 3 ++-
2 files changed, 16 insertions(+), 24 deletions(-)
diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c
index 434d9ae..f460637 100644
--- a/drivers/scsi/libiscsi.c
+++ b/drivers/scsi/libiscsi.c
@@ -505,6 +505,7 @@ static void iscsi_free_task(struct iscsi_task *task)
struct iscsi_conn *conn = task->conn;
struct iscsi_session *session = conn->session;
struct scsi_cmnd *sc = task->sc;
+ int oldstate = task->state;
ISCSI_DBG_SESSION(session, "freeing task itt 0x%x state %d sc %p\n",
task->itt, task->state, task->sc);
@@ -525,10 +526,10 @@ static void iscsi_free_task(struct iscsi_task *task)
/* SCSI eh reuses commands to verify us */
sc->SCp.ptr = NULL;
/*
- * queue command may call this to free the task, but
- * not have setup the sc callback
+ * queue command may call this to free the task, so
+ * it will decide how to return sc to scsi-ml.
*/
- if (sc->scsi_done)
+ if (oldstate != ISCSI_TASK_REQUEUE_SCSIQ)
sc->scsi_done(sc);
}
}
@@ -571,7 +572,8 @@ static void iscsi_complete_task(struct iscsi_task *task, int state)
task->itt, task->state, task->sc);
if (task->state == ISCSI_TASK_COMPLETED ||
task->state == ISCSI_TASK_ABRT_TMF ||
- task->state == ISCSI_TASK_ABRT_SESS_RECOV)
+ task->state == ISCSI_TASK_ABRT_SESS_RECOV ||
+ task->state == ISCSI_TASK_REQUEUE_SCSIQ)
return;
WARN_ON_ONCE(task->state == ISCSI_TASK_FREE);
task->state = state;
@@ -1599,27 +1601,23 @@ enum {
FAILURE_SESSION_NOT_READY,
};
-static int iscsi_queuecommand_lck(struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
+int iscsi_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *sc)
{
struct iscsi_cls_session *cls_session;
- struct Scsi_Host *host;
struct iscsi_host *ihost;
int reason = 0;
struct iscsi_session *session;
struct iscsi_conn *conn;
struct iscsi_task *task = NULL;
- sc->scsi_done = done;
sc->result = 0;
sc->SCp.ptr = NULL;
- host = sc->device->host;
ihost = shost_priv(host);
- spin_unlock(host->host_lock);
cls_session = starget_to_session(scsi_target(sc->device));
session = cls_session->dd_data;
- spin_lock(&session->lock);
+ spin_lock_bh(&session->lock);
reason = iscsi_session_chkready(cls_session);
if (reason) {
@@ -1705,25 +1703,21 @@ static int iscsi_queuecommand_lck(struct scsi_cmnd *sc, void (*done)(struct scsi
}
session->queued_cmdsn++;
- spin_unlock(&session->lock);
- spin_lock(host->host_lock);
+ spin_unlock_bh(&session->lock);
return 0;
prepd_reject:
- sc->scsi_done = NULL;
- iscsi_complete_task(task, ISCSI_TASK_COMPLETED);
+ iscsi_complete_task(task, ISCSI_TASK_REQUEUE_SCSIQ);
reject:
- spin_unlock(&session->lock);
+ spin_unlock_bh(&session->lock);
ISCSI_DBG_SESSION(session, "cmd 0x%x rejected (%d)\n",
sc->cmnd[0], reason);
- spin_lock(host->host_lock);
return SCSI_MLQUEUE_TARGET_BUSY;
prepd_fault:
- sc->scsi_done = NULL;
- iscsi_complete_task(task, ISCSI_TASK_COMPLETED);
+ iscsi_complete_task(task, ISCSI_TASK_REQUEUE_SCSIQ);
fault:
- spin_unlock(&session->lock);
+ spin_unlock_bh(&session->lock);
ISCSI_DBG_SESSION(session, "iscsi: cmd 0x%x is not queued (%d)\n",
sc->cmnd[0], reason);
if (!scsi_bidi_cmnd(sc))
@@ -1732,12 +1726,9 @@ fault:
scsi_out(sc)->resid = scsi_out(sc)->length;
scsi_in(sc)->resid = scsi_in(sc)->length;
}
- done(sc);
- spin_lock(host->host_lock);
+ sc->scsi_done(sc);
return 0;
}
-
-DEF_SCSI_QCMD(iscsi_queuecommand)
EXPORT_SYMBOL_GPL(iscsi_queuecommand);
int iscsi_change_queue_depth(struct scsi_device *sdev, int depth, int reason)
diff --git a/include/scsi/libiscsi.h b/include/scsi/libiscsi.h
index b81d969..f1ff600 100644
--- a/include/scsi/libiscsi.h
+++ b/include/scsi/libiscsi.h
@@ -89,6 +89,7 @@ enum {
ISCSI_TASK_RUNNING,
ISCSI_TASK_ABRT_TMF, /* aborted due to TMF */
ISCSI_TASK_ABRT_SESS_RECOV, /* aborted due to session recovery */
+ ISCSI_TASK_REQUEUE_SCSIQ, /* qcmd requeueing to scsi-ml */
};
struct iscsi_r2t_info {
@@ -341,7 +342,7 @@ extern int iscsi_eh_abort(struct scsi_cmnd *sc);
extern int iscsi_eh_recover_target(struct scsi_cmnd *sc);
extern int iscsi_eh_session_reset(struct scsi_cmnd *sc);
extern int iscsi_eh_device_reset(struct scsi_cmnd *sc);
-extern int iscsi_queuecommand(struct Scsi_Host *h, struct scsi_cmnd *sc);
+extern int iscsi_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *sc);
/*
* iSCSI host helpers.
--
1.7.0.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 5/5] libiscsi: use bh locking instead of irq with session lock
2010-12-31 8:22 iscsi changes for 2.6.38 michaelc
` (3 preceding siblings ...)
2010-12-31 8:22 ` [PATCH 4/5] libiscsi: do not take host lock in queuecommand michaelc
@ 2010-12-31 8:22 ` michaelc
4 siblings, 0 replies; 6+ messages in thread
From: michaelc @ 2010-12-31 8:22 UTC (permalink / raw)
To: linux-scsi; +Cc: Mike Christie
From: Mike Christie <michaelc@cs.wisc.edu>
The session lock is taken in threads, timers, and bottom halves
like softirqs and tasklets. All the code but
iscsi_conn/session_failure take the session lock with the spin_lock_bh
call. This was done because I thought some offload drivers
would be calling these functions from a irq. They never did,
so this patch has iscsi_conn/session_failure use the bh
locking.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
---
drivers/scsi/libiscsi.c | 14 ++++++--------
1 files changed, 6 insertions(+), 8 deletions(-)
diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c
index f460637..b0481a8 100644
--- a/drivers/scsi/libiscsi.c
+++ b/drivers/scsi/libiscsi.c
@@ -1337,17 +1337,16 @@ void iscsi_session_failure(struct iscsi_session *session,
{
struct iscsi_conn *conn;
struct device *dev;
- unsigned long flags;
- spin_lock_irqsave(&session->lock, flags);
+ spin_lock_bh(&session->lock);
conn = session->leadconn;
if (session->state == ISCSI_STATE_TERMINATE || !conn) {
- spin_unlock_irqrestore(&session->lock, flags);
+ spin_unlock_bh(&session->lock);
return;
}
dev = get_device(&conn->cls_conn->dev);
- spin_unlock_irqrestore(&session->lock, flags);
+ spin_unlock_bh(&session->lock);
if (!dev)
return;
/*
@@ -1366,17 +1365,16 @@ EXPORT_SYMBOL_GPL(iscsi_session_failure);
void iscsi_conn_failure(struct iscsi_conn *conn, enum iscsi_err err)
{
struct iscsi_session *session = conn->session;
- unsigned long flags;
- spin_lock_irqsave(&session->lock, flags);
+ spin_lock_bh(&session->lock);
if (session->state == ISCSI_STATE_FAILED) {
- spin_unlock_irqrestore(&session->lock, flags);
+ spin_unlock_bh(&session->lock);
return;
}
if (conn->stop_stage == 0)
session->state = ISCSI_STATE_FAILED;
- spin_unlock_irqrestore(&session->lock, flags);
+ spin_unlock_bh(&session->lock);
set_bit(ISCSI_SUSPEND_BIT, &conn->suspend_tx);
set_bit(ISCSI_SUSPEND_BIT, &conn->suspend_rx);
--
1.7.0.1
^ permalink raw reply related [flat|nested] 6+ messages in thread