From: Mike Christie <[EMAIL PROTECTED]>
fc_fcp.c must lock between a the fcoe recv threads and
its timer, so it should be using spin_lock_bh instead
of just spin_lock.
It looks like this should also be safe to call into send_frame
with the scsi pkt lock held using the bh lock call, because
the spin_lock_bh code will handle the case where the network
layer will also do spin_lock_bh/spin_unlock_bh calls.
Note that it looks like other drivers are not going to use fc_fcp.c,
so I only take into consideration fcoe's recv threads for this
patch and not any other possible drivers.
Signed-off-by: Mike Christie <[EMAIL PROTECTED]>
---
drivers/scsi/libfc/fc_fcp.c | 42 +++++++++++++++++-------------------------
1 files changed, 17 insertions(+), 25 deletions(-)
diff --git a/drivers/scsi/libfc/fc_fcp.c b/drivers/scsi/libfc/fc_fcp.c
index a5f7aba..6fb7d1b 100644
--- a/drivers/scsi/libfc/fc_fcp.c
+++ b/drivers/scsi/libfc/fc_fcp.c
@@ -253,17 +253,9 @@ static void fc_fcp_pkt_release(struct fc_fcp_pkt *sp)
*/
static inline int fc_fcp_lock_pkt(struct fc_fcp_pkt *fsp)
{
- /*
- * TODO mnc: locking is not right. This can be called
- * from a timer context so we need to stop bottom halves from the
- * thread caller.
- *
- * It can also be called while sending packets, which can result
- * in bh's being enabled and disabled.
- */
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
if (!fsp->cmd) {
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
FC_DBG("Invalid scsi cmd pointer on fcp packet.\n");
return -EINVAL;
}
@@ -274,7 +266,7 @@ static inline int fc_fcp_lock_pkt(struct fc_fcp_pkt *fsp)
static inline void fc_fcp_unlock_pkt(struct fc_fcp_pkt *fsp)
{
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
fc_fcp_pkt_release(fsp);
}
@@ -1020,10 +1012,10 @@ static int fc_fcp_pkt_abort(struct fc_lport *lp, struct
fc_fcp_pkt *fsp)
init_completion(&fsp->tm_done);
fsp->wait_for_comp = 1;
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
rc = wait_for_completion_timeout(&fsp->tm_done,
msecs_to_jiffies(FC_SCSI_TM_TOV));
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
if (fsp->seq_ptr) {
lp->tt.exch_done(fsp->seq_ptr);
@@ -1057,7 +1049,7 @@ static void fc_lun_reset_send(unsigned long data)
struct fc_rport *rport;
struct fc_rport_libfc_priv *rp;
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
if (fsp->state & FC_SRB_COMPL)
goto unlock;
@@ -1088,7 +1080,7 @@ retry:
setup_timer(&fsp->timer, fc_lun_reset_send, (unsigned long)fsp);
fc_fcp_timer_set(fsp, FC_SCSI_REC_TOV);
unlock:
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
}
static void fc_fcp_cleanup_lun_reset(struct fc_fcp_pkt *fsp)
@@ -1126,13 +1118,13 @@ static int fc_lun_reset(struct fc_lport *lp, struct
fc_fcp_pkt *fsp,
*/
rc = wait_for_completion_timeout(&fsp->tm_done, FC_SCSI_TM_TOV);
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
fsp->state |= FC_SRB_COMPL;
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
del_timer_sync(&fsp->timer);
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
if (fsp->seq_ptr) {
/* TODO:
* if the exch resp function is running and trying to grab
@@ -1144,7 +1136,7 @@ static int fc_lun_reset(struct fc_lport *lp, struct
fc_fcp_pkt *fsp,
fsp->seq_ptr = NULL;
}
fsp->wait_for_comp = 0;
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
if (!rc) {
FC_DBG("lun reset failed\n");
@@ -1167,7 +1159,7 @@ static void fc_tm_done(struct fc_seq *sp, struct fc_frame
*fp, void *arg)
{
struct fc_fcp_pkt *fsp = arg;
- spin_lock(&fsp->scsi_pkt_lock);
+ spin_lock_bh(&fsp->scsi_pkt_lock);
/*
* raced with eh timeout handler.
*
@@ -1176,7 +1168,7 @@ static void fc_tm_done(struct fc_seq *sp, struct fc_frame
*fp, void *arg)
*/
if ((fsp->state & FC_SRB_COMPL) || !fsp->seq_ptr ||
!fsp->wait_for_comp) {
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
return;
}
@@ -1185,7 +1177,7 @@ static void fc_tm_done(struct fc_seq *sp, struct fc_frame
*fp, void *arg)
* If there is an error just let it timeout.
* scsi-eh will escalate for us.
*/
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
return;
}
@@ -1193,7 +1185,7 @@ static void fc_tm_done(struct fc_seq *sp, struct fc_frame
*fp, void *arg)
fsp->seq_ptr = NULL;
fsp->lp->tt.exch_done(sp);
fc_frame_free(fp);
- spin_unlock(&fsp->scsi_pkt_lock);
+ spin_unlock_bh(&fsp->scsi_pkt_lock);
}
static void fc_fcp_cleanup_io(struct fc_fcp_pkt *fsp)
@@ -1790,9 +1782,9 @@ static void fc_io_compl(struct fc_fcp_pkt *sp)
sp->state |= FC_SRB_COMPL;
if (!(sp->state & FC_SRB_FCP_PROCESSING_TMO)) {
- spin_unlock(&sp->scsi_pkt_lock);
+ spin_unlock_bh(&sp->scsi_pkt_lock);
del_timer_sync(&sp->timer);
- spin_lock(&sp->scsi_pkt_lock);
+ spin_lock_bh(&sp->scsi_pkt_lock);
}
lp = sp->lp;
--
1.5.4.1
_______________________________________________
devel mailing list
[email protected]
http://www.open-fcoe.org/mailman/listinfo/devel
