* [PATCH 1/4] fcoe: Fix deadlock between fip's recv_work and rtnl
2011-08-25 19:40 [PATCH 0/4] libfc and fcoe updates for scsi-misc Robert Love
@ 2011-08-25 19:40 ` Robert Love
2011-08-25 19:40 ` [PATCH 2/4] libfc: fix fc_eh_host_reset Robert Love
` (2 subsequent siblings)
3 siblings, 0 replies; 5+ messages in thread
From: Robert Love @ 2011-08-25 19:40 UTC (permalink / raw)
To: linux-scsi; +Cc: Ross Brattain, Yi Zou
The rtnl cannot be held durrng the fcoe_interface_put.
If it is the last reference on the fcoe_interface the
fcoe_ctlr_destroy will be called as a part of the
cleanup, ultimately calling cancel_work_sync(&fip->recv_work);
If we are processing a flogi response we will be in
the recv_work context and we will lock the rtnl to
add a new unicast MAC address. This is how the deadlock
can occur.
The fix is simply to move the rtnl_lock/unlock into
fcoe_interface_cleanup so that it can be unlocked before
fcoe_interface_put is called.
Here is the lockdep report:
Jul 21 11:26:35 bubba [ 223.870702]
ul 21 11:26:35 bubba [ 223.870704] =======================================================
Jul 21 11:26:35 bubba [ 223.871255] [ INFO: possible circular locking dependency detected ]
Jul 21 11:26:35 bubba [ 223.871530] 3.0.0-rc7+ #1
Jul 21 11:26:35 bubba [ 223.871797] -------------------------------------------------------
Jul 21 11:26:35 bubba [ 223.872072] lockdeptest.sh/3464 is trying to acquire lock:
Jul 21 11:26:35 bubba [ 223.872345] ((&fip->recv_work)
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff810531f1>] wait_on_work+0x0/0xbd
Jul 21 11:26:35 bubba [ 223.873022]
Jul 21 11:26:35 bubba [ 223.873023] but task is already holding lock:
Jul 21 11:26:35 bubba [ 223.873555] (rtnl_mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff813e8233>] rtnl_lock+0x12/0x14
Jul 21 11:26:35 bubba [ 223.874229]
Jul 21 11:26:35 bubba [ 223.874230] which lock already depends on the new lock.
Jul 21 11:26:35 bubba [ 223.874231]
Jul 21 11:26:35 bubba [ 223.875032]
Jul 21 11:26:35 bubba [ 223.875033] the existing dependency chain (in reverse order) is:
Jul 21 11:26:35 bubba [ 223.875573]
Jul 21 11:26:35 bubba [ 223.875573] -> #1
Jul 21 11:26:35 bubba (rtnl_mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba :
Jul 21 11:26:35 bubba [ 223.876301]
Jul 21 11:26:35 bubba [<ffffffff8106c14a>] lock_acquire+0xd2/0xf7
Jul 21 11:26:35 bubba [ 223.876645]
Jul 21 11:26:35 bubba [<ffffffff8151d975>] __mutex_lock_common+0x47/0x30d
Jul 21 11:26:35 bubba [ 223.876991]
Jul 21 11:26:35 bubba [<ffffffff8151dd36>] mutex_lock_nested+0x3b/0x40
Jul 21 11:26:35 bubba [ 223.877334]
Jul 21 11:26:35 bubba [<ffffffff813e8233>] rtnl_lock+0x12/0x14
Jul 21 11:26:35 bubba [ 223.877675]
Jul 21 11:26:35 bubba [<ffffffffa003d5a0>] fcoe_update_src_mac+0x2b/0x80 [fcoe]
Jul 21 11:26:35 bubba [ 223.878022]
Jul 21 11:26:35 bubba [<ffffffffa003d698>] fcoe_flogi_resp+0x5e/0x79 [fcoe]
Jul 21 11:26:35 bubba [ 223.878366]
Jul 21 11:26:35 bubba [<ffffffffa001566f>] fc_exch_recv+0x7f5/0x9da [libfc]
Jul 21 11:26:35 bubba [ 223.878713]
Jul 21 11:26:35 bubba [<ffffffffa00327d8>] fcoe_ctlr_recv_work+0x71f/0x10dc [libfcoe]
Jul 21 11:26:35 bubba [ 223.879258]
Jul 21 11:26:35 bubba [<ffffffff81053761>] process_one_work+0x1d7/0x347
Jul 21 11:26:35 bubba [ 223.879601]
Jul 21 11:26:35 bubba [<ffffffff81054ade>] worker_thread+0xf8/0x17c
Jul 21 11:26:35 bubba [ 223.879944]
Jul 21 11:26:35 bubba [<ffffffff81058184>] kthread+0x7d/0x85
Jul 21 11:26:35 bubba [ 223.880287]
Jul 21 11:26:35 bubba [<ffffffff81526414>] kernel_thread_helper+0x4/0x10
Jul 21 11:26:35 bubba [ 223.880634]
Jul 21 11:26:35 bubba [ 223.880635] -> #0
Jul 21 11:26:35 bubba ((&fip->recv_work)
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba :
Jul 21 11:26:35 bubba [ 223.881357]
Jul 21 11:26:35 bubba [<ffffffff8106b93e>] __lock_acquire+0xb1d/0xe2c
Jul 21 11:26:35 bubba [ 223.881695]
Jul 21 11:26:35 bubba [<ffffffff8106c14a>] lock_acquire+0xd2/0xf7
Jul 21 11:26:35 bubba [ 223.882033]
Jul 21 11:26:35 bubba [<ffffffff81053241>] wait_on_work+0x50/0xbd
Jul 21 11:26:35 bubba [ 223.882378]
Jul 21 11:26:35 bubba [<ffffffff81053b32>] __cancel_work_timer+0xb6/0xf4
Jul 21 11:26:35 bubba [ 223.882718]
Jul 21 11:26:35 bubba [<ffffffff81053b8a>] cancel_work_sync+0xb/0xd
Jul 21 11:26:35 bubba [ 223.883057]
Jul 21 11:26:35 bubba [<ffffffffa00317e6>] fcoe_ctlr_destroy+0x1d/0x67 [libfcoe]
Jul 21 11:26:35 bubba [ 223.883399]
Jul 21 11:26:35 bubba [<ffffffffa003e51e>] fcoe_interface_release+0x21/0x45 [fcoe]
Jul 21 11:26:35 bubba [ 223.883940]
Jul 21 11:26:35 bubba [<ffffffff811fbbe6>] kref_put+0x43/0x4d
Jul 21 11:26:35 bubba [ 223.884280]
Jul 21 11:26:35 bubba [<ffffffffa003ebba>] fcoe_interface_put+0x17/0x19 [fcoe]
Jul 21 11:26:35 bubba [ 223.884624]
Jul 21 11:26:35 bubba [<ffffffffa003f2a6>] fcoe_interface_cleanup+0x188/0x193 [fcoe]
Jul 21 11:26:35 bubba [ 223.885163]
Jul 21 11:26:35 bubba [<ffffffffa003f303>] fcoe_destroy+0x52/0x72 [fcoe]
Jul 21 11:26:35 bubba [ 223.885502]
Jul 21 11:26:35 bubba [<ffffffffa00340a4>] fcoe_transport_destroy+0xab/0x110 [libfcoe]
Jul 21 11:26:35 bubba [ 223.886045]
Jul 21 11:26:35 bubba [<ffffffff81056153>] param_attr_store+0x43/0x62
Jul 21 11:26:35 bubba [ 223.886385]
Jul 21 11:26:35 bubba [<ffffffff8105602d>] module_attr_store+0x21/0x25
Jul 21 11:26:35 bubba [ 223.886728]
Jul 21 11:26:35 bubba [<ffffffff8114c23d>] sysfs_write_file+0x103/0x13f
Jul 21 11:26:35 bubba [ 223.887068]
Jul 21 11:26:35 bubba [<ffffffff810f3e7b>] vfs_write+0xa7/0xfa
Jul 21 11:26:35 bubba [ 223.887406]
Jul 21 11:26:35 bubba [<ffffffff810f4073>] sys_write+0x45/0x69
Jul 21 11:26:35 bubba [ 223.887742]
Jul 21 11:26:35 bubba [<ffffffff815252bb>] system_call_fastpath+0x16/0x1b
Jul 21 11:26:35 bubba [ 223.888083]
Jul 21 11:26:35 bubba [ 223.888084] other info that might help us debug this:
Jul 21 11:26:35 bubba [ 223.888085]
Jul 21 11:26:35 bubba [ 223.888879] Possible unsafe locking scenario:
Jul 21 11:26:35 bubba [ 223.888881]
Jul 21 11:26:35 bubba [ 223.889411] CPU0 CPU1
Jul 21 11:26:35 bubba [ 223.889683] ---- ----
Jul 21 11:26:35 bubba [ 223.889955] lock(
Jul 21 11:26:35 bubba rtnl_mutex
Jul 21 11:26:35 bubba );
Jul 21 11:26:35 bubba [ 223.890349] lock(
Jul 21 11:26:35 bubba (&fip->recv_work)
Jul 21 11:26:35 bubba );
Jul 21 11:26:35 bubba [ 223.890751] lock(
Jul 21 11:26:35 bubba rtnl_mutex
Jul 21 11:26:35 bubba );
Jul 21 11:26:35 bubba [ 223.891154] lock(
Jul 21 11:26:35 bubba (&fip->recv_work)
Jul 21 11:26:35 bubba );
Jul 21 11:26:35 bubba [ 223.891549]
Jul 21 11:26:35 bubba [ 223.891550] *** DEADLOCK ***
Jul 21 11:26:35 bubba [ 223.891551]
Jul 21 11:26:35 bubba [ 223.892347] 6 locks held by lockdeptest.sh/3464:
Jul 21 11:26:35 bubba [ 223.892621] #0:
Jul 21 11:26:35 bubba (&buffer->mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff8114c171>] sysfs_write_file+0x37/0x13f
Jul 21 11:26:35 bubba [ 223.893359] #1:
Jul 21 11:26:35 bubba (s_active
Jul 21 11:26:35 bubba ){++++.+}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff8114c21c>] sysfs_write_file+0xe2/0x13f
Jul 21 11:26:35 bubba [ 223.894094] #2:
Jul 21 11:26:35 bubba (param_lock
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff81056146>] param_attr_store+0x36/0x62
Jul 21 11:26:35 bubba [ 223.894835] #3:
Jul 21 11:26:35 bubba (ft_mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffffa0034017>] fcoe_transport_destroy+0x1e/0x110 [libfcoe]
Jul 21 11:26:35 bubba [ 223.895574] #4:
Jul 21 11:26:35 bubba (fcoe_config_mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffffa003f2c9>] fcoe_destroy+0x18/0x72 [fcoe]
Jul 21 11:26:35 bubba [ 223.896314] #5:
Jul 21 11:26:35 bubba (rtnl_mutex
Jul 21 11:26:35 bubba ){+.+.+.}
Jul 21 11:26:35 bubba , at:
Jul 21 11:26:35 bubba [<ffffffff813e8233>] rtnl_lock+0x12/0x14
Jul 21 11:26:35 bubba [ 223.897047]
Jul 21 11:26:35 bubba [ 223.897048] stack backtrace:
Jul 21 11:26:35 bubba [ 223.897578] Pid: 3464, comm: lockdeptest.sh Not tainted 3.0.0-rc7+ #1
Jul 21 11:26:35 bubba [ 223.897853] Call Trace:
Jul 21 11:26:35 bubba [ 223.898128] [<ffffffff81068e16>] print_circular_bug+0x1f8/0x209
Jul 21 11:26:35 bubba [ 223.898416] [<ffffffff8106b93e>] __lock_acquire+0xb1d/0xe2c
Jul 21 11:26:35 bubba [ 223.898699] [<ffffffff810531f1>] ? wait_on_cpu_work+0xe6/0xe6
Jul 21 11:26:35 bubba [ 223.898982] [<ffffffff8106c14a>] lock_acquire+0xd2/0xf7
Jul 21 11:26:35 bubba [ 223.899263] [<ffffffff810531f1>] ? wait_on_cpu_work+0xe6/0xe6
Jul 21 11:26:35 bubba [ 223.899547] [<ffffffff8104a097>] ? mod_timer+0x8f/0x98
Jul 21 11:26:35 bubba [ 223.899827] [<ffffffff81053241>] wait_on_work+0x50/0xbd
Jul 21 11:26:35 bubba [ 223.900108] [<ffffffff810531f1>] ? wait_on_cpu_work+0xe6/0xe6
Jul 21 11:26:35 bubba [ 223.900390] [<ffffffff81053b32>] __cancel_work_timer+0xb6/0xf4
Jul 21 11:26:35 bubba [ 223.900671] [<ffffffff81053b8a>] cancel_work_sync+0xb/0xd
Jul 21 11:26:35 bubba [ 223.900953] [<ffffffffa00317e6>] fcoe_ctlr_destroy+0x1d/0x67 [libfcoe]
Jul 21 11:26:35 bubba [ 223.901237] [<ffffffffa003e51e>] fcoe_interface_release+0x21/0x45 [fcoe]
Jul 21 11:26:35 bubba [ 223.901522] [<ffffffffa003e4fd>] ? fcoe_enable+0x6b/0x6b [fcoe]
Jul 21 11:26:35 bubba [ 223.901803] [<ffffffff811fbbe6>] kref_put+0x43/0x4d
Jul 21 11:26:35 bubba [ 223.902083] [<ffffffffa003ebba>] fcoe_interface_put+0x17/0x19 [fcoe]
Jul 21 11:26:35 bubba [ 223.902367] [<ffffffffa003f2a6>] fcoe_interface_cleanup+0x188/0x193 [fcoe]
Jul 21 11:26:35 bubba [ 223.902653] [<ffffffff8151dd36>] ? mutex_lock_nested+0x3b/0x40
Jul 21 11:26:35 bubba [ 223.902939] [<ffffffffa003f303>] fcoe_destroy+0x52/0x72 [fcoe]
Jul 21 11:26:35 bubba [ 223.903223] [<ffffffffa00340a4>] fcoe_transport_destroy+0xab/0x110 [libfcoe]
Jul 21 11:26:35 bubba [ 223.903508] [<ffffffff81056153>] param_attr_store+0x43/0x62
Jul 21 11:26:35 bubba [ 223.903792] [<ffffffff8105602d>] module_attr_store+0x21/0x25
Jul 21 11:26:35 bubba [ 223.904075] [<ffffffff8114c23d>] sysfs_write_file+0x103/0x13f
Jul 21 11:26:35 bubba [ 223.904357] [<ffffffff810f3e7b>] vfs_write+0xa7/0xfa
Jul 21 11:26:35 bubba [ 223.904642] [<ffffffff810f51d6>] ? fget_light+0x35/0x96
Jul 21 11:26:35 bubba [ 223.904923] [<ffffffff810f4073>] sys_write+0x45/0x69
Jul 21 11:26:35 bubba [ 223.905204] [<ffffffff815252bb>] system_call_fastpath+0x16/0x1b
Jul 21 11:26:36 bubba [ 223.964438] ixgbe 0000:05:00.0: eth3: detected SFP+: 5
Jul 21 11:26:37 bubba [ 225.196702] ixgbe 0000:05:00.0: eth3: NIC Link is Up 10 Gbps, Flow Control: None
Signed-off-by: Robert Love <robert.w.love@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Reviewed-by: Yi Zou <yi.zou@intel.com>
---
drivers/scsi/fcoe/fcoe.c | 13 ++++++++-----
1 files changed, 8 insertions(+), 5 deletions(-)
diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c
index c30fa27..b5538e8 100644
--- a/drivers/scsi/fcoe/fcoe.c
+++ b/drivers/scsi/fcoe/fcoe.c
@@ -431,6 +431,8 @@ void fcoe_interface_cleanup(struct fcoe_interface *fcoe)
u8 flogi_maddr[ETH_ALEN];
const struct net_device_ops *ops;
+ rtnl_lock();
+
/*
* Don't listen for Ethernet packets anymore.
* synchronize_net() ensures that the packet handlers are not running
@@ -460,6 +462,8 @@ void fcoe_interface_cleanup(struct fcoe_interface *fcoe)
" specific feature for LLD.\n");
}
+ rtnl_unlock();
+
/* Release the self-reference taken during fcoe_interface_create() */
fcoe_interface_put(fcoe);
}
@@ -1933,11 +1937,8 @@ static void fcoe_destroy_work(struct work_struct *work)
fcoe_if_destroy(port->lport);
/* Do not tear down the fcoe interface for NPIV port */
- if (!npiv) {
- rtnl_lock();
+ if (!npiv)
fcoe_interface_cleanup(fcoe);
- rtnl_unlock();
- }
mutex_unlock(&fcoe_config_mutex);
}
@@ -1991,8 +1992,9 @@ static int fcoe_create(struct net_device *netdev, enum fip_state fip_mode)
printk(KERN_ERR "fcoe: Failed to create interface (%s)\n",
netdev->name);
rc = -EIO;
+ rtnl_unlock();
fcoe_interface_cleanup(fcoe);
- goto out_nodev;
+ goto out_nortnl;
}
/* Make this the "master" N_Port */
@@ -2009,6 +2011,7 @@ static int fcoe_create(struct net_device *netdev, enum fip_state fip_mode)
out_nodev:
rtnl_unlock();
+out_nortnl:
mutex_unlock(&fcoe_config_mutex);
return rc;
}
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 2/4] libfc: fix fc_eh_host_reset
2011-08-25 19:40 [PATCH 0/4] libfc and fcoe updates for scsi-misc Robert Love
2011-08-25 19:40 ` [PATCH 1/4] fcoe: Fix deadlock between fip's recv_work and rtnl Robert Love
@ 2011-08-25 19:40 ` Robert Love
2011-08-25 19:40 ` [PATCH 3/4] libfc: block SCSI eh thread for blocked rports Robert Love
2011-08-25 19:41 ` [PATCH 4/4] libfc: fix referencing to fc_fcp_pkt from the frame pointer via fr_fsp() Robert Love
3 siblings, 0 replies; 5+ messages in thread
From: Robert Love @ 2011-08-25 19:40 UTC (permalink / raw)
To: linux-scsi; +Cc: Ross Brattain, Vasu Dev, Yi Zou
From: Vasu Dev <vasu.dev@intel.com>
Current fc_eh_host_reset leaves lport offline
permanently due to FLOGI response getting
handled by LOGO response from last reset as both
had same exchange id.
So fix this by having end to end exches clean-up
using exchange abort along exches reset
done from fc_eh_host_reset. This would avoid
exchanges collision between the sessions across
the reset. In this case implicit login should have
done that but no aborting support for FIP
frames, so just wait till lport->r_a_tov before
restarting next flogi to ensure all exchanges
are good to use again for next session.
Below is the trace of LOGO from older session
coming ahead of FLOGI response with same exche id
0x203:-
617 86.435165 4e.00.0b -> ff.ff.fc FC ELS LOGO 0x203
618 86.435195 4e.00.0b -> b6.02.00 FC ELS LOGO 0x213
619 86.435220 4e.00.0b -> 18.03.00 FC ELS LOGO 0x223
620 86.435244 4e.00.0b -> 18.02.00 FC ELS LOGO 0x233
621 86.435267 4e.00.0b -> 18.01.00 FC ELS LOGO 0x243
622 86.435349 00.00.00 -> ff.ff.fe FC ELS FLOGI 0x203
623 86.435549 ff.ff.fc -> 4e.00.0b FC ELS ACC (LOGO) 0x203
624 86.438721 ff.ff.fe -> 4e.00.0b FC ELS ACC (FLOGI) 0x203
625 86.442059 18.03.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x223
626 86.443683 b6.02.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x213
627 86.447693 18.01.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x243
628 86.453499 18.02.00 -> 4e.00.0b FC ELS ACC (LOGO) 0x233
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Reviewed-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
---
drivers/scsi/libfc/fc_exch.c | 51 ++++++++++++++++++++++++++++-------------
drivers/scsi/libfc/fc_lport.c | 11 ++++++++-
2 files changed, 45 insertions(+), 17 deletions(-)
diff --git a/drivers/scsi/libfc/fc_exch.c b/drivers/scsi/libfc/fc_exch.c
index 01ff082..744fefe 100644
--- a/drivers/scsi/libfc/fc_exch.c
+++ b/drivers/scsi/libfc/fc_exch.c
@@ -494,6 +494,9 @@ static int fc_seq_send(struct fc_lport *lport, struct fc_seq *sp,
*/
error = lport->tt.frame_send(lport, fp);
+ if (fh->fh_type == FC_TYPE_BLS)
+ return error;
+
/*
* Update the exchange and sequence flags,
* assuming all frames for the sequence have been sent.
@@ -575,42 +578,35 @@ static void fc_seq_set_resp(struct fc_seq *sp,
}
/**
- * fc_seq_exch_abort() - Abort an exchange and sequence
- * @req_sp: The sequence to be aborted
+ * fc_exch_abort_locked() - Abort an exchange
+ * @ep: The exchange to be aborted
* @timer_msec: The period of time to wait before aborting
*
- * Generally called because of a timeout or an abort from the upper layer.
+ * Locking notes: Called with exch lock held
+ *
+ * Return value: 0 on success else error code
*/
-static int fc_seq_exch_abort(const struct fc_seq *req_sp,
- unsigned int timer_msec)
+static int fc_exch_abort_locked(struct fc_exch *ep,
+ unsigned int timer_msec)
{
struct fc_seq *sp;
- struct fc_exch *ep;
struct fc_frame *fp;
int error;
- ep = fc_seq_exch(req_sp);
-
- spin_lock_bh(&ep->ex_lock);
if (ep->esb_stat & (ESB_ST_COMPLETE | ESB_ST_ABNORMAL) ||
- ep->state & (FC_EX_DONE | FC_EX_RST_CLEANUP)) {
- spin_unlock_bh(&ep->ex_lock);
+ ep->state & (FC_EX_DONE | FC_EX_RST_CLEANUP))
return -ENXIO;
- }
/*
* Send the abort on a new sequence if possible.
*/
sp = fc_seq_start_next_locked(&ep->seq);
- if (!sp) {
- spin_unlock_bh(&ep->ex_lock);
+ if (!sp)
return -ENOMEM;
- }
ep->esb_stat |= ESB_ST_SEQ_INIT | ESB_ST_ABNORMAL;
if (timer_msec)
fc_exch_timer_set_locked(ep, timer_msec);
- spin_unlock_bh(&ep->ex_lock);
/*
* If not logged into the fabric, don't send ABTS but leave
@@ -633,6 +629,28 @@ static int fc_seq_exch_abort(const struct fc_seq *req_sp,
}
/**
+ * fc_seq_exch_abort() - Abort an exchange and sequence
+ * @req_sp: The sequence to be aborted
+ * @timer_msec: The period of time to wait before aborting
+ *
+ * Generally called because of a timeout or an abort from the upper layer.
+ *
+ * Return value: 0 on success else error code
+ */
+static int fc_seq_exch_abort(const struct fc_seq *req_sp,
+ unsigned int timer_msec)
+{
+ struct fc_exch *ep;
+ int error;
+
+ ep = fc_seq_exch(req_sp);
+ spin_lock_bh(&ep->ex_lock);
+ error = fc_exch_abort_locked(ep, timer_msec);
+ spin_unlock_bh(&ep->ex_lock);
+ return error;
+}
+
+/**
* fc_exch_timeout() - Handle exchange timer expiration
* @work: The work_struct identifying the exchange that timed out
*/
@@ -1715,6 +1733,7 @@ static void fc_exch_reset(struct fc_exch *ep)
int rc = 1;
spin_lock_bh(&ep->ex_lock);
+ fc_exch_abort_locked(ep, 0);
ep->state |= FC_EX_RST_CLEANUP;
if (cancel_delayed_work(&ep->timeout_work))
atomic_dec(&ep->ex_refcnt); /* drop hold for timer */
diff --git a/drivers/scsi/libfc/fc_lport.c b/drivers/scsi/libfc/fc_lport.c
index e55ed9c..628f347 100644
--- a/drivers/scsi/libfc/fc_lport.c
+++ b/drivers/scsi/libfc/fc_lport.c
@@ -88,6 +88,7 @@
*/
#include <linux/timer.h>
+#include <linux/delay.h>
#include <linux/slab.h>
#include <asm/unaligned.h>
@@ -1029,8 +1030,16 @@ static void fc_lport_enter_reset(struct fc_lport *lport)
FCH_EVT_LIPRESET, 0);
fc_vports_linkchange(lport);
fc_lport_reset_locked(lport);
- if (lport->link_up)
+ if (lport->link_up) {
+ /*
+ * Wait upto resource allocation time out before
+ * doing re-login since incomplete FIP exchanged
+ * from last session may collide with exchanges
+ * in new session.
+ */
+ msleep(lport->r_a_tov);
fc_lport_enter_flogi(lport);
+ }
}
/**
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH 3/4] libfc: block SCSI eh thread for blocked rports
2011-08-25 19:40 [PATCH 0/4] libfc and fcoe updates for scsi-misc Robert Love
2011-08-25 19:40 ` [PATCH 1/4] fcoe: Fix deadlock between fip's recv_work and rtnl Robert Love
2011-08-25 19:40 ` [PATCH 2/4] libfc: fix fc_eh_host_reset Robert Love
@ 2011-08-25 19:40 ` Robert Love
2011-08-25 19:41 ` [PATCH 4/4] libfc: fix referencing to fc_fcp_pkt from the frame pointer via fr_fsp() Robert Love
3 siblings, 0 replies; 5+ messages in thread
From: Robert Love @ 2011-08-25 19:40 UTC (permalink / raw)
To: linux-scsi; +Cc: Ross Brattain, Vasu Dev, Yi Zou
From: Vasu Dev <vasu.dev@intel.com>
Call fc_block_scsi_eh() in all fcoe eh to blocks
the scsi_eh thread for blocked rports.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Reviewed-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
---
drivers/scsi/libfc/fc_fcp.c | 11 +++++++++--
1 files changed, 9 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/libfc/fc_fcp.c b/drivers/scsi/libfc/fc_fcp.c
index afb63c8..4c41ee8 100644
--- a/drivers/scsi/libfc/fc_fcp.c
+++ b/drivers/scsi/libfc/fc_fcp.c
@@ -2019,6 +2019,11 @@ int fc_eh_abort(struct scsi_cmnd *sc_cmd)
struct fc_fcp_internal *si;
int rc = FAILED;
unsigned long flags;
+ int rval;
+
+ rval = fc_block_scsi_eh(sc_cmd);
+ if (rval)
+ return rval;
lport = shost_priv(sc_cmd->device->host);
if (lport->state != LPORT_ST_READY)
@@ -2068,9 +2073,9 @@ int fc_eh_device_reset(struct scsi_cmnd *sc_cmd)
int rc = FAILED;
int rval;
- rval = fc_remote_port_chkready(rport);
+ rval = fc_block_scsi_eh(sc_cmd);
if (rval)
- goto out;
+ return rval;
lport = shost_priv(sc_cmd->device->host);
@@ -2116,6 +2121,8 @@ int fc_eh_host_reset(struct scsi_cmnd *sc_cmd)
FC_SCSI_DBG(lport, "Resetting host\n");
+ fc_block_scsi_eh(sc_cmd);
+
lport->tt.lport_reset(lport);
wait_tmo = jiffies + FC_HOST_RESET_TIMEOUT;
while (!fc_fcp_lport_queue_ready(lport) && time_before(jiffies,
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 4/4] libfc: fix referencing to fc_fcp_pkt from the frame pointer via fr_fsp()
2011-08-25 19:40 [PATCH 0/4] libfc and fcoe updates for scsi-misc Robert Love
` (2 preceding siblings ...)
2011-08-25 19:40 ` [PATCH 3/4] libfc: block SCSI eh thread for blocked rports Robert Love
@ 2011-08-25 19:41 ` Robert Love
3 siblings, 0 replies; 5+ messages in thread
From: Robert Love @ 2011-08-25 19:41 UTC (permalink / raw)
To: linux-scsi; +Cc: Ross Brattain, Bhanu Prakash Gollapudi, Yi Zou
From: Yi Zou <yi.zou@intel.com>
In commit 6a716a8, while releasing the DDP context in case frame_send() failed,
the frame may already be freed, so we should store the pointer to fc_fcp_pkt and
release the DDP context using the locally stored fsp instead of getting fsp from
the fr_fsp(fp) on a frame.
Signed-off-by: Yi Zou <yi.zou@intel.com>
Reported-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
---
drivers/scsi/libfc/fc_exch.c | 8 ++++++--
1 files changed, 6 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/libfc/fc_exch.c b/drivers/scsi/libfc/fc_exch.c
index 744fefe..d261e98 100644
--- a/drivers/scsi/libfc/fc_exch.c
+++ b/drivers/scsi/libfc/fc_exch.c
@@ -1981,6 +1981,7 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport,
struct fc_exch *ep;
struct fc_seq *sp = NULL;
struct fc_frame_header *fh;
+ struct fc_fcp_pkt *fsp = NULL;
int rc = 1;
ep = fc_exch_alloc(lport, fp);
@@ -2003,8 +2004,10 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport,
fc_exch_setup_hdr(ep, fp, ep->f_ctl);
sp->cnt++;
- if (ep->xid <= lport->lro_xid && fh->fh_r_ctl == FC_RCTL_DD_UNSOL_CMD)
+ if (ep->xid <= lport->lro_xid && fh->fh_r_ctl == FC_RCTL_DD_UNSOL_CMD) {
+ fsp = fr_fsp(fp);
fc_fcp_ddp_setup(fr_fsp(fp), ep->xid);
+ }
if (unlikely(lport->tt.frame_send(lport, fp)))
goto err;
@@ -2018,7 +2021,8 @@ static struct fc_seq *fc_exch_seq_send(struct fc_lport *lport,
spin_unlock_bh(&ep->ex_lock);
return sp;
err:
- fc_fcp_ddp_done(fr_fsp(fp));
+ if (fsp)
+ fc_fcp_ddp_done(fsp);
rc = fc_exch_done_locked(ep);
spin_unlock_bh(&ep->ex_lock);
if (!rc)
^ permalink raw reply related [flat|nested] 5+ messages in thread