2009-03-18 02:42:13 +08:00
|
|
|
/*
|
2009-03-18 02:42:40 +08:00
|
|
|
* Copyright (c) 2008-2009 Cisco Systems, Inc. All rights reserved.
|
|
|
|
|
* Copyright (c) 2009 Intel Corporation. All rights reserved.
|
2009-03-18 02:42:13 +08:00
|
|
|
*
|
|
|
|
|
* This program is free software; you can redistribute it and/or modify it
|
|
|
|
|
* under the terms and conditions of the GNU General Public License,
|
|
|
|
|
* version 2, as published by the Free Software Foundation.
|
|
|
|
|
*
|
|
|
|
|
* This program is distributed in the hope it will be useful, but WITHOUT
|
|
|
|
|
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
|
|
|
|
|
* more details.
|
|
|
|
|
*
|
|
|
|
|
* You should have received a copy of the GNU General Public License along with
|
|
|
|
|
* this program; if not, write to the Free Software Foundation, Inc.,
|
|
|
|
|
* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
|
|
|
|
|
*
|
|
|
|
|
* Maintained at www.Open-FCoE.org
|
|
|
|
|
*/
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
#include <linux/types.h>
|
2009-03-18 02:42:13 +08:00
|
|
|
#include <linux/module.h>
|
2009-03-18 02:42:40 +08:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
|
#include <linux/list.h>
|
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
|
#include <linux/timer.h>
|
2009-03-18 02:42:18 +08:00
|
|
|
#include <linux/netdevice.h>
|
2009-03-18 02:42:40 +08:00
|
|
|
#include <linux/etherdevice.h>
|
|
|
|
|
#include <linux/ethtool.h>
|
|
|
|
|
#include <linux/if_ether.h>
|
|
|
|
|
#include <linux/if_vlan.h>
|
|
|
|
|
#include <linux/errno.h>
|
|
|
|
|
#include <linux/bitops.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
|
|
|
#include <linux/slab.h>
|
2009-03-18 02:42:40 +08:00
|
|
|
#include <net/rtnetlink.h>
|
|
|
|
|
|
|
|
|
|
#include <scsi/fc/fc_els.h>
|
|
|
|
|
#include <scsi/fc/fc_fs.h>
|
|
|
|
|
#include <scsi/fc/fc_fip.h>
|
|
|
|
|
#include <scsi/fc/fc_encaps.h>
|
|
|
|
|
#include <scsi/fc/fc_fcoe.h>
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
#include <scsi/fc/fc_fcp.h>
|
2009-03-18 02:42:18 +08:00
|
|
|
|
|
|
|
|
#include <scsi/libfc.h>
|
2009-03-18 02:42:40 +08:00
|
|
|
#include <scsi/libfcoe.h>
|
2009-03-18 02:42:13 +08:00
|
|
|
|
2011-01-29 08:04:45 +08:00
|
|
|
#include "libfcoe.h"
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
#define FCOE_CTLR_MIN_FKA 500 /* min keep alive (mS) */
|
|
|
|
|
#define FCOE_CTLR_DEF_FKA FIP_DEF_FKA /* default keep alive (mS) */
|
|
|
|
|
|
|
|
|
|
static void fcoe_ctlr_timeout(unsigned long);
|
2010-03-13 08:08:23 +08:00
|
|
|
static void fcoe_ctlr_timer_work(struct work_struct *);
|
2009-03-18 02:42:40 +08:00
|
|
|
static void fcoe_ctlr_recv_work(struct work_struct *);
|
2010-12-01 08:19:56 +08:00
|
|
|
static int fcoe_ctlr_flogi_retry(struct fcoe_ctlr *);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
static void fcoe_ctlr_vn_start(struct fcoe_ctlr *);
|
|
|
|
|
static int fcoe_ctlr_vn_recv(struct fcoe_ctlr *, struct sk_buff *);
|
|
|
|
|
static void fcoe_ctlr_vn_timeout(struct fcoe_ctlr *);
|
|
|
|
|
static int fcoe_ctlr_vn_lookup(struct fcoe_ctlr *, u32, u8 *);
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
static u8 fcoe_all_fcfs[ETH_ALEN] = FIP_ALL_FCF_MACS;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
static u8 fcoe_all_enode[ETH_ALEN] = FIP_ALL_ENODE_MACS;
|
|
|
|
|
static u8 fcoe_all_vn2vn[ETH_ALEN] = FIP_ALL_VN2VN_MACS;
|
|
|
|
|
static u8 fcoe_all_p2p[ETH_ALEN] = FIP_ALL_P2P_MACS;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2011-01-29 08:05:00 +08:00
|
|
|
static const char * const fcoe_ctlr_states[] = {
|
2010-07-21 06:20:24 +08:00
|
|
|
[FIP_ST_DISABLED] = "DISABLED",
|
|
|
|
|
[FIP_ST_LINK_WAIT] = "LINK_WAIT",
|
|
|
|
|
[FIP_ST_AUTO] = "AUTO",
|
|
|
|
|
[FIP_ST_NON_FIP] = "NON_FIP",
|
|
|
|
|
[FIP_ST_ENABLED] = "ENABLED",
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
[FIP_ST_VNMP_START] = "VNMP_START",
|
|
|
|
|
[FIP_ST_VNMP_PROBE1] = "VNMP_PROBE1",
|
|
|
|
|
[FIP_ST_VNMP_PROBE2] = "VNMP_PROBE2",
|
|
|
|
|
[FIP_ST_VNMP_CLAIM] = "VNMP_CLAIM",
|
|
|
|
|
[FIP_ST_VNMP_UP] = "VNMP_UP",
|
2010-07-21 06:20:24 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static const char *fcoe_ctlr_state(enum fip_state state)
|
|
|
|
|
{
|
|
|
|
|
const char *cp = "unknown";
|
|
|
|
|
|
|
|
|
|
if (state < ARRAY_SIZE(fcoe_ctlr_states))
|
|
|
|
|
cp = fcoe_ctlr_states[state];
|
|
|
|
|
if (!cp)
|
|
|
|
|
cp = "unknown";
|
|
|
|
|
return cp;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_set_state() - Set and do debug printing for the new FIP state.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @state: The new state
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_set_state(struct fcoe_ctlr *fip, enum fip_state state)
|
|
|
|
|
{
|
|
|
|
|
if (state == fip->state)
|
|
|
|
|
return;
|
|
|
|
|
if (fip->lp)
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "state %s -> %s\n",
|
|
|
|
|
fcoe_ctlr_state(fip->state), fcoe_ctlr_state(state));
|
|
|
|
|
fip->state = state;
|
|
|
|
|
}
|
|
|
|
|
|
2009-11-04 03:47:45 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_mtu_valid() - Check if a FCF's MTU is valid
|
|
|
|
|
* @fcf: The FCF to check
|
|
|
|
|
*
|
2009-03-18 02:42:40 +08:00
|
|
|
* Return non-zero if FCF fcoe_size has been validated.
|
|
|
|
|
*/
|
|
|
|
|
static inline int fcoe_ctlr_mtu_valid(const struct fcoe_fcf *fcf)
|
|
|
|
|
{
|
|
|
|
|
return (fcf->flags & FIP_FL_SOL) != 0;
|
|
|
|
|
}
|
|
|
|
|
|
2009-11-04 03:47:45 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_fcf_usable() - Check if a FCF is usable
|
|
|
|
|
* @fcf: The FCF to check
|
|
|
|
|
*
|
2009-03-18 02:42:40 +08:00
|
|
|
* Return non-zero if the FCF is usable.
|
|
|
|
|
*/
|
|
|
|
|
static inline int fcoe_ctlr_fcf_usable(struct fcoe_fcf *fcf)
|
|
|
|
|
{
|
|
|
|
|
u16 flags = FIP_FL_SOL | FIP_FL_AVAIL;
|
|
|
|
|
|
|
|
|
|
return (fcf->flags & flags) == flags;
|
|
|
|
|
}
|
|
|
|
|
|
2010-07-21 06:20:40 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_map_dest() - Set flag and OUI for mapping destination addresses
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_map_dest(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
if (fip->mode == FIP_MODE_VN2VN)
|
|
|
|
|
hton24(fip->dest_addr, FIP_VN_FC_MAP);
|
|
|
|
|
else
|
|
|
|
|
hton24(fip->dest_addr, FIP_DEF_FC_MAP);
|
|
|
|
|
hton24(fip->dest_addr + 3, 0);
|
|
|
|
|
fip->map_dest = 1;
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_init() - Initialize the FCoE Controller instance
|
|
|
|
|
* @fip: The FCoE controller to initialize
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
2010-07-21 06:19:58 +08:00
|
|
|
void fcoe_ctlr_init(struct fcoe_ctlr *fip, enum fip_state mode)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_LINK_WAIT);
|
2010-07-21 06:19:58 +08:00
|
|
|
fip->mode = mode;
|
2009-03-18 02:42:40 +08:00
|
|
|
INIT_LIST_HEAD(&fip->fcfs);
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_init(&fip->ctlr_mutex);
|
2010-12-01 08:19:56 +08:00
|
|
|
spin_lock_init(&fip->ctlr_lock);
|
2009-03-18 02:42:40 +08:00
|
|
|
fip->flogi_oxid = FC_XID_UNKNOWN;
|
|
|
|
|
setup_timer(&fip->timer, fcoe_ctlr_timeout, (unsigned long)fip);
|
2010-03-13 08:08:23 +08:00
|
|
|
INIT_WORK(&fip->timer_work, fcoe_ctlr_timer_work);
|
2009-03-18 02:42:40 +08:00
|
|
|
INIT_WORK(&fip->recv_work, fcoe_ctlr_recv_work);
|
|
|
|
|
skb_queue_head_init(&fip->fip_recv_list);
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_init);
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_reset_fcfs() - Reset and free all FCFs for a controller
|
|
|
|
|
* @fip: The FCoE controller whose FCFs are to be reset
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Called with &fcoe_ctlr lock held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_reset_fcfs(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
struct fcoe_fcf *next;
|
|
|
|
|
|
|
|
|
|
fip->sel_fcf = NULL;
|
|
|
|
|
list_for_each_entry_safe(fcf, next, &fip->fcfs, list) {
|
|
|
|
|
list_del(&fcf->list);
|
|
|
|
|
kfree(fcf);
|
|
|
|
|
}
|
|
|
|
|
fip->fcf_count = 0;
|
|
|
|
|
fip->sel_time = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_destroy() - Disable and tear down a FCoE controller
|
|
|
|
|
* @fip: The FCoE controller to tear down
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* This is called by FCoE drivers before freeing the &fcoe_ctlr.
|
|
|
|
|
*
|
|
|
|
|
* The receive handler will have been deleted before this to guarantee
|
|
|
|
|
* that no more recv_work will be scheduled.
|
|
|
|
|
*
|
|
|
|
|
* The timer routine will simply return once we set FIP_ST_DISABLED.
|
|
|
|
|
* This guarantees that no further timeouts or work will be scheduled.
|
|
|
|
|
*/
|
|
|
|
|
void fcoe_ctlr_destroy(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
2009-08-26 04:59:08 +08:00
|
|
|
cancel_work_sync(&fip->recv_work);
|
2009-11-04 03:48:22 +08:00
|
|
|
skb_queue_purge(&fip->fip_recv_list);
|
2009-08-26 04:59:08 +08:00
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_DISABLED);
|
2009-03-18 02:42:40 +08:00
|
|
|
fcoe_ctlr_reset_fcfs(fip);
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
del_timer_sync(&fip->timer);
|
2010-03-13 08:08:23 +08:00
|
|
|
cancel_work_sync(&fip->timer_work);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_destroy);
|
|
|
|
|
|
2010-12-01 08:19:40 +08:00
|
|
|
/**
|
2010-12-01 08:19:56 +08:00
|
|
|
* fcoe_ctlr_announce() - announce new FCF selection
|
2010-12-01 08:19:40 +08:00
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Also sets the destination MAC for FCoE and control packets
|
2010-12-01 08:19:56 +08:00
|
|
|
*
|
|
|
|
|
* Called with neither ctlr_mutex nor ctlr_lock held.
|
2010-12-01 08:19:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_announce(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
2010-12-01 08:19:56 +08:00
|
|
|
struct fcoe_fcf *sel;
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
spin_lock_bh(&fip->ctlr_lock);
|
|
|
|
|
|
|
|
|
|
kfree_skb(fip->flogi_req);
|
|
|
|
|
fip->flogi_req = NULL;
|
|
|
|
|
list_for_each_entry(fcf, &fip->fcfs, list)
|
|
|
|
|
fcf->flogi_sent = 0;
|
|
|
|
|
|
|
|
|
|
spin_unlock_bh(&fip->ctlr_lock);
|
|
|
|
|
sel = fip->sel_fcf;
|
2010-12-01 08:19:40 +08:00
|
|
|
|
|
|
|
|
if (sel && !compare_ether_addr(sel->fcf_mac, fip->dest_addr))
|
2010-12-01 08:19:56 +08:00
|
|
|
goto unlock;
|
2010-12-01 08:19:40 +08:00
|
|
|
if (!is_zero_ether_addr(fip->dest_addr)) {
|
|
|
|
|
printk(KERN_NOTICE "libfcoe: host%d: "
|
|
|
|
|
"FIP Fibre-Channel Forwarder MAC %pM deselected\n",
|
|
|
|
|
fip->lp->host->host_no, fip->dest_addr);
|
|
|
|
|
memset(fip->dest_addr, 0, ETH_ALEN);
|
|
|
|
|
}
|
|
|
|
|
if (sel) {
|
|
|
|
|
printk(KERN_INFO "libfcoe: host%d: FIP selected "
|
|
|
|
|
"Fibre-Channel Forwarder MAC %pM\n",
|
|
|
|
|
fip->lp->host->host_no, sel->fcf_mac);
|
|
|
|
|
memcpy(fip->dest_addr, sel->fcf_mac, ETH_ALEN);
|
|
|
|
|
fip->map_dest = 0;
|
|
|
|
|
}
|
2010-12-01 08:19:56 +08:00
|
|
|
unlock:
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2010-12-01 08:19:40 +08:00
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_fcoe_size() - Return the maximum FCoE size required for VN_Port
|
|
|
|
|
* @fip: The FCoE controller to get the maximum FCoE size from
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Returns the maximum packet size including the FCoE header and trailer,
|
|
|
|
|
* but not including any Ethernet or VLAN headers.
|
|
|
|
|
*/
|
|
|
|
|
static inline u32 fcoe_ctlr_fcoe_size(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
/*
|
|
|
|
|
* Determine the max FCoE frame size allowed, including
|
|
|
|
|
* FCoE header and trailer.
|
|
|
|
|
* Note: lp->mfs is currently the payload size, not the frame size.
|
|
|
|
|
*/
|
|
|
|
|
return fip->lp->mfs + sizeof(struct fc_frame_header) +
|
|
|
|
|
sizeof(struct fcoe_hdr) + sizeof(struct fcoe_crc_eof);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_solicit() - Send a FIP solicitation
|
|
|
|
|
* @fip: The FCoE controller to send the solicitation on
|
|
|
|
|
* @fcf: The destination FCF (if NULL, a multicast solicitation is sent)
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_solicit(struct fcoe_ctlr *fip, struct fcoe_fcf *fcf)
|
|
|
|
|
{
|
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
struct fip_sol {
|
|
|
|
|
struct ethhdr eth;
|
|
|
|
|
struct fip_header fip;
|
|
|
|
|
struct {
|
|
|
|
|
struct fip_mac_desc mac;
|
|
|
|
|
struct fip_wwn_desc wwnn;
|
|
|
|
|
struct fip_size_desc size;
|
2011-01-29 08:05:00 +08:00
|
|
|
} __packed desc;
|
|
|
|
|
} __packed * sol;
|
2009-03-18 02:42:40 +08:00
|
|
|
u32 fcoe_size;
|
|
|
|
|
|
|
|
|
|
skb = dev_alloc_skb(sizeof(*sol));
|
|
|
|
|
if (!skb)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
sol = (struct fip_sol *)skb->data;
|
|
|
|
|
|
|
|
|
|
memset(sol, 0, sizeof(*sol));
|
|
|
|
|
memcpy(sol->eth.h_dest, fcf ? fcf->fcf_mac : fcoe_all_fcfs, ETH_ALEN);
|
|
|
|
|
memcpy(sol->eth.h_source, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
sol->eth.h_proto = htons(ETH_P_FIP);
|
|
|
|
|
|
|
|
|
|
sol->fip.fip_ver = FIP_VER_ENCAPS(FIP_VER);
|
|
|
|
|
sol->fip.fip_op = htons(FIP_OP_DISC);
|
|
|
|
|
sol->fip.fip_subcode = FIP_SC_SOL;
|
|
|
|
|
sol->fip.fip_dl_len = htons(sizeof(sol->desc) / FIP_BPW);
|
|
|
|
|
sol->fip.fip_flags = htons(FIP_FL_FPMA);
|
2009-05-17 20:33:28 +08:00
|
|
|
if (fip->spma)
|
|
|
|
|
sol->fip.fip_flags |= htons(FIP_FL_SPMA);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
sol->desc.mac.fd_desc.fip_dtype = FIP_DT_MAC;
|
|
|
|
|
sol->desc.mac.fd_desc.fip_dlen = sizeof(sol->desc.mac) / FIP_BPW;
|
|
|
|
|
memcpy(sol->desc.mac.fd_mac, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
|
|
|
|
|
sol->desc.wwnn.fd_desc.fip_dtype = FIP_DT_NAME;
|
|
|
|
|
sol->desc.wwnn.fd_desc.fip_dlen = sizeof(sol->desc.wwnn) / FIP_BPW;
|
|
|
|
|
put_unaligned_be64(fip->lp->wwnn, &sol->desc.wwnn.fd_wwn);
|
|
|
|
|
|
|
|
|
|
fcoe_size = fcoe_ctlr_fcoe_size(fip);
|
|
|
|
|
sol->desc.size.fd_desc.fip_dtype = FIP_DT_FCOE_SIZE;
|
|
|
|
|
sol->desc.size.fd_desc.fip_dlen = sizeof(sol->desc.size) / FIP_BPW;
|
|
|
|
|
sol->desc.size.fd_size = htons(fcoe_size);
|
|
|
|
|
|
|
|
|
|
skb_put(skb, sizeof(*sol));
|
2009-05-07 01:52:18 +08:00
|
|
|
skb->protocol = htons(ETH_P_FIP);
|
2011-11-19 05:35:56 +08:00
|
|
|
skb->priority = fip->priority;
|
2009-03-18 02:42:40 +08:00
|
|
|
skb_reset_mac_header(skb);
|
|
|
|
|
skb_reset_network_header(skb);
|
|
|
|
|
fip->send(fip, skb);
|
|
|
|
|
|
|
|
|
|
if (!fcf)
|
|
|
|
|
fip->sol_time = jiffies;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_link_up() - Start FCoE controller
|
|
|
|
|
* @fip: The FCoE controller to start
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Called from the LLD when the network link is ready.
|
|
|
|
|
*/
|
|
|
|
|
void fcoe_ctlr_link_up(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fip->state == FIP_ST_NON_FIP || fip->state == FIP_ST_AUTO) {
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
fc_linkup(fip->lp);
|
|
|
|
|
} else if (fip->state == FIP_ST_LINK_WAIT) {
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, fip->mode);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
switch (fip->mode) {
|
|
|
|
|
default:
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "invalid mode %d\n", fip->mode);
|
|
|
|
|
/* fall-through */
|
|
|
|
|
case FIP_MODE_AUTO:
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "%s", "setting AUTO mode.\n");
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
/* fall-through */
|
|
|
|
|
case FIP_MODE_FABRIC:
|
|
|
|
|
case FIP_MODE_NON_FIP:
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
fc_linkup(fip->lp);
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_MODE_VN2VN:
|
|
|
|
|
fcoe_ctlr_vn_start(fip);
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
fc_linkup(fip->lp);
|
|
|
|
|
break;
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
} else
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_link_up);
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_reset() - Reset a FCoE controller
|
|
|
|
|
* @fip: The FCoE controller to reset
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
static void fcoe_ctlr_reset(struct fcoe_ctlr *fip)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
fcoe_ctlr_reset_fcfs(fip);
|
|
|
|
|
del_timer(&fip->timer);
|
|
|
|
|
fip->ctlr_ka_time = 0;
|
|
|
|
|
fip->port_ka_time = 0;
|
|
|
|
|
fip->sol_time = 0;
|
|
|
|
|
fip->flogi_oxid = FC_XID_UNKNOWN;
|
2010-07-21 06:20:40 +08:00
|
|
|
fcoe_ctlr_map_dest(fip);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_link_down() - Stop a FCoE controller
|
|
|
|
|
* @fip: The FCoE controller to be stopped
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Returns non-zero if the link was up and now isn't.
|
|
|
|
|
*
|
|
|
|
|
* Called from the LLD when the network link is not ready.
|
|
|
|
|
* There may be multiple calls while the link is down.
|
|
|
|
|
*/
|
|
|
|
|
int fcoe_ctlr_link_down(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
int link_dropped;
|
|
|
|
|
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "link down.\n");
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
fcoe_ctlr_reset(fip);
|
2010-03-13 08:08:23 +08:00
|
|
|
link_dropped = fip->state != FIP_ST_LINK_WAIT;
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_LINK_WAIT);
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
|
|
|
|
|
if (link_dropped)
|
|
|
|
|
fc_linkdown(fip->lp);
|
|
|
|
|
return link_dropped;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_link_down);
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_send_keep_alive() - Send a keep-alive to the selected FCF
|
|
|
|
|
* @fip: The FCoE controller to send the FKA on
|
|
|
|
|
* @lport: libfc fc_lport to send from
|
|
|
|
|
* @ports: 0 for controller keep-alive, 1 for port keep-alive
|
|
|
|
|
* @sa: The source MAC address
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* A controller keep-alive is sent every fka_period (typically 8 seconds).
|
|
|
|
|
* The source MAC is the native MAC address.
|
|
|
|
|
*
|
|
|
|
|
* A port keep-alive is sent every 90 seconds while logged in.
|
|
|
|
|
* The source MAC is the assigned mapped source address.
|
|
|
|
|
* The destination is the FCF's F-port.
|
|
|
|
|
*/
|
2009-11-04 03:46:29 +08:00
|
|
|
static void fcoe_ctlr_send_keep_alive(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_lport *lport,
|
|
|
|
|
int ports, u8 *sa)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
struct fip_kal {
|
|
|
|
|
struct ethhdr eth;
|
|
|
|
|
struct fip_header fip;
|
|
|
|
|
struct fip_mac_desc mac;
|
2011-01-29 08:05:00 +08:00
|
|
|
} __packed * kal;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fip_vn_desc *vn;
|
|
|
|
|
u32 len;
|
|
|
|
|
struct fc_lport *lp;
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
|
|
|
|
|
fcf = fip->sel_fcf;
|
|
|
|
|
lp = fip->lp;
|
2010-06-12 07:43:33 +08:00
|
|
|
if (!fcf || (ports && !lp->port_id))
|
2009-03-18 02:42:40 +08:00
|
|
|
return;
|
|
|
|
|
|
2009-11-04 03:50:16 +08:00
|
|
|
len = sizeof(*kal) + ports * sizeof(*vn);
|
2009-03-18 02:42:40 +08:00
|
|
|
skb = dev_alloc_skb(len);
|
|
|
|
|
if (!skb)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
kal = (struct fip_kal *)skb->data;
|
|
|
|
|
memset(kal, 0, len);
|
|
|
|
|
memcpy(kal->eth.h_dest, fcf->fcf_mac, ETH_ALEN);
|
|
|
|
|
memcpy(kal->eth.h_source, sa, ETH_ALEN);
|
|
|
|
|
kal->eth.h_proto = htons(ETH_P_FIP);
|
|
|
|
|
|
|
|
|
|
kal->fip.fip_ver = FIP_VER_ENCAPS(FIP_VER);
|
|
|
|
|
kal->fip.fip_op = htons(FIP_OP_CTRL);
|
|
|
|
|
kal->fip.fip_subcode = FIP_SC_KEEP_ALIVE;
|
|
|
|
|
kal->fip.fip_dl_len = htons((sizeof(kal->mac) +
|
2009-11-04 03:47:45 +08:00
|
|
|
ports * sizeof(*vn)) / FIP_BPW);
|
2009-03-18 02:42:40 +08:00
|
|
|
kal->fip.fip_flags = htons(FIP_FL_FPMA);
|
2009-05-17 20:33:28 +08:00
|
|
|
if (fip->spma)
|
|
|
|
|
kal->fip.fip_flags |= htons(FIP_FL_SPMA);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
kal->mac.fd_desc.fip_dtype = FIP_DT_MAC;
|
|
|
|
|
kal->mac.fd_desc.fip_dlen = sizeof(kal->mac) / FIP_BPW;
|
|
|
|
|
memcpy(kal->mac.fd_mac, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
if (ports) {
|
|
|
|
|
vn = (struct fip_vn_desc *)(kal + 1);
|
|
|
|
|
vn->fd_desc.fip_dtype = FIP_DT_VN_ID;
|
|
|
|
|
vn->fd_desc.fip_dlen = sizeof(*vn) / FIP_BPW;
|
2009-11-04 03:46:29 +08:00
|
|
|
memcpy(vn->fd_mac, fip->get_src_addr(lport), ETH_ALEN);
|
2010-05-08 06:18:57 +08:00
|
|
|
hton24(vn->fd_fc_id, lport->port_id);
|
|
|
|
|
put_unaligned_be64(lport->wwpn, &vn->fd_wwpn);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
skb_put(skb, len);
|
2009-05-07 01:52:18 +08:00
|
|
|
skb->protocol = htons(ETH_P_FIP);
|
2011-11-19 05:35:56 +08:00
|
|
|
skb->priority = fip->priority;
|
2009-03-18 02:42:40 +08:00
|
|
|
skb_reset_mac_header(skb);
|
|
|
|
|
skb_reset_network_header(skb);
|
|
|
|
|
fip->send(fip, skb);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_encaps() - Encapsulate an ELS frame for FIP, without sending it
|
|
|
|
|
* @fip: The FCoE controller for the ELS frame
|
|
|
|
|
* @dtype: The FIP descriptor type for the frame
|
|
|
|
|
* @skb: The FCoE ELS frame including FC header but no FCoE headers
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
* @d_id: The destination port ID.
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Returns non-zero error code on failure.
|
|
|
|
|
*
|
|
|
|
|
* The caller must check that the length is a multiple of 4.
|
|
|
|
|
*
|
|
|
|
|
* The @skb must have enough headroom (28 bytes) and tailroom (8 bytes).
|
|
|
|
|
* Headroom includes the FIP encapsulation description, FIP header, and
|
|
|
|
|
* Ethernet header. The tailroom is for the FIP MAC descriptor.
|
|
|
|
|
*/
|
2009-11-04 03:46:29 +08:00
|
|
|
static int fcoe_ctlr_encaps(struct fcoe_ctlr *fip, struct fc_lport *lport,
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
u8 dtype, struct sk_buff *skb, u32 d_id)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fip_encaps_head {
|
|
|
|
|
struct ethhdr eth;
|
|
|
|
|
struct fip_header fip;
|
|
|
|
|
struct fip_encaps encaps;
|
2011-01-29 08:05:00 +08:00
|
|
|
} __packed * cap;
|
2010-07-21 06:20:35 +08:00
|
|
|
struct fc_frame_header *fh;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fip_mac_desc *mac;
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
size_t dlen;
|
2009-07-30 08:03:55 +08:00
|
|
|
u16 fip_flags;
|
2010-07-21 06:20:35 +08:00
|
|
|
u8 op;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-07-21 06:20:35 +08:00
|
|
|
fh = (struct fc_frame_header *)skb->data;
|
|
|
|
|
op = *(u8 *)(fh + 1);
|
2009-03-18 02:42:40 +08:00
|
|
|
dlen = sizeof(struct fip_encaps) + skb->len; /* len before push */
|
|
|
|
|
cap = (struct fip_encaps_head *)skb_push(skb, sizeof(*cap));
|
|
|
|
|
memset(cap, 0, sizeof(*cap));
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
|
|
|
|
|
if (lport->point_to_multipoint) {
|
|
|
|
|
if (fcoe_ctlr_vn_lookup(fip, d_id, cap->eth.h_dest))
|
|
|
|
|
return -ENODEV;
|
2010-07-21 06:20:35 +08:00
|
|
|
fip_flags = 0;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
} else {
|
|
|
|
|
fcf = fip->sel_fcf;
|
|
|
|
|
if (!fcf)
|
|
|
|
|
return -ENODEV;
|
|
|
|
|
fip_flags = fcf->flags;
|
|
|
|
|
fip_flags &= fip->spma ? FIP_FL_SPMA | FIP_FL_FPMA :
|
|
|
|
|
FIP_FL_FPMA;
|
|
|
|
|
if (!fip_flags)
|
|
|
|
|
return -ENODEV;
|
|
|
|
|
memcpy(cap->eth.h_dest, fcf->fcf_mac, ETH_ALEN);
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
memcpy(cap->eth.h_source, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
cap->eth.h_proto = htons(ETH_P_FIP);
|
|
|
|
|
|
|
|
|
|
cap->fip.fip_ver = FIP_VER_ENCAPS(FIP_VER);
|
|
|
|
|
cap->fip.fip_op = htons(FIP_OP_LS);
|
2010-07-21 06:20:35 +08:00
|
|
|
if (op == ELS_LS_ACC || op == ELS_LS_RJT)
|
|
|
|
|
cap->fip.fip_subcode = FIP_SC_REP;
|
|
|
|
|
else
|
|
|
|
|
cap->fip.fip_subcode = FIP_SC_REQ;
|
2009-07-30 08:03:55 +08:00
|
|
|
cap->fip.fip_flags = htons(fip_flags);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
cap->encaps.fd_desc.fip_dtype = dtype;
|
|
|
|
|
cap->encaps.fd_desc.fip_dlen = dlen / FIP_BPW;
|
|
|
|
|
|
2010-07-21 06:20:35 +08:00
|
|
|
if (op != ELS_LS_RJT) {
|
|
|
|
|
dlen += sizeof(*mac);
|
|
|
|
|
mac = (struct fip_mac_desc *)skb_put(skb, sizeof(*mac));
|
|
|
|
|
memset(mac, 0, sizeof(*mac));
|
|
|
|
|
mac->fd_desc.fip_dtype = FIP_DT_MAC;
|
|
|
|
|
mac->fd_desc.fip_dlen = sizeof(*mac) / FIP_BPW;
|
|
|
|
|
if (dtype != FIP_DT_FLOGI && dtype != FIP_DT_FDISC) {
|
|
|
|
|
memcpy(mac->fd_mac, fip->get_src_addr(lport), ETH_ALEN);
|
|
|
|
|
} else if (fip->mode == FIP_MODE_VN2VN) {
|
|
|
|
|
hton24(mac->fd_mac, FIP_VN_FC_MAP);
|
|
|
|
|
hton24(mac->fd_mac + 3, fip->port_id);
|
|
|
|
|
} else if (fip_flags & FIP_FL_SPMA) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FLOGI/FDISC sent with SPMA\n");
|
|
|
|
|
memcpy(mac->fd_mac, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
} else {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FLOGI/FDISC sent with FPMA\n");
|
|
|
|
|
/* FPMA only FLOGI. Must leave the MAC desc zeroed. */
|
|
|
|
|
}
|
2010-04-10 05:22:17 +08:00
|
|
|
}
|
2010-07-21 06:20:35 +08:00
|
|
|
cap->fip.fip_dl_len = htons(dlen / FIP_BPW);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2009-05-07 01:52:18 +08:00
|
|
|
skb->protocol = htons(ETH_P_FIP);
|
2011-11-19 05:35:56 +08:00
|
|
|
skb->priority = fip->priority;
|
2009-03-18 02:42:40 +08:00
|
|
|
skb_reset_mac_header(skb);
|
|
|
|
|
skb_reset_network_header(skb);
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_els_send() - Send an ELS frame encapsulated by FIP if appropriate.
|
|
|
|
|
* @fip: FCoE controller.
|
2009-11-04 03:46:29 +08:00
|
|
|
* @lport: libfc fc_lport to send from
|
2009-03-18 02:42:40 +08:00
|
|
|
* @skb: FCoE ELS frame including FC header but no FCoE headers.
|
|
|
|
|
*
|
|
|
|
|
* Returns a non-zero error code if the frame should not be sent.
|
|
|
|
|
* Returns zero if the caller should send the frame with FCoE encapsulation.
|
|
|
|
|
*
|
|
|
|
|
* The caller must check that the length is a multiple of 4.
|
|
|
|
|
* The SKB must have enough headroom (28 bytes) and tailroom (8 bytes).
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
* The the skb must also be an fc_frame.
|
2010-12-01 08:19:56 +08:00
|
|
|
*
|
|
|
|
|
* This is called from the lower-level driver with spinlocks held,
|
|
|
|
|
* so we must not take a mutex here.
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
2009-11-04 03:46:29 +08:00
|
|
|
int fcoe_ctlr_els_send(struct fcoe_ctlr *fip, struct fc_lport *lport,
|
|
|
|
|
struct sk_buff *skb)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
struct fc_frame *fp;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fc_frame_header *fh;
|
|
|
|
|
u16 old_xid;
|
|
|
|
|
u8 op;
|
2009-11-04 03:46:29 +08:00
|
|
|
u8 mac[ETH_ALEN];
|
2009-03-18 02:42:40 +08:00
|
|
|
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
fp = container_of(skb, struct fc_frame, skb);
|
2009-03-18 02:42:40 +08:00
|
|
|
fh = (struct fc_frame_header *)skb->data;
|
|
|
|
|
op = *(u8 *)(fh + 1);
|
|
|
|
|
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (op == ELS_FLOGI && fip->mode != FIP_MODE_VN2VN) {
|
2009-03-18 02:42:40 +08:00
|
|
|
old_xid = fip->flogi_oxid;
|
|
|
|
|
fip->flogi_oxid = ntohs(fh->fh_ox_id);
|
|
|
|
|
if (fip->state == FIP_ST_AUTO) {
|
|
|
|
|
if (old_xid == FC_XID_UNKNOWN)
|
|
|
|
|
fip->flogi_count = 0;
|
|
|
|
|
fip->flogi_count++;
|
|
|
|
|
if (fip->flogi_count < 3)
|
|
|
|
|
goto drop;
|
2010-07-21 06:20:40 +08:00
|
|
|
fcoe_ctlr_map_dest(fip);
|
2009-03-18 02:42:40 +08:00
|
|
|
return 0;
|
|
|
|
|
}
|
2009-05-07 01:52:12 +08:00
|
|
|
if (fip->state == FIP_ST_NON_FIP)
|
2010-07-21 06:20:40 +08:00
|
|
|
fcoe_ctlr_map_dest(fip);
|
2009-05-07 01:52:12 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (fip->state == FIP_ST_NON_FIP)
|
|
|
|
|
return 0;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (!fip->sel_fcf && fip->mode != FIP_MODE_VN2VN)
|
2009-11-04 03:48:32 +08:00
|
|
|
goto drop;
|
2009-05-07 01:52:12 +08:00
|
|
|
switch (op) {
|
|
|
|
|
case ELS_FLOGI:
|
2009-03-18 02:42:40 +08:00
|
|
|
op = FIP_DT_FLOGI;
|
2010-12-01 08:19:56 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN)
|
|
|
|
|
break;
|
|
|
|
|
spin_lock_bh(&fip->ctlr_lock);
|
|
|
|
|
kfree_skb(fip->flogi_req);
|
|
|
|
|
fip->flogi_req = skb;
|
|
|
|
|
fip->flogi_req_send = 1;
|
|
|
|
|
spin_unlock_bh(&fip->ctlr_lock);
|
|
|
|
|
schedule_work(&fip->timer_work);
|
|
|
|
|
return -EINPROGRESS;
|
2009-03-18 02:42:40 +08:00
|
|
|
case ELS_FDISC:
|
|
|
|
|
if (ntoh24(fh->fh_s_id))
|
|
|
|
|
return 0;
|
|
|
|
|
op = FIP_DT_FDISC;
|
|
|
|
|
break;
|
|
|
|
|
case ELS_LOGO:
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN) {
|
|
|
|
|
if (fip->state != FIP_ST_VNMP_UP)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
if (ntoh24(fh->fh_d_id) == FC_FID_FLOGI)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
} else {
|
|
|
|
|
if (fip->state != FIP_ST_ENABLED)
|
|
|
|
|
return 0;
|
|
|
|
|
if (ntoh24(fh->fh_d_id) != FC_FID_FLOGI)
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
op = FIP_DT_LOGO;
|
|
|
|
|
break;
|
|
|
|
|
case ELS_LS_ACC:
|
|
|
|
|
/*
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
* If non-FIP, we may have gotten an SID by accepting an FLOGI
|
2009-03-18 02:42:40 +08:00
|
|
|
* from a point-to-point connection. Switch to using
|
|
|
|
|
* the source mac based on the SID. The destination
|
2011-03-31 09:57:33 +08:00
|
|
|
* MAC in this case would have been set by receiving the
|
2009-03-18 02:42:40 +08:00
|
|
|
* FLOGI.
|
|
|
|
|
*/
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->state == FIP_ST_NON_FIP) {
|
|
|
|
|
if (fip->flogi_oxid == FC_XID_UNKNOWN)
|
|
|
|
|
return 0;
|
|
|
|
|
fip->flogi_oxid = FC_XID_UNKNOWN;
|
|
|
|
|
fc_fcoe_set_mac(mac, fh->fh_d_id);
|
|
|
|
|
fip->update_mac(lport, mac);
|
|
|
|
|
}
|
|
|
|
|
/* fall through */
|
|
|
|
|
case ELS_LS_RJT:
|
|
|
|
|
op = fr_encaps(fp);
|
|
|
|
|
if (op)
|
|
|
|
|
break;
|
2009-03-18 02:42:40 +08:00
|
|
|
return 0;
|
|
|
|
|
default:
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->state != FIP_ST_ENABLED &&
|
|
|
|
|
fip->state != FIP_ST_VNMP_UP)
|
2009-03-18 02:42:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "els_send op %u d_id %x\n",
|
|
|
|
|
op, ntoh24(fh->fh_d_id));
|
|
|
|
|
if (fcoe_ctlr_encaps(fip, lport, op, skb, ntoh24(fh->fh_d_id)))
|
2009-03-18 02:42:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
fip->send(fip, skb);
|
|
|
|
|
return -EINPROGRESS;
|
|
|
|
|
drop:
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_els_send);
|
|
|
|
|
|
2009-11-04 03:47:45 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_age_fcfs() - Reset and free all old FCFs for a controller
|
|
|
|
|
* @fip: The FCoE controller to free FCFs on
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
2010-03-13 08:08:55 +08:00
|
|
|
* Called with lock held and preemption disabled.
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
2010-06-12 07:44:10 +08:00
|
|
|
* An FCF is considered old if we have missed two advertisements.
|
|
|
|
|
* That is, there have been no valid advertisement from it for 2.5
|
|
|
|
|
* times its keep-alive period.
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* In addition, determine the time when an FCF selection can occur.
|
2009-11-21 06:55:08 +08:00
|
|
|
*
|
|
|
|
|
* Also, increment the MissDiscAdvCount when no advertisement is received
|
|
|
|
|
* for the corresponding FCF for 1.5 * FKA_ADV_PERIOD (FC-BB-5 LESB).
|
2010-06-12 07:44:10 +08:00
|
|
|
*
|
|
|
|
|
* Returns the time in jiffies for the next call.
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
2010-06-12 07:44:10 +08:00
|
|
|
static unsigned long fcoe_ctlr_age_fcfs(struct fcoe_ctlr *fip)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
struct fcoe_fcf *next;
|
2010-06-12 07:44:10 +08:00
|
|
|
unsigned long next_timer = jiffies + msecs_to_jiffies(FIP_VN_KA_PERIOD);
|
|
|
|
|
unsigned long deadline;
|
2009-03-18 02:42:40 +08:00
|
|
|
unsigned long sel_time = 0;
|
2010-03-13 08:08:55 +08:00
|
|
|
struct fcoe_dev_stats *stats;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
stats = per_cpu_ptr(fip->lp->dev_stats, get_cpu());
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
list_for_each_entry_safe(fcf, next, &fip->fcfs, list) {
|
2010-06-12 07:44:10 +08:00
|
|
|
deadline = fcf->time + fcf->fka_period + fcf->fka_period / 2;
|
|
|
|
|
if (fip->sel_fcf == fcf) {
|
|
|
|
|
if (time_after(jiffies, deadline)) {
|
|
|
|
|
stats->MissDiscAdvCount++;
|
|
|
|
|
printk(KERN_INFO "libfcoe: host%d: "
|
|
|
|
|
"Missing Discovery Advertisement "
|
|
|
|
|
"for fab %16.16llx count %lld\n",
|
|
|
|
|
fip->lp->host->host_no, fcf->fabric_name,
|
|
|
|
|
stats->MissDiscAdvCount);
|
|
|
|
|
} else if (time_after(next_timer, deadline))
|
|
|
|
|
next_timer = deadline;
|
2009-11-21 06:55:08 +08:00
|
|
|
}
|
2010-06-12 07:44:10 +08:00
|
|
|
|
|
|
|
|
deadline += fcf->fka_period;
|
2010-06-12 07:44:15 +08:00
|
|
|
if (time_after_eq(jiffies, deadline)) {
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fip->sel_fcf == fcf)
|
|
|
|
|
fip->sel_fcf = NULL;
|
|
|
|
|
list_del(&fcf->list);
|
|
|
|
|
WARN_ON(!fip->fcf_count);
|
|
|
|
|
fip->fcf_count--;
|
|
|
|
|
kfree(fcf);
|
2010-03-13 08:08:55 +08:00
|
|
|
stats->VLinkFailureCount++;
|
2010-06-12 07:44:10 +08:00
|
|
|
} else {
|
|
|
|
|
if (time_after(next_timer, deadline))
|
|
|
|
|
next_timer = deadline;
|
|
|
|
|
if (fcoe_ctlr_mtu_valid(fcf) &&
|
|
|
|
|
(!sel_time || time_before(sel_time, fcf->time)))
|
|
|
|
|
sel_time = fcf->time;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
}
|
2010-07-21 06:19:47 +08:00
|
|
|
put_cpu();
|
2010-06-12 07:44:15 +08:00
|
|
|
if (sel_time && !fip->sel_fcf && !fip->sel_time) {
|
2009-03-18 02:42:40 +08:00
|
|
|
sel_time += msecs_to_jiffies(FCOE_CTLR_START_DELAY);
|
|
|
|
|
fip->sel_time = sel_time;
|
|
|
|
|
}
|
2010-06-12 07:44:15 +08:00
|
|
|
|
2010-06-12 07:44:10 +08:00
|
|
|
return next_timer;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_parse_adv() - Decode a FIP advertisement into a new FCF entry
|
2009-11-04 03:48:16 +08:00
|
|
|
* @fip: The FCoE controller receiving the advertisement
|
2009-11-04 03:47:45 +08:00
|
|
|
* @skb: The received FIP advertisement frame
|
|
|
|
|
* @fcf: The resulting FCF entry
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Returns zero on a valid parsed advertisement,
|
|
|
|
|
* otherwise returns non zero value.
|
|
|
|
|
*/
|
2009-11-04 03:48:16 +08:00
|
|
|
static int fcoe_ctlr_parse_adv(struct fcoe_ctlr *fip,
|
|
|
|
|
struct sk_buff *skb, struct fcoe_fcf *fcf)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fip_header *fiph;
|
|
|
|
|
struct fip_desc *desc = NULL;
|
|
|
|
|
struct fip_wwn_desc *wwn;
|
|
|
|
|
struct fip_fab_desc *fab;
|
|
|
|
|
struct fip_fka_desc *fka;
|
|
|
|
|
unsigned long t;
|
|
|
|
|
size_t rlen;
|
|
|
|
|
size_t dlen;
|
2010-06-12 07:44:25 +08:00
|
|
|
u32 desc_mask;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
memset(fcf, 0, sizeof(*fcf));
|
|
|
|
|
fcf->fka_period = msecs_to_jiffies(FCOE_CTLR_DEF_FKA);
|
|
|
|
|
|
|
|
|
|
fiph = (struct fip_header *)skb->data;
|
|
|
|
|
fcf->flags = ntohs(fiph->fip_flags);
|
|
|
|
|
|
2010-06-12 07:44:25 +08:00
|
|
|
/*
|
|
|
|
|
* mask of required descriptors. validating each one clears its bit.
|
|
|
|
|
*/
|
|
|
|
|
desc_mask = BIT(FIP_DT_PRI) | BIT(FIP_DT_MAC) | BIT(FIP_DT_NAME) |
|
|
|
|
|
BIT(FIP_DT_FAB) | BIT(FIP_DT_FKA);
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
rlen = ntohs(fiph->fip_dl_len) * 4;
|
|
|
|
|
if (rlen + sizeof(*fiph) > skb->len)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
desc = (struct fip_desc *)(fiph + 1);
|
|
|
|
|
while (rlen > 0) {
|
|
|
|
|
dlen = desc->fip_dlen * FIP_BPW;
|
|
|
|
|
if (dlen < sizeof(*desc) || dlen > rlen)
|
|
|
|
|
return -EINVAL;
|
2010-06-12 07:44:25 +08:00
|
|
|
/* Drop Adv if there are duplicate critical descriptors */
|
|
|
|
|
if ((desc->fip_dtype < 32) &&
|
|
|
|
|
!(desc_mask & 1U << desc->fip_dtype)) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "Duplicate Critical "
|
|
|
|
|
"Descriptors in FIP adv\n");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
switch (desc->fip_dtype) {
|
|
|
|
|
case FIP_DT_PRI:
|
|
|
|
|
if (dlen != sizeof(struct fip_pri_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
fcf->pri = ((struct fip_pri_desc *)desc)->fd_pri;
|
2010-06-12 07:44:25 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_PRI);
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
case FIP_DT_MAC:
|
|
|
|
|
if (dlen != sizeof(struct fip_mac_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
memcpy(fcf->fcf_mac,
|
|
|
|
|
((struct fip_mac_desc *)desc)->fd_mac,
|
|
|
|
|
ETH_ALEN);
|
|
|
|
|
if (!is_valid_ether_addr(fcf->fcf_mac)) {
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"Invalid MAC addr %pM in FIP adv\n",
|
|
|
|
|
fcf->fcf_mac);
|
2009-03-18 02:42:40 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
2010-06-12 07:44:25 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_MAC);
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
case FIP_DT_NAME:
|
|
|
|
|
if (dlen != sizeof(struct fip_wwn_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
wwn = (struct fip_wwn_desc *)desc;
|
|
|
|
|
fcf->switch_name = get_unaligned_be64(&wwn->fd_wwn);
|
2010-06-12 07:44:25 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_NAME);
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
case FIP_DT_FAB:
|
|
|
|
|
if (dlen != sizeof(struct fip_fab_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
fab = (struct fip_fab_desc *)desc;
|
|
|
|
|
fcf->fabric_name = get_unaligned_be64(&fab->fd_wwn);
|
|
|
|
|
fcf->vfid = ntohs(fab->fd_vfid);
|
|
|
|
|
fcf->fc_map = ntoh24(fab->fd_map);
|
2010-06-12 07:44:25 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_FAB);
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
case FIP_DT_FKA:
|
|
|
|
|
if (dlen != sizeof(struct fip_fka_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
fka = (struct fip_fka_desc *)desc;
|
2009-11-21 06:54:57 +08:00
|
|
|
if (fka->fd_flags & FIP_FKA_ADV_D)
|
|
|
|
|
fcf->fd_flags = 1;
|
2009-03-18 02:42:40 +08:00
|
|
|
t = ntohl(fka->fd_fka_period);
|
|
|
|
|
if (t >= FCOE_CTLR_MIN_FKA)
|
|
|
|
|
fcf->fka_period = msecs_to_jiffies(t);
|
2010-06-12 07:44:25 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_FKA);
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
case FIP_DT_MAP_OUI:
|
|
|
|
|
case FIP_DT_FCOE_SIZE:
|
|
|
|
|
case FIP_DT_FLOGI:
|
|
|
|
|
case FIP_DT_FDISC:
|
|
|
|
|
case FIP_DT_LOGO:
|
|
|
|
|
case FIP_DT_ELP:
|
|
|
|
|
default:
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "unexpected descriptor type %x "
|
2009-06-11 06:31:05 +08:00
|
|
|
"in FIP adv\n", desc->fip_dtype);
|
2009-03-18 02:42:40 +08:00
|
|
|
/* standard says ignore unknown descriptors >= 128 */
|
|
|
|
|
if (desc->fip_dtype < FIP_DT_VENDOR_BASE)
|
|
|
|
|
return -EINVAL;
|
2010-06-12 07:43:38 +08:00
|
|
|
break;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
desc = (struct fip_desc *)((char *)desc + dlen);
|
|
|
|
|
rlen -= dlen;
|
|
|
|
|
}
|
|
|
|
|
if (!fcf->fc_map || (fcf->fc_map & 0x10000))
|
|
|
|
|
return -EINVAL;
|
2010-07-21 06:21:33 +08:00
|
|
|
if (!fcf->switch_name)
|
2009-03-18 02:42:40 +08:00
|
|
|
return -EINVAL;
|
2010-06-12 07:44:25 +08:00
|
|
|
if (desc_mask) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "adv missing descriptors mask %x\n",
|
|
|
|
|
desc_mask);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
len_err:
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP length error in descriptor type %x len %zu\n",
|
2009-06-11 06:31:05 +08:00
|
|
|
desc->fip_dtype, dlen);
|
2009-03-18 02:42:40 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv_adv() - Handle an incoming advertisement
|
|
|
|
|
* @fip: The FCoE controller receiving the advertisement
|
|
|
|
|
* @skb: The received FIP packet
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_recv_adv(struct fcoe_ctlr *fip, struct sk_buff *skb)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
struct fcoe_fcf new;
|
|
|
|
|
struct fcoe_fcf *found;
|
|
|
|
|
unsigned long sol_tov = msecs_to_jiffies(FCOE_CTRL_SOL_TOV);
|
|
|
|
|
int first = 0;
|
|
|
|
|
int mtu_valid;
|
|
|
|
|
|
2009-11-04 03:48:16 +08:00
|
|
|
if (fcoe_ctlr_parse_adv(fip, skb, &new))
|
2009-03-18 02:42:40 +08:00
|
|
|
return;
|
|
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
first = list_empty(&fip->fcfs);
|
|
|
|
|
found = NULL;
|
|
|
|
|
list_for_each_entry(fcf, &fip->fcfs, list) {
|
|
|
|
|
if (fcf->switch_name == new.switch_name &&
|
|
|
|
|
fcf->fabric_name == new.fabric_name &&
|
|
|
|
|
fcf->fc_map == new.fc_map &&
|
|
|
|
|
compare_ether_addr(fcf->fcf_mac, new.fcf_mac) == 0) {
|
|
|
|
|
found = fcf;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (!found) {
|
|
|
|
|
if (fip->fcf_count >= FCOE_CTLR_FCF_LIMIT)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
fcf = kmalloc(sizeof(*fcf), GFP_ATOMIC);
|
|
|
|
|
if (!fcf)
|
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
|
|
fip->fcf_count++;
|
|
|
|
|
memcpy(fcf, &new, sizeof(new));
|
|
|
|
|
list_add(&fcf->list, &fip->fcfs);
|
|
|
|
|
} else {
|
|
|
|
|
/*
|
2010-06-12 07:44:20 +08:00
|
|
|
* Update the FCF's keep-alive descriptor flags.
|
|
|
|
|
* Other flag changes from new advertisements are
|
|
|
|
|
* ignored after a solicited advertisement is
|
|
|
|
|
* received and the FCF is selectable (usable).
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
2010-06-12 07:44:20 +08:00
|
|
|
fcf->fd_flags = new.fd_flags;
|
|
|
|
|
if (!fcoe_ctlr_fcf_usable(fcf))
|
|
|
|
|
fcf->flags = new.flags;
|
|
|
|
|
|
2010-06-12 07:44:15 +08:00
|
|
|
if (fcf == fip->sel_fcf && !fcf->fd_flags) {
|
2009-03-18 02:42:40 +08:00
|
|
|
fip->ctlr_ka_time -= fcf->fka_period;
|
|
|
|
|
fip->ctlr_ka_time += new.fka_period;
|
|
|
|
|
if (time_before(fip->ctlr_ka_time, fip->timer.expires))
|
|
|
|
|
mod_timer(&fip->timer, fip->ctlr_ka_time);
|
2010-06-12 07:44:20 +08:00
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
fcf->fka_period = new.fka_period;
|
|
|
|
|
memcpy(fcf->fcf_mac, new.fcf_mac, ETH_ALEN);
|
|
|
|
|
}
|
|
|
|
|
mtu_valid = fcoe_ctlr_mtu_valid(fcf);
|
|
|
|
|
fcf->time = jiffies;
|
2010-12-01 08:20:02 +08:00
|
|
|
if (!found)
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "New FCF fab %16.16llx mac %pM\n",
|
|
|
|
|
fcf->fabric_name, fcf->fcf_mac);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If this advertisement is not solicited and our max receive size
|
|
|
|
|
* hasn't been verified, send a solicited advertisement.
|
|
|
|
|
*/
|
|
|
|
|
if (!mtu_valid)
|
|
|
|
|
fcoe_ctlr_solicit(fip, fcf);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If its been a while since we did a solicit, and this is
|
|
|
|
|
* the first advertisement we've received, do a multicast
|
|
|
|
|
* solicitation to gather as many advertisements as we can
|
|
|
|
|
* before selection occurs.
|
|
|
|
|
*/
|
|
|
|
|
if (first && time_after(jiffies, fip->sol_time + sol_tov))
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
|
|
|
|
|
2010-12-01 08:20:07 +08:00
|
|
|
/*
|
|
|
|
|
* Put this FCF at the head of the list for priority among equals.
|
|
|
|
|
* This helps in the case of an NPV switch which insists we use
|
|
|
|
|
* the FCF that answers multicast solicitations, not the others that
|
|
|
|
|
* are sending periodic multicast advertisements.
|
|
|
|
|
*/
|
2011-04-02 07:06:09 +08:00
|
|
|
if (mtu_valid)
|
|
|
|
|
list_move(&fcf->list, &fip->fcfs);
|
2010-12-01 08:20:07 +08:00
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
/*
|
|
|
|
|
* If this is the first validated FCF, note the time and
|
|
|
|
|
* set a timer to trigger selection.
|
|
|
|
|
*/
|
2010-06-12 07:44:15 +08:00
|
|
|
if (mtu_valid && !fip->sel_fcf && fcoe_ctlr_fcf_usable(fcf)) {
|
2009-03-18 02:42:40 +08:00
|
|
|
fip->sel_time = jiffies +
|
2009-11-04 03:47:45 +08:00
|
|
|
msecs_to_jiffies(FCOE_CTLR_START_DELAY);
|
2009-03-18 02:42:40 +08:00
|
|
|
if (!timer_pending(&fip->timer) ||
|
|
|
|
|
time_before(fip->sel_time, fip->timer.expires))
|
|
|
|
|
mod_timer(&fip->timer, fip->sel_time);
|
|
|
|
|
}
|
|
|
|
|
out:
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv_els() - Handle an incoming FIP encapsulated ELS frame
|
|
|
|
|
* @fip: The FCoE controller which received the packet
|
|
|
|
|
* @skb: The received FIP packet
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_recv_els(struct fcoe_ctlr *fip, struct sk_buff *skb)
|
|
|
|
|
{
|
2009-11-04 03:47:45 +08:00
|
|
|
struct fc_lport *lport = fip->lp;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fip_header *fiph;
|
2009-11-04 03:46:29 +08:00
|
|
|
struct fc_frame *fp = (struct fc_frame *)skb;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fc_frame_header *fh = NULL;
|
|
|
|
|
struct fip_desc *desc;
|
|
|
|
|
struct fip_encaps *els;
|
|
|
|
|
struct fcoe_dev_stats *stats;
|
|
|
|
|
enum fip_desc_type els_dtype = 0;
|
|
|
|
|
u8 els_op;
|
|
|
|
|
u8 sub;
|
|
|
|
|
u8 granted_mac[ETH_ALEN] = { 0 };
|
|
|
|
|
size_t els_len = 0;
|
|
|
|
|
size_t rlen;
|
|
|
|
|
size_t dlen;
|
2010-06-12 07:44:36 +08:00
|
|
|
u32 desc_mask = 0;
|
|
|
|
|
u32 desc_cnt = 0;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
fiph = (struct fip_header *)skb->data;
|
|
|
|
|
sub = fiph->fip_subcode;
|
|
|
|
|
if (sub != FIP_SC_REQ && sub != FIP_SC_REP)
|
|
|
|
|
goto drop;
|
|
|
|
|
|
|
|
|
|
rlen = ntohs(fiph->fip_dl_len) * 4;
|
|
|
|
|
if (rlen + sizeof(*fiph) > skb->len)
|
|
|
|
|
goto drop;
|
|
|
|
|
|
|
|
|
|
desc = (struct fip_desc *)(fiph + 1);
|
|
|
|
|
while (rlen > 0) {
|
2010-06-12 07:44:36 +08:00
|
|
|
desc_cnt++;
|
2009-03-18 02:42:40 +08:00
|
|
|
dlen = desc->fip_dlen * FIP_BPW;
|
|
|
|
|
if (dlen < sizeof(*desc) || dlen > rlen)
|
|
|
|
|
goto drop;
|
2010-06-12 07:44:25 +08:00
|
|
|
/* Drop ELS if there are duplicate critical descriptors */
|
|
|
|
|
if (desc->fip_dtype < 32) {
|
2010-06-12 07:44:36 +08:00
|
|
|
if (desc_mask & 1U << desc->fip_dtype) {
|
2010-06-12 07:44:25 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "Duplicate Critical "
|
|
|
|
|
"Descriptors in FIP ELS\n");
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
2010-06-12 07:44:36 +08:00
|
|
|
desc_mask |= (1 << desc->fip_dtype);
|
2010-06-12 07:44:25 +08:00
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
switch (desc->fip_dtype) {
|
|
|
|
|
case FIP_DT_MAC:
|
2010-06-12 07:44:36 +08:00
|
|
|
if (desc_cnt == 1) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP descriptors "
|
|
|
|
|
"received out of order\n");
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
if (dlen != sizeof(struct fip_mac_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
memcpy(granted_mac,
|
|
|
|
|
((struct fip_mac_desc *)desc)->fd_mac,
|
|
|
|
|
ETH_ALEN);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_FLOGI:
|
|
|
|
|
case FIP_DT_FDISC:
|
|
|
|
|
case FIP_DT_LOGO:
|
|
|
|
|
case FIP_DT_ELP:
|
2010-06-12 07:44:36 +08:00
|
|
|
if (desc_cnt != 1) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP descriptors "
|
|
|
|
|
"received out of order\n");
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fh)
|
|
|
|
|
goto drop;
|
|
|
|
|
if (dlen < sizeof(*els) + sizeof(*fh) + 1)
|
|
|
|
|
goto len_err;
|
|
|
|
|
els_len = dlen - sizeof(*els);
|
|
|
|
|
els = (struct fip_encaps *)desc;
|
|
|
|
|
fh = (struct fc_frame_header *)(els + 1);
|
|
|
|
|
els_dtype = desc->fip_dtype;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "unexpected descriptor type %x "
|
2009-06-11 06:31:05 +08:00
|
|
|
"in FIP adv\n", desc->fip_dtype);
|
2009-03-18 02:42:40 +08:00
|
|
|
/* standard says ignore unknown descriptors >= 128 */
|
|
|
|
|
if (desc->fip_dtype < FIP_DT_VENDOR_BASE)
|
|
|
|
|
goto drop;
|
2010-06-12 07:44:36 +08:00
|
|
|
if (desc_cnt <= 2) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP descriptors "
|
|
|
|
|
"received out of order\n");
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
2010-06-12 07:43:38 +08:00
|
|
|
break;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
desc = (struct fip_desc *)((char *)desc + dlen);
|
|
|
|
|
rlen -= dlen;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!fh)
|
|
|
|
|
goto drop;
|
|
|
|
|
els_op = *(u8 *)(fh + 1);
|
|
|
|
|
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if ((els_dtype == FIP_DT_FLOGI || els_dtype == FIP_DT_FDISC) &&
|
2010-12-01 08:19:56 +08:00
|
|
|
sub == FIP_SC_REP && fip->mode != FIP_MODE_VN2VN) {
|
|
|
|
|
if (els_op == ELS_LS_ACC) {
|
|
|
|
|
if (!is_valid_ether_addr(granted_mac)) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"Invalid MAC address %pM in FIP ELS\n",
|
|
|
|
|
granted_mac);
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
|
|
|
|
memcpy(fr_cb(fp)->granted_mac, granted_mac, ETH_ALEN);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
|
2010-12-01 08:19:56 +08:00
|
|
|
if (fip->flogi_oxid == ntohs(fh->fh_ox_id)) {
|
|
|
|
|
fip->flogi_oxid = FC_XID_UNKNOWN;
|
|
|
|
|
if (els_dtype == FIP_DT_FLOGI)
|
|
|
|
|
fcoe_ctlr_announce(fip);
|
|
|
|
|
}
|
|
|
|
|
} else if (els_dtype == FIP_DT_FLOGI &&
|
|
|
|
|
!fcoe_ctlr_flogi_retry(fip))
|
|
|
|
|
goto drop; /* retrying FLOGI so drop reject */
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-06-12 07:44:36 +08:00
|
|
|
if ((desc_cnt == 0) || ((els_op != ELS_LS_RJT) &&
|
|
|
|
|
(!(1U << FIP_DT_MAC & desc_mask)))) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "Missing critical descriptors "
|
|
|
|
|
"in FIP ELS\n");
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
/*
|
|
|
|
|
* Convert skb into an fc_frame containing only the ELS.
|
|
|
|
|
*/
|
|
|
|
|
skb_pull(skb, (u8 *)fh - skb->data);
|
|
|
|
|
skb_trim(skb, els_len);
|
|
|
|
|
fp = (struct fc_frame *)skb;
|
|
|
|
|
fc_frame_init(fp);
|
|
|
|
|
fr_sof(fp) = FC_SOF_I3;
|
|
|
|
|
fr_eof(fp) = FC_EOF_T;
|
2009-11-04 03:47:45 +08:00
|
|
|
fr_dev(fp) = lport;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
fr_encaps(fp) = els_dtype;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-03-13 08:08:55 +08:00
|
|
|
stats = per_cpu_ptr(lport->dev_stats, get_cpu());
|
2009-03-18 02:42:40 +08:00
|
|
|
stats->RxFrames++;
|
|
|
|
|
stats->RxWords += skb->len / FIP_BPW;
|
2010-03-13 08:08:55 +08:00
|
|
|
put_cpu();
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2009-11-04 03:47:45 +08:00
|
|
|
fc_exch_recv(lport, fp);
|
2009-03-18 02:42:40 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
len_err:
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP length error in descriptor type %x len %zu\n",
|
2009-06-11 06:31:05 +08:00
|
|
|
desc->fip_dtype, dlen);
|
2009-03-18 02:42:40 +08:00
|
|
|
drop:
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv_els() - Handle an incoming link reset frame
|
|
|
|
|
* @fip: The FCoE controller that received the frame
|
|
|
|
|
* @fh: The received FIP header
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* There may be multiple VN_Port descriptors.
|
|
|
|
|
* The overall length has already been checked.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_recv_clr_vlink(struct fcoe_ctlr *fip,
|
2009-11-04 03:47:45 +08:00
|
|
|
struct fip_header *fh)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fip_desc *desc;
|
|
|
|
|
struct fip_mac_desc *mp;
|
|
|
|
|
struct fip_wwn_desc *wp;
|
|
|
|
|
struct fip_vn_desc *vp;
|
|
|
|
|
size_t rlen;
|
|
|
|
|
size_t dlen;
|
|
|
|
|
struct fcoe_fcf *fcf = fip->sel_fcf;
|
2009-11-04 03:47:45 +08:00
|
|
|
struct fc_lport *lport = fip->lp;
|
2010-06-12 07:44:31 +08:00
|
|
|
struct fc_lport *vn_port = NULL;
|
|
|
|
|
u32 desc_mask;
|
2011-05-17 07:45:24 +08:00
|
|
|
int num_vlink_desc;
|
|
|
|
|
int reset_phys_port = 0;
|
|
|
|
|
struct fip_vn_desc **vlink_desc_arr = NULL;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "Clear Virtual Link received\n");
|
2010-05-08 06:18:30 +08:00
|
|
|
|
2010-05-08 06:18:41 +08:00
|
|
|
if (!fcf || !lport->port_id)
|
2009-03-18 02:42:40 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* mask of required descriptors. Validating each one clears its bit.
|
|
|
|
|
*/
|
2011-05-17 07:45:24 +08:00
|
|
|
desc_mask = BIT(FIP_DT_MAC) | BIT(FIP_DT_NAME);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
rlen = ntohs(fh->fip_dl_len) * FIP_BPW;
|
|
|
|
|
desc = (struct fip_desc *)(fh + 1);
|
2011-05-17 07:45:24 +08:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Actually need to subtract 'sizeof(*mp) - sizeof(*wp)' from 'rlen'
|
|
|
|
|
* before determining max Vx_Port descriptor but a buggy FCF could have
|
|
|
|
|
* omited either or both MAC Address and Name Identifier descriptors
|
|
|
|
|
*/
|
|
|
|
|
num_vlink_desc = rlen / sizeof(*vp);
|
|
|
|
|
if (num_vlink_desc)
|
|
|
|
|
vlink_desc_arr = kmalloc(sizeof(vp) * num_vlink_desc,
|
|
|
|
|
GFP_ATOMIC);
|
|
|
|
|
if (!vlink_desc_arr)
|
|
|
|
|
return;
|
|
|
|
|
num_vlink_desc = 0;
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
while (rlen >= sizeof(*desc)) {
|
|
|
|
|
dlen = desc->fip_dlen * FIP_BPW;
|
|
|
|
|
if (dlen > rlen)
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2010-06-12 07:44:25 +08:00
|
|
|
/* Drop CVL if there are duplicate critical descriptors */
|
|
|
|
|
if ((desc->fip_dtype < 32) &&
|
2011-05-17 07:45:24 +08:00
|
|
|
(desc->fip_dtype != FIP_DT_VN_ID) &&
|
2010-06-12 07:44:25 +08:00
|
|
|
!(desc_mask & 1U << desc->fip_dtype)) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "Duplicate Critical "
|
|
|
|
|
"Descriptors in FIP CVL\n");
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2010-06-12 07:44:25 +08:00
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
switch (desc->fip_dtype) {
|
|
|
|
|
case FIP_DT_MAC:
|
|
|
|
|
mp = (struct fip_mac_desc *)desc;
|
|
|
|
|
if (dlen < sizeof(*mp))
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2009-03-18 02:42:40 +08:00
|
|
|
if (compare_ether_addr(mp->fd_mac, fcf->fcf_mac))
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2009-03-18 02:42:40 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_MAC);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_NAME:
|
|
|
|
|
wp = (struct fip_wwn_desc *)desc;
|
|
|
|
|
if (dlen < sizeof(*wp))
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2009-03-18 02:42:40 +08:00
|
|
|
if (get_unaligned_be64(&wp->fd_wwn) != fcf->switch_name)
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2009-03-18 02:42:40 +08:00
|
|
|
desc_mask &= ~BIT(FIP_DT_NAME);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_VN_ID:
|
|
|
|
|
vp = (struct fip_vn_desc *)desc;
|
|
|
|
|
if (dlen < sizeof(*vp))
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
|
|
|
|
vlink_desc_arr[num_vlink_desc++] = vp;
|
|
|
|
|
vn_port = fc_vport_id_lookup(lport,
|
|
|
|
|
ntoh24(vp->fd_fc_id));
|
|
|
|
|
if (vn_port && (vn_port == lport)) {
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
per_cpu_ptr(lport->dev_stats,
|
|
|
|
|
get_cpu())->VLinkFailureCount++;
|
|
|
|
|
put_cpu();
|
|
|
|
|
fcoe_ctlr_reset(fip);
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2010-06-12 07:44:31 +08:00
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
/* standard says ignore unknown descriptors >= 128 */
|
|
|
|
|
if (desc->fip_dtype < FIP_DT_VENDOR_BASE)
|
2011-05-17 07:45:24 +08:00
|
|
|
goto err;
|
2009-03-18 02:42:40 +08:00
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
desc = (struct fip_desc *)((char *)desc + dlen);
|
|
|
|
|
rlen -= dlen;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* reset only if all required descriptors were present and valid.
|
|
|
|
|
*/
|
2011-05-17 07:45:24 +08:00
|
|
|
if (desc_mask)
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "missing descriptors mask %x\n",
|
|
|
|
|
desc_mask);
|
2011-05-17 07:45:24 +08:00
|
|
|
else if (!num_vlink_desc) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "CVL: no Vx_Port descriptor found\n");
|
|
|
|
|
/*
|
|
|
|
|
* No Vx_Port description. Clear all NPIV ports,
|
|
|
|
|
* followed by physical port
|
|
|
|
|
*/
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
per_cpu_ptr(lport->dev_stats,
|
|
|
|
|
get_cpu())->VLinkFailureCount++;
|
|
|
|
|
put_cpu();
|
|
|
|
|
fcoe_ctlr_reset(fip);
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
|
2012-03-10 06:49:53 +08:00
|
|
|
mutex_lock(&lport->lp_mutex);
|
|
|
|
|
list_for_each_entry(vn_port, &lport->vports, list)
|
|
|
|
|
fc_lport_reset(vn_port);
|
|
|
|
|
mutex_unlock(&lport->lp_mutex);
|
|
|
|
|
|
2011-05-17 07:45:24 +08:00
|
|
|
fc_lport_reset(fip->lp);
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
2009-03-18 02:42:40 +08:00
|
|
|
} else {
|
2011-05-17 07:45:24 +08:00
|
|
|
int i;
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
|
2011-05-17 07:45:24 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "performing Clear Virtual Link\n");
|
|
|
|
|
for (i = 0; i < num_vlink_desc; i++) {
|
|
|
|
|
vp = vlink_desc_arr[i];
|
|
|
|
|
vn_port = fc_vport_id_lookup(lport,
|
|
|
|
|
ntoh24(vp->fd_fc_id));
|
|
|
|
|
if (!vn_port)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* 'port_id' is already validated, check MAC address and
|
|
|
|
|
* wwpn
|
|
|
|
|
*/
|
|
|
|
|
if (compare_ether_addr(fip->get_src_addr(vn_port),
|
|
|
|
|
vp->fd_mac) != 0 ||
|
|
|
|
|
get_unaligned_be64(&vp->fd_wwpn) !=
|
|
|
|
|
vn_port->wwpn)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
if (vn_port == lport)
|
|
|
|
|
/*
|
|
|
|
|
* Physical port, defer processing till all
|
|
|
|
|
* listed NPIV ports are cleared
|
|
|
|
|
*/
|
|
|
|
|
reset_phys_port = 1;
|
|
|
|
|
else /* NPIV port */
|
|
|
|
|
fc_lport_reset(vn_port);
|
|
|
|
|
}
|
2010-06-12 07:44:31 +08:00
|
|
|
|
2011-05-17 07:45:24 +08:00
|
|
|
if (reset_phys_port) {
|
2010-06-12 07:44:31 +08:00
|
|
|
fc_lport_reset(fip->lp);
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
2011-05-17 07:45:24 +08:00
|
|
|
|
|
|
|
|
err:
|
|
|
|
|
kfree(vlink_desc_arr);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv() - Receive a FIP packet
|
|
|
|
|
* @fip: The FCoE controller that received the packet
|
|
|
|
|
* @skb: The received FIP packet
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
2009-11-04 03:48:22 +08:00
|
|
|
* This may be called from either NET_RX_SOFTIRQ or IRQ.
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
void fcoe_ctlr_recv(struct fcoe_ctlr *fip, struct sk_buff *skb)
|
|
|
|
|
{
|
2009-11-04 03:48:22 +08:00
|
|
|
skb_queue_tail(&fip->fip_recv_list, skb);
|
2009-03-18 02:42:40 +08:00
|
|
|
schedule_work(&fip->recv_work);
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_recv);
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv_handler() - Receive a FIP frame
|
|
|
|
|
* @fip: The FCoE controller that received the frame
|
|
|
|
|
* @skb: The received FIP frame
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Returns non-zero if the frame is dropped.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_recv_handler(struct fcoe_ctlr *fip, struct sk_buff *skb)
|
|
|
|
|
{
|
|
|
|
|
struct fip_header *fiph;
|
|
|
|
|
struct ethhdr *eh;
|
|
|
|
|
enum fip_state state;
|
|
|
|
|
u16 op;
|
|
|
|
|
u8 sub;
|
|
|
|
|
|
|
|
|
|
if (skb_linearize(skb))
|
|
|
|
|
goto drop;
|
|
|
|
|
if (skb->len < sizeof(*fiph))
|
|
|
|
|
goto drop;
|
|
|
|
|
eh = eth_hdr(skb);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN) {
|
|
|
|
|
if (compare_ether_addr(eh->h_dest, fip->ctl_src_addr) &&
|
|
|
|
|
compare_ether_addr(eh->h_dest, fcoe_all_vn2vn) &&
|
|
|
|
|
compare_ether_addr(eh->h_dest, fcoe_all_p2p))
|
|
|
|
|
goto drop;
|
|
|
|
|
} else if (compare_ether_addr(eh->h_dest, fip->ctl_src_addr) &&
|
|
|
|
|
compare_ether_addr(eh->h_dest, fcoe_all_enode))
|
2009-03-18 02:42:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
fiph = (struct fip_header *)skb->data;
|
|
|
|
|
op = ntohs(fiph->fip_op);
|
|
|
|
|
sub = fiph->fip_subcode;
|
|
|
|
|
|
|
|
|
|
if (FIP_VER_DECAPS(fiph->fip_ver) != FIP_VER)
|
|
|
|
|
goto drop;
|
|
|
|
|
if (ntohs(fiph->fip_dl_len) * FIP_BPW + sizeof(*fiph) > skb->len)
|
|
|
|
|
goto drop;
|
|
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
state = fip->state;
|
|
|
|
|
if (state == FIP_ST_AUTO) {
|
|
|
|
|
fip->map_dest = 0;
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_ENABLED);
|
2009-03-18 02:42:40 +08:00
|
|
|
state = FIP_ST_ENABLED;
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "Using FIP mode\n");
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
|
|
|
|
|
if (fip->mode == FIP_MODE_VN2VN && op == FIP_OP_VN2VN)
|
|
|
|
|
return fcoe_ctlr_vn_recv(fip, skb);
|
|
|
|
|
|
|
|
|
|
if (state != FIP_ST_ENABLED && state != FIP_ST_VNMP_UP &&
|
|
|
|
|
state != FIP_ST_VNMP_CLAIM)
|
2009-03-18 02:42:40 +08:00
|
|
|
goto drop;
|
|
|
|
|
|
|
|
|
|
if (op == FIP_OP_LS) {
|
|
|
|
|
fcoe_ctlr_recv_els(fip, skb); /* consumes skb */
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
|
|
|
|
|
if (state != FIP_ST_ENABLED)
|
|
|
|
|
goto drop;
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
if (op == FIP_OP_DISC && sub == FIP_SC_ADV)
|
|
|
|
|
fcoe_ctlr_recv_adv(fip, skb);
|
|
|
|
|
else if (op == FIP_OP_CTRL && sub == FIP_SC_CLR_VLINK)
|
|
|
|
|
fcoe_ctlr_recv_clr_vlink(fip, fiph);
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
return 0;
|
|
|
|
|
drop:
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_select() - Select the best FCF (if possible)
|
|
|
|
|
* @fip: The FCoE controller
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
2010-12-01 08:20:12 +08:00
|
|
|
* Returns the selected FCF, or NULL if none are usable.
|
|
|
|
|
*
|
2009-03-18 02:42:40 +08:00
|
|
|
* If there are conflicting advertisements, no FCF can be chosen.
|
|
|
|
|
*
|
2010-12-01 08:19:56 +08:00
|
|
|
* If there is already a selected FCF, this will choose a better one or
|
|
|
|
|
* an equivalent one that hasn't already been sent a FLOGI.
|
|
|
|
|
*
|
2009-03-18 02:42:40 +08:00
|
|
|
* Called with lock held.
|
|
|
|
|
*/
|
2010-12-01 08:20:12 +08:00
|
|
|
static struct fcoe_fcf *fcoe_ctlr_select(struct fcoe_ctlr *fip)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
2010-12-01 08:19:56 +08:00
|
|
|
struct fcoe_fcf *best = fip->sel_fcf;
|
2010-12-01 08:19:51 +08:00
|
|
|
struct fcoe_fcf *first;
|
|
|
|
|
|
|
|
|
|
first = list_first_entry(&fip->fcfs, struct fcoe_fcf, list);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
list_for_each_entry(fcf, &fip->fcfs, list) {
|
2010-12-01 08:20:02 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "consider FCF fab %16.16llx "
|
|
|
|
|
"VFID %d mac %pM map %x val %d "
|
|
|
|
|
"sent %u pri %u\n",
|
|
|
|
|
fcf->fabric_name, fcf->vfid, fcf->fcf_mac,
|
|
|
|
|
fcf->fc_map, fcoe_ctlr_mtu_valid(fcf),
|
|
|
|
|
fcf->flogi_sent, fcf->pri);
|
2010-12-01 08:19:51 +08:00
|
|
|
if (fcf->fabric_name != first->fabric_name ||
|
|
|
|
|
fcf->vfid != first->vfid ||
|
|
|
|
|
fcf->fc_map != first->fc_map) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "Conflicting fabric, VFID, "
|
|
|
|
|
"or FC-MAP\n");
|
|
|
|
|
return NULL;
|
|
|
|
|
}
|
2010-12-01 08:19:56 +08:00
|
|
|
if (fcf->flogi_sent)
|
|
|
|
|
continue;
|
2009-03-18 02:42:40 +08:00
|
|
|
if (!fcoe_ctlr_fcf_usable(fcf)) {
|
2010-04-10 05:23:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "FCF for fab %16.16llx "
|
|
|
|
|
"map %x %svalid %savailable\n",
|
|
|
|
|
fcf->fabric_name, fcf->fc_map,
|
|
|
|
|
(fcf->flags & FIP_FL_SOL) ? "" : "in",
|
|
|
|
|
(fcf->flags & FIP_FL_AVAIL) ?
|
|
|
|
|
"" : "un");
|
2009-03-18 02:42:40 +08:00
|
|
|
continue;
|
|
|
|
|
}
|
2010-12-01 08:19:56 +08:00
|
|
|
if (!best || fcf->pri < best->pri || best->flogi_sent)
|
2009-03-18 02:42:40 +08:00
|
|
|
best = fcf;
|
|
|
|
|
}
|
|
|
|
|
fip->sel_fcf = best;
|
2010-12-01 08:19:46 +08:00
|
|
|
if (best) {
|
2010-12-01 08:20:02 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip, "using FCF mac %pM\n", best->fcf_mac);
|
2010-12-01 08:19:46 +08:00
|
|
|
fip->port_ka_time = jiffies +
|
|
|
|
|
msecs_to_jiffies(FIP_VN_KA_PERIOD);
|
|
|
|
|
fip->ctlr_ka_time = jiffies + best->fka_period;
|
|
|
|
|
if (time_before(fip->ctlr_ka_time, fip->timer.expires))
|
|
|
|
|
mod_timer(&fip->timer, fip->ctlr_ka_time);
|
|
|
|
|
}
|
2010-12-01 08:20:12 +08:00
|
|
|
return best;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
2010-12-01 08:19:56 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_flogi_send_locked() - send FIP-encapsulated FLOGI to current FCF
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Returns non-zero error if it could not be sent.
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex and ctlr_lock held.
|
|
|
|
|
* Caller must verify that fip->sel_fcf is not NULL.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_flogi_send_locked(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
struct sk_buff *skb_orig;
|
|
|
|
|
struct fc_frame_header *fh;
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
skb_orig = fip->flogi_req;
|
|
|
|
|
if (!skb_orig)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Clone and send the FLOGI request. If clone fails, use original.
|
|
|
|
|
*/
|
|
|
|
|
skb = skb_clone(skb_orig, GFP_ATOMIC);
|
|
|
|
|
if (!skb) {
|
|
|
|
|
skb = skb_orig;
|
|
|
|
|
fip->flogi_req = NULL;
|
|
|
|
|
}
|
|
|
|
|
fh = (struct fc_frame_header *)skb->data;
|
|
|
|
|
error = fcoe_ctlr_encaps(fip, fip->lp, FIP_DT_FLOGI, skb,
|
|
|
|
|
ntoh24(fh->fh_d_id));
|
|
|
|
|
if (error) {
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
return error;
|
|
|
|
|
}
|
|
|
|
|
fip->send(fip, skb);
|
|
|
|
|
fip->sel_fcf->flogi_sent = 1;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_flogi_retry() - resend FLOGI request to a new FCF if possible
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Returns non-zero error code if there's no FLOGI request to retry or
|
|
|
|
|
* no alternate FCF available.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_flogi_retry(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
int error;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
spin_lock_bh(&fip->ctlr_lock);
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "re-sending FLOGI - reselect\n");
|
2010-12-01 08:20:12 +08:00
|
|
|
fcf = fcoe_ctlr_select(fip);
|
2010-12-01 08:19:56 +08:00
|
|
|
if (!fcf || fcf->flogi_sent) {
|
|
|
|
|
kfree_skb(fip->flogi_req);
|
|
|
|
|
fip->flogi_req = NULL;
|
|
|
|
|
error = -ENOENT;
|
|
|
|
|
} else {
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
|
|
|
|
error = fcoe_ctlr_flogi_send_locked(fip);
|
|
|
|
|
}
|
|
|
|
|
spin_unlock_bh(&fip->ctlr_lock);
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
return error;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_flogi_send() - Handle sending of FIP FLOGI.
|
|
|
|
|
* @fip: The FCoE controller that timed out
|
|
|
|
|
*
|
|
|
|
|
* Done here because fcoe_ctlr_els_send() can't get mutex.
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held. The caller must not hold ctlr_lock.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_flogi_send(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_fcf *fcf;
|
|
|
|
|
|
|
|
|
|
spin_lock_bh(&fip->ctlr_lock);
|
|
|
|
|
fcf = fip->sel_fcf;
|
|
|
|
|
if (!fcf || !fip->flogi_req_send)
|
|
|
|
|
goto unlock;
|
|
|
|
|
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "sending FLOGI\n");
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* If this FLOGI is being sent due to a timeout retry
|
|
|
|
|
* to the same FCF as before, select a different FCF if possible.
|
|
|
|
|
*/
|
|
|
|
|
if (fcf->flogi_sent) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "sending FLOGI - reselect\n");
|
2010-12-01 08:20:12 +08:00
|
|
|
fcf = fcoe_ctlr_select(fip);
|
2010-12-01 08:19:56 +08:00
|
|
|
if (!fcf || fcf->flogi_sent) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "sending FLOGI - clearing\n");
|
|
|
|
|
list_for_each_entry(fcf, &fip->fcfs, list)
|
|
|
|
|
fcf->flogi_sent = 0;
|
2010-12-01 08:20:12 +08:00
|
|
|
fcf = fcoe_ctlr_select(fip);
|
2010-12-01 08:19:56 +08:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (fcf) {
|
|
|
|
|
fcoe_ctlr_flogi_send_locked(fip);
|
|
|
|
|
fip->flogi_req_send = 0;
|
|
|
|
|
} else /* XXX */
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "No FCF selected - defer send\n");
|
|
|
|
|
unlock:
|
|
|
|
|
spin_unlock_bh(&fip->ctlr_lock);
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:40 +08:00
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_timeout() - FIP timeout handler
|
|
|
|
|
* @arg: The FCoE controller that timed out
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_timeout(unsigned long arg)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_ctlr *fip = (struct fcoe_ctlr *)arg;
|
2010-07-21 06:19:47 +08:00
|
|
|
|
|
|
|
|
schedule_work(&fip->timer_work);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_timer_work() - Worker thread function for timer work
|
|
|
|
|
* @work: Handle to a FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Ages FCFs. Triggers FCF selection if possible.
|
|
|
|
|
* Sends keep-alives and resets.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_timer_work(struct work_struct *work)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_ctlr *fip;
|
|
|
|
|
struct fc_lport *vport;
|
|
|
|
|
u8 *mac;
|
|
|
|
|
u8 reset = 0;
|
|
|
|
|
u8 send_ctlr_ka = 0;
|
|
|
|
|
u8 send_port_ka = 0;
|
2009-03-18 02:42:40 +08:00
|
|
|
struct fcoe_fcf *sel;
|
|
|
|
|
struct fcoe_fcf *fcf;
|
2010-06-12 07:44:10 +08:00
|
|
|
unsigned long next_timer;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
fip = container_of(work, struct fcoe_ctlr, timer_work);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN)
|
|
|
|
|
return fcoe_ctlr_vn_timeout(fip);
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fip->state == FIP_ST_DISABLED) {
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
fcf = fip->sel_fcf;
|
2010-06-12 07:44:10 +08:00
|
|
|
next_timer = fcoe_ctlr_age_fcfs(fip);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
sel = fip->sel_fcf;
|
2010-06-12 07:44:10 +08:00
|
|
|
if (!sel && fip->sel_time) {
|
|
|
|
|
if (time_after_eq(jiffies, fip->sel_time)) {
|
2010-12-01 08:20:12 +08:00
|
|
|
sel = fcoe_ctlr_select(fip);
|
2010-06-12 07:44:10 +08:00
|
|
|
fip->sel_time = 0;
|
|
|
|
|
} else if (time_after(next_timer, fip->sel_time))
|
|
|
|
|
next_timer = fip->sel_time;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
2010-12-01 08:19:56 +08:00
|
|
|
if (sel && fip->flogi_req_send)
|
|
|
|
|
fcoe_ctlr_flogi_send(fip);
|
|
|
|
|
else if (!sel && fcf)
|
|
|
|
|
reset = 1;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2009-11-21 06:54:57 +08:00
|
|
|
if (sel && !sel->fd_flags) {
|
2009-03-18 02:42:40 +08:00
|
|
|
if (time_after_eq(jiffies, fip->ctlr_ka_time)) {
|
|
|
|
|
fip->ctlr_ka_time = jiffies + sel->fka_period;
|
2010-07-21 06:19:47 +08:00
|
|
|
send_ctlr_ka = 1;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
if (time_after(next_timer, fip->ctlr_ka_time))
|
|
|
|
|
next_timer = fip->ctlr_ka_time;
|
|
|
|
|
|
|
|
|
|
if (time_after_eq(jiffies, fip->port_ka_time)) {
|
2010-01-22 02:16:00 +08:00
|
|
|
fip->port_ka_time = jiffies +
|
2009-11-04 03:47:45 +08:00
|
|
|
msecs_to_jiffies(FIP_VN_KA_PERIOD);
|
2010-07-21 06:19:47 +08:00
|
|
|
send_port_ka = 1;
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
if (time_after(next_timer, fip->port_ka_time))
|
|
|
|
|
next_timer = fip->port_ka_time;
|
|
|
|
|
}
|
2010-06-12 07:44:10 +08:00
|
|
|
if (!list_empty(&fip->fcfs))
|
|
|
|
|
mod_timer(&fip->timer, next_timer);
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2010-06-12 07:43:44 +08:00
|
|
|
if (reset) {
|
[SCSI] libfcoe: FIP should report link to libfc whether selected or not
The fnic driver with FIP is reporting link up, even though it's down.
When the interface is shut down by the switch, we receive a clear
virtual link, and set the state reported to libfc as down, although
we still report it up. Clearly wrong. That causes the subsequent
link down event not to be reported, and /sys shows the host "Online".
Currently, in FIP mode, if an FCF times out, then link to libfc
is reported as down, to stop FLOGIs. That interferes with the LLD
link down being reported.
Users really need to know the physical link information, to diagnose
cabling issues, so physical link status should be reported to libfc.
If the selected FCF needs to be reported, that should be done
separately, in a later patch.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2009-11-04 03:48:27 +08:00
|
|
|
fc_lport_reset(fip->lp);
|
2010-06-12 07:43:44 +08:00
|
|
|
/* restart things with a solicitation */
|
|
|
|
|
fcoe_ctlr_solicit(fip, NULL);
|
|
|
|
|
}
|
2009-11-04 03:46:29 +08:00
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
if (send_ctlr_ka)
|
2009-11-04 03:46:29 +08:00
|
|
|
fcoe_ctlr_send_keep_alive(fip, NULL, 0, fip->ctl_src_addr);
|
2010-07-21 06:19:47 +08:00
|
|
|
|
|
|
|
|
if (send_port_ka) {
|
2009-11-04 03:46:29 +08:00
|
|
|
mutex_lock(&fip->lp->lp_mutex);
|
|
|
|
|
mac = fip->get_src_addr(fip->lp);
|
|
|
|
|
fcoe_ctlr_send_keep_alive(fip, fip->lp, 1, mac);
|
|
|
|
|
list_for_each_entry(vport, &fip->lp->vports, list) {
|
|
|
|
|
mac = fip->get_src_addr(vport);
|
|
|
|
|
fcoe_ctlr_send_keep_alive(fip, vport, 1, mac);
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&fip->lp->lp_mutex);
|
|
|
|
|
}
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_ctlr_recv_work() - Worker thread function for receiving FIP frames
|
|
|
|
|
* @recv_work: Handle to a FCoE controller
|
2009-03-18 02:42:40 +08:00
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_recv_work(struct work_struct *recv_work)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_ctlr *fip;
|
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
|
|
|
|
|
fip = container_of(recv_work, struct fcoe_ctlr, recv_work);
|
2009-11-04 03:48:22 +08:00
|
|
|
while ((skb = skb_dequeue(&fip->fip_recv_list)))
|
2009-03-18 02:42:40 +08:00
|
|
|
fcoe_ctlr_recv_handler(fip, skb);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
2009-11-04 03:49:16 +08:00
|
|
|
* fcoe_ctlr_recv_flogi() - Snoop pre-FIP receipt of FLOGI response
|
2009-11-04 03:47:45 +08:00
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @fp: The FC frame to snoop
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Snoop potential response to FLOGI or even incoming FLOGI.
|
|
|
|
|
*
|
|
|
|
|
* The caller has checked that we are waiting for login as indicated
|
|
|
|
|
* by fip->flogi_oxid != FC_XID_UNKNOWN.
|
|
|
|
|
*
|
|
|
|
|
* The caller is responsible for freeing the frame.
|
2009-11-04 03:49:16 +08:00
|
|
|
* Fill in the granted_mac address.
|
2009-03-18 02:42:40 +08:00
|
|
|
*
|
|
|
|
|
* Return non-zero if the frame should not be delivered to libfc.
|
|
|
|
|
*/
|
2009-11-04 03:46:29 +08:00
|
|
|
int fcoe_ctlr_recv_flogi(struct fcoe_ctlr *fip, struct fc_lport *lport,
|
2009-11-04 03:49:16 +08:00
|
|
|
struct fc_frame *fp)
|
2009-03-18 02:42:40 +08:00
|
|
|
{
|
|
|
|
|
struct fc_frame_header *fh;
|
|
|
|
|
u8 op;
|
2009-11-04 03:49:16 +08:00
|
|
|
u8 *sa;
|
2009-03-18 02:42:40 +08:00
|
|
|
|
2009-11-04 03:49:16 +08:00
|
|
|
sa = eth_hdr(&fp->skb)->h_source;
|
2009-03-18 02:42:40 +08:00
|
|
|
fh = fc_frame_header_get(fp);
|
|
|
|
|
if (fh->fh_type != FC_TYPE_ELS)
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
op = fc_frame_payload_op(fp);
|
|
|
|
|
if (op == ELS_LS_ACC && fh->fh_r_ctl == FC_RCTL_ELS_REP &&
|
|
|
|
|
fip->flogi_oxid == ntohs(fh->fh_ox_id)) {
|
|
|
|
|
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fip->state != FIP_ST_AUTO && fip->state != FIP_ST_NON_FIP) {
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_NON_FIP);
|
2009-11-04 03:48:16 +08:00
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"received FLOGI LS_ACC using non-FIP mode\n");
|
2009-03-18 02:42:40 +08:00
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* FLOGI accepted.
|
|
|
|
|
* If the src mac addr is FC_OUI-based, then we mark the
|
|
|
|
|
* address_mode flag to use FC_OUI-based Ethernet DA.
|
|
|
|
|
* Otherwise we use the FCoE gateway addr
|
|
|
|
|
*/
|
|
|
|
|
if (!compare_ether_addr(sa, (u8[6])FC_FCOE_FLOGI_MAC)) {
|
2010-07-21 06:20:40 +08:00
|
|
|
fcoe_ctlr_map_dest(fip);
|
2009-03-18 02:42:40 +08:00
|
|
|
} else {
|
|
|
|
|
memcpy(fip->dest_addr, sa, ETH_ALEN);
|
|
|
|
|
fip->map_dest = 0;
|
|
|
|
|
}
|
|
|
|
|
fip->flogi_oxid = FC_XID_UNKNOWN;
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-11-04 03:49:16 +08:00
|
|
|
fc_fcoe_set_mac(fr_cb(fp)->granted_mac, fh->fh_d_id);
|
2009-03-18 02:42:40 +08:00
|
|
|
} else if (op == ELS_FLOGI && fh->fh_r_ctl == FC_RCTL_ELS_REQ && sa) {
|
|
|
|
|
/*
|
|
|
|
|
* Save source MAC for point-to-point responses.
|
|
|
|
|
*/
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
if (fip->state == FIP_ST_AUTO || fip->state == FIP_ST_NON_FIP) {
|
|
|
|
|
memcpy(fip->dest_addr, sa, ETH_ALEN);
|
|
|
|
|
fip->map_dest = 0;
|
2010-03-13 08:07:57 +08:00
|
|
|
if (fip->state == FIP_ST_AUTO)
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "received non-FIP FLOGI. "
|
|
|
|
|
"Setting non-FIP mode\n");
|
2010-07-21 06:20:24 +08:00
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_NON_FIP);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
2010-07-21 06:19:47 +08:00
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
2009-03-18 02:42:40 +08:00
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL(fcoe_ctlr_recv_flogi);
|
|
|
|
|
|
2009-03-18 02:42:18 +08:00
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_wwn_from_mac() - Converts a 48-bit IEEE MAC address to a 64-bit FC WWN
|
|
|
|
|
* @mac: The MAC address to convert
|
|
|
|
|
* @scheme: The scheme to use when converting
|
|
|
|
|
* @port: The port indicator for converting
|
2009-03-18 02:42:18 +08:00
|
|
|
*
|
|
|
|
|
* Returns: u64 fc world wide name
|
|
|
|
|
*/
|
|
|
|
|
u64 fcoe_wwn_from_mac(unsigned char mac[MAX_ADDR_LEN],
|
|
|
|
|
unsigned int scheme, unsigned int port)
|
|
|
|
|
{
|
|
|
|
|
u64 wwn;
|
|
|
|
|
u64 host_mac;
|
|
|
|
|
|
|
|
|
|
/* The MAC is in NO, so flip only the low 48 bits */
|
|
|
|
|
host_mac = ((u64) mac[0] << 40) |
|
|
|
|
|
((u64) mac[1] << 32) |
|
|
|
|
|
((u64) mac[2] << 24) |
|
|
|
|
|
((u64) mac[3] << 16) |
|
|
|
|
|
((u64) mac[4] << 8) |
|
|
|
|
|
(u64) mac[5];
|
|
|
|
|
|
|
|
|
|
WARN_ON(host_mac >= (1ULL << 48));
|
|
|
|
|
wwn = host_mac | ((u64) scheme << 60);
|
|
|
|
|
switch (scheme) {
|
|
|
|
|
case 1:
|
|
|
|
|
WARN_ON(port != 0);
|
|
|
|
|
break;
|
|
|
|
|
case 2:
|
|
|
|
|
WARN_ON(port >= 0xfff);
|
|
|
|
|
wwn |= (u64) port << 48;
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
WARN_ON(1);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return wwn;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(fcoe_wwn_from_mac);
|
|
|
|
|
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_rport() - return the fcoe_rport for a given fc_rport_priv
|
|
|
|
|
* @rdata: libfc remote port
|
|
|
|
|
*/
|
|
|
|
|
static inline struct fcoe_rport *fcoe_ctlr_rport(struct fc_rport_priv *rdata)
|
|
|
|
|
{
|
|
|
|
|
return (struct fcoe_rport *)(rdata + 1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_send() - Send a FIP VN2VN Probe Request or Reply.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @sub: sub-opcode for probe request, reply, or advertisement.
|
|
|
|
|
* @dest: The destination Ethernet MAC address
|
|
|
|
|
* @min_len: minimum size of the Ethernet payload to be sent
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_send(struct fcoe_ctlr *fip,
|
|
|
|
|
enum fip_vn2vn_subcode sub,
|
|
|
|
|
const u8 *dest, size_t min_len)
|
|
|
|
|
{
|
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
struct fip_frame {
|
|
|
|
|
struct ethhdr eth;
|
|
|
|
|
struct fip_header fip;
|
|
|
|
|
struct fip_mac_desc mac;
|
|
|
|
|
struct fip_wwn_desc wwnn;
|
|
|
|
|
struct fip_vn_desc vn;
|
2011-01-29 08:05:00 +08:00
|
|
|
} __packed * frame;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
struct fip_fc4_feat *ff;
|
|
|
|
|
struct fip_size_desc *size;
|
|
|
|
|
u32 fcp_feat;
|
|
|
|
|
size_t len;
|
|
|
|
|
size_t dlen;
|
|
|
|
|
|
|
|
|
|
len = sizeof(*frame);
|
|
|
|
|
dlen = 0;
|
|
|
|
|
if (sub == FIP_SC_VN_CLAIM_NOTIFY || sub == FIP_SC_VN_CLAIM_REP) {
|
|
|
|
|
dlen = sizeof(struct fip_fc4_feat) +
|
|
|
|
|
sizeof(struct fip_size_desc);
|
|
|
|
|
len += dlen;
|
|
|
|
|
}
|
|
|
|
|
dlen += sizeof(frame->mac) + sizeof(frame->wwnn) + sizeof(frame->vn);
|
|
|
|
|
len = max(len, min_len + sizeof(struct ethhdr));
|
|
|
|
|
|
|
|
|
|
skb = dev_alloc_skb(len);
|
|
|
|
|
if (!skb)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
frame = (struct fip_frame *)skb->data;
|
|
|
|
|
memset(frame, 0, len);
|
|
|
|
|
memcpy(frame->eth.h_dest, dest, ETH_ALEN);
|
|
|
|
|
memcpy(frame->eth.h_source, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
frame->eth.h_proto = htons(ETH_P_FIP);
|
|
|
|
|
|
|
|
|
|
frame->fip.fip_ver = FIP_VER_ENCAPS(FIP_VER);
|
|
|
|
|
frame->fip.fip_op = htons(FIP_OP_VN2VN);
|
|
|
|
|
frame->fip.fip_subcode = sub;
|
|
|
|
|
frame->fip.fip_dl_len = htons(dlen / FIP_BPW);
|
|
|
|
|
|
|
|
|
|
frame->mac.fd_desc.fip_dtype = FIP_DT_MAC;
|
|
|
|
|
frame->mac.fd_desc.fip_dlen = sizeof(frame->mac) / FIP_BPW;
|
|
|
|
|
memcpy(frame->mac.fd_mac, fip->ctl_src_addr, ETH_ALEN);
|
|
|
|
|
|
|
|
|
|
frame->wwnn.fd_desc.fip_dtype = FIP_DT_NAME;
|
|
|
|
|
frame->wwnn.fd_desc.fip_dlen = sizeof(frame->wwnn) / FIP_BPW;
|
|
|
|
|
put_unaligned_be64(fip->lp->wwnn, &frame->wwnn.fd_wwn);
|
|
|
|
|
|
|
|
|
|
frame->vn.fd_desc.fip_dtype = FIP_DT_VN_ID;
|
|
|
|
|
frame->vn.fd_desc.fip_dlen = sizeof(frame->vn) / FIP_BPW;
|
|
|
|
|
hton24(frame->vn.fd_mac, FIP_VN_FC_MAP);
|
|
|
|
|
hton24(frame->vn.fd_mac + 3, fip->port_id);
|
|
|
|
|
hton24(frame->vn.fd_fc_id, fip->port_id);
|
|
|
|
|
put_unaligned_be64(fip->lp->wwpn, &frame->vn.fd_wwpn);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* For claims, add FC-4 features.
|
|
|
|
|
* TBD: Add interface to get fc-4 types and features from libfc.
|
|
|
|
|
*/
|
|
|
|
|
if (sub == FIP_SC_VN_CLAIM_NOTIFY || sub == FIP_SC_VN_CLAIM_REP) {
|
|
|
|
|
ff = (struct fip_fc4_feat *)(frame + 1);
|
|
|
|
|
ff->fd_desc.fip_dtype = FIP_DT_FC4F;
|
|
|
|
|
ff->fd_desc.fip_dlen = sizeof(*ff) / FIP_BPW;
|
|
|
|
|
ff->fd_fts = fip->lp->fcts;
|
|
|
|
|
|
|
|
|
|
fcp_feat = 0;
|
|
|
|
|
if (fip->lp->service_params & FCP_SPPF_INIT_FCN)
|
|
|
|
|
fcp_feat |= FCP_FEAT_INIT;
|
|
|
|
|
if (fip->lp->service_params & FCP_SPPF_TARG_FCN)
|
|
|
|
|
fcp_feat |= FCP_FEAT_TARG;
|
|
|
|
|
fcp_feat <<= (FC_TYPE_FCP * 4) % 32;
|
|
|
|
|
ff->fd_ff.fd_feat[FC_TYPE_FCP * 4 / 32] = htonl(fcp_feat);
|
|
|
|
|
|
|
|
|
|
size = (struct fip_size_desc *)(ff + 1);
|
|
|
|
|
size->fd_desc.fip_dtype = FIP_DT_FCOE_SIZE;
|
|
|
|
|
size->fd_desc.fip_dlen = sizeof(*size) / FIP_BPW;
|
|
|
|
|
size->fd_size = htons(fcoe_ctlr_fcoe_size(fip));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
skb_put(skb, len);
|
|
|
|
|
skb->protocol = htons(ETH_P_FIP);
|
2011-11-19 05:35:56 +08:00
|
|
|
skb->priority = fip->priority;
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
skb_reset_mac_header(skb);
|
|
|
|
|
skb_reset_network_header(skb);
|
|
|
|
|
|
|
|
|
|
fip->send(fip, skb);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_rport_callback - Event handler for rport events.
|
|
|
|
|
* @lport: The lport which is receiving the event
|
|
|
|
|
* @rdata: remote port private data
|
2011-03-31 09:57:33 +08:00
|
|
|
* @event: The event that occurred
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
*
|
|
|
|
|
* Locking Note: The rport lock must not be held when calling this function.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_rport_callback(struct fc_lport *lport,
|
|
|
|
|
struct fc_rport_priv *rdata,
|
|
|
|
|
enum fc_rport_event event)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_ctlr *fip = lport->disc.priv;
|
|
|
|
|
struct fcoe_rport *frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "vn_rport_callback %x event %d\n",
|
|
|
|
|
rdata->ids.port_id, event);
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
switch (event) {
|
|
|
|
|
case RPORT_EV_READY:
|
|
|
|
|
frport->login_count = 0;
|
|
|
|
|
break;
|
|
|
|
|
case RPORT_EV_LOGO:
|
|
|
|
|
case RPORT_EV_FAILED:
|
|
|
|
|
case RPORT_EV_STOP:
|
|
|
|
|
frport->login_count++;
|
|
|
|
|
if (frport->login_count > FCOE_CTLR_VN2VN_LOGIN_LIMIT) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"rport FLOGI limited port_id %6.6x\n",
|
|
|
|
|
rdata->ids.port_id);
|
|
|
|
|
lport->tt.rport_logoff(rdata);
|
|
|
|
|
}
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static struct fc_rport_operations fcoe_ctlr_vn_rport_ops = {
|
|
|
|
|
.event_callback = fcoe_ctlr_vn_rport_callback,
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_disc_stop_locked() - stop discovery in VN2VN mode
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_disc_stop_locked(struct fc_lport *lport)
|
|
|
|
|
{
|
|
|
|
|
mutex_lock(&lport->disc.disc_mutex);
|
|
|
|
|
lport->disc.disc_callback = NULL;
|
|
|
|
|
mutex_unlock(&lport->disc.disc_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_disc_stop() - stop discovery in VN2VN mode
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called through the local port template for discovery.
|
|
|
|
|
* Called without the ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_disc_stop(struct fc_lport *lport)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_ctlr *fip = lport->disc.priv;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
fcoe_ctlr_disc_stop_locked(lport);
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_disc_stop_final() - stop discovery for shutdown in VN2VN mode
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called through the local port template for discovery.
|
|
|
|
|
* Called without the ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_disc_stop_final(struct fc_lport *lport)
|
|
|
|
|
{
|
|
|
|
|
fcoe_ctlr_disc_stop(lport);
|
|
|
|
|
lport->tt.rport_flush_queue();
|
|
|
|
|
synchronize_rcu();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_restart() - VN2VN probe restart with new port_id
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called with fcoe_ctlr lock held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_restart(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
unsigned long wait;
|
|
|
|
|
u32 port_id;
|
|
|
|
|
|
|
|
|
|
fcoe_ctlr_disc_stop_locked(fip->lp);
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Get proposed port ID.
|
|
|
|
|
* If this is the first try after link up, use any previous port_id.
|
|
|
|
|
* If there was none, use the low bits of the port_name.
|
|
|
|
|
* On subsequent tries, get the next random one.
|
|
|
|
|
* Don't use reserved IDs, use another non-zero value, just as random.
|
|
|
|
|
*/
|
|
|
|
|
port_id = fip->port_id;
|
|
|
|
|
if (fip->probe_tries)
|
|
|
|
|
port_id = prandom32(&fip->rnd_state) & 0xffff;
|
|
|
|
|
else if (!port_id)
|
|
|
|
|
port_id = fip->lp->wwpn & 0xffff;
|
|
|
|
|
if (!port_id || port_id == 0xffff)
|
|
|
|
|
port_id = 1;
|
|
|
|
|
fip->port_id = port_id;
|
|
|
|
|
|
|
|
|
|
if (fip->probe_tries < FIP_VN_RLIM_COUNT) {
|
|
|
|
|
fip->probe_tries++;
|
|
|
|
|
wait = random32() % FIP_VN_PROBE_WAIT;
|
|
|
|
|
} else
|
|
|
|
|
wait = FIP_VN_RLIM_INT;
|
|
|
|
|
mod_timer(&fip->timer, jiffies + msecs_to_jiffies(wait));
|
|
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_VNMP_START);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_start() - Start in VN2VN mode
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called with fcoe_ctlr lock held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_start(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
fip->probe_tries = 0;
|
|
|
|
|
prandom32_seed(&fip->rnd_state, fip->lp->wwpn);
|
|
|
|
|
fcoe_ctlr_vn_restart(fip);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_parse - parse probe request or response
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @skb: incoming packet
|
|
|
|
|
* @rdata: buffer for resulting parsed VN entry plus fcoe_rport
|
|
|
|
|
*
|
|
|
|
|
* Returns non-zero error number on error.
|
|
|
|
|
* Does not consume the packet.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_vn_parse(struct fcoe_ctlr *fip,
|
|
|
|
|
struct sk_buff *skb,
|
|
|
|
|
struct fc_rport_priv *rdata)
|
|
|
|
|
{
|
|
|
|
|
struct fip_header *fiph;
|
|
|
|
|
struct fip_desc *desc = NULL;
|
|
|
|
|
struct fip_mac_desc *macd = NULL;
|
|
|
|
|
struct fip_wwn_desc *wwn = NULL;
|
|
|
|
|
struct fip_vn_desc *vn = NULL;
|
|
|
|
|
struct fip_size_desc *size = NULL;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
size_t rlen;
|
|
|
|
|
size_t dlen;
|
|
|
|
|
u32 desc_mask = 0;
|
|
|
|
|
u32 dtype;
|
|
|
|
|
u8 sub;
|
|
|
|
|
|
|
|
|
|
memset(rdata, 0, sizeof(*rdata) + sizeof(*frport));
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
|
|
|
|
|
fiph = (struct fip_header *)skb->data;
|
|
|
|
|
frport->flags = ntohs(fiph->fip_flags);
|
|
|
|
|
|
|
|
|
|
sub = fiph->fip_subcode;
|
|
|
|
|
switch (sub) {
|
|
|
|
|
case FIP_SC_VN_PROBE_REQ:
|
|
|
|
|
case FIP_SC_VN_PROBE_REP:
|
|
|
|
|
case FIP_SC_VN_BEACON:
|
|
|
|
|
desc_mask = BIT(FIP_DT_MAC) | BIT(FIP_DT_NAME) |
|
|
|
|
|
BIT(FIP_DT_VN_ID);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_SC_VN_CLAIM_NOTIFY:
|
|
|
|
|
case FIP_SC_VN_CLAIM_REP:
|
|
|
|
|
desc_mask = BIT(FIP_DT_MAC) | BIT(FIP_DT_NAME) |
|
|
|
|
|
BIT(FIP_DT_VN_ID) | BIT(FIP_DT_FC4F) |
|
|
|
|
|
BIT(FIP_DT_FCOE_SIZE);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "vn_parse unknown subcode %u\n", sub);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
rlen = ntohs(fiph->fip_dl_len) * 4;
|
|
|
|
|
if (rlen + sizeof(*fiph) > skb->len)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
desc = (struct fip_desc *)(fiph + 1);
|
|
|
|
|
while (rlen > 0) {
|
|
|
|
|
dlen = desc->fip_dlen * FIP_BPW;
|
|
|
|
|
if (dlen < sizeof(*desc) || dlen > rlen)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
|
|
dtype = desc->fip_dtype;
|
|
|
|
|
if (dtype < 32) {
|
|
|
|
|
if (!(desc_mask & BIT(dtype))) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"unexpected or duplicated desc "
|
|
|
|
|
"desc type %u in "
|
|
|
|
|
"FIP VN2VN subtype %u\n",
|
|
|
|
|
dtype, sub);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
desc_mask &= ~BIT(dtype);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
switch (dtype) {
|
|
|
|
|
case FIP_DT_MAC:
|
|
|
|
|
if (dlen != sizeof(struct fip_mac_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
macd = (struct fip_mac_desc *)desc;
|
|
|
|
|
if (!is_valid_ether_addr(macd->fd_mac)) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"Invalid MAC addr %pM in FIP VN2VN\n",
|
|
|
|
|
macd->fd_mac);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
memcpy(frport->enode_mac, macd->fd_mac, ETH_ALEN);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_NAME:
|
|
|
|
|
if (dlen != sizeof(struct fip_wwn_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
wwn = (struct fip_wwn_desc *)desc;
|
|
|
|
|
rdata->ids.node_name = get_unaligned_be64(&wwn->fd_wwn);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_VN_ID:
|
|
|
|
|
if (dlen != sizeof(struct fip_vn_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
vn = (struct fip_vn_desc *)desc;
|
|
|
|
|
memcpy(frport->vn_mac, vn->fd_mac, ETH_ALEN);
|
|
|
|
|
rdata->ids.port_id = ntoh24(vn->fd_fc_id);
|
|
|
|
|
rdata->ids.port_name = get_unaligned_be64(&vn->fd_wwpn);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_FC4F:
|
|
|
|
|
if (dlen != sizeof(struct fip_fc4_feat))
|
|
|
|
|
goto len_err;
|
|
|
|
|
break;
|
|
|
|
|
case FIP_DT_FCOE_SIZE:
|
|
|
|
|
if (dlen != sizeof(struct fip_size_desc))
|
|
|
|
|
goto len_err;
|
|
|
|
|
size = (struct fip_size_desc *)desc;
|
|
|
|
|
frport->fcoe_len = ntohs(size->fd_size);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "unexpected descriptor type %x "
|
|
|
|
|
"in FIP probe\n", dtype);
|
|
|
|
|
/* standard says ignore unknown descriptors >= 128 */
|
|
|
|
|
if (dtype < FIP_DT_VENDOR_BASE)
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
desc = (struct fip_desc *)((char *)desc + dlen);
|
|
|
|
|
rlen -= dlen;
|
|
|
|
|
}
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
len_err:
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "FIP length error in descriptor type %x len %zu\n",
|
|
|
|
|
dtype, dlen);
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_send_claim() - send multicast FIP VN2VN Claim Notification.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_send_claim(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_CLAIM_NOTIFY, fcoe_all_vn2vn, 0);
|
|
|
|
|
fip->sol_time = jiffies;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_probe_req() - handle incoming VN2VN probe request.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @rdata: parsed remote port with frport from the probe request
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_probe_req(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_rport_priv *rdata)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_rport *frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
|
|
|
|
|
if (rdata->ids.port_id != fip->port_id)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
switch (fip->state) {
|
|
|
|
|
case FIP_ST_VNMP_CLAIM:
|
|
|
|
|
case FIP_ST_VNMP_UP:
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REP,
|
|
|
|
|
frport->enode_mac, 0);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_PROBE1:
|
|
|
|
|
case FIP_ST_VNMP_PROBE2:
|
|
|
|
|
/*
|
|
|
|
|
* Decide whether to reply to the Probe.
|
|
|
|
|
* Our selected address is never a "recorded" one, so
|
|
|
|
|
* only reply if our WWPN is greater and the
|
|
|
|
|
* Probe's REC bit is not set.
|
|
|
|
|
* If we don't reply, we will change our address.
|
|
|
|
|
*/
|
|
|
|
|
if (fip->lp->wwpn > rdata->ids.port_name &&
|
|
|
|
|
!(frport->flags & FIP_FL_REC_OR_P2P)) {
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REP,
|
|
|
|
|
frport->enode_mac, 0);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
/* fall through */
|
|
|
|
|
case FIP_ST_VNMP_START:
|
|
|
|
|
fcoe_ctlr_vn_restart(fip);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_probe_reply() - handle incoming VN2VN probe reply.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @rdata: parsed remote port with frport from the probe request
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_probe_reply(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_rport_priv *rdata)
|
|
|
|
|
{
|
|
|
|
|
if (rdata->ids.port_id != fip->port_id)
|
|
|
|
|
return;
|
|
|
|
|
switch (fip->state) {
|
|
|
|
|
case FIP_ST_VNMP_START:
|
|
|
|
|
case FIP_ST_VNMP_PROBE1:
|
|
|
|
|
case FIP_ST_VNMP_PROBE2:
|
|
|
|
|
case FIP_ST_VNMP_CLAIM:
|
|
|
|
|
fcoe_ctlr_vn_restart(fip);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_UP:
|
|
|
|
|
fcoe_ctlr_vn_send_claim(fip);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_add() - Add a VN2VN entry to the list, based on a claim reply.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @new: newly-parsed remote port with frport as a template for new rdata
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_add(struct fcoe_ctlr *fip, struct fc_rport_priv *new)
|
|
|
|
|
{
|
|
|
|
|
struct fc_lport *lport = fip->lp;
|
|
|
|
|
struct fc_rport_priv *rdata;
|
|
|
|
|
struct fc_rport_identifiers *ids;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
u32 port_id;
|
|
|
|
|
|
|
|
|
|
port_id = new->ids.port_id;
|
|
|
|
|
if (port_id == fip->port_id)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&lport->disc.disc_mutex);
|
|
|
|
|
rdata = lport->tt.rport_create(lport, port_id);
|
|
|
|
|
if (!rdata) {
|
|
|
|
|
mutex_unlock(&lport->disc.disc_mutex);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
rdata->ops = &fcoe_ctlr_vn_rport_ops;
|
|
|
|
|
rdata->disc_id = lport->disc.disc_id;
|
|
|
|
|
|
|
|
|
|
ids = &rdata->ids;
|
|
|
|
|
if ((ids->port_name != -1 && ids->port_name != new->ids.port_name) ||
|
|
|
|
|
(ids->node_name != -1 && ids->node_name != new->ids.node_name))
|
|
|
|
|
lport->tt.rport_logoff(rdata);
|
|
|
|
|
ids->port_name = new->ids.port_name;
|
|
|
|
|
ids->node_name = new->ids.node_name;
|
|
|
|
|
mutex_unlock(&lport->disc.disc_mutex);
|
|
|
|
|
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "vn_add rport %6.6x %s\n",
|
|
|
|
|
port_id, frport->fcoe_len ? "old" : "new");
|
|
|
|
|
*frport = *fcoe_ctlr_rport(new);
|
|
|
|
|
frport->time = 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_lookup() - Find VN remote port's MAC address
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @port_id: The port_id of the remote VN_node
|
|
|
|
|
* @mac: buffer which will hold the VN_NODE destination MAC address, if found.
|
|
|
|
|
*
|
|
|
|
|
* Returns non-zero error if no remote port found.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_vn_lookup(struct fcoe_ctlr *fip, u32 port_id, u8 *mac)
|
|
|
|
|
{
|
|
|
|
|
struct fc_lport *lport = fip->lp;
|
|
|
|
|
struct fc_rport_priv *rdata;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
int ret = -1;
|
|
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
|
rdata = lport->tt.rport_lookup(lport, port_id);
|
|
|
|
|
if (rdata) {
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
memcpy(mac, frport->enode_mac, ETH_ALEN);
|
|
|
|
|
ret = 0;
|
|
|
|
|
}
|
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
return ret;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_claim_notify() - handle received FIP VN2VN Claim Notification
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
* @new: newly-parsed remote port with frport as a template for new rdata
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_claim_notify(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_rport_priv *new)
|
|
|
|
|
{
|
|
|
|
|
struct fcoe_rport *frport = fcoe_ctlr_rport(new);
|
|
|
|
|
|
|
|
|
|
if (frport->flags & FIP_FL_REC_OR_P2P) {
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REQ, fcoe_all_vn2vn, 0);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
switch (fip->state) {
|
|
|
|
|
case FIP_ST_VNMP_START:
|
|
|
|
|
case FIP_ST_VNMP_PROBE1:
|
|
|
|
|
case FIP_ST_VNMP_PROBE2:
|
|
|
|
|
if (new->ids.port_id == fip->port_id)
|
|
|
|
|
fcoe_ctlr_vn_restart(fip);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_CLAIM:
|
|
|
|
|
case FIP_ST_VNMP_UP:
|
|
|
|
|
if (new->ids.port_id == fip->port_id) {
|
|
|
|
|
if (new->ids.port_name > fip->lp->wwpn) {
|
|
|
|
|
fcoe_ctlr_vn_restart(fip);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
fcoe_ctlr_vn_send_claim(fip);
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_CLAIM_REP, frport->enode_mac,
|
|
|
|
|
min((u32)frport->fcoe_len,
|
|
|
|
|
fcoe_ctlr_fcoe_size(fip)));
|
|
|
|
|
fcoe_ctlr_vn_add(fip, new);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_claim_resp() - handle received Claim Response
|
|
|
|
|
* @fip: The FCoE controller that received the frame
|
|
|
|
|
* @new: newly-parsed remote port with frport from the Claim Response
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_claim_resp(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_rport_priv *new)
|
|
|
|
|
{
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "claim resp from from rport %x - state %s\n",
|
|
|
|
|
new->ids.port_id, fcoe_ctlr_state(fip->state));
|
|
|
|
|
if (fip->state == FIP_ST_VNMP_UP || fip->state == FIP_ST_VNMP_CLAIM)
|
|
|
|
|
fcoe_ctlr_vn_add(fip, new);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_beacon() - handle received beacon.
|
|
|
|
|
* @fip: The FCoE controller that received the frame
|
|
|
|
|
* @new: newly-parsed remote port with frport from the Beacon
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_beacon(struct fcoe_ctlr *fip,
|
|
|
|
|
struct fc_rport_priv *new)
|
|
|
|
|
{
|
|
|
|
|
struct fc_lport *lport = fip->lp;
|
|
|
|
|
struct fc_rport_priv *rdata;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
|
|
|
|
|
frport = fcoe_ctlr_rport(new);
|
|
|
|
|
if (frport->flags & FIP_FL_REC_OR_P2P) {
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REQ, fcoe_all_vn2vn, 0);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
mutex_lock(&lport->disc.disc_mutex);
|
|
|
|
|
rdata = lport->tt.rport_lookup(lport, new->ids.port_id);
|
|
|
|
|
if (rdata)
|
|
|
|
|
kref_get(&rdata->kref);
|
|
|
|
|
mutex_unlock(&lport->disc.disc_mutex);
|
|
|
|
|
if (rdata) {
|
|
|
|
|
if (rdata->ids.node_name == new->ids.node_name &&
|
|
|
|
|
rdata->ids.port_name == new->ids.port_name) {
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
if (!frport->time && fip->state == FIP_ST_VNMP_UP)
|
|
|
|
|
lport->tt.rport_login(rdata);
|
|
|
|
|
frport->time = jiffies;
|
|
|
|
|
}
|
|
|
|
|
kref_put(&rdata->kref, lport->tt.rport_destroy);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
if (fip->state != FIP_ST_VNMP_UP)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* Beacon from a new neighbor.
|
|
|
|
|
* Send a claim notify if one hasn't been sent recently.
|
|
|
|
|
* Don't add the neighbor yet.
|
|
|
|
|
*/
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "beacon from new rport %x. sending claim notify\n",
|
|
|
|
|
new->ids.port_id);
|
|
|
|
|
if (time_after(jiffies,
|
|
|
|
|
fip->sol_time + msecs_to_jiffies(FIP_VN_ANN_WAIT)))
|
|
|
|
|
fcoe_ctlr_vn_send_claim(fip);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_age() - Check for VN_ports without recent beacons
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Called with ctlr_mutex held.
|
|
|
|
|
* Called only in state FIP_ST_VNMP_UP.
|
|
|
|
|
* Returns the soonest time for next age-out or a time far in the future.
|
|
|
|
|
*/
|
|
|
|
|
static unsigned long fcoe_ctlr_vn_age(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct fc_lport *lport = fip->lp;
|
|
|
|
|
struct fc_rport_priv *rdata;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
unsigned long next_time;
|
|
|
|
|
unsigned long deadline;
|
|
|
|
|
|
|
|
|
|
next_time = jiffies + msecs_to_jiffies(FIP_VN_BEACON_INT * 10);
|
|
|
|
|
mutex_lock(&lport->disc.disc_mutex);
|
|
|
|
|
list_for_each_entry_rcu(rdata, &lport->disc.rports, peers) {
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
if (!frport->time)
|
|
|
|
|
continue;
|
|
|
|
|
deadline = frport->time +
|
|
|
|
|
msecs_to_jiffies(FIP_VN_BEACON_INT * 25 / 10);
|
|
|
|
|
if (time_after_eq(jiffies, deadline)) {
|
|
|
|
|
frport->time = 0;
|
|
|
|
|
LIBFCOE_FIP_DBG(fip,
|
|
|
|
|
"port %16.16llx fc_id %6.6x beacon expired\n",
|
|
|
|
|
rdata->ids.port_name, rdata->ids.port_id);
|
|
|
|
|
lport->tt.rport_logoff(rdata);
|
|
|
|
|
} else if (time_before(deadline, next_time))
|
|
|
|
|
next_time = deadline;
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&lport->disc.disc_mutex);
|
|
|
|
|
return next_time;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_recv() - Receive a FIP frame
|
|
|
|
|
* @fip: The FCoE controller that received the frame
|
|
|
|
|
* @skb: The received FIP frame
|
|
|
|
|
*
|
|
|
|
|
* Returns non-zero if the frame is dropped.
|
|
|
|
|
* Always consumes the frame.
|
|
|
|
|
*/
|
|
|
|
|
static int fcoe_ctlr_vn_recv(struct fcoe_ctlr *fip, struct sk_buff *skb)
|
|
|
|
|
{
|
|
|
|
|
struct fip_header *fiph;
|
|
|
|
|
enum fip_vn2vn_subcode sub;
|
2010-10-09 08:12:41 +08:00
|
|
|
struct {
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
struct fc_rport_priv rdata;
|
|
|
|
|
struct fcoe_rport frport;
|
|
|
|
|
} buf;
|
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
|
|
fiph = (struct fip_header *)skb->data;
|
|
|
|
|
sub = fiph->fip_subcode;
|
|
|
|
|
|
|
|
|
|
rc = fcoe_ctlr_vn_parse(fip, skb, &buf.rdata);
|
|
|
|
|
if (rc) {
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "vn_recv vn_parse error %d\n", rc);
|
|
|
|
|
goto drop;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
switch (sub) {
|
|
|
|
|
case FIP_SC_VN_PROBE_REQ:
|
|
|
|
|
fcoe_ctlr_vn_probe_req(fip, &buf.rdata);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_SC_VN_PROBE_REP:
|
|
|
|
|
fcoe_ctlr_vn_probe_reply(fip, &buf.rdata);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_SC_VN_CLAIM_NOTIFY:
|
|
|
|
|
fcoe_ctlr_vn_claim_notify(fip, &buf.rdata);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_SC_VN_CLAIM_REP:
|
|
|
|
|
fcoe_ctlr_vn_claim_resp(fip, &buf.rdata);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_SC_VN_BEACON:
|
|
|
|
|
fcoe_ctlr_vn_beacon(fip, &buf.rdata);
|
|
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
LIBFCOE_FIP_DBG(fip, "vn_recv unknown subcode %d\n", sub);
|
|
|
|
|
rc = -1;
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
drop:
|
|
|
|
|
kfree_skb(skb);
|
|
|
|
|
return rc;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_disc_recv - discovery receive handler for VN2VN mode.
|
2010-07-21 06:21:12 +08:00
|
|
|
* @lport: The local port
|
|
|
|
|
* @fp: The received frame
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
*
|
|
|
|
|
* This should never be called since we don't see RSCNs or other
|
|
|
|
|
* fabric-generated ELSes.
|
|
|
|
|
*/
|
2010-07-21 06:21:12 +08:00
|
|
|
static void fcoe_ctlr_disc_recv(struct fc_lport *lport, struct fc_frame *fp)
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
{
|
|
|
|
|
struct fc_seq_els_data rjt_data;
|
|
|
|
|
|
|
|
|
|
rjt_data.reason = ELS_RJT_UNSUP;
|
|
|
|
|
rjt_data.explan = ELS_EXPL_NONE;
|
2010-07-21 06:21:12 +08:00
|
|
|
lport->tt.seq_els_rsp_send(fp, ELS_LS_RJT, &rjt_data);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
fc_frame_free(fp);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_disc_recv - start discovery for VN2VN mode.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* This sets a flag indicating that remote ports should be created
|
|
|
|
|
* and started for the peers we discover. We use the disc_callback
|
|
|
|
|
* pointer as that flag. Peers already discovered are created here.
|
|
|
|
|
*
|
|
|
|
|
* The lport lock is held during this call. The callback must be done
|
|
|
|
|
* later, without holding either the lport or discovery locks.
|
|
|
|
|
* The fcoe_ctlr lock may also be held during this call.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_disc_start(void (*callback)(struct fc_lport *,
|
|
|
|
|
enum fc_disc_event),
|
|
|
|
|
struct fc_lport *lport)
|
|
|
|
|
{
|
|
|
|
|
struct fc_disc *disc = &lport->disc;
|
|
|
|
|
struct fcoe_ctlr *fip = disc->priv;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&disc->disc_mutex);
|
|
|
|
|
disc->disc_callback = callback;
|
|
|
|
|
disc->disc_id = (disc->disc_id + 2) | 1;
|
|
|
|
|
disc->pending = 1;
|
|
|
|
|
schedule_work(&fip->timer_work);
|
|
|
|
|
mutex_unlock(&disc->disc_mutex);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_disc() - report FIP VN_port discovery results after claim state.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*
|
|
|
|
|
* Starts the FLOGI and PLOGI login process to each discovered rport for which
|
|
|
|
|
* we've received at least one beacon.
|
|
|
|
|
* Performs the discovery complete callback.
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_disc(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
struct fc_lport *lport = fip->lp;
|
|
|
|
|
struct fc_disc *disc = &lport->disc;
|
|
|
|
|
struct fc_rport_priv *rdata;
|
|
|
|
|
struct fcoe_rport *frport;
|
|
|
|
|
void (*callback)(struct fc_lport *, enum fc_disc_event);
|
|
|
|
|
|
|
|
|
|
mutex_lock(&disc->disc_mutex);
|
|
|
|
|
callback = disc->pending ? disc->disc_callback : NULL;
|
|
|
|
|
disc->pending = 0;
|
|
|
|
|
list_for_each_entry_rcu(rdata, &disc->rports, peers) {
|
|
|
|
|
frport = fcoe_ctlr_rport(rdata);
|
|
|
|
|
if (frport->time)
|
|
|
|
|
lport->tt.rport_login(rdata);
|
|
|
|
|
}
|
|
|
|
|
mutex_unlock(&disc->disc_mutex);
|
|
|
|
|
if (callback)
|
|
|
|
|
callback(lport, DISC_EV_SUCCESS);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
|
* fcoe_ctlr_vn_timeout - timer work function for VN2VN mode.
|
|
|
|
|
* @fip: The FCoE controller
|
|
|
|
|
*/
|
|
|
|
|
static void fcoe_ctlr_vn_timeout(struct fcoe_ctlr *fip)
|
|
|
|
|
{
|
|
|
|
|
unsigned long next_time;
|
|
|
|
|
u8 mac[ETH_ALEN];
|
|
|
|
|
u32 new_port_id = 0;
|
|
|
|
|
|
|
|
|
|
mutex_lock(&fip->ctlr_mutex);
|
|
|
|
|
switch (fip->state) {
|
|
|
|
|
case FIP_ST_VNMP_START:
|
|
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_VNMP_PROBE1);
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REQ, fcoe_all_vn2vn, 0);
|
|
|
|
|
next_time = jiffies + msecs_to_jiffies(FIP_VN_PROBE_WAIT);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_PROBE1:
|
|
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_VNMP_PROBE2);
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_PROBE_REQ, fcoe_all_vn2vn, 0);
|
|
|
|
|
next_time = jiffies + msecs_to_jiffies(FIP_VN_ANN_WAIT);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_PROBE2:
|
|
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_VNMP_CLAIM);
|
|
|
|
|
new_port_id = fip->port_id;
|
|
|
|
|
hton24(mac, FIP_VN_FC_MAP);
|
|
|
|
|
hton24(mac + 3, new_port_id);
|
2010-07-21 06:20:40 +08:00
|
|
|
fcoe_ctlr_map_dest(fip);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
fip->update_mac(fip->lp, mac);
|
|
|
|
|
fcoe_ctlr_vn_send_claim(fip);
|
|
|
|
|
next_time = jiffies + msecs_to_jiffies(FIP_VN_ANN_WAIT);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_CLAIM:
|
|
|
|
|
/*
|
|
|
|
|
* This may be invoked either by starting discovery so don't
|
|
|
|
|
* go to the next state unless it's been long enough.
|
|
|
|
|
*/
|
|
|
|
|
next_time = fip->sol_time + msecs_to_jiffies(FIP_VN_ANN_WAIT);
|
|
|
|
|
if (time_after_eq(jiffies, next_time)) {
|
|
|
|
|
fcoe_ctlr_set_state(fip, FIP_ST_VNMP_UP);
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_BEACON,
|
|
|
|
|
fcoe_all_vn2vn, 0);
|
|
|
|
|
next_time = jiffies + msecs_to_jiffies(FIP_VN_ANN_WAIT);
|
|
|
|
|
fip->port_ka_time = next_time;
|
|
|
|
|
}
|
|
|
|
|
fcoe_ctlr_vn_disc(fip);
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_VNMP_UP:
|
|
|
|
|
next_time = fcoe_ctlr_vn_age(fip);
|
|
|
|
|
if (time_after_eq(jiffies, fip->port_ka_time)) {
|
|
|
|
|
fcoe_ctlr_vn_send(fip, FIP_SC_VN_BEACON,
|
|
|
|
|
fcoe_all_vn2vn, 0);
|
|
|
|
|
fip->port_ka_time = jiffies +
|
|
|
|
|
msecs_to_jiffies(FIP_VN_BEACON_INT +
|
|
|
|
|
(random32() % FIP_VN_BEACON_FUZZ));
|
|
|
|
|
}
|
|
|
|
|
if (time_before(fip->port_ka_time, next_time))
|
|
|
|
|
next_time = fip->port_ka_time;
|
|
|
|
|
break;
|
|
|
|
|
case FIP_ST_LINK_WAIT:
|
|
|
|
|
goto unlock;
|
|
|
|
|
default:
|
2010-12-01 08:19:09 +08:00
|
|
|
WARN(1, "unexpected state %d\n", fip->state);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
goto unlock;
|
|
|
|
|
}
|
|
|
|
|
mod_timer(&fip->timer, next_time);
|
|
|
|
|
unlock:
|
|
|
|
|
mutex_unlock(&fip->ctlr_mutex);
|
|
|
|
|
|
|
|
|
|
/* If port ID is new, notify local port after dropping ctlr_mutex */
|
|
|
|
|
if (new_port_id)
|
|
|
|
|
fc_lport_set_local_id(fip->lp, new_port_id);
|
|
|
|
|
}
|
|
|
|
|
|
2009-03-18 02:42:18 +08:00
|
|
|
/**
|
2009-11-04 03:47:45 +08:00
|
|
|
* fcoe_libfc_config() - Sets up libfc related properties for local port
|
|
|
|
|
* @lp: The local port to configure libfc for
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
* @fip: The FCoE controller in use by the local port
|
2009-11-04 03:47:45 +08:00
|
|
|
* @tt: The libfc function template
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
* @init_fcp: If non-zero, the FCP portion of libfc should be initialized
|
2009-03-18 02:42:18 +08:00
|
|
|
*
|
|
|
|
|
* Returns : 0 for success
|
|
|
|
|
*/
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
int fcoe_libfc_config(struct fc_lport *lport, struct fcoe_ctlr *fip,
|
|
|
|
|
const struct libfc_function_template *tt, int init_fcp)
|
2009-03-18 02:42:18 +08:00
|
|
|
{
|
|
|
|
|
/* Set the function pointers set by the LLDD */
|
2009-11-04 03:47:45 +08:00
|
|
|
memcpy(&lport->tt, tt, sizeof(*tt));
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (init_fcp && fc_fcp_init(lport))
|
2009-03-18 02:42:18 +08:00
|
|
|
return -ENOMEM;
|
2009-11-04 03:47:45 +08:00
|
|
|
fc_exch_init(lport);
|
|
|
|
|
fc_elsct_init(lport);
|
|
|
|
|
fc_lport_init(lport);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN)
|
|
|
|
|
lport->rport_priv_size = sizeof(struct fcoe_rport);
|
2009-11-04 03:47:45 +08:00
|
|
|
fc_rport_init(lport);
|
[SCSI] libfcoe: fcoe: fnic: add FIP VN2VN point-to-multipoint support
The FC-BB-6 committee is proposing a new FIP usage model called
VN_port to VN_port mode. It allows VN_ports to discover each other
over a loss-free L2 Ethernet without any FCF or Fibre-channel fabric
services. This is point-to-multipoint. There is also a variant
of this called point-to-point which provides for making sure there
is just one pair of ports operating over the Ethernet fabric.
We add these new states: VNMP_START, _PROBE1, _PROBE2, _CLAIM, and _UP.
These usually go quickly in that sequence. After waiting a random
amount of time up to 100 ms in START, we select a pseudo-random
proposed locally-unique port ID and send out probes in states PROBE1
and PROBE2, 100 ms apart. If no probe responses are heard, we
proceed to CLAIM state 400 ms later and send a claim notification.
We wait another 400 ms to receive claim responses, which give us
a list of the other nodes on the network, including their FC-4
capabilities. After another 400 ms we go to VNMP_UP state and
should start interoperating with any of the nodes for whic we
receivec claim responses. More details are in the spec.j
Add the new mode as FIP_MODE_VN2VN. The driver must specify
explicitly that it wants to operate in this mode. There is
no automatic detection between point-to-multipoint and fabric
mode, and the local port initialization is affected, so it isn't
anticipated that there will ever be any such automatic switchover.
It may eventually be possible to have both fabric and VN2VN
modes on the same L2 network, which may be done by two separate
local VN_ports (lports).
When in VN2VN mode, FIP replaces libfc's fabric-oriented discovery
module with its own simple code that adds remote ports as they
are discovered from incoming claim notifications and responses.
These hooks are placed by fcoe_disc_init().
A linear list of discovered vn_ports is maintained under the
fcoe_ctlr struct. It is expected to be short for now, and
accessed infrequently. It is kept under RCU for lock-ordering
reasons. The lport and/or rport mutexes may be held when we
need to lookup a fcoe_vnport during an ELS send.
Change fcoe_ctlr_encaps() to lookup the destination vn_port in
the list of peers for the destination MAC address of the
FIP-encapsulated frame.
Add a new function fcoe_disc_init() to initialize just the
discovery portion of libfcoe for VN2VN mode.
Signed-off-by: Joe Eykholt <jeykholt@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
2010-07-21 06:20:30 +08:00
|
|
|
if (fip->mode == FIP_MODE_VN2VN) {
|
|
|
|
|
lport->point_to_multipoint = 1;
|
|
|
|
|
lport->tt.disc_recv_req = fcoe_ctlr_disc_recv;
|
|
|
|
|
lport->tt.disc_start = fcoe_ctlr_disc_start;
|
|
|
|
|
lport->tt.disc_stop = fcoe_ctlr_disc_stop;
|
|
|
|
|
lport->tt.disc_stop_final = fcoe_ctlr_disc_stop_final;
|
|
|
|
|
mutex_init(&lport->disc.disc_mutex);
|
|
|
|
|
INIT_LIST_HEAD(&lport->disc.rports);
|
|
|
|
|
lport->disc.priv = fip;
|
|
|
|
|
} else {
|
|
|
|
|
fc_disc_init(lport);
|
|
|
|
|
}
|
2009-03-18 02:42:18 +08:00
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
EXPORT_SYMBOL_GPL(fcoe_libfc_config);
|
