1
0
Fork 0
alistair23-linux/drivers/scsi/libfc/fc_disc.c

860 lines
22 KiB
C
Raw Normal View History

/*
* Copyright(c) 2007 - 2008 Intel Corporation. All rights reserved.
*
* 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
*/
/*
* Target Discovery
*
* This block discovers all FC-4 remote ports, including FCP initiators. It
* also handles RSCN events and re-discovery if necessary.
*/
/*
* DISC LOCKING
*
* The disc mutex is can be locked when acquiring rport locks, but may not
* be held when acquiring the lport lock. Refer to fc_lport.c for more
* details.
*/
#include <linux/timer.h>
#include <linux/err.h>
#include <asm/unaligned.h>
#include <scsi/fc/fc_gs.h>
#include <scsi/libfc.h>
#define FC_DISC_RETRY_LIMIT 3 /* max retries */
#define FC_DISC_RETRY_DELAY 500UL /* (msecs) delay */
#define FC_DISC_DELAY 3
static void fc_disc_gpn_ft_req(struct fc_disc *);
static void fc_disc_gpn_ft_resp(struct fc_seq *, struct fc_frame *, void *);
static int fc_disc_new_target(struct fc_disc *, struct fc_rport *,
struct fc_rport_identifiers *);
static void fc_disc_del_target(struct fc_disc *, struct fc_rport *);
static void fc_disc_done(struct fc_disc *);
static void fc_disc_timeout(struct work_struct *);
static void fc_disc_single(struct fc_disc *, struct fc_disc_port *);
static void fc_disc_restart(struct fc_disc *);
/**
* fc_disc_lookup_rport() - lookup a remote port by port_id
* @lport: Fibre Channel host port instance
* @port_id: remote port port_id to match
*/
struct fc_rport *fc_disc_lookup_rport(const struct fc_lport *lport,
u32 port_id)
{
const struct fc_disc *disc = &lport->disc;
struct fc_rport *rport, *found = NULL;
struct fc_rport_libfc_priv *rdata;
int disc_found = 0;
list_for_each_entry(rdata, &disc->rports, peers) {
rport = PRIV_TO_RPORT(rdata);
if (rport->port_id == port_id) {
disc_found = 1;
found = rport;
break;
}
}
if (!disc_found)
found = NULL;
return found;
}
/**
* fc_disc_stop_rports() - delete all the remote ports associated with the lport
* @disc: The discovery job to stop rports on
*
* Locking Note: This function expects that the lport mutex is locked before
* calling it.
*/
void fc_disc_stop_rports(struct fc_disc *disc)
{
struct fc_lport *lport;
struct fc_rport *rport;
struct fc_rport_libfc_priv *rdata, *next;
lport = disc->lport;
mutex_lock(&disc->disc_mutex);
list_for_each_entry_safe(rdata, next, &disc->rports, peers) {
rport = PRIV_TO_RPORT(rdata);
list_del(&rdata->peers);
lport->tt.rport_logoff(rport);
}
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
list_for_each_entry_safe(rdata, next, &disc->rogue_rports, peers) {
rport = PRIV_TO_RPORT(rdata);
lport->tt.rport_logoff(rport);
}
mutex_unlock(&disc->disc_mutex);
}
/**
* fc_disc_rport_callback() - Event handler for rport events
* @lport: The lport which is receiving the event
* @rport: The rport which the event has occured on
* @event: The event that occured
*
* Locking Note: The rport lock should not be held when calling
* this function.
*/
static void fc_disc_rport_callback(struct fc_lport *lport,
struct fc_rport *rport,
enum fc_rport_event event)
{
struct fc_rport_libfc_priv *rdata = rport->dd_data;
struct fc_disc *disc = &lport->disc;
FC_DISC_DBG(disc, "Received a %d event for port (%6x)\n", event,
rport->port_id);
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
switch (event) {
case RPORT_EV_CREATED:
if (disc) {
mutex_lock(&disc->disc_mutex);
list_add_tail(&rdata->peers, &disc->rports);
mutex_unlock(&disc->disc_mutex);
}
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
break;
case RPORT_EV_LOGO:
case RPORT_EV_FAILED:
case RPORT_EV_STOP:
mutex_lock(&disc->disc_mutex);
mutex_lock(&rdata->rp_mutex);
if (rdata->trans_state == FC_PORTSTATE_ROGUE)
list_del(&rdata->peers);
mutex_unlock(&rdata->rp_mutex);
mutex_unlock(&disc->disc_mutex);
break;
default:
break;
}
}
/**
* fc_disc_recv_rscn_req() - Handle Registered State Change Notification (RSCN)
* @sp: Current sequence of the RSCN exchange
* @fp: RSCN Frame
* @lport: Fibre Channel host port instance
*
* Locking Note: This function expects that the disc_mutex is locked
* before it is called.
*/
static void fc_disc_recv_rscn_req(struct fc_seq *sp, struct fc_frame *fp,
struct fc_disc *disc)
{
struct fc_lport *lport;
struct fc_rport *rport;
struct fc_rport_libfc_priv *rdata;
struct fc_els_rscn *rp;
struct fc_els_rscn_page *pp;
struct fc_seq_els_data rjt_data;
unsigned int len;
int redisc = 0;
enum fc_els_rscn_ev_qual ev_qual;
enum fc_els_rscn_addr_fmt fmt;
LIST_HEAD(disc_ports);
struct fc_disc_port *dp, *next;
lport = disc->lport;
FC_DISC_DBG(disc, "Received an RSCN event\n");
/* make sure the frame contains an RSCN message */
rp = fc_frame_payload_get(fp, sizeof(*rp));
if (!rp)
goto reject;
/* make sure the page length is as expected (4 bytes) */
if (rp->rscn_page_len != sizeof(*pp))
goto reject;
/* get the RSCN payload length */
len = ntohs(rp->rscn_plen);
if (len < sizeof(*rp))
goto reject;
/* make sure the frame contains the expected payload */
rp = fc_frame_payload_get(fp, len);
if (!rp)
goto reject;
/* payload must be a multiple of the RSCN page size */
len -= sizeof(*rp);
if (len % sizeof(*pp))
goto reject;
for (pp = (void *)(rp + 1); len > 0; len -= sizeof(*pp), pp++) {
ev_qual = pp->rscn_page_flags >> ELS_RSCN_EV_QUAL_BIT;
ev_qual &= ELS_RSCN_EV_QUAL_MASK;
fmt = pp->rscn_page_flags >> ELS_RSCN_ADDR_FMT_BIT;
fmt &= ELS_RSCN_ADDR_FMT_MASK;
/*
* if we get an address format other than port
* (area, domain, fabric), then do a full discovery
*/
switch (fmt) {
case ELS_ADDR_FMT_PORT:
FC_DISC_DBG(disc, "Port address format for port "
"(%6x)\n", ntoh24(pp->rscn_fid));
dp = kzalloc(sizeof(*dp), GFP_KERNEL);
if (!dp) {
redisc = 1;
break;
}
dp->lp = lport;
dp->ids.port_id = ntoh24(pp->rscn_fid);
dp->ids.port_name = -1;
dp->ids.node_name = -1;
dp->ids.roles = FC_RPORT_ROLE_UNKNOWN;
list_add_tail(&dp->peers, &disc_ports);
break;
case ELS_ADDR_FMT_AREA:
case ELS_ADDR_FMT_DOM:
case ELS_ADDR_FMT_FAB:
default:
FC_DISC_DBG(disc, "Address format is (%d)\n", fmt);
redisc = 1;
break;
}
}
lport->tt.seq_els_rsp_send(sp, ELS_LS_ACC, NULL);
if (redisc) {
FC_DISC_DBG(disc, "RSCN received: rediscovering\n");
fc_disc_restart(disc);
} else {
FC_DISC_DBG(disc, "RSCN received: not rediscovering. "
"redisc %d state %d in_prog %d\n",
redisc, lport->state, disc->pending);
list_for_each_entry_safe(dp, next, &disc_ports, peers) {
list_del(&dp->peers);
rport = lport->tt.rport_lookup(lport, dp->ids.port_id);
if (rport) {
rdata = rport->dd_data;
list_del(&rdata->peers);
lport->tt.rport_logoff(rport);
}
fc_disc_single(disc, dp);
}
}
fc_frame_free(fp);
return;
reject:
FC_DISC_DBG(disc, "Received a bad RSCN frame\n");
rjt_data.fp = NULL;
rjt_data.reason = ELS_RJT_LOGIC;
rjt_data.explan = ELS_EXPL_NONE;
lport->tt.seq_els_rsp_send(sp, ELS_LS_RJT, &rjt_data);
fc_frame_free(fp);
}
/**
* fc_disc_recv_req() - Handle incoming requests
* @sp: Current sequence of the request exchange
* @fp: The frame
* @lport: The FC local port
*
* Locking Note: This function is called from the EM and will lock
* the disc_mutex before calling the handler for the
* request.
*/
static void fc_disc_recv_req(struct fc_seq *sp, struct fc_frame *fp,
struct fc_lport *lport)
{
u8 op;
struct fc_disc *disc = &lport->disc;
op = fc_frame_payload_op(fp);
switch (op) {
case ELS_RSCN:
mutex_lock(&disc->disc_mutex);
fc_disc_recv_rscn_req(sp, fp, disc);
mutex_unlock(&disc->disc_mutex);
break;
default:
FC_DISC_DBG(disc, "Received an unsupported request, "
"the opcode is (%x)\n", op);
break;
}
}
/**
* fc_disc_restart() - Restart discovery
* @lport: FC discovery context
*
* Locking Note: This function expects that the disc mutex
* is already locked.
*/
static void fc_disc_restart(struct fc_disc *disc)
{
struct fc_rport *rport;
struct fc_rport_libfc_priv *rdata, *next;
struct fc_lport *lport = disc->lport;
FC_DISC_DBG(disc, "Restarting discovery\n");
list_for_each_entry_safe(rdata, next, &disc->rports, peers) {
rport = PRIV_TO_RPORT(rdata);
list_del(&rdata->peers);
lport->tt.rport_logoff(rport);
}
disc->requested = 1;
if (!disc->pending)
fc_disc_gpn_ft_req(disc);
}
/**
* fc_disc_start() - Fibre Channel Target discovery
* @lport: FC local port
*
* Returns non-zero if discovery cannot be started.
*/
static void fc_disc_start(void (*disc_callback)(struct fc_lport *,
enum fc_disc_event),
struct fc_lport *lport)
{
struct fc_rport *rport;
struct fc_rport_identifiers ids;
struct fc_disc *disc = &lport->disc;
/*
* At this point we may have a new disc job or an existing
* one. Either way, let's lock when we make changes to it
* and send the GPN_FT request.
*/
mutex_lock(&disc->disc_mutex);
disc->disc_callback = disc_callback;
/*
* If not ready, or already running discovery, just set request flag.
*/
disc->requested = 1;
if (disc->pending) {
mutex_unlock(&disc->disc_mutex);
return;
}
/*
* Handle point-to-point mode as a simple discovery
* of the remote port. Yucky, yucky, yuck, yuck!
*/
rport = disc->lport->ptp_rp;
if (rport) {
ids.port_id = rport->port_id;
ids.port_name = rport->port_name;
ids.node_name = rport->node_name;
ids.roles = FC_RPORT_ROLE_UNKNOWN;
get_device(&rport->dev);
if (!fc_disc_new_target(disc, rport, &ids)) {
disc->event = DISC_EV_SUCCESS;
fc_disc_done(disc);
}
put_device(&rport->dev);
} else {
fc_disc_gpn_ft_req(disc); /* get ports by FC-4 type */
}
mutex_unlock(&disc->disc_mutex);
}
static struct fc_rport_operations fc_disc_rport_ops = {
.event_callback = fc_disc_rport_callback,
};
/**
* fc_disc_new_target() - Handle new target found by discovery
* @lport: FC local port
* @rport: The previous FC remote port (NULL if new remote port)
* @ids: Identifiers for the new FC remote port
*
* Locking Note: This function expects that the disc_mutex is locked
* before it is called.
*/
static int fc_disc_new_target(struct fc_disc *disc,
struct fc_rport *rport,
struct fc_rport_identifiers *ids)
{
struct fc_lport *lport = disc->lport;
struct fc_rport_libfc_priv *rdata;
int error = 0;
if (rport && ids->port_name) {
if (rport->port_name == -1) {
/*
* Set WWN and fall through to notify of create.
*/
fc_rport_set_name(rport, ids->port_name,
rport->node_name);
} else if (rport->port_name != ids->port_name) {
/*
* This is a new port with the same FCID as
* a previously-discovered port. Presumably the old
* port logged out and a new port logged in and was
* assigned the same FCID. This should be rare.
* Delete the old one and fall thru to re-create.
*/
fc_disc_del_target(disc, rport);
rport = NULL;
}
}
if (((ids->port_name != -1) || (ids->port_id != -1)) &&
ids->port_id != fc_host_port_id(lport->host) &&
ids->port_name != lport->wwpn) {
if (!rport) {
rport = lport->tt.rport_lookup(lport, ids->port_id);
if (!rport) {
struct fc_disc_port dp;
dp.lp = lport;
dp.ids.port_id = ids->port_id;
dp.ids.port_name = ids->port_name;
dp.ids.node_name = ids->node_name;
dp.ids.roles = ids->roles;
rport = lport->tt.rport_create(&dp);
}
if (!rport)
error = -ENOMEM;
}
if (rport) {
rdata = rport->dd_data;
rdata->ops = &fc_disc_rport_ops;
rdata->rp_state = RPORT_ST_INIT;
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
list_add_tail(&rdata->peers, &disc->rogue_rports);
lport->tt.rport_login(rport);
}
}
return error;
}
/**
* fc_disc_del_target() - Delete a target
* @disc: FC discovery context
* @rport: The remote port to be removed
*/
static void fc_disc_del_target(struct fc_disc *disc, struct fc_rport *rport)
{
struct fc_lport *lport = disc->lport;
struct fc_rport_libfc_priv *rdata = rport->dd_data;
list_del(&rdata->peers);
lport->tt.rport_logoff(rport);
}
/**
* fc_disc_done() - Discovery has been completed
* @disc: FC discovery context
* Locking Note: This function expects that the disc mutex is locked before
* it is called. The discovery callback is then made with the lock released,
* and the lock is re-taken before returning from this function
*/
static void fc_disc_done(struct fc_disc *disc)
{
struct fc_lport *lport = disc->lport;
enum fc_disc_event event;
FC_DISC_DBG(disc, "Discovery complete\n");
event = disc->event;
disc->event = DISC_EV_NONE;
if (disc->requested)
fc_disc_gpn_ft_req(disc);
else
disc->pending = 0;
mutex_unlock(&disc->disc_mutex);
disc->disc_callback(lport, event);
mutex_lock(&disc->disc_mutex);
}
/**
* fc_disc_error() - Handle error on dNS request
* @disc: FC discovery context
* @fp: The frame pointer
*/
static void fc_disc_error(struct fc_disc *disc, struct fc_frame *fp)
{
struct fc_lport *lport = disc->lport;
unsigned long delay = 0;
FC_DISC_DBG(disc, "Error %ld, retries %d/%d\n",
PTR_ERR(fp), disc->retry_count,
FC_DISC_RETRY_LIMIT);
if (!fp || PTR_ERR(fp) == -FC_EX_TIMEOUT) {
/*
* Memory allocation failure, or the exchange timed out,
* retry after delay.
*/
if (disc->retry_count < FC_DISC_RETRY_LIMIT) {
/* go ahead and retry */
if (!fp)
delay = msecs_to_jiffies(FC_DISC_RETRY_DELAY);
else {
delay = msecs_to_jiffies(lport->e_d_tov);
/* timeout faster first time */
if (!disc->retry_count)
delay /= 4;
}
disc->retry_count++;
schedule_delayed_work(&disc->disc_work, delay);
} else {
/* exceeded retries */
disc->event = DISC_EV_FAILED;
fc_disc_done(disc);
}
}
}
/**
* fc_disc_gpn_ft_req() - Send Get Port Names by FC-4 type (GPN_FT) request
* @lport: FC discovery context
*
* Locking Note: This function expects that the disc_mutex is locked
* before it is called.
*/
static void fc_disc_gpn_ft_req(struct fc_disc *disc)
{
struct fc_frame *fp;
struct fc_lport *lport = disc->lport;
WARN_ON(!fc_lport_test_ready(lport));
disc->pending = 1;
disc->requested = 0;
disc->buf_len = 0;
disc->seq_count = 0;
fp = fc_frame_alloc(lport,
sizeof(struct fc_ct_hdr) +
sizeof(struct fc_ns_gid_ft));
if (!fp)
goto err;
if (lport->tt.elsct_send(lport, NULL, fp,
FC_NS_GPN_FT,
fc_disc_gpn_ft_resp,
disc, lport->e_d_tov))
return;
err:
fc_disc_error(disc, fp);
}
/**
* fc_disc_gpn_ft_parse() - Parse the list of IDs and names resulting from a request
* @lport: Fibre Channel host port instance
* @buf: GPN_FT response buffer
* @len: size of response buffer
*/
static int fc_disc_gpn_ft_parse(struct fc_disc *disc, void *buf, size_t len)
{
struct fc_lport *lport;
struct fc_gpn_ft_resp *np;
char *bp;
size_t plen;
size_t tlen;
int error = 0;
struct fc_disc_port dp;
struct fc_rport *rport;
struct fc_rport_libfc_priv *rdata;
lport = disc->lport;
/*
* Handle partial name record left over from previous call.
*/
bp = buf;
plen = len;
np = (struct fc_gpn_ft_resp *)bp;
tlen = disc->buf_len;
if (tlen) {
WARN_ON(tlen >= sizeof(*np));
plen = sizeof(*np) - tlen;
WARN_ON(plen <= 0);
WARN_ON(plen >= sizeof(*np));
if (plen > len)
plen = len;
np = &disc->partial_buf;
memcpy((char *)np + tlen, bp, plen);
/*
* Set bp so that the loop below will advance it to the
* first valid full name element.
*/
bp -= tlen;
len += tlen;
plen += tlen;
disc->buf_len = (unsigned char) plen;
if (plen == sizeof(*np))
disc->buf_len = 0;
}
/*
* Handle full name records, including the one filled from above.
* Normally, np == bp and plen == len, but from the partial case above,
* bp, len describe the overall buffer, and np, plen describe the
* partial buffer, which if would usually be full now.
* After the first time through the loop, things return to "normal".
*/
while (plen >= sizeof(*np)) {
dp.lp = lport;
dp.ids.port_id = ntoh24(np->fp_fid);
dp.ids.port_name = ntohll(np->fp_wwpn);
dp.ids.node_name = -1;
dp.ids.roles = FC_RPORT_ROLE_UNKNOWN;
if ((dp.ids.port_id != fc_host_port_id(lport->host)) &&
(dp.ids.port_name != lport->wwpn)) {
rport = lport->tt.rport_create(&dp);
if (rport) {
rdata = rport->dd_data;
rdata->ops = &fc_disc_rport_ops;
rdata->local_port = lport;
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
list_add_tail(&rdata->peers,
&disc->rogue_rports);
lport->tt.rport_login(rport);
} else
printk(KERN_WARNING "libfc: Failed to allocate "
"memory for the newly discovered port "
"(%6x)\n", dp.ids.port_id);
}
if (np->fp_flags & FC_NS_FID_LAST) {
disc->event = DISC_EV_SUCCESS;
fc_disc_done(disc);
len = 0;
break;
}
len -= sizeof(*np);
bp += sizeof(*np);
np = (struct fc_gpn_ft_resp *)bp;
plen = len;
}
/*
* Save any partial record at the end of the buffer for next time.
*/
if (error == 0 && len > 0 && len < sizeof(*np)) {
if (np != &disc->partial_buf) {
FC_DISC_DBG(disc, "Partial buffer remains "
"for discovery\n");
memcpy(&disc->partial_buf, np, len);
}
disc->buf_len = (unsigned char) len;
} else {
disc->buf_len = 0;
}
return error;
}
/**
* fc_disc_timeout() - Retry handler for the disc component
* @work: Structure holding disc obj that needs retry discovery
*
* Handle retry of memory allocation for remote ports.
*/
static void fc_disc_timeout(struct work_struct *work)
{
struct fc_disc *disc = container_of(work,
struct fc_disc,
disc_work.work);
mutex_lock(&disc->disc_mutex);
if (disc->requested && !disc->pending)
fc_disc_gpn_ft_req(disc);
mutex_unlock(&disc->disc_mutex);
}
/**
* fc_disc_gpn_ft_resp() - Handle a response frame from Get Port Names (GPN_FT)
* @sp: Current sequence of GPN_FT exchange
* @fp: response frame
* @lp_arg: Fibre Channel host port instance
*
* Locking Note: This function is called without disc mutex held, and
* should do all its processing with the mutex held
*/
static void fc_disc_gpn_ft_resp(struct fc_seq *sp, struct fc_frame *fp,
void *disc_arg)
{
struct fc_disc *disc = disc_arg;
struct fc_ct_hdr *cp;
struct fc_frame_header *fh;
unsigned int seq_cnt;
void *buf = NULL;
unsigned int len;
int error;
mutex_lock(&disc->disc_mutex);
FC_DISC_DBG(disc, "Received a GPN_FT response\n");
if (IS_ERR(fp)) {
fc_disc_error(disc, fp);
mutex_unlock(&disc->disc_mutex);
return;
}
WARN_ON(!fc_frame_is_linear(fp)); /* buffer must be contiguous */
fh = fc_frame_header_get(fp);
len = fr_len(fp) - sizeof(*fh);
seq_cnt = ntohs(fh->fh_seq_cnt);
if (fr_sof(fp) == FC_SOF_I3 && seq_cnt == 0 &&
disc->seq_count == 0) {
cp = fc_frame_payload_get(fp, sizeof(*cp));
if (!cp) {
FC_DISC_DBG(disc, "GPN_FT response too short, len %d\n",
fr_len(fp));
} else if (ntohs(cp->ct_cmd) == FC_FS_ACC) {
/* Accepted, parse the response. */
buf = cp + 1;
len -= sizeof(*cp);
} else if (ntohs(cp->ct_cmd) == FC_FS_RJT) {
FC_DISC_DBG(disc, "GPN_FT rejected reason %x exp %x "
"(check zoning)\n", cp->ct_reason,
cp->ct_explan);
disc->event = DISC_EV_FAILED;
fc_disc_done(disc);
} else {
FC_DISC_DBG(disc, "GPN_FT unexpected response code "
"%x\n", ntohs(cp->ct_cmd));
}
} else if (fr_sof(fp) == FC_SOF_N3 &&
seq_cnt == disc->seq_count) {
buf = fh + 1;
} else {
FC_DISC_DBG(disc, "GPN_FT unexpected frame - out of sequence? "
"seq_cnt %x expected %x sof %x eof %x\n",
seq_cnt, disc->seq_count, fr_sof(fp), fr_eof(fp));
}
if (buf) {
error = fc_disc_gpn_ft_parse(disc, buf, len);
if (error)
fc_disc_error(disc, fp);
else
disc->seq_count++;
}
fc_frame_free(fp);
mutex_unlock(&disc->disc_mutex);
}
/**
* fc_disc_single() - Discover the directory information for a single target
* @lport: FC local port
* @dp: The port to rediscover
*
* Locking Note: This function expects that the disc_mutex is locked
* before it is called.
*/
static void fc_disc_single(struct fc_disc *disc, struct fc_disc_port *dp)
{
struct fc_lport *lport;
struct fc_rport *new_rport;
struct fc_rport_libfc_priv *rdata;
lport = disc->lport;
if (dp->ids.port_id == fc_host_port_id(lport->host))
goto out;
new_rport = lport->tt.rport_create(dp);
if (new_rport) {
rdata = new_rport->dd_data;
rdata->ops = &fc_disc_rport_ops;
kfree(dp);
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
list_add_tail(&rdata->peers, &disc->rogue_rports);
lport->tt.rport_login(new_rport);
}
return;
out:
kfree(dp);
}
/**
* fc_disc_stop() - Stop discovery for a given lport
* @lport: The lport that discovery should stop for
*/
void fc_disc_stop(struct fc_lport *lport)
{
struct fc_disc *disc = &lport->disc;
if (disc) {
cancel_delayed_work_sync(&disc->disc_work);
fc_disc_stop_rports(disc);
}
}
/**
* fc_disc_stop_final() - Stop discovery for a given lport
* @lport: The lport that discovery should stop for
*
* This function will block until discovery has been
* completely stopped and all rports have been deleted.
*/
void fc_disc_stop_final(struct fc_lport *lport)
{
fc_disc_stop(lport);
lport->tt.rport_flush_queue();
}
/**
* fc_disc_init() - Initialize the discovery block
* @lport: FC local port
*/
int fc_disc_init(struct fc_lport *lport)
{
struct fc_disc *disc;
if (!lport->tt.disc_start)
lport->tt.disc_start = fc_disc_start;
if (!lport->tt.disc_stop)
lport->tt.disc_stop = fc_disc_stop;
if (!lport->tt.disc_stop_final)
lport->tt.disc_stop_final = fc_disc_stop_final;
if (!lport->tt.disc_recv_req)
lport->tt.disc_recv_req = fc_disc_recv_req;
if (!lport->tt.rport_lookup)
lport->tt.rport_lookup = fc_disc_lookup_rport;
disc = &lport->disc;
INIT_DELAYED_WORK(&disc->disc_work, fc_disc_timeout);
mutex_init(&disc->disc_mutex);
INIT_LIST_HEAD(&disc->rports);
[SCSI] libfc: Track rogue remote ports Rogue ports are currently not tracked on any list. The only reference to them is through any outstanding exchanges pending on the rogue ports. If the module is removed while a retry is set on a rogue port (say a Plogi retry for instance), this retry is not cancelled because there is no reference to the rogue port in the discovery rports list. Thus the local port can clean itself up, delete the exchange pool, and then the rogue port timeout can fire and try to start up another exchange. This patch tracks the rogue ports in a new list disc->rogue_rports. Creating a new list instead of using the disc->rports list keeps remote port code change to a minimum. 1) Whenever a rogue port is created, it is immediately added to the disc->rogue_rports list. 2) When the rogues port goes to ready, it is removed from the rogue list and the real remote port is added to the disc->rports list 3) The removal of the rogue from the disc->rogue_rports list is done in the context of the fc_rport_work() workQ thread in discovery callback. 4) Real rports are removed from the disc->rports list like before. Lookup is done only in the real rports list. This avoids making large changes to the remote port code. 5) In fc_disc_stop_rports, the rogues list is traversed in addition to the real list to stop the rogue ports and issue logoffs on them. This way, rogue ports get cleaned up when the local port goes away. 6) rogue remote ports are not removed from the list right away, but removed late in fc_rport_work() context, multiple threads can find the same remote port in the list and call rport_logoff(). Rport_logoff() only continues with the logoff if port is not in NONE state, thus preventing multiple logoffs and multiple list deletions. 7) Since the rport is removed from the disc list at a later stage (in the disc callback), incoming frames can find the rport even if rport_logoff() has been called on the rport. When rport_logoff() is called, the rport state is set to NONE, and we are trying to cancel all exchanges and retries on that port. While in this state, if an incoming Plogi/Prli/Logo or other frames match the rport, we should not reply because the rport is in the NONE state. Just drop the frame, since the rport will be deleted soon in the disc callback (fc_rport_work) 8) In fc_disc_single(), remove rport lookup and call to fc_disc_del_target. fc_disc_single() is called from recv_rscn_req() where rport lookup and rport_logoff is already done. Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-21 17:27:04 -06:00
INIT_LIST_HEAD(&disc->rogue_rports);
disc->lport = lport;
disc->delay = FC_DISC_DELAY;
disc->event = DISC_EV_NONE;
return 0;
}
EXPORT_SYMBOL(fc_disc_init);