linux/drivers/scsi/bfa/rport.c

2677 lines
67 KiB
C
Raw Normal View History

/*
* Copyright (c) 2005-2009 Brocade Communications Systems, Inc.
* All rights reserved
* www.brocade.com
*
* Linux driver for Brocade Fibre Channel Host Bus Adapter.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License (GPL) Version 2 as
* published by the Free Software Foundation
*
* This program is distributed in the hope that 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.
*/
/**
* rport.c Remote port implementation.
*/
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 08:04:11 +00:00
#include <linux/slab.h>
#include <bfa.h>
#include <bfa_svc.h>
#include "fcbuild.h"
#include "fcs_vport.h"
#include "fcs_lport.h"
#include "fcs_rport.h"
#include "fcs_fcpim.h"
#include "fcs_fcptm.h"
#include "fcs_trcmod.h"
#include "fcs_fcxp.h"
#include "fcs.h"
#include <fcb/bfa_fcb_rport.h>
#include <aen/bfa_aen_rport.h>
BFA_TRC_FILE(FCS, RPORT);
/* In millisecs */
static u32 bfa_fcs_rport_del_timeout =
BFA_FCS_RPORT_DEF_DEL_TIMEOUT * 1000;
/*
* forward declarations
*/
static struct bfa_fcs_rport_s *bfa_fcs_rport_alloc(struct bfa_fcs_port_s *port,
wwn_t pwwn, u32 rpid);
static void bfa_fcs_rport_free(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_hal_online(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_online_action(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_offline_action(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_update(struct bfa_fcs_rport_s *rport,
struct fc_logi_s *plogi);
static void bfa_fcs_rport_fc4_pause(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_fc4_resume(struct bfa_fcs_rport_s *rport);
static void bfa_fcs_rport_timeout(void *arg);
static void bfa_fcs_rport_send_plogi(void *rport_cbarg,
struct bfa_fcxp_s *fcxp_alloced);
static void bfa_fcs_rport_send_plogiacc(void *rport_cbarg,
struct bfa_fcxp_s *fcxp_alloced);
static void bfa_fcs_rport_plogi_response(void *fcsarg,
struct bfa_fcxp_s *fcxp,
void *cbarg,
bfa_status_t req_status,
u32 rsp_len,
u32 resid_len,
struct fchs_s *rsp_fchs);
static void bfa_fcs_rport_send_adisc(void *rport_cbarg,
struct bfa_fcxp_s *fcxp_alloced);
static void bfa_fcs_rport_adisc_response(void *fcsarg,
struct bfa_fcxp_s *fcxp,
void *cbarg,
bfa_status_t req_status,
u32 rsp_len,
u32 resid_len,
struct fchs_s *rsp_fchs);
static void bfa_fcs_rport_send_gidpn(void *rport_cbarg,
struct bfa_fcxp_s *fcxp_alloced);
static void bfa_fcs_rport_gidpn_response(void *fcsarg,
struct bfa_fcxp_s *fcxp,
void *cbarg,
bfa_status_t req_status,
u32 rsp_len,
u32 resid_len,
struct fchs_s *rsp_fchs);
static void bfa_fcs_rport_send_logo(void *rport_cbarg,
struct bfa_fcxp_s *fcxp_alloced);
static void bfa_fcs_rport_send_logo_acc(void *rport_cbarg);
static void bfa_fcs_rport_process_prli(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u16 len);
static void bfa_fcs_rport_send_ls_rjt(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u8 reason_code,
u8 reason_code_expl);
static void bfa_fcs_rport_process_adisc(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u16 len);
static void bfa_fcs_rport_send_prlo_acc(struct bfa_fcs_rport_s *rport);
/**
* fcs_rport_sm FCS rport state machine events
*/
enum rport_event {
RPSM_EVENT_PLOGI_SEND = 1, /* new rport; start with PLOGI */
RPSM_EVENT_PLOGI_RCVD = 2, /* Inbound PLOGI from remote port */
RPSM_EVENT_PLOGI_COMP = 3, /* PLOGI completed to rport */
RPSM_EVENT_LOGO_RCVD = 4, /* LOGO from remote device */
RPSM_EVENT_LOGO_IMP = 5, /* implicit logo for SLER */
RPSM_EVENT_FCXP_SENT = 6, /* Frame from has been sent */
RPSM_EVENT_DELETE = 7, /* RPORT delete request */
RPSM_EVENT_SCN = 8, /* state change notification */
RPSM_EVENT_ACCEPTED = 9,/* Good response from remote device */
RPSM_EVENT_FAILED = 10, /* Request to rport failed. */
RPSM_EVENT_TIMEOUT = 11, /* Rport SM timeout event */
RPSM_EVENT_HCB_ONLINE = 12, /* BFA rport online callback */
RPSM_EVENT_HCB_OFFLINE = 13, /* BFA rport offline callback */
RPSM_EVENT_FC4_OFFLINE = 14, /* FC-4 offline complete */
RPSM_EVENT_ADDRESS_CHANGE = 15, /* Rport's PID has changed */
RPSM_EVENT_ADDRESS_DISC = 16, /* Need to Discover rport's PID */
RPSM_EVENT_PRLO_RCVD = 17, /* PRLO from remote device */
};
static void bfa_fcs_rport_sm_uninit(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_plogi_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_plogiacc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_plogi_retry(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_plogi(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_hal_online(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_online(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsquery_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsquery(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_adisc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_adisc(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_fc4_logorcv(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_fc4_logosend(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_fc4_offline(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_hcb_offline(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_hcb_logorcv(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_hcb_logosend(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_logo_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_offline(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsdisc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsdisc_retry(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsdisc_sent(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static void bfa_fcs_rport_sm_nsdisc_sent(struct bfa_fcs_rport_s *rport,
enum rport_event event);
static struct bfa_sm_table_s rport_sm_table[] = {
{BFA_SM(bfa_fcs_rport_sm_uninit), BFA_RPORT_UNINIT},
{BFA_SM(bfa_fcs_rport_sm_plogi_sending), BFA_RPORT_PLOGI},
{BFA_SM(bfa_fcs_rport_sm_plogiacc_sending), BFA_RPORT_ONLINE},
{BFA_SM(bfa_fcs_rport_sm_plogi_retry), BFA_RPORT_PLOGI_RETRY},
{BFA_SM(bfa_fcs_rport_sm_plogi), BFA_RPORT_PLOGI},
{BFA_SM(bfa_fcs_rport_sm_hal_online), BFA_RPORT_ONLINE},
{BFA_SM(bfa_fcs_rport_sm_online), BFA_RPORT_ONLINE},
{BFA_SM(bfa_fcs_rport_sm_nsquery_sending), BFA_RPORT_NSQUERY},
{BFA_SM(bfa_fcs_rport_sm_nsquery), BFA_RPORT_NSQUERY},
{BFA_SM(bfa_fcs_rport_sm_adisc_sending), BFA_RPORT_ADISC},
{BFA_SM(bfa_fcs_rport_sm_adisc), BFA_RPORT_ADISC},
{BFA_SM(bfa_fcs_rport_sm_fc4_logorcv), BFA_RPORT_LOGORCV},
{BFA_SM(bfa_fcs_rport_sm_fc4_logosend), BFA_RPORT_LOGO},
{BFA_SM(bfa_fcs_rport_sm_fc4_offline), BFA_RPORT_OFFLINE},
{BFA_SM(bfa_fcs_rport_sm_hcb_offline), BFA_RPORT_OFFLINE},
{BFA_SM(bfa_fcs_rport_sm_hcb_logorcv), BFA_RPORT_LOGORCV},
{BFA_SM(bfa_fcs_rport_sm_hcb_logosend), BFA_RPORT_LOGO},
{BFA_SM(bfa_fcs_rport_sm_logo_sending), BFA_RPORT_LOGO},
{BFA_SM(bfa_fcs_rport_sm_offline), BFA_RPORT_OFFLINE},
{BFA_SM(bfa_fcs_rport_sm_nsdisc_sending), BFA_RPORT_NSDISC},
{BFA_SM(bfa_fcs_rport_sm_nsdisc_retry), BFA_RPORT_NSDISC},
{BFA_SM(bfa_fcs_rport_sm_nsdisc_sent), BFA_RPORT_NSDISC},
};
/**
* Beginning state.
*/
static void
bfa_fcs_rport_sm_uninit(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_PLOGI_SEND:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi_sending);
rport->plogi_retries = 0;
bfa_fcs_rport_send_plogi(rport, NULL);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_fcs_rport_hal_online(rport);
break;
case RPSM_EVENT_ADDRESS_CHANGE:
case RPSM_EVENT_ADDRESS_DISC:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* PLOGI is being sent.
*/
static void
bfa_fcs_rport_sm_plogi_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_SCN:
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* PLOGI is being sent.
*/
static void
bfa_fcs_rport_sm_plogiacc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_fcs_rport_hal_online(rport);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_SCN:
/**
* Ignore, SCN is possibly online notification.
*/
break;
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_HCB_OFFLINE:
/**
* Ignore BFA callback, on a PLOGI receive we call bfa offline.
*/
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* PLOGI is sent.
*/
static void
bfa_fcs_rport_sm_plogi_retry(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_SCN:
bfa_timer_stop(&rport->timer);
/*
* !! fall through !!
*/
case RPSM_EVENT_TIMEOUT:
if (rport->plogi_retries < BFA_FCS_RPORT_MAX_RETRIES) {
rport->plogi_retries++;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi_sending);
bfa_fcs_rport_send_plogi(rport, NULL);
} else {
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
}
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_LOGO_RCVD:
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_timer_stop(&rport->timer);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_stop(&rport->timer);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_hal_online(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* PLOGI is sent.
*/
static void
bfa_fcs_rport_sm_plogi(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_ACCEPTED:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
rport->plogi_retries = 0;
bfa_fcs_rport_hal_online(rport);
break;
case RPSM_EVENT_LOGO_RCVD:
bfa_fcs_rport_send_logo_acc(rport);
/*
* !! fall through !!
*/
case RPSM_EVENT_PRLO_RCVD:
if (rport->prlo == BFA_TRUE)
bfa_fcs_rport_send_prlo_acc(rport);
bfa_fcxp_discard(rport->fcxp);
/*
* !! fall through !!
*/
case RPSM_EVENT_FAILED:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi_retry);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
BFA_FCS_RETRY_TIMEOUT);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_discard(rport->fcxp);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_fcxp_discard(rport->fcxp);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_SCN:
/**
* Ignore SCN - wait for PLOGI response.
*/
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_hal_online(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* PLOGI is complete. Awaiting BFA rport online callback. FC-4s
* are offline.
*/
static void
bfa_fcs_rport_sm_hal_online(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_HCB_ONLINE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_online);
bfa_fcs_rport_online_action(rport);
break;
case RPSM_EVENT_PRLO_RCVD:
break;
case RPSM_EVENT_LOGO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_logorcv);
bfa_rport_offline(rport->bfa_rport);
break;
case RPSM_EVENT_LOGO_IMP:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_offline);
bfa_rport_offline(rport->bfa_rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_rport_offline(rport->bfa_rport);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_logosend);
bfa_rport_offline(rport->bfa_rport);
break;
case RPSM_EVENT_SCN:
/**
* @todo
* Ignore SCN - PLOGI just completed, FC-4 login should detect
* device failures.
*/
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is ONLINE. FC-4s active.
*/
static void
bfa_fcs_rport_sm_online(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_SCN:
/**
* Pause FC-4 activity till rport is authenticated.
* In switched fabrics, check presence of device in nameserver
* first.
*/
bfa_fcs_rport_fc4_pause(rport);
if (bfa_fcs_fabric_is_switched(rport->port->fabric)) {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsquery_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
} else {
bfa_sm_set_state(rport, bfa_fcs_rport_sm_adisc_sending);
bfa_fcs_rport_send_adisc(rport, NULL);
}
break;
case RPSM_EVENT_PLOGI_RCVD:
case RPSM_EVENT_LOGO_IMP:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logorcv);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_PLOGI_COMP:
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* An SCN event is received in ONLINE state. NS query is being sent
* prior to ADISC authentication with rport. FC-4s are paused.
*/
static void
bfa_fcs_rport_sm_nsquery_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsquery);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_SCN:
/**
* ignore SCN, wait for response to query itself
*/
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logorcv);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_PLOGI_RCVD:
case RPSM_EVENT_ADDRESS_CHANGE:
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* An SCN event is received in ONLINE state. NS query is sent to rport.
* FC-4s are paused.
*/
static void
bfa_fcs_rport_sm_nsquery(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_ACCEPTED:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_adisc_sending);
bfa_fcs_rport_send_adisc(rport, NULL);
break;
case RPSM_EVENT_FAILED:
rport->ns_retries++;
if (rport->ns_retries < BFA_FCS_RPORT_MAX_RETRIES) {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsquery_sending);
bfa_fcs_rport_send_gidpn(rport, NULL);
} else {
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcs_rport_offline_action(rport);
}
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_SCN:
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logorcv);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_PLOGI_COMP:
case RPSM_EVENT_ADDRESS_CHANGE:
case RPSM_EVENT_PLOGI_RCVD:
case RPSM_EVENT_LOGO_IMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* An SCN event is received in ONLINE state. ADISC is being sent for
* authenticating with rport. FC-4s are paused.
*/
static void
bfa_fcs_rport_sm_adisc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_adisc);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_LOGO_IMP:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logorcv);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_SCN:
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_offline_action(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* An SCN event is received in ONLINE state. ADISC is to rport.
* FC-4s are paused.
*/
static void
bfa_fcs_rport_sm_adisc(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_ACCEPTED:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_online);
bfa_fcs_rport_fc4_resume(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
/**
* Too complex to cleanup FC-4 & rport and then acc to PLOGI.
* At least go offline when a PLOGI is received.
*/
bfa_fcxp_discard(rport->fcxp);
/*
* !!! fall through !!!
*/
case RPSM_EVENT_FAILED:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_SCN:
/**
* already processing RSCN
*/
break;
case RPSM_EVENT_LOGO_IMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_offline);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logorcv);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_offline_action(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport has sent LOGO. Awaiting FC-4 offline completion callback.
*/
static void
bfa_fcs_rport_sm_fc4_logorcv(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FC4_OFFLINE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_logorcv);
bfa_rport_offline(rport->bfa_rport);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_ADDRESS_CHANGE:
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* LOGO needs to be sent to rport. Awaiting FC-4 offline completion
* callback.
*/
static void
bfa_fcs_rport_sm_fc4_logosend(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FC4_OFFLINE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_logosend);
bfa_rport_offline(rport->bfa_rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is going offline. Awaiting FC-4 offline completion callback.
*/
static void
bfa_fcs_rport_sm_fc4_offline(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FC4_OFFLINE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_offline);
bfa_rport_offline(rport->bfa_rport);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_LOGO_IMP:
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_ADDRESS_CHANGE:
/**
* rport is already going offline.
* SCN - ignore and wait till transitioning to offline state
*/
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_fc4_logosend);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is offline. FC-4s are offline. Awaiting BFA rport offline
* callback.
*/
static void
bfa_fcs_rport_sm_hcb_offline(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_HCB_OFFLINE:
case RPSM_EVENT_ADDRESS_CHANGE:
if (bfa_fcs_port_is_online(rport->port)) {
if (bfa_fcs_fabric_is_switched(rport->port->fabric)) {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
} else {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_plogi_sending);
rport->plogi_retries = 0;
bfa_fcs_rport_send_plogi(rport, NULL);
}
} else {
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
}
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
/**
* Ignore, already offline.
*/
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is offline. FC-4s are offline. Awaiting BFA rport offline
* callback to send LOGO accept.
*/
static void
bfa_fcs_rport_sm_hcb_logorcv(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_HCB_OFFLINE:
case RPSM_EVENT_ADDRESS_CHANGE:
if (rport->pid && (rport->prlo == BFA_TRUE))
bfa_fcs_rport_send_prlo_acc(rport);
if (rport->pid && (rport->prlo == BFA_FALSE))
bfa_fcs_rport_send_logo_acc(rport);
/*
* If the lport is online and if the rport is not a well known
* address port, we try to re-discover the r-port.
*/
if (bfa_fcs_port_is_online(rport->port)
&& (!BFA_FCS_PID_IS_WKA(rport->pid))) {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
} else {
/*
* if it is not a well known address, reset the pid to
*
*/
if (!BFA_FCS_PID_IS_WKA(rport->pid))
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
}
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_logosend);
break;
case RPSM_EVENT_LOGO_IMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hcb_offline);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
/**
* Ignore - already processing a LOGO.
*/
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is being deleted. FC-4s are offline. Awaiting BFA rport offline
* callback to send LOGO.
*/
static void
bfa_fcs_rport_sm_hcb_logosend(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_HCB_OFFLINE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_logo_sending);
bfa_fcs_rport_send_logo(rport, NULL);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_ADDRESS_CHANGE:
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is being deleted. FC-4s are offline. LOGO is being sent.
*/
static void
bfa_fcs_rport_sm_logo_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
/*
* Once LOGO is sent, we donot wait for the response
*/
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_ADDRESS_CHANGE:
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_free(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport is offline. FC-4s are offline. BFA rport is offline.
* Timer active to delete stale rport.
*/
static void
bfa_fcs_rport_sm_offline(struct bfa_fcs_rport_s *rport, enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_TIMEOUT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
bfa_timer_stop(&rport->timer);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_LOGO_IMP:
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_hal_online(rport);
break;
case RPSM_EVENT_PLOGI_SEND:
bfa_timer_stop(&rport->timer);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi_sending);
rport->plogi_retries = 0;
bfa_fcs_rport_send_plogi(rport, NULL);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport address has changed. Nameserver discovery request is being sent.
*/
static void
bfa_fcs_rport_sm_nsdisc_sending(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_FCXP_SENT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sent);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_LOGO_RCVD:
case RPSM_EVENT_PRLO_RCVD:
case RPSM_EVENT_PLOGI_SEND:
break;
case RPSM_EVENT_ADDRESS_CHANGE:
rport->ns_retries = 0; /* reset the retry count */
break;
case RPSM_EVENT_LOGO_IMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_fcxp_walloc_cancel(rport->fcs->bfa, &rport->fcxp_wqe);
bfa_fcs_rport_hal_online(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Nameserver discovery failed. Waiting for timeout to retry.
*/
static void
bfa_fcs_rport_sm_nsdisc_retry(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_TIMEOUT:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_SCN:
case RPSM_EVENT_ADDRESS_CHANGE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_nsdisc_sending);
bfa_timer_stop(&rport->timer);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_stop(&rport->timer);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_LOGO_RCVD:
bfa_fcs_rport_send_logo_acc(rport);
break;
case RPSM_EVENT_PRLO_RCVD:
bfa_fcs_rport_send_prlo_acc(rport);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_timer_stop(&rport->timer);
bfa_fcs_rport_hal_online(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* Rport address has changed. Nameserver discovery request is sent.
*/
static void
bfa_fcs_rport_sm_nsdisc_sent(struct bfa_fcs_rport_s *rport,
enum rport_event event)
{
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->pid);
bfa_trc(rport->fcs, event);
switch (event) {
case RPSM_EVENT_ACCEPTED:
case RPSM_EVENT_ADDRESS_CHANGE:
if (rport->pid) {
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogi_sending);
bfa_fcs_rport_send_plogi(rport, NULL);
} else {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsdisc_sending);
rport->ns_retries = 0;
bfa_fcs_rport_send_gidpn(rport, NULL);
}
break;
case RPSM_EVENT_FAILED:
rport->ns_retries++;
if (rport->ns_retries < BFA_FCS_RPORT_MAX_RETRIES) {
bfa_sm_set_state(rport,
bfa_fcs_rport_sm_nsdisc_sending);
bfa_fcs_rport_send_gidpn(rport, NULL);
} else {
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
};
break;
case RPSM_EVENT_DELETE:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_free(rport);
break;
case RPSM_EVENT_PLOGI_RCVD:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_plogiacc_sending);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_send_plogiacc(rport, NULL);
break;
case RPSM_EVENT_LOGO_IMP:
rport->pid = 0;
bfa_sm_set_state(rport, bfa_fcs_rport_sm_offline);
bfa_fcxp_discard(rport->fcxp);
bfa_timer_start(rport->fcs->bfa, &rport->timer,
bfa_fcs_rport_timeout, rport,
bfa_fcs_rport_del_timeout);
break;
case RPSM_EVENT_PRLO_RCVD:
bfa_fcs_rport_send_prlo_acc(rport);
break;
case RPSM_EVENT_SCN:
/**
* ignore, wait for NS query response
*/
break;
case RPSM_EVENT_LOGO_RCVD:
/**
* Not logged-in yet. Accept LOGO.
*/
bfa_fcs_rport_send_logo_acc(rport);
break;
case RPSM_EVENT_PLOGI_COMP:
bfa_sm_set_state(rport, bfa_fcs_rport_sm_hal_online);
bfa_fcxp_discard(rport->fcxp);
bfa_fcs_rport_hal_online(rport);
break;
default:
bfa_sm_fault(rport->fcs, event);
}
}
/**
* fcs_rport_private FCS RPORT provate functions
*/
static void
bfa_fcs_rport_send_plogi(void *rport_cbarg, struct bfa_fcxp_s *fcxp_alloced)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
int len;
struct bfa_fcxp_s *fcxp;
bfa_trc(rport->fcs, rport->pwwn);
fcxp = fcxp_alloced ? fcxp_alloced : bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp) {
bfa_fcxp_alloc_wait(port->fcs->bfa, &rport->fcxp_wqe,
bfa_fcs_rport_send_plogi, rport);
return;
}
rport->fcxp = fcxp;
len = fc_plogi_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rport->pid,
bfa_fcs_port_get_fcid(port), 0,
port->port_cfg.pwwn, port->port_cfg.nwwn,
bfa_fcport_get_maxfrsize(port->fcs->bfa));
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, bfa_fcs_rport_plogi_response,
(void *)rport, FC_MAX_PDUSZ, FC_ELS_TOV);
rport->stats.plogis++;
bfa_sm_send_event(rport, RPSM_EVENT_FCXP_SENT);
}
static void
bfa_fcs_rport_plogi_response(void *fcsarg, struct bfa_fcxp_s *fcxp, void *cbarg,
bfa_status_t req_status, u32 rsp_len,
u32 resid_len, struct fchs_s *rsp_fchs)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
struct fc_logi_s *plogi_rsp;
struct fc_ls_rjt_s *ls_rjt;
struct bfa_fcs_rport_s *twin;
struct list_head *qe;
bfa_trc(rport->fcs, rport->pwwn);
/*
* Sanity Checks
*/
if (req_status != BFA_STATUS_OK) {
bfa_trc(rport->fcs, req_status);
rport->stats.plogi_failed++;
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
return;
}
plogi_rsp = (struct fc_logi_s *) BFA_FCXP_RSP_PLD(fcxp);
/**
* Check for failure first.
*/
if (plogi_rsp->els_cmd.els_code != FC_ELS_ACC) {
ls_rjt = (struct fc_ls_rjt_s *) BFA_FCXP_RSP_PLD(fcxp);
bfa_trc(rport->fcs, ls_rjt->reason_code);
bfa_trc(rport->fcs, ls_rjt->reason_code_expl);
rport->stats.plogi_rejects++;
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
return;
}
/**
* PLOGI is complete. Make sure this device is not one of the known
* device with a new FC port address.
*/
list_for_each(qe, &rport->port->rport_q) {
twin = (struct bfa_fcs_rport_s *)qe;
if (twin == rport)
continue;
if (!rport->pwwn && (plogi_rsp->port_name == twin->pwwn)) {
bfa_trc(rport->fcs, twin->pid);
bfa_trc(rport->fcs, rport->pid);
/*
* Update plogi stats in twin
*/
twin->stats.plogis += rport->stats.plogis;
twin->stats.plogi_rejects += rport->stats.plogi_rejects;
twin->stats.plogi_timeouts +=
rport->stats.plogi_timeouts;
twin->stats.plogi_failed += rport->stats.plogi_failed;
twin->stats.plogi_rcvd += rport->stats.plogi_rcvd;
twin->stats.plogi_accs++;
bfa_fcs_rport_delete(rport);
bfa_fcs_rport_update(twin, plogi_rsp);
twin->pid = rsp_fchs->s_id;
bfa_sm_send_event(twin, RPSM_EVENT_PLOGI_COMP);
return;
}
}
/**
* Normal login path -- no evil twins.
*/
rport->stats.plogi_accs++;
bfa_fcs_rport_update(rport, plogi_rsp);
bfa_sm_send_event(rport, RPSM_EVENT_ACCEPTED);
}
static void
bfa_fcs_rport_send_plogiacc(void *rport_cbarg, struct bfa_fcxp_s *fcxp_alloced)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
int len;
struct bfa_fcxp_s *fcxp;
bfa_trc(rport->fcs, rport->pwwn);
bfa_trc(rport->fcs, rport->reply_oxid);
fcxp = fcxp_alloced ? fcxp_alloced : bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp) {
bfa_fcxp_alloc_wait(port->fcs->bfa, &rport->fcxp_wqe,
bfa_fcs_rport_send_plogiacc, rport);
return;
}
rport->fcxp = fcxp;
len = fc_plogi_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rport->pid,
bfa_fcs_port_get_fcid(port), rport->reply_oxid,
port->port_cfg.pwwn, port->port_cfg.nwwn,
bfa_fcport_get_maxfrsize(port->fcs->bfa));
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, NULL, FC_MAX_PDUSZ, 0);
bfa_sm_send_event(rport, RPSM_EVENT_FCXP_SENT);
}
static void
bfa_fcs_rport_send_adisc(void *rport_cbarg, struct bfa_fcxp_s *fcxp_alloced)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
int len;
struct bfa_fcxp_s *fcxp;
bfa_trc(rport->fcs, rport->pwwn);
fcxp = fcxp_alloced ? fcxp_alloced : bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp) {
bfa_fcxp_alloc_wait(port->fcs->bfa, &rport->fcxp_wqe,
bfa_fcs_rport_send_adisc, rport);
return;
}
rport->fcxp = fcxp;
len = fc_adisc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rport->pid,
bfa_fcs_port_get_fcid(port), 0,
port->port_cfg.pwwn, port->port_cfg.nwwn);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, bfa_fcs_rport_adisc_response,
rport, FC_MAX_PDUSZ, FC_ELS_TOV);
rport->stats.adisc_sent++;
bfa_sm_send_event(rport, RPSM_EVENT_FCXP_SENT);
}
static void
bfa_fcs_rport_adisc_response(void *fcsarg, struct bfa_fcxp_s *fcxp, void *cbarg,
bfa_status_t req_status, u32 rsp_len,
u32 resid_len, struct fchs_s *rsp_fchs)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
void *pld = bfa_fcxp_get_rspbuf(fcxp);
struct fc_ls_rjt_s *ls_rjt;
if (req_status != BFA_STATUS_OK) {
bfa_trc(rport->fcs, req_status);
rport->stats.adisc_failed++;
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
return;
}
if (fc_adisc_rsp_parse((struct fc_adisc_s *)pld, rsp_len, rport->pwwn,
rport->nwwn) == FC_PARSE_OK) {
rport->stats.adisc_accs++;
bfa_sm_send_event(rport, RPSM_EVENT_ACCEPTED);
return;
}
rport->stats.adisc_rejects++;
ls_rjt = pld;
bfa_trc(rport->fcs, ls_rjt->els_cmd.els_code);
bfa_trc(rport->fcs, ls_rjt->reason_code);
bfa_trc(rport->fcs, ls_rjt->reason_code_expl);
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
}
static void
bfa_fcs_rport_send_gidpn(void *rport_cbarg, struct bfa_fcxp_s *fcxp_alloced)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
struct bfa_fcxp_s *fcxp;
int len;
bfa_trc(rport->fcs, rport->pid);
fcxp = fcxp_alloced ? fcxp_alloced : bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp) {
bfa_fcxp_alloc_wait(port->fcs->bfa, &rport->fcxp_wqe,
bfa_fcs_rport_send_gidpn, rport);
return;
}
rport->fcxp = fcxp;
len = fc_gidpn_build(&fchs, bfa_fcxp_get_reqbuf(fcxp),
bfa_fcs_port_get_fcid(port), 0, rport->pwwn);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, bfa_fcs_rport_gidpn_response,
(void *)rport, FC_MAX_PDUSZ, FC_FCCT_TOV);
bfa_sm_send_event(rport, RPSM_EVENT_FCXP_SENT);
}
static void
bfa_fcs_rport_gidpn_response(void *fcsarg, struct bfa_fcxp_s *fcxp, void *cbarg,
bfa_status_t req_status, u32 rsp_len,
u32 resid_len, struct fchs_s *rsp_fchs)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
struct bfa_fcs_rport_s *twin;
struct list_head *qe;
struct ct_hdr_s *cthdr;
struct fcgs_gidpn_resp_s *gidpn_rsp;
bfa_trc(rport->fcs, rport->pwwn);
cthdr = (struct ct_hdr_s *) BFA_FCXP_RSP_PLD(fcxp);
cthdr->cmd_rsp_code = bfa_os_ntohs(cthdr->cmd_rsp_code);
if (cthdr->cmd_rsp_code == CT_RSP_ACCEPT) {
/*
* Check if the pid is the same as before.
*/
gidpn_rsp = (struct fcgs_gidpn_resp_s *) (cthdr + 1);
if (gidpn_rsp->dap == rport->pid) {
/*
* Device is online
*/
bfa_sm_send_event(rport, RPSM_EVENT_ACCEPTED);
} else {
/*
* Device's PID has changed. We need to cleanup and
* re-login. If there is another device with the the
* newly discovered pid, send an scn notice so that its
* new pid can be discovered.
*/
list_for_each(qe, &rport->port->rport_q) {
twin = (struct bfa_fcs_rport_s *)qe;
if (twin == rport)
continue;
if (gidpn_rsp->dap == twin->pid) {
bfa_trc(rport->fcs, twin->pid);
bfa_trc(rport->fcs, rport->pid);
twin->pid = 0;
bfa_sm_send_event(twin,
RPSM_EVENT_ADDRESS_CHANGE);
}
}
rport->pid = gidpn_rsp->dap;
bfa_sm_send_event(rport, RPSM_EVENT_ADDRESS_CHANGE);
}
return;
}
/*
* Reject Response
*/
switch (cthdr->reason_code) {
case CT_RSN_LOGICAL_BUSY:
/*
* Need to retry
*/
bfa_sm_send_event(rport, RPSM_EVENT_TIMEOUT);
break;
case CT_RSN_UNABLE_TO_PERF:
/*
* device doesn't exist : Start timer to cleanup this later.
*/
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
break;
default:
bfa_sm_send_event(rport, RPSM_EVENT_FAILED);
break;
}
}
/**
* Called to send a logout to the rport.
*/
static void
bfa_fcs_rport_send_logo(void *rport_cbarg, struct bfa_fcxp_s *fcxp_alloced)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port;
struct fchs_s fchs;
struct bfa_fcxp_s *fcxp;
u16 len;
bfa_trc(rport->fcs, rport->pid);
port = rport->port;
fcxp = fcxp_alloced ? fcxp_alloced : bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp) {
bfa_fcxp_alloc_wait(port->fcs->bfa, &rport->fcxp_wqe,
bfa_fcs_rport_send_logo, rport);
return;
}
rport->fcxp = fcxp;
len = fc_logo_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rport->pid,
bfa_fcs_port_get_fcid(port), 0,
bfa_fcs_port_get_pwwn(port));
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, rport, FC_MAX_PDUSZ,
FC_ELS_TOV);
rport->stats.logos++;
bfa_fcxp_discard(rport->fcxp);
bfa_sm_send_event(rport, RPSM_EVENT_FCXP_SENT);
}
/**
* Send ACC for a LOGO received.
*/
static void
bfa_fcs_rport_send_logo_acc(void *rport_cbarg)
{
struct bfa_fcs_rport_s *rport = rport_cbarg;
struct bfa_fcs_port_s *port;
struct fchs_s fchs;
struct bfa_fcxp_s *fcxp;
u16 len;
bfa_trc(rport->fcs, rport->pid);
port = rport->port;
fcxp = bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp)
return;
rport->stats.logo_rcvd++;
len = fc_logo_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rport->pid,
bfa_fcs_port_get_fcid(port), rport->reply_oxid);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, NULL, FC_MAX_PDUSZ, 0);
}
/**
* This routine will be called by bfa_timer on timer timeouts.
*
* param[in] rport - pointer to bfa_fcs_port_ns_t.
* param[out] rport_status - pointer to return vport status in
*
* return
* void
*
* Special Considerations:
*
* note
*/
static void
bfa_fcs_rport_timeout(void *arg)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)arg;
rport->stats.plogi_timeouts++;
bfa_sm_send_event(rport, RPSM_EVENT_TIMEOUT);
}
static void
bfa_fcs_rport_process_prli(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u16 len)
{
struct bfa_fcxp_s *fcxp;
struct fchs_s fchs;
struct bfa_fcs_port_s *port = rport->port;
struct fc_prli_s *prli;
bfa_trc(port->fcs, rx_fchs->s_id);
bfa_trc(port->fcs, rx_fchs->d_id);
rport->stats.prli_rcvd++;
if (BFA_FCS_VPORT_IS_TARGET_MODE(port)) {
/*
* Target Mode : Let the fcptm handle it
*/
bfa_fcs_tin_rx_prli(rport->tin, rx_fchs, len);
return;
}
/*
* We are either in Initiator or ipfc Mode
*/
prli = (struct fc_prli_s *) (rx_fchs + 1);
if (prli->parampage.servparams.initiator) {
bfa_trc(rport->fcs, prli->parampage.type);
rport->scsi_function = BFA_RPORT_INITIATOR;
bfa_fcs_itnim_is_initiator(rport->itnim);
} else {
/*
* @todo: PRLI from a target ?
*/
bfa_trc(port->fcs, rx_fchs->s_id);
rport->scsi_function = BFA_RPORT_TARGET;
}
fcxp = bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp)
return;
len = fc_prli_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rx_fchs->s_id,
bfa_fcs_port_get_fcid(port), rx_fchs->ox_id,
port->port_cfg.roles);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, NULL, FC_MAX_PDUSZ, 0);
}
static void
bfa_fcs_rport_process_rpsc(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u16 len)
{
struct bfa_fcxp_s *fcxp;
struct fchs_s fchs;
struct bfa_fcs_port_s *port = rport->port;
struct fc_rpsc_speed_info_s speeds;
struct bfa_pport_attr_s pport_attr;
bfa_trc(port->fcs, rx_fchs->s_id);
bfa_trc(port->fcs, rx_fchs->d_id);
rport->stats.rpsc_rcvd++;
speeds.port_speed_cap =
RPSC_SPEED_CAP_1G | RPSC_SPEED_CAP_2G | RPSC_SPEED_CAP_4G |
RPSC_SPEED_CAP_8G;
/*
* get curent speed from pport attributes from BFA
*/
bfa_fcport_get_attr(port->fcs->bfa, &pport_attr);
speeds.port_op_speed = fc_bfa_speed_to_rpsc_operspeed(pport_attr.speed);
fcxp = bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp)
return;
len = fc_rpsc_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rx_fchs->s_id,
bfa_fcs_port_get_fcid(port), rx_fchs->ox_id,
&speeds);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, NULL, FC_MAX_PDUSZ, 0);
}
static void
bfa_fcs_rport_process_adisc(struct bfa_fcs_rport_s *rport,
struct fchs_s *rx_fchs, u16 len)
{
struct bfa_fcxp_s *fcxp;
struct fchs_s fchs;
struct bfa_fcs_port_s *port = rport->port;
struct fc_adisc_s *adisc;
bfa_trc(port->fcs, rx_fchs->s_id);
bfa_trc(port->fcs, rx_fchs->d_id);
rport->stats.adisc_rcvd++;
if (BFA_FCS_VPORT_IS_TARGET_MODE(port)) {
/*
* @todo : Target Mode handling
*/
bfa_trc(port->fcs, rx_fchs->d_id);
bfa_assert(0);
return;
}
adisc = (struct fc_adisc_s *) (rx_fchs + 1);
/*
* Accept if the itnim for this rport is online. Else reject the ADISC
*/
if (bfa_fcs_itnim_get_online_state(rport->itnim) == BFA_STATUS_OK) {
fcxp = bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp)
return;
len = fc_adisc_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp),
rx_fchs->s_id,
bfa_fcs_port_get_fcid(port),
rx_fchs->ox_id, port->port_cfg.pwwn,
port->port_cfg.nwwn);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag,
BFA_FALSE, FC_CLASS_3, len, &fchs, NULL, NULL,
FC_MAX_PDUSZ, 0);
} else {
rport->stats.adisc_rejected++;
bfa_fcs_rport_send_ls_rjt(rport, rx_fchs,
FC_LS_RJT_RSN_UNABLE_TO_PERF_CMD,
FC_LS_RJT_EXP_LOGIN_REQUIRED);
}
}
static void
bfa_fcs_rport_hal_online(struct bfa_fcs_rport_s *rport)
{
struct bfa_fcs_port_s *port = rport->port;
struct bfa_rport_info_s rport_info;
rport_info.pid = rport->pid;
rport_info.local_pid = port->pid;
rport_info.lp_tag = port->lp_tag;
rport_info.vf_id = port->fabric->vf_id;
rport_info.vf_en = port->fabric->is_vf;
rport_info.fc_class = rport->fc_cos;
rport_info.cisc = rport->cisc;
rport_info.max_frmsz = rport->maxfrsize;
bfa_rport_online(rport->bfa_rport, &rport_info);
}
static void
bfa_fcs_rport_fc4_pause(struct bfa_fcs_rport_s *rport)
{
if (bfa_fcs_port_is_initiator(rport->port))
bfa_fcs_itnim_pause(rport->itnim);
if (bfa_fcs_port_is_target(rport->port))
bfa_fcs_tin_pause(rport->tin);
}
static void
bfa_fcs_rport_fc4_resume(struct bfa_fcs_rport_s *rport)
{
if (bfa_fcs_port_is_initiator(rport->port))
bfa_fcs_itnim_resume(rport->itnim);
if (bfa_fcs_port_is_target(rport->port))
bfa_fcs_tin_resume(rport->tin);
}
static struct bfa_fcs_rport_s *
bfa_fcs_rport_alloc(struct bfa_fcs_port_s *port, wwn_t pwwn, u32 rpid)
{
struct bfa_fcs_s *fcs = port->fcs;
struct bfa_fcs_rport_s *rport;
struct bfad_rport_s *rport_drv;
/**
* allocate rport
*/
if (bfa_fcb_rport_alloc(fcs->bfad, &rport, &rport_drv)
!= BFA_STATUS_OK) {
bfa_trc(fcs, rpid);
return NULL;
}
/*
* Initialize r-port
*/
rport->port = port;
rport->fcs = fcs;
rport->rp_drv = rport_drv;
rport->pid = rpid;
rport->pwwn = pwwn;
/**
* allocate BFA rport
*/
rport->bfa_rport = bfa_rport_create(port->fcs->bfa, rport);
if (!rport->bfa_rport) {
bfa_trc(fcs, rpid);
kfree(rport_drv);
return NULL;
}
/**
* allocate FC-4s
*/
bfa_assert(bfa_fcs_port_is_initiator(port) ^
bfa_fcs_port_is_target(port));
if (bfa_fcs_port_is_initiator(port)) {
rport->itnim = bfa_fcs_itnim_create(rport);
if (!rport->itnim) {
bfa_trc(fcs, rpid);
bfa_rport_delete(rport->bfa_rport);
kfree(rport_drv);
return NULL;
}
}
if (bfa_fcs_port_is_target(port)) {
rport->tin = bfa_fcs_tin_create(rport);
if (!rport->tin) {
bfa_trc(fcs, rpid);
bfa_rport_delete(rport->bfa_rport);
kfree(rport_drv);
return NULL;
}
}
bfa_fcs_port_add_rport(port, rport);
bfa_sm_set_state(rport, bfa_fcs_rport_sm_uninit);
/*
* Initialize the Rport Features(RPF) Sub Module
*/
if (!BFA_FCS_PID_IS_WKA(rport->pid))
bfa_fcs_rpf_init(rport);
return rport;
}
static void
bfa_fcs_rport_free(struct bfa_fcs_rport_s *rport)
{
struct bfa_fcs_port_s *port = rport->port;
/**
* - delete FC-4s
* - delete BFA rport
* - remove from queue of rports
*/
if (bfa_fcs_port_is_initiator(port))
bfa_fcs_itnim_delete(rport->itnim);
if (bfa_fcs_port_is_target(port))
bfa_fcs_tin_delete(rport->tin);
bfa_rport_delete(rport->bfa_rport);
bfa_fcs_port_del_rport(port, rport);
kfree(rport->rp_drv);
}
static void
bfa_fcs_rport_aen_post(struct bfa_fcs_rport_s *rport,
enum bfa_rport_aen_event event,
struct bfa_rport_aen_data_s *data)
{
union bfa_aen_data_u aen_data;
struct bfa_log_mod_s *logmod = rport->fcs->logm;
wwn_t lpwwn = bfa_fcs_port_get_pwwn(rport->port);
wwn_t rpwwn = rport->pwwn;
char lpwwn_ptr[BFA_STRING_32];
char rpwwn_ptr[BFA_STRING_32];
char *prio_str[] = { "unknown", "high", "medium", "low" };
wwn2str(lpwwn_ptr, lpwwn);
wwn2str(rpwwn_ptr, rpwwn);
switch (event) {
case BFA_RPORT_AEN_ONLINE:
case BFA_RPORT_AEN_OFFLINE:
case BFA_RPORT_AEN_DISCONNECT:
bfa_log(logmod, BFA_LOG_CREATE_ID(BFA_AEN_CAT_RPORT, event),
rpwwn_ptr, lpwwn_ptr);
break;
case BFA_RPORT_AEN_QOS_PRIO:
aen_data.rport.priv.qos = data->priv.qos;
bfa_log(logmod, BFA_AEN_RPORT_QOS_PRIO,
prio_str[aen_data.rport.priv.qos.qos_priority],
rpwwn_ptr, lpwwn_ptr);
break;
case BFA_RPORT_AEN_QOS_FLOWID:
aen_data.rport.priv.qos = data->priv.qos;
bfa_log(logmod, BFA_AEN_RPORT_QOS_FLOWID,
aen_data.rport.priv.qos.qos_flow_id, rpwwn_ptr,
lpwwn_ptr);
break;
default:
break;
}
aen_data.rport.vf_id = rport->port->fabric->vf_id;
aen_data.rport.ppwwn =
bfa_fcs_port_get_pwwn(bfa_fcs_get_base_port(rport->fcs));
aen_data.rport.lpwwn = lpwwn;
aen_data.rport.rpwwn = rpwwn;
}
static void
bfa_fcs_rport_online_action(struct bfa_fcs_rport_s *rport)
{
struct bfa_fcs_port_s *port = rport->port;
rport->stats.onlines++;
if (bfa_fcs_port_is_initiator(port)) {
bfa_fcs_itnim_rport_online(rport->itnim);
if (!BFA_FCS_PID_IS_WKA(rport->pid))
bfa_fcs_rpf_rport_online(rport);
};
if (bfa_fcs_port_is_target(port))
bfa_fcs_tin_rport_online(rport->tin);
/*
* Don't post events for well known addresses
*/
if (!BFA_FCS_PID_IS_WKA(rport->pid))
bfa_fcs_rport_aen_post(rport, BFA_RPORT_AEN_ONLINE, NULL);
}
static void
bfa_fcs_rport_offline_action(struct bfa_fcs_rport_s *rport)
{
struct bfa_fcs_port_s *port = rport->port;
rport->stats.offlines++;
/*
* Don't post events for well known addresses
*/
if (!BFA_FCS_PID_IS_WKA(rport->pid)) {
if (bfa_fcs_port_is_online(rport->port) == BFA_TRUE) {
bfa_fcs_rport_aen_post(rport, BFA_RPORT_AEN_DISCONNECT,
NULL);
} else {
bfa_fcs_rport_aen_post(rport, BFA_RPORT_AEN_OFFLINE,
NULL);
}
}
if (bfa_fcs_port_is_initiator(port)) {
bfa_fcs_itnim_rport_offline(rport->itnim);
if (!BFA_FCS_PID_IS_WKA(rport->pid))
bfa_fcs_rpf_rport_offline(rport);
}
if (bfa_fcs_port_is_target(port))
bfa_fcs_tin_rport_offline(rport->tin);
}
/**
* Update rport parameters from PLOGI or PLOGI accept.
*/
static void
bfa_fcs_rport_update(struct bfa_fcs_rport_s *rport, struct fc_logi_s *plogi)
{
struct bfa_fcs_port_s *port = rport->port;
/**
* - port name
* - node name
*/
rport->pwwn = plogi->port_name;
rport->nwwn = plogi->node_name;
/**
* - class of service
*/
rport->fc_cos = 0;
if (plogi->class3.class_valid)
rport->fc_cos = FC_CLASS_3;
if (plogi->class2.class_valid)
rport->fc_cos |= FC_CLASS_2;
/**
* - CISC
* - MAX receive frame size
*/
rport->cisc = plogi->csp.cisc;
rport->maxfrsize = bfa_os_ntohs(plogi->class3.rxsz);
bfa_trc(port->fcs, bfa_os_ntohs(plogi->csp.bbcred));
bfa_trc(port->fcs, port->fabric->bb_credit);
/**
* Direct Attach P2P mode :
* This is to handle a bug (233476) in IBM targets in Direct Attach
* Mode. Basically, in FLOGI Accept the target would have erroneously
* set the BB Credit to the value used in the FLOGI sent by the HBA.
* It uses the correct value (its own BB credit) in PLOGI.
*/
if ((!bfa_fcs_fabric_is_switched(port->fabric))
&& (bfa_os_ntohs(plogi->csp.bbcred) < port->fabric->bb_credit)) {
bfa_trc(port->fcs, bfa_os_ntohs(plogi->csp.bbcred));
bfa_trc(port->fcs, port->fabric->bb_credit);
port->fabric->bb_credit = bfa_os_ntohs(plogi->csp.bbcred);
bfa_fcport_set_tx_bbcredit(port->fcs->bfa,
port->fabric->bb_credit);
}
}
/**
* Called to handle LOGO received from an existing remote port.
*/
static void
bfa_fcs_rport_process_logo(struct bfa_fcs_rport_s *rport, struct fchs_s *fchs)
{
rport->reply_oxid = fchs->ox_id;
bfa_trc(rport->fcs, rport->reply_oxid);
rport->prlo = BFA_FALSE;
rport->stats.logo_rcvd++;
bfa_sm_send_event(rport, RPSM_EVENT_LOGO_RCVD);
}
/**
* fcs_rport_public FCS rport public interfaces
*/
/**
* Called by bport/vport to create a remote port instance for a discovered
* remote device.
*
* @param[in] port - base port or vport
* @param[in] rpid - remote port ID
*
* @return None
*/
struct bfa_fcs_rport_s *
bfa_fcs_rport_create(struct bfa_fcs_port_s *port, u32 rpid)
{
struct bfa_fcs_rport_s *rport;
bfa_trc(port->fcs, rpid);
rport = bfa_fcs_rport_alloc(port, WWN_NULL, rpid);
if (!rport)
return NULL;
bfa_sm_send_event(rport, RPSM_EVENT_PLOGI_SEND);
return rport;
}
/**
* Called to create a rport for which only the wwn is known.
*
* @param[in] port - base port
* @param[in] rpwwn - remote port wwn
*
* @return None
*/
struct bfa_fcs_rport_s *
bfa_fcs_rport_create_by_wwn(struct bfa_fcs_port_s *port, wwn_t rpwwn)
{
struct bfa_fcs_rport_s *rport;
bfa_trc(port->fcs, rpwwn);
rport = bfa_fcs_rport_alloc(port, rpwwn, 0);
if (!rport)
return NULL;
bfa_sm_send_event(rport, RPSM_EVENT_ADDRESS_DISC);
return rport;
}
/**
* Called by bport in private loop topology to indicate that a
* rport has been discovered and plogi has been completed.
*
* @param[in] port - base port or vport
* @param[in] rpid - remote port ID
*/
void
bfa_fcs_rport_start(struct bfa_fcs_port_s *port, struct fchs_s *fchs,
struct fc_logi_s *plogi)
{
struct bfa_fcs_rport_s *rport;
rport = bfa_fcs_rport_alloc(port, WWN_NULL, fchs->s_id);
if (!rport)
return;
bfa_fcs_rport_update(rport, plogi);
bfa_sm_send_event(rport, RPSM_EVENT_PLOGI_COMP);
}
/**
* Called by bport/vport to handle PLOGI received from a new remote port.
* If an existing rport does a plogi, it will be handled separately.
*/
void
bfa_fcs_rport_plogi_create(struct bfa_fcs_port_s *port, struct fchs_s *fchs,
struct fc_logi_s *plogi)
{
struct bfa_fcs_rport_s *rport;
rport = bfa_fcs_rport_alloc(port, plogi->port_name, fchs->s_id);
if (!rport)
return;
bfa_fcs_rport_update(rport, plogi);
rport->reply_oxid = fchs->ox_id;
bfa_trc(rport->fcs, rport->reply_oxid);
rport->stats.plogi_rcvd++;
bfa_sm_send_event(rport, RPSM_EVENT_PLOGI_RCVD);
}
static int
wwn_compare(wwn_t wwn1, wwn_t wwn2)
{
u8 *b1 = (u8 *) &wwn1;
u8 *b2 = (u8 *) &wwn2;
int i;
for (i = 0; i < sizeof(wwn_t); i++) {
if (b1[i] < b2[i])
return -1;
if (b1[i] > b2[i])
return 1;
}
return 0;
}
/**
* Called by bport/vport to handle PLOGI received from an existing
* remote port.
*/
void
bfa_fcs_rport_plogi(struct bfa_fcs_rport_s *rport, struct fchs_s *rx_fchs,
struct fc_logi_s *plogi)
{
/**
* @todo Handle P2P and initiator-initiator.
*/
bfa_fcs_rport_update(rport, plogi);
rport->reply_oxid = rx_fchs->ox_id;
bfa_trc(rport->fcs, rport->reply_oxid);
/**
* In Switched fabric topology,
* PLOGI to each other. If our pwwn is smaller, ignore it,
* if it is not a well known address.
* If the link topology is N2N,
* this Plogi should be accepted.
*/
if ((wwn_compare(rport->port->port_cfg.pwwn, rport->pwwn) == -1)
&& (bfa_fcs_fabric_is_switched(rport->port->fabric))
&& (!BFA_FCS_PID_IS_WKA(rport->pid))) {
bfa_trc(rport->fcs, rport->pid);
return;
}
rport->stats.plogi_rcvd++;
bfa_sm_send_event(rport, RPSM_EVENT_PLOGI_RCVD);
}
/**
* Called by bport/vport to delete a remote port instance.
*
* Rport delete is called under the following conditions:
* - vport is deleted
* - vf is deleted
* - explicit request from OS to delete rport (vmware)
*/
void
bfa_fcs_rport_delete(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_DELETE);
}
/**
* Called by bport/vport to when a target goes offline.
*
*/
void
bfa_fcs_rport_offline(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_LOGO_IMP);
}
/**
* Called by bport in n2n when a target (attached port) becomes online.
*
*/
void
bfa_fcs_rport_online(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_PLOGI_SEND);
}
/**
* Called by bport/vport to notify SCN for the remote port
*/
void
bfa_fcs_rport_scn(struct bfa_fcs_rport_s *rport)
{
rport->stats.rscns++;
bfa_sm_send_event(rport, RPSM_EVENT_SCN);
}
/**
* Called by fcpim to notify that the ITN cleanup is done.
*/
void
bfa_fcs_rport_itnim_ack(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_FC4_OFFLINE);
}
/**
* Called by fcptm to notify that the ITN cleanup is done.
*/
void
bfa_fcs_rport_tin_ack(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_FC4_OFFLINE);
}
/**
* This routine BFA callback for bfa_rport_online() call.
*
* param[in] cb_arg - rport struct.
*
* return
* void
*
* Special Considerations:
*
* note
*/
void
bfa_cb_rport_online(void *cbarg)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
bfa_trc(rport->fcs, rport->pwwn);
bfa_sm_send_event(rport, RPSM_EVENT_HCB_ONLINE);
}
/**
* This routine BFA callback for bfa_rport_offline() call.
*
* param[in] rport -
*
* return
* void
*
* Special Considerations:
*
* note
*/
void
bfa_cb_rport_offline(void *cbarg)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
bfa_trc(rport->fcs, rport->pwwn);
bfa_sm_send_event(rport, RPSM_EVENT_HCB_OFFLINE);
}
/**
* This routine is a static BFA callback when there is a QoS flow_id
* change notification
*
* @param[in] rport -
*
* @return void
*
* Special Considerations:
*
* @note
*/
void
bfa_cb_rport_qos_scn_flowid(void *cbarg,
struct bfa_rport_qos_attr_s old_qos_attr,
struct bfa_rport_qos_attr_s new_qos_attr)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
struct bfa_rport_aen_data_s aen_data;
bfa_trc(rport->fcs, rport->pwwn);
aen_data.priv.qos = new_qos_attr;
bfa_fcs_rport_aen_post(rport, BFA_RPORT_AEN_QOS_FLOWID, &aen_data);
}
/**
* This routine is a static BFA callback when there is a QoS priority
* change notification
*
* @param[in] rport -
*
* @return void
*
* Special Considerations:
*
* @note
*/
void
bfa_cb_rport_qos_scn_prio(void *cbarg, struct bfa_rport_qos_attr_s old_qos_attr,
struct bfa_rport_qos_attr_s new_qos_attr)
{
struct bfa_fcs_rport_s *rport = (struct bfa_fcs_rport_s *)cbarg;
struct bfa_rport_aen_data_s aen_data;
bfa_trc(rport->fcs, rport->pwwn);
aen_data.priv.qos = new_qos_attr;
bfa_fcs_rport_aen_post(rport, BFA_RPORT_AEN_QOS_PRIO, &aen_data);
}
/**
* Called to process any unsolicted frames from this remote port
*/
void
bfa_fcs_rport_logo_imp(struct bfa_fcs_rport_s *rport)
{
bfa_sm_send_event(rport, RPSM_EVENT_LOGO_IMP);
}
/**
* Called to process any unsolicted frames from this remote port
*/
void
bfa_fcs_rport_uf_recv(struct bfa_fcs_rport_s *rport, struct fchs_s *fchs,
u16 len)
{
struct bfa_fcs_port_s *port = rport->port;
struct fc_els_cmd_s *els_cmd;
bfa_trc(rport->fcs, fchs->s_id);
bfa_trc(rport->fcs, fchs->d_id);
bfa_trc(rport->fcs, fchs->type);
if (fchs->type != FC_TYPE_ELS)
return;
els_cmd = (struct fc_els_cmd_s *) (fchs + 1);
bfa_trc(rport->fcs, els_cmd->els_code);
switch (els_cmd->els_code) {
case FC_ELS_LOGO:
bfa_fcs_rport_process_logo(rport, fchs);
break;
case FC_ELS_ADISC:
bfa_fcs_rport_process_adisc(rport, fchs, len);
break;
case FC_ELS_PRLO:
if (bfa_fcs_port_is_initiator(port))
bfa_fcs_fcpim_uf_recv(rport->itnim, fchs, len);
if (bfa_fcs_port_is_target(port))
bfa_fcs_fcptm_uf_recv(rport->tin, fchs, len);
break;
case FC_ELS_PRLI:
bfa_fcs_rport_process_prli(rport, fchs, len);
break;
case FC_ELS_RPSC:
bfa_fcs_rport_process_rpsc(rport, fchs, len);
break;
default:
bfa_fcs_rport_send_ls_rjt(rport, fchs,
FC_LS_RJT_RSN_CMD_NOT_SUPP,
FC_LS_RJT_EXP_NO_ADDL_INFO);
break;
}
}
/* Send best case acc to prlo */
static void
bfa_fcs_rport_send_prlo_acc(struct bfa_fcs_rport_s *rport)
{
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
struct bfa_fcxp_s *fcxp;
int len;
bfa_trc(rport->fcs, rport->pid);
fcxp = bfa_fcs_fcxp_alloc(port->fcs);
if (!fcxp)
return;
len = fc_prlo_acc_build(&fchs, bfa_fcxp_get_reqbuf(fcxp),
rport->pid, bfa_fcs_port_get_fcid(port),
rport->reply_oxid, 0);
bfa_fcxp_send(fcxp, rport->bfa_rport, port->fabric->vf_id,
port->lp_tag, BFA_FALSE, FC_CLASS_3, len, &fchs,
NULL, NULL, FC_MAX_PDUSZ, 0);
}
/*
* Send a LS reject
*/
static void
bfa_fcs_rport_send_ls_rjt(struct bfa_fcs_rport_s *rport, struct fchs_s *rx_fchs,
u8 reason_code, u8 reason_code_expl)
{
struct bfa_fcs_port_s *port = rport->port;
struct fchs_s fchs;
struct bfa_fcxp_s *fcxp;
int len;
bfa_trc(rport->fcs, rx_fchs->s_id);
fcxp = bfa_fcs_fcxp_alloc(rport->fcs);
if (!fcxp)
return;
len = fc_ls_rjt_build(&fchs, bfa_fcxp_get_reqbuf(fcxp), rx_fchs->s_id,
bfa_fcs_port_get_fcid(port), rx_fchs->ox_id,
reason_code, reason_code_expl);
bfa_fcxp_send(fcxp, NULL, port->fabric->vf_id, port->lp_tag, BFA_FALSE,
FC_CLASS_3, len, &fchs, NULL, NULL, FC_MAX_PDUSZ, 0);
}
/**
* Return state of rport.
*/
int
bfa_fcs_rport_get_state(struct bfa_fcs_rport_s *rport)
{
return bfa_sm_to_state(rport_sm_table, rport->sm);
}
/**
* Called by the Driver to set rport delete/ageout timeout
*
* param[in] rport timeout value in seconds.
*
* return None
*/
void
bfa_fcs_rport_set_del_timeout(u8 rport_tmo)
{
/*
* convert to Millisecs
*/
if (rport_tmo > 0)
bfa_fcs_rport_del_timeout = rport_tmo * 1000;
}
void
bfa_fcs_rport_prlo(struct bfa_fcs_rport_s *rport, uint16_t ox_id)
{
bfa_trc(rport->fcs, rport->pid);
rport->prlo = BFA_TRUE;
rport->reply_oxid = ox_id;
bfa_sm_send_event(rport, RPSM_EVENT_PRLO_RCVD);
}