scsi: elx: libefc: Extended link Service I/O handling
authorJames Smart <jsmart2021@gmail.com>
Tue, 1 Jun 2021 23:54:56 +0000 (16:54 -0700)
committerMartin K. Petersen <martin.petersen@oracle.com>
Wed, 16 Jun 2021 03:39:30 +0000 (23:39 -0400)
Add library interface definitions for:

 - Functions to build and send ELS/CT/BLS commands and responses.

Link: https://lore.kernel.org/r/20210601235512.20104-16-jsmart2021@gmail.com
Reviewed-by: Daniel Wagner <dwagner@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Co-developed-by: Ram Vegesna <ram.vegesna@broadcom.com>
Signed-off-by: Ram Vegesna <ram.vegesna@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/elx/libefc/efc_els.c [new file with mode: 0644]
drivers/scsi/elx/libefc/efc_els.h [new file with mode: 0644]

diff --git a/drivers/scsi/elx/libefc/efc_els.c b/drivers/scsi/elx/libefc/efc_els.c
new file mode 100644 (file)
index 0000000..24db0ac
--- /dev/null
@@ -0,0 +1,1098 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021 Broadcom. All Rights Reserved. The term
+ * “Broadcom” refers to Broadcom Inc. and/or its subsidiaries.
+ */
+
+/*
+ * Functions to build and send ELS/CT/BLS commands and responses.
+ */
+
+#include "efc.h"
+#include "efc_els.h"
+#include "../libefc_sli/sli4.h"
+
+#define EFC_LOG_ENABLE_ELS_TRACE(efc)          \
+               (((efc) != NULL) ? (((efc)->logmask & (1U << 1)) != 0) : 0)
+
+#define node_els_trace()  \
+       do { \
+               if (EFC_LOG_ENABLE_ELS_TRACE(efc)) \
+                       efc_log_info(efc, "[%s] %-20s\n", \
+                               node->display_name, __func__); \
+       } while (0)
+
+#define els_io_printf(els, fmt, ...) \
+       efc_log_err((struct efc *)els->node->efc,\
+                     "[%s] %-8s " fmt, \
+                     els->node->display_name,\
+                     els->display_name, ##__VA_ARGS__)
+
+#define EFC_ELS_RSP_LEN                        1024
+#define EFC_ELS_GID_PT_RSP_LEN         8096
+
+struct efc_els_io_req *
+efc_els_io_alloc(struct efc_node *node, u32 reqlen)
+{
+       return efc_els_io_alloc_size(node, reqlen, EFC_ELS_RSP_LEN);
+}
+
+struct efc_els_io_req *
+efc_els_io_alloc_size(struct efc_node *node, u32 reqlen, u32 rsplen)
+{
+       struct efc *efc;
+       struct efc_els_io_req *els;
+       unsigned long flags = 0;
+
+       efc = node->efc;
+
+       spin_lock_irqsave(&node->els_ios_lock, flags);
+
+       if (!node->els_io_enabled) {
+               efc_log_err(efc, "els io alloc disabled\n");
+               spin_unlock_irqrestore(&node->els_ios_lock, flags);
+               return NULL;
+       }
+
+       els = mempool_alloc(efc->els_io_pool, GFP_ATOMIC);
+       if (!els) {
+               atomic_add_return(1, &efc->els_io_alloc_failed_count);
+               spin_unlock_irqrestore(&node->els_ios_lock, flags);
+               return NULL;
+       }
+
+       /* initialize refcount */
+       kref_init(&els->ref);
+       els->release = _efc_els_io_free;
+
+       /* populate generic io fields */
+       els->node = node;
+
+       /* now allocate DMA for request and response */
+       els->io.req.size = reqlen;
+       els->io.req.virt = dma_alloc_coherent(&efc->pci->dev, els->io.req.size,
+                                             &els->io.req.phys, GFP_DMA);
+       if (!els->io.req.virt) {
+               mempool_free(els, efc->els_io_pool);
+               spin_unlock_irqrestore(&node->els_ios_lock, flags);
+               return NULL;
+       }
+
+       els->io.rsp.size = rsplen;
+       els->io.rsp.virt = dma_alloc_coherent(&efc->pci->dev, els->io.rsp.size,
+                                             &els->io.rsp.phys, GFP_DMA);
+       if (!els->io.rsp.virt) {
+               dma_free_coherent(&efc->pci->dev, els->io.req.size,
+                                 els->io.req.virt, els->io.req.phys);
+               mempool_free(els, efc->els_io_pool);
+               els = NULL;
+       }
+
+       if (els) {
+               /* initialize fields */
+               els->els_retries_remaining = EFC_FC_ELS_DEFAULT_RETRIES;
+
+               /* add els structure to ELS IO list */
+               INIT_LIST_HEAD(&els->list_entry);
+               list_add_tail(&els->list_entry, &node->els_ios_list);
+       }
+
+       spin_unlock_irqrestore(&node->els_ios_lock, flags);
+       return els;
+}
+
+void
+efc_els_io_free(struct efc_els_io_req *els)
+{
+       kref_put(&els->ref, els->release);
+}
+
+void
+_efc_els_io_free(struct kref *arg)
+{
+       struct efc_els_io_req *els =
+                               container_of(arg, struct efc_els_io_req, ref);
+       struct efc *efc;
+       struct efc_node *node;
+       int send_empty_event = false;
+       unsigned long flags = 0;
+
+       node = els->node;
+       efc = node->efc;
+
+       spin_lock_irqsave(&node->els_ios_lock, flags);
+
+       list_del(&els->list_entry);
+       /* Send list empty event if the IO allocator
+        * is disabled, and the list is empty
+        * If node->els_io_enabled was not checked,
+        * the event would be posted continually
+        */
+       send_empty_event = (!node->els_io_enabled &&
+                          list_empty(&node->els_ios_list));
+
+       spin_unlock_irqrestore(&node->els_ios_lock, flags);
+
+       /* free ELS request and response buffers */
+       dma_free_coherent(&efc->pci->dev, els->io.rsp.size,
+                         els->io.rsp.virt, els->io.rsp.phys);
+       dma_free_coherent(&efc->pci->dev, els->io.req.size,
+                         els->io.req.virt, els->io.req.phys);
+
+       mempool_free(els, efc->els_io_pool);
+
+       if (send_empty_event)
+               efc_scsi_io_list_empty(node->efc, node);
+}
+
+static void
+efc_els_retry(struct efc_els_io_req *els);
+
+static void
+efc_els_delay_timer_cb(struct timer_list *t)
+{
+       struct efc_els_io_req *els = from_timer(els, t, delay_timer);
+
+       /* Retry delay timer expired, retry the ELS request */
+       efc_els_retry(els);
+}
+
+static int
+efc_els_req_cb(void *arg, u32 length, int status, u32 ext_status)
+{
+       struct efc_els_io_req *els;
+       struct efc_node *node;
+       struct efc *efc;
+       struct efc_node_cb cbdata;
+       u32 reason_code;
+
+       els = arg;
+       node = els->node;
+       efc = node->efc;
+
+       if (status)
+               els_io_printf(els, "status x%x ext x%x\n", status, ext_status);
+
+       /* set the response len element of els->rsp */
+       els->io.rsp.len = length;
+
+       cbdata.status = status;
+       cbdata.ext_status = ext_status;
+       cbdata.header = NULL;
+       cbdata.els_rsp = els->io.rsp;
+
+       /* set the response len element of els->rsp */
+       cbdata.rsp_len = length;
+
+       /* FW returns the number of bytes received on the link in
+        * the WCQE, not the amount placed in the buffer; use this info to
+        * check if there was an overrun.
+        */
+       if (length > els->io.rsp.size) {
+               efc_log_warn(efc,
+                            "ELS response returned len=%d > buflen=%zu\n",
+                            length, els->io.rsp.size);
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_FAIL, &cbdata);
+               return 0;
+       }
+
+       /* Post event to ELS IO object */
+       switch (status) {
+       case SLI4_FC_WCQE_STATUS_SUCCESS:
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_OK, &cbdata);
+               break;
+
+       case SLI4_FC_WCQE_STATUS_LS_RJT:
+               reason_code = (ext_status >> 16) & 0xff;
+
+               /* delay and retry if reason code is Logical Busy */
+               switch (reason_code) {
+               case ELS_RJT_BUSY:
+                       els->node->els_req_cnt--;
+                       els_io_printf(els,
+                                     "LS_RJT Logical Busy, delay and retry\n");
+                       timer_setup(&els->delay_timer,
+                                   efc_els_delay_timer_cb, 0);
+                       mod_timer(&els->delay_timer,
+                                 jiffies + msecs_to_jiffies(5000));
+                       break;
+               default:
+                       efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_RJT,
+                                          &cbdata);
+                       break;
+               }
+               break;
+
+       case SLI4_FC_WCQE_STATUS_LOCAL_REJECT:
+               switch (ext_status) {
+               case SLI4_FC_LOCAL_REJECT_SEQUENCE_TIMEOUT:
+                       efc_els_retry(els);
+                       break;
+               default:
+                       efc_log_err(efc, "LOCAL_REJECT with ext status:%x\n",
+                                   ext_status);
+                       efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_FAIL,
+                                          &cbdata);
+                       break;
+               }
+               break;
+       default:        /* Other error */
+               efc_log_warn(efc, "els req failed status x%x, ext_status x%x\n",
+                            status, ext_status);
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_FAIL, &cbdata);
+               break;
+       }
+
+       return 0;
+}
+
+void efc_disc_io_complete(struct efc_disc_io *io, u32 len, u32 status,
+                         u32 ext_status)
+{
+       struct efc_els_io_req *els =
+                               container_of(io, struct efc_els_io_req, io);
+
+       WARN_ON_ONCE(!els->cb);
+
+       ((efc_hw_srrs_cb_t)els->cb) (els, len, status, ext_status);
+}
+
+static int efc_els_send_req(struct efc_node *node, struct efc_els_io_req *els,
+                           enum efc_disc_io_type io_type)
+{
+       int rc = 0;
+       struct efc *efc = node->efc;
+       struct efc_node_cb cbdata;
+
+       /* update ELS request counter */
+       els->node->els_req_cnt++;
+
+       /* Prepare the IO request details */
+       els->io.io_type = io_type;
+       els->io.xmit_len = els->io.req.size;
+       els->io.rsp_len = els->io.rsp.size;
+       els->io.rpi = node->rnode.indicator;
+       els->io.vpi = node->nport->indicator;
+       els->io.s_id = node->nport->fc_id;
+       els->io.d_id = node->rnode.fc_id;
+
+       if (node->rnode.attached)
+               els->io.rpi_registered = true;
+
+       els->cb = efc_els_req_cb;
+
+       rc = efc->tt.send_els(efc, &els->io);
+       if (!rc)
+               return rc;
+
+       cbdata.status = EFC_STATUS_INVALID;
+       cbdata.ext_status = EFC_STATUS_INVALID;
+       cbdata.els_rsp = els->io.rsp;
+       efc_log_err(efc, "efc_els_send failed: %d\n", rc);
+       efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_FAIL, &cbdata);
+
+       return rc;
+}
+
+static void
+efc_els_retry(struct efc_els_io_req *els)
+{
+       struct efc *efc;
+       struct efc_node_cb cbdata;
+       u32 rc;
+
+       efc = els->node->efc;
+       cbdata.status = EFC_STATUS_INVALID;
+       cbdata.ext_status = EFC_STATUS_INVALID;
+       cbdata.els_rsp = els->io.rsp;
+
+       if (els->els_retries_remaining) {
+               els->els_retries_remaining--;
+               rc = efc->tt.send_els(efc, &els->io);
+       } else {
+               rc = -EIO;
+       }
+
+       if (rc) {
+               efc_log_err(efc, "ELS retries exhausted\n");
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_REQ_FAIL, &cbdata);
+       }
+}
+
+static int
+efc_els_acc_cb(void *arg, u32 length, int status, u32 ext_status)
+{
+       struct efc_els_io_req *els;
+       struct efc_node *node;
+       struct efc *efc;
+       struct efc_node_cb cbdata;
+
+       els = arg;
+       node = els->node;
+       efc = node->efc;
+
+       cbdata.status = status;
+       cbdata.ext_status = ext_status;
+       cbdata.header = NULL;
+       cbdata.els_rsp = els->io.rsp;
+
+       /* Post node event */
+       switch (status) {
+       case SLI4_FC_WCQE_STATUS_SUCCESS:
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_CMPL_OK, &cbdata);
+               break;
+
+       default:        /* Other error */
+               efc_log_warn(efc, "[%s] %-8s failed status x%x, ext x%x\n",
+                            node->display_name, els->display_name,
+                            status, ext_status);
+               efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_CMPL_FAIL, &cbdata);
+               break;
+       }
+
+       return 0;
+}
+
+static int
+efc_els_send_rsp(struct efc_els_io_req *els, u32 rsplen)
+{
+       int rc = 0;
+       struct efc_node_cb cbdata;
+       struct efc_node *node = els->node;
+       struct efc *efc = node->efc;
+
+       /* increment ELS completion counter */
+       node->els_cmpl_cnt++;
+
+       els->io.io_type = EFC_DISC_IO_ELS_RESP;
+       els->cb = efc_els_acc_cb;
+
+       /* Prepare the IO request details */
+       els->io.xmit_len = rsplen;
+       els->io.rsp_len = els->io.rsp.size;
+       els->io.rpi = node->rnode.indicator;
+       els->io.vpi = node->nport->indicator;
+       if (node->nport->fc_id != U32_MAX)
+               els->io.s_id = node->nport->fc_id;
+       else
+               els->io.s_id = els->io.iparam.els.s_id;
+       els->io.d_id = node->rnode.fc_id;
+
+       if (node->attached)
+               els->io.rpi_registered = true;
+
+       rc = efc->tt.send_els(efc, &els->io);
+       if (!rc)
+               return rc;
+
+       cbdata.status = EFC_STATUS_INVALID;
+       cbdata.ext_status = EFC_STATUS_INVALID;
+       cbdata.els_rsp = els->io.rsp;
+       efc_els_io_cleanup(els, EFC_EVT_SRRS_ELS_CMPL_FAIL, &cbdata);
+
+       return rc;
+}
+
+int
+efc_send_plogi(struct efc_node *node)
+{
+       struct efc_els_io_req *els;
+       struct efc *efc = node->efc;
+       struct fc_els_flogi  *plogi;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*plogi));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+       els->display_name = "plogi";
+
+       /* Build PLOGI request */
+       plogi = els->io.req.virt;
+
+       memcpy(plogi, node->nport->service_params, sizeof(*plogi));
+
+       plogi->fl_cmd = ELS_PLOGI;
+       memset(plogi->_fl_resvd, 0, sizeof(plogi->_fl_resvd));
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_flogi(struct efc_node *node)
+{
+       struct efc_els_io_req *els;
+       struct efc *efc;
+       struct fc_els_flogi  *flogi;
+
+       efc = node->efc;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*flogi));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "flogi";
+
+       /* Build FLOGI request */
+       flogi = els->io.req.virt;
+
+       memcpy(flogi, node->nport->service_params, sizeof(*flogi));
+       flogi->fl_cmd = ELS_FLOGI;
+       memset(flogi->_fl_resvd, 0, sizeof(flogi->_fl_resvd));
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_fdisc(struct efc_node *node)
+{
+       struct efc_els_io_req *els;
+       struct efc *efc;
+       struct fc_els_flogi *fdisc;
+
+       efc = node->efc;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*fdisc));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "fdisc";
+
+       /* Build FDISC request */
+       fdisc = els->io.req.virt;
+
+       memcpy(fdisc, node->nport->service_params, sizeof(*fdisc));
+       fdisc->fl_cmd = ELS_FDISC;
+       memset(fdisc->_fl_resvd, 0, sizeof(fdisc->_fl_resvd));
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_prli(struct efc_node *node)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els;
+       struct {
+               struct fc_els_prli prli;
+               struct fc_els_spp spp;
+       } *pp;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*pp));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "prli";
+
+       /* Build PRLI request */
+       pp = els->io.req.virt;
+
+       memset(pp, 0, sizeof(*pp));
+
+       pp->prli.prli_cmd = ELS_PRLI;
+       pp->prli.prli_spp_len = 16;
+       pp->prli.prli_len = cpu_to_be16(sizeof(*pp));
+       pp->spp.spp_type = FC_TYPE_FCP;
+       pp->spp.spp_type_ext = 0;
+       pp->spp.spp_flags = FC_SPP_EST_IMG_PAIR;
+       pp->spp.spp_params = cpu_to_be32(FCP_SPPF_RD_XRDY_DIS |
+                              (node->nport->enable_ini ?
+                              FCP_SPPF_INIT_FCN : 0) |
+                              (node->nport->enable_tgt ?
+                              FCP_SPPF_TARG_FCN : 0));
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_logo(struct efc_node *node)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els;
+       struct fc_els_logo *logo;
+       struct fc_els_flogi  *sparams;
+
+       node_els_trace();
+
+       sparams = (struct fc_els_flogi *)node->nport->service_params;
+
+       els = efc_els_io_alloc(node, sizeof(*logo));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "logo";
+
+       /* Build LOGO request */
+
+       logo = els->io.req.virt;
+
+       memset(logo, 0, sizeof(*logo));
+       logo->fl_cmd = ELS_LOGO;
+       hton24(logo->fl_n_port_id, node->rnode.nport->fc_id);
+       logo->fl_n_port_wwn = sparams->fl_wwpn;
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_adisc(struct efc_node *node)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els;
+       struct fc_els_adisc *adisc;
+       struct fc_els_flogi  *sparams;
+       struct efc_nport *nport = node->nport;
+
+       node_els_trace();
+
+       sparams = (struct fc_els_flogi *)node->nport->service_params;
+
+       els = efc_els_io_alloc(node, sizeof(*adisc));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "adisc";
+
+       /* Build ADISC request */
+
+       adisc = els->io.req.virt;
+
+       memset(adisc, 0, sizeof(*adisc));
+       adisc->adisc_cmd = ELS_ADISC;
+       hton24(adisc->adisc_hard_addr, nport->fc_id);
+       adisc->adisc_wwpn = sparams->fl_wwpn;
+       adisc->adisc_wwnn = sparams->fl_wwnn;
+       hton24(adisc->adisc_port_id, node->rnode.nport->fc_id);
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_scr(struct efc_node *node)
+{
+       struct efc_els_io_req *els;
+       struct efc *efc = node->efc;
+       struct fc_els_scr *req;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*req));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "scr";
+
+       req = els->io.req.virt;
+
+       memset(req, 0, sizeof(*req));
+       req->scr_cmd = ELS_SCR;
+       req->scr_reg_func = ELS_SCRF_FULL;
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_ELS_REQ);
+}
+
+int
+efc_send_ls_rjt(struct efc_node *node, u32 ox_id, u32 reason_code,
+               u32 reason_code_expl, u32 vendor_unique)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct fc_els_ls_rjt *rjt;
+
+       els = efc_els_io_alloc(node, sizeof(*rjt));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       node_els_trace();
+
+       els->display_name = "ls_rjt";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       rjt = els->io.req.virt;
+       memset(rjt, 0, sizeof(*rjt));
+
+       rjt->er_cmd = ELS_LS_RJT;
+       rjt->er_reason = reason_code;
+       rjt->er_explan = reason_code_expl;
+
+       return efc_els_send_rsp(els, sizeof(*rjt));
+}
+
+int
+efc_send_plogi_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct fc_els_flogi  *plogi;
+       struct fc_els_flogi  *req = (struct fc_els_flogi *)node->service_params;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*plogi));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "plogi_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       plogi = els->io.req.virt;
+
+       /* copy our port's service parameters to payload */
+       memcpy(plogi, node->nport->service_params, sizeof(*plogi));
+       plogi->fl_cmd = ELS_LS_ACC;
+       memset(plogi->_fl_resvd, 0, sizeof(plogi->_fl_resvd));
+
+       /* Set Application header support bit if requested */
+       if (req->fl_csp.sp_features & cpu_to_be16(FC_SP_FT_BCAST))
+               plogi->fl_csp.sp_features |= cpu_to_be16(FC_SP_FT_BCAST);
+
+       return efc_els_send_rsp(els, sizeof(*plogi));
+}
+
+int
+efc_send_flogi_p2p_acc(struct efc_node *node, u32 ox_id, u32 s_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct fc_els_flogi  *flogi;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*flogi));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "flogi_p2p_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+       els->io.iparam.els.s_id = s_id;
+
+       flogi = els->io.req.virt;
+
+       /* copy our port's service parameters to payload */
+       memcpy(flogi, node->nport->service_params, sizeof(*flogi));
+       flogi->fl_cmd = ELS_LS_ACC;
+       memset(flogi->_fl_resvd, 0, sizeof(flogi->_fl_resvd));
+
+       memset(flogi->fl_cssp, 0, sizeof(flogi->fl_cssp));
+
+       return efc_els_send_rsp(els, sizeof(*flogi));
+}
+
+int
+efc_send_prli_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct {
+               struct fc_els_prli prli;
+               struct fc_els_spp spp;
+       } *pp;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*pp));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "prli_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       pp = els->io.req.virt;
+       memset(pp, 0, sizeof(*pp));
+
+       pp->prli.prli_cmd = ELS_LS_ACC;
+       pp->prli.prli_spp_len = 0x10;
+       pp->prli.prli_len = cpu_to_be16(sizeof(*pp));
+       pp->spp.spp_type = FC_TYPE_FCP;
+       pp->spp.spp_type_ext = 0;
+       pp->spp.spp_flags = FC_SPP_EST_IMG_PAIR | FC_SPP_RESP_ACK;
+
+       pp->spp.spp_params = cpu_to_be32(FCP_SPPF_RD_XRDY_DIS |
+                                       (node->nport->enable_ini ?
+                                        FCP_SPPF_INIT_FCN : 0) |
+                                       (node->nport->enable_tgt ?
+                                        FCP_SPPF_TARG_FCN : 0));
+
+       return efc_els_send_rsp(els, sizeof(*pp));
+}
+
+int
+efc_send_prlo_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct {
+               struct fc_els_prlo prlo;
+               struct fc_els_spp spp;
+       } *pp;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*pp));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "prlo_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       pp = els->io.req.virt;
+       memset(pp, 0, sizeof(*pp));
+       pp->prlo.prlo_cmd = ELS_LS_ACC;
+       pp->prlo.prlo_obs = 0x10;
+       pp->prlo.prlo_len = cpu_to_be16(sizeof(*pp));
+
+       pp->spp.spp_type = FC_TYPE_FCP;
+       pp->spp.spp_type_ext = 0;
+       pp->spp.spp_flags = FC_SPP_RESP_ACK;
+
+       return efc_els_send_rsp(els, sizeof(*pp));
+}
+
+int
+efc_send_ls_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct fc_els_ls_acc *acc;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*acc));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "ls_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       acc = els->io.req.virt;
+       memset(acc, 0, sizeof(*acc));
+
+       acc->la_cmd = ELS_LS_ACC;
+
+       return efc_els_send_rsp(els, sizeof(*acc));
+}
+
+int
+efc_send_logo_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc_els_io_req *els = NULL;
+       struct efc *efc = node->efc;
+       struct fc_els_ls_acc *logo;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*logo));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "logo_acc";
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       logo = els->io.req.virt;
+       memset(logo, 0, sizeof(*logo));
+
+       logo->la_cmd = ELS_LS_ACC;
+
+       return efc_els_send_rsp(els, sizeof(*logo));
+}
+
+int
+efc_send_adisc_acc(struct efc_node *node, u32 ox_id)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els = NULL;
+       struct fc_els_adisc *adisc;
+       struct fc_els_flogi  *sparams;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*adisc));
+       if (!els) {
+               efc_log_err(efc, "els IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->display_name = "adisc_acc";
+
+       /* Go ahead and send the ELS_ACC */
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+       els->io.iparam.els.ox_id = ox_id;
+
+       sparams = (struct fc_els_flogi  *)node->nport->service_params;
+       adisc = els->io.req.virt;
+       memset(adisc, 0, sizeof(*adisc));
+       adisc->adisc_cmd = ELS_LS_ACC;
+       adisc->adisc_wwpn = sparams->fl_wwpn;
+       adisc->adisc_wwnn = sparams->fl_wwnn;
+       hton24(adisc->adisc_port_id, node->rnode.nport->fc_id);
+
+       return efc_els_send_rsp(els, sizeof(*adisc));
+}
+
+static inline void
+fcct_build_req_header(struct fc_ct_hdr  *hdr, u16 cmd, u16 max_size)
+{
+       hdr->ct_rev = FC_CT_REV;
+       hdr->ct_fs_type = FC_FST_DIR;
+       hdr->ct_fs_subtype = FC_NS_SUBTYPE;
+       hdr->ct_options = 0;
+       hdr->ct_cmd = cpu_to_be16(cmd);
+       /* words */
+       hdr->ct_mr_size = cpu_to_be16(max_size / (sizeof(u32)));
+       hdr->ct_reason = 0;
+       hdr->ct_explan = 0;
+       hdr->ct_vendor = 0;
+}
+
+int
+efc_ns_send_rftid(struct efc_node *node)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els;
+       struct {
+               struct fc_ct_hdr hdr;
+               struct fc_ns_rft_id rftid;
+       } *ct;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*ct));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->io.iparam.ct.r_ctl = FC_RCTL_ELS_REQ;
+       els->io.iparam.ct.type = FC_TYPE_CT;
+       els->io.iparam.ct.df_ctl = 0;
+       els->io.iparam.ct.timeout = EFC_FC_ELS_SEND_DEFAULT_TIMEOUT;
+
+       els->display_name = "rftid";
+
+       ct = els->io.req.virt;
+       memset(ct, 0, sizeof(*ct));
+       fcct_build_req_header(&ct->hdr, FC_NS_RFT_ID,
+                             sizeof(struct fc_ns_rft_id));
+
+       hton24(ct->rftid.fr_fid.fp_fid, node->rnode.nport->fc_id);
+       ct->rftid.fr_fts.ff_type_map[FC_TYPE_FCP / FC_NS_BPW] =
+               cpu_to_be32(1 << (FC_TYPE_FCP % FC_NS_BPW));
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_CT_REQ);
+}
+
+int
+efc_ns_send_rffid(struct efc_node *node)
+{
+       struct efc *efc = node->efc;
+       struct efc_els_io_req *els;
+       struct {
+               struct fc_ct_hdr hdr;
+               struct fc_ns_rff_id rffid;
+       } *ct;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc(node, sizeof(*ct));
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->io.iparam.ct.r_ctl = FC_RCTL_ELS_REQ;
+       els->io.iparam.ct.type = FC_TYPE_CT;
+       els->io.iparam.ct.df_ctl = 0;
+       els->io.iparam.ct.timeout = EFC_FC_ELS_SEND_DEFAULT_TIMEOUT;
+
+       els->display_name = "rffid";
+       ct = els->io.req.virt;
+
+       memset(ct, 0, sizeof(*ct));
+       fcct_build_req_header(&ct->hdr, FC_NS_RFF_ID,
+                             sizeof(struct fc_ns_rff_id));
+
+       hton24(ct->rffid.fr_fid.fp_fid, node->rnode.nport->fc_id);
+       if (node->nport->enable_ini)
+               ct->rffid.fr_feat |= FCP_FEAT_INIT;
+       if (node->nport->enable_tgt)
+               ct->rffid.fr_feat |= FCP_FEAT_TARG;
+       ct->rffid.fr_type = FC_TYPE_FCP;
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_CT_REQ);
+}
+
+int
+efc_ns_send_gidpt(struct efc_node *node)
+{
+       struct efc_els_io_req *els = NULL;
+       struct efc *efc = node->efc;
+       struct {
+               struct fc_ct_hdr hdr;
+               struct fc_ns_gid_pt gidpt;
+       } *ct;
+
+       node_els_trace();
+
+       els = efc_els_io_alloc_size(node, sizeof(*ct), EFC_ELS_GID_PT_RSP_LEN);
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       els->io.iparam.ct.r_ctl = FC_RCTL_ELS_REQ;
+       els->io.iparam.ct.type = FC_TYPE_CT;
+       els->io.iparam.ct.df_ctl = 0;
+       els->io.iparam.ct.timeout = EFC_FC_ELS_SEND_DEFAULT_TIMEOUT;
+
+       els->display_name = "gidpt";
+
+       ct = els->io.req.virt;
+
+       memset(ct, 0, sizeof(*ct));
+       fcct_build_req_header(&ct->hdr, FC_NS_GID_PT,
+                             sizeof(struct fc_ns_gid_pt));
+
+       ct->gidpt.fn_pt_type = FC_TYPE_FCP;
+
+       return efc_els_send_req(node, els, EFC_DISC_IO_CT_REQ);
+}
+
+void
+efc_els_io_cleanup(struct efc_els_io_req *els, int evt, void *arg)
+{
+       /* don't want further events that could come; e.g. abort requests
+        * from the node state machine; thus, disable state machine
+        */
+       els->els_req_free = true;
+       efc_node_post_els_resp(els->node, evt, arg);
+
+       efc_els_io_free(els);
+}
+
+static int
+efc_ct_acc_cb(void *arg, u32 length, int status, u32 ext_status)
+{
+       struct efc_els_io_req *els = arg;
+
+       efc_els_io_free(els);
+
+       return 0;
+}
+
+int
+efc_send_ct_rsp(struct efc *efc, struct efc_node *node, u16 ox_id,
+               struct fc_ct_hdr *ct_hdr, u32 cmd_rsp_code,
+               u32 reason_code, u32 reason_code_explanation)
+{
+       struct efc_els_io_req *els = NULL;
+       struct fc_ct_hdr  *rsp = NULL;
+
+       els = efc_els_io_alloc(node, 256);
+       if (!els) {
+               efc_log_err(efc, "IO alloc failed\n");
+               return -EIO;
+       }
+
+       rsp = els->io.rsp.virt;
+
+       *rsp = *ct_hdr;
+
+       fcct_build_req_header(rsp, cmd_rsp_code, 0);
+       rsp->ct_reason = reason_code;
+       rsp->ct_explan = reason_code_explanation;
+
+       els->display_name = "ct_rsp";
+       els->cb = efc_ct_acc_cb;
+
+       /* Prepare the IO request details */
+       els->io.io_type = EFC_DISC_IO_CT_RESP;
+       els->io.xmit_len = sizeof(*rsp);
+
+       els->io.rpi = node->rnode.indicator;
+       els->io.d_id = node->rnode.fc_id;
+
+       memset(&els->io.iparam, 0, sizeof(els->io.iparam));
+
+       els->io.iparam.ct.ox_id = ox_id;
+       els->io.iparam.ct.r_ctl = 3;
+       els->io.iparam.ct.type = FC_TYPE_CT;
+       els->io.iparam.ct.df_ctl = 0;
+       els->io.iparam.ct.timeout = 5;
+
+       if (efc->tt.send_els(efc, &els->io)) {
+               efc_els_io_free(els);
+               return -EIO;
+       }
+       return 0;
+}
+
+int
+efc_send_bls_acc(struct efc_node *node, struct fc_frame_header *hdr)
+{
+       struct sli_bls_params bls;
+       struct fc_ba_acc *acc;
+       struct efc *efc = node->efc;
+
+       memset(&bls, 0, sizeof(bls));
+       bls.ox_id = be16_to_cpu(hdr->fh_ox_id);
+       bls.rx_id = be16_to_cpu(hdr->fh_rx_id);
+       bls.s_id = ntoh24(hdr->fh_d_id);
+       bls.d_id = node->rnode.fc_id;
+       bls.rpi = node->rnode.indicator;
+       bls.vpi = node->nport->indicator;
+
+       acc = (void *)bls.payload;
+       acc->ba_ox_id = cpu_to_be16(bls.ox_id);
+       acc->ba_rx_id = cpu_to_be16(bls.rx_id);
+       acc->ba_high_seq_cnt = cpu_to_be16(U16_MAX);
+
+       return efc->tt.send_bls(efc, FC_RCTL_BA_ACC, &bls);
+}
diff --git a/drivers/scsi/elx/libefc/efc_els.h b/drivers/scsi/elx/libefc/efc_els.h
new file mode 100644 (file)
index 0000000..3c4f820
--- /dev/null
@@ -0,0 +1,107 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2021 Broadcom. All Rights Reserved. The term
+ * “Broadcom” refers to Broadcom Inc. and/or its subsidiaries.
+ */
+
+#ifndef __EFC_ELS_H__
+#define __EFC_ELS_H__
+
+#define EFC_STATUS_INVALID     INT_MAX
+#define EFC_ELS_IO_POOL_SZ     1024
+
+struct efc_els_io_req {
+       struct list_head        list_entry;
+       struct kref             ref;
+       void                    (*release)(struct kref *arg);
+       struct efc_node         *node;
+       void                    *cb;
+       u32                     els_retries_remaining;
+       bool                    els_req_free;
+       struct timer_list       delay_timer;
+
+       const char              *display_name;
+
+       struct efc_disc_io      io;
+};
+
+typedef int(*efc_hw_srrs_cb_t)(void *arg, u32 length, int status,
+                              u32 ext_status);
+
+void _efc_els_io_free(struct kref *arg);
+struct efc_els_io_req *
+efc_els_io_alloc(struct efc_node *node, u32 reqlen);
+struct efc_els_io_req *
+efc_els_io_alloc_size(struct efc_node *node, u32 reqlen, u32 rsplen);
+void efc_els_io_free(struct efc_els_io_req *els);
+
+/* ELS command send */
+typedef void (*els_cb_t)(struct efc_node *node,
+                        struct efc_node_cb *cbdata, void *arg);
+int
+efc_send_plogi(struct efc_node *node);
+int
+efc_send_flogi(struct efc_node *node);
+int
+efc_send_fdisc(struct efc_node *node);
+int
+efc_send_prli(struct efc_node *node);
+int
+efc_send_prlo(struct efc_node *node);
+int
+efc_send_logo(struct efc_node *node);
+int
+efc_send_adisc(struct efc_node *node);
+int
+efc_send_pdisc(struct efc_node *node);
+int
+efc_send_scr(struct efc_node *node);
+int
+efc_ns_send_rftid(struct efc_node *node);
+int
+efc_ns_send_rffid(struct efc_node *node);
+int
+efc_ns_send_gidpt(struct efc_node *node);
+void
+efc_els_io_cleanup(struct efc_els_io_req *els, int evt, void *arg);
+
+/* ELS acc send */
+int
+efc_send_ls_acc(struct efc_node *node, u32 ox_id);
+int
+efc_send_ls_rjt(struct efc_node *node, u32 ox_id, u32 reason_cod,
+               u32 reason_code_expl, u32 vendor_unique);
+int
+efc_send_flogi_p2p_acc(struct efc_node *node, u32 ox_id, u32 s_id);
+int
+efc_send_flogi_acc(struct efc_node *node, u32 ox_id, u32 is_fport);
+int
+efc_send_plogi_acc(struct efc_node *node, u32 ox_id);
+int
+efc_send_prli_acc(struct efc_node *node, u32 ox_id);
+int
+efc_send_logo_acc(struct efc_node *node, u32 ox_id);
+int
+efc_send_prlo_acc(struct efc_node *node, u32 ox_id);
+int
+efc_send_adisc_acc(struct efc_node *node, u32 ox_id);
+
+int
+efc_bls_send_acc_hdr(struct efc *efc, struct efc_node *node,
+                    struct fc_frame_header *hdr);
+int
+efc_bls_send_rjt_hdr(struct efc_els_io_req *io, struct fc_frame_header *hdr);
+
+int
+efc_els_io_list_empty(struct efc_node *node, struct list_head *list);
+
+/* CT */
+int
+efc_send_ct_rsp(struct efc *efc, struct efc_node *node, u16 ox_id,
+               struct fc_ct_hdr *ct_hdr, u32 cmd_rsp_code, u32 reason_code,
+               u32 reason_code_explanation);
+
+int
+efc_send_bls_acc(struct efc_node *node, struct fc_frame_header *hdr);
+
+#endif /* __EFC_ELS_H__ */