@@ -16,4 +16,5 @@ fnic-y := \
vnic_intr.o \
vnic_rq.o \
vnic_wq_copy.o \
- vnic_wq.o
+ vnic_wq.o \
+ fnic_pci_subsys_devid.o
@@ -9,11 +9,23 @@
#include "fdls_fc.h"
#include "fnic_fdls.h"
#include <scsi/fc/fc_fcp.h>
+#include <scsi/scsi_transport_fc.h>
#include <linux/utsname.h>
#define FC_FC4_TYPE_SCSI 0x08
+#define PORT_SPEED_BIT_8 8
+#define PORT_SPEED_BIT_9 9
+#define PORT_SPEED_BIT_14 14
+#define PORT_SPEED_BIT_15 15
static void fdls_send_rpn_id(struct fnic_iport_s *iport);
+static void fdls_fdmi_register_hba(struct fnic_iport_s *iport);
+static void fdls_fdmi_register_pa(struct fnic_iport_s *iport);
+#define FDLS_FDMI_PLOGI_PENDING 0x1
+#define FDLS_FDMI_REG_HBA_PENDING 0x2
+#define FDLS_FDMI_RPA_PENDING 0x4
+#define FDLS_FDMI_ABORT_PENDING 0x8
+#define FDLS_FDMI_MAX_RETRY 3
/* Frame initialization */
/*
@@ -84,6 +96,70 @@ struct fc_std_els_prli fnic_std_prli_req = {
.spp_params = cpu_to_be32(0xA2)}
};
+/*
+ * Variables:
+ * sid, port_id, port_name
+ */
+struct fc_std_fdmi_rhba fnic_std_fdmi_rhba = {
+ .fchdr = {.fh_r_ctl = FC_RCTL_DD_UNSOL_CTL,
+ .fh_d_id = {0xFF, 0XFF, 0XFA},
+ .fh_type = FC_TYPE_CT, .fh_f_ctl = {FNIC_ELS_REQ_FCTL, 0, 0},
+ .fh_rx_id = 0xFFFF},
+ .fc_std_ct_hdr = {.ct_rev = FC_CT_REV, .ct_fs_type = FC_FST_MGMT,
+ .ct_fs_subtype = FC_FDMI_SUBTYPE,
+ .ct_cmd = cpu_to_be16(FC_FDMI_RHBA)},
+ .num_ports = FNIC_FDMI_NUM_PORTS,
+ .num_hba_attributes = FNIC_FDMI_NUM_HBA_ATTRS,
+ .type_nn = FNIC_FDMI_TYPE_NODE_NAME,
+ .length_nn = FNIC_FDMI_NN_LEN,
+ .type_manu = FNIC_FDMI_TYPE_MANUFACTURER,
+ .length_manu = FNIC_FDMI_MANU_LEN,
+ .manufacturer = FNIC_FDMI_MANUFACTURER,
+ .type_serial = FNIC_FDMI_TYPE_SERIAL_NUMBER,
+ .length_serial = FNIC_FDMI_SERIAL_LEN,
+ .type_model = FNIC_FDMI_TYPE_MODEL,
+ .length_model = FNIC_FDMI_MODEL_LEN,
+ .type_model_des = FNIC_FDMI_TYPE_MODEL_DES,
+ .length_model_des = FNIC_FDMI_MODEL_DES_LEN,
+ .model_description = FNIC_FDMI_MODEL_DESCRIPTION,
+ .type_hw_ver = FNIC_FDMI_TYPE_HARDWARE_VERSION,
+ .length_hw_ver = FNIC_FDMI_HW_VER_LEN,
+ .type_dr_ver = FNIC_FDMI_TYPE_DRIVER_VERSION,
+ .length_dr_ver = FNIC_FDMI_DR_VER_LEN,
+ .type_rom_ver = FNIC_FDMI_TYPE_ROM_VERSION,
+ .length_rom_ver = FNIC_FDMI_ROM_VER_LEN,
+ .type_fw_ver = FNIC_FDMI_TYPE_FIRMWARE_VERSION,
+ .length_fw_ver = FNIC_FDMI_FW_VER_LEN,
+};
+
+/*
+ * Variables
+ *sid, port_id, port_name
+ */
+struct fc_std_fdmi_rpa fnic_std_fdmi_rpa = {
+ .fchdr = {.fh_r_ctl = FC_RCTL_DD_UNSOL_CTL,
+ .fh_d_id = {0xFF, 0xFF, 0xFA},
+ .fh_type = FC_TYPE_CT, .fh_f_ctl = {FNIC_ELS_REQ_FCTL, 0, 0},
+ .fh_rx_id = 0xFFFF},
+ .fc_std_ct_hdr = {.ct_rev = FC_CT_REV, .ct_fs_type = FC_FST_MGMT,
+ .ct_fs_subtype = FC_FDMI_SUBTYPE,
+ .ct_cmd = cpu_to_be16(FC_FDMI_RPA)},
+ .num_port_attributes = FNIC_FDMI_NUM_PORT_ATTRS,
+ .type_fc4 = FNIC_FDMI_TYPE_FC4_TYPES,
+ .length_fc4 = FNIC_FDMI_FC4_LEN,
+ .type_supp_speed = FNIC_FDMI_TYPE_SUPPORTED_SPEEDS,
+ .length_supp_speed = FNIC_FDMI_SUPP_SPEED_LEN,
+ .type_cur_speed = FNIC_FDMI_TYPE_CURRENT_SPEED,
+ .length_cur_speed = FNIC_FDMI_CUR_SPEED_LEN,
+ .type_max_frame_size = FNIC_FDMI_TYPE_MAX_FRAME_SIZE,
+ .length_max_frame_size = FNIC_FDMI_MFS_LEN,
+ .max_frame_size = FNIC_FDMI_MFS,
+ .type_os_name = FNIC_FDMI_TYPE_OS_NAME,
+ .length_os_name = FNIC_FDMI_OS_NAME_LEN,
+ .type_host_name = FNIC_FDMI_TYPE_HOST_NAME,
+ .length_host_name = FNIC_FDMI_HN_LEN,
+};
+
/*
* Variables:
* fh_s_id, port_id, port_name
@@ -223,6 +299,7 @@ static void fdls_target_restart_nexus(struct fnic_tport_s *tport);
static void fdls_start_tport_timer(struct fnic_iport_s *iport,
struct fnic_tport_s *tport, int timeout);
static void fdls_tport_timer_callback(struct timer_list *t);
+static void fdls_send_fdmi_plogi(struct fnic_iport_s *iport);
static void fdls_start_fabric_timer(struct fnic_iport_s *iport,
int timeout);
static void
@@ -336,6 +413,15 @@ uint16_t fdls_alloc_fabric_oxid(struct fnic_iport_s *iport,
* separately.
*/
switch (exp_rsp_type) {
+ case FNIC_FDMI_PLOGI_RSP:
+ oxid_pool->active_oxid_fdmi_plogi = oxid;
+ break;
+ case FNIC_FDMI_REG_HBA_RSP:
+ oxid_pool->active_oxid_fdmi_rhba = oxid;
+ break;
+ case FNIC_FDMI_RPA_RSP:
+ oxid_pool->active_oxid_fdmi_rpa = oxid;
+ break;
default:
oxid_pool->active_oxid_fabric_req = oxid;
break;
@@ -371,6 +457,21 @@ static inline void fdls_schedule_fabric_oxid_free(struct fnic_iport_s
iport->fabric_oxid_pool.active_oxid_fabric_req);
}
+static inline void fdls_schedule_fdmi_oxid_free(struct fnic_iport_s *iport)
+{
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_PLOGI_PENDING)
+ fdls_schedule_oxid_free(&iport->fdmi_oxid_pool.meta,
+ iport->fdmi_oxid_pool.active_oxid_fdmi_plogi);
+
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_REG_HBA_PENDING)
+ fdls_schedule_oxid_free(&iport->fdmi_oxid_pool.meta,
+ iport->fdmi_oxid_pool.active_oxid_fdmi_rhba);
+
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_RPA_PENDING)
+ fdls_schedule_oxid_free(&iport->fdmi_oxid_pool.meta,
+ iport->fdmi_oxid_pool.active_oxid_fdmi_rpa);
+}
+
static inline void fdls_schedule_tgt_oxid_free(struct fnic_iport_s *iport,
struct fnic_tgt_oxid_pool_s
*oxid_pool, uint16_t oxid)
@@ -405,6 +506,14 @@ static int fdls_is_oxid_in_fabric_range(uint16_t oxid)
(oxid_unmasked <= FDLS_FABRIC_OXID_POOL_END));
}
+static int fdls_is_oxid_in_fdmi_range(uint16_t oxid)
+{
+ uint16_t oxid_unmasked = FDLS_OXID_RSP_TYPE_UNMASKED(oxid);
+
+ return ((oxid_unmasked >= FDLS_FDMI_OXID_POOL_BASE) &&
+ (oxid_unmasked <= FDLS_FDMI_OXID_POOL_END));
+}
+
void fdls_init_tgt_oxid_pool(struct fnic_iport_s *iport)
{
memset(&iport->plogi_oxid_pool, 0, sizeof(iport->plogi_oxid_pool));
@@ -670,6 +779,42 @@ static void fdls_send_fabric_abts(struct fnic_iport_s *iport)
iport->fabric.timer_pending = 1;
}
+static void fdls_send_fdmi_abts(struct fnic_iport_s *iport)
+{
+ uint8_t fcid[3];
+ struct fc_frame_header fabric_abort = fc_std_fabric_abts;
+ struct fc_frame_header *fabric_abts = &fabric_abort;
+ struct fnic_fabric_oxid_pool_s *oxid_pool = &iport->fdmi_oxid_pool;
+ int fdmi_tov;
+ uint16_t oxid;
+
+ hton24(fcid, 0XFFFFFA);
+
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_PLOGI_PENDING) {
+ oxid = htons(oxid_pool->active_oxid_fdmi_plogi);
+ FNIC_STD_SET_OX_ID(fabric_abts, oxid);
+ fnic_send_fcoe_frame(iport, fabric_abts,
+ sizeof(struct fc_frame_header));
+ } else {
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_REG_HBA_PENDING) {
+ oxid = htons(oxid_pool->active_oxid_fdmi_rhba);
+ FNIC_STD_SET_OX_ID(fabric_abts, oxid);
+ fnic_send_fcoe_frame(iport, fabric_abts,
+ sizeof(struct fc_frame_header));
+ }
+ if (iport->fabric.fdmi_pending & FDLS_FDMI_RPA_PENDING) {
+ oxid = htons(oxid_pool->active_oxid_fdmi_rpa);
+ FNIC_STD_SET_OX_ID(fabric_abts, oxid);
+ fnic_send_fcoe_frame(iport, fabric_abts,
+ sizeof(struct fc_frame_header));
+ }
+ }
+
+ fdmi_tov = jiffies + msecs_to_jiffies(2 * iport->e_d_tov);
+ mod_timer(&iport->fabric.fdmi_timer, round_jiffies(fdmi_tov));
+ iport->fabric.fdmi_pending |= FDLS_FDMI_ABORT_PENDING;
+}
+
static void fdls_send_fabric_flogi(struct fnic_iport_s *iport)
{
struct fc_std_flogi flogi;
@@ -737,6 +882,47 @@ static void fdls_send_fabric_plogi(struct fnic_iport_s *iport)
fdls_start_fabric_timer(iport, 2 * iport->e_d_tov);
}
+static void fdls_send_fdmi_plogi(struct fnic_iport_s *iport)
+{
+ struct fc_std_flogi plogi;
+ struct fc_frame_header *fchdr = &plogi.fchdr;
+ uint8_t fcid[3];
+ struct fnic *fnic = iport->fnic;
+ u64 fdmi_tov;
+ uint16_t oxid;
+
+ memcpy(&plogi, &fnic_std_plogi_req, sizeof(plogi));
+
+ oxid = fdls_alloc_fabric_oxid(iport, &iport->fdmi_oxid_pool,
+ FNIC_FDMI_PLOGI_RSP);
+ if (oxid == 0xFFFF) {
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Failed to allocate OXID to send fdmi plogi %p",
+ iport);
+ return;
+ }
+
+ hton24(fcid, iport->fcid);
+
+ FNIC_STD_SET_S_ID(fchdr, fcid);
+ hton24(fcid, 0XFFFFFA);
+ FNIC_STD_SET_D_ID(fchdr, fcid);
+ FNIC_STD_SET_OX_ID(fchdr, htons(oxid));
+ FNIC_LOGI_SET_NPORT_NAME(&plogi.els, iport->wwpn);
+ FNIC_LOGI_SET_NODE_NAME(&plogi.els, iport->wwnn);
+ FNIC_LOGI_SET_RDF_SIZE(&plogi.els, iport->max_payload_size);
+
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "fcid: 0x%x: FDLS send FDMI PLOGI with oxid:%x",
+ iport->fcid, oxid);
+
+ fnic_send_fcoe_frame(iport, &plogi, sizeof(struct fc_std_flogi));
+
+ fdmi_tov = jiffies + msecs_to_jiffies(2 * iport->e_d_tov);
+ mod_timer(&iport->fabric.fdmi_timer, round_jiffies(fdmi_tov));
+ iport->fabric.fdmi_pending = FDLS_FDMI_PLOGI_PENDING;
+}
+
static void fdls_send_rpn_id(struct fnic_iport_s *iport)
{
struct fc_std_rpn_id rpn_id;
@@ -1375,6 +1561,126 @@ struct fnic_tport_s *fnic_find_tport_by_wwpn(struct fnic_iport_s *iport,
return NULL;
}
+static void fdls_fdmi_register_hba(struct fnic_iport_s *iport)
+{
+ struct fc_std_fdmi_rhba fdmi_rhba;
+ uint8_t fcid[3];
+ uint16_t len;
+ int err;
+ struct fnic *fnic = iport->fnic;
+ struct vnic_devcmd_fw_info *fw_info = NULL;
+ uint16_t oxid;
+
+ memcpy(&fdmi_rhba, &fnic_std_fdmi_rhba,
+ sizeof(struct fc_std_fdmi_rhba));
+
+ hton24(fcid, iport->fcid);
+ FNIC_STD_SET_S_ID((&fdmi_rhba.fchdr), fcid);
+
+ oxid = fdls_alloc_fabric_oxid(iport, &iport->fdmi_oxid_pool,
+ FNIC_FDMI_REG_HBA_RSP);
+ if (oxid == 0xFFFF) {
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Failed to allocate OXID to send fdmi reg hba %p",
+ iport);
+ return;
+ }
+ FNIC_STD_SET_OX_ID(&fdmi_rhba.fchdr, htons(oxid));
+
+ fdmi_rhba.hba_identifier = get_unaligned_be64(&iport->wwpn);
+ fdmi_rhba.port_name = get_unaligned_be64(&iport->wwpn);
+ fdmi_rhba.node_name = get_unaligned_be64(&iport->wwnn);
+
+ err = vnic_dev_fw_info(fnic->vdev, &fw_info);
+ if (!err) {
+ snprintf(fdmi_rhba.serial_num, sizeof(fdmi_rhba.serial_num) - 1,
+ "%s", fw_info->hw_serial_number);
+ snprintf(fdmi_rhba.hardware_ver,
+ sizeof(fdmi_rhba.hardware_ver) - 1, "%s",
+ fw_info->hw_version);
+ strscpy(fdmi_rhba.firmware_ver, fw_info->fw_version,
+ sizeof(fdmi_rhba.firmware_ver) - 1);
+
+ len = ARRAY_SIZE(fdmi_rhba.model);
+ if (fnic->subsys_desc_len >= len)
+ fnic->subsys_desc_len = len - 1;
+ memcpy(&fdmi_rhba.model, fnic->subsys_desc,
+ fnic->subsys_desc_len);
+ fdmi_rhba.model[fnic->subsys_desc_len] = 0x00;
+ }
+
+ snprintf(fdmi_rhba.driver_ver, sizeof(fdmi_rhba.driver_ver) - 1, "%s",
+ DRV_VERSION);
+ snprintf(fdmi_rhba.rom_ver, sizeof(fdmi_rhba.rom_ver) - 1, "%s", "N/A");
+
+ fnic_send_fcoe_frame(iport, &fdmi_rhba,
+ sizeof(struct fc_std_fdmi_rhba));
+ iport->fabric.fdmi_pending |= FDLS_FDMI_REG_HBA_PENDING;
+}
+
+static void fdls_fdmi_register_pa(struct fnic_iport_s *iport)
+{
+ struct fc_std_fdmi_rpa fdmi_rpa;
+
+ uint8_t fcid[3];
+ struct fnic *fnic = iport->fnic;
+ u32 port_speed_bm;
+ u32 port_speed = vnic_dev_port_speed(fnic->vdev);
+ uint16_t oxid;
+
+ memcpy(&fdmi_rpa, &fnic_std_fdmi_rpa, sizeof(struct fc_std_fdmi_rpa));
+ hton24(fcid, iport->fcid);
+ FNIC_STD_SET_S_ID((&fdmi_rpa.fchdr), fcid);
+
+ oxid = fdls_alloc_fabric_oxid(iport, &iport->fdmi_oxid_pool,
+ FNIC_FDMI_RPA_RSP);
+ if (oxid == 0xFFFF) {
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Failed to allocate OXID to send fdmi rpa %p",
+ iport);
+ return;
+ }
+ FNIC_STD_SET_OX_ID(&fdmi_rpa.fchdr, htons(oxid));
+
+ fdmi_rpa.port_name = get_unaligned_be64(&iport->wwpn);
+
+ /* MDS does not support GIGE speed.
+ * Bit shift standard definitions from scsi_transport_fc.h to
+ * match FC spec.
+ */
+ switch (port_speed) {
+ case DCEM_PORTSPEED_10G:
+ case DCEM_PORTSPEED_20G:
+ /* There is no bit for 20G */
+ port_speed_bm = FC_PORTSPEED_10GBIT << PORT_SPEED_BIT_14;
+ break;
+ case DCEM_PORTSPEED_25G:
+ port_speed_bm = FC_PORTSPEED_25GBIT << PORT_SPEED_BIT_8;
+ break;
+ case DCEM_PORTSPEED_40G:
+ case DCEM_PORTSPEED_4x10G:
+ port_speed_bm = FC_PORTSPEED_40GBIT << PORT_SPEED_BIT_9;
+ break;
+ case DCEM_PORTSPEED_100G:
+ port_speed_bm = FC_PORTSPEED_100GBIT << PORT_SPEED_BIT_8;
+ break;
+ default:
+ port_speed_bm = FC_PORTSPEED_1GBIT << PORT_SPEED_BIT_15;
+ break;
+ }
+ fdmi_rpa.supported_speed = htonl(port_speed_bm);
+ fdmi_rpa.current_speed = htonl(port_speed_bm);
+ fdmi_rpa.fc4_type[2] = 1;
+ snprintf(fdmi_rpa.os_name, sizeof(fdmi_rpa.os_name) - 1, "host%d",
+ fnic->lport->host->host_no);
+ sprintf(fc_host_system_hostname(fnic->lport->host), "%s", utsname()->nodename);
+ snprintf(fdmi_rpa.host_name, sizeof(fdmi_rpa.host_name) - 1, "%s",
+ fc_host_system_hostname(fnic->lport->host));
+
+ fnic_send_fcoe_frame(iport, &fdmi_rpa, sizeof(struct fc_std_fdmi_rpa));
+ iport->fabric.fdmi_pending |= FDLS_FDMI_RPA_PENDING;
+}
+
void fdls_fabric_timer_callback(struct timer_list *t)
{
struct fnic_fdls_fabric_s *fabric = from_timer(fabric, t, retry_timer);
@@ -1547,6 +1853,43 @@ void fdls_fabric_timer_callback(struct timer_list *t)
spin_unlock_irqrestore(&fnic->fnic_lock, flags);
}
+void fdls_fdmi_timer_callback(struct timer_list *t)
+{
+ struct fnic_fdls_fabric_s *fabric = from_timer(fabric, t, fdmi_timer);
+ struct fnic_iport_s *iport =
+ container_of(fabric, struct fnic_iport_s, fabric);
+ struct fnic *fnic = iport->fnic;
+ unsigned long flags;
+
+ spin_lock_irqsave(&fnic->fnic_lock, flags);
+
+ if (!iport->fabric.fdmi_pending) {
+ /* timer expired after fdmi responses received. */
+ spin_unlock_irqrestore(&fnic->fnic_lock, flags);
+ return;
+ }
+
+ /* if not abort pending, send an abort */
+ if (!(iport->fabric.fdmi_pending & FDLS_FDMI_ABORT_PENDING)) {
+ fdls_send_fdmi_abts(iport);
+ spin_unlock_irqrestore(&fnic->fnic_lock, flags);
+ return;
+ }
+
+ /* Abort timed out */
+ fdls_schedule_fdmi_oxid_free(iport);
+
+ iport->fabric.fdmi_pending = 0;
+ /* If max retries not exhaused, start over from fdmi plogi */
+ if (iport->fabric.fdmi_retry < FDLS_FDMI_MAX_RETRY) {
+ iport->fabric.fdmi_retry++;
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "retry fdmi timer %d", iport->fabric.fdmi_retry);
+ fdls_send_fdmi_plogi(iport);
+ }
+ spin_unlock_irqrestore(&fnic->fnic_lock, flags);
+}
+
static void fdls_send_delete_tport_msg(struct fnic_tport_s *tport)
{
struct fnic_iport_s *iport = (struct fnic_iport_s *) tport->iport;
@@ -1696,6 +2039,15 @@ static void fnic_fdls_start_plogi(struct fnic_iport_s *iport)
fdls_send_fabric_plogi(iport);
fdls_set_state((&iport->fabric), FDLS_STATE_FABRIC_PLOGI);
iport->fabric.flags &= ~FNIC_FDLS_FABRIC_ABORT_ISSUED;
+
+ if ((fnic_fdmi_support == 1) && (!(iport->flags & FNIC_FDMI_ACTIVE))) {
+ /* we can do FDMI at the same time */
+ iport->fabric.fdmi_retry = 0;
+ timer_setup(&iport->fabric.fdmi_timer, fdls_fdmi_timer_callback,
+ 0);
+ fdls_send_fdmi_plogi(iport);
+ iport->flags |= FNIC_FDMI_ACTIVE;
+ }
}
static void
@@ -2783,6 +3135,109 @@ fdls_process_fabric_plogi_rsp(struct fnic_iport_s *iport,
}
}
+static void fdls_process_fdmi_plogi_rsp(struct fnic_iport_s *iport,
+ struct fc_frame_header *fchdr)
+{
+ struct fc_std_flogi *plogi_rsp = (struct fc_std_flogi *)fchdr;
+ struct fc_std_els_rsp *els_rjt = (struct fc_std_els_rsp *)fchdr;
+ struct fnic *fnic = iport->fnic;
+ u64 fdmi_tov;
+
+ iport->fabric.fdmi_pending &= ~FDLS_FDMI_PLOGI_PENDING;
+ fdls_free_fabric_oxid(iport, &iport->fabric_oxid_pool,
+ ntohs(FNIC_STD_GET_OX_ID(fchdr)));
+
+ if (ntoh24(fchdr->fh_s_id) == 0XFFFFFA) {
+ del_timer_sync(&iport->fabric.fdmi_timer);
+ iport->fabric.fdmi_pending = 0;
+ switch (plogi_rsp->els.fl_cmd) {
+ case ELS_LS_ACC:
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "FDLS process fdmi PLOGI response status: ELS_LS_ACC\n");
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Sending fdmi registration for port 0x%x\n",
+ iport->fcid);
+
+ fdls_fdmi_register_hba(iport);
+ fdls_fdmi_register_pa(iport);
+ fdmi_tov = jiffies + msecs_to_jiffies(5000);
+ mod_timer(&iport->fabric.fdmi_timer,
+ round_jiffies(fdmi_tov));
+ break;
+ case ELS_LS_RJT:
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Fabric FDMI PLOGI returned ELS_LS_RJT reason: 0x%x",
+ els_rjt->u.rej.er_reason);
+
+ if (((els_rjt->u.rej.er_reason == ELS_RJT_BUSY)
+ || (els_rjt->u.rej.er_reason == ELS_RJT_UNAB))
+ && (iport->fabric.fdmi_retry < 7)) {
+ iport->fabric.fdmi_retry++;
+ fdls_send_fdmi_plogi(iport);
+ }
+ break;
+ default:
+ break;
+ }
+ }
+}
+
+static void fdls_process_fdmi_reg_ack(struct fnic_iport_s *iport,
+ struct fc_frame_header *fchdr,
+ int rsp_type)
+{
+ struct fnic *fnic = iport->fnic;
+
+ if (!iport->fabric.fdmi_pending) {
+ FNIC_FCS_DBG(KERN_ERR, fnic->lport->host, fnic->fnic_num,
+ "Received FDMI ack while not waiting:%x\n",
+ ntohs(FNIC_STD_GET_OX_ID(fchdr)));
+ return;
+ }
+
+ if (rsp_type == FNIC_FDMI_REG_HBA_RSP)
+ iport->fabric.fdmi_pending &= ~FDLS_FDMI_REG_HBA_PENDING;
+ else
+ iport->fabric.fdmi_pending &= ~FDLS_FDMI_RPA_PENDING;
+
+ fdls_free_fabric_oxid(iport, &iport->fdmi_oxid_pool,
+ ntohs(FNIC_STD_GET_OX_ID(fchdr)));
+
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "iport fcid: 0x%x: Received FDMI registration ack\n",
+ iport->fcid);
+
+ if (!iport->fabric.fdmi_pending) {
+ del_timer_sync(&iport->fabric.fdmi_timer);
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "iport fcid: 0x%x: Canceling FDMI timer\n",
+ iport->fcid);
+ }
+}
+
+static void fdls_process_fdmi_abts_rsp(struct fnic_iport_s *iport,
+ struct fc_frame_header *fchdr)
+{
+ uint32_t s_id;
+ struct fnic *fnic = iport->fnic;
+
+ s_id = ntoh24(FNIC_STD_GET_S_ID(fchdr));
+
+ if (!(s_id != 0xFFFFFA)) {
+ FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
+ "Received abts rsp with invalid SID: 0x%x. Dropping frame",
+ s_id);
+ return;
+ }
+
+ del_timer_sync(&iport->fabric.fdmi_timer);
+ iport->fabric.fdmi_pending &= ~FDLS_FDMI_ABORT_PENDING;
+
+ fdls_free_fabric_oxid(iport, &iport->fdmi_oxid_pool,
+ ntohs(FNIC_STD_GET_OX_ID(fchdr)));
+ fdls_send_fdmi_plogi(iport);
+}
+
static void
fdls_process_fabric_abts_rsp(struct fnic_iport_s *iport,
struct fc_frame_header *fchdr)
@@ -3779,6 +4234,9 @@ fnic_fdls_validate_and_get_frame_type(struct fnic_iport_s *iport,
break;
case FNIC_FABRIC_LOGO_RSP:
+ case FNIC_FDMI_PLOGI_RSP:
+ case FNIC_FDMI_REG_HBA_RSP:
+ case FNIC_FDMI_RPA_RSP:
break;
default:
/* Drop the Rx frame and log/stats it */
@@ -3821,6 +4279,9 @@ void fnic_fdls_recv_frame(struct fnic_iport_s *iport, void *rx_frame,
case FNIC_FABRIC_PLOGI_RSP:
fdls_process_fabric_plogi_rsp(iport, fchdr);
break;
+ case FNIC_FDMI_PLOGI_RSP:
+ fdls_process_fdmi_plogi_rsp(iport, fchdr);
+ break;
case FNIC_FABRIC_RPN_RSP:
fdls_process_rpn_id_rsp(iport, fchdr);
break;
@@ -3860,6 +4321,9 @@ void fnic_fdls_recv_frame(struct fnic_iport_s *iport, void *rx_frame,
if (fdls_is_oxid_in_fabric_range(oxid) &&
(iport->fabric.flags & FNIC_FDLS_FABRIC_ABORT_ISSUED)) {
fdls_process_fabric_abts_rsp(iport, fchdr);
+ } else if (fdls_is_oxid_in_fdmi_range(oxid) &&
+ iport->fabric.fdmi_pending) {
+ fdls_process_fdmi_abts_rsp(iport, fchdr);
} else {
fdls_process_tgt_abts_rsp(iport, fchdr);
}
@@ -3889,6 +4353,10 @@ void fnic_fdls_recv_frame(struct fnic_iport_s *iport, void *rx_frame,
case FNIC_ELS_RLS:
fdls_process_rls_req(iport, fchdr);
break;
+ case FNIC_FDMI_REG_HBA_RSP:
+ case FNIC_FDMI_RPA_RSP:
+ fdls_process_fdmi_reg_ack(iport, fchdr, frame_type);
+ break;
default:
FNIC_FCS_DBG(KERN_INFO, fnic->lport->host, fnic->fnic_num,
"s_id: 0x%x d_did: 0x%x", s_id, d_id);
@@ -3897,3 +4365,4 @@ void fnic_fdls_recv_frame(struct fnic_iport_s *iport, void *rx_frame,
break;
}
}
+
@@ -82,6 +82,72 @@
/* Retry supported by rport (returned by PRLI service parameters) */
#define FNIC_FC_RP_FLAGS_RETRY 0x1
+/* Cisco vendor id */
+#define PCI_VENDOR_ID_CISCO 0x1137
+#define PCI_DEVICE_ID_CISCO_VIC_FC 0x0045 /* fc vnic */
+
+/* sereno pcie switch */
+#define PCI_DEVICE_ID_CISCO_SERENO 0x004e
+#define PCI_DEVICE_ID_CISCO_CRUZ 0x007a /* Cruz */
+#define PCI_DEVICE_ID_CISCO_BODEGA 0x0131 /* Bodega */
+#define PCI_DEVICE_ID_CISCO_BEVERLY 0x025f /* Beverly */
+
+/* Sereno */
+#define PCI_SUBDEVICE_ID_CISCO_VASONA 0x004f /* vasona mezz */
+#define PCI_SUBDEVICE_ID_CISCO_COTATI 0x0084 /* cotati mlom */
+#define PCI_SUBDEVICE_ID_CISCO_LEXINGTON 0x0085 /* lexington pcie */
+#define PCI_SUBDEVICE_ID_CISCO_ICEHOUSE 0x00cd /* Icehouse */
+#define PCI_SUBDEVICE_ID_CISCO_KIRKWOODLAKE 0x00ce /* KirkwoodLake pcie */
+#define PCI_SUBDEVICE_ID_CISCO_SUSANVILLE 0x012e /* Susanville MLOM */
+#define PCI_SUBDEVICE_ID_CISCO_TORRANCE 0x0139 /* Torrance MLOM */
+
+/* Cruz */
+#define PCI_SUBDEVICE_ID_CISCO_CALISTOGA 0x012c /* Calistoga MLOM */
+#define PCI_SUBDEVICE_ID_CISCO_MOUNTAINVIEW 0x0137 /* Cruz Mezz */
+/* Cruz MountTian SIOC */
+#define PCI_SUBDEVICE_ID_CISCO_MOUNTTIAN 0x014b
+#define PCI_SUBDEVICE_ID_CISCO_CLEARLAKE 0x014d /* ClearLake pcie */
+/* Cruz MountTian2 SIOC */
+#define PCI_SUBDEVICE_ID_CISCO_MOUNTTIAN2 0x0157
+#define PCI_SUBDEVICE_ID_CISCO_CLAREMONT 0x015d /* Claremont MLOM */
+
+/* Bodega */
+/* VIC 1457 PCIe mLOM */
+#define PCI_SUBDEVICE_ID_CISCO_BRADBURY 0x0218
+#define PCI_SUBDEVICE_ID_CISCO_BRENTWOOD 0x0217 /* VIC 1455 PCIe */
+/* VIC 1487 PCIe mLOM */
+#define PCI_SUBDEVICE_ID_CISCO_BURLINGAME 0x021a
+#define PCI_SUBDEVICE_ID_CISCO_BAYSIDE 0x0219 /* VIC 1485 PCIe */
+/* VIC 1440 Mezz mLOM */
+#define PCI_SUBDEVICE_ID_CISCO_BAKERSFIELD 0x0215
+#define PCI_SUBDEVICE_ID_CISCO_BOONVILLE 0x0216 /* VIC 1480 Mezz */
+#define PCI_SUBDEVICE_ID_CISCO_BENICIA 0x024a /* VIC 1495 */
+#define PCI_SUBDEVICE_ID_CISCO_BEAUMONT 0x024b /* VIC 1497 */
+#define PCI_SUBDEVICE_ID_CISCO_BRISBANE 0x02af /* VIC 1467 */
+#define PCI_SUBDEVICE_ID_CISCO_BENTON 0x02b0 /* VIC 1477 */
+#define PCI_SUBDEVICE_ID_CISCO_TWIN_RIVER 0x02cf /* VIC 14425 */
+#define PCI_SUBDEVICE_ID_CISCO_TWIN_PEAK 0x02d0 /* VIC 14825 */
+
+/* Beverly */
+#define PCI_SUBDEVICE_ID_CISCO_BERN 0x02de /* VIC 15420 */
+#define PCI_SUBDEVICE_ID_CISCO_STOCKHOLM 0x02dd /* VIC 15428 */
+#define PCI_SUBDEVICE_ID_CISCO_KRAKOW 0x02dc /* VIC 15411 */
+#define PCI_SUBDEVICE_ID_CISCO_LUCERNE 0x02db /* VIC 15231 */
+#define PCI_SUBDEVICE_ID_CISCO_TURKU 0x02e8 /* VIC 15238 */
+#define PCI_SUBDEVICE_ID_CISCO_TURKU_PLUS 0x02f3 /* VIC 15237 */
+#define PCI_SUBDEVICE_ID_CISCO_ZURICH 0x02df /* VIC 15230 */
+#define PCI_SUBDEVICE_ID_CISCO_RIGA 0x02e0 /* VIC 15427 */
+#define PCI_SUBDEVICE_ID_CISCO_GENEVA 0x02e1 /* VIC 15422 */
+#define PCI_SUBDEVICE_ID_CISCO_HELSINKI 0x02e4 /* VIC 15235 */
+#define PCI_SUBDEVICE_ID_CISCO_GOTHENBURG 0x02f2 /* VIC 15425 */
+
+struct fnic_pcie_device {
+ u32 device;
+ u8 *desc;
+ u32 subsystem_device;
+ u8 *subsys_desc;
+};
+
/*
* fnic private data per SCSI command.
* These fields are locked by the hashed io_req_lock.
@@ -134,6 +200,7 @@ static inline u64 fnic_flags_and_state(struct scsi_cmnd *cmd)
#define fnic_clear_state_flags(fnicp, st_flags) \
__fnic_set_state_flags(fnicp, st_flags, 1)
+extern unsigned int fnic_fdmi_support;
extern unsigned int fnic_log_level;
extern unsigned int io_completions;
extern struct workqueue_struct *fnic_event_queue;
@@ -336,6 +403,9 @@ struct fnic {
struct work_struct tport_work;
struct list_head tport_event_list;
+ char subsys_desc[14];
+ int subsys_desc_len;
+
/*** FIP related data members -- start ***/
void (*set_vlan)(struct fnic *, u16 vlan);
struct work_struct fip_frame_work;
@@ -433,5 +503,7 @@ fnic_chk_state_flags_locked(struct fnic *fnic, unsigned long st_flags)
void __fnic_set_state_flags(struct fnic *, unsigned long, unsigned long);
void fnic_dump_fchost_stats(struct Scsi_Host *, struct fc_host_statistics *);
void fnic_free_txq(struct list_head *head);
+int fnic_get_desc_by_devid(struct pci_dev *pdev, char **desc,
+ char **subsys_desc);
#endif /* _FNIC_H_ */
@@ -427,4 +427,3 @@ struct fnic_tport_s *fnic_find_tport_by_wwpn(struct fnic_iport_s *iport,
uint64_t wwpn);
#endif /* _FNIC_FDLS_H_ */
-
@@ -62,6 +62,9 @@ unsigned int fnic_log_level;
module_param(fnic_log_level, int, S_IRUGO|S_IWUSR);
MODULE_PARM_DESC(fnic_log_level, "bit mask of fnic logging levels");
+unsigned int fnic_fdmi_support = 1;
+module_param(fnic_fdmi_support, int, 0644);
+MODULE_PARM_DESC(fnic_fdmi_support, "FDMI support");
unsigned int io_completions = FNIC_DFLT_IO_COMPLETIONS;
module_param(io_completions, int, S_IRUGO|S_IWUSR);
@@ -607,6 +610,8 @@ static int fnic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
int i;
unsigned long flags;
int hwq;
+ char *desc, *subsys_desc;
+ int len;
/*
* Allocate SCSI Host and set up association between host,
@@ -640,6 +645,23 @@ static int fnic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
fnic->fnic_num = fnic_id;
fnic_stats_debugfs_init(fnic);
+ /* Find model name from PCIe subsys ID */
+ if (fnic_get_desc_by_devid(pdev, &desc, &subsys_desc) == 0) {
+ dev_info(&fnic->pdev->dev, "Model: %s\n", subsys_desc);
+
+ /* Update FDMI model */
+ fnic->subsys_desc_len = strlen(subsys_desc);
+ len = ARRAY_SIZE(fnic->subsys_desc);
+ if (fnic->subsys_desc_len > len)
+ fnic->subsys_desc_len = len;
+ memcpy(fnic->subsys_desc, subsys_desc, fnic->subsys_desc_len);
+ dev_info(&fnic->pdev->dev, "FDMI Model: %s\n", fnic->subsys_desc);
+ } else {
+ fnic->subsys_desc_len = 0;
+ dev_info(&fnic->pdev->dev, "Model: %s subsys_id: 0x%04x\n", "Unknown",
+ pdev->subsystem_device);
+ }
+
err = pci_enable_device(pdev);
if (err) {
dev_err(&fnic->pdev->dev, "Cannot enable PCI device, aborting.\n");
@@ -1014,6 +1036,9 @@ static void fnic_remove(struct pci_dev *pdev)
fnic_fcoe_evlist_free(fnic);
}
+ if ((fnic_fdmi_support == 1) && (fnic->iport.fabric.fdmi_pending > 0))
+ del_timer_sync(&fnic->iport.fabric.fdmi_timer);
+
/*
* Log off the fabric. This stops all remote ports, dns port,
* logs off the fabric. This flushes all rport, disc, lport work
@@ -1201,3 +1226,4 @@ static void __exit fnic_cleanup_module(void)
module_init(fnic_init_module);
module_exit(fnic_cleanup_module);
+
new file mode 100644
@@ -0,0 +1,131 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2008 Cisco Systems, Inc. All rights reserved.
+ * Copyright 2007 Nuova Systems, Inc. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/mempool.h>
+#include <linux/string.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/spinlock.h>
+#include <linux/workqueue.h>
+#include <linux/kthread.h>
+#include <linux/if_ether.h>
+#include "fnic.h"
+
+static struct fnic_pcie_device fnic_pcie_device_table[] = {
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno", PCI_SUBDEVICE_ID_CISCO_VASONA,
+ "VIC 1280"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno", PCI_SUBDEVICE_ID_CISCO_COTATI,
+ "VIC 1240"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno",
+ PCI_SUBDEVICE_ID_CISCO_LEXINGTON, "VIC 1225"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno", PCI_SUBDEVICE_ID_CISCO_ICEHOUSE,
+ "VIC 1285"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno",
+ PCI_SUBDEVICE_ID_CISCO_KIRKWOODLAKE, "VIC 1225T"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno",
+ PCI_SUBDEVICE_ID_CISCO_SUSANVILLE, "VIC 1227"},
+ {PCI_DEVICE_ID_CISCO_SERENO, "Sereno", PCI_SUBDEVICE_ID_CISCO_TORRANCE,
+ "VIC 1227T"},
+
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_CALISTOGA,
+ "VIC 1340"},
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_MOUNTAINVIEW,
+ "VIC 1380"},
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_MOUNTTIAN,
+ "C3260-SIOC"},
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_CLEARLAKE,
+ "VIC 1385"},
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_MOUNTTIAN2,
+ "C3260-SIOC"},
+ {PCI_DEVICE_ID_CISCO_CRUZ, "Cruz", PCI_SUBDEVICE_ID_CISCO_CLAREMONT,
+ "VIC 1387"},
+
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BRADBURY,
+ "VIC 1457"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_BRENTWOOD, "VIC 1455"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_BURLINGAME, "VIC 1487"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BAYSIDE,
+ "VIC 1485"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_BAKERSFIELD, "VIC 1440"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_BOONVILLE, "VIC 1480"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BENICIA,
+ "VIC 1495"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BEAUMONT,
+ "VIC 1497"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BRISBANE,
+ "VIC 1467"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega", PCI_SUBDEVICE_ID_CISCO_BENTON,
+ "VIC 1477"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_TWIN_RIVER, "VIC 14425"},
+ {PCI_DEVICE_ID_CISCO_BODEGA, "Bodega",
+ PCI_SUBDEVICE_ID_CISCO_TWIN_PEAK, "VIC 14825"},
+
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_BERN,
+ "VIC 15420"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly",
+ PCI_SUBDEVICE_ID_CISCO_STOCKHOLM, "VIC 15428"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_KRAKOW,
+ "VIC 15411"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly",
+ PCI_SUBDEVICE_ID_CISCO_LUCERNE, "VIC 15231"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_TURKU,
+ "VIC 15238"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_GENEVA,
+ "VIC 15422"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly",
+ PCI_SUBDEVICE_ID_CISCO_HELSINKI, "VIC 15235"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly",
+ PCI_SUBDEVICE_ID_CISCO_GOTHENBURG, "VIC 15425"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly",
+ PCI_SUBDEVICE_ID_CISCO_TURKU_PLUS, "VIC 15237"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_ZURICH,
+ "VIC 15230"},
+ {PCI_DEVICE_ID_CISCO_BEVERLY, "Beverly", PCI_SUBDEVICE_ID_CISCO_RIGA,
+ "VIC 15427"},
+
+ {0,}
+};
+
+int fnic_get_desc_by_devid(struct pci_dev *pdev, char **desc,
+ char **subsys_desc)
+{
+ unsigned short device = PCI_DEVICE_ID_CISCO_VIC_FC;
+ int max = ARRAY_SIZE(fnic_pcie_device_table);
+ struct fnic_pcie_device *t = fnic_pcie_device_table;
+ int index = 0;
+
+ if (pdev->device != device)
+ return 1;
+
+ while (t->device != 0) {
+ if (memcmp
+ ((char *) &pdev->subsystem_device,
+ (char *) &t->subsystem_device, sizeof(short)) == 0)
+ break;
+ t++;
+ index++;
+ }
+
+ if (index >= max - 1) {
+ *desc = NULL;
+ *subsys_desc = NULL;
+ return 1;
+ }
+
+ *desc = fnic_pcie_device_table[index].desc;
+ *subsys_desc = fnic_pcie_device_table[index].subsys_desc;
+ return 0;
+}