diff --git a/qla2x00t-32gbit/qla_bsg.c b/qla2x00t-32gbit/qla_bsg.c index b37851583..aa991c86c 100644 --- a/qla2x00t-32gbit/qla_bsg.c +++ b/qla2x00t-32gbit/qla_bsg.c @@ -50,10 +50,11 @@ void qla2x00_bsg_job_done(srb_t *sp, int res) #endif struct fc_bsg_reply *bsg_reply = bsg_job->reply; + sp->free(sp); + bsg_reply->result = res; bsg_job_done(bsg_job, bsg_reply->result, bsg_reply->reply_payload_rcv_len); - sp->free(sp); } void qla2x00_bsg_sp_free(srb_t *sp) @@ -2725,6 +2726,10 @@ qla2x00_get_host_stats(struct bsg_job *bsg_job) } data = kzalloc(response_len, GFP_KERNEL); + if (!data) { + ret = -ENOMEM; + goto host_stat_out; + } ret = qla2xxx_get_ini_stats(fc_bsg_to_shost(bsg_job), req_data->stat_type, data, response_len); diff --git a/qla2x00t-32gbit/qla_dbg.c b/qla2x00t-32gbit/qla_dbg.c index 144a893e7..f2d05592c 100644 --- a/qla2x00t-32gbit/qla_dbg.c +++ b/qla2x00t-32gbit/qla_dbg.c @@ -113,8 +113,13 @@ qla27xx_dump_mpi_ram(struct qla_hw_data *ha, uint32_t addr, uint32_t *ram, uint32_t stat; ulong i, j, timer = 6000000; int rval = QLA_FUNCTION_FAILED; + scsi_qla_host_t *vha = pci_get_drvdata(ha->pdev); clear_bit(MBX_INTERRUPT, &ha->mbx_cmd_flags); + + if (qla_pci_disconnected(vha, reg)) + return rval; + for (i = 0; i < ram_dwords; i += dwords, addr += dwords) { if (i + dwords > ram_dwords) dwords = ram_dwords - i; @@ -138,6 +143,9 @@ qla27xx_dump_mpi_ram(struct qla_hw_data *ha, uint32_t addr, uint32_t *ram, while (timer--) { udelay(5); + if (qla_pci_disconnected(vha, reg)) + return rval; + stat = rd_reg_dword(®->host_status); /* Check for pending interrupts. */ if (!(stat & HSRX_RISC_INT)) @@ -192,9 +200,13 @@ qla24xx_dump_ram(struct qla_hw_data *ha, uint32_t addr, __be32 *ram, uint32_t dwords = qla2x00_gid_list_size(ha) / 4; uint32_t stat; ulong i, j, timer = 6000000; + scsi_qla_host_t *vha = pci_get_drvdata(ha->pdev); clear_bit(MBX_INTERRUPT, &ha->mbx_cmd_flags); + if (qla_pci_disconnected(vha, reg)) + return rval; + for (i = 0; i < ram_dwords; i += dwords, addr += dwords) { if (i + dwords > ram_dwords) dwords = ram_dwords - i; @@ -216,8 +228,10 @@ qla24xx_dump_ram(struct qla_hw_data *ha, uint32_t addr, __be32 *ram, ha->flags.mbox_int = 0; while (timer--) { udelay(5); - stat = rd_reg_dword(®->host_status); + if (qla_pci_disconnected(vha, reg)) + return rval; + stat = rd_reg_dword(®->host_status); /* Check for pending interrupts. */ if (!(stat & HSRX_RISC_INT)) continue; diff --git a/qla2x00t-32gbit/qla_dbg.h b/qla2x00t-32gbit/qla_dbg.h index 2e59e75c6..9eb708e5e 100644 --- a/qla2x00t-32gbit/qla_dbg.h +++ b/qla2x00t-32gbit/qla_dbg.h @@ -308,7 +308,7 @@ struct qla2xxx_fw_dump { }; #define QL_MSGHDR "qla2xxx" -#define QL_DBG_DEFAULT1_MASK 0x1e400000 +#define QL_DBG_DEFAULT1_MASK 0x1e600000 #define ql_log_fatal 0 /* display fatal errors */ #define ql_log_warn 1 /* display critical errors */ diff --git a/qla2x00t-32gbit/qla_def.h b/qla2x00t-32gbit/qla_def.h index 7df2eee7b..edd938e54 100644 --- a/qla2x00t-32gbit/qla_def.h +++ b/qla2x00t-32gbit/qla_def.h @@ -418,6 +418,7 @@ typedef union { } b; } port_id_t; #define INVALID_PORT_ID 0xFFFFFF +#define ISP_REG16_DISCONNECT 0xFFFF static inline le_id_t be_id_to_le(be_id_t id) { @@ -1553,7 +1554,7 @@ struct init_sf_cb { * BIT_12 = Remote Write Optimization (1 - Enabled, 0 - Disabled) * BIT 11-0 = Reserved */ - uint16_t flags; + __le16 flags; uint8_t reserved1[32]; uint16_t discard_OHRB_timeout_value; uint16_t remote_write_opt_queue_num; @@ -3842,7 +3843,7 @@ struct qlt_hw_data { __le32 __iomem *atio_q_in; __le32 __iomem *atio_q_out; - struct qla_tgt_func_tmpl *tgt_ops; + const struct qla_tgt_func_tmpl *tgt_ops; struct qla_tgt_vp_map *tgt_vp_map; int saved_set; @@ -3884,6 +3885,13 @@ struct qla_hw_data_stat { u32 num_mpi_reset; }; +/* refer to pcie_do_recovery reference */ +typedef enum { + QLA_PCI_RESUME, + QLA_PCI_ERR_DETECTED, + QLA_PCI_MMIO_ENABLED, + QLA_PCI_SLOT_RESET, +} pci_error_state_t; /* * Qlogic host adapter specific data structure. */ @@ -4634,6 +4642,7 @@ struct qla_hw_data { #define DEFAULT_ZIO_THRESHOLD 5 struct qla_hw_data_stat stat; + pci_error_state_t pci_error_state; }; struct active_regions { @@ -4754,7 +4763,7 @@ typedef struct scsi_qla_host { #define FX00_CRITEMP_RECOVERY 25 #define FX00_HOST_INFO_RESEND 26 #define QPAIR_ONLINE_CHECK_NEEDED 27 -#define SET_NVME_ZIO_THRESHOLD_NEEDED 28 +#define DO_EEH_RECOVERY 28 #define DETECT_SFP_CHANGE 29 #define N2N_LOGIN_NEEDED 30 #define IOCB_WORK_ACTIVE 31 diff --git a/qla2x00t-32gbit/qla_gbl.h b/qla2x00t-32gbit/qla_gbl.h index 6c7c7a6b2..b56753ced 100644 --- a/qla2x00t-32gbit/qla_gbl.h +++ b/qla2x00t-32gbit/qla_gbl.h @@ -228,6 +228,7 @@ extern int qla2x00_post_uevent_work(struct scsi_qla_host *, u32); extern int qla2x00_post_uevent_work(struct scsi_qla_host *, u32); extern void qla2x00_disable_board_on_pci_error(struct work_struct *); +extern void qla_eeh_work(struct work_struct *); extern void qla2x00_sp_compl(srb_t *sp, int); extern void qla2xxx_qpair_sp_free_dma(srb_t *sp); extern void qla2xxx_qpair_sp_compl(srb_t *sp, int); @@ -239,6 +240,8 @@ int qla24xx_post_relogin_work(struct scsi_qla_host *vha); void qla2x00_wait_for_sess_deletion(scsi_qla_host_t *); void qla24xx_process_purex_rdp(struct scsi_qla_host *vha, struct purex_item *pkt); +void qla_pci_set_eeh_busy(struct scsi_qla_host *); +void qla_schedule_eeh_work(struct scsi_qla_host *); /* * Global Functions in qla_mid.c source file. diff --git a/qla2x00t-32gbit/qla_gs.c b/qla2x00t-32gbit/qla_gs.c index 517d358b0..5b6e04a91 100644 --- a/qla2x00t-32gbit/qla_gs.c +++ b/qla2x00t-32gbit/qla_gs.c @@ -1247,7 +1247,7 @@ qla2x00_sns_gnn_id(scsi_qla_host_t *vha, sw_info_t *list) } /** - * qla2x00_snd_rft_id() - SNS Register FC-4 TYPEs (RFT_ID) supported by the HBA. + * qla2x00_sns_rft_id() - SNS Register FC-4 TYPEs (RFT_ID) supported by the HBA. * @vha: HA context * * This command uses the old Exectute SNS Command mailbox routine. @@ -1479,7 +1479,7 @@ qla2x00_update_ms_fdmi_iocb(scsi_qla_host_t *vha, uint32_t req_size) } /** - * qla2x00_prep_ct_req() - Prepare common CT request fields for SNS query. + * qla2x00_prep_ct_fdmi_req() - Prepare common CT request fields for SNS query. * @p: CT request buffer * @cmd: GS command * @rsp_size: response size in bytes @@ -1582,7 +1582,7 @@ qla25xx_fdmi_port_speed_currently(struct qla_hw_data *ha) } /** - * qla2x00_hba_attributes() perform HBA attributes registration + * qla2x00_hba_attributes() - perform HBA attributes registration * @vha: HA context * @entries: number of entries to use * @callopt: Option to issue extended or standard FDMI @@ -1837,7 +1837,7 @@ done: } /** - * qla2x00_port_attributes() perform Port attributes registration + * qla2x00_port_attributes() - perform Port attributes registration * @vha: HA context * @entries: number of entries to use * @callopt: Option to issue extended or standard FDMI @@ -2272,7 +2272,7 @@ qla2x00_fdmi_dhba(scsi_qla_host_t *vha) } /** - * qla2x00_fdmi_rprt() perform RPRT registration + * qla2x00_fdmi_rprt() - perform RPRT registration * @vha: HA context * @callopt: Option to issue extended or standard FDMI * command parameter @@ -3443,6 +3443,10 @@ void qla24xx_async_gnnft_done(scsi_qla_host_t *vha, srb_t *sp) list_for_each_entry(fcport, &vha->vp_fcports, list) { if ((fcport->flags & FCF_FABRIC_DEVICE) != 0) { fcport->scan_state = QLA_FCPORT_SCAN; + if (fcport->loop_id == FC_NO_LOOP_ID) + fcport->logout_on_delete = 0; + else + fcport->logout_on_delete = 1; } } goto login_logout; diff --git a/qla2x00t-32gbit/qla_init.c b/qla2x00t-32gbit/qla_init.c index 068579863..7a0832e8e 100644 --- a/qla2x00t-32gbit/qla_init.c +++ b/qla2x00t-32gbit/qla_init.c @@ -718,6 +718,7 @@ static void qla24xx_handle_gnl_done_event(scsi_qla_host_t *vha, ql_dbg(ql_dbg_disc, vha, 0x20e0, "%s %8phC login gen changed\n", __func__, fcport->port_name); + set_bit(RELOGIN_NEEDED, &vha->dpc_flags); return; } @@ -1194,6 +1195,9 @@ static int qla24xx_post_prli_work(struct scsi_qla_host *vha, fc_port_t *fcport) { struct qla_work_evt *e; + if (vha->host->active_mode == MODE_TARGET) + return QLA_FUNCTION_FAILED; + e = qla2x00_alloc_work(vha, QLA_EVT_PRLI); if (!e) return QLA_FUNCTION_FAILED; @@ -2766,6 +2770,49 @@ qla81xx_reset_mpi(scsi_qla_host_t *vha) return qla81xx_write_mpi_register(vha, mb); } +static int +qla_chk_risc_recovery(scsi_qla_host_t *vha) +{ + struct qla_hw_data *ha = vha->hw; + struct device_reg_24xx __iomem *reg = &ha->iobase->isp24; + __le16 __iomem *mbptr = ®->mailbox0; + int i; + u16 mb[32]; + int rc = QLA_SUCCESS; + + if (!IS_QLA27XX(ha) && !IS_QLA28XX(ha)) + return rc; + + /* this check is only valid after RISC reset */ + mb[0] = rd_reg_word(mbptr); + mbptr++; + if (mb[0] == 0xf) { + rc = QLA_FUNCTION_FAILED; + + for (i = 1; i < 32; i++) { + mb[i] = rd_reg_word(mbptr); + mbptr++; + } + + ql_log(ql_log_warn, vha, 0x1015, + "RISC reset failed. mb[0-7] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n", + mb[0], mb[1], mb[2], mb[3], mb[4], mb[5], mb[6], mb[7]); + ql_log(ql_log_warn, vha, 0x1015, + "RISC reset failed. mb[8-15] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n", + mb[8], mb[9], mb[10], mb[11], mb[12], mb[13], mb[14], + mb[15]); + ql_log(ql_log_warn, vha, 0x1015, + "RISC reset failed. mb[16-23] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n", + mb[16], mb[17], mb[18], mb[19], mb[20], mb[21], mb[22], + mb[23]); + ql_log(ql_log_warn, vha, 0x1015, + "RISC reset failed. mb[24-31] %04xh %04xh %04xh %04xh %04xh %04xh %04xh %04xh\n", + mb[24], mb[25], mb[26], mb[27], mb[28], mb[29], mb[30], + mb[31]); + } + return rc; +} + /** * qla24xx_reset_risc() - Perform full reset of ISP24xx RISC. * @vha: HA context @@ -2782,6 +2829,7 @@ qla24xx_reset_risc(scsi_qla_host_t *vha) uint16_t wd; static int abts_cnt; /* ISP abort retry counts */ int rval = QLA_SUCCESS; + int print = 1; spin_lock_irqsave(&ha->hardware_lock, flags); @@ -2870,17 +2918,26 @@ qla24xx_reset_risc(scsi_qla_host_t *vha) rd_reg_dword(®->hccr); wrt_reg_dword(®->hccr, HCCRX_CLR_RISC_RESET); + mdelay(10); rd_reg_dword(®->hccr); - rd_reg_word(®->mailbox0); - for (cnt = 60; rd_reg_word(®->mailbox0) != 0 && - rval == QLA_SUCCESS; cnt--) { + wd = rd_reg_word(®->mailbox0); + for (cnt = 300; wd != 0 && rval == QLA_SUCCESS; cnt--) { barrier(); - if (cnt) - udelay(5); - else + if (cnt) { + mdelay(1); + if (print && qla_chk_risc_recovery(vha)) + print = 0; + + wd = rd_reg_word(®->mailbox0); + } else { rval = QLA_FUNCTION_TIMEOUT; + + ql_log(ql_log_warn, vha, 0x015e, + "RISC reset timeout\n"); + } } + if (rval == QLA_SUCCESS) set_bit(RISC_RDY_AFT_RESET, &ha->fw_dump_cap_flags); @@ -5512,13 +5569,14 @@ qla2x00_reg_remote_port(scsi_qla_host_t *vha, fc_port_t *fcport) if (fcport->port_type & FCT_NVME_DISCOVERY) rport_ids.roles |= FC_PORT_ROLE_NVME_DISCOVERY; + fc_remote_port_rolechg(rport, rport_ids.roles); + ql_dbg(ql_dbg_disc, vha, 0x20ee, - "%s %8phN. rport %p is %s mode\n", - __func__, fcport->port_name, rport, + "%s: %8phN. rport %ld:0:%d (%p) is %s mode\n", + __func__, fcport->port_name, vha->host_no, + rport->scsi_target_id, rport, (fcport->port_type == FCT_TARGET) ? "tgt" : ((fcport->port_type & FCT_NVME) ? "nvme" : "ini")); - - fc_remote_port_rolechg(rport, rport_ids.roles); } /* @@ -6877,22 +6935,18 @@ qla2x00_abort_isp_cleanup(scsi_qla_host_t *vha) } spin_unlock_irqrestore(&ha->vport_slock, flags); - if (!ha->flags.eeh_busy) { - /* Make sure for ISP 82XX IO DMA is complete */ - if (IS_P3P_TYPE(ha)) { - qla82xx_chip_reset_cleanup(vha); - ql_log(ql_log_info, vha, 0x00b4, - "Done chip reset cleanup.\n"); + /* Make sure for ISP 82XX IO DMA is complete */ + if (IS_P3P_TYPE(ha)) { + qla82xx_chip_reset_cleanup(vha); + ql_log(ql_log_info, vha, 0x00b4, + "Done chip reset cleanup.\n"); - /* Done waiting for pending commands. - * Reset the online flag. - */ - vha->flags.online = 0; - } - - /* Requeue all commands in outstanding command list. */ - qla2x00_abort_all_cmds(vha, DID_RESET << 16); + /* Done waiting for pending commands. Reset online flag */ + vha->flags.online = 0; } + + /* Requeue all commands in outstanding command list. */ + qla2x00_abort_all_cmds(vha, DID_RESET << 16); /* memory barrier */ wmb(); } @@ -6923,6 +6977,12 @@ qla2x00_abort_isp(scsi_qla_host_t *vha) if (vha->hw->flags.port_isolated) return status; + if (qla2x00_isp_reg_stat(ha)) { + ql_log(ql_log_info, vha, 0x803f, + "ISP Abort - ISP reg disconnect, exiting.\n"); + return status; + } + if (test_and_clear_bit(ISP_ABORT_TO_ROM, &vha->dpc_flags)) { ha->flags.chip_reset_done = 1; vha->flags.online = 1; @@ -6962,8 +7022,18 @@ qla2x00_abort_isp(scsi_qla_host_t *vha) ha->isp_ops->get_flash_version(vha, req->ring); + if (qla2x00_isp_reg_stat(ha)) { + ql_log(ql_log_info, vha, 0x803f, + "ISP Abort - ISP reg disconnect pre nvram config, exiting.\n"); + return status; + } ha->isp_ops->nvram_config(vha); + if (qla2x00_isp_reg_stat(ha)) { + ql_log(ql_log_info, vha, 0x803f, + "ISP Abort - ISP reg disconnect post nvmram config, exiting.\n"); + return status; + } if (!qla2x00_restart_isp(vha)) { clear_bit(RESET_MARKER_NEEDED, &vha->dpc_flags); diff --git a/qla2x00t-32gbit/qla_inline.h b/qla2x00t-32gbit/qla_inline.h index ca33521e2..6ad5f8a22 100644 --- a/qla2x00t-32gbit/qla_inline.h +++ b/qla2x00t-32gbit/qla_inline.h @@ -440,3 +440,49 @@ qla_put_iocbs(struct qla_qpair *qp, struct iocb_resource *iores) } iores->res_type = RESOURCE_NONE; } + +#define ISP_REG_DISCONNECT 0xffffffffU +/************************************************************************** + * qla2x00_isp_reg_stat + * + * Description: + * Read the host status register of ISP before aborting the command. + * + * Input: + * ha = pointer to host adapter structure. + * + * + * Returns: + * Either true or false. + * + * Note: Return true if there is register disconnect. + **************************************************************************/ +static inline +uint32_t qla2x00_isp_reg_stat(struct qla_hw_data *ha) +{ + struct device_reg_24xx __iomem *reg = &ha->iobase->isp24; + struct device_reg_82xx __iomem *reg82 = &ha->iobase->isp82; + + if (IS_P3P_TYPE(ha)) + return ((rd_reg_dword(®82->host_int)) == ISP_REG_DISCONNECT); + else + return ((rd_reg_dword(®->host_status)) == + ISP_REG_DISCONNECT); +} + +static inline +bool qla_pci_disconnected(struct scsi_qla_host *vha, + struct device_reg_24xx __iomem *reg) +{ + uint32_t stat; + bool ret = false; + + stat = rd_reg_dword(®->host_status); + if (stat == 0xffffffff) { + ql_log(ql_log_info, vha, 0x8041, + "detected PCI disconnect.\n"); + qla_schedule_eeh_work(vha); + ret = true; + } + return ret; +} diff --git a/qla2x00t-32gbit/qla_iocb.c b/qla2x00t-32gbit/qla_iocb.c index 14ada875f..a6e0e8fb4 100644 --- a/qla2x00t-32gbit/qla_iocb.c +++ b/qla2x00t-32gbit/qla_iocb.c @@ -491,7 +491,7 @@ qla2x00_start_iocbs(struct scsi_qla_host *vha, struct req_que *req) } /** - * qla2x00_marker() - Send a marker IOCB to the firmware. + * __qla2x00_marker() - Send a marker IOCB to the firmware. * @vha: HA context * @qpair: queue pair pointer * @loop_id: loop ID @@ -1606,12 +1606,14 @@ qla24xx_start_scsi(srb_t *sp) uint16_t req_cnt; uint16_t tot_dsds; struct req_que *req = NULL; + struct rsp_que *rsp; struct scsi_cmnd *cmd = GET_CMD_SP(sp); struct scsi_qla_host *vha = sp->vha; struct qla_hw_data *ha = vha->hw; /* Setup device pointers. */ req = vha->req; + rsp = req->rsp; /* So we know we haven't pci_map'ed anything yet */ tot_dsds = 0; @@ -1649,8 +1651,14 @@ qla24xx_start_scsi(srb_t *sp) goto queuing_error; if (req->cnt < (req_cnt + 2)) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } + if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -1713,6 +1721,11 @@ qla24xx_start_scsi(srb_t *sp) /* Set chip new ring index. */ wrt_reg_dword(req->req_q_in, req->ring_index); + /* Manage unprocessed RIO/ZIO commands in response queue. */ + if (vha->flags.process_response_queue && + rsp->ring_ptr->signature != RESPONSE_PROCESSED) + qla24xx_process_response_queue(vha, rsp); + spin_unlock_irqrestore(&ha->hardware_lock, flags); return QLA_SUCCESS; @@ -1841,8 +1854,13 @@ qla24xx_dif_start_scsi(srb_t *sp) goto queuing_error; if (req->cnt < (req_cnt + 2)) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -1903,6 +1921,11 @@ qla24xx_dif_start_scsi(srb_t *sp) /* Set chip new ring index. */ wrt_reg_dword(req->req_q_in, req->ring_index); + /* Manage unprocessed RIO/ZIO commands in response queue. */ + if (vha->flags.process_response_queue && + rsp->ring_ptr->signature != RESPONSE_PROCESSED) + qla24xx_process_response_queue(vha, rsp); + spin_unlock_irqrestore(&ha->hardware_lock, flags); return QLA_SUCCESS; @@ -1916,6 +1939,7 @@ queuing_error: qla_put_iocbs(sp->qpair, &sp->iores); spin_unlock_irqrestore(&ha->hardware_lock, flags); + return QLA_FUNCTION_FAILED; } @@ -1937,6 +1961,7 @@ qla2xxx_start_scsi_mq(srb_t *sp) uint16_t req_cnt; uint16_t tot_dsds; struct req_que *req = NULL; + struct rsp_que *rsp; struct scsi_cmnd *cmd = GET_CMD_SP(sp); struct scsi_qla_host *vha = sp->fcport->vha; struct qla_hw_data *ha = vha->hw; @@ -1947,6 +1972,7 @@ qla2xxx_start_scsi_mq(srb_t *sp) /* Setup qpair pointers */ req = qpair->req; + rsp = qpair->rsp; /* So we know we haven't pci_map'ed anything yet */ tot_dsds = 0; @@ -1983,8 +2009,14 @@ qla2xxx_start_scsi_mq(srb_t *sp) goto queuing_error; if (req->cnt < (req_cnt + 2)) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } + if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -2047,6 +2079,11 @@ qla2xxx_start_scsi_mq(srb_t *sp) /* Set chip new ring index. */ wrt_reg_dword(req->req_q_in, req->ring_index); + /* Manage unprocessed RIO/ZIO commands in response queue. */ + if (vha->flags.process_response_queue && + rsp->ring_ptr->signature != RESPONSE_PROCESSED) + qla24xx_process_response_queue(vha, rsp); + spin_unlock_irqrestore(&qpair->qp_lock, flags); return QLA_SUCCESS; @@ -2190,8 +2227,14 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp) goto queuing_error; if (req->cnt < (req_cnt + 2)) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } + if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -2268,6 +2311,7 @@ queuing_error: qla_put_iocbs(sp->qpair, &sp->iores); spin_unlock_irqrestore(&qpair->qp_lock, flags); + return QLA_FUNCTION_FAILED; } @@ -2312,6 +2356,11 @@ __qla2x00_alloc_iocbs(struct qla_qpair *qpair, srb_t *sp) cnt = qla2x00_debounce_register( ISP_REQ_Q_OUT(ha, ®->isp)); + if (!qpair->use_shadow_reg && cnt == ISP_REG16_DISCONNECT) { + qla_schedule_eeh_work(vha); + return NULL; + } + if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -2385,7 +2434,8 @@ qla24xx_prli_iocb(srb_t *sp, struct logio_entry_24xx *logio) cpu_to_le32(NVME_PRLI_SP_FIRST_BURST); if (sp->vha->flags.nvme2_enabled) { /* Set service parameter BIT_7 for NVME CONF support */ - logio->io_parameter[0] |= NVME_PRLI_SP_CONF; + logio->io_parameter[0] |= + cpu_to_le32(NVME_PRLI_SP_CONF); /* Set service parameter BIT_8 for SLER support */ logio->io_parameter[0] |= cpu_to_le32(NVME_PRLI_SP_SLER); @@ -3738,6 +3788,9 @@ qla2x00_start_sp(srb_t *sp) void *pkt; unsigned long flags; + if (vha->hw->flags.eeh_busy) + return -EIO; + spin_lock_irqsave(qp->qp_lock_ptr, flags); pkt = __qla2x00_alloc_iocbs(sp->qpair, sp); if (!pkt) { @@ -3959,8 +4012,14 @@ qla2x00_start_bidir(srb_t *sp, struct scsi_qla_host *vha, uint32_t tot_dsds) /* Check for room on request queue. */ if (req->cnt < req_cnt + 2) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } + if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; else @@ -3999,5 +4058,6 @@ qla2x00_start_bidir(srb_t *sp, struct scsi_qla_host *vha, uint32_t tot_dsds) qla2x00_start_iocbs(vha, req); queuing_error: spin_unlock_irqrestore(&ha->hardware_lock, flags); + return rval; } diff --git a/qla2x00t-32gbit/qla_isr.c b/qla2x00t-32gbit/qla_isr.c index 4f9d19d8b..443ed5bed 100644 --- a/qla2x00t-32gbit/qla_isr.c +++ b/qla2x00t-32gbit/qla_isr.c @@ -282,12 +282,7 @@ qla2x00_check_reg32_for_disconnect(scsi_qla_host_t *vha, uint32_t reg) if (!test_and_set_bit(PFLG_DISCONNECTED, &vha->pci_flags) && !test_bit(PFLG_DRIVER_REMOVING, &vha->pci_flags) && !test_bit(PFLG_DRIVER_PROBING, &vha->pci_flags)) { - /* - * Schedule this (only once) on the default system - * workqueue so that all the adapter workqueues and the - * DPC thread can be shutdown cleanly. - */ - schedule_work(&vha->hw->board_disable); + qla_schedule_eeh_work(vha); } return true; } else @@ -1669,8 +1664,6 @@ global_port_update: case MBA_TEMPERATURE_ALERT: ql_dbg(ql_dbg_async, vha, 0x505e, "TEMPERATURE ALERT: %04x %04x %04x\n", mb[1], mb[2], mb[3]); - if (mb[1] == 0x12) - schedule_work(&ha->board_disable); break; case MBA_TRANS_INSERT: @@ -3475,7 +3468,7 @@ qla24xx_abort_iocb_entry(scsi_qla_host_t *vha, struct req_que *req, return; abt = &sp->u.iocb_cmd; - abt->u.abt.comp_status = le16_to_cpu(pkt->comp_status); + abt->u.abt.comp_status = pkt->comp_status; orig_sp = sp->cmd_sp; /* Need to pass original sp */ if (orig_sp) @@ -4109,11 +4102,11 @@ msix_failed: if (USER_CTRL_IRQ(ha) || !ha->mqiobase) { /* user wants to control IRQ setting for target mode */ ret = pci_alloc_irq_vectors(ha->pdev, min_vecs, - min((u16)ha->msix_count, (u16)num_online_cpus()), + min((u16)ha->msix_count, (u16)(num_online_cpus() + min_vecs)), PCI_IRQ_MSIX); } else ret = pci_alloc_irq_vectors_affinity(ha->pdev, min_vecs, - min((u16)ha->msix_count, (u16)num_online_cpus()), + min((u16)ha->msix_count, (u16)(num_online_cpus() + min_vecs)), PCI_IRQ_MSIX | PCI_IRQ_AFFINITY, &desc); #endif diff --git a/qla2x00t-32gbit/qla_mbx.c b/qla2x00t-32gbit/qla_mbx.c index 06c99963b..0bcd8afdc 100644 --- a/qla2x00t-32gbit/qla_mbx.c +++ b/qla2x00t-32gbit/qla_mbx.c @@ -102,7 +102,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) int rval, i; unsigned long flags = 0; device_reg_t *reg; - uint8_t abort_active; + uint8_t abort_active, eeh_delay; uint8_t io_lock_on; uint16_t command = 0; uint16_t *iptr; @@ -136,7 +136,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) "PCI error, exiting.\n"); return QLA_FUNCTION_TIMEOUT; } - + eeh_delay = 0; reg = ha->iobase; io_lock_on = base_vha->flags.init_done; @@ -159,10 +159,10 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) } /* check if ISP abort is active and return cmd with timeout */ - if ((test_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags) || - test_bit(ISP_ABORT_RETRY, &base_vha->dpc_flags) || - test_bit(ISP_ABORT_NEEDED, &base_vha->dpc_flags)) && - !is_rom_cmd(mcp->mb[0])) { + if (((test_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags) || + test_bit(ISP_ABORT_RETRY, &base_vha->dpc_flags) || + test_bit(ISP_ABORT_NEEDED, &base_vha->dpc_flags)) && + !is_rom_cmd(mcp->mb[0])) || ha->flags.eeh_busy) { ql_log(ql_log_info, vha, 0x1005, "Cmd 0x%x aborted with timeout since ISP Abort is pending\n", mcp->mb[0]); @@ -185,7 +185,11 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) return QLA_FUNCTION_TIMEOUT; } atomic_dec(&ha->num_pend_mbx_stage1); - if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { + if (ha->flags.purge_mbox || chip_reset != ha->chip_reset || + ha->flags.eeh_busy) { + ql_log(ql_log_warn, vha, 0xd035, + "Error detected: purge[%d] eeh[%d] cmd=0x%x, Exiting.\n", + ha->flags.purge_mbox, ha->flags.eeh_busy, mcp->mb[0]); rval = QLA_ABORTED; goto premature_exit; } @@ -265,6 +269,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) if (!wait_for_completion_timeout(&ha->mbx_intr_comp, mcp->tov * HZ)) { if (chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ? 1 : 0; + spin_lock_irqsave(&ha->hardware_lock, flags); ha->flags.mbox_busy = 0; spin_unlock_irqrestore(&ha->hardware_lock, @@ -282,6 +288,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) } else if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ? 1 : 0; + spin_lock_irqsave(&ha->hardware_lock, flags); ha->flags.mbox_busy = 0; spin_unlock_irqrestore(&ha->hardware_lock, flags); @@ -323,6 +331,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) while (!ha->flags.mbox_int) { if (ha->flags.purge_mbox || chip_reset != ha->chip_reset) { + eeh_delay = ha->flags.eeh_busy ? 1 : 0; + spin_lock_irqsave(&ha->hardware_lock, flags); ha->flags.mbox_busy = 0; spin_unlock_irqrestore(&ha->hardware_lock, @@ -531,7 +541,8 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) clear_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); /* Allow next mbx cmd to come in. */ complete(&ha->mbx_cmd_comp); - if (ha->isp_ops->abort_isp(vha)) { + if (ha->isp_ops->abort_isp(vha) && + !ha->flags.eeh_busy) { /* Failed. retry later. */ set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); @@ -584,6 +595,17 @@ mbx_done: ql_dbg(ql_dbg_mbx, base_vha, 0x1021, "Done %s.\n", __func__); } + i = 500; + while (i && eeh_delay && (ha->pci_error_state < QLA_PCI_SLOT_RESET)) { + /* + * The caller of this mailbox encounter pci error. + * Hold the thread until PCIE link reset complete to make + * sure caller does not unmap dma while recovery is + * in progress. + */ + msleep(1); + i--; + } return rval; } diff --git a/qla2x00t-32gbit/qla_mr.c b/qla2x00t-32gbit/qla_mr.c index 608e06777..16d00e3e3 100644 --- a/qla2x00t-32gbit/qla_mr.c +++ b/qla2x00t-32gbit/qla_mr.c @@ -516,7 +516,7 @@ qlafx00_pci_config(scsi_qla_host_t *vha) } /** - * qlafx00_warm_reset() - Perform warm reset of iSA(CPUs being reset on SOC). + * qlafx00_soc_cpu_reset() - Perform warm reset of iSA(CPUs being reset on SOC). * @vha: HA context * */ @@ -2875,7 +2875,7 @@ qlafx00_async_event(scsi_qla_host_t *vha) } /** - * qlafx00x_mbx_completion() - Process mailbox command completions. + * qlafx00_mbx_completion() - Process mailbox command completions. * @vha: SCSI driver HA context * @mb0: value to be written into mailbox register 0 */ diff --git a/qla2x00t-32gbit/qla_nvme.c b/qla2x00t-32gbit/qla_nvme.c index f6f8e6738..726b185c1 100644 --- a/qla2x00t-32gbit/qla_nvme.c +++ b/qla2x00t-32gbit/qla_nvme.c @@ -411,8 +411,13 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp) } req_cnt = qla24xx_calc_iocbs(vha, tot_dsds); if (req->cnt < (req_cnt + 2)) { - cnt = IS_SHADOW_REG_CAPABLE(ha) ? *req->out_ptr : - rd_reg_dword_relaxed(req->req_q_out); + if (IS_SHADOW_REG_CAPABLE(ha)) { + cnt = *req->out_ptr; + } else { + cnt = rd_reg_dword_relaxed(req->req_q_out); + if (qla2x00_check_reg16_for_disconnect(vha, cnt)) + goto queuing_error; + } if (req->ring_index < cnt) req->cnt = cnt - req->ring_index; @@ -549,6 +554,7 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp) queuing_error: spin_unlock_irqrestore(&qpair->qp_lock, flags); + return rval; } diff --git a/qla2x00t-32gbit/qla_nx2.c b/qla2x00t-32gbit/qla_nx2.c index 68a16c95d..5ceecc964 100644 --- a/qla2x00t-32gbit/qla_nx2.c +++ b/qla2x00t-32gbit/qla_nx2.c @@ -2028,7 +2028,7 @@ exit_error: } /** - * qla4_8xxx_check_temp - Check the ISP82XX temperature. + * qla8044_check_temp - Check the ISP82XX temperature. * @vha: adapter block pointer. * * Note: The caller should not hold the idc lock. @@ -2226,19 +2226,16 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, if (opcode & QLA82XX_DBG_OPCODE_WR) { qla8044_wr_reg_indirect(vha, crb_addr, crb_entry->value_1); - opcode &= ~QLA82XX_DBG_OPCODE_WR; } if (opcode & QLA82XX_DBG_OPCODE_RW) { qla8044_rd_reg_indirect(vha, crb_addr, &read_value); qla8044_wr_reg_indirect(vha, crb_addr, read_value); - opcode &= ~QLA82XX_DBG_OPCODE_RW; } if (opcode & QLA82XX_DBG_OPCODE_AND) { qla8044_rd_reg_indirect(vha, crb_addr, &read_value); read_value &= crb_entry->value_2; - opcode &= ~QLA82XX_DBG_OPCODE_AND; if (opcode & QLA82XX_DBG_OPCODE_OR) { read_value |= crb_entry->value_3; opcode &= ~QLA82XX_DBG_OPCODE_OR; @@ -2249,7 +2246,6 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, qla8044_rd_reg_indirect(vha, crb_addr, &read_value); read_value |= crb_entry->value_3; qla8044_wr_reg_indirect(vha, crb_addr, read_value); - opcode &= ~QLA82XX_DBG_OPCODE_OR; } if (opcode & QLA82XX_DBG_OPCODE_POLL) { poll_time = crb_entry->crb_strd.poll_timeout; @@ -2269,7 +2265,6 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, crb_addr, &read_value); } } while (1); - opcode &= ~QLA82XX_DBG_OPCODE_POLL; } if (opcode & QLA82XX_DBG_OPCODE_RDSTATE) { @@ -2283,7 +2278,6 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, qla8044_rd_reg_indirect(vha, addr, &read_value); index = crb_entry->crb_ctrl.state_index_v; tmplt_hdr->saved_state_array[index] = read_value; - opcode &= ~QLA82XX_DBG_OPCODE_RDSTATE; } if (opcode & QLA82XX_DBG_OPCODE_WRSTATE) { @@ -2303,7 +2297,6 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, } qla8044_wr_reg_indirect(vha, addr, read_value); - opcode &= ~QLA82XX_DBG_OPCODE_WRSTATE; } if (opcode & QLA82XX_DBG_OPCODE_MDSTATE) { @@ -2316,7 +2309,6 @@ qla8044_minidump_process_control(struct scsi_qla_host *vha, read_value |= crb_entry->value_3; read_value += crb_entry->value_1; tmplt_hdr->saved_state_array[index] = read_value; - opcode &= ~QLA82XX_DBG_OPCODE_MDSTATE; } crb_addr += crb_entry->crb_strd.addr_stride; } diff --git a/qla2x00t-32gbit/qla_os.c b/qla2x00t-32gbit/qla_os.c index ba6b00b01..7fc147b68 100644 --- a/qla2x00t-32gbit/qla_os.c +++ b/qla2x00t-32gbit/qla_os.c @@ -1008,6 +1008,13 @@ qla2xxx_mqueuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd, goto qc24_fail_command; } + if (!qpair->online) { + ql_dbg(ql_dbg_io, vha, 0x3077, + "qpair not online. eeh_busy=%d.\n", ha->flags.eeh_busy); + cmd->result = DID_NO_CONNECT << 16; + goto qc24_fail_command; + } + if (!fcport || fcport->deleted) { cmd->result = DID_IMM_RETRY << 16; goto qc24_fail_command; @@ -1050,8 +1057,6 @@ qla2xxx_mqueuecommand(struct Scsi_Host *host, struct scsi_cmnd *cmd, if (rval != QLA_SUCCESS) { ql_dbg(ql_dbg_io + ql_dbg_verbose, vha, 0x3078, "Start scsi failed rval=%d for cmd=%p.\n", rval, cmd); - if (rval == QLA_INTERFACE_ERROR) - goto qc24_free_sp_fail_command; goto qc24_host_busy_free_sp; } @@ -1063,11 +1068,6 @@ qc24_host_busy_free_sp: qc24_target_busy: return SCSI_MLQUEUE_TARGET_BUSY; -qc24_free_sp_fail_command: - sp->free(sp); - CMD_SP(cmd) = NULL; - qla2xxx_rel_qpair_sp(sp->qpair, sp); - qc24_fail_command: cmd->scsi_done(cmd); @@ -1244,35 +1244,6 @@ qla2x00_wait_for_chip_reset(scsi_qla_host_t *vha) return return_status; } -#define ISP_REG_DISCONNECT 0xffffffffU -/************************************************************************** -* qla2x00_isp_reg_stat -* -* Description: -* Read the host status register of ISP before aborting the command. -* -* Input: -* ha = pointer to host adapter structure. -* -* -* Returns: -* Either true or false. -* -* Note: Return true if there is register disconnect. -**************************************************************************/ -static inline -uint32_t qla2x00_isp_reg_stat(struct qla_hw_data *ha) -{ - struct device_reg_24xx __iomem *reg = &ha->iobase->isp24; - struct device_reg_82xx __iomem *reg82 = &ha->iobase->isp82; - - if (IS_P3P_TYPE(ha)) - return ((rd_reg_dword(®82->host_int)) == ISP_REG_DISCONNECT); - else - return ((rd_reg_dword(®->host_status)) == - ISP_REG_DISCONNECT); -} - /************************************************************************** * qla2xxx_eh_abort * @@ -1306,6 +1277,7 @@ qla2xxx_eh_abort(struct scsi_cmnd *cmd) if (qla2x00_isp_reg_stat(ha)) { ql_log(ql_log_info, vha, 0x8042, "PCI/Register disconnect, exiting.\n"); + qla_pci_set_eeh_busy(vha); return FAILED; } @@ -1499,6 +1471,7 @@ qla2xxx_eh_device_reset(struct scsi_cmnd *cmd) if (qla2x00_isp_reg_stat(ha)) { ql_log(ql_log_info, vha, 0x803e, "PCI/Register disconnect, exiting.\n"); + qla_pci_set_eeh_busy(vha); return FAILED; } @@ -1515,6 +1488,7 @@ qla2xxx_eh_target_reset(struct scsi_cmnd *cmd) if (qla2x00_isp_reg_stat(ha)) { ql_log(ql_log_info, vha, 0x803f, "PCI/Register disconnect, exiting.\n"); + qla_pci_set_eeh_busy(vha); return FAILED; } @@ -1550,6 +1524,7 @@ qla2xxx_eh_bus_reset(struct scsi_cmnd *cmd) if (qla2x00_isp_reg_stat(ha)) { ql_log(ql_log_info, vha, 0x8040, "PCI/Register disconnect, exiting.\n"); + qla_pci_set_eeh_busy(vha); return FAILED; } @@ -1627,7 +1602,7 @@ qla2xxx_eh_host_reset(struct scsi_cmnd *cmd) if (qla2x00_isp_reg_stat(ha)) { ql_log(ql_log_info, vha, 0x8041, "PCI/Register disconnect, exiting.\n"); - schedule_work(&ha->board_disable); + qla_pci_set_eeh_busy(vha); return SUCCESS; } @@ -4286,11 +4261,10 @@ qla2x00_mem_alloc(struct qla_hw_data *ha, uint16_t req_len, uint16_t rsp_len, /* Get consistent memory allocated for Special Features-CB. */ if (IS_QLA27XX(ha) || IS_QLA28XX(ha)) { - ha->sf_init_cb = dma_pool_alloc(ha->s_dma_pool, GFP_KERNEL, + ha->sf_init_cb = dma_pool_zalloc(ha->s_dma_pool, GFP_KERNEL, &ha->sf_init_cb_dma); if (!ha->sf_init_cb) goto fail_sf_init_cb; - memset(ha->sf_init_cb, 0, sizeof(struct init_sf_cb)); ql_dbg_pci(ql_dbg_init, ha->pdev, 0x0199, "sf_init_cb=%p.\n", ha->sf_init_cb); } @@ -4692,8 +4666,7 @@ qla2x00_free_fw_dump(struct qla_hw_data *ha) dma_free_coherent(&ha->pdev->dev, EFT_SIZE, ha->eft, ha->eft_dma); - if (ha->fw_dump) - vfree(ha->fw_dump); + vfree(ha->fw_dump); ha->fce = NULL; ha->fce_dma = 0; @@ -4707,8 +4680,7 @@ qla2x00_free_fw_dump(struct qla_hw_data *ha) ha->fw_dump_len = 0; for (j = 0; j < 2; j++, fwdt++) { - if (fwdt->template) - vfree(fwdt->template); + vfree(fwdt->template); fwdt->template = NULL; fwdt->length = 0; } @@ -6728,6 +6700,9 @@ qla2x00_do_dpc(void *data) schedule(); + if (test_and_clear_bit(DO_EEH_RECOVERY, &base_vha->dpc_flags)) + qla_pci_set_eeh_busy(base_vha); + if (!base_vha->flags.init_done || ha->flags.mbox_busy) goto end_loop; @@ -7021,26 +6996,21 @@ intr_on_check: mutex_unlock(&ha->mq_lock); } - if (test_and_clear_bit(SET_NVME_ZIO_THRESHOLD_NEEDED, - &base_vha->dpc_flags)) { - ql_log(ql_log_info, base_vha, 0xffffff, - "nvme: SET ZIO Activity exchange threshold to %d.\n", - ha->nvme_last_rptd_aen); - if (qla27xx_set_zio_threshold(base_vha, - ha->nvme_last_rptd_aen)) { - ql_log(ql_log_info, base_vha, 0xffffff, - "nvme: Unable to SET ZIO Activity exchange threshold to %d.\n", - ha->nvme_last_rptd_aen); - } - } - if (test_and_clear_bit(SET_ZIO_THRESHOLD_NEEDED, - &base_vha->dpc_flags)) { + &base_vha->dpc_flags)) { + u16 threshold = ha->nvme_last_rptd_aen + ha->last_zio_threshold; + + if (threshold > ha->orig_fw_xcb_count) + threshold = ha->orig_fw_xcb_count; + ql_log(ql_log_info, base_vha, 0xffffff, - "SET ZIO Activity exchange threshold to %d.\n", - ha->last_zio_threshold); - qla27xx_set_zio_threshold(base_vha, - ha->last_zio_threshold); + "SET ZIO Activity exchange threshold to %d.\n", + threshold); + if (qla27xx_set_zio_threshold(base_vha, threshold)) { + ql_log(ql_log_info, base_vha, 0xffffff, + "Unable to SET ZIO Activity exchange threshold to %d.\n", + threshold); + } } if (!IS_QLAFX00(ha)) @@ -7270,14 +7240,13 @@ qla2x00_timer(struct timer_list *t) index = atomic_read(&ha->nvme_active_aen_cnt); if (!vha->vp_idx && (index != ha->nvme_last_rptd_aen) && - (index >= DEFAULT_ZIO_THRESHOLD) && ha->zio_mode == QLA_ZIO_MODE_6 && !ha->flags.host_shutting_down) { + ha->nvme_last_rptd_aen = atomic_read(&ha->nvme_active_aen_cnt); ql_log(ql_log_info, vha, 0x3002, "nvme: Sched: Set ZIO exchange threshold to %d.\n", ha->nvme_last_rptd_aen); - ha->nvme_last_rptd_aen = atomic_read(&ha->nvme_active_aen_cnt); - set_bit(SET_NVME_ZIO_THRESHOLD_NEEDED, &vha->dpc_flags); + set_bit(SET_ZIO_THRESHOLD_NEEDED, &vha->dpc_flags); start_dpc++; } @@ -7450,6 +7419,8 @@ static void qla_pci_error_cleanup(scsi_qla_host_t *vha) int i; unsigned long flags; + ql_dbg(ql_dbg_aer, vha, 0x9000, + "%s\n", __func__); ha->chip_reset++; ha->base_qpair->chip_reset = ha->chip_reset; @@ -7459,28 +7430,16 @@ static void qla_pci_error_cleanup(scsi_qla_host_t *vha) ha->base_qpair->chip_reset; } - /* purge MBox commands */ - if (atomic_read(&ha->num_pend_mbx_stage3)) { - clear_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags); - complete(&ha->mbx_intr_comp); - } - - i = 0; - - while (atomic_read(&ha->num_pend_mbx_stage3) || - atomic_read(&ha->num_pend_mbx_stage2) || - atomic_read(&ha->num_pend_mbx_stage1)) { - msleep(20); - i++; - if (i > 50) - break; - } - - ha->flags.purge_mbox = 0; + /* + * purge mailbox might take a while. Slot Reset/chip reset + * will take care of the purge + */ mutex_lock(&ha->mq_lock); + ha->base_qpair->online = 0; list_for_each_entry(qpair, &base_vha->qp_list, qp_list_elem) qpair->online = 0; + wmb(); mutex_unlock(&ha->mq_lock); qla2x00_mark_all_devices_lost(vha); @@ -7517,14 +7476,17 @@ qla2xxx_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state) { scsi_qla_host_t *vha = pci_get_drvdata(pdev); struct qla_hw_data *ha = vha->hw; + pci_ers_result_t ret = PCI_ERS_RESULT_NEED_RESET; - ql_dbg(ql_dbg_aer, vha, 0x9000, - "PCI error detected, state %x.\n", state); + ql_log(ql_log_warn, vha, 0x9000, + "PCI error detected, state %x.\n", state); + ha->pci_error_state = QLA_PCI_ERR_DETECTED; if (!atomic_read(&pdev->enable_cnt)) { ql_log(ql_log_info, vha, 0xffff, "PCI device is disabled,state %x\n", state); - return PCI_ERS_RESULT_NEED_RESET; + ret = PCI_ERS_RESULT_NEED_RESET; + goto out; } switch (state) { @@ -7534,11 +7496,12 @@ qla2xxx_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state) set_bit(QPAIR_ONLINE_CHECK_NEEDED, &vha->dpc_flags); qla2xxx_wake_dpc(vha); } - return PCI_ERS_RESULT_CAN_RECOVER; + ret = PCI_ERS_RESULT_CAN_RECOVER; + break; case pci_channel_io_frozen: - ha->flags.eeh_busy = 1; - qla_pci_error_cleanup(vha); - return PCI_ERS_RESULT_NEED_RESET; + qla_pci_set_eeh_busy(vha); + ret = PCI_ERS_RESULT_NEED_RESET; + break; case pci_channel_io_perm_failure: ha->flags.pci_channel_io_perm_failure = 1; qla2x00_abort_all_cmds(vha, DID_NO_CONNECT << 16); @@ -7546,9 +7509,12 @@ qla2xxx_pci_error_detected(struct pci_dev *pdev, pci_channel_state_t state) set_bit(QPAIR_ONLINE_CHECK_NEEDED, &vha->dpc_flags); qla2xxx_wake_dpc(vha); } - return PCI_ERS_RESULT_DISCONNECT; + ret = PCI_ERS_RESULT_DISCONNECT; } - return PCI_ERS_RESULT_NEED_RESET; +out: + ql_dbg(ql_dbg_aer, vha, 0x600d, + "PCI error detected returning [%x].\n", ret); + return ret; } static pci_ers_result_t @@ -7562,6 +7528,10 @@ qla2xxx_pci_mmio_enabled(struct pci_dev *pdev) struct device_reg_2xxx __iomem *reg = &ha->iobase->isp; struct device_reg_24xx __iomem *reg24 = &ha->iobase->isp24; + ql_log(ql_log_warn, base_vha, 0x9000, + "mmio enabled\n"); + + ha->pci_error_state = QLA_PCI_MMIO_ENABLED; if (IS_QLA82XX(ha)) return PCI_ERS_RESULT_RECOVERED; @@ -7585,10 +7555,11 @@ qla2xxx_pci_mmio_enabled(struct pci_dev *pdev) ql_log(ql_log_info, base_vha, 0x9003, "RISC paused -- mmio_enabled, Dumping firmware.\n"); qla2xxx_dump_fw(base_vha); - - return PCI_ERS_RESULT_NEED_RESET; - } else - return PCI_ERS_RESULT_RECOVERED; + } + /* set PCI_ERS_RESULT_NEED_RESET to trigger call to qla2xxx_pci_slot_reset */ + ql_dbg(ql_dbg_aer, base_vha, 0x600d, + "mmio enabled returning.\n"); + return PCI_ERS_RESULT_NEED_RESET; } static pci_ers_result_t @@ -7600,9 +7571,10 @@ qla2xxx_pci_slot_reset(struct pci_dev *pdev) int rc; struct qla_qpair *qpair = NULL; - ql_dbg(ql_dbg_aer, base_vha, 0x9004, - "Slot Reset.\n"); + ql_log(ql_log_warn, base_vha, 0x9004, + "Slot Reset.\n"); + ha->pci_error_state = QLA_PCI_SLOT_RESET; /* Workaround: qla2xxx driver which access hardware earlier * needs error state to be pci_channel_io_online. * Otherwise mailbox command timesout. @@ -7636,16 +7608,24 @@ qla2xxx_pci_slot_reset(struct pci_dev *pdev) qpair->online = 1; mutex_unlock(&ha->mq_lock); + ha->flags.eeh_busy = 0; base_vha->flags.online = 1; set_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags); - if (ha->isp_ops->abort_isp(base_vha) == QLA_SUCCESS) - ret = PCI_ERS_RESULT_RECOVERED; + ha->isp_ops->abort_isp(base_vha); clear_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags); + if (qla2x00_isp_reg_stat(ha)) { + ha->flags.eeh_busy = 1; + qla_pci_error_cleanup(base_vha); + ql_log(ql_log_warn, base_vha, 0x9005, + "Device unable to recover from PCI error.\n"); + } else { + ret = PCI_ERS_RESULT_RECOVERED; + } exit_slot_reset: ql_dbg(ql_dbg_aer, base_vha, 0x900e, - "slot_reset return %x.\n", ret); + "Slot Reset returning %x.\n", ret); return ret; } @@ -7657,16 +7637,55 @@ qla2xxx_pci_resume(struct pci_dev *pdev) struct qla_hw_data *ha = base_vha->hw; int ret; - ql_dbg(ql_dbg_aer, base_vha, 0x900f, - "pci_resume.\n"); + ql_log(ql_log_warn, base_vha, 0x900f, + "Pci Resume.\n"); - ha->flags.eeh_busy = 0; ret = qla2x00_wait_for_hba_online(base_vha); if (ret != QLA_SUCCESS) { ql_log(ql_log_fatal, base_vha, 0x9002, "The device failed to resume I/O from slot/link_reset.\n"); } + ha->pci_error_state = QLA_PCI_RESUME; + ql_dbg(ql_dbg_aer, base_vha, 0x600d, + "Pci Resume returning.\n"); +} + +void qla_pci_set_eeh_busy(struct scsi_qla_host *vha) +{ + struct qla_hw_data *ha = vha->hw; + struct scsi_qla_host *base_vha = pci_get_drvdata(ha->pdev); + bool do_cleanup = false; + unsigned long flags; + + if (ha->flags.eeh_busy) + return; + + spin_lock_irqsave(&base_vha->work_lock, flags); + if (!ha->flags.eeh_busy) { + ha->flags.eeh_busy = 1; + do_cleanup = true; + } + spin_unlock_irqrestore(&base_vha->work_lock, flags); + + if (do_cleanup) + qla_pci_error_cleanup(base_vha); +} + +/* + * this routine will schedule a task to pause IO from interrupt context + * if caller sees a PCIE error event (register read = 0xf's) + */ +void qla_schedule_eeh_work(struct scsi_qla_host *vha) +{ + struct qla_hw_data *ha = vha->hw; + struct scsi_qla_host *base_vha = pci_get_drvdata(ha->pdev); + + if (ha->flags.eeh_busy) + return; + + set_bit(DO_EEH_RECOVERY, &base_vha->dpc_flags); + qla2xxx_wake_dpc(base_vha); } #if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0) @@ -7774,6 +7793,7 @@ struct scsi_host_template qla2xxx_driver_template = { .eh_timed_out = fc_eh_timed_out, #endif .eh_abort_handler = qla2xxx_eh_abort, + .eh_should_retry_cmd = fc_eh_should_retry_cmd, .eh_device_reset_handler = qla2xxx_eh_device_reset, .eh_target_reset_handler = qla2xxx_eh_target_reset, .eh_bus_reset_handler = qla2xxx_eh_bus_reset, diff --git a/qla2x00t-32gbit/qla_sup.c b/qla2x00t-32gbit/qla_sup.c index f771fabcb..060c89237 100644 --- a/qla2x00t-32gbit/qla_sup.c +++ b/qla2x00t-32gbit/qla_sup.c @@ -2621,10 +2621,11 @@ qla24xx_read_optrom_data(struct scsi_qla_host *vha, void *buf, } static int -qla28xx_extract_sfub_and_verify(struct scsi_qla_host *vha, uint32_t *buf, +qla28xx_extract_sfub_and_verify(struct scsi_qla_host *vha, __le32 *buf, uint32_t len, uint32_t buf_size_without_sfub, uint8_t *sfub_buf) { - uint32_t *p, check_sum = 0; + uint32_t check_sum = 0; + __le32 *p; int i; p = buf + buf_size_without_sfub; @@ -2790,8 +2791,8 @@ qla28xx_write_flash_data(scsi_qla_host_t *vha, uint32_t *dwptr, uint32_t faddr, goto done; } - rval = qla28xx_extract_sfub_and_verify(vha, dwptr, dwords, - buf_size_without_sfub, (uint8_t *)sfub); + rval = qla28xx_extract_sfub_and_verify(vha, (__le32 *)dwptr, + dwords, buf_size_without_sfub, (uint8_t *)sfub); if (rval != QLA_SUCCESS) goto done; diff --git a/qla2x00t-32gbit/qla_target.c b/qla2x00t-32gbit/qla_target.c index 8ba4e207c..659bdbb6e 100644 --- a/qla2x00t-32gbit/qla_target.c +++ b/qla2x00t-32gbit/qla_target.c @@ -1044,7 +1044,12 @@ void qlt_free_session_done(struct work_struct *work) } msleep(100); cnt++; - if (cnt > 200) + /* + * Driver timeout is set to 22 Sec, update count value to loop + * long enough for log-out to complete before advancing. Otherwise, + * straddling logout can interfere with re-login attempt. + */ + if (cnt > 230) break; } @@ -6527,7 +6532,7 @@ static void qlt_lport_dump(struct scsi_qla_host *vha, u64 wwpn, } /** - * qla_tgt_lport_register - register lport with external module + * qlt_lport_register - register lport with external module * * @target_lport_ptr: pointer for tcm_qla2xxx specific lport data * @phys_wwpn: physical port WWPN @@ -6603,7 +6608,7 @@ int qlt_lport_register(void *target_lport_ptr, u64 phys_wwpn, EXPORT_SYMBOL(qlt_lport_register); /** - * qla_tgt_lport_deregister - Degister lport + * qlt_lport_deregister - Degister lport * * @vha: Registered scsi_qla_host pointer */ diff --git a/qla2x00t-32gbit/qla_target.h b/qla2x00t-32gbit/qla_target.h index 8afdc4b64..aaf8b6216 100644 --- a/qla2x00t-32gbit/qla_target.h +++ b/qla2x00t-32gbit/qla_target.h @@ -143,7 +143,6 @@ (min(1270, ((ql) > 0) ? (QLA_TGT_DATASEGS_PER_CMD_24XX + \ QLA_TGT_DATASEGS_PER_CONT_24XX*((ql) - 1)) : 0)) #endif -#endif #define GET_TARGET_ID(ha, iocb) ((HAS_EXTENDED_IDS(ha)) \ ? le16_to_cpu((iocb)->u.isp2x.target.extended) \ @@ -271,6 +270,7 @@ struct ctio_to_2xxx { #ifndef CTIO_RET_TYPE #define CTIO_RET_TYPE 0x17 /* CTIO return entry */ #define ATIO_TYPE7 0x06 /* Accept target I/O entry for 24xx */ +#endif struct fcp_hdr { uint8_t r_ctl; diff --git a/qla2x00t-32gbit/qla_version.h b/qla2x00t-32gbit/qla_version.h index 72c648442..da11829fa 100644 --- a/qla2x00t-32gbit/qla_version.h +++ b/qla2x00t-32gbit/qla_version.h @@ -6,9 +6,9 @@ /* * Driver version */ -#define QLA2XXX_VERSION "10.02.00.105-k" +#define QLA2XXX_VERSION "10.02.00.106-k" #define QLA_DRIVER_MAJOR_VER 10 #define QLA_DRIVER_MINOR_VER 2 #define QLA_DRIVER_PATCH_VER 0 -#define QLA_DRIVER_BETA_VER 105 +#define QLA_DRIVER_BETA_VER 106