diff options
author | Saurav Kashyap <skashyap@marvell.com> | 2021-01-11 01:31:28 -0800 |
---|---|---|
committer | Martin K. Petersen <martin.petersen@oracle.com> | 2021-01-13 00:25:19 -0500 |
commit | dbf1f53cfd238090c69f92725b91208b97eb53fe (patch) | |
tree | d46e4bcfebe1e440572a2071d26541dfb8f9a697 | |
parent | ac341c2d2f1f72bc48e7950bb648466d015e657d (diff) |
scsi: qla2xxx: Implementation to get and manage host, target stats and initiator port
This statistics will help in debugging process and checking specific error
counts. It also provides a capability to isolate the port or bring it out
of isolation.
Link: https://lore.kernel.org/r/20210111093134.1206-2-njavali@marvell.com
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
-rw-r--r-- | drivers/scsi/qla2xxx/qla_attr.c | 9 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_bsg.c | 342 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_bsg.h | 5 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_def.h | 71 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_gbl.h | 23 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_gs.c | 1 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_init.c | 216 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_isr.c | 22 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_mbx.c | 9 | ||||
-rw-r--r-- | drivers/scsi/qla2xxx/qla_os.c | 20 |
10 files changed, 718 insertions, 0 deletions
diff --git a/drivers/scsi/qla2xxx/qla_attr.c b/drivers/scsi/qla2xxx/qla_attr.c index ab45ac1e5a72..63391c9be05d 100644 --- a/drivers/scsi/qla2xxx/qla_attr.c +++ b/drivers/scsi/qla2xxx/qla_attr.c @@ -710,6 +710,12 @@ qla2x00_sysfs_write_reset(struct file *filp, struct kobject *kobj, ql_log(ql_log_info, vha, 0x706e, "Issuing ISP reset.\n"); + if (vha->hw->flags.port_isolated) { + ql_log(ql_log_info, vha, 0x706e, + "Port is isolated, returning.\n"); + return -EINVAL; + } + scsi_block_requests(vha->host); if (IS_QLA82XX(ha)) { ha->flags.isp82xx_no_md_cap = 1; @@ -2717,6 +2723,9 @@ qla2x00_issue_lip(struct Scsi_Host *shost) if (IS_QLAFX00(vha->hw)) return 0; + if (vha->hw->flags.port_isolated) + return 0; + qla2x00_loop_reset(vha); return 0; } diff --git a/drivers/scsi/qla2xxx/qla_bsg.c b/drivers/scsi/qla2xxx/qla_bsg.c index 23b604832a54..e45da05383cd 100644 --- a/drivers/scsi/qla2xxx/qla_bsg.c +++ b/drivers/scsi/qla2xxx/qla_bsg.c @@ -4,6 +4,7 @@ * Copyright (c) 2003-2014 QLogic Corporation */ #include "qla_def.h" +#include "qla_gbl.h" #include <linux/kthread.h> #include <linux/vmalloc.h> @@ -2445,6 +2446,323 @@ qla2x00_get_flash_image_status(struct bsg_job *bsg_job) } static int +qla2x00_manage_host_stats(struct bsg_job *bsg_job) +{ + scsi_qla_host_t *vha = shost_priv(fc_bsg_to_shost(bsg_job)); + struct fc_bsg_reply *bsg_reply = bsg_job->reply; + struct ql_vnd_mng_host_stats_param *req_data; + struct ql_vnd_mng_host_stats_resp rsp_data; + u32 req_data_len; + int ret = 0; + + if (!vha->flags.online) { + ql_log(ql_log_warn, vha, 0x0000, "Host is not online.\n"); + return -EIO; + } + + req_data_len = bsg_job->request_payload.payload_len; + + if (req_data_len != sizeof(struct ql_vnd_mng_host_stats_param)) { + ql_log(ql_log_warn, vha, 0x0000, "req_data_len invalid.\n"); + return -EIO; + } + + req_data = kzalloc(sizeof(*req_data), GFP_KERNEL); + if (!req_data) { + ql_log(ql_log_warn, vha, 0x0000, "req_data memory allocation failure.\n"); + return -ENOMEM; + } + + /* Copy the request buffer in req_data */ + sg_copy_to_buffer(bsg_job->request_payload.sg_list, + bsg_job->request_payload.sg_cnt, req_data, + req_data_len); + + switch (req_data->action) { + case QLA_STOP: + ret = qla2xxx_stop_stats(vha->host, req_data->stat_type); + break; + case QLA_START: + ret = qla2xxx_start_stats(vha->host, req_data->stat_type); + break; + case QLA_CLEAR: + ret = qla2xxx_reset_stats(vha->host, req_data->stat_type); + break; + default: + ql_log(ql_log_warn, vha, 0x0000, "Invalid action.\n"); + ret = -EIO; + break; + } + + kfree(req_data); + + /* Prepare response */ + rsp_data.status = ret; + bsg_job->reply_payload.payload_len = sizeof(struct ql_vnd_mng_host_stats_resp); + + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_OK; + bsg_reply->reply_payload_rcv_len = + sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, + &rsp_data, + sizeof(struct ql_vnd_mng_host_stats_resp)); + + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + + return ret; +} + +static int +qla2x00_get_host_stats(struct bsg_job *bsg_job) +{ + scsi_qla_host_t *vha = shost_priv(fc_bsg_to_shost(bsg_job)); + struct fc_bsg_reply *bsg_reply = bsg_job->reply; + struct ql_vnd_stats_param *req_data; + struct ql_vnd_host_stats_resp rsp_data; + u32 req_data_len; + int ret = 0; + u64 ini_entry_count = 0; + u64 entry_count = 0; + u64 tgt_num = 0; + u64 tmp_stat_type = 0; + u64 response_len = 0; + void *data; + + req_data_len = bsg_job->request_payload.payload_len; + + if (req_data_len != sizeof(struct ql_vnd_stats_param)) { + ql_log(ql_log_warn, vha, 0x0000, "req_data_len invalid.\n"); + return -EIO; + } + + req_data = kzalloc(sizeof(*req_data), GFP_KERNEL); + if (!req_data) { + ql_log(ql_log_warn, vha, 0x0000, "req_data memory allocation failure.\n"); + return -ENOMEM; + } + + /* Copy the request buffer in req_data */ + sg_copy_to_buffer(bsg_job->request_payload.sg_list, + bsg_job->request_payload.sg_cnt, req_data, req_data_len); + + /* Copy stat type to work on it */ + tmp_stat_type = req_data->stat_type; + + if (tmp_stat_type & QLA2XX_TGT_SHT_LNK_DOWN) { + /* Num of tgts connected to this host */ + tgt_num = qla2x00_get_num_tgts(vha); + /* unset BIT_17 */ + tmp_stat_type &= ~(1 << 17); + } + + /* Total ini stats */ + ini_entry_count = qla2x00_count_set_bits(tmp_stat_type); + + /* Total number of entries */ + entry_count = ini_entry_count + tgt_num; + + response_len = sizeof(struct ql_vnd_host_stats_resp) + + (sizeof(struct ql_vnd_stat_entry) * entry_count); + + if (response_len > bsg_job->reply_payload.payload_len) { + rsp_data.status = EXT_STATUS_BUFFER_TOO_SMALL; + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_BUFFER_TOO_SMALL; + bsg_job->reply_payload.payload_len = sizeof(struct ql_vnd_mng_host_stats_resp); + + bsg_reply->reply_payload_rcv_len = + sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, &rsp_data, + sizeof(struct ql_vnd_mng_host_stats_resp)); + + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + goto host_stat_out; + } + + data = kzalloc(response_len, GFP_KERNEL); + + ret = qla2xxx_get_ini_stats(fc_bsg_to_shost(bsg_job), req_data->stat_type, + data, response_len); + + rsp_data.status = EXT_STATUS_OK; + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_OK; + + bsg_reply->reply_payload_rcv_len = sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, + data, response_len); + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + + kfree(data); +host_stat_out: + kfree(req_data); + return ret; +} + +static struct fc_rport * +qla2xxx_find_rport(scsi_qla_host_t *vha, uint32_t tgt_num) +{ + fc_port_t *fcport = NULL; + + list_for_each_entry(fcport, &vha->vp_fcports, list) { + if (fcport->rport->number == tgt_num) + return fcport->rport; + } + return NULL; +} + +static int +qla2x00_get_tgt_stats(struct bsg_job *bsg_job) +{ + scsi_qla_host_t *vha = shost_priv(fc_bsg_to_shost(bsg_job)); + struct fc_bsg_reply *bsg_reply = bsg_job->reply; + struct ql_vnd_tgt_stats_param *req_data; + u32 req_data_len; + int ret = 0; + u64 response_len = 0; + struct ql_vnd_tgt_stats_resp *data = NULL; + struct fc_rport *rport = NULL; + + if (!vha->flags.online) { + ql_log(ql_log_warn, vha, 0x0000, "Host is not online.\n"); + return -EIO; + } + + req_data_len = bsg_job->request_payload.payload_len; + + if (req_data_len != sizeof(struct ql_vnd_stat_entry)) { + ql_log(ql_log_warn, vha, 0x0000, "req_data_len invalid.\n"); + return -EIO; + } + + req_data = kzalloc(sizeof(*req_data), GFP_KERNEL); + if (!req_data) { + ql_log(ql_log_warn, vha, 0x0000, "req_data memory allocation failure.\n"); + return -ENOMEM; + } + + /* Copy the request buffer in req_data */ + sg_copy_to_buffer(bsg_job->request_payload.sg_list, + bsg_job->request_payload.sg_cnt, + req_data, req_data_len); + + response_len = sizeof(struct ql_vnd_tgt_stats_resp) + + sizeof(struct ql_vnd_stat_entry); + + /* structure + size for one entry */ + data = kzalloc(response_len, GFP_KERNEL); + if (!data) { + kfree(req_data); + return -ENOMEM; + } + + if (response_len > bsg_job->reply_payload.payload_len) { + data->status = EXT_STATUS_BUFFER_TOO_SMALL; + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_BUFFER_TOO_SMALL; + bsg_job->reply_payload.payload_len = sizeof(struct ql_vnd_mng_host_stats_resp); + + bsg_reply->reply_payload_rcv_len = + sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, &data, + sizeof(struct ql_vnd_tgt_stats_resp)); + + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + goto tgt_stat_out; + } + + rport = qla2xxx_find_rport(vha, req_data->tgt_id); + if (!rport) { + ql_log(ql_log_warn, vha, 0x0000, "target %d not found.\n", req_data->tgt_id); + ret = EXT_STATUS_INVALID_PARAM; + data->status = EXT_STATUS_INVALID_PARAM; + goto reply; + } + + ret = qla2xxx_get_tgt_stats(fc_bsg_to_shost(bsg_job), req_data->stat_type, + rport, (void *)data, response_len); + + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_OK; +reply: + bsg_reply->reply_payload_rcv_len = + sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, data, + response_len); + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + +tgt_stat_out: + kfree(data); + kfree(req_data); + + return ret; +} + +static int +qla2x00_manage_host_port(struct bsg_job *bsg_job) +{ + scsi_qla_host_t *vha = shost_priv(fc_bsg_to_shost(bsg_job)); + struct fc_bsg_reply *bsg_reply = bsg_job->reply; + struct ql_vnd_mng_host_port_param *req_data; + struct ql_vnd_mng_host_port_resp rsp_data; + u32 req_data_len; + int ret = 0; + + req_data_len = bsg_job->request_payload.payload_len; + + if (req_data_len != sizeof(struct ql_vnd_mng_host_port_param)) { + ql_log(ql_log_warn, vha, 0x0000, "req_data_len invalid.\n"); + return -EIO; + } + + req_data = kzalloc(sizeof(*req_data), GFP_KERNEL); + if (!req_data) { + ql_log(ql_log_warn, vha, 0x0000, "req_data memory allocation failure.\n"); + return -ENOMEM; + } + + /* Copy the request buffer in req_data */ + sg_copy_to_buffer(bsg_job->request_payload.sg_list, + bsg_job->request_payload.sg_cnt, req_data, req_data_len); + + switch (req_data->action) { + case QLA_ENABLE: + ret = qla2xxx_enable_port(vha->host); + break; + case QLA_DISABLE: + ret = qla2xxx_disable_port(vha->host); + break; + default: + ql_log(ql_log_warn, vha, 0x0000, "Invalid action.\n"); + ret = -EIO; + break; + } + + kfree(req_data); + + /* Prepare response */ + rsp_data.status = ret; + bsg_reply->reply_data.vendor_reply.vendor_rsp[0] = EXT_STATUS_OK; + bsg_job->reply_payload.payload_len = sizeof(struct ql_vnd_mng_host_port_resp); + + bsg_reply->reply_payload_rcv_len = + sg_copy_from_buffer(bsg_job->reply_payload.sg_list, + bsg_job->reply_payload.sg_cnt, &rsp_data, + sizeof(struct ql_vnd_mng_host_port_resp)); + bsg_reply->result = DID_OK; + bsg_job_done(bsg_job, bsg_reply->result, + bsg_reply->reply_payload_rcv_len); + + return ret; +} + +static int qla2x00_process_vendor_specific(struct bsg_job *bsg_job) { struct fc_bsg_request *bsg_request = bsg_job->request; @@ -2520,6 +2838,18 @@ qla2x00_process_vendor_specific(struct bsg_job *bsg_job) case QL_VND_SS_GET_FLASH_IMAGE_STATUS: return qla2x00_get_flash_image_status(bsg_job); + case QL_VND_MANAGE_HOST_STATS: + return qla2x00_manage_host_stats(bsg_job); + + case QL_VND_GET_HOST_STATS: + return qla2x00_get_host_stats(bsg_job); + + case QL_VND_GET_TGT_STATS: + return qla2x00_get_tgt_stats(bsg_job); + + case QL_VND_MANAGE_HOST_PORT: + return qla2x00_manage_host_port(bsg_job); + default: return -ENOSYS; } @@ -2547,6 +2877,17 @@ qla24xx_bsg_request(struct bsg_job *bsg_job) vha = shost_priv(host); } + /* Disable port will bring down the chip, allow enable command */ + if (bsg_request->rqst_data.h_vendor.vendor_cmd[0] == QL_VND_MANAGE_HOST_PORT || + bsg_request->rqst_data.h_vendor.vendor_cmd[0] == QL_VND_GET_HOST_STATS) + goto skip_chip_chk; + + if (vha->hw->flags.port_isolated) { + bsg_reply->result = DID_ERROR; + /* operation not permitted */ + return -EPERM; + } + if (qla2x00_chip_is_down(vha)) { ql_dbg(ql_dbg_user, vha, 0x709f, "BSG: ISP abort active/needed -- cmd=%d.\n", @@ -2554,6 +2895,7 @@ qla24xx_bsg_request(struct bsg_job *bsg_job) return -EBUSY; } +skip_chip_chk: ql_dbg(ql_dbg_user, vha, 0x7000, "Entered %s msgcode=0x%x.\n", __func__, bsg_request->msgcode); diff --git a/drivers/scsi/qla2xxx/qla_bsg.h b/drivers/scsi/qla2xxx/qla_bsg.h index 1a09b5512267..0274e99e4a12 100644 --- a/drivers/scsi/qla2xxx/qla_bsg.h +++ b/drivers/scsi/qla2xxx/qla_bsg.h @@ -31,6 +31,10 @@ #define QL_VND_DPORT_DIAGNOSTICS 0x19 #define QL_VND_GET_PRIV_STATS_EX 0x1A #define QL_VND_SS_GET_FLASH_IMAGE_STATUS 0x1E +#define QL_VND_MANAGE_HOST_STATS 0x23 +#define QL_VND_GET_HOST_STATS 0x24 +#define QL_VND_GET_TGT_STATS 0x25 +#define QL_VND_MANAGE_HOST_PORT 0x26 /* BSG Vendor specific subcode returns */ #define EXT_STATUS_OK 0 @@ -40,6 +44,7 @@ #define EXT_STATUS_DATA_OVERRUN 7 #define EXT_STATUS_DATA_UNDERRUN 8 #define EXT_STATUS_MAILBOX 11 +#define EXT_STATUS_BUFFER_TOO_SMALL 16 #define EXT_STATUS_NO_MEMORY 17 #define EXT_STATUS_DEVICE_OFFLINE 22 diff --git a/drivers/scsi/qla2xxx/qla_def.h b/drivers/scsi/qla2xxx/qla_def.h index 30c7e5e63851..ca67be8b62ec 100644 --- a/drivers/scsi/qla2xxx/qla_def.h +++ b/drivers/scsi/qla2xxx/qla_def.h @@ -2557,6 +2557,10 @@ typedef struct fc_port { u16 n2n_chip_reset; struct dentry *dfs_rport_dir; + + u64 tgt_short_link_down_cnt; + u64 tgt_link_down_time; + u64 dev_loss_tmo; } fc_port_t; enum { @@ -3922,6 +3926,7 @@ struct qla_hw_data { uint32_t scm_enabled:1; uint32_t max_req_queue_warned:1; uint32_t plogi_template_valid:1; + uint32_t port_isolated:1; } flags; uint16_t max_exchg; @@ -4851,6 +4856,13 @@ typedef struct scsi_qla_host { uint8_t scm_fabric_connection_flags; unsigned int irq_offset; + + u64 hw_err_cnt; + u64 interface_err_cnt; + u64 cmd_timeout_cnt; + u64 reset_cmd_err_cnt; + u64 link_down_time; + u64 short_link_down_cnt; } scsi_qla_host_t; struct qla27xx_image_status { @@ -5174,6 +5186,65 @@ struct sff_8247_a0 { #define PRLI_PHASE(_cls) \ ((_cls == DSC_LS_PRLI_PEND) || (_cls == DSC_LS_PRLI_COMP)) +enum ql_vnd_host_stat_action { + QLA_STOP = 0, + QLA_START, + QLA_CLEAR, +}; + +struct ql_vnd_mng_host_stats_param { + u32 stat_type; + enum ql_vnd_host_stat_action action; +} __packed; + +struct ql_vnd_mng_host_stats_resp { + u32 status; +} __packed; + +struct ql_vnd_stats_param { + u32 stat_type; +} __packed; + +struct ql_vnd_tgt_stats_param { + s32 tgt_id; + u32 stat_type; +} __packed; + +enum ql_vnd_host_port_action { + QLA_ENABLE = 0, + QLA_DISABLE, +}; + +struct ql_vnd_mng_host_port_param { + enum ql_vnd_host_port_action action; +} __packed; + +struct ql_vnd_mng_host_port_resp { + u32 status; +} __packed; + +struct ql_vnd_stat_entry { + u32 stat_type; /* Failure type */ + u32 tgt_num; /* Target Num */ + u64 cnt; /* Counter value */ +} __packed; + +struct ql_vnd_stats { + u64 entry_count; /* Num of entries */ + u64 rservd; + struct ql_vnd_stat_entry entry[0]; /* Place holder of entries */ +} __packed; + +struct ql_vnd_host_stats_resp { + u32 status; + struct ql_vnd_stats stats; +} __packed; + +struct ql_vnd_tgt_stats_resp { + u32 status; + struct ql_vnd_stats stats; +} __packed; + #include "qla_target.h" #include "qla_gbl.h" #include "qla_dbg.h" diff --git a/drivers/scsi/qla2xxx/qla_gbl.h b/drivers/scsi/qla2xxx/qla_gbl.h index e39b4f2da73a..708f82311b83 100644 --- a/drivers/scsi/qla2xxx/qla_gbl.h +++ b/drivers/scsi/qla2xxx/qla_gbl.h @@ -945,4 +945,27 @@ extern void qla2x00_dfs_remove_rport(scsi_qla_host_t *vha, struct fc_port *fp); /* nvme.c */ void qla_nvme_unregister_remote_port(struct fc_port *fcport); void qla_handle_els_plogi_done(scsi_qla_host_t *vha, struct event_arg *ea); + +#define QLA2XX_HW_ERROR BIT_0 +#define QLA2XX_SHT_LNK_DWN BIT_1 +#define QLA2XX_INT_ERR BIT_2 +#define QLA2XX_CMD_TIMEOUT BIT_3 +#define QLA2XX_RESET_CMD_ERR BIT_4 +#define QLA2XX_TGT_SHT_LNK_DOWN BIT_17 + +#define QLA2XX_MAX_LINK_DOWN_TIME 100 + +int qla2xxx_start_stats(struct Scsi_Host *shost, u32 flags); +int qla2xxx_stop_stats(struct Scsi_Host *shost, u32 flags); +int qla2xxx_reset_stats(struct Scsi_Host *shost, u32 flags); + +int qla2xxx_get_ini_stats(struct Scsi_Host *shost, u32 flags, void *data, u64 size); +int qla2xxx_get_tgt_stats(struct Scsi_Host *shost, u32 flags, + struct fc_rport *rport, void *data, u64 size); +int qla2xxx_disable_port(struct Scsi_Host *shost); +int qla2xxx_enable_port(struct Scsi_Host *shost); + +uint64_t qla2x00_get_num_tgts(scsi_qla_host_t *vha); +uint64_t qla2x00_count_set_bits(u32 num); + #endif /* _QLA_GBL_H */ diff --git a/drivers/scsi/qla2xxx/qla_gs.c b/drivers/scsi/qla2xxx/qla_gs.c index 391ac75e3de3..517d358b0031 100644 --- a/drivers/scsi/qla2xxx/qla_gs.c +++ b/drivers/scsi/qla2xxx/qla_gs.c @@ -3563,6 +3563,7 @@ login_logout: __func__, __LINE__, fcport->port_name); + fcport->tgt_link_down_time = 0; qlt_schedule_sess_for_deletion(fcport); continue; } diff --git a/drivers/scsi/qla2xxx/qla_init.c b/drivers/scsi/qla2xxx/qla_init.c index dcc0f0d823db..665a35810270 100644 --- a/drivers/scsi/qla2xxx/qla_init.c +++ b/drivers/scsi/qla2xxx/qla_init.c @@ -4993,6 +4993,9 @@ qla2x00_alloc_fcport(scsi_qla_host_t *vha, gfp_t flags) fcport->login_retry = vha->hw->login_retry_count; fcport->chip_reset = vha->hw->base_qpair->chip_reset; fcport->logout_on_delete = 1; + fcport->tgt_link_down_time = QLA2XX_MAX_LINK_DOWN_TIME; + fcport->tgt_short_link_down_cnt = 0; + fcport->dev_loss_tmo = 0; if (!fcport->ct_desc.ct_sns) { ql_log(ql_log_warn, vha, 0xd049, @@ -5490,6 +5493,7 @@ qla2x00_reg_remote_port(scsi_qla_host_t *vha, fc_port_t *fcport) spin_lock_irqsave(fcport->vha->host->host_lock, flags); *((fc_port_t **)rport->dd_data) = fcport; spin_unlock_irqrestore(fcport->vha->host->host_lock, flags); + fcport->dev_loss_tmo = rport->dev_loss_tmo; rport->supported_classes = fcport->supported_classes; @@ -5548,6 +5552,11 @@ qla2x00_update_fcport(scsi_qla_host_t *vha, fc_port_t *fcport) fcport->logout_on_delete = 1; fcport->n2n_chip_reset = fcport->n2n_link_reset_cnt = 0; + if (fcport->tgt_link_down_time < fcport->dev_loss_tmo) { + fcport->tgt_short_link_down_cnt++; + fcport->tgt_link_down_time = QLA2XX_MAX_LINK_DOWN_TIME; + } + switch (vha->hw->current_topology) { case ISP_CFG_N: case ISP_CFG_NL: @@ -6908,6 +6917,9 @@ qla2x00_abort_isp(scsi_qla_host_t *vha) if (vha->flags.online) { qla2x00_abort_isp_cleanup(vha); + if (vha->hw->flags.port_isolated) + return status; + if (test_and_clear_bit(ISP_ABORT_TO_ROM, &vha->dpc_flags)) { ha->flags.chip_reset_done = 1; vha->flags.online = 1; @@ -7029,6 +7041,11 @@ qla2x00_abort_isp(scsi_qla_host_t *vha) } + if (vha->hw->flags.port_isolated) { + qla2x00_abort_isp_cleanup(vha); + return status; + } + if (!status) { ql_dbg(ql_dbg_taskm, vha, 0x8022, "%s succeeded.\n", __func__); qla2x00_configure_hba(vha); @@ -9171,3 +9188,202 @@ int qla2xxx_delete_qpair(struct scsi_qla_host *vha, struct qla_qpair *qpair) fail: return ret; } + +uint64_t +qla2x00_count_set_bits(uint32_t num) +{ + /* Brian Kernighan's Algorithm */ + u64 count = 0; + + while (num) { + num &= (num - 1); + count++; + } + return count; +} + +uint64_t +qla2x00_get_num_tgts(scsi_qla_host_t *vha) +{ + fc_port_t *f, *tf; + u64 count = 0; + + f = NULL; + tf = NULL; + + list_for_each_entry_safe(f, tf, &vha->vp_fcports, list) { + if (f->port_type != FCT_TARGET) + continue; + count++; + } + return count; +} + +int qla2xxx_reset_stats(struct Scsi_Host *host, u32 flags) +{ + scsi_qla_host_t *vha = shost_priv(host); + fc_port_t *fcport = NULL; + unsigned long int_flags; + + if (flags & QLA2XX_HW_ERROR) + vha->hw_err_cnt = 0; + if (flags & QLA2XX_SHT_LNK_DWN) + vha->short_link_down_cnt = 0; + if (flags & QLA2XX_INT_ERR) + vha->interface_err_cnt = 0; + if (flags & QLA2XX_CMD_TIMEOUT) + vha->cmd_timeout_cnt = 0; + if (flags & QLA2XX_RESET_CMD_ERR) + vha->reset_cmd_err_cnt = 0; + if (flags & QLA2XX_TGT_SHT_LNK_DOWN) { + spin_lock_irqsave(&vha->hw->tgt.sess_lock, int_flags); + list_for_each_entry(fcport, &vha->vp_fcports, list) { + fcport->tgt_short_link_down_cnt = 0; + fcport->tgt_link_down_time = QLA2XX_MAX_LINK_DOWN_TIME; + } + spin_unlock_irqrestore(&vha->hw->tgt.sess_lock, int_flags); + } + vha->link_down_time = QLA2XX_MAX_LINK_DOWN_TIME; + return 0; +} + +int qla2xxx_start_stats(struct Scsi_Host *host, u32 flags) +{ + return qla2xxx_reset_stats(host, flags); +} + +int qla2xxx_stop_stats(struct Scsi_Host *host, u32 flags) +{ + return qla2xxx_reset_stats(host, flags); +} + +int qla2xxx_get_ini_stats(struct Scsi_Host *host, u32 flags, + void *data, u64 size) +{ + scsi_qla_host_t *vha = shost_priv(host); + struct ql_vnd_host_stats_resp *resp = (struct ql_vnd_host_stats_resp *)data; + struct ql_vnd_stats *rsp_data = &resp->stats; + u64 ini_entry_count = 0; + u64 i = 0; + u64 entry_count = 0; + u64 num_tgt = 0; + u32 tmp_stat_type = 0; + fc_port_t *fcport = NULL; + unsigned long int_flags; + + /* Copy stat type to work on it */ + tmp_stat_type = flags; + + if (tmp_stat_type & BIT_17) { + num_tgt = qla2x00_get_num_tgts(vha); + /* unset BIT_17 */ + tmp_stat_type &= ~(1 << 17); + } + ini_entry_count = qla2x00_count_set_bits(tmp_stat_type); + + entry_count = ini_entry_count + num_tgt; + + rsp_data->entry_count = entry_count; + + i = 0; + if (flags & QLA2XX_HW_ERROR) { + rsp_data->entry[i].stat_type = QLA2XX_HW_ERROR; + rsp_data->entry[i].tgt_num = 0x0; + rsp_data->entry[i].cnt = vha->hw_err_cnt; + i++; + } + + if (flags & QLA2XX_SHT_LNK_DWN) { + rsp_data->entry[i].stat_type = QLA2XX_SHT_LNK_DWN; + rsp_data->entry[i].tgt_num = 0x0; + rsp_data->entry[i].cnt = vha->short_link_down_cnt; + i++; + } + + if (flags & QLA2XX_INT_ERR) { + rsp_data->entry[i].stat_type = QLA2XX_INT_ERR; + rsp_data->entry[i].tgt_num = 0x0; + rsp_data->entry[i].cnt = vha->interface_err_cnt; + i++; + } + + if (flags & QLA2XX_CMD_TIMEOUT) { + rsp_data->entry[i].stat_type = QLA2XX_CMD_TIMEOUT; + rsp_data->entry[i].tgt_num = 0x0; + rsp_data->entry[i].cnt = vha->cmd_timeout_cnt; + i++; + } + + if (flags & QLA2XX_RESET_CMD_ERR) { + rsp_data->entry[i].stat_type = QLA2XX_RESET_CMD_ERR; + rsp_data->entry[i].tgt_num = 0x0; + rsp_data->entry[i].cnt = vha->reset_cmd_err_cnt; + i++; + } + + /* i will continue from previous loop, as target + * entries are after initiator + */ + if (flags & QLA2XX_TGT_SHT_LNK_DOWN) { + spin_lock_irqsave(&vha->hw->tgt.sess_lock, int_flags); + list_for_each_entry(fcport, &vha->vp_fcports, list) { + if (fcport->port_type != FCT_TARGET) + continue; + if (!fcport->rport) + continue; + rsp_data->entry[i].stat_type = QLA2XX_TGT_SHT_LNK_DOWN; + rsp_data->entry[i].tgt_num = fcport->rport->number; + rsp_data->entry[i].cnt = fcport->tgt_short_link_down_cnt; + i++; + } + spin_unlock_irqrestore(&vha->hw->tgt.sess_lock, int_flags); + } + resp->status = EXT_STATUS_OK; + + return 0; +} + +int qla2xxx_get_tgt_stats(struct Scsi_Host *host, u32 flags, + struct fc_rport *rport, void *data, u64 size) +{ + struct ql_vnd_tgt_stats_resp *tgt_data = data; + fc_port_t *fcport = *(fc_port_t **)rport->dd_data; + + tgt_data->status = 0; + tgt_data->stats.entry_count = 1; + tgt_data->stats.entry[0].stat_type = flags; + tgt_data->stats.entry[0].tgt_num = rport->number; + tgt_data->stats.entry[0].cnt = fcport->tgt_short_link_down_cnt; + + return 0; +} + +int qla2xxx_disable_port(struct Scsi_Host *host) +{ + scsi_qla_host_t *vha = shost_priv(host); + + vha->hw->flags.port_isolated = 1; + + if (qla2x00_chip_is_down(vha)) + return 0; + + if (vha->flags.online) { + qla2x00_abort_isp_cleanup(vha); + qla2x00_wait_for_sess_deletion(vha); + } + + return 0; +} + +int qla2xxx_enable_port(struct Scsi_Host *host) +{ + scsi_qla_host_t *vha = shost_priv(host); + + vha->hw->flags.port_isolated = 0; + /* Set the flag to 1, so that isp_abort can proceed */ + vha->flags.online = 1; + set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); + qla2xxx_wake_dpc(vha); + + return 0; +} diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c index f9142dbec112..9cf8326ab9fc 100644 --- a/drivers/scsi/qla2xxx/qla_isr.c +++ b/drivers/scsi/qla2xxx/qla_isr.c @@ -1059,6 +1059,9 @@ skip_rio: case MBA_SYSTEM_ERR: /* System Error */ mbx = 0; + + vha->hw_err_cnt++; + if (IS_QLA81XX(ha) || IS_QLA83XX(ha) || IS_QLA27XX(ha) || IS_QLA28XX(ha)) { u16 m[4]; @@ -1112,6 +1115,8 @@ skip_rio: ql_log(ql_log_warn, vha, 0x5006, "ISP Request Transfer Error (%x).\n", mb[1]); + vha->hw_err_cnt++; + set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); break; @@ -1119,6 +1124,8 @@ skip_rio: ql_log(ql_log_warn, vha, 0x5007, "ISP Response Transfer Error (%x).\n", mb[1]); + vha->hw_err_cnt++; + set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); break; @@ -1176,12 +1183,18 @@ skip_rio: vha->flags.management_server_logged_in = 0; qla2x00_post_aen_work(vha, FCH_EVT_LINKUP, ha->link_data_rate); + if (vha->link_down_time < vha->hw->port_down_retry_count) { + vha->short_link_down_cnt++; + vha->link_down_time = QLA2XX_MAX_LINK_DOWN_TIME; + } + break; case MBA_LOOP_DOWN: /* Loop Down Event */ SAVE_TOPO(ha); ha->flags.lip_ae = 0; ha->current_topology = 0; + vha->link_down_time = 0; mbx = (IS_QLA81XX(ha) || IS_QLA8031(ha)) ? rd_reg_word(®24->mailbox4) : 0; @@ -1503,6 +1516,7 @@ global_port_update: ql_dbg(ql_dbg_async, vha, 0x5016, "Discard RND Frame -- %04x %04x %04x.\n", mb[1], mb[2], mb[3]); + vha->interface_err_cnt++; break; case MBA_TRACE_NOTIFICATION: @@ -1592,6 +1606,7 @@ global_port_update: case MBA_IDC_AEN: if (IS_QLA27XX(ha) || IS_QLA28XX(ha)) { + vha->hw_err_cnt++; qla27xx_handle_8200_aen(vha, mb); } else if (IS_QLA83XX(ha)) { mb[4] = rd_reg_word(®24->mailbox4); @@ -3101,6 +3116,8 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt) "Dropped frame(s) detected (0x%x of 0x%x bytes).\n", resid, scsi_bufflen(cp)); + vha->interface_err_cnt++; + res = DID_ERROR << 16 | lscsi_status; goto check_scsi_status; } @@ -3126,6 +3143,8 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt) "Dropped frame(s) detected (0x%x of 0x%x bytes).\n", resid, scsi_bufflen(cp)); + vha->interface_err_cnt++; + res = DID_ERROR << 16 | lscsi_status; goto check_scsi_status; } else { @@ -3208,6 +3227,7 @@ check_scsi_status: case CS_TRANSPORT: res = DID_ERROR << 16; + vha->hw_err_cnt++; if (!IS_PI_SPLIT_DET_CAPABLE(ha)) break; @@ -3228,6 +3248,7 @@ check_scsi_status: ql_dump_buffer(ql_dbg_tgt + ql_dbg_verbose, vha, 0xe0ee, pkt, sizeof(*sts24)); res = DID_ERROR << 16; + vha->hw_err_cnt++; break; default: res = DID_ERROR << 16; @@ -3839,6 +3860,7 @@ qla24xx_msix_default(int irq, void *dev_id) hccr); qla2xxx_check_risc_status(vha); + vha->hw_err_cnt++; ha->isp_ops->fw_dump(vha); set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); diff --git a/drivers/scsi/qla2xxx/qla_mbx.c b/drivers/scsi/qla2xxx/qla_mbx.c index d7d4ab65009c..f438cdedca23 100644 --- a/drivers/scsi/qla2xxx/qla_mbx.c +++ b/drivers/scsi/qla2xxx/qla_mbx.c @@ -180,6 +180,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) ql_log(ql_log_warn, vha, 0xd035, "Cmd access timeout, cmd=0x%x, Exiting.\n", mcp->mb[0]); + vha->hw_err_cnt++; atomic_dec(&ha->num_pend_mbx_stage1); return QLA_FUNCTION_TIMEOUT; } @@ -307,6 +308,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) atomic_dec(&ha->num_pend_mbx_stage2); ql_dbg(ql_dbg_mbx, vha, 0x1012, "Pending mailbox timeout, exiting.\n"); + vha->hw_err_cnt++; rval = QLA_FUNCTION_TIMEOUT; goto premature_exit; } @@ -418,6 +420,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) "mb[0-3]=[0x%x 0x%x 0x%x 0x%x] mb7 0x%x host_status 0x%x hccr 0x%x\n", command, ictrl, jiffies, mb[0], mb[1], mb[2], mb[3], mb[7], host_status, hccr); + vha->hw_err_cnt++; } else { mb[0] = RD_MAILBOX_REG(ha, ®->isp, 0); @@ -425,6 +428,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) ql_dbg(ql_dbg_mbx + ql_dbg_buffer, vha, 0x1119, "MBX Command timeout for cmd %x, iocontrol=%x jiffies=%lx " "mb[0]=0x%x\n", command, ictrl, jiffies, mb[0]); + vha->hw_err_cnt++; } ql_dump_regs(ql_dbg_mbx + ql_dbg_buffer, vha, 0x1019); @@ -497,6 +501,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) "mb[0]=0x%x, eeh_busy=0x%x. Scheduling ISP " "abort.\n", command, mcp->mb[0], ha->flags.eeh_busy); + vha->hw_err_cnt++; set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); qla2xxx_wake_dpc(vha); } @@ -521,6 +526,7 @@ qla2x00_mailbox_command(scsi_qla_host_t *vha, mbx_cmd_t *mcp) "Mailbox cmd timeout occurred, cmd=0x%x, " "mb[0]=0x%x. Scheduling ISP abort ", command, mcp->mb[0]); + vha->hw_err_cnt++; set_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags); clear_bit(ISP_ABORT_NEEDED, &vha->dpc_flags); /* Allow next mbx cmd to come in. */ @@ -625,6 +631,7 @@ qla2x00_load_ram(scsi_qla_host_t *vha, dma_addr_t req_dma, uint32_t risc_addr, ql_dbg(ql_dbg_mbx, vha, 0x1023, "Failed=%x mb[0]=%x mb[1]=%x.\n", rval, mcp->mb[0], mcp->mb[1]); + vha->hw_err_cnt++; } else { ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1024, "Done %s.\n", __func__); @@ -736,6 +743,7 @@ again: ql_dbg(ql_dbg_mbx, vha, 0x1026, "Failed=%x mb[0]=%x.\n", rval, mcp->mb[0]); + vha->hw_err_cnt++; return rval; } @@ -1313,6 +1321,7 @@ qla2x00_mbx_reg_test(scsi_qla_host_t *vha) if (rval != QLA_SUCCESS) { /*EMPTY*/ ql_dbg(ql_dbg_mbx, vha, 0x1033, "Failed=%x.\n", rval); + vha->hw_err_cnt++; } else { /*EMPTY*/ ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1034, diff --git a/drivers/scsi/qla2xxx/qla_os.c b/drivers/scsi/qla2xxx/qla_os.c index f80abe28f35a..a760cb38e487 100644 --- a/drivers/scsi/qla2xxx/qla_os.c +++ b/drivers/scsi/qla2xxx/qla_os.c @@ -1274,6 +1274,8 @@ qla2xxx_eh_abort(struct scsi_cmnd *cmd) sp = scsi_cmd_priv(cmd); qpair = sp->qpair; + vha->cmd_timeout_cnt++; + if ((sp->fcport && sp->fcport->deleted) || !qpair) return SUCCESS; @@ -1442,6 +1444,7 @@ eh_reset_failed: "%s RESET FAILED: %s nexus=%ld:%d:%llu cmd=%p.\n", name, reset_errors[err], vha->host_no, cmd->device->id, cmd->device->lun, cmd); + vha->reset_cmd_err_cnt++; return FAILED; } @@ -3141,6 +3144,10 @@ qla2x00_probe_one(struct pci_dev *pdev, const struct pci_device_id *id) ha->mr.fcport.supported_classes = FC_COS_UNSPECIFIED; ha->mr.fcport.scan_state = 1; + qla2xxx_reset_stats(host, QLA2XX_HW_ERROR | QLA2XX_SHT_LNK_DWN | + QLA2XX_INT_ERR | QLA2XX_CMD_TIMEOUT | + QLA2XX_RESET_CMD_ERR | QLA2XX_TGT_SHT_LNK_DOWN); + /* Set the SG table size based on ISP type */ if (!IS_FWI2_CAPABLE(ha)) { if (IS_QLA2100(ha)) @@ -5090,6 +5097,7 @@ void qla24xx_create_new_sess(struct scsi_qla_host *vha, struct qla_work_evt *e) fcport->d_id = e->u.new_sess.id; fcport->flags |= FCF_FABRIC_DEVICE; fcport->fw_login_state = DSC_LS_PLOGI_PEND; + fcport->tgt_short_link_down_cnt = 0; memcpy(fcport->port_name, e->u.new_sess.port_name, WWN_SIZE); @@ -7061,6 +7069,8 @@ qla2x00_timer(struct timer_list *t) uint16_t w; struct qla_hw_data *ha = vha->hw; struct req_que *req; + unsigned long flags; + fc_port_t *fcport = NULL; if (ha->flags.eeh_busy) { ql_dbg(ql_dbg_timer, vha, 0x6000, @@ -7092,6 +7102,16 @@ qla2x00_timer(struct timer_list *t) if (!vha->vp_idx && IS_QLAFX00(ha)) qlafx00_timer_routine(vha); + if (vha->link_down_time < QLA2XX_MAX_LINK_DOWN_TIME) + vha->link_down_time++; + + spin_lock_irqsave(&vha->hw->tgt.sess_lock, flags); + list_for_each_entry(fcport, &vha->vp_fcports, list) { + if (fcport->tgt_link_down_time < QLA2XX_MAX_LINK_DOWN_TIME) + fcport->tgt_link_down_time++; + } + spin_unlock_irqrestore(&vha->hw->tgt.sess_lock, flags); + /* Loop down handler. */ if (atomic_read(&vha->loop_down_timer) > 0 && !(test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags)) && |