@@ -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;
@@ -396,6 +396,7 @@ 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)
{
@@ -3857,6 +3858,13 @@ 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.
*/
@@ -4607,6 +4615,7 @@ 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 {
@@ -4727,6 +4736,7 @@ typedef struct scsi_qla_host {
#define FX00_CRITEMP_RECOVERY 25
#define FX00_HOST_INFO_RESEND 26
#define QPAIR_ONLINE_CHECK_NEEDED 27
+#define DO_EEH_RECOVERY 28
#define DETECT_SFP_CHANGE 29
#define N2N_LOGIN_NEEDED 30
#define IOCB_WORK_ACTIVE 31
@@ -224,6 +224,7 @@ 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);
@@ -235,6 +236,8 @@ 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.
@@ -6932,22 +6932,18 @@ 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");
-
- /* Done waiting for pending commands.
- * Reset the online flag.
- */
- vha->flags.online = 0;
- }
+ /* 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");
- /* 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();
}
@@ -6978,6 +6974,12 @@ 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;
@@ -7017,8 +7019,18 @@ 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);
@@ -432,3 +432,49 @@ 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;
+}
@@ -1645,8 +1645,14 @@ 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
@@ -1842,8 +1848,13 @@ 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
@@ -1922,6 +1933,7 @@ qla24xx_dif_start_scsi(srb_t *sp)
qla_put_iocbs(sp->qpair, &sp->iores);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
+
return QLA_FUNCTION_FAILED;
}
@@ -1991,8 +2003,14 @@ 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
@@ -2203,8 +2221,14 @@ 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
@@ -2281,6 +2305,7 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp)
qla_put_iocbs(sp->qpair, &sp->iores);
spin_unlock_irqrestore(&qpair->qp_lock, flags);
+
return QLA_FUNCTION_FAILED;
}
@@ -2325,6 +2350,11 @@ __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
@@ -3739,6 +3769,9 @@ 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) {
@@ -3956,8 +3989,14 @@ 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
@@ -3996,5 +4035,6 @@ 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;
}
@@ -270,12 +270,7 @@ 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
@@ -1657,8 +1652,6 @@ qla2x00_async_event(scsi_qla_host_t *vha, struct rsp_que *rsp, uint16_t *mb)
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:
@@ -161,7 +161,8 @@ 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)) &&
+ test_bit(ISP_ABORT_NEEDED, &base_vha->dpc_flags) ||
+ ha->flags.eeh_busy) &&
!is_rom_cmd(mcp->mb[0])) {
ql_log(ql_log_info, vha, 0x1005,
"Cmd 0x%x aborted with timeout since ISP Abort is pending\n",
@@ -398,8 +398,13 @@ 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;
@@ -536,6 +541,7 @@ static inline int qla2x00_start_nvme_mq(srb_t *sp)
queuing_error:
spin_unlock_irqrestore(&qpair->qp_lock, flags);
+
return rval;
}
@@ -971,6 +971,13 @@ 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;
@@ -1200,35 +1207,6 @@ 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
*
@@ -1262,6 +1240,7 @@ 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;
}
@@ -1455,6 +1434,7 @@ 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;
}
@@ -1471,6 +1451,7 @@ 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;
}
@@ -1506,6 +1487,7 @@ 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;
}
@@ -1583,7 +1565,7 @@ 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;
}
@@ -6669,6 +6651,9 @@ 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;
@@ -7385,6 +7370,8 @@ 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;
@@ -7394,28 +7381,16 @@ 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);
@@ -7452,14 +7427,17 @@ 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) {
@@ -7469,11 +7447,12 @@ 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);
@@ -7481,9 +7460,12 @@ 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
@@ -7497,6 +7479,10 @@ 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;
@@ -7520,10 +7506,11 @@ 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
@@ -7535,9 +7522,10 @@ 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.
@@ -7571,16 +7559,24 @@ 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;
}
@@ -7592,16 +7588,55 @@ 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);
}
static void