linux/drivers/scsi/qla4xxx/ql4_isr.c
FUJITA Tomonori b80ca4f7ee [SCSI] replace sizeof sense_buffer with SCSI_SENSE_BUFFERSIZE
This replaces sizeof sense_buffer with SCSI_SENSE_BUFFERSIZE in
several LLDs. It's a preparation for the future changes to remove
sense_buffer array in scsi_cmnd structure.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-23 11:29:27 -06:00

819 lines
22 KiB
C

/*
* QLogic iSCSI HBA Driver
* Copyright (c) 2003-2006 QLogic Corporation
*
* See LICENSE.qla4xxx for copyright and licensing details.
*/
#include "ql4_def.h"
#include "ql4_glbl.h"
#include "ql4_dbg.h"
#include "ql4_inline.h"
/**
* qla2x00_process_completed_request() - Process a Fast Post response.
* @ha: SCSI driver HA context
* @index: SRB index
**/
static void qla4xxx_process_completed_request(struct scsi_qla_host *ha,
uint32_t index)
{
struct srb *srb;
srb = qla4xxx_del_from_active_array(ha, index);
if (srb) {
/* Save ISP completion status */
srb->cmd->result = DID_OK << 16;
qla4xxx_srb_compl(ha, srb);
} else {
DEBUG2(printk("scsi%ld: Invalid ISP SCSI completion handle = "
"%d\n", ha->host_no, index));
set_bit(DPC_RESET_HA, &ha->dpc_flags);
}
}
/**
* qla4xxx_status_entry - processes status IOCBs
* @ha: Pointer to host adapter structure.
* @sts_entry: Pointer to status entry structure.
**/
static void qla4xxx_status_entry(struct scsi_qla_host *ha,
struct status_entry *sts_entry)
{
uint8_t scsi_status;
struct scsi_cmnd *cmd;
struct srb *srb;
struct ddb_entry *ddb_entry;
uint32_t residual;
uint16_t sensebytecnt;
if (sts_entry->completionStatus == SCS_COMPLETE &&
sts_entry->scsiStatus == 0) {
qla4xxx_process_completed_request(ha,
le32_to_cpu(sts_entry->
handle));
return;
}
srb = qla4xxx_del_from_active_array(ha, le32_to_cpu(sts_entry->handle));
if (!srb) {
/* FIXMEdg: Don't we need to reset ISP in this case??? */
DEBUG2(printk(KERN_WARNING "scsi%ld: %s: Status Entry invalid "
"handle 0x%x, sp=%p. This cmd may have already "
"been completed.\n", ha->host_no, __func__,
le32_to_cpu(sts_entry->handle), srb));
return;
}
cmd = srb->cmd;
if (cmd == NULL) {
DEBUG2(printk("scsi%ld: %s: Command already returned back to "
"OS pkt->handle=%d srb=%p srb->state:%d\n",
ha->host_no, __func__, sts_entry->handle,
srb, srb->state));
dev_warn(&ha->pdev->dev, "Command is NULL:"
" already returned to OS (srb=%p)\n", srb);
return;
}
ddb_entry = srb->ddb;
if (ddb_entry == NULL) {
cmd->result = DID_NO_CONNECT << 16;
goto status_entry_exit;
}
residual = le32_to_cpu(sts_entry->residualByteCnt);
/* Translate ISP error to a Linux SCSI error. */
scsi_status = sts_entry->scsiStatus;
switch (sts_entry->completionStatus) {
case SCS_COMPLETE:
if (scsi_status == 0) {
cmd->result = DID_OK << 16;
break;
}
if (sts_entry->iscsiFlags & ISCSI_FLAG_RESIDUAL_OVER) {
cmd->result = DID_ERROR << 16;
break;
}
if (sts_entry->iscsiFlags &ISCSI_FLAG_RESIDUAL_UNDER) {
scsi_set_resid(cmd, residual);
if (!scsi_status && ((scsi_bufflen(cmd) - residual) <
cmd->underflow)) {
cmd->result = DID_ERROR << 16;
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: "
"Mid-layer Data underrun0, "
"xferlen = 0x%x, "
"residual = 0x%x\n", ha->host_no,
cmd->device->channel,
cmd->device->id,
cmd->device->lun, __func__,
scsi_bufflen(cmd), residual));
break;
}
}
cmd->result = DID_OK << 16 | scsi_status;
if (scsi_status != SCSI_CHECK_CONDITION)
break;
/* Copy Sense Data into sense buffer. */
memset(cmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
sensebytecnt = le16_to_cpu(sts_entry->senseDataByteCnt);
if (sensebytecnt == 0)
break;
memcpy(cmd->sense_buffer, sts_entry->senseData,
min_t(uint16_t, sensebytecnt, SCSI_SENSE_BUFFERSIZE));
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: sense key = %x, "
"ASC/ASCQ = %02x/%02x\n", ha->host_no,
cmd->device->channel, cmd->device->id,
cmd->device->lun, __func__,
sts_entry->senseData[2] & 0x0f,
sts_entry->senseData[12],
sts_entry->senseData[13]));
srb->flags |= SRB_GOT_SENSE;
break;
case SCS_INCOMPLETE:
/* Always set the status to DID_ERROR, since
* all conditions result in that status anyway */
cmd->result = DID_ERROR << 16;
break;
case SCS_RESET_OCCURRED:
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: Device RESET occurred\n",
ha->host_no, cmd->device->channel,
cmd->device->id, cmd->device->lun, __func__));
cmd->result = DID_RESET << 16;
break;
case SCS_ABORTED:
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: Abort occurred\n",
ha->host_no, cmd->device->channel,
cmd->device->id, cmd->device->lun, __func__));
cmd->result = DID_RESET << 16;
break;
case SCS_TIMEOUT:
DEBUG2(printk(KERN_INFO "scsi%ld:%d:%d:%d: Timeout\n",
ha->host_no, cmd->device->channel,
cmd->device->id, cmd->device->lun));
cmd->result = DID_BUS_BUSY << 16;
/*
* Mark device missing so that we won't continue to send
* I/O to this device. We should get a ddb state change
* AEN soon.
*/
if (atomic_read(&ddb_entry->state) == DDB_STATE_ONLINE)
qla4xxx_mark_device_missing(ha, ddb_entry);
break;
case SCS_DATA_UNDERRUN:
case SCS_DATA_OVERRUN:
if ((sts_entry->iscsiFlags & ISCSI_FLAG_RESIDUAL_OVER) ||
(sts_entry->completionStatus == SCS_DATA_OVERRUN)) {
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: " "Data overrun, "
"residual = 0x%x\n", ha->host_no,
cmd->device->channel, cmd->device->id,
cmd->device->lun, __func__, residual));
cmd->result = DID_ERROR << 16;
break;
}
scsi_set_resid(cmd, residual);
/*
* If there is scsi_status, it takes precedense over
* underflow condition.
*/
if (scsi_status != 0) {
cmd->result = DID_OK << 16 | scsi_status;
if (scsi_status != SCSI_CHECK_CONDITION)
break;
/* Copy Sense Data into sense buffer. */
memset(cmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
sensebytecnt =
le16_to_cpu(sts_entry->senseDataByteCnt);
if (sensebytecnt == 0)
break;
memcpy(cmd->sense_buffer, sts_entry->senseData,
min_t(uint16_t, sensebytecnt, SCSI_SENSE_BUFFERSIZE));
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: sense key = %x, "
"ASC/ASCQ = %02x/%02x\n", ha->host_no,
cmd->device->channel, cmd->device->id,
cmd->device->lun, __func__,
sts_entry->senseData[2] & 0x0f,
sts_entry->senseData[12],
sts_entry->senseData[13]));
} else {
/*
* If RISC reports underrun and target does not
* report it then we must have a lost frame, so
* tell upper layer to retry it by reporting a
* bus busy.
*/
if ((sts_entry->iscsiFlags &
ISCSI_FLAG_RESIDUAL_UNDER) == 0) {
cmd->result = DID_BUS_BUSY << 16;
} else if ((scsi_bufflen(cmd) - residual) <
cmd->underflow) {
/*
* Handle mid-layer underflow???
*
* For kernels less than 2.4, the driver must
* return an error if an underflow is detected.
* For kernels equal-to and above 2.4, the
* mid-layer will appearantly handle the
* underflow by detecting the residual count --
* unfortunately, we do not see where this is
* actually being done. In the interim, we
* will return DID_ERROR.
*/
DEBUG2(printk("scsi%ld:%d:%d:%d: %s: "
"Mid-layer Data underrun1, "
"xferlen = 0x%x, "
"residual = 0x%x\n", ha->host_no,
cmd->device->channel,
cmd->device->id,
cmd->device->lun, __func__,
scsi_bufflen(cmd), residual));
cmd->result = DID_ERROR << 16;
} else {
cmd->result = DID_OK << 16;
}
}
break;
case SCS_DEVICE_LOGGED_OUT:
case SCS_DEVICE_UNAVAILABLE:
/*
* Mark device missing so that we won't continue to
* send I/O to this device. We should get a ddb
* state change AEN soon.
*/
if (atomic_read(&ddb_entry->state) == DDB_STATE_ONLINE)
qla4xxx_mark_device_missing(ha, ddb_entry);
cmd->result = DID_BUS_BUSY << 16;
break;
case SCS_QUEUE_FULL:
/*
* SCSI Mid-Layer handles device queue full
*/
cmd->result = DID_OK << 16 | sts_entry->scsiStatus;
DEBUG2(printk("scsi%ld:%d:%d: %s: QUEUE FULL detected "
"compl=%02x, scsi=%02x, state=%02x, iFlags=%02x,"
" iResp=%02x\n", ha->host_no, cmd->device->id,
cmd->device->lun, __func__,
sts_entry->completionStatus,
sts_entry->scsiStatus, sts_entry->state_flags,
sts_entry->iscsiFlags,
sts_entry->iscsiResponse));
break;
default:
cmd->result = DID_ERROR << 16;
break;
}
status_entry_exit:
/* complete the request */
srb->cc_stat = sts_entry->completionStatus;
qla4xxx_srb_compl(ha, srb);
}
/**
* qla4xxx_process_response_queue - process response queue completions
* @ha: Pointer to host adapter structure.
*
* This routine process response queue completions in interrupt context.
* Hardware_lock locked upon entry
**/
static void qla4xxx_process_response_queue(struct scsi_qla_host * ha)
{
uint32_t count = 0;
struct srb *srb = NULL;
struct status_entry *sts_entry;
/* Process all responses from response queue */
while ((ha->response_in =
(uint16_t)le32_to_cpu(ha->shadow_regs->rsp_q_in)) !=
ha->response_out) {
sts_entry = (struct status_entry *) ha->response_ptr;
count++;
/* Advance pointers for next entry */
if (ha->response_out == (RESPONSE_QUEUE_DEPTH - 1)) {
ha->response_out = 0;
ha->response_ptr = ha->response_ring;
} else {
ha->response_out++;
ha->response_ptr++;
}
/* process entry */
switch (sts_entry->hdr.entryType) {
case ET_STATUS:
/*
* Common status - Single completion posted in single
* IOSB.
*/
qla4xxx_status_entry(ha, sts_entry);
break;
case ET_PASSTHRU_STATUS:
break;
case ET_STATUS_CONTINUATION:
/* Just throw away the status continuation entries */
DEBUG2(printk("scsi%ld: %s: Status Continuation entry "
"- ignoring\n", ha->host_no, __func__));
break;
case ET_COMMAND:
/* ISP device queue is full. Command not
* accepted by ISP. Queue command for
* later */
srb = qla4xxx_del_from_active_array(ha,
le32_to_cpu(sts_entry->
handle));
if (srb == NULL)
goto exit_prq_invalid_handle;
DEBUG2(printk("scsi%ld: %s: FW device queue full, "
"srb %p\n", ha->host_no, __func__, srb));
/* ETRY normally by sending it back with
* DID_BUS_BUSY */
srb->cmd->result = DID_BUS_BUSY << 16;
qla4xxx_srb_compl(ha, srb);
break;
case ET_CONTINUE:
/* Just throw away the continuation entries */
DEBUG2(printk("scsi%ld: %s: Continuation entry - "
"ignoring\n", ha->host_no, __func__));
break;
default:
/*
* Invalid entry in response queue, reset RISC
* firmware.
*/
DEBUG2(printk("scsi%ld: %s: Invalid entry %x in "
"response queue \n", ha->host_no,
__func__,
sts_entry->hdr.entryType));
goto exit_prq_error;
}
}
/*
* Done with responses, update the ISP For QLA4010, this also clears
* the interrupt.
*/
writel(ha->response_out, &ha->reg->rsp_q_out);
readl(&ha->reg->rsp_q_out);
return;
exit_prq_invalid_handle:
DEBUG2(printk("scsi%ld: %s: Invalid handle(srb)=%p type=%x IOCS=%x\n",
ha->host_no, __func__, srb, sts_entry->hdr.entryType,
sts_entry->completionStatus));
exit_prq_error:
writel(ha->response_out, &ha->reg->rsp_q_out);
readl(&ha->reg->rsp_q_out);
set_bit(DPC_RESET_HA, &ha->dpc_flags);
}
/**
* qla4xxx_isr_decode_mailbox - decodes mailbox status
* @ha: Pointer to host adapter structure.
* @mailbox_status: Mailbox status.
*
* This routine decodes the mailbox status during the ISR.
* Hardware_lock locked upon entry. runs in interrupt context.
**/
static void qla4xxx_isr_decode_mailbox(struct scsi_qla_host * ha,
uint32_t mbox_status)
{
int i;
uint32_t mbox_stat2, mbox_stat3;
if ((mbox_status == MBOX_STS_BUSY) ||
(mbox_status == MBOX_STS_INTERMEDIATE_COMPLETION) ||
(mbox_status >> 12 == MBOX_COMPLETION_STATUS)) {
ha->mbox_status[0] = mbox_status;
if (test_bit(AF_MBOX_COMMAND, &ha->flags)) {
/*
* Copy all mailbox registers to a temporary
* location and set mailbox command done flag
*/
for (i = 1; i < ha->mbox_status_count; i++)
ha->mbox_status[i] =
readl(&ha->reg->mailbox[i]);
set_bit(AF_MBOX_COMMAND_DONE, &ha->flags);
}
} else if (mbox_status >> 12 == MBOX_ASYNC_EVENT_STATUS) {
/* Immediately process the AENs that don't require much work.
* Only queue the database_changed AENs */
if (ha->aen_log.count < MAX_AEN_ENTRIES) {
for (i = 0; i < MBOX_AEN_REG_COUNT; i++)
ha->aen_log.entry[ha->aen_log.count].mbox_sts[i] =
readl(&ha->reg->mailbox[i]);
ha->aen_log.count++;
}
switch (mbox_status) {
case MBOX_ASTS_SYSTEM_ERROR:
/* Log Mailbox registers */
if (ql4xdontresethba) {
DEBUG2(printk("%s:Dont Reset HBA\n",
__func__));
} else {
set_bit(AF_GET_CRASH_RECORD, &ha->flags);
set_bit(DPC_RESET_HA, &ha->dpc_flags);
}
break;
case MBOX_ASTS_REQUEST_TRANSFER_ERROR:
case MBOX_ASTS_RESPONSE_TRANSFER_ERROR:
case MBOX_ASTS_NVRAM_INVALID:
case MBOX_ASTS_IP_ADDRESS_CHANGED:
case MBOX_ASTS_DHCP_LEASE_EXPIRED:
DEBUG2(printk("scsi%ld: AEN %04x, ERROR Status, "
"Reset HA\n", ha->host_no, mbox_status));
set_bit(DPC_RESET_HA, &ha->dpc_flags);
break;
case MBOX_ASTS_LINK_UP:
DEBUG2(printk("scsi%ld: AEN %04x Adapter LINK UP\n",
ha->host_no, mbox_status));
set_bit(AF_LINK_UP, &ha->flags);
break;
case MBOX_ASTS_LINK_DOWN:
DEBUG2(printk("scsi%ld: AEN %04x Adapter LINK DOWN\n",
ha->host_no, mbox_status));
clear_bit(AF_LINK_UP, &ha->flags);
break;
case MBOX_ASTS_HEARTBEAT:
ha->seconds_since_last_heartbeat = 0;
break;
case MBOX_ASTS_DHCP_LEASE_ACQUIRED:
DEBUG2(printk("scsi%ld: AEN %04x DHCP LEASE "
"ACQUIRED\n", ha->host_no, mbox_status));
set_bit(DPC_GET_DHCP_IP_ADDR, &ha->dpc_flags);
break;
case MBOX_ASTS_PROTOCOL_STATISTIC_ALARM:
case MBOX_ASTS_SCSI_COMMAND_PDU_REJECTED: /* Target
* mode
* only */
case MBOX_ASTS_UNSOLICITED_PDU_RECEIVED: /* Connection mode */
case MBOX_ASTS_IPSEC_SYSTEM_FATAL_ERROR:
case MBOX_ASTS_SUBNET_STATE_CHANGE:
/* No action */
DEBUG2(printk("scsi%ld: AEN %04x\n", ha->host_no,
mbox_status));
break;
case MBOX_ASTS_IP_ADDR_STATE_CHANGED:
mbox_stat2 = readl(&ha->reg->mailbox[2]);
mbox_stat3 = readl(&ha->reg->mailbox[3]);
if ((mbox_stat3 == 5) && (mbox_stat2 == 3))
set_bit(DPC_GET_DHCP_IP_ADDR, &ha->dpc_flags);
else if ((mbox_stat3 == 2) && (mbox_stat2 == 5))
set_bit(DPC_RESET_HA, &ha->dpc_flags);
break;
case MBOX_ASTS_MAC_ADDRESS_CHANGED:
case MBOX_ASTS_DNS:
/* No action */
DEBUG2(printk(KERN_INFO "scsi%ld: AEN %04x, "
"mbox_sts[1]=%04x, mbox_sts[2]=%04x\n",
ha->host_no, mbox_status,
readl(&ha->reg->mailbox[1]),
readl(&ha->reg->mailbox[2])));
break;
case MBOX_ASTS_SELF_TEST_FAILED:
case MBOX_ASTS_LOGIN_FAILED:
/* No action */
DEBUG2(printk("scsi%ld: AEN %04x, mbox_sts[1]=%04x, "
"mbox_sts[2]=%04x, mbox_sts[3]=%04x\n",
ha->host_no, mbox_status,
readl(&ha->reg->mailbox[1]),
readl(&ha->reg->mailbox[2]),
readl(&ha->reg->mailbox[3])));
break;
case MBOX_ASTS_DATABASE_CHANGED:
/* Queue AEN information and process it in the DPC
* routine */
if (ha->aen_q_count > 0) {
/* decrement available counter */
ha->aen_q_count--;
for (i = 1; i < MBOX_AEN_REG_COUNT; i++)
ha->aen_q[ha->aen_in].mbox_sts[i] =
readl(&ha->reg->mailbox[i]);
ha->aen_q[ha->aen_in].mbox_sts[0] = mbox_status;
/* print debug message */
DEBUG2(printk("scsi%ld: AEN[%d] %04x queued"
" mb1:0x%x mb2:0x%x mb3:0x%x mb4:0x%x\n",
ha->host_no, ha->aen_in,
mbox_status,
ha->aen_q[ha->aen_in].mbox_sts[1],
ha->aen_q[ha->aen_in].mbox_sts[2],
ha->aen_q[ha->aen_in].mbox_sts[3],
ha->aen_q[ha->aen_in]. mbox_sts[4]));
/* advance pointer */
ha->aen_in++;
if (ha->aen_in == MAX_AEN_ENTRIES)
ha->aen_in = 0;
/* The DPC routine will process the aen */
set_bit(DPC_AEN, &ha->dpc_flags);
} else {
DEBUG2(printk("scsi%ld: %s: aen %04x, queue "
"overflowed! AEN LOST!!\n",
ha->host_no, __func__,
mbox_status));
DEBUG2(printk("scsi%ld: DUMP AEN QUEUE\n",
ha->host_no));
for (i = 0; i < MAX_AEN_ENTRIES; i++) {
DEBUG2(printk("AEN[%d] %04x %04x %04x "
"%04x\n", i,
ha->aen_q[i].mbox_sts[0],
ha->aen_q[i].mbox_sts[1],
ha->aen_q[i].mbox_sts[2],
ha->aen_q[i].mbox_sts[3]));
}
}
break;
default:
DEBUG2(printk(KERN_WARNING
"scsi%ld: AEN %04x UNKNOWN\n",
ha->host_no, mbox_status));
break;
}
} else {
DEBUG2(printk("scsi%ld: Unknown mailbox status %08X\n",
ha->host_no, mbox_status));
ha->mbox_status[0] = mbox_status;
}
}
/**
* qla4xxx_interrupt_service_routine - isr
* @ha: pointer to host adapter structure.
*
* This is the main interrupt service routine.
* hardware_lock locked upon entry. runs in interrupt context.
**/
void qla4xxx_interrupt_service_routine(struct scsi_qla_host * ha,
uint32_t intr_status)
{
/* Process response queue interrupt. */
if (intr_status & CSR_SCSI_COMPLETION_INTR)
qla4xxx_process_response_queue(ha);
/* Process mailbox/asynch event interrupt.*/
if (intr_status & CSR_SCSI_PROCESSOR_INTR) {
qla4xxx_isr_decode_mailbox(ha,
readl(&ha->reg->mailbox[0]));
/* Clear Mailbox Interrupt */
writel(set_rmask(CSR_SCSI_PROCESSOR_INTR),
&ha->reg->ctrl_status);
readl(&ha->reg->ctrl_status);
}
}
/**
* qla4xxx_intr_handler - hardware interrupt handler.
* @irq: Unused
* @dev_id: Pointer to host adapter structure
**/
irqreturn_t qla4xxx_intr_handler(int irq, void *dev_id)
{
struct scsi_qla_host *ha;
uint32_t intr_status;
unsigned long flags = 0;
uint8_t reqs_count = 0;
ha = (struct scsi_qla_host *) dev_id;
if (!ha) {
DEBUG2(printk(KERN_INFO
"qla4xxx: Interrupt with NULL host ptr\n"));
return IRQ_NONE;
}
spin_lock_irqsave(&ha->hardware_lock, flags);
ha->isr_count++;
/*
* Repeatedly service interrupts up to a maximum of
* MAX_REQS_SERVICED_PER_INTR
*/
while (1) {
/*
* Read interrupt status
*/
if (le32_to_cpu(ha->shadow_regs->rsp_q_in) !=
ha->response_out)
intr_status = CSR_SCSI_COMPLETION_INTR;
else
intr_status = readl(&ha->reg->ctrl_status);
if ((intr_status &
(CSR_SCSI_RESET_INTR|CSR_FATAL_ERROR|INTR_PENDING)) ==
0) {
if (reqs_count == 0)
ha->spurious_int_count++;
break;
}
if (intr_status & CSR_FATAL_ERROR) {
DEBUG2(printk(KERN_INFO "scsi%ld: Fatal Error, "
"Status 0x%04x\n", ha->host_no,
readl(isp_port_error_status (ha))));
/* Issue Soft Reset to clear this error condition.
* This will prevent the RISC from repeatedly
* interrupting the driver; thus, allowing the DPC to
* get scheduled to continue error recovery.
* NOTE: Disabling RISC interrupts does not work in
* this case, as CSR_FATAL_ERROR overrides
* CSR_SCSI_INTR_ENABLE */
if ((readl(&ha->reg->ctrl_status) &
CSR_SCSI_RESET_INTR) == 0) {
writel(set_rmask(CSR_SOFT_RESET),
&ha->reg->ctrl_status);
readl(&ha->reg->ctrl_status);
}
writel(set_rmask(CSR_FATAL_ERROR),
&ha->reg->ctrl_status);
readl(&ha->reg->ctrl_status);
__qla4xxx_disable_intrs(ha);
set_bit(DPC_RESET_HA, &ha->dpc_flags);
break;
} else if (intr_status & CSR_SCSI_RESET_INTR) {
clear_bit(AF_ONLINE, &ha->flags);
__qla4xxx_disable_intrs(ha);
writel(set_rmask(CSR_SCSI_RESET_INTR),
&ha->reg->ctrl_status);
readl(&ha->reg->ctrl_status);
if (!ql4_mod_unload)
set_bit(DPC_RESET_HA_INTR, &ha->dpc_flags);
break;
} else if (intr_status & INTR_PENDING) {
qla4xxx_interrupt_service_routine(ha, intr_status);
ha->total_io_count++;
if (++reqs_count == MAX_REQS_SERVICED_PER_INTR)
break;
intr_status = 0;
}
}
spin_unlock_irqrestore(&ha->hardware_lock, flags);
return IRQ_HANDLED;
}
/**
* qla4xxx_process_aen - processes AENs generated by firmware
* @ha: pointer to host adapter structure.
* @process_aen: type of AENs to process
*
* Processes specific types of Asynchronous Events generated by firmware.
* The type of AENs to process is specified by process_aen and can be
* PROCESS_ALL_AENS 0
* FLUSH_DDB_CHANGED_AENS 1
* RELOGIN_DDB_CHANGED_AENS 2
**/
void qla4xxx_process_aen(struct scsi_qla_host * ha, uint8_t process_aen)
{
uint32_t mbox_sts[MBOX_AEN_REG_COUNT];
struct aen *aen;
int i;
unsigned long flags;
spin_lock_irqsave(&ha->hardware_lock, flags);
while (ha->aen_out != ha->aen_in) {
aen = &ha->aen_q[ha->aen_out];
/* copy aen information to local structure */
for (i = 0; i < MBOX_AEN_REG_COUNT; i++)
mbox_sts[i] = aen->mbox_sts[i];
ha->aen_q_count++;
ha->aen_out++;
if (ha->aen_out == MAX_AEN_ENTRIES)
ha->aen_out = 0;
spin_unlock_irqrestore(&ha->hardware_lock, flags);
DEBUG2(printk("qla4xxx(%ld): AEN[%d]=0x%08x, mbx1=0x%08x mbx2=0x%08x"
" mbx3=0x%08x mbx4=0x%08x\n", ha->host_no,
(ha->aen_out ? (ha->aen_out-1): (MAX_AEN_ENTRIES-1)),
mbox_sts[0], mbox_sts[1], mbox_sts[2],
mbox_sts[3], mbox_sts[4]));
switch (mbox_sts[0]) {
case MBOX_ASTS_DATABASE_CHANGED:
if (process_aen == FLUSH_DDB_CHANGED_AENS) {
DEBUG2(printk("scsi%ld: AEN[%d] %04x, index "
"[%d] state=%04x FLUSHED!\n",
ha->host_no, ha->aen_out,
mbox_sts[0], mbox_sts[2],
mbox_sts[3]));
break;
} else if (process_aen == RELOGIN_DDB_CHANGED_AENS) {
/* for use during init time, we only want to
* relogin non-active ddbs */
struct ddb_entry *ddb_entry;
ddb_entry =
/* FIXME: name length? */
qla4xxx_lookup_ddb_by_fw_index(ha,
mbox_sts[2]);
if (!ddb_entry)
break;
ddb_entry->dev_scan_wait_to_complete_relogin =
0;
ddb_entry->dev_scan_wait_to_start_relogin =
jiffies +
((ddb_entry->default_time2wait +
4) * HZ);
DEBUG2(printk("scsi%ld: ddb index [%d] initate"
" RELOGIN after %d seconds\n",
ha->host_no,
ddb_entry->fw_ddb_index,
ddb_entry->default_time2wait +
4));
break;
}
if (mbox_sts[1] == 0) { /* Global DB change. */
qla4xxx_reinitialize_ddb_list(ha);
} else if (mbox_sts[1] == 1) { /* Specific device. */
qla4xxx_process_ddb_changed(ha, mbox_sts[2],
mbox_sts[3]);
}
break;
}
spin_lock_irqsave(&ha->hardware_lock, flags);
}
spin_unlock_irqrestore(&ha->hardware_lock, flags);
}