linux/drivers/scsi/qla2xxx/qla_mr.c

3475 lines
91 KiB
C

/*
* QLogic Fibre Channel HBA Driver
* Copyright (c) 2003-2014 QLogic Corporation
*
* See LICENSE.qla2xxx for copyright and licensing details.
*/
#include "qla_def.h"
#include <linux/delay.h>
#include <linux/pci.h>
#include <linux/ratelimit.h>
#include <linux/vmalloc.h>
#include <scsi/scsi_tcq.h>
#include <linux/utsname.h>
/* QLAFX00 specific Mailbox implementation functions */
/*
* qlafx00_mailbox_command
* Issue mailbox command and waits for completion.
*
* Input:
* ha = adapter block pointer.
* mcp = driver internal mbx struct pointer.
*
* Output:
* mb[MAX_MAILBOX_REGISTER_COUNT] = returned mailbox data.
*
* Returns:
* 0 : QLA_SUCCESS = cmd performed success
* 1 : QLA_FUNCTION_FAILED (error encountered)
* 6 : QLA_FUNCTION_TIMEOUT (timeout condition encountered)
*
* Context:
* Kernel context.
*/
static int
qlafx00_mailbox_command(scsi_qla_host_t *vha, struct mbx_cmd_32 *mcp)
{
int rval;
unsigned long flags = 0;
device_reg_t *reg;
uint8_t abort_active;
uint8_t io_lock_on;
uint16_t command = 0;
uint32_t *iptr;
uint32_t __iomem *optr;
uint32_t cnt;
uint32_t mboxes;
unsigned long wait_time;
struct qla_hw_data *ha = vha->hw;
scsi_qla_host_t *base_vha = pci_get_drvdata(ha->pdev);
if (ha->pdev->error_state > pci_channel_io_frozen) {
ql_log(ql_log_warn, vha, 0x115c,
"error_state is greater than pci_channel_io_frozen, "
"exiting.\n");
return QLA_FUNCTION_TIMEOUT;
}
if (vha->device_flags & DFLG_DEV_FAILED) {
ql_log(ql_log_warn, vha, 0x115f,
"Device in failed state, exiting.\n");
return QLA_FUNCTION_TIMEOUT;
}
reg = ha->iobase;
io_lock_on = base_vha->flags.init_done;
rval = QLA_SUCCESS;
abort_active = test_bit(ABORT_ISP_ACTIVE, &base_vha->dpc_flags);
if (ha->flags.pci_channel_io_perm_failure) {
ql_log(ql_log_warn, vha, 0x1175,
"Perm failure on EEH timeout MBX, exiting.\n");
return QLA_FUNCTION_TIMEOUT;
}
if (ha->flags.isp82xx_fw_hung) {
/* Setting Link-Down error */
mcp->mb[0] = MBS_LINK_DOWN_ERROR;
ql_log(ql_log_warn, vha, 0x1176,
"FW hung = %d.\n", ha->flags.isp82xx_fw_hung);
rval = QLA_FUNCTION_FAILED;
goto premature_exit;
}
/*
* Wait for active mailbox commands to finish by waiting at most tov
* seconds. This is to serialize actual issuing of mailbox cmds during
* non ISP abort time.
*/
if (!wait_for_completion_timeout(&ha->mbx_cmd_comp, mcp->tov * HZ)) {
/* Timeout occurred. Return error. */
ql_log(ql_log_warn, vha, 0x1177,
"Cmd access timeout, cmd=0x%x, Exiting.\n",
mcp->mb[0]);
return QLA_FUNCTION_TIMEOUT;
}
ha->flags.mbox_busy = 1;
/* Save mailbox command for debug */
ha->mcp32 = mcp;
ql_dbg(ql_dbg_mbx, vha, 0x1178,
"Prepare to issue mbox cmd=0x%x.\n", mcp->mb[0]);
spin_lock_irqsave(&ha->hardware_lock, flags);
/* Load mailbox registers. */
optr = (uint32_t __iomem *)&reg->ispfx00.mailbox0;
iptr = mcp->mb;
command = mcp->mb[0];
mboxes = mcp->out_mb;
for (cnt = 0; cnt < ha->mbx_count; cnt++) {
if (mboxes & BIT_0)
WRT_REG_DWORD(optr, *iptr);
mboxes >>= 1;
optr++;
iptr++;
}
/* Issue set host interrupt command to send cmd out. */
ha->flags.mbox_int = 0;
clear_bit(MBX_INTERRUPT, &ha->mbx_cmd_flags);
ql_dump_buffer(ql_dbg_mbx + ql_dbg_buffer, vha, 0x1172,
(uint8_t *)mcp->mb, 16);
ql_dump_buffer(ql_dbg_mbx + ql_dbg_buffer, vha, 0x1173,
((uint8_t *)mcp->mb + 0x10), 16);
ql_dump_buffer(ql_dbg_mbx + ql_dbg_buffer, vha, 0x1174,
((uint8_t *)mcp->mb + 0x20), 8);
/* Unlock mbx registers and wait for interrupt */
ql_dbg(ql_dbg_mbx, vha, 0x1179,
"Going to unlock irq & waiting for interrupts. "
"jiffies=%lx.\n", jiffies);
/* Wait for mbx cmd completion until timeout */
if ((!abort_active && io_lock_on) || IS_NOPOLLING_TYPE(ha)) {
set_bit(MBX_INTR_WAIT, &ha->mbx_cmd_flags);
QLAFX00_SET_HST_INTR(ha, ha->mbx_intr_code);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
wait_for_completion_timeout(&ha->mbx_intr_comp, mcp->tov * HZ);
} else {
ql_dbg(ql_dbg_mbx, vha, 0x112c,
"Cmd=%x Polling Mode.\n", command);
QLAFX00_SET_HST_INTR(ha, ha->mbx_intr_code);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
wait_time = jiffies + mcp->tov * HZ; /* wait at most tov secs */
while (!ha->flags.mbox_int) {
if (time_after(jiffies, wait_time))
break;
/* Check for pending interrupts. */
qla2x00_poll(ha->rsp_q_map[0]);
if (!ha->flags.mbox_int &&
!(IS_QLA2200(ha) &&
command == MBC_LOAD_RISC_RAM_EXTENDED))
usleep_range(10000, 11000);
} /* while */
ql_dbg(ql_dbg_mbx, vha, 0x112d,
"Waited %d sec.\n",
(uint)((jiffies - (wait_time - (mcp->tov * HZ)))/HZ));
}
/* Check whether we timed out */
if (ha->flags.mbox_int) {
uint32_t *iptr2;
ql_dbg(ql_dbg_mbx, vha, 0x112e,
"Cmd=%x completed.\n", command);
/* Got interrupt. Clear the flag. */
ha->flags.mbox_int = 0;
clear_bit(MBX_INTERRUPT, &ha->mbx_cmd_flags);
if (ha->mailbox_out32[0] != MBS_COMMAND_COMPLETE)
rval = QLA_FUNCTION_FAILED;
/* Load return mailbox registers. */
iptr2 = mcp->mb;
iptr = (uint32_t *)&ha->mailbox_out32[0];
mboxes = mcp->in_mb;
for (cnt = 0; cnt < ha->mbx_count; cnt++) {
if (mboxes & BIT_0)
*iptr2 = *iptr;
mboxes >>= 1;
iptr2++;
iptr++;
}
} else {
rval = QLA_FUNCTION_TIMEOUT;
}
ha->flags.mbox_busy = 0;
/* Clean up */
ha->mcp32 = NULL;
if ((abort_active || !io_lock_on) && !IS_NOPOLLING_TYPE(ha)) {
ql_dbg(ql_dbg_mbx, vha, 0x113a,
"checking for additional resp interrupt.\n");
/* polling mode for non isp_abort commands. */
qla2x00_poll(ha->rsp_q_map[0]);
}
if (rval == QLA_FUNCTION_TIMEOUT &&
mcp->mb[0] != MBC_GEN_SYSTEM_ERROR) {
if (!io_lock_on || (mcp->flags & IOCTL_CMD) ||
ha->flags.eeh_busy) {
/* not in dpc. schedule it for dpc to take over. */
ql_dbg(ql_dbg_mbx, vha, 0x115d,
"Timeout, schedule isp_abort_needed.\n");
if (!test_bit(ISP_ABORT_NEEDED, &vha->dpc_flags) &&
!test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags) &&
!test_bit(ISP_ABORT_RETRY, &vha->dpc_flags)) {
ql_log(ql_log_info, base_vha, 0x115e,
"Mailbox cmd timeout occurred, cmd=0x%x, "
"mb[0]=0x%x, eeh_busy=0x%x. Scheduling ISP "
"abort.\n", command, mcp->mb[0],
ha->flags.eeh_busy);
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
}
} else if (!abort_active) {
/* call abort directly since we are in the DPC thread */
ql_dbg(ql_dbg_mbx, vha, 0x1160,
"Timeout, calling abort_isp.\n");
if (!test_bit(ISP_ABORT_NEEDED, &vha->dpc_flags) &&
!test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags) &&
!test_bit(ISP_ABORT_RETRY, &vha->dpc_flags)) {
ql_log(ql_log_info, base_vha, 0x1161,
"Mailbox cmd timeout occurred, cmd=0x%x, "
"mb[0]=0x%x. Scheduling ISP abort ",
command, mcp->mb[0]);
set_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags);
clear_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
if (ha->isp_ops->abort_isp(vha)) {
/* Failed. retry later. */
set_bit(ISP_ABORT_NEEDED,
&vha->dpc_flags);
}
clear_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags);
ql_dbg(ql_dbg_mbx, vha, 0x1162,
"Finished abort_isp.\n");
}
}
}
premature_exit:
/* Allow next mbx cmd to come in. */
complete(&ha->mbx_cmd_comp);
if (rval) {
ql_log(ql_log_warn, base_vha, 0x1163,
"**** Failed mbx[0]=%x, mb[1]=%x, mb[2]=%x, "
"mb[3]=%x, cmd=%x ****.\n",
mcp->mb[0], mcp->mb[1], mcp->mb[2], mcp->mb[3], command);
} else {
ql_dbg(ql_dbg_mbx, base_vha, 0x1164, "Done %s.\n", __func__);
}
return rval;
}
/*
* qlafx00_driver_shutdown
* Indicate a driver shutdown to firmware.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* local function return status code.
*
* Context:
* Kernel context.
*/
int
qlafx00_driver_shutdown(scsi_qla_host_t *vha, int tmo)
{
int rval;
struct mbx_cmd_32 mc;
struct mbx_cmd_32 *mcp = &mc;
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1166,
"Entered %s.\n", __func__);
mcp->mb[0] = MBC_MR_DRV_SHUTDOWN;
mcp->out_mb = MBX_0;
mcp->in_mb = MBX_0;
if (tmo)
mcp->tov = tmo;
else
mcp->tov = MBX_TOV_SECONDS;
mcp->flags = 0;
rval = qlafx00_mailbox_command(vha, mcp);
if (rval != QLA_SUCCESS) {
ql_dbg(ql_dbg_mbx, vha, 0x1167,
"Failed=%x.\n", rval);
} else {
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1168,
"Done %s.\n", __func__);
}
return rval;
}
/*
* qlafx00_get_firmware_state
* Get adapter firmware state.
*
* Input:
* ha = adapter block pointer.
* TARGET_QUEUE_LOCK must be released.
* ADAPTER_STATE_LOCK must be released.
*
* Returns:
* qla7xxx local function return status code.
*
* Context:
* Kernel context.
*/
static int
qlafx00_get_firmware_state(scsi_qla_host_t *vha, uint32_t *states)
{
int rval;
struct mbx_cmd_32 mc;
struct mbx_cmd_32 *mcp = &mc;
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1169,
"Entered %s.\n", __func__);
mcp->mb[0] = MBC_GET_FIRMWARE_STATE;
mcp->out_mb = MBX_0;
mcp->in_mb = MBX_1|MBX_0;
mcp->tov = MBX_TOV_SECONDS;
mcp->flags = 0;
rval = qlafx00_mailbox_command(vha, mcp);
/* Return firmware states. */
states[0] = mcp->mb[1];
if (rval != QLA_SUCCESS) {
ql_dbg(ql_dbg_mbx, vha, 0x116a,
"Failed=%x mb[0]=%x.\n", rval, mcp->mb[0]);
} else {
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x116b,
"Done %s.\n", __func__);
}
return rval;
}
/*
* qlafx00_init_firmware
* Initialize adapter firmware.
*
* Input:
* ha = adapter block pointer.
* dptr = Initialization control block pointer.
* size = size of initialization control block.
* TARGET_QUEUE_LOCK must be released.
* ADAPTER_STATE_LOCK must be released.
*
* Returns:
* qlafx00 local function return status code.
*
* Context:
* Kernel context.
*/
int
qlafx00_init_firmware(scsi_qla_host_t *vha, uint16_t size)
{
int rval;
struct mbx_cmd_32 mc;
struct mbx_cmd_32 *mcp = &mc;
struct qla_hw_data *ha = vha->hw;
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x116c,
"Entered %s.\n", __func__);
mcp->mb[0] = MBC_INITIALIZE_FIRMWARE;
mcp->mb[1] = 0;
mcp->mb[2] = MSD(ha->init_cb_dma);
mcp->mb[3] = LSD(ha->init_cb_dma);
mcp->out_mb = MBX_3|MBX_2|MBX_1|MBX_0;
mcp->in_mb = MBX_0;
mcp->buf_size = size;
mcp->flags = MBX_DMA_OUT;
mcp->tov = MBX_TOV_SECONDS;
rval = qlafx00_mailbox_command(vha, mcp);
if (rval != QLA_SUCCESS) {
ql_dbg(ql_dbg_mbx, vha, 0x116d,
"Failed=%x mb[0]=%x.\n", rval, mcp->mb[0]);
} else {
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x116e,
"Done %s.\n", __func__);
}
return rval;
}
/*
* qlafx00_mbx_reg_test
*/
static int
qlafx00_mbx_reg_test(scsi_qla_host_t *vha)
{
int rval;
struct mbx_cmd_32 mc;
struct mbx_cmd_32 *mcp = &mc;
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x116f,
"Entered %s.\n", __func__);
mcp->mb[0] = MBC_MAILBOX_REGISTER_TEST;
mcp->mb[1] = 0xAAAA;
mcp->mb[2] = 0x5555;
mcp->mb[3] = 0xAA55;
mcp->mb[4] = 0x55AA;
mcp->mb[5] = 0xA5A5;
mcp->mb[6] = 0x5A5A;
mcp->mb[7] = 0x2525;
mcp->mb[8] = 0xBBBB;
mcp->mb[9] = 0x6666;
mcp->mb[10] = 0xBB66;
mcp->mb[11] = 0x66BB;
mcp->mb[12] = 0xB6B6;
mcp->mb[13] = 0x6B6B;
mcp->mb[14] = 0x3636;
mcp->mb[15] = 0xCCCC;
mcp->out_mb = MBX_15|MBX_14|MBX_13|MBX_12|MBX_11|MBX_10|MBX_9|MBX_8|
MBX_7|MBX_6|MBX_5|MBX_4|MBX_3|MBX_2|MBX_1|MBX_0;
mcp->in_mb = MBX_15|MBX_14|MBX_13|MBX_12|MBX_11|MBX_10|MBX_9|MBX_8|
MBX_7|MBX_6|MBX_5|MBX_4|MBX_3|MBX_2|MBX_1|MBX_0;
mcp->buf_size = 0;
mcp->flags = MBX_DMA_OUT;
mcp->tov = MBX_TOV_SECONDS;
rval = qlafx00_mailbox_command(vha, mcp);
if (rval == QLA_SUCCESS) {
if (mcp->mb[17] != 0xAAAA || mcp->mb[18] != 0x5555 ||
mcp->mb[19] != 0xAA55 || mcp->mb[20] != 0x55AA)
rval = QLA_FUNCTION_FAILED;
if (mcp->mb[21] != 0xA5A5 || mcp->mb[22] != 0x5A5A ||
mcp->mb[23] != 0x2525 || mcp->mb[24] != 0xBBBB)
rval = QLA_FUNCTION_FAILED;
if (mcp->mb[25] != 0x6666 || mcp->mb[26] != 0xBB66 ||
mcp->mb[27] != 0x66BB || mcp->mb[28] != 0xB6B6)
rval = QLA_FUNCTION_FAILED;
if (mcp->mb[29] != 0x6B6B || mcp->mb[30] != 0x3636 ||
mcp->mb[31] != 0xCCCC)
rval = QLA_FUNCTION_FAILED;
}
if (rval != QLA_SUCCESS) {
ql_dbg(ql_dbg_mbx, vha, 0x1170,
"Failed=%x mb[0]=%x.\n", rval, mcp->mb[0]);
} else {
ql_dbg(ql_dbg_mbx + ql_dbg_verbose, vha, 0x1171,
"Done %s.\n", __func__);
}
return rval;
}
/**
* qlafx00_pci_config() - Setup ISPFx00 PCI configuration registers.
* @ha: HA context
*
* Returns 0 on success.
*/
int
qlafx00_pci_config(scsi_qla_host_t *vha)
{
uint16_t w;
struct qla_hw_data *ha = vha->hw;
pci_set_master(ha->pdev);
pci_try_set_mwi(ha->pdev);
pci_read_config_word(ha->pdev, PCI_COMMAND, &w);
w |= (PCI_COMMAND_PARITY | PCI_COMMAND_SERR);
w &= ~PCI_COMMAND_INTX_DISABLE;
pci_write_config_word(ha->pdev, PCI_COMMAND, w);
/* PCIe -- adjust Maximum Read Request Size (2048). */
if (pci_is_pcie(ha->pdev))
pcie_set_readrq(ha->pdev, 2048);
ha->chip_revision = ha->pdev->revision;
return QLA_SUCCESS;
}
/**
* qlafx00_warm_reset() - Perform warm reset of iSA(CPUs being reset on SOC).
* @ha: HA context
*
*/
static inline void
qlafx00_soc_cpu_reset(scsi_qla_host_t *vha)
{
unsigned long flags = 0;
struct qla_hw_data *ha = vha->hw;
int i, core;
uint32_t cnt;
uint32_t reg_val;
spin_lock_irqsave(&ha->hardware_lock, flags);
QLAFX00_SET_HBA_SOC_REG(ha, 0x80004, 0);
QLAFX00_SET_HBA_SOC_REG(ha, 0x82004, 0);
/* stop the XOR DMA engines */
QLAFX00_SET_HBA_SOC_REG(ha, 0x60920, 0x02);
QLAFX00_SET_HBA_SOC_REG(ha, 0x60924, 0x02);
QLAFX00_SET_HBA_SOC_REG(ha, 0xf0920, 0x02);
QLAFX00_SET_HBA_SOC_REG(ha, 0xf0924, 0x02);
/* stop the IDMA engines */
reg_val = QLAFX00_GET_HBA_SOC_REG(ha, 0x60840);
reg_val &= ~(1<<12);
QLAFX00_SET_HBA_SOC_REG(ha, 0x60840, reg_val);
reg_val = QLAFX00_GET_HBA_SOC_REG(ha, 0x60844);
reg_val &= ~(1<<12);
QLAFX00_SET_HBA_SOC_REG(ha, 0x60844, reg_val);
reg_val = QLAFX00_GET_HBA_SOC_REG(ha, 0x60848);
reg_val &= ~(1<<12);
QLAFX00_SET_HBA_SOC_REG(ha, 0x60848, reg_val);
reg_val = QLAFX00_GET_HBA_SOC_REG(ha, 0x6084C);
reg_val &= ~(1<<12);
QLAFX00_SET_HBA_SOC_REG(ha, 0x6084C, reg_val);
for (i = 0; i < 100000; i++) {
if ((QLAFX00_GET_HBA_SOC_REG(ha, 0xd0000) & 0x10000000) == 0 &&
(QLAFX00_GET_HBA_SOC_REG(ha, 0x10600) & 0x1) == 0)
break;
udelay(100);
}
/* Set all 4 cores in reset */
for (i = 0; i < 4; i++) {
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_SW_RST_CONTROL_REG_CORE0 + 8*i), (0xF01));
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_SW_RST_CONTROL_REG_CORE0 + 4 + 8*i), (0x01010101));
}
/* Reset all units in Fabric */
QLAFX00_SET_HBA_SOC_REG(ha, SOC_FABRIC_RST_CONTROL_REG, (0x011f0101));
/* */
QLAFX00_SET_HBA_SOC_REG(ha, 0x10610, 1);
QLAFX00_SET_HBA_SOC_REG(ha, 0x10600, 0);
/* Set all 4 core Memory Power Down Registers */
for (i = 0; i < 5; i++) {
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_PWR_MANAGEMENT_PWR_DOWN_REG + 4*i), (0x0));
}
/* Reset all interrupt control registers */
for (i = 0; i < 115; i++) {
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_INTERRUPT_SOURCE_I_CONTROL_REG + 4*i), (0x0));
}
/* Reset Timers control registers. per core */
for (core = 0; core < 4; core++)
for (i = 0; i < 8; i++)
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_CORE_TIMER_REG + 0x100*core + 4*i), (0x0));
/* Reset per core IRQ ack register */
for (core = 0; core < 4; core++)
QLAFX00_SET_HBA_SOC_REG(ha,
(SOC_IRQ_ACK_REG + 0x100*core), (0x3FF));
/* Set Fabric control and config to defaults */
QLAFX00_SET_HBA_SOC_REG(ha, SOC_FABRIC_CONTROL_REG, (0x2));
QLAFX00_SET_HBA_SOC_REG(ha, SOC_FABRIC_CONFIG_REG, (0x3));
/* Kick in Fabric units */
QLAFX00_SET_HBA_SOC_REG(ha, SOC_FABRIC_RST_CONTROL_REG, (0x0));
/* Kick in Core0 to start boot process */
QLAFX00_SET_HBA_SOC_REG(ha, SOC_SW_RST_CONTROL_REG_CORE0, (0xF00));
spin_unlock_irqrestore(&ha->hardware_lock, flags);
/* Wait 10secs for soft-reset to complete. */
for (cnt = 10; cnt; cnt--) {
msleep(1000);
barrier();
}
}
/**
* qlafx00_soft_reset() - Soft Reset ISPFx00.
* @ha: HA context
*
* Returns 0 on success.
*/
void
qlafx00_soft_reset(scsi_qla_host_t *vha)
{
struct qla_hw_data *ha = vha->hw;
if (unlikely(pci_channel_offline(ha->pdev) &&
ha->flags.pci_channel_io_perm_failure))
return;
ha->isp_ops->disable_intrs(ha);
qlafx00_soc_cpu_reset(vha);
}
/**
* qlafx00_chip_diag() - Test ISPFx00 for proper operation.
* @ha: HA context
*
* Returns 0 on success.
*/
int
qlafx00_chip_diag(scsi_qla_host_t *vha)
{
int rval = 0;
struct qla_hw_data *ha = vha->hw;
struct req_que *req = ha->req_q_map[0];
ha->fw_transfer_size = REQUEST_ENTRY_SIZE * req->length;
rval = qlafx00_mbx_reg_test(vha);
if (rval) {
ql_log(ql_log_warn, vha, 0x1165,
"Failed mailbox send register test\n");
} else {
/* Flag a successful rval */
rval = QLA_SUCCESS;
}
return rval;
}
void
qlafx00_config_rings(struct scsi_qla_host *vha)
{
struct qla_hw_data *ha = vha->hw;
struct device_reg_fx00 __iomem *reg = &ha->iobase->ispfx00;
WRT_REG_DWORD(&reg->req_q_in, 0);
WRT_REG_DWORD(&reg->req_q_out, 0);
WRT_REG_DWORD(&reg->rsp_q_in, 0);
WRT_REG_DWORD(&reg->rsp_q_out, 0);
/* PCI posting */
RD_REG_DWORD(&reg->rsp_q_out);
}
char *
qlafx00_pci_info_str(struct scsi_qla_host *vha, char *str)
{
struct qla_hw_data *ha = vha->hw;
if (pci_is_pcie(ha->pdev)) {
strcpy(str, "PCIe iSA");
return str;
}
return str;
}
char *
qlafx00_fw_version_str(struct scsi_qla_host *vha, char *str, size_t size)
{
struct qla_hw_data *ha = vha->hw;
snprintf(str, size, "%s", ha->mr.fw_version);
return str;
}
void
qlafx00_enable_intrs(struct qla_hw_data *ha)
{
unsigned long flags = 0;
spin_lock_irqsave(&ha->hardware_lock, flags);
ha->interrupts_on = 1;
QLAFX00_ENABLE_ICNTRL_REG(ha);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
}
void
qlafx00_disable_intrs(struct qla_hw_data *ha)
{
unsigned long flags = 0;
spin_lock_irqsave(&ha->hardware_lock, flags);
ha->interrupts_on = 0;
QLAFX00_DISABLE_ICNTRL_REG(ha);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
}
int
qlafx00_abort_target(fc_port_t *fcport, uint64_t l, int tag)
{
return qla2x00_async_tm_cmd(fcport, TCF_TARGET_RESET, l, tag);
}
int
qlafx00_lun_reset(fc_port_t *fcport, uint64_t l, int tag)
{
return qla2x00_async_tm_cmd(fcport, TCF_LUN_RESET, l, tag);
}
int
qlafx00_loop_reset(scsi_qla_host_t *vha)
{
int ret;
struct fc_port *fcport;
struct qla_hw_data *ha = vha->hw;
if (ql2xtargetreset) {
list_for_each_entry(fcport, &vha->vp_fcports, list) {
if (fcport->port_type != FCT_TARGET)
continue;
ret = ha->isp_ops->target_reset(fcport, 0, 0);
if (ret != QLA_SUCCESS) {
ql_dbg(ql_dbg_taskm, vha, 0x803d,
"Bus Reset failed: Reset=%d "
"d_id=%x.\n", ret, fcport->d_id.b24);
}
}
}
return QLA_SUCCESS;
}
int
qlafx00_iospace_config(struct qla_hw_data *ha)
{
if (pci_request_selected_regions(ha->pdev, ha->bars,
QLA2XXX_DRIVER_NAME)) {
ql_log_pci(ql_log_fatal, ha->pdev, 0x014e,
"Failed to reserve PIO/MMIO regions (%s), aborting.\n",
pci_name(ha->pdev));
goto iospace_error_exit;
}
/* Use MMIO operations for all accesses. */
if (!(pci_resource_flags(ha->pdev, 0) & IORESOURCE_MEM)) {
ql_log_pci(ql_log_warn, ha->pdev, 0x014f,
"Invalid pci I/O region size (%s).\n",
pci_name(ha->pdev));
goto iospace_error_exit;
}
if (pci_resource_len(ha->pdev, 0) < BAR0_LEN_FX00) {
ql_log_pci(ql_log_warn, ha->pdev, 0x0127,
"Invalid PCI mem BAR0 region size (%s), aborting\n",
pci_name(ha->pdev));
goto iospace_error_exit;
}
ha->cregbase =
ioremap_nocache(pci_resource_start(ha->pdev, 0), BAR0_LEN_FX00);
if (!ha->cregbase) {
ql_log_pci(ql_log_fatal, ha->pdev, 0x0128,
"cannot remap MMIO (%s), aborting\n", pci_name(ha->pdev));
goto iospace_error_exit;
}
if (!(pci_resource_flags(ha->pdev, 2) & IORESOURCE_MEM)) {
ql_log_pci(ql_log_warn, ha->pdev, 0x0129,
"region #2 not an MMIO resource (%s), aborting\n",
pci_name(ha->pdev));
goto iospace_error_exit;
}
if (pci_resource_len(ha->pdev, 2) < BAR2_LEN_FX00) {
ql_log_pci(ql_log_warn, ha->pdev, 0x012a,
"Invalid PCI mem BAR2 region size (%s), aborting\n",
pci_name(ha->pdev));
goto iospace_error_exit;
}
ha->iobase =
ioremap_nocache(pci_resource_start(ha->pdev, 2), BAR2_LEN_FX00);
if (!ha->iobase) {
ql_log_pci(ql_log_fatal, ha->pdev, 0x012b,
"cannot remap MMIO (%s), aborting\n", pci_name(ha->pdev));
goto iospace_error_exit;
}
/* Determine queue resources */
ha->max_req_queues = ha->max_rsp_queues = 1;
ql_log_pci(ql_log_info, ha->pdev, 0x012c,
"Bars 0x%x, iobase0 0x%p, iobase2 0x%p\n",
ha->bars, ha->cregbase, ha->iobase);
return 0;
iospace_error_exit:
return -ENOMEM;
}
static void
qlafx00_save_queue_ptrs(struct scsi_qla_host *vha)
{
struct qla_hw_data *ha = vha->hw;
struct req_que *req = ha->req_q_map[0];
struct rsp_que *rsp = ha->rsp_q_map[0];
req->length_fx00 = req->length;
req->ring_fx00 = req->ring;
req->dma_fx00 = req->dma;
rsp->length_fx00 = rsp->length;
rsp->ring_fx00 = rsp->ring;
rsp->dma_fx00 = rsp->dma;
ql_dbg(ql_dbg_init, vha, 0x012d,
"req: %p, ring_fx00: %p, length_fx00: 0x%x,"
"req->dma_fx00: 0x%llx\n", req, req->ring_fx00,
req->length_fx00, (u64)req->dma_fx00);
ql_dbg(ql_dbg_init, vha, 0x012e,
"rsp: %p, ring_fx00: %p, length_fx00: 0x%x,"
"rsp->dma_fx00: 0x%llx\n", rsp, rsp->ring_fx00,
rsp->length_fx00, (u64)rsp->dma_fx00);
}
static int
qlafx00_config_queues(struct scsi_qla_host *vha)
{
struct qla_hw_data *ha = vha->hw;
struct req_que *req = ha->req_q_map[0];
struct rsp_que *rsp = ha->rsp_q_map[0];
dma_addr_t bar2_hdl = pci_resource_start(ha->pdev, 2);
req->length = ha->req_que_len;
req->ring = (void *)ha->iobase + ha->req_que_off;
req->dma = bar2_hdl + ha->req_que_off;
if ((!req->ring) || (req->length == 0)) {
ql_log_pci(ql_log_info, ha->pdev, 0x012f,
"Unable to allocate memory for req_ring\n");
return QLA_FUNCTION_FAILED;
}
ql_dbg(ql_dbg_init, vha, 0x0130,
"req: %p req_ring pointer %p req len 0x%x "
"req off 0x%x\n, req->dma: 0x%llx",
req, req->ring, req->length,
ha->req_que_off, (u64)req->dma);
rsp->length = ha->rsp_que_len;
rsp->ring = (void *)ha->iobase + ha->rsp_que_off;
rsp->dma = bar2_hdl + ha->rsp_que_off;
if ((!rsp->ring) || (rsp->length == 0)) {
ql_log_pci(ql_log_info, ha->pdev, 0x0131,
"Unable to allocate memory for rsp_ring\n");
return QLA_FUNCTION_FAILED;
}
ql_dbg(ql_dbg_init, vha, 0x0132,
"rsp: %p rsp_ring pointer %p rsp len 0x%x "
"rsp off 0x%x, rsp->dma: 0x%llx\n",
rsp, rsp->ring, rsp->length,
ha->rsp_que_off, (u64)rsp->dma);
return QLA_SUCCESS;
}
static int
qlafx00_init_fw_ready(scsi_qla_host_t *vha)
{
int rval = 0;
unsigned long wtime;
uint16_t wait_time; /* Wait time */
struct qla_hw_data *ha = vha->hw;
struct device_reg_fx00 __iomem *reg = &ha->iobase->ispfx00;
uint32_t aenmbx, aenmbx7 = 0;
uint32_t pseudo_aen;
uint32_t state[5];
bool done = false;
/* 30 seconds wait - Adjust if required */
wait_time = 30;
pseudo_aen = RD_REG_DWORD(&reg->pseudoaen);
if (pseudo_aen == 1) {
aenmbx7 = RD_REG_DWORD(&reg->initval7);
ha->mbx_intr_code = MSW(aenmbx7);
ha->rqstq_intr_code = LSW(aenmbx7);
rval = qlafx00_driver_shutdown(vha, 10);
if (rval != QLA_SUCCESS)
qlafx00_soft_reset(vha);
}
/* wait time before firmware ready */
wtime = jiffies + (wait_time * HZ);
do {
aenmbx = RD_REG_DWORD(&reg->aenmailbox0);
barrier();
ql_dbg(ql_dbg_mbx, vha, 0x0133,
"aenmbx: 0x%x\n", aenmbx);
switch (aenmbx) {
case MBA_FW_NOT_STARTED:
case MBA_FW_STARTING:
break;
case MBA_SYSTEM_ERR:
case MBA_REQ_TRANSFER_ERR:
case MBA_RSP_TRANSFER_ERR:
case MBA_FW_INIT_FAILURE:
qlafx00_soft_reset(vha);
break;
case MBA_FW_RESTART_CMPLT:
/* Set the mbx and rqstq intr code */
aenmbx7 = RD_REG_DWORD(&reg->aenmailbox7);
ha->mbx_intr_code = MSW(aenmbx7);
ha->rqstq_intr_code = LSW(aenmbx7);
ha->req_que_off = RD_REG_DWORD(&reg->aenmailbox1);
ha->rsp_que_off = RD_REG_DWORD(&reg->aenmailbox3);
ha->req_que_len = RD_REG_DWORD(&reg->aenmailbox5);
ha->rsp_que_len = RD_REG_DWORD(&reg->aenmailbox6);
WRT_REG_DWORD(&reg->aenmailbox0, 0);
RD_REG_DWORD_RELAXED(&reg->aenmailbox0);
ql_dbg(ql_dbg_init, vha, 0x0134,
"f/w returned mbx_intr_code: 0x%x, "
"rqstq_intr_code: 0x%x\n",
ha->mbx_intr_code, ha->rqstq_intr_code);
QLAFX00_CLR_INTR_REG(ha, QLAFX00_HST_INT_STS_BITS);
rval = QLA_SUCCESS;
done = true;
break;
default:
if ((aenmbx & 0xFF00) == MBA_FW_INIT_INPROGRESS)
break;
/* If fw is apparently not ready. In order to continue,
* we might need to issue Mbox cmd, but the problem is
* that the DoorBell vector values that come with the
* 8060 AEN are most likely gone by now (and thus no
* bell would be rung on the fw side when mbox cmd is
* issued). We have to therefore grab the 8060 AEN
* shadow regs (filled in by FW when the last 8060
* AEN was being posted).
* Do the following to determine what is needed in
* order to get the FW ready:
* 1. reload the 8060 AEN values from the shadow regs
* 2. clear int status to get rid of possible pending
* interrupts
* 3. issue Get FW State Mbox cmd to determine fw state
* Set the mbx and rqstq intr code from Shadow Regs
*/
aenmbx7 = RD_REG_DWORD(&reg->initval7);
ha->mbx_intr_code = MSW(aenmbx7);
ha->rqstq_intr_code = LSW(aenmbx7);
ha->req_que_off = RD_REG_DWORD(&reg->initval1);
ha->rsp_que_off = RD_REG_DWORD(&reg->initval3);
ha->req_que_len = RD_REG_DWORD(&reg->initval5);
ha->rsp_que_len = RD_REG_DWORD(&reg->initval6);
ql_dbg(ql_dbg_init, vha, 0x0135,
"f/w returned mbx_intr_code: 0x%x, "
"rqstq_intr_code: 0x%x\n",
ha->mbx_intr_code, ha->rqstq_intr_code);
QLAFX00_CLR_INTR_REG(ha, QLAFX00_HST_INT_STS_BITS);
/* Get the FW state */
rval = qlafx00_get_firmware_state(vha, state);
if (rval != QLA_SUCCESS) {
/* Retry if timer has not expired */
break;
}
if (state[0] == FSTATE_FX00_CONFIG_WAIT) {
/* Firmware is waiting to be
* initialized by driver
*/
rval = QLA_SUCCESS;
done = true;
break;
}
/* Issue driver shutdown and wait until f/w recovers.
* Driver should continue to poll until 8060 AEN is
* received indicating firmware recovery.
*/
ql_dbg(ql_dbg_init, vha, 0x0136,
"Sending Driver shutdown fw_state 0x%x\n",
state[0]);
rval = qlafx00_driver_shutdown(vha, 10);
if (rval != QLA_SUCCESS) {
rval = QLA_FUNCTION_FAILED;
break;
}
msleep(500);
wtime = jiffies + (wait_time * HZ);
break;
}
if (!done) {
if (time_after_eq(jiffies, wtime)) {
ql_dbg(ql_dbg_init, vha, 0x0137,
"Init f/w failed: aen[7]: 0x%x\n",
RD_REG_DWORD(&reg->aenmailbox7));
rval = QLA_FUNCTION_FAILED;
done = true;
break;
}
/* Delay for a while */
msleep(500);
}
} while (!done);
if (rval)
ql_dbg(ql_dbg_init, vha, 0x0138,
"%s **** FAILED ****.\n", __func__);
else
ql_dbg(ql_dbg_init, vha, 0x0139,
"%s **** SUCCESS ****.\n", __func__);
return rval;
}
/*
* qlafx00_fw_ready() - Waits for firmware ready.
* @ha: HA context
*
* Returns 0 on success.
*/
int
qlafx00_fw_ready(scsi_qla_host_t *vha)
{
int rval;
unsigned long wtime;
uint16_t wait_time; /* Wait time if loop is coming ready */
uint32_t state[5];
rval = QLA_SUCCESS;
wait_time = 10;
/* wait time before firmware ready */
wtime = jiffies + (wait_time * HZ);
/* Wait for ISP to finish init */
if (!vha->flags.init_done)
ql_dbg(ql_dbg_init, vha, 0x013a,
"Waiting for init to complete...\n");
do {
rval = qlafx00_get_firmware_state(vha, state);
if (rval == QLA_SUCCESS) {
if (state[0] == FSTATE_FX00_INITIALIZED) {
ql_dbg(ql_dbg_init, vha, 0x013b,
"fw_state=%x\n", state[0]);
rval = QLA_SUCCESS;
break;
}
}
rval = QLA_FUNCTION_FAILED;
if (time_after_eq(jiffies, wtime))
break;
/* Delay for a while */
msleep(500);
ql_dbg(ql_dbg_init, vha, 0x013c,
"fw_state=%x curr time=%lx.\n", state[0], jiffies);
} while (1);
if (rval)
ql_dbg(ql_dbg_init, vha, 0x013d,
"Firmware ready **** FAILED ****.\n");
else
ql_dbg(ql_dbg_init, vha, 0x013e,
"Firmware ready **** SUCCESS ****.\n");
return rval;
}
static int
qlafx00_find_all_targets(scsi_qla_host_t *vha,
struct list_head *new_fcports)
{
int rval;
uint16_t tgt_id;
fc_port_t *fcport, *new_fcport;
int found;
struct qla_hw_data *ha = vha->hw;
rval = QLA_SUCCESS;
if (!test_bit(LOOP_RESYNC_ACTIVE, &vha->dpc_flags))
return QLA_FUNCTION_FAILED;
if ((atomic_read(&vha->loop_down_timer) ||
STATE_TRANSITION(vha))) {
atomic_set(&vha->loop_down_timer, 0);
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
return QLA_FUNCTION_FAILED;
}
ql_dbg(ql_dbg_disc + ql_dbg_init, vha, 0x2088,
"Listing Target bit map...\n");
ql_dump_buffer(ql_dbg_disc + ql_dbg_init, vha,
0x2089, (uint8_t *)ha->gid_list, 32);
/* Allocate temporary rmtport for any new rmtports discovered. */
new_fcport = qla2x00_alloc_fcport(vha, GFP_KERNEL);
if (new_fcport == NULL)
return QLA_MEMORY_ALLOC_FAILED;
for_each_set_bit(tgt_id, (void *)ha->gid_list,
QLAFX00_TGT_NODE_LIST_SIZE) {
/* Send get target node info */
new_fcport->tgt_id = tgt_id;
rval = qlafx00_fx_disc(vha, new_fcport,
FXDISC_GET_TGT_NODE_INFO);
if (rval != QLA_SUCCESS) {
ql_log(ql_log_warn, vha, 0x208a,
"Target info scan failed -- assuming zero-entry "
"result...\n");
continue;
}
/* Locate matching device in database. */
found = 0;
list_for_each_entry(fcport, &vha->vp_fcports, list) {
if (memcmp(new_fcport->port_name,
fcport->port_name, WWN_SIZE))
continue;
found++;
/*
* If tgt_id is same and state FCS_ONLINE, nothing
* changed.
*/
if (fcport->tgt_id == new_fcport->tgt_id &&
atomic_read(&fcport->state) == FCS_ONLINE)
break;
/*
* Tgt ID changed or device was marked to be updated.
*/
ql_dbg(ql_dbg_disc + ql_dbg_init, vha, 0x208b,
"TGT-ID Change(%s): Present tgt id: "
"0x%x state: 0x%x "
"wwnn = %llx wwpn = %llx.\n",
__func__, fcport->tgt_id,
atomic_read(&fcport->state),
(unsigned long long)wwn_to_u64(fcport->node_name),
(unsigned long long)wwn_to_u64(fcport->port_name));
ql_log(ql_log_info, vha, 0x208c,
"TGT-ID Announce(%s): Discovered tgt "
"id 0x%x wwnn = %llx "
"wwpn = %llx.\n", __func__, new_fcport->tgt_id,
(unsigned long long)
wwn_to_u64(new_fcport->node_name),
(unsigned long long)
wwn_to_u64(new_fcport->port_name));
if (atomic_read(&fcport->state) != FCS_ONLINE) {
fcport->old_tgt_id = fcport->tgt_id;
fcport->tgt_id = new_fcport->tgt_id;
ql_log(ql_log_info, vha, 0x208d,
"TGT-ID: New fcport Added: %p\n", fcport);
qla2x00_update_fcport(vha, fcport);
} else {
ql_log(ql_log_info, vha, 0x208e,
" Existing TGT-ID %x did not get "
" offline event from firmware.\n",
fcport->old_tgt_id);
qla2x00_mark_device_lost(vha, fcport, 0, 0);
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
kfree(new_fcport);
return rval;
}
break;
}
if (found)
continue;
/* If device was not in our fcports list, then add it. */
list_add_tail(&new_fcport->list, new_fcports);
/* Allocate a new replacement fcport. */
new_fcport = qla2x00_alloc_fcport(vha, GFP_KERNEL);
if (new_fcport == NULL)
return QLA_MEMORY_ALLOC_FAILED;
}
kfree(new_fcport);
return rval;
}
/*
* qlafx00_configure_all_targets
* Setup target devices with node ID's.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* 0 = success.
* BIT_0 = error
*/
static int
qlafx00_configure_all_targets(scsi_qla_host_t *vha)
{
int rval;
fc_port_t *fcport, *rmptemp;
LIST_HEAD(new_fcports);
rval = qlafx00_fx_disc(vha, &vha->hw->mr.fcport,
FXDISC_GET_TGT_NODE_LIST);
if (rval != QLA_SUCCESS) {
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
return rval;
}
rval = qlafx00_find_all_targets(vha, &new_fcports);
if (rval != QLA_SUCCESS) {
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
return rval;
}
/*
* Delete all previous devices marked lost.
*/
list_for_each_entry(fcport, &vha->vp_fcports, list) {
if (test_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags))
break;
if (atomic_read(&fcport->state) == FCS_DEVICE_LOST) {
if (fcport->port_type != FCT_INITIATOR)
qla2x00_mark_device_lost(vha, fcport, 0, 0);
}
}
/*
* Add the new devices to our devices list.
*/
list_for_each_entry_safe(fcport, rmptemp, &new_fcports, list) {
if (test_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags))
break;
qla2x00_update_fcport(vha, fcport);
list_move_tail(&fcport->list, &vha->vp_fcports);
ql_log(ql_log_info, vha, 0x208f,
"Attach new target id 0x%x wwnn = %llx "
"wwpn = %llx.\n",
fcport->tgt_id,
(unsigned long long)wwn_to_u64(fcport->node_name),
(unsigned long long)wwn_to_u64(fcport->port_name));
}
/* Free all new device structures not processed. */
list_for_each_entry_safe(fcport, rmptemp, &new_fcports, list) {
list_del(&fcport->list);
kfree(fcport);
}
return rval;
}
/*
* qlafx00_configure_devices
* Updates Fibre Channel Device Database with what is actually on loop.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* 0 = success.
* 1 = error.
* 2 = database was full and device was not configured.
*/
int
qlafx00_configure_devices(scsi_qla_host_t *vha)
{
int rval;
unsigned long flags, save_flags;
rval = QLA_SUCCESS;
save_flags = flags = vha->dpc_flags;
ql_dbg(ql_dbg_disc, vha, 0x2090,
"Configure devices -- dpc flags =0x%lx\n", flags);
rval = qlafx00_configure_all_targets(vha);
if (rval == QLA_SUCCESS) {
if (test_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags)) {
rval = QLA_FUNCTION_FAILED;
} else {
atomic_set(&vha->loop_state, LOOP_READY);
ql_log(ql_log_info, vha, 0x2091,
"Device Ready\n");
}
}
if (rval) {
ql_dbg(ql_dbg_disc, vha, 0x2092,
"%s *** FAILED ***.\n", __func__);
} else {
ql_dbg(ql_dbg_disc, vha, 0x2093,
"%s: exiting normally.\n", __func__);
}
return rval;
}
static void
qlafx00_abort_isp_cleanup(scsi_qla_host_t *vha, bool critemp)
{
struct qla_hw_data *ha = vha->hw;
fc_port_t *fcport;
vha->flags.online = 0;
ha->mr.fw_hbt_en = 0;
if (!critemp) {
ha->flags.chip_reset_done = 0;
clear_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
vha->qla_stats.total_isp_aborts++;
ql_log(ql_log_info, vha, 0x013f,
"Performing ISP error recovery - ha = %p.\n", ha);
ha->isp_ops->reset_chip(vha);
}
if (atomic_read(&vha->loop_state) != LOOP_DOWN) {
atomic_set(&vha->loop_state, LOOP_DOWN);
atomic_set(&vha->loop_down_timer,
QLAFX00_LOOP_DOWN_TIME);
} else {
if (!atomic_read(&vha->loop_down_timer))
atomic_set(&vha->loop_down_timer,
QLAFX00_LOOP_DOWN_TIME);
}
/* Clear all async request states across all VPs. */
list_for_each_entry(fcport, &vha->vp_fcports, list) {
fcport->flags = 0;
if (atomic_read(&fcport->state) == FCS_ONLINE)
qla2x00_set_fcport_state(fcport, FCS_DEVICE_LOST);
}
if (!ha->flags.eeh_busy) {
if (critemp) {
qla2x00_abort_all_cmds(vha, DID_NO_CONNECT << 16);
} else {
/* Requeue all commands in outstanding command list. */
qla2x00_abort_all_cmds(vha, DID_RESET << 16);
}
}
qla2x00_free_irqs(vha);
if (critemp)
set_bit(FX00_CRITEMP_RECOVERY, &vha->dpc_flags);
else
set_bit(FX00_RESET_RECOVERY, &vha->dpc_flags);
/* Clear the Interrupts */
QLAFX00_CLR_INTR_REG(ha, QLAFX00_HST_INT_STS_BITS);
ql_log(ql_log_info, vha, 0x0140,
"%s Done done - ha=%p.\n", __func__, ha);
}
/**
* qlafx00_init_response_q_entries() - Initializes response queue entries.
* @ha: HA context
*
* Beginning of request ring has initialization control block already built
* by nvram config routine.
*
* Returns 0 on success.
*/
void
qlafx00_init_response_q_entries(struct rsp_que *rsp)
{
uint16_t cnt;
response_t *pkt;
rsp->ring_ptr = rsp->ring;
rsp->ring_index = 0;
rsp->status_srb = NULL;
pkt = rsp->ring_ptr;
for (cnt = 0; cnt < rsp->length; cnt++) {
pkt->signature = RESPONSE_PROCESSED;
WRT_REG_DWORD((void __iomem *)&pkt->signature,
RESPONSE_PROCESSED);
pkt++;
}
}
int
qlafx00_rescan_isp(scsi_qla_host_t *vha)
{
uint32_t status = QLA_FUNCTION_FAILED;
struct qla_hw_data *ha = vha->hw;
struct device_reg_fx00 __iomem *reg = &ha->iobase->ispfx00;
uint32_t aenmbx7;
qla2x00_request_irqs(ha, ha->rsp_q_map[0]);
aenmbx7 = RD_REG_DWORD(&reg->aenmailbox7);
ha->mbx_intr_code = MSW(aenmbx7);
ha->rqstq_intr_code = LSW(aenmbx7);
ha->req_que_off = RD_REG_DWORD(&reg->aenmailbox1);
ha->rsp_que_off = RD_REG_DWORD(&reg->aenmailbox3);
ha->req_que_len = RD_REG_DWORD(&reg->aenmailbox5);
ha->rsp_que_len = RD_REG_DWORD(&reg->aenmailbox6);
ql_dbg(ql_dbg_disc, vha, 0x2094,
"fw returned mbx_intr_code: 0x%x, rqstq_intr_code: 0x%x "
" Req que offset 0x%x Rsp que offset 0x%x\n",
ha->mbx_intr_code, ha->rqstq_intr_code,
ha->req_que_off, ha->rsp_que_len);
/* Clear the Interrupts */
QLAFX00_CLR_INTR_REG(ha, QLAFX00_HST_INT_STS_BITS);
status = qla2x00_init_rings(vha);
if (!status) {
vha->flags.online = 1;
/* if no cable then assume it's good */
if ((vha->device_flags & DFLG_NO_CABLE))
status = 0;
/* Register system information */
if (qlafx00_fx_disc(vha,
&vha->hw->mr.fcport, FXDISC_REG_HOST_INFO))
ql_dbg(ql_dbg_disc, vha, 0x2095,
"failed to register host info\n");
}
scsi_unblock_requests(vha->host);
return status;
}
void
qlafx00_timer_routine(scsi_qla_host_t *vha)
{
struct qla_hw_data *ha = vha->hw;
uint32_t fw_heart_beat;
uint32_t aenmbx0;
struct device_reg_fx00 __iomem *reg = &ha->iobase->ispfx00;
uint32_t tempc;
/* Check firmware health */
if (ha->mr.fw_hbt_cnt)
ha->mr.fw_hbt_cnt--;
else {
if ((!ha->flags.mr_reset_hdlr_active) &&
(!test_bit(UNLOADING, &vha->dpc_flags)) &&
(!test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags)) &&
(ha->mr.fw_hbt_en)) {
fw_heart_beat = RD_REG_DWORD(&reg->fwheartbeat);
if (fw_heart_beat != ha->mr.old_fw_hbt_cnt) {
ha->mr.old_fw_hbt_cnt = fw_heart_beat;
ha->mr.fw_hbt_miss_cnt = 0;
} else {
ha->mr.fw_hbt_miss_cnt++;
if (ha->mr.fw_hbt_miss_cnt ==
QLAFX00_HEARTBEAT_MISS_CNT) {
set_bit(ISP_ABORT_NEEDED,
&vha->dpc_flags);
qla2xxx_wake_dpc(vha);
ha->mr.fw_hbt_miss_cnt = 0;
}
}
}
ha->mr.fw_hbt_cnt = QLAFX00_HEARTBEAT_INTERVAL;
}
if (test_bit(FX00_RESET_RECOVERY, &vha->dpc_flags)) {
/* Reset recovery to be performed in timer routine */
aenmbx0 = RD_REG_DWORD(&reg->aenmailbox0);
if (ha->mr.fw_reset_timer_exp) {
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
ha->mr.fw_reset_timer_exp = 0;
} else if (aenmbx0 == MBA_FW_RESTART_CMPLT) {
/* Wake up DPC to rescan the targets */
set_bit(FX00_TARGET_SCAN, &vha->dpc_flags);
clear_bit(FX00_RESET_RECOVERY, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
ha->mr.fw_reset_timer_tick = QLAFX00_RESET_INTERVAL;
} else if ((aenmbx0 == MBA_FW_STARTING) &&
(!ha->mr.fw_hbt_en)) {
ha->mr.fw_hbt_en = 1;
} else if (!ha->mr.fw_reset_timer_tick) {
if (aenmbx0 == ha->mr.old_aenmbx0_state)
ha->mr.fw_reset_timer_exp = 1;
ha->mr.fw_reset_timer_tick = QLAFX00_RESET_INTERVAL;
} else if (aenmbx0 == 0xFFFFFFFF) {
uint32_t data0, data1;
data0 = QLAFX00_RD_REG(ha,
QLAFX00_BAR1_BASE_ADDR_REG);
data1 = QLAFX00_RD_REG(ha,
QLAFX00_PEX0_WIN0_BASE_ADDR_REG);
data0 &= 0xffff0000;
data1 &= 0x0000ffff;
QLAFX00_WR_REG(ha,
QLAFX00_PEX0_WIN0_BASE_ADDR_REG,
(data0 | data1));
} else if ((aenmbx0 & 0xFF00) == MBA_FW_POLL_STATE) {
ha->mr.fw_reset_timer_tick =
QLAFX00_MAX_RESET_INTERVAL;
} else if (aenmbx0 == MBA_FW_RESET_FCT) {
ha->mr.fw_reset_timer_tick =
QLAFX00_MAX_RESET_INTERVAL;
}
if (ha->mr.old_aenmbx0_state != aenmbx0) {
ha->mr.old_aenmbx0_state = aenmbx0;
ha->mr.fw_reset_timer_tick = QLAFX00_RESET_INTERVAL;
}
ha->mr.fw_reset_timer_tick--;
}
if (test_bit(FX00_CRITEMP_RECOVERY, &vha->dpc_flags)) {
/*
* Critical temperature recovery to be
* performed in timer routine
*/
if (ha->mr.fw_critemp_timer_tick == 0) {
tempc = QLAFX00_GET_TEMPERATURE(ha);
ql_dbg(ql_dbg_timer, vha, 0x6012,
"ISPFx00(%s): Critical temp timer, "
"current SOC temperature: %d\n",
__func__, tempc);
if (tempc < ha->mr.critical_temperature) {
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
clear_bit(FX00_CRITEMP_RECOVERY,
&vha->dpc_flags);
qla2xxx_wake_dpc(vha);
}
ha->mr.fw_critemp_timer_tick =
QLAFX00_CRITEMP_INTERVAL;
} else {
ha->mr.fw_critemp_timer_tick--;
}
}
if (ha->mr.host_info_resend) {
/*
* Incomplete host info might be sent to firmware
* durinng system boot - info should be resend
*/
if (ha->mr.hinfo_resend_timer_tick == 0) {
ha->mr.host_info_resend = false;
set_bit(FX00_HOST_INFO_RESEND, &vha->dpc_flags);
ha->mr.hinfo_resend_timer_tick =
QLAFX00_HINFO_RESEND_INTERVAL;
qla2xxx_wake_dpc(vha);
} else {
ha->mr.hinfo_resend_timer_tick--;
}
}
}
/*
* qlfx00a_reset_initialize
* Re-initialize after a iSA device reset.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* 0 = success
*/
int
qlafx00_reset_initialize(scsi_qla_host_t *vha)
{
struct qla_hw_data *ha = vha->hw;
if (vha->device_flags & DFLG_DEV_FAILED) {
ql_dbg(ql_dbg_init, vha, 0x0142,
"Device in failed state\n");
return QLA_SUCCESS;
}
ha->flags.mr_reset_hdlr_active = 1;
if (vha->flags.online) {
scsi_block_requests(vha->host);
qlafx00_abort_isp_cleanup(vha, false);
}
ql_log(ql_log_info, vha, 0x0143,
"(%s): succeeded.\n", __func__);
ha->flags.mr_reset_hdlr_active = 0;
return QLA_SUCCESS;
}
/*
* qlafx00_abort_isp
* Resets ISP and aborts all outstanding commands.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* 0 = success
*/
int
qlafx00_abort_isp(scsi_qla_host_t *vha)
{
struct qla_hw_data *ha = vha->hw;
if (vha->flags.online) {
if (unlikely(pci_channel_offline(ha->pdev) &&
ha->flags.pci_channel_io_perm_failure)) {
clear_bit(ISP_ABORT_RETRY, &vha->dpc_flags);
return QLA_SUCCESS;
}
scsi_block_requests(vha->host);
qlafx00_abort_isp_cleanup(vha, false);
} else {
scsi_block_requests(vha->host);
clear_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
vha->qla_stats.total_isp_aborts++;
ha->isp_ops->reset_chip(vha);
set_bit(FX00_RESET_RECOVERY, &vha->dpc_flags);
/* Clear the Interrupts */
QLAFX00_CLR_INTR_REG(ha, QLAFX00_HST_INT_STS_BITS);
}
ql_log(ql_log_info, vha, 0x0145,
"(%s): succeeded.\n", __func__);
return QLA_SUCCESS;
}
static inline fc_port_t*
qlafx00_get_fcport(struct scsi_qla_host *vha, int tgt_id)
{
fc_port_t *fcport;
/* Check for matching device in remote port list. */
list_for_each_entry(fcport, &vha->vp_fcports, list) {
if (fcport->tgt_id == tgt_id) {
ql_dbg(ql_dbg_async, vha, 0x5072,
"Matching fcport(%p) found with TGT-ID: 0x%x "
"and Remote TGT_ID: 0x%x\n",
fcport, fcport->tgt_id, tgt_id);
return fcport;
}
}
return NULL;
}
static void
qlafx00_tgt_detach(struct scsi_qla_host *vha, int tgt_id)
{
fc_port_t *fcport;
ql_log(ql_log_info, vha, 0x5073,
"Detach TGT-ID: 0x%x\n", tgt_id);
fcport = qlafx00_get_fcport(vha, tgt_id);
if (!fcport)
return;
qla2x00_mark_device_lost(vha, fcport, 0, 0);
return;
}
int
qlafx00_process_aen(struct scsi_qla_host *vha, struct qla_work_evt *evt)
{
int rval = 0;
uint32_t aen_code, aen_data;
aen_code = FCH_EVT_VENDOR_UNIQUE;
aen_data = evt->u.aenfx.evtcode;
switch (evt->u.aenfx.evtcode) {
case QLAFX00_MBA_PORT_UPDATE: /* Port database update */
if (evt->u.aenfx.mbx[1] == 0) {
if (evt->u.aenfx.mbx[2] == 1) {
if (!vha->flags.fw_tgt_reported)
vha->flags.fw_tgt_reported = 1;
atomic_set(&vha->loop_down_timer, 0);
atomic_set(&vha->loop_state, LOOP_UP);
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
} else if (evt->u.aenfx.mbx[2] == 2) {
qlafx00_tgt_detach(vha, evt->u.aenfx.mbx[3]);
}
} else if (evt->u.aenfx.mbx[1] == 0xffff) {
if (evt->u.aenfx.mbx[2] == 1) {
if (!vha->flags.fw_tgt_reported)
vha->flags.fw_tgt_reported = 1;
set_bit(LOOP_RESYNC_NEEDED, &vha->dpc_flags);
} else if (evt->u.aenfx.mbx[2] == 2) {
vha->device_flags |= DFLG_NO_CABLE;
qla2x00_mark_all_devices_lost(vha, 1);
}
}
break;
case QLAFX00_MBA_LINK_UP:
aen_code = FCH_EVT_LINKUP;
aen_data = 0;
break;
case QLAFX00_MBA_LINK_DOWN:
aen_code = FCH_EVT_LINKDOWN;
aen_data = 0;
break;
case QLAFX00_MBA_TEMP_CRIT: /* Critical temperature event */
ql_log(ql_log_info, vha, 0x5082,
"Process critical temperature event "
"aenmb[0]: %x\n",
evt->u.aenfx.evtcode);
scsi_block_requests(vha->host);
qlafx00_abort_isp_cleanup(vha, true);
scsi_unblock_requests(vha->host);
break;
}
fc_host_post_event(vha->host, fc_get_event_number(),
aen_code, aen_data);
return rval;
}
static void
qlafx00_update_host_attr(scsi_qla_host_t *vha, struct port_info_data *pinfo)
{
u64 port_name = 0, node_name = 0;
port_name = (unsigned long long)wwn_to_u64(pinfo->port_name);
node_name = (unsigned long long)wwn_to_u64(pinfo->node_name);
fc_host_node_name(vha->host) = node_name;
fc_host_port_name(vha->host) = port_name;
if (!pinfo->port_type)
vha->hw->current_topology = ISP_CFG_F;
if (pinfo->link_status == QLAFX00_LINK_STATUS_UP)
atomic_set(&vha->loop_state, LOOP_READY);
else if (pinfo->link_status == QLAFX00_LINK_STATUS_DOWN)
atomic_set(&vha->loop_state, LOOP_DOWN);
vha->hw->link_data_rate = (uint16_t)pinfo->link_config;
}
static void
qla2x00_fxdisc_iocb_timeout(void *data)
{
srb_t *sp = (srb_t *)data;
struct srb_iocb *lio = &sp->u.iocb_cmd;
complete(&lio->u.fxiocb.fxiocb_comp);
}
static void
qla2x00_fxdisc_sp_done(void *data, void *ptr, int res)
{
srb_t *sp = (srb_t *)ptr;
struct srb_iocb *lio = &sp->u.iocb_cmd;
complete(&lio->u.fxiocb.fxiocb_comp);
}
int
qlafx00_fx_disc(scsi_qla_host_t *vha, fc_port_t *fcport, uint16_t fx_type)
{
srb_t *sp;
struct srb_iocb *fdisc;
int rval = QLA_FUNCTION_FAILED;
struct qla_hw_data *ha = vha->hw;
struct host_system_info *phost_info;
struct register_host_info *preg_hsi;
struct new_utsname *p_sysid = NULL;
struct timeval tv;
sp = qla2x00_get_sp(vha, fcport, GFP_KERNEL);
if (!sp)
goto done;
fdisc = &sp->u.iocb_cmd;
switch (fx_type) {
case FXDISC_GET_CONFIG_INFO:
fdisc->u.fxiocb.flags =
SRB_FXDISC_RESP_DMA_VALID;
fdisc->u.fxiocb.rsp_len = sizeof(struct config_info_data);
break;
case FXDISC_GET_PORT_INFO:
fdisc->u.fxiocb.flags =
SRB_FXDISC_RESP_DMA_VALID | SRB_FXDISC_REQ_DWRD_VALID;
fdisc->u.fxiocb.rsp_len = QLAFX00_PORT_DATA_INFO;
fdisc->u.fxiocb.req_data = cpu_to_le32(fcport->port_id);
break;
case FXDISC_GET_TGT_NODE_INFO:
fdisc->u.fxiocb.flags =
SRB_FXDISC_RESP_DMA_VALID | SRB_FXDISC_REQ_DWRD_VALID;
fdisc->u.fxiocb.rsp_len = QLAFX00_TGT_NODE_INFO;
fdisc->u.fxiocb.req_data = cpu_to_le32(fcport->tgt_id);
break;
case FXDISC_GET_TGT_NODE_LIST:
fdisc->u.fxiocb.flags =
SRB_FXDISC_RESP_DMA_VALID | SRB_FXDISC_REQ_DWRD_VALID;
fdisc->u.fxiocb.rsp_len = QLAFX00_TGT_NODE_LIST_SIZE;
break;
case FXDISC_REG_HOST_INFO:
fdisc->u.fxiocb.flags = SRB_FXDISC_REQ_DMA_VALID;
fdisc->u.fxiocb.req_len = sizeof(struct register_host_info);
p_sysid = utsname();
if (!p_sysid) {
ql_log(ql_log_warn, vha, 0x303c,
"Not able to get the system information\n");
goto done_free_sp;
}
break;
case FXDISC_ABORT_IOCTL:
default:
break;
}
if (fdisc->u.fxiocb.flags & SRB_FXDISC_REQ_DMA_VALID) {
fdisc->u.fxiocb.req_addr = dma_alloc_coherent(&ha->pdev->dev,
fdisc->u.fxiocb.req_len,
&fdisc->u.fxiocb.req_dma_handle, GFP_KERNEL);
if (!fdisc->u.fxiocb.req_addr)
goto done_free_sp;
if (fx_type == FXDISC_REG_HOST_INFO) {
preg_hsi = (struct register_host_info *)
fdisc->u.fxiocb.req_addr;
phost_info = &preg_hsi->hsi;
memset(preg_hsi, 0, sizeof(struct register_host_info));
phost_info->os_type = OS_TYPE_LINUX;
strncpy(phost_info->sysname,
p_sysid->sysname, SYSNAME_LENGTH);
strncpy(phost_info->nodename,
p_sysid->nodename, NODENAME_LENGTH);
if (!strcmp(phost_info->nodename, "(none)"))
ha->mr.host_info_resend = true;
strncpy(phost_info->release,
p_sysid->release, RELEASE_LENGTH);
strncpy(phost_info->version,
p_sysid->version, VERSION_LENGTH);
strncpy(phost_info->machine,
p_sysid->machine, MACHINE_LENGTH);
strncpy(phost_info->domainname,
p_sysid->domainname, DOMNAME_LENGTH);
strncpy(phost_info->hostdriver,
QLA2XXX_VERSION, VERSION_LENGTH);
do_gettimeofday(&tv);
preg_hsi->utc = (uint64_t)tv.tv_sec;
ql_dbg(ql_dbg_init, vha, 0x0149,
"ISP%04X: Host registration with firmware\n",
ha->pdev->device);
ql_dbg(ql_dbg_init, vha, 0x014a,
"os_type = '%d', sysname = '%s', nodname = '%s'\n",
phost_info->os_type,
phost_info->sysname,
phost_info->nodename);
ql_dbg(ql_dbg_init, vha, 0x014b,
"release = '%s', version = '%s'\n",
phost_info->release,
phost_info->version);
ql_dbg(ql_dbg_init, vha, 0x014c,
"machine = '%s' "
"domainname = '%s', hostdriver = '%s'\n",
phost_info->machine,
phost_info->domainname,
phost_info->hostdriver);
ql_dump_buffer(ql_dbg_init + ql_dbg_disc, vha, 0x014d,
(uint8_t *)phost_info,
sizeof(struct host_system_info));
}
}
if (fdisc->u.fxiocb.flags & SRB_FXDISC_RESP_DMA_VALID) {
fdisc->u.fxiocb.rsp_addr = dma_alloc_coherent(&ha->pdev->dev,
fdisc->u.fxiocb.rsp_len,
&fdisc->u.fxiocb.rsp_dma_handle, GFP_KERNEL);
if (!fdisc->u.fxiocb.rsp_addr)
goto done_unmap_req;
}
sp->type = SRB_FXIOCB_DCMD;
sp->name = "fxdisc";
qla2x00_init_timer(sp, FXDISC_TIMEOUT);
fdisc->timeout = qla2x00_fxdisc_iocb_timeout;
fdisc->u.fxiocb.req_func_type = cpu_to_le16(fx_type);
sp->done = qla2x00_fxdisc_sp_done;
rval = qla2x00_start_sp(sp);
if (rval != QLA_SUCCESS)
goto done_unmap_dma;
wait_for_completion(&fdisc->u.fxiocb.fxiocb_comp);
if (fx_type == FXDISC_GET_CONFIG_INFO) {
struct config_info_data *pinfo =
(struct config_info_data *) fdisc->u.fxiocb.rsp_addr;
strcpy(vha->hw->model_number, pinfo->model_num);
strcpy(vha->hw->model_desc, pinfo->model_description);
memcpy(&vha->hw->mr.symbolic_name, pinfo->symbolic_name,
sizeof(vha->hw->mr.symbolic_name));
memcpy(&vha->hw->mr.serial_num, pinfo->serial_num,
sizeof(vha->hw->mr.serial_num));
memcpy(&vha->hw->mr.hw_version, pinfo->hw_version,
sizeof(vha->hw->mr.hw_version));
memcpy(&vha->hw->mr.fw_version, pinfo->fw_version,
sizeof(vha->hw->mr.fw_version));
strim(vha->hw->mr.fw_version);
memcpy(&vha->hw->mr.uboot_version, pinfo->uboot_version,
sizeof(vha->hw->mr.uboot_version));
memcpy(&vha->hw->mr.fru_serial_num, pinfo->fru_serial_num,
sizeof(vha->hw->mr.fru_serial_num));
vha->hw->mr.critical_temperature =
(pinfo->nominal_temp_value) ?
pinfo->nominal_temp_value : QLAFX00_CRITEMP_THRSHLD;
ha->mr.extended_io_enabled = (pinfo->enabled_capabilities &
QLAFX00_EXTENDED_IO_EN_MASK) != 0;
} else if (fx_type == FXDISC_GET_PORT_INFO) {
struct port_info_data *pinfo =
(struct port_info_data *) fdisc->u.fxiocb.rsp_addr;
memcpy(vha->node_name, pinfo->node_name, WWN_SIZE);
memcpy(vha->port_name, pinfo->port_name, WWN_SIZE);
vha->d_id.b.domain = pinfo->port_id[0];
vha->d_id.b.area = pinfo->port_id[1];
vha->d_id.b.al_pa = pinfo->port_id[2];
qlafx00_update_host_attr(vha, pinfo);
ql_dump_buffer(ql_dbg_init + ql_dbg_buffer, vha, 0x0141,
(uint8_t *)pinfo, 16);
} else if (fx_type == FXDISC_GET_TGT_NODE_INFO) {
struct qlafx00_tgt_node_info *pinfo =
(struct qlafx00_tgt_node_info *) fdisc->u.fxiocb.rsp_addr;
memcpy(fcport->node_name, pinfo->tgt_node_wwnn, WWN_SIZE);
memcpy(fcport->port_name, pinfo->tgt_node_wwpn, WWN_SIZE);
fcport->port_type = FCT_TARGET;
ql_dump_buffer(ql_dbg_init + ql_dbg_buffer, vha, 0x0144,
(uint8_t *)pinfo, 16);
} else if (fx_type == FXDISC_GET_TGT_NODE_LIST) {
struct qlafx00_tgt_node_info *pinfo =
(struct qlafx00_tgt_node_info *) fdisc->u.fxiocb.rsp_addr;
ql_dump_buffer(ql_dbg_init + ql_dbg_buffer, vha, 0x0146,
(uint8_t *)pinfo, 16);
memcpy(vha->hw->gid_list, pinfo, QLAFX00_TGT_NODE_LIST_SIZE);
} else if (fx_type == FXDISC_ABORT_IOCTL)
fdisc->u.fxiocb.result =
(fdisc->u.fxiocb.result ==
cpu_to_le32(QLAFX00_IOCTL_ICOB_ABORT_SUCCESS)) ?
cpu_to_le32(QLA_SUCCESS) : cpu_to_le32(QLA_FUNCTION_FAILED);
rval = le32_to_cpu(fdisc->u.fxiocb.result);
done_unmap_dma:
if (fdisc->u.fxiocb.rsp_addr)
dma_free_coherent(&ha->pdev->dev, fdisc->u.fxiocb.rsp_len,
fdisc->u.fxiocb.rsp_addr, fdisc->u.fxiocb.rsp_dma_handle);
done_unmap_req:
if (fdisc->u.fxiocb.req_addr)
dma_free_coherent(&ha->pdev->dev, fdisc->u.fxiocb.req_len,
fdisc->u.fxiocb.req_addr, fdisc->u.fxiocb.req_dma_handle);
done_free_sp:
sp->free(vha, sp);
done:
return rval;
}
/*
* qlafx00_initialize_adapter
* Initialize board.
*
* Input:
* ha = adapter block pointer.
*
* Returns:
* 0 = success
*/
int
qlafx00_initialize_adapter(scsi_qla_host_t *vha)
{
int rval;
struct qla_hw_data *ha = vha->hw;
uint32_t tempc;
/* Clear adapter flags. */
vha->flags.online = 0;
ha->flags.chip_reset_done = 0;
vha->flags.reset_active = 0;
ha->flags.pci_channel_io_perm_failure = 0;
ha->flags.eeh_busy = 0;
atomic_set(&vha->loop_down_timer, LOOP_DOWN_TIME);
atomic_set(&vha->loop_state, LOOP_DOWN);
vha->device_flags = DFLG_NO_CABLE;
vha->dpc_flags = 0;
vha->flags.management_server_logged_in = 0;
ha->isp_abort_cnt = 0;
ha->beacon_blink_led = 0;
set_bit(0, ha->req_qid_map);
set_bit(0, ha->rsp_qid_map);
ql_dbg(ql_dbg_init, vha, 0x0147,
"Configuring PCI space...\n");
rval = ha->isp_ops->pci_config(vha);
if (rval) {
ql_log(ql_log_warn, vha, 0x0148,
"Unable to configure PCI space.\n");
return rval;
}
rval = qlafx00_init_fw_ready(vha);
if (rval != QLA_SUCCESS)
return rval;
qlafx00_save_queue_ptrs(vha);
rval = qlafx00_config_queues(vha);
if (rval != QLA_SUCCESS)
return rval;
/*
* Allocate the array of outstanding commands
* now that we know the firmware resources.
*/
rval = qla2x00_alloc_outstanding_cmds(ha, vha->req);
if (rval != QLA_SUCCESS)
return rval;
rval = qla2x00_init_rings(vha);
ha->flags.chip_reset_done = 1;
tempc = QLAFX00_GET_TEMPERATURE(ha);
ql_dbg(ql_dbg_init, vha, 0x0152,
"ISPFx00(%s): Critical temp timer, current SOC temperature: 0x%x\n",
__func__, tempc);
return rval;
}
uint32_t
qlafx00_fw_state_show(struct device *dev, struct device_attribute *attr,
char *buf)
{
scsi_qla_host_t *vha = shost_priv(class_to_shost(dev));
int rval = QLA_FUNCTION_FAILED;
uint32_t state[1];
if (qla2x00_reset_active(vha))
ql_log(ql_log_warn, vha, 0x70ce,
"ISP reset active.\n");
else if (!vha->hw->flags.eeh_busy) {
rval = qlafx00_get_firmware_state(vha, state);
}
if (rval != QLA_SUCCESS)
memset(state, -1, sizeof(state));
return state[0];
}
void
qlafx00_get_host_speed(struct Scsi_Host *shost)
{
struct qla_hw_data *ha = ((struct scsi_qla_host *)
(shost_priv(shost)))->hw;
u32 speed = FC_PORTSPEED_UNKNOWN;
switch (ha->link_data_rate) {
case QLAFX00_PORT_SPEED_2G:
speed = FC_PORTSPEED_2GBIT;
break;
case QLAFX00_PORT_SPEED_4G:
speed = FC_PORTSPEED_4GBIT;
break;
case QLAFX00_PORT_SPEED_8G:
speed = FC_PORTSPEED_8GBIT;
break;
case QLAFX00_PORT_SPEED_10G:
speed = FC_PORTSPEED_10GBIT;
break;
}
fc_host_speed(shost) = speed;
}
/** QLAFX00 specific ISR implementation functions */
static inline void
qlafx00_handle_sense(srb_t *sp, uint8_t *sense_data, uint32_t par_sense_len,
uint32_t sense_len, struct rsp_que *rsp, int res)
{
struct scsi_qla_host *vha = sp->fcport->vha;
struct scsi_cmnd *cp = GET_CMD_SP(sp);
uint32_t track_sense_len;
SET_FW_SENSE_LEN(sp, sense_len);
if (sense_len >= SCSI_SENSE_BUFFERSIZE)
sense_len = SCSI_SENSE_BUFFERSIZE;
SET_CMD_SENSE_LEN(sp, sense_len);
SET_CMD_SENSE_PTR(sp, cp->sense_buffer);
track_sense_len = sense_len;
if (sense_len > par_sense_len)
sense_len = par_sense_len;
memcpy(cp->sense_buffer, sense_data, sense_len);
SET_FW_SENSE_LEN(sp, GET_FW_SENSE_LEN(sp) - sense_len);
SET_CMD_SENSE_PTR(sp, cp->sense_buffer + sense_len);
track_sense_len -= sense_len;
SET_CMD_SENSE_LEN(sp, track_sense_len);
ql_dbg(ql_dbg_io, vha, 0x304d,
"sense_len=0x%x par_sense_len=0x%x track_sense_len=0x%x.\n",
sense_len, par_sense_len, track_sense_len);
if (GET_FW_SENSE_LEN(sp) > 0) {
rsp->status_srb = sp;
cp->result = res;
}
if (sense_len) {
ql_dbg(ql_dbg_io + ql_dbg_buffer, vha, 0x3039,
"Check condition Sense data, nexus%ld:%d:%llu cmd=%p.\n",
sp->fcport->vha->host_no, cp->device->id, cp->device->lun,
cp);
ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x3049,
cp->sense_buffer, sense_len);
}
}
static void
qlafx00_tm_iocb_entry(scsi_qla_host_t *vha, struct req_que *req,
struct tsk_mgmt_entry_fx00 *pkt, srb_t *sp,
__le16 sstatus, __le16 cpstatus)
{
struct srb_iocb *tmf;
tmf = &sp->u.iocb_cmd;
if (cpstatus != cpu_to_le16((uint16_t)CS_COMPLETE) ||
(sstatus & cpu_to_le16((uint16_t)SS_RESPONSE_INFO_LEN_VALID)))
cpstatus = cpu_to_le16((uint16_t)CS_INCOMPLETE);
tmf->u.tmf.comp_status = cpstatus;
sp->done(vha, sp, 0);
}
static void
qlafx00_abort_iocb_entry(scsi_qla_host_t *vha, struct req_que *req,
struct abort_iocb_entry_fx00 *pkt)
{
const char func[] = "ABT_IOCB";
srb_t *sp;
struct srb_iocb *abt;
sp = qla2x00_get_sp_from_handle(vha, func, req, pkt);
if (!sp)
return;
abt = &sp->u.iocb_cmd;
abt->u.abt.comp_status = pkt->tgt_id_sts;
sp->done(vha, sp, 0);
}
static void
qlafx00_ioctl_iosb_entry(scsi_qla_host_t *vha, struct req_que *req,
struct ioctl_iocb_entry_fx00 *pkt)
{
const char func[] = "IOSB_IOCB";
srb_t *sp;
struct fc_bsg_job *bsg_job;
struct srb_iocb *iocb_job;
int res;
struct qla_mt_iocb_rsp_fx00 fstatus;
uint8_t *fw_sts_ptr;
sp = qla2x00_get_sp_from_handle(vha, func, req, pkt);
if (!sp)
return;
if (sp->type == SRB_FXIOCB_DCMD) {
iocb_job = &sp->u.iocb_cmd;
iocb_job->u.fxiocb.seq_number = pkt->seq_no;
iocb_job->u.fxiocb.fw_flags = pkt->fw_iotcl_flags;
iocb_job->u.fxiocb.result = pkt->status;
if (iocb_job->u.fxiocb.flags & SRB_FXDISC_RSP_DWRD_VALID)
iocb_job->u.fxiocb.req_data =
pkt->dataword_r;
} else {
bsg_job = sp->u.bsg_job;
memset(&fstatus, 0, sizeof(struct qla_mt_iocb_rsp_fx00));
fstatus.reserved_1 = pkt->reserved_0;
fstatus.func_type = pkt->comp_func_num;
fstatus.ioctl_flags = pkt->fw_iotcl_flags;
fstatus.ioctl_data = pkt->dataword_r;
fstatus.adapid = pkt->adapid;
fstatus.reserved_2 = pkt->dataword_r_extra;
fstatus.res_count = pkt->residuallen;
fstatus.status = pkt->status;
fstatus.seq_number = pkt->seq_no;
memcpy(fstatus.reserved_3,
pkt->reserved_2, 20 * sizeof(uint8_t));
fw_sts_ptr = ((uint8_t *)bsg_job->req->sense) +
sizeof(struct fc_bsg_reply);
memcpy(fw_sts_ptr, (uint8_t *)&fstatus,
sizeof(struct qla_mt_iocb_rsp_fx00));
bsg_job->reply_len = sizeof(struct fc_bsg_reply) +
sizeof(struct qla_mt_iocb_rsp_fx00) + sizeof(uint8_t);
ql_dump_buffer(ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x5080,
(uint8_t *)pkt, sizeof(struct ioctl_iocb_entry_fx00));
ql_dump_buffer(ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x5074,
(uint8_t *)fw_sts_ptr, sizeof(struct qla_mt_iocb_rsp_fx00));
res = bsg_job->reply->result = DID_OK << 16;
bsg_job->reply->reply_payload_rcv_len =
bsg_job->reply_payload.payload_len;
}
sp->done(vha, sp, res);
}
/**
* qlafx00_status_entry() - Process a Status IOCB entry.
* @ha: SCSI driver HA context
* @pkt: Entry pointer
*/
static void
qlafx00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
{
srb_t *sp;
fc_port_t *fcport;
struct scsi_cmnd *cp;
struct sts_entry_fx00 *sts;
__le16 comp_status;
__le16 scsi_status;
uint16_t ox_id;
__le16 lscsi_status;
int32_t resid;
uint32_t sense_len, par_sense_len, rsp_info_len, resid_len,
fw_resid_len;
uint8_t *rsp_info = NULL, *sense_data = NULL;
struct qla_hw_data *ha = vha->hw;
uint32_t hindex, handle;
uint16_t que;
struct req_que *req;
int logit = 1;
int res = 0;
sts = (struct sts_entry_fx00 *) pkt;
comp_status = sts->comp_status;
scsi_status = sts->scsi_status & cpu_to_le16((uint16_t)SS_MASK);
hindex = sts->handle;
handle = LSW(hindex);
que = MSW(hindex);
req = ha->req_q_map[que];
/* Validate handle. */
if (handle < req->num_outstanding_cmds)
sp = req->outstanding_cmds[handle];
else
sp = NULL;
if (sp == NULL) {
ql_dbg(ql_dbg_io, vha, 0x3034,
"Invalid status handle (0x%x).\n", handle);
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
return;
}
if (sp->type == SRB_TM_CMD) {
req->outstanding_cmds[handle] = NULL;
qlafx00_tm_iocb_entry(vha, req, pkt, sp,
scsi_status, comp_status);
return;
}
/* Fast path completion. */
if (comp_status == CS_COMPLETE && scsi_status == 0) {
qla2x00_process_completed_request(vha, req, handle);
return;
}
req->outstanding_cmds[handle] = NULL;
cp = GET_CMD_SP(sp);
if (cp == NULL) {
ql_dbg(ql_dbg_io, vha, 0x3048,
"Command already returned (0x%x/%p).\n",
handle, sp);
return;
}
lscsi_status = scsi_status & cpu_to_le16((uint16_t)STATUS_MASK);
fcport = sp->fcport;
ox_id = 0;
sense_len = par_sense_len = rsp_info_len = resid_len =
fw_resid_len = 0;
if (scsi_status & cpu_to_le16((uint16_t)SS_SENSE_LEN_VALID))
sense_len = sts->sense_len;
if (scsi_status & cpu_to_le16(((uint16_t)SS_RESIDUAL_UNDER
| (uint16_t)SS_RESIDUAL_OVER)))
resid_len = le32_to_cpu(sts->residual_len);
if (comp_status == cpu_to_le16((uint16_t)CS_DATA_UNDERRUN))
fw_resid_len = le32_to_cpu(sts->residual_len);
rsp_info = sense_data = sts->data;
par_sense_len = sizeof(sts->data);
/* Check for overrun. */
if (comp_status == CS_COMPLETE &&
scsi_status & cpu_to_le16((uint16_t)SS_RESIDUAL_OVER))
comp_status = cpu_to_le16((uint16_t)CS_DATA_OVERRUN);
/*
* Based on Host and scsi status generate status code for Linux
*/
switch (le16_to_cpu(comp_status)) {
case CS_COMPLETE:
case CS_QUEUE_FULL:
if (scsi_status == 0) {
res = DID_OK << 16;
break;
}
if (scsi_status & cpu_to_le16(((uint16_t)SS_RESIDUAL_UNDER
| (uint16_t)SS_RESIDUAL_OVER))) {
resid = resid_len;
scsi_set_resid(cp, resid);
if (!lscsi_status &&
((unsigned)(scsi_bufflen(cp) - resid) <
cp->underflow)) {
ql_dbg(ql_dbg_io, fcport->vha, 0x3050,
"Mid-layer underflow "
"detected (0x%x of 0x%x bytes).\n",
resid, scsi_bufflen(cp));
res = DID_ERROR << 16;
break;
}
}
res = DID_OK << 16 | le16_to_cpu(lscsi_status);
if (lscsi_status ==
cpu_to_le16((uint16_t)SAM_STAT_TASK_SET_FULL)) {
ql_dbg(ql_dbg_io, fcport->vha, 0x3051,
"QUEUE FULL detected.\n");
break;
}
logit = 0;
if (lscsi_status != cpu_to_le16((uint16_t)SS_CHECK_CONDITION))
break;
memset(cp->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
if (!(scsi_status & cpu_to_le16((uint16_t)SS_SENSE_LEN_VALID)))
break;
qlafx00_handle_sense(sp, sense_data, par_sense_len, sense_len,
rsp, res);
break;
case CS_DATA_UNDERRUN:
/* Use F/W calculated residual length. */
if (IS_FWI2_CAPABLE(ha) || IS_QLAFX00(ha))
resid = fw_resid_len;
else
resid = resid_len;
scsi_set_resid(cp, resid);
if (scsi_status & cpu_to_le16((uint16_t)SS_RESIDUAL_UNDER)) {
if ((IS_FWI2_CAPABLE(ha) || IS_QLAFX00(ha))
&& fw_resid_len != resid_len) {
ql_dbg(ql_dbg_io, fcport->vha, 0x3052,
"Dropped frame(s) detected "
"(0x%x of 0x%x bytes).\n",
resid, scsi_bufflen(cp));
res = DID_ERROR << 16 |
le16_to_cpu(lscsi_status);
goto check_scsi_status;
}
if (!lscsi_status &&
((unsigned)(scsi_bufflen(cp) - resid) <
cp->underflow)) {
ql_dbg(ql_dbg_io, fcport->vha, 0x3053,
"Mid-layer underflow "
"detected (0x%x of 0x%x bytes, "
"cp->underflow: 0x%x).\n",
resid, scsi_bufflen(cp), cp->underflow);
res = DID_ERROR << 16;
break;
}
} else if (lscsi_status !=
cpu_to_le16((uint16_t)SAM_STAT_TASK_SET_FULL) &&
lscsi_status != cpu_to_le16((uint16_t)SAM_STAT_BUSY)) {
/*
* scsi status of task set and busy are considered
* to be task not completed.
*/
ql_dbg(ql_dbg_io, fcport->vha, 0x3054,
"Dropped frame(s) detected (0x%x "
"of 0x%x bytes).\n", resid,
scsi_bufflen(cp));
res = DID_ERROR << 16 | le16_to_cpu(lscsi_status);
goto check_scsi_status;
} else {
ql_dbg(ql_dbg_io, fcport->vha, 0x3055,
"scsi_status: 0x%x, lscsi_status: 0x%x\n",
scsi_status, lscsi_status);
}
res = DID_OK << 16 | le16_to_cpu(lscsi_status);
logit = 0;
check_scsi_status:
/*
* Check to see if SCSI Status is non zero. If so report SCSI
* Status.
*/
if (lscsi_status != 0) {
if (lscsi_status ==
cpu_to_le16((uint16_t)SAM_STAT_TASK_SET_FULL)) {
ql_dbg(ql_dbg_io, fcport->vha, 0x3056,
"QUEUE FULL detected.\n");
logit = 1;
break;
}
if (lscsi_status !=
cpu_to_le16((uint16_t)SS_CHECK_CONDITION))
break;
memset(cp->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
if (!(scsi_status &
cpu_to_le16((uint16_t)SS_SENSE_LEN_VALID)))
break;
qlafx00_handle_sense(sp, sense_data, par_sense_len,
sense_len, rsp, res);
}
break;
case CS_PORT_LOGGED_OUT:
case CS_PORT_CONFIG_CHG:
case CS_PORT_BUSY:
case CS_INCOMPLETE:
case CS_PORT_UNAVAILABLE:
case CS_TIMEOUT:
case CS_RESET:
/*
* We are going to have the fc class block the rport
* while we try to recover so instruct the mid layer
* to requeue until the class decides how to handle this.
*/
res = DID_TRANSPORT_DISRUPTED << 16;
ql_dbg(ql_dbg_io, fcport->vha, 0x3057,
"Port down status: port-state=0x%x.\n",
atomic_read(&fcport->state));
if (atomic_read(&fcport->state) == FCS_ONLINE)
qla2x00_mark_device_lost(fcport->vha, fcport, 1, 1);
break;
case CS_ABORTED:
res = DID_RESET << 16;
break;
default:
res = DID_ERROR << 16;
break;
}
if (logit)
ql_dbg(ql_dbg_io, fcport->vha, 0x3058,
"FCP command status: 0x%x-0x%x (0x%x) nexus=%ld:%d:%llu "
"tgt_id: 0x%x lscsi_status: 0x%x cdb=%10phN len=0x%x "
"rsp_info=0x%x resid=0x%x fw_resid=0x%x sense_len=0x%x, "
"par_sense_len=0x%x, rsp_info_len=0x%x\n",
comp_status, scsi_status, res, vha->host_no,
cp->device->id, cp->device->lun, fcport->tgt_id,
lscsi_status, cp->cmnd, scsi_bufflen(cp),
rsp_info_len, resid_len, fw_resid_len, sense_len,
par_sense_len, rsp_info_len);
if (rsp->status_srb == NULL)
sp->done(ha, sp, res);
}
/**
* qlafx00_status_cont_entry() - Process a Status Continuations entry.
* @ha: SCSI driver HA context
* @pkt: Entry pointer
*
* Extended sense data.
*/
static void
qlafx00_status_cont_entry(struct rsp_que *rsp, sts_cont_entry_t *pkt)
{
uint8_t sense_sz = 0;
struct qla_hw_data *ha = rsp->hw;
struct scsi_qla_host *vha = pci_get_drvdata(ha->pdev);
srb_t *sp = rsp->status_srb;
struct scsi_cmnd *cp;
uint32_t sense_len;
uint8_t *sense_ptr;
if (!sp) {
ql_dbg(ql_dbg_io, vha, 0x3037,
"no SP, sp = %p\n", sp);
return;
}
if (!GET_FW_SENSE_LEN(sp)) {
ql_dbg(ql_dbg_io, vha, 0x304b,
"no fw sense data, sp = %p\n", sp);
return;
}
cp = GET_CMD_SP(sp);
if (cp == NULL) {
ql_log(ql_log_warn, vha, 0x303b,
"cmd is NULL: already returned to OS (sp=%p).\n", sp);
rsp->status_srb = NULL;
return;
}
if (!GET_CMD_SENSE_LEN(sp)) {
ql_dbg(ql_dbg_io, vha, 0x304c,
"no sense data, sp = %p\n", sp);
} else {
sense_len = GET_CMD_SENSE_LEN(sp);
sense_ptr = GET_CMD_SENSE_PTR(sp);
ql_dbg(ql_dbg_io, vha, 0x304f,
"sp=%p sense_len=0x%x sense_ptr=%p.\n",
sp, sense_len, sense_ptr);
if (sense_len > sizeof(pkt->data))
sense_sz = sizeof(pkt->data);
else
sense_sz = sense_len;
/* Move sense data. */
ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x304e,
(uint8_t *)pkt, sizeof(sts_cont_entry_t));
memcpy(sense_ptr, pkt->data, sense_sz);
ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x304a,
sense_ptr, sense_sz);
sense_len -= sense_sz;
sense_ptr += sense_sz;
SET_CMD_SENSE_PTR(sp, sense_ptr);
SET_CMD_SENSE_LEN(sp, sense_len);
}
sense_len = GET_FW_SENSE_LEN(sp);
sense_len = (sense_len > sizeof(pkt->data)) ?
(sense_len - sizeof(pkt->data)) : 0;
SET_FW_SENSE_LEN(sp, sense_len);
/* Place command on done queue. */
if (sense_len == 0) {
rsp->status_srb = NULL;
sp->done(ha, sp, cp->result);
}
}
/**
* qlafx00_multistatus_entry() - Process Multi response queue entries.
* @ha: SCSI driver HA context
*/
static void
qlafx00_multistatus_entry(struct scsi_qla_host *vha,
struct rsp_que *rsp, void *pkt)
{
srb_t *sp;
struct multi_sts_entry_fx00 *stsmfx;
struct qla_hw_data *ha = vha->hw;
uint32_t handle, hindex, handle_count, i;
uint16_t que;
struct req_que *req;
__le32 *handle_ptr;
stsmfx = (struct multi_sts_entry_fx00 *) pkt;
handle_count = stsmfx->handle_count;
if (handle_count > MAX_HANDLE_COUNT) {
ql_dbg(ql_dbg_io, vha, 0x3035,
"Invalid handle count (0x%x).\n", handle_count);
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
return;
}
handle_ptr = &stsmfx->handles[0];
for (i = 0; i < handle_count; i++) {
hindex = le32_to_cpu(*handle_ptr);
handle = LSW(hindex);
que = MSW(hindex);
req = ha->req_q_map[que];
/* Validate handle. */
if (handle < req->num_outstanding_cmds)
sp = req->outstanding_cmds[handle];
else
sp = NULL;
if (sp == NULL) {
ql_dbg(ql_dbg_io, vha, 0x3044,
"Invalid status handle (0x%x).\n", handle);
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
return;
}
qla2x00_process_completed_request(vha, req, handle);
handle_ptr++;
}
}
/**
* qlafx00_error_entry() - Process an error entry.
* @ha: SCSI driver HA context
* @pkt: Entry pointer
*/
static void
qlafx00_error_entry(scsi_qla_host_t *vha, struct rsp_que *rsp,
struct sts_entry_fx00 *pkt, uint8_t estatus, uint8_t etype)
{
srb_t *sp;
struct qla_hw_data *ha = vha->hw;
const char func[] = "ERROR-IOCB";
uint16_t que = 0;
struct req_que *req = NULL;
int res = DID_ERROR << 16;
ql_dbg(ql_dbg_async, vha, 0x507f,
"type of error status in response: 0x%x\n", estatus);
req = ha->req_q_map[que];
sp = qla2x00_get_sp_from_handle(vha, func, req, pkt);
if (sp) {
sp->done(ha, sp, res);
return;
}
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
}
/**
* qlafx00_process_response_queue() - Process response queue entries.
* @ha: SCSI driver HA context
*/
static void
qlafx00_process_response_queue(struct scsi_qla_host *vha,
struct rsp_que *rsp)
{
struct sts_entry_fx00 *pkt;
response_t *lptr;
uint16_t lreq_q_in = 0;
uint16_t lreq_q_out = 0;
lreq_q_in = RD_REG_DWORD(rsp->rsp_q_in);
lreq_q_out = rsp->ring_index;
while (lreq_q_in != lreq_q_out) {
lptr = rsp->ring_ptr;
memcpy_fromio(rsp->rsp_pkt, (void __iomem *)lptr,
sizeof(rsp->rsp_pkt));
pkt = (struct sts_entry_fx00 *)rsp->rsp_pkt;
rsp->ring_index++;
lreq_q_out++;
if (rsp->ring_index == rsp->length) {
lreq_q_out = 0;
rsp->ring_index = 0;
rsp->ring_ptr = rsp->ring;
} else {
rsp->ring_ptr++;
}
if (pkt->entry_status != 0 &&
pkt->entry_type != IOCTL_IOSB_TYPE_FX00) {
qlafx00_error_entry(vha, rsp,
(struct sts_entry_fx00 *)pkt, pkt->entry_status,
pkt->entry_type);
continue;
}
switch (pkt->entry_type) {
case STATUS_TYPE_FX00:
qlafx00_status_entry(vha, rsp, pkt);
break;
case STATUS_CONT_TYPE_FX00:
qlafx00_status_cont_entry(rsp, (sts_cont_entry_t *)pkt);
break;
case MULTI_STATUS_TYPE_FX00:
qlafx00_multistatus_entry(vha, rsp, pkt);
break;
case ABORT_IOCB_TYPE_FX00:
qlafx00_abort_iocb_entry(vha, rsp->req,
(struct abort_iocb_entry_fx00 *)pkt);
break;
case IOCTL_IOSB_TYPE_FX00:
qlafx00_ioctl_iosb_entry(vha, rsp->req,
(struct ioctl_iocb_entry_fx00 *)pkt);
break;
default:
/* Type Not Supported. */
ql_dbg(ql_dbg_async, vha, 0x5081,
"Received unknown response pkt type %x "
"entry status=%x.\n",
pkt->entry_type, pkt->entry_status);
break;
}
}
/* Adjust ring index */
WRT_REG_DWORD(rsp->rsp_q_out, rsp->ring_index);
}
/**
* qlafx00_async_event() - Process aynchronous events.
* @ha: SCSI driver HA context
*/
static void
qlafx00_async_event(scsi_qla_host_t *vha)
{
struct qla_hw_data *ha = vha->hw;
struct device_reg_fx00 __iomem *reg;
int data_size = 1;
reg = &ha->iobase->ispfx00;
/* Setup to process RIO completion. */
switch (ha->aenmb[0]) {
case QLAFX00_MBA_SYSTEM_ERR: /* System Error */
ql_log(ql_log_warn, vha, 0x5079,
"ISP System Error - mbx1=%x\n", ha->aenmb[0]);
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
break;
case QLAFX00_MBA_SHUTDOWN_RQSTD: /* Shutdown requested */
ql_dbg(ql_dbg_async, vha, 0x5076,
"Asynchronous FW shutdown requested.\n");
set_bit(ISP_ABORT_NEEDED, &vha->dpc_flags);
qla2xxx_wake_dpc(vha);
break;
case QLAFX00_MBA_PORT_UPDATE: /* Port database update */
ha->aenmb[1] = RD_REG_DWORD(&reg->aenmailbox1);
ha->aenmb[2] = RD_REG_DWORD(&reg->aenmailbox2);
ha->aenmb[3] = RD_REG_DWORD(&reg->aenmailbox3);
ql_dbg(ql_dbg_async, vha, 0x5077,
"Asynchronous port Update received "
"aenmb[0]: %x, aenmb[1]: %x, aenmb[2]: %x, aenmb[3]: %x\n",
ha->aenmb[0], ha->aenmb[1], ha->aenmb[2], ha->aenmb[3]);
data_size = 4;
break;
case QLAFX00_MBA_TEMP_OVER: /* Over temperature event */
ql_log(ql_log_info, vha, 0x5085,
"Asynchronous over temperature event received "
"aenmb[0]: %x\n",
ha->aenmb[0]);
break;
case QLAFX00_MBA_TEMP_NORM: /* Normal temperature event */
ql_log(ql_log_info, vha, 0x5086,
"Asynchronous normal temperature event received "
"aenmb[0]: %x\n",
ha->aenmb[0]);
break;
case QLAFX00_MBA_TEMP_CRIT: /* Critical temperature event */
ql_log(ql_log_info, vha, 0x5083,
"Asynchronous critical temperature event received "
"aenmb[0]: %x\n",
ha->aenmb[0]);
break;
default:
ha->aenmb[1] = RD_REG_WORD(&reg->aenmailbox1);
ha->aenmb[2] = RD_REG_WORD(&reg->aenmailbox2);
ha->aenmb[3] = RD_REG_WORD(&reg->aenmailbox3);
ha->aenmb[4] = RD_REG_WORD(&reg->aenmailbox4);
ha->aenmb[5] = RD_REG_WORD(&reg->aenmailbox5);
ha->aenmb[6] = RD_REG_WORD(&reg->aenmailbox6);
ha->aenmb[7] = RD_REG_WORD(&reg->aenmailbox7);
ql_dbg(ql_dbg_async, vha, 0x5078,
"AEN:%04x %04x %04x %04x :%04x %04x %04x %04x\n",
ha->aenmb[0], ha->aenmb[1], ha->aenmb[2], ha->aenmb[3],
ha->aenmb[4], ha->aenmb[5], ha->aenmb[6], ha->aenmb[7]);
break;
}
qlafx00_post_aenfx_work(vha, ha->aenmb[0],
(uint32_t *)ha->aenmb, data_size);
}
/**
*
* qlafx00x_mbx_completion() - Process mailbox command completions.
* @ha: SCSI driver HA context
* @mb16: Mailbox16 register
*/
static void
qlafx00_mbx_completion(scsi_qla_host_t *vha, uint32_t mb0)
{
uint16_t cnt;
uint32_t __iomem *wptr;
struct qla_hw_data *ha = vha->hw;
struct device_reg_fx00 __iomem *reg = &ha->iobase->ispfx00;
if (!ha->mcp32)
ql_dbg(ql_dbg_async, vha, 0x507e, "MBX pointer ERROR.\n");
/* Load return mailbox registers. */
ha->flags.mbox_int = 1;
ha->mailbox_out32[0] = mb0;
wptr = (uint32_t __iomem *)&reg->mailbox17;
for (cnt = 1; cnt < ha->mbx_count; cnt++) {
ha->mailbox_out32[cnt] = RD_REG_DWORD(wptr);
wptr++;
}
}
/**
* qlafx00_intr_handler() - Process interrupts for the ISPFX00.
* @irq:
* @dev_id: SCSI driver HA context
*
* Called by system whenever the host adapter generates an interrupt.
*
* Returns handled flag.
*/
irqreturn_t
qlafx00_intr_handler(int irq, void *dev_id)
{
scsi_qla_host_t *vha;
struct qla_hw_data *ha;
struct device_reg_fx00 __iomem *reg;
int status;
unsigned long iter;
uint32_t stat;
uint32_t mb[8];
struct rsp_que *rsp;
unsigned long flags;
uint32_t clr_intr = 0;
uint32_t intr_stat = 0;
rsp = (struct rsp_que *) dev_id;
if (!rsp) {
ql_log(ql_log_info, NULL, 0x507d,
"%s: NULL response queue pointer.\n", __func__);
return IRQ_NONE;
}
ha = rsp->hw;
reg = &ha->iobase->ispfx00;
status = 0;
if (unlikely(pci_channel_offline(ha->pdev)))
return IRQ_HANDLED;
spin_lock_irqsave(&ha->hardware_lock, flags);
vha = pci_get_drvdata(ha->pdev);
for (iter = 50; iter--; clr_intr = 0) {
stat = QLAFX00_RD_INTR_REG(ha);
if (qla2x00_check_reg32_for_disconnect(vha, stat))
break;
intr_stat = stat & QLAFX00_HST_INT_STS_BITS;
if (!intr_stat)
break;
if (stat & QLAFX00_INTR_MB_CMPLT) {
mb[0] = RD_REG_WORD(&reg->mailbox16);
qlafx00_mbx_completion(vha, mb[0]);
status |= MBX_INTERRUPT;
clr_intr |= QLAFX00_INTR_MB_CMPLT;
}
if (intr_stat & QLAFX00_INTR_ASYNC_CMPLT) {
ha->aenmb[0] = RD_REG_WORD(&reg->aenmailbox0);
qlafx00_async_event(vha);
clr_intr |= QLAFX00_INTR_ASYNC_CMPLT;
}
if (intr_stat & QLAFX00_INTR_RSP_CMPLT) {
qlafx00_process_response_queue(vha, rsp);
clr_intr |= QLAFX00_INTR_RSP_CMPLT;
}
QLAFX00_CLR_INTR_REG(ha, clr_intr);
QLAFX00_RD_INTR_REG(ha);
}
qla2x00_handle_mbx_completion(ha, status);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
return IRQ_HANDLED;
}
/** QLAFX00 specific IOCB implementation functions */
static inline cont_a64_entry_t *
qlafx00_prep_cont_type1_iocb(struct req_que *req,
cont_a64_entry_t *lcont_pkt)
{
cont_a64_entry_t *cont_pkt;
/* Adjust ring index. */
req->ring_index++;
if (req->ring_index == req->length) {
req->ring_index = 0;
req->ring_ptr = req->ring;
} else {
req->ring_ptr++;
}
cont_pkt = (cont_a64_entry_t *)req->ring_ptr;
/* Load packet defaults. */
lcont_pkt->entry_type = CONTINUE_A64_TYPE_FX00;
return cont_pkt;
}
static inline void
qlafx00_build_scsi_iocbs(srb_t *sp, struct cmd_type_7_fx00 *cmd_pkt,
uint16_t tot_dsds, struct cmd_type_7_fx00 *lcmd_pkt)
{
uint16_t avail_dsds;
__le32 *cur_dsd;
scsi_qla_host_t *vha;
struct scsi_cmnd *cmd;
struct scatterlist *sg;
int i, cont;
struct req_que *req;
cont_a64_entry_t lcont_pkt;
cont_a64_entry_t *cont_pkt;
vha = sp->fcport->vha;
req = vha->req;
cmd = GET_CMD_SP(sp);
cont = 0;
cont_pkt = NULL;
/* Update entry type to indicate Command Type 3 IOCB */
lcmd_pkt->entry_type = FX00_COMMAND_TYPE_7;
/* No data transfer */
if (!scsi_bufflen(cmd) || cmd->sc_data_direction == DMA_NONE) {
lcmd_pkt->byte_count = __constant_cpu_to_le32(0);
return;
}
/* Set transfer direction */
if (cmd->sc_data_direction == DMA_TO_DEVICE) {
lcmd_pkt->cntrl_flags = TMF_WRITE_DATA;
vha->qla_stats.output_bytes += scsi_bufflen(cmd);
} else if (cmd->sc_data_direction == DMA_FROM_DEVICE) {
lcmd_pkt->cntrl_flags = TMF_READ_DATA;
vha->qla_stats.input_bytes += scsi_bufflen(cmd);
}
/* One DSD is available in the Command Type 3 IOCB */
avail_dsds = 1;
cur_dsd = (__le32 *)&lcmd_pkt->dseg_0_address;
/* Load data segments */
scsi_for_each_sg(cmd, sg, tot_dsds, i) {
dma_addr_t sle_dma;
/* Allocate additional continuation packets? */
if (avail_dsds == 0) {
/*
* Five DSDs are available in the Continuation
* Type 1 IOCB.
*/
memset(&lcont_pkt, 0, REQUEST_ENTRY_SIZE);
cont_pkt =
qlafx00_prep_cont_type1_iocb(req, &lcont_pkt);
cur_dsd = (__le32 *)lcont_pkt.dseg_0_address;
avail_dsds = 5;
cont = 1;
}
sle_dma = sg_dma_address(sg);
*cur_dsd++ = cpu_to_le32(LSD(sle_dma));
*cur_dsd++ = cpu_to_le32(MSD(sle_dma));
*cur_dsd++ = cpu_to_le32(sg_dma_len(sg));
avail_dsds--;
if (avail_dsds == 0 && cont == 1) {
cont = 0;
memcpy_toio((void __iomem *)cont_pkt, &lcont_pkt,
REQUEST_ENTRY_SIZE);
}
}
if (avail_dsds != 0 && cont == 1) {
memcpy_toio((void __iomem *)cont_pkt, &lcont_pkt,
REQUEST_ENTRY_SIZE);
}
}
/**
* qlafx00_start_scsi() - Send a SCSI command to the ISP
* @sp: command to send to the ISP
*
* Returns non-zero if a failure occurred, else zero.
*/
int
qlafx00_start_scsi(srb_t *sp)
{
int ret, nseg;
unsigned long flags;
uint32_t index;
uint32_t handle;
uint16_t cnt;
uint16_t req_cnt;
uint16_t tot_dsds;
struct req_que *req = NULL;
struct rsp_que *rsp = NULL;
struct scsi_cmnd *cmd = GET_CMD_SP(sp);
struct scsi_qla_host *vha = sp->fcport->vha;
struct qla_hw_data *ha = vha->hw;
struct cmd_type_7_fx00 *cmd_pkt;
struct cmd_type_7_fx00 lcmd_pkt;
struct scsi_lun llun;
char tag[2];
/* Setup device pointers. */
ret = 0;
rsp = ha->rsp_q_map[0];
req = vha->req;
/* So we know we haven't pci_map'ed anything yet */
tot_dsds = 0;
/* Acquire ring specific lock */
spin_lock_irqsave(&ha->hardware_lock, flags);
/* Check for room in outstanding command list. */
handle = req->current_outstanding_cmd;
for (index = 1; index < req->num_outstanding_cmds; index++) {
handle++;
if (handle == req->num_outstanding_cmds)
handle = 1;
if (!req->outstanding_cmds[handle])
break;
}
if (index == req->num_outstanding_cmds)
goto queuing_error;
/* Map the sg table so we have an accurate count of sg entries needed */
if (scsi_sg_count(cmd)) {
nseg = dma_map_sg(&ha->pdev->dev, scsi_sglist(cmd),
scsi_sg_count(cmd), cmd->sc_data_direction);
if (unlikely(!nseg))
goto queuing_error;
} else
nseg = 0;
tot_dsds = nseg;
req_cnt = qla24xx_calc_iocbs(vha, tot_dsds);
if (req->cnt < (req_cnt + 2)) {
cnt = RD_REG_DWORD_RELAXED(req->req_q_out);
if (req->ring_index < cnt)
req->cnt = cnt - req->ring_index;
else
req->cnt = req->length -
(req->ring_index - cnt);
if (req->cnt < (req_cnt + 2))
goto queuing_error;
}
/* Build command packet. */
req->current_outstanding_cmd = handle;
req->outstanding_cmds[handle] = sp;
sp->handle = handle;
cmd->host_scribble = (unsigned char *)(unsigned long)handle;
req->cnt -= req_cnt;
cmd_pkt = (struct cmd_type_7_fx00 *)req->ring_ptr;
memset(&lcmd_pkt, 0, REQUEST_ENTRY_SIZE);
lcmd_pkt.handle = MAKE_HANDLE(req->id, sp->handle);
lcmd_pkt.reserved_0 = 0;
lcmd_pkt.port_path_ctrl = 0;
lcmd_pkt.reserved_1 = 0;
lcmd_pkt.dseg_count = cpu_to_le16(tot_dsds);
lcmd_pkt.tgt_idx = cpu_to_le16(sp->fcport->tgt_id);
int_to_scsilun(cmd->device->lun, &llun);
host_to_adap((uint8_t *)&llun, (uint8_t *)&lcmd_pkt.lun,
sizeof(lcmd_pkt.lun));
/* Update tagged queuing modifier -- default is TSK_SIMPLE (0). */
if (scsi_populate_tag_msg(cmd, tag)) {
switch (tag[0]) {
case HEAD_OF_QUEUE_TAG:
lcmd_pkt.task = TSK_HEAD_OF_QUEUE;
break;
case ORDERED_QUEUE_TAG:
lcmd_pkt.task = TSK_ORDERED;
break;
}
}
/* Load SCSI command packet. */
host_to_adap(cmd->cmnd, lcmd_pkt.fcp_cdb, sizeof(lcmd_pkt.fcp_cdb));
lcmd_pkt.byte_count = cpu_to_le32((uint32_t)scsi_bufflen(cmd));
/* Build IOCB segments */
qlafx00_build_scsi_iocbs(sp, cmd_pkt, tot_dsds, &lcmd_pkt);
/* Set total data segment count. */
lcmd_pkt.entry_count = (uint8_t)req_cnt;
/* Specify response queue number where completion should happen */
lcmd_pkt.entry_status = (uint8_t) rsp->id;
ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x302e,
(uint8_t *)cmd->cmnd, cmd->cmd_len);
ql_dump_buffer(ql_dbg_io + ql_dbg_buffer, vha, 0x3032,
(uint8_t *)&lcmd_pkt, REQUEST_ENTRY_SIZE);
memcpy_toio((void __iomem *)cmd_pkt, &lcmd_pkt, REQUEST_ENTRY_SIZE);
wmb();
/* Adjust ring index. */
req->ring_index++;
if (req->ring_index == req->length) {
req->ring_index = 0;
req->ring_ptr = req->ring;
} else
req->ring_ptr++;
sp->flags |= SRB_DMA_VALID;
/* Set chip new ring index. */
WRT_REG_DWORD(req->req_q_in, req->ring_index);
QLAFX00_SET_HST_INTR(ha, ha->rqstq_intr_code);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
return QLA_SUCCESS;
queuing_error:
if (tot_dsds)
scsi_dma_unmap(cmd);
spin_unlock_irqrestore(&ha->hardware_lock, flags);
return QLA_FUNCTION_FAILED;
}
void
qlafx00_tm_iocb(srb_t *sp, struct tsk_mgmt_entry_fx00 *ptm_iocb)
{
struct srb_iocb *fxio = &sp->u.iocb_cmd;
scsi_qla_host_t *vha = sp->fcport->vha;
struct req_que *req = vha->req;
struct tsk_mgmt_entry_fx00 tm_iocb;
struct scsi_lun llun;
memset(&tm_iocb, 0, sizeof(struct tsk_mgmt_entry_fx00));
tm_iocb.entry_type = TSK_MGMT_IOCB_TYPE_FX00;
tm_iocb.entry_count = 1;
tm_iocb.handle = cpu_to_le32(MAKE_HANDLE(req->id, sp->handle));
tm_iocb.reserved_0 = 0;
tm_iocb.tgt_id = cpu_to_le16(sp->fcport->tgt_id);
tm_iocb.control_flags = cpu_to_le32(fxio->u.tmf.flags);
if (tm_iocb.control_flags == cpu_to_le32((uint32_t)TCF_LUN_RESET)) {
int_to_scsilun(fxio->u.tmf.lun, &llun);
host_to_adap((uint8_t *)&llun, (uint8_t *)&tm_iocb.lun,
sizeof(struct scsi_lun));
}
memcpy((void *)ptm_iocb, &tm_iocb,
sizeof(struct tsk_mgmt_entry_fx00));
wmb();
}
void
qlafx00_abort_iocb(srb_t *sp, struct abort_iocb_entry_fx00 *pabt_iocb)
{
struct srb_iocb *fxio = &sp->u.iocb_cmd;
scsi_qla_host_t *vha = sp->fcport->vha;
struct req_que *req = vha->req;
struct abort_iocb_entry_fx00 abt_iocb;
memset(&abt_iocb, 0, sizeof(struct abort_iocb_entry_fx00));
abt_iocb.entry_type = ABORT_IOCB_TYPE_FX00;
abt_iocb.entry_count = 1;
abt_iocb.handle = cpu_to_le32(MAKE_HANDLE(req->id, sp->handle));
abt_iocb.abort_handle =
cpu_to_le32(MAKE_HANDLE(req->id, fxio->u.abt.cmd_hndl));
abt_iocb.tgt_id_sts = cpu_to_le16(sp->fcport->tgt_id);
abt_iocb.req_que_no = cpu_to_le16(req->id);
memcpy((void *)pabt_iocb, &abt_iocb,
sizeof(struct abort_iocb_entry_fx00));
wmb();
}
void
qlafx00_fxdisc_iocb(srb_t *sp, struct fxdisc_entry_fx00 *pfxiocb)
{
struct srb_iocb *fxio = &sp->u.iocb_cmd;
struct qla_mt_iocb_rqst_fx00 *piocb_rqst;
struct fc_bsg_job *bsg_job;
struct fxdisc_entry_fx00 fx_iocb;
uint8_t entry_cnt = 1;
memset(&fx_iocb, 0, sizeof(struct fxdisc_entry_fx00));
fx_iocb.entry_type = FX00_IOCB_TYPE;
fx_iocb.handle = cpu_to_le32(sp->handle);
fx_iocb.entry_count = entry_cnt;
if (sp->type == SRB_FXIOCB_DCMD) {
fx_iocb.func_num =
sp->u.iocb_cmd.u.fxiocb.req_func_type;
fx_iocb.adapid = fxio->u.fxiocb.adapter_id;
fx_iocb.adapid_hi = fxio->u.fxiocb.adapter_id_hi;
fx_iocb.reserved_0 = fxio->u.fxiocb.reserved_0;
fx_iocb.reserved_1 = fxio->u.fxiocb.reserved_1;
fx_iocb.dataword_extra = fxio->u.fxiocb.req_data_extra;
if (fxio->u.fxiocb.flags & SRB_FXDISC_REQ_DMA_VALID) {
fx_iocb.req_dsdcnt = cpu_to_le16(1);
fx_iocb.req_xfrcnt =
cpu_to_le16(fxio->u.fxiocb.req_len);
fx_iocb.dseg_rq_address[0] =
cpu_to_le32(LSD(fxio->u.fxiocb.req_dma_handle));
fx_iocb.dseg_rq_address[1] =
cpu_to_le32(MSD(fxio->u.fxiocb.req_dma_handle));
fx_iocb.dseg_rq_len =
cpu_to_le32(fxio->u.fxiocb.req_len);
}
if (fxio->u.fxiocb.flags & SRB_FXDISC_RESP_DMA_VALID) {
fx_iocb.rsp_dsdcnt = cpu_to_le16(1);
fx_iocb.rsp_xfrcnt =
cpu_to_le16(fxio->u.fxiocb.rsp_len);
fx_iocb.dseg_rsp_address[0] =
cpu_to_le32(LSD(fxio->u.fxiocb.rsp_dma_handle));
fx_iocb.dseg_rsp_address[1] =
cpu_to_le32(MSD(fxio->u.fxiocb.rsp_dma_handle));
fx_iocb.dseg_rsp_len =
cpu_to_le32(fxio->u.fxiocb.rsp_len);
}
if (fxio->u.fxiocb.flags & SRB_FXDISC_REQ_DWRD_VALID) {
fx_iocb.dataword = fxio->u.fxiocb.req_data;
}
fx_iocb.flags = fxio->u.fxiocb.flags;
} else {
struct scatterlist *sg;
bsg_job = sp->u.bsg_job;
piocb_rqst = (struct qla_mt_iocb_rqst_fx00 *)
&bsg_job->request->rqst_data.h_vendor.vendor_cmd[1];
fx_iocb.func_num = piocb_rqst->func_type;
fx_iocb.adapid = piocb_rqst->adapid;
fx_iocb.adapid_hi = piocb_rqst->adapid_hi;
fx_iocb.reserved_0 = piocb_rqst->reserved_0;
fx_iocb.reserved_1 = piocb_rqst->reserved_1;
fx_iocb.dataword_extra = piocb_rqst->dataword_extra;
fx_iocb.dataword = piocb_rqst->dataword;
fx_iocb.req_xfrcnt = piocb_rqst->req_len;
fx_iocb.rsp_xfrcnt = piocb_rqst->rsp_len;
if (piocb_rqst->flags & SRB_FXDISC_REQ_DMA_VALID) {
int avail_dsds, tot_dsds;
cont_a64_entry_t lcont_pkt;
cont_a64_entry_t *cont_pkt = NULL;
__le32 *cur_dsd;
int index = 0, cont = 0;
fx_iocb.req_dsdcnt =
cpu_to_le16(bsg_job->request_payload.sg_cnt);
tot_dsds =
bsg_job->request_payload.sg_cnt;
cur_dsd = (__le32 *)&fx_iocb.dseg_rq_address[0];
avail_dsds = 1;
for_each_sg(bsg_job->request_payload.sg_list, sg,
tot_dsds, index) {
dma_addr_t sle_dma;
/* Allocate additional continuation packets? */
if (avail_dsds == 0) {
/*
* Five DSDs are available in the Cont.
* Type 1 IOCB.
*/
memset(&lcont_pkt, 0,
REQUEST_ENTRY_SIZE);
cont_pkt =
qlafx00_prep_cont_type1_iocb(
sp->fcport->vha->req,
&lcont_pkt);
cur_dsd = (__le32 *)
lcont_pkt.dseg_0_address;
avail_dsds = 5;
cont = 1;
entry_cnt++;
}
sle_dma = sg_dma_address(sg);
*cur_dsd++ = cpu_to_le32(LSD(sle_dma));
*cur_dsd++ = cpu_to_le32(MSD(sle_dma));
*cur_dsd++ = cpu_to_le32(sg_dma_len(sg));
avail_dsds--;
if (avail_dsds == 0 && cont == 1) {
cont = 0;
memcpy_toio(
(void __iomem *)cont_pkt,
&lcont_pkt, REQUEST_ENTRY_SIZE);
ql_dump_buffer(
ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x3042,
(uint8_t *)&lcont_pkt,
REQUEST_ENTRY_SIZE);
}
}
if (avail_dsds != 0 && cont == 1) {
memcpy_toio((void __iomem *)cont_pkt,
&lcont_pkt, REQUEST_ENTRY_SIZE);
ql_dump_buffer(ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x3043,
(uint8_t *)&lcont_pkt, REQUEST_ENTRY_SIZE);
}
}
if (piocb_rqst->flags & SRB_FXDISC_RESP_DMA_VALID) {
int avail_dsds, tot_dsds;
cont_a64_entry_t lcont_pkt;
cont_a64_entry_t *cont_pkt = NULL;
__le32 *cur_dsd;
int index = 0, cont = 0;
fx_iocb.rsp_dsdcnt =
cpu_to_le16(bsg_job->reply_payload.sg_cnt);
tot_dsds = bsg_job->reply_payload.sg_cnt;
cur_dsd = (__le32 *)&fx_iocb.dseg_rsp_address[0];
avail_dsds = 1;
for_each_sg(bsg_job->reply_payload.sg_list, sg,
tot_dsds, index) {
dma_addr_t sle_dma;
/* Allocate additional continuation packets? */
if (avail_dsds == 0) {
/*
* Five DSDs are available in the Cont.
* Type 1 IOCB.
*/
memset(&lcont_pkt, 0,
REQUEST_ENTRY_SIZE);
cont_pkt =
qlafx00_prep_cont_type1_iocb(
sp->fcport->vha->req,
&lcont_pkt);
cur_dsd = (__le32 *)
lcont_pkt.dseg_0_address;
avail_dsds = 5;
cont = 1;
entry_cnt++;
}
sle_dma = sg_dma_address(sg);
*cur_dsd++ = cpu_to_le32(LSD(sle_dma));
*cur_dsd++ = cpu_to_le32(MSD(sle_dma));
*cur_dsd++ = cpu_to_le32(sg_dma_len(sg));
avail_dsds--;
if (avail_dsds == 0 && cont == 1) {
cont = 0;
memcpy_toio((void __iomem *)cont_pkt,
&lcont_pkt,
REQUEST_ENTRY_SIZE);
ql_dump_buffer(
ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x3045,
(uint8_t *)&lcont_pkt,
REQUEST_ENTRY_SIZE);
}
}
if (avail_dsds != 0 && cont == 1) {
memcpy_toio((void __iomem *)cont_pkt,
&lcont_pkt, REQUEST_ENTRY_SIZE);
ql_dump_buffer(ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x3046,
(uint8_t *)&lcont_pkt, REQUEST_ENTRY_SIZE);
}
}
if (piocb_rqst->flags & SRB_FXDISC_REQ_DWRD_VALID)
fx_iocb.dataword = piocb_rqst->dataword;
fx_iocb.flags = piocb_rqst->flags;
fx_iocb.entry_count = entry_cnt;
}
ql_dump_buffer(ql_dbg_user + ql_dbg_verbose,
sp->fcport->vha, 0x3047,
(uint8_t *)&fx_iocb, sizeof(struct fxdisc_entry_fx00));
memcpy_toio((void __iomem *)pfxiocb, &fx_iocb,
sizeof(struct fxdisc_entry_fx00));
wmb();
}