blob: ceff1ec13f9ea9ea056da947d3939c51f4797522 [file] [log] [blame]
// SPDX-License-Identifier: GPL-2.0
/*
* driver for Microchip PQI-based storage controllers
* Copyright (c) 2019-2023 Microchip Technology Inc. and its subsidiaries
* Copyright (c) 2016-2018 Microsemi Corporation
* Copyright (c) 2016 PMC-Sierra, Inc.
*
* Questions/Comments/Bugfixes to storagedev@microchip.com
*
*/
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/interrupt.h>
#include <linux/sched.h>
#include <linux/rtc.h>
#include <linux/bcd.h>
#include <linux/reboot.h>
#include <linux/cciss_ioctl.h>
#include <linux/blk-mq-pci.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_eh.h>
#include <scsi/scsi_transport_sas.h>
#include <asm/unaligned.h>
#include "smartpqi.h"
#include "smartpqi_sis.h"
#if !defined(BUILD_TIMESTAMP)
#define BUILD_TIMESTAMP
#endif
#define DRIVER_VERSION "2.1.26-030"
#define DRIVER_MAJOR 2
#define DRIVER_MINOR 1
#define DRIVER_RELEASE 26
#define DRIVER_REVISION 30
#define DRIVER_NAME "Microchip SmartPQI Driver (v" \
DRIVER_VERSION BUILD_TIMESTAMP ")"
#define DRIVER_NAME_SHORT "smartpqi"
#define PQI_EXTRA_SGL_MEMORY (12 * sizeof(struct pqi_sg_descriptor))
#define PQI_POST_RESET_DELAY_SECS 5
#define PQI_POST_OFA_RESET_DELAY_UPON_TIMEOUT_SECS 10
#define PQI_NO_COMPLETION ((void *)-1)
MODULE_AUTHOR("Microchip");
MODULE_DESCRIPTION("Driver for Microchip Smart Family Controller version "
DRIVER_VERSION);
MODULE_VERSION(DRIVER_VERSION);
MODULE_LICENSE("GPL");
struct pqi_cmd_priv {
int this_residual;
};
static struct pqi_cmd_priv *pqi_cmd_priv(struct scsi_cmnd *cmd)
{
return scsi_cmd_priv(cmd);
}
static void pqi_verify_structures(void);
static void pqi_take_ctrl_offline(struct pqi_ctrl_info *ctrl_info,
enum pqi_ctrl_shutdown_reason ctrl_shutdown_reason);
static void pqi_ctrl_offline_worker(struct work_struct *work);
static int pqi_scan_scsi_devices(struct pqi_ctrl_info *ctrl_info);
static void pqi_scan_start(struct Scsi_Host *shost);
static void pqi_start_io(struct pqi_ctrl_info *ctrl_info,
struct pqi_queue_group *queue_group, enum pqi_io_path path,
struct pqi_io_request *io_request);
static int pqi_submit_raid_request_synchronous(struct pqi_ctrl_info *ctrl_info,
struct pqi_iu_header *request, unsigned int flags,
struct pqi_raid_error_info *error_info);
static int pqi_aio_submit_io(struct pqi_ctrl_info *ctrl_info,
struct scsi_cmnd *scmd, u32 aio_handle, u8 *cdb,
unsigned int cdb_length, struct pqi_queue_group *queue_group,
struct pqi_encryption_info *encryption_info, bool raid_bypass, bool io_high_prio);
static int pqi_aio_submit_r1_write_io(struct pqi_ctrl_info *ctrl_info,
struct scsi_cmnd *scmd, struct pqi_queue_group *queue_group,
struct pqi_encryption_info *encryption_info, struct pqi_scsi_dev *device,
struct pqi_scsi_dev_raid_map_data *rmd);
static int pqi_aio_submit_r56_write_io(struct pqi_ctrl_info *ctrl_info,
struct scsi_cmnd *scmd, struct pqi_queue_group *queue_group,
struct pqi_encryption_info *encryption_info, struct pqi_scsi_dev *device,
struct pqi_scsi_dev_raid_map_data *rmd);
static void pqi_ofa_ctrl_quiesce(struct pqi_ctrl_info *ctrl_info);
static void pqi_ofa_ctrl_unquiesce(struct pqi_ctrl_info *ctrl_info);
static int pqi_ofa_ctrl_restart(struct pqi_ctrl_info *ctrl_info, unsigned int delay_secs);
static void pqi_ofa_setup_host_buffer(struct pqi_ctrl_info *ctrl_info);
static void pqi_ofa_free_host_buffer(struct pqi_ctrl_info *ctrl_info);
static int pqi_ofa_host_memory_update(struct pqi_ctrl_info *ctrl_info);
static int pqi_device_wait_for_pending_io(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device, u8 lun, unsigned long timeout_msecs);
static void pqi_fail_all_outstanding_requests(struct pqi_ctrl_info *ctrl_info);
static void pqi_tmf_worker(struct work_struct *work);
/* for flags argument to pqi_submit_raid_request_synchronous() */
#define PQI_SYNC_FLAGS_INTERRUPTABLE 0x1
static struct scsi_transport_template *pqi_sas_transport_template;
static atomic_t pqi_controller_count = ATOMIC_INIT(0);
enum pqi_lockup_action {
NONE,
REBOOT,
PANIC
};
static enum pqi_lockup_action pqi_lockup_action = NONE;
static struct {
enum pqi_lockup_action action;
char *name;
} pqi_lockup_actions[] = {
{
.action = NONE,
.name = "none",
},
{
.action = REBOOT,
.name = "reboot",
},
{
.action = PANIC,
.name = "panic",
},
};
static unsigned int pqi_supported_event_types[] = {
PQI_EVENT_TYPE_HOTPLUG,
PQI_EVENT_TYPE_HARDWARE,
PQI_EVENT_TYPE_PHYSICAL_DEVICE,
PQI_EVENT_TYPE_LOGICAL_DEVICE,
PQI_EVENT_TYPE_OFA,
PQI_EVENT_TYPE_AIO_STATE_CHANGE,
PQI_EVENT_TYPE_AIO_CONFIG_CHANGE,
};
static int pqi_disable_device_id_wildcards;
module_param_named(disable_device_id_wildcards,
pqi_disable_device_id_wildcards, int, 0644);
MODULE_PARM_DESC(disable_device_id_wildcards,
"Disable device ID wildcards.");
static int pqi_disable_heartbeat;
module_param_named(disable_heartbeat,
pqi_disable_heartbeat, int, 0644);
MODULE_PARM_DESC(disable_heartbeat,
"Disable heartbeat.");
static int pqi_disable_ctrl_shutdown;
module_param_named(disable_ctrl_shutdown,
pqi_disable_ctrl_shutdown, int, 0644);
MODULE_PARM_DESC(disable_ctrl_shutdown,
"Disable controller shutdown when controller locked up.");
static char *pqi_lockup_action_param;
module_param_named(lockup_action,
pqi_lockup_action_param, charp, 0644);
MODULE_PARM_DESC(lockup_action, "Action to take when controller locked up.\n"
"\t\tSupported: none, reboot, panic\n"
"\t\tDefault: none");
static int pqi_expose_ld_first;
module_param_named(expose_ld_first,
pqi_expose_ld_first, int, 0644);
MODULE_PARM_DESC(expose_ld_first, "Expose logical drives before physical drives.");
static int pqi_hide_vsep;
module_param_named(hide_vsep,
pqi_hide_vsep, int, 0644);
MODULE_PARM_DESC(hide_vsep, "Hide the virtual SEP for direct attached drives.");
static int pqi_disable_managed_interrupts;
module_param_named(disable_managed_interrupts,
pqi_disable_managed_interrupts, int, 0644);
MODULE_PARM_DESC(disable_managed_interrupts,
"Disable the kernel automatically assigning SMP affinity to IRQs.");
static unsigned int pqi_ctrl_ready_timeout_secs;
module_param_named(ctrl_ready_timeout,
pqi_ctrl_ready_timeout_secs, uint, 0644);
MODULE_PARM_DESC(ctrl_ready_timeout,
"Timeout in seconds for driver to wait for controller ready.");
static char *raid_levels[] = {
"RAID-0",
"RAID-4",
"RAID-1(1+0)",
"RAID-5",
"RAID-5+1",
"RAID-6",
"RAID-1(Triple)",
};
static char *pqi_raid_level_to_string(u8 raid_level)
{
if (raid_level < ARRAY_SIZE(raid_levels))
return raid_levels[raid_level];
return "RAID UNKNOWN";
}
#define SA_RAID_0 0
#define SA_RAID_4 1
#define SA_RAID_1 2 /* also used for RAID 10 */
#define SA_RAID_5 3 /* also used for RAID 50 */
#define SA_RAID_51 4
#define SA_RAID_6 5 /* also used for RAID 60 */
#define SA_RAID_TRIPLE 6 /* also used for RAID 1+0 Triple */
#define SA_RAID_MAX SA_RAID_TRIPLE
#define SA_RAID_UNKNOWN 0xff
static inline void pqi_scsi_done(struct scsi_cmnd *scmd)
{
pqi_prep_for_scsi_done(scmd);
scsi_done(scmd);
}
static inline void pqi_disable_write_same(struct scsi_device *sdev)
{
sdev->no_write_same = 1;
}
static inline bool pqi_scsi3addr_equal(u8 *scsi3addr1, u8 *scsi3addr2)
{
return memcmp(scsi3addr1, scsi3addr2, 8) == 0;
}
static inline bool pqi_is_logical_device(struct pqi_scsi_dev *device)
{
return !device->is_physical_device;
}
static inline bool pqi_is_external_raid_addr(u8 *scsi3addr)
{
return scsi3addr[2] != 0;
}
static inline bool pqi_ctrl_offline(struct pqi_ctrl_info *ctrl_info)
{
return !ctrl_info->controller_online;
}
static inline void pqi_check_ctrl_health(struct pqi_ctrl_info *ctrl_info)
{
if (ctrl_info->controller_online)
if (!sis_is_firmware_running(ctrl_info))
pqi_take_ctrl_offline(ctrl_info, PQI_FIRMWARE_KERNEL_NOT_UP);
}
static inline bool pqi_is_hba_lunid(u8 *scsi3addr)
{
return pqi_scsi3addr_equal(scsi3addr, RAID_CTLR_LUNID);
}
#define PQI_DRIVER_SCRATCH_PQI_MODE 0x1
#define PQI_DRIVER_SCRATCH_FW_TRIAGE_SUPPORTED 0x2
static inline enum pqi_ctrl_mode pqi_get_ctrl_mode(struct pqi_ctrl_info *ctrl_info)
{
return sis_read_driver_scratch(ctrl_info) & PQI_DRIVER_SCRATCH_PQI_MODE ? PQI_MODE : SIS_MODE;
}
static inline void pqi_save_ctrl_mode(struct pqi_ctrl_info *ctrl_info,
enum pqi_ctrl_mode mode)
{
u32 driver_scratch;
driver_scratch = sis_read_driver_scratch(ctrl_info);
if (mode == PQI_MODE)
driver_scratch |= PQI_DRIVER_SCRATCH_PQI_MODE;
else
driver_scratch &= ~PQI_DRIVER_SCRATCH_PQI_MODE;
sis_write_driver_scratch(ctrl_info, driver_scratch);
}
static inline bool pqi_is_fw_triage_supported(struct pqi_ctrl_info *ctrl_info)
{
return (sis_read_driver_scratch(ctrl_info) & PQI_DRIVER_SCRATCH_FW_TRIAGE_SUPPORTED) != 0;
}
static inline void pqi_save_fw_triage_setting(struct pqi_ctrl_info *ctrl_info, bool is_supported)
{
u32 driver_scratch;
driver_scratch = sis_read_driver_scratch(ctrl_info);
if (is_supported)
driver_scratch |= PQI_DRIVER_SCRATCH_FW_TRIAGE_SUPPORTED;
else
driver_scratch &= ~PQI_DRIVER_SCRATCH_FW_TRIAGE_SUPPORTED;
sis_write_driver_scratch(ctrl_info, driver_scratch);
}
static inline void pqi_ctrl_block_scan(struct pqi_ctrl_info *ctrl_info)
{
ctrl_info->scan_blocked = true;
mutex_lock(&ctrl_info->scan_mutex);
}
static inline void pqi_ctrl_unblock_scan(struct pqi_ctrl_info *ctrl_info)
{
ctrl_info->scan_blocked = false;
mutex_unlock(&ctrl_info->scan_mutex);
}
static inline bool pqi_ctrl_scan_blocked(struct pqi_ctrl_info *ctrl_info)
{
return ctrl_info->scan_blocked;
}
static inline void pqi_ctrl_block_device_reset(struct pqi_ctrl_info *ctrl_info)
{
mutex_lock(&ctrl_info->lun_reset_mutex);
}
static inline void pqi_ctrl_unblock_device_reset(struct pqi_ctrl_info *ctrl_info)
{
mutex_unlock(&ctrl_info->lun_reset_mutex);
}
static inline void pqi_scsi_block_requests(struct pqi_ctrl_info *ctrl_info)
{
struct Scsi_Host *shost;
unsigned int num_loops;
int msecs_sleep;
shost = ctrl_info->scsi_host;
scsi_block_requests(shost);
num_loops = 0;
msecs_sleep = 20;
while (scsi_host_busy(shost)) {
num_loops++;
if (num_loops == 10)
msecs_sleep = 500;
msleep(msecs_sleep);
}
}
static inline void pqi_scsi_unblock_requests(struct pqi_ctrl_info *ctrl_info)
{
scsi_unblock_requests(ctrl_info->scsi_host);
}
static inline void pqi_ctrl_busy(struct pqi_ctrl_info *ctrl_info)
{
atomic_inc(&ctrl_info->num_busy_threads);
}
static inline void pqi_ctrl_unbusy(struct pqi_ctrl_info *ctrl_info)
{
atomic_dec(&ctrl_info->num_busy_threads);
}
static inline bool pqi_ctrl_blocked(struct pqi_ctrl_info *ctrl_info)
{
return ctrl_info->block_requests;
}
static inline void pqi_ctrl_block_requests(struct pqi_ctrl_info *ctrl_info)
{
ctrl_info->block_requests = true;
}
static inline void pqi_ctrl_unblock_requests(struct pqi_ctrl_info *ctrl_info)
{
ctrl_info->block_requests = false;
wake_up_all(&ctrl_info->block_requests_wait);
}
static void pqi_wait_if_ctrl_blocked(struct pqi_ctrl_info *ctrl_info)
{
if (!pqi_ctrl_blocked(ctrl_info))
return;
atomic_inc(&ctrl_info->num_blocked_threads);
wait_event(ctrl_info->block_requests_wait,
!pqi_ctrl_blocked(ctrl_info));
atomic_dec(&ctrl_info->num_blocked_threads);
}
#define PQI_QUIESCE_WARNING_TIMEOUT_SECS 10
static inline void pqi_ctrl_wait_until_quiesced(struct pqi_ctrl_info *ctrl_info)
{
unsigned long start_jiffies;
unsigned long warning_timeout;
bool displayed_warning;
displayed_warning = false;
start_jiffies = jiffies;
warning_timeout = (PQI_QUIESCE_WARNING_TIMEOUT_SECS * HZ) + start_jiffies;
while (atomic_read(&ctrl_info->num_busy_threads) >
atomic_read(&ctrl_info->num_blocked_threads)) {
if (time_after(jiffies, warning_timeout)) {
dev_warn(&ctrl_info->pci_dev->dev,
"waiting %u seconds for driver activity to quiesce\n",
jiffies_to_msecs(jiffies - start_jiffies) / 1000);
displayed_warning = true;
warning_timeout = (PQI_QUIESCE_WARNING_TIMEOUT_SECS * HZ) + jiffies;
}
usleep_range(1000, 2000);
}
if (displayed_warning)
dev_warn(&ctrl_info->pci_dev->dev,
"driver activity quiesced after waiting for %u seconds\n",
jiffies_to_msecs(jiffies - start_jiffies) / 1000);
}
static inline bool pqi_device_offline(struct pqi_scsi_dev *device)
{
return device->device_offline;
}
static inline void pqi_ctrl_ofa_start(struct pqi_ctrl_info *ctrl_info)
{
mutex_lock(&ctrl_info->ofa_mutex);
}
static inline void pqi_ctrl_ofa_done(struct pqi_ctrl_info *ctrl_info)
{
mutex_unlock(&ctrl_info->ofa_mutex);
}
static inline void pqi_wait_until_ofa_finished(struct pqi_ctrl_info *ctrl_info)
{
mutex_lock(&ctrl_info->ofa_mutex);
mutex_unlock(&ctrl_info->ofa_mutex);
}
static inline bool pqi_ofa_in_progress(struct pqi_ctrl_info *ctrl_info)
{
return mutex_is_locked(&ctrl_info->ofa_mutex);
}
static inline void pqi_device_remove_start(struct pqi_scsi_dev *device)
{
device->in_remove = true;
}
static inline bool pqi_device_in_remove(struct pqi_scsi_dev *device)
{
return device->in_remove;
}
static inline void pqi_device_reset_start(struct pqi_scsi_dev *device, u8 lun)
{
device->in_reset[lun] = true;
}
static inline void pqi_device_reset_done(struct pqi_scsi_dev *device, u8 lun)
{
device->in_reset[lun] = false;
}
static inline bool pqi_device_in_reset(struct pqi_scsi_dev *device, u8 lun)
{
return device->in_reset[lun];
}
static inline int pqi_event_type_to_event_index(unsigned int event_type)
{
int index;
for (index = 0; index < ARRAY_SIZE(pqi_supported_event_types); index++)
if (event_type == pqi_supported_event_types[index])
return index;
return -1;
}
static inline bool pqi_is_supported_event(unsigned int event_type)
{
return pqi_event_type_to_event_index(event_type) != -1;
}
static inline void pqi_schedule_rescan_worker_with_delay(struct pqi_ctrl_info *ctrl_info,
unsigned long delay)
{
if (pqi_ctrl_offline(ctrl_info))
return;
schedule_delayed_work(&ctrl_info->rescan_work, delay);
}
static inline void pqi_schedule_rescan_worker(struct pqi_ctrl_info *ctrl_info)
{
pqi_schedule_rescan_worker_with_delay(ctrl_info, 0);
}
#define PQI_RESCAN_WORK_DELAY (10 * HZ)
static inline void pqi_schedule_rescan_worker_delayed(struct pqi_ctrl_info *ctrl_info)
{
pqi_schedule_rescan_worker_with_delay(ctrl_info, PQI_RESCAN_WORK_DELAY);
}
static inline void pqi_cancel_rescan_worker(struct pqi_ctrl_info *ctrl_info)
{
cancel_delayed_work_sync(&ctrl_info->rescan_work);
}
static inline u32 pqi_read_heartbeat_counter(struct pqi_ctrl_info *ctrl_info)
{
if (!ctrl_info->heartbeat_counter)
return 0;
return readl(ctrl_info->heartbeat_counter);
}
static inline u8 pqi_read_soft_reset_status(struct pqi_ctrl_info *ctrl_info)
{
return readb(ctrl_info->soft_reset_status);
}
static inline void pqi_clear_soft_reset_status(struct pqi_ctrl_info *ctrl_info)
{
u8 status;
status = pqi_read_soft_reset_status(ctrl_info);
status &= ~PQI_SOFT_RESET_ABORT;
writeb(status, ctrl_info->soft_reset_status);
}
static inline bool pqi_is_io_high_priority(struct pqi_scsi_dev *device, struct scsi_cmnd *scmd)
{
bool io_high_prio;
int priority_class;
io_high_prio = false;
if (device->ncq_prio_enable) {
priority_class =
IOPRIO_PRIO_CLASS(req_get_ioprio(scsi_cmd_to_rq(scmd)));
if (priority_class == IOPRIO_CLASS_RT) {
/* Set NCQ priority for read/write commands. */
switch (scmd->cmnd[0]) {
case WRITE_16:
case READ_16:
case WRITE_12:
case READ_12:
case WRITE_10:
case READ_10:
case WRITE_6:
case READ_6:
io_high_prio = true;
break;
}
}
}
return io_high_prio;
}
static int pqi_map_single(struct pci_dev *pci_dev,
struct pqi_sg_descriptor *sg_descriptor, void *buffer,
size_t buffer_length, enum dma_data_direction data_direction)
{
dma_addr_t bus_address;
if (!buffer || buffer_length == 0 || data_direction == DMA_NONE)
return 0;
bus_address = dma_map_single(&pci_dev->dev, buffer, buffer_length,
data_direction);
if (dma_mapping_error(&pci_dev->dev, bus_address))
return -ENOMEM;
put_unaligned_le64((u64)bus_address, &sg_descriptor->address);
put_unaligned_le32(buffer_length, &sg_descriptor->length);
put_unaligned_le32(CISS_SG_LAST, &sg_descriptor->flags);
return 0;
}
static void pqi_pci_unmap(struct pci_dev *pci_dev,
struct pqi_sg_descriptor *descriptors, int num_descriptors,
enum dma_data_direction data_direction)
{
int i;
if (data_direction == DMA_NONE)
return;
for (i = 0; i < num_descriptors; i++)
dma_unmap_single(&pci_dev->dev,
(dma_addr_t)get_unaligned_le64(&descriptors[i].address),
get_unaligned_le32(&descriptors[i].length),
data_direction);
}
static int pqi_build_raid_path_request(struct pqi_ctrl_info *ctrl_info,
struct pqi_raid_path_request *request, u8 cmd,
u8 *scsi3addr, void *buffer, size_t buffer_length,
u16 vpd_page, enum dma_data_direction *dir)
{
u8 *cdb;
size_t cdb_length = buffer_length;
memset(request, 0, sizeof(*request));
request->header.iu_type = PQI_REQUEST_IU_RAID_PATH_IO;
put_unaligned_le16(offsetof(struct pqi_raid_path_request,
sg_descriptors[1]) - PQI_REQUEST_HEADER_LENGTH,
&request->header.iu_length);
put_unaligned_le32(buffer_length, &request->buffer_length);
memcpy(request->lun_number, scsi3addr, sizeof(request->lun_number));
request->task_attribute = SOP_TASK_ATTRIBUTE_SIMPLE;
request->additional_cdb_bytes_usage = SOP_ADDITIONAL_CDB_BYTES_0;
cdb = request->cdb;
switch (cmd) {
case INQUIRY:
request->data_direction = SOP_READ_FLAG;
cdb[0] = INQUIRY;
if (vpd_page & VPD_PAGE) {
cdb[1] = 0x1;
cdb[2] = (u8)vpd_page;
}
cdb[4] = (u8)cdb_length;
break;
case CISS_REPORT_LOG:
case CISS_REPORT_PHYS:
request->data_direction = SOP_READ_FLAG;
cdb[0] = cmd;
if (cmd == CISS_REPORT_PHYS) {
if (ctrl_info->rpl_extended_format_4_5_supported)
cdb[1] = CISS_REPORT_PHYS_FLAG_EXTENDED_FORMAT_4;
else
cdb[1] = CISS_REPORT_PHYS_FLAG_EXTENDED_FORMAT_2;
} else {
cdb[1] = ctrl_info->ciss_report_log_flags;
}
put_unaligned_be32(cdb_length, &cdb[6]);
break;
case CISS_GET_RAID_MAP:
request->data_direction = SOP_READ_FLAG;
cdb[0] = CISS_READ;
cdb[1] = CISS_GET_RAID_MAP;
put_unaligned_be32(cdb_length, &cdb[6]);
break;
case SA_FLUSH_CACHE:
request->header.driver_flags = PQI_DRIVER_NONBLOCKABLE_REQUEST;
request->data_direction = SOP_WRITE_FLAG;
cdb[0] = BMIC_WRITE;
cdb[6] = BMIC_FLUSH_CACHE;
put_unaligned_be16(cdb_length, &cdb[7]);
break;
case BMIC_SENSE_DIAG_OPTIONS:
cdb_length = 0;
fallthrough;
case BMIC_IDENTIFY_CONTROLLER:
case BMIC_IDENTIFY_PHYSICAL_DEVICE:
case BMIC_SENSE_SUBSYSTEM_INFORMATION:
case BMIC_SENSE_FEATURE:
request->data_direction = SOP_READ_FLAG;
cdb[0] = BMIC_READ;
cdb[6] = cmd;
put_unaligned_be16(cdb_length, &cdb[7]);
break;
case BMIC_SET_DIAG_OPTIONS:
cdb_length = 0;
fallthrough;
case BMIC_WRITE_HOST_WELLNESS:
request->data_direction = SOP_WRITE_FLAG;
cdb[0] = BMIC_WRITE;
cdb[6] = cmd;
put_unaligned_be16(cdb_length, &cdb[7]);
break;
case BMIC_CSMI_PASSTHRU:
request->data_direction = SOP_BIDIRECTIONAL;
cdb[0] = BMIC_WRITE;
cdb[5] = CSMI_CC_SAS_SMP_PASSTHRU;
cdb[6] = cmd;
put_unaligned_be16(cdb_length, &cdb[7]);
break;
default:
dev_err(&ctrl_info->pci_dev->dev, "unknown command 0x%c\n", cmd);
break;
}
switch (request->data_direction) {
case SOP_READ_FLAG:
*dir = DMA_FROM_DEVICE;
break;
case SOP_WRITE_FLAG:
*dir = DMA_TO_DEVICE;
break;
case SOP_NO_DIRECTION_FLAG:
*dir = DMA_NONE;
break;
default:
*dir = DMA_BIDIRECTIONAL;
break;
}
return pqi_map_single(ctrl_info->pci_dev, &request->sg_descriptors[0],
buffer, buffer_length, *dir);
}
static inline void pqi_reinit_io_request(struct pqi_io_request *io_request)
{
io_request->scmd = NULL;
io_request->status = 0;
io_request->error_info = NULL;
io_request->raid_bypass = false;
}
static inline struct pqi_io_request *pqi_alloc_io_request(struct pqi_ctrl_info *ctrl_info, struct scsi_cmnd *scmd)
{
struct pqi_io_request *io_request;
u16 i;
if (scmd) { /* SML I/O request */
u32 blk_tag = blk_mq_unique_tag(scsi_cmd_to_rq(scmd));
i = blk_mq_unique_tag_to_tag(blk_tag);
io_request = &ctrl_info->io_request_pool[i];
if (atomic_inc_return(&io_request->refcount) > 1) {
atomic_dec(&io_request->refcount);
return NULL;
}
} else { /* IOCTL or driver internal request */
/*
* benignly racy - may have to wait for an open slot.
* command slot range is scsi_ml_can_queue -
* [scsi_ml_can_queue + (PQI_RESERVED_IO_SLOTS - 1)]
*/
i = 0;
while (1) {
io_request = &ctrl_info->io_request_pool[ctrl_info->scsi_ml_can_queue + i];
if (atomic_inc_return(&io_request->refcount) == 1)
break;
atomic_dec(&io_request->refcount);
i = (i + 1) % PQI_RESERVED_IO_SLOTS;
}
}
if (io_request)
pqi_reinit_io_request(io_request);
return io_request;
}
static void pqi_free_io_request(struct pqi_io_request *io_request)
{
atomic_dec(&io_request->refcount);
}
static int pqi_send_scsi_raid_request(struct pqi_ctrl_info *ctrl_info, u8 cmd,
u8 *scsi3addr, void *buffer, size_t buffer_length, u16 vpd_page,
struct pqi_raid_error_info *error_info)
{
int rc;
struct pqi_raid_path_request request;
enum dma_data_direction dir;
rc = pqi_build_raid_path_request(ctrl_info, &request, cmd, scsi3addr,
buffer, buffer_length, vpd_page, &dir);
if (rc)
return rc;
rc = pqi_submit_raid_request_synchronous(ctrl_info, &request.header, 0, error_info);
pqi_pci_unmap(ctrl_info->pci_dev, request.sg_descriptors, 1, dir);
return rc;
}
/* helper functions for pqi_send_scsi_raid_request */
static inline int pqi_send_ctrl_raid_request(struct pqi_ctrl_info *ctrl_info,
u8 cmd, void *buffer, size_t buffer_length)
{
return pqi_send_scsi_raid_request(ctrl_info, cmd, RAID_CTLR_LUNID,
buffer, buffer_length, 0, NULL);
}
static inline int pqi_send_ctrl_raid_with_error(struct pqi_ctrl_info *ctrl_info,
u8 cmd, void *buffer, size_t buffer_length,
struct pqi_raid_error_info *error_info)
{
return pqi_send_scsi_raid_request(ctrl_info, cmd, RAID_CTLR_LUNID,
buffer, buffer_length, 0, error_info);
}
static inline int pqi_identify_controller(struct pqi_ctrl_info *ctrl_info,
struct bmic_identify_controller *buffer)
{
return pqi_send_ctrl_raid_request(ctrl_info, BMIC_IDENTIFY_CONTROLLER,
buffer, sizeof(*buffer));
}
static inline int pqi_sense_subsystem_info(struct pqi_ctrl_info *ctrl_info,
struct bmic_sense_subsystem_info *sense_info)
{
return pqi_send_ctrl_raid_request(ctrl_info,
BMIC_SENSE_SUBSYSTEM_INFORMATION, sense_info,
sizeof(*sense_info));
}
static inline int pqi_scsi_inquiry(struct pqi_ctrl_info *ctrl_info,
u8 *scsi3addr, u16 vpd_page, void *buffer, size_t buffer_length)
{
return pqi_send_scsi_raid_request(ctrl_info, INQUIRY, scsi3addr,
buffer, buffer_length, vpd_page, NULL);
}
static int pqi_identify_physical_device(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device,
struct bmic_identify_physical_device *buffer, size_t buffer_length)
{
int rc;
enum dma_data_direction dir;
u16 bmic_device_index;
struct pqi_raid_path_request request;
rc = pqi_build_raid_path_request(ctrl_info, &request,
BMIC_IDENTIFY_PHYSICAL_DEVICE, RAID_CTLR_LUNID, buffer,
buffer_length, 0, &dir);
if (rc)
return rc;
bmic_device_index = CISS_GET_DRIVE_NUMBER(device->scsi3addr);
request.cdb[2] = (u8)bmic_device_index;
request.cdb[9] = (u8)(bmic_device_index >> 8);
rc = pqi_submit_raid_request_synchronous(ctrl_info, &request.header, 0, NULL);
pqi_pci_unmap(ctrl_info->pci_dev, request.sg_descriptors, 1, dir);
return rc;
}
static inline u32 pqi_aio_limit_to_bytes(__le16 *limit)
{
u32 bytes;
bytes = get_unaligned_le16(limit);
if (bytes == 0)
bytes = ~0;
else
bytes *= 1024;
return bytes;
}
#pragma pack(1)
struct bmic_sense_feature_buffer {
struct bmic_sense_feature_buffer_header header;
struct bmic_sense_feature_io_page_aio_subpage aio_subpage;
};
#pragma pack()
#define MINIMUM_AIO_SUBPAGE_BUFFER_LENGTH \
offsetofend(struct bmic_sense_feature_buffer, \
aio_subpage.max_write_raid_1_10_3drive)
#define MINIMUM_AIO_SUBPAGE_LENGTH \
(offsetofend(struct bmic_sense_feature_io_page_aio_subpage, \
max_write_raid_1_10_3drive) - \
sizeof_field(struct bmic_sense_feature_io_page_aio_subpage, header))
static int pqi_get_advanced_raid_bypass_config(struct pqi_ctrl_info *ctrl_info)
{
int rc;
enum dma_data_direction dir;
struct pqi_raid_path_request request;
struct bmic_sense_feature_buffer *buffer;
buffer = kmalloc(sizeof(*buffer), GFP_KERNEL);
if (!buffer)
return -ENOMEM;
rc = pqi_build_raid_path_request(ctrl_info, &request, BMIC_SENSE_FEATURE, RAID_CTLR_LUNID,
buffer, sizeof(*buffer), 0, &dir);
if (rc)
goto error;
request.cdb[2] = BMIC_SENSE_FEATURE_IO_PAGE;
request.cdb[3] = BMIC_SENSE_FEATURE_IO_PAGE_AIO_SUBPAGE;
rc = pqi_submit_raid_request_synchronous(ctrl_info, &request.header, 0, NULL);
pqi_pci_unmap(ctrl_info->pci_dev, request.sg_descriptors, 1, dir);
if (rc)
goto error;
if (buffer->header.page_code != BMIC_SENSE_FEATURE_IO_PAGE ||
buffer->header.subpage_code !=
BMIC_SENSE_FEATURE_IO_PAGE_AIO_SUBPAGE ||
get_unaligned_le16(&buffer->header.buffer_length) <
MINIMUM_AIO_SUBPAGE_BUFFER_LENGTH ||
buffer->aio_subpage.header.page_code !=
BMIC_SENSE_FEATURE_IO_PAGE ||
buffer->aio_subpage.header.subpage_code !=
BMIC_SENSE_FEATURE_IO_PAGE_AIO_SUBPAGE ||
get_unaligned_le16(&buffer->aio_subpage.header.page_length) <
MINIMUM_AIO_SUBPAGE_LENGTH) {
goto error;
}
ctrl_info->max_transfer_encrypted_sas_sata =
pqi_aio_limit_to_bytes(
&buffer->aio_subpage.max_transfer_encrypted_sas_sata);
ctrl_info->max_transfer_encrypted_nvme =
pqi_aio_limit_to_bytes(
&buffer->aio_subpage.max_transfer_encrypted_nvme);
ctrl_info->max_write_raid_5_6 =
pqi_aio_limit_to_bytes(
&buffer->aio_subpage.max_write_raid_5_6);
ctrl_info->max_write_raid_1_10_2drive =
pqi_aio_limit_to_bytes(
&buffer->aio_subpage.max_write_raid_1_10_2drive);
ctrl_info->max_write_raid_1_10_3drive =
pqi_aio_limit_to_bytes(
&buffer->aio_subpage.max_write_raid_1_10_3drive);
error:
kfree(buffer);
return rc;
}
static int pqi_flush_cache(struct pqi_ctrl_info *ctrl_info,
enum bmic_flush_cache_shutdown_event shutdown_event)
{
int rc;
struct bmic_flush_cache *flush_cache;
flush_cache = kzalloc(sizeof(*flush_cache), GFP_KERNEL);
if (!flush_cache)
return -ENOMEM;
flush_cache->shutdown_event = shutdown_event;
rc = pqi_send_ctrl_raid_request(ctrl_info, SA_FLUSH_CACHE, flush_cache,
sizeof(*flush_cache));
kfree(flush_cache);
return rc;
}
int pqi_csmi_smp_passthru(struct pqi_ctrl_info *ctrl_info,
struct bmic_csmi_smp_passthru_buffer *buffer, size_t buffer_length,
struct pqi_raid_error_info *error_info)
{
return pqi_send_ctrl_raid_with_error(ctrl_info, BMIC_CSMI_PASSTHRU,
buffer, buffer_length, error_info);
}
#define PQI_FETCH_PTRAID_DATA (1 << 31)
static int pqi_set_diag_rescan(struct pqi_ctrl_info *ctrl_info)
{
int rc;
struct bmic_diag_options *diag;
diag = kzalloc(sizeof(*diag), GFP_KERNEL);
if (!diag)
return -ENOMEM;
rc = pqi_send_ctrl_raid_request(ctrl_info, BMIC_SENSE_DIAG_OPTIONS,
diag, sizeof(*diag));
if (rc)
goto out;
diag->options |= cpu_to_le32(PQI_FETCH_PTRAID_DATA);
rc = pqi_send_ctrl_raid_request(ctrl_info, BMIC_SET_DIAG_OPTIONS, diag,
sizeof(*diag));
out:
kfree(diag);
return rc;
}
static inline int pqi_write_host_wellness(struct pqi_ctrl_info *ctrl_info,
void *buffer, size_t buffer_length)
{
return pqi_send_ctrl_raid_request(ctrl_info, BMIC_WRITE_HOST_WELLNESS,
buffer, buffer_length);
}
#pragma pack(1)
struct bmic_host_wellness_driver_version {
u8 start_tag[4];
u8 driver_version_tag[2];
__le16 driver_version_length;
char driver_version[32];
u8 dont_write_tag[2];
u8 end_tag[2];
};
#pragma pack()
static int pqi_write_driver_version_to_host_wellness(
struct pqi_ctrl_info *ctrl_info)
{
int rc;
struct bmic_host_wellness_driver_version *buffer;
size_t buffer_length;
buffer_length = sizeof(*buffer);
buffer = kmalloc(buffer_length, GFP_KERNEL);
if (!buffer)
return -ENOMEM;
buffer->start_tag[0] = '<';
buffer->start_tag[1] = 'H';
buffer->start_tag[2] = 'W';
buffer->start_tag[3] = '>';
buffer->driver_version_tag[0] = 'D';
buffer->driver_version_tag[1] = 'V';
put_unaligned_le16(sizeof(buffer->driver_version),
&buffer->driver_version_length);
strncpy(buffer->driver_version, "Linux " DRIVER_VERSION,
sizeof(buffer->driver_version) - 1);
buffer->driver_version[sizeof(buffer->driver_version) - 1] = '\0';
buffer->dont_write_tag[0] = 'D';
buffer->dont_write_tag[1] = 'W';
buffer->end_tag[0] = 'Z';
buffer->end_tag[1] = 'Z';
rc = pqi_write_host_wellness(ctrl_info, buffer, buffer_length);
kfree(buffer);
return rc;
}
#pragma pack(1)
struct bmic_host_wellness_time {
u8 start_tag[4];
u8 time_tag[2];
__le16 time_length;
u8 time[8];
u8 dont_write_tag[2];
u8 end_tag[2];
};
#pragma pack()
static int pqi_write_current_time_to_host_wellness(
struct pqi_ctrl_info *ctrl_info)
{
int rc;
struct bmic_host_wellness_time *buffer;
size_t buffer_length;
time64_t local_time;
unsigned int year;
struct tm tm;
buffer_length = sizeof(*buffer);
buffer = kmalloc(buffer_length, GFP_KERNEL);
if (!buffer)
return -ENOMEM;
buffer->start_tag[0] = '<';
buffer->start_tag[1] = 'H';
buffer->start_tag[2] = 'W';
buffer->start_tag[3] = '>';
buffer->time_tag[0] = 'T';
buffer->time_tag[1] = 'D';
put_unaligned_le16(sizeof(buffer->time),
&buffer->time_length);
local_time = ktime_get_real_seconds();
time64_to_tm(local_time, -sys_tz.tz_minuteswest * 60, &tm);
year = tm.tm_year + 1900;
buffer->time[0] = bin2bcd(tm.tm_hour);
buffer->time[1] = bin2bcd(tm.tm_min);
buffer->time[2] = bin2bcd(tm.tm_sec);
buffer->time[3] = 0;
buffer->time[4] = bin2bcd(tm.tm_mon + 1);
buffer->time[5] = bin2bcd(tm.tm_mday);
buffer->time[6] = bin2bcd(year / 100);
buffer->time[7] = bin2bcd(year % 100);
buffer->dont_write_tag[0] = 'D';
buffer->dont_write_tag[1] = 'W';
buffer->end_tag[0] = 'Z';
buffer->end_tag[1] = 'Z';
rc = pqi_write_host_wellness(ctrl_info, buffer, buffer_length);
kfree(buffer);
return rc;
}
#define PQI_UPDATE_TIME_WORK_INTERVAL (24UL * 60 * 60 * HZ)
static void pqi_update_time_worker(struct work_struct *work)
{
int rc;
struct pqi_ctrl_info *ctrl_info;
ctrl_info = container_of(to_delayed_work(work), struct pqi_ctrl_info,
update_time_work);
rc = pqi_write_current_time_to_host_wellness(ctrl_info);
if (rc)
dev_warn(&ctrl_info->pci_dev->dev,
"error updating time on controller\n");
schedule_delayed_work(&ctrl_info->update_time_work,
PQI_UPDATE_TIME_WORK_INTERVAL);
}
static inline void pqi_schedule_update_time_worker(struct pqi_ctrl_info *ctrl_info)
{
schedule_delayed_work(&ctrl_info->update_time_work, 0);
}
static inline void pqi_cancel_update_time_worker(struct pqi_ctrl_info *ctrl_info)
{
cancel_delayed_work_sync(&ctrl_info->update_time_work);
}
static inline int pqi_report_luns(struct pqi_ctrl_info *ctrl_info, u8 cmd, void *buffer,
size_t buffer_length)
{
return pqi_send_ctrl_raid_request(ctrl_info, cmd, buffer, buffer_length);
}
static int pqi_report_phys_logical_luns(struct pqi_ctrl_info *ctrl_info, u8 cmd, void **buffer)
{
int rc;
size_t lun_list_length;
size_t lun_data_length;
size_t new_lun_list_length;
void *lun_data = NULL;
struct report_lun_header *report_lun_header;
report_lun_header = kmalloc(sizeof(*report_lun_header), GFP_KERNEL);
if (!report_lun_header) {
rc = -ENOMEM;
goto out;
}
rc = pqi_report_luns(ctrl_info, cmd, report_lun_header, sizeof(*report_lun_header));
if (rc)
goto out;
lun_list_length = get_unaligned_be32(&report_lun_header->list_length);
again:
lun_data_length = sizeof(struct report_lun_header) + lun_list_length;
lun_data = kmalloc(lun_data_length, GFP_KERNEL);
if (!lun_data) {
rc = -ENOMEM;
goto out;
}
if (lun_list_length == 0) {
memcpy(lun_data, report_lun_header, sizeof(*report_lun_header));
goto out;
}
rc = pqi_report_luns(ctrl_info, cmd, lun_data, lun_data_length);
if (rc)
goto out;
new_lun_list_length =
get_unaligned_be32(&((struct report_lun_header *)lun_data)->list_length);
if (new_lun_list_length > lun_list_length) {
lun_list_length = new_lun_list_length;
kfree(lun_data);
goto again;
}
out:
kfree(report_lun_header);
if (rc) {
kfree(lun_data);
lun_data = NULL;
}
*buffer = lun_data;
return rc;
}
static inline int pqi_report_phys_luns(struct pqi_ctrl_info *ctrl_info, void **buffer)
{
int rc;
unsigned int i;
u8 rpl_response_format;
u32 num_physicals;
void *rpl_list;
struct report_lun_header *rpl_header;
struct report_phys_lun_8byte_wwid_list *rpl_8byte_wwid_list;
struct report_phys_lun_16byte_wwid_list *rpl_16byte_wwid_list;
rc = pqi_report_phys_logical_luns(ctrl_info, CISS_REPORT_PHYS, &rpl_list);
if (rc)
return rc;
if (ctrl_info->rpl_extended_format_4_5_supported) {
rpl_header = rpl_list;
rpl_response_format = rpl_header->flags & CISS_REPORT_PHYS_FLAG_EXTENDED_FORMAT_MASK;
if (rpl_response_format == CISS_REPORT_PHYS_FLAG_EXTENDED_FORMAT_4) {
*buffer = rpl_list;
return 0;
} else if (rpl_response_format != CISS_REPORT_PHYS_FLAG_EXTENDED_FORMAT_2) {
dev_err(&ctrl_info->pci_dev->dev,
"RPL returned unsupported data format %u\n",
rpl_response_format);
return -EINVAL;
} else {
dev_warn(&ctrl_info->pci_dev->dev,
"RPL returned extended format 2 instead of 4\n");
}
}
rpl_8byte_wwid_list = rpl_list;
num_physicals = get_unaligned_be32(&rpl_8byte_wwid_list->header.list_length) / sizeof(rpl_8byte_wwid_list->lun_entries[0]);
rpl_16byte_wwid_list = kmalloc(struct_size(rpl_16byte_wwid_list, lun_entries,
num_physicals), GFP_KERNEL);
if (!rpl_16byte_wwid_list)
return -ENOMEM;
put_unaligned_be32(num_physicals * sizeof(struct report_phys_lun_16byte_wwid),
&rpl_16byte_wwid_list->header.list_length);
rpl_16byte_wwid_list->header.flags = rpl_8byte_wwid_list->header.flags;
for (i = 0; i < num_physicals; i++) {
memcpy(&rpl_16byte_wwid_list->lun_entries[i].lunid, &rpl_8byte_wwid_list->lun_entries[i].lunid, sizeof(rpl_8byte_wwid_list->lun_entries[i].lunid));
memcpy(&rpl_16byte_wwid_list->lun_entries[i].wwid[0], &rpl_8byte_wwid_list->lun_entries[i].wwid, sizeof(rpl_8byte_wwid_list->lun_entries[i].wwid));
memset(&rpl_16byte_wwid_list->lun_entries[i].wwid[8], 0, 8);
rpl_16byte_wwid_list->lun_entries[i].device_type = rpl_8byte_wwid_list->lun_entries[i].device_type;
rpl_16byte_wwid_list->lun_entries[i].device_flags = rpl_8byte_wwid_list->lun_entries[i].device_flags;
rpl_16byte_wwid_list->lun_entries[i].lun_count = rpl_8byte_wwid_list->lun_entries[i].lun_count;
rpl_16byte_wwid_list->lun_entries[i].redundant_paths = rpl_8byte_wwid_list->lun_entries[i].redundant_paths;
rpl_16byte_wwid_list->lun_entries[i].aio_handle = rpl_8byte_wwid_list->lun_entries[i].aio_handle;
}
kfree(rpl_8byte_wwid_list);
*buffer = rpl_16byte_wwid_list;
return 0;
}
static inline int pqi_report_logical_luns(struct pqi_ctrl_info *ctrl_info, void **buffer)
{
return pqi_report_phys_logical_luns(ctrl_info, CISS_REPORT_LOG, buffer);
}
static int pqi_get_device_lists(struct pqi_ctrl_info *ctrl_info,
struct report_phys_lun_16byte_wwid_list **physdev_list,
struct report_log_lun_list **logdev_list)
{
int rc;
size_t logdev_list_length;
size_t logdev_data_length;
struct report_log_lun_list *internal_logdev_list;
struct report_log_lun_list *logdev_data;
struct report_lun_header report_lun_header;
rc = pqi_report_phys_luns(ctrl_info, (void **)physdev_list);
if (rc)
dev_err(&ctrl_info->pci_dev->dev,
"report physical LUNs failed\n");
rc = pqi_report_logical_luns(ctrl_info, (void **)logdev_list);
if (rc)
dev_err(&ctrl_info->pci_dev->dev,
"report logical LUNs failed\n");
/*
* Tack the controller itself onto the end of the logical device list
* by adding a list entry that is all zeros.
*/
logdev_data = *logdev_list;
if (logdev_data) {
logdev_list_length =
get_unaligned_be32(&logdev_data->header.list_length);
} else {
memset(&report_lun_header, 0, sizeof(report_lun_header));
logdev_data =
(struct report_log_lun_list *)&report_lun_header;
logdev_list_length = 0;
}
logdev_data_length = sizeof(struct report_lun_header) +
logdev_list_length;
internal_logdev_list = kmalloc(logdev_data_length +
sizeof(struct report_log_lun), GFP_KERNEL);
if (!internal_logdev_list) {
kfree(*logdev_list);
*logdev_list = NULL;
return -ENOMEM;
}
memcpy(internal_logdev_list, logdev_data, logdev_data_length);
memset((u8 *)internal_logdev_list + logdev_data_length, 0,
sizeof(struct report_log_lun));
put_unaligned_be32(logdev_list_length +
sizeof(struct report_log_lun),
&internal_logdev_list->header.list_length);
kfree(*logdev_list);
*logdev_list = internal_logdev_list;
return 0;
}
static inline void pqi_set_bus_target_lun(struct pqi_scsi_dev *device,
int bus, int target, int lun)
{
device->bus = bus;
device->target = target;
device->lun = lun;
}
static void pqi_assign_bus_target_lun(struct pqi_scsi_dev *device)
{
u8 *scsi3addr;
u32 lunid;
int bus;
int target;
int lun;
scsi3addr = device->scsi3addr;
lunid = get_unaligned_le32(scsi3addr);
if (pqi_is_hba_lunid(scsi3addr)) {
/* The specified device is the controller. */
pqi_set_bus_target_lun(device, PQI_HBA_BUS, 0, lunid & 0x3fff);
device->target_lun_valid = true;
return;
}
if (pqi_is_logical_device(device)) {
if (device->is_external_raid_device) {
bus = PQI_EXTERNAL_RAID_VOLUME_BUS;
target = (lunid >> 16) & 0x3fff;
lun = lunid & 0xff;
} else {
bus = PQI_RAID_VOLUME_BUS;
target = 0;
lun = lunid & 0x3fff;
}
pqi_set_bus_target_lun(device, bus, target, lun);
device->target_lun_valid = true;
return;
}
/*
* Defer target and LUN assignment for non-controller physical devices
* because the SAS transport layer will make these assignments later.
*/
pqi_set_bus_target_lun(device, PQI_PHYSICAL_DEVICE_BUS, 0, 0);
}
static void pqi_get_raid_level(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
u8 raid_level;
u8 *buffer;
raid_level = SA_RAID_UNKNOWN;
buffer = kmalloc(64, GFP_KERNEL);
if (buffer) {
rc = pqi_scsi_inquiry(ctrl_info, device->scsi3addr,
VPD_PAGE | CISS_VPD_LV_DEVICE_GEOMETRY, buffer, 64);
if (rc == 0) {
raid_level = buffer[8];
if (raid_level > SA_RAID_MAX)
raid_level = SA_RAID_UNKNOWN;
}
kfree(buffer);
}
device->raid_level = raid_level;
}
static int pqi_validate_raid_map(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device, struct raid_map *raid_map)
{
char *err_msg;
u32 raid_map_size;
u32 r5or6_blocks_per_row;
raid_map_size = get_unaligned_le32(&raid_map->structure_size);
if (raid_map_size < offsetof(struct raid_map, disk_data)) {
err_msg = "RAID map too small";
goto bad_raid_map;
}
if (device->raid_level == SA_RAID_1) {
if (get_unaligned_le16(&raid_map->layout_map_count) != 2) {
err_msg = "invalid RAID-1 map";
goto bad_raid_map;
}
} else if (device->raid_level == SA_RAID_TRIPLE) {
if (get_unaligned_le16(&raid_map->layout_map_count) != 3) {
err_msg = "invalid RAID-1(Triple) map";
goto bad_raid_map;
}
} else if ((device->raid_level == SA_RAID_5 ||
device->raid_level == SA_RAID_6) &&
get_unaligned_le16(&raid_map->layout_map_count) > 1) {
/* RAID 50/60 */
r5or6_blocks_per_row =
get_unaligned_le16(&raid_map->strip_size) *
get_unaligned_le16(&raid_map->data_disks_per_row);
if (r5or6_blocks_per_row == 0) {
err_msg = "invalid RAID-5 or RAID-6 map";
goto bad_raid_map;
}
}
return 0;
bad_raid_map:
dev_warn(&ctrl_info->pci_dev->dev,
"logical device %08x%08x %s\n",
*((u32 *)&device->scsi3addr),
*((u32 *)&device->scsi3addr[4]), err_msg);
return -EINVAL;
}
static int pqi_get_raid_map(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
u32 raid_map_size;
struct raid_map *raid_map;
raid_map = kmalloc(sizeof(*raid_map), GFP_KERNEL);
if (!raid_map)
return -ENOMEM;
rc = pqi_send_scsi_raid_request(ctrl_info, CISS_GET_RAID_MAP,
device->scsi3addr, raid_map, sizeof(*raid_map), 0, NULL);
if (rc)
goto error;
raid_map_size = get_unaligned_le32(&raid_map->structure_size);
if (raid_map_size > sizeof(*raid_map)) {
kfree(raid_map);
raid_map = kmalloc(raid_map_size, GFP_KERNEL);
if (!raid_map)
return -ENOMEM;
rc = pqi_send_scsi_raid_request(ctrl_info, CISS_GET_RAID_MAP,
device->scsi3addr, raid_map, raid_map_size, 0, NULL);
if (rc)
goto error;
if (get_unaligned_le32(&raid_map->structure_size)
!= raid_map_size) {
dev_warn(&ctrl_info->pci_dev->dev,
"requested %u bytes, received %u bytes\n",
raid_map_size,
get_unaligned_le32(&raid_map->structure_size));
rc = -EINVAL;
goto error;
}
}
rc = pqi_validate_raid_map(ctrl_info, device, raid_map);
if (rc)
goto error;
device->raid_map = raid_map;
return 0;
error:
kfree(raid_map);
return rc;
}
static void pqi_set_max_transfer_encrypted(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
if (!ctrl_info->lv_drive_type_mix_valid) {
device->max_transfer_encrypted = ~0;
return;
}
switch (LV_GET_DRIVE_TYPE_MIX(device->scsi3addr)) {
case LV_DRIVE_TYPE_MIX_SAS_HDD_ONLY:
case LV_DRIVE_TYPE_MIX_SATA_HDD_ONLY:
case LV_DRIVE_TYPE_MIX_SAS_OR_SATA_SSD_ONLY:
case LV_DRIVE_TYPE_MIX_SAS_SSD_ONLY:
case LV_DRIVE_TYPE_MIX_SATA_SSD_ONLY:
case LV_DRIVE_TYPE_MIX_SAS_ONLY:
case LV_DRIVE_TYPE_MIX_SATA_ONLY:
device->max_transfer_encrypted =
ctrl_info->max_transfer_encrypted_sas_sata;
break;
case LV_DRIVE_TYPE_MIX_NVME_ONLY:
device->max_transfer_encrypted =
ctrl_info->max_transfer_encrypted_nvme;
break;
case LV_DRIVE_TYPE_MIX_UNKNOWN:
case LV_DRIVE_TYPE_MIX_NO_RESTRICTION:
default:
device->max_transfer_encrypted =
min(ctrl_info->max_transfer_encrypted_sas_sata,
ctrl_info->max_transfer_encrypted_nvme);
break;
}
}
static void pqi_get_raid_bypass_status(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
u8 *buffer;
u8 bypass_status;
buffer = kmalloc(64, GFP_KERNEL);
if (!buffer)
return;
rc = pqi_scsi_inquiry(ctrl_info, device->scsi3addr,
VPD_PAGE | CISS_VPD_LV_BYPASS_STATUS, buffer, 64);
if (rc)
goto out;
#define RAID_BYPASS_STATUS 4
#define RAID_BYPASS_CONFIGURED 0x1
#define RAID_BYPASS_ENABLED 0x2
bypass_status = buffer[RAID_BYPASS_STATUS];
device->raid_bypass_configured =
(bypass_status & RAID_BYPASS_CONFIGURED) != 0;
if (device->raid_bypass_configured &&
(bypass_status & RAID_BYPASS_ENABLED) &&
pqi_get_raid_map(ctrl_info, device) == 0) {
device->raid_bypass_enabled = true;
if (get_unaligned_le16(&device->raid_map->flags) &
RAID_MAP_ENCRYPTION_ENABLED)
pqi_set_max_transfer_encrypted(ctrl_info, device);
}
out:
kfree(buffer);
}
/*
* Use vendor-specific VPD to determine online/offline status of a volume.
*/
static void pqi_get_volume_status(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
size_t page_length;
u8 volume_status = CISS_LV_STATUS_UNAVAILABLE;
bool volume_offline = true;
u32 volume_flags;
struct ciss_vpd_logical_volume_status *vpd;
vpd = kmalloc(sizeof(*vpd), GFP_KERNEL);
if (!vpd)
goto no_buffer;
rc = pqi_scsi_inquiry(ctrl_info, device->scsi3addr,
VPD_PAGE | CISS_VPD_LV_STATUS, vpd, sizeof(*vpd));
if (rc)
goto out;
if (vpd->page_code != CISS_VPD_LV_STATUS)
goto out;
page_length = offsetof(struct ciss_vpd_logical_volume_status,
volume_status) + vpd->page_length;
if (page_length < sizeof(*vpd))
goto out;
volume_status = vpd->volume_status;
volume_flags = get_unaligned_be32(&vpd->flags);
volume_offline = (volume_flags & CISS_LV_FLAGS_NO_HOST_IO) != 0;
out:
kfree(vpd);
no_buffer:
device->volume_status = volume_status;
device->volume_offline = volume_offline;
}
#define PQI_DEVICE_NCQ_PRIO_SUPPORTED 0x01
#define PQI_DEVICE_PHY_MAP_SUPPORTED 0x10
#define PQI_DEVICE_ERASE_IN_PROGRESS 0x10
static int pqi_get_physical_device_info(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device,
struct bmic_identify_physical_device *id_phys)
{
int rc;
memset(id_phys, 0, sizeof(*id_phys));
rc = pqi_identify_physical_device(ctrl_info, device,
id_phys, sizeof(*id_phys));
if (rc) {
device->queue_depth = PQI_PHYSICAL_DISK_DEFAULT_MAX_QUEUE_DEPTH;
return rc;
}
scsi_sanitize_inquiry_string(&id_phys->model[0], 8);
scsi_sanitize_inquiry_string(&id_phys->model[8], 16);
memcpy(device->vendor, &id_phys->model[0], sizeof(device->vendor));
memcpy(device->model, &id_phys->model[8], sizeof(device->model));
device->box_index = id_phys->box_index;
device->phys_box_on_bus = id_phys->phys_box_on_bus;
device->phy_connected_dev_type = id_phys->phy_connected_dev_type[0];
device->queue_depth =
get_unaligned_le16(&id_phys->current_queue_depth_limit);
device->active_path_index = id_phys->active_path_number;
device->path_map = id_phys->redundant_path_present_map;
memcpy(&device->box,
&id_phys->alternate_paths_phys_box_on_port,
sizeof(device->box));
memcpy(&device->phys_connector,
&id_phys->alternate_paths_phys_connector,
sizeof(device->phys_connector));
device->bay = id_phys->phys_bay_in_box;
device->lun_count = id_phys->multi_lun_device_lun_count;
if ((id_phys->even_more_flags & PQI_DEVICE_PHY_MAP_SUPPORTED) &&
id_phys->phy_count)
device->phy_id =
id_phys->phy_to_phy_map[device->active_path_index];
else
device->phy_id = 0xFF;
device->ncq_prio_support =
((get_unaligned_le32(&id_phys->misc_drive_flags) >> 16) &
PQI_DEVICE_NCQ_PRIO_SUPPORTED);
device->erase_in_progress = !!(get_unaligned_le16(&id_phys->extra_physical_drive_flags) & PQI_DEVICE_ERASE_IN_PROGRESS);
return 0;
}
static int pqi_get_logical_device_info(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
u8 *buffer;
buffer = kmalloc(64, GFP_KERNEL);
if (!buffer)
return -ENOMEM;
/* Send an inquiry to the device to see what it is. */
rc = pqi_scsi_inquiry(ctrl_info, device->scsi3addr, 0, buffer, 64);
if (rc)
goto out;
scsi_sanitize_inquiry_string(&buffer[8], 8);
scsi_sanitize_inquiry_string(&buffer[16], 16);
device->devtype = buffer[0] & 0x1f;
memcpy(device->vendor, &buffer[8], sizeof(device->vendor));
memcpy(device->model, &buffer[16], sizeof(device->model));
if (device->devtype == TYPE_DISK) {
if (device->is_external_raid_device) {
device->raid_level = SA_RAID_UNKNOWN;
device->volume_status = CISS_LV_OK;
device->volume_offline = false;
} else {
pqi_get_raid_level(ctrl_info, device);
pqi_get_raid_bypass_status(ctrl_info, device);
pqi_get_volume_status(ctrl_info, device);
}
}
out:
kfree(buffer);
return rc;
}
/*
* Prevent adding drive to OS for some corner cases such as a drive
* undergoing a sanitize (erase) operation. Some OSes will continue to poll
* the drive until the sanitize completes, which can take hours,
* resulting in long bootup delays. Commands such as TUR, READ_CAP
* are allowed, but READ/WRITE cause check condition. So the OS
* cannot check/read the partition table.
* Note: devices that have completed sanitize must be re-enabled
* using the management utility.
*/
static inline bool pqi_keep_device_offline(struct pqi_scsi_dev *device)
{
return device->erase_in_progress;
}
static int pqi_get_device_info_phys_logical(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device,
struct bmic_identify_physical_device *id_phys)
{
int rc;
if (device->is_expander_smp_device)
return 0;
if (pqi_is_logical_device(device))
rc = pqi_get_logical_device_info(ctrl_info, device);
else
rc = pqi_get_physical_device_info(ctrl_info, device, id_phys);
return rc;
}
static int pqi_get_device_info(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device,
struct bmic_identify_physical_device *id_phys)
{
int rc;
rc = pqi_get_device_info_phys_logical(ctrl_info, device, id_phys);
if (rc == 0 && device->lun_count == 0)
device->lun_count = 1;
return rc;
}
static void pqi_show_volume_status(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
char *status;
static const char unknown_state_str[] =
"Volume is in an unknown state (%u)";
char unknown_state_buffer[sizeof(unknown_state_str) + 10];
switch (device->volume_status) {
case CISS_LV_OK:
status = "Volume online";
break;
case CISS_LV_FAILED:
status = "Volume failed";
break;
case CISS_LV_NOT_CONFIGURED:
status = "Volume not configured";
break;
case CISS_LV_DEGRADED:
status = "Volume degraded";
break;
case CISS_LV_READY_FOR_RECOVERY:
status = "Volume ready for recovery operation";
break;
case CISS_LV_UNDERGOING_RECOVERY:
status = "Volume undergoing recovery";
break;
case CISS_LV_WRONG_PHYSICAL_DRIVE_REPLACED:
status = "Wrong physical drive was replaced";
break;
case CISS_LV_PHYSICAL_DRIVE_CONNECTION_PROBLEM:
status = "A physical drive not properly connected";
break;
case CISS_LV_HARDWARE_OVERHEATING:
status = "Hardware is overheating";
break;
case CISS_LV_HARDWARE_HAS_OVERHEATED:
status = "Hardware has overheated";
break;
case CISS_LV_UNDERGOING_EXPANSION:
status = "Volume undergoing expansion";
break;
case CISS_LV_NOT_AVAILABLE:
status = "Volume waiting for transforming volume";
break;
case CISS_LV_QUEUED_FOR_EXPANSION:
status = "Volume queued for expansion";
break;
case CISS_LV_DISABLED_SCSI_ID_CONFLICT:
status = "Volume disabled due to SCSI ID conflict";
break;
case CISS_LV_EJECTED:
status = "Volume has been ejected";
break;
case CISS_LV_UNDERGOING_ERASE:
status = "Volume undergoing background erase";
break;
case CISS_LV_READY_FOR_PREDICTIVE_SPARE_REBUILD:
status = "Volume ready for predictive spare rebuild";
break;
case CISS_LV_UNDERGOING_RPI:
status = "Volume undergoing rapid parity initialization";
break;
case CISS_LV_PENDING_RPI:
status = "Volume queued for rapid parity initialization";
break;
case CISS_LV_ENCRYPTED_NO_KEY:
status = "Encrypted volume inaccessible - key not present";
break;
case CISS_LV_UNDERGOING_ENCRYPTION:
status = "Volume undergoing encryption process";
break;
case CISS_LV_UNDERGOING_ENCRYPTION_REKEYING:
status = "Volume undergoing encryption re-keying process";
break;
case CISS_LV_ENCRYPTED_IN_NON_ENCRYPTED_CONTROLLER:
status = "Volume encrypted but encryption is disabled";
break;
case CISS_LV_PENDING_ENCRYPTION:
status = "Volume pending migration to encrypted state";
break;
case CISS_LV_PENDING_ENCRYPTION_REKEYING:
status = "Volume pending encryption rekeying";
break;
case CISS_LV_NOT_SUPPORTED:
status = "Volume not supported on this controller";
break;
case CISS_LV_STATUS_UNAVAILABLE:
status = "Volume status not available";
break;
default:
snprintf(unknown_state_buffer, sizeof(unknown_state_buffer),
unknown_state_str, device->volume_status);
status = unknown_state_buffer;
break;
}
dev_info(&ctrl_info->pci_dev->dev,
"scsi %d:%d:%d:%d %s\n",
ctrl_info->scsi_host->host_no,
device->bus, device->target, device->lun, status);
}
static void pqi_rescan_worker(struct work_struct *work)
{
struct pqi_ctrl_info *ctrl_info;
ctrl_info = container_of(to_delayed_work(work), struct pqi_ctrl_info,
rescan_work);
pqi_scan_scsi_devices(ctrl_info);
}
static int pqi_add_device(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
int rc;
if (pqi_is_logical_device(device))
rc = scsi_add_device(ctrl_info->scsi_host, device->bus,
device->target, device->lun);
else
rc = pqi_add_sas_device(ctrl_info->sas_host, device);
return rc;
}
#define PQI_REMOVE_DEVICE_PENDING_IO_TIMEOUT_MSECS (20 * 1000)
static inline void pqi_remove_device(struct pqi_ctrl_info *ctrl_info, struct pqi_scsi_dev *device)
{
int rc;
int lun;
for (lun = 0; lun < device->lun_count; lun++) {
rc = pqi_device_wait_for_pending_io(ctrl_info, device, lun,
PQI_REMOVE_DEVICE_PENDING_IO_TIMEOUT_MSECS);
if (rc)
dev_err(&ctrl_info->pci_dev->dev,
"scsi %d:%d:%d:%d removing device with %d outstanding command(s)\n",
ctrl_info->scsi_host->host_no, device->bus,
device->target, lun,
atomic_read(&device->scsi_cmds_outstanding[lun]));
}
if (pqi_is_logical_device(device))
scsi_remove_device(device->sdev);
else
pqi_remove_sas_device(device);
pqi_device_remove_start(device);
}
/* Assumes the SCSI device list lock is held. */
static struct pqi_scsi_dev *pqi_find_scsi_dev(struct pqi_ctrl_info *ctrl_info,
int bus, int target, int lun)
{
struct pqi_scsi_dev *device;
list_for_each_entry(device, &ctrl_info->scsi_device_list, scsi_device_list_entry)
if (device->bus == bus && device->target == target && device->lun == lun)
return device;
return NULL;
}
static inline bool pqi_device_equal(struct pqi_scsi_dev *dev1, struct pqi_scsi_dev *dev2)
{
if (dev1->is_physical_device != dev2->is_physical_device)
return false;
if (dev1->is_physical_device)
return memcmp(dev1->wwid, dev2->wwid, sizeof(dev1->wwid)) == 0;
return memcmp(dev1->volume_id, dev2->volume_id, sizeof(dev1->volume_id)) == 0;
}
enum pqi_find_result {
DEVICE_NOT_FOUND,
DEVICE_CHANGED,
DEVICE_SAME,
};
static enum pqi_find_result pqi_scsi_find_entry(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device_to_find, struct pqi_scsi_dev **matching_device)
{
struct pqi_scsi_dev *device;
list_for_each_entry(device, &ctrl_info->scsi_device_list, scsi_device_list_entry) {
if (pqi_scsi3addr_equal(device_to_find->scsi3addr, device->scsi3addr)) {
*matching_device = device;
if (pqi_device_equal(device_to_find, device)) {
if (device_to_find->volume_offline)
return DEVICE_CHANGED;
return DEVICE_SAME;
}
return DEVICE_CHANGED;
}
}
return DEVICE_NOT_FOUND;
}
static inline const char *pqi_device_type(struct pqi_scsi_dev *device)
{
if (device->is_expander_smp_device)
return "Enclosure SMP ";
return scsi_device_type(device->devtype);
}
#define PQI_DEV_INFO_BUFFER_LENGTH 128
static void pqi_dev_info(struct pqi_ctrl_info *ctrl_info,
char *action, struct pqi_scsi_dev *device)
{
ssize_t count;
char buffer[PQI_DEV_INFO_BUFFER_LENGTH];
count = scnprintf(buffer, PQI_DEV_INFO_BUFFER_LENGTH,
"%d:%d:", ctrl_info->scsi_host->host_no, device->bus);
if (device->target_lun_valid)
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
"%d:%d",
device->target,
device->lun);
else
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
"-:-");
if (pqi_is_logical_device(device))
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
" %08x%08x",
*((u32 *)&device->scsi3addr),
*((u32 *)&device->scsi3addr[4]));
else
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
" %016llx%016llx",
get_unaligned_be64(&device->wwid[0]),
get_unaligned_be64(&device->wwid[8]));
count += scnprintf(buffer + count, PQI_DEV_INFO_BUFFER_LENGTH - count,
" %s %.8s %.16s ",
pqi_device_type(device),
device->vendor,
device->model);
if (pqi_is_logical_device(device)) {
if (device->devtype == TYPE_DISK)
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
"SSDSmartPathCap%c En%c %-12s",
device->raid_bypass_configured ? '+' : '-',
device->raid_bypass_enabled ? '+' : '-',
pqi_raid_level_to_string(device->raid_level));
} else {
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
"AIO%c", device->aio_enabled ? '+' : '-');
if (device->devtype == TYPE_DISK ||
device->devtype == TYPE_ZBC)
count += scnprintf(buffer + count,
PQI_DEV_INFO_BUFFER_LENGTH - count,
" qd=%-6d", device->queue_depth);
}
dev_info(&ctrl_info->pci_dev->dev, "%s %s\n", action, buffer);
}
static bool pqi_raid_maps_equal(struct raid_map *raid_map1, struct raid_map *raid_map2)
{
u32 raid_map1_size;
u32 raid_map2_size;
if (raid_map1 == NULL || raid_map2 == NULL)
return raid_map1 == raid_map2;
raid_map1_size = get_unaligned_le32(&raid_map1->structure_size);
raid_map2_size = get_unaligned_le32(&raid_map2->structure_size);
if (raid_map1_size != raid_map2_size)
return false;
return memcmp(raid_map1, raid_map2, raid_map1_size) == 0;
}
/* Assumes the SCSI device list lock is held. */
static void pqi_scsi_update_device(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *existing_device, struct pqi_scsi_dev *new_device)
{
existing_device->device_type = new_device->device_type;
existing_device->bus = new_device->bus;
if (new_device->target_lun_valid) {
existing_device->target = new_device->target;
existing_device->lun = new_device->lun;
existing_device->target_lun_valid = true;
}
/* By definition, the scsi3addr and wwid fields are already the same. */
existing_device->is_physical_device = new_device->is_physical_device;
memcpy(existing_device->vendor, new_device->vendor, sizeof(existing_device->vendor));
memcpy(existing_device->model, new_device->model, sizeof(existing_device->model));
existing_device->sas_address = new_device->sas_address;
existing_device->queue_depth = new_device->queue_depth;
existing_device->device_offline = false;
existing_device->lun_count = new_device->lun_count;
if (pqi_is_logical_device(existing_device)) {
existing_device->is_external_raid_device = new_device->is_external_raid_device;
if (existing_device->devtype == TYPE_DISK) {
existing_device->raid_level = new_device->raid_level;
existing_device->volume_status = new_device->volume_status;
memset(existing_device->next_bypass_group, 0, sizeof(existing_device->next_bypass_group));
if (!pqi_raid_maps_equal(existing_device->raid_map, new_device->raid_map)) {
kfree(existing_device->raid_map);
existing_device->raid_map = new_device->raid_map;
/* To prevent this from being freed later. */
new_device->raid_map = NULL;
}
existing_device->raid_bypass_configured = new_device->raid_bypass_configured;
existing_device->raid_bypass_enabled = new_device->raid_bypass_enabled;
}
} else {
existing_device->aio_enabled = new_device->aio_enabled;
existing_device->aio_handle = new_device->aio_handle;
existing_device->is_expander_smp_device = new_device->is_expander_smp_device;
existing_device->active_path_index = new_device->active_path_index;
existing_device->phy_id = new_device->phy_id;
existing_device->path_map = new_device->path_map;
existing_device->bay = new_device->bay;
existing_device->box_index = new_device->box_index;
existing_device->phys_box_on_bus = new_device->phys_box_on_bus;
existing_device->phy_connected_dev_type = new_device->phy_connected_dev_type;
memcpy(existing_device->box, new_device->box, sizeof(existing_device->box));
memcpy(existing_device->phys_connector, new_device->phys_connector, sizeof(existing_device->phys_connector));
}
}
static inline void pqi_free_device(struct pqi_scsi_dev *device)
{
if (device) {
kfree(device->raid_map);
kfree(device);
}
}
/*
* Called when exposing a new device to the OS fails in order to re-adjust
* our internal SCSI device list to match the SCSI ML's view.
*/
static inline void pqi_fixup_botched_add(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device)
{
unsigned long flags;
spin_lock_irqsave(&ctrl_info->scsi_device_list_lock, flags);
list_del(&device->scsi_device_list_entry);
spin_unlock_irqrestore(&ctrl_info->scsi_device_list_lock, flags);
/* Allow the device structure to be freed later. */
device->keep_device = false;
}
static inline bool pqi_is_device_added(struct pqi_scsi_dev *device)
{
if (device->is_expander_smp_device)
return device->sas_port != NULL;
return device->sdev != NULL;
}
static inline void pqi_init_device_tmf_work(struct pqi_scsi_dev *device)
{
unsigned int lun;
struct pqi_tmf_work *tmf_work;
for (lun = 0, tmf_work = device->tmf_work; lun < PQI_MAX_LUNS_PER_DEVICE; lun++, tmf_work++)
INIT_WORK(&tmf_work->work_struct, pqi_tmf_worker);
}
static inline bool pqi_volume_rescan_needed(struct pqi_scsi_dev *device)
{
if (pqi_device_in_remove(device))
return false;
if (device->sdev == NULL)
return false;
if (!scsi_device_online(device->sdev))
return false;
return device->rescan;
}
static void pqi_update_device_list(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *new_device_list[], unsigned int num_new_devices)
{
int rc;
unsigned int i;
unsigned long flags;
enum pqi_find_result find_result;
struct pqi_scsi_dev *device;
struct pqi_scsi_dev *next;
struct pqi_scsi_dev *matching_device;
LIST_HEAD(add_list);
LIST_HEAD(delete_list);
/*
* The idea here is to do as little work as possible while holding the
* spinlock. That's why we go to great pains to defer anything other
* than updating the internal device list until after we release the
* spinlock.
*/
spin_lock_irqsave(&ctrl_info->scsi_device_list_lock, flags);
/* Assume that all devices in the existing list have gone away. */
list_for_each_entry(device, &ctrl_info->scsi_device_list, scsi_device_list_entry)
device->device_gone = true;
for (i = 0; i < num_new_devices; i++) {
device = new_device_list[i];
find_result = pqi_scsi_find_entry(ctrl_info, device,
&matching_device);
switch (find_result) {
case DEVICE_SAME:
/*
* The newly found device is already in the existing
* device list.
*/
device->new_device = false;
matching_device->device_gone = false;
pqi_scsi_update_device(ctrl_info, matching_device, device);
break;
case DEVICE_NOT_FOUND:
/*
* The newly found device is NOT in the existing device
* list.
*/
device->new_device = true;
break;
case DEVICE_CHANGED:
/*
* The original device has gone away and we need to add
* the new device.
*/
device->new_device = true;
break;
}
}
/* Process all devices that have gone away. */
list_for_each_entry_safe(device, next, &ctrl_info->scsi_device_list,
scsi_device_list_entry) {
if (device->device_gone) {
list_del(&device->scsi_device_list_entry);
list_add_tail(&device->delete_list_entry, &delete_list);
}
}
/* Process all new devices. */
for (i = 0; i < num_new_devices; i++) {
device = new_device_list[i];
if (!device->new_device)
continue;
if (device->volume_offline)
continue;
list_add_tail(&device->scsi_device_list_entry,
&ctrl_info->scsi_device_list);
list_add_tail(&device->add_list_entry, &add_list);
/* To prevent this device structure from being freed later. */
device->keep_device = true;
pqi_init_device_tmf_work(device);
}
spin_unlock_irqrestore(&ctrl_info->scsi_device_list_lock, flags);
/*
* If OFA is in progress and there are devices that need to be deleted,
* allow any pending reset operations to continue and unblock any SCSI
* requests before removal.
*/
if (pqi_ofa_in_progress(ctrl_info)) {
list_for_each_entry_safe(device, next, &delete_list, delete_list_entry)
if (pqi_is_device_added(device))
pqi_device_remove_start(device);
pqi_ctrl_unblock_device_reset(ctrl_info);
pqi_scsi_unblock_requests(ctrl_info);
}
/* Remove all devices that have gone away. */
list_for_each_entry_safe(device, next, &delete_list, delete_list_entry) {
if (device->volume_offline) {
pqi_dev_info(ctrl_info, "offline", device);
pqi_show_volume_status(ctrl_info, device);
} else {
pqi_dev_info(ctrl_info, "removed", device);
}
if (pqi_is_device_added(device))
pqi_remove_device(ctrl_info, device);
list_del(&device->delete_list_entry);
pqi_free_device(device);
}
/*
* Notify the SML of any existing device changes such as;
* queue depth, device size.
*/
list_for_each_entry(device, &ctrl_info->scsi_device_list, scsi_device_list_entry) {
if (device->sdev && device->queue_depth != device->advertised_queue_depth) {
device->advertised_queue_depth = device->queue_depth;
scsi_change_queue_depth(device->sdev, device->advertised_queue_depth);
spin_lock_irqsave(&ctrl_info->scsi_device_list_lock, flags);
if (pqi_volume_rescan_needed(device)) {
device->rescan = false;
spin_unlock_irqrestore(&ctrl_info->scsi_device_list_lock, flags);
scsi_rescan_device(device->sdev);
} else {
spin_unlock_irqrestore(&ctrl_info->scsi_device_list_lock, flags);
}
}
}
/* Expose any new devices. */
list_for_each_entry_safe(device, next, &add_list, add_list_entry) {
if (!pqi_is_device_added(device)) {
rc = pqi_add_device(ctrl_info, device);
if (rc == 0) {
pqi_dev_info(ctrl_info, "added", device);
} else {
dev_warn(&ctrl_info->pci_dev->dev,
"scsi %d:%d:%d:%d addition failed, device not added\n",
ctrl_info->scsi_host->host_no,
device->bus, device->target,
device->lun);
pqi_fixup_botched_add(ctrl_info, device);
}
}
}
}
static inline bool pqi_is_supported_device(struct pqi_scsi_dev *device)
{
/*
* Only support the HBA controller itself as a RAID
* controller. If it's a RAID controller other than
* the HBA itself (an external RAID controller, for
* example), we don't support it.
*/
if (device->device_type == SA_DEVICE_TYPE_CONTROLLER &&
!pqi_is_hba_lunid(device->scsi3addr))
return false;
return true;
}
static inline bool pqi_skip_device(u8 *scsi3addr)
{
/* Ignore all masked devices. */
if (MASKED_DEVICE(scsi3addr))
return true;
return false;
}
static inline void pqi_mask_device(u8 *scsi3addr)
{
scsi3addr[3] |= 0xc0;
}
static inline bool pqi_is_multipath_device(struct pqi_scsi_dev *device)
{
if (pqi_is_logical_device(device))
return false;
return (device->path_map & (device->path_map - 1)) != 0;
}
static inline bool pqi_expose_device(struct pqi_scsi_dev *device)
{
return !device->is_physical_device || !pqi_skip_device(device->scsi3addr);
}
static int pqi_update_scsi_devices(struct pqi_ctrl_info *ctrl_info)
{
int i;
int rc;
LIST_HEAD(new_device_list_head);
struct report_phys_lun_16byte_wwid_list *physdev_list = NULL;
struct report_log_lun_list *logdev_list = NULL;
struct report_phys_lun_16byte_wwid *phys_lun;
struct report_log_lun *log_lun;
struct bmic_identify_physical_device *id_phys = NULL;
u32 num_physicals;
u32 num_logicals;
struct pqi_scsi_dev **new_device_list = NULL;
struct pqi_scsi_dev *device;
struct pqi_scsi_dev *next;
unsigned int num_new_devices;
unsigned int num_valid_devices;
bool is_physical_device;
u8 *scsi3addr;
unsigned int physical_index;
unsigned int logical_index;
static char *out_of_memory_msg =
"failed to allocate memory, device discovery stopped";
rc = pqi_get_device_lists(ctrl_info, &physdev_list, &logdev_list);
if (rc)
goto out;
if (physdev_list)
num_physicals =
get_unaligned_be32(&physdev_list->header.list_length)
/ sizeof(physdev_list->lun_entries[0]);
else
num_physicals = 0;
if (logdev_list)
num_logicals =
get_unaligned_be32(&logdev_list->header.list_length)
/ sizeof(logdev_list->lun_entries[0]);
else
num_logicals = 0;
if (num_physicals) {
/*
* We need this buffer for calls to pqi_get_physical_disk_info()
* below. We allocate it here instead of inside
* pqi_get_physical_disk_info() because it's a fairly large
* buffer.
*/
id_phys = kmalloc(sizeof(*id_phys), GFP_KERNEL);
if (!id_phys) {
dev_warn(&ctrl_info->pci_dev->dev, "%s\n",
out_of_memory_msg);
rc = -ENOMEM;
goto out;
}
if (pqi_hide_vsep) {
for (i = num_physicals - 1; i >= 0; i--) {
phys_lun = &physdev_list->lun_entries[i];
if (CISS_GET_DRIVE_NUMBER(phys_lun->lunid) == PQI_VSEP_CISS_BTL) {
pqi_mask_device(phys_lun->lunid);
break;
}
}
}
}
if (num_logicals &&
(logdev_list->header.flags & CISS_REPORT_LOG_FLAG_DRIVE_TYPE_MIX))
ctrl_info->lv_drive_type_mix_valid = true;
num_new_devices = num_physicals + num_logicals;
new_device_list = kmalloc_array(num_new_devices,
sizeof(*new_device_list),
GFP_KERNEL);
if (!new_device_list) {
dev_warn(&ctrl_info->pci_dev->dev, "%s\n", out_of_memory_msg);
rc = -ENOMEM;
goto out;
}
for (i = 0; i < num_new_devices; i++) {
device = kzalloc(sizeof(*device), GFP_KERNEL);
if (!device) {
dev_warn(&ctrl_info->pci_dev->dev, "%s\n",
out_of_memory_msg);
rc = -ENOMEM;
goto out;
}
list_add_tail(&device->new_device_list_entry,
&new_device_list_head);
}
device = NULL;
num_valid_devices = 0;
physical_index = 0;
logical_index = 0;
for (i = 0; i < num_new_devices; i++) {
if ((!pqi_expose_ld_first && i < num_physicals) ||
(pqi_expose_ld_first && i >= num_logicals)) {
is_physical_device = true;
phys_lun = &physdev_list->lun_entries[physical_index++];
log_lun = NULL;
scsi3addr = phys_lun->lunid;
} else {
is_physical_device = false;
phys_lun = NULL;
log_lun = &logdev_list->lun_entries[logical_index++];
scsi3addr = log_lun->lunid;
}
if (is_physical_device && pqi_skip_device(scsi3addr))
continue;
if (device)
device = list_next_entry(device, new_device_list_entry);
else
device = list_first_entry(&new_device_list_head,
struct pqi_scsi_dev, new_device_list_entry);
memcpy(device->scsi3addr, scsi3addr, sizeof(device->scsi3addr));
device->is_physical_device = is_physical_device;
if (is_physical_device) {
device->device_type = phys_lun->device_type;
if (device->device_type == SA_DEVICE_TYPE_EXPANDER_SMP)
device->is_expander_smp_device = true;
} else {
device->is_external_raid_device =
pqi_is_external_raid_addr(scsi3addr);
}
if (!pqi_is_supported_device(device))
continue;
/* Gather information about the device. */
rc = pqi_get_device_info(ctrl_info, device, id_phys);
if (rc == -ENOMEM) {
dev_warn(&ctrl_info->pci_dev->dev, "%s\n",
out_of_memory_msg);
goto out;
}
if (rc) {
if (device->is_physical_device)
dev_warn(&ctrl_info->pci_dev->dev,
"obtaining device info failed, skipping physical device %016llx%016llx\n",
get_unaligned_be64(&phys_lun->wwid[0]),
get_unaligned_be64(&phys_lun->wwid[8]));
else
dev_warn(&ctrl_info->pci_dev->dev,
"obtaining device info failed, skipping logical device %08x%08x\n",
*((u32 *)&device->scsi3addr),
*((u32 *)&device->scsi3addr[4]));
rc = 0;
continue;
}
/* Do not present disks that the OS cannot fully probe. */
if (pqi_keep_device_offline(device))
continue;
pqi_assign_bus_target_lun(device);
if (device->is_physical_device) {
memcpy(device->wwid, phys_lun->wwid, sizeof(device->wwid));
if ((phys_lun->device_flags &
CISS_REPORT_PHYS_DEV_FLAG_AIO_ENABLED) &&
phys_lun->aio_handle) {
device->aio_enabled = true;
device->aio_handle =
phys_lun->aio_handle;
}
} else {
memcpy(device->volume_id, log_lun->volume_id,
sizeof(device->volume_id));
}
device->sas_address = get_unaligned_be64(&device->wwid[0]);
new_device_list[num_valid_devices++] = device;
}
pqi_update_device_list(ctrl_info, new_device_list, num_valid_devices);
out:
list_for_each_entry_safe(device, next, &new_device_list_head,
new_device_list_entry) {
if (device->keep_device)
continue;
list_del(&device->new_device_list_entry);
pqi_free_device(device);
}
kfree(new_device_list);
kfree(physdev_list);
kfree(logdev_list);
kfree(id_phys);
return rc;
}
static int pqi_scan_scsi_devices(struct pqi_ctrl_info *ctrl_info)
{
int rc;
int mutex_acquired;
if (pqi_ctrl_offline(ctrl_info))
return -ENXIO;
mutex_acquired = mutex_trylock(&ctrl_info->scan_mutex);
if (!mutex_acquired) {
if (pqi_ctrl_scan_blocked(ctrl_info))
return -EBUSY;
pqi_schedule_rescan_worker_delayed(ctrl_info);
return -EINPROGRESS;
}
rc = pqi_update_scsi_devices(ctrl_info);
if (rc && !pqi_ctrl_scan_blocked(ctrl_info))
pqi_schedule_rescan_worker_delayed(ctrl_info);
mutex_unlock(&ctrl_info->scan_mutex);
return rc;
}
static void pqi_scan_start(struct Scsi_Host *shost)
{
struct pqi_ctrl_info *ctrl_info;
ctrl_info = shost_to_hba(shost);
pqi_scan_scsi_devices(ctrl_info);
}
/* Returns TRUE if scan is finished. */
static int pqi_scan_finished(struct Scsi_Host *shost,
unsigned long elapsed_time)
{
struct pqi_ctrl_info *ctrl_info;
ctrl_info = shost_priv(shost);
return !mutex_is_locked(&ctrl_info->scan_mutex);
}
static inline void pqi_set_encryption_info(struct pqi_encryption_info *encryption_info,
struct raid_map *raid_map, u64 first_block)
{
u32 volume_blk_size;
/*
* Set the encryption tweak values based on logical block address.
* If the block size is 512, the tweak value is equal to the LBA.
* For other block sizes, tweak value is (LBA * block size) / 512.
*/
volume_blk_size = get_unaligned_le32(&raid_map->volume_blk_size);
if (volume_blk_size != 512)
first_block = (first_block * volume_blk_size) / 512;
encryption_info->data_encryption_key_index =
get_unaligned_le16(&raid_map->data_encryption_key_index);
encryption_info->encrypt_tweak_lower = lower_32_bits(first_block);
encryption_info->encrypt_tweak_upper = upper_32_bits(first_block);
}
/*
* Attempt to perform RAID bypass mapping for a logical volume I/O.
*/
static bool pqi_aio_raid_level_supported(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev_raid_map_data *rmd)
{
bool is_supported = true;
switch (rmd->raid_level) {
case SA_RAID_0:
break;
case SA_RAID_1:
if (rmd->is_write && (!ctrl_info->enable_r1_writes ||
rmd->data_length > ctrl_info->max_write_raid_1_10_2drive))
is_supported = false;
break;
case SA_RAID_TRIPLE:
if (rmd->is_write && (!ctrl_info->enable_r1_writes ||
rmd->data_length > ctrl_info->max_write_raid_1_10_3drive))
is_supported = false;
break;
case SA_RAID_5:
if (rmd->is_write && (!ctrl_info->enable_r5_writes ||
rmd->data_length > ctrl_info->max_write_raid_5_6))
is_supported = false;
break;
case SA_RAID_6:
if (rmd->is_write && (!ctrl_info->enable_r6_writes ||
rmd->data_length > ctrl_info->max_write_raid_5_6))
is_supported = false;
break;
default:
is_supported = false;
break;
}
return is_supported;
}
#define PQI_RAID_BYPASS_INELIGIBLE 1
static int pqi_get_aio_lba_and_block_count(struct scsi_cmnd *scmd,
struct pqi_scsi_dev_raid_map_data *rmd)
{
/* Check for valid opcode, get LBA and block count. */
switch (scmd->cmnd[0]) {
case WRITE_6:
rmd->is_write = true;
fallthrough;
case READ_6:
rmd->first_block = (u64)(((scmd->cmnd[1] & 0x1f) << 16) |
(scmd->cmnd[2] << 8) | scmd->cmnd[3]);
rmd->block_cnt = (u32)scmd->cmnd[4];
if (rmd->block_cnt == 0)
rmd->block_cnt = 256;
break;
case WRITE_10:
rmd->is_write = true;
fallthrough;
case READ_10:
rmd->first_block = (u64)get_unaligned_be32(&scmd->cmnd[2]);
rmd->block_cnt = (u32)get_unaligned_be16(&scmd->cmnd[7]);
break;
case WRITE_12:
rmd->is_write = true;
fallthrough;
case READ_12:
rmd->first_block = (u64)get_unaligned_be32(&scmd->cmnd[2]);
rmd->block_cnt = get_unaligned_be32(&scmd->cmnd[6]);
break;
case WRITE_16:
rmd->is_write = true;
fallthrough;
case READ_16:
rmd->first_block = get_unaligned_be64(&scmd->cmnd[2]);
rmd->block_cnt = get_unaligned_be32(&scmd->cmnd[10]);
break;
default:
/* Process via normal I/O path. */
return PQI_RAID_BYPASS_INELIGIBLE;
}
put_unaligned_le32(scsi_bufflen(scmd), &rmd->data_length);
return 0;
}
static int pci_get_aio_common_raid_map_values(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev_raid_map_data *rmd, struct raid_map *raid_map)
{
#if BITS_PER_LONG == 32
u64 tmpdiv;
#endif
rmd->last_block = rmd->first_block + rmd->block_cnt - 1;
/* Check for invalid block or wraparound. */
if (rmd->last_block >=
get_unaligned_le64(&raid_map->volume_blk_cnt) ||
rmd->last_block < rmd->first_block)
return PQI_RAID_BYPASS_INELIGIBLE;
rmd->data_disks_per_row =
get_unaligned_le16(&raid_map->data_disks_per_row);
rmd->strip_size = get_unaligned_le16(&raid_map->strip_size);
rmd->layout_map_count = get_unaligned_le16(&raid_map->layout_map_count);
/* Calculate stripe information for the request. */
rmd->blocks_per_row = rmd->data_disks_per_row * rmd->strip_size;
if (rmd->blocks_per_row == 0) /* Used as a divisor in many calculations */
return PQI_RAID_BYPASS_INELIGIBLE;
#if BITS_PER_LONG == 32
tmpdiv = rmd->first_block;
do_div(tmpdiv, rmd->blocks_per_row);
rmd->first_row = tmpdiv;
tmpdiv = rmd->last_block;
do_div(tmpdiv, rmd->blocks_per_row);
rmd->last_row = tmpdiv;
rmd->first_row_offset = (u32)(rmd->first_block - (rmd->first_row * rmd->blocks_per_row));
rmd->last_row_offset = (u32)(rmd->last_block - (rmd->last_row * rmd->blocks_per_row));
tmpdiv = rmd->first_row_offset;
do_div(tmpdiv, rmd->strip_size);
rmd->first_column = tmpdiv;
tmpdiv = rmd->last_row_offset;
do_div(tmpdiv, rmd->strip_size);
rmd->last_column = tmpdiv;
#else
rmd->first_row = rmd->first_block / rmd->blocks_per_row;
rmd->last_row = rmd->last_block / rmd->blocks_per_row;
rmd->first_row_offset = (u32)(rmd->first_block -
(rmd->first_row * rmd->blocks_per_row));
rmd->last_row_offset = (u32)(rmd->last_block - (rmd->last_row *
rmd->blocks_per_row));
rmd->first_column = rmd->first_row_offset / rmd->strip_size;
rmd->last_column = rmd->last_row_offset / rmd->strip_size;
#endif
/* If this isn't a single row/column then give to the controller. */
if (rmd->first_row != rmd->last_row ||
rmd->first_column != rmd->last_column)
return PQI_RAID_BYPASS_INELIGIBLE;
/* Proceeding with driver mapping. */
rmd->total_disks_per_row = rmd->data_disks_per_row +
get_unaligned_le16(&raid_map->metadata_disks_per_row);
rmd->map_row = ((u32)(rmd->first_row >>
raid_map->parity_rotation_shift)) %
get_unaligned_le16(&raid_map->row_cnt);
rmd->map_index = (rmd->map_row * rmd->total_disks_per_row) +
rmd->first_column;
return 0;
}
static int pqi_calc_aio_r5_or_r6(struct pqi_scsi_dev_raid_map_data *rmd,
struct raid_map *raid_map)
{
#if BITS_PER_LONG == 32
u64 tmpdiv;
#endif
if (rmd->blocks_per_row == 0) /* Used as a divisor in many calculations */
return PQI_RAID_BYPASS_INELIGIBLE;
/* RAID 50/60 */
/* Verify first and last block are in same RAID group. */
rmd->stripesize = rmd->blocks_per_row * rmd->layout_map_count;
#if BITS_PER_LONG == 32
tmpdiv = rmd->first_block;
rmd->first_group = do_div(tmpdiv, rmd->stripesize);
tmpdiv = rmd->first_group;
do_div(tmpdiv, rmd->blocks_per_row);
rmd->first_group = tmpdiv;
tmpdiv = rmd->last_block;
rmd->last_group = do_div(tmpdiv, rmd->stripesize);
tmpdiv = rmd->last_group;
do_div(tmpdiv, rmd->blocks_per_row);
rmd->last_group = tmpdiv;
#else
rmd->first_group = (rmd->first_block % rmd->stripesize) / rmd->blocks_per_row;
rmd->last_group = (rmd->last_block % rmd->stripesize) / rmd->blocks_per_row;
#endif
if (rmd->first_group != rmd->last_group)
return PQI_RAID_BYPASS_INELIGIBLE;
/* Verify request is in a single row of RAID 5/6. */
#if BITS_PER_LONG == 32
tmpdiv = rmd->first_block;
do_div(tmpdiv, rmd->stripesize);
rmd->first_row = tmpdiv;
rmd->r5or6_first_row = tmpdiv;
tmpdiv = rmd->last_block;
do_div(tmpdiv, rmd->stripesize);
rmd->r5or6_last_row = tmpdiv;
#else
rmd->first_row = rmd->r5or6_first_row =
rmd->first_block / rmd->stripesize;
rmd->r5or6_last_row = rmd->last_block / rmd->stripesize;
#endif
if (rmd->r5or6_first_row != rmd->r5or6_last_row)
return PQI_RAID_BYPASS_INELIGIBLE;
/* Verify request is in a single column. */
#if BITS_PER_LONG == 32
tmpdiv = rmd->first_block;
rmd->first_row_offset = do_div(tmpdiv, rmd->stripesize);
tmpdiv = rmd->first_row_offset;
rmd->first_row_offset = (u32)do_div(tmpdiv, rmd->blocks_per_row);
rmd->r5or6_first_row_offset = rmd->first_row_offset;
tmpdiv = rmd->last_block;
rmd->r5or6_last_row_offset = do_div(tmpdiv, rmd->stripesize);
tmpdiv = rmd->r5or6_last_row_offset;
rmd->r5or6_last_row_offset = do_div(tmpdiv, rmd->blocks_per_row);
tmpdiv = rmd->r5or6_first_row_offset;
do_div(tmpdiv, rmd->strip_size);
rmd->first_column = rmd->r5or6_first_column = tmpdiv;
tmpdiv = rmd->r5or6_last_row_offset;
do_div(tmpdiv, rmd->strip_size);
rmd->r5or6_last_column = tmpdiv;
#else
rmd->first_row_offset = rmd->r5or6_first_row_offset =
(u32)((rmd->first_block % rmd->stripesize) %
rmd->blocks_per_row);
rmd->r5or6_last_row_offset =
(u32)((rmd->last_block % rmd->stripesize) %
rmd->blocks_per_row);
rmd->first_column =
rmd->r5or6_first_row_offset / rmd->strip_size;
rmd->r5or6_first_column = rmd->first_column;
rmd->r5or6_last_column = rmd->r5or6_last_row_offset / rmd->strip_size;
#endif
if (rmd->r5or6_first_column != rmd->r5or6_last_column)
return PQI_RAID_BYPASS_INELIGIBLE;
/* Request is eligible. */
rmd->map_row =
((u32)(rmd->first_row >> raid_map->parity_rotation_shift)) %
get_unaligned_le16(&raid_map->row_cnt);
rmd->map_index = (rmd->first_group *
(get_unaligned_le16(&raid_map->row_cnt) *
rmd->total_disks_per_row)) +
(rmd->map_row * rmd->total_disks_per_row) + rmd->first_column;
if (rmd->is_write) {
u32 index;
/*
* p_parity_it_nexus and q_parity_it_nexus are pointers to the
* parity entries inside the device's raid_map.
*
* A device's RAID map is bounded by: number of RAID disks squared.
*
* The devices RAID map size is checked during device
* initialization.
*/
index = DIV_ROUND_UP(rmd->map_index + 1, rmd->total_disks_per_row);
index *= rmd->total_disks_per_row;
index -= get_unaligned_le16(&raid_map->metadata_disks_per_row);
rmd->p_parity_it_nexus = raid_map->disk_data[index].aio_handle;
if (rmd->raid_level == SA_RAID_6) {
rmd->q_parity_it_nexus = raid_map->disk_data[index + 1].aio_handle;
rmd->xor_mult = raid_map->disk_data[rmd->map_index].xor_mult[1];
}
#if BITS_PER_LONG == 32
tmpdiv = rmd->first_block;
do_div(tmpdiv, rmd->blocks_per_row);
rmd->row = tmpdiv;
#else
rmd->row = rmd->first_block / rmd->blocks_per_row;
#endif
}
return 0;
}
static void pqi_set_aio_cdb(struct pqi_scsi_dev_raid_map_data *rmd)
{
/* Build the new CDB for the physical disk I/O. */
if (rmd->disk_block > 0xffffffff) {
rmd->cdb[0] = rmd->is_write ? WRITE_16 : READ_16;
rmd->cdb[1] = 0;
put_unaligned_be64(rmd->disk_block, &rmd->cdb[2]);
put_unaligned_be32(rmd->disk_block_cnt, &rmd->cdb[10]);
rmd->cdb[14] = 0;
rmd->cdb[15] = 0;
rmd->cdb_length = 16;
} else {
rmd->cdb[0] = rmd->is_write ? WRITE_10 : READ_10;
rmd->cdb[1] = 0;
put_unaligned_be32((u32)rmd->disk_block, &rmd->cdb[2]);
rmd->cdb[6] = 0;
put_unaligned_be16((u16)rmd->disk_block_cnt, &rmd->cdb[7]);
rmd->cdb[9] = 0;
rmd->cdb_length = 10;
}
}
static void pqi_calc_aio_r1_nexus(struct raid_map *raid_map,
struct pqi_scsi_dev_raid_map_data *rmd)
{
u32 index;
u32 group;
group = rmd->map_index / rmd->data_disks_per_row;
index = rmd->map_index - (group * rmd->data_disks_per_row);
rmd->it_nexus[0] = raid_map->disk_data[index].aio_handle;
index += rmd->data_disks_per_row;
rmd->it_nexus[1] = raid_map->disk_data[index].aio_handle;
if (rmd->layout_map_count > 2) {
index += rmd->data_disks_per_row;
rmd->it_nexus[2] = raid_map->disk_data[index].aio_handle;
}
rmd->num_it_nexus_entries = rmd->layout_map_count;
}
static int pqi_raid_bypass_submit_scsi_cmd(struct pqi_ctrl_info *ctrl_info,
struct pqi_scsi_dev *device, struct scsi_cmnd *scmd,
struct pqi_queue_group *queue_group)
{
int rc;
struct raid_map *raid_map;
u32 group;
u32 next_bypass_group;
struct pqi_encryption_info *encryption_info_ptr;
struct pqi_encryption_info encryption_info;
struct pqi_scsi_dev_raid_map_data rmd = { 0 };
rc = pqi_get_aio_lba_and_block_count(scmd, &rmd);
if (rc)
return PQI_RAID_BYPASS_INELIGIBLE;
rmd.raid_level = device->raid_level;
if (!pqi_aio_raid_level_supported(ctrl_info, &rmd))
return PQI_RAID_BYPASS_INELIGIBLE;
if (unlikely(rmd.block_cnt == 0))
return PQI_RAID_BYPASS_INELIGIBLE;
raid_map = device->raid_map;
rc = pci_get_aio_common_raid_map_values(ctrl_info, &rmd, raid_map);
if (rc)
return PQI_RAID_BYPASS_INELIGIBLE;
if (device->raid_level == SA_RAID_1 ||
device->raid_level == SA_RAID_TRIPLE) {
if (rmd.is_write) {
pqi_calc_aio_r1_nexus(raid_map, &rmd);
} else {
group = device->next_bypass_group[rmd.map_index];
next_bypass_group = group + 1;
if (next_bypass_group >= rmd.layout_map_count)
next_bypass_group = 0;
device->next_bypass_group[rmd.map_index] = next_bypass_group;
rmd.map_index += group * rmd.data_disks_per_row;
}
} else if ((device->raid_level == SA_RAID_5 ||
device->raid_level == SA_RAID_6) &&
(rmd.layout_map_count > 1 || rmd.is_write)) {
rc = pqi_calc_aio_r5_or_r6(&rmd, raid_map);
if (rc)
return PQI_RAID_BYPASS_INELIGIBLE;
}
if (unlikely(rmd.map_index >= RAID_MAP_MAX_ENTRIES))
return PQI_RAID_BYPASS_INELIGIBLE;
rmd.aio_handle = raid_map->disk_data[rmd.map_index].aio_handle;
rmd.disk_block = get_unaligned_le64(&raid_map->disk_starting_blk) +
rmd.first_row * rmd.strip_size +
(rmd.first_row_offset - rmd.first_column * rmd.strip_size);
rmd.disk_block_cnt = rmd.block_cnt;
/* Handle differing logical/physical block sizes. */
if (raid_map->phys_blk_shift) {
rmd.disk_block <<= raid_map->phys_blk_shift;
rmd.disk_block_cnt <<= raid_map->phys_blk_shift;
}
if (unlikely(rmd.disk_block_cnt > 0xffff))
return PQI_RAID_BYPASS_INELIGIBLE;
pqi_set_aio_cdb(&rmd);
if (get_unaligned_le16(&raid_map->flags) & RAID_MAP_ENCRYPTION_ENABLED) {
if (rmd.data_length > device->max_transfer_encrypted)