freebsd-skq/sys/dev/smartpqi/smartpqi_request.c

791 lines
23 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2018 Microsemi Corporation.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/* $FreeBSD$ */
#include "smartpqi_includes.h"
#define SG_FLAG_LAST 0x40000000
#define SG_FLAG_CHAIN 0x80000000
/* Subroutine to find out embedded sgl count in IU */
static inline
uint32_t pqisrc_embedded_sgl_count(uint32_t elem_alloted)
{
uint32_t embedded_sgl_count = MAX_EMBEDDED_SG_IN_FIRST_IU;
DBG_FUNC(" IN ");
/**
calculate embedded sgl count using num_elem_alloted for IO
**/
if(elem_alloted - 1)
embedded_sgl_count += ((elem_alloted - 1) * MAX_EMBEDDED_SG_IN_IU);
DBG_IO("embedded_sgl_count :%d\n",embedded_sgl_count);
DBG_FUNC(" OUT ");
return embedded_sgl_count;
}
/* Subroutine to find out contiguous free elem in IU */
static inline
uint32_t pqisrc_contiguous_free_elem(uint32_t pi, uint32_t ci, uint32_t elem_in_q)
{
uint32_t contiguous_free_elem = 0;
DBG_FUNC(" IN ");
if(pi >= ci) {
contiguous_free_elem = (elem_in_q - pi);
if(ci == 0)
contiguous_free_elem -= 1;
} else {
contiguous_free_elem = (ci - pi - 1);
}
DBG_FUNC(" OUT ");
return contiguous_free_elem;
}
/* Subroutine to find out num of elements need for the request */
static uint32_t
pqisrc_num_elem_needed(pqisrc_softstate_t *softs, uint32_t SG_Count)
{
uint32_t num_sg;
uint32_t num_elem_required = 1;
DBG_FUNC(" IN ");
DBG_IO("SGL_Count :%d",SG_Count);
/********
If SG_Count greater than max sg per IU i.e 4 or 68
(4 is with out spanning or 68 is with spanning) chaining is required.
OR, If SG_Count <= MAX_EMBEDDED_SG_IN_FIRST_IU then,
on these two cases one element is enough.
********/
if(SG_Count > softs->max_sg_per_iu || SG_Count <= MAX_EMBEDDED_SG_IN_FIRST_IU)
return num_elem_required;
/*
SGL Count Other Than First IU
*/
num_sg = SG_Count - MAX_EMBEDDED_SG_IN_FIRST_IU;
num_elem_required += PQISRC_DIV_ROUND_UP(num_sg, MAX_EMBEDDED_SG_IN_IU);
DBG_FUNC(" OUT ");
return num_elem_required;
}
/* Subroutine to build SG list for the IU submission*/
static
boolean_t pqisrc_build_sgl(sgt_t *sg_array, rcb_t *rcb, iu_header_t *iu_hdr,
uint32_t num_elem_alloted)
{
uint32_t i;
uint32_t num_sg = OS_GET_IO_SG_COUNT(rcb);
sgt_t *sgt = sg_array;
sgt_t *sg_chain = NULL;
boolean_t partial = false;
DBG_FUNC(" IN ");
DBG_IO("SGL_Count :%d",num_sg);
if (0 == num_sg) {
goto out;
}
if (num_sg <= pqisrc_embedded_sgl_count(num_elem_alloted)) {
for (i = 0; i < num_sg; i++, sgt++) {
sgt->addr= OS_GET_IO_SG_ADDR(rcb,i);
sgt->len= OS_GET_IO_SG_LEN(rcb,i);
sgt->flags= 0;
}
sg_array[num_sg - 1].flags = SG_FLAG_LAST;
} else {
/**
SGL Chaining
**/
sg_chain = rcb->sg_chain_virt;
sgt->addr = rcb->sg_chain_dma;
sgt->len = num_sg * sizeof(sgt_t);
sgt->flags = SG_FLAG_CHAIN;
sgt = sg_chain;
for (i = 0; i < num_sg; i++, sgt++) {
sgt->addr = OS_GET_IO_SG_ADDR(rcb,i);
sgt->len = OS_GET_IO_SG_LEN(rcb,i);
sgt->flags = 0;
}
sg_chain[num_sg - 1].flags = SG_FLAG_LAST;
num_sg = 1;
partial = true;
}
out:
iu_hdr->iu_length = num_sg * sizeof(sgt_t);
DBG_FUNC(" OUT ");
return partial;
}
/*Subroutine used to Build the RAID request */
static void
pqisrc_build_raid_io(pqisrc_softstate_t *softs, rcb_t *rcb,
pqisrc_raid_req_t *raid_req, uint32_t num_elem_alloted)
{
DBG_FUNC(" IN ");
raid_req->header.iu_type = PQI_IU_TYPE_RAID_PATH_IO_REQUEST;
raid_req->header.comp_feature = 0;
raid_req->response_queue_id = OS_GET_IO_RESP_QID(softs, rcb);
raid_req->work_area[0] = 0;
raid_req->work_area[1] = 0;
raid_req->request_id = rcb->tag;
raid_req->nexus_id = 0;
raid_req->buffer_length = GET_SCSI_BUFFLEN(rcb);
memcpy(raid_req->lun_number, rcb->dvp->scsi3addr,
sizeof(raid_req->lun_number));
raid_req->protocol_spec = 0;
raid_req->data_direction = rcb->data_dir;
raid_req->reserved1 = 0;
raid_req->fence = 0;
raid_req->error_index = raid_req->request_id;
raid_req->reserved2 = 0;
raid_req->task_attribute = OS_GET_TASK_ATTR(rcb);
raid_req->command_priority = 0;
raid_req->reserved3 = 0;
raid_req->reserved4 = 0;
raid_req->reserved5 = 0;
/* As cdb and additional_cdb_bytes are contiguous,
update them in a single statement */
memcpy(raid_req->cdb, rcb->cdbp, rcb->cmdlen);
#if 0
DBG_IO("CDB :");
for(i = 0; i < rcb->cmdlen ; i++)
DBG_IO(" 0x%x \n ",raid_req->cdb[i]);
#endif
switch (rcb->cmdlen) {
case 6:
case 10:
case 12:
case 16:
raid_req->additional_cdb_bytes_usage =
PQI_ADDITIONAL_CDB_BYTES_0;
break;
case 20:
raid_req->additional_cdb_bytes_usage =
PQI_ADDITIONAL_CDB_BYTES_4;
break;
case 24:
raid_req->additional_cdb_bytes_usage =
PQI_ADDITIONAL_CDB_BYTES_8;
break;
case 28:
raid_req->additional_cdb_bytes_usage =
PQI_ADDITIONAL_CDB_BYTES_12;
break;
case 32:
default: /* todo:review again */
raid_req->additional_cdb_bytes_usage =
PQI_ADDITIONAL_CDB_BYTES_16;
break;
}
/* Frame SGL Descriptor */
raid_req->partial = pqisrc_build_sgl(&raid_req->sg_descriptors[0], rcb,
&raid_req->header, num_elem_alloted);
raid_req->header.iu_length +=
offsetof(pqisrc_raid_req_t, sg_descriptors) - sizeof(iu_header_t);
#if 0
DBG_IO("raid_req->header.iu_type : 0x%x", raid_req->header.iu_type);
DBG_IO("raid_req->response_queue_id :%d\n"raid_req->response_queue_id);
DBG_IO("raid_req->request_id : 0x%x", raid_req->request_id);
DBG_IO("raid_req->buffer_length : 0x%x", raid_req->buffer_length);
DBG_IO("raid_req->task_attribute : 0x%x", raid_req->task_attribute);
DBG_IO("raid_req->lun_number : 0x%x", raid_req->lun_number);
DBG_IO("raid_req->error_index : 0x%x", raid_req->error_index);
DBG_IO("raid_req->sg_descriptors[0].addr : %p", (void*)raid_req->sg_descriptors[0].addr);
DBG_IO("raid_req->sg_descriptors[0].len : 0x%x", raid_req->sg_descriptors[0].len);
DBG_IO("raid_req->sg_descriptors[0].flags : 0%x", raid_req->sg_descriptors[0].flags);
#endif
rcb->success_cmp_callback = pqisrc_process_io_response_success;
rcb->error_cmp_callback = pqisrc_process_raid_response_error;
rcb->resp_qid = raid_req->response_queue_id;
DBG_FUNC(" OUT ");
}
/*Subroutine used to Build the AIO request */
static void
pqisrc_build_aio_io(pqisrc_softstate_t *softs, rcb_t *rcb,
pqi_aio_req_t *aio_req, uint32_t num_elem_alloted)
{
DBG_FUNC(" IN ");
aio_req->header.iu_type = PQI_IU_TYPE_AIO_PATH_IO_REQUEST;
aio_req->header.comp_feature = 0;
aio_req->response_queue_id = OS_GET_IO_RESP_QID(softs, rcb);
aio_req->work_area[0] = 0;
aio_req->work_area[1] = 0;
aio_req->req_id = rcb->tag;
aio_req->res1[0] = 0;
aio_req->res1[1] = 0;
aio_req->nexus = rcb->ioaccel_handle;
aio_req->buf_len = GET_SCSI_BUFFLEN(rcb);
aio_req->data_dir = rcb->data_dir;
aio_req->mem_type = 0;
aio_req->fence = 0;
aio_req->res2 = 0;
aio_req->task_attr = OS_GET_TASK_ATTR(rcb);
aio_req->cmd_prio = 0;
aio_req->res3 = 0;
aio_req->err_idx = aio_req->req_id;
aio_req->cdb_len = rcb->cmdlen;
if(rcb->cmdlen > sizeof(aio_req->cdb))
rcb->cmdlen = sizeof(aio_req->cdb);
memcpy(aio_req->cdb, rcb->cdbp, rcb->cmdlen);
#if 0
DBG_IO("CDB : \n");
for(int i = 0; i < rcb->cmdlen ; i++)
DBG_IO(" 0x%x \n",aio_req->cdb[i]);
#endif
memset(aio_req->lun,0,sizeof(aio_req->lun));
memset(aio_req->res4,0,sizeof(aio_req->res4));
if(rcb->encrypt_enable == true) {
aio_req->encrypt_enable = true;
aio_req->encrypt_key_index = LE_16(rcb->enc_info.data_enc_key_index);
aio_req->encrypt_twk_low = LE_32(rcb->enc_info.encrypt_tweak_lower);
aio_req->encrypt_twk_high = LE_32(rcb->enc_info.encrypt_tweak_upper);
} else {
aio_req->encrypt_enable = 0;
aio_req->encrypt_key_index = 0;
aio_req->encrypt_twk_high = 0;
aio_req->encrypt_twk_low = 0;
}
/* Frame SGL Descriptor */
aio_req->partial = pqisrc_build_sgl(&aio_req->sg_desc[0], rcb,
&aio_req->header, num_elem_alloted);
aio_req->num_sg = aio_req->header.iu_length / sizeof(sgt_t);
DBG_INFO("aio_req->num_sg :%d",aio_req->num_sg);
aio_req->header.iu_length += offsetof(pqi_aio_req_t, sg_desc) -
sizeof(iu_header_t);
#if 0
DBG_IO("aio_req->header.iu_type : 0x%x \n",aio_req->header.iu_type);
DBG_IO("aio_req->resp_qid :0x%x",aio_req->resp_qid);
DBG_IO("aio_req->req_id : 0x%x \n",aio_req->req_id);
DBG_IO("aio_req->nexus : 0x%x \n",aio_req->nexus);
DBG_IO("aio_req->buf_len : 0x%x \n",aio_req->buf_len);
DBG_IO("aio_req->data_dir : 0x%x \n",aio_req->data_dir);
DBG_IO("aio_req->task_attr : 0x%x \n",aio_req->task_attr);
DBG_IO("aio_req->err_idx : 0x%x \n",aio_req->err_idx);
DBG_IO("aio_req->num_sg :%d",aio_req->num_sg);
DBG_IO("aio_req->sg_desc[0].addr : %p \n", (void*)aio_req->sg_desc[0].addr);
DBG_IO("aio_req->sg_desc[0].len : 0%x \n", aio_req->sg_desc[0].len);
DBG_IO("aio_req->sg_desc[0].flags : 0%x \n", aio_req->sg_desc[0].flags);
#endif
rcb->success_cmp_callback = pqisrc_process_io_response_success;
rcb->error_cmp_callback = pqisrc_process_aio_response_error;
rcb->resp_qid = aio_req->response_queue_id;
DBG_FUNC(" OUT ");
}
/*Function used to build and send RAID/AIO */
int pqisrc_build_send_io(pqisrc_softstate_t *softs,rcb_t *rcb)
{
ib_queue_t *ib_q_array = softs->op_aio_ib_q;
ib_queue_t *ib_q = NULL;
char *ib_iu = NULL;
IO_PATH_T io_path = AIO_PATH;
uint32_t TraverseCount = 0;
int first_qindex = OS_GET_IO_REQ_QINDEX(softs, rcb);
int qindex = first_qindex;
uint32_t num_op_ib_q = softs->num_op_aio_ibq;
uint32_t num_elem_needed;
uint32_t num_elem_alloted = 0;
pqi_scsi_dev_t *devp = rcb->dvp;
uint8_t raidbypass_cdb[16];
DBG_FUNC(" IN ");
rcb->cdbp = OS_GET_CDBP(rcb);
if(IS_AIO_PATH(devp)) {
/** IO for Physical Drive **/
/** Send in AIO PATH**/
rcb->ioaccel_handle = devp->ioaccel_handle;
} else {
int ret = PQI_STATUS_FAILURE;
/** IO for RAID Volume **/
if (devp->offload_enabled) {
/** ByPass IO ,Send in AIO PATH **/
ret = pqisrc_send_scsi_cmd_raidbypass(softs,
devp, rcb, raidbypass_cdb);
}
if (PQI_STATUS_FAILURE == ret) {
/** Send in RAID PATH **/
io_path = RAID_PATH;
num_op_ib_q = softs->num_op_raid_ibq;
ib_q_array = softs->op_raid_ib_q;
} else {
rcb->cdbp = raidbypass_cdb;
}
}
num_elem_needed = pqisrc_num_elem_needed(softs, OS_GET_IO_SG_COUNT(rcb));
DBG_IO("num_elem_needed :%d",num_elem_needed);
do {
uint32_t num_elem_available;
ib_q = (ib_q_array + qindex);
PQI_LOCK(&ib_q->lock);
num_elem_available = pqisrc_contiguous_free_elem(ib_q->pi_local,
*(ib_q->ci_virt_addr), ib_q->num_elem);
DBG_IO("num_elem_avialable :%d\n",num_elem_available);
if(num_elem_available >= num_elem_needed) {
num_elem_alloted = num_elem_needed;
break;
}
DBG_IO("Current queue is busy! Hop to next queue\n");
PQI_UNLOCK(&ib_q->lock);
qindex = (qindex + 1) % num_op_ib_q;
if(qindex == first_qindex) {
if (num_elem_needed == 1)
break;
TraverseCount += 1;
num_elem_needed = 1;
}
}while(TraverseCount < 2);
DBG_IO("num_elem_alloted :%d",num_elem_alloted);
if (num_elem_alloted == 0) {
DBG_WARN("OUT: IB Queues were full\n");
return PQI_STATUS_QFULL;
}
/* Get IB Queue Slot address to build IU */
ib_iu = ib_q->array_virt_addr + (ib_q->pi_local * ib_q->elem_size);
if(io_path == AIO_PATH) {
/** Build AIO structure **/
pqisrc_build_aio_io(softs, rcb, (pqi_aio_req_t*)ib_iu,
num_elem_alloted);
} else {
/** Build RAID structure **/
pqisrc_build_raid_io(softs, rcb, (pqisrc_raid_req_t*)ib_iu,
num_elem_alloted);
}
rcb->req_pending = true;
/* Update the local PI */
ib_q->pi_local = (ib_q->pi_local + num_elem_alloted) % ib_q->num_elem;
DBG_INFO("ib_q->pi_local : %x\n", ib_q->pi_local);
DBG_INFO("*ib_q->ci_virt_addr: %x\n",*(ib_q->ci_virt_addr));
/* Inform the fw about the new IU */
PCI_MEM_PUT32(softs, ib_q->pi_register_abs, ib_q->pi_register_offset, ib_q->pi_local);
PQI_UNLOCK(&ib_q->lock);
DBG_FUNC(" OUT ");
return PQI_STATUS_SUCCESS;
}
/* Subroutine used to set encryption info as part of RAID bypass IO*/
static inline void pqisrc_set_enc_info(
struct pqi_enc_info *enc_info, struct raid_map *raid_map,
uint64_t first_block)
{
uint32_t 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_LE32((uint8_t *)&raid_map->volume_blk_size);
if (volume_blk_size != 512)
first_block = (first_block * volume_blk_size) / 512;
enc_info->data_enc_key_index =
GET_LE16((uint8_t *)&raid_map->data_encryption_key_index);
enc_info->encrypt_tweak_upper = ((uint32_t)(((first_block) >> 16) >> 16));
enc_info->encrypt_tweak_lower = ((uint32_t)(first_block));
}
/*
* Attempt to perform offload RAID mapping for a logical volume I/O.
*/
#define HPSA_RAID_0 0
#define HPSA_RAID_4 1
#define HPSA_RAID_1 2 /* also used for RAID 10 */
#define HPSA_RAID_5 3 /* also used for RAID 50 */
#define HPSA_RAID_51 4
#define HPSA_RAID_6 5 /* also used for RAID 60 */
#define HPSA_RAID_ADM 6 /* also used for RAID 1+0 ADM */
#define HPSA_RAID_MAX HPSA_RAID_ADM
#define HPSA_RAID_UNKNOWN 0xff
/* Subroutine used to parse the scsi opcode and build the CDB for RAID bypass*/
int check_for_scsi_opcode(uint8_t *cdb, boolean_t *is_write, uint64_t *fst_blk,
uint32_t *blk_cnt) {
switch (cdb[0]) {
case SCMD_WRITE_6:
*is_write = true;
case SCMD_READ_6:
*fst_blk = (uint64_t)(((cdb[1] & 0x1F) << 16) |
(cdb[2] << 8) | cdb[3]);
*blk_cnt = (uint32_t)cdb[4];
if (*blk_cnt == 0)
*blk_cnt = 256;
break;
case SCMD_WRITE_10:
*is_write = true;
case SCMD_READ_10:
*fst_blk = (uint64_t)GET_BE32(&cdb[2]);
*blk_cnt = (uint32_t)GET_BE16(&cdb[7]);
break;
case SCMD_WRITE_12:
*is_write = true;
case SCMD_READ_12:
*fst_blk = (uint64_t)GET_BE32(&cdb[2]);
*blk_cnt = GET_BE32(&cdb[6]);
break;
case SCMD_WRITE_16:
*is_write = true;
case SCMD_READ_16:
*fst_blk = GET_BE64(&cdb[2]);
*blk_cnt = GET_BE32(&cdb[10]);
break;
default:
/* Process via normal I/O path. */
return PQI_STATUS_FAILURE;
}
return PQI_STATUS_SUCCESS;
}
/*
* Function used to build and send RAID bypass request to the adapter
*/
int pqisrc_send_scsi_cmd_raidbypass(pqisrc_softstate_t *softs,
pqi_scsi_dev_t *device, rcb_t *rcb, uint8_t *cdb)
{
struct raid_map *raid_map;
boolean_t is_write = false;
uint32_t map_idx;
uint64_t fst_blk, lst_blk;
uint32_t blk_cnt, blks_per_row;
uint64_t fst_row, lst_row;
uint32_t fst_row_offset, lst_row_offset;
uint32_t fst_col, lst_col;
uint32_t r5or6_blks_per_row;
uint64_t r5or6_fst_row, r5or6_lst_row;
uint32_t r5or6_fst_row_offset, r5or6_lst_row_offset;
uint32_t r5or6_fst_col, r5or6_lst_col;
uint16_t data_disks_per_row, total_disks_per_row;
uint16_t layout_map_count;
uint32_t stripesz;
uint16_t strip_sz;
uint32_t fst_grp, lst_grp, cur_grp;
uint32_t map_row;
uint64_t disk_block;
uint32_t disk_blk_cnt;
uint8_t cdb_length;
int offload_to_mirror;
int i;
DBG_FUNC(" IN \n");
DBG_IO("!!!!!\n");
/* Check for eligible opcode, get LBA and block count. */
memcpy(cdb, OS_GET_CDBP(rcb), rcb->cmdlen);
for(i = 0; i < rcb->cmdlen ; i++)
DBG_IO(" CDB [ %d ] : %x\n",i,cdb[i]);
if(check_for_scsi_opcode(cdb, &is_write,
&fst_blk, &blk_cnt) == PQI_STATUS_FAILURE)
return PQI_STATUS_FAILURE;
/* Check for write to non-RAID-0. */
if (is_write && device->raid_level != SA_RAID_0)
2020-03-30 16:04:25 +00:00
return PQI_STATUS_FAILURE;
if(blk_cnt == 0)
return PQI_STATUS_FAILURE;
lst_blk = fst_blk + blk_cnt - 1;
raid_map = device->raid_map;
/* Check for invalid block or wraparound. */
if (lst_blk >= GET_LE64((uint8_t *)&raid_map->volume_blk_cnt) ||
lst_blk < fst_blk)
return PQI_STATUS_FAILURE;
data_disks_per_row = GET_LE16((uint8_t *)&raid_map->data_disks_per_row);
strip_sz = GET_LE16((uint8_t *)(&raid_map->strip_size));
layout_map_count = GET_LE16((uint8_t *)(&raid_map->layout_map_count));
/* Calculate stripe information for the request. */
blks_per_row = data_disks_per_row * strip_sz;
if (!blks_per_row)
return PQI_STATUS_FAILURE;
/* use __udivdi3 ? */
fst_row = fst_blk / blks_per_row;
lst_row = lst_blk / blks_per_row;
fst_row_offset = (uint32_t)(fst_blk - (fst_row * blks_per_row));
lst_row_offset = (uint32_t)(lst_blk - (lst_row * blks_per_row));
fst_col = fst_row_offset / strip_sz;
lst_col = lst_row_offset / strip_sz;
/* If this isn't a single row/column then give to the controller. */
if (fst_row != lst_row || fst_col != lst_col)
return PQI_STATUS_FAILURE;
/* Proceeding with driver mapping. */
total_disks_per_row = data_disks_per_row +
GET_LE16((uint8_t *)(&raid_map->metadata_disks_per_row));
map_row = ((uint32_t)(fst_row >> raid_map->parity_rotation_shift)) %
GET_LE16((uint8_t *)(&raid_map->row_cnt));
map_idx = (map_row * total_disks_per_row) + fst_col;
/* RAID 1 */
if (device->raid_level == SA_RAID_1) {
if (device->offload_to_mirror)
map_idx += data_disks_per_row;
device->offload_to_mirror = !device->offload_to_mirror;
} else if (device->raid_level == SA_RAID_ADM) {
/* RAID ADM */
/*
* Handles N-way mirrors (R1-ADM) and R10 with # of drives
* divisible by 3.
*/
offload_to_mirror = device->offload_to_mirror;
if (offload_to_mirror == 0) {
/* use physical disk in the first mirrored group. */
map_idx %= data_disks_per_row;
} else {
do {
/*
* Determine mirror group that map_idx
* indicates.
*/
cur_grp = map_idx / data_disks_per_row;
if (offload_to_mirror != cur_grp) {
if (cur_grp <
layout_map_count - 1) {
/*
* Select raid index from
* next group.
*/
map_idx += data_disks_per_row;
cur_grp++;
} else {
/*
* Select raid index from first
* group.
*/
map_idx %= data_disks_per_row;
cur_grp = 0;
}
}
} while (offload_to_mirror != cur_grp);
}
/* Set mirror group to use next time. */
offload_to_mirror =
(offload_to_mirror >= layout_map_count - 1) ?
0 : offload_to_mirror + 1;
if(offload_to_mirror >= layout_map_count)
return PQI_STATUS_FAILURE;
device->offload_to_mirror = offload_to_mirror;
/*
* Avoid direct use of device->offload_to_mirror within this
* function since multiple threads might simultaneously
* increment it beyond the range of device->layout_map_count -1.
*/
} else if ((device->raid_level == SA_RAID_5 ||
device->raid_level == SA_RAID_6) && layout_map_count > 1) {
/* RAID 50/60 */
/* Verify first and last block are in same RAID group */
r5or6_blks_per_row = strip_sz * data_disks_per_row;
stripesz = r5or6_blks_per_row * layout_map_count;
fst_grp = (fst_blk % stripesz) / r5or6_blks_per_row;
lst_grp = (lst_blk % stripesz) / r5or6_blks_per_row;
if (fst_grp != lst_grp)
return PQI_STATUS_FAILURE;
/* Verify request is in a single row of RAID 5/6 */
fst_row = r5or6_fst_row =
fst_blk / stripesz;
r5or6_lst_row = lst_blk / stripesz;
if (r5or6_fst_row != r5or6_lst_row)
return PQI_STATUS_FAILURE;
/* Verify request is in a single column */
fst_row_offset = r5or6_fst_row_offset =
(uint32_t)((fst_blk % stripesz) %
r5or6_blks_per_row);
r5or6_lst_row_offset =
(uint32_t)((lst_blk % stripesz) %
r5or6_blks_per_row);
fst_col = r5or6_fst_row_offset / strip_sz;
r5or6_fst_col = fst_col;
r5or6_lst_col = r5or6_lst_row_offset / strip_sz;
if (r5or6_fst_col != r5or6_lst_col)
return PQI_STATUS_FAILURE;
/* Request is eligible */
map_row =
((uint32_t)(fst_row >> raid_map->parity_rotation_shift)) %
GET_LE16((uint8_t *)(&raid_map->row_cnt));
map_idx = (fst_grp *
(GET_LE16((uint8_t *)(&raid_map->row_cnt)) *
total_disks_per_row)) +
(map_row * total_disks_per_row) + fst_col;
}
if (map_idx >= RAID_MAP_MAX_ENTRIES)
return PQI_STATUS_FAILURE;
rcb->ioaccel_handle = raid_map->dev_data[map_idx].ioaccel_handle;
disk_block = GET_LE64((uint8_t *)(&raid_map->disk_starting_blk)) +
fst_row * strip_sz +
(fst_row_offset - fst_col * strip_sz);
disk_blk_cnt = blk_cnt;
/* Handle differing logical/physical block sizes. */
if (raid_map->phys_blk_shift) {
disk_block <<= raid_map->phys_blk_shift;
disk_blk_cnt <<= raid_map->phys_blk_shift;
}
if (disk_blk_cnt > 0xffff)
return PQI_STATUS_FAILURE;
/* Build the new CDB for the physical disk I/O. */
if (disk_block > 0xffffffff) {
cdb[0] = is_write ? SCMD_WRITE_16 : SCMD_READ_16;
cdb[1] = 0;
PUT_BE64(disk_block, &cdb[2]);
PUT_BE32(disk_blk_cnt, &cdb[10]);
cdb[14] = 0;
cdb[15] = 0;
cdb_length = 16;
} else {
cdb[0] = is_write ? SCMD_WRITE_10 : SCMD_READ_10;
cdb[1] = 0;
PUT_BE32(disk_block, &cdb[2]);
cdb[6] = 0;
PUT_BE16(disk_blk_cnt, &cdb[7]);
cdb[9] = 0;
cdb_length = 10;
}
if (GET_LE16((uint8_t *)(&raid_map->flags)) &
RAID_MAP_ENCRYPTION_ENABLED) {
pqisrc_set_enc_info(&rcb->enc_info, raid_map,
fst_blk);
rcb->encrypt_enable = true;
} else {
rcb->encrypt_enable = false;
}
rcb->cmdlen = cdb_length;
DBG_FUNC("OUT");
return PQI_STATUS_SUCCESS;
}
/* Function used to submit a TMF to the adater */
int pqisrc_send_tmf(pqisrc_softstate_t *softs, pqi_scsi_dev_t *devp,
rcb_t *rcb, int req_id, int tmf_type)
{
int rval = PQI_STATUS_SUCCESS;
pqi_tmf_req_t tmf_req;
memset(&tmf_req, 0, sizeof(pqi_tmf_req_t));
DBG_FUNC("IN");
tmf_req.header.iu_type = PQI_REQUEST_IU_TASK_MANAGEMENT;
tmf_req.header.iu_length = sizeof(tmf_req) - sizeof(iu_header_t);
tmf_req.req_id = rcb->tag;
memcpy(tmf_req.lun, devp->scsi3addr, sizeof(tmf_req.lun));
tmf_req.tmf = tmf_type;
tmf_req.req_id_to_manage = req_id;
tmf_req.resp_qid = OS_GET_TMF_RESP_QID(softs, rcb);
tmf_req.obq_id_to_manage = rcb->resp_qid;
rcb->req_pending = true;
rval = pqisrc_submit_cmnd(softs,
&softs->op_raid_ib_q[OS_GET_TMF_REQ_QINDEX(softs, rcb)], &tmf_req);
if (rval != PQI_STATUS_SUCCESS) {
DBG_ERR("Unable to submit command rval=%d\n", rval);
return rval;
}
rval = pqisrc_wait_on_condition(softs, rcb);
if (rval != PQI_STATUS_SUCCESS){
DBG_ERR("Task Management tmf_type : %d timeout\n", tmf_type);
rcb->status = REQUEST_FAILED;
}
if (rcb->status != REQUEST_SUCCESS) {
DBG_ERR_BTL(devp, "Task Management failed tmf_type:%d "
"stat:0x%x\n", tmf_type, rcb->status);
rval = PQI_STATUS_FAILURE;
}
DBG_FUNC("OUT");
return rval;
}