linux-zen-desktop/drivers/scsi/elx/libefc_sli/sli4.c

5160 lines
131 KiB
C
Raw Normal View History

2023-08-30 17:31:07 +02:00
// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2021 Broadcom. All Rights Reserved. The term
* Broadcom refers to Broadcom Inc. and/or its subsidiaries.
*/
/*
* All common (i.e. transport-independent) SLI-4 functions are implemented
* in this file.
*/
#include "sli4.h"
static struct sli4_asic_entry_t sli4_asic_table[] = {
{ SLI4_ASIC_REV_B0, SLI4_ASIC_GEN_5},
{ SLI4_ASIC_REV_D0, SLI4_ASIC_GEN_5},
{ SLI4_ASIC_REV_A3, SLI4_ASIC_GEN_6},
{ SLI4_ASIC_REV_A0, SLI4_ASIC_GEN_6},
{ SLI4_ASIC_REV_A1, SLI4_ASIC_GEN_6},
{ SLI4_ASIC_REV_A3, SLI4_ASIC_GEN_6},
{ SLI4_ASIC_REV_A1, SLI4_ASIC_GEN_7},
{ SLI4_ASIC_REV_A0, SLI4_ASIC_GEN_7},
};
/* Convert queue type enum (SLI_QTYPE_*) into a string */
static char *SLI4_QNAME[] = {
"Event Queue",
"Completion Queue",
"Mailbox Queue",
"Work Queue",
"Receive Queue",
"Undefined"
};
/**
* sli_config_cmd_init() - Write a SLI_CONFIG command to the provided buffer.
*
* @sli4: SLI context pointer.
* @buf: Destination buffer for the command.
* @length: Length in bytes of attached command.
* @dma: DMA buffer for non-embedded commands.
* Return: Command payload buffer.
*/
static void *
sli_config_cmd_init(struct sli4 *sli4, void *buf, u32 length,
struct efc_dma *dma)
{
struct sli4_cmd_sli_config *config;
u32 flags;
if (length > sizeof(config->payload.embed) && !dma) {
efc_log_err(sli4, "Too big for an embedded cmd with len(%d)\n",
length);
return NULL;
}
memset(buf, 0, SLI4_BMBX_SIZE);
config = buf;
config->hdr.command = SLI4_MBX_CMD_SLI_CONFIG;
if (!dma) {
flags = SLI4_SLICONF_EMB;
config->dw1_flags = cpu_to_le32(flags);
config->payload_len = cpu_to_le32(length);
return config->payload.embed;
}
flags = SLI4_SLICONF_PMDCMD_VAL_1;
flags &= ~SLI4_SLICONF_EMB;
config->dw1_flags = cpu_to_le32(flags);
config->payload.mem.addr.low = cpu_to_le32(lower_32_bits(dma->phys));
config->payload.mem.addr.high = cpu_to_le32(upper_32_bits(dma->phys));
config->payload.mem.length =
cpu_to_le32(dma->size & SLI4_SLICONF_PMD_LEN);
config->payload_len = cpu_to_le32(dma->size);
/* save pointer to DMA for BMBX dumping purposes */
sli4->bmbx_non_emb_pmd = dma;
return dma->virt;
}
/**
* sli_cmd_common_create_cq() - Write a COMMON_CREATE_CQ V2 command.
*
* @sli4: SLI context pointer.
* @buf: Destination buffer for the command.
* @qmem: DMA memory for queue.
* @eq_id: EQ id assosiated with this cq.
* Return: status -EIO/0.
*/
static int
sli_cmd_common_create_cq(struct sli4 *sli4, void *buf, struct efc_dma *qmem,
u16 eq_id)
{
struct sli4_rqst_cmn_create_cq_v2 *cqv2 = NULL;
u32 p;
uintptr_t addr;
u32 num_pages = 0;
size_t cmd_size = 0;
u32 page_size = 0;
u32 n_cqe = 0;
u32 dw5_flags = 0;
u16 dw6w1_arm = 0;
__le32 len;
/* First calculate number of pages and the mailbox cmd length */
n_cqe = qmem->size / SLI4_CQE_BYTES;
switch (n_cqe) {
case 256:
case 512:
case 1024:
case 2048:
page_size = SZ_4K;
break;
case 4096:
page_size = SZ_8K;
break;
default:
return -EIO;
}
num_pages = sli_page_count(qmem->size, page_size);
cmd_size = SLI4_RQST_CMDSZ(cmn_create_cq_v2)
+ SZ_DMAADDR * num_pages;
cqv2 = sli_config_cmd_init(sli4, buf, cmd_size, NULL);
if (!cqv2)
return -EIO;
len = SLI4_RQST_PYLD_LEN_VAR(cmn_create_cq_v2, SZ_DMAADDR * num_pages);
sli_cmd_fill_hdr(&cqv2->hdr, SLI4_CMN_CREATE_CQ, SLI4_SUBSYSTEM_COMMON,
CMD_V2, len);
cqv2->page_size = page_size / SLI_PAGE_SIZE;
/* valid values for number of pages: 1, 2, 4, 8 (sec 4.4.3) */
cqv2->num_pages = cpu_to_le16(num_pages);
if (!num_pages || num_pages > SLI4_CREATE_CQV2_MAX_PAGES)
return -EIO;
switch (num_pages) {
case 1:
dw5_flags |= SLI4_CQ_CNT_VAL(256);
break;
case 2:
dw5_flags |= SLI4_CQ_CNT_VAL(512);
break;
case 4:
dw5_flags |= SLI4_CQ_CNT_VAL(1024);
break;
case 8:
dw5_flags |= SLI4_CQ_CNT_VAL(LARGE);
cqv2->cqe_count = cpu_to_le16(n_cqe);
break;
default:
efc_log_err(sli4, "num_pages %d not valid\n", num_pages);
return -EIO;
}
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
dw5_flags |= SLI4_CREATE_CQV2_AUTOVALID;
dw5_flags |= SLI4_CREATE_CQV2_EVT;
dw5_flags |= SLI4_CREATE_CQV2_VALID;
cqv2->dw5_flags = cpu_to_le32(dw5_flags);
cqv2->dw6w1_arm = cpu_to_le16(dw6w1_arm);
cqv2->eq_id = cpu_to_le16(eq_id);
for (p = 0, addr = qmem->phys; p < num_pages; p++, addr += page_size) {
cqv2->page_phys_addr[p].low = cpu_to_le32(lower_32_bits(addr));
cqv2->page_phys_addr[p].high = cpu_to_le32(upper_32_bits(addr));
}
return 0;
}
static int
sli_cmd_common_create_eq(struct sli4 *sli4, void *buf, struct efc_dma *qmem)
{
struct sli4_rqst_cmn_create_eq *eq;
u32 p;
uintptr_t addr;
u16 num_pages;
u32 dw5_flags = 0;
u32 dw6_flags = 0, ver;
eq = sli_config_cmd_init(sli4, buf, SLI4_CFG_PYLD_LENGTH(cmn_create_eq),
NULL);
if (!eq)
return -EIO;
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
ver = CMD_V2;
else
ver = CMD_V0;
sli_cmd_fill_hdr(&eq->hdr, SLI4_CMN_CREATE_EQ, SLI4_SUBSYSTEM_COMMON,
ver, SLI4_RQST_PYLD_LEN(cmn_create_eq));
/* valid values for number of pages: 1, 2, 4 (sec 4.4.3) */
num_pages = qmem->size / SLI_PAGE_SIZE;
eq->num_pages = cpu_to_le16(num_pages);
switch (num_pages) {
case 1:
dw5_flags |= SLI4_EQE_SIZE_4;
dw6_flags |= SLI4_EQ_CNT_VAL(1024);
break;
case 2:
dw5_flags |= SLI4_EQE_SIZE_4;
dw6_flags |= SLI4_EQ_CNT_VAL(2048);
break;
case 4:
dw5_flags |= SLI4_EQE_SIZE_4;
dw6_flags |= SLI4_EQ_CNT_VAL(4096);
break;
default:
efc_log_err(sli4, "num_pages %d not valid\n", num_pages);
return -EIO;
}
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
dw5_flags |= SLI4_CREATE_EQ_AUTOVALID;
dw5_flags |= SLI4_CREATE_EQ_VALID;
dw6_flags &= (~SLI4_CREATE_EQ_ARM);
eq->dw5_flags = cpu_to_le32(dw5_flags);
eq->dw6_flags = cpu_to_le32(dw6_flags);
eq->dw7_delaymulti = cpu_to_le32(SLI4_CREATE_EQ_DELAYMULTI);
for (p = 0, addr = qmem->phys; p < num_pages;
p++, addr += SLI_PAGE_SIZE) {
eq->page_address[p].low = cpu_to_le32(lower_32_bits(addr));
eq->page_address[p].high = cpu_to_le32(upper_32_bits(addr));
}
return 0;
}
static int
sli_cmd_common_create_mq_ext(struct sli4 *sli4, void *buf, struct efc_dma *qmem,
u16 cq_id)
{
struct sli4_rqst_cmn_create_mq_ext *mq;
u32 p;
uintptr_t addr;
u32 num_pages;
u16 dw6w1_flags = 0;
mq = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(cmn_create_mq_ext), NULL);
if (!mq)
return -EIO;
sli_cmd_fill_hdr(&mq->hdr, SLI4_CMN_CREATE_MQ_EXT,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_create_mq_ext));
/* valid values for number of pages: 1, 2, 4, 8 (sec 4.4.12) */
num_pages = qmem->size / SLI_PAGE_SIZE;
mq->num_pages = cpu_to_le16(num_pages);
switch (num_pages) {
case 1:
dw6w1_flags |= SLI4_MQE_SIZE_16;
break;
case 2:
dw6w1_flags |= SLI4_MQE_SIZE_32;
break;
case 4:
dw6w1_flags |= SLI4_MQE_SIZE_64;
break;
case 8:
dw6w1_flags |= SLI4_MQE_SIZE_128;
break;
default:
efc_log_info(sli4, "num_pages %d not valid\n", num_pages);
return -EIO;
}
mq->async_event_bitmap = cpu_to_le32(SLI4_ASYNC_EVT_FC_ALL);
if (sli4->params.mq_create_version) {
mq->cq_id_v1 = cpu_to_le16(cq_id);
mq->hdr.dw3_version = cpu_to_le32(CMD_V1);
} else {
dw6w1_flags |= (cq_id << SLI4_CREATE_MQEXT_CQID_SHIFT);
}
mq->dw7_val = cpu_to_le32(SLI4_CREATE_MQEXT_VAL);
mq->dw6w1_flags = cpu_to_le16(dw6w1_flags);
for (p = 0, addr = qmem->phys; p < num_pages;
p++, addr += SLI_PAGE_SIZE) {
mq->page_phys_addr[p].low = cpu_to_le32(lower_32_bits(addr));
mq->page_phys_addr[p].high = cpu_to_le32(upper_32_bits(addr));
}
return 0;
}
int
sli_cmd_wq_create(struct sli4 *sli4, void *buf, struct efc_dma *qmem, u16 cq_id)
{
struct sli4_rqst_wq_create *wq;
u32 p;
uintptr_t addr;
u32 page_size = 0;
u32 n_wqe = 0;
u16 num_pages;
wq = sli_config_cmd_init(sli4, buf, SLI4_CFG_PYLD_LENGTH(wq_create),
NULL);
if (!wq)
return -EIO;
sli_cmd_fill_hdr(&wq->hdr, SLI4_OPC_WQ_CREATE, SLI4_SUBSYSTEM_FC,
CMD_V1, SLI4_RQST_PYLD_LEN(wq_create));
n_wqe = qmem->size / sli4->wqe_size;
switch (qmem->size) {
case 4096:
case 8192:
case 16384:
case 32768:
page_size = SZ_4K;
break;
case 65536:
page_size = SZ_8K;
break;
case 131072:
page_size = SZ_16K;
break;
case 262144:
page_size = SZ_32K;
break;
case 524288:
page_size = SZ_64K;
break;
default:
return -EIO;
}
/* valid values for number of pages(num_pages): 1-8 */
num_pages = sli_page_count(qmem->size, page_size);
wq->num_pages = cpu_to_le16(num_pages);
if (!num_pages || num_pages > SLI4_WQ_CREATE_MAX_PAGES)
return -EIO;
wq->cq_id = cpu_to_le16(cq_id);
wq->page_size = page_size / SLI_PAGE_SIZE;
if (sli4->wqe_size == SLI4_WQE_EXT_BYTES)
wq->wqe_size_byte |= SLI4_WQE_EXT_SIZE;
else
wq->wqe_size_byte |= SLI4_WQE_SIZE;
wq->wqe_count = cpu_to_le16(n_wqe);
for (p = 0, addr = qmem->phys; p < num_pages; p++, addr += page_size) {
wq->page_phys_addr[p].low = cpu_to_le32(lower_32_bits(addr));
wq->page_phys_addr[p].high = cpu_to_le32(upper_32_bits(addr));
}
return 0;
}
static int
sli_cmd_rq_create_v1(struct sli4 *sli4, void *buf, struct efc_dma *qmem,
u16 cq_id, u16 buffer_size)
{
struct sli4_rqst_rq_create_v1 *rq;
u32 p;
uintptr_t addr;
u32 num_pages;
rq = sli_config_cmd_init(sli4, buf, SLI4_CFG_PYLD_LENGTH(rq_create_v1),
NULL);
if (!rq)
return -EIO;
sli_cmd_fill_hdr(&rq->hdr, SLI4_OPC_RQ_CREATE, SLI4_SUBSYSTEM_FC,
CMD_V1, SLI4_RQST_PYLD_LEN(rq_create_v1));
/* Disable "no buffer warnings" to avoid Lancer bug */
rq->dim_dfd_dnb |= SLI4_RQ_CREATE_V1_DNB;
/* valid values for number of pages: 1-8 (sec 4.5.6) */
num_pages = sli_page_count(qmem->size, SLI_PAGE_SIZE);
rq->num_pages = cpu_to_le16(num_pages);
if (!num_pages ||
num_pages > SLI4_RQ_CREATE_V1_MAX_PAGES) {
efc_log_info(sli4, "num_pages %d not valid, max %d\n",
num_pages, SLI4_RQ_CREATE_V1_MAX_PAGES);
return -EIO;
}
/*
* RQE count is the total number of entries (note not lg2(# entries))
*/
rq->rqe_count = cpu_to_le16(qmem->size / SLI4_RQE_SIZE);
rq->rqe_size_byte |= SLI4_RQE_SIZE_8;
rq->page_size = SLI4_RQ_PAGE_SIZE_4096;
if (buffer_size < sli4->rq_min_buf_size ||
buffer_size > sli4->rq_max_buf_size) {
efc_log_err(sli4, "buffer_size %d out of range (%d-%d)\n",
buffer_size, sli4->rq_min_buf_size,
sli4->rq_max_buf_size);
return -EIO;
}
rq->buffer_size = cpu_to_le32(buffer_size);
rq->cq_id = cpu_to_le16(cq_id);
for (p = 0, addr = qmem->phys;
p < num_pages;
p++, addr += SLI_PAGE_SIZE) {
rq->page_phys_addr[p].low = cpu_to_le32(lower_32_bits(addr));
rq->page_phys_addr[p].high = cpu_to_le32(upper_32_bits(addr));
}
return 0;
}
static int
sli_cmd_rq_create_v2(struct sli4 *sli4, u32 num_rqs,
struct sli4_queue *qs[], u32 base_cq_id,
u32 header_buffer_size,
u32 payload_buffer_size, struct efc_dma *dma)
{
struct sli4_rqst_rq_create_v2 *req = NULL;
u32 i, p, offset = 0;
u32 payload_size, page_count;
uintptr_t addr;
u32 num_pages;
__le32 len;
page_count = sli_page_count(qs[0]->dma.size, SLI_PAGE_SIZE) * num_rqs;
/* Payload length must accommodate both request and response */
payload_size = max(SLI4_RQST_CMDSZ(rq_create_v2) +
SZ_DMAADDR * page_count,
sizeof(struct sli4_rsp_cmn_create_queue_set));
dma->size = payload_size;
dma->virt = dma_alloc_coherent(&sli4->pci->dev, dma->size,
&dma->phys, GFP_KERNEL);
if (!dma->virt)
return -EIO;
memset(dma->virt, 0, payload_size);
req = sli_config_cmd_init(sli4, sli4->bmbx.virt, payload_size, dma);
if (!req)
return -EIO;
len = SLI4_RQST_PYLD_LEN_VAR(rq_create_v2, SZ_DMAADDR * page_count);
sli_cmd_fill_hdr(&req->hdr, SLI4_OPC_RQ_CREATE, SLI4_SUBSYSTEM_FC,
CMD_V2, len);
/* Fill Payload fields */
req->dim_dfd_dnb |= SLI4_RQCREATEV2_DNB;
num_pages = sli_page_count(qs[0]->dma.size, SLI_PAGE_SIZE);
req->num_pages = cpu_to_le16(num_pages);
req->rqe_count = cpu_to_le16(qs[0]->dma.size / SLI4_RQE_SIZE);
req->rqe_size_byte |= SLI4_RQE_SIZE_8;
req->page_size = SLI4_RQ_PAGE_SIZE_4096;
req->rq_count = num_rqs;
req->base_cq_id = cpu_to_le16(base_cq_id);
req->hdr_buffer_size = cpu_to_le16(header_buffer_size);
req->payload_buffer_size = cpu_to_le16(payload_buffer_size);
for (i = 0; i < num_rqs; i++) {
for (p = 0, addr = qs[i]->dma.phys; p < num_pages;
p++, addr += SLI_PAGE_SIZE) {
req->page_phys_addr[offset].low =
cpu_to_le32(lower_32_bits(addr));
req->page_phys_addr[offset].high =
cpu_to_le32(upper_32_bits(addr));
offset++;
}
}
return 0;
}
static void
__sli_queue_destroy(struct sli4 *sli4, struct sli4_queue *q)
{
if (!q->dma.size)
return;
dma_free_coherent(&sli4->pci->dev, q->dma.size,
q->dma.virt, q->dma.phys);
memset(&q->dma, 0, sizeof(struct efc_dma));
}
int
__sli_queue_init(struct sli4 *sli4, struct sli4_queue *q, u32 qtype,
size_t size, u32 n_entries, u32 align)
{
if (q->dma.virt) {
efc_log_err(sli4, "%s failed\n", __func__);
return -EIO;
}
memset(q, 0, sizeof(struct sli4_queue));
q->dma.size = size * n_entries;
q->dma.virt = dma_alloc_coherent(&sli4->pci->dev, q->dma.size,
&q->dma.phys, GFP_KERNEL);
if (!q->dma.virt) {
memset(&q->dma, 0, sizeof(struct efc_dma));
efc_log_err(sli4, "%s allocation failed\n", SLI4_QNAME[qtype]);
return -EIO;
}
memset(q->dma.virt, 0, size * n_entries);
spin_lock_init(&q->lock);
q->type = qtype;
q->size = size;
q->length = n_entries;
if (q->type == SLI4_QTYPE_EQ || q->type == SLI4_QTYPE_CQ) {
/* For prism, phase will be flipped after
* a sweep through eq and cq
*/
q->phase = 1;
}
/* Limit to hwf the queue size per interrupt */
q->proc_limit = n_entries / 2;
if (q->type == SLI4_QTYPE_EQ)
q->posted_limit = q->length / 2;
else
q->posted_limit = 64;
return 0;
}
int
sli_fc_rq_alloc(struct sli4 *sli4, struct sli4_queue *q,
u32 n_entries, u32 buffer_size,
struct sli4_queue *cq, bool is_hdr)
{
if (__sli_queue_init(sli4, q, SLI4_QTYPE_RQ, SLI4_RQE_SIZE,
n_entries, SLI_PAGE_SIZE))
return -EIO;
if (sli_cmd_rq_create_v1(sli4, sli4->bmbx.virt, &q->dma, cq->id,
buffer_size))
goto error;
if (__sli_create_queue(sli4, q))
goto error;
if (is_hdr && q->id & 1) {
efc_log_info(sli4, "bad header RQ_ID %d\n", q->id);
goto error;
} else if (!is_hdr && (q->id & 1) == 0) {
efc_log_info(sli4, "bad data RQ_ID %d\n", q->id);
goto error;
}
if (is_hdr)
q->u.flag |= SLI4_QUEUE_FLAG_HDR;
else
q->u.flag &= ~SLI4_QUEUE_FLAG_HDR;
return 0;
error:
__sli_queue_destroy(sli4, q);
return -EIO;
}
int
sli_fc_rq_set_alloc(struct sli4 *sli4, u32 num_rq_pairs,
struct sli4_queue *qs[], u32 base_cq_id,
u32 n_entries, u32 header_buffer_size,
u32 payload_buffer_size)
{
u32 i;
struct efc_dma dma = {0};
struct sli4_rsp_cmn_create_queue_set *rsp = NULL;
void __iomem *db_regaddr = NULL;
u32 num_rqs = num_rq_pairs * 2;
for (i = 0; i < num_rqs; i++) {
if (__sli_queue_init(sli4, qs[i], SLI4_QTYPE_RQ,
SLI4_RQE_SIZE, n_entries,
SLI_PAGE_SIZE)) {
goto error;
}
}
if (sli_cmd_rq_create_v2(sli4, num_rqs, qs, base_cq_id,
header_buffer_size, payload_buffer_size,
&dma)) {
goto error;
}
if (sli_bmbx_command(sli4)) {
efc_log_err(sli4, "bootstrap mailbox write failed RQSet\n");
goto error;
}
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
db_regaddr = sli4->reg[1] + SLI4_IF6_RQ_DB_REG;
else
db_regaddr = sli4->reg[0] + SLI4_RQ_DB_REG;
rsp = dma.virt;
if (rsp->hdr.status) {
efc_log_err(sli4, "bad create RQSet status=%#x addl=%#x\n",
rsp->hdr.status, rsp->hdr.additional_status);
goto error;
}
for (i = 0; i < num_rqs; i++) {
qs[i]->id = i + le16_to_cpu(rsp->q_id);
if ((qs[i]->id & 1) == 0)
qs[i]->u.flag |= SLI4_QUEUE_FLAG_HDR;
else
qs[i]->u.flag &= ~SLI4_QUEUE_FLAG_HDR;
qs[i]->db_regaddr = db_regaddr;
}
dma_free_coherent(&sli4->pci->dev, dma.size, dma.virt, dma.phys);
return 0;
error:
for (i = 0; i < num_rqs; i++)
__sli_queue_destroy(sli4, qs[i]);
if (dma.virt)
dma_free_coherent(&sli4->pci->dev, dma.size, dma.virt,
dma.phys);
return -EIO;
}
static int
sli_res_sli_config(struct sli4 *sli4, void *buf)
{
struct sli4_cmd_sli_config *sli_config = buf;
/* sanity check */
if (!buf || sli_config->hdr.command !=
SLI4_MBX_CMD_SLI_CONFIG) {
efc_log_err(sli4, "bad parameter buf=%p cmd=%#x\n", buf,
buf ? sli_config->hdr.command : -1);
return -EIO;
}
if (le16_to_cpu(sli_config->hdr.status))
return le16_to_cpu(sli_config->hdr.status);
if (le32_to_cpu(sli_config->dw1_flags) & SLI4_SLICONF_EMB)
return sli_config->payload.embed[4];
efc_log_info(sli4, "external buffers not supported\n");
return -EIO;
}
int
__sli_create_queue(struct sli4 *sli4, struct sli4_queue *q)
{
struct sli4_rsp_cmn_create_queue *res_q = NULL;
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail %s\n",
SLI4_QNAME[q->type]);
return -EIO;
}
if (sli_res_sli_config(sli4, sli4->bmbx.virt)) {
efc_log_err(sli4, "bad status create %s\n",
SLI4_QNAME[q->type]);
return -EIO;
}
res_q = (void *)((u8 *)sli4->bmbx.virt +
offsetof(struct sli4_cmd_sli_config, payload));
if (res_q->hdr.status) {
efc_log_err(sli4, "bad create %s status=%#x addl=%#x\n",
SLI4_QNAME[q->type], res_q->hdr.status,
res_q->hdr.additional_status);
return -EIO;
}
q->id = le16_to_cpu(res_q->q_id);
switch (q->type) {
case SLI4_QTYPE_EQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
q->db_regaddr = sli4->reg[1] + SLI4_IF6_EQ_DB_REG;
else
q->db_regaddr = sli4->reg[0] + SLI4_EQCQ_DB_REG;
break;
case SLI4_QTYPE_CQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
q->db_regaddr = sli4->reg[1] + SLI4_IF6_CQ_DB_REG;
else
q->db_regaddr = sli4->reg[0] + SLI4_EQCQ_DB_REG;
break;
case SLI4_QTYPE_MQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
q->db_regaddr = sli4->reg[1] + SLI4_IF6_MQ_DB_REG;
else
q->db_regaddr = sli4->reg[0] + SLI4_MQ_DB_REG;
break;
case SLI4_QTYPE_RQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
q->db_regaddr = sli4->reg[1] + SLI4_IF6_RQ_DB_REG;
else
q->db_regaddr = sli4->reg[0] + SLI4_RQ_DB_REG;
break;
case SLI4_QTYPE_WQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
q->db_regaddr = sli4->reg[1] + SLI4_IF6_WQ_DB_REG;
else
q->db_regaddr = sli4->reg[0] + SLI4_IO_WQ_DB_REG;
break;
default:
break;
}
return 0;
}
int
sli_get_queue_entry_size(struct sli4 *sli4, u32 qtype)
{
u32 size = 0;
switch (qtype) {
case SLI4_QTYPE_EQ:
size = sizeof(u32);
break;
case SLI4_QTYPE_CQ:
size = 16;
break;
case SLI4_QTYPE_MQ:
size = 256;
break;
case SLI4_QTYPE_WQ:
size = sli4->wqe_size;
break;
case SLI4_QTYPE_RQ:
size = SLI4_RQE_SIZE;
break;
default:
efc_log_info(sli4, "unknown queue type %d\n", qtype);
return -1;
}
return size;
}
int
sli_queue_alloc(struct sli4 *sli4, u32 qtype,
struct sli4_queue *q, u32 n_entries,
struct sli4_queue *assoc)
{
int size;
u32 align = 0;
/* get queue size */
size = sli_get_queue_entry_size(sli4, qtype);
if (size < 0)
return -EIO;
align = SLI_PAGE_SIZE;
if (__sli_queue_init(sli4, q, qtype, size, n_entries, align))
return -EIO;
switch (qtype) {
case SLI4_QTYPE_EQ:
if (!sli_cmd_common_create_eq(sli4, sli4->bmbx.virt, &q->dma) &&
!__sli_create_queue(sli4, q))
return 0;
break;
case SLI4_QTYPE_CQ:
if (!sli_cmd_common_create_cq(sli4, sli4->bmbx.virt, &q->dma,
assoc ? assoc->id : 0) &&
!__sli_create_queue(sli4, q))
return 0;
break;
case SLI4_QTYPE_MQ:
assoc->u.flag |= SLI4_QUEUE_FLAG_MQ;
if (!sli_cmd_common_create_mq_ext(sli4, sli4->bmbx.virt,
&q->dma, assoc->id) &&
!__sli_create_queue(sli4, q))
return 0;
break;
case SLI4_QTYPE_WQ:
if (!sli_cmd_wq_create(sli4, sli4->bmbx.virt, &q->dma,
assoc ? assoc->id : 0) &&
!__sli_create_queue(sli4, q))
return 0;
break;
default:
efc_log_info(sli4, "unknown queue type %d\n", qtype);
}
__sli_queue_destroy(sli4, q);
return -EIO;
}
static int sli_cmd_cq_set_create(struct sli4 *sli4,
struct sli4_queue *qs[], u32 num_cqs,
struct sli4_queue *eqs[],
struct efc_dma *dma)
{
struct sli4_rqst_cmn_create_cq_set_v0 *req = NULL;
uintptr_t addr;
u32 i, offset = 0, page_bytes = 0, payload_size;
u32 p = 0, page_size = 0, n_cqe = 0, num_pages_cq;
u32 dw5_flags = 0;
u16 dw6w1_flags = 0;
__le32 req_len;
n_cqe = qs[0]->dma.size / SLI4_CQE_BYTES;
switch (n_cqe) {
case 256:
case 512:
case 1024:
case 2048:
page_size = 1;
break;
case 4096:
page_size = 2;
break;
default:
return -EIO;
}
page_bytes = page_size * SLI_PAGE_SIZE;
num_pages_cq = sli_page_count(qs[0]->dma.size, page_bytes);
payload_size = max(SLI4_RQST_CMDSZ(cmn_create_cq_set_v0) +
(SZ_DMAADDR * num_pages_cq * num_cqs),
sizeof(struct sli4_rsp_cmn_create_queue_set));
dma->size = payload_size;
dma->virt = dma_alloc_coherent(&sli4->pci->dev, dma->size,
&dma->phys, GFP_KERNEL);
if (!dma->virt)
return -EIO;
memset(dma->virt, 0, payload_size);
req = sli_config_cmd_init(sli4, sli4->bmbx.virt, payload_size, dma);
if (!req)
return -EIO;
req_len = SLI4_RQST_PYLD_LEN_VAR(cmn_create_cq_set_v0,
SZ_DMAADDR * num_pages_cq * num_cqs);
sli_cmd_fill_hdr(&req->hdr, SLI4_CMN_CREATE_CQ_SET, SLI4_SUBSYSTEM_FC,
CMD_V0, req_len);
req->page_size = page_size;
req->num_pages = cpu_to_le16(num_pages_cq);
switch (num_pages_cq) {
case 1:
dw5_flags |= SLI4_CQ_CNT_VAL(256);
break;
case 2:
dw5_flags |= SLI4_CQ_CNT_VAL(512);
break;
case 4:
dw5_flags |= SLI4_CQ_CNT_VAL(1024);
break;
case 8:
dw5_flags |= SLI4_CQ_CNT_VAL(LARGE);
dw6w1_flags |= (n_cqe & SLI4_CREATE_CQSETV0_CQE_COUNT);
break;
default:
efc_log_info(sli4, "num_pages %d not valid\n", num_pages_cq);
return -EIO;
}
dw5_flags |= SLI4_CREATE_CQSETV0_EVT;
dw5_flags |= SLI4_CREATE_CQSETV0_VALID;
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
dw5_flags |= SLI4_CREATE_CQSETV0_AUTOVALID;
dw6w1_flags &= ~SLI4_CREATE_CQSETV0_ARM;
req->dw5_flags = cpu_to_le32(dw5_flags);
req->dw6w1_flags = cpu_to_le16(dw6w1_flags);
req->num_cq_req = cpu_to_le16(num_cqs);
/* Fill page addresses of all the CQs. */
for (i = 0; i < num_cqs; i++) {
req->eq_id[i] = cpu_to_le16(eqs[i]->id);
for (p = 0, addr = qs[i]->dma.phys; p < num_pages_cq;
p++, addr += page_bytes) {
req->page_phys_addr[offset].low =
cpu_to_le32(lower_32_bits(addr));
req->page_phys_addr[offset].high =
cpu_to_le32(upper_32_bits(addr));
offset++;
}
}
return 0;
}
int
sli_cq_alloc_set(struct sli4 *sli4, struct sli4_queue *qs[],
u32 num_cqs, u32 n_entries, struct sli4_queue *eqs[])
{
u32 i;
struct efc_dma dma = {0};
struct sli4_rsp_cmn_create_queue_set *res;
void __iomem *db_regaddr;
/* Align the queue DMA memory */
for (i = 0; i < num_cqs; i++) {
if (__sli_queue_init(sli4, qs[i], SLI4_QTYPE_CQ, SLI4_CQE_BYTES,
n_entries, SLI_PAGE_SIZE))
goto error;
}
if (sli_cmd_cq_set_create(sli4, qs, num_cqs, eqs, &dma))
goto error;
if (sli_bmbx_command(sli4))
goto error;
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
db_regaddr = sli4->reg[1] + SLI4_IF6_CQ_DB_REG;
else
db_regaddr = sli4->reg[0] + SLI4_EQCQ_DB_REG;
res = dma.virt;
if (res->hdr.status) {
efc_log_err(sli4, "bad create CQSet status=%#x addl=%#x\n",
res->hdr.status, res->hdr.additional_status);
goto error;
}
/* Check if we got all requested CQs. */
if (le16_to_cpu(res->num_q_allocated) != num_cqs) {
efc_log_crit(sli4, "Requested count CQs doesn't match.\n");
goto error;
}
/* Fill the resp cq ids. */
for (i = 0; i < num_cqs; i++) {
qs[i]->id = le16_to_cpu(res->q_id) + i;
qs[i]->db_regaddr = db_regaddr;
}
dma_free_coherent(&sli4->pci->dev, dma.size, dma.virt, dma.phys);
return 0;
error:
for (i = 0; i < num_cqs; i++)
__sli_queue_destroy(sli4, qs[i]);
if (dma.virt)
dma_free_coherent(&sli4->pci->dev, dma.size, dma.virt,
dma.phys);
return -EIO;
}
static int
sli_cmd_common_destroy_q(struct sli4 *sli4, u8 opc, u8 subsystem, u16 q_id)
{
struct sli4_rqst_cmn_destroy_q *req;
/* Payload length must accommodate both request and response */
req = sli_config_cmd_init(sli4, sli4->bmbx.virt,
SLI4_CFG_PYLD_LENGTH(cmn_destroy_q), NULL);
if (!req)
return -EIO;
sli_cmd_fill_hdr(&req->hdr, opc, subsystem,
CMD_V0, SLI4_RQST_PYLD_LEN(cmn_destroy_q));
req->q_id = cpu_to_le16(q_id);
return 0;
}
int
sli_queue_free(struct sli4 *sli4, struct sli4_queue *q,
u32 destroy_queues, u32 free_memory)
{
int rc = 0;
u8 opcode, subsystem;
struct sli4_rsp_hdr *res;
if (!q) {
efc_log_err(sli4, "bad parameter sli4=%p q=%p\n", sli4, q);
return -EIO;
}
if (!destroy_queues)
goto free_mem;
switch (q->type) {
case SLI4_QTYPE_EQ:
opcode = SLI4_CMN_DESTROY_EQ;
subsystem = SLI4_SUBSYSTEM_COMMON;
break;
case SLI4_QTYPE_CQ:
opcode = SLI4_CMN_DESTROY_CQ;
subsystem = SLI4_SUBSYSTEM_COMMON;
break;
case SLI4_QTYPE_MQ:
opcode = SLI4_CMN_DESTROY_MQ;
subsystem = SLI4_SUBSYSTEM_COMMON;
break;
case SLI4_QTYPE_WQ:
opcode = SLI4_OPC_WQ_DESTROY;
subsystem = SLI4_SUBSYSTEM_FC;
break;
case SLI4_QTYPE_RQ:
opcode = SLI4_OPC_RQ_DESTROY;
subsystem = SLI4_SUBSYSTEM_FC;
break;
default:
efc_log_info(sli4, "bad queue type %d\n", q->type);
rc = -EIO;
goto free_mem;
}
rc = sli_cmd_common_destroy_q(sli4, opcode, subsystem, q->id);
if (rc)
goto free_mem;
rc = sli_bmbx_command(sli4);
if (rc)
goto free_mem;
rc = sli_res_sli_config(sli4, sli4->bmbx.virt);
if (rc)
goto free_mem;
res = (void *)((u8 *)sli4->bmbx.virt +
offsetof(struct sli4_cmd_sli_config, payload));
if (res->status) {
efc_log_err(sli4, "destroy %s st=%#x addl=%#x\n",
SLI4_QNAME[q->type], res->status,
res->additional_status);
rc = -EIO;
goto free_mem;
}
free_mem:
if (free_memory)
__sli_queue_destroy(sli4, q);
return rc;
}
int
sli_queue_eq_arm(struct sli4 *sli4, struct sli4_queue *q, bool arm)
{
u32 val;
unsigned long flags = 0;
u32 a = arm ? SLI4_EQCQ_ARM : SLI4_EQCQ_UNARM;
spin_lock_irqsave(&q->lock, flags);
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
val = sli_format_if6_eq_db_data(q->n_posted, q->id, a);
else
val = sli_format_eq_db_data(q->n_posted, q->id, a);
writel(val, q->db_regaddr);
q->n_posted = 0;
spin_unlock_irqrestore(&q->lock, flags);
return 0;
}
int
sli_queue_arm(struct sli4 *sli4, struct sli4_queue *q, bool arm)
{
u32 val = 0;
unsigned long flags = 0;
u32 a = arm ? SLI4_EQCQ_ARM : SLI4_EQCQ_UNARM;
spin_lock_irqsave(&q->lock, flags);
switch (q->type) {
case SLI4_QTYPE_EQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
val = sli_format_if6_eq_db_data(q->n_posted, q->id, a);
else
val = sli_format_eq_db_data(q->n_posted, q->id, a);
writel(val, q->db_regaddr);
q->n_posted = 0;
break;
case SLI4_QTYPE_CQ:
if (sli4->if_type == SLI4_INTF_IF_TYPE_6)
val = sli_format_if6_cq_db_data(q->n_posted, q->id, a);
else
val = sli_format_cq_db_data(q->n_posted, q->id, a);
writel(val, q->db_regaddr);
q->n_posted = 0;
break;
default:
efc_log_info(sli4, "should only be used for EQ/CQ, not %s\n",
SLI4_QNAME[q->type]);
}
spin_unlock_irqrestore(&q->lock, flags);
return 0;
}
int
sli_wq_write(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
u32 qindex;
u32 val = 0;
qindex = q->index;
qe += q->index * q->size;
if (sli4->params.perf_wq_id_association)
sli_set_wq_id_association(entry, q->id);
memcpy(qe, entry, q->size);
val = sli_format_wq_db_data(q->id);
writel(val, q->db_regaddr);
q->index = (q->index + 1) & (q->length - 1);
return qindex;
}
int
sli_mq_write(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
u32 qindex;
u32 val = 0;
unsigned long flags;
spin_lock_irqsave(&q->lock, flags);
qindex = q->index;
qe += q->index * q->size;
memcpy(qe, entry, q->size);
val = sli_format_mq_db_data(q->id);
writel(val, q->db_regaddr);
q->index = (q->index + 1) & (q->length - 1);
spin_unlock_irqrestore(&q->lock, flags);
return qindex;
}
int
sli_rq_write(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
u32 qindex;
u32 val = 0;
qindex = q->index;
qe += q->index * q->size;
memcpy(qe, entry, q->size);
/*
* In RQ-pair, an RQ either contains the FC header
* (i.e. is_hdr == TRUE) or the payload.
*
* Don't ring doorbell for payload RQ
*/
if (!(q->u.flag & SLI4_QUEUE_FLAG_HDR))
goto skip;
val = sli_format_rq_db_data(q->id);
writel(val, q->db_regaddr);
skip:
q->index = (q->index + 1) & (q->length - 1);
return qindex;
}
int
sli_eq_read(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
unsigned long flags = 0;
u16 wflags = 0;
spin_lock_irqsave(&q->lock, flags);
qe += q->index * q->size;
/* Check if eqe is valid */
wflags = le16_to_cpu(((struct sli4_eqe *)qe)->dw0w0_flags);
if ((wflags & SLI4_EQE_VALID) != q->phase) {
spin_unlock_irqrestore(&q->lock, flags);
return -EIO;
}
if (sli4->if_type != SLI4_INTF_IF_TYPE_6) {
wflags &= ~SLI4_EQE_VALID;
((struct sli4_eqe *)qe)->dw0w0_flags = cpu_to_le16(wflags);
}
memcpy(entry, qe, q->size);
q->index = (q->index + 1) & (q->length - 1);
q->n_posted++;
/*
* For prism, the phase value will be used
* to check the validity of eq/cq entries.
* The value toggles after a complete sweep
* through the queue.
*/
if (sli4->if_type == SLI4_INTF_IF_TYPE_6 && q->index == 0)
q->phase ^= (u16)0x1;
spin_unlock_irqrestore(&q->lock, flags);
return 0;
}
int
sli_cq_read(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
unsigned long flags = 0;
u32 dwflags = 0;
bool valid_bit_set;
spin_lock_irqsave(&q->lock, flags);
qe += q->index * q->size;
/* Check if cqe is valid */
dwflags = le32_to_cpu(((struct sli4_mcqe *)qe)->dw3_flags);
valid_bit_set = (dwflags & SLI4_MCQE_VALID) != 0;
if (valid_bit_set != q->phase) {
spin_unlock_irqrestore(&q->lock, flags);
return -EIO;
}
if (sli4->if_type != SLI4_INTF_IF_TYPE_6) {
dwflags &= ~SLI4_MCQE_VALID;
((struct sli4_mcqe *)qe)->dw3_flags = cpu_to_le32(dwflags);
}
memcpy(entry, qe, q->size);
q->index = (q->index + 1) & (q->length - 1);
q->n_posted++;
/*
* For prism, the phase value will be used
* to check the validity of eq/cq entries.
* The value toggles after a complete sweep
* through the queue.
*/
if (sli4->if_type == SLI4_INTF_IF_TYPE_6 && q->index == 0)
q->phase ^= (u16)0x1;
spin_unlock_irqrestore(&q->lock, flags);
return 0;
}
int
sli_mq_read(struct sli4 *sli4, struct sli4_queue *q, u8 *entry)
{
u8 *qe = q->dma.virt;
unsigned long flags = 0;
spin_lock_irqsave(&q->lock, flags);
qe += q->u.r_idx * q->size;
/* Check if mqe is valid */
if (q->index == q->u.r_idx) {
spin_unlock_irqrestore(&q->lock, flags);
return -EIO;
}
memcpy(entry, qe, q->size);
q->u.r_idx = (q->u.r_idx + 1) & (q->length - 1);
spin_unlock_irqrestore(&q->lock, flags);
return 0;
}
int
sli_eq_parse(struct sli4 *sli4, u8 *buf, u16 *cq_id)
{
struct sli4_eqe *eqe = (void *)buf;
int rc = 0;
u16 flags = 0;
u16 majorcode;
u16 minorcode;
if (!buf || !cq_id) {
efc_log_err(sli4, "bad parameters sli4=%p buf=%p cq_id=%p\n",
sli4, buf, cq_id);
return -EIO;
}
flags = le16_to_cpu(eqe->dw0w0_flags);
majorcode = (flags & SLI4_EQE_MJCODE) >> 1;
minorcode = (flags & SLI4_EQE_MNCODE) >> 4;
switch (majorcode) {
case SLI4_MAJOR_CODE_STANDARD:
*cq_id = le16_to_cpu(eqe->resource_id);
break;
case SLI4_MAJOR_CODE_SENTINEL:
efc_log_info(sli4, "sentinel EQE\n");
rc = SLI4_EQE_STATUS_EQ_FULL;
break;
default:
efc_log_info(sli4, "Unsupported EQE: major %x minor %x\n",
majorcode, minorcode);
rc = -EIO;
}
return rc;
}
int
sli_cq_parse(struct sli4 *sli4, struct sli4_queue *cq, u8 *cqe,
enum sli4_qentry *etype, u16 *q_id)
{
int rc = 0;
if (!cq || !cqe || !etype) {
efc_log_err(sli4, "bad params sli4=%p cq=%p cqe=%p etype=%p q_id=%p\n",
sli4, cq, cqe, etype, q_id);
return -EINVAL;
}
/* Parse a CQ entry to retrieve the event type and the queue id */
if (cq->u.flag & SLI4_QUEUE_FLAG_MQ) {
struct sli4_mcqe *mcqe = (void *)cqe;
if (le32_to_cpu(mcqe->dw3_flags) & SLI4_MCQE_AE) {
*etype = SLI4_QENTRY_ASYNC;
} else {
*etype = SLI4_QENTRY_MQ;
rc = sli_cqe_mq(sli4, mcqe);
}
*q_id = -1;
} else {
rc = sli_fc_cqe_parse(sli4, cq, cqe, etype, q_id);
}
return rc;
}
int
sli_abort_wqe(struct sli4 *sli, void *buf, enum sli4_abort_type type,
bool send_abts, u32 ids, u32 mask, u16 tag, u16 cq_id)
{
struct sli4_abort_wqe *abort = buf;
memset(buf, 0, sli->wqe_size);
switch (type) {
case SLI4_ABORT_XRI:
abort->criteria = SLI4_ABORT_CRITERIA_XRI_TAG;
if (mask) {
efc_log_warn(sli, "%#x aborting XRI %#x warning non-zero mask",
mask, ids);
mask = 0;
}
break;
case SLI4_ABORT_ABORT_ID:
abort->criteria = SLI4_ABORT_CRITERIA_ABORT_TAG;
break;
case SLI4_ABORT_REQUEST_ID:
abort->criteria = SLI4_ABORT_CRITERIA_REQUEST_TAG;
break;
default:
efc_log_info(sli, "unsupported type %#x\n", type);
return -EIO;
}
abort->ia_ir_byte |= send_abts ? 0 : 1;
/* Suppress ABTS retries */
abort->ia_ir_byte |= SLI4_ABRT_WQE_IR;
abort->t_mask = cpu_to_le32(mask);
abort->t_tag = cpu_to_le32(ids);
abort->command = SLI4_WQE_ABORT;
abort->request_tag = cpu_to_le16(tag);
abort->dw10w0_flags = cpu_to_le16(SLI4_ABRT_WQE_QOSD);
abort->cq_id = cpu_to_le16(cq_id);
abort->cmdtype_wqec_byte |= SLI4_CMD_ABORT_WQE;
return 0;
}
int
sli_els_request64_wqe(struct sli4 *sli, void *buf, struct efc_dma *sgl,
struct sli_els_params *params)
{
struct sli4_els_request64_wqe *els = buf;
struct sli4_sge *sge = sgl->virt;
bool is_fabric = false;
struct sli4_bde *bptr;
memset(buf, 0, sli->wqe_size);
bptr = &els->els_request_payload;
if (sli->params.sgl_pre_registered) {
els->qosd_xbl_hlm_iod_dbde_wqes &= ~SLI4_REQ_WQE_XBL;
els->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_REQ_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(params->xmit_len & SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
} else {
els->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_REQ_WQE_XBL;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
((2 * sizeof(struct sli4_sge)) &
SLI4_BDE_LEN_MASK));
bptr->u.blp.low = cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high = cpu_to_le32(upper_32_bits(sgl->phys));
}
els->els_request_payload_length = cpu_to_le32(params->xmit_len);
els->max_response_payload_length = cpu_to_le32(params->rsp_len);
els->xri_tag = cpu_to_le16(params->xri);
els->timer = params->timeout;
els->class_byte |= SLI4_GENERIC_CLASS_CLASS_3;
els->command = SLI4_WQE_ELS_REQUEST64;
els->request_tag = cpu_to_le16(params->tag);
els->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_REQ_WQE_IOD;
els->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_REQ_WQE_QOSD;
/* figure out the ELS_ID value from the request buffer */
switch (params->cmd) {
case ELS_LOGO:
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_LOGO << SLI4_REQ_WQE_ELSID_SHFT;
if (params->rpi_registered) {
els->ct_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->rpi);
} else {
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
}
if (params->d_id == FC_FID_FLOGI)
is_fabric = true;
break;
case ELS_FDISC:
if (params->d_id == FC_FID_FLOGI)
is_fabric = true;
if (params->s_id == 0) {
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_FDISC << SLI4_REQ_WQE_ELSID_SHFT;
is_fabric = true;
} else {
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_OTHER << SLI4_REQ_WQE_ELSID_SHFT;
}
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
els->sid_sp_dword |= cpu_to_le32(1 << SLI4_REQ_WQE_SP_SHFT);
break;
case ELS_FLOGI:
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
/*
* Set SP here ... we haven't done a REG_VPI yet
* need to maybe not set this when we have
* completed VFI/VPI registrations ...
*
* Use the FC_ID of the SPORT if it has been allocated,
* otherwise use an S_ID of zero.
*/
els->sid_sp_dword |= cpu_to_le32(1 << SLI4_REQ_WQE_SP_SHFT);
if (params->s_id != U32_MAX)
els->sid_sp_dword |= cpu_to_le32(params->s_id);
break;
case ELS_PLOGI:
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_PLOGI << SLI4_REQ_WQE_ELSID_SHFT;
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
break;
case ELS_SCR:
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_OTHER << SLI4_REQ_WQE_ELSID_SHFT;
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
break;
default:
els->cmdtype_elsid_byte |=
SLI4_ELS_REQUEST64_OTHER << SLI4_REQ_WQE_ELSID_SHFT;
if (params->rpi_registered) {
els->ct_byte |= (SLI4_GENERIC_CONTEXT_RPI <<
SLI4_REQ_WQE_CT_SHFT);
els->context_tag = cpu_to_le16(params->vpi);
} else {
els->ct_byte |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_REQ_WQE_CT_SHFT;
els->context_tag = cpu_to_le16(params->vpi);
}
break;
}
if (is_fabric)
els->cmdtype_elsid_byte |= SLI4_ELS_REQUEST64_CMD_FABRIC;
else
els->cmdtype_elsid_byte |= SLI4_ELS_REQUEST64_CMD_NON_FABRIC;
els->cq_id = cpu_to_le16(SLI4_CQ_DEFAULT);
if (((els->ct_byte & SLI4_REQ_WQE_CT) >> SLI4_REQ_WQE_CT_SHFT) !=
SLI4_GENERIC_CONTEXT_RPI)
els->remote_id_dword = cpu_to_le32(params->d_id);
if (((els->ct_byte & SLI4_REQ_WQE_CT) >> SLI4_REQ_WQE_CT_SHFT) ==
SLI4_GENERIC_CONTEXT_VPI)
els->temporary_rpi = cpu_to_le16(params->rpi);
return 0;
}
int
sli_fcp_icmnd64_wqe(struct sli4 *sli, void *buf, struct efc_dma *sgl, u16 xri,
u16 tag, u16 cq_id, u32 rpi, u32 rnode_fcid, u8 timeout)
{
struct sli4_fcp_icmnd64_wqe *icmnd = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
u32 len;
memset(buf, 0, sli->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &icmnd->bde;
if (sli->params.sgl_pre_registered) {
icmnd->qosd_xbl_hlm_iod_dbde_wqes &= ~SLI4_ICMD_WQE_XBL;
icmnd->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_ICMD_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[0].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
} else {
icmnd->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_ICMD_WQE_XBL;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
(sgl->size & SLI4_BDE_LEN_MASK));
bptr->u.blp.low = cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high = cpu_to_le32(upper_32_bits(sgl->phys));
}
len = le32_to_cpu(sge[0].buffer_length) +
le32_to_cpu(sge[1].buffer_length);
icmnd->payload_offset_length = cpu_to_le16(len);
icmnd->xri_tag = cpu_to_le16(xri);
icmnd->context_tag = cpu_to_le16(rpi);
icmnd->timer = timeout;
/* WQE word 4 contains read transfer length */
icmnd->class_pu_byte |= 2 << SLI4_ICMD_WQE_PU_SHFT;
icmnd->class_pu_byte |= SLI4_GENERIC_CLASS_CLASS_3;
icmnd->command = SLI4_WQE_FCP_ICMND64;
icmnd->dif_ct_bs_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_ICMD_WQE_CT_SHFT;
icmnd->abort_tag = cpu_to_le32(xri);
icmnd->request_tag = cpu_to_le16(tag);
icmnd->len_loc1_byte |= SLI4_ICMD_WQE_LEN_LOC_BIT1;
icmnd->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_ICMD_WQE_LEN_LOC_BIT2;
icmnd->cmd_type_byte |= SLI4_CMD_FCP_ICMND64_WQE;
icmnd->cq_id = cpu_to_le16(cq_id);
return 0;
}
int
sli_fcp_iread64_wqe(struct sli4 *sli, void *buf, struct efc_dma *sgl,
u32 first_data_sge, u32 xfer_len, u16 xri, u16 tag,
u16 cq_id, u32 rpi, u32 rnode_fcid,
u8 dif, u8 bs, u8 timeout)
{
struct sli4_fcp_iread64_wqe *iread = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
u32 sge_flags, len;
memset(buf, 0, sli->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &iread->bde;
if (sli->params.sgl_pre_registered) {
iread->qosd_xbl_hlm_iod_dbde_wqes &= ~SLI4_IR_WQE_XBL;
iread->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IR_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[0].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.blp.low = sge[0].buffer_address_low;
bptr->u.blp.high = sge[0].buffer_address_high;
} else {
iread->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IR_WQE_XBL;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
(sgl->size & SLI4_BDE_LEN_MASK));
bptr->u.blp.low =
cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high =
cpu_to_le32(upper_32_bits(sgl->phys));
/*
* fill out fcp_cmnd buffer len and change resp buffer to be of
* type "skip" (note: response will still be written to sge[1]
* if necessary)
*/
len = le32_to_cpu(sge[0].buffer_length);
iread->fcp_cmd_buffer_length = cpu_to_le16(len);
sge_flags = le32_to_cpu(sge[1].dw2_flags);
sge_flags &= (~SLI4_SGE_TYPE_MASK);
sge_flags |= (SLI4_SGE_TYPE_SKIP << SLI4_SGE_TYPE_SHIFT);
sge[1].dw2_flags = cpu_to_le32(sge_flags);
}
len = le32_to_cpu(sge[0].buffer_length) +
le32_to_cpu(sge[1].buffer_length);
iread->payload_offset_length = cpu_to_le16(len);
iread->total_transfer_length = cpu_to_le32(xfer_len);
iread->xri_tag = cpu_to_le16(xri);
iread->context_tag = cpu_to_le16(rpi);
iread->timer = timeout;
/* WQE word 4 contains read transfer length */
iread->class_pu_byte |= 2 << SLI4_IR_WQE_PU_SHFT;
iread->class_pu_byte |= SLI4_GENERIC_CLASS_CLASS_3;
iread->command = SLI4_WQE_FCP_IREAD64;
iread->dif_ct_bs_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_IR_WQE_CT_SHFT;
iread->dif_ct_bs_byte |= dif;
iread->dif_ct_bs_byte |= bs << SLI4_IR_WQE_BS_SHFT;
iread->abort_tag = cpu_to_le32(xri);
iread->request_tag = cpu_to_le16(tag);
iread->len_loc1_byte |= SLI4_IR_WQE_LEN_LOC_BIT1;
iread->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IR_WQE_LEN_LOC_BIT2;
iread->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IR_WQE_IOD;
iread->cmd_type_byte |= SLI4_CMD_FCP_IREAD64_WQE;
iread->cq_id = cpu_to_le16(cq_id);
if (sli->params.perf_hint) {
bptr = &iread->first_data_bde;
bptr->bde_type_buflen = cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[first_data_sge].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low =
sge[first_data_sge].buffer_address_low;
bptr->u.data.high =
sge[first_data_sge].buffer_address_high;
}
return 0;
}
int
sli_fcp_iwrite64_wqe(struct sli4 *sli, void *buf, struct efc_dma *sgl,
u32 first_data_sge, u32 xfer_len,
u32 first_burst, u16 xri, u16 tag,
u16 cq_id, u32 rpi,
u32 rnode_fcid,
u8 dif, u8 bs, u8 timeout)
{
struct sli4_fcp_iwrite64_wqe *iwrite = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
u32 sge_flags, min, len;
memset(buf, 0, sli->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &iwrite->bde;
if (sli->params.sgl_pre_registered) {
iwrite->qosd_xbl_hlm_iod_dbde_wqes &= ~SLI4_IWR_WQE_XBL;
iwrite->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IWR_WQE_DBDE;
bptr->bde_type_buflen = cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[0].buffer_length) & SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
} else {
iwrite->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IWR_WQE_XBL;
bptr->bde_type_buflen = cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(sgl->size & SLI4_BDE_LEN_MASK));
bptr->u.blp.low = cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high = cpu_to_le32(upper_32_bits(sgl->phys));
/*
* fill out fcp_cmnd buffer len and change resp buffer to be of
* type "skip" (note: response will still be written to sge[1]
* if necessary)
*/
len = le32_to_cpu(sge[0].buffer_length);
iwrite->fcp_cmd_buffer_length = cpu_to_le16(len);
sge_flags = le32_to_cpu(sge[1].dw2_flags);
sge_flags &= ~SLI4_SGE_TYPE_MASK;
sge_flags |= (SLI4_SGE_TYPE_SKIP << SLI4_SGE_TYPE_SHIFT);
sge[1].dw2_flags = cpu_to_le32(sge_flags);
}
len = le32_to_cpu(sge[0].buffer_length) +
le32_to_cpu(sge[1].buffer_length);
iwrite->payload_offset_length = cpu_to_le16(len);
iwrite->total_transfer_length = cpu_to_le16(xfer_len);
min = (xfer_len < first_burst) ? xfer_len : first_burst;
iwrite->initial_transfer_length = cpu_to_le16(min);
iwrite->xri_tag = cpu_to_le16(xri);
iwrite->context_tag = cpu_to_le16(rpi);
iwrite->timer = timeout;
/* WQE word 4 contains read transfer length */
iwrite->class_pu_byte |= 2 << SLI4_IWR_WQE_PU_SHFT;
iwrite->class_pu_byte |= SLI4_GENERIC_CLASS_CLASS_3;
iwrite->command = SLI4_WQE_FCP_IWRITE64;
iwrite->dif_ct_bs_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_IWR_WQE_CT_SHFT;
iwrite->dif_ct_bs_byte |= dif;
iwrite->dif_ct_bs_byte |= bs << SLI4_IWR_WQE_BS_SHFT;
iwrite->abort_tag = cpu_to_le32(xri);
iwrite->request_tag = cpu_to_le16(tag);
iwrite->len_loc1_byte |= SLI4_IWR_WQE_LEN_LOC_BIT1;
iwrite->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_IWR_WQE_LEN_LOC_BIT2;
iwrite->cmd_type_byte |= SLI4_CMD_FCP_IWRITE64_WQE;
iwrite->cq_id = cpu_to_le16(cq_id);
if (sli->params.perf_hint) {
bptr = &iwrite->first_data_bde;
bptr->bde_type_buflen = cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[first_data_sge].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[first_data_sge].buffer_address_low;
bptr->u.data.high = sge[first_data_sge].buffer_address_high;
}
return 0;
}
int
sli_fcp_treceive64_wqe(struct sli4 *sli, void *buf, struct efc_dma *sgl,
u32 first_data_sge, u16 cq_id, u8 dif, u8 bs,
struct sli_fcp_tgt_params *params)
{
struct sli4_fcp_treceive64_wqe *trecv = buf;
struct sli4_fcp_128byte_wqe *trecv_128 = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
memset(buf, 0, sli->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &trecv->bde;
if (sli->params.sgl_pre_registered) {
trecv->qosd_xbl_hlm_iod_dbde_wqes &= ~SLI4_TRCV_WQE_XBL;
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[0].buffer_length)
& SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
trecv->payload_offset_length = sge[0].buffer_length;
} else {
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_XBL;
/* if data is a single physical address, use a BDE */
if (!dif &&
params->xmit_len <= le32_to_cpu(sge[2].buffer_length)) {
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[2].buffer_length)
& SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[2].buffer_address_low;
bptr->u.data.high = sge[2].buffer_address_high;
} else {
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
(sgl->size & SLI4_BDE_LEN_MASK));
bptr->u.blp.low = cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high =
cpu_to_le32(upper_32_bits(sgl->phys));
}
}
trecv->relative_offset = cpu_to_le32(params->offset);
if (params->flags & SLI4_IO_CONTINUATION)
trecv->eat_xc_ccpe |= SLI4_TRCV_WQE_XC;
trecv->xri_tag = cpu_to_le16(params->xri);
trecv->context_tag = cpu_to_le16(params->rpi);
/* WQE uses relative offset */
trecv->class_ar_pu_byte |= 1 << SLI4_TRCV_WQE_PU_SHFT;
if (params->flags & SLI4_IO_AUTO_GOOD_RESPONSE)
trecv->class_ar_pu_byte |= SLI4_TRCV_WQE_AR;
trecv->command = SLI4_WQE_FCP_TRECEIVE64;
trecv->class_ar_pu_byte |= SLI4_GENERIC_CLASS_CLASS_3;
trecv->dif_ct_bs_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_TRCV_WQE_CT_SHFT;
trecv->dif_ct_bs_byte |= bs << SLI4_TRCV_WQE_BS_SHFT;
trecv->remote_xid = cpu_to_le16(params->ox_id);
trecv->request_tag = cpu_to_le16(params->tag);
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_IOD;
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_LEN_LOC_BIT2;
trecv->cmd_type_byte |= SLI4_CMD_FCP_TRECEIVE64_WQE;
trecv->cq_id = cpu_to_le16(cq_id);
trecv->fcp_data_receive_length = cpu_to_le32(params->xmit_len);
if (sli->params.perf_hint) {
bptr = &trecv->first_data_bde;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[first_data_sge].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[first_data_sge].buffer_address_low;
bptr->u.data.high = sge[first_data_sge].buffer_address_high;
}
/* The upper 7 bits of csctl is the priority */
if (params->cs_ctl & SLI4_MASK_CCP) {
trecv->eat_xc_ccpe |= SLI4_TRCV_WQE_CCPE;
trecv->ccp = (params->cs_ctl & SLI4_MASK_CCP);
}
if (params->app_id && sli->wqe_size == SLI4_WQE_EXT_BYTES &&
!(trecv->eat_xc_ccpe & SLI4_TRSP_WQE_EAT)) {
trecv->lloc1_appid |= SLI4_TRCV_WQE_APPID;
trecv->qosd_xbl_hlm_iod_dbde_wqes |= SLI4_TRCV_WQE_WQES;
trecv_128->dw[31] = params->app_id;
}
return 0;
}
int
sli_fcp_cont_treceive64_wqe(struct sli4 *sli, void *buf,
struct efc_dma *sgl, u32 first_data_sge,
u16 sec_xri, u16 cq_id, u8 dif, u8 bs,
struct sli_fcp_tgt_params *params)
{
int rc;
rc = sli_fcp_treceive64_wqe(sli, buf, sgl, first_data_sge,
cq_id, dif, bs, params);
if (!rc) {
struct sli4_fcp_treceive64_wqe *trecv = buf;
trecv->command = SLI4_WQE_FCP_CONT_TRECEIVE64;
trecv->dword5.sec_xri_tag = cpu_to_le16(sec_xri);
}
return rc;
}
int
sli_fcp_trsp64_wqe(struct sli4 *sli4, void *buf, struct efc_dma *sgl,
u16 cq_id, u8 port_owned, struct sli_fcp_tgt_params *params)
{
struct sli4_fcp_trsp64_wqe *trsp = buf;
struct sli4_fcp_128byte_wqe *trsp_128 = buf;
memset(buf, 0, sli4->wqe_size);
if (params->flags & SLI4_IO_AUTO_GOOD_RESPONSE) {
trsp->class_ag_byte |= SLI4_TRSP_WQE_AG;
} else {
struct sli4_sge *sge = sgl->virt;
struct sli4_bde *bptr;
if (sli4->params.sgl_pre_registered || port_owned)
trsp->qosd_xbl_hlm_dbde_wqes |= SLI4_TRSP_WQE_DBDE;
else
trsp->qosd_xbl_hlm_dbde_wqes |= SLI4_TRSP_WQE_XBL;
bptr = &trsp->bde;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[0].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
trsp->fcp_response_length = cpu_to_le32(params->xmit_len);
}
if (params->flags & SLI4_IO_CONTINUATION)
trsp->eat_xc_ccpe |= SLI4_TRSP_WQE_XC;
trsp->xri_tag = cpu_to_le16(params->xri);
trsp->rpi = cpu_to_le16(params->rpi);
trsp->command = SLI4_WQE_FCP_TRSP64;
trsp->class_ag_byte |= SLI4_GENERIC_CLASS_CLASS_3;
trsp->remote_xid = cpu_to_le16(params->ox_id);
trsp->request_tag = cpu_to_le16(params->tag);
if (params->flags & SLI4_IO_DNRX)
trsp->ct_dnrx_byte |= SLI4_TRSP_WQE_DNRX;
else
trsp->ct_dnrx_byte &= ~SLI4_TRSP_WQE_DNRX;
trsp->lloc1_appid |= 0x1;
trsp->cq_id = cpu_to_le16(cq_id);
trsp->cmd_type_byte = SLI4_CMD_FCP_TRSP64_WQE;
/* The upper 7 bits of csctl is the priority */
if (params->cs_ctl & SLI4_MASK_CCP) {
trsp->eat_xc_ccpe |= SLI4_TRSP_WQE_CCPE;
trsp->ccp = (params->cs_ctl & SLI4_MASK_CCP);
}
if (params->app_id && sli4->wqe_size == SLI4_WQE_EXT_BYTES &&
!(trsp->eat_xc_ccpe & SLI4_TRSP_WQE_EAT)) {
trsp->lloc1_appid |= SLI4_TRSP_WQE_APPID;
trsp->qosd_xbl_hlm_dbde_wqes |= SLI4_TRSP_WQE_WQES;
trsp_128->dw[31] = params->app_id;
}
return 0;
}
int
sli_fcp_tsend64_wqe(struct sli4 *sli4, void *buf, struct efc_dma *sgl,
u32 first_data_sge, u16 cq_id, u8 dif, u8 bs,
struct sli_fcp_tgt_params *params)
{
struct sli4_fcp_tsend64_wqe *tsend = buf;
struct sli4_fcp_128byte_wqe *tsend_128 = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
memset(buf, 0, sli4->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli4, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &tsend->bde;
if (sli4->params.sgl_pre_registered) {
tsend->ll_qd_xbl_hlm_iod_dbde &= ~SLI4_TSEND_WQE_XBL;
tsend->ll_qd_xbl_hlm_iod_dbde |= SLI4_TSEND_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[2].buffer_length) &
SLI4_BDE_LEN_MASK));
/* TSEND64_WQE specifies first two SGE are skipped (3rd is
* valid)
*/
bptr->u.data.low = sge[2].buffer_address_low;
bptr->u.data.high = sge[2].buffer_address_high;
} else {
tsend->ll_qd_xbl_hlm_iod_dbde |= SLI4_TSEND_WQE_XBL;
/* if data is a single physical address, use a BDE */
if (!dif &&
params->xmit_len <= le32_to_cpu(sge[2].buffer_length)) {
tsend->ll_qd_xbl_hlm_iod_dbde |= SLI4_TSEND_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[2].buffer_length) &
SLI4_BDE_LEN_MASK));
/*
* TSEND64_WQE specifies first two SGE are skipped
* (i.e. 3rd is valid)
*/
bptr->u.data.low =
sge[2].buffer_address_low;
bptr->u.data.high =
sge[2].buffer_address_high;
} else {
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
(sgl->size &
SLI4_BDE_LEN_MASK));
bptr->u.blp.low =
cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high =
cpu_to_le32(upper_32_bits(sgl->phys));
}
}
tsend->relative_offset = cpu_to_le32(params->offset);
if (params->flags & SLI4_IO_CONTINUATION)
tsend->dw10byte2 |= SLI4_TSEND_XC;
tsend->xri_tag = cpu_to_le16(params->xri);
tsend->rpi = cpu_to_le16(params->rpi);
/* WQE uses relative offset */
tsend->class_pu_ar_byte |= 1 << SLI4_TSEND_WQE_PU_SHFT;
if (params->flags & SLI4_IO_AUTO_GOOD_RESPONSE)
tsend->class_pu_ar_byte |= SLI4_TSEND_WQE_AR;
tsend->command = SLI4_WQE_FCP_TSEND64;
tsend->class_pu_ar_byte |= SLI4_GENERIC_CLASS_CLASS_3;
tsend->ct_byte |= SLI4_GENERIC_CONTEXT_RPI << SLI4_TSEND_CT_SHFT;
tsend->ct_byte |= dif;
tsend->ct_byte |= bs << SLI4_TSEND_BS_SHFT;
tsend->remote_xid = cpu_to_le16(params->ox_id);
tsend->request_tag = cpu_to_le16(params->tag);
tsend->ll_qd_xbl_hlm_iod_dbde |= SLI4_TSEND_LEN_LOC_BIT2;
tsend->cq_id = cpu_to_le16(cq_id);
tsend->cmd_type_byte |= SLI4_CMD_FCP_TSEND64_WQE;
tsend->fcp_data_transmit_length = cpu_to_le32(params->xmit_len);
if (sli4->params.perf_hint) {
bptr = &tsend->first_data_bde;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(le32_to_cpu(sge[first_data_sge].buffer_length) &
SLI4_BDE_LEN_MASK));
bptr->u.data.low =
sge[first_data_sge].buffer_address_low;
bptr->u.data.high =
sge[first_data_sge].buffer_address_high;
}
/* The upper 7 bits of csctl is the priority */
if (params->cs_ctl & SLI4_MASK_CCP) {
tsend->dw10byte2 |= SLI4_TSEND_CCPE;
tsend->ccp = (params->cs_ctl & SLI4_MASK_CCP);
}
if (params->app_id && sli4->wqe_size == SLI4_WQE_EXT_BYTES &&
!(tsend->dw10byte2 & SLI4_TSEND_EAT)) {
tsend->dw10byte0 |= SLI4_TSEND_APPID_VALID;
tsend->ll_qd_xbl_hlm_iod_dbde |= SLI4_TSEND_WQES;
tsend_128->dw[31] = params->app_id;
}
return 0;
}
int
sli_gen_request64_wqe(struct sli4 *sli4, void *buf, struct efc_dma *sgl,
struct sli_ct_params *params)
{
struct sli4_gen_request64_wqe *gen = buf;
struct sli4_sge *sge = NULL;
struct sli4_bde *bptr;
memset(buf, 0, sli4->wqe_size);
if (!sgl || !sgl->virt) {
efc_log_err(sli4, "bad parameter sgl=%p virt=%p\n",
sgl, sgl ? sgl->virt : NULL);
return -EIO;
}
sge = sgl->virt;
bptr = &gen->bde;
if (sli4->params.sgl_pre_registered) {
gen->dw10flags1 &= ~SLI4_GEN_REQ64_WQE_XBL;
gen->dw10flags1 |= SLI4_GEN_REQ64_WQE_DBDE;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(params->xmit_len & SLI4_BDE_LEN_MASK));
bptr->u.data.low = sge[0].buffer_address_low;
bptr->u.data.high = sge[0].buffer_address_high;
} else {
gen->dw10flags1 |= SLI4_GEN_REQ64_WQE_XBL;
bptr->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(BLP)) |
((2 * sizeof(struct sli4_sge)) &
SLI4_BDE_LEN_MASK));
bptr->u.blp.low =
cpu_to_le32(lower_32_bits(sgl->phys));
bptr->u.blp.high =
cpu_to_le32(upper_32_bits(sgl->phys));
}
gen->request_payload_length = cpu_to_le32(params->xmit_len);
gen->max_response_payload_length = cpu_to_le32(params->rsp_len);
gen->df_ctl = params->df_ctl;
gen->type = params->type;
gen->r_ctl = params->r_ctl;
gen->xri_tag = cpu_to_le16(params->xri);
gen->ct_byte = SLI4_GENERIC_CONTEXT_RPI << SLI4_GEN_REQ64_CT_SHFT;
gen->context_tag = cpu_to_le16(params->rpi);
gen->class_byte = SLI4_GENERIC_CLASS_CLASS_3;
gen->command = SLI4_WQE_GEN_REQUEST64;
gen->timer = params->timeout;
gen->request_tag = cpu_to_le16(params->tag);
gen->dw10flags1 |= SLI4_GEN_REQ64_WQE_IOD;
gen->dw10flags0 |= SLI4_GEN_REQ64_WQE_QOSD;
gen->cmd_type_byte = SLI4_CMD_GEN_REQUEST64_WQE;
gen->cq_id = cpu_to_le16(SLI4_CQ_DEFAULT);
return 0;
}
int
sli_send_frame_wqe(struct sli4 *sli, void *buf, u8 sof, u8 eof, u32 *hdr,
struct efc_dma *payload, u32 req_len, u8 timeout, u16 xri,
u16 req_tag)
{
struct sli4_send_frame_wqe *sf = buf;
memset(buf, 0, sli->wqe_size);
sf->dw10flags1 |= SLI4_SF_WQE_DBDE;
sf->bde.bde_type_buflen = cpu_to_le32(req_len &
SLI4_BDE_LEN_MASK);
sf->bde.u.data.low = cpu_to_le32(lower_32_bits(payload->phys));
sf->bde.u.data.high = cpu_to_le32(upper_32_bits(payload->phys));
/* Copy FC header */
sf->fc_header_0_1[0] = cpu_to_le32(hdr[0]);
sf->fc_header_0_1[1] = cpu_to_le32(hdr[1]);
sf->fc_header_2_5[0] = cpu_to_le32(hdr[2]);
sf->fc_header_2_5[1] = cpu_to_le32(hdr[3]);
sf->fc_header_2_5[2] = cpu_to_le32(hdr[4]);
sf->fc_header_2_5[3] = cpu_to_le32(hdr[5]);
sf->frame_length = cpu_to_le32(req_len);
sf->xri_tag = cpu_to_le16(xri);
sf->dw7flags0 &= ~SLI4_SF_PU;
sf->context_tag = 0;
sf->ct_byte &= ~SLI4_SF_CT;
sf->command = SLI4_WQE_SEND_FRAME;
sf->dw7flags0 |= SLI4_GENERIC_CLASS_CLASS_3;
sf->timer = timeout;
sf->request_tag = cpu_to_le16(req_tag);
sf->eof = eof;
sf->sof = sof;
sf->dw10flags1 &= ~SLI4_SF_QOSD;
sf->dw10flags0 |= SLI4_SF_LEN_LOC_BIT1;
sf->dw10flags2 &= ~SLI4_SF_XC;
sf->dw10flags1 |= SLI4_SF_XBL;
sf->cmd_type_byte |= SLI4_CMD_SEND_FRAME_WQE;
sf->cq_id = cpu_to_le16(0xffff);
return 0;
}
int
sli_xmit_bls_rsp64_wqe(struct sli4 *sli, void *buf,
struct sli_bls_payload *payload,
struct sli_bls_params *params)
{
struct sli4_xmit_bls_rsp_wqe *bls = buf;
u32 dw_ridflags = 0;
/*
* Callers can either specify RPI or S_ID, but not both
*/
if (params->rpi_registered && params->s_id != U32_MAX) {
efc_log_info(sli, "S_ID specified for attached remote node %d\n",
params->rpi);
return -EIO;
}
memset(buf, 0, sli->wqe_size);
if (payload->type == SLI4_SLI_BLS_ACC) {
bls->payload_word0 =
cpu_to_le32((payload->u.acc.seq_id_last << 16) |
(payload->u.acc.seq_id_validity << 24));
bls->high_seq_cnt = payload->u.acc.high_seq_cnt;
bls->low_seq_cnt = payload->u.acc.low_seq_cnt;
} else if (payload->type == SLI4_SLI_BLS_RJT) {
bls->payload_word0 =
cpu_to_le32(*((u32 *)&payload->u.rjt));
dw_ridflags |= SLI4_BLS_RSP_WQE_AR;
} else {
efc_log_info(sli, "bad BLS type %#x\n", payload->type);
return -EIO;
}
bls->ox_id = payload->ox_id;
bls->rx_id = payload->rx_id;
if (params->rpi_registered) {
bls->dw8flags0 |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_BLS_RSP_WQE_CT_SHFT;
bls->context_tag = cpu_to_le16(params->rpi);
} else {
bls->dw8flags0 |=
SLI4_GENERIC_CONTEXT_VPI << SLI4_BLS_RSP_WQE_CT_SHFT;
bls->context_tag = cpu_to_le16(params->vpi);
if (params->s_id != U32_MAX)
bls->local_n_port_id_dword |=
cpu_to_le32(params->s_id & 0x00ffffff);
else
bls->local_n_port_id_dword |=
cpu_to_le32(params->s_id & 0x00ffffff);
dw_ridflags = (dw_ridflags & ~SLI4_BLS_RSP_RID) |
(params->d_id & SLI4_BLS_RSP_RID);
bls->temporary_rpi = cpu_to_le16(params->rpi);
}
bls->xri_tag = cpu_to_le16(params->xri);
bls->dw8flags1 |= SLI4_GENERIC_CLASS_CLASS_3;
bls->command = SLI4_WQE_XMIT_BLS_RSP;
bls->request_tag = cpu_to_le16(params->tag);
bls->dw11flags1 |= SLI4_BLS_RSP_WQE_QOSD;
bls->remote_id_dword = cpu_to_le32(dw_ridflags);
bls->cq_id = cpu_to_le16(SLI4_CQ_DEFAULT);
bls->dw12flags0 |= SLI4_CMD_XMIT_BLS_RSP64_WQE;
return 0;
}
int
sli_xmit_els_rsp64_wqe(struct sli4 *sli, void *buf, struct efc_dma *rsp,
struct sli_els_params *params)
{
struct sli4_xmit_els_rsp64_wqe *els = buf;
memset(buf, 0, sli->wqe_size);
if (sli->params.sgl_pre_registered)
els->flags2 |= SLI4_ELS_DBDE;
else
els->flags2 |= SLI4_ELS_XBL;
els->els_response_payload.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(params->rsp_len & SLI4_BDE_LEN_MASK));
els->els_response_payload.u.data.low =
cpu_to_le32(lower_32_bits(rsp->phys));
els->els_response_payload.u.data.high =
cpu_to_le32(upper_32_bits(rsp->phys));
els->els_response_payload_length = cpu_to_le32(params->rsp_len);
els->xri_tag = cpu_to_le16(params->xri);
els->class_byte |= SLI4_GENERIC_CLASS_CLASS_3;
els->command = SLI4_WQE_ELS_RSP64;
els->request_tag = cpu_to_le16(params->tag);
els->ox_id = cpu_to_le16(params->ox_id);
els->flags2 |= SLI4_ELS_QOSD;
els->cmd_type_wqec = SLI4_ELS_REQUEST64_CMD_GEN;
els->cq_id = cpu_to_le16(SLI4_CQ_DEFAULT);
if (params->rpi_registered) {
els->ct_byte |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_ELS_CT_OFFSET;
els->context_tag = cpu_to_le16(params->rpi);
return 0;
}
els->ct_byte |= SLI4_GENERIC_CONTEXT_VPI << SLI4_ELS_CT_OFFSET;
els->context_tag = cpu_to_le16(params->vpi);
els->rid_dw = cpu_to_le32(params->d_id & SLI4_ELS_RID);
els->temporary_rpi = cpu_to_le16(params->rpi);
if (params->s_id != U32_MAX) {
els->sid_dw |=
cpu_to_le32(SLI4_ELS_SP | (params->s_id & SLI4_ELS_SID));
}
return 0;
}
int
sli_xmit_sequence64_wqe(struct sli4 *sli4, void *buf, struct efc_dma *payload,
struct sli_ct_params *params)
{
struct sli4_xmit_sequence64_wqe *xmit = buf;
memset(buf, 0, sli4->wqe_size);
if (!payload || !payload->virt) {
efc_log_err(sli4, "bad parameter sgl=%p virt=%p\n",
payload, payload ? payload->virt : NULL);
return -EIO;
}
if (sli4->params.sgl_pre_registered)
xmit->dw10w0 |= cpu_to_le16(SLI4_SEQ_WQE_DBDE);
else
xmit->dw10w0 |= cpu_to_le16(SLI4_SEQ_WQE_XBL);
xmit->bde.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(params->rsp_len & SLI4_BDE_LEN_MASK));
xmit->bde.u.data.low =
cpu_to_le32(lower_32_bits(payload->phys));
xmit->bde.u.data.high =
cpu_to_le32(upper_32_bits(payload->phys));
xmit->sequence_payload_len = cpu_to_le32(params->rsp_len);
xmit->remote_n_port_id_dword |= cpu_to_le32(params->d_id & 0x00ffffff);
xmit->relative_offset = 0;
/* sequence initiative - this matches what is seen from
* FC switches in response to FCGS commands
*/
xmit->dw5flags0 &= (~SLI4_SEQ_WQE_SI);
xmit->dw5flags0 &= (~SLI4_SEQ_WQE_FT);/* force transmit */
xmit->dw5flags0 &= (~SLI4_SEQ_WQE_XO);/* exchange responder */
xmit->dw5flags0 |= SLI4_SEQ_WQE_LS;/* last in seqence */
xmit->df_ctl = params->df_ctl;
xmit->type = params->type;
xmit->r_ctl = params->r_ctl;
xmit->xri_tag = cpu_to_le16(params->xri);
xmit->context_tag = cpu_to_le16(params->rpi);
xmit->dw7flags0 &= ~SLI4_SEQ_WQE_DIF;
xmit->dw7flags0 |=
SLI4_GENERIC_CONTEXT_RPI << SLI4_SEQ_WQE_CT_SHIFT;
xmit->dw7flags0 &= ~SLI4_SEQ_WQE_BS;
xmit->command = SLI4_WQE_XMIT_SEQUENCE64;
xmit->dw7flags1 |= SLI4_GENERIC_CLASS_CLASS_3;
xmit->dw7flags1 &= ~SLI4_SEQ_WQE_PU;
xmit->timer = params->timeout;
xmit->abort_tag = 0;
xmit->request_tag = cpu_to_le16(params->tag);
xmit->remote_xid = cpu_to_le16(params->ox_id);
xmit->dw10w0 |=
cpu_to_le16(SLI4_ELS_REQUEST64_DIR_READ << SLI4_SEQ_WQE_IOD_SHIFT);
xmit->cmd_type_wqec_byte |= SLI4_CMD_XMIT_SEQUENCE64_WQE;
xmit->dw10w0 |= cpu_to_le16(2 << SLI4_SEQ_WQE_LEN_LOC_SHIFT);
xmit->cq_id = cpu_to_le16(0xFFFF);
return 0;
}
int
sli_requeue_xri_wqe(struct sli4 *sli4, void *buf, u16 xri, u16 tag, u16 cq_id)
{
struct sli4_requeue_xri_wqe *requeue = buf;
memset(buf, 0, sli4->wqe_size);
requeue->command = SLI4_WQE_REQUEUE_XRI;
requeue->xri_tag = cpu_to_le16(xri);
requeue->request_tag = cpu_to_le16(tag);
requeue->flags2 |= cpu_to_le16(SLI4_REQU_XRI_WQE_XC);
requeue->flags1 |= cpu_to_le16(SLI4_REQU_XRI_WQE_QOSD);
requeue->cq_id = cpu_to_le16(cq_id);
requeue->cmd_type_wqec_byte = SLI4_CMD_REQUEUE_XRI_WQE;
return 0;
}
int
sli_fc_process_link_attention(struct sli4 *sli4, void *acqe)
{
struct sli4_link_attention *link_attn = acqe;
struct sli4_link_event event = { 0 };
efc_log_info(sli4, "link=%d attn_type=%#x top=%#x speed=%#x pfault=%#x\n",
link_attn->link_number, link_attn->attn_type,
link_attn->topology, link_attn->port_speed,
link_attn->port_fault);
efc_log_info(sli4, "shared_lnk_status=%#x logl_lnk_speed=%#x evttag=%#x\n",
link_attn->shared_link_status,
le16_to_cpu(link_attn->logical_link_speed),
le32_to_cpu(link_attn->event_tag));
if (!sli4->link)
return -EIO;
event.medium = SLI4_LINK_MEDIUM_FC;
switch (link_attn->attn_type) {
case SLI4_LNK_ATTN_TYPE_LINK_UP:
event.status = SLI4_LINK_STATUS_UP;
break;
case SLI4_LNK_ATTN_TYPE_LINK_DOWN:
event.status = SLI4_LINK_STATUS_DOWN;
break;
case SLI4_LNK_ATTN_TYPE_NO_HARD_ALPA:
efc_log_info(sli4, "attn_type: no hard alpa\n");
event.status = SLI4_LINK_STATUS_NO_ALPA;
break;
default:
efc_log_info(sli4, "attn_type: unknown\n");
break;
}
switch (link_attn->event_type) {
case SLI4_EVENT_LINK_ATTENTION:
break;
case SLI4_EVENT_SHARED_LINK_ATTENTION:
efc_log_info(sli4, "event_type: FC shared link event\n");
break;
default:
efc_log_info(sli4, "event_type: unknown\n");
break;
}
switch (link_attn->topology) {
case SLI4_LNK_ATTN_P2P:
event.topology = SLI4_LINK_TOPO_NON_FC_AL;
break;
case SLI4_LNK_ATTN_FC_AL:
event.topology = SLI4_LINK_TOPO_FC_AL;
break;
case SLI4_LNK_ATTN_INTERNAL_LOOPBACK:
efc_log_info(sli4, "topology Internal loopback\n");
event.topology = SLI4_LINK_TOPO_LOOPBACK_INTERNAL;
break;
case SLI4_LNK_ATTN_SERDES_LOOPBACK:
efc_log_info(sli4, "topology serdes loopback\n");
event.topology = SLI4_LINK_TOPO_LOOPBACK_EXTERNAL;
break;
default:
efc_log_info(sli4, "topology: unknown\n");
break;
}
event.speed = link_attn->port_speed * 1000;
sli4->link(sli4->link_arg, (void *)&event);
return 0;
}
int
sli_fc_cqe_parse(struct sli4 *sli4, struct sli4_queue *cq,
u8 *cqe, enum sli4_qentry *etype, u16 *r_id)
{
u8 code = cqe[SLI4_CQE_CODE_OFFSET];
int rc;
switch (code) {
case SLI4_CQE_CODE_WORK_REQUEST_COMPLETION:
{
struct sli4_fc_wcqe *wcqe = (void *)cqe;
*etype = SLI4_QENTRY_WQ;
*r_id = le16_to_cpu(wcqe->request_tag);
rc = wcqe->status;
/* Flag errors except for FCP_RSP_FAILURE */
if (rc && rc != SLI4_FC_WCQE_STATUS_FCP_RSP_FAILURE) {
efc_log_info(sli4, "WCQE: status=%#x hw_status=%#x tag=%#x\n",
wcqe->status, wcqe->hw_status,
le16_to_cpu(wcqe->request_tag));
efc_log_info(sli4, "w1=%#x w2=%#x xb=%d\n",
le32_to_cpu(wcqe->wqe_specific_1),
le32_to_cpu(wcqe->wqe_specific_2),
(wcqe->flags & SLI4_WCQE_XB));
efc_log_info(sli4, " %08X %08X %08X %08X\n",
((u32 *)cqe)[0], ((u32 *)cqe)[1],
((u32 *)cqe)[2], ((u32 *)cqe)[3]);
}
break;
}
case SLI4_CQE_CODE_RQ_ASYNC:
{
struct sli4_fc_async_rcqe *rcqe = (void *)cqe;
*etype = SLI4_QENTRY_RQ;
*r_id = le16_to_cpu(rcqe->fcfi_rq_id_word) & SLI4_RACQE_RQ_ID;
rc = rcqe->status;
break;
}
case SLI4_CQE_CODE_RQ_ASYNC_V1:
{
struct sli4_fc_async_rcqe_v1 *rcqe = (void *)cqe;
*etype = SLI4_QENTRY_RQ;
*r_id = le16_to_cpu(rcqe->rq_id);
rc = rcqe->status;
break;
}
case SLI4_CQE_CODE_OPTIMIZED_WRITE_CMD:
{
struct sli4_fc_optimized_write_cmd_cqe *optcqe = (void *)cqe;
*etype = SLI4_QENTRY_OPT_WRITE_CMD;
*r_id = le16_to_cpu(optcqe->rq_id);
rc = optcqe->status;
break;
}
case SLI4_CQE_CODE_OPTIMIZED_WRITE_DATA:
{
struct sli4_fc_optimized_write_data_cqe *dcqe = (void *)cqe;
*etype = SLI4_QENTRY_OPT_WRITE_DATA;
*r_id = le16_to_cpu(dcqe->xri);
rc = dcqe->status;
/* Flag errors */
if (rc != SLI4_FC_WCQE_STATUS_SUCCESS) {
efc_log_info(sli4, "Optimized DATA CQE: status=%#x\n",
dcqe->status);
efc_log_info(sli4, "hstat=%#x xri=%#x dpl=%#x w3=%#x xb=%d\n",
dcqe->hw_status, le16_to_cpu(dcqe->xri),
le32_to_cpu(dcqe->total_data_placed),
((u32 *)cqe)[3],
(dcqe->flags & SLI4_OCQE_XB));
}
break;
}
case SLI4_CQE_CODE_RQ_COALESCING:
{
struct sli4_fc_coalescing_rcqe *rcqe = (void *)cqe;
*etype = SLI4_QENTRY_RQ;
*r_id = le16_to_cpu(rcqe->rq_id);
rc = rcqe->status;
break;
}
case SLI4_CQE_CODE_XRI_ABORTED:
{
struct sli4_fc_xri_aborted_cqe *xa = (void *)cqe;
*etype = SLI4_QENTRY_XABT;
*r_id = le16_to_cpu(xa->xri);
rc = 0;
break;
}
case SLI4_CQE_CODE_RELEASE_WQE:
{
struct sli4_fc_wqec *wqec = (void *)cqe;
*etype = SLI4_QENTRY_WQ_RELEASE;
*r_id = le16_to_cpu(wqec->wq_id);
rc = 0;
break;
}
default:
efc_log_info(sli4, "CQE completion code %d not handled\n",
code);
*etype = SLI4_QENTRY_MAX;
*r_id = U16_MAX;
rc = -EINVAL;
}
return rc;
}
u32
sli_fc_response_length(struct sli4 *sli4, u8 *cqe)
{
struct sli4_fc_wcqe *wcqe = (void *)cqe;
return le32_to_cpu(wcqe->wqe_specific_1);
}
u32
sli_fc_io_length(struct sli4 *sli4, u8 *cqe)
{
struct sli4_fc_wcqe *wcqe = (void *)cqe;
return le32_to_cpu(wcqe->wqe_specific_1);
}
int
sli_fc_els_did(struct sli4 *sli4, u8 *cqe, u32 *d_id)
{
struct sli4_fc_wcqe *wcqe = (void *)cqe;
*d_id = 0;
if (wcqe->status)
return -EIO;
*d_id = le32_to_cpu(wcqe->wqe_specific_2) & 0x00ffffff;
return 0;
}
u32
sli_fc_ext_status(struct sli4 *sli4, u8 *cqe)
{
struct sli4_fc_wcqe *wcqe = (void *)cqe;
u32 mask;
switch (wcqe->status) {
case SLI4_FC_WCQE_STATUS_FCP_RSP_FAILURE:
mask = U32_MAX;
break;
case SLI4_FC_WCQE_STATUS_LOCAL_REJECT:
case SLI4_FC_WCQE_STATUS_CMD_REJECT:
mask = 0xff;
break;
case SLI4_FC_WCQE_STATUS_NPORT_RJT:
case SLI4_FC_WCQE_STATUS_FABRIC_RJT:
case SLI4_FC_WCQE_STATUS_NPORT_BSY:
case SLI4_FC_WCQE_STATUS_FABRIC_BSY:
case SLI4_FC_WCQE_STATUS_LS_RJT:
mask = U32_MAX;
break;
case SLI4_FC_WCQE_STATUS_DI_ERROR:
mask = U32_MAX;
break;
default:
mask = 0;
}
return le32_to_cpu(wcqe->wqe_specific_2) & mask;
}
int
sli_fc_rqe_rqid_and_index(struct sli4 *sli4, u8 *cqe, u16 *rq_id, u32 *index)
{
int rc = -EIO;
u8 code = 0;
u16 rq_element_index;
*rq_id = 0;
*index = U32_MAX;
code = cqe[SLI4_CQE_CODE_OFFSET];
/* Retrieve the RQ index from the completion */
if (code == SLI4_CQE_CODE_RQ_ASYNC) {
struct sli4_fc_async_rcqe *rcqe = (void *)cqe;
*rq_id = le16_to_cpu(rcqe->fcfi_rq_id_word) & SLI4_RACQE_RQ_ID;
rq_element_index =
le16_to_cpu(rcqe->rq_elmt_indx_word) & SLI4_RACQE_RQ_EL_INDX;
*index = rq_element_index;
if (rcqe->status == SLI4_FC_ASYNC_RQ_SUCCESS) {
rc = 0;
} else {
rc = rcqe->status;
efc_log_info(sli4, "status=%02x (%s) rq_id=%d\n",
rcqe->status,
sli_fc_get_status_string(rcqe->status),
le16_to_cpu(rcqe->fcfi_rq_id_word) &
SLI4_RACQE_RQ_ID);
efc_log_info(sli4, "pdpl=%x sof=%02x eof=%02x hdpl=%x\n",
le16_to_cpu(rcqe->data_placement_length),
rcqe->sof_byte, rcqe->eof_byte,
rcqe->hdpl_byte & SLI4_RACQE_HDPL);
}
} else if (code == SLI4_CQE_CODE_RQ_ASYNC_V1) {
struct sli4_fc_async_rcqe_v1 *rcqe_v1 = (void *)cqe;
*rq_id = le16_to_cpu(rcqe_v1->rq_id);
rq_element_index =
(le16_to_cpu(rcqe_v1->rq_elmt_indx_word) &
SLI4_RACQE_RQ_EL_INDX);
*index = rq_element_index;
if (rcqe_v1->status == SLI4_FC_ASYNC_RQ_SUCCESS) {
rc = 0;
} else {
rc = rcqe_v1->status;
efc_log_info(sli4, "status=%02x (%s) rq_id=%d, index=%x\n",
rcqe_v1->status,
sli_fc_get_status_string(rcqe_v1->status),
le16_to_cpu(rcqe_v1->rq_id), rq_element_index);
efc_log_info(sli4, "pdpl=%x sof=%02x eof=%02x hdpl=%x\n",
le16_to_cpu(rcqe_v1->data_placement_length),
rcqe_v1->sof_byte, rcqe_v1->eof_byte,
rcqe_v1->hdpl_byte & SLI4_RACQE_HDPL);
}
} else if (code == SLI4_CQE_CODE_OPTIMIZED_WRITE_CMD) {
struct sli4_fc_optimized_write_cmd_cqe *optcqe = (void *)cqe;
*rq_id = le16_to_cpu(optcqe->rq_id);
*index = le16_to_cpu(optcqe->w1) & SLI4_OCQE_RQ_EL_INDX;
if (optcqe->status == SLI4_FC_ASYNC_RQ_SUCCESS) {
rc = 0;
} else {
rc = optcqe->status;
efc_log_info(sli4, "stat=%02x (%s) rqid=%d, idx=%x pdpl=%x\n",
optcqe->status,
sli_fc_get_status_string(optcqe->status),
le16_to_cpu(optcqe->rq_id), *index,
le16_to_cpu(optcqe->data_placement_length));
efc_log_info(sli4, "hdpl=%x oox=%d agxr=%d xri=0x%x rpi=%x\n",
(optcqe->hdpl_vld & SLI4_OCQE_HDPL),
(optcqe->flags1 & SLI4_OCQE_OOX),
(optcqe->flags1 & SLI4_OCQE_AGXR),
optcqe->xri, le16_to_cpu(optcqe->rpi));
}
} else if (code == SLI4_CQE_CODE_RQ_COALESCING) {
struct sli4_fc_coalescing_rcqe *rcqe = (void *)cqe;
rq_element_index = (le16_to_cpu(rcqe->rq_elmt_indx_word) &
SLI4_RCQE_RQ_EL_INDX);
*rq_id = le16_to_cpu(rcqe->rq_id);
if (rcqe->status == SLI4_FC_COALESCE_RQ_SUCCESS) {
*index = rq_element_index;
rc = 0;
} else {
*index = U32_MAX;
rc = rcqe->status;
efc_log_info(sli4, "stat=%02x (%s) rq_id=%d, idx=%x\n",
rcqe->status,
sli_fc_get_status_string(rcqe->status),
le16_to_cpu(rcqe->rq_id), rq_element_index);
efc_log_info(sli4, "rq_id=%#x sdpl=%x\n",
le16_to_cpu(rcqe->rq_id),
le16_to_cpu(rcqe->seq_placement_length));
}
} else {
struct sli4_fc_async_rcqe *rcqe = (void *)cqe;
*index = U32_MAX;
rc = rcqe->status;
efc_log_info(sli4, "status=%02x rq_id=%d, index=%x pdpl=%x\n",
rcqe->status,
le16_to_cpu(rcqe->fcfi_rq_id_word) & SLI4_RACQE_RQ_ID,
(le16_to_cpu(rcqe->rq_elmt_indx_word) & SLI4_RACQE_RQ_EL_INDX),
le16_to_cpu(rcqe->data_placement_length));
efc_log_info(sli4, "sof=%02x eof=%02x hdpl=%x\n",
rcqe->sof_byte, rcqe->eof_byte,
rcqe->hdpl_byte & SLI4_RACQE_HDPL);
}
return rc;
}
static int
sli_bmbx_wait(struct sli4 *sli4, u32 msec)
{
u32 val;
unsigned long end;
/* Wait for the bootstrap mailbox to report "ready" */
end = jiffies + msecs_to_jiffies(msec);
do {
val = readl(sli4->reg[0] + SLI4_BMBX_REG);
if (val & SLI4_BMBX_RDY)
return 0;
usleep_range(1000, 2000);
} while (time_before(jiffies, end));
return -EIO;
}
static int
sli_bmbx_write(struct sli4 *sli4)
{
u32 val;
/* write buffer location to bootstrap mailbox register */
val = sli_bmbx_write_hi(sli4->bmbx.phys);
writel(val, (sli4->reg[0] + SLI4_BMBX_REG));
if (sli_bmbx_wait(sli4, SLI4_BMBX_DELAY_US)) {
efc_log_crit(sli4, "BMBX WRITE_HI failed\n");
return -EIO;
}
val = sli_bmbx_write_lo(sli4->bmbx.phys);
writel(val, (sli4->reg[0] + SLI4_BMBX_REG));
/* wait for SLI Port to set ready bit */
return sli_bmbx_wait(sli4, SLI4_BMBX_TIMEOUT_MSEC);
}
int
sli_bmbx_command(struct sli4 *sli4)
{
void *cqe = (u8 *)sli4->bmbx.virt + SLI4_BMBX_SIZE;
if (sli_fw_error_status(sli4) > 0) {
efc_log_crit(sli4, "Chip is in an error state -Mailbox command rejected");
efc_log_crit(sli4, " status=%#x error1=%#x error2=%#x\n",
sli_reg_read_status(sli4),
sli_reg_read_err1(sli4),
sli_reg_read_err2(sli4));
return -EIO;
}
/* Submit a command to the bootstrap mailbox and check the status */
if (sli_bmbx_write(sli4)) {
efc_log_crit(sli4, "bmbx write fail phys=%pad reg=%#x\n",
&sli4->bmbx.phys, readl(sli4->reg[0] + SLI4_BMBX_REG));
return -EIO;
}
/* check completion queue entry status */
if (le32_to_cpu(((struct sli4_mcqe *)cqe)->dw3_flags) &
SLI4_MCQE_VALID) {
return sli_cqe_mq(sli4, cqe);
}
efc_log_crit(sli4, "invalid or wrong type\n");
return -EIO;
}
int
sli_cmd_config_link(struct sli4 *sli4, void *buf)
{
struct sli4_cmd_config_link *config_link = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
config_link->hdr.command = SLI4_MBX_CMD_CONFIG_LINK;
/* Port interprets zero in a field as "use default value" */
return 0;
}
int
sli_cmd_down_link(struct sli4 *sli4, void *buf)
{
struct sli4_mbox_command_header *hdr = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
hdr->command = SLI4_MBX_CMD_DOWN_LINK;
/* Port interprets zero in a field as "use default value" */
return 0;
}
int
sli_cmd_dump_type4(struct sli4 *sli4, void *buf, u16 wki)
{
struct sli4_cmd_dump4 *cmd = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
cmd->hdr.command = SLI4_MBX_CMD_DUMP;
cmd->type_dword = cpu_to_le32(0x4);
cmd->wki_selection = cpu_to_le16(wki);
return 0;
}
int
sli_cmd_common_read_transceiver_data(struct sli4 *sli4, void *buf, u32 page_num,
struct efc_dma *dma)
{
struct sli4_rqst_cmn_read_transceiver_data *req = NULL;
u32 psize;
if (!dma)
psize = SLI4_CFG_PYLD_LENGTH(cmn_read_transceiver_data);
else
psize = dma->size;
req = sli_config_cmd_init(sli4, buf, psize, dma);
if (!req)
return -EIO;
sli_cmd_fill_hdr(&req->hdr, SLI4_CMN_READ_TRANS_DATA,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_read_transceiver_data));
req->page_number = cpu_to_le32(page_num);
req->port = cpu_to_le32(sli4->port_number);
return 0;
}
int
sli_cmd_read_link_stats(struct sli4 *sli4, void *buf, u8 req_ext_counters,
u8 clear_overflow_flags,
u8 clear_all_counters)
{
struct sli4_cmd_read_link_stats *cmd = buf;
u32 flags;
memset(buf, 0, SLI4_BMBX_SIZE);
cmd->hdr.command = SLI4_MBX_CMD_READ_LNK_STAT;
flags = 0;
if (req_ext_counters)
flags |= SLI4_READ_LNKSTAT_REC;
if (clear_all_counters)
flags |= SLI4_READ_LNKSTAT_CLRC;
if (clear_overflow_flags)
flags |= SLI4_READ_LNKSTAT_CLOF;
cmd->dw1_flags = cpu_to_le32(flags);
return 0;
}
int
sli_cmd_read_status(struct sli4 *sli4, void *buf, u8 clear_counters)
{
struct sli4_cmd_read_status *cmd = buf;
u32 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
cmd->hdr.command = SLI4_MBX_CMD_READ_STATUS;
if (clear_counters)
flags |= SLI4_READSTATUS_CLEAR_COUNTERS;
else
flags &= ~SLI4_READSTATUS_CLEAR_COUNTERS;
cmd->dw1_flags = cpu_to_le32(flags);
return 0;
}
int
sli_cmd_init_link(struct sli4 *sli4, void *buf, u32 speed, u8 reset_alpa)
{
struct sli4_cmd_init_link *init_link = buf;
u32 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
init_link->hdr.command = SLI4_MBX_CMD_INIT_LINK;
init_link->sel_reset_al_pa_dword =
cpu_to_le32(reset_alpa);
flags &= ~SLI4_INIT_LINK_F_LOOPBACK;
init_link->link_speed_sel_code = cpu_to_le32(speed);
switch (speed) {
case SLI4_LINK_SPEED_1G:
case SLI4_LINK_SPEED_2G:
case SLI4_LINK_SPEED_4G:
case SLI4_LINK_SPEED_8G:
case SLI4_LINK_SPEED_16G:
case SLI4_LINK_SPEED_32G:
case SLI4_LINK_SPEED_64G:
flags |= SLI4_INIT_LINK_F_FIXED_SPEED;
break;
case SLI4_LINK_SPEED_10G:
efc_log_info(sli4, "unsupported FC speed %d\n", speed);
init_link->flags0 = cpu_to_le32(flags);
return -EIO;
}
switch (sli4->topology) {
case SLI4_READ_CFG_TOPO_FC:
/* Attempt P2P but failover to FC-AL */
flags |= SLI4_INIT_LINK_F_FAIL_OVER;
flags |= SLI4_INIT_LINK_F_P2P_FAIL_OVER;
break;
case SLI4_READ_CFG_TOPO_FC_AL:
flags |= SLI4_INIT_LINK_F_FCAL_ONLY;
if (speed == SLI4_LINK_SPEED_16G ||
speed == SLI4_LINK_SPEED_32G) {
efc_log_info(sli4, "unsupported FC-AL speed %d\n",
speed);
init_link->flags0 = cpu_to_le32(flags);
return -EIO;
}
break;
case SLI4_READ_CFG_TOPO_NON_FC_AL:
flags |= SLI4_INIT_LINK_F_P2P_ONLY;
break;
default:
efc_log_info(sli4, "unsupported topology %#x\n", sli4->topology);
init_link->flags0 = cpu_to_le32(flags);
return -EIO;
}
flags &= ~SLI4_INIT_LINK_F_UNFAIR;
flags &= ~SLI4_INIT_LINK_F_NO_LIRP;
flags &= ~SLI4_INIT_LINK_F_LOOP_VALID_CHK;
flags &= ~SLI4_INIT_LINK_F_NO_LISA;
flags &= ~SLI4_INIT_LINK_F_PICK_HI_ALPA;
init_link->flags0 = cpu_to_le32(flags);
return 0;
}
int
sli_cmd_init_vfi(struct sli4 *sli4, void *buf, u16 vfi, u16 fcfi, u16 vpi)
{
struct sli4_cmd_init_vfi *init_vfi = buf;
u16 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
init_vfi->hdr.command = SLI4_MBX_CMD_INIT_VFI;
init_vfi->vfi = cpu_to_le16(vfi);
init_vfi->fcfi = cpu_to_le16(fcfi);
/*
* If the VPI is valid, initialize it at the same time as
* the VFI
*/
if (vpi != U16_MAX) {
flags |= SLI4_INIT_VFI_FLAG_VP;
init_vfi->flags0_word = cpu_to_le16(flags);
init_vfi->vpi = cpu_to_le16(vpi);
}
return 0;
}
int
sli_cmd_init_vpi(struct sli4 *sli4, void *buf, u16 vpi, u16 vfi)
{
struct sli4_cmd_init_vpi *init_vpi = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
init_vpi->hdr.command = SLI4_MBX_CMD_INIT_VPI;
init_vpi->vpi = cpu_to_le16(vpi);
init_vpi->vfi = cpu_to_le16(vfi);
return 0;
}
int
sli_cmd_post_xri(struct sli4 *sli4, void *buf, u16 xri_base, u16 xri_count)
{
struct sli4_cmd_post_xri *post_xri = buf;
u16 xri_count_flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
post_xri->hdr.command = SLI4_MBX_CMD_POST_XRI;
post_xri->xri_base = cpu_to_le16(xri_base);
xri_count_flags = xri_count & SLI4_POST_XRI_COUNT;
xri_count_flags |= SLI4_POST_XRI_FLAG_ENX;
xri_count_flags |= SLI4_POST_XRI_FLAG_VAL;
post_xri->xri_count_flags = cpu_to_le16(xri_count_flags);
return 0;
}
int
sli_cmd_release_xri(struct sli4 *sli4, void *buf, u8 num_xri)
{
struct sli4_cmd_release_xri *release_xri = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
release_xri->hdr.command = SLI4_MBX_CMD_RELEASE_XRI;
release_xri->xri_count_word = cpu_to_le16(num_xri &
SLI4_RELEASE_XRI_COUNT);
return 0;
}
static int
sli_cmd_read_config(struct sli4 *sli4, void *buf)
{
struct sli4_cmd_read_config *read_config = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
read_config->hdr.command = SLI4_MBX_CMD_READ_CONFIG;
return 0;
}
int
sli_cmd_read_nvparms(struct sli4 *sli4, void *buf)
{
struct sli4_cmd_read_nvparms *read_nvparms = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
read_nvparms->hdr.command = SLI4_MBX_CMD_READ_NVPARMS;
return 0;
}
int
sli_cmd_write_nvparms(struct sli4 *sli4, void *buf, u8 *wwpn, u8 *wwnn,
u8 hard_alpa, u32 preferred_d_id)
{
struct sli4_cmd_write_nvparms *write_nvparms = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
write_nvparms->hdr.command = SLI4_MBX_CMD_WRITE_NVPARMS;
memcpy(write_nvparms->wwpn, wwpn, 8);
memcpy(write_nvparms->wwnn, wwnn, 8);
write_nvparms->hard_alpa_d_id =
cpu_to_le32((preferred_d_id << 8) | hard_alpa);
return 0;
}
static int
sli_cmd_read_rev(struct sli4 *sli4, void *buf, struct efc_dma *vpd)
{
struct sli4_cmd_read_rev *read_rev = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
read_rev->hdr.command = SLI4_MBX_CMD_READ_REV;
if (vpd && vpd->size) {
read_rev->flags0_word |= cpu_to_le16(SLI4_READ_REV_FLAG_VPD);
read_rev->available_length_dword =
cpu_to_le32(vpd->size &
SLI4_READ_REV_AVAILABLE_LENGTH);
read_rev->hostbuf.low =
cpu_to_le32(lower_32_bits(vpd->phys));
read_rev->hostbuf.high =
cpu_to_le32(upper_32_bits(vpd->phys));
}
return 0;
}
int
sli_cmd_read_sparm64(struct sli4 *sli4, void *buf, struct efc_dma *dma, u16 vpi)
{
struct sli4_cmd_read_sparm64 *read_sparm64 = buf;
if (vpi == U16_MAX) {
efc_log_err(sli4, "special VPI not supported!!!\n");
return -EIO;
}
if (!dma || !dma->phys) {
efc_log_err(sli4, "bad DMA buffer\n");
return -EIO;
}
memset(buf, 0, SLI4_BMBX_SIZE);
read_sparm64->hdr.command = SLI4_MBX_CMD_READ_SPARM64;
read_sparm64->bde_64.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(dma->size & SLI4_BDE_LEN_MASK));
read_sparm64->bde_64.u.data.low =
cpu_to_le32(lower_32_bits(dma->phys));
read_sparm64->bde_64.u.data.high =
cpu_to_le32(upper_32_bits(dma->phys));
read_sparm64->vpi = cpu_to_le16(vpi);
return 0;
}
int
sli_cmd_read_topology(struct sli4 *sli4, void *buf, struct efc_dma *dma)
{
struct sli4_cmd_read_topology *read_topo = buf;
if (!dma || !dma->size)
return -EIO;
if (dma->size < SLI4_MIN_LOOP_MAP_BYTES) {
efc_log_err(sli4, "loop map buffer too small %zx\n", dma->size);
return -EIO;
}
memset(buf, 0, SLI4_BMBX_SIZE);
read_topo->hdr.command = SLI4_MBX_CMD_READ_TOPOLOGY;
memset(dma->virt, 0, dma->size);
read_topo->bde_loop_map.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(dma->size & SLI4_BDE_LEN_MASK));
read_topo->bde_loop_map.u.data.low =
cpu_to_le32(lower_32_bits(dma->phys));
read_topo->bde_loop_map.u.data.high =
cpu_to_le32(upper_32_bits(dma->phys));
return 0;
}
int
sli_cmd_reg_fcfi(struct sli4 *sli4, void *buf, u16 index,
struct sli4_cmd_rq_cfg *rq_cfg)
{
struct sli4_cmd_reg_fcfi *reg_fcfi = buf;
u32 i;
memset(buf, 0, SLI4_BMBX_SIZE);
reg_fcfi->hdr.command = SLI4_MBX_CMD_REG_FCFI;
reg_fcfi->fcf_index = cpu_to_le16(index);
for (i = 0; i < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; i++) {
switch (i) {
case 0:
reg_fcfi->rqid0 = rq_cfg[0].rq_id;
break;
case 1:
reg_fcfi->rqid1 = rq_cfg[1].rq_id;
break;
case 2:
reg_fcfi->rqid2 = rq_cfg[2].rq_id;
break;
case 3:
reg_fcfi->rqid3 = rq_cfg[3].rq_id;
break;
}
reg_fcfi->rq_cfg[i].r_ctl_mask = rq_cfg[i].r_ctl_mask;
reg_fcfi->rq_cfg[i].r_ctl_match = rq_cfg[i].r_ctl_match;
reg_fcfi->rq_cfg[i].type_mask = rq_cfg[i].type_mask;
reg_fcfi->rq_cfg[i].type_match = rq_cfg[i].type_match;
}
return 0;
}
int
sli_cmd_reg_fcfi_mrq(struct sli4 *sli4, void *buf, u8 mode, u16 fcf_index,
u8 rq_selection_policy, u8 mrq_bit_mask, u16 num_mrqs,
struct sli4_cmd_rq_cfg *rq_cfg)
{
struct sli4_cmd_reg_fcfi_mrq *reg_fcfi_mrq = buf;
u32 i;
u32 mrq_flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
reg_fcfi_mrq->hdr.command = SLI4_MBX_CMD_REG_FCFI_MRQ;
if (mode == SLI4_CMD_REG_FCFI_SET_FCFI_MODE) {
reg_fcfi_mrq->fcf_index = cpu_to_le16(fcf_index);
goto done;
}
reg_fcfi_mrq->dw8_vlan = cpu_to_le32(SLI4_REGFCFI_MRQ_MODE);
for (i = 0; i < SLI4_CMD_REG_FCFI_NUM_RQ_CFG; i++) {
reg_fcfi_mrq->rq_cfg[i].r_ctl_mask = rq_cfg[i].r_ctl_mask;
reg_fcfi_mrq->rq_cfg[i].r_ctl_match = rq_cfg[i].r_ctl_match;
reg_fcfi_mrq->rq_cfg[i].type_mask = rq_cfg[i].type_mask;
reg_fcfi_mrq->rq_cfg[i].type_match = rq_cfg[i].type_match;
switch (i) {
case 3:
reg_fcfi_mrq->rqid3 = rq_cfg[i].rq_id;
break;
case 2:
reg_fcfi_mrq->rqid2 = rq_cfg[i].rq_id;
break;
case 1:
reg_fcfi_mrq->rqid1 = rq_cfg[i].rq_id;
break;
case 0:
reg_fcfi_mrq->rqid0 = rq_cfg[i].rq_id;
break;
}
}
mrq_flags = num_mrqs & SLI4_REGFCFI_MRQ_MASK_NUM_PAIRS;
mrq_flags |= (mrq_bit_mask << 8);
mrq_flags |= (rq_selection_policy << 12);
reg_fcfi_mrq->dw9_mrqflags = cpu_to_le32(mrq_flags);
done:
return 0;
}
int
sli_cmd_reg_rpi(struct sli4 *sli4, void *buf, u32 rpi, u32 vpi, u32 fc_id,
struct efc_dma *dma, u8 update, u8 enable_t10_pi)
{
struct sli4_cmd_reg_rpi *reg_rpi = buf;
u32 rportid_flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
reg_rpi->hdr.command = SLI4_MBX_CMD_REG_RPI;
reg_rpi->rpi = cpu_to_le16(rpi);
rportid_flags = fc_id & SLI4_REGRPI_REMOTE_N_PORTID;
if (update)
rportid_flags |= SLI4_REGRPI_UPD;
else
rportid_flags &= ~SLI4_REGRPI_UPD;
if (enable_t10_pi)
rportid_flags |= SLI4_REGRPI_ETOW;
else
rportid_flags &= ~SLI4_REGRPI_ETOW;
reg_rpi->dw2_rportid_flags = cpu_to_le32(rportid_flags);
reg_rpi->bde_64.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(SLI4_REG_RPI_BUF_LEN & SLI4_BDE_LEN_MASK));
reg_rpi->bde_64.u.data.low =
cpu_to_le32(lower_32_bits(dma->phys));
reg_rpi->bde_64.u.data.high =
cpu_to_le32(upper_32_bits(dma->phys));
reg_rpi->vpi = cpu_to_le16(vpi);
return 0;
}
int
sli_cmd_reg_vfi(struct sli4 *sli4, void *buf, size_t size,
u16 vfi, u16 fcfi, struct efc_dma dma,
u16 vpi, __be64 sli_wwpn, u32 fc_id)
{
struct sli4_cmd_reg_vfi *reg_vfi = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
reg_vfi->hdr.command = SLI4_MBX_CMD_REG_VFI;
reg_vfi->vfi = cpu_to_le16(vfi);
reg_vfi->fcfi = cpu_to_le16(fcfi);
reg_vfi->sparm.bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(SLI4_REG_RPI_BUF_LEN & SLI4_BDE_LEN_MASK));
reg_vfi->sparm.u.data.low =
cpu_to_le32(lower_32_bits(dma.phys));
reg_vfi->sparm.u.data.high =
cpu_to_le32(upper_32_bits(dma.phys));
reg_vfi->e_d_tov = cpu_to_le32(sli4->e_d_tov);
reg_vfi->r_a_tov = cpu_to_le32(sli4->r_a_tov);
reg_vfi->dw0w1_flags |= cpu_to_le16(SLI4_REGVFI_VP);
reg_vfi->vpi = cpu_to_le16(vpi);
memcpy(reg_vfi->wwpn, &sli_wwpn, sizeof(reg_vfi->wwpn));
reg_vfi->dw10_lportid_flags = cpu_to_le32(fc_id);
return 0;
}
int
sli_cmd_reg_vpi(struct sli4 *sli4, void *buf, u32 fc_id, __be64 sli_wwpn,
u16 vpi, u16 vfi, bool update)
{
struct sli4_cmd_reg_vpi *reg_vpi = buf;
u32 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
reg_vpi->hdr.command = SLI4_MBX_CMD_REG_VPI;
flags = (fc_id & SLI4_REGVPI_LOCAL_N_PORTID);
if (update)
flags |= SLI4_REGVPI_UPD;
else
flags &= ~SLI4_REGVPI_UPD;
reg_vpi->dw2_lportid_flags = cpu_to_le32(flags);
memcpy(reg_vpi->wwpn, &sli_wwpn, sizeof(reg_vpi->wwpn));
reg_vpi->vpi = cpu_to_le16(vpi);
reg_vpi->vfi = cpu_to_le16(vfi);
return 0;
}
static int
sli_cmd_request_features(struct sli4 *sli4, void *buf, u32 features_mask,
bool query)
{
struct sli4_cmd_request_features *req_features = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
req_features->hdr.command = SLI4_MBX_CMD_RQST_FEATURES;
if (query)
req_features->dw1_qry = cpu_to_le32(SLI4_REQFEAT_QRY);
req_features->cmd = cpu_to_le32(features_mask);
return 0;
}
int
sli_cmd_unreg_fcfi(struct sli4 *sli4, void *buf, u16 indicator)
{
struct sli4_cmd_unreg_fcfi *unreg_fcfi = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
unreg_fcfi->hdr.command = SLI4_MBX_CMD_UNREG_FCFI;
unreg_fcfi->fcfi = cpu_to_le16(indicator);
return 0;
}
int
sli_cmd_unreg_rpi(struct sli4 *sli4, void *buf, u16 indicator,
enum sli4_resource which, u32 fc_id)
{
struct sli4_cmd_unreg_rpi *unreg_rpi = buf;
u32 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
unreg_rpi->hdr.command = SLI4_MBX_CMD_UNREG_RPI;
switch (which) {
case SLI4_RSRC_RPI:
flags |= SLI4_UNREG_RPI_II_RPI;
if (fc_id == U32_MAX)
break;
flags |= SLI4_UNREG_RPI_DP;
unreg_rpi->dw2_dest_n_portid =
cpu_to_le32(fc_id & SLI4_UNREG_RPI_DEST_N_PORTID_MASK);
break;
case SLI4_RSRC_VPI:
flags |= SLI4_UNREG_RPI_II_VPI;
break;
case SLI4_RSRC_VFI:
flags |= SLI4_UNREG_RPI_II_VFI;
break;
case SLI4_RSRC_FCFI:
flags |= SLI4_UNREG_RPI_II_FCFI;
break;
default:
efc_log_info(sli4, "unknown type %#x\n", which);
return -EIO;
}
unreg_rpi->dw1w1_flags = cpu_to_le16(flags);
unreg_rpi->index = cpu_to_le16(indicator);
return 0;
}
int
sli_cmd_unreg_vfi(struct sli4 *sli4, void *buf, u16 index, u32 which)
{
struct sli4_cmd_unreg_vfi *unreg_vfi = buf;
memset(buf, 0, SLI4_BMBX_SIZE);
unreg_vfi->hdr.command = SLI4_MBX_CMD_UNREG_VFI;
switch (which) {
case SLI4_UNREG_TYPE_DOMAIN:
unreg_vfi->index = cpu_to_le16(index);
break;
case SLI4_UNREG_TYPE_FCF:
unreg_vfi->index = cpu_to_le16(index);
break;
case SLI4_UNREG_TYPE_ALL:
unreg_vfi->index = cpu_to_le16(U32_MAX);
break;
default:
return -EIO;
}
if (which != SLI4_UNREG_TYPE_DOMAIN)
unreg_vfi->dw2_flags = cpu_to_le16(SLI4_UNREG_VFI_II_FCFI);
return 0;
}
int
sli_cmd_unreg_vpi(struct sli4 *sli4, void *buf, u16 indicator, u32 which)
{
struct sli4_cmd_unreg_vpi *unreg_vpi = buf;
u32 flags = 0;
memset(buf, 0, SLI4_BMBX_SIZE);
unreg_vpi->hdr.command = SLI4_MBX_CMD_UNREG_VPI;
unreg_vpi->index = cpu_to_le16(indicator);
switch (which) {
case SLI4_UNREG_TYPE_PORT:
flags |= SLI4_UNREG_VPI_II_VPI;
break;
case SLI4_UNREG_TYPE_DOMAIN:
flags |= SLI4_UNREG_VPI_II_VFI;
break;
case SLI4_UNREG_TYPE_FCF:
flags |= SLI4_UNREG_VPI_II_FCFI;
break;
case SLI4_UNREG_TYPE_ALL:
/* override indicator */
unreg_vpi->index = cpu_to_le16(U32_MAX);
flags |= SLI4_UNREG_VPI_II_FCFI;
break;
default:
return -EIO;
}
unreg_vpi->dw2w0_flags = cpu_to_le16(flags);
return 0;
}
static int
sli_cmd_common_modify_eq_delay(struct sli4 *sli4, void *buf,
struct sli4_queue *q, int num_q, u32 shift,
u32 delay_mult)
{
struct sli4_rqst_cmn_modify_eq_delay *req = NULL;
int i;
req = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(cmn_modify_eq_delay), NULL);
if (!req)
return -EIO;
sli_cmd_fill_hdr(&req->hdr, SLI4_CMN_MODIFY_EQ_DELAY,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_modify_eq_delay));
req->num_eq = cpu_to_le32(num_q);
for (i = 0; i < num_q; i++) {
req->eq_delay_record[i].eq_id = cpu_to_le32(q[i].id);
req->eq_delay_record[i].phase = cpu_to_le32(shift);
req->eq_delay_record[i].delay_multiplier =
cpu_to_le32(delay_mult);
}
return 0;
}
void
sli4_cmd_lowlevel_set_watchdog(struct sli4 *sli4, void *buf,
size_t size, u16 timeout)
{
struct sli4_rqst_lowlevel_set_watchdog *req = NULL;
req = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(lowlevel_set_watchdog), NULL);
if (!req)
return;
sli_cmd_fill_hdr(&req->hdr, SLI4_OPC_LOWLEVEL_SET_WATCHDOG,
SLI4_SUBSYSTEM_LOWLEVEL, CMD_V0,
SLI4_RQST_PYLD_LEN(lowlevel_set_watchdog));
req->watchdog_timeout = cpu_to_le16(timeout);
}
static int
sli_cmd_common_get_cntl_attributes(struct sli4 *sli4, void *buf,
struct efc_dma *dma)
{
struct sli4_rqst_hdr *hdr = NULL;
hdr = sli_config_cmd_init(sli4, buf, SLI4_RQST_CMDSZ(hdr), dma);
if (!hdr)
return -EIO;
hdr->opcode = SLI4_CMN_GET_CNTL_ATTRIBUTES;
hdr->subsystem = SLI4_SUBSYSTEM_COMMON;
hdr->request_length = cpu_to_le32(dma->size);
return 0;
}
static int
sli_cmd_common_get_cntl_addl_attributes(struct sli4 *sli4, void *buf,
struct efc_dma *dma)
{
struct sli4_rqst_hdr *hdr = NULL;
hdr = sli_config_cmd_init(sli4, buf, SLI4_RQST_CMDSZ(hdr), dma);
if (!hdr)
return -EIO;
hdr->opcode = SLI4_CMN_GET_CNTL_ADDL_ATTRS;
hdr->subsystem = SLI4_SUBSYSTEM_COMMON;
hdr->request_length = cpu_to_le32(dma->size);
return 0;
}
int
sli_cmd_common_nop(struct sli4 *sli4, void *buf, uint64_t context)
{
struct sli4_rqst_cmn_nop *nop = NULL;
nop = sli_config_cmd_init(sli4, buf, SLI4_CFG_PYLD_LENGTH(cmn_nop),
NULL);
if (!nop)
return -EIO;
sli_cmd_fill_hdr(&nop->hdr, SLI4_CMN_NOP, SLI4_SUBSYSTEM_COMMON,
CMD_V0, SLI4_RQST_PYLD_LEN(cmn_nop));
memcpy(&nop->context, &context, sizeof(context));
return 0;
}
int
sli_cmd_common_get_resource_extent_info(struct sli4 *sli4, void *buf, u16 rtype)
{
struct sli4_rqst_cmn_get_resource_extent_info *ext = NULL;
ext = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_get_resource_extent_info), NULL);
if (!ext)
return -EIO;
sli_cmd_fill_hdr(&ext->hdr, SLI4_CMN_GET_RSC_EXTENT_INFO,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_get_resource_extent_info));
ext->resource_type = cpu_to_le16(rtype);
return 0;
}
int
sli_cmd_common_get_sli4_parameters(struct sli4 *sli4, void *buf)
{
struct sli4_rqst_hdr *hdr = NULL;
hdr = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(cmn_get_sli4_params), NULL);
if (!hdr)
return -EIO;
hdr->opcode = SLI4_CMN_GET_SLI4_PARAMS;
hdr->subsystem = SLI4_SUBSYSTEM_COMMON;
hdr->request_length = SLI4_RQST_PYLD_LEN(cmn_get_sli4_params);
return 0;
}
static int
sli_cmd_common_get_port_name(struct sli4 *sli4, void *buf)
{
struct sli4_rqst_cmn_get_port_name *pname;
pname = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(cmn_get_port_name), NULL);
if (!pname)
return -EIO;
sli_cmd_fill_hdr(&pname->hdr, SLI4_CMN_GET_PORT_NAME,
SLI4_SUBSYSTEM_COMMON, CMD_V1,
SLI4_RQST_PYLD_LEN(cmn_get_port_name));
/* Set the port type value (ethernet=0, FC=1) for V1 commands */
pname->port_type = SLI4_PORT_TYPE_FC;
return 0;
}
int
sli_cmd_common_write_object(struct sli4 *sli4, void *buf, u16 noc,
u16 eof, u32 desired_write_length,
u32 offset, char *obj_name,
struct efc_dma *dma)
{
struct sli4_rqst_cmn_write_object *wr_obj = NULL;
struct sli4_bde *bde;
u32 dwflags = 0;
wr_obj = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_write_object) + sizeof(*bde), NULL);
if (!wr_obj)
return -EIO;
sli_cmd_fill_hdr(&wr_obj->hdr, SLI4_CMN_WRITE_OBJECT,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN_VAR(cmn_write_object, sizeof(*bde)));
if (noc)
dwflags |= SLI4_RQ_DES_WRITE_LEN_NOC;
if (eof)
dwflags |= SLI4_RQ_DES_WRITE_LEN_EOF;
dwflags |= (desired_write_length & SLI4_RQ_DES_WRITE_LEN);
wr_obj->desired_write_len_dword = cpu_to_le32(dwflags);
wr_obj->write_offset = cpu_to_le32(offset);
strncpy(wr_obj->object_name, obj_name, sizeof(wr_obj->object_name) - 1);
wr_obj->host_buffer_descriptor_count = cpu_to_le32(1);
bde = (struct sli4_bde *)wr_obj->host_buffer_descriptor;
/* Setup to transfer xfer_size bytes to device */
bde->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(desired_write_length & SLI4_BDE_LEN_MASK));
bde->u.data.low = cpu_to_le32(lower_32_bits(dma->phys));
bde->u.data.high = cpu_to_le32(upper_32_bits(dma->phys));
return 0;
}
int
sli_cmd_common_delete_object(struct sli4 *sli4, void *buf, char *obj_name)
{
struct sli4_rqst_cmn_delete_object *req = NULL;
req = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_delete_object), NULL);
if (!req)
return -EIO;
sli_cmd_fill_hdr(&req->hdr, SLI4_CMN_DELETE_OBJECT,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_delete_object));
strncpy(req->object_name, obj_name, sizeof(req->object_name) - 1);
return 0;
}
int
sli_cmd_common_read_object(struct sli4 *sli4, void *buf, u32 desired_read_len,
u32 offset, char *obj_name, struct efc_dma *dma)
{
struct sli4_rqst_cmn_read_object *rd_obj = NULL;
struct sli4_bde *bde;
rd_obj = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_read_object) + sizeof(*bde), NULL);
if (!rd_obj)
return -EIO;
sli_cmd_fill_hdr(&rd_obj->hdr, SLI4_CMN_READ_OBJECT,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN_VAR(cmn_read_object, sizeof(*bde)));
rd_obj->desired_read_length_dword =
cpu_to_le32(desired_read_len & SLI4_REQ_DESIRE_READLEN);
rd_obj->read_offset = cpu_to_le32(offset);
strncpy(rd_obj->object_name, obj_name, sizeof(rd_obj->object_name) - 1);
rd_obj->host_buffer_descriptor_count = cpu_to_le32(1);
bde = (struct sli4_bde *)rd_obj->host_buffer_descriptor;
/* Setup to transfer xfer_size bytes to device */
bde->bde_type_buflen =
cpu_to_le32((SLI4_BDE_TYPE_VAL(64)) |
(desired_read_len & SLI4_BDE_LEN_MASK));
if (dma) {
bde->u.data.low = cpu_to_le32(lower_32_bits(dma->phys));
bde->u.data.high = cpu_to_le32(upper_32_bits(dma->phys));
} else {
bde->u.data.low = 0;
bde->u.data.high = 0;
}
return 0;
}
int
sli_cmd_dmtf_exec_clp_cmd(struct sli4 *sli4, void *buf, struct efc_dma *cmd,
struct efc_dma *resp)
{
struct sli4_rqst_dmtf_exec_clp_cmd *clp_cmd = NULL;
clp_cmd = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(dmtf_exec_clp_cmd), NULL);
if (!clp_cmd)
return -EIO;
sli_cmd_fill_hdr(&clp_cmd->hdr, DMTF_EXEC_CLP_CMD, SLI4_SUBSYSTEM_DMTF,
CMD_V0, SLI4_RQST_PYLD_LEN(dmtf_exec_clp_cmd));
clp_cmd->cmd_buf_length = cpu_to_le32(cmd->size);
clp_cmd->cmd_buf_addr_low = cpu_to_le32(lower_32_bits(cmd->phys));
clp_cmd->cmd_buf_addr_high = cpu_to_le32(upper_32_bits(cmd->phys));
clp_cmd->resp_buf_length = cpu_to_le32(resp->size);
clp_cmd->resp_buf_addr_low = cpu_to_le32(lower_32_bits(resp->phys));
clp_cmd->resp_buf_addr_high = cpu_to_le32(upper_32_bits(resp->phys));
return 0;
}
int
sli_cmd_common_set_dump_location(struct sli4 *sli4, void *buf, bool query,
bool is_buffer_list,
struct efc_dma *buffer, u8 fdb)
{
struct sli4_rqst_cmn_set_dump_location *set_dump_loc = NULL;
u32 buffer_length_flag = 0;
set_dump_loc = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_set_dump_location), NULL);
if (!set_dump_loc)
return -EIO;
sli_cmd_fill_hdr(&set_dump_loc->hdr, SLI4_CMN_SET_DUMP_LOCATION,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_set_dump_location));
if (is_buffer_list)
buffer_length_flag |= SLI4_CMN_SET_DUMP_BLP;
if (query)
buffer_length_flag |= SLI4_CMN_SET_DUMP_QRY;
if (fdb)
buffer_length_flag |= SLI4_CMN_SET_DUMP_FDB;
if (buffer) {
set_dump_loc->buf_addr_low =
cpu_to_le32(lower_32_bits(buffer->phys));
set_dump_loc->buf_addr_high =
cpu_to_le32(upper_32_bits(buffer->phys));
buffer_length_flag |=
buffer->len & SLI4_CMN_SET_DUMP_BUFFER_LEN;
} else {
set_dump_loc->buf_addr_low = 0;
set_dump_loc->buf_addr_high = 0;
set_dump_loc->buffer_length_dword = 0;
}
set_dump_loc->buffer_length_dword = cpu_to_le32(buffer_length_flag);
return 0;
}
int
sli_cmd_common_set_features(struct sli4 *sli4, void *buf, u32 feature,
u32 param_len, void *parameter)
{
struct sli4_rqst_cmn_set_features *cmd = NULL;
cmd = sli_config_cmd_init(sli4, buf,
SLI4_RQST_CMDSZ(cmn_set_features), NULL);
if (!cmd)
return -EIO;
sli_cmd_fill_hdr(&cmd->hdr, SLI4_CMN_SET_FEATURES,
SLI4_SUBSYSTEM_COMMON, CMD_V0,
SLI4_RQST_PYLD_LEN(cmn_set_features));
cmd->feature = cpu_to_le32(feature);
cmd->param_len = cpu_to_le32(param_len);
memcpy(cmd->params, parameter, param_len);
return 0;
}
int
sli_cqe_mq(struct sli4 *sli4, void *buf)
{
struct sli4_mcqe *mcqe = buf;
u32 dwflags = le32_to_cpu(mcqe->dw3_flags);
/*
* Firmware can split mbx completions into two MCQEs: first with only
* the "consumed" bit set and a second with the "complete" bit set.
* Thus, ignore MCQE unless "complete" is set.
*/
if (!(dwflags & SLI4_MCQE_COMPLETED))
return SLI4_MCQE_STATUS_NOT_COMPLETED;
if (le16_to_cpu(mcqe->completion_status)) {
efc_log_info(sli4, "status(st=%#x ext=%#x con=%d cmp=%d ae=%d val=%d)\n",
le16_to_cpu(mcqe->completion_status),
le16_to_cpu(mcqe->extended_status),
(dwflags & SLI4_MCQE_CONSUMED),
(dwflags & SLI4_MCQE_COMPLETED),
(dwflags & SLI4_MCQE_AE),
(dwflags & SLI4_MCQE_VALID));
}
return le16_to_cpu(mcqe->completion_status);
}
int
sli_cqe_async(struct sli4 *sli4, void *buf)
{
struct sli4_acqe *acqe = buf;
int rc = -EIO;
if (!buf) {
efc_log_err(sli4, "bad parameter sli4=%p buf=%p\n", sli4, buf);
return -EIO;
}
switch (acqe->event_code) {
case SLI4_ACQE_EVENT_CODE_LINK_STATE:
efc_log_info(sli4, "Unsupported by FC link, evt code:%#x\n",
acqe->event_code);
break;
case SLI4_ACQE_EVENT_CODE_GRP_5:
efc_log_info(sli4, "ACQE GRP5\n");
break;
case SLI4_ACQE_EVENT_CODE_SLI_PORT_EVENT:
efc_log_info(sli4, "ACQE SLI Port, type=0x%x, data1,2=0x%08x,0x%08x\n",
acqe->event_type,
le32_to_cpu(acqe->event_data[0]),
le32_to_cpu(acqe->event_data[1]));
break;
case SLI4_ACQE_EVENT_CODE_FC_LINK_EVENT:
rc = sli_fc_process_link_attention(sli4, buf);
break;
default:
efc_log_info(sli4, "ACQE unknown=%#x\n", acqe->event_code);
}
return rc;
}
bool
sli_fw_ready(struct sli4 *sli4)
{
u32 val;
/* Determine if the chip FW is in a ready state */
val = sli_reg_read_status(sli4);
return (val & SLI4_PORT_STATUS_RDY) ? 1 : 0;
}
static bool
sli_wait_for_fw_ready(struct sli4 *sli4, u32 timeout_ms)
{
unsigned long end;
end = jiffies + msecs_to_jiffies(timeout_ms);
do {
if (sli_fw_ready(sli4))
return true;
usleep_range(1000, 2000);
} while (time_before(jiffies, end));
return false;
}
static bool
sli_sliport_reset(struct sli4 *sli4)
{
bool rc;
u32 val;
val = SLI4_PORT_CTRL_IP;
/* Initialize port, endian */
writel(val, (sli4->reg[0] + SLI4_PORT_CTRL_REG));
rc = sli_wait_for_fw_ready(sli4, SLI4_FW_READY_TIMEOUT_MSEC);
if (!rc)
efc_log_crit(sli4, "port failed to become ready after initialization\n");
return rc;
}
static bool
sli_fw_init(struct sli4 *sli4)
{
/*
* Is firmware ready for operation?
*/
if (!sli_wait_for_fw_ready(sli4, SLI4_FW_READY_TIMEOUT_MSEC)) {
efc_log_crit(sli4, "FW status is NOT ready\n");
return false;
}
/*
* Reset port to a known state
*/
return sli_sliport_reset(sli4);
}
static int
sli_request_features(struct sli4 *sli4, u32 *features, bool query)
{
struct sli4_cmd_request_features *req_features = sli4->bmbx.virt;
if (sli_cmd_request_features(sli4, sli4->bmbx.virt, *features, query)) {
efc_log_err(sli4, "bad REQUEST_FEATURES write\n");
return -EIO;
}
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail\n");
return -EIO;
}
if (le16_to_cpu(req_features->hdr.status)) {
efc_log_err(sli4, "REQUEST_FEATURES bad status %#x\n",
le16_to_cpu(req_features->hdr.status));
return -EIO;
}
*features = le32_to_cpu(req_features->resp);
return 0;
}
void
sli_calc_max_qentries(struct sli4 *sli4)
{
enum sli4_qtype q;
u32 qentries;
for (q = SLI4_QTYPE_EQ; q < SLI4_QTYPE_MAX; q++) {
sli4->qinfo.max_qentries[q] =
sli_convert_mask_to_count(sli4->qinfo.count_method[q],
sli4->qinfo.count_mask[q]);
}
/* single, contiguous DMA allocations will be called for each queue
* of size (max_qentries * queue entry size); since these can be large,
* check against the OS max DMA allocation size
*/
for (q = SLI4_QTYPE_EQ; q < SLI4_QTYPE_MAX; q++) {
qentries = sli4->qinfo.max_qentries[q];
efc_log_info(sli4, "[%s]: max_qentries from %d to %d\n",
SLI4_QNAME[q],
sli4->qinfo.max_qentries[q], qentries);
sli4->qinfo.max_qentries[q] = qentries;
}
}
static int
sli_get_read_config(struct sli4 *sli4)
{
struct sli4_rsp_read_config *conf = sli4->bmbx.virt;
u32 i, total;
u32 *base;
if (sli_cmd_read_config(sli4, sli4->bmbx.virt)) {
efc_log_err(sli4, "bad READ_CONFIG write\n");
return -EIO;
}
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox fail (READ_CONFIG)\n");
return -EIO;
}
if (le16_to_cpu(conf->hdr.status)) {
efc_log_err(sli4, "READ_CONFIG bad status %#x\n",
le16_to_cpu(conf->hdr.status));
return -EIO;
}
sli4->params.has_extents =
le32_to_cpu(conf->ext_dword) & SLI4_READ_CFG_RESP_RESOURCE_EXT;
if (sli4->params.has_extents) {
efc_log_err(sli4, "extents not supported\n");
return -EIO;
}
base = sli4->ext[0].base;
if (!base) {
int size = SLI4_RSRC_MAX * sizeof(u32);
base = kzalloc(size, GFP_KERNEL);
if (!base)
return -EIO;
}
for (i = 0; i < SLI4_RSRC_MAX; i++) {
sli4->ext[i].number = 1;
sli4->ext[i].n_alloc = 0;
sli4->ext[i].base = &base[i];
}
sli4->ext[SLI4_RSRC_VFI].base[0] = le16_to_cpu(conf->vfi_base);
sli4->ext[SLI4_RSRC_VFI].size = le16_to_cpu(conf->vfi_count);
sli4->ext[SLI4_RSRC_VPI].base[0] = le16_to_cpu(conf->vpi_base);
sli4->ext[SLI4_RSRC_VPI].size = le16_to_cpu(conf->vpi_count);
sli4->ext[SLI4_RSRC_RPI].base[0] = le16_to_cpu(conf->rpi_base);
sli4->ext[SLI4_RSRC_RPI].size = le16_to_cpu(conf->rpi_count);
sli4->ext[SLI4_RSRC_XRI].base[0] = le16_to_cpu(conf->xri_base);
sli4->ext[SLI4_RSRC_XRI].size = le16_to_cpu(conf->xri_count);
sli4->ext[SLI4_RSRC_FCFI].base[0] = 0;
sli4->ext[SLI4_RSRC_FCFI].size = le16_to_cpu(conf->fcfi_count);
for (i = 0; i < SLI4_RSRC_MAX; i++) {
total = sli4->ext[i].number * sli4->ext[i].size;
sli4->ext[i].use_map = bitmap_zalloc(total, GFP_KERNEL);
if (!sli4->ext[i].use_map) {
efc_log_err(sli4, "bitmap memory allocation failed %d\n",
i);
return -EIO;
}
sli4->ext[i].map_size = total;
}
sli4->topology = (le32_to_cpu(conf->topology_dword) &
SLI4_READ_CFG_RESP_TOPOLOGY) >> 24;
switch (sli4->topology) {
case SLI4_READ_CFG_TOPO_FC:
efc_log_info(sli4, "FC (unknown)\n");
break;
case SLI4_READ_CFG_TOPO_NON_FC_AL:
efc_log_info(sli4, "FC (direct attach)\n");
break;
case SLI4_READ_CFG_TOPO_FC_AL:
efc_log_info(sli4, "FC (arbitrated loop)\n");
break;
default:
efc_log_info(sli4, "bad topology %#x\n", sli4->topology);
}
sli4->e_d_tov = le16_to_cpu(conf->e_d_tov);
sli4->r_a_tov = le16_to_cpu(conf->r_a_tov);
sli4->link_module_type = le16_to_cpu(conf->lmt);
sli4->qinfo.max_qcount[SLI4_QTYPE_EQ] = le16_to_cpu(conf->eq_count);
sli4->qinfo.max_qcount[SLI4_QTYPE_CQ] = le16_to_cpu(conf->cq_count);
sli4->qinfo.max_qcount[SLI4_QTYPE_WQ] = le16_to_cpu(conf->wq_count);
sli4->qinfo.max_qcount[SLI4_QTYPE_RQ] = le16_to_cpu(conf->rq_count);
/*
* READ_CONFIG doesn't give the max number of MQ. Applications
* will typically want 1, but we may need another at some future
* date. Dummy up a "max" MQ count here.
*/
sli4->qinfo.max_qcount[SLI4_QTYPE_MQ] = SLI4_USER_MQ_COUNT;
return 0;
}
static int
sli_get_sli4_parameters(struct sli4 *sli4)
{
struct sli4_rsp_cmn_get_sli4_params *parms;
u32 dw_loopback;
u32 dw_eq_pg_cnt;
u32 dw_cq_pg_cnt;
u32 dw_mq_pg_cnt;
u32 dw_wq_pg_cnt;
u32 dw_rq_pg_cnt;
u32 dw_sgl_pg_cnt;
if (sli_cmd_common_get_sli4_parameters(sli4, sli4->bmbx.virt))
return -EIO;
parms = (struct sli4_rsp_cmn_get_sli4_params *)
(((u8 *)sli4->bmbx.virt) +
offsetof(struct sli4_cmd_sli_config, payload.embed));
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail\n");
return -EIO;
}
if (parms->hdr.status) {
efc_log_err(sli4, "COMMON_GET_SLI4_PARAMETERS bad status %#x",
parms->hdr.status);
efc_log_err(sli4, "additional status %#x\n",
parms->hdr.additional_status);
return -EIO;
}
dw_loopback = le32_to_cpu(parms->dw16_loopback_scope);
dw_eq_pg_cnt = le32_to_cpu(parms->dw6_eq_page_cnt);
dw_cq_pg_cnt = le32_to_cpu(parms->dw8_cq_page_cnt);
dw_mq_pg_cnt = le32_to_cpu(parms->dw10_mq_page_cnt);
dw_wq_pg_cnt = le32_to_cpu(parms->dw12_wq_page_cnt);
dw_rq_pg_cnt = le32_to_cpu(parms->dw14_rq_page_cnt);
sli4->params.auto_reg = (dw_loopback & SLI4_PARAM_AREG);
sli4->params.auto_xfer_rdy = (dw_loopback & SLI4_PARAM_AGXF);
sli4->params.hdr_template_req = (dw_loopback & SLI4_PARAM_HDRR);
sli4->params.t10_dif_inline_capable = (dw_loopback & SLI4_PARAM_TIMM);
sli4->params.t10_dif_separate_capable = (dw_loopback & SLI4_PARAM_TSMM);
sli4->params.mq_create_version = GET_Q_CREATE_VERSION(dw_mq_pg_cnt);
sli4->params.cq_create_version = GET_Q_CREATE_VERSION(dw_cq_pg_cnt);
sli4->rq_min_buf_size = le16_to_cpu(parms->min_rq_buffer_size);
sli4->rq_max_buf_size = le32_to_cpu(parms->max_rq_buffer_size);
sli4->qinfo.qpage_count[SLI4_QTYPE_EQ] =
(dw_eq_pg_cnt & SLI4_PARAM_EQ_PAGE_CNT_MASK);
sli4->qinfo.qpage_count[SLI4_QTYPE_CQ] =
(dw_cq_pg_cnt & SLI4_PARAM_CQ_PAGE_CNT_MASK);
sli4->qinfo.qpage_count[SLI4_QTYPE_MQ] =
(dw_mq_pg_cnt & SLI4_PARAM_MQ_PAGE_CNT_MASK);
sli4->qinfo.qpage_count[SLI4_QTYPE_WQ] =
(dw_wq_pg_cnt & SLI4_PARAM_WQ_PAGE_CNT_MASK);
sli4->qinfo.qpage_count[SLI4_QTYPE_RQ] =
(dw_rq_pg_cnt & SLI4_PARAM_RQ_PAGE_CNT_MASK);
/* save count methods and masks for each queue type */
sli4->qinfo.count_mask[SLI4_QTYPE_EQ] =
le16_to_cpu(parms->eqe_count_mask);
sli4->qinfo.count_method[SLI4_QTYPE_EQ] =
GET_Q_CNT_METHOD(dw_eq_pg_cnt);
sli4->qinfo.count_mask[SLI4_QTYPE_CQ] =
le16_to_cpu(parms->cqe_count_mask);
sli4->qinfo.count_method[SLI4_QTYPE_CQ] =
GET_Q_CNT_METHOD(dw_cq_pg_cnt);
sli4->qinfo.count_mask[SLI4_QTYPE_MQ] =
le16_to_cpu(parms->mqe_count_mask);
sli4->qinfo.count_method[SLI4_QTYPE_MQ] =
GET_Q_CNT_METHOD(dw_mq_pg_cnt);
sli4->qinfo.count_mask[SLI4_QTYPE_WQ] =
le16_to_cpu(parms->wqe_count_mask);
sli4->qinfo.count_method[SLI4_QTYPE_WQ] =
GET_Q_CNT_METHOD(dw_wq_pg_cnt);
sli4->qinfo.count_mask[SLI4_QTYPE_RQ] =
le16_to_cpu(parms->rqe_count_mask);
sli4->qinfo.count_method[SLI4_QTYPE_RQ] =
GET_Q_CNT_METHOD(dw_rq_pg_cnt);
/* now calculate max queue entries */
sli_calc_max_qentries(sli4);
dw_sgl_pg_cnt = le32_to_cpu(parms->dw18_sgl_page_cnt);
/* max # of pages */
sli4->max_sgl_pages = (dw_sgl_pg_cnt & SLI4_PARAM_SGL_PAGE_CNT_MASK);
/* bit map of available sizes */
sli4->sgl_page_sizes = (dw_sgl_pg_cnt &
SLI4_PARAM_SGL_PAGE_SZS_MASK) >> 8;
/* ignore HLM here. Use value from REQUEST_FEATURES */
sli4->sge_supported_length = le32_to_cpu(parms->sge_supported_length);
sli4->params.sgl_pre_reg_required = (dw_loopback & SLI4_PARAM_SGLR);
/* default to using pre-registered SGL's */
sli4->params.sgl_pre_registered = true;
sli4->params.perf_hint = dw_loopback & SLI4_PARAM_PHON;
sli4->params.perf_wq_id_association = (dw_loopback & SLI4_PARAM_PHWQ);
sli4->rq_batch = (le16_to_cpu(parms->dw15w1_rq_db_window) &
SLI4_PARAM_RQ_DB_WINDOW_MASK) >> 12;
/* Use the highest available WQE size. */
if (((dw_wq_pg_cnt & SLI4_PARAM_WQE_SZS_MASK) >> 8) &
SLI4_128BYTE_WQE_SUPPORT)
sli4->wqe_size = SLI4_WQE_EXT_BYTES;
else
sli4->wqe_size = SLI4_WQE_BYTES;
return 0;
}
static int
sli_get_ctrl_attributes(struct sli4 *sli4)
{
struct sli4_rsp_cmn_get_cntl_attributes *attr;
struct sli4_rsp_cmn_get_cntl_addl_attributes *add_attr;
struct efc_dma data;
u32 psize;
/*
* Issue COMMON_GET_CNTL_ATTRIBUTES to get port_number. Temporarily
* uses VPD DMA buffer as the response won't fit in the embedded
* buffer.
*/
memset(sli4->vpd_data.virt, 0, sli4->vpd_data.size);
if (sli_cmd_common_get_cntl_attributes(sli4, sli4->bmbx.virt,
&sli4->vpd_data)) {
efc_log_err(sli4, "bad COMMON_GET_CNTL_ATTRIBUTES write\n");
return -EIO;
}
attr = sli4->vpd_data.virt;
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail\n");
return -EIO;
}
if (attr->hdr.status) {
efc_log_err(sli4, "COMMON_GET_CNTL_ATTRIBUTES bad status %#x",
attr->hdr.status);
efc_log_err(sli4, "additional status %#x\n",
attr->hdr.additional_status);
return -EIO;
}
sli4->port_number = attr->port_num_type_flags & SLI4_CNTL_ATTR_PORTNUM;
memcpy(sli4->bios_version_string, attr->bios_version_str,
sizeof(sli4->bios_version_string));
/* get additional attributes */
psize = sizeof(struct sli4_rsp_cmn_get_cntl_addl_attributes);
data.size = psize;
data.virt = dma_alloc_coherent(&sli4->pci->dev, data.size,
&data.phys, GFP_KERNEL);
if (!data.virt) {
memset(&data, 0, sizeof(struct efc_dma));
efc_log_err(sli4, "Failed to allocate memory for GET_CNTL_ADDL_ATTR\n");
return -EIO;
}
if (sli_cmd_common_get_cntl_addl_attributes(sli4, sli4->bmbx.virt,
&data)) {
efc_log_err(sli4, "bad GET_CNTL_ADDL_ATTR write\n");
dma_free_coherent(&sli4->pci->dev, data.size,
data.virt, data.phys);
return -EIO;
}
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "mailbox fail (GET_CNTL_ADDL_ATTR)\n");
dma_free_coherent(&sli4->pci->dev, data.size,
data.virt, data.phys);
return -EIO;
}
add_attr = data.virt;
if (add_attr->hdr.status) {
efc_log_err(sli4, "GET_CNTL_ADDL_ATTR bad status %#x\n",
add_attr->hdr.status);
dma_free_coherent(&sli4->pci->dev, data.size,
data.virt, data.phys);
return -EIO;
}
memcpy(sli4->ipl_name, add_attr->ipl_file_name, sizeof(sli4->ipl_name));
efc_log_info(sli4, "IPL:%s\n", (char *)sli4->ipl_name);
dma_free_coherent(&sli4->pci->dev, data.size, data.virt,
data.phys);
memset(&data, 0, sizeof(struct efc_dma));
return 0;
}
static int
sli_get_fw_rev(struct sli4 *sli4)
{
struct sli4_cmd_read_rev *read_rev = sli4->bmbx.virt;
if (sli_cmd_read_rev(sli4, sli4->bmbx.virt, &sli4->vpd_data))
return -EIO;
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail (READ_REV)\n");
return -EIO;
}
if (le16_to_cpu(read_rev->hdr.status)) {
efc_log_err(sli4, "READ_REV bad status %#x\n",
le16_to_cpu(read_rev->hdr.status));
return -EIO;
}
sli4->fw_rev[0] = le32_to_cpu(read_rev->first_fw_id);
memcpy(sli4->fw_name[0], read_rev->first_fw_name,
sizeof(sli4->fw_name[0]));
sli4->fw_rev[1] = le32_to_cpu(read_rev->second_fw_id);
memcpy(sli4->fw_name[1], read_rev->second_fw_name,
sizeof(sli4->fw_name[1]));
sli4->hw_rev[0] = le32_to_cpu(read_rev->first_hw_rev);
sli4->hw_rev[1] = le32_to_cpu(read_rev->second_hw_rev);
sli4->hw_rev[2] = le32_to_cpu(read_rev->third_hw_rev);
efc_log_info(sli4, "FW1:%s (%08x) / FW2:%s (%08x)\n",
read_rev->first_fw_name, le32_to_cpu(read_rev->first_fw_id),
read_rev->second_fw_name, le32_to_cpu(read_rev->second_fw_id));
efc_log_info(sli4, "HW1: %08x / HW2: %08x\n",
le32_to_cpu(read_rev->first_hw_rev),
le32_to_cpu(read_rev->second_hw_rev));
/* Check that all VPD data was returned */
if (le32_to_cpu(read_rev->returned_vpd_length) !=
le32_to_cpu(read_rev->actual_vpd_length)) {
efc_log_info(sli4, "VPD length: avail=%d return=%d actual=%d\n",
le32_to_cpu(read_rev->available_length_dword) &
SLI4_READ_REV_AVAILABLE_LENGTH,
le32_to_cpu(read_rev->returned_vpd_length),
le32_to_cpu(read_rev->actual_vpd_length));
}
sli4->vpd_length = le32_to_cpu(read_rev->returned_vpd_length);
return 0;
}
static int
sli_get_config(struct sli4 *sli4)
{
struct sli4_rsp_cmn_get_port_name *port_name;
struct sli4_cmd_read_nvparms *read_nvparms;
/*
* Read the device configuration
*/
if (sli_get_read_config(sli4))
return -EIO;
if (sli_get_sli4_parameters(sli4))
return -EIO;
if (sli_get_ctrl_attributes(sli4))
return -EIO;
if (sli_cmd_common_get_port_name(sli4, sli4->bmbx.virt))
return -EIO;
port_name = (struct sli4_rsp_cmn_get_port_name *)
(((u8 *)sli4->bmbx.virt) +
offsetof(struct sli4_cmd_sli_config, payload.embed));
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox fail (GET_PORT_NAME)\n");
return -EIO;
}
sli4->port_name[0] = port_name->port_name[sli4->port_number];
sli4->port_name[1] = '\0';
if (sli_get_fw_rev(sli4))
return -EIO;
if (sli_cmd_read_nvparms(sli4, sli4->bmbx.virt)) {
efc_log_err(sli4, "bad READ_NVPARMS write\n");
return -EIO;
}
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox fail (READ_NVPARMS)\n");
return -EIO;
}
read_nvparms = sli4->bmbx.virt;
if (le16_to_cpu(read_nvparms->hdr.status)) {
efc_log_err(sli4, "READ_NVPARMS bad status %#x\n",
le16_to_cpu(read_nvparms->hdr.status));
return -EIO;
}
memcpy(sli4->wwpn, read_nvparms->wwpn, sizeof(sli4->wwpn));
memcpy(sli4->wwnn, read_nvparms->wwnn, sizeof(sli4->wwnn));
efc_log_info(sli4, "WWPN %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x\n",
sli4->wwpn[0], sli4->wwpn[1], sli4->wwpn[2], sli4->wwpn[3],
sli4->wwpn[4], sli4->wwpn[5], sli4->wwpn[6], sli4->wwpn[7]);
efc_log_info(sli4, "WWNN %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x\n",
sli4->wwnn[0], sli4->wwnn[1], sli4->wwnn[2], sli4->wwnn[3],
sli4->wwnn[4], sli4->wwnn[5], sli4->wwnn[6], sli4->wwnn[7]);
return 0;
}
int
sli_setup(struct sli4 *sli4, void *os, struct pci_dev *pdev,
void __iomem *reg[])
{
u32 intf = U32_MAX;
u32 pci_class_rev = 0;
u32 rev_id = 0;
u32 family = 0;
u32 asic_id = 0;
u32 i;
struct sli4_asic_entry_t *asic;
memset(sli4, 0, sizeof(struct sli4));
sli4->os = os;
sli4->pci = pdev;
for (i = 0; i < 6; i++)
sli4->reg[i] = reg[i];
/*
* Read the SLI_INTF register to discover the register layout
* and other capability information
*/
if (pci_read_config_dword(pdev, SLI4_INTF_REG, &intf))
return -EIO;
if ((intf & SLI4_INTF_VALID_MASK) != (u32)SLI4_INTF_VALID_VALUE) {
efc_log_err(sli4, "SLI_INTF is not valid\n");
return -EIO;
}
/* driver only support SLI-4 */
if ((intf & SLI4_INTF_REV_MASK) != SLI4_INTF_REV_S4) {
efc_log_err(sli4, "Unsupported SLI revision (intf=%#x)\n", intf);
return -EIO;
}
sli4->sli_family = intf & SLI4_INTF_FAMILY_MASK;
sli4->if_type = intf & SLI4_INTF_IF_TYPE_MASK;
efc_log_info(sli4, "status=%#x error1=%#x error2=%#x\n",
sli_reg_read_status(sli4),
sli_reg_read_err1(sli4),
sli_reg_read_err2(sli4));
/*
* set the ASIC type and revision
*/
if (pci_read_config_dword(pdev, PCI_CLASS_REVISION, &pci_class_rev))
return -EIO;
rev_id = pci_class_rev & 0xff;
family = sli4->sli_family;
if (family == SLI4_FAMILY_CHECK_ASIC_TYPE) {
if (!pci_read_config_dword(pdev, SLI4_ASIC_ID_REG, &asic_id))
family = asic_id & SLI4_ASIC_GEN_MASK;
}
for (i = 0, asic = sli4_asic_table; i < ARRAY_SIZE(sli4_asic_table);
i++, asic++) {
if (rev_id == asic->rev_id && family == asic->family) {
sli4->asic_type = family;
sli4->asic_rev = rev_id;
break;
}
}
/* Fail if no matching asic type/rev was found */
if (!sli4->asic_type) {
efc_log_err(sli4, "no matching asic family/rev found: %02x/%02x\n",
family, rev_id);
return -EIO;
}
/*
* The bootstrap mailbox is equivalent to a MQ with a single 256 byte
* entry, a CQ with a single 16 byte entry, and no event queue.
* Alignment must be 16 bytes as the low order address bits in the
* address register are also control / status.
*/
sli4->bmbx.size = SLI4_BMBX_SIZE + sizeof(struct sli4_mcqe);
sli4->bmbx.virt = dma_alloc_coherent(&pdev->dev, sli4->bmbx.size,
&sli4->bmbx.phys, GFP_KERNEL);
if (!sli4->bmbx.virt) {
memset(&sli4->bmbx, 0, sizeof(struct efc_dma));
efc_log_err(sli4, "bootstrap mailbox allocation failed\n");
return -EIO;
}
if (sli4->bmbx.phys & SLI4_BMBX_MASK_LO) {
efc_log_err(sli4, "bad alignment for bootstrap mailbox\n");
return -EIO;
}
efc_log_info(sli4, "bmbx v=%p p=0x%x %08x s=%zd\n", sli4->bmbx.virt,
upper_32_bits(sli4->bmbx.phys),
lower_32_bits(sli4->bmbx.phys), sli4->bmbx.size);
/* 4096 is arbitrary. What should this value actually be? */
sli4->vpd_data.size = 4096;
sli4->vpd_data.virt = dma_alloc_coherent(&pdev->dev,
sli4->vpd_data.size,
&sli4->vpd_data.phys,
GFP_KERNEL);
if (!sli4->vpd_data.virt) {
memset(&sli4->vpd_data, 0, sizeof(struct efc_dma));
/* Note that failure isn't fatal in this specific case */
efc_log_info(sli4, "VPD buffer allocation failed\n");
}
if (!sli_fw_init(sli4)) {
efc_log_err(sli4, "FW initialization failed\n");
return -EIO;
}
/*
* Set one of fcpi(initiator), fcpt(target), fcpc(combined) to true
* in addition to any other desired features
*/
sli4->features = (SLI4_REQFEAT_IAAB | SLI4_REQFEAT_NPIV |
SLI4_REQFEAT_DIF | SLI4_REQFEAT_VF |
SLI4_REQFEAT_FCPC | SLI4_REQFEAT_IAAR |
SLI4_REQFEAT_HLM | SLI4_REQFEAT_PERFH |
SLI4_REQFEAT_RXSEQ | SLI4_REQFEAT_RXRI |
SLI4_REQFEAT_MRQP);
/* use performance hints if available */
if (sli4->params.perf_hint)
sli4->features |= SLI4_REQFEAT_PERFH;
if (sli_request_features(sli4, &sli4->features, true))
return -EIO;
if (sli_get_config(sli4))
return -EIO;
return 0;
}
int
sli_init(struct sli4 *sli4)
{
if (sli4->params.has_extents) {
efc_log_info(sli4, "extend allocation not supported\n");
return -EIO;
}
sli4->features &= (~SLI4_REQFEAT_HLM);
sli4->features &= (~SLI4_REQFEAT_RXSEQ);
sli4->features &= (~SLI4_REQFEAT_RXRI);
if (sli_request_features(sli4, &sli4->features, false))
return -EIO;
return 0;
}
int
sli_reset(struct sli4 *sli4)
{
u32 i;
if (!sli_fw_init(sli4)) {
efc_log_crit(sli4, "FW initialization failed\n");
return -EIO;
}
kfree(sli4->ext[0].base);
sli4->ext[0].base = NULL;
for (i = 0; i < SLI4_RSRC_MAX; i++) {
bitmap_free(sli4->ext[i].use_map);
sli4->ext[i].use_map = NULL;
sli4->ext[i].base = NULL;
}
return sli_get_config(sli4);
}
int
sli_fw_reset(struct sli4 *sli4)
{
/*
* Firmware must be ready before issuing the reset.
*/
if (!sli_wait_for_fw_ready(sli4, SLI4_FW_READY_TIMEOUT_MSEC)) {
efc_log_crit(sli4, "FW status is NOT ready\n");
return -EIO;
}
/* Lancer uses PHYDEV_CONTROL */
writel(SLI4_PHYDEV_CTRL_FRST, (sli4->reg[0] + SLI4_PHYDEV_CTRL_REG));
/* wait for the FW to become ready after the reset */
if (!sli_wait_for_fw_ready(sli4, SLI4_FW_READY_TIMEOUT_MSEC)) {
efc_log_crit(sli4, "Failed to be ready after firmware reset\n");
return -EIO;
}
return 0;
}
void
sli_teardown(struct sli4 *sli4)
{
u32 i;
kfree(sli4->ext[0].base);
sli4->ext[0].base = NULL;
for (i = 0; i < SLI4_RSRC_MAX; i++) {
sli4->ext[i].base = NULL;
bitmap_free(sli4->ext[i].use_map);
sli4->ext[i].use_map = NULL;
}
if (!sli_sliport_reset(sli4))
efc_log_err(sli4, "FW deinitialization failed\n");
dma_free_coherent(&sli4->pci->dev, sli4->vpd_data.size,
sli4->vpd_data.virt, sli4->vpd_data.phys);
memset(&sli4->vpd_data, 0, sizeof(struct efc_dma));
dma_free_coherent(&sli4->pci->dev, sli4->bmbx.size,
sli4->bmbx.virt, sli4->bmbx.phys);
memset(&sli4->bmbx, 0, sizeof(struct efc_dma));
}
int
sli_callback(struct sli4 *sli4, enum sli4_callback which,
void *func, void *arg)
{
if (!func) {
efc_log_err(sli4, "bad parameter sli4=%p which=%#x func=%p\n",
sli4, which, func);
return -EIO;
}
switch (which) {
case SLI4_CB_LINK:
sli4->link = func;
sli4->link_arg = arg;
break;
default:
efc_log_info(sli4, "unknown callback %#x\n", which);
return -EIO;
}
return 0;
}
int
sli_eq_modify_delay(struct sli4 *sli4, struct sli4_queue *eq,
u32 num_eq, u32 shift, u32 delay_mult)
{
sli_cmd_common_modify_eq_delay(sli4, sli4->bmbx.virt, eq, num_eq,
shift, delay_mult);
if (sli_bmbx_command(sli4)) {
efc_log_crit(sli4, "bootstrap mailbox write fail (MODIFY EQ DELAY)\n");
return -EIO;
}
if (sli_res_sli_config(sli4, sli4->bmbx.virt)) {
efc_log_err(sli4, "bad status MODIFY EQ DELAY\n");
return -EIO;
}
return 0;
}
int
sli_resource_alloc(struct sli4 *sli4, enum sli4_resource rtype,
u32 *rid, u32 *index)
{
int rc = 0;
u32 size;
u32 ext_idx;
u32 item_idx;
u32 position;
*rid = U32_MAX;
*index = U32_MAX;
switch (rtype) {
case SLI4_RSRC_VFI:
case SLI4_RSRC_VPI:
case SLI4_RSRC_RPI:
case SLI4_RSRC_XRI:
position =
find_first_zero_bit(sli4->ext[rtype].use_map,
sli4->ext[rtype].map_size);
if (position >= sli4->ext[rtype].map_size) {
efc_log_err(sli4, "out of resource %d (alloc=%d)\n",
rtype, sli4->ext[rtype].n_alloc);
rc = -EIO;
break;
}
set_bit(position, sli4->ext[rtype].use_map);
*index = position;
size = sli4->ext[rtype].size;
ext_idx = *index / size;
item_idx = *index % size;
*rid = sli4->ext[rtype].base[ext_idx] + item_idx;
sli4->ext[rtype].n_alloc++;
break;
default:
rc = -EIO;
}
return rc;
}
int
sli_resource_free(struct sli4 *sli4, enum sli4_resource rtype, u32 rid)
{
int rc = -EIO;
u32 x;
u32 size, *base;
switch (rtype) {
case SLI4_RSRC_VFI:
case SLI4_RSRC_VPI:
case SLI4_RSRC_RPI:
case SLI4_RSRC_XRI:
/*
* Figure out which extent contains the resource ID. I.e. find
* the extent such that
* extent->base <= resource ID < extent->base + extent->size
*/
base = sli4->ext[rtype].base;
size = sli4->ext[rtype].size;
/*
* In the case of FW reset, this may be cleared
* but the force_free path will still attempt to
* free the resource. Prevent a NULL pointer access.
*/
if (!base)
break;
for (x = 0; x < sli4->ext[rtype].number; x++) {
if ((rid < base[x] || (rid >= (base[x] + size))))
continue;
rid -= base[x];
clear_bit((x * size) + rid, sli4->ext[rtype].use_map);
rc = 0;
break;
}
break;
default:
break;
}
return rc;
}
int
sli_resource_reset(struct sli4 *sli4, enum sli4_resource rtype)
{
int rc = -EIO;
u32 i;
switch (rtype) {
case SLI4_RSRC_VFI:
case SLI4_RSRC_VPI:
case SLI4_RSRC_RPI:
case SLI4_RSRC_XRI:
for (i = 0; i < sli4->ext[rtype].map_size; i++)
clear_bit(i, sli4->ext[rtype].use_map);
rc = 0;
break;
default:
break;
}
return rc;
}
int sli_raise_ue(struct sli4 *sli4, u8 dump)
{
u32 val = 0;
if (dump == SLI4_FUNC_DESC_DUMP) {
val = SLI4_PORT_CTRL_FDD | SLI4_PORT_CTRL_IP;
writel(val, (sli4->reg[0] + SLI4_PORT_CTRL_REG));
} else {
val = SLI4_PHYDEV_CTRL_FRST;
if (dump == SLI4_CHIP_LEVEL_DUMP)
val |= SLI4_PHYDEV_CTRL_DD;
writel(val, (sli4->reg[0] + SLI4_PHYDEV_CTRL_REG));
}
return 0;
}
int sli_dump_is_ready(struct sli4 *sli4)
{
int rc = SLI4_DUMP_READY_STATUS_NOT_READY;
u32 port_val;
u32 bmbx_val;
/*
* Ensure that the port is ready AND the mailbox is
* ready before signaling that the dump is ready to go.
*/
port_val = sli_reg_read_status(sli4);
bmbx_val = readl(sli4->reg[0] + SLI4_BMBX_REG);
if ((bmbx_val & SLI4_BMBX_RDY) &&
(port_val & SLI4_PORT_STATUS_RDY)) {
if (port_val & SLI4_PORT_STATUS_DIP)
rc = SLI4_DUMP_READY_STATUS_DD_PRESENT;
else if (port_val & SLI4_PORT_STATUS_FDP)
rc = SLI4_DUMP_READY_STATUS_FDB_PRESENT;
}
return rc;
}
bool sli_reset_required(struct sli4 *sli4)
{
u32 val;
val = sli_reg_read_status(sli4);
return (val & SLI4_PORT_STATUS_RN);
}
int
sli_cmd_post_sgl_pages(struct sli4 *sli4, void *buf, u16 xri,
u32 xri_count, struct efc_dma *page0[],
struct efc_dma *page1[], struct efc_dma *dma)
{
struct sli4_rqst_post_sgl_pages *post = NULL;
u32 i;
__le32 req_len;
post = sli_config_cmd_init(sli4, buf,
SLI4_CFG_PYLD_LENGTH(post_sgl_pages), dma);
if (!post)
return -EIO;
/* payload size calculation */
/* 4 = xri_start + xri_count */
/* xri_count = # of XRI's registered */
/* sizeof(uint64_t) = physical address size */
/* 2 = # of physical addresses per page set */
req_len = cpu_to_le32(4 + (xri_count * (sizeof(uint64_t) * 2)));
sli_cmd_fill_hdr(&post->hdr, SLI4_OPC_POST_SGL_PAGES, SLI4_SUBSYSTEM_FC,
CMD_V0, req_len);
post->xri_start = cpu_to_le16(xri);
post->xri_count = cpu_to_le16(xri_count);
for (i = 0; i < xri_count; i++) {
post->page_set[i].page0_low =
cpu_to_le32(lower_32_bits(page0[i]->phys));
post->page_set[i].page0_high =
cpu_to_le32(upper_32_bits(page0[i]->phys));
}
if (page1) {
for (i = 0; i < xri_count; i++) {
post->page_set[i].page1_low =
cpu_to_le32(lower_32_bits(page1[i]->phys));
post->page_set[i].page1_high =
cpu_to_le32(upper_32_bits(page1[i]->phys));
}
}
return 0;
}
int
sli_cmd_post_hdr_templates(struct sli4 *sli4, void *buf, struct efc_dma *dma,
u16 rpi, struct efc_dma *payload_dma)
{
struct sli4_rqst_post_hdr_templates *req = NULL;
uintptr_t phys = 0;
u32 i = 0;
u32 page_count, payload_size;
page_count = sli_page_count(dma->size, SLI_PAGE_SIZE);
payload_size = ((sizeof(struct sli4_rqst_post_hdr_templates) +
(page_count * SZ_DMAADDR)) - sizeof(struct sli4_rqst_hdr));
if (page_count > 16) {
/*
* We can't fit more than 16 descriptors into an embedded mbox
* command, it has to be non-embedded
*/
payload_dma->size = payload_size;
payload_dma->virt = dma_alloc_coherent(&sli4->pci->dev,
payload_dma->size,
&payload_dma->phys, GFP_KERNEL);
if (!payload_dma->virt) {
memset(payload_dma, 0, sizeof(struct efc_dma));
efc_log_err(sli4, "mbox payload memory allocation fail\n");
return -EIO;
}
req = sli_config_cmd_init(sli4, buf, payload_size, payload_dma);
} else {
req = sli_config_cmd_init(sli4, buf, payload_size, NULL);
}
if (!req)
return -EIO;
if (rpi == U16_MAX)
rpi = sli4->ext[SLI4_RSRC_RPI].base[0];
sli_cmd_fill_hdr(&req->hdr, SLI4_OPC_POST_HDR_TEMPLATES,
SLI4_SUBSYSTEM_FC, CMD_V0,
SLI4_RQST_PYLD_LEN(post_hdr_templates));
req->rpi_offset = cpu_to_le16(rpi);
req->page_count = cpu_to_le16(page_count);
phys = dma->phys;
for (i = 0; i < page_count; i++) {
req->page_descriptor[i].low = cpu_to_le32(lower_32_bits(phys));
req->page_descriptor[i].high = cpu_to_le32(upper_32_bits(phys));
phys += SLI_PAGE_SIZE;
}
return 0;
}
u32
sli_fc_get_rpi_requirements(struct sli4 *sli4, u32 n_rpi)
{
u32 bytes = 0;
/* Check if header templates needed */
if (sli4->params.hdr_template_req)
/* round up to a page */
bytes = round_up(n_rpi * SLI4_HDR_TEMPLATE_SIZE, SLI_PAGE_SIZE);
return bytes;
}
const char *
sli_fc_get_status_string(u32 status)
{
static struct {
u32 code;
const char *label;
} lookup[] = {
{SLI4_FC_WCQE_STATUS_SUCCESS, "SUCCESS"},
{SLI4_FC_WCQE_STATUS_FCP_RSP_FAILURE, "FCP_RSP_FAILURE"},
{SLI4_FC_WCQE_STATUS_REMOTE_STOP, "REMOTE_STOP"},
{SLI4_FC_WCQE_STATUS_LOCAL_REJECT, "LOCAL_REJECT"},
{SLI4_FC_WCQE_STATUS_NPORT_RJT, "NPORT_RJT"},
{SLI4_FC_WCQE_STATUS_FABRIC_RJT, "FABRIC_RJT"},
{SLI4_FC_WCQE_STATUS_NPORT_BSY, "NPORT_BSY"},
{SLI4_FC_WCQE_STATUS_FABRIC_BSY, "FABRIC_BSY"},
{SLI4_FC_WCQE_STATUS_LS_RJT, "LS_RJT"},
{SLI4_FC_WCQE_STATUS_CMD_REJECT, "CMD_REJECT"},
{SLI4_FC_WCQE_STATUS_FCP_TGT_LENCHECK, "FCP_TGT_LENCHECK"},
{SLI4_FC_WCQE_STATUS_RQ_BUF_LEN_EXCEEDED, "BUF_LEN_EXCEEDED"},
{SLI4_FC_WCQE_STATUS_RQ_INSUFF_BUF_NEEDED,
"RQ_INSUFF_BUF_NEEDED"},
{SLI4_FC_WCQE_STATUS_RQ_INSUFF_FRM_DISC, "RQ_INSUFF_FRM_DESC"},
{SLI4_FC_WCQE_STATUS_RQ_DMA_FAILURE, "RQ_DMA_FAILURE"},
{SLI4_FC_WCQE_STATUS_FCP_RSP_TRUNCATE, "FCP_RSP_TRUNCATE"},
{SLI4_FC_WCQE_STATUS_DI_ERROR, "DI_ERROR"},
{SLI4_FC_WCQE_STATUS_BA_RJT, "BA_RJT"},
{SLI4_FC_WCQE_STATUS_RQ_INSUFF_XRI_NEEDED,
"RQ_INSUFF_XRI_NEEDED"},
{SLI4_FC_WCQE_STATUS_RQ_INSUFF_XRI_DISC, "INSUFF_XRI_DISC"},
{SLI4_FC_WCQE_STATUS_RX_ERROR_DETECT, "RX_ERROR_DETECT"},
{SLI4_FC_WCQE_STATUS_RX_ABORT_REQUEST, "RX_ABORT_REQUEST"},
};
u32 i;
for (i = 0; i < ARRAY_SIZE(lookup); i++) {
if (status == lookup[i].code)
return lookup[i].label;
}
return "unknown";
}