Commit 826da486 authored by Shai Malin's avatar Shai Malin Committed by David S. Miller

qed: Add NVMeTCP Offload IO Level FW Initializations

This patch introduces the NVMeTCP FW initializations which is used
to initialize the IO level configuration into a per IO HW
resource ("task") as part of the IO path flow.

This includes:
- Write IO FW initialization
- Read IO FW initialization.
- IC-Req and IC-Resp FW exchange.
- FW Cleanup flow (Flush IO).
Acked-by: default avatarIgor Russkikh <irusskikh@marvell.com>
Signed-off-by: default avatarPrabhakar Kushwaha <pkushwaha@marvell.com>
Signed-off-by: default avatarOmkar Kulkarni <okulkarni@marvell.com>
Signed-off-by: default avatarShai Malin <smalin@marvell.com>
Signed-off-by: default avatarMichal Kalderon <mkalderon@marvell.com>
Signed-off-by: default avatarAriel Elior <aelior@marvell.com>
Reviewed-by: default avatarHannes Reinecke <hare@suse.de>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent ab47bdfd
......@@ -28,7 +28,10 @@ qed-$(CONFIG_QED_ISCSI) += qed_iscsi.o
qed-$(CONFIG_QED_LL2) += qed_ll2.o
qed-$(CONFIG_QED_OOO) += qed_ooo.o
qed-$(CONFIG_QED_NVMETCP) += qed_nvmetcp.o
qed-$(CONFIG_QED_NVMETCP) += \
qed_nvmetcp.o \
qed_nvmetcp_fw_funcs.o \
qed_nvmetcp_ip_services.o
qed-$(CONFIG_QED_RDMA) += \
qed_iwarp.o \
......
......@@ -27,6 +27,7 @@
#include "qed_mcp.h"
#include "qed_sp.h"
#include "qed_reg_addr.h"
#include "qed_nvmetcp_fw_funcs.h"
static int qed_nvmetcp_async_event(struct qed_hwfn *p_hwfn, u8 fw_event_code,
u16 echo, union event_ring_data *data,
......@@ -809,7 +810,11 @@ static const struct qed_nvmetcp_ops qed_nvmetcp_ops_pass = {
.remove_src_tcp_port_filter = &qed_llh_remove_src_tcp_port_filter,
.add_dst_tcp_port_filter = &qed_llh_add_dst_tcp_port_filter,
.remove_dst_tcp_port_filter = &qed_llh_remove_dst_tcp_port_filter,
.clear_all_filters = &qed_llh_clear_all_filters
.clear_all_filters = &qed_llh_clear_all_filters,
.init_read_io = &init_nvmetcp_host_read_task,
.init_write_io = &init_nvmetcp_host_write_task,
.init_icreq_exchange = &init_nvmetcp_init_conn_req_task,
.init_task_cleanup = &init_cleanup_task_nvmetcp
};
const struct qed_nvmetcp_ops *qed_get_nvmetcp_ops(void)
......
// SPDX-License-Identifier: (GPL-2.0-only OR BSD-3-Clause)
/* Copyright 2021 Marvell. All rights reserved. */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/types.h>
#include <asm/byteorder.h>
#include <linux/qed/common_hsi.h>
#include <linux/qed/storage_common.h>
#include <linux/qed/nvmetcp_common.h>
#include <linux/qed/qed_nvmetcp_if.h>
#include "qed_nvmetcp_fw_funcs.h"
#define NVMETCP_NUM_SGES_IN_CACHE 0x4
bool nvmetcp_is_slow_sgl(u16 num_sges, bool small_mid_sge)
{
return (num_sges > SCSI_NUM_SGES_SLOW_SGL_THR && small_mid_sge);
}
void init_scsi_sgl_context(struct scsi_sgl_params *ctx_sgl_params,
struct scsi_cached_sges *ctx_data_desc,
struct storage_sgl_task_params *sgl_params)
{
u8 num_sges_to_init = (u8)(sgl_params->num_sges > NVMETCP_NUM_SGES_IN_CACHE ?
NVMETCP_NUM_SGES_IN_CACHE : sgl_params->num_sges);
u8 sge_index;
/* sgl params */
ctx_sgl_params->sgl_addr.lo = cpu_to_le32(sgl_params->sgl_phys_addr.lo);
ctx_sgl_params->sgl_addr.hi = cpu_to_le32(sgl_params->sgl_phys_addr.hi);
ctx_sgl_params->sgl_total_length = cpu_to_le32(sgl_params->total_buffer_size);
ctx_sgl_params->sgl_num_sges = cpu_to_le16(sgl_params->num_sges);
for (sge_index = 0; sge_index < num_sges_to_init; sge_index++) {
ctx_data_desc->sge[sge_index].sge_addr.lo =
cpu_to_le32(sgl_params->sgl[sge_index].sge_addr.lo);
ctx_data_desc->sge[sge_index].sge_addr.hi =
cpu_to_le32(sgl_params->sgl[sge_index].sge_addr.hi);
ctx_data_desc->sge[sge_index].sge_len =
cpu_to_le32(sgl_params->sgl[sge_index].sge_len);
}
}
static inline u32 calc_rw_task_size(struct nvmetcp_task_params *task_params,
enum nvmetcp_task_type task_type)
{
u32 io_size;
if (task_type == NVMETCP_TASK_TYPE_HOST_WRITE)
io_size = task_params->tx_io_size;
else
io_size = task_params->rx_io_size;
if (unlikely(!io_size))
return 0;
return io_size;
}
static inline void init_sqe(struct nvmetcp_task_params *task_params,
struct storage_sgl_task_params *sgl_task_params,
enum nvmetcp_task_type task_type)
{
if (!task_params->sqe)
return;
memset(task_params->sqe, 0, sizeof(*task_params->sqe));
task_params->sqe->task_id = cpu_to_le16(task_params->itid);
switch (task_type) {
case NVMETCP_TASK_TYPE_HOST_WRITE: {
u32 buf_size = 0;
u32 num_sges = 0;
SET_FIELD(task_params->sqe->contlen_cdbsize,
NVMETCP_WQE_CDB_SIZE_OR_NVMETCP_CMD, 1);
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_WQE_TYPE,
NVMETCP_WQE_TYPE_NORMAL);
if (task_params->tx_io_size) {
if (task_params->send_write_incapsule)
buf_size = calc_rw_task_size(task_params, task_type);
if (nvmetcp_is_slow_sgl(sgl_task_params->num_sges,
sgl_task_params->small_mid_sge))
num_sges = NVMETCP_WQE_NUM_SGES_SLOWIO;
else
num_sges = min((u16)sgl_task_params->num_sges,
(u16)SCSI_NUM_SGES_SLOW_SGL_THR);
}
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_NUM_SGES, num_sges);
SET_FIELD(task_params->sqe->contlen_cdbsize, NVMETCP_WQE_CONT_LEN, buf_size);
} break;
case NVMETCP_TASK_TYPE_HOST_READ: {
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_WQE_TYPE,
NVMETCP_WQE_TYPE_NORMAL);
SET_FIELD(task_params->sqe->contlen_cdbsize,
NVMETCP_WQE_CDB_SIZE_OR_NVMETCP_CMD, 1);
} break;
case NVMETCP_TASK_TYPE_INIT_CONN_REQUEST: {
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_WQE_TYPE,
NVMETCP_WQE_TYPE_MIDDLE_PATH);
if (task_params->tx_io_size) {
SET_FIELD(task_params->sqe->contlen_cdbsize, NVMETCP_WQE_CONT_LEN,
task_params->tx_io_size);
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_NUM_SGES,
min((u16)sgl_task_params->num_sges,
(u16)SCSI_NUM_SGES_SLOW_SGL_THR));
}
} break;
case NVMETCP_TASK_TYPE_CLEANUP:
SET_FIELD(task_params->sqe->flags, NVMETCP_WQE_WQE_TYPE,
NVMETCP_WQE_TYPE_TASK_CLEANUP);
default:
break;
}
}
/* The following function initializes of NVMeTCP task params */
static inline void
init_nvmetcp_task_params(struct e5_nvmetcp_task_context *context,
struct nvmetcp_task_params *task_params,
enum nvmetcp_task_type task_type)
{
context->ystorm_st_context.state.cccid = task_params->host_cccid;
SET_FIELD(context->ustorm_st_context.error_flags, USTORM_NVMETCP_TASK_ST_CTX_NVME_TCP, 1);
context->ustorm_st_context.nvme_tcp_opaque_lo = cpu_to_le32(task_params->opq.lo);
context->ustorm_st_context.nvme_tcp_opaque_hi = cpu_to_le32(task_params->opq.hi);
}
/* The following function initializes default values to all tasks */
static inline void
init_default_nvmetcp_task(struct nvmetcp_task_params *task_params,
void *pdu_header, void *nvme_cmd,
enum nvmetcp_task_type task_type)
{
struct e5_nvmetcp_task_context *context = task_params->context;
const u8 val_byte = context->mstorm_ag_context.cdu_validation;
u8 dw_index;
memset(context, 0, sizeof(*context));
init_nvmetcp_task_params(context, task_params,
(enum nvmetcp_task_type)task_type);
/* Swapping requirements used below, will be removed in future FW versions */
if (task_type == NVMETCP_TASK_TYPE_HOST_WRITE ||
task_type == NVMETCP_TASK_TYPE_HOST_READ) {
for (dw_index = 0;
dw_index < QED_NVMETCP_CMN_HDR_SIZE / sizeof(u32);
dw_index++)
context->ystorm_st_context.pdu_hdr.task_hdr.reg[dw_index] =
cpu_to_le32(__swab32(((u32 *)pdu_header)[dw_index]));
for (dw_index = QED_NVMETCP_CMN_HDR_SIZE / sizeof(u32);
dw_index < QED_NVMETCP_CMD_HDR_SIZE / sizeof(u32);
dw_index++)
context->ystorm_st_context.pdu_hdr.task_hdr.reg[dw_index] =
cpu_to_le32(__swab32(((u32 *)nvme_cmd)[dw_index - 2]));
} else {
for (dw_index = 0;
dw_index < QED_NVMETCP_NON_IO_HDR_SIZE / sizeof(u32);
dw_index++)
context->ystorm_st_context.pdu_hdr.task_hdr.reg[dw_index] =
cpu_to_le32(__swab32(((u32 *)pdu_header)[dw_index]));
}
/* M-Storm Context: */
context->mstorm_ag_context.cdu_validation = val_byte;
context->mstorm_st_context.task_type = (u8)(task_type);
context->mstorm_ag_context.task_cid = cpu_to_le16(task_params->conn_icid);
/* Ustorm Context: */
SET_FIELD(context->ustorm_ag_context.flags1, E5_USTORM_NVMETCP_TASK_AG_CTX_R2T2RECV, 1);
context->ustorm_st_context.task_type = (u8)(task_type);
context->ustorm_st_context.cq_rss_number = task_params->cq_rss_number;
context->ustorm_ag_context.icid = cpu_to_le16(task_params->conn_icid);
}
/* The following function initializes the U-Storm Task Contexts */
static inline void
init_ustorm_task_contexts(struct ustorm_nvmetcp_task_st_ctx *ustorm_st_context,
struct e5_ustorm_nvmetcp_task_ag_ctx *ustorm_ag_context,
u32 remaining_recv_len,
u32 expected_data_transfer_len, u8 num_sges,
bool tx_dif_conn_err_en)
{
/* Remaining data to be received in bytes. Used in validations*/
ustorm_st_context->rem_rcv_len = cpu_to_le32(remaining_recv_len);
ustorm_ag_context->exp_data_acked = cpu_to_le32(expected_data_transfer_len);
ustorm_st_context->exp_data_transfer_len = cpu_to_le32(expected_data_transfer_len);
SET_FIELD(ustorm_st_context->reg1_map, REG1_NUM_SGES, num_sges);
SET_FIELD(ustorm_ag_context->flags2, E5_USTORM_NVMETCP_TASK_AG_CTX_DIF_ERROR_CF_EN,
tx_dif_conn_err_en ? 1 : 0);
}
/* The following function initializes Local Completion Contexts: */
static inline void
set_local_completion_context(struct e5_nvmetcp_task_context *context)
{
SET_FIELD(context->ystorm_st_context.state.flags,
YSTORM_NVMETCP_TASK_STATE_LOCAL_COMP, 1);
SET_FIELD(context->ustorm_st_context.flags,
USTORM_NVMETCP_TASK_ST_CTX_LOCAL_COMP, 1);
}
/* Common Fastpath task init function: */
static inline void
init_rw_nvmetcp_task(struct nvmetcp_task_params *task_params,
enum nvmetcp_task_type task_type,
void *pdu_header, void *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params)
{
struct e5_nvmetcp_task_context *context = task_params->context;
u32 task_size = calc_rw_task_size(task_params, task_type);
bool slow_io = false;
u8 num_sges = 0;
init_default_nvmetcp_task(task_params, pdu_header, nvme_cmd, task_type);
/* Tx/Rx: */
if (task_params->tx_io_size) {
/* if data to transmit: */
init_scsi_sgl_context(&context->ystorm_st_context.state.sgl_params,
&context->ystorm_st_context.state.data_desc,
sgl_task_params);
slow_io = nvmetcp_is_slow_sgl(sgl_task_params->num_sges,
sgl_task_params->small_mid_sge);
num_sges =
(u8)(!slow_io ? min((u32)sgl_task_params->num_sges,
(u32)SCSI_NUM_SGES_SLOW_SGL_THR) :
NVMETCP_WQE_NUM_SGES_SLOWIO);
if (slow_io) {
SET_FIELD(context->ystorm_st_context.state.flags,
YSTORM_NVMETCP_TASK_STATE_SLOW_IO, 1);
}
} else if (task_params->rx_io_size) {
/* if data to receive: */
init_scsi_sgl_context(&context->mstorm_st_context.sgl_params,
&context->mstorm_st_context.data_desc,
sgl_task_params);
num_sges =
(u8)(!nvmetcp_is_slow_sgl(sgl_task_params->num_sges,
sgl_task_params->small_mid_sge) ?
min((u32)sgl_task_params->num_sges,
(u32)SCSI_NUM_SGES_SLOW_SGL_THR) :
NVMETCP_WQE_NUM_SGES_SLOWIO);
context->mstorm_st_context.rem_task_size = cpu_to_le32(task_size);
}
/* Ustorm context: */
init_ustorm_task_contexts(&context->ustorm_st_context,
&context->ustorm_ag_context,
/* Remaining Receive length is the Task Size */
task_size,
/* The size of the transmitted task */
task_size,
/* num_sges */
num_sges,
false);
/* Set exp_data_acked */
if (task_type == NVMETCP_TASK_TYPE_HOST_WRITE) {
if (task_params->send_write_incapsule)
context->ustorm_ag_context.exp_data_acked = task_size;
else
context->ustorm_ag_context.exp_data_acked = 0;
} else if (task_type == NVMETCP_TASK_TYPE_HOST_READ) {
context->ustorm_ag_context.exp_data_acked = 0;
}
context->ustorm_ag_context.exp_cont_len = 0;
init_sqe(task_params, sgl_task_params, task_type);
}
static void
init_common_initiator_read_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params)
{
init_rw_nvmetcp_task(task_params, NVMETCP_TASK_TYPE_HOST_READ,
cmd_pdu_header, nvme_cmd, sgl_task_params);
}
void init_nvmetcp_host_read_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params)
{
init_common_initiator_read_task(task_params, (void *)cmd_pdu_header,
(void *)nvme_cmd, sgl_task_params);
}
static void
init_common_initiator_write_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params)
{
init_rw_nvmetcp_task(task_params, NVMETCP_TASK_TYPE_HOST_WRITE,
cmd_pdu_header, nvme_cmd, sgl_task_params);
}
void init_nvmetcp_host_write_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params)
{
init_common_initiator_write_task(task_params, (void *)cmd_pdu_header,
(void *)nvme_cmd, sgl_task_params);
}
static void
init_common_login_request_task(struct nvmetcp_task_params *task_params,
void *login_req_pdu_header,
struct storage_sgl_task_params *tx_sgl_task_params,
struct storage_sgl_task_params *rx_sgl_task_params)
{
struct e5_nvmetcp_task_context *context = task_params->context;
init_default_nvmetcp_task(task_params, (void *)login_req_pdu_header, NULL,
NVMETCP_TASK_TYPE_INIT_CONN_REQUEST);
/* Ustorm Context: */
init_ustorm_task_contexts(&context->ustorm_st_context,
&context->ustorm_ag_context,
/* Remaining Receive length is the Task Size */
task_params->rx_io_size ?
rx_sgl_task_params->total_buffer_size : 0,
/* The size of the transmitted task */
task_params->tx_io_size ?
tx_sgl_task_params->total_buffer_size : 0,
0, /* num_sges */
0); /* tx_dif_conn_err_en */
/* SGL context: */
if (task_params->tx_io_size)
init_scsi_sgl_context(&context->ystorm_st_context.state.sgl_params,
&context->ystorm_st_context.state.data_desc,
tx_sgl_task_params);
if (task_params->rx_io_size)
init_scsi_sgl_context(&context->mstorm_st_context.sgl_params,
&context->mstorm_st_context.data_desc,
rx_sgl_task_params);
context->mstorm_st_context.rem_task_size =
cpu_to_le32(task_params->rx_io_size ?
rx_sgl_task_params->total_buffer_size : 0);
init_sqe(task_params, tx_sgl_task_params, NVMETCP_TASK_TYPE_INIT_CONN_REQUEST);
}
/* The following function initializes Login task in Host mode: */
void init_nvmetcp_init_conn_req_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_icreq_pdu *init_conn_req_pdu_hdr,
struct storage_sgl_task_params *tx_sgl_task_params,
struct storage_sgl_task_params *rx_sgl_task_params)
{
init_common_login_request_task(task_params, init_conn_req_pdu_hdr,
tx_sgl_task_params, rx_sgl_task_params);
}
void init_cleanup_task_nvmetcp(struct nvmetcp_task_params *task_params)
{
init_sqe(task_params, NULL, NVMETCP_TASK_TYPE_CLEANUP);
}
/* SPDX-License-Identifier: (GPL-2.0-only OR BSD-3-Clause) */
/* Copyright 2021 Marvell. All rights reserved. */
#ifndef _QED_NVMETCP_FW_FUNCS_H
#define _QED_NVMETCP_FW_FUNCS_H
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/pci.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/types.h>
#include <asm/byteorder.h>
#include <linux/qed/common_hsi.h>
#include <linux/qed/storage_common.h>
#include <linux/qed/nvmetcp_common.h>
#include <linux/qed/qed_nvmetcp_if.h>
#if IS_ENABLED(CONFIG_QED_NVMETCP)
void init_nvmetcp_host_read_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params);
void init_nvmetcp_host_write_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params);
void init_nvmetcp_init_conn_req_task(struct nvmetcp_task_params *task_params,
struct nvme_tcp_icreq_pdu *init_conn_req_pdu_hdr,
struct storage_sgl_task_params *tx_sgl_task_params,
struct storage_sgl_task_params *rx_sgl_task_params);
void init_cleanup_task_nvmetcp(struct nvmetcp_task_params *task_params);
#else /* IS_ENABLED(CONFIG_QED_NVMETCP) */
#endif /* IS_ENABLED(CONFIG_QED_NVMETCP) */
#endif /* _QED_NVMETCP_FW_FUNCS_H */
......@@ -5,6 +5,7 @@
#define __NVMETCP_COMMON__
#include "tcp_common.h"
#include <linux/nvme-tcp.h>
#define NVMETCP_SLOW_PATH_LAYER_CODE (6)
#define NVMETCP_WQE_NUM_SGES_SLOWIO (0xf)
......
......@@ -9,6 +9,9 @@
#include <linux/qed/nvmetcp_common.h>
#define QED_NVMETCP_MAX_IO_SIZE 0x800000
#define QED_NVMETCP_CMN_HDR_SIZE (sizeof(struct nvme_tcp_hdr))
#define QED_NVMETCP_CMD_HDR_SIZE (sizeof(struct nvme_tcp_cmd_pdu))
#define QED_NVMETCP_NON_IO_HDR_SIZE ((QED_NVMETCP_CMN_HDR_SIZE + 16))
typedef int (*nvmetcp_event_cb_t) (void *context,
u8 fw_event_code, void *fw_handle);
......@@ -213,6 +216,23 @@ struct qed_nvmetcp_ops {
void (*remove_dst_tcp_port_filter)(struct qed_dev *cdev, u16 dest_port);
void (*clear_all_filters)(struct qed_dev *cdev);
void (*init_read_io)(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params);
void (*init_write_io)(struct nvmetcp_task_params *task_params,
struct nvme_tcp_cmd_pdu *cmd_pdu_header,
struct nvme_command *nvme_cmd,
struct storage_sgl_task_params *sgl_task_params);
void (*init_icreq_exchange)(struct nvmetcp_task_params *task_params,
struct nvme_tcp_icreq_pdu *init_conn_req_pdu_hdr,
struct storage_sgl_task_params *tx_sgl_task_params,
struct storage_sgl_task_params *rx_sgl_task_params);
void (*init_task_cleanup)(struct nvmetcp_task_params *task_params);
};
const struct qed_nvmetcp_ops *qed_get_nvmetcp_ops(void);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment