forked from Minki/linux
9db5fbe1a4
In a recent change to the SPI subsystem [1], a new `delay` struct was added
to replace the `delay_usecs`. This change replaces the current
`delay_usecs` with `delay` for this driver.
The `spi_transfer_delay_exec()` function [in the SPI framework] makes sure
that both `delay_usecs` & `delay` are used (in this order to preserve
backwards compatibility).
[1] commit bebcfd272d
("spi: introduce `delay` field for
`spi_transfer` + spi_transfer_delay_exec()")
Signed-off-by: Sergiu Cuciurean <sergiu.cuciurean@analog.com>
Link: https://lore.kernel.org/r/20200227130336.27042-1-sergiu.cuciurean@analog.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
534 lines
12 KiB
C
534 lines
12 KiB
C
// SPDX-License-Identifier: GPL-2.0-only
|
|
/*
|
|
* Copyright (c) 2011-2016 Synaptics Incorporated
|
|
* Copyright (c) 2011 Unixphere
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
#include <linux/module.h>
|
|
#include <linux/rmi.h>
|
|
#include <linux/slab.h>
|
|
#include <linux/spi/spi.h>
|
|
#include <linux/of.h>
|
|
#include "rmi_driver.h"
|
|
|
|
#define RMI_SPI_DEFAULT_XFER_BUF_SIZE 64
|
|
|
|
#define RMI_PAGE_SELECT_REGISTER 0x00FF
|
|
#define RMI_SPI_PAGE(addr) (((addr) >> 8) & 0x80)
|
|
#define RMI_SPI_XFER_SIZE_LIMIT 255
|
|
|
|
#define BUFFER_SIZE_INCREMENT 32
|
|
|
|
enum rmi_spi_op {
|
|
RMI_SPI_WRITE = 0,
|
|
RMI_SPI_READ,
|
|
RMI_SPI_V2_READ_UNIFIED,
|
|
RMI_SPI_V2_READ_SPLIT,
|
|
RMI_SPI_V2_WRITE,
|
|
};
|
|
|
|
struct rmi_spi_cmd {
|
|
enum rmi_spi_op op;
|
|
u16 addr;
|
|
};
|
|
|
|
struct rmi_spi_xport {
|
|
struct rmi_transport_dev xport;
|
|
struct spi_device *spi;
|
|
|
|
struct mutex page_mutex;
|
|
int page;
|
|
|
|
u8 *rx_buf;
|
|
u8 *tx_buf;
|
|
int xfer_buf_size;
|
|
|
|
struct spi_transfer *rx_xfers;
|
|
struct spi_transfer *tx_xfers;
|
|
int rx_xfer_count;
|
|
int tx_xfer_count;
|
|
};
|
|
|
|
static int rmi_spi_manage_pools(struct rmi_spi_xport *rmi_spi, int len)
|
|
{
|
|
struct spi_device *spi = rmi_spi->spi;
|
|
int buf_size = rmi_spi->xfer_buf_size
|
|
? rmi_spi->xfer_buf_size : RMI_SPI_DEFAULT_XFER_BUF_SIZE;
|
|
struct spi_transfer *xfer_buf;
|
|
void *buf;
|
|
void *tmp;
|
|
|
|
while (buf_size < len)
|
|
buf_size *= 2;
|
|
|
|
if (buf_size > RMI_SPI_XFER_SIZE_LIMIT)
|
|
buf_size = RMI_SPI_XFER_SIZE_LIMIT;
|
|
|
|
tmp = rmi_spi->rx_buf;
|
|
buf = devm_kcalloc(&spi->dev, buf_size, 2,
|
|
GFP_KERNEL | GFP_DMA);
|
|
if (!buf)
|
|
return -ENOMEM;
|
|
|
|
rmi_spi->rx_buf = buf;
|
|
rmi_spi->tx_buf = &rmi_spi->rx_buf[buf_size];
|
|
rmi_spi->xfer_buf_size = buf_size;
|
|
|
|
if (tmp)
|
|
devm_kfree(&spi->dev, tmp);
|
|
|
|
if (rmi_spi->xport.pdata.spi_data.read_delay_us)
|
|
rmi_spi->rx_xfer_count = buf_size;
|
|
else
|
|
rmi_spi->rx_xfer_count = 1;
|
|
|
|
if (rmi_spi->xport.pdata.spi_data.write_delay_us)
|
|
rmi_spi->tx_xfer_count = buf_size;
|
|
else
|
|
rmi_spi->tx_xfer_count = 1;
|
|
|
|
/*
|
|
* Allocate a pool of spi_transfer buffers for devices which need
|
|
* per byte delays.
|
|
*/
|
|
tmp = rmi_spi->rx_xfers;
|
|
xfer_buf = devm_kcalloc(&spi->dev,
|
|
rmi_spi->rx_xfer_count + rmi_spi->tx_xfer_count,
|
|
sizeof(struct spi_transfer),
|
|
GFP_KERNEL);
|
|
if (!xfer_buf)
|
|
return -ENOMEM;
|
|
|
|
rmi_spi->rx_xfers = xfer_buf;
|
|
rmi_spi->tx_xfers = &xfer_buf[rmi_spi->rx_xfer_count];
|
|
|
|
if (tmp)
|
|
devm_kfree(&spi->dev, tmp);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int rmi_spi_xfer(struct rmi_spi_xport *rmi_spi,
|
|
const struct rmi_spi_cmd *cmd, const u8 *tx_buf,
|
|
int tx_len, u8 *rx_buf, int rx_len)
|
|
{
|
|
struct spi_device *spi = rmi_spi->spi;
|
|
struct rmi_device_platform_data_spi *spi_data =
|
|
&rmi_spi->xport.pdata.spi_data;
|
|
struct spi_message msg;
|
|
struct spi_transfer *xfer;
|
|
int ret = 0;
|
|
int len;
|
|
int cmd_len = 0;
|
|
int total_tx_len;
|
|
int i;
|
|
u16 addr = cmd->addr;
|
|
|
|
spi_message_init(&msg);
|
|
|
|
switch (cmd->op) {
|
|
case RMI_SPI_WRITE:
|
|
case RMI_SPI_READ:
|
|
cmd_len += 2;
|
|
break;
|
|
case RMI_SPI_V2_READ_UNIFIED:
|
|
case RMI_SPI_V2_READ_SPLIT:
|
|
case RMI_SPI_V2_WRITE:
|
|
cmd_len += 4;
|
|
break;
|
|
}
|
|
|
|
total_tx_len = cmd_len + tx_len;
|
|
len = max(total_tx_len, rx_len);
|
|
|
|
if (len > RMI_SPI_XFER_SIZE_LIMIT)
|
|
return -EINVAL;
|
|
|
|
if (rmi_spi->xfer_buf_size < len) {
|
|
ret = rmi_spi_manage_pools(rmi_spi, len);
|
|
if (ret < 0)
|
|
return ret;
|
|
}
|
|
|
|
if (addr == 0)
|
|
/*
|
|
* SPI needs an address. Use 0x7FF if we want to keep
|
|
* reading from the last position of the register pointer.
|
|
*/
|
|
addr = 0x7FF;
|
|
|
|
switch (cmd->op) {
|
|
case RMI_SPI_WRITE:
|
|
rmi_spi->tx_buf[0] = (addr >> 8);
|
|
rmi_spi->tx_buf[1] = addr & 0xFF;
|
|
break;
|
|
case RMI_SPI_READ:
|
|
rmi_spi->tx_buf[0] = (addr >> 8) | 0x80;
|
|
rmi_spi->tx_buf[1] = addr & 0xFF;
|
|
break;
|
|
case RMI_SPI_V2_READ_UNIFIED:
|
|
break;
|
|
case RMI_SPI_V2_READ_SPLIT:
|
|
break;
|
|
case RMI_SPI_V2_WRITE:
|
|
rmi_spi->tx_buf[0] = 0x40;
|
|
rmi_spi->tx_buf[1] = (addr >> 8) & 0xFF;
|
|
rmi_spi->tx_buf[2] = addr & 0xFF;
|
|
rmi_spi->tx_buf[3] = tx_len;
|
|
break;
|
|
}
|
|
|
|
if (tx_buf)
|
|
memcpy(&rmi_spi->tx_buf[cmd_len], tx_buf, tx_len);
|
|
|
|
if (rmi_spi->tx_xfer_count > 1) {
|
|
for (i = 0; i < total_tx_len; i++) {
|
|
xfer = &rmi_spi->tx_xfers[i];
|
|
memset(xfer, 0, sizeof(struct spi_transfer));
|
|
xfer->tx_buf = &rmi_spi->tx_buf[i];
|
|
xfer->len = 1;
|
|
xfer->delay.value = spi_data->write_delay_us;
|
|
xfer->delay.unit = SPI_DELAY_UNIT_USECS;
|
|
spi_message_add_tail(xfer, &msg);
|
|
}
|
|
} else {
|
|
xfer = rmi_spi->tx_xfers;
|
|
memset(xfer, 0, sizeof(struct spi_transfer));
|
|
xfer->tx_buf = rmi_spi->tx_buf;
|
|
xfer->len = total_tx_len;
|
|
spi_message_add_tail(xfer, &msg);
|
|
}
|
|
|
|
rmi_dbg(RMI_DEBUG_XPORT, &spi->dev, "%s: cmd: %s tx_buf len: %d tx_buf: %*ph\n",
|
|
__func__, cmd->op == RMI_SPI_WRITE ? "WRITE" : "READ",
|
|
total_tx_len, total_tx_len, rmi_spi->tx_buf);
|
|
|
|
if (rx_buf) {
|
|
if (rmi_spi->rx_xfer_count > 1) {
|
|
for (i = 0; i < rx_len; i++) {
|
|
xfer = &rmi_spi->rx_xfers[i];
|
|
memset(xfer, 0, sizeof(struct spi_transfer));
|
|
xfer->rx_buf = &rmi_spi->rx_buf[i];
|
|
xfer->len = 1;
|
|
xfer->delay.value = spi_data->read_delay_us;
|
|
xfer->delay.unit = SPI_DELAY_UNIT_USECS;
|
|
spi_message_add_tail(xfer, &msg);
|
|
}
|
|
} else {
|
|
xfer = rmi_spi->rx_xfers;
|
|
memset(xfer, 0, sizeof(struct spi_transfer));
|
|
xfer->rx_buf = rmi_spi->rx_buf;
|
|
xfer->len = rx_len;
|
|
spi_message_add_tail(xfer, &msg);
|
|
}
|
|
}
|
|
|
|
ret = spi_sync(spi, &msg);
|
|
if (ret < 0) {
|
|
dev_err(&spi->dev, "spi xfer failed: %d\n", ret);
|
|
return ret;
|
|
}
|
|
|
|
if (rx_buf) {
|
|
memcpy(rx_buf, rmi_spi->rx_buf, rx_len);
|
|
rmi_dbg(RMI_DEBUG_XPORT, &spi->dev, "%s: (%d) %*ph\n",
|
|
__func__, rx_len, rx_len, rx_buf);
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* rmi_set_page - Set RMI page
|
|
* @xport: The pointer to the rmi_transport_dev struct
|
|
* @page: The new page address.
|
|
*
|
|
* RMI devices have 16-bit addressing, but some of the transport
|
|
* implementations (like SMBus) only have 8-bit addressing. So RMI implements
|
|
* a page address at 0xff of every page so we can reliable page addresses
|
|
* every 256 registers.
|
|
*
|
|
* The page_mutex lock must be held when this function is entered.
|
|
*
|
|
* Returns zero on success, non-zero on failure.
|
|
*/
|
|
static int rmi_set_page(struct rmi_spi_xport *rmi_spi, u8 page)
|
|
{
|
|
struct rmi_spi_cmd cmd;
|
|
int ret;
|
|
|
|
cmd.op = RMI_SPI_WRITE;
|
|
cmd.addr = RMI_PAGE_SELECT_REGISTER;
|
|
|
|
ret = rmi_spi_xfer(rmi_spi, &cmd, &page, 1, NULL, 0);
|
|
|
|
if (ret)
|
|
rmi_spi->page = page;
|
|
|
|
return ret;
|
|
}
|
|
|
|
static int rmi_spi_write_block(struct rmi_transport_dev *xport, u16 addr,
|
|
const void *buf, size_t len)
|
|
{
|
|
struct rmi_spi_xport *rmi_spi =
|
|
container_of(xport, struct rmi_spi_xport, xport);
|
|
struct rmi_spi_cmd cmd;
|
|
int ret;
|
|
|
|
mutex_lock(&rmi_spi->page_mutex);
|
|
|
|
if (RMI_SPI_PAGE(addr) != rmi_spi->page) {
|
|
ret = rmi_set_page(rmi_spi, RMI_SPI_PAGE(addr));
|
|
if (ret)
|
|
goto exit;
|
|
}
|
|
|
|
cmd.op = RMI_SPI_WRITE;
|
|
cmd.addr = addr;
|
|
|
|
ret = rmi_spi_xfer(rmi_spi, &cmd, buf, len, NULL, 0);
|
|
|
|
exit:
|
|
mutex_unlock(&rmi_spi->page_mutex);
|
|
return ret;
|
|
}
|
|
|
|
static int rmi_spi_read_block(struct rmi_transport_dev *xport, u16 addr,
|
|
void *buf, size_t len)
|
|
{
|
|
struct rmi_spi_xport *rmi_spi =
|
|
container_of(xport, struct rmi_spi_xport, xport);
|
|
struct rmi_spi_cmd cmd;
|
|
int ret;
|
|
|
|
mutex_lock(&rmi_spi->page_mutex);
|
|
|
|
if (RMI_SPI_PAGE(addr) != rmi_spi->page) {
|
|
ret = rmi_set_page(rmi_spi, RMI_SPI_PAGE(addr));
|
|
if (ret)
|
|
goto exit;
|
|
}
|
|
|
|
cmd.op = RMI_SPI_READ;
|
|
cmd.addr = addr;
|
|
|
|
ret = rmi_spi_xfer(rmi_spi, &cmd, NULL, 0, buf, len);
|
|
|
|
exit:
|
|
mutex_unlock(&rmi_spi->page_mutex);
|
|
return ret;
|
|
}
|
|
|
|
static const struct rmi_transport_ops rmi_spi_ops = {
|
|
.write_block = rmi_spi_write_block,
|
|
.read_block = rmi_spi_read_block,
|
|
};
|
|
|
|
#ifdef CONFIG_OF
|
|
static int rmi_spi_of_probe(struct spi_device *spi,
|
|
struct rmi_device_platform_data *pdata)
|
|
{
|
|
struct device *dev = &spi->dev;
|
|
int retval;
|
|
|
|
retval = rmi_of_property_read_u32(dev,
|
|
&pdata->spi_data.read_delay_us,
|
|
"spi-rx-delay-us", 1);
|
|
if (retval)
|
|
return retval;
|
|
|
|
retval = rmi_of_property_read_u32(dev,
|
|
&pdata->spi_data.write_delay_us,
|
|
"spi-tx-delay-us", 1);
|
|
if (retval)
|
|
return retval;
|
|
|
|
return 0;
|
|
}
|
|
|
|
static const struct of_device_id rmi_spi_of_match[] = {
|
|
{ .compatible = "syna,rmi4-spi" },
|
|
{},
|
|
};
|
|
MODULE_DEVICE_TABLE(of, rmi_spi_of_match);
|
|
#else
|
|
static inline int rmi_spi_of_probe(struct spi_device *spi,
|
|
struct rmi_device_platform_data *pdata)
|
|
{
|
|
return -ENODEV;
|
|
}
|
|
#endif
|
|
|
|
static void rmi_spi_unregister_transport(void *data)
|
|
{
|
|
struct rmi_spi_xport *rmi_spi = data;
|
|
|
|
rmi_unregister_transport_device(&rmi_spi->xport);
|
|
}
|
|
|
|
static int rmi_spi_probe(struct spi_device *spi)
|
|
{
|
|
struct rmi_spi_xport *rmi_spi;
|
|
struct rmi_device_platform_data *pdata;
|
|
struct rmi_device_platform_data *spi_pdata = spi->dev.platform_data;
|
|
int error;
|
|
|
|
if (spi->master->flags & SPI_MASTER_HALF_DUPLEX)
|
|
return -EINVAL;
|
|
|
|
rmi_spi = devm_kzalloc(&spi->dev, sizeof(struct rmi_spi_xport),
|
|
GFP_KERNEL);
|
|
if (!rmi_spi)
|
|
return -ENOMEM;
|
|
|
|
pdata = &rmi_spi->xport.pdata;
|
|
|
|
if (spi->dev.of_node) {
|
|
error = rmi_spi_of_probe(spi, pdata);
|
|
if (error)
|
|
return error;
|
|
} else if (spi_pdata) {
|
|
*pdata = *spi_pdata;
|
|
}
|
|
|
|
if (pdata->spi_data.bits_per_word)
|
|
spi->bits_per_word = pdata->spi_data.bits_per_word;
|
|
|
|
if (pdata->spi_data.mode)
|
|
spi->mode = pdata->spi_data.mode;
|
|
|
|
error = spi_setup(spi);
|
|
if (error < 0) {
|
|
dev_err(&spi->dev, "spi_setup failed!\n");
|
|
return error;
|
|
}
|
|
|
|
pdata->irq = spi->irq;
|
|
|
|
rmi_spi->spi = spi;
|
|
mutex_init(&rmi_spi->page_mutex);
|
|
|
|
rmi_spi->xport.dev = &spi->dev;
|
|
rmi_spi->xport.proto_name = "spi";
|
|
rmi_spi->xport.ops = &rmi_spi_ops;
|
|
|
|
spi_set_drvdata(spi, rmi_spi);
|
|
|
|
error = rmi_spi_manage_pools(rmi_spi, RMI_SPI_DEFAULT_XFER_BUF_SIZE);
|
|
if (error)
|
|
return error;
|
|
|
|
/*
|
|
* Setting the page to zero will (a) make sure the PSR is in a
|
|
* known state, and (b) make sure we can talk to the device.
|
|
*/
|
|
error = rmi_set_page(rmi_spi, 0);
|
|
if (error) {
|
|
dev_err(&spi->dev, "Failed to set page select to 0.\n");
|
|
return error;
|
|
}
|
|
|
|
dev_info(&spi->dev, "registering SPI-connected sensor\n");
|
|
|
|
error = rmi_register_transport_device(&rmi_spi->xport);
|
|
if (error) {
|
|
dev_err(&spi->dev, "failed to register sensor: %d\n", error);
|
|
return error;
|
|
}
|
|
|
|
error = devm_add_action_or_reset(&spi->dev,
|
|
rmi_spi_unregister_transport,
|
|
rmi_spi);
|
|
if (error)
|
|
return error;
|
|
|
|
return 0;
|
|
}
|
|
|
|
#ifdef CONFIG_PM_SLEEP
|
|
static int rmi_spi_suspend(struct device *dev)
|
|
{
|
|
struct spi_device *spi = to_spi_device(dev);
|
|
struct rmi_spi_xport *rmi_spi = spi_get_drvdata(spi);
|
|
int ret;
|
|
|
|
ret = rmi_driver_suspend(rmi_spi->xport.rmi_dev, true);
|
|
if (ret)
|
|
dev_warn(dev, "Failed to resume device: %d\n", ret);
|
|
|
|
return ret;
|
|
}
|
|
|
|
static int rmi_spi_resume(struct device *dev)
|
|
{
|
|
struct spi_device *spi = to_spi_device(dev);
|
|
struct rmi_spi_xport *rmi_spi = spi_get_drvdata(spi);
|
|
int ret;
|
|
|
|
ret = rmi_driver_resume(rmi_spi->xport.rmi_dev, true);
|
|
if (ret)
|
|
dev_warn(dev, "Failed to resume device: %d\n", ret);
|
|
|
|
return ret;
|
|
}
|
|
#endif
|
|
|
|
#ifdef CONFIG_PM
|
|
static int rmi_spi_runtime_suspend(struct device *dev)
|
|
{
|
|
struct spi_device *spi = to_spi_device(dev);
|
|
struct rmi_spi_xport *rmi_spi = spi_get_drvdata(spi);
|
|
int ret;
|
|
|
|
ret = rmi_driver_suspend(rmi_spi->xport.rmi_dev, false);
|
|
if (ret)
|
|
dev_warn(dev, "Failed to resume device: %d\n", ret);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int rmi_spi_runtime_resume(struct device *dev)
|
|
{
|
|
struct spi_device *spi = to_spi_device(dev);
|
|
struct rmi_spi_xport *rmi_spi = spi_get_drvdata(spi);
|
|
int ret;
|
|
|
|
ret = rmi_driver_resume(rmi_spi->xport.rmi_dev, false);
|
|
if (ret)
|
|
dev_warn(dev, "Failed to resume device: %d\n", ret);
|
|
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
static const struct dev_pm_ops rmi_spi_pm = {
|
|
SET_SYSTEM_SLEEP_PM_OPS(rmi_spi_suspend, rmi_spi_resume)
|
|
SET_RUNTIME_PM_OPS(rmi_spi_runtime_suspend, rmi_spi_runtime_resume,
|
|
NULL)
|
|
};
|
|
|
|
static const struct spi_device_id rmi_id[] = {
|
|
{ "rmi4_spi", 0 },
|
|
{ }
|
|
};
|
|
MODULE_DEVICE_TABLE(spi, rmi_id);
|
|
|
|
static struct spi_driver rmi_spi_driver = {
|
|
.driver = {
|
|
.name = "rmi4_spi",
|
|
.pm = &rmi_spi_pm,
|
|
.of_match_table = of_match_ptr(rmi_spi_of_match),
|
|
},
|
|
.id_table = rmi_id,
|
|
.probe = rmi_spi_probe,
|
|
};
|
|
|
|
module_spi_driver(rmi_spi_driver);
|
|
|
|
MODULE_AUTHOR("Christopher Heiny <cheiny@synaptics.com>");
|
|
MODULE_AUTHOR("Andrew Duggan <aduggan@synaptics.com>");
|
|
MODULE_DESCRIPTION("RMI SPI driver");
|
|
MODULE_LICENSE("GPL");
|