spidev_probe() 함수에 printk() 로 내용을 찍어보아도,, 부팅시 나오지가 않네요..

spidev_init() 함수에 찍었을때는 잘 나오더군요..

제공된 소스 그대로인데,, 수정을 가해야 할까요...??

 

/*
 * spidev.c -- simple synchronous userspace interface to SPI devices
 *
 * Copyright (C) 2006 SWAPP
 * Andrea Paterniani <a.paterniani@swapp-eng.it>
 * Copyright (C) 2007 David Brownell (simplification, cleanup)
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/ioctl.h>
#include <linux/fs.h>
#include <linux/device.h>
#include <linux/err.h>
#include <linux/list.h>
#include <linux/errno.h>
#include <linux/mutex.h>
#include <linux/slab.h>

#include <linux/spi/spi.h>
#include <linux/spi/spidev.h>

#include <asm/uaccess.h>


/*
 * This supports acccess to SPI devices using normal userspace I/O calls.
 * Note that while traditional UNIX/POSIX I/O semantics are half duplex,
 * and often mask message boundaries, full SPI support requires full duplex
 * transfers.  There are several kinds of internal message boundaries to
 * handle chipselect management and other protocol options.
 *
 * SPI has a character major number assigned.  We allocate minor numbers
 * dynamically using a bitmask.  You must use hotplug tools, such as udev
 * (or mdev with busybox) to create and destroy the /dev/spidevB.C device
 * nodes, since there is no fixed association of minor numbers with any
 * particular SPI bus or device.
 */
#define SPIDEV_MAJOR   153 /* assigned */
#define N_SPI_MINORS   32 /* ... up to 256 */

static DECLARE_BITMAP(minors, N_SPI_MINORS);


/* Bit masks for spi_device.mode management.  Note that incorrect
 * settings for some settings can cause *lots* of trouble for other
 * devices on a shared bus:
 *
 *  - CS_HIGH ... this device will be active when it shouldn't be
 *  - 3WIRE ... when active, it won't behave as it should
 *  - NO_CS ... there will be no explicit message boundaries; this
 * is completely incompatible with the shared bus model
 *  - READY ... transfers may proceed when they shouldn't.
 *
 * REVISIT should changing those flags be privileged?
 */
#define SPI_MODE_MASK  (SPI_CPHA | SPI_CPOL | SPI_CS_HIGH \
    | SPI_LSB_FIRST | SPI_3WIRE | SPI_LOOP \
    | SPI_NO_CS | SPI_READY)

struct spidev_data {
 dev_t   devt;
 spinlock_t  spi_lock;
 struct spi_device *spi;
 struct list_head device_entry;

 /* buffer is NULL unless this device is open (users > 0) */
 struct mutex  buf_lock;
 unsigned  users;
 u8   *buffer;
};

static LIST_HEAD(device_list);
static DEFINE_MUTEX(device_list_lock);

static unsigned bufsiz = 4096;
module_param(bufsiz, uint, S_IRUGO);
MODULE_PARM_DESC(bufsiz, "data bytes in biggest supported SPI message");

/*-------------------------------------------------------------------------*/

/*
 * We can't use the standard synchronous wrappers for file I/O; we
 * need to protect against async removal of the underlying spi_device.
 */
static void spidev_complete(void *arg)
{
 complete(arg);
}

static ssize_t
spidev_sync(struct spidev_data *spidev, struct spi_message *message)
{
 DECLARE_COMPLETION_ONSTACK(done);
 int status;

 message->complete = spidev_complete;
 message->context = &done;

 spin_lock_irq(&spidev->spi_lock);
 if (spidev->spi == NULL)
  status = -ESHUTDOWN;
 else
  status = spi_async(spidev->spi, message);
 spin_unlock_irq(&spidev->spi_lock);

 if (status == 0) {
  wait_for_completion(&done);
  status = message->status;
  if (status == 0)
   status = message->actual_length;
 }
 return status;
}

static inline ssize_t
spidev_sync_write(struct spidev_data *spidev, size_t len)
{
 struct spi_transfer t = {
   .tx_buf  = spidev->buffer,
   .len  = len,
  };
 struct spi_message m;

 spi_message_init(&m);
 spi_message_add_tail(&t, &m);
 return spidev_sync(spidev, &m);
}

static inline ssize_t
spidev_sync_read(struct spidev_data *spidev, size_t len)
{
 struct spi_transfer t = {
   .rx_buf  = spidev->buffer,
   .len  = len,
  };
 struct spi_message m;

 spi_message_init(&m);
 spi_message_add_tail(&t, &m);
 return spidev_sync(spidev, &m);
}

/*-------------------------------------------------------------------------*/

/* Read-only message with current device setup */
static ssize_t
spidev_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos)
{
 struct spidev_data *spidev;
 ssize_t   status = 0;

 /* chipselect only toggles at start or end of operation */
 if (count > bufsiz)
  return -EMSGSIZE;

 spidev = filp->private_data;

 mutex_lock(&spidev->buf_lock);
 status = spidev_sync_read(spidev, count);
 if (status > 0) {
  unsigned long missing;

  missing = copy_to_user(buf, spidev->buffer, status);
  if (missing == status)
   status = -EFAULT;
  else
   status = status - missing;
 }
 mutex_unlock(&spidev->buf_lock);

 return status;
}

/* Write-only message with current device setup */
static ssize_t
spidev_write(struct file *filp, const char __user *buf,
  size_t count, loff_t *f_pos)
{
 struct spidev_data *spidev;
 ssize_t   status = 0;
 unsigned long  missing;

 /* chipselect only toggles at start or end of operation */
 if (count > bufsiz)
  return -EMSGSIZE;

 spidev = filp->private_data;

 mutex_lock(&spidev->buf_lock);
 missing = copy_from_user(spidev->buffer, buf, count);
 if (missing == 0) {
  status = spidev_sync_write(spidev, count);
 } else
  status = -EFAULT;
 mutex_unlock(&spidev->buf_lock);

 return status;
}

static int spidev_message(struct spidev_data *spidev,
  struct spi_ioc_transfer *u_xfers, unsigned n_xfers)
{
 struct spi_message msg;
 struct spi_transfer *k_xfers;
 struct spi_transfer *k_tmp;
 struct spi_ioc_transfer *u_tmp;
 unsigned  n, total;
 u8   *buf;
 int   status = -EFAULT;

 spi_message_init(&msg);
 k_xfers = kcalloc(n_xfers, sizeof(*k_tmp), GFP_KERNEL);
 if (k_xfers == NULL)
  return -ENOMEM;

 /* Construct spi_message, copying any tx data to bounce buffer.
  * We walk the array of user-provided transfers, using each one
  * to initialize a kernel version of the same transfer.
  */
 buf = spidev->buffer;
 total = 0;
 for (n = n_xfers, k_tmp = k_xfers, u_tmp = u_xfers;
   n;
   n--, k_tmp++, u_tmp++) {
  k_tmp->len = u_tmp->len;

  total += k_tmp->len;
  if (total > bufsiz) {
   status = -EMSGSIZE;
   goto done;
  }

  if (u_tmp->rx_buf) {
   k_tmp->rx_buf = buf;
   if (!access_ok(VERIFY_WRITE, (u8 __user *)
      (uintptr_t) u_tmp->rx_buf,
      u_tmp->len))
    goto done;
  }
  if (u_tmp->tx_buf) {
   k_tmp->tx_buf = buf;
   if (copy_from_user(buf, (const u8 __user *)
      (uintptr_t) u_tmp->tx_buf,
     u_tmp->len))
    goto done;
  }
  buf += k_tmp->len;

  k_tmp->cs_change = !!u_tmp->cs_change;
  k_tmp->bits_per_word = u_tmp->bits_per_word;
  k_tmp->delay_usecs = u_tmp->delay_usecs;
  k_tmp->speed_hz = u_tmp->speed_hz;
#ifdef VERBOSE
  dev_dbg(&spidev->spi->dev,
   "  xfer len %zd %s%s%s%dbits %u usec %uHz\n",
   u_tmp->len,
   u_tmp->rx_buf ? "rx " : "",
   u_tmp->tx_buf ? "tx " : "",
   u_tmp->cs_change ? "cs " : "",
   u_tmp->bits_per_word ? : spidev->spi->bits_per_word,
   u_tmp->delay_usecs,
   u_tmp->speed_hz ? : spidev->spi->max_speed_hz);
#endif
  spi_message_add_tail(k_tmp, &msg);
 }

 status = spidev_sync(spidev, &msg);
 if (status < 0)
  goto done;

 /* copy any rx data out of bounce buffer */
 buf = spidev->buffer;
 for (n = n_xfers, u_tmp = u_xfers; n; n--, u_tmp++) {
  if (u_tmp->rx_buf) {
   if (__copy_to_user((u8 __user *)
     (uintptr_t) u_tmp->rx_buf, buf,
     u_tmp->len)) {
    status = -EFAULT;
    goto done;
   }
  }
  buf += u_tmp->len;
 }
 status = total;

done:
 kfree(k_xfers);
 return status;
}

static long
spidev_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
{
 int   err = 0;
 int   retval = 0;
 struct spidev_data *spidev;
 struct spi_device *spi;
 u32   tmp;
 unsigned  n_ioc;
 struct spi_ioc_transfer *ioc;

 /* Check type and command number */
 if (_IOC_TYPE(cmd) != SPI_IOC_MAGIC)
  return -ENOTTY;

 /* Check access direction once here; don't repeat below.
  * IOC_DIR is from the user perspective, while access_ok is
  * from the kernel perspective; so they look reversed.
  */
 if (_IOC_DIR(cmd) & _IOC_READ)
  err = !access_ok(VERIFY_WRITE,
    (void __user *)arg, _IOC_SIZE(cmd));
 if (err == 0 && _IOC_DIR(cmd) & _IOC_WRITE)
  err = !access_ok(VERIFY_READ,
    (void __user *)arg, _IOC_SIZE(cmd));
 if (err)
  return -EFAULT;

 /* guard against device removal before, or while,
  * we issue this ioctl.
  */
 spidev = filp->private_data;
 spin_lock_irq(&spidev->spi_lock);
 spi = spi_dev_get(spidev->spi);
 spin_unlock_irq(&spidev->spi_lock);

 if (spi == NULL)
  return -ESHUTDOWN;

 /* use the buffer lock here for triple duty:
  *  - prevent I/O (from us) so calling spi_setup() is safe;
  *  - prevent concurrent SPI_IOC_WR_* from morphing
  *    data fields while SPI_IOC_RD_* reads them;
  *  - SPI_IOC_MESSAGE needs the buffer locked "normally".
  */
 mutex_lock(&spidev->buf_lock);

 switch (cmd) {
 /* read requests */
 case SPI_IOC_RD_MODE:
  retval = __put_user(spi->mode & SPI_MODE_MASK,
     (__u8 __user *)arg);
  break;
 case SPI_IOC_RD_LSB_FIRST:
  retval = __put_user((spi->mode & SPI_LSB_FIRST) ?  1 : 0,
     (__u8 __user *)arg);
  break;
 case SPI_IOC_RD_BITS_PER_WORD:
  retval = __put_user(spi->bits_per_word, (__u8 __user *)arg);
  break;
 case SPI_IOC_RD_MAX_SPEED_HZ:
  retval = __put_user(spi->max_speed_hz, (__u32 __user *)arg);
  break;

 /* write requests */
 case SPI_IOC_WR_MODE:
  retval = __get_user(tmp, (u8 __user *)arg);
  if (retval == 0) {
   u8 save = spi->mode;

   if (tmp & ~SPI_MODE_MASK) {
    retval = -EINVAL;
    break;
   }

   tmp |= spi->mode & ~SPI_MODE_MASK;
   spi->mode = (u8)tmp;
   retval = spi_setup(spi);
   if (retval < 0)
    spi->mode = save;
   else
    dev_dbg(&spi->dev, "spi mode %02x\n", tmp);
  }
  break;
 case SPI_IOC_WR_LSB_FIRST:
  retval = __get_user(tmp, (__u8 __user *)arg);
  if (retval == 0) {
   u8 save = spi->mode;

   if (tmp)
    spi->mode |= SPI_LSB_FIRST;
   else
    spi->mode &= ~SPI_LSB_FIRST;
   retval = spi_setup(spi);
   if (retval < 0)
    spi->mode = save;
   else
    dev_dbg(&spi->dev, "%csb first\n",
      tmp ? 'l' : 'm');
  }
  break;
 case SPI_IOC_WR_BITS_PER_WORD:
  retval = __get_user(tmp, (__u8 __user *)arg);
  if (retval == 0) {
   u8 save = spi->bits_per_word;

   spi->bits_per_word = tmp;
   retval = spi_setup(spi);
   if (retval < 0)
    spi->bits_per_word = save;
   else
    dev_dbg(&spi->dev, "%d bits per word\n", tmp);
  }
  break;
 case SPI_IOC_WR_MAX_SPEED_HZ:
  retval = __get_user(tmp, (__u32 __user *)arg);
  if (retval == 0) {
   u32 save = spi->max_speed_hz;

   spi->max_speed_hz = tmp;
   retval = spi_setup(spi);
   if (retval < 0)
    spi->max_speed_hz = save;
   else
    dev_dbg(&spi->dev, "%d Hz (max)\n", tmp);
  }
  break;

 default:
  /* segmented and/or full-duplex I/O request */
  if (_IOC_NR(cmd) != _IOC_NR(SPI_IOC_MESSAGE(0))
    || _IOC_DIR(cmd) != _IOC_WRITE) {
   retval = -ENOTTY;
   break;
  }

  tmp = _IOC_SIZE(cmd);
  if ((tmp % sizeof(struct spi_ioc_transfer)) != 0) {
   retval = -EINVAL;
   break;
  }
  n_ioc = tmp / sizeof(struct spi_ioc_transfer);
  if (n_ioc == 0)
   break;

  /* copy into scratch area */
  ioc = kmalloc(tmp, GFP_KERNEL);
  if (!ioc) {
   retval = -ENOMEM;
   break;
  }
  if (__copy_from_user(ioc, (void __user *)arg, tmp)) {
   kfree(ioc);
   retval = -EFAULT;
   break;
  }

  /* translate to spi_message, execute */
  retval = spidev_message(spidev, ioc, n_ioc);
  kfree(ioc);
  break;
 }

 mutex_unlock(&spidev->buf_lock);
 spi_dev_put(spi);
 return retval;
}

static int spidev_open(struct inode *inode, struct file *filp)
{
 struct spidev_data *spidev;
 int   status = -ENXIO;

 mutex_lock(&device_list_lock);

 list_for_each_entry(spidev, &device_list, device_entry) {
  if (spidev->devt == inode->i_rdev) {
   status = 0;
   break;
  }
 }
 if (status == 0) {
  if (!spidev->buffer) {
   spidev->buffer = kmalloc(bufsiz, GFP_KERNEL);
   if (!spidev->buffer) {
    dev_dbg(&spidev->spi->dev, "open/ENOMEM\n");
    status = -ENOMEM;
   }
  }
  if (status == 0) {
   spidev->users++;
   filp->private_data = spidev;
   nonseekable_open(inode, filp);
  }
 } else
  pr_debug("spidev: nothing for minor %d\n", iminor(inode));

 mutex_unlock(&device_list_lock);
 return status;
}

static int spidev_release(struct inode *inode, struct file *filp)
{
 struct spidev_data *spidev;
 int   status = 0;

 mutex_lock(&device_list_lock);
 spidev = filp->private_data;
 filp->private_data = NULL;

 /* last close? */
 spidev->users--;
 if (!spidev->users) {
  int  dofree;

  kfree(spidev->buffer);
  spidev->buffer = NULL;

  /* ... after we unbound from the underlying device? */
  spin_lock_irq(&spidev->spi_lock);
  dofree = (spidev->spi == NULL);
  spin_unlock_irq(&spidev->spi_lock);

  if (dofree)
   kfree(spidev);
 }
 mutex_unlock(&device_list_lock);

 return status;
}

static const struct file_operations spidev_fops = {
 .owner = THIS_MODULE,
 /* REVISIT switch to aio primitives, so that userspace
  * gets more complete API coverage.  It'll simplify things
  * too, except for the locking.
  */
 .write = spidev_write,
 .read =  spidev_read,
 .unlocked_ioctl = spidev_ioctl,
 .open =  spidev_open,
 .release = spidev_release,
};

/*-------------------------------------------------------------------------*/

/* The main reason to have this class is to make mdev/udev create the
 * /dev/spidevB.C character device nodes exposing our userspace API.
 * It also simplifies memory management.
 */

static struct class *spidev_class;

/*-------------------------------------------------------------------------*/

static int __devinit spidev_probe(struct spi_device *spi)
{
 struct spidev_data *spidev;
 int   status;
 unsigned long  minor;

 /* Allocate driver data */
 spidev = kzalloc(sizeof(*spidev), GFP_KERNEL);
 if (!spidev)
  return -ENOMEM;

 /* Initialize the driver data */
 spidev->spi = spi;
 spin_lock_init(&spidev->spi_lock);
 mutex_init(&spidev->buf_lock);

 INIT_LIST_HEAD(&spidev->device_entry);

 /* If we can allocate a minor number, hook up this device.
  * Reusing minors is fine so long as udev or mdev is working.
  */
 mutex_lock(&device_list_lock);
 minor = find_first_zero_bit(minors, N_SPI_MINORS);
 if (minor < N_SPI_MINORS) {
  struct device *dev;

  spidev->devt = MKDEV(SPIDEV_MAJOR, minor);
  dev = device_create(spidev_class, &spi->dev, spidev->devt,
        spidev, "spidev%d.%d",
        spi->master->bus_num, spi->chip_select);
  status = IS_ERR(dev) ? PTR_ERR(dev) : 0;
 } else {
  dev_dbg(&spi->dev, "no minor number available!\n");
  status = -ENODEV;
 }
 if (status == 0) {
  set_bit(minor, minors);
  list_add(&spidev->device_entry, &device_list);
 }
 mutex_unlock(&device_list_lock);

 if (status == 0)
  spi_set_drvdata(spi, spidev);
 else
  kfree(spidev);

 return status;
}

static int __devexit spidev_remove(struct spi_device *spi)
{
 struct spidev_data *spidev = spi_get_drvdata(spi);

 /* make sure ops on existing fds can abort cleanly */
 spin_lock_irq(&spidev->spi_lock);
 spidev->spi = NULL;
 spi_set_drvdata(spi, NULL);
 spin_unlock_irq(&spidev->spi_lock);

 /* prevent new opens */
 mutex_lock(&device_list_lock);
 list_del(&spidev->device_entry);
 device_destroy(spidev_class, spidev->devt);
 clear_bit(MINOR(spidev->devt), minors);
 if (spidev->users == 0)
  kfree(spidev);
 mutex_unlock(&device_list_lock);

 return 0;
}

static struct spi_driver spidev_spi_driver = {
 .driver = {
  .name =  "spidev",
  .owner = THIS_MODULE,
 },
 .probe = spidev_probe,
 .remove = __devexit_p(spidev_remove),

 /* NOTE:  suspend/resume methods are not necessary here.
  * We don't do anything except pass the requests to/from
  * the underlying controller.  The refrigerator handles
  * most issues; the controller driver handles the rest.
  */
};

/*-------------------------------------------------------------------------*/

static int __init spidev_init(void)
{
 int status;

 /* Claim our 256 reserved device numbers.  Then register a class
  * that will key udev/mdev to add/remove /dev nodes.  Last, register
  * the driver which manages those device numbers.
  */
 BUILD_BUG_ON(N_SPI_MINORS > 256);
 status = register_chrdev(SPIDEV_MAJOR, "spi", &spidev_fops);
 if (status < 0)
  return status;

 spidev_class = class_create(THIS_MODULE, "spidev");
 if (IS_ERR(spidev_class)) {
  unregister_chrdev(SPIDEV_MAJOR, spidev_spi_driver.driver.name);
  return PTR_ERR(spidev_class);
 }

 status = spi_register_driver(&spidev_spi_driver);
 if (status < 0) {
  class_destroy(spidev_class);
  unregister_chrdev(SPIDEV_MAJOR, spidev_spi_driver.driver.name);
 }
 return status;
}
module_init(spidev_init);

static void __exit spidev_exit(void)
{
 spi_unregister_driver(&spidev_spi_driver);
 class_destroy(spidev_class);
 unregister_chrdev(SPIDEV_MAJOR, spidev_spi_driver.driver.name);
}
module_exit(spidev_exit);

MODULE_AUTHOR("Andrea Paterniani, <a.paterniani@swapp-eng.it>");
MODULE_DESCRIPTION("User mode SPI device interface");
MODULE_LICENSE("GPL");
MODULE_ALIAS("spi:spidev");