如何在 Linux kernel v3.2 中创建一个简单的 sysfs 类属性
声明:本页面是StackOverFlow热门问题的中英对照翻译,遵循CC BY-SA 4.0协议,如果您需要使用它,必须同样遵循CC BY-SA许可,注明原文地址和作者信息,同时你必须将它归于原作者(不是我):StackOverFlow
原文地址: http://stackoverflow.com/questions/16245100/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
How to create a simple sysfs class attribute in Linux kernel v3.2
提问by Vilhelm Gray
I'm learning how to use sysfsin my Linux modules, but I'm having the hardest time finding current documentation on these topics. The Linux Device Drivers 3rd Editionbook I've been using seems to be rather dated in this area unfortunately (e.g. the class_device
structure appears to be completely gone in current Linux versions).
我正在学习如何在我的 Linux 模块中使用sysfs,但我最难找到有关这些主题的最新文档。不幸的是,我一直在使用的Linux Device Drivers 3rd Edition书在这方面似乎已经过时了(例如,class_device
在当前的 Linux 版本中,该结构似乎完全消失了)。
I'm simply trying to get an attribute to appear, under the respective sysfsclass for my module, that will allow me to read the value of a module variable from kernel space.
我只是想让一个属性出现在我模块的相应sysfs类下,这将允许我从内核空间读取模块变量的值。
In my code, I have a class created that allows udev to create a device node at /dev/foofor my module:
在我的代码中,我创建了一个类,允许 udev 在/dev/foo为我的模块创建一个设备节点:
dev_t foo_dev;
alloc_chrdev_region(&foo_dev, 0, 1, "bar");
struct class *bar = class_create(THIS_MODULE, "bar");
device_create(bar, NULL, foo_dev, NULL, "foo");
struct cdev foo_dev_file;
cdev_init(&foo_dev_file, &fops); /* fops defined earlier */
cdev_add(&foo_dev_file, foo_dev, 1);
When I insert the module I get a sysfsclass directory created and populated with some default attributes at /sys/class/bar/foo/. How can I create attributes that show up under this new directory?
当我插入模块时,我会在/sys/class/bar/foo/处创建一个sysfs类目录并填充一些默认属性。如何创建显示在此新目录下的属性?
I have the concepts down pretty well I believe -- create attribute structure, define sysfs_ops
functions, etc -- my problem is that I don't know which particular kernel structure to use (class_attribute
?), nor how to make these attributes appear under the right sysfsdirectory.
我相信我已经很好地理解了这些概念 - 创建属性结构,定义sysfs_ops
函数等 - 我的问题是我不知道要使用哪个特定的内核结构(class_attribute
?),也不知道如何使这些属性出现在右侧sysfs目录。
Would anyone point me to a tutorial or article detailing the process for current Linux kernels?
有没有人会告诉我详细介绍当前 Linux 内核过程的教程或文章?
采纳答案by Daniel Santos
Even though my knowledge is still fairly low on the topic, I'm going to post an answer just because of the age of this question. If somebody else has a better answer, please post!:)
尽管我对这个主题的了解仍然很低,但我还是会因为这个问题的年龄而发布一个答案。如果其他人有更好的答案,请发布!:)
First off, I'm going to assume that you've read that whole chapter (specifically about kobjects & ksets). So just about every struct in the device driver model has these cutely included in them. If you want to manipulate the kobject for the class its self (not sure if that's wise or not), that's your struct class
's dev_kobj
member.
首先,我假设您已经阅读了整章(特别是关于 kobjects 和 ksets)。因此,几乎设备驱动程序模型中的每个结构体都可爱地包含了这些。如果你想为自己的类操作 kobject(不确定这是否明智),那就是你struct class
的dev_kobj
成员。
However, you want to manipulate the attributes of that class. I believeyou do this by defining a (usually static), NULL-terminated array of them as follows and then assigning its address to the struct class
's class_attrs
member (taken from drivers/uwb/driver.c
):
但是,您想要操作该类的属性。我相信您可以通过如下定义一个(通常是静态的)、以 NULL 结尾的数组,然后将其地址分配给struct class
的class_attrs
成员(取自drivers/uwb/driver.c
):
static struct class_attribute uwb_class_attrs[] = {
__ATTR(beacon_timeout_ms, S_IWUSR | S_IRUGO,
beacon_timeout_ms_show, beacon_timeout_ms_store),
__ATTR_NULL,
};
/** Device model classes */
struct class uwb_rc_class = {
.name = "uwb_rc",
.class_attrs = uwb_class_attrs,
};
When I don't know how to use something, I usually git grep
the repository for somebody else who has used it and try to learn from it that way. It would seem that this is why they tend to say kernel "hackers" and not "developers".
当我不知道如何使用某些东西时,我通常会git grep
为使用过它的其他人提供存储库,并尝试以这种方式从中学习。似乎这就是为什么他们倾向于说内核“黑客”而不是“开发人员”的原因。
回答by mehmet riza oz
There is a good tutorial in the link below
下面的链接中有一个很好的教程
http://pete.akeo.ie/2011/08/writing-linux-device-driver-for-kernels.html
http://pete.akeo.ie/2011/08/writing-linux-device-driver-for-kernels.html
parrot_driver.c:
parrot_driver.c:
/*
* Linux 2.6 and 3.0 'parrot' sample device driver
*
* Copyright (c) 2011, Pete Batard <[email protected]>
*
*
* 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, see <http://www.gnu.org/licenses/>.
*/
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/device.h>
#include <linux/types.h>
#include <linux/mutex.h>
#include <linux/kfifo.h>
#include "parrot_driver.h"
/* Module information */
MODULE_AUTHOR(AUTHOR);
MODULE_DESCRIPTION(DESCRIPTION);
MODULE_VERSION(VERSION);
MODULE_LICENSE("GPL");
/* Device variables */
static struct class* parrot_class = NULL;
static struct device* parrot_device = NULL;
static int parrot_major;
/* Flag used with the one_shot mode */
static bool message_read;
/* A mutex will ensure that only one process accesses our device */
static DEFINE_MUTEX(parrot_device_mutex);
/* Use a Kernel FIFO for read operations */
static DECLARE_KFIFO(parrot_msg_fifo, char, PARROT_MSG_FIFO_SIZE);
/* This table keeps track of each message length in the FIFO */
static unsigned int parrot_msg_len[PARROT_MSG_FIFO_MAX];
/* Read and write index for the table above */
static int parrot_msg_idx_rd, parrot_msg_idx_wr;
/* Module parameters that can be provided on insmod */
static bool debug = false; /* print extra debug info */
module_param(debug, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(debug, "enable debug info (default: false)");
static bool one_shot = true; /* only read a single message after open() */
module_param(one_shot, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(debug, "disable the readout of multiple messages at once (default: true)");
static int parrot_device_open(struct inode* inode, struct file* filp)
{
dbg("");
/* Our sample device does not allow write access */
if ( ((filp->f_flags & O_ACCMODE) == O_WRONLY)
|| ((filp->f_flags & O_ACCMODE) == O_RDWR) ) {
warn("write access is prohibited\n");
return -EACCES;
}
/* Ensure that only one process has access to our device at any one time
* For more info on concurrent accesses, see http://lwn.net/images/pdf/LDD3/ch05.pdf */
if (!mutex_trylock(&parrot_device_mutex)) {
warn("another process is accessing the device\n");
return -EBUSY;
}
message_read = false;
return 0;
}
static int parrot_device_close(struct inode* inode, struct file* filp)
{
dbg("");
mutex_unlock(&parrot_device_mutex);
return 0;
}
static ssize_t parrot_device_read(struct file* filp, char __user *buffer, size_t length, loff_t* offset)
{
int retval;
unsigned int copied;
/* The default from 'cat' is to issue multiple reads until the FIFO is depleted
* one_shot avoids that */
if (one_shot && message_read) return 0;
dbg("");
if (kfifo_is_empty(&parrot_msg_fifo)) {
dbg("no message in fifo\n");
return 0;
}
retval = kfifo_to_user(&parrot_msg_fifo, buffer, parrot_msg_len[parrot_msg_idx_rd], &copied);
/* Ignore short reads (but warn about them) */
if (parrot_msg_len[parrot_msg_idx_rd] != copied) {
warn("short read detected\n");
}
/* loop into the message length table */
parrot_msg_idx_rd = (parrot_msg_idx_rd+1)%PARROT_MSG_FIFO_MAX;
message_read = true;
return retval ? retval : copied;
}
/* The file_operation scructure tells the kernel which device operations are handled.
* For a list of available file operations, see http://lwn.net/images/pdf/LDD3/ch03.pdf */
static struct file_operations fops = {
.read = parrot_device_read,
.open = parrot_device_open,
.release = parrot_device_close
};
/* Placing data into the read FIFO is done through sysfs */
static ssize_t sys_add_to_fifo(struct device* dev, struct device_attribute* attr, const char* buf, size_t count)
{
unsigned int copied;
dbg("");
if (kfifo_avail(&parrot_msg_fifo) < count) {
warn("not enough space left on fifo\n");
return -ENOSPC;
}
if ((parrot_msg_idx_wr+1)%PARROT_MSG_FIFO_MAX == parrot_msg_idx_rd) {
/* We've looped into our message length table */
warn("message length table is full\n");
return -ENOSPC;
}
/* The buffer is already in kernel space, so no need for ..._from_user() */
copied = kfifo_in(&parrot_msg_fifo, buf, count);
parrot_msg_len[parrot_msg_idx_wr] = copied;
if (copied != count) {
warn("short write detected\n");
}
parrot_msg_idx_wr = (parrot_msg_idx_wr+1)%PARROT_MSG_FIFO_MAX;
return copied;
}
/* This sysfs entry resets the FIFO */
static ssize_t sys_reset(struct device* dev, struct device_attribute* attr, const char* buf, size_t count)
{
dbg("");
/* Ideally, we would have a mutex around the FIFO, to ensure that we don't reset while in use.
* To keep this sample simple, and because this is a sysfs operation, we don't do that */
kfifo_reset(&parrot_msg_fifo);
parrot_msg_idx_rd = parrot_msg_idx_wr = 0;
return count;
}
/* Declare the sysfs entries. The macros create instances of dev_attr_fifo and dev_attr_reset */
static DEVICE_ATTR(fifo, S_IWUSR, NULL, sys_add_to_fifo);
static DEVICE_ATTR(reset, S_IWUSR, NULL, sys_reset);
/* Module initialization and release */
static int __init parrot_module_init(void)
{
int retval;
dbg("");
/* First, see if we can dynamically allocate a major for our device */
parrot_major = register_chrdev(0, DEVICE_NAME, &fops);
if (parrot_major < 0) {
err("failed to register device: error %d\n", parrot_major);
retval = parrot_major;
goto failed_chrdevreg;
}
/* We can either tie our device to a bus (existing, or one that we create)
* or use a "virtual" device class. For this example, we choose the latter */
parrot_class = class_create(THIS_MODULE, CLASS_NAME);
if (IS_ERR(parrot_class)) {
err("failed to register device class '%s'\n", CLASS_NAME);
retval = PTR_ERR(parrot_class);
goto failed_classreg;
}
/* With a class, the easiest way to instantiate a device is to call device_create() */
parrot_device = device_create(parrot_class, NULL, MKDEV(parrot_major, 0), NULL, CLASS_NAME "_" DEVICE_NAME);
if (IS_ERR(parrot_device)) {
err("failed to create device '%s_%s'\n", CLASS_NAME, DEVICE_NAME);
retval = PTR_ERR(parrot_device);
goto failed_devreg;
}
/* Now we can create the sysfs endpoints (don't care about errors).
* dev_attr_fifo and dev_attr_reset come from the DEVICE_ATTR(...) earlier */
retval = device_create_file(parrot_device, &dev_attr_fifo);
if (retval < 0) {
warn("failed to create write /sys endpoint - continuing without\n");
}
retval = device_create_file(parrot_device, &dev_attr_reset);
if (retval < 0) {
warn("failed to create reset /sys endpoint - continuing without\n");
}
mutex_init(&parrot_device_mutex);
/* This device uses a Kernel FIFO for its read operation */
INIT_KFIFO(parrot_msg_fifo);
parrot_msg_idx_rd = parrot_msg_idx_wr = 0;
return 0;
failed_devreg:
class_unregister(parrot_class);
class_destroy(parrot_class);
failed_classreg:
unregister_chrdev(parrot_major, DEVICE_NAME);
failed_chrdevreg:
return -1;
}
static void __exit parrot_module_exit(void)
{
dbg("");
device_remove_file(parrot_device, &dev_attr_fifo);
device_remove_file(parrot_device, &dev_attr_reset);
device_destroy(parrot_class, MKDEV(parrot_major, 0));
class_unregister(parrot_class);
class_destroy(parrot_class);
unregister_chrdev(parrot_major, DEVICE_NAME);
}
/* Let the kernel know the calls for module init and exit */
module_init(parrot_module_init);
module_exit(parrot_module_exit);
parrot_driver.h:
parrot_driver.h:
/*
* Linux 2.6 and 3.0 'parrot' sample device driver
*
* Copyright (c) 2011, Pete Batard <[email protected]>
*
*
* 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, see <http://www.gnu.org/licenses/>.
*/
#define DEVICE_NAME "device"
#define CLASS_NAME "parrot"
#define PARROT_MSG_FIFO_SIZE 1024
#define PARROT_MSG_FIFO_MAX 128
#define AUTHOR "Pete Batard <[email protected]>"
#define DESCRIPTION "'parrot' sample device driver"
#define VERSION "0.3"
/* We'll use our own macros for printk */
#define dbg(format, arg...) do { if (debug) pr_info(CLASS_NAME ": %s: " format , __FUNCTION__ , ## arg); } while (0)
#define err(format, arg...) pr_err(CLASS_NAME ": " format, ## arg)
#define info(format, arg...) pr_info(CLASS_NAME ": " format, ## arg)
#define warn(format, arg...) pr_warn(CLASS_NAME ": " format, ## arg)