(1)块设备只能以块为单位接受输入和返回输出,而字符设备则以字节为单位。大多数设备是字符设备,因为它们不需要缓冲而且不以固定块大小进行操作。
(2)块设备对于 I/O 请求有对应的缓冲区,因此它们可以选择以什么顺序进行响应,字符设备无须缓冲且被直接读写。对于存储设备而言调整读写的顺序作用巨大,因为在读写连续的扇区比分离的扇区更快。
(3)字符设备只能被顺序读写,而块设备可以随机访问。虽然块设备可随机访问,但是对于磁盘这类机械设备而言,顺序地组织块设备的访问可以提高性能,而对 SD 卡、RamDisk 等块设备而言,不存在机械上的原因,进行这样的调整没有必要。
2. 块设备例程
将ldd3(linxu device driver 3)中块设备驱动部分的代码在kerne 3.19.0 编译时,会出现很多问题,主要是由于ldd3示例代码使用的内核版本较低(2.6.10版本),对于块设备子系统,很多接口都已经发生了改变,修改后的驱动代码如下ramdisk.c(亲测可用):
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/vmalloc.h>
#include <linux/blkdev.h>
#include <linux/hdreg.h>
#define RAMDISK_NAME "ramdisk"
#define RAMDISK_MAX_DEVICE 1
#define RAMDISK_MAX_PARTITIONS 4
#define RAMDISK_SECTOR_SIZE 512
#define RAMDISK_SECTORS 16
#define RAMDISK_HEADS 4
#define RAMDISK_CYLINDERS 256
#define RAMDISK_SECTOR_TOTAL (RAMDISK_SECTORS * RAMDISK_HEADS * RAMDISK_CYLINDERS * 50) //16383 sectors
#define RAMDISK_SIZE (RAMDISK_SECTOR_SIZE * RAMDISK_SECTOR_TOTAL) //8MB = 16383 x 512k
typedef struct {
unsigned char* data;
struct request_queue* queue;
struct gendisk* gd;
}RAMDISK_DEV;
static char* sdisk[RAMDISK_MAX_DEVICE] = {NULL};
static RAMDISK_DEV* rdev[RAMDISK_MAX_DEVICE] = {NULL};
static dev_t ramdisk_major;
static int ramdisk_space_init(void)
{
int i;
int err = 0;
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
sdisk[i] = vmalloc(RAMDISK_SIZE);
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
if(!sdisk[i]){
err = -ENOMEM;
return err;
}
memset(sdisk[i], 0, RAMDISK_SIZE);
}
return err;
}
static void ramdisk_space_clean(void)
{
int i;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
vfree(sdisk[i]);
}
}
static int ramdisk_open(struct block_device* bdev, fmode_t mode)
{
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
return 0;
}
static void ramdisk_release(struct gendisk*gd, fmode_t mode)
{
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
return;
}
static int ramdisk_ioctl(struct block_device* bdev, fmode_t mode, unsigned int cmd, unsigned long arg)
{
int err;
struct hd_geometry geo;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
switch(cmd)
{
case HDIO_GETGEO:
err = !access_ok(VERIFY_WRITE, arg, sizeof(geo));
if(err)
return -EFAULT;
geo.cylinders = RAMDISK_CYLINDERS;
geo.heads = RAMDISK_HEADS;
geo.sectors = RAMDISK_SECTORS;
geo.start = get_start_sect(bdev);
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
if(copy_to_user((void*)arg, &geo, sizeof(geo)))
return -EFAULT;
return 0;
}
return -ENOTTY;
}
static struct block_device_operations ramdisk_fops = {
.owner = THIS_MODULE,
.open = ramdisk_open,
.release = ramdisk_release,
.ioctl = ramdisk_ioctl,
};
static int ramdisk_make_request(struct request_queue* q, struct bio* bio)
{
char* pRHdata;
char* pBuffer;
struct bio_vec bvec;
int err = 0;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
struct block_device* bdev = bio->bi_bdev;
RAMDISK_DEV* pdev = bdev->bd_disk->private_data;
if(((bio->bi_iter.bi_sector * RAMDISK_SECTOR_SIZE) + bio->bi_iter.bi_size) > RAMDISK_SIZE){
err = -EIO;
return err;
}
pRHdata = pdev->data + (bio->bi_iter.bi_sector * RAMDISK_SECTOR_SIZE);
bio_for_each_segment(bvec, bio, bio->bi_iter){
pBuffer = kmap(bvec.bv_page) + bvec.bv_offset;
switch(bio_data_dir(bio)){
case READ:
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
memcpy(pBuffer, pRHdata, bvec.bv_len);
flush_dcache_page(bvec.bv_page);
break;
case WRITE:
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
flush_dcache_page(bvec.bv_page);
memcpy(pRHdata, pBuffer, bvec.bv_len);
break;
default:
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
kunmap(bvec.bv_page);
goto out;
}
kunmap(bvec.bv_page);
pRHdata += bvec.bv_len;
}
out:
bio_endio(bio, err);
return 0;
}
static int alloc_ramdev(void)
{
int i;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
rdev[i] = kzalloc(sizeof(RAMDISK_DEV), GFP_KERNEL);
if(!rdev[i]){
return -ENOMEM;
}
}
return 0;
}
static void clean_ramdev(void)
{
int i;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
if(rdev[i])
kfree(rdev[i]);
}
}
static int __init ramdisk_init(void)
{
int i;
ramdisk_space_init();
alloc_ramdev();
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
ramdisk_major = register_blkdev(0, RAMDISK_NAME);
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
rdev[i]->data = sdisk[i];
rdev[i]->queue = blk_alloc_queue(GFP_KERNEL);
blk_queue_make_request(rdev[i]->queue, ramdisk_make_request);
rdev[i]->gd = alloc_disk(RAMDISK_MAX_PARTITIONS);
rdev[i]->gd->major = ramdisk_major;
rdev[i]->gd->first_minor = i * RAMDISK_MAX_PARTITIONS;
rdev[i]->gd->fops = &ramdisk_fops;
rdev[i]->gd->queue = rdev[i]->queue;
rdev[i]->gd->private_data = rdev[i];
sprintf(rdev[i]->gd->disk_name, "ramdisk%c", 'A' +i);
rdev[i]->gd->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
set_capacity(rdev[i]->gd, RAMDISK_SECTOR_TOTAL);
add_disk(rdev[i]->gd);
}
return 0;
}
static void __exit ramdisk_exit(void)
{
int i;
printk(KERN_EMERG "%s*********Start*******%d \n", __FUNCTION__, __LINE__);
for(i = 0; i < RAMDISK_MAX_DEVICE; i++){
del_gendisk(rdev[i]->gd);
put_disk(rdev[i]->gd);
blk_cleanup_queue(rdev[i]->queue);
}
clean_ramdev();
ramdisk_space_clean();
unregister_blkdev(ramdisk_major, RAMDISK_NAME);
}
module_init(ramdisk_init);
module_exit(ramdisk_exit);
MODULE_AUTHOR("zkj");
MODULE_DESCRIPTION("block device");
MODULE_LICENSE("GPL");
Makefile参考:
KVERS = $(shell uname -r)
CURDIR = $(shell pwd)
# Kernel modules
obj-m += ramdisk.o
# Specify flags for the module compilation.
#EXTRA_CFLAGS=-g -O0
build: kernel_modules
kernel_modules:
# make -C /lib/modules/$(KVERS)/build M=$(CURDIR) modules
make -C ../linux-3.16 M=$(CURDIR) modules
clean:
# make -C /lib/modules/$(KVERS)/build M=$(CURDIR) clean
make -C ../linux-3.16 M=$(CURDIR) clean
Terminal下编译(arm平台):
ubuntu@ubuntu:~/WorkSpace/kernel-src/ ramdisk$ make CROSS_COMPILE=arm-linux- ARCH=arm内核模块加载:
insmod ramdisk.ko
查看加载情况:
#sudo insmod ramdisk.ko
#
下面内容参考:http://blog.csdn.net/rocky_zhm/article/details/50372243
用lsmod看看。这里我们注意到,该模块的Used by为0,因为它既没有被其他模块使用,也没有被mount。
# lsmod
Module Size Used by
simp_blkdev 16784008 0
...
#
如果当前系统支持udev,在调用add_disk()函数时即插即用机制会自动为我们在/dev/目录下建立设备文件。
设备文件的名称为我们在gendisk.disk_name中设置的simp_blkdev,主、从设备号也是我们在程序中设定的72和0。
如果当前系统不支持udev,那么很不幸,你需要自己用mknod /dev/simp_blkdev b 72 0来创建设备文件了。
# ls -l /dev/ramdiskA
brw-rw---- 1 root disk 251, 0 12月 21 14:23 /dev/ramdiskA
创建块设备中的分区
sudo fdisk /dev/ramdiskADevice contains neither a valid DOS partition table, nor Sun, SGI or OSF disklabel
Building a new DOS disklabel with disk identifier 0x7670efa4.
Changes will remain in memory only, until you decide to write them.
After that, of course, the previous content won't be recoverable.
Warning: invalid flag 0x0000 of partition table 4 will be corrected by w(rite)
Command (m for help): n // 创建一个新分区
Partition type:
p primary (0 primary, 0 extended, 4 free)
e extended
Select (default p): p //选择分区类型,主分区(p)
Partition number (1-4, default 1): 1 // 分区编号,默认为1
First sector (2048-819199, default 2048): 2048 //分区起始sector默认为2048
Last sector, +sectors or +size{K,M,G} (2048-819199, default 819199): 200000 //设定分区大小
Command (m for help): w //保存设定的分区信息
The partition table has been altered!
Calling ioctl() to re-read partition table.
Syncing disks.
在块设备中创建文件系统,这里我们创建常用的ext3。
当然,作为通用的块设备,创建其他类型的文件系统也没问题。
# mkfs.ext3 /dev/ramdiskA
mke2fs 1.39 (29-May-2006)
Filesystem label=
OS type: Linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
4096 inodes, 16384 blocks
819 blocks (5.00%) reserved for the super user
First data block=1
Maximum filesystem blocks=16777216
2 block groups
8192 blocks per group, 8192 fragments per group
2048 inodes per group
Superblock backups stored on blocks:
8193
Writing inode tables: done
Creating journal (1024 blocks): done
Writing superblocks and filesystem accounting information: done
This filesystem will be automatically checked every 38 mounts or
180 days, whichever comes first. Use tune2fs -c or -i to override.
#
如果这是第一次使用,建议创建一个目录用来mount这个设备中的文件系统。
当然,这不是必需的。如果你对mount之类的用法很熟,你完全能够自己决定在这里干什么,甚至把这个设备mount成root。
# mkdir -p /mnt/temp1
#
把建立好文件系统的块设备mount到刚才建立的目录中
# mount /dev/ramdiskA /mnt/temp1
#
看看现在的mount表
# mount
...
/dev/ramdiskA on /mnt/temp1 type ext3 (rw)
#
看看现在的模块引用计数,从刚才的0变成1了,
原因是我们mount了。
# lsmod
Module Size Used by
ramdiskA 16784008 1
...
#
看看文件系统的内容,有个mkfs时自动建立的lost+found目录。
# ls /mnt/temp1
lost+found
#
随便拷点东西进去
# cp /etc/init.d/* /mnt/temp1
#
再看看
# ls /mnt/temp1
acpid conman functions irqbalance mdmpd NetworkManagerDispatcher rdisc sendmail winbind
anacron cpuspeed gpm kdump messagebus nfs readahead_early setroubleshoot wpa_supplicant
apmd crond haldaemon killall microcode_ctl nfslock readahead_later single xfs
atd cups halt krb524 multipathd nscd restorecond smartd xinetd
auditd cups-config-daemon hidd kudzu netconsole ntpd rhnsd smb ypbind
autofs dhcdbd ip6tables lost+found netfs pand rpcgssd sshd yum-updatesd
avahi-daemon dund ipmi lvm2-monitor netplugd pcscd rpcidmapd syslog
avahi-dnsconfd firstboot iptables mcstrans network portmap rpcsvcgssd vmware
bluetooth frecord irda mdmonitor NetworkManager psacct saslauthd vncserver
#
现在这个块设备的使用情况是
# df
文件系统 1K-块 已用 可用 已用% 挂载点
...
/dev/ramdiskA 15863 1440 13604 10% /mnt/temp1
#
再全删了玩玩
# rm -rf /mnt/temp1/*
#
看看删完了没有
# ls /mnt/temp1
#
好了,大概玩够了,我们把文件系统umount掉
# umount /mnt/temp1
#
模块的引用计数应该还原成0了吧
# lsmod
Module Size Used by
ramdiskA 16784008 0
...
#
最后一步,移除模块
# sudo rmmod ramdiskA
#