一、剖析一个块设备
块设备最小的可寻址单元是扇区。 扇区大小一般是2的整数倍,最常见的是512字节。
因为各种软件的用途不同,所以他们都会用到自己的最小逻辑可寻址单元----块。块只能基于文件系统,是一种抽象。
而扇区是设备的最小可寻址单元,所以块不能比扇区还小,只能数倍于扇区大小。此外内核要求块是2的整数倍。
至此,块最终要求是:必须是扇区大小的2的整数倍,并且要小于页面大小。所以通常块是512字节、1KB或4KB。
扇区:设备最小寻址单元。别名:硬扇区、设备块
块:文件系统的最小寻址单元。别名:文件块、I/O块
块包含一个或多个扇区,但大小不能超过一个页。所以页可以容纳一个或多个内存中的块。
二、缓冲区和缓冲区头
当一个块被调入内存时,它要存储在一个缓冲区中。每个缓冲区与一个块对应,相当于磁盘块在内存中的表示。
每个缓冲区都有一个对应的描述符。该描述符用buffer_head结构体表示,称作缓冲区头,在文件<linux/buffer_head.h>中定义。
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
struct buffer_head { unsigned long b_state; /* buffer state bitmap (see above)缓冲区状态标志位 */ struct buffer_head *b_this_page;/* circular list of page's buffers页面中的缓冲区 */ struct page *b_page; /* the page this bh is mapped to 存储缓冲区的页面*/ sector_t b_blocknr; /* start block number 其实块号*/ size_t b_size; /* size of mapping 映像的大小*/ char *b_data; /* pointer to data within the page 页面内的数据指针*/ struct block_device *b_bdev; /* 相关联的块设备 */ bh_end_io_t *b_end_io; /* I/O completion I/O完成方法*/ void *b_private; /* reserved for b_end_io io完成方法*/ struct list_head b_assoc_buffers; /* associated with another mapping 相关的映射链表*/ struct address_space *b_assoc_map; /* mapping this buffer is associated with 相关的地址空间*/ atomic_t b_count; /* users using this buffer_head 缓冲区使用计数*/ };
b_count域表示缓冲区的使用记数,可通过两个定义在文件<linux/buffer_head.h>中内联函数对此域进行增减。
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
static inline void get_bh(struct buffer_head *bh) { atomic_inc(&bh->b_count); }
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
static inline void put_bh(struct buffer_head *bh) { atomic_dec(&bh->b_count); }
合法的标志存放在bh_state_bits枚举中,该枚举在<linux/buffer_head.h>中定义:
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
enum bh_state_bits { BH_Uptodate, /* Contains valid data 包含可用数据*/ BH_Dirty, /* Is dirty 脏数据*/ BH_Lock, /* Is locked */ BH_Req, /* Has been submitted for I/O 有IO操作请求*/ BH_Uptodate_Lock,/* Used by the first bh in a page, to serialise * IO completion of other buffers in the page 被IO操作占用 */ BH_Mapped, /* Has a disk mapping 映射了磁盘块*/ BH_New, /* Disk mapping was newly created by get_block 缓冲区用get_block刚刚获取的,还不能使用*/ BH_Async_Read, /* Is under end_buffer_async_read I/O 正在被异步读*/ BH_Async_Write, /* Is under end_buffer_async_write I/O 正在被异步写*/ BH_Delay, /* Buffer is not yet allocated on disk 尚未与磁盘关联*/ BH_Boundary, /* Block is followed by a discontiguity 缓冲区在连续块边缘*/ BH_Write_EIO, /* I/O error on write 写入时遇到错误*/ BH_Unwritten, /* Buffer is allocated on disk but not written */ BH_Quiet, /* Buffer Error Prinks to be quiet 禁止错误*/ BH_Meta, /* Buffer contains metadata */ BH_Prio, /* Buffer should be submitted with REQ_PRIO */ BH_Defer_Completion, /* Defer AIO completion to workqueue */ BH_PrivateStart,/* not a state bit, but the first bit available * for private allocation by other entities */ };
bh_state_bits还有一个特殊标志,BH_PrivateStart,该标志不是可用状态标志。
三、bio结构体
内核中块I/O操作基本容器由bio结构表示,定义在文件<linux/bio.h>中。
该结构体代表了正在现场的(活动的)以片段(segment)链表形式组织的块I/O操作。
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
/* * main unit of I/O for the block layer and lower layers (ie drivers and * stacking drivers) */ struct bio { struct bio *bi_next; /* request queue link 请求链表*/ struct block_device *bi_bdev; /* 相关的块设备 */ unsigned int bi_flags; /* status, command, etc 状态和命令标志 */ int bi_error; unsigned long bi_rw; /* bottom bits READ/WRITE, * top bits priority 读还是写 */ struct bvec_iter bi_iter; /* Number of segments in this BIO after * physical address coalescing is performed. 结合后的片段数目 */ unsigned int bi_phys_segments; /* * To keep track of the max segment size, we account for the * sizes of the first and last mergeable segments in this bio. 第一个可合并的段大小,最后一个可合并的段大小 */ unsigned int bi_seg_front_size; unsigned int bi_seg_back_size; atomic_t __bi_remaining; bio_end_io_t *bi_end_io; /* bi io完成方法 */ void *bi_private; /* 拥有者的私有方法 */ #ifdef CONFIG_BLK_CGROUP /* * Optional ioc and css associated with this bio. Put on bio * release. Read comment on top of bio_associate_current(). */ struct io_context *bi_ioc; struct cgroup_subsys_state *bi_css; #endif union { #if defined(CONFIG_BLK_DEV_INTEGRITY) struct bio_integrity_payload *bi_integrity; /* data integrity */ #endif }; unsigned short bi_vcnt; /* how many bio_vec's bio链表个数*/ /* * Everything starting with bi_max_vecs will be preserved by bio_reset() */ unsigned short bi_max_vecs; /* max bvl_vecs we can hold */ atomic_t __bi_cnt; /* pin count */ struct bio_vec *bi_io_vec; /* the actual vec list bio链表*/ struct bio_set *bi_pool; /* * We can inline a number of vecs at the end of the bio, to avoid * double allocations for a small number of bio_vecs. This member * MUST obviously be kept at the very end of the bio.内嵌bio向量 */ struct bio_vec bi_inline_vecs[0]; };
bi_io_vec是bio_vec的链表指针,共有bi_vcnt个,当前指向bi_idx位置。bi_vcnt域用来描述bi_io_vec所指向的vio_vec数组中的向量数目。
3.1 I/O向量
bi_io_vec域指向一个bio_vec结构体数组,该结构体链表包含一个特定I/O操作所需要使用的所有片段。
每个bio_vec结构都是一个形式为<page, offset, len>的向量,它描述的是一个特定的片段。片段所在的物理页、块在物理页中的偏移位置、从给定偏移量开始的块长度。
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
/* * was unsigned short, but we might as well be ready for > 64kB I/O pages */ struct bio_vec { /* 指向这个缓冲区所驻留的物理页 */ struct page *bv_page; /* 这个缓冲区以字节为单位的大小 */ unsigned int bv_len; /* 缓冲区所驻留的页中以字节为单位的偏移量 */ unsigned int bv_offset; };
总而言之,每一个I/O请求都通过一个bio结构体表示。每个请求包含一个或多个块,这些块存储在bio_vec结构体数组中。
bi_cnt域记录bio结构体的使用计数,如果该值减为0,就撤销该bio结构体。并释放占用的内存。
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
void bio_get(struct bio *bio) void bio_put(struct bio *bio)
bi_private域,是一个私有域。只有创建了bio结构的拥有者可以读写。
利用bio结构体代替buffer_head结构体有以下好处:
- bio结构体很容易处理高端内存,因为它处理的是物理页而不是直接指针。
- bio结构体既可以代表普通页I/O,同时也可以代表直接I/O
- bio结构体便于执行分散--集中块I/O操作,操作中的数据可取自多个物理页面。
- bio结构体相比缓冲区头属于轻量级的结构体。因为它只需要包含块I/O操作所需的信息就行了,不用包含与缓冲区本身相关的不必要信息。
四、请求队列
块设备将他们挂起在块I/O请求保存在请求队列中,该队列有reques_queue结构体表示,定义在文件<linux/blkdev.h>中,包含要给双向请求链表以及相关控制信息。
队列中的请求由request表示,定义在<linux/blkdev.h>中。
bio使用例程:
![](https://images.cnblogs.com/OutliningIndicators/ContractedBlock.gif)
#include <linux/module.h> #include <linux/init.h> #include <linux/genhd.h> #include <linux/string.h> #include <uapi/linux/major.h> #include <linux/blkdev.h> #define SIMP_BLKDEV_DISK_NAME "simp_blkdev" #define SIMP_BLKDEV_DEVICEMAJOR COMPAQ_SMART2_MAJOR7; #define SIMP_BLKDEV_BYTES (16*1024*1024) static struct gendisk *simp_blkdev_disk; static struct request_queue *simp_blkdev_queue; unsigned char simp_blkdev_data[SIMP_BLKDEV_BYTES]; static void simp_blkdev_do_request(struct request_queue *q); struct block_device_operations simp_blkdev_fops = { .owner = THIS_MODULE, }; static void simp_blkdev_do_request(struct request_queue *q) { struct request *req; struct bio *req_bio; struct bio_vec *bvec; char *disk_mem; char *buffer; int i = 0; //get request from queue while((req = blk_fetch_request(q))!=NULL) { //request ok? if((blk_rq_pos(req) << 9) + blk_rq_cur_bytes(req) > SIMP_BLKDEV_BYTES) { printk(KERN_ERR SIMP_BLKDEV_DISK_NAME ":bad request:block=%llu,count=%u ", (unsigned long long)blk_rq_pos(req), blk_rq_cur_bytes(req)); blk_end_request_all(req, -EIO); continue; } //get memory position disk_mem = simp_blkdev_data + (blk_rq_pos(req) << 9); req_bio = req->bio; switch(rq_data_dir(req)) { case READ: while(req_bio != NULL) { for(i=0;i<req_bio->bi_vcnt;i++) { bvec = &(req_bio->bi_io_vec[i]); buffer = kmap(bvec->bv_page) + bvec->bv_offset; memcpy(buffer, disk_mem, bvec->bv_len); kunmap(bvec->bv_page); disk_mem += bvec->bv_len; } req_bio = req_bio->bi_next; } __blk_end_request_all(req, 0); break; case WRITE: while(req_bio != NULL) { for(i=0;i<req_bio->bi_vcnt;i++) { bvec = &(req_bio->bi_io_vec[i]); buffer = kmap(bvec->bv_page) + bvec->bv_offset; memcpy(disk_mem, buffer, bvec->bv_len); kunmap(bvec->bv_page); disk_mem += bvec->bv_len; } req_bio = req_bio->bi_next; } __blk_end_request_all(req, 0); break; default: break; } } } static int __init init_base(void) { int ret; printk("----Hello World---- "); /* init queue */ simp_blkdev_queue = blk_init_queue(simp_blkdev_do_request, NULL); if(!simp_blkdev_queue) { ret = -ENOMEM; goto err_init_queue; } /* init gendisk */ simp_blkdev_disk = alloc_disk(1); if(!simp_blkdev_disk) { ret = -ENOMEM; goto err_alloc_disk; } strcpy(simp_blkdev_disk->disk_name, SIMP_BLKDEV_DISK_NAME); simp_blkdev_disk->major = SIMP_BLKDEV_DEVICEMAJOR; simp_blkdev_disk->first_minor = 0; simp_blkdev_disk->fops = &simp_blkdev_fops; simp_blkdev_disk->queue = simp_blkdev_queue; set_capacity(simp_blkdev_disk, SIMP_BLKDEV_BYTES>>9); add_disk(simp_blkdev_disk); return 0; err_alloc_disk: blk_cleanup_queue(simp_blkdev_queue); err_init_queue: return ret; } static void __exit exit_base(void) { blk_cleanup_queue(simp_blkdev_queue); put_disk(simp_blkdev_disk); del_gendisk(simp_blkdev_disk); printk("----exit ---- "); } module_init(init_base); module_exit(exit_base); MODULE_LICENSE("GPL"); MODULE_AUTHOR("chen"); MODULE_DESCRIPTION("bio test");
参考资料:https://lwn.net/Articles/736534/
https://www.xuebuyuan.com/3227084.html
https://blog.csdn.net/cxy_chen/article/details/80998510
http://bbs.chinaunix.net/thread-2017377-1-1.html