[Linux] 技术干货 | 漫游Linux块IO

SYSCALL_DEFINE3(write, unsigned int, fd, const char __user *, buf,
size_t, count)
{
struct fd f = fdget_pos(fd);
ssize_t ret = -EBADF;
if (f.file) {
loff_t pos = file_pos_read(f.file);
ret = vfs_write(f.file, buf, count, &pos);
if (ret >= 0)
file_pos_write(f.file, pos);
fdput_pos(f);
}
return ret;
}

const struct file_operations def_blk_fops = {
.open= blkdev_open,
  ... ...
.read= do_sync_read,
.write= do_sync_write,
  ... ...
};

const struct file_operations_extend  ext4_file_operations = {
.kabi_fops = {
  ... ...
.read= do_sync_read,
.write= do_sync_write,
  ... ...
.open= ext4_file_open,
  ... ...
};

ssize_t __generic_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
 unsigned long nr_segs, loff_t *ppos)
  ... ...
if (io_is_direct(file)) {
  ... ...
written = generic_file_direct_write(iocb, iov, &nr_segs, pos,
ppos, count, ocount);
  ... ...
} else {
written = generic_file_buffered_write(iocb, iov, nr_segs,
pos, ppos, count, written);
}
  ... ...

do_blockdev_direct_IO 
-> dio_bio_submit 
-> submit_bio

• bio是Linux通用块层和底层驱动的IO基本单位，可以看到它的最重要的几个属性，一个bio就可以表示一个完整的IO操作：
  
• 
  
  

struct bio {
sector_tbi_sector; //io的起始扇区
... ...
struct block_device*bi_bdev;//对应的块设备
... ...
bio_end_io_t*bi_end_io;//io结束的回调函数
... ...
struct bio_vec*bi_io_vec;//内存page列表
... ...
};

• request代表一个独立的IO请求，是通用块层和驱动层进行IO传递的结构，它容纳了一组连续的bio。通用块层提供了很多IO调度策略，将多个bio合并生成一个request，以提高IO的效率。
  
  

struct request_queue {
... ...
struct elevator_queue*elevator;//调度器
request_fn_proc*request_fn;//请求处理函数
make_request_fn*make_request_fn;//请求入队函数
... ...
softirq_done_fn*softirq_done_fn;//软中断处理
struct device*dev;
unsigned longnr_requests;
... ...
};

void blk_queue_bio(struct request_queue *q, struct bio *bio)
{
  ... ...
el_ret = elv_merge(q, &req, bio);//尝试将bio合并到已有的request中
... ...
req = get_request(q, rw_flags, bio, 0);//无法合并，申请新的request
  ... ...
init_request_from_bio(req, bio);
}
void blk_flush_plug_list(struct blk_plug *plug, bool from_schedule)
{
  ... ...
__elv_add_request(q, rq, ELEVATOR_INSERT_SORT_MERGE);//将request插入request_queue的elevator调度器
  ... ...
}

#/sys/block/sdx/queue
scheduler#调度器配置
nr_requests#队列深度
max_sectors_kb#最大IO大小

static void scsi_request_fn(struct request_queue *q)
{
for (;;) {
... ...
req = blk_peek_request(q);//从request_queue中取出request
... ...
    cmd->scsi_done = scsi_done;//指定cmd完成后回调
rtn = scsi_dispatch_cmd(cmd);//下发将request对应的scsi_cmd
... ...
}
}
int scsi_dispatch_cmd(struct scsi_cmnd *cmd)
{
... ...
rtn = host->hostt->queuecommand(host, cmd);
  ... ...
}

struct request_queue *scsi_alloc_queue(struct scsi_device *sdev)
{
  ... ...
q = __scsi_alloc_queue(sdev->host, scsi_request_fn);
  ... ...
blk_queue_softirq_done(q, scsi_softirq_done);
... ...
}

void __blk_complete_request(struct request *req)
{
... ...
raise_softirq_irqoff(BLOCK_SOFTIRQ);
... ...
}

-> scsi_finish_command
  -> scsi_io_completion
  -> scsi_end_request
  -> blk_update_request
  -> req_bio_endio
  -> bio_endio