很多情況下開發(fā)者調測程序需要在Linux下獲取具體的IO的狀況，目前常用的IO觀察工具用vmstat和iostat，具體功能上說當然是iostat更勝一籌，在IO統(tǒng)計上時間點上更具體精細。但二者都是在全局上看到IO，宏觀上的數(shù)據(jù)對于判斷IO到哪個文件上毫無幫助，這個時候block_dump的作用就顯現(xiàn)出來了。
需要先停掉syslog功能，因為具體IO數(shù)據(jù)要通過printk輸出，如果syslog存在，則會往message產(chǎn)生大量IO，干擾正常結果
1 2 | suse:~ # service syslog stopShutting down syslog services done |
然后啟動block_dump
1 | suse:~ # echo 1 > /proc/sys/vm/block_dump |
先說效果：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | suse:~ # dmesg | taildmesg(3414): dirtied inode 9594 (LC_MONETARY) on sda1dmesg(3414): dirtied inode 9238 (LC_COLLATE) on sda1dmesg(3414): dirtied inode 9241 (LC_TIME) on sda1dmesg(3414): dirtied inode 9606 (LC_NUMERIC) on sda1dmesg(3414): dirtied inode 9350 (LC_CTYPE) on sda1kjournald(506): WRITE block 3683672 on sda1kjournald(506): WRITE block 3683680 on sda1kjournald(506): WRITE block 3683688 on sda1kjournald(506): WRITE block 3683696 on sda1kjournald(506): WRITE block 3683704 on sda1kjournald(506): WRITE block 3683712 on sda1kjournald(506): WRITE block 3683720 on sda1kjournald(506): WRITE block 3683728 on sda1kjournald(506): WRITE block 3683736 on sda1kjournald(506): WRITE block 3683744 on sda1 |
通過dmesg信息可以看到IO正在寫那些文件，有進程號，inode號，文件名和磁盤設備名；但每個文件寫了多少呢，僅僅通過dirtied inode就看不出來了，還需要分析WRITE block，后面的數(shù)字并不是真正的塊號，而是內(nèi)核IO層獲取的扇區(qū)號，除以8即為塊號，然后根據(jù)debugfs工具的icheck和ncheck選項，就可以獲取該文件系統(tǒng)塊屬于哪個具體文件，具體請google之。
block_dump的原理其實很簡單，內(nèi)核在IO層根據(jù)標志block_dump在IO提交給磁盤的關口卡主過關的每一個BIO，將它們的數(shù)據(jù)打出來：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 | void submit_bio(int rw, struct bio *bio){ int count = bio_sectors(bio); bio->bi_rw |= rw;/* * If it's a regular read/write or a barrier with data attached, * go through the normal accounting stuff before submission. */ if (bio_has_data(bio) && !(rw & REQ_DISCARD)) { if (rw & WRITE) { count_vm_events(PGPGOUT, count); } else { task_io_account_read(bio->bi_size); count_vm_events(PGPGIN, count); } if (unlikely(block_dump)) { char b[BDEVNAME_SIZE]; printk(KERN_DEBUG "%s(%d): %s block %Lu on %s (%u sectors)n", current->comm, task_pid_nr(current), (rw & WRITE) ? "WRITE" : "READ", (unsigned long long)bio->bi_sector, bdevname(bio->bi_bdev, b), count); } } generic_make_request(bio);} |
具體WRITE block塊號和文件系統(tǒng)塊號之間的對應關系在submit_bh函數(shù)中決定
1 | bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9); |
inode的block_dump實現(xiàn)是通過block_dump___mark_inode_dirty搞定的，這次把關口架在inode臟數(shù)據(jù)寫回的路上，把每個過關的inode信息打出來：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 | void __mark_inode_dirty(struct inode *inode, int flags){if (unlikely(block_dump))block_dump___mark_inode_dirty(inode);}static noinline void block_dump___mark_inode_dirty(struct inode *inode){if (inode->i_ino || strcmp(inode->i_sb->s_id, "bdev")) {struct dentry *dentry;const char *name = "?";dentry = d_find_alias(inode);if (dentry) {spin_lock(&dentry->d_lock);name = (const char *) dentry->d_name.name;}printk(KERN_DEBUG"%s(%d): dirtied inode %lu (%s) on %sn",current->comm, task_pid_nr(current), inode->i_ino,name, inode->i_sb->s_id);if (dentry) {spin_unlock(&dentry->d_lock);dput(dentry);} } |
1.內(nèi)核由很多合適的關口來截獲獲取的IO信息，不改動內(nèi)核，也可以用jprobe搶劫很多東西。
2.debugfs在大量的block–>file轉換過程總太慢，自己用ext2fs寫一個，效率應該能提高很多。
