linux内核奇遇记之md源代码解读之8阵列同步二

linux内核奇遇记之md源代码解读之八阵列同步二
linux内核奇遇记之md源代码解读之八阵列同步二转载请注明出处：http://blog.csdn.net/liumangxiong
在上一小节里讲到启动同步线程：

7518         wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));7519 7520         /* tell personality that we are finished */7521         mddev->pers->sync_request(mddev, max_sectors, &skipped, 1);7522 7523         if (!test_bit(MD_RECOVERY_CHECK, &mddev->recovery) &&7524             mddev->curr_resync > 2) {7525                 if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {7526                         if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {7527                                 if (mddev->curr_resync >= mddev->recovery_cp) {7528                                         printk(KERN_INFO7529                                                "md: checkpointing %s of %s.\n",7530                                                desc, mdname(mddev));7531                                         mddev->recovery_cp =7532                                                 mddev->curr_resync_completed;7533                                 }7534                         } else7535                                 mddev->recovery_cp = MaxSector;7536                 } else {7537                         if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery))7538                                 mddev->curr_resync = MaxSector;7539                         rcu_read_lock();7540                         rdev_for_each_rcu(rdev, mddev)7541                                 if (rdev->raid_disk >= 0 &&7542                                     mddev->delta_disks >= 0 &&7543                                     !test_bit(Faulty, &rdev->flags) &&7544                                     !test_bit(In_sync, &rdev->flags) &&7545                                     rdev->recovery_offset < mddev->curr_resync)7546                                         rdev->recovery_offset = mddev->curr_resync;7547                         rcu_read_unlock();7548                 }7549         }7550  skip:7551         set_bit(MD_CHANGE_DEVS, &mddev->flags);7552 7553         if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {7554                 /* We completed so min/max setting can be forgotten if used. */7555                 if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))7556                         mddev->resync_min = 0;7557                 mddev->resync_max = MaxSector;7558         } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))7559                 mddev->resync_min = mddev->curr_resync_completed;7560         mddev->curr_resync = 0;7561         wake_up(&resync_wait);7562         set_bit(MD_RECOVERY_DONE, &mddev->recovery);7563         md_wakeup_thread(mddev->thread);7564         return;

7518行，等待所有同步请求返回7521行，根据同步结果更新bitmap，回收资源7525行，如果同步中断则设置recovery_cp为同步完成点，正常完成则设置为MaxSector。7551行，设置改变状态7553-7560行，恢复同步值7561行，唤醒同磁盘分区同步等待线程7562行，设置同步完成标志7563行，唤醒主线程。如果到这里就认为同步完成了，那就大错特错了。记得有一句话讲，每一个阶段的结束就是下一个阶段的起点。看到md_wake_up我们就想到事情又有了一个新起点。在唤醒主线程之后，主线程会调用上一小节中的md_check_recovery来清理现场，最终调用到7774行的reap_sync_thread函数。对于同步来说，这个函数做了以下事情：1）回收同步线程2）更新超级块3）更新mddev标志其实阵列的同步很简单，下一小节讲raid5同步过程sync_request函数。转载请注明出处：http://blog.csdn.net/liumangxiong

查看更多 下一篇