Finally we can kill the 'sync_supers' kernel thread along with the '->write_super()' superblock operation because all the users are gone. Now every file-system is supposed to self-manage own superblock and its dirty state. The nice thing about killing this thread is that it improves power management. Indeed, 'sync_supers' is a source of monotonic system wake-ups - it woke up every 5 seconds no matter what - even if there were no dirty superblocks and even if there were no file-systems using this service (e.g., btrfs and journalled ext4 do not need it). So it was wasting power most of the time. And because the thread was in the core of the kernel, all systems had to have it. So I am quite happy to make it go away. Interestingly, this thread is a left-over from the pdflush kernel thread which was a self-forking kernel thread responsible for all the write-back in old Linux kernels. It was turned into per-block device BDI threads, and 'sync_supers' was a left-over. Thus, R.I.P, pdflush as well.
Unified · Split
| @@ -536,46 +536,6 @@ void drop_super(struct super_block *sb) · 535 unmodified lines | ||
| 536 | 536 | |
| 537 | 537 | EXPORT_SYMBOL(drop_super); |
| 538 | 538 | |
| 539 | −/** | |
| 540 | − * sync_supers - helper for periodic superblock writeback | |
| 541 | − * | |
| 542 | − * Call the write_super method if present on all dirty superblocks in | |
| 543 | − * the system. This is for the periodic writeback used by most older | |
| 544 | − * filesystems. For data integrity superblock writeback use | |
| 545 | − * sync_filesystems() instead. | |
| 546 | − * | |
| 547 | − * Note: check the dirty flag before waiting, so we don't | |
| 548 | − * hold up the sync while mounting a device. (The newly | |
| 549 | − * mounted device won't need syncing.) | |
| 550 | − */ | |
| 551 | −void sync_supers(void) | |
| 552 | −{ | |
| 553 | − struct super_block *sb, *p = NULL; | |
| 554 | − | |
| 555 | − spin_lock(&sb_lock); | |
| 556 | − list_for_each_entry(sb, &super_blocks, s_list) { | |
| 557 | − if (hlist_unhashed(&sb->s_instances)) | |
| 558 | − continue; | |
| 559 | − if (sb->s_op->write_super && sb->s_dirt) { | |
| 560 | − sb->s_count++; | |
| 561 | − spin_unlock(&sb_lock); | |
| 562 | − | |
| 563 | − down_read(&sb->s_umount); | |
| 564 | − if (sb->s_root && sb->s_dirt && (sb->s_flags & MS_BORN)) | |
| 565 | − sb->s_op->write_super(sb); | |
| 566 | − up_read(&sb->s_umount); | |
| 567 | − | |
| 568 | − spin_lock(&sb_lock); | |
| 569 | − if (p) | |
| 570 | − __put_super(p); | |
| 571 | − p = sb; | |
| 572 | − } | |
| 573 | − } | |
| 574 | − if (p) | |
| 575 | − __put_super(p); | |
| 576 | − spin_unlock(&sb_lock); | |
| 577 | −} | |
| 578 | − | |
| 579 | 539 | /** |
| 580 | 540 | * iterate_supers - call function for all active superblocks |
| 581 | 541 | * @f: function to call |
| @@ -124,7 +124,6 @@ void bdi_start_writeback(struct backing_dev_info *bdi, long nr_pages, · 123 unmodified lines | ||
| 124 | 124 | void bdi_start_background_writeback(struct backing_dev_info *bdi); |
| 125 | 125 | int bdi_writeback_thread(void *data); |
| 126 | 126 | int bdi_has_dirty_io(struct backing_dev_info *bdi); |
| 127 | −void bdi_arm_supers_timer(void); | |
| 128 | 127 | void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi); |
| 129 | 128 | void bdi_lock_two(struct bdi_writeback *wb1, struct bdi_writeback *wb2); |
| 130 | 129 | |
| @@ -1491,7 +1491,6 @@ struct sb_writers { · 1490 unmodified lines | ||
| 1491 | 1491 | struct super_block { |
| 1492 | 1492 | struct list_head s_list; /* Keep this first */ |
| 1493 | 1493 | dev_t s_dev; /* search index; _not_ kdev_t */ |
| 1494 | − unsigned char s_dirt; | |
| 1495 | 1494 | unsigned char s_blocksize_bits; |
| 1496 | 1495 | unsigned long s_blocksize; |
| 1497 | 1496 | loff_t s_maxbytes; /* Max file size */ |
| @@ -1861,7 +1860,6 @@ struct super_operations { · 363 unmodified lines | ||
| 1861 | 1860 | int (*drop_inode) (struct inode *); |
| 1862 | 1861 | void (*evict_inode) (struct inode *); |
| 1863 | 1862 | void (*put_super) (struct super_block *); |
| 1864 | − void (*write_super) (struct super_block *); | |
| 1865 | 1863 | int (*sync_fs)(struct super_block *sb, int wait); |
| 1866 | 1864 | int (*freeze_fs) (struct super_block *); |
| 1867 | 1865 | int (*unfreeze_fs) (struct super_block *); |
| @@ -2397,7 +2395,6 @@ extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end, · 529 unmodified lines | ||
| 2397 | 2395 | int datasync); |
| 2398 | 2396 | extern int vfs_fsync(struct file *file, int datasync); |
| 2399 | 2397 | extern int generic_write_sync(struct file *file, loff_t pos, loff_t count); |
| 2400 | −extern void sync_supers(void); | |
| 2401 | 2398 | extern void emergency_sync(void); |
| 2402 | 2399 | extern void emergency_remount(void); |
| 2403 | 2400 | #ifdef CONFIG_BLOCK |
| @@ -39,12 +39,6 @@ DEFINE_SPINLOCK(bdi_lock); · 38 unmodified lines | ||
| 39 | 39 | LIST_HEAD(bdi_list); |
| 40 | 40 | LIST_HEAD(bdi_pending_list); |
| 41 | 41 | |
| 42 | −static struct task_struct *sync_supers_tsk; | |
| 43 | −static struct timer_list sync_supers_timer; | |
| 44 | − | |
| 45 | −static int bdi_sync_supers(void *); | |
| 46 | −static void sync_supers_timer_fn(unsigned long); | |
| 47 | − | |
| 48 | 42 | void bdi_lock_two(struct bdi_writeback *wb1, struct bdi_writeback *wb2) |
| 49 | 43 | { |
| 50 | 44 | if (wb1 < wb2) { |
| @@ -250,12 +244,6 @@ static int __init default_bdi_init(void) · 199 unmodified lines | ||
| 250 | 244 | { |
| 251 | 245 | int err; |
| 252 | 246 | |
| 253 | − sync_supers_tsk = kthread_run(bdi_sync_supers, NULL, "sync_supers"); | |
| 254 | − BUG_ON(IS_ERR(sync_supers_tsk)); | |
| 255 | − | |
| 256 | − setup_timer(&sync_supers_timer, sync_supers_timer_fn, 0); | |
| 257 | − bdi_arm_supers_timer(); | |
| 258 | − | |
| 259 | 247 | err = bdi_init(&default_backing_dev_info); |
| 260 | 248 | if (!err) |
| 261 | 249 | bdi_register(&default_backing_dev_info, NULL, "default"); |
| @@ -270,46 +258,6 @@ int bdi_has_dirty_io(struct backing_dev_info *bdi) · 8 unmodified lines | ||
| 270 | 258 | return wb_has_dirty_io(&bdi->wb); |
| 271 | 259 | } |
| 272 | 260 | |
| 273 | −/* | |
| 274 | − * kupdated() used to do this. We cannot do it from the bdi_forker_thread() | |
| 275 | − * or we risk deadlocking on ->s_umount. The longer term solution would be | |
| 276 | − * to implement sync_supers_bdi() or similar and simply do it from the | |
| 277 | − * bdi writeback thread individually. | |
| 278 | − */ | |
| 279 | −static int bdi_sync_supers(void *unused) | |
| 280 | −{ | |
| 281 | − set_user_nice(current, 0); | |
| 282 | − | |
| 283 | − while (!kthread_should_stop()) { | |
| 284 | − set_current_state(TASK_INTERRUPTIBLE); | |
| 285 | − schedule(); | |
| 286 | − | |
| 287 | − /* | |
| 288 | − * Do this periodically, like kupdated() did before. | |
| 289 | − */ | |
| 290 | − sync_supers(); | |
| 291 | − } | |
| 292 | − | |
| 293 | − return 0; | |
| 294 | −} | |
| 295 | − | |
| 296 | −void bdi_arm_supers_timer(void) | |
| 297 | −{ | |
| 298 | − unsigned long next; | |
| 299 | − | |
| 300 | − if (!dirty_writeback_interval) | |
| 301 | − return; | |
| 302 | − | |
| 303 | − next = msecs_to_jiffies(dirty_writeback_interval * 10) + jiffies; | |
| 304 | − mod_timer(&sync_supers_timer, round_jiffies_up(next)); | |
| 305 | −} | |
| 306 | − | |
| 307 | −static void sync_supers_timer_fn(unsigned long unused) | |
| 308 | −{ | |
| 309 | − wake_up_process(sync_supers_tsk); | |
| 310 | − bdi_arm_supers_timer(); | |
| 311 | −} | |
| 312 | − | |
| 313 | 261 | static void wakeup_timer_fn(unsigned long data) |
| 314 | 262 | { |
| 315 | 263 | struct backing_dev_info *bdi = (struct backing_dev_info *)data; |
| @@ -1532,7 +1532,6 @@ int dirty_writeback_centisecs_handler(ctl_table *table, int write, · 1531 unmodified lines | ||
| 1532 | 1532 | void __user *buffer, size_t *length, loff_t *ppos) |
| 1533 | 1533 | { |
| 1534 | 1534 | proc_dointvec(table, write, buffer, length, ppos); |
| 1535 | − bdi_arm_supers_timer(); | |
| 1536 | 1535 | return 0; |
| 1537 | 1536 | } |
| 1538 | 1537 | |