aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
Diffstat (limited to 'fs')
-rw-r--r--fs/btrfs/async-thread.c81
-rw-r--r--fs/btrfs/async-thread.h10
-rw-r--r--fs/btrfs/btrfs_inode.h12
-rw-r--r--fs/btrfs/ctree.h15
-rw-r--r--fs/btrfs/disk-io.c48
-rw-r--r--fs/btrfs/extent-tree.c235
-rw-r--r--fs/btrfs/extent_io.c42
-rw-r--r--fs/btrfs/extent_io.h18
-rw-r--r--fs/btrfs/file.c3
-rw-r--r--fs/btrfs/inode.c111
-rw-r--r--fs/btrfs/ioctl.c7
-rw-r--r--fs/btrfs/ordered-data.c6
-rw-r--r--fs/btrfs/relocation.c4
-rw-r--r--fs/btrfs/tree-log.c12
-rw-r--r--fs/ecryptfs/Kconfig3
-rw-r--r--fs/ecryptfs/main.c7
-rw-r--r--fs/file.c1
-rw-r--r--fs/nfs/client.c2
-rw-r--r--fs/nfs/nfs4namespace.c12
-rw-r--r--fs/nfs/nfs4renewd.c6
-rw-r--r--fs/nfs/super.c36
-rw-r--r--fs/proc/kcore.c1
-rw-r--r--fs/proc/page.c5
-rw-r--r--fs/romfs/storage.c4
-rw-r--r--fs/xfs/linux-2.6/xfs_aops.c38
-rw-r--r--fs/xfs/linux-2.6/xfs_file.c9
-rw-r--r--fs/xfs/linux-2.6/xfs_iops.c41
-rw-r--r--fs/xfs/linux-2.6/xfs_lrw.c2
-rw-r--r--fs/xfs/linux-2.6/xfs_super.c59
-rw-r--r--fs/xfs/linux-2.6/xfs_sync.c36
-rw-r--r--fs/xfs/xfs_dfrag.c8
-rw-r--r--fs/xfs/xfs_dir2_leaf.c4
-rw-r--r--fs/xfs/xfs_inode.c4
-rw-r--r--fs/xfs/xfs_inode.h2
-rw-r--r--fs/xfs/xfs_inode_item.c18
-rw-r--r--fs/xfs/xfs_itable.c21
-rw-r--r--fs/xfs/xfs_vnodeops.c6
37 files changed, 633 insertions, 296 deletions
diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index 282ca085c2fb..c0861e781cdb 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -64,6 +64,51 @@ struct btrfs_worker_thread {
64}; 64};
65 65
66/* 66/*
67 * btrfs_start_workers uses kthread_run, which can block waiting for memory
68 * for a very long time. It will actually throttle on page writeback,
69 * and so it may not make progress until after our btrfs worker threads
70 * process all of the pending work structs in their queue
71 *
72 * This means we can't use btrfs_start_workers from inside a btrfs worker
73 * thread that is used as part of cleaning dirty memory, which pretty much
74 * involves all of the worker threads.
75 *
76 * Instead we have a helper queue who never has more than one thread
77 * where we scheduler thread start operations. This worker_start struct
78 * is used to contain the work and hold a pointer to the queue that needs
79 * another worker.
80 */
81struct worker_start {
82 struct btrfs_work work;
83 struct btrfs_workers *queue;
84};
85
86static void start_new_worker_func(struct btrfs_work *work)
87{
88 struct worker_start *start;
89 start = container_of(work, struct worker_start, work);
90 btrfs_start_workers(start->queue, 1);
91 kfree(start);
92}
93
94static int start_new_worker(struct btrfs_workers *queue)
95{
96 struct worker_start *start;
97 int ret;
98
99 start = kzalloc(sizeof(*start), GFP_NOFS);
100 if (!start)
101 return -ENOMEM;
102
103 start->work.func = start_new_worker_func;
104 start->queue = queue;
105 ret = btrfs_queue_worker(queue->atomic_worker_start, &start->work);
106 if (ret)
107 kfree(start);
108 return ret;
109}
110
111/*
67 * helper function to move a thread onto the idle list after it 112 * helper function to move a thread onto the idle list after it
68 * has finished some requests. 113 * has finished some requests.
69 */ 114 */
@@ -118,11 +163,13 @@ static void check_pending_worker_creates(struct btrfs_worker_thread *worker)
118 goto out; 163 goto out;
119 164
120 workers->atomic_start_pending = 0; 165 workers->atomic_start_pending = 0;
121 if (workers->num_workers >= workers->max_workers) 166 if (workers->num_workers + workers->num_workers_starting >=
167 workers->max_workers)
122 goto out; 168 goto out;
123 169
170 workers->num_workers_starting += 1;
124 spin_unlock_irqrestore(&workers->lock, flags); 171 spin_unlock_irqrestore(&workers->lock, flags);
125 btrfs_start_workers(workers, 1); 172 start_new_worker(workers);
126 return; 173 return;
127 174
128out: 175out:
@@ -390,9 +437,11 @@ int btrfs_stop_workers(struct btrfs_workers *workers)
390/* 437/*
391 * simple init on struct btrfs_workers 438 * simple init on struct btrfs_workers
392 */ 439 */
393void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max) 440void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
441 struct btrfs_workers *async_helper)
394{ 442{
395 workers->num_workers = 0; 443 workers->num_workers = 0;
444 workers->num_workers_starting = 0;
396 INIT_LIST_HEAD(&workers->worker_list); 445 INIT_LIST_HEAD(&workers->worker_list);
397 INIT_LIST_HEAD(&workers->idle_list); 446 INIT_LIST_HEAD(&workers->idle_list);
398 INIT_LIST_HEAD(&workers->order_list); 447 INIT_LIST_HEAD(&workers->order_list);
@@ -404,14 +453,15 @@ void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max)
404 workers->name = name; 453 workers->name = name;
405 workers->ordered = 0; 454 workers->ordered = 0;
406 workers->atomic_start_pending = 0; 455 workers->atomic_start_pending = 0;
407 workers->atomic_worker_start = 0; 456 workers->atomic_worker_start = async_helper;
408} 457}
409 458
410/* 459/*
411 * starts new worker threads. This does not enforce the max worker 460 * starts new worker threads. This does not enforce the max worker
412 * count in case you need to temporarily go past it. 461 * count in case you need to temporarily go past it.
413 */ 462 */
414int btrfs_start_workers(struct btrfs_workers *workers, int num_workers) 463static int __btrfs_start_workers(struct btrfs_workers *workers,
464 int num_workers)
415{ 465{
416 struct btrfs_worker_thread *worker; 466 struct btrfs_worker_thread *worker;
417 int ret = 0; 467 int ret = 0;
@@ -444,6 +494,8 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
444 list_add_tail(&worker->worker_list, &workers->idle_list); 494 list_add_tail(&worker->worker_list, &workers->idle_list);
445 worker->idle = 1; 495 worker->idle = 1;
446 workers->num_workers++; 496 workers->num_workers++;
497 workers->num_workers_starting--;
498 WARN_ON(workers->num_workers_starting < 0);
447 spin_unlock_irq(&workers->lock); 499 spin_unlock_irq(&workers->lock);
448 } 500 }
449 return 0; 501 return 0;
@@ -452,6 +504,14 @@ fail:
452 return ret; 504 return ret;
453} 505}
454 506
507int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
508{
509 spin_lock_irq(&workers->lock);
510 workers->num_workers_starting += num_workers;
511 spin_unlock_irq(&workers->lock);
512 return __btrfs_start_workers(workers, num_workers);
513}
514
455/* 515/*
456 * run through the list and find a worker thread that doesn't have a lot 516 * run through the list and find a worker thread that doesn't have a lot
457 * to do right now. This can return null if we aren't yet at the thread 517 * to do right now. This can return null if we aren't yet at the thread
@@ -461,7 +521,10 @@ static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers)
461{ 521{
462 struct btrfs_worker_thread *worker; 522 struct btrfs_worker_thread *worker;
463 struct list_head *next; 523 struct list_head *next;
464 int enforce_min = workers->num_workers < workers->max_workers; 524 int enforce_min;
525
526 enforce_min = (workers->num_workers + workers->num_workers_starting) <
527 workers->max_workers;
465 528
466 /* 529 /*
467 * if we find an idle thread, don't move it to the end of the 530 * if we find an idle thread, don't move it to the end of the
@@ -509,15 +572,17 @@ again:
509 worker = next_worker(workers); 572 worker = next_worker(workers);
510 573
511 if (!worker) { 574 if (!worker) {
512 if (workers->num_workers >= workers->max_workers) { 575 if (workers->num_workers + workers->num_workers_starting >=
576 workers->max_workers) {
513 goto fallback; 577 goto fallback;
514 } else if (workers->atomic_worker_start) { 578 } else if (workers->atomic_worker_start) {
515 workers->atomic_start_pending = 1; 579 workers->atomic_start_pending = 1;
516 goto fallback; 580 goto fallback;
517 } else { 581 } else {
582 workers->num_workers_starting++;
518 spin_unlock_irqrestore(&workers->lock, flags); 583 spin_unlock_irqrestore(&workers->lock, flags);
519 /* we're below the limit, start another worker */ 584 /* we're below the limit, start another worker */
520 btrfs_start_workers(workers, 1); 585 __btrfs_start_workers(workers, 1);
521 goto again; 586 goto again;
522 } 587 }
523 } 588 }
diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h
index fc089b95ec14..5077746cf85e 100644
--- a/fs/btrfs/async-thread.h
+++ b/fs/btrfs/async-thread.h
@@ -64,6 +64,8 @@ struct btrfs_workers {
64 /* current number of running workers */ 64 /* current number of running workers */
65 int num_workers; 65 int num_workers;
66 66
67 int num_workers_starting;
68
67 /* max number of workers allowed. changed by btrfs_start_workers */ 69 /* max number of workers allowed. changed by btrfs_start_workers */
68 int max_workers; 70 int max_workers;
69 71
@@ -78,9 +80,10 @@ struct btrfs_workers {
78 80
79 /* 81 /*
80 * are we allowed to sleep while starting workers or are we required 82 * are we allowed to sleep while starting workers or are we required
81 * to start them at a later time? 83 * to start them at a later time? If we can't sleep, this indicates
84 * which queue we need to use to schedule thread creation.
82 */ 85 */
83 int atomic_worker_start; 86 struct btrfs_workers *atomic_worker_start;
84 87
85 /* list with all the work threads. The workers on the idle thread 88 /* list with all the work threads. The workers on the idle thread
86 * may be actively servicing jobs, but they haven't yet hit the 89 * may be actively servicing jobs, but they haven't yet hit the
@@ -109,7 +112,8 @@ struct btrfs_workers {
109int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work); 112int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work);
110int btrfs_start_workers(struct btrfs_workers *workers, int num_workers); 113int btrfs_start_workers(struct btrfs_workers *workers, int num_workers);
111int btrfs_stop_workers(struct btrfs_workers *workers); 114int btrfs_stop_workers(struct btrfs_workers *workers);
112void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max); 115void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
116 struct btrfs_workers *async_starter);
113int btrfs_requeue_work(struct btrfs_work *work); 117int btrfs_requeue_work(struct btrfs_work *work);
114void btrfs_set_work_high_prio(struct btrfs_work *work); 118void btrfs_set_work_high_prio(struct btrfs_work *work);
115#endif 119#endif
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index a54d354cefcb..c71abec0ab90 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -128,12 +128,14 @@ struct btrfs_inode {
128 u64 last_unlink_trans; 128 u64 last_unlink_trans;
129 129
130 /* 130 /*
131 * These two counters are for delalloc metadata reservations. We keep 131 * Counters to keep track of the number of extent item's we may use due
132 * track of how many extents we've accounted for vs how many extents we 132 * to delalloc and such. outstanding_extents is the number of extent
133 * have. 133 * items we think we'll end up using, and reserved_extents is the number
134 * of extent items we've reserved metadata for.
134 */ 135 */
135 int delalloc_reserved_extents; 136 spinlock_t accounting_lock;
136 int delalloc_extents; 137 int reserved_extents;
138 int outstanding_extents;
137 139
138 /* 140 /*
139 * ordered_data_close is set by truncate when a file that used 141 * ordered_data_close is set by truncate when a file that used
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index dd8ced9814c4..1bb897ecdeeb 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -691,14 +691,17 @@ struct btrfs_space_info {
691 691
692 struct list_head list; 692 struct list_head list;
693 693
694 /* for controlling how we free up space for allocations */
695 wait_queue_head_t allocate_wait;
696 wait_queue_head_t flush_wait;
697 int allocating_chunk;
698 int flushing;
699
694 /* for block groups in our same type */ 700 /* for block groups in our same type */
695 struct list_head block_groups; 701 struct list_head block_groups;
696 spinlock_t lock; 702 spinlock_t lock;
697 struct rw_semaphore groups_sem; 703 struct rw_semaphore groups_sem;
698 atomic_t caching_threads; 704 atomic_t caching_threads;
699
700 int allocating_chunk;
701 wait_queue_head_t wait;
702}; 705};
703 706
704/* 707/*
@@ -907,6 +910,7 @@ struct btrfs_fs_info {
907 * A third pool does submit_bio to avoid deadlocking with the other 910 * A third pool does submit_bio to avoid deadlocking with the other
908 * two 911 * two
909 */ 912 */
913 struct btrfs_workers generic_worker;
910 struct btrfs_workers workers; 914 struct btrfs_workers workers;
911 struct btrfs_workers delalloc_workers; 915 struct btrfs_workers delalloc_workers;
912 struct btrfs_workers endio_workers; 916 struct btrfs_workers endio_workers;
@@ -914,6 +918,7 @@ struct btrfs_fs_info {
914 struct btrfs_workers endio_meta_write_workers; 918 struct btrfs_workers endio_meta_write_workers;
915 struct btrfs_workers endio_write_workers; 919 struct btrfs_workers endio_write_workers;
916 struct btrfs_workers submit_workers; 920 struct btrfs_workers submit_workers;
921 struct btrfs_workers enospc_workers;
917 /* 922 /*
918 * fixup workers take dirty pages that didn't properly go through 923 * fixup workers take dirty pages that didn't properly go through
919 * the cow mechanism and make them safe to write. It happens 924 * the cow mechanism and make them safe to write. It happens
@@ -1005,6 +1010,8 @@ struct btrfs_root {
1005 atomic_t log_commit[2]; 1010 atomic_t log_commit[2];
1006 unsigned long log_transid; 1011 unsigned long log_transid;
1007 unsigned long log_batch; 1012 unsigned long log_batch;
1013 pid_t log_start_pid;
1014 bool log_multiple_pids;
1008 1015
1009 u64 objectid; 1016 u64 objectid;
1010 u64 last_trans; 1017 u64 last_trans;
@@ -2323,7 +2330,7 @@ int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode);
2323void btrfs_orphan_cleanup(struct btrfs_root *root); 2330void btrfs_orphan_cleanup(struct btrfs_root *root);
2324int btrfs_cont_expand(struct inode *inode, loff_t size); 2331int btrfs_cont_expand(struct inode *inode, loff_t size);
2325int btrfs_invalidate_inodes(struct btrfs_root *root); 2332int btrfs_invalidate_inodes(struct btrfs_root *root);
2326extern struct dentry_operations btrfs_dentry_operations; 2333extern const struct dentry_operations btrfs_dentry_operations;
2327 2334
2328/* ioctl.c */ 2335/* ioctl.c */
2329long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 2336long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index af0435f79fa6..100551a66c46 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1746,21 +1746,25 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1746 err = -EINVAL; 1746 err = -EINVAL;
1747 goto fail_iput; 1747 goto fail_iput;
1748 } 1748 }
1749printk("thread pool is %d\n", fs_info->thread_pool_size); 1749
1750 /* 1750 btrfs_init_workers(&fs_info->generic_worker,
1751 * we need to start all the end_io workers up front because the 1751 "genwork", 1, NULL);
1752 * queue work function gets called at interrupt time, and so it 1752
1753 * cannot dynamically grow.
1754 */
1755 btrfs_init_workers(&fs_info->workers, "worker", 1753 btrfs_init_workers(&fs_info->workers, "worker",
1756 fs_info->thread_pool_size); 1754 fs_info->thread_pool_size,
1755 &fs_info->generic_worker);
1757 1756
1758 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", 1757 btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
1759 fs_info->thread_pool_size); 1758 fs_info->thread_pool_size,
1759 &fs_info->generic_worker);
1760 1760
1761 btrfs_init_workers(&fs_info->submit_workers, "submit", 1761 btrfs_init_workers(&fs_info->submit_workers, "submit",
1762 min_t(u64, fs_devices->num_devices, 1762 min_t(u64, fs_devices->num_devices,
1763 fs_info->thread_pool_size)); 1763 fs_info->thread_pool_size),
1764 &fs_info->generic_worker);
1765 btrfs_init_workers(&fs_info->enospc_workers, "enospc",
1766 fs_info->thread_pool_size,
1767 &fs_info->generic_worker);
1764 1768
1765 /* a higher idle thresh on the submit workers makes it much more 1769 /* a higher idle thresh on the submit workers makes it much more
1766 * likely that bios will be send down in a sane order to the 1770 * likely that bios will be send down in a sane order to the
@@ -1774,15 +1778,20 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
1774 fs_info->delalloc_workers.idle_thresh = 2; 1778 fs_info->delalloc_workers.idle_thresh = 2;
1775 fs_info->delalloc_workers.ordered = 1; 1779 fs_info->delalloc_workers.ordered = 1;
1776 1780
1777 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); 1781 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
1782 &fs_info->generic_worker);
1778 btrfs_init_workers(&fs_info->endio_workers, "endio", 1783 btrfs_init_workers(&fs_info->endio_workers, "endio",
1779 fs_info->thread_pool_size); 1784 fs_info->thread_pool_size,
1785 &fs_info->generic_worker);
1780 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta", 1786 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
1781 fs_info->thread_pool_size); 1787 fs_info->thread_pool_size,
1788 &fs_info->generic_worker);
1782 btrfs_init_workers(&fs_info->endio_meta_write_workers, 1789 btrfs_init_workers(&fs_info->endio_meta_write_workers,
1783 "endio-meta-write", fs_info->thread_pool_size); 1790 "endio-meta-write", fs_info->thread_pool_size,
1791 &fs_info->generic_worker);
1784 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", 1792 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
1785 fs_info->thread_pool_size); 1793 fs_info->thread_pool_size,
1794 &fs_info->generic_worker);
1786 1795
1787 /* 1796 /*
1788 * endios are largely parallel and should have a very 1797 * endios are largely parallel and should have a very
@@ -1794,12 +1803,8 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
1794 fs_info->endio_write_workers.idle_thresh = 2; 1803 fs_info->endio_write_workers.idle_thresh = 2;
1795 fs_info->endio_meta_write_workers.idle_thresh = 2; 1804 fs_info->endio_meta_write_workers.idle_thresh = 2;
1796 1805
1797 fs_info->endio_workers.atomic_worker_start = 1;
1798 fs_info->endio_meta_workers.atomic_worker_start = 1;
1799 fs_info->endio_write_workers.atomic_worker_start = 1;
1800 fs_info->endio_meta_write_workers.atomic_worker_start = 1;
1801
1802 btrfs_start_workers(&fs_info->workers, 1); 1806 btrfs_start_workers(&fs_info->workers, 1);
1807 btrfs_start_workers(&fs_info->generic_worker, 1);
1803 btrfs_start_workers(&fs_info->submit_workers, 1); 1808 btrfs_start_workers(&fs_info->submit_workers, 1);
1804 btrfs_start_workers(&fs_info->delalloc_workers, 1); 1809 btrfs_start_workers(&fs_info->delalloc_workers, 1);
1805 btrfs_start_workers(&fs_info->fixup_workers, 1); 1810 btrfs_start_workers(&fs_info->fixup_workers, 1);
@@ -1807,6 +1812,7 @@ printk("thread pool is %d\n", fs_info->thread_pool_size);
1807 btrfs_start_workers(&fs_info->endio_meta_workers, 1); 1812 btrfs_start_workers(&fs_info->endio_meta_workers, 1);
1808 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1); 1813 btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
1809 btrfs_start_workers(&fs_info->endio_write_workers, 1); 1814 btrfs_start_workers(&fs_info->endio_write_workers, 1);
1815 btrfs_start_workers(&fs_info->enospc_workers, 1);
1810 1816
1811 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 1817 fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
1812 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 1818 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
@@ -2012,6 +2018,7 @@ fail_chunk_root:
2012 free_extent_buffer(chunk_root->node); 2018 free_extent_buffer(chunk_root->node);
2013 free_extent_buffer(chunk_root->commit_root); 2019 free_extent_buffer(chunk_root->commit_root);
2014fail_sb_buffer: 2020fail_sb_buffer:
2021 btrfs_stop_workers(&fs_info->generic_worker);
2015 btrfs_stop_workers(&fs_info->fixup_workers); 2022 btrfs_stop_workers(&fs_info->fixup_workers);
2016 btrfs_stop_workers(&fs_info->delalloc_workers); 2023 btrfs_stop_workers(&fs_info->delalloc_workers);
2017 btrfs_stop_workers(&fs_info->workers); 2024 btrfs_stop_workers(&fs_info->workers);
@@ -2020,6 +2027,7 @@ fail_sb_buffer:
2020 btrfs_stop_workers(&fs_info->endio_meta_write_workers); 2027 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2021 btrfs_stop_workers(&fs_info->endio_write_workers); 2028 btrfs_stop_workers(&fs_info->endio_write_workers);
2022 btrfs_stop_workers(&fs_info->submit_workers); 2029 btrfs_stop_workers(&fs_info->submit_workers);
2030 btrfs_stop_workers(&fs_info->enospc_workers);
2023fail_iput: 2031fail_iput:
2024 invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 2032 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
2025 iput(fs_info->btree_inode); 2033 iput(fs_info->btree_inode);
@@ -2437,6 +2445,7 @@ int close_ctree(struct btrfs_root *root)
2437 2445
2438 iput(fs_info->btree_inode); 2446 iput(fs_info->btree_inode);
2439 2447
2448 btrfs_stop_workers(&fs_info->generic_worker);
2440 btrfs_stop_workers(&fs_info->fixup_workers); 2449 btrfs_stop_workers(&fs_info->fixup_workers);
2441 btrfs_stop_workers(&fs_info->delalloc_workers); 2450 btrfs_stop_workers(&fs_info->delalloc_workers);
2442 btrfs_stop_workers(&fs_info->workers); 2451 btrfs_stop_workers(&fs_info->workers);
@@ -2445,6 +2454,7 @@ int close_ctree(struct btrfs_root *root)
2445 btrfs_stop_workers(&fs_info->endio_meta_write_workers); 2454 btrfs_stop_workers(&fs_info->endio_meta_write_workers);
2446 btrfs_stop_workers(&fs_info->endio_write_workers); 2455 btrfs_stop_workers(&fs_info->endio_write_workers);
2447 btrfs_stop_workers(&fs_info->submit_workers); 2456 btrfs_stop_workers(&fs_info->submit_workers);
2457 btrfs_stop_workers(&fs_info->enospc_workers);
2448 2458
2449 btrfs_close_devices(fs_info->fs_devices); 2459 btrfs_close_devices(fs_info->fs_devices);
2450 btrfs_mapping_tree_free(&fs_info->mapping_tree); 2460 btrfs_mapping_tree_free(&fs_info->mapping_tree);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 359a754c782c..d0c4d584efad 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2824,14 +2824,17 @@ int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
2824 num_items); 2824 num_items);
2825 2825
2826 spin_lock(&meta_sinfo->lock); 2826 spin_lock(&meta_sinfo->lock);
2827 if (BTRFS_I(inode)->delalloc_reserved_extents <= 2827 spin_lock(&BTRFS_I(inode)->accounting_lock);
2828 BTRFS_I(inode)->delalloc_extents) { 2828 if (BTRFS_I(inode)->reserved_extents <=
2829 BTRFS_I(inode)->outstanding_extents) {
2830 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2829 spin_unlock(&meta_sinfo->lock); 2831 spin_unlock(&meta_sinfo->lock);
2830 return 0; 2832 return 0;
2831 } 2833 }
2834 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2832 2835
2833 BTRFS_I(inode)->delalloc_reserved_extents--; 2836 BTRFS_I(inode)->reserved_extents--;
2834 BUG_ON(BTRFS_I(inode)->delalloc_reserved_extents < 0); 2837 BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
2835 2838
2836 if (meta_sinfo->bytes_delalloc < num_bytes) { 2839 if (meta_sinfo->bytes_delalloc < num_bytes) {
2837 bug = true; 2840 bug = true;
@@ -2864,6 +2867,107 @@ static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
2864 meta_sinfo->force_delalloc = 0; 2867 meta_sinfo->force_delalloc = 0;
2865} 2868}
2866 2869
2870struct async_flush {
2871 struct btrfs_root *root;
2872 struct btrfs_space_info *info;
2873 struct btrfs_work work;
2874};
2875
2876static noinline void flush_delalloc_async(struct btrfs_work *work)
2877{
2878 struct async_flush *async;
2879 struct btrfs_root *root;
2880 struct btrfs_space_info *info;
2881
2882 async = container_of(work, struct async_flush, work);
2883 root = async->root;
2884 info = async->info;
2885
2886 btrfs_start_delalloc_inodes(root);
2887 wake_up(&info->flush_wait);
2888 btrfs_wait_ordered_extents(root, 0);
2889
2890 spin_lock(&info->lock);
2891 info->flushing = 0;
2892 spin_unlock(&info->lock);
2893 wake_up(&info->flush_wait);
2894
2895 kfree(async);
2896}
2897
2898static void wait_on_flush(struct btrfs_space_info *info)
2899{
2900 DEFINE_WAIT(wait);
2901 u64 used;
2902
2903 while (1) {
2904 prepare_to_wait(&info->flush_wait, &wait,
2905 TASK_UNINTERRUPTIBLE);
2906 spin_lock(&info->lock);
2907 if (!info->flushing) {
2908 spin_unlock(&info->lock);
2909 break;
2910 }
2911
2912 used = info->bytes_used + info->bytes_reserved +
2913 info->bytes_pinned + info->bytes_readonly +
2914 info->bytes_super + info->bytes_root +
2915 info->bytes_may_use + info->bytes_delalloc;
2916 if (used < info->total_bytes) {
2917 spin_unlock(&info->lock);
2918 break;
2919 }
2920 spin_unlock(&info->lock);
2921 schedule();
2922 }
2923 finish_wait(&info->flush_wait, &wait);
2924}
2925
2926static void flush_delalloc(struct btrfs_root *root,
2927 struct btrfs_space_info *info)
2928{
2929 struct async_flush *async;
2930 bool wait = false;
2931
2932 spin_lock(&info->lock);
2933
2934 if (!info->flushing) {
2935 info->flushing = 1;
2936 init_waitqueue_head(&info->flush_wait);
2937 } else {
2938 wait = true;
2939 }
2940
2941 spin_unlock(&info->lock);
2942
2943 if (wait) {
2944 wait_on_flush(info);
2945 return;
2946 }
2947
2948 async = kzalloc(sizeof(*async), GFP_NOFS);
2949 if (!async)
2950 goto flush;
2951
2952 async->root = root;
2953 async->info = info;
2954 async->work.func = flush_delalloc_async;
2955
2956 btrfs_queue_worker(&root->fs_info->enospc_workers,
2957 &async->work);
2958 wait_on_flush(info);
2959 return;
2960
2961flush:
2962 btrfs_start_delalloc_inodes(root);
2963 btrfs_wait_ordered_extents(root, 0);
2964
2965 spin_lock(&info->lock);
2966 info->flushing = 0;
2967 spin_unlock(&info->lock);
2968 wake_up(&info->flush_wait);
2969}
2970
2867static int maybe_allocate_chunk(struct btrfs_root *root, 2971static int maybe_allocate_chunk(struct btrfs_root *root,
2868 struct btrfs_space_info *info) 2972 struct btrfs_space_info *info)
2869{ 2973{
@@ -2894,7 +2998,7 @@ static int maybe_allocate_chunk(struct btrfs_root *root,
2894 if (!info->allocating_chunk) { 2998 if (!info->allocating_chunk) {
2895 info->force_alloc = 1; 2999 info->force_alloc = 1;
2896 info->allocating_chunk = 1; 3000 info->allocating_chunk = 1;
2897 init_waitqueue_head(&info->wait); 3001 init_waitqueue_head(&info->allocate_wait);
2898 } else { 3002 } else {
2899 wait = true; 3003 wait = true;
2900 } 3004 }
@@ -2902,7 +3006,7 @@ static int maybe_allocate_chunk(struct btrfs_root *root,
2902 spin_unlock(&info->lock); 3006 spin_unlock(&info->lock);
2903 3007
2904 if (wait) { 3008 if (wait) {
2905 wait_event(info->wait, 3009 wait_event(info->allocate_wait,
2906 !info->allocating_chunk); 3010 !info->allocating_chunk);
2907 return 1; 3011 return 1;
2908 } 3012 }
@@ -2923,7 +3027,7 @@ out:
2923 spin_lock(&info->lock); 3027 spin_lock(&info->lock);
2924 info->allocating_chunk = 0; 3028 info->allocating_chunk = 0;
2925 spin_unlock(&info->lock); 3029 spin_unlock(&info->lock);
2926 wake_up(&info->wait); 3030 wake_up(&info->allocate_wait);
2927 3031
2928 if (ret) 3032 if (ret)
2929 return 0; 3033 return 0;
@@ -2981,21 +3085,20 @@ again:
2981 filemap_flush(inode->i_mapping); 3085 filemap_flush(inode->i_mapping);
2982 goto again; 3086 goto again;
2983 } else if (flushed == 3) { 3087 } else if (flushed == 3) {
2984 btrfs_start_delalloc_inodes(root); 3088 flush_delalloc(root, meta_sinfo);
2985 btrfs_wait_ordered_extents(root, 0);
2986 goto again; 3089 goto again;
2987 } 3090 }
2988 spin_lock(&meta_sinfo->lock); 3091 spin_lock(&meta_sinfo->lock);
2989 meta_sinfo->bytes_delalloc -= num_bytes; 3092 meta_sinfo->bytes_delalloc -= num_bytes;
2990 spin_unlock(&meta_sinfo->lock); 3093 spin_unlock(&meta_sinfo->lock);
2991 printk(KERN_ERR "enospc, has %d, reserved %d\n", 3094 printk(KERN_ERR "enospc, has %d, reserved %d\n",
2992 BTRFS_I(inode)->delalloc_extents, 3095 BTRFS_I(inode)->outstanding_extents,
2993 BTRFS_I(inode)->delalloc_reserved_extents); 3096 BTRFS_I(inode)->reserved_extents);
2994 dump_space_info(meta_sinfo, 0, 0); 3097 dump_space_info(meta_sinfo, 0, 0);
2995 return -ENOSPC; 3098 return -ENOSPC;
2996 } 3099 }
2997 3100
2998 BTRFS_I(inode)->delalloc_reserved_extents++; 3101 BTRFS_I(inode)->reserved_extents++;
2999 check_force_delalloc(meta_sinfo); 3102 check_force_delalloc(meta_sinfo);
3000 spin_unlock(&meta_sinfo->lock); 3103 spin_unlock(&meta_sinfo->lock);
3001 3104
@@ -3094,8 +3197,7 @@ again:
3094 } 3197 }
3095 3198
3096 if (retries == 2) { 3199 if (retries == 2) {
3097 btrfs_start_delalloc_inodes(root); 3200 flush_delalloc(root, meta_sinfo);
3098 btrfs_wait_ordered_extents(root, 0);
3099 goto again; 3201 goto again;
3100 } 3202 }
3101 spin_lock(&meta_sinfo->lock); 3203 spin_lock(&meta_sinfo->lock);
@@ -4029,6 +4131,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
4029 int loop = 0; 4131 int loop = 0;
4030 bool found_uncached_bg = false; 4132 bool found_uncached_bg = false;
4031 bool failed_cluster_refill = false; 4133 bool failed_cluster_refill = false;
4134 bool failed_alloc = false;
4032 4135
4033 WARN_ON(num_bytes < root->sectorsize); 4136 WARN_ON(num_bytes < root->sectorsize);
4034 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); 4137 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
@@ -4233,14 +4336,23 @@ refill_cluster:
4233 4336
4234 offset = btrfs_find_space_for_alloc(block_group, search_start, 4337 offset = btrfs_find_space_for_alloc(block_group, search_start,
4235 num_bytes, empty_size); 4338 num_bytes, empty_size);
4236 if (!offset && (cached || (!cached && 4339 /*
4237 loop == LOOP_CACHING_NOWAIT))) { 4340 * If we didn't find a chunk, and we haven't failed on this
4238 goto loop; 4341 * block group before, and this block group is in the middle of
4239 } else if (!offset && (!cached && 4342 * caching and we are ok with waiting, then go ahead and wait
4240 loop > LOOP_CACHING_NOWAIT)) { 4343 * for progress to be made, and set failed_alloc to true.
4344 *
4345 * If failed_alloc is true then we've already waited on this
4346 * block group once and should move on to the next block group.
4347 */
4348 if (!offset && !failed_alloc && !cached &&
4349 loop > LOOP_CACHING_NOWAIT) {
4241 wait_block_group_cache_progress(block_group, 4350 wait_block_group_cache_progress(block_group,
4242 num_bytes + empty_size); 4351 num_bytes + empty_size);
4352 failed_alloc = true;
4243 goto have_block_group; 4353 goto have_block_group;
4354 } else if (!offset) {
4355 goto loop;
4244 } 4356 }
4245checks: 4357checks:
4246 search_start = stripe_align(root, offset); 4358 search_start = stripe_align(root, offset);
@@ -4288,6 +4400,7 @@ checks:
4288 break; 4400 break;
4289loop: 4401loop:
4290 failed_cluster_refill = false; 4402 failed_cluster_refill = false;
4403 failed_alloc = false;
4291 btrfs_put_block_group(block_group); 4404 btrfs_put_block_group(block_group);
4292 } 4405 }
4293 up_read(&space_info->groups_sem); 4406 up_read(&space_info->groups_sem);
@@ -4799,6 +4912,7 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4799 u64 bytenr; 4912 u64 bytenr;
4800 u64 generation; 4913 u64 generation;
4801 u64 refs; 4914 u64 refs;
4915 u64 flags;
4802 u64 last = 0; 4916 u64 last = 0;
4803 u32 nritems; 4917 u32 nritems;
4804 u32 blocksize; 4918 u32 blocksize;
@@ -4836,15 +4950,19 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4836 generation <= root->root_key.offset) 4950 generation <= root->root_key.offset)
4837 continue; 4951 continue;
4838 4952
4953 /* We don't lock the tree block, it's OK to be racy here */
4954 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
4955 &refs, &flags);
4956 BUG_ON(ret);
4957 BUG_ON(refs == 0);
4958
4839 if (wc->stage == DROP_REFERENCE) { 4959 if (wc->stage == DROP_REFERENCE) {
4840 ret = btrfs_lookup_extent_info(trans, root,
4841 bytenr, blocksize,
4842 &refs, NULL);
4843 BUG_ON(ret);
4844 BUG_ON(refs == 0);
4845 if (refs == 1) 4960 if (refs == 1)
4846 goto reada; 4961 goto reada;
4847 4962
4963 if (wc->level == 1 &&
4964 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4965 continue;
4848 if (!wc->update_ref || 4966 if (!wc->update_ref ||
4849 generation <= root->root_key.offset) 4967 generation <= root->root_key.offset)
4850 continue; 4968 continue;
@@ -4853,6 +4971,10 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4853 &wc->update_progress); 4971 &wc->update_progress);
4854 if (ret < 0) 4972 if (ret < 0)
4855 continue; 4973 continue;
4974 } else {
4975 if (wc->level == 1 &&
4976 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4977 continue;
4856 } 4978 }
4857reada: 4979reada:
4858 ret = readahead_tree_block(root, bytenr, blocksize, 4980 ret = readahead_tree_block(root, bytenr, blocksize,
@@ -4876,7 +4998,7 @@ reada:
4876static noinline int walk_down_proc(struct btrfs_trans_handle *trans, 4998static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4877 struct btrfs_root *root, 4999 struct btrfs_root *root,
4878 struct btrfs_path *path, 5000 struct btrfs_path *path,
4879 struct walk_control *wc) 5001 struct walk_control *wc, int lookup_info)
4880{ 5002{
4881 int level = wc->level; 5003 int level = wc->level;
4882 struct extent_buffer *eb = path->nodes[level]; 5004 struct extent_buffer *eb = path->nodes[level];
@@ -4891,8 +5013,9 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4891 * when reference count of tree block is 1, it won't increase 5013 * when reference count of tree block is 1, it won't increase
4892 * again. once full backref flag is set, we never clear it. 5014 * again. once full backref flag is set, we never clear it.
4893 */ 5015 */
4894 if ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) || 5016 if (lookup_info &&
4895 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag))) { 5017 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
5018 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
4896 BUG_ON(!path->locks[level]); 5019 BUG_ON(!path->locks[level]);
4897 ret = btrfs_lookup_extent_info(trans, root, 5020 ret = btrfs_lookup_extent_info(trans, root,
4898 eb->start, eb->len, 5021 eb->start, eb->len,
@@ -4953,7 +5076,7 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4953static noinline int do_walk_down(struct btrfs_trans_handle *trans, 5076static noinline int do_walk_down(struct btrfs_trans_handle *trans,
4954 struct btrfs_root *root, 5077 struct btrfs_root *root,
4955 struct btrfs_path *path, 5078 struct btrfs_path *path,
4956 struct walk_control *wc) 5079 struct walk_control *wc, int *lookup_info)
4957{ 5080{
4958 u64 bytenr; 5081 u64 bytenr;
4959 u64 generation; 5082 u64 generation;
@@ -4973,8 +5096,10 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
4973 * for the subtree 5096 * for the subtree
4974 */ 5097 */
4975 if (wc->stage == UPDATE_BACKREF && 5098 if (wc->stage == UPDATE_BACKREF &&
4976 generation <= root->root_key.offset) 5099 generation <= root->root_key.offset) {
5100 *lookup_info = 1;
4977 return 1; 5101 return 1;
5102 }
4978 5103
4979 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]); 5104 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
4980 blocksize = btrfs_level_size(root, level - 1); 5105 blocksize = btrfs_level_size(root, level - 1);
@@ -4987,14 +5112,19 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
4987 btrfs_tree_lock(next); 5112 btrfs_tree_lock(next);
4988 btrfs_set_lock_blocking(next); 5113 btrfs_set_lock_blocking(next);
4989 5114
4990 if (wc->stage == DROP_REFERENCE) { 5115 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
4991 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize, 5116 &wc->refs[level - 1],
4992 &wc->refs[level - 1], 5117 &wc->flags[level - 1]);
4993 &wc->flags[level - 1]); 5118 BUG_ON(ret);
4994 BUG_ON(ret); 5119 BUG_ON(wc->refs[level - 1] == 0);
4995 BUG_ON(wc->refs[level - 1] == 0); 5120 *lookup_info = 0;
4996 5121
5122 if (wc->stage == DROP_REFERENCE) {
4997 if (wc->refs[level - 1] > 1) { 5123 if (wc->refs[level - 1] > 1) {
5124 if (level == 1 &&
5125 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5126 goto skip;
5127
4998 if (!wc->update_ref || 5128 if (!wc->update_ref ||
4999 generation <= root->root_key.offset) 5129 generation <= root->root_key.offset)
5000 goto skip; 5130 goto skip;
@@ -5008,12 +5138,17 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5008 wc->stage = UPDATE_BACKREF; 5138 wc->stage = UPDATE_BACKREF;
5009 wc->shared_level = level - 1; 5139 wc->shared_level = level - 1;
5010 } 5140 }
5141 } else {
5142 if (level == 1 &&
5143 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5144 goto skip;
5011 } 5145 }
5012 5146
5013 if (!btrfs_buffer_uptodate(next, generation)) { 5147 if (!btrfs_buffer_uptodate(next, generation)) {
5014 btrfs_tree_unlock(next); 5148 btrfs_tree_unlock(next);
5015 free_extent_buffer(next); 5149 free_extent_buffer(next);
5016 next = NULL; 5150 next = NULL;
5151 *lookup_info = 1;
5017 } 5152 }
5018 5153
5019 if (!next) { 5154 if (!next) {
@@ -5036,21 +5171,22 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5036skip: 5171skip:
5037 wc->refs[level - 1] = 0; 5172 wc->refs[level - 1] = 0;
5038 wc->flags[level - 1] = 0; 5173 wc->flags[level - 1] = 0;
5174 if (wc->stage == DROP_REFERENCE) {
5175 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5176 parent = path->nodes[level]->start;
5177 } else {
5178 BUG_ON(root->root_key.objectid !=
5179 btrfs_header_owner(path->nodes[level]));
5180 parent = 0;
5181 }
5039 5182
5040 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) { 5183 ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
5041 parent = path->nodes[level]->start; 5184 root->root_key.objectid, level - 1, 0);
5042 } else { 5185 BUG_ON(ret);
5043 BUG_ON(root->root_key.objectid !=
5044 btrfs_header_owner(path->nodes[level]));
5045 parent = 0;
5046 } 5186 }
5047
5048 ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
5049 root->root_key.objectid, level - 1, 0);
5050 BUG_ON(ret);
5051
5052 btrfs_tree_unlock(next); 5187 btrfs_tree_unlock(next);
5053 free_extent_buffer(next); 5188 free_extent_buffer(next);
5189 *lookup_info = 1;
5054 return 1; 5190 return 1;
5055} 5191}
5056 5192
@@ -5164,6 +5300,7 @@ static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5164 struct walk_control *wc) 5300 struct walk_control *wc)
5165{ 5301{
5166 int level = wc->level; 5302 int level = wc->level;
5303 int lookup_info = 1;
5167 int ret; 5304 int ret;
5168 5305
5169 while (level >= 0) { 5306 while (level >= 0) {
@@ -5171,14 +5308,14 @@ static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5171 btrfs_header_nritems(path->nodes[level])) 5308 btrfs_header_nritems(path->nodes[level]))
5172 break; 5309 break;
5173 5310
5174 ret = walk_down_proc(trans, root, path, wc); 5311 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5175 if (ret > 0) 5312 if (ret > 0)
5176 break; 5313 break;
5177 5314
5178 if (level == 0) 5315 if (level == 0)
5179 break; 5316 break;
5180 5317
5181 ret = do_walk_down(trans, root, path, wc); 5318 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5182 if (ret > 0) { 5319 if (ret > 0) {
5183 path->slots[level]++; 5320 path->slots[level]++;
5184 continue; 5321 continue;
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index de1793ba004a..96577e8bf9fd 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -460,7 +460,8 @@ static int clear_state_bit(struct extent_io_tree *tree,
460 struct extent_state *state, int bits, int wake, 460 struct extent_state *state, int bits, int wake,
461 int delete) 461 int delete)
462{ 462{
463 int ret = state->state & bits; 463 int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING;
464 int ret = state->state & bits_to_clear;
464 465
465 if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) { 466 if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
466 u64 range = state->end - state->start + 1; 467 u64 range = state->end - state->start + 1;
@@ -468,7 +469,7 @@ static int clear_state_bit(struct extent_io_tree *tree,
468 tree->dirty_bytes -= range; 469 tree->dirty_bytes -= range;
469 } 470 }
470 clear_state_cb(tree, state, bits); 471 clear_state_cb(tree, state, bits);
471 state->state &= ~bits; 472 state->state &= ~bits_to_clear;
472 if (wake) 473 if (wake)
473 wake_up(&state->wq); 474 wake_up(&state->wq);
474 if (delete || state->state == 0) { 475 if (delete || state->state == 0) {
@@ -956,7 +957,8 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
956 gfp_t mask) 957 gfp_t mask)
957{ 958{
958 return clear_extent_bit(tree, start, end, 959 return clear_extent_bit(tree, start, end,
959 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, 960 EXTENT_DIRTY | EXTENT_DELALLOC |
961 EXTENT_DO_ACCOUNTING, 0, 0,
960 NULL, mask); 962 NULL, mask);
961} 963}
962 964
@@ -1401,12 +1403,7 @@ out_failed:
1401int extent_clear_unlock_delalloc(struct inode *inode, 1403int extent_clear_unlock_delalloc(struct inode *inode,
1402 struct extent_io_tree *tree, 1404 struct extent_io_tree *tree,
1403 u64 start, u64 end, struct page *locked_page, 1405 u64 start, u64 end, struct page *locked_page,
1404 int unlock_pages, 1406 unsigned long op)
1405 int clear_unlock,
1406 int clear_delalloc, int clear_dirty,
1407 int set_writeback,
1408 int end_writeback,
1409 int set_private2)
1410{ 1407{
1411 int ret; 1408 int ret;
1412 struct page *pages[16]; 1409 struct page *pages[16];
@@ -1416,17 +1413,21 @@ int extent_clear_unlock_delalloc(struct inode *inode,
1416 int i; 1413 int i;
1417 int clear_bits = 0; 1414 int clear_bits = 0;
1418 1415
1419 if (clear_unlock) 1416 if (op & EXTENT_CLEAR_UNLOCK)
1420 clear_bits |= EXTENT_LOCKED; 1417 clear_bits |= EXTENT_LOCKED;
1421 if (clear_dirty) 1418 if (op & EXTENT_CLEAR_DIRTY)
1422 clear_bits |= EXTENT_DIRTY; 1419 clear_bits |= EXTENT_DIRTY;
1423 1420
1424 if (clear_delalloc) 1421 if (op & EXTENT_CLEAR_DELALLOC)
1425 clear_bits |= EXTENT_DELALLOC; 1422 clear_bits |= EXTENT_DELALLOC;
1426 1423
1424 if (op & EXTENT_CLEAR_ACCOUNTING)
1425 clear_bits |= EXTENT_DO_ACCOUNTING;
1426
1427 clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS); 1427 clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1428 if (!(unlock_pages || clear_dirty || set_writeback || end_writeback || 1428 if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
1429 set_private2)) 1429 EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
1430 EXTENT_SET_PRIVATE2)))
1430 return 0; 1431 return 0;
1431 1432
1432 while (nr_pages > 0) { 1433 while (nr_pages > 0) {
@@ -1435,20 +1436,20 @@ int extent_clear_unlock_delalloc(struct inode *inode,
1435 nr_pages, ARRAY_SIZE(pages)), pages); 1436 nr_pages, ARRAY_SIZE(pages)), pages);
1436 for (i = 0; i < ret; i++) { 1437 for (i = 0; i < ret; i++) {
1437 1438
1438 if (set_private2) 1439 if (op & EXTENT_SET_PRIVATE2)
1439 SetPagePrivate2(pages[i]); 1440 SetPagePrivate2(pages[i]);
1440 1441
1441 if (pages[i] == locked_page) { 1442 if (pages[i] == locked_page) {
1442 page_cache_release(pages[i]); 1443 page_cache_release(pages[i]);
1443 continue; 1444 continue;
1444 } 1445 }
1445 if (clear_dirty) 1446 if (op & EXTENT_CLEAR_DIRTY)
1446 clear_page_dirty_for_io(pages[i]); 1447 clear_page_dirty_for_io(pages[i]);
1447 if (set_writeback) 1448 if (op & EXTENT_SET_WRITEBACK)
1448 set_page_writeback(pages[i]); 1449 set_page_writeback(pages[i]);
1449 if (end_writeback) 1450 if (op & EXTENT_END_WRITEBACK)
1450 end_page_writeback(pages[i]); 1451 end_page_writeback(pages[i]);
1451 if (unlock_pages) 1452 if (op & EXTENT_CLEAR_UNLOCK_PAGE)
1452 unlock_page(pages[i]); 1453 unlock_page(pages[i]);
1453 page_cache_release(pages[i]); 1454 page_cache_release(pages[i]);
1454 } 1455 }
@@ -2714,7 +2715,8 @@ int extent_invalidatepage(struct extent_io_tree *tree,
2714 lock_extent(tree, start, end, GFP_NOFS); 2715 lock_extent(tree, start, end, GFP_NOFS);
2715 wait_on_page_writeback(page); 2716 wait_on_page_writeback(page);
2716 clear_extent_bit(tree, start, end, 2717 clear_extent_bit(tree, start, end,
2717 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, 2718 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2719 EXTENT_DO_ACCOUNTING,
2718 1, 1, NULL, GFP_NOFS); 2720 1, 1, NULL, GFP_NOFS);
2719 return 0; 2721 return 0;
2720} 2722}
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 4794ec891fed..36de250a7b2b 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -15,6 +15,7 @@
15#define EXTENT_BUFFER_FILLED (1 << 8) 15#define EXTENT_BUFFER_FILLED (1 << 8)
16#define EXTENT_BOUNDARY (1 << 9) 16#define EXTENT_BOUNDARY (1 << 9)
17#define EXTENT_NODATASUM (1 << 10) 17#define EXTENT_NODATASUM (1 << 10)
18#define EXTENT_DO_ACCOUNTING (1 << 11)
18#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK) 19#define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
19 20
20/* flags for bio submission */ 21/* flags for bio submission */
@@ -25,6 +26,16 @@
25#define EXTENT_BUFFER_BLOCKING 1 26#define EXTENT_BUFFER_BLOCKING 1
26#define EXTENT_BUFFER_DIRTY 2 27#define EXTENT_BUFFER_DIRTY 2
27 28
29/* these are flags for extent_clear_unlock_delalloc */
30#define EXTENT_CLEAR_UNLOCK_PAGE 0x1
31#define EXTENT_CLEAR_UNLOCK 0x2
32#define EXTENT_CLEAR_DELALLOC 0x4
33#define EXTENT_CLEAR_DIRTY 0x8
34#define EXTENT_SET_WRITEBACK 0x10
35#define EXTENT_END_WRITEBACK 0x20
36#define EXTENT_SET_PRIVATE2 0x40
37#define EXTENT_CLEAR_ACCOUNTING 0x80
38
28/* 39/*
29 * page->private values. Every page that is controlled by the extent 40 * page->private values. Every page that is controlled by the extent
30 * map has page->private set to one. 41 * map has page->private set to one.
@@ -288,10 +299,5 @@ int extent_range_uptodate(struct extent_io_tree *tree,
288int extent_clear_unlock_delalloc(struct inode *inode, 299int extent_clear_unlock_delalloc(struct inode *inode,
289 struct extent_io_tree *tree, 300 struct extent_io_tree *tree,
290 u64 start, u64 end, struct page *locked_page, 301 u64 start, u64 end, struct page *locked_page,
291 int unlock_page, 302 unsigned long op);
292 int clear_unlock,
293 int clear_delalloc, int clear_dirty,
294 int set_writeback,
295 int end_writeback,
296 int set_private2);
297#endif 303#endif
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index f19e1259a971..2d623aa0625f 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -878,7 +878,8 @@ again:
878 btrfs_put_ordered_extent(ordered); 878 btrfs_put_ordered_extent(ordered);
879 879
880 clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos, 880 clear_extent_bits(&BTRFS_I(inode)->io_tree, start_pos,
881 last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC, 881 last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
882 EXTENT_DO_ACCOUNTING,
882 GFP_NOFS); 883 GFP_NOFS);
883 unlock_extent(&BTRFS_I(inode)->io_tree, 884 unlock_extent(&BTRFS_I(inode)->io_tree,
884 start_pos, last_pos - 1, GFP_NOFS); 885 start_pos, last_pos - 1, GFP_NOFS);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 112e5aa85892..9e138b793dc7 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -424,9 +424,12 @@ again:
424 * and free up our temp pages. 424 * and free up our temp pages.
425 */ 425 */
426 extent_clear_unlock_delalloc(inode, 426 extent_clear_unlock_delalloc(inode,
427 &BTRFS_I(inode)->io_tree, 427 &BTRFS_I(inode)->io_tree,
428 start, end, NULL, 1, 0, 428 start, end, NULL,
429 0, 1, 1, 1, 0); 429 EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
430 EXTENT_CLEAR_DELALLOC |
431 EXTENT_CLEAR_ACCOUNTING |
432 EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
430 ret = 0; 433 ret = 0;
431 goto free_pages_out; 434 goto free_pages_out;
432 } 435 }
@@ -637,11 +640,14 @@ static noinline int submit_compressed_extents(struct inode *inode,
637 * clear dirty, set writeback and unlock the pages. 640 * clear dirty, set writeback and unlock the pages.
638 */ 641 */
639 extent_clear_unlock_delalloc(inode, 642 extent_clear_unlock_delalloc(inode,
640 &BTRFS_I(inode)->io_tree, 643 &BTRFS_I(inode)->io_tree,
641 async_extent->start, 644 async_extent->start,
642 async_extent->start + 645 async_extent->start +
643 async_extent->ram_size - 1, 646 async_extent->ram_size - 1,
644 NULL, 1, 1, 0, 1, 1, 0, 0); 647 NULL, EXTENT_CLEAR_UNLOCK_PAGE |
648 EXTENT_CLEAR_UNLOCK |
649 EXTENT_CLEAR_DELALLOC |
650 EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
645 651
646 ret = btrfs_submit_compressed_write(inode, 652 ret = btrfs_submit_compressed_write(inode,
647 async_extent->start, 653 async_extent->start,
@@ -712,9 +718,15 @@ static noinline int cow_file_range(struct inode *inode,
712 start, end, 0, NULL); 718 start, end, 0, NULL);
713 if (ret == 0) { 719 if (ret == 0) {
714 extent_clear_unlock_delalloc(inode, 720 extent_clear_unlock_delalloc(inode,
715 &BTRFS_I(inode)->io_tree, 721 &BTRFS_I(inode)->io_tree,
716 start, end, NULL, 1, 1, 722 start, end, NULL,
717 1, 1, 1, 1, 0); 723 EXTENT_CLEAR_UNLOCK_PAGE |
724 EXTENT_CLEAR_UNLOCK |
725 EXTENT_CLEAR_DELALLOC |
726 EXTENT_CLEAR_ACCOUNTING |
727 EXTENT_CLEAR_DIRTY |
728 EXTENT_SET_WRITEBACK |
729 EXTENT_END_WRITEBACK);
718 *nr_written = *nr_written + 730 *nr_written = *nr_written +
719 (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE; 731 (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
720 *page_started = 1; 732 *page_started = 1;
@@ -738,6 +750,8 @@ static noinline int cow_file_range(struct inode *inode,
738 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0); 750 btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
739 751
740 while (disk_num_bytes > 0) { 752 while (disk_num_bytes > 0) {
753 unsigned long op;
754
741 cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent); 755 cur_alloc_size = min(disk_num_bytes, root->fs_info->max_extent);
742 ret = btrfs_reserve_extent(trans, root, cur_alloc_size, 756 ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
743 root->sectorsize, 0, alloc_hint, 757 root->sectorsize, 0, alloc_hint,
@@ -789,10 +803,13 @@ static noinline int cow_file_range(struct inode *inode,
789 * Do set the Private2 bit so we know this page was properly 803 * Do set the Private2 bit so we know this page was properly
790 * setup for writepage 804 * setup for writepage
791 */ 805 */
806 op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
807 op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
808 EXTENT_SET_PRIVATE2;
809
792 extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, 810 extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
793 start, start + ram_size - 1, 811 start, start + ram_size - 1,
794 locked_page, unlock, 1, 812 locked_page, op);
795 1, 0, 0, 0, 1);
796 disk_num_bytes -= cur_alloc_size; 813 disk_num_bytes -= cur_alloc_size;
797 num_bytes -= cur_alloc_size; 814 num_bytes -= cur_alloc_size;
798 alloc_hint = ins.objectid + ins.offset; 815 alloc_hint = ins.objectid + ins.offset;
@@ -864,8 +881,8 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page,
864 u64 cur_end; 881 u64 cur_end;
865 int limit = 10 * 1024 * 1042; 882 int limit = 10 * 1024 * 1042;
866 883
867 clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED | 884 clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
868 EXTENT_DELALLOC, 1, 0, NULL, GFP_NOFS); 885 1, 0, NULL, GFP_NOFS);
869 while (start < end) { 886 while (start < end) {
870 async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS); 887 async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
871 async_cow->inode = inode; 888 async_cow->inode = inode;
@@ -1006,6 +1023,7 @@ next_slot:
1006 1023
1007 if (found_key.offset > cur_offset) { 1024 if (found_key.offset > cur_offset) {
1008 extent_end = found_key.offset; 1025 extent_end = found_key.offset;
1026 extent_type = 0;
1009 goto out_check; 1027 goto out_check;
1010 } 1028 }
1011 1029
@@ -1112,8 +1130,10 @@ out_check:
1112 BUG_ON(ret); 1130 BUG_ON(ret);
1113 1131
1114 extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree, 1132 extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1115 cur_offset, cur_offset + num_bytes - 1, 1133 cur_offset, cur_offset + num_bytes - 1,
1116 locked_page, 1, 1, 1, 0, 0, 0, 1); 1134 locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
1135 EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
1136 EXTENT_SET_PRIVATE2);
1117 cur_offset = extent_end; 1137 cur_offset = extent_end;
1118 if (cur_offset > end) 1138 if (cur_offset > end)
1119 break; 1139 break;
@@ -1178,15 +1198,17 @@ static int btrfs_split_extent_hook(struct inode *inode,
1178 root->fs_info->max_extent); 1198 root->fs_info->max_extent);
1179 1199
1180 /* 1200 /*
1181 * if we break a large extent up then leave delalloc_extents be, 1201 * if we break a large extent up then leave oustanding_extents
1182 * since we've already accounted for the large extent. 1202 * be, since we've already accounted for the large extent.
1183 */ 1203 */
1184 if (div64_u64(new_size + root->fs_info->max_extent - 1, 1204 if (div64_u64(new_size + root->fs_info->max_extent - 1,
1185 root->fs_info->max_extent) < num_extents) 1205 root->fs_info->max_extent) < num_extents)
1186 return 0; 1206 return 0;
1187 } 1207 }
1188 1208
1189 BTRFS_I(inode)->delalloc_extents++; 1209 spin_lock(&BTRFS_I(inode)->accounting_lock);
1210 BTRFS_I(inode)->outstanding_extents++;
1211 spin_unlock(&BTRFS_I(inode)->accounting_lock);
1190 1212
1191 return 0; 1213 return 0;
1192} 1214}
@@ -1217,7 +1239,9 @@ static int btrfs_merge_extent_hook(struct inode *inode,
1217 1239
1218 /* we're not bigger than the max, unreserve the space and go */ 1240 /* we're not bigger than the max, unreserve the space and go */
1219 if (new_size <= root->fs_info->max_extent) { 1241 if (new_size <= root->fs_info->max_extent) {
1220 BTRFS_I(inode)->delalloc_extents--; 1242 spin_lock(&BTRFS_I(inode)->accounting_lock);
1243 BTRFS_I(inode)->outstanding_extents--;
1244 spin_unlock(&BTRFS_I(inode)->accounting_lock);
1221 return 0; 1245 return 0;
1222 } 1246 }
1223 1247
@@ -1231,7 +1255,9 @@ static int btrfs_merge_extent_hook(struct inode *inode,
1231 root->fs_info->max_extent) > num_extents) 1255 root->fs_info->max_extent) > num_extents)
1232 return 0; 1256 return 0;
1233 1257
1234 BTRFS_I(inode)->delalloc_extents--; 1258 spin_lock(&BTRFS_I(inode)->accounting_lock);
1259 BTRFS_I(inode)->outstanding_extents--;
1260 spin_unlock(&BTRFS_I(inode)->accounting_lock);
1235 1261
1236 return 0; 1262 return 0;
1237} 1263}
@@ -1253,7 +1279,9 @@ static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
1253 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { 1279 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
1254 struct btrfs_root *root = BTRFS_I(inode)->root; 1280 struct btrfs_root *root = BTRFS_I(inode)->root;
1255 1281
1256 BTRFS_I(inode)->delalloc_extents++; 1282 spin_lock(&BTRFS_I(inode)->accounting_lock);
1283 BTRFS_I(inode)->outstanding_extents++;
1284 spin_unlock(&BTRFS_I(inode)->accounting_lock);
1257 btrfs_delalloc_reserve_space(root, inode, end - start + 1); 1285 btrfs_delalloc_reserve_space(root, inode, end - start + 1);
1258 spin_lock(&root->fs_info->delalloc_lock); 1286 spin_lock(&root->fs_info->delalloc_lock);
1259 BTRFS_I(inode)->delalloc_bytes += end - start + 1; 1287 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
@@ -1281,8 +1309,12 @@ static int btrfs_clear_bit_hook(struct inode *inode,
1281 if ((state->state & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) { 1309 if ((state->state & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
1282 struct btrfs_root *root = BTRFS_I(inode)->root; 1310 struct btrfs_root *root = BTRFS_I(inode)->root;
1283 1311
1284 BTRFS_I(inode)->delalloc_extents--; 1312 if (bits & EXTENT_DO_ACCOUNTING) {
1285 btrfs_unreserve_metadata_for_delalloc(root, inode, 1); 1313 spin_lock(&BTRFS_I(inode)->accounting_lock);
1314 BTRFS_I(inode)->outstanding_extents--;
1315 spin_unlock(&BTRFS_I(inode)->accounting_lock);
1316 btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
1317 }
1286 1318
1287 spin_lock(&root->fs_info->delalloc_lock); 1319 spin_lock(&root->fs_info->delalloc_lock);
1288 if (state->end - state->start + 1 > 1320 if (state->end - state->start + 1 >
@@ -3598,12 +3630,14 @@ static int btrfs_dentry_delete(struct dentry *dentry)
3598{ 3630{
3599 struct btrfs_root *root; 3631 struct btrfs_root *root;
3600 3632
3601 if (!dentry->d_inode) 3633 if (!dentry->d_inode && !IS_ROOT(dentry))
3602 return 0; 3634 dentry = dentry->d_parent;
3603 3635
3604 root = BTRFS_I(dentry->d_inode)->root; 3636 if (dentry->d_inode) {
3605 if (btrfs_root_refs(&root->root_item) == 0) 3637 root = BTRFS_I(dentry->d_inode)->root;
3606 return 1; 3638 if (btrfs_root_refs(&root->root_item) == 0)
3639 return 1;
3640 }
3607 return 0; 3641 return 0;
3608} 3642}
3609 3643
@@ -4808,7 +4842,8 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
4808 */ 4842 */
4809 clear_extent_bit(tree, page_start, page_end, 4843 clear_extent_bit(tree, page_start, page_end,
4810 EXTENT_DIRTY | EXTENT_DELALLOC | 4844 EXTENT_DIRTY | EXTENT_DELALLOC |
4811 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS); 4845 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
4846 NULL, GFP_NOFS);
4812 /* 4847 /*
4813 * whoever cleared the private bit is responsible 4848 * whoever cleared the private bit is responsible
4814 * for the finish_ordered_io 4849 * for the finish_ordered_io
@@ -4821,8 +4856,8 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
4821 lock_extent(tree, page_start, page_end, GFP_NOFS); 4856 lock_extent(tree, page_start, page_end, GFP_NOFS);
4822 } 4857 }
4823 clear_extent_bit(tree, page_start, page_end, 4858 clear_extent_bit(tree, page_start, page_end,
4824 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC, 4859 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
4825 1, 1, NULL, GFP_NOFS); 4860 EXTENT_DO_ACCOUNTING, 1, 1, NULL, GFP_NOFS);
4826 __btrfs_releasepage(page, GFP_NOFS); 4861 __btrfs_releasepage(page, GFP_NOFS);
4827 4862
4828 ClearPageChecked(page); 4863 ClearPageChecked(page);
@@ -4917,7 +4952,8 @@ again:
4917 * prepare_pages in the normal write path. 4952 * prepare_pages in the normal write path.
4918 */ 4953 */
4919 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 4954 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end,
4920 EXTENT_DIRTY | EXTENT_DELALLOC, GFP_NOFS); 4955 EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
4956 GFP_NOFS);
4921 4957
4922 ret = btrfs_set_extent_delalloc(inode, page_start, page_end); 4958 ret = btrfs_set_extent_delalloc(inode, page_start, page_end);
4923 if (ret) { 4959 if (ret) {
@@ -5065,8 +5101,9 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
5065 return NULL; 5101 return NULL;
5066 ei->last_trans = 0; 5102 ei->last_trans = 0;
5067 ei->logged_trans = 0; 5103 ei->logged_trans = 0;
5068 ei->delalloc_extents = 0; 5104 ei->outstanding_extents = 0;
5069 ei->delalloc_reserved_extents = 0; 5105 ei->reserved_extents = 0;
5106 spin_lock_init(&ei->accounting_lock);
5070 btrfs_ordered_inode_tree_init(&ei->ordered_tree); 5107 btrfs_ordered_inode_tree_init(&ei->ordered_tree);
5071 INIT_LIST_HEAD(&ei->i_orphan); 5108 INIT_LIST_HEAD(&ei->i_orphan);
5072 INIT_LIST_HEAD(&ei->ordered_operations); 5109 INIT_LIST_HEAD(&ei->ordered_operations);
@@ -5805,6 +5842,6 @@ static const struct inode_operations btrfs_symlink_inode_operations = {
5805 .removexattr = btrfs_removexattr, 5842 .removexattr = btrfs_removexattr,
5806}; 5843};
5807 5844
5808struct dentry_operations btrfs_dentry_operations = { 5845const struct dentry_operations btrfs_dentry_operations = {
5809 .d_delete = btrfs_dentry_delete, 5846 .d_delete = btrfs_dentry_delete,
5810}; 5847};
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 9a780c8d0ac8..cdbb054102b9 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -830,6 +830,7 @@ out_up_write:
830out_unlock: 830out_unlock:
831 mutex_unlock(&inode->i_mutex); 831 mutex_unlock(&inode->i_mutex);
832 if (!err) { 832 if (!err) {
833 shrink_dcache_sb(root->fs_info->sb);
833 btrfs_invalidate_inodes(dest); 834 btrfs_invalidate_inodes(dest);
834 d_delete(dentry); 835 d_delete(dentry);
835 } 836 }
@@ -1122,8 +1123,10 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1122 datao += off - key.offset; 1123 datao += off - key.offset;
1123 datal -= off - key.offset; 1124 datal -= off - key.offset;
1124 } 1125 }
1125 if (key.offset + datao + datal > off + len) 1126
1126 datal = off + len - key.offset - datao; 1127 if (key.offset + datal > off + len)
1128 datal = off + len - key.offset;
1129
1127 /* disko == 0 means it's a hole */ 1130 /* disko == 0 means it's a hole */
1128 if (!disko) 1131 if (!disko)
1129 datao = 0; 1132 datao = 0;
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 897fba835f89..5799bc46a309 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -306,6 +306,12 @@ int btrfs_remove_ordered_extent(struct inode *inode,
306 tree->last = NULL; 306 tree->last = NULL;
307 set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags); 307 set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
308 308
309 spin_lock(&BTRFS_I(inode)->accounting_lock);
310 BTRFS_I(inode)->outstanding_extents--;
311 spin_unlock(&BTRFS_I(inode)->accounting_lock);
312 btrfs_unreserve_metadata_for_delalloc(BTRFS_I(inode)->root,
313 inode, 1);
314
309 spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock); 315 spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
310 list_del_init(&entry->root_extent_list); 316 list_del_init(&entry->root_extent_list);
311 317
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 361ad323faac..cfcc93c93a7b 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -3518,7 +3518,7 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
3518 BUG_ON(!rc->block_group); 3518 BUG_ON(!rc->block_group);
3519 3519
3520 btrfs_init_workers(&rc->workers, "relocate", 3520 btrfs_init_workers(&rc->workers, "relocate",
3521 fs_info->thread_pool_size); 3521 fs_info->thread_pool_size, NULL);
3522 3522
3523 rc->extent_root = extent_root; 3523 rc->extent_root = extent_root;
3524 btrfs_prepare_block_group_relocation(extent_root, rc->block_group); 3524 btrfs_prepare_block_group_relocation(extent_root, rc->block_group);
@@ -3701,7 +3701,7 @@ int btrfs_recover_relocation(struct btrfs_root *root)
3701 mapping_tree_init(&rc->reloc_root_tree); 3701 mapping_tree_init(&rc->reloc_root_tree);
3702 INIT_LIST_HEAD(&rc->reloc_roots); 3702 INIT_LIST_HEAD(&rc->reloc_roots);
3703 btrfs_init_workers(&rc->workers, "relocate", 3703 btrfs_init_workers(&rc->workers, "relocate",
3704 root->fs_info->thread_pool_size); 3704 root->fs_info->thread_pool_size, NULL);
3705 rc->extent_root = root->fs_info->extent_root; 3705 rc->extent_root = root->fs_info->extent_root;
3706 3706
3707 set_reloc_control(rc); 3707 set_reloc_control(rc);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 7827841b55cb..4edfdc2acc5f 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -137,11 +137,20 @@ static int start_log_trans(struct btrfs_trans_handle *trans,
137 137
138 mutex_lock(&root->log_mutex); 138 mutex_lock(&root->log_mutex);
139 if (root->log_root) { 139 if (root->log_root) {
140 if (!root->log_start_pid) {
141 root->log_start_pid = current->pid;
142 root->log_multiple_pids = false;
143 } else if (root->log_start_pid != current->pid) {
144 root->log_multiple_pids = true;
145 }
146
140 root->log_batch++; 147 root->log_batch++;
141 atomic_inc(&root->log_writers); 148 atomic_inc(&root->log_writers);
142 mutex_unlock(&root->log_mutex); 149 mutex_unlock(&root->log_mutex);
143 return 0; 150 return 0;
144 } 151 }
152 root->log_multiple_pids = false;
153 root->log_start_pid = current->pid;
145 mutex_lock(&root->fs_info->tree_log_mutex); 154 mutex_lock(&root->fs_info->tree_log_mutex);
146 if (!root->fs_info->log_root_tree) { 155 if (!root->fs_info->log_root_tree) {
147 ret = btrfs_init_log_root_tree(trans, root->fs_info); 156 ret = btrfs_init_log_root_tree(trans, root->fs_info);
@@ -1985,7 +1994,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
1985 if (atomic_read(&root->log_commit[(index1 + 1) % 2])) 1994 if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
1986 wait_log_commit(trans, root, root->log_transid - 1); 1995 wait_log_commit(trans, root, root->log_transid - 1);
1987 1996
1988 while (1) { 1997 while (root->log_multiple_pids) {
1989 unsigned long batch = root->log_batch; 1998 unsigned long batch = root->log_batch;
1990 mutex_unlock(&root->log_mutex); 1999 mutex_unlock(&root->log_mutex);
1991 schedule_timeout_uninterruptible(1); 2000 schedule_timeout_uninterruptible(1);
@@ -2011,6 +2020,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
2011 root->log_batch = 0; 2020 root->log_batch = 0;
2012 root->log_transid++; 2021 root->log_transid++;
2013 log->log_transid = root->log_transid; 2022 log->log_transid = root->log_transid;
2023 root->log_start_pid = 0;
2014 smp_mb(); 2024 smp_mb();
2015 /* 2025 /*
2016 * log tree has been flushed to disk, new modifications of 2026 * log tree has been flushed to disk, new modifications of
diff --git a/fs/ecryptfs/Kconfig b/fs/ecryptfs/Kconfig
index 8aadb99b7634..1cd6d9d3e29a 100644
--- a/fs/ecryptfs/Kconfig
+++ b/fs/ecryptfs/Kconfig
@@ -1,8 +1,9 @@
1config ECRYPT_FS 1config ECRYPT_FS
2 tristate "eCrypt filesystem layer support (EXPERIMENTAL)" 2 tristate "eCrypt filesystem layer support (EXPERIMENTAL)"
3 depends on EXPERIMENTAL && KEYS && NET 3 depends on EXPERIMENTAL && KEYS && CRYPTO
4 select CRYPTO_ECB 4 select CRYPTO_ECB
5 select CRYPTO_CBC 5 select CRYPTO_CBC
6 select CRYPTO_MD5
6 help 7 help
7 Encrypted filesystem that operates on the VFS layer. See 8 Encrypted filesystem that operates on the VFS layer. See
8 <file:Documentation/filesystems/ecryptfs.txt> to learn more about 9 <file:Documentation/filesystems/ecryptfs.txt> to learn more about
diff --git a/fs/ecryptfs/main.c b/fs/ecryptfs/main.c
index 101fe4c7b1ee..c6ac85d6c701 100644
--- a/fs/ecryptfs/main.c
+++ b/fs/ecryptfs/main.c
@@ -35,6 +35,7 @@
35#include <linux/key.h> 35#include <linux/key.h>
36#include <linux/parser.h> 36#include <linux/parser.h>
37#include <linux/fs_stack.h> 37#include <linux/fs_stack.h>
38#include <linux/ima.h>
38#include "ecryptfs_kernel.h" 39#include "ecryptfs_kernel.h"
39 40
40/** 41/**
@@ -118,6 +119,7 @@ int ecryptfs_init_persistent_file(struct dentry *ecryptfs_dentry)
118 const struct cred *cred = current_cred(); 119 const struct cred *cred = current_cred();
119 struct ecryptfs_inode_info *inode_info = 120 struct ecryptfs_inode_info *inode_info =
120 ecryptfs_inode_to_private(ecryptfs_dentry->d_inode); 121 ecryptfs_inode_to_private(ecryptfs_dentry->d_inode);
122 int opened_lower_file = 0;
121 int rc = 0; 123 int rc = 0;
122 124
123 mutex_lock(&inode_info->lower_file_mutex); 125 mutex_lock(&inode_info->lower_file_mutex);
@@ -134,9 +136,12 @@ int ecryptfs_init_persistent_file(struct dentry *ecryptfs_dentry)
134 "for lower_dentry [0x%p] and lower_mnt [0x%p]; " 136 "for lower_dentry [0x%p] and lower_mnt [0x%p]; "
135 "rc = [%d]\n", lower_dentry, lower_mnt, rc); 137 "rc = [%d]\n", lower_dentry, lower_mnt, rc);
136 inode_info->lower_file = NULL; 138 inode_info->lower_file = NULL;
137 } 139 } else
140 opened_lower_file = 1;
138 } 141 }
139 mutex_unlock(&inode_info->lower_file_mutex); 142 mutex_unlock(&inode_info->lower_file_mutex);
143 if (opened_lower_file)
144 ima_counts_get(inode_info->lower_file);
140 return rc; 145 return rc;
141} 146}
142 147
diff --git a/fs/file.c b/fs/file.c
index f313314f996f..87e129030ab1 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -10,6 +10,7 @@
10#include <linux/fs.h> 10#include <linux/fs.h>
11#include <linux/mm.h> 11#include <linux/mm.h>
12#include <linux/time.h> 12#include <linux/time.h>
13#include <linux/sched.h>
13#include <linux/slab.h> 14#include <linux/slab.h>
14#include <linux/vmalloc.h> 15#include <linux/vmalloc.h>
15#include <linux/file.h> 16#include <linux/file.h>
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 63976c0ccc25..99ea196f071f 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -1180,7 +1180,7 @@ static int nfs4_init_client(struct nfs_client *clp,
1180 1, flags & NFS_MOUNT_NORESVPORT); 1180 1, flags & NFS_MOUNT_NORESVPORT);
1181 if (error < 0) 1181 if (error < 0)
1182 goto error; 1182 goto error;
1183 memcpy(clp->cl_ipaddr, ip_addr, sizeof(clp->cl_ipaddr)); 1183 strlcpy(clp->cl_ipaddr, ip_addr, sizeof(clp->cl_ipaddr));
1184 1184
1185 error = nfs_idmap_new(clp); 1185 error = nfs_idmap_new(clp);
1186 if (error < 0) { 1186 if (error < 0) {
diff --git a/fs/nfs/nfs4namespace.c b/fs/nfs/nfs4namespace.c
index 2636c26d56fa..fa3408f20112 100644
--- a/fs/nfs/nfs4namespace.c
+++ b/fs/nfs/nfs4namespace.c
@@ -121,7 +121,7 @@ static struct vfsmount *try_location(struct nfs_clone_mount *mountdata,
121 121
122 mnt_path = nfs4_pathname_string(&location->rootpath, page2, PAGE_SIZE); 122 mnt_path = nfs4_pathname_string(&location->rootpath, page2, PAGE_SIZE);
123 if (IS_ERR(mnt_path)) 123 if (IS_ERR(mnt_path))
124 return mnt; 124 return ERR_CAST(mnt_path);
125 mountdata->mnt_path = mnt_path; 125 mountdata->mnt_path = mnt_path;
126 maxbuflen = mnt_path - 1 - page2; 126 maxbuflen = mnt_path - 1 - page2;
127 127
@@ -132,15 +132,15 @@ static struct vfsmount *try_location(struct nfs_clone_mount *mountdata,
132 if (buf->len <= 0 || buf->len >= maxbuflen) 132 if (buf->len <= 0 || buf->len >= maxbuflen)
133 continue; 133 continue;
134 134
135 mountdata->addr = (struct sockaddr *)&addr;
136
137 if (memchr(buf->data, IPV6_SCOPE_DELIMITER, buf->len)) 135 if (memchr(buf->data, IPV6_SCOPE_DELIMITER, buf->len))
138 continue; 136 continue;
139 mountdata->addrlen = nfs_parse_server_name(buf->data, 137
140 buf->len, 138 mountdata->addrlen = nfs_parse_server_name(buf->data, buf->len,
141 mountdata->addr, mountdata->addrlen); 139 (struct sockaddr *)&addr, sizeof(addr));
142 if (mountdata->addrlen == 0) 140 if (mountdata->addrlen == 0)
143 continue; 141 continue;
142
143 mountdata->addr = (struct sockaddr *)&addr;
144 rpc_set_port(mountdata->addr, NFS_PORT); 144 rpc_set_port(mountdata->addr, NFS_PORT);
145 145
146 memcpy(page2, buf->data, buf->len); 146 memcpy(page2, buf->data, buf->len);
diff --git a/fs/nfs/nfs4renewd.c b/fs/nfs/nfs4renewd.c
index e27c6cef18f2..0156c01c212c 100644
--- a/fs/nfs/nfs4renewd.c
+++ b/fs/nfs/nfs4renewd.c
@@ -127,12 +127,6 @@ nfs4_schedule_state_renewal(struct nfs_client *clp)
127} 127}
128 128
129void 129void
130nfs4_renewd_prepare_shutdown(struct nfs_server *server)
131{
132 cancel_delayed_work(&server->nfs_client->cl_renewd);
133}
134
135void
136nfs4_kill_renewd(struct nfs_client *clp) 130nfs4_kill_renewd(struct nfs_client *clp)
137{ 131{
138 cancel_delayed_work_sync(&clp->cl_renewd); 132 cancel_delayed_work_sync(&clp->cl_renewd);
diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index 29786d3b9326..a2c18acb8568 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -728,22 +728,24 @@ static void nfs_umount_begin(struct super_block *sb)
728 unlock_kernel(); 728 unlock_kernel();
729} 729}
730 730
731static struct nfs_parsed_mount_data *nfs_alloc_parsed_mount_data(int flags) 731static struct nfs_parsed_mount_data *nfs_alloc_parsed_mount_data(unsigned int version)
732{ 732{
733 struct nfs_parsed_mount_data *data; 733 struct nfs_parsed_mount_data *data;
734 734
735 data = kzalloc(sizeof(*data), GFP_KERNEL); 735 data = kzalloc(sizeof(*data), GFP_KERNEL);
736 if (data) { 736 if (data) {
737 data->flags = flags;
738 data->rsize = NFS_MAX_FILE_IO_SIZE; 737 data->rsize = NFS_MAX_FILE_IO_SIZE;
739 data->wsize = NFS_MAX_FILE_IO_SIZE; 738 data->wsize = NFS_MAX_FILE_IO_SIZE;
740 data->acregmin = NFS_DEF_ACREGMIN; 739 data->acregmin = NFS_DEF_ACREGMIN;
741 data->acregmax = NFS_DEF_ACREGMAX; 740 data->acregmax = NFS_DEF_ACREGMAX;
742 data->acdirmin = NFS_DEF_ACDIRMIN; 741 data->acdirmin = NFS_DEF_ACDIRMIN;
743 data->acdirmax = NFS_DEF_ACDIRMAX; 742 data->acdirmax = NFS_DEF_ACDIRMAX;
743 data->mount_server.port = NFS_UNSPEC_PORT;
744 data->nfs_server.port = NFS_UNSPEC_PORT; 744 data->nfs_server.port = NFS_UNSPEC_PORT;
745 data->nfs_server.protocol = XPRT_TRANSPORT_TCP;
745 data->auth_flavors[0] = RPC_AUTH_UNIX; 746 data->auth_flavors[0] = RPC_AUTH_UNIX;
746 data->auth_flavor_len = 1; 747 data->auth_flavor_len = 1;
748 data->version = version;
747 data->minorversion = 0; 749 data->minorversion = 0;
748 } 750 }
749 return data; 751 return data;
@@ -776,15 +778,13 @@ static int nfs_verify_server_address(struct sockaddr *addr)
776 * Select between a default port value and a user-specified port value. 778 * Select between a default port value and a user-specified port value.
777 * If a zero value is set, then autobind will be used. 779 * If a zero value is set, then autobind will be used.
778 */ 780 */
779static void nfs_set_default_port(struct sockaddr *sap, const int parsed_port, 781static void nfs_set_port(struct sockaddr *sap, int *port,
780 const unsigned short default_port) 782 const unsigned short default_port)
781{ 783{
782 unsigned short port = default_port; 784 if (*port == NFS_UNSPEC_PORT)
785 *port = default_port;
783 786
784 if (parsed_port != NFS_UNSPEC_PORT) 787 rpc_set_port(sap, *port);
785 port = parsed_port;
786
787 rpc_set_port(sap, port);
788} 788}
789 789
790/* 790/*
@@ -1475,7 +1475,7 @@ static int nfs_try_mount(struct nfs_parsed_mount_data *args,
1475 args->mount_server.addrlen = args->nfs_server.addrlen; 1475 args->mount_server.addrlen = args->nfs_server.addrlen;
1476 } 1476 }
1477 request.salen = args->mount_server.addrlen; 1477 request.salen = args->mount_server.addrlen;
1478 nfs_set_default_port(request.sap, args->mount_server.port, 0); 1478 nfs_set_port(request.sap, &args->mount_server.port, 0);
1479 1479
1480 /* 1480 /*
1481 * Now ask the mount server to map our export path 1481 * Now ask the mount server to map our export path
@@ -1711,8 +1711,6 @@ static int nfs_validate_mount_data(void *options,
1711 1711
1712 if (!(data->flags & NFS_MOUNT_TCP)) 1712 if (!(data->flags & NFS_MOUNT_TCP))
1713 args->nfs_server.protocol = XPRT_TRANSPORT_UDP; 1713 args->nfs_server.protocol = XPRT_TRANSPORT_UDP;
1714 else
1715 args->nfs_server.protocol = XPRT_TRANSPORT_TCP;
1716 /* N.B. caller will free nfs_server.hostname in all cases */ 1714 /* N.B. caller will free nfs_server.hostname in all cases */
1717 args->nfs_server.hostname = kstrdup(data->hostname, GFP_KERNEL); 1715 args->nfs_server.hostname = kstrdup(data->hostname, GFP_KERNEL);
1718 args->namlen = data->namlen; 1716 args->namlen = data->namlen;
@@ -1767,7 +1765,7 @@ static int nfs_validate_mount_data(void *options,
1767 goto out_v4_not_compiled; 1765 goto out_v4_not_compiled;
1768#endif 1766#endif
1769 1767
1770 nfs_set_default_port(sap, args->nfs_server.port, 0); 1768 nfs_set_port(sap, &args->nfs_server.port, 0);
1771 1769
1772 nfs_set_mount_transport_protocol(args); 1770 nfs_set_mount_transport_protocol(args);
1773 1771
@@ -1848,9 +1846,10 @@ nfs_compare_remount_data(struct nfs_server *nfss,
1848 data->acdirmin != nfss->acdirmin / HZ || 1846 data->acdirmin != nfss->acdirmin / HZ ||
1849 data->acdirmax != nfss->acdirmax / HZ || 1847 data->acdirmax != nfss->acdirmax / HZ ||
1850 data->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) || 1848 data->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) ||
1849 data->nfs_server.port != nfss->port ||
1851 data->nfs_server.addrlen != nfss->nfs_client->cl_addrlen || 1850 data->nfs_server.addrlen != nfss->nfs_client->cl_addrlen ||
1852 memcmp(&data->nfs_server.address, &nfss->nfs_client->cl_addr, 1851 !rpc_cmp_addr((struct sockaddr *)&data->nfs_server.address,
1853 data->nfs_server.addrlen) != 0) 1852 (struct sockaddr *)&nfss->nfs_client->cl_addr))
1854 return -EINVAL; 1853 return -EINVAL;
1855 1854
1856 return 0; 1855 return 0;
@@ -1893,6 +1892,7 @@ nfs_remount(struct super_block *sb, int *flags, char *raw_data)
1893 data->acdirmin = nfss->acdirmin / HZ; 1892 data->acdirmin = nfss->acdirmin / HZ;
1894 data->acdirmax = nfss->acdirmax / HZ; 1893 data->acdirmax = nfss->acdirmax / HZ;
1895 data->timeo = 10U * nfss->client->cl_timeout->to_initval / HZ; 1894 data->timeo = 10U * nfss->client->cl_timeout->to_initval / HZ;
1895 data->nfs_server.port = nfss->port;
1896 data->nfs_server.addrlen = nfss->nfs_client->cl_addrlen; 1896 data->nfs_server.addrlen = nfss->nfs_client->cl_addrlen;
1897 memcpy(&data->nfs_server.address, &nfss->nfs_client->cl_addr, 1897 memcpy(&data->nfs_server.address, &nfss->nfs_client->cl_addr,
1898 data->nfs_server.addrlen); 1898 data->nfs_server.addrlen);
@@ -2106,7 +2106,7 @@ static int nfs_get_sb(struct file_system_type *fs_type,
2106 }; 2106 };
2107 int error = -ENOMEM; 2107 int error = -ENOMEM;
2108 2108
2109 data = nfs_alloc_parsed_mount_data(NFS_MOUNT_VER3 | NFS_MOUNT_TCP); 2109 data = nfs_alloc_parsed_mount_data(3);
2110 mntfh = kzalloc(sizeof(*mntfh), GFP_KERNEL); 2110 mntfh = kzalloc(sizeof(*mntfh), GFP_KERNEL);
2111 if (data == NULL || mntfh == NULL) 2111 if (data == NULL || mntfh == NULL)
2112 goto out_free_fh; 2112 goto out_free_fh;
@@ -2331,7 +2331,7 @@ static int nfs4_validate_text_mount_data(void *options,
2331{ 2331{
2332 struct sockaddr *sap = (struct sockaddr *)&args->nfs_server.address; 2332 struct sockaddr *sap = (struct sockaddr *)&args->nfs_server.address;
2333 2333
2334 nfs_set_default_port(sap, args->nfs_server.port, NFS_PORT); 2334 nfs_set_port(sap, &args->nfs_server.port, NFS_PORT);
2335 2335
2336 nfs_validate_transport_protocol(args); 2336 nfs_validate_transport_protocol(args);
2337 2337
@@ -2376,7 +2376,6 @@ static int nfs4_validate_mount_data(void *options,
2376 if (data == NULL) 2376 if (data == NULL)
2377 goto out_no_data; 2377 goto out_no_data;
2378 2378
2379 args->version = 4;
2380 switch (data->version) { 2379 switch (data->version) {
2381 case 1: 2380 case 1:
2382 if (data->host_addrlen > sizeof(args->nfs_server.address)) 2381 if (data->host_addrlen > sizeof(args->nfs_server.address))
@@ -2660,7 +2659,7 @@ static int nfs4_get_sb(struct file_system_type *fs_type,
2660 struct nfs_parsed_mount_data *data; 2659 struct nfs_parsed_mount_data *data;
2661 int error = -ENOMEM; 2660 int error = -ENOMEM;
2662 2661
2663 data = nfs_alloc_parsed_mount_data(0); 2662 data = nfs_alloc_parsed_mount_data(4);
2664 if (data == NULL) 2663 if (data == NULL)
2665 goto out_free_data; 2664 goto out_free_data;
2666 2665
@@ -2690,7 +2689,6 @@ static void nfs4_kill_super(struct super_block *sb)
2690 dprintk("--> %s\n", __func__); 2689 dprintk("--> %s\n", __func__);
2691 nfs_super_return_all_delegations(sb); 2690 nfs_super_return_all_delegations(sb);
2692 kill_anon_super(sb); 2691 kill_anon_super(sb);
2693 nfs4_renewd_prepare_shutdown(server);
2694 nfs_fscache_release_super_cookie(sb); 2692 nfs_fscache_release_super_cookie(sb);
2695 nfs_free_server(server); 2693 nfs_free_server(server);
2696 dprintk("<-- %s\n", __func__); 2694 dprintk("<-- %s\n", __func__);
diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
index 56013371f9f3..a44a7897fd4d 100644
--- a/fs/proc/kcore.c
+++ b/fs/proc/kcore.c
@@ -23,7 +23,6 @@
23#include <asm/io.h> 23#include <asm/io.h>
24#include <linux/list.h> 24#include <linux/list.h>
25#include <linux/ioport.h> 25#include <linux/ioport.h>
26#include <linux/mm.h>
27#include <linux/memory.h> 26#include <linux/memory.h>
28#include <asm/sections.h> 27#include <asm/sections.h>
29 28
diff --git a/fs/proc/page.c b/fs/proc/page.c
index 2281c2cbfe2b..5033ce0d254b 100644
--- a/fs/proc/page.c
+++ b/fs/proc/page.c
@@ -94,6 +94,7 @@ static const struct file_operations proc_kpagecount_operations = {
94#define KPF_COMPOUND_TAIL 16 94#define KPF_COMPOUND_TAIL 16
95#define KPF_HUGE 17 95#define KPF_HUGE 17
96#define KPF_UNEVICTABLE 18 96#define KPF_UNEVICTABLE 18
97#define KPF_HWPOISON 19
97#define KPF_NOPAGE 20 98#define KPF_NOPAGE 20
98 99
99#define KPF_KSM 21 100#define KPF_KSM 21
@@ -180,6 +181,10 @@ static u64 get_uflags(struct page *page)
180 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable); 181 u |= kpf_copy_bit(k, KPF_UNEVICTABLE, PG_unevictable);
181 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked); 182 u |= kpf_copy_bit(k, KPF_MLOCKED, PG_mlocked);
182 183
184#ifdef CONFIG_MEMORY_FAILURE
185 u |= kpf_copy_bit(k, KPF_HWPOISON, PG_hwpoison);
186#endif
187
183#ifdef CONFIG_IA64_UNCACHED_ALLOCATOR 188#ifdef CONFIG_IA64_UNCACHED_ALLOCATOR
184 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached); 189 u |= kpf_copy_bit(k, KPF_UNCACHED, PG_uncached);
185#endif 190#endif
diff --git a/fs/romfs/storage.c b/fs/romfs/storage.c
index b3208adf8e71..71e2b4d50a0a 100644
--- a/fs/romfs/storage.c
+++ b/fs/romfs/storage.c
@@ -253,11 +253,11 @@ ssize_t romfs_dev_strnlen(struct super_block *sb,
253 253
254#ifdef CONFIG_ROMFS_ON_MTD 254#ifdef CONFIG_ROMFS_ON_MTD
255 if (sb->s_mtd) 255 if (sb->s_mtd)
256 return romfs_mtd_strnlen(sb, pos, limit); 256 return romfs_mtd_strnlen(sb, pos, maxlen);
257#endif 257#endif
258#ifdef CONFIG_ROMFS_ON_BLOCK 258#ifdef CONFIG_ROMFS_ON_BLOCK
259 if (sb->s_bdev) 259 if (sb->s_bdev)
260 return romfs_blk_strnlen(sb, pos, limit); 260 return romfs_blk_strnlen(sb, pos, maxlen);
261#endif 261#endif
262 return -EIO; 262 return -EIO;
263} 263}
diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c
index 381854461b28..c2e30eea74dc 100644
--- a/fs/xfs/linux-2.6/xfs_aops.c
+++ b/fs/xfs/linux-2.6/xfs_aops.c
@@ -186,19 +186,37 @@ xfs_destroy_ioend(
186} 186}
187 187
188/* 188/*
189 * If the end of the current ioend is beyond the current EOF,
190 * return the new EOF value, otherwise zero.
191 */
192STATIC xfs_fsize_t
193xfs_ioend_new_eof(
194 xfs_ioend_t *ioend)
195{
196 xfs_inode_t *ip = XFS_I(ioend->io_inode);
197 xfs_fsize_t isize;
198 xfs_fsize_t bsize;
199
200 bsize = ioend->io_offset + ioend->io_size;
201 isize = MAX(ip->i_size, ip->i_new_size);
202 isize = MIN(isize, bsize);
203 return isize > ip->i_d.di_size ? isize : 0;
204}
205
206/*
189 * Update on-disk file size now that data has been written to disk. 207 * Update on-disk file size now that data has been written to disk.
190 * The current in-memory file size is i_size. If a write is beyond 208 * The current in-memory file size is i_size. If a write is beyond
191 * eof i_new_size will be the intended file size until i_size is 209 * eof i_new_size will be the intended file size until i_size is
192 * updated. If this write does not extend all the way to the valid 210 * updated. If this write does not extend all the way to the valid
193 * file size then restrict this update to the end of the write. 211 * file size then restrict this update to the end of the write.
194 */ 212 */
213
195STATIC void 214STATIC void
196xfs_setfilesize( 215xfs_setfilesize(
197 xfs_ioend_t *ioend) 216 xfs_ioend_t *ioend)
198{ 217{
199 xfs_inode_t *ip = XFS_I(ioend->io_inode); 218 xfs_inode_t *ip = XFS_I(ioend->io_inode);
200 xfs_fsize_t isize; 219 xfs_fsize_t isize;
201 xfs_fsize_t bsize;
202 220
203 ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG); 221 ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG);
204 ASSERT(ioend->io_type != IOMAP_READ); 222 ASSERT(ioend->io_type != IOMAP_READ);
@@ -206,16 +224,10 @@ xfs_setfilesize(
206 if (unlikely(ioend->io_error)) 224 if (unlikely(ioend->io_error))
207 return; 225 return;
208 226
209 bsize = ioend->io_offset + ioend->io_size;
210
211 xfs_ilock(ip, XFS_ILOCK_EXCL); 227 xfs_ilock(ip, XFS_ILOCK_EXCL);
212 228 isize = xfs_ioend_new_eof(ioend);
213 isize = MAX(ip->i_size, ip->i_new_size); 229 if (isize) {
214 isize = MIN(isize, bsize);
215
216 if (ip->i_d.di_size < isize) {
217 ip->i_d.di_size = isize; 230 ip->i_d.di_size = isize;
218 ip->i_update_core = 1;
219 xfs_mark_inode_dirty_sync(ip); 231 xfs_mark_inode_dirty_sync(ip);
220 } 232 }
221 233
@@ -404,10 +416,16 @@ xfs_submit_ioend_bio(
404 struct bio *bio) 416 struct bio *bio)
405{ 417{
406 atomic_inc(&ioend->io_remaining); 418 atomic_inc(&ioend->io_remaining);
407
408 bio->bi_private = ioend; 419 bio->bi_private = ioend;
409 bio->bi_end_io = xfs_end_bio; 420 bio->bi_end_io = xfs_end_bio;
410 421
422 /*
423 * If the I/O is beyond EOF we mark the inode dirty immediately
424 * but don't update the inode size until I/O completion.
425 */
426 if (xfs_ioend_new_eof(ioend))
427 xfs_mark_inode_dirty_sync(XFS_I(ioend->io_inode));
428
411 submit_bio(WRITE, bio); 429 submit_bio(WRITE, bio);
412 ASSERT(!bio_flagged(bio, BIO_EOPNOTSUPP)); 430 ASSERT(!bio_flagged(bio, BIO_EOPNOTSUPP));
413 bio_put(bio); 431 bio_put(bio);
diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
index 629370974e57..eff61e2732af 100644
--- a/fs/xfs/linux-2.6/xfs_file.c
+++ b/fs/xfs/linux-2.6/xfs_file.c
@@ -176,14 +176,7 @@ xfs_file_fsync(
176 struct dentry *dentry, 176 struct dentry *dentry,
177 int datasync) 177 int datasync)
178{ 178{
179 struct inode *inode = dentry->d_inode; 179 struct xfs_inode *ip = XFS_I(dentry->d_inode);
180 struct xfs_inode *ip = XFS_I(inode);
181 int error;
182
183 /* capture size updates in I/O completion before writing the inode. */
184 error = filemap_fdatawait(inode->i_mapping);
185 if (error)
186 return error;
187 180
188 xfs_iflags_clear(ip, XFS_ITRUNCATED); 181 xfs_iflags_clear(ip, XFS_ITRUNCATED);
189 return -xfs_fsync(ip); 182 return -xfs_fsync(ip);
diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
index da0159d99f82..cd42ef78f6b5 100644
--- a/fs/xfs/linux-2.6/xfs_iops.c
+++ b/fs/xfs/linux-2.6/xfs_iops.c
@@ -57,19 +57,22 @@
57#include <linux/fiemap.h> 57#include <linux/fiemap.h>
58 58
59/* 59/*
60 * Bring the atime in the XFS inode uptodate. 60 * Bring the timestamps in the XFS inode uptodate.
61 * Used before logging the inode to disk or when the Linux inode goes away. 61 *
62 * Used before writing the inode to disk.
62 */ 63 */
63void 64void
64xfs_synchronize_atime( 65xfs_synchronize_times(
65 xfs_inode_t *ip) 66 xfs_inode_t *ip)
66{ 67{
67 struct inode *inode = VFS_I(ip); 68 struct inode *inode = VFS_I(ip);
68 69
69 if (!(inode->i_state & I_CLEAR)) { 70 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
70 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec; 71 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
71 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec; 72 ip->i_d.di_ctime.t_sec = (__int32_t)inode->i_ctime.tv_sec;
72 } 73 ip->i_d.di_ctime.t_nsec = (__int32_t)inode->i_ctime.tv_nsec;
74 ip->i_d.di_mtime.t_sec = (__int32_t)inode->i_mtime.tv_sec;
75 ip->i_d.di_mtime.t_nsec = (__int32_t)inode->i_mtime.tv_nsec;
73} 76}
74 77
75/* 78/*
@@ -106,32 +109,20 @@ xfs_ichgtime(
106 if ((flags & XFS_ICHGTIME_MOD) && 109 if ((flags & XFS_ICHGTIME_MOD) &&
107 !timespec_equal(&inode->i_mtime, &tv)) { 110 !timespec_equal(&inode->i_mtime, &tv)) {
108 inode->i_mtime = tv; 111 inode->i_mtime = tv;
109 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
110 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
111 sync_it = 1; 112 sync_it = 1;
112 } 113 }
113 if ((flags & XFS_ICHGTIME_CHG) && 114 if ((flags & XFS_ICHGTIME_CHG) &&
114 !timespec_equal(&inode->i_ctime, &tv)) { 115 !timespec_equal(&inode->i_ctime, &tv)) {
115 inode->i_ctime = tv; 116 inode->i_ctime = tv;
116 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
117 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
118 sync_it = 1; 117 sync_it = 1;
119 } 118 }
120 119
121 /* 120 /*
122 * We update the i_update_core field _after_ changing 121 * Update complete - now make sure everyone knows that the inode
123 * the timestamps in order to coordinate properly with 122 * is dirty.
124 * xfs_iflush() so that we don't lose timestamp updates.
125 * This keeps us from having to hold the inode lock
126 * while doing this. We use the SYNCHRONIZE macro to
127 * ensure that the compiler does not reorder the update
128 * of i_update_core above the timestamp updates above.
129 */ 123 */
130 if (sync_it) { 124 if (sync_it)
131 SYNCHRONIZE();
132 ip->i_update_core = 1;
133 xfs_mark_inode_dirty_sync(ip); 125 xfs_mark_inode_dirty_sync(ip);
134 }
135} 126}
136 127
137/* 128/*
@@ -506,10 +497,8 @@ xfs_vn_getattr(
506 stat->gid = ip->i_d.di_gid; 497 stat->gid = ip->i_d.di_gid;
507 stat->ino = ip->i_ino; 498 stat->ino = ip->i_ino;
508 stat->atime = inode->i_atime; 499 stat->atime = inode->i_atime;
509 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec; 500 stat->mtime = inode->i_mtime;
510 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; 501 stat->ctime = inode->i_ctime;
511 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
512 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
513 stat->blocks = 502 stat->blocks =
514 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); 503 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
515 504
diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c
index 49e4a6aea73c..072050f8d346 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.c
+++ b/fs/xfs/linux-2.6/xfs_lrw.c
@@ -667,7 +667,7 @@ start:
667 xip->i_new_size = new_size; 667 xip->i_new_size = new_size;
668 668
669 if (likely(!(ioflags & IO_INVIS))) 669 if (likely(!(ioflags & IO_INVIS)))
670 xfs_ichgtime(xip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 670 file_update_time(file);
671 671
672 /* 672 /*
673 * If the offset is beyond the size of the file, we have a couple 673 * If the offset is beyond the size of the file, we have a couple
diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index bdd41c8c342f..18a4b8e11df2 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -977,6 +977,28 @@ xfs_fs_inode_init_once(
977} 977}
978 978
979/* 979/*
980 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
981 * we catch unlogged VFS level updates to the inode. Care must be taken
982 * here - the transaction code calls mark_inode_dirty_sync() to mark the
983 * VFS inode dirty in a transaction and clears the i_update_core field;
984 * it must clear the field after calling mark_inode_dirty_sync() to
985 * correctly indicate that the dirty state has been propagated into the
986 * inode log item.
987 *
988 * We need the barrier() to maintain correct ordering between unlogged
989 * updates and the transaction commit code that clears the i_update_core
990 * field. This requires all updates to be completed before marking the
991 * inode dirty.
992 */
993STATIC void
994xfs_fs_dirty_inode(
995 struct inode *inode)
996{
997 barrier();
998 XFS_I(inode)->i_update_core = 1;
999}
1000
1001/*
980 * Attempt to flush the inode, this will actually fail 1002 * Attempt to flush the inode, this will actually fail
981 * if the inode is pinned, but we dirty the inode again 1003 * if the inode is pinned, but we dirty the inode again
982 * at the point when it is unpinned after a log write, 1004 * at the point when it is unpinned after a log write,
@@ -1126,7 +1148,7 @@ xfs_fs_put_super(
1126} 1148}
1127 1149
1128STATIC int 1150STATIC int
1129xfs_fs_sync_super( 1151xfs_fs_sync_fs(
1130 struct super_block *sb, 1152 struct super_block *sb,
1131 int wait) 1153 int wait)
1132{ 1154{
@@ -1134,23 +1156,23 @@ xfs_fs_sync_super(
1134 int error; 1156 int error;
1135 1157
1136 /* 1158 /*
1137 * Treat a sync operation like a freeze. This is to work 1159 * Not much we can do for the first async pass. Writing out the
1138 * around a race in sync_inodes() which works in two phases 1160 * superblock would be counter-productive as we are going to redirty
1139 * - an asynchronous flush, which can write out an inode 1161 * when writing out other data and metadata (and writing out a single
1140 * without waiting for file size updates to complete, and a 1162 * block is quite fast anyway).
1141 * synchronous flush, which wont do anything because the 1163 *
1142 * async flush removed the inode's dirty flag. Also 1164 * Try to asynchronously kick off quota syncing at least.
1143 * sync_inodes() will not see any files that just have
1144 * outstanding transactions to be flushed because we don't
1145 * dirty the Linux inode until after the transaction I/O
1146 * completes.
1147 */ 1165 */
1148 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) 1166 if (!wait) {
1149 error = xfs_quiesce_data(mp); 1167 xfs_qm_sync(mp, SYNC_TRYLOCK);
1150 else 1168 return 0;
1151 error = xfs_sync_fsdata(mp, 0); 1169 }
1170
1171 error = xfs_quiesce_data(mp);
1172 if (error)
1173 return -error;
1152 1174
1153 if (unlikely(laptop_mode)) { 1175 if (laptop_mode) {
1154 int prev_sync_seq = mp->m_sync_seq; 1176 int prev_sync_seq = mp->m_sync_seq;
1155 1177
1156 /* 1178 /*
@@ -1169,7 +1191,7 @@ xfs_fs_sync_super(
1169 mp->m_sync_seq != prev_sync_seq); 1191 mp->m_sync_seq != prev_sync_seq);
1170 } 1192 }
1171 1193
1172 return -error; 1194 return 0;
1173} 1195}
1174 1196
1175STATIC int 1197STATIC int
@@ -1539,10 +1561,11 @@ xfs_fs_get_sb(
1539static const struct super_operations xfs_super_operations = { 1561static const struct super_operations xfs_super_operations = {
1540 .alloc_inode = xfs_fs_alloc_inode, 1562 .alloc_inode = xfs_fs_alloc_inode,
1541 .destroy_inode = xfs_fs_destroy_inode, 1563 .destroy_inode = xfs_fs_destroy_inode,
1564 .dirty_inode = xfs_fs_dirty_inode,
1542 .write_inode = xfs_fs_write_inode, 1565 .write_inode = xfs_fs_write_inode,
1543 .clear_inode = xfs_fs_clear_inode, 1566 .clear_inode = xfs_fs_clear_inode,
1544 .put_super = xfs_fs_put_super, 1567 .put_super = xfs_fs_put_super,
1545 .sync_fs = xfs_fs_sync_super, 1568 .sync_fs = xfs_fs_sync_fs,
1546 .freeze_fs = xfs_fs_freeze, 1569 .freeze_fs = xfs_fs_freeze,
1547 .statfs = xfs_fs_statfs, 1570 .statfs = xfs_fs_statfs,
1548 .remount_fs = xfs_fs_remount, 1571 .remount_fs = xfs_fs_remount,
diff --git a/fs/xfs/linux-2.6/xfs_sync.c b/fs/xfs/linux-2.6/xfs_sync.c
index 320be6aea492..961df0a22c78 100644
--- a/fs/xfs/linux-2.6/xfs_sync.c
+++ b/fs/xfs/linux-2.6/xfs_sync.c
@@ -309,11 +309,15 @@ xfs_sync_attr(
309STATIC int 309STATIC int
310xfs_commit_dummy_trans( 310xfs_commit_dummy_trans(
311 struct xfs_mount *mp, 311 struct xfs_mount *mp,
312 uint log_flags) 312 uint flags)
313{ 313{
314 struct xfs_inode *ip = mp->m_rootip; 314 struct xfs_inode *ip = mp->m_rootip;
315 struct xfs_trans *tp; 315 struct xfs_trans *tp;
316 int error; 316 int error;
317 int log_flags = XFS_LOG_FORCE;
318
319 if (flags & SYNC_WAIT)
320 log_flags |= XFS_LOG_SYNC;
317 321
318 /* 322 /*
319 * Put a dummy transaction in the log to tell recovery 323 * Put a dummy transaction in the log to tell recovery
@@ -331,13 +335,12 @@ xfs_commit_dummy_trans(
331 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); 335 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
332 xfs_trans_ihold(tp, ip); 336 xfs_trans_ihold(tp, ip);
333 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); 337 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
334 /* XXX(hch): ignoring the error here.. */
335 error = xfs_trans_commit(tp, 0); 338 error = xfs_trans_commit(tp, 0);
336
337 xfs_iunlock(ip, XFS_ILOCK_EXCL); 339 xfs_iunlock(ip, XFS_ILOCK_EXCL);
338 340
341 /* the log force ensures this transaction is pushed to disk */
339 xfs_log_force(mp, 0, log_flags); 342 xfs_log_force(mp, 0, log_flags);
340 return 0; 343 return error;
341} 344}
342 345
343int 346int
@@ -385,7 +388,20 @@ xfs_sync_fsdata(
385 else 388 else
386 XFS_BUF_ASYNC(bp); 389 XFS_BUF_ASYNC(bp);
387 390
388 return xfs_bwrite(mp, bp); 391 error = xfs_bwrite(mp, bp);
392 if (error)
393 return error;
394
395 /*
396 * If this is a data integrity sync make sure all pending buffers
397 * are flushed out for the log coverage check below.
398 */
399 if (flags & SYNC_WAIT)
400 xfs_flush_buftarg(mp->m_ddev_targp, 1);
401
402 if (xfs_log_need_covered(mp))
403 error = xfs_commit_dummy_trans(mp, flags);
404 return error;
389 405
390 out_brelse: 406 out_brelse:
391 xfs_buf_relse(bp); 407 xfs_buf_relse(bp);
@@ -419,14 +435,16 @@ xfs_quiesce_data(
419 /* push non-blocking */ 435 /* push non-blocking */
420 xfs_sync_data(mp, 0); 436 xfs_sync_data(mp, 0);
421 xfs_qm_sync(mp, SYNC_TRYLOCK); 437 xfs_qm_sync(mp, SYNC_TRYLOCK);
422 xfs_filestream_flush(mp);
423 438
424 /* push and block */ 439 /* push and block till complete */
425 xfs_sync_data(mp, SYNC_WAIT); 440 xfs_sync_data(mp, SYNC_WAIT);
426 xfs_qm_sync(mp, SYNC_WAIT); 441 xfs_qm_sync(mp, SYNC_WAIT);
427 442
443 /* drop inode references pinned by filestreams */
444 xfs_filestream_flush(mp);
445
428 /* write superblock and hoover up shutdown errors */ 446 /* write superblock and hoover up shutdown errors */
429 error = xfs_sync_fsdata(mp, 0); 447 error = xfs_sync_fsdata(mp, SYNC_WAIT);
430 448
431 /* flush data-only devices */ 449 /* flush data-only devices */
432 if (mp->m_rtdev_targp) 450 if (mp->m_rtdev_targp)
@@ -570,8 +588,6 @@ xfs_sync_worker(
570 /* dgc: errors ignored here */ 588 /* dgc: errors ignored here */
571 error = xfs_qm_sync(mp, SYNC_TRYLOCK); 589 error = xfs_qm_sync(mp, SYNC_TRYLOCK);
572 error = xfs_sync_fsdata(mp, SYNC_TRYLOCK); 590 error = xfs_sync_fsdata(mp, SYNC_TRYLOCK);
573 if (xfs_log_need_covered(mp))
574 error = xfs_commit_dummy_trans(mp, XFS_LOG_FORCE);
575 } 591 }
576 mp->m_sync_seq++; 592 mp->m_sync_seq++;
577 wake_up(&mp->m_wait_single_sync_task); 593 wake_up(&mp->m_wait_single_sync_task);
diff --git a/fs/xfs/xfs_dfrag.c b/fs/xfs/xfs_dfrag.c
index 7465f9ee125f..ab89a7e94a0f 100644
--- a/fs/xfs/xfs_dfrag.c
+++ b/fs/xfs/xfs_dfrag.c
@@ -206,10 +206,10 @@ xfs_swap_extents(
206 * process that the file was not changed out from 206 * process that the file was not changed out from
207 * under it. 207 * under it.
208 */ 208 */
209 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || 209 if ((sbp->bs_ctime.tv_sec != VFS_I(ip)->i_ctime.tv_sec) ||
210 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || 210 (sbp->bs_ctime.tv_nsec != VFS_I(ip)->i_ctime.tv_nsec) ||
211 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || 211 (sbp->bs_mtime.tv_sec != VFS_I(ip)->i_mtime.tv_sec) ||
212 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { 212 (sbp->bs_mtime.tv_nsec != VFS_I(ip)->i_mtime.tv_nsec)) {
213 error = XFS_ERROR(EBUSY); 213 error = XFS_ERROR(EBUSY);
214 goto out_unlock; 214 goto out_unlock;
215 } 215 }
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index fa913e459442..41ad537c49e9 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -854,6 +854,7 @@ xfs_dir2_leaf_getdents(
854 */ 854 */
855 ra_want = howmany(bufsize + mp->m_dirblksize, 855 ra_want = howmany(bufsize + mp->m_dirblksize,
856 mp->m_sb.sb_blocksize) - 1; 856 mp->m_sb.sb_blocksize) - 1;
857 ASSERT(ra_want >= 0);
857 858
858 /* 859 /*
859 * If we don't have as many as we want, and we haven't 860 * If we don't have as many as we want, and we haven't
@@ -1088,7 +1089,8 @@ xfs_dir2_leaf_getdents(
1088 */ 1089 */
1089 ptr += length; 1090 ptr += length;
1090 curoff += length; 1091 curoff += length;
1091 bufsize -= length; 1092 /* bufsize may have just been a guess; don't go negative */
1093 bufsize = bufsize > length ? bufsize - length : 0;
1092 } 1094 }
1093 1095
1094 /* 1096 /*
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index c1dc7ef5a1d8..b92a4fa2a0a1 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -3068,9 +3068,9 @@ xfs_iflush_int(
3068 SYNCHRONIZE(); 3068 SYNCHRONIZE();
3069 3069
3070 /* 3070 /*
3071 * Make sure to get the latest atime from the Linux inode. 3071 * Make sure to get the latest timestamps from the Linux inode.
3072 */ 3072 */
3073 xfs_synchronize_atime(ip); 3073 xfs_synchronize_times(ip);
3074 3074
3075 if (XFS_TEST_ERROR(be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC, 3075 if (XFS_TEST_ERROR(be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC,
3076 mp, XFS_ERRTAG_IFLUSH_1, XFS_RANDOM_IFLUSH_1)) { 3076 mp, XFS_ERRTAG_IFLUSH_1, XFS_RANDOM_IFLUSH_1)) {
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 0b38b9a869ec..41555de1d1db 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -504,7 +504,7 @@ void xfs_ichgtime(xfs_inode_t *, int);
504void xfs_lock_inodes(xfs_inode_t **, int, uint); 504void xfs_lock_inodes(xfs_inode_t **, int, uint);
505void xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint); 505void xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint);
506 506
507void xfs_synchronize_atime(xfs_inode_t *); 507void xfs_synchronize_times(xfs_inode_t *);
508void xfs_mark_inode_dirty_sync(xfs_inode_t *); 508void xfs_mark_inode_dirty_sync(xfs_inode_t *);
509 509
510#if defined(XFS_INODE_TRACE) 510#if defined(XFS_INODE_TRACE)
diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c
index 47d5b663c37e..9794b876d6ff 100644
--- a/fs/xfs/xfs_inode_item.c
+++ b/fs/xfs/xfs_inode_item.c
@@ -232,6 +232,15 @@ xfs_inode_item_format(
232 nvecs = 1; 232 nvecs = 1;
233 233
234 /* 234 /*
235 * Make sure the linux inode is dirty. We do this before
236 * clearing i_update_core as the VFS will call back into
237 * XFS here and set i_update_core, so we need to dirty the
238 * inode first so that the ordering of i_update_core and
239 * unlogged modifications still works as described below.
240 */
241 xfs_mark_inode_dirty_sync(ip);
242
243 /*
235 * Clear i_update_core if the timestamps (or any other 244 * Clear i_update_core if the timestamps (or any other
236 * non-transactional modification) need flushing/logging 245 * non-transactional modification) need flushing/logging
237 * and we're about to log them with the rest of the core. 246 * and we're about to log them with the rest of the core.
@@ -263,14 +272,9 @@ xfs_inode_item_format(
263 } 272 }
264 273
265 /* 274 /*
266 * Make sure to get the latest atime from the Linux inode. 275 * Make sure to get the latest timestamps from the Linux inode.
267 */ 276 */
268 xfs_synchronize_atime(ip); 277 xfs_synchronize_times(ip);
269
270 /*
271 * make sure the linux inode is dirty
272 */
273 xfs_mark_inode_dirty_sync(ip);
274 278
275 vecp->i_addr = (xfs_caddr_t)&ip->i_d; 279 vecp->i_addr = (xfs_caddr_t)&ip->i_d;
276 vecp->i_len = sizeof(struct xfs_icdinode); 280 vecp->i_len = sizeof(struct xfs_icdinode);
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index b68f9107e26c..62efab2f3839 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -59,6 +59,7 @@ xfs_bulkstat_one_iget(
59{ 59{
60 xfs_icdinode_t *dic; /* dinode core info pointer */ 60 xfs_icdinode_t *dic; /* dinode core info pointer */
61 xfs_inode_t *ip; /* incore inode pointer */ 61 xfs_inode_t *ip; /* incore inode pointer */
62 struct inode *inode;
62 int error; 63 int error;
63 64
64 error = xfs_iget(mp, NULL, ino, 65 error = xfs_iget(mp, NULL, ino,
@@ -72,6 +73,7 @@ xfs_bulkstat_one_iget(
72 ASSERT(ip->i_imap.im_blkno != 0); 73 ASSERT(ip->i_imap.im_blkno != 0);
73 74
74 dic = &ip->i_d; 75 dic = &ip->i_d;
76 inode = VFS_I(ip);
75 77
76 /* xfs_iget returns the following without needing 78 /* xfs_iget returns the following without needing
77 * further change. 79 * further change.
@@ -83,16 +85,19 @@ xfs_bulkstat_one_iget(
83 buf->bs_uid = dic->di_uid; 85 buf->bs_uid = dic->di_uid;
84 buf->bs_gid = dic->di_gid; 86 buf->bs_gid = dic->di_gid;
85 buf->bs_size = dic->di_size; 87 buf->bs_size = dic->di_size;
88
86 /* 89 /*
87 * We are reading the atime from the Linux inode because the 90 * We need to read the timestamps from the Linux inode because
88 * dinode might not be uptodate. 91 * the VFS keeps writing directly into the inode structure instead
92 * of telling us about the updates.
89 */ 93 */
90 buf->bs_atime.tv_sec = VFS_I(ip)->i_atime.tv_sec; 94 buf->bs_atime.tv_sec = inode->i_atime.tv_sec;
91 buf->bs_atime.tv_nsec = VFS_I(ip)->i_atime.tv_nsec; 95 buf->bs_atime.tv_nsec = inode->i_atime.tv_nsec;
92 buf->bs_mtime.tv_sec = dic->di_mtime.t_sec; 96 buf->bs_mtime.tv_sec = inode->i_mtime.tv_sec;
93 buf->bs_mtime.tv_nsec = dic->di_mtime.t_nsec; 97 buf->bs_mtime.tv_nsec = inode->i_mtime.tv_nsec;
94 buf->bs_ctime.tv_sec = dic->di_ctime.t_sec; 98 buf->bs_ctime.tv_sec = inode->i_ctime.tv_sec;
95 buf->bs_ctime.tv_nsec = dic->di_ctime.t_nsec; 99 buf->bs_ctime.tv_nsec = inode->i_ctime.tv_nsec;
100
96 buf->bs_xflags = xfs_ip2xflags(ip); 101 buf->bs_xflags = xfs_ip2xflags(ip);
97 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog; 102 buf->bs_extsize = dic->di_extsize << mp->m_sb.sb_blocklog;
98 buf->bs_extents = dic->di_nextents; 103 buf->bs_extents = dic->di_nextents;
diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
index a434f287962d..b572f7e840e0 100644
--- a/fs/xfs/xfs_vnodeops.c
+++ b/fs/xfs/xfs_vnodeops.c
@@ -2476,12 +2476,6 @@ xfs_reclaim(
2476 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0); 2476 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
2477 2477
2478 /* 2478 /*
2479 * Make sure the atime in the XFS inode is correct before freeing the
2480 * Linux inode.
2481 */
2482 xfs_synchronize_atime(ip);
2483
2484 /*
2485 * If we have nothing to flush with this inode then complete the 2479 * If we have nothing to flush with this inode then complete the
2486 * teardown now, otherwise break the link between the xfs inode and the 2480 * teardown now, otherwise break the link between the xfs inode and the
2487 * linux inode and clean up the xfs inode later. This avoids flushing 2481 * linux inode and clean up the xfs inode later. This avoids flushing
generated by cgit v1.2.2 (git 2.25.0) at 2025-12-30 01:07:04 -0500