xref: /openbmc/linux/fs/btrfs/ordered-data.h (revision 31e67366)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #ifndef BTRFS_ORDERED_DATA_H
7 #define BTRFS_ORDERED_DATA_H
8 
9 /* one of these per inode */
10 struct btrfs_ordered_inode_tree {
11 	spinlock_t lock;
12 	struct rb_root tree;
13 	struct rb_node *last;
14 };
15 
16 struct btrfs_ordered_sum {
17 	/* bytenr is the start of this extent on disk */
18 	u64 bytenr;
19 
20 	/*
21 	 * this is the length in bytes covered by the sums array below.
22 	 */
23 	int len;
24 	struct list_head list;
25 	/* last field is a variable length array of csums */
26 	u8 sums[];
27 };
28 
29 /*
30  * Bits for btrfs_ordered_extent::flags.
31  *
32  * BTRFS_ORDERED_IO_DONE is set when all of the blocks are written.
33  * It is used to make sure metadata is inserted into the tree only once
34  * per extent.
35  *
36  * BTRFS_ORDERED_COMPLETE is set when the extent is removed from the
37  * rbtree, just before waking any waiters.  It is used to indicate the
38  * IO is done and any metadata is inserted into the tree.
39  */
40 enum {
41 	/*
42 	 * Different types for direct io, one and only one of the 4 type can
43 	 * be set when creating ordered extent.
44 	 *
45 	 * REGULAR:	For regular non-compressed COW write
46 	 * NOCOW:	For NOCOW write into existing non-hole extent
47 	 * PREALLOC:	For NOCOW write into preallocated extent
48 	 * COMPRESSED:	For compressed COW write
49 	 */
50 	BTRFS_ORDERED_REGULAR,
51 	BTRFS_ORDERED_NOCOW,
52 	BTRFS_ORDERED_PREALLOC,
53 	BTRFS_ORDERED_COMPRESSED,
54 
55 	/*
56 	 * Extra bit for direct io, can only be set for
57 	 * REGULAR/NOCOW/PREALLOC. No direct io for compressed extent.
58 	 */
59 	BTRFS_ORDERED_DIRECT,
60 
61 	/* Extra status bits for ordered extents */
62 
63 	/* set when all the pages are written */
64 	BTRFS_ORDERED_IO_DONE,
65 	/* set when removed from the tree */
66 	BTRFS_ORDERED_COMPLETE,
67 	/* We had an io error when writing this out */
68 	BTRFS_ORDERED_IOERR,
69 	/* Set when we have to truncate an extent */
70 	BTRFS_ORDERED_TRUNCATED,
71 	/* Used during fsync to track already logged extents */
72 	BTRFS_ORDERED_LOGGED,
73 	/* We have already logged all the csums of the ordered extent */
74 	BTRFS_ORDERED_LOGGED_CSUM,
75 	/* We wait for this extent to complete in the current transaction */
76 	BTRFS_ORDERED_PENDING,
77 };
78 
79 struct btrfs_ordered_extent {
80 	/* logical offset in the file */
81 	u64 file_offset;
82 
83 	/*
84 	 * These fields directly correspond to the same fields in
85 	 * btrfs_file_extent_item.
86 	 */
87 	u64 disk_bytenr;
88 	u64 num_bytes;
89 	u64 disk_num_bytes;
90 
91 	/* number of bytes that still need writing */
92 	u64 bytes_left;
93 
94 	/*
95 	 * the end of the ordered extent which is behind it but
96 	 * didn't update disk_i_size. Please see the comment of
97 	 * btrfs_ordered_update_i_size();
98 	 */
99 	u64 outstanding_isize;
100 
101 	/*
102 	 * If we get truncated we need to adjust the file extent we enter for
103 	 * this ordered extent so that we do not expose stale data.
104 	 */
105 	u64 truncated_len;
106 
107 	/* flags (described above) */
108 	unsigned long flags;
109 
110 	/* compression algorithm */
111 	int compress_type;
112 
113 	/* Qgroup reserved space */
114 	int qgroup_rsv;
115 
116 	/* reference count */
117 	refcount_t refs;
118 
119 	/* the inode we belong to */
120 	struct inode *inode;
121 
122 	/* list of checksums for insertion when the extent io is done */
123 	struct list_head list;
124 
125 	/* used for fast fsyncs */
126 	struct list_head log_list;
127 
128 	/* used to wait for the BTRFS_ORDERED_COMPLETE bit */
129 	wait_queue_head_t wait;
130 
131 	/* our friendly rbtree entry */
132 	struct rb_node rb_node;
133 
134 	/* a per root list of all the pending ordered extents */
135 	struct list_head root_extent_list;
136 
137 	struct btrfs_work work;
138 
139 	struct completion completion;
140 	struct btrfs_work flush_work;
141 	struct list_head work_list;
142 
143 	/*
144 	 * Used to reverse-map physical address returned from ZONE_APPEND write
145 	 * command in a workqueue context
146 	 */
147 	u64 physical;
148 	struct gendisk *disk;
149 	u8 partno;
150 };
151 
152 /*
153  * calculates the total size you need to allocate for an ordered sum
154  * structure spanning 'bytes' in the file
155  */
156 static inline int btrfs_ordered_sum_size(struct btrfs_fs_info *fs_info,
157 					 unsigned long bytes)
158 {
159 	int num_sectors = (int)DIV_ROUND_UP(bytes, fs_info->sectorsize);
160 
161 	return sizeof(struct btrfs_ordered_sum) + num_sectors * fs_info->csum_size;
162 }
163 
164 static inline void
165 btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t)
166 {
167 	spin_lock_init(&t->lock);
168 	t->tree = RB_ROOT;
169 	t->last = NULL;
170 }
171 
172 void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry);
173 void btrfs_remove_ordered_extent(struct btrfs_inode *btrfs_inode,
174 				struct btrfs_ordered_extent *entry);
175 bool btrfs_dec_test_ordered_pending(struct btrfs_inode *inode,
176 				    struct btrfs_ordered_extent **cached,
177 				    u64 file_offset, u64 io_size, int uptodate);
178 bool btrfs_dec_test_first_ordered_pending(struct btrfs_inode *inode,
179 				   struct btrfs_ordered_extent **finished_ret,
180 				   u64 *file_offset, u64 io_size,
181 				   int uptodate);
182 int btrfs_add_ordered_extent(struct btrfs_inode *inode, u64 file_offset,
183 			     u64 disk_bytenr, u64 num_bytes, u64 disk_num_bytes,
184 			     int type);
185 int btrfs_add_ordered_extent_dio(struct btrfs_inode *inode, u64 file_offset,
186 				 u64 disk_bytenr, u64 num_bytes,
187 				 u64 disk_num_bytes, int type);
188 int btrfs_add_ordered_extent_compress(struct btrfs_inode *inode, u64 file_offset,
189 				      u64 disk_bytenr, u64 num_bytes,
190 				      u64 disk_num_bytes, int compress_type);
191 void btrfs_add_ordered_sum(struct btrfs_ordered_extent *entry,
192 			   struct btrfs_ordered_sum *sum);
193 struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct btrfs_inode *inode,
194 							 u64 file_offset);
195 void btrfs_start_ordered_extent(struct btrfs_ordered_extent *entry, int wait);
196 int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len);
197 struct btrfs_ordered_extent *
198 btrfs_lookup_first_ordered_extent(struct btrfs_inode *inode, u64 file_offset);
199 struct btrfs_ordered_extent *btrfs_lookup_ordered_range(
200 		struct btrfs_inode *inode,
201 		u64 file_offset,
202 		u64 len);
203 void btrfs_get_ordered_extents_for_logging(struct btrfs_inode *inode,
204 					   struct list_head *list);
205 u64 btrfs_wait_ordered_extents(struct btrfs_root *root, u64 nr,
206 			       const u64 range_start, const u64 range_len);
207 void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, u64 nr,
208 			      const u64 range_start, const u64 range_len);
209 void btrfs_lock_and_flush_ordered_range(struct btrfs_inode *inode, u64 start,
210 					u64 end,
211 					struct extent_state **cached_state);
212 int btrfs_split_ordered_extent(struct btrfs_ordered_extent *ordered, u64 pre,
213 			       u64 post);
214 int __init ordered_data_init(void);
215 void __cold ordered_data_exit(void);
216 
217 #endif
218