forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-05-10 093a6c67005148ae32a5c9e4553491b9f5c2457b 
 kernel/fs/btrfs/btrfs_inode.h
....@@ -7,6 +7,7 @@
77 #define BTRFS_INODE_H
88 
99 #include <linux/hash.h>
10
+#include <linux/refcount.h>
1011 #include "extent_map.h"
1112 #include "extent_io.h"
1213 #include "ordered-data.h"
....@@ -20,16 +21,36 @@
2021  * new data the application may have written before commit.
2122  */
2223 enum {
23
-	BTRFS_INODE_ORDERED_DATA_CLOSE = 0,
24
+	BTRFS_INODE_FLUSH_ON_CLOSE,
2425 	BTRFS_INODE_DUMMY,
2526 	BTRFS_INODE_IN_DEFRAG,
2627 	BTRFS_INODE_HAS_ASYNC_EXTENT,
28
+	 /*
29
+	  * Always set under the VFS' inode lock, otherwise it can cause races
30
+	  * during fsync (we start as a fast fsync and then end up in a full
31
+	  * fsync racing with ordered extent completion).
32
+	  */
2733 	BTRFS_INODE_NEEDS_FULL_SYNC,
2834 	BTRFS_INODE_COPY_EVERYTHING,
2935 	BTRFS_INODE_IN_DELALLOC_LIST,
30
-	BTRFS_INODE_READDIO_NEED_LOCK,
3136 	BTRFS_INODE_HAS_PROPS,
3237 	BTRFS_INODE_SNAPSHOT_FLUSH,
38
+	/*
39
+	 * Set and used when logging an inode and it serves to signal that an
40
+	 * inode does not have xattrs, so subsequent fsyncs can avoid searching
41
+	 * for xattrs to log. This bit must be cleared whenever a xattr is added
42
+	 * to an inode.
43
+	 */
44
+	BTRFS_INODE_NO_XATTRS,
45
+	/*
46
+	 * Set when we are in a context where we need to start a transaction and
47
+	 * have dirty pages with the respective file range locked. This is to
48
+	 * ensure that when reserving space for the transaction, if we are low
49
+	 * on available space and need to flush delalloc, we will not flush
50
+	 * delalloc for this inode, because that could result in a deadlock (on
51
+	 * the file range, inode's io_tree).
52
+	 */
53
+	BTRFS_INODE_NO_DELALLOC_FLUSH,
3354 };
3455 
3556 /* in memory btrfs inode */
....@@ -60,11 +81,14 @@
6081 	 */
6182 	struct extent_io_tree io_failure_tree;
6283 
84
+	/*
85
+	 * Keep track of where the inode has extent items mapped in order to
86
+	 * make sure the i_size adjustments are accurate
87
+	 */
88
+	struct extent_io_tree file_extent_tree;
89
+
6390 	/* held while logging the inode in tree-log.c */
6491 	struct mutex log_mutex;
65
-
66
-	/* held while doing delalloc reservations */
67
-	struct mutex delalloc_mutex;
6892 
6993 	/* used to order data wrt metadata */
7094 	struct btrfs_ordered_inode_tree ordered_tree;
....@@ -148,10 +172,15 @@
148172 	u64 last_unlink_trans;
149173 
150174 	/*
151
-	 * Track the transaction id of the last transaction used to create a
152
-	 * hard link for the inode. This is used by the log tree (fsync).
175
+	 * The id/generation of the last transaction where this inode was
176
+	 * either the source or the destination of a clone/dedupe operation.
177
+	 * Used when logging an inode to know if there are shared extents that
178
+	 * need special care when logging checksum items, to avoid duplicate
179
+	 * checksum items in a log (which can lead to a corruption where we end
180
+	 * up with missing checksum ranges after log replay).
181
+	 * Protected by the vfs inode lock.
153182 	 */
154
-	u64 last_link_trans;
183
+	u64 last_reflink_trans;
155184 
156185 	/*
157186 	 * Number of bytes outstanding that are going to need csums.  This is
....@@ -203,7 +232,10 @@
203232 	struct inode vfs_inode;
204233 };
205234 
206
-extern unsigned char btrfs_filetype_table[];
235
+static inline u32 btrfs_inode_sectorsize(const struct btrfs_inode *inode)
236
+{
237
+	return inode->root->fs_info->sectorsize;
238
+}
207239 
208240 static inline struct btrfs_inode *BTRFS_I(const struct inode *inode)
209241 {
....@@ -213,7 +245,7 @@
213245 static inline unsigned long btrfs_inode_hash(u64 objectid,
214246 					     const struct btrfs_root *root)
215247 {
216
-	u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
248
+	u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);
217249 
218250 #if BITS_PER_LONG == 32
219251 	h = (h >> 32) ^ (h & 0xffffffff);
....@@ -260,6 +292,11 @@
260292 	return false;
261293 }
262294 
295
+static inline bool is_data_inode(struct inode *inode)
296
+{
297
+	return btrfs_ino(BTRFS_I(inode)) != BTRFS_BTREE_INODE_OBJECTID;
298
+}
299
+
263300 static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,
264301 						 int mod)
265302 {
....@@ -269,6 +306,21 @@
269306 		return;
270307 	trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),
271308 						  mod);
309
+}
310
+
311
+/*
312
+ * Called every time after doing a buffered, direct IO or memory mapped write.
313
+ *
314
+ * This is to ensure that if we write to a file that was previously fsynced in
315
+ * the current transaction, then try to fsync it again in the same transaction,
316
+ * we will know that there were changes in the file and that it needs to be
317
+ * logged.
318
+ */
319
+static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)
320
+{
321
+	spin_lock(&inode->lock);
322
+	inode->last_sub_trans = inode->root->log_transid;
323
+	spin_unlock(&inode->lock);
272324 }
273325 
274326 static inline int btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)
....@@ -293,69 +345,53 @@
293345 	return ret;
294346 }
295347 
296
-#define BTRFS_DIO_ORIG_BIO_SUBMITTED	0x1
297
-
298348 struct btrfs_dio_private {
299349 	struct inode *inode;
300
-	unsigned long flags;
301350 	u64 logical_offset;
302351 	u64 disk_bytenr;
303352 	u64 bytes;
304
-	void *private;
305353 
306
-	/* number of bios pending for this dio */
307
-	atomic_t pending_bios;
308
-
309
-	/* IO errors */
310
-	int errors;
311
-
312
-	/* orig_bio is our btrfs_io_bio */
313
-	struct bio *orig_bio;
354
+	/*
355
+	 * References to this structure. There is one reference per in-flight
356
+	 * bio plus one while we're still setting up.
357
+	 */
358
+	refcount_t refs;
314359 
315360 	/* dio_bio came from fs/direct-io.c */
316361 	struct bio *dio_bio;
317362 
318
-	/*
319
-	 * The original bio may be split to several sub-bios, this is
320
-	 * done during endio of sub-bios
321
-	 */
322
-	blk_status_t (*subio_endio)(struct inode *, struct btrfs_io_bio *,
323
-			blk_status_t);
363
+	/* Array of checksums */
364
+	u8 csums[];
324365 };
325366 
326
-/*
327
- * Disable DIO read nolock optimization, so new dio readers will be forced
328
- * to grab i_mutex. It is used to avoid the endless truncate due to
329
- * nonlocked dio read.
330
- */
331
-static inline void btrfs_inode_block_unlocked_dio(struct btrfs_inode *inode)
332
-{
333
-	set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
334
-	smp_mb();
335
-}
336
-
337
-static inline void btrfs_inode_resume_unlocked_dio(struct btrfs_inode *inode)
338
-{
339
-	smp_mb__before_atomic();
340
-	clear_bit(BTRFS_INODE_READDIO_NEED_LOCK, &inode->runtime_flags);
341
-}
367
+/* Array of bytes with variable length, hexadecimal format 0x1234 */
368
+#define CSUM_FMT				"0x%*phN"
369
+#define CSUM_FMT_VALUE(size, bytes)		size, bytes
342370 
343371 static inline void btrfs_print_data_csum_error(struct btrfs_inode *inode,
344
-		u64 logical_start, u32 csum, u32 csum_expected, int mirror_num)
372
+		u64 logical_start, u8 *csum, u8 *csum_expected, int mirror_num)
345373 {
346374 	struct btrfs_root *root = inode->root;
375
+	struct btrfs_super_block *sb = root->fs_info->super_copy;
376
+	const u16 csum_size = btrfs_super_csum_size(sb);
347377 
348378 	/* Output minus objectid, which is more meaningful */
349
-	if (root->objectid >= BTRFS_LAST_FREE_OBJECTID)
379
+	if (root->root_key.objectid >= BTRFS_LAST_FREE_OBJECTID)
350380 		btrfs_warn_rl(root->fs_info,
351
-	"csum failed root %lld ino %lld off %llu csum 0x%08x expected csum 0x%08x mirror %d",
352
-			root->objectid, btrfs_ino(inode),
353
-			logical_start, csum, csum_expected, mirror_num);
381
+"csum failed root %lld ino %lld off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
382
+			root->root_key.objectid, btrfs_ino(inode),
383
+			logical_start,
384
+			CSUM_FMT_VALUE(csum_size, csum),
385
+			CSUM_FMT_VALUE(csum_size, csum_expected),
386
+			mirror_num);
354387 	else
355388 		btrfs_warn_rl(root->fs_info,
356
-	"csum failed root %llu ino %llu off %llu csum 0x%08x expected csum 0x%08x mirror %d",
357
-			root->objectid, btrfs_ino(inode),
358
-			logical_start, csum, csum_expected, mirror_num);
389
+"csum failed root %llu ino %llu off %llu csum " CSUM_FMT " expected csum " CSUM_FMT " mirror %d",
390
+			root->root_key.objectid, btrfs_ino(inode),
391
+			logical_start,
392
+			CSUM_FMT_VALUE(csum_size, csum),
393
+			CSUM_FMT_VALUE(csum_size, csum_expected),
394
+			mirror_num);
359395 }
360396 
361397 #endif