forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-10-09 244b2c5ca8b14627e4a17755e5922221e121c771 
 kernel/fs/incfs/data_mgmt.h
....@@ -10,49 +10,71 @@
1010 #include <linux/types.h>
1111 #include <linux/mutex.h>
1212 #include <linux/spinlock.h>
13
+#include <linux/rcupdate.h>
1314 #include <linux/completion.h>
1415 #include <linux/wait.h>
16
+#include <linux/zstd.h>
1517 #include <crypto/hash.h>
18
+#include <linux/rwsem.h>
1619 
1720 #include <uapi/linux/incrementalfs.h>
1821 
1922 #include "internal.h"
23
+#include "pseudo_files.h"
2024 
2125 #define SEGMENTS_PER_FILE 3
2226 
2327 enum LOG_RECORD_TYPE {
2428 	FULL,
2529 	SAME_FILE,
30
+	SAME_FILE_CLOSE_BLOCK,
31
+	SAME_FILE_CLOSE_BLOCK_SHORT,
2632 	SAME_FILE_NEXT_BLOCK,
2733 	SAME_FILE_NEXT_BLOCK_SHORT,
2834 };
2935 
3036 struct full_record {
31
-	enum LOG_RECORD_TYPE type : 2; /* FULL */
32
-	u32 block_index : 30;
37
+	enum LOG_RECORD_TYPE type : 3; /* FULL */
38
+	u32 block_index : 29;
3339 	incfs_uuid_t file_id;
3440 	u64 absolute_ts_us;
35
-} __packed; /* 28 bytes */
41
+	uid_t uid;
42
+} __packed; /* 32 bytes */
3643 
37
-struct same_file_record {
38
-	enum LOG_RECORD_TYPE type : 2; /* SAME_FILE */
39
-	u32 block_index : 30;
40
-	u32 relative_ts_us; /* max 2^32 us ~= 1 hour (1:11:30) */
41
-} __packed; /* 12 bytes */
44
+struct same_file {
45
+	enum LOG_RECORD_TYPE type : 3; /* SAME_FILE */
46
+	u32 block_index : 29;
47
+	uid_t uid;
48
+	u16 relative_ts_us; /* max 2^16 us ~= 64 ms */
49
+} __packed; /* 10 bytes */
4250 
43
-struct same_file_next_block {
44
-	enum LOG_RECORD_TYPE type : 2; /* SAME_FILE_NEXT_BLOCK */
45
-	u32 relative_ts_us : 30; /* max 2^30 us ~= 15 min (17:50) */
51
+struct same_file_close_block {
52
+	enum LOG_RECORD_TYPE type : 3; /* SAME_FILE_CLOSE_BLOCK */
53
+	u16 relative_ts_us : 13; /* max 2^13 us ~= 8 ms */
54
+	s16 block_index_delta;
4655 } __packed; /* 4 bytes */
4756 
48
-struct same_file_next_block_short {
49
-	enum LOG_RECORD_TYPE type : 2; /* SAME_FILE_NEXT_BLOCK_SHORT */
50
-	u16 relative_ts_us : 14; /* max 2^14 us ~= 16 ms */
57
+struct same_file_close_block_short {
58
+	enum LOG_RECORD_TYPE type : 3; /* SAME_FILE_CLOSE_BLOCK_SHORT */
59
+	u8 relative_ts_tens_us : 5; /* max 2^5*10 us ~= 320 us */
60
+	s8 block_index_delta;
5161 } __packed; /* 2 bytes */
62
+
63
+struct same_file_next_block {
64
+	enum LOG_RECORD_TYPE type : 3; /* SAME_FILE_NEXT_BLOCK */
65
+	u16 relative_ts_us : 13; /* max 2^13 us ~= 8 ms */
66
+} __packed; /* 2 bytes */
67
+
68
+struct same_file_next_block_short {
69
+	enum LOG_RECORD_TYPE type : 3; /* SAME_FILE_NEXT_BLOCK_SHORT */
70
+	u8 relative_ts_tens_us : 5; /* max 2^5*10 us ~= 320 us */
71
+} __packed; /* 1 byte */
5272 
5373 union log_record {
5474 	struct full_record full_record;
55
-	struct same_file_record same_file_record;
75
+	struct same_file same_file;
76
+	struct same_file_close_block same_file_close_block;
77
+	struct same_file_close_block_short same_file_close_block_short;
5678 	struct same_file_next_block same_file_next_block;
5779 	struct same_file_next_block_short same_file_next_block_short;
5880 };
....@@ -99,8 +121,8 @@
99121 	unsigned int readahead_pages;
100122 	unsigned int read_log_pages;
101123 	unsigned int read_log_wakeup_count;
102
-	bool no_backing_file_cache;
103
-	bool no_backing_file_readahead;
124
+	bool report_uid;
125
+	char *sysfs_name;
104126 };
105127 
106128 struct mount_info {
....@@ -109,6 +131,14 @@
109131 	struct path mi_backing_dir_path;
110132 
111133 	struct dentry *mi_index_dir;
134
+	/* For stacking mounts, if true, this indicates if the index dir needs
135
+	 * to be freed for this SB otherwise it was created by lower level SB */
136
+	bool mi_index_free;
137
+
138
+	struct dentry *mi_incomplete_dir;
139
+	/* For stacking mounts, if true, this indicates if the incomplete dir
140
+	 * needs to be freed for this SB. Similar to mi_index_free */
141
+	bool mi_incomplete_free;
112142 
113143 	const struct cred *mi_owner;
114144 
....@@ -123,13 +153,13 @@
123153 	wait_queue_head_t mi_pending_reads_notif_wq;
124154 
125155 	/*
126
-	 * Protects:
156
+	 * Protects - RCU safe:
127157 	 *  - reads_list_head
128158 	 *  - mi_pending_reads_count
129159 	 *  - mi_last_pending_read_number
130160 	 *  - data_file_segment.reads_list_head
131161 	 */
132
-	struct mutex mi_pending_reads_mutex;
162
+	spinlock_t pending_read_lock;
133163 
134164 	/* List of active pending_read objects */
135165 	struct list_head mi_reads_list_head;
....@@ -146,11 +176,67 @@
146176 	/* Temporary buffer for read logger. */
147177 	struct read_log mi_log;
148178 
149
-	void *log_xattr;
150
-	size_t log_xattr_size;
179
+	/* SELinux needs special xattrs on our pseudo files */
180
+	struct mem_range pseudo_file_xattr[PSEUDO_FILE_COUNT];
151181 
152
-	void *pending_read_xattr;
153
-	size_t pending_read_xattr_size;
182
+	/* A queue of waiters who want to be notified about blocks_written */
183
+	wait_queue_head_t mi_blocks_written_notif_wq;
184
+
185
+	/* Number of blocks written since mount */
186
+	atomic_t mi_blocks_written;
187
+
188
+	/* Per UID read timeouts */
189
+	spinlock_t mi_per_uid_read_timeouts_lock;
190
+	struct incfs_per_uid_read_timeouts *mi_per_uid_read_timeouts;
191
+	int mi_per_uid_read_timeouts_size;
192
+
193
+	/* zstd workspace */
194
+	struct mutex mi_zstd_workspace_mutex;
195
+	void *mi_zstd_workspace;
196
+	ZSTD_DStream *mi_zstd_stream;
197
+	struct delayed_work mi_zstd_cleanup_work;
198
+
199
+	/* sysfs node */
200
+	struct incfs_sysfs_node *mi_sysfs_node;
201
+
202
+	/* Last error information */
203
+	struct mutex	mi_le_mutex;
204
+	incfs_uuid_t	mi_le_file_id;
205
+	u64		mi_le_time_us;
206
+	u32		mi_le_page;
207
+	u32		mi_le_errno;
208
+	uid_t		mi_le_uid;
209
+
210
+	/* Number of reads timed out */
211
+	u32 mi_reads_failed_timed_out;
212
+
213
+	/* Number of reads failed because hash verification failed */
214
+	u32 mi_reads_failed_hash_verification;
215
+
216
+	/* Number of reads failed for another reason */
217
+	u32 mi_reads_failed_other;
218
+
219
+	/* Number of reads delayed because page had to be fetched */
220
+	u32 mi_reads_delayed_pending;
221
+
222
+	/* Total time waiting for pages to be fetched */
223
+	u64 mi_reads_delayed_pending_us;
224
+
225
+	/*
226
+	 * Number of reads delayed because of per-uid min_time_us or
227
+	 * min_pending_time_us settings
228
+	 */
229
+	u32 mi_reads_delayed_min;
230
+
231
+	/* Total time waiting because of per-uid min_time_us or
232
+	 * min_pending_time_us settings.
233
+	 *
234
+	 * Note that if a read is initially delayed because we have to wait for
235
+	 * the page, then further delayed because of min_pending_time_us
236
+	 * setting, this counter gets incremented by only the further delay
237
+	 * time.
238
+	 */
239
+	u64 mi_reads_delayed_min_us;
154240 };
155241 
156242 struct data_file_block {
....@@ -172,17 +258,20 @@
172258 
173259 	int serial_number;
174260 
261
+	uid_t uid;
262
+
175263 	struct list_head mi_reads_list;
176264 
177265 	struct list_head segment_reads_list;
266
+
267
+	struct rcu_head rcu;
178268 };
179269 
180270 struct data_file_segment {
181271 	wait_queue_head_t new_data_arrival_wq;
182272 
183273 	/* Protects reads and writes from the blockmap */
184
-	/* Good candidate for read/write mutex */
185
-	struct mutex blockmap_mutex;
274
+	struct rw_semaphore rwsem;
186275 
187276 	/* List of active pending_read objects belonging to this segment */
188277 	/* Protected by mount_info.pending_reads_mutex */
....@@ -232,11 +321,50 @@
232321 	/* Total number of blocks, data + hash */
233322 	int df_total_block_count;
234323 
235
-	struct file_attr n_attr;
324
+	/* For mapped files, the offset into the actual file */
325
+	loff_t df_mapped_offset;
236326 
327
+	/* Number of data blocks written to file */
328
+	atomic_t df_data_blocks_written;
329
+
330
+	/* Number of data blocks in the status block */
331
+	u32 df_initial_data_blocks_written;
332
+
333
+	/* Number of hash blocks written to file */
334
+	atomic_t df_hash_blocks_written;
335
+
336
+	/* Number of hash blocks in the status block */
337
+	u32 df_initial_hash_blocks_written;
338
+
339
+	/* Offset to status metadata header */
340
+	loff_t df_status_offset;
341
+
342
+	/*
343
+	 * Mutex acquired while enabling verity. Note that df_hash_tree is set
344
+	 * by enable verity.
345
+	 *
346
+	 * The backing file mutex bc_mutex  may be taken while this mutex is
347
+	 * held.
348
+	 */
349
+	struct mutex df_enable_verity;
350
+
351
+	/*
352
+	 * Set either at construction time or during enabling verity. In the
353
+	 * latter case, set via smp_store_release, so use smp_load_acquire to
354
+	 * read it.
355
+	 */
237356 	struct mtree *df_hash_tree;
238357 
358
+	/* Guaranteed set if df_hash_tree is set. */
239359 	struct incfs_df_signature *df_signature;
360
+
361
+	/*
362
+	 * The verity file digest, set when verity is enabled and the file has
363
+	 * been opened
364
+	 */
365
+	struct mem_range df_verity_file_digest;
366
+
367
+	struct incfs_df_verity_signature *df_verity_signature;
240368 };
241369 
242370 struct dir_file {
....@@ -259,6 +387,23 @@
259387 	struct path backing_path;
260388 };
261389 
390
+enum FILL_PERMISSION {
391
+	CANT_FILL = 0,
392
+	CAN_FILL = 1,
393
+};
394
+
395
+struct incfs_file_data {
396
+	/* Does this file handle have INCFS_IOC_FILL_BLOCKS permission */
397
+	enum FILL_PERMISSION fd_fill_permission;
398
+
399
+	/* If INCFS_IOC_GET_FILLED_BLOCKS has been called, where are we */
400
+	int fd_get_block_pos;
401
+
402
+	/* And how many filled blocks are there up to that point */
403
+	int fd_filled_data_blocks;
404
+	int fd_filled_hash_blocks;
405
+};
406
+
262407 struct mount_info *incfs_alloc_mount_info(struct super_block *sb,
263408 					  struct mount_options *options,
264409 					  struct path *backing_dir_path);
....@@ -268,25 +413,37 @@
268413 
269414 void incfs_free_mount_info(struct mount_info *mi);
270415 
416
+char *file_id_to_str(incfs_uuid_t id);
417
+struct dentry *incfs_lookup_dentry(struct dentry *parent, const char *name);
271418 struct data_file *incfs_open_data_file(struct mount_info *mi, struct file *bf);
272419 void incfs_free_data_file(struct data_file *df);
273
-
274
-int incfs_scan_metadata_chain(struct data_file *df);
275420 
276421 struct dir_file *incfs_open_dir_file(struct mount_info *mi, struct file *bf);
277422 void incfs_free_dir_file(struct dir_file *dir);
278423 
424
+struct incfs_read_data_file_timeouts {
425
+	u32 min_time_us;
426
+	u32 min_pending_time_us;
427
+	u32 max_pending_time_us;
428
+};
429
+
279430 ssize_t incfs_read_data_file_block(struct mem_range dst, struct file *f,
280
-				   int index, int timeout_ms,
281
-				   struct mem_range tmp);
431
+			int index, struct mem_range tmp,
432
+			struct incfs_read_data_file_timeouts *timeouts,
433
+			unsigned int *delayed_min_us);
434
+
435
+ssize_t incfs_read_merkle_tree_blocks(struct mem_range dst,
436
+				      struct data_file *df, size_t offset);
282437 
283438 int incfs_get_filled_blocks(struct data_file *df,
439
+			    struct incfs_file_data *fd,
284440 			    struct incfs_get_filled_blocks_args *arg);
285441 
286442 int incfs_read_file_signature(struct data_file *df, struct mem_range dst);
287443 
288444 int incfs_process_new_data_block(struct data_file *df,
289
-				 struct incfs_fill_block *block, u8 *data);
445
+				 struct incfs_fill_block *block, u8 *data,
446
+				 bool *complete);
290447 
291448 int incfs_process_new_hash_block(struct data_file *df,
292449 				 struct incfs_fill_block *block, u8 *data);
....@@ -300,11 +457,13 @@
300457  */
301458 int incfs_collect_pending_reads(struct mount_info *mi, int sn_lowerbound,
302459 				struct incfs_pending_read_info *reads,
303
-				int reads_size);
460
+				struct incfs_pending_read_info2 *reads2,
461
+				int reads_size, int *new_max_sn);
304462 
305463 int incfs_collect_logged_reads(struct mount_info *mi,
306464 			       struct read_log_state *start_state,
307465 			       struct incfs_pending_read_info *reads,
466
+			       struct incfs_pending_read_info2 *reads2,
308467 			       int reads_size);
309468 struct read_log_state incfs_get_log_state(struct mount_info *mi);
310469 int incfs_get_uncollected_logs_count(struct mount_info *mi,
....@@ -315,7 +474,7 @@
315474 	if (!inode)
316475 		return NULL;
317476 
318
-	if (inode->i_sb->s_magic != (long) INCFS_MAGIC_NUMBER) {
477
+	if (inode->i_sb->s_magic != INCFS_MAGIC_NUMBER) {
319478 		/* This inode doesn't belong to us. */
320479 		pr_warn_once("incfs: %s on an alien inode.", __func__);
321480 		return NULL;
....@@ -388,7 +547,5 @@
388547 		return 0;
389548 	return 1 + (size - 1) / INCFS_DATA_FILE_BLOCK_SIZE;
390549 }
391
-
392
-bool incfs_equal_ranges(struct mem_range lhs, struct mem_range rhs);
393550 
394551 #endif /* _INCFS_DATA_MGMT_H */