brintos

brintos / linux-shallow public Read only

0
0
Text · 116.9 KiB · 12fc20a Raw
3420 lines · c
1/* SPDX-License-Identifier: GPL-2.0 */2/*3 * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for4 * licensing and copyright details5 */6 7#include <linux/reiserfs_fs.h>8 9#include <linux/slab.h>10#include <linux/interrupt.h>11#include <linux/sched.h>12#include <linux/bug.h>13#include <linux/workqueue.h>14#include <linux/unaligned.h>15#include <linux/bitops.h>16#include <linux/proc_fs.h>17#include <linux/buffer_head.h>18 19/* the 32 bit compat definitions with int argument */20#define REISERFS_IOC32_UNPACK		_IOW(0xCD, 1, int)21#define REISERFS_IOC32_GETVERSION	FS_IOC32_GETVERSION22#define REISERFS_IOC32_SETVERSION	FS_IOC32_SETVERSION23 24struct reiserfs_journal_list;25 26/* bitmasks for i_flags field in reiserfs-specific part of inode */27typedef enum {28	/*29	 * this says what format of key do all items (but stat data) of30	 * an object have.  If this is set, that format is 3.6 otherwise - 3.531	 */32	i_item_key_version_mask = 0x0001,33 34	/*35	 * If this is unset, object has 3.5 stat data, otherwise,36	 * it has 3.6 stat data with 64bit size, 32bit nlink etc.37	 */38	i_stat_data_version_mask = 0x0002,39 40	/* file might need tail packing on close */41	i_pack_on_close_mask = 0x0004,42 43	/* don't pack tail of file */44	i_nopack_mask = 0x0008,45 46	/*47	 * If either of these are set, "safe link" was created for this48	 * file during truncate or unlink. Safe link is used to avoid49	 * leakage of disk space on crash with some files open, but unlinked.50	 */51	i_link_saved_unlink_mask = 0x0010,52	i_link_saved_truncate_mask = 0x0020,53 54	i_has_xattr_dir = 0x0040,55	i_data_log = 0x0080,56} reiserfs_inode_flags;57 58struct reiserfs_inode_info {59	__u32 i_key[4];		/* key is still 4 32 bit integers */60 61	/*62	 * transient inode flags that are never stored on disk. Bitmasks63	 * for this field are defined above.64	 */65	__u32 i_flags;66 67	/* offset of first byte stored in direct item. */68	__u32 i_first_direct_byte;69 70	/* copy of persistent inode flags read from sd_attrs. */71	__u32 i_attrs;72 73	/* first unused block of a sequence of unused blocks */74	int i_prealloc_block;75	int i_prealloc_count;	/* length of that sequence */76 77	/* per-transaction list of inodes which  have preallocated blocks */78	struct list_head i_prealloc_list;79 80	/*81	 * new_packing_locality is created; new blocks for the contents82	 * of this directory should be displaced83	 */84	unsigned new_packing_locality:1;85 86	/*87	 * we use these for fsync or O_SYNC to decide which transaction88	 * needs to be committed in order for this inode to be properly89	 * flushed90	 */91	unsigned int i_trans_id;92 93	struct reiserfs_journal_list *i_jl;94	atomic_t openers;95	struct mutex tailpack;96#ifdef CONFIG_REISERFS_FS_XATTR97	struct rw_semaphore i_xattr_sem;98#endif99#ifdef CONFIG_QUOTA100	struct dquot __rcu *i_dquot[MAXQUOTAS];101#endif102 103	struct inode vfs_inode;104};105 106typedef enum {107	reiserfs_attrs_cleared = 0x00000001,108} reiserfs_super_block_flags;109 110/*111 * struct reiserfs_super_block accessors/mutators since this is a disk112 * structure, it will always be in little endian format.113 */114#define sb_block_count(sbp)         (le32_to_cpu((sbp)->s_v1.s_block_count))115#define set_sb_block_count(sbp,v)   ((sbp)->s_v1.s_block_count = cpu_to_le32(v))116#define sb_free_blocks(sbp)         (le32_to_cpu((sbp)->s_v1.s_free_blocks))117#define set_sb_free_blocks(sbp,v)   ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v))118#define sb_root_block(sbp)          (le32_to_cpu((sbp)->s_v1.s_root_block))119#define set_sb_root_block(sbp,v)    ((sbp)->s_v1.s_root_block = cpu_to_le32(v))120 121#define sb_jp_journal_1st_block(sbp)  \122              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block))123#define set_sb_jp_journal_1st_block(sbp,v) \124              ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v))125#define sb_jp_journal_dev(sbp) \126              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev))127#define set_sb_jp_journal_dev(sbp,v) \128              ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v))129#define sb_jp_journal_size(sbp) \130              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size))131#define set_sb_jp_journal_size(sbp,v) \132              ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v))133#define sb_jp_journal_trans_max(sbp) \134              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max))135#define set_sb_jp_journal_trans_max(sbp,v) \136              ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v))137#define sb_jp_journal_magic(sbp) \138              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic))139#define set_sb_jp_journal_magic(sbp,v) \140              ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v))141#define sb_jp_journal_max_batch(sbp) \142              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch))143#define set_sb_jp_journal_max_batch(sbp,v) \144              ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v))145#define sb_jp_jourmal_max_commit_age(sbp) \146              (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age))147#define set_sb_jp_journal_max_commit_age(sbp,v) \148              ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v))149 150#define sb_blocksize(sbp)          (le16_to_cpu((sbp)->s_v1.s_blocksize))151#define set_sb_blocksize(sbp,v)    ((sbp)->s_v1.s_blocksize = cpu_to_le16(v))152#define sb_oid_maxsize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_maxsize))153#define set_sb_oid_maxsize(sbp,v)  ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v))154#define sb_oid_cursize(sbp)        (le16_to_cpu((sbp)->s_v1.s_oid_cursize))155#define set_sb_oid_cursize(sbp,v)  ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v))156#define sb_umount_state(sbp)       (le16_to_cpu((sbp)->s_v1.s_umount_state))157#define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v))158#define sb_fs_state(sbp)           (le16_to_cpu((sbp)->s_v1.s_fs_state))159#define set_sb_fs_state(sbp,v)     ((sbp)->s_v1.s_fs_state = cpu_to_le16(v))160#define sb_hash_function_code(sbp) \161              (le32_to_cpu((sbp)->s_v1.s_hash_function_code))162#define set_sb_hash_function_code(sbp,v) \163              ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v))164#define sb_tree_height(sbp)        (le16_to_cpu((sbp)->s_v1.s_tree_height))165#define set_sb_tree_height(sbp,v)  ((sbp)->s_v1.s_tree_height = cpu_to_le16(v))166#define sb_bmap_nr(sbp)            (le16_to_cpu((sbp)->s_v1.s_bmap_nr))167#define set_sb_bmap_nr(sbp,v)      ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v))168#define sb_version(sbp)            (le16_to_cpu((sbp)->s_v1.s_version))169#define set_sb_version(sbp,v)      ((sbp)->s_v1.s_version = cpu_to_le16(v))170 171#define sb_mnt_count(sbp)	   (le16_to_cpu((sbp)->s_mnt_count))172#define set_sb_mnt_count(sbp, v)   ((sbp)->s_mnt_count = cpu_to_le16(v))173 174#define sb_reserved_for_journal(sbp) \175              (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal))176#define set_sb_reserved_for_journal(sbp,v) \177              ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v))178 179/* LOGGING -- */180 181/*182 * These all interelate for performance.183 *184 * If the journal block count is smaller than n transactions, you lose speed.185 * I don't know what n is yet, I'm guessing 8-16.186 *187 * typical transaction size depends on the application, how often fsync is188 * called, and how many metadata blocks you dirty in a 30 second period.189 * The more small files (<16k) you use, the larger your transactions will190 * be.191 *192 * If your journal fills faster than dirty buffers get flushed to disk, it193 * must flush them before allowing the journal to wrap, which slows things194 * down.  If you need high speed meta data updates, the journal should be195 * big enough to prevent wrapping before dirty meta blocks get to disk.196 *197 * If the batch max is smaller than the transaction max, you'll waste space198 * at the end of the journal because journal_end sets the next transaction199 * to start at 0 if the next transaction has any chance of wrapping.200 *201 * The large the batch max age, the better the speed, and the more meta202 * data changes you'll lose after a crash.203 */204 205/* don't mess with these for a while */206/* we have a node size define somewhere in reiserfs_fs.h. -Hans */207#define JOURNAL_BLOCK_SIZE  4096	/* BUG gotta get rid of this */208#define JOURNAL_MAX_CNODE   1500	/* max cnodes to allocate. */209#define JOURNAL_HASH_SIZE 8192210 211/* number of copies of the bitmaps to have floating.  Must be >= 2 */212#define JOURNAL_NUM_BITMAPS 5213 214/*215 * One of these for every block in every transaction216 * Each one is in two hash tables.  First, a hash of the current transaction,217 * and after journal_end, a hash of all the in memory transactions.218 * next and prev are used by the current transaction (journal_hash).219 * hnext and hprev are used by journal_list_hash.  If a block is in more220 * than one transaction, the journal_list_hash links it in multiple times.221 * This allows flush_journal_list to remove just the cnode belonging to a222 * given transaction.223 */224struct reiserfs_journal_cnode {225	struct buffer_head *bh;	/* real buffer head */226	struct super_block *sb;	/* dev of real buffer head */227 228	/* block number of real buffer head, == 0 when buffer on disk */229	__u32 blocknr;230 231	unsigned long state;232 233	/* journal list this cnode lives in */234	struct reiserfs_journal_list *jlist;235 236	struct reiserfs_journal_cnode *next;	/* next in transaction list */237	struct reiserfs_journal_cnode *prev;	/* prev in transaction list */238	struct reiserfs_journal_cnode *hprev;	/* prev in hash list */239	struct reiserfs_journal_cnode *hnext;	/* next in hash list */240};241 242struct reiserfs_bitmap_node {243	int id;244	char *data;245	struct list_head list;246};247 248struct reiserfs_list_bitmap {249	struct reiserfs_journal_list *journal_list;250	struct reiserfs_bitmap_node **bitmaps;251};252 253/*254 * one of these for each transaction.  The most important part here is the255 * j_realblock.  this list of cnodes is used to hash all the blocks in all256 * the commits, to mark all the real buffer heads dirty once all the commits257 * hit the disk, and to make sure every real block in a transaction is on258 * disk before allowing the log area to be overwritten259 */260struct reiserfs_journal_list {261	unsigned long j_start;262	unsigned long j_state;263	unsigned long j_len;264	atomic_t j_nonzerolen;265	atomic_t j_commit_left;266 267	/* all commits older than this on disk */268	atomic_t j_older_commits_done;269 270	struct mutex j_commit_mutex;271	unsigned int j_trans_id;272	time64_t j_timestamp; /* write-only but useful for crash dump analysis */273	struct reiserfs_list_bitmap *j_list_bitmap;274	struct buffer_head *j_commit_bh;	/* commit buffer head */275	struct reiserfs_journal_cnode *j_realblock;276	struct reiserfs_journal_cnode *j_freedlist;	/* list of buffers that were freed during this trans.  free each of these on flush */277	/* time ordered list of all active transactions */278	struct list_head j_list;279 280	/*281	 * time ordered list of all transactions we haven't tried282	 * to flush yet283	 */284	struct list_head j_working_list;285 286	/* list of tail conversion targets in need of flush before commit */287	struct list_head j_tail_bh_list;288 289	/* list of data=ordered buffers in need of flush before commit */290	struct list_head j_bh_list;291	int j_refcount;292};293 294struct reiserfs_journal {295	struct buffer_head **j_ap_blocks;	/* journal blocks on disk */296	/* newest journal block */297	struct reiserfs_journal_cnode *j_last;298 299	/* oldest journal block.  start here for traverse */300	struct reiserfs_journal_cnode *j_first;301 302	struct file *j_bdev_file;303 304	/* first block on s_dev of reserved area journal */305	int j_1st_reserved_block;306 307	unsigned long j_state;308	unsigned int j_trans_id;309	unsigned long j_mount_id;310 311	/* start of current waiting commit (index into j_ap_blocks) */312	unsigned long j_start;313	unsigned long j_len;	/* length of current waiting commit */314 315	/* number of buffers requested by journal_begin() */316	unsigned long j_len_alloc;317 318	atomic_t j_wcount;	/* count of writers for current commit */319 320	/* batch count. allows turning X transactions into 1 */321	unsigned long j_bcount;322 323	/* first unflushed transactions offset */324	unsigned long j_first_unflushed_offset;325 326	/* last fully flushed journal timestamp */327	unsigned j_last_flush_trans_id;328 329	struct buffer_head *j_header_bh;330 331	time64_t j_trans_start_time;	/* time this transaction started */332	struct mutex j_mutex;333	struct mutex j_flush_mutex;334 335	/* wait for current transaction to finish before starting new one */336	wait_queue_head_t j_join_wait;337 338	atomic_t j_jlock;		/* lock for j_join_wait */339	int j_list_bitmap_index;	/* number of next list bitmap to use */340 341	/* no more journal begins allowed. MUST sleep on j_join_wait */342	int j_must_wait;343 344	/* next journal_end will flush all journal list */345	int j_next_full_flush;346 347	/* next journal_end will flush all async commits */348	int j_next_async_flush;349 350	int j_cnode_used;	/* number of cnodes on the used list */351	int j_cnode_free;	/* number of cnodes on the free list */352 353	/* max number of blocks in a transaction.  */354	unsigned int j_trans_max;355 356	/* max number of blocks to batch into a trans */357	unsigned int j_max_batch;358 359	/* in seconds, how old can an async commit be */360	unsigned int j_max_commit_age;361 362	/* in seconds, how old can a transaction be */363	unsigned int j_max_trans_age;364 365	/* the default for the max commit age */366	unsigned int j_default_max_commit_age;367 368	struct reiserfs_journal_cnode *j_cnode_free_list;369 370	/* orig pointer returned from vmalloc */371	struct reiserfs_journal_cnode *j_cnode_free_orig;372 373	struct reiserfs_journal_list *j_current_jl;374	int j_free_bitmap_nodes;375	int j_used_bitmap_nodes;376 377	int j_num_lists;	/* total number of active transactions */378	int j_num_work_lists;	/* number that need attention from kreiserfsd */379 380	/* debugging to make sure things are flushed in order */381	unsigned int j_last_flush_id;382 383	/* debugging to make sure things are committed in order */384	unsigned int j_last_commit_id;385 386	struct list_head j_bitmap_nodes;387	struct list_head j_dirty_buffers;388	spinlock_t j_dirty_buffers_lock;	/* protects j_dirty_buffers */389 390	/* list of all active transactions */391	struct list_head j_journal_list;392 393	/* lists that haven't been touched by writeback attempts */394	struct list_head j_working_list;395 396	/* hash table for real buffer heads in current trans */397	struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE];398 399	/* hash table for all the real buffer heads in all the transactions */400	struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE];401 402	/* array of bitmaps to record the deleted blocks */403	struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS];404 405	/* list of inodes which have preallocated blocks */406	struct list_head j_prealloc_list;407	int j_persistent_trans;408	unsigned long j_max_trans_size;409	unsigned long j_max_batch_size;410 411	int j_errno;412 413	/* when flushing ordered buffers, throttle new ordered writers */414	struct delayed_work j_work;415	struct super_block *j_work_sb;416	atomic_t j_async_throttle;417};418 419enum journal_state_bits {420	J_WRITERS_BLOCKED = 1,	/* set when new writers not allowed */421	J_WRITERS_QUEUED,    /* set when log is full due to too many writers */422	J_ABORTED,           /* set when log is aborted */423};424 425/* ick.  magic string to find desc blocks in the journal */426#define JOURNAL_DESC_MAGIC "ReIsErLB"427 428typedef __u32(*hashf_t) (const signed char *, int);429 430struct reiserfs_bitmap_info {431	__u32 free_count;432};433 434struct proc_dir_entry;435 436#if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO )437typedef unsigned long int stat_cnt_t;438typedef struct reiserfs_proc_info_data {439	spinlock_t lock;440	int exiting;441	int max_hash_collisions;442 443	stat_cnt_t breads;444	stat_cnt_t bread_miss;445	stat_cnt_t search_by_key;446	stat_cnt_t search_by_key_fs_changed;447	stat_cnt_t search_by_key_restarted;448 449	stat_cnt_t insert_item_restarted;450	stat_cnt_t paste_into_item_restarted;451	stat_cnt_t cut_from_item_restarted;452	stat_cnt_t delete_solid_item_restarted;453	stat_cnt_t delete_item_restarted;454 455	stat_cnt_t leaked_oid;456	stat_cnt_t leaves_removable;457 458	/*459	 * balances per level.460	 * Use explicit 5 as MAX_HEIGHT is not visible yet.461	 */462	stat_cnt_t balance_at[5];	/* XXX */463	/* sbk == search_by_key */464	stat_cnt_t sbk_read_at[5];	/* XXX */465	stat_cnt_t sbk_fs_changed[5];466	stat_cnt_t sbk_restarted[5];467	stat_cnt_t items_at[5];	/* XXX */468	stat_cnt_t free_at[5];	/* XXX */469	stat_cnt_t can_node_be_removed[5];	/* XXX */470	long int lnum[5];	/* XXX */471	long int rnum[5];	/* XXX */472	long int lbytes[5];	/* XXX */473	long int rbytes[5];	/* XXX */474	stat_cnt_t get_neighbors[5];475	stat_cnt_t get_neighbors_restart[5];476	stat_cnt_t need_l_neighbor[5];477	stat_cnt_t need_r_neighbor[5];478 479	stat_cnt_t free_block;480	struct __scan_bitmap_stats {481		stat_cnt_t call;482		stat_cnt_t wait;483		stat_cnt_t bmap;484		stat_cnt_t retry;485		stat_cnt_t in_journal_hint;486		stat_cnt_t in_journal_nohint;487		stat_cnt_t stolen;488	} scan_bitmap;489	struct __journal_stats {490		stat_cnt_t in_journal;491		stat_cnt_t in_journal_bitmap;492		stat_cnt_t in_journal_reusable;493		stat_cnt_t lock_journal;494		stat_cnt_t lock_journal_wait;495		stat_cnt_t journal_being;496		stat_cnt_t journal_relock_writers;497		stat_cnt_t journal_relock_wcount;498		stat_cnt_t mark_dirty;499		stat_cnt_t mark_dirty_already;500		stat_cnt_t mark_dirty_notjournal;501		stat_cnt_t restore_prepared;502		stat_cnt_t prepare;503		stat_cnt_t prepare_retry;504	} journal;505} reiserfs_proc_info_data_t;506#else507typedef struct reiserfs_proc_info_data {508} reiserfs_proc_info_data_t;509#endif510 511/* Number of quota types we support */512#define REISERFS_MAXQUOTAS 2513 514/* reiserfs union of in-core super block data */515struct reiserfs_sb_info {516	/* Buffer containing the super block */517	struct buffer_head *s_sbh;518 519	/* Pointer to the on-disk super block in the buffer */520	struct reiserfs_super_block *s_rs;521	struct reiserfs_bitmap_info *s_ap_bitmap;522 523	/* pointer to journal information */524	struct reiserfs_journal *s_journal;525 526	unsigned short s_mount_state;	/* reiserfs state (valid, invalid) */527 528	/* Serialize writers access, replace the old bkl */529	struct mutex lock;530 531	/* Owner of the lock (can be recursive) */532	struct task_struct *lock_owner;533 534	/* Depth of the lock, start from -1 like the bkl */535	int lock_depth;536 537	struct workqueue_struct *commit_wq;538 539	/* Comment? -Hans */540	void (*end_io_handler) (struct buffer_head *, int);541 542	/*543	 * pointer to function which is used to sort names in directory.544	 * Set on mount545	 */546	hashf_t s_hash_function;547 548	/* reiserfs's mount options are set here */549	unsigned long s_mount_opt;550 551	/* This is a structure that describes block allocator options */552	struct {553		/* Bitfield for enable/disable kind of options */554		unsigned long bits;555 556		/*557		 * size started from which we consider file558		 * to be a large one (in blocks)559		 */560		unsigned long large_file_size;561 562		int border;	/* percentage of disk, border takes */563 564		/*565		 * Minimal file size (in blocks) starting566		 * from which we do preallocations567		 */568		int preallocmin;569 570		/*571		 * Number of blocks we try to prealloc when file572		 * reaches preallocmin size (in blocks) or prealloc_list573		 is empty.574		 */575		int preallocsize;576	} s_alloc_options;577 578	/* Comment? -Hans */579	wait_queue_head_t s_wait;580	/* increased by one every time the  tree gets re-balanced */581	atomic_t s_generation_counter;582 583	/* File system properties. Currently holds on-disk FS format */584	unsigned long s_properties;585 586	/* session statistics */587	int s_disk_reads;588	int s_disk_writes;589	int s_fix_nodes;590	int s_do_balance;591	int s_unneeded_left_neighbor;592	int s_good_search_by_key_reada;593	int s_bmaps;594	int s_bmaps_without_search;595	int s_direct2indirect;596	int s_indirect2direct;597 598	/*599	 * set up when it's ok for reiserfs_read_inode2() to read from600	 * disk inode with nlink==0. Currently this is only used during601	 * finish_unfinished() processing at mount time602	 */603	int s_is_unlinked_ok;604 605	reiserfs_proc_info_data_t s_proc_info_data;606	struct proc_dir_entry *procdir;607 608	/* amount of blocks reserved for further allocations */609	int reserved_blocks;610 611 612	/* this lock on now only used to protect reserved_blocks variable */613	spinlock_t bitmap_lock;614	struct dentry *priv_root;	/* root of /.reiserfs_priv */615	struct dentry *xattr_root;	/* root of /.reiserfs_priv/xattrs */616	int j_errno;617 618	int work_queued;              /* non-zero delayed work is queued */619	struct delayed_work old_work; /* old transactions flush delayed work */620	spinlock_t old_work_lock;     /* protects old_work and work_queued */621 622#ifdef CONFIG_QUOTA623	char *s_qf_names[REISERFS_MAXQUOTAS];624	int s_jquota_fmt;625#endif626	char *s_jdev;		/* Stored jdev for mount option showing */627#ifdef CONFIG_REISERFS_CHECK628 629	/*630	 * Detects whether more than one copy of tb exists per superblock631	 * as a means of checking whether do_balance is executing632	 * concurrently against another tree reader/writer on a same633	 * mount point.634	 */635	struct tree_balance *cur_tb;636#endif637};638 639/* Definitions of reiserfs on-disk properties: */640#define REISERFS_3_5 0641#define REISERFS_3_6 1642#define REISERFS_OLD_FORMAT 2643 644/* Mount options */645enum reiserfs_mount_options {646	/* large tails will be created in a session */647	REISERFS_LARGETAIL,648	/*649	 * small (for files less than block size) tails will650	 * be created in a session651	 */652	REISERFS_SMALLTAIL,653 654	/* replay journal and return 0. Use by fsck */655	REPLAYONLY,656 657	/*658	 * -o conv: causes conversion of old format super block to the659	 * new format. If not specified - old partition will be dealt660	 * with in a manner of 3.5.x661	 */662	REISERFS_CONVERT,663 664	/*665	 * -o hash={tea, rupasov, r5, detect} is meant for properly mounting666	 * reiserfs disks from 3.5.19 or earlier.  99% of the time, this667	 * option is not required.  If the normal autodection code can't668	 * determine which hash to use (because both hashes had the same669	 * value for a file) use this option to force a specific hash.670	 * It won't allow you to override the existing hash on the FS, so671	 * if you have a tea hash disk, and mount with -o hash=rupasov,672	 * the mount will fail.673	 */674	FORCE_TEA_HASH,		/* try to force tea hash on mount */675	FORCE_RUPASOV_HASH,	/* try to force rupasov hash on mount */676	FORCE_R5_HASH,		/* try to force rupasov hash on mount */677	FORCE_HASH_DETECT,	/* try to detect hash function on mount */678 679	REISERFS_DATA_LOG,680	REISERFS_DATA_ORDERED,681	REISERFS_DATA_WRITEBACK,682 683	/*684	 * used for testing experimental features, makes benchmarking new685	 * features with and without more convenient, should never be used by686	 * users in any code shipped to users (ideally)687	 */688 689	REISERFS_NO_BORDER,690	REISERFS_NO_UNHASHED_RELOCATION,691	REISERFS_HASHED_RELOCATION,692	REISERFS_ATTRS,693	REISERFS_XATTRS_USER,694	REISERFS_POSIXACL,695	REISERFS_EXPOSE_PRIVROOT,696	REISERFS_BARRIER_NONE,697	REISERFS_BARRIER_FLUSH,698 699	/* Actions on error */700	REISERFS_ERROR_PANIC,701	REISERFS_ERROR_RO,702	REISERFS_ERROR_CONTINUE,703 704	REISERFS_USRQUOTA,	/* User quota option specified */705	REISERFS_GRPQUOTA,	/* Group quota option specified */706 707	REISERFS_TEST1,708	REISERFS_TEST2,709	REISERFS_TEST3,710	REISERFS_TEST4,711	REISERFS_UNSUPPORTED_OPT,712};713 714#define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH))715#define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH))716#define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH))717#define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT))718#define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER))719#define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION))720#define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION))721#define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4))722 723#define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL))724#define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL))725#define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY))726#define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS))727#define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5))728#define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT))729#define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG))730#define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED))731#define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK))732#define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER))733#define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL))734#define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT))735#define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s))736#define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE))737#define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH))738 739#define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC))740#define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO))741 742void reiserfs_file_buffer(struct buffer_head *bh, int list);743extern struct file_system_type reiserfs_fs_type;744int reiserfs_resize(struct super_block *, unsigned long);745 746#define CARRY_ON                0747#define SCHEDULE_OCCURRED       1748 749#define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh)750#define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal)751#define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block)752#define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free)753#define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap)754 755#define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->)756 757#define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal)))758static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal759						*journal)760{761	return test_bit(J_ABORTED, &journal->j_state);762}763 764/*765 * Locking primitives. The write lock is a per superblock766 * special mutex that has properties close to the Big Kernel Lock767 * which was used in the previous locking scheme.768 */769void reiserfs_write_lock(struct super_block *s);770void reiserfs_write_unlock(struct super_block *s);771int __must_check reiserfs_write_unlock_nested(struct super_block *s);772void reiserfs_write_lock_nested(struct super_block *s, int depth);773 774#ifdef CONFIG_REISERFS_CHECK775void reiserfs_lock_check_recursive(struct super_block *s);776#else777static inline void reiserfs_lock_check_recursive(struct super_block *s) { }778#endif779 780/*781 * Several mutexes depend on the write lock.782 * However sometimes we want to relax the write lock while we hold783 * these mutexes, according to the release/reacquire on schedule()784 * properties of the Bkl that were used.785 * Reiserfs performances and locking were based on this scheme.786 * Now that the write lock is a mutex and not the bkl anymore, doing so787 * may result in a deadlock:788 *789 * A acquire write_lock790 * A acquire j_commit_mutex791 * A release write_lock and wait for something792 * B acquire write_lock793 * B can't acquire j_commit_mutex and sleep794 * A can't acquire write lock anymore795 * deadlock796 *797 * What we do here is avoiding such deadlock by playing the same game798 * than the Bkl: if we can't acquire a mutex that depends on the write lock,799 * we release the write lock, wait a bit and then retry.800 *801 * The mutexes concerned by this hack are:802 * - The commit mutex of a journal list803 * - The flush mutex804 * - The journal lock805 * - The inode mutex806 */807static inline void reiserfs_mutex_lock_safe(struct mutex *m,808					    struct super_block *s)809{810	int depth;811 812	depth = reiserfs_write_unlock_nested(s);813	mutex_lock(m);814	reiserfs_write_lock_nested(s, depth);815}816 817static inline void818reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass,819				struct super_block *s)820{821	int depth;822 823	depth = reiserfs_write_unlock_nested(s);824	mutex_lock_nested(m, subclass);825	reiserfs_write_lock_nested(s, depth);826}827 828static inline void829reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s)830{831       int depth;832       depth = reiserfs_write_unlock_nested(s);833       down_read(sem);834       reiserfs_write_lock_nested(s, depth);835}836 837/*838 * When we schedule, we usually want to also release the write lock,839 * according to the previous bkl based locking scheme of reiserfs.840 */841static inline void reiserfs_cond_resched(struct super_block *s)842{843	if (need_resched()) {844		int depth;845 846		depth = reiserfs_write_unlock_nested(s);847		schedule();848		reiserfs_write_lock_nested(s, depth);849	}850}851 852struct fid;853 854/*855 * in reading the #defines, it may help to understand that they employ856 *  the following abbreviations:857 *858 *  B = Buffer859 *  I = Item header860 *  H = Height within the tree (should be changed to LEV)861 *  N = Number of the item in the node862 *  STAT = stat data863 *  DEH = Directory Entry Header864 *  EC = Entry Count865 *  E = Entry number866 *  UL = Unsigned Long867 *  BLKH = BLocK Header868 *  UNFM = UNForMatted node869 *  DC = Disk Child870 *  P = Path871 *872 *  These #defines are named by concatenating these abbreviations,873 *  where first comes the arguments, and last comes the return value,874 *  of the macro.875 */876 877#define USE_INODE_GENERATION_COUNTER878 879#define REISERFS_PREALLOCATE880#define DISPLACE_NEW_PACKING_LOCALITIES881#define PREALLOCATION_SIZE 9882 883/* n must be power of 2 */884#define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u))885 886/*887 * to be ok for alpha and others we have to align structures to 8 byte888 * boundary.889 * FIXME: do not change 4 by anything else: there is code which relies on that890 */891#define ROUND_UP(x) _ROUND_UP(x,8LL)892 893/*894 * debug levels.  Right now, CONFIG_REISERFS_CHECK means print all debug895 * messages.896 */897#define REISERFS_DEBUG_CODE 5	/* extra messages to help find/debug errors */898 899void __reiserfs_warning(struct super_block *s, const char *id,900			 const char *func, const char *fmt, ...);901#define reiserfs_warning(s, id, fmt, args...) \902	 __reiserfs_warning(s, id, __func__, fmt, ##args)903/* assertions handling */904 905/* always check a condition and panic if it's false. */906#define __RASSERT(cond, scond, format, args...)			\907do {									\908	if (!(cond))							\909		reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \910			       __FILE__ ":%i:%s: " format "\n",		\911			       __LINE__, __func__ , ##args);		\912} while (0)913 914#define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args)915 916#if defined( CONFIG_REISERFS_CHECK )917#define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args)918#else919#define RFALSE( cond, format, args... ) do {;} while( 0 )920#endif921 922#define CONSTF __attribute_const__923/*924 * Disk Data Structures925 */926 927/***************************************************************************928 *                             SUPER BLOCK                                 *929 ***************************************************************************/930 931/*932 * Structure of super block on disk, a version of which in RAM is often933 * accessed as REISERFS_SB(s)->s_rs. The version in RAM is part of a larger934 * structure containing fields never written to disk.935 */936#define UNSET_HASH 0	/* Detect hash on disk */937#define TEA_HASH  1938#define YURA_HASH 2939#define R5_HASH   3940#define DEFAULT_HASH R5_HASH941 942struct journal_params {943	/* where does journal start from on its * device */944	__le32 jp_journal_1st_block;945 946	/* journal device st_rdev */947	__le32 jp_journal_dev;948 949	/* size of the journal */950	__le32 jp_journal_size;951 952	/* max number of blocks in a transaction. */953	__le32 jp_journal_trans_max;954 955	/*956	 * random value made on fs creation957	 * (this was sb_journal_block_count)958	 */959	__le32 jp_journal_magic;960 961	/* max number of blocks to batch into a trans */962	__le32 jp_journal_max_batch;963 964	/* in seconds, how old can an async  commit be */965	__le32 jp_journal_max_commit_age;966 967	/* in seconds, how old can a transaction be */968	__le32 jp_journal_max_trans_age;969};970 971/* this is the super from 3.5.X, where X >= 10 */972struct reiserfs_super_block_v1 {973	__le32 s_block_count;	/* blocks count         */974	__le32 s_free_blocks;	/* free blocks count    */975	__le32 s_root_block;	/* root block number    */976	struct journal_params s_journal;977	__le16 s_blocksize;	/* block size */978 979	/* max size of object id array, see get_objectid() commentary  */980	__le16 s_oid_maxsize;981	__le16 s_oid_cursize;	/* current size of object id array */982 983	/* this is set to 1 when filesystem was umounted, to 2 - when not */984	__le16 s_umount_state;985 986	/*987	 * reiserfs magic string indicates that file system is reiserfs:988	 * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs"989	 */990	char s_magic[10];991 992	/*993	 * it is set to used by fsck to mark which994	 * phase of rebuilding is done995	 */996	__le16 s_fs_state;997	/*998	 * indicate, what hash function is being use999	 * to sort names in a directory1000	 */1001	__le32 s_hash_function_code;1002	__le16 s_tree_height;	/* height of disk tree */1003 1004	/*1005	 * amount of bitmap blocks needed to address1006	 * each block of file system1007	 */1008	__le16 s_bmap_nr;1009 1010	/*1011	 * this field is only reliable on filesystem with non-standard journal1012	 */1013	__le16 s_version;1014 1015	/*1016	 * size in blocks of journal area on main device, we need to1017	 * keep after making fs with non-standard journal1018	 */1019	__le16 s_reserved_for_journal;1020} __attribute__ ((__packed__));1021 1022#define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1))1023 1024/* this is the on disk super block */1025struct reiserfs_super_block {1026	struct reiserfs_super_block_v1 s_v1;1027	__le32 s_inode_generation;1028 1029	/* Right now used only by inode-attributes, if enabled */1030	__le32 s_flags;1031 1032	unsigned char s_uuid[16];	/* filesystem unique identifier */1033	unsigned char s_label[16];	/* filesystem volume label */1034	__le16 s_mnt_count;		/* Count of mounts since last fsck */1035	__le16 s_max_mnt_count;		/* Maximum mounts before check */1036	__le32 s_lastcheck;		/* Timestamp of last fsck */1037	__le32 s_check_interval;	/* Interval between checks */1038 1039	/*1040	 * zero filled by mkreiserfs and reiserfs_convert_objectid_map_v1()1041	 * so any additions must be updated there as well. */1042	char s_unused[76];1043} __attribute__ ((__packed__));1044 1045#define SB_SIZE (sizeof(struct reiserfs_super_block))1046 1047#define REISERFS_VERSION_1 01048#define REISERFS_VERSION_2 21049 1050/* on-disk super block fields converted to cpu form */1051#define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs)1052#define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1))1053#define SB_BLOCKSIZE(s) \1054        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize))1055#define SB_BLOCK_COUNT(s) \1056        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count))1057#define SB_FREE_BLOCKS(s) \1058        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks))1059#define SB_REISERFS_MAGIC(s) \1060        (SB_V1_DISK_SUPER_BLOCK(s)->s_magic)1061#define SB_ROOT_BLOCK(s) \1062        le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block))1063#define SB_TREE_HEIGHT(s) \1064        le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height))1065#define SB_REISERFS_STATE(s) \1066        le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state))1067#define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version))1068#define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr))1069 1070#define PUT_SB_BLOCK_COUNT(s, val) \1071   do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0)1072#define PUT_SB_FREE_BLOCKS(s, val) \1073   do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0)1074#define PUT_SB_ROOT_BLOCK(s, val) \1075   do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0)1076#define PUT_SB_TREE_HEIGHT(s, val) \1077   do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0)1078#define PUT_SB_REISERFS_STATE(s, val) \1079   do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0)1080#define PUT_SB_VERSION(s, val) \1081   do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0)1082#define PUT_SB_BMAP_NR(s, val) \1083   do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0)1084 1085#define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal)1086#define SB_ONDISK_JOURNAL_SIZE(s) \1087         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size))1088#define SB_ONDISK_JOURNAL_1st_BLOCK(s) \1089         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block))1090#define SB_ONDISK_JOURNAL_DEVICE(s) \1091         le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))1092#define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \1093         le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))1094 1095#define is_block_in_log_or_reserved_area(s, block) \1096         block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \1097         && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) +  \1098         ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \1099         SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s)))1100 1101int is_reiserfs_3_5(struct reiserfs_super_block *rs);1102int is_reiserfs_3_6(struct reiserfs_super_block *rs);1103int is_reiserfs_jr(struct reiserfs_super_block *rs);1104 1105/*1106 * ReiserFS leaves the first 64k unused, so that partition labels have1107 * enough space.  If someone wants to write a fancy bootloader that1108 * needs more than 64k, let us know, and this will be increased in size.1109 * This number must be larger than the largest block size on any1110 * platform, or code will break.  -Hans1111 */1112#define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024)1113#define REISERFS_FIRST_BLOCK unused_define1114#define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES1115 1116/* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */1117#define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024)1118 1119/* reiserfs internal error code (used by search_by_key and fix_nodes)) */1120#define CARRY_ON      01121#define REPEAT_SEARCH -11122#define IO_ERROR      -21123#define NO_DISK_SPACE -31124#define NO_BALANCING_NEEDED  (-4)1125#define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5)1126#define QUOTA_EXCEEDED -61127 1128typedef __u32 b_blocknr_t;1129typedef __le32 unp_t;1130 1131struct unfm_nodeinfo {1132	unp_t unfm_nodenum;1133	unsigned short unfm_freespace;1134};1135 1136/* there are two formats of keys: 3.5 and 3.6 */1137#define KEY_FORMAT_3_5 01138#define KEY_FORMAT_3_6 11139 1140/* there are two stat datas */1141#define STAT_DATA_V1 01142#define STAT_DATA_V2 11143 1144static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode)1145{1146	return container_of(inode, struct reiserfs_inode_info, vfs_inode);1147}1148 1149static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb)1150{1151	return sb->s_fs_info;1152}1153 1154/*1155 * Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u161156 * which overflows on large file systems.1157 */1158static inline __u32 reiserfs_bmap_count(struct super_block *sb)1159{1160	return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1;1161}1162 1163static inline int bmap_would_wrap(unsigned bmap_nr)1164{1165	return bmap_nr > ((1LL << 16) - 1);1166}1167 1168extern const struct xattr_handler * const reiserfs_xattr_handlers[];1169 1170/*1171 * this says about version of key of all items (but stat data) the1172 * object consists of1173 */1174#define get_inode_item_key_version( inode )                                    \1175    ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5)1176 1177#define set_inode_item_key_version( inode, version )                           \1178         ({ if((version)==KEY_FORMAT_3_6)                                      \1179                REISERFS_I(inode)->i_flags |= i_item_key_version_mask;      \1180            else                                                               \1181                REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; })1182 1183#define get_inode_sd_version(inode)                                            \1184    ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1)1185 1186#define set_inode_sd_version(inode, version)                                   \1187         ({ if((version)==STAT_DATA_V2)                                        \1188                REISERFS_I(inode)->i_flags |= i_stat_data_version_mask;     \1189            else                                                               \1190                REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; })1191 1192/*1193 * This is an aggressive tail suppression policy, I am hoping it1194 * improves our benchmarks. The principle behind it is that percentage1195 * space saving is what matters, not absolute space saving.  This is1196 * non-intuitive, but it helps to understand it if you consider that the1197 * cost to access 4 blocks is not much more than the cost to access 11198 * block, if you have to do a seek and rotate.  A tail risks a1199 * non-linear disk access that is significant as a percentage of total1200 * time cost for a 4 block file and saves an amount of space that is1201 * less significant as a percentage of space, or so goes the hypothesis.1202 * -Hans1203 */1204#define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \1205(\1206  (!(n_tail_size)) || \1207  (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \1208   ( (n_file_size) >= (n_block_size) * 4 ) || \1209   ( ( (n_file_size) >= (n_block_size) * 3 ) && \1210     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \1211   ( ( (n_file_size) >= (n_block_size) * 2 ) && \1212     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \1213   ( ( (n_file_size) >= (n_block_size) ) && \1214     ( (n_tail_size) >=   (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \1215)1216 1217/*1218 * Another strategy for tails, this one means only create a tail if all the1219 * file would fit into one DIRECT item.1220 * Primary intention for this one is to increase performance by decreasing1221 * seeking.1222*/1223#define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \1224(\1225  (!(n_tail_size)) || \1226  (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \1227)1228 1229/*1230 * values for s_umount_state field1231 */1232#define REISERFS_VALID_FS    11233#define REISERFS_ERROR_FS    21234 1235/*1236 * there are 5 item types currently1237 */1238#define TYPE_STAT_DATA 01239#define TYPE_INDIRECT 11240#define TYPE_DIRECT 21241#define TYPE_DIRENTRY 31242#define TYPE_MAXTYPE 31243#define TYPE_ANY 15		/* FIXME: comment is required */1244 1245/***************************************************************************1246 *                       KEY & ITEM HEAD                                   *1247 ***************************************************************************/1248 1249/* * directories use this key as well as old files */1250struct offset_v1 {1251	__le32 k_offset;1252	__le32 k_uniqueness;1253} __attribute__ ((__packed__));1254 1255struct offset_v2 {1256	__le64 v;1257} __attribute__ ((__packed__));1258 1259static inline __u16 offset_v2_k_type(const struct offset_v2 *v2)1260{1261	__u8 type = le64_to_cpu(v2->v) >> 60;1262	return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY;1263}1264 1265static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type)1266{1267	v2->v =1268	    (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60);1269}1270 1271static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2)1272{1273	return le64_to_cpu(v2->v) & (~0ULL >> 4);1274}1275 1276static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset)1277{1278	offset &= (~0ULL >> 4);1279	v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset);1280}1281 1282/*1283 * Key of an item determines its location in the S+tree, and1284 * is composed of 4 components1285 */1286struct reiserfs_key {1287	/* packing locality: by default parent directory object id */1288	__le32 k_dir_id;1289 1290	__le32 k_objectid;	/* object identifier */1291	union {1292		struct offset_v1 k_offset_v1;1293		struct offset_v2 k_offset_v2;1294	} __attribute__ ((__packed__)) u;1295} __attribute__ ((__packed__));1296 1297struct in_core_key {1298	/* packing locality: by default parent directory object id */1299	__u32 k_dir_id;1300	__u32 k_objectid;	/* object identifier */1301	__u64 k_offset;1302	__u8 k_type;1303};1304 1305struct cpu_key {1306	struct in_core_key on_disk_key;1307	int version;1308	/* 3 in all cases but direct2indirect and indirect2direct conversion */1309	int key_length;1310};1311 1312/*1313 * Our function for comparing keys can compare keys of different1314 * lengths.  It takes as a parameter the length of the keys it is to1315 * compare.  These defines are used in determining what is to be passed1316 * to it as that parameter.1317 */1318#define REISERFS_FULL_KEY_LEN     41319#define REISERFS_SHORT_KEY_LEN    21320 1321/* The result of the key compare */1322#define FIRST_GREATER 11323#define SECOND_GREATER -11324#define KEYS_IDENTICAL 01325#define KEY_FOUND 11326#define KEY_NOT_FOUND 01327 1328#define KEY_SIZE (sizeof(struct reiserfs_key))1329 1330/* return values for search_by_key and clones */1331#define ITEM_FOUND 11332#define ITEM_NOT_FOUND 01333#define ENTRY_FOUND 11334#define ENTRY_NOT_FOUND 01335#define DIRECTORY_NOT_FOUND -11336#define REGULAR_FILE_FOUND -21337#define DIRECTORY_FOUND -31338#define BYTE_FOUND 11339#define BYTE_NOT_FOUND 01340#define FILE_NOT_FOUND -11341 1342#define POSITION_FOUND 11343#define POSITION_NOT_FOUND 01344 1345/* return values for reiserfs_find_entry and search_by_entry_key */1346#define NAME_FOUND 11347#define NAME_NOT_FOUND 01348#define GOTO_PREVIOUS_ITEM 21349#define NAME_FOUND_INVISIBLE 31350 1351/*1352 * Everything in the filesystem is stored as a set of items.  The1353 * item head contains the key of the item, its free space (for1354 * indirect items) and specifies the location of the item itself1355 * within the block.1356 */1357 1358struct item_head {1359	/*1360	 * Everything in the tree is found by searching for it based on1361	 * its key.1362	 */1363	struct reiserfs_key ih_key;1364	union {1365		/*1366		 * The free space in the last unformatted node of an1367		 * indirect item if this is an indirect item.  This1368		 * equals 0xFFFF iff this is a direct item or stat data1369		 * item. Note that the key, not this field, is used to1370		 * determine the item type, and thus which field this1371		 * union contains.1372		 */1373		__le16 ih_free_space_reserved;1374 1375		/*1376		 * Iff this is a directory item, this field equals the1377		 * number of directory entries in the directory item.1378		 */1379		__le16 ih_entry_count;1380	} __attribute__ ((__packed__)) u;1381	__le16 ih_item_len;	/* total size of the item body */1382 1383	/* an offset to the item body within the block */1384	__le16 ih_item_location;1385 1386	/*1387	 * 0 for all old items, 2 for new ones. Highest bit is set by fsck1388	 * temporary, cleaned after all done1389	 */1390	__le16 ih_version;1391} __attribute__ ((__packed__));1392/* size of item header     */1393#define IH_SIZE (sizeof(struct item_head))1394 1395#define ih_free_space(ih)            le16_to_cpu((ih)->u.ih_free_space_reserved)1396#define ih_version(ih)               le16_to_cpu((ih)->ih_version)1397#define ih_entry_count(ih)           le16_to_cpu((ih)->u.ih_entry_count)1398#define ih_location(ih)              le16_to_cpu((ih)->ih_item_location)1399#define ih_item_len(ih)              le16_to_cpu((ih)->ih_item_len)1400 1401#define put_ih_free_space(ih, val)   do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0)1402#define put_ih_version(ih, val)      do { (ih)->ih_version = cpu_to_le16(val); } while (0)1403#define put_ih_entry_count(ih, val)  do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0)1404#define put_ih_location(ih, val)     do { (ih)->ih_item_location = cpu_to_le16(val); } while (0)1405#define put_ih_item_len(ih, val)     do { (ih)->ih_item_len = cpu_to_le16(val); } while (0)1406 1407#define unreachable_item(ih) (ih_version(ih) & (1 << 15))1408 1409#define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih))1410#define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val)))1411 1412/*1413 * these operate on indirect items, where you've got an array of ints1414 * at a possibly unaligned location.  These are a noop on ia321415 *1416 * p is the array of __u32, i is the index into the array, v is the value1417 * to store there.1418 */1419#define get_block_num(p, i) get_unaligned_le32((p) + (i))1420#define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i))1421 1422/* * in old version uniqueness field shows key type */1423#define V1_SD_UNIQUENESS 01424#define V1_INDIRECT_UNIQUENESS 0xfffffffe1425#define V1_DIRECT_UNIQUENESS 0xffffffff1426#define V1_DIRENTRY_UNIQUENESS 5001427#define V1_ANY_UNIQUENESS 555	/* FIXME: comment is required */1428 1429/* here are conversion routines */1430static inline int uniqueness2type(__u32 uniqueness) CONSTF;1431static inline int uniqueness2type(__u32 uniqueness)1432{1433	switch ((int)uniqueness) {1434	case V1_SD_UNIQUENESS:1435		return TYPE_STAT_DATA;1436	case V1_INDIRECT_UNIQUENESS:1437		return TYPE_INDIRECT;1438	case V1_DIRECT_UNIQUENESS:1439		return TYPE_DIRECT;1440	case V1_DIRENTRY_UNIQUENESS:1441		return TYPE_DIRENTRY;1442	case V1_ANY_UNIQUENESS:1443	default:1444		return TYPE_ANY;1445	}1446}1447 1448static inline __u32 type2uniqueness(int type) CONSTF;1449static inline __u32 type2uniqueness(int type)1450{1451	switch (type) {1452	case TYPE_STAT_DATA:1453		return V1_SD_UNIQUENESS;1454	case TYPE_INDIRECT:1455		return V1_INDIRECT_UNIQUENESS;1456	case TYPE_DIRECT:1457		return V1_DIRECT_UNIQUENESS;1458	case TYPE_DIRENTRY:1459		return V1_DIRENTRY_UNIQUENESS;1460	case TYPE_ANY:1461	default:1462		return V1_ANY_UNIQUENESS;1463	}1464}1465 1466/*1467 * key is pointer to on disk key which is stored in le, result is cpu,1468 * there is no way to get version of object from key, so, provide1469 * version to these defines1470 */1471static inline loff_t le_key_k_offset(int version,1472				     const struct reiserfs_key *key)1473{1474	return (version == KEY_FORMAT_3_5) ?1475	    le32_to_cpu(key->u.k_offset_v1.k_offset) :1476	    offset_v2_k_offset(&(key->u.k_offset_v2));1477}1478 1479static inline loff_t le_ih_k_offset(const struct item_head *ih)1480{1481	return le_key_k_offset(ih_version(ih), &(ih->ih_key));1482}1483 1484static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key)1485{1486	if (version == KEY_FORMAT_3_5) {1487		loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness);1488		return uniqueness2type(val);1489	} else1490		return offset_v2_k_type(&(key->u.k_offset_v2));1491}1492 1493static inline loff_t le_ih_k_type(const struct item_head *ih)1494{1495	return le_key_k_type(ih_version(ih), &(ih->ih_key));1496}1497 1498static inline void set_le_key_k_offset(int version, struct reiserfs_key *key,1499				       loff_t offset)1500{1501	if (version == KEY_FORMAT_3_5)1502		key->u.k_offset_v1.k_offset = cpu_to_le32(offset);1503	else1504		set_offset_v2_k_offset(&key->u.k_offset_v2, offset);1505}1506 1507static inline void add_le_key_k_offset(int version, struct reiserfs_key *key,1508				       loff_t offset)1509{1510	set_le_key_k_offset(version, key,1511			    le_key_k_offset(version, key) + offset);1512}1513 1514static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset)1515{1516	add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);1517}1518 1519static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset)1520{1521	set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset);1522}1523 1524static inline void set_le_key_k_type(int version, struct reiserfs_key *key,1525				     int type)1526{1527	if (version == KEY_FORMAT_3_5) {1528		type = type2uniqueness(type);1529		key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type);1530	} else1531	       set_offset_v2_k_type(&key->u.k_offset_v2, type);1532}1533 1534static inline void set_le_ih_k_type(struct item_head *ih, int type)1535{1536	set_le_key_k_type(ih_version(ih), &(ih->ih_key), type);1537}1538 1539static inline int is_direntry_le_key(int version, struct reiserfs_key *key)1540{1541	return le_key_k_type(version, key) == TYPE_DIRENTRY;1542}1543 1544static inline int is_direct_le_key(int version, struct reiserfs_key *key)1545{1546	return le_key_k_type(version, key) == TYPE_DIRECT;1547}1548 1549static inline int is_indirect_le_key(int version, struct reiserfs_key *key)1550{1551	return le_key_k_type(version, key) == TYPE_INDIRECT;1552}1553 1554static inline int is_statdata_le_key(int version, struct reiserfs_key *key)1555{1556	return le_key_k_type(version, key) == TYPE_STAT_DATA;1557}1558 1559/* item header has version.  */1560static inline int is_direntry_le_ih(struct item_head *ih)1561{1562	return is_direntry_le_key(ih_version(ih), &ih->ih_key);1563}1564 1565static inline int is_direct_le_ih(struct item_head *ih)1566{1567	return is_direct_le_key(ih_version(ih), &ih->ih_key);1568}1569 1570static inline int is_indirect_le_ih(struct item_head *ih)1571{1572	return is_indirect_le_key(ih_version(ih), &ih->ih_key);1573}1574 1575static inline int is_statdata_le_ih(struct item_head *ih)1576{1577	return is_statdata_le_key(ih_version(ih), &ih->ih_key);1578}1579 1580/* key is pointer to cpu key, result is cpu */1581static inline loff_t cpu_key_k_offset(const struct cpu_key *key)1582{1583	return key->on_disk_key.k_offset;1584}1585 1586static inline loff_t cpu_key_k_type(const struct cpu_key *key)1587{1588	return key->on_disk_key.k_type;1589}1590 1591static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset)1592{1593	key->on_disk_key.k_offset = offset;1594}1595 1596static inline void set_cpu_key_k_type(struct cpu_key *key, int type)1597{1598	key->on_disk_key.k_type = type;1599}1600 1601static inline void cpu_key_k_offset_dec(struct cpu_key *key)1602{1603	key->on_disk_key.k_offset--;1604}1605 1606#define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY)1607#define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT)1608#define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT)1609#define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA)1610 1611/* are these used ? */1612#define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key)))1613#define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key)))1614#define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key)))1615#define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key)))1616 1617#define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \1618    (!COMP_SHORT_KEYS(ih, key) && \1619	  I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize))1620 1621/* maximal length of item */1622#define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE)1623#define MIN_ITEM_LEN 11624 1625/* object identifier for root dir */1626#define REISERFS_ROOT_OBJECTID 21627#define REISERFS_ROOT_PARENT_OBJECTID 11628 1629extern struct reiserfs_key root_key;1630 1631/*1632 * Picture represents a leaf of the S+tree1633 *  ______________________________________________________1634 * |      |  Array of     |                   |           |1635 * |Block |  Object-Item  |      F r e e      |  Objects- |1636 * | head |  Headers      |     S p a c e     |   Items   |1637 * |______|_______________|___________________|___________|1638 */1639 1640/*1641 * Header of a disk block.  More precisely, header of a formatted leaf1642 * or internal node, and not the header of an unformatted node.1643 */1644struct block_head {1645	__le16 blk_level;	/* Level of a block in the tree. */1646	__le16 blk_nr_item;	/* Number of keys/items in a block. */1647	__le16 blk_free_space;	/* Block free space in bytes. */1648	__le16 blk_reserved;1649	/* dump this in v4/planA */1650 1651	/* kept only for compatibility */1652	struct reiserfs_key blk_right_delim_key;1653};1654 1655#define BLKH_SIZE                     (sizeof(struct block_head))1656#define blkh_level(p_blkh)            (le16_to_cpu((p_blkh)->blk_level))1657#define blkh_nr_item(p_blkh)          (le16_to_cpu((p_blkh)->blk_nr_item))1658#define blkh_free_space(p_blkh)       (le16_to_cpu((p_blkh)->blk_free_space))1659#define blkh_reserved(p_blkh)         (le16_to_cpu((p_blkh)->blk_reserved))1660#define set_blkh_level(p_blkh,val)    ((p_blkh)->blk_level = cpu_to_le16(val))1661#define set_blkh_nr_item(p_blkh,val)  ((p_blkh)->blk_nr_item = cpu_to_le16(val))1662#define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val))1663#define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val))1664#define blkh_right_delim_key(p_blkh)  ((p_blkh)->blk_right_delim_key)1665#define set_blkh_right_delim_key(p_blkh,val)  ((p_blkh)->blk_right_delim_key = val)1666 1667/* values for blk_level field of the struct block_head */1668 1669/*1670 * When node gets removed from the tree its blk_level is set to FREE_LEVEL.1671 * It is then  used to see whether the node is still in the tree1672 */1673#define FREE_LEVEL 01674 1675#define DISK_LEAF_NODE_LEVEL  1	/* Leaf node level. */1676 1677/*1678 * Given the buffer head of a formatted node, resolve to the1679 * block head of that node.1680 */1681#define B_BLK_HEAD(bh)			((struct block_head *)((bh)->b_data))1682/* Number of items that are in buffer. */1683#define B_NR_ITEMS(bh)			(blkh_nr_item(B_BLK_HEAD(bh)))1684#define B_LEVEL(bh)			(blkh_level(B_BLK_HEAD(bh)))1685#define B_FREE_SPACE(bh)		(blkh_free_space(B_BLK_HEAD(bh)))1686 1687#define PUT_B_NR_ITEMS(bh, val)		do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0)1688#define PUT_B_LEVEL(bh, val)		do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0)1689#define PUT_B_FREE_SPACE(bh, val)	do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0)1690 1691/* Get right delimiting key. -- little endian */1692#define B_PRIGHT_DELIM_KEY(bh)		(&(blk_right_delim_key(B_BLK_HEAD(bh))))1693 1694/* Does the buffer contain a disk leaf. */1695#define B_IS_ITEMS_LEVEL(bh)		(B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL)1696 1697/* Does the buffer contain a disk internal node */1698#define B_IS_KEYS_LEVEL(bh)      (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \1699					    && B_LEVEL(bh) <= MAX_HEIGHT)1700 1701/***************************************************************************1702 *                             STAT DATA                                   *1703 ***************************************************************************/1704 1705/*1706 * old stat data is 32 bytes long. We are going to distinguish new one by1707 * different size1708*/1709struct stat_data_v1 {1710	__le16 sd_mode;		/* file type, permissions */1711	__le16 sd_nlink;	/* number of hard links */1712	__le16 sd_uid;		/* owner */1713	__le16 sd_gid;		/* group */1714	__le32 sd_size;		/* file size */1715	__le32 sd_atime;	/* time of last access */1716	__le32 sd_mtime;	/* time file was last modified  */1717 1718	/*1719	 * time inode (stat data) was last changed1720	 * (except changes to sd_atime and sd_mtime)1721	 */1722	__le32 sd_ctime;1723	union {1724		__le32 sd_rdev;1725		__le32 sd_blocks;	/* number of blocks file uses */1726	} __attribute__ ((__packed__)) u;1727 1728	/*1729	 * first byte of file which is stored in a direct item: except that if1730	 * it equals 1 it is a symlink and if it equals ~(__u32)0 there is no1731	 * direct item.  The existence of this field really grates on me.1732	 * Let's replace it with a macro based on sd_size and our tail1733	 * suppression policy.  Someday.  -Hans1734	 */1735	__le32 sd_first_direct_byte;1736} __attribute__ ((__packed__));1737 1738#define SD_V1_SIZE              (sizeof(struct stat_data_v1))1739#define stat_data_v1(ih)        (ih_version (ih) == KEY_FORMAT_3_5)1740#define sd_v1_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))1741#define set_sd_v1_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))1742#define sd_v1_nlink(sdp)        (le16_to_cpu((sdp)->sd_nlink))1743#define set_sd_v1_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le16(v))1744#define sd_v1_uid(sdp)          (le16_to_cpu((sdp)->sd_uid))1745#define set_sd_v1_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le16(v))1746#define sd_v1_gid(sdp)          (le16_to_cpu((sdp)->sd_gid))1747#define set_sd_v1_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le16(v))1748#define sd_v1_size(sdp)         (le32_to_cpu((sdp)->sd_size))1749#define set_sd_v1_size(sdp,v)   ((sdp)->sd_size = cpu_to_le32(v))1750#define sd_v1_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))1751#define set_sd_v1_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))1752#define sd_v1_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))1753#define set_sd_v1_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))1754#define sd_v1_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))1755#define set_sd_v1_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))1756#define sd_v1_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))1757#define set_sd_v1_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))1758#define sd_v1_blocks(sdp)       (le32_to_cpu((sdp)->u.sd_blocks))1759#define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v))1760#define sd_v1_first_direct_byte(sdp) \1761                                (le32_to_cpu((sdp)->sd_first_direct_byte))1762#define set_sd_v1_first_direct_byte(sdp,v) \1763                                ((sdp)->sd_first_direct_byte = cpu_to_le32(v))1764 1765/* inode flags stored in sd_attrs (nee sd_reserved) */1766 1767/*1768 * we want common flags to have the same values as in ext2,1769 * so chattr(1) will work without problems1770 */1771#define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL1772#define REISERFS_APPEND_FL    FS_APPEND_FL1773#define REISERFS_SYNC_FL      FS_SYNC_FL1774#define REISERFS_NOATIME_FL   FS_NOATIME_FL1775#define REISERFS_NODUMP_FL    FS_NODUMP_FL1776#define REISERFS_SECRM_FL     FS_SECRM_FL1777#define REISERFS_UNRM_FL      FS_UNRM_FL1778#define REISERFS_COMPR_FL     FS_COMPR_FL1779#define REISERFS_NOTAIL_FL    FS_NOTAIL_FL1780 1781/* persistent flags that file inherits from the parent directory */1782#define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL |	\1783				REISERFS_SYNC_FL |	\1784				REISERFS_NOATIME_FL |	\1785				REISERFS_NODUMP_FL |	\1786				REISERFS_SECRM_FL |	\1787				REISERFS_COMPR_FL |	\1788				REISERFS_NOTAIL_FL )1789 1790/*1791 * Stat Data on disk (reiserfs version of UFS disk inode minus the1792 * address blocks)1793 */1794struct stat_data {1795	__le16 sd_mode;		/* file type, permissions */1796	__le16 sd_attrs;	/* persistent inode flags */1797	__le32 sd_nlink;	/* number of hard links */1798	__le64 sd_size;		/* file size */1799	__le32 sd_uid;		/* owner */1800	__le32 sd_gid;		/* group */1801	__le32 sd_atime;	/* time of last access */1802	__le32 sd_mtime;	/* time file was last modified  */1803 1804	/*1805	 * time inode (stat data) was last changed1806	 * (except changes to sd_atime and sd_mtime)1807	 */1808	__le32 sd_ctime;1809	__le32 sd_blocks;1810	union {1811		__le32 sd_rdev;1812		__le32 sd_generation;1813	} __attribute__ ((__packed__)) u;1814} __attribute__ ((__packed__));1815 1816/* this is 44 bytes long */1817#define SD_SIZE (sizeof(struct stat_data))1818#define SD_V2_SIZE              SD_SIZE1819#define stat_data_v2(ih)        (ih_version (ih) == KEY_FORMAT_3_6)1820#define sd_v2_mode(sdp)         (le16_to_cpu((sdp)->sd_mode))1821#define set_sd_v2_mode(sdp,v)   ((sdp)->sd_mode = cpu_to_le16(v))1822/* sd_reserved */1823/* set_sd_reserved */1824#define sd_v2_nlink(sdp)        (le32_to_cpu((sdp)->sd_nlink))1825#define set_sd_v2_nlink(sdp,v)  ((sdp)->sd_nlink = cpu_to_le32(v))1826#define sd_v2_size(sdp)         (le64_to_cpu((sdp)->sd_size))1827#define set_sd_v2_size(sdp,v)   ((sdp)->sd_size = cpu_to_le64(v))1828#define sd_v2_uid(sdp)          (le32_to_cpu((sdp)->sd_uid))1829#define set_sd_v2_uid(sdp,v)    ((sdp)->sd_uid = cpu_to_le32(v))1830#define sd_v2_gid(sdp)          (le32_to_cpu((sdp)->sd_gid))1831#define set_sd_v2_gid(sdp,v)    ((sdp)->sd_gid = cpu_to_le32(v))1832#define sd_v2_atime(sdp)        (le32_to_cpu((sdp)->sd_atime))1833#define set_sd_v2_atime(sdp,v)  ((sdp)->sd_atime = cpu_to_le32(v))1834#define sd_v2_mtime(sdp)        (le32_to_cpu((sdp)->sd_mtime))1835#define set_sd_v2_mtime(sdp,v)  ((sdp)->sd_mtime = cpu_to_le32(v))1836#define sd_v2_ctime(sdp)        (le32_to_cpu((sdp)->sd_ctime))1837#define set_sd_v2_ctime(sdp,v)  ((sdp)->sd_ctime = cpu_to_le32(v))1838#define sd_v2_blocks(sdp)       (le32_to_cpu((sdp)->sd_blocks))1839#define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v))1840#define sd_v2_rdev(sdp)         (le32_to_cpu((sdp)->u.sd_rdev))1841#define set_sd_v2_rdev(sdp,v)   ((sdp)->u.sd_rdev = cpu_to_le32(v))1842#define sd_v2_generation(sdp)   (le32_to_cpu((sdp)->u.sd_generation))1843#define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v))1844#define sd_v2_attrs(sdp)         (le16_to_cpu((sdp)->sd_attrs))1845#define set_sd_v2_attrs(sdp,v)   ((sdp)->sd_attrs = cpu_to_le16(v))1846 1847/***************************************************************************1848 *                      DIRECTORY STRUCTURE                                *1849 ***************************************************************************/1850/*1851 * Picture represents the structure of directory items1852 * ________________________________________________1853 * |  Array of     |   |     |        |       |   |1854 * | directory     |N-1| N-2 | ....   |   1st |0th|1855 * | entry headers |   |     |        |       |   |1856 * |_______________|___|_____|________|_______|___|1857 *                  <----   directory entries         ------>1858 *1859 * First directory item has k_offset component 1. We store "." and ".."1860 * in one item, always, we never split "." and ".." into differing1861 * items.  This makes, among other things, the code for removing1862 * directories simpler.1863 */1864#define SD_OFFSET  01865#define SD_UNIQUENESS 01866#define DOT_OFFSET 11867#define DOT_DOT_OFFSET 21868#define DIRENTRY_UNIQUENESS 5001869 1870#define FIRST_ITEM_OFFSET 11871 1872/*1873 * Q: How to get key of object pointed to by entry from entry?1874 *1875 * A: Each directory entry has its header. This header has deh_dir_id1876 *    and deh_objectid fields, those are key of object, entry points to1877 */1878 1879/*1880 * NOT IMPLEMENTED:1881 * Directory will someday contain stat data of object1882 */1883 1884struct reiserfs_de_head {1885	__le32 deh_offset;	/* third component of the directory entry key */1886 1887	/*1888	 * objectid of the parent directory of the object, that is referenced1889	 * by directory entry1890	 */1891	__le32 deh_dir_id;1892 1893	/* objectid of the object, that is referenced by directory entry */1894	__le32 deh_objectid;1895	__le16 deh_location;	/* offset of name in the whole item */1896 1897	/*1898	 * whether 1) entry contains stat data (for future), and1899	 * 2) whether entry is hidden (unlinked)1900	 */1901	__le16 deh_state;1902} __attribute__ ((__packed__));1903#define DEH_SIZE                  sizeof(struct reiserfs_de_head)1904#define deh_offset(p_deh)         (le32_to_cpu((p_deh)->deh_offset))1905#define deh_dir_id(p_deh)         (le32_to_cpu((p_deh)->deh_dir_id))1906#define deh_objectid(p_deh)       (le32_to_cpu((p_deh)->deh_objectid))1907#define deh_location(p_deh)       (le16_to_cpu((p_deh)->deh_location))1908#define deh_state(p_deh)          (le16_to_cpu((p_deh)->deh_state))1909 1910#define put_deh_offset(p_deh,v)   ((p_deh)->deh_offset = cpu_to_le32((v)))1911#define put_deh_dir_id(p_deh,v)   ((p_deh)->deh_dir_id = cpu_to_le32((v)))1912#define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v)))1913#define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v)))1914#define put_deh_state(p_deh,v)    ((p_deh)->deh_state = cpu_to_le16((v)))1915 1916/* empty directory contains two entries "." and ".." and their headers */1917#define EMPTY_DIR_SIZE \1918(DEH_SIZE * 2 + ROUND_UP (sizeof(".") - 1) + ROUND_UP (sizeof("..") - 1))1919 1920/* old format directories have this size when empty */1921#define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3)1922 1923#define DEH_Statdata 0		/* not used now */1924#define DEH_Visible 21925 1926/* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */1927#if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__)1928#   define ADDR_UNALIGNED_BITS  (3)1929#endif1930 1931/*1932 * These are only used to manipulate deh_state.1933 * Because of this, we'll use the ext2_ bit routines,1934 * since they are little endian1935 */1936#ifdef ADDR_UNALIGNED_BITS1937 1938#   define aligned_address(addr)           ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1)))1939#   define unaligned_offset(addr)          (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3)1940 1941#   define set_bit_unaligned(nr, addr)	\1942	__test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))1943#   define clear_bit_unaligned(nr, addr)	\1944	__test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))1945#   define test_bit_unaligned(nr, addr)	\1946	test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr))1947 1948#else1949 1950#   define set_bit_unaligned(nr, addr)	__test_and_set_bit_le(nr, addr)1951#   define clear_bit_unaligned(nr, addr)	__test_and_clear_bit_le(nr, addr)1952#   define test_bit_unaligned(nr, addr)	test_bit_le(nr, addr)1953 1954#endif1955 1956#define mark_de_with_sd(deh)        set_bit_unaligned (DEH_Statdata, &((deh)->deh_state))1957#define mark_de_without_sd(deh)     clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state))1958#define mark_de_visible(deh)	    set_bit_unaligned (DEH_Visible, &((deh)->deh_state))1959#define mark_de_hidden(deh)	    clear_bit_unaligned (DEH_Visible, &((deh)->deh_state))1960 1961#define de_with_sd(deh)		    test_bit_unaligned (DEH_Statdata, &((deh)->deh_state))1962#define de_visible(deh)	    	    test_bit_unaligned (DEH_Visible, &((deh)->deh_state))1963#define de_hidden(deh)	    	    !test_bit_unaligned (DEH_Visible, &((deh)->deh_state))1964 1965extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid,1966				   __le32 par_dirid, __le32 par_objid);1967extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid,1968				__le32 par_dirid, __le32 par_objid);1969 1970/* two entries per block (at least) */1971#define REISERFS_MAX_NAME(block_size) 2551972 1973/*1974 * this structure is used for operations on directory entries. It is1975 * not a disk structure.1976 *1977 * When reiserfs_find_entry or search_by_entry_key find directory1978 * entry, they return filled reiserfs_dir_entry structure1979 */1980struct reiserfs_dir_entry {1981	struct buffer_head *de_bh;1982	int de_item_num;1983	struct item_head *de_ih;1984	int de_entry_num;1985	struct reiserfs_de_head *de_deh;1986	int de_entrylen;1987	int de_namelen;1988	char *de_name;1989	unsigned long *de_gen_number_bit_string;1990 1991	__u32 de_dir_id;1992	__u32 de_objectid;1993 1994	struct cpu_key de_entry_key;1995};1996 1997/*1998 * these defines are useful when a particular member of1999 * a reiserfs_dir_entry is needed2000 */2001 2002/* pointer to file name, stored in entry */2003#define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \2004				(ih_item_body(bh, ih) + deh_location(deh))2005 2006/* length of name */2007#define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \2008(I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0))2009 2010/* hash value occupies bits from 7 up to 30 */2011#define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL)2012/* generation number occupies 7 bits starting from 0 up to 6 */2013#define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL)2014#define MAX_GENERATION_NUMBER  1272015 2016#define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number))2017 2018/*2019 * Picture represents an internal node of the reiserfs tree2020 *  ______________________________________________________2021 * |      |  Array of     |  Array of         |  Free     |2022 * |block |    keys       |  pointers         | space     |2023 * | head |      N        |      N+1          |           |2024 * |______|_______________|___________________|___________|2025 */2026 2027/***************************************************************************2028 *                      DISK CHILD                                         *2029 ***************************************************************************/2030/*2031 * Disk child pointer:2032 * The pointer from an internal node of the tree to a node that is on disk.2033 */2034struct disk_child {2035	__le32 dc_block_number;	/* Disk child's block number. */2036	__le16 dc_size;		/* Disk child's used space.   */2037	__le16 dc_reserved;2038};2039 2040#define DC_SIZE (sizeof(struct disk_child))2041#define dc_block_number(dc_p)	(le32_to_cpu((dc_p)->dc_block_number))2042#define dc_size(dc_p)		(le16_to_cpu((dc_p)->dc_size))2043#define put_dc_block_number(dc_p, val)   do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0)2044#define put_dc_size(dc_p, val)   do { (dc_p)->dc_size = cpu_to_le16(val); } while(0)2045 2046/* Get disk child by buffer header and position in the tree node. */2047#define B_N_CHILD(bh, n_pos)  ((struct disk_child *)\2048((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos)))2049 2050/* Get disk child number by buffer header and position in the tree node. */2051#define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos)))2052#define PUT_B_N_CHILD_NUM(bh, n_pos, val) \2053				(put_dc_block_number(B_N_CHILD(bh, n_pos), val))2054 2055 /* maximal value of field child_size in structure disk_child */2056 /* child size is the combined size of all items and their headers */2057#define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE ))2058 2059/* amount of used space in buffer (not including block head) */2060#define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur)))2061 2062/* max and min number of keys in internal node */2063#define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) )2064#define MIN_NR_KEY(bh)    (MAX_NR_KEY(bh)/2)2065 2066/***************************************************************************2067 *                      PATH STRUCTURES AND DEFINES                        *2068 ***************************************************************************/2069 2070/*2071 * search_by_key fills up the path from the root to the leaf as it descends2072 * the tree looking for the key.  It uses reiserfs_bread to try to find2073 * buffers in the cache given their block number.  If it does not find2074 * them in the cache it reads them from disk.  For each node search_by_key2075 * finds using reiserfs_bread it then uses bin_search to look through that2076 * node.  bin_search will find the position of the block_number of the next2077 * node if it is looking through an internal node.  If it is looking through2078 * a leaf node bin_search will find the position of the item which has key2079 * either equal to given key, or which is the maximal key less than the2080 * given key.2081 */2082 2083struct path_element {2084	/* Pointer to the buffer at the path in the tree. */2085	struct buffer_head *pe_buffer;2086	/* Position in the tree node which is placed in the buffer above. */2087	int pe_position;2088};2089 2090/*2091 * maximal height of a tree. don't change this without2092 * changing JOURNAL_PER_BALANCE_CNT2093 */2094#define MAX_HEIGHT 52095 2096/* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */2097#define EXTENDED_MAX_HEIGHT         72098 2099/* Must be equal to at least 2. */2100#define FIRST_PATH_ELEMENT_OFFSET   22101 2102/* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */2103#define ILLEGAL_PATH_ELEMENT_OFFSET 12104 2105/* this MUST be MAX_HEIGHT + 1. See about FEB below */2106#define MAX_FEB_SIZE 62107 2108/*2109 * We need to keep track of who the ancestors of nodes are.  When we2110 * perform a search we record which nodes were visited while2111 * descending the tree looking for the node we searched for. This list2112 * of nodes is called the path.  This information is used while2113 * performing balancing.  Note that this path information may become2114 * invalid, and this means we must check it when using it to see if it2115 * is still valid. You'll need to read search_by_key and the comments2116 * in it, especially about decrement_counters_in_path(), to understand2117 * this structure.2118 *2119 * Paths make the code so much harder to work with and debug.... An2120 * enormous number of bugs are due to them, and trying to write or modify2121 * code that uses them just makes my head hurt.  They are based on an2122 * excessive effort to avoid disturbing the precious VFS code.:-( The2123 * gods only know how we are going to SMP the code that uses them.2124 * znodes are the way!2125 */2126 2127#define PATH_READA	0x1	/* do read ahead */2128#define PATH_READA_BACK 0x2	/* read backwards */2129 2130struct treepath {2131	int path_length;	/* Length of the array above.   */2132	int reada;2133	/* Array of the path elements.  */2134	struct path_element path_elements[EXTENDED_MAX_HEIGHT];2135	int pos_in_item;2136};2137 2138#define pos_in_item(path) ((path)->pos_in_item)2139 2140#define INITIALIZE_PATH(var) \2141struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,}2142 2143/* Get path element by path and path position. */2144#define PATH_OFFSET_PELEMENT(path, n_offset)  ((path)->path_elements + (n_offset))2145 2146/* Get buffer header at the path by path and path position. */2147#define PATH_OFFSET_PBUFFER(path, n_offset)   (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer)2148 2149/* Get position in the element at the path by path and path position. */2150#define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position)2151 2152#define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length))2153 2154/*2155 * you know, to the person who didn't write this the macro name does not2156 * at first suggest what it does.  Maybe POSITION_FROM_PATH_END? Or2157 * maybe we should just focus on dumping paths... -Hans2158 */2159#define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length))2160 2161/*2162 * in do_balance leaf has h == 0 in contrast with path structure,2163 * where root has level == 0. That is why we need these defines2164 */2165 2166/* tb->S[h] */2167#define PATH_H_PBUFFER(path, h) \2168			PATH_OFFSET_PBUFFER(path, path->path_length - (h))2169 2170/* tb->F[h] or tb->S[0]->b_parent */2171#define PATH_H_PPARENT(path, h) PATH_H_PBUFFER(path, (h) + 1)2172 2173#define PATH_H_POSITION(path, h) \2174			PATH_OFFSET_POSITION(path, path->path_length - (h))2175 2176/* tb->S[h]->b_item_order */2177#define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1)2178 2179#define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h))2180 2181static inline void *reiserfs_node_data(const struct buffer_head *bh)2182{2183	return bh->b_data + sizeof(struct block_head);2184}2185 2186/* get key from internal node */2187static inline struct reiserfs_key *internal_key(struct buffer_head *bh,2188						int item_num)2189{2190	struct reiserfs_key *key = reiserfs_node_data(bh);2191 2192	return &key[item_num];2193}2194 2195/* get the item header from leaf node */2196static inline struct item_head *item_head(const struct buffer_head *bh,2197					  int item_num)2198{2199	struct item_head *ih = reiserfs_node_data(bh);2200 2201	return &ih[item_num];2202}2203 2204/* get the key from leaf node */2205static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh,2206					    int item_num)2207{2208	return &item_head(bh, item_num)->ih_key;2209}2210 2211static inline void *ih_item_body(const struct buffer_head *bh,2212				 const struct item_head *ih)2213{2214	return bh->b_data + ih_location(ih);2215}2216 2217/* get item body from leaf node */2218static inline void *item_body(const struct buffer_head *bh, int item_num)2219{2220	return ih_item_body(bh, item_head(bh, item_num));2221}2222 2223static inline struct item_head *tp_item_head(const struct treepath *path)2224{2225	return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));2226}2227 2228static inline void *tp_item_body(const struct treepath *path)2229{2230	return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path));2231}2232 2233#define get_last_bh(path) PATH_PLAST_BUFFER(path)2234#define get_item_pos(path) PATH_LAST_POSITION(path)2235#define item_moved(ih,path) comp_items(ih, path)2236#define path_changed(ih,path) comp_items (ih, path)2237 2238/* array of the entry headers */2239 /* get item body */2240#define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih)))2241 2242/*2243 * length of the directory entry in directory item. This define2244 * calculates length of i-th directory entry using directory entry2245 * locations from dir entry head. When it calculates length of 0-th2246 * directory entry, it uses length of whole item in place of entry2247 * location of the non-existent following entry in the calculation.2248 * See picture above.2249 */2250static inline int entry_length(const struct buffer_head *bh,2251			       const struct item_head *ih, int pos_in_item)2252{2253	struct reiserfs_de_head *deh;2254 2255	deh = B_I_DEH(bh, ih) + pos_in_item;2256	if (pos_in_item)2257		return deh_location(deh - 1) - deh_location(deh);2258 2259	return ih_item_len(ih) - deh_location(deh);2260}2261 2262/***************************************************************************2263 *                       MISC                                              *2264 ***************************************************************************/2265 2266/* Size of pointer to the unformatted node. */2267#define UNFM_P_SIZE (sizeof(unp_t))2268#define UNFM_P_SHIFT 22269 2270/* in in-core inode key is stored on le form */2271#define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key))2272 2273#define MAX_UL_INT 0xffffffff2274#define MAX_INT    0x7ffffff2275#define MAX_US_INT 0xffff2276 2277// reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset2278static inline loff_t max_reiserfs_offset(struct inode *inode)2279{2280	if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5)2281		return (loff_t) U32_MAX;2282 2283	return (loff_t) ((~(__u64) 0) >> 4);2284}2285 2286#define MAX_KEY_OBJECTID	MAX_UL_INT2287 2288#define MAX_B_NUM  MAX_UL_INT2289#define MAX_FC_NUM MAX_US_INT2290 2291/* the purpose is to detect overflow of an unsigned short */2292#define REISERFS_LINK_MAX (MAX_US_INT - 1000)2293 2294/*2295 * The following defines are used in reiserfs_insert_item2296 * and reiserfs_append_item2297 */2298#define REISERFS_KERNEL_MEM		0	/* kernel memory mode */2299#define REISERFS_USER_MEM		1	/* user memory mode */2300 2301#define fs_generation(s) (REISERFS_SB(s)->s_generation_counter)2302#define get_generation(s) atomic_read (&fs_generation(s))2303#define FILESYSTEM_CHANGED_TB(tb)  (get_generation((tb)->tb_sb) != (tb)->fs_gen)2304#define __fs_changed(gen,s) (gen != get_generation (s))2305#define fs_changed(gen,s)		\2306({					\2307	reiserfs_cond_resched(s);	\2308	__fs_changed(gen, s);		\2309})2310 2311/***************************************************************************2312 *                  FIXATE NODES                                           *2313 ***************************************************************************/2314 2315#define VI_TYPE_LEFT_MERGEABLE 12316#define VI_TYPE_RIGHT_MERGEABLE 22317 2318/*2319 * To make any changes in the tree we always first find node, that2320 * contains item to be changed/deleted or place to insert a new2321 * item. We call this node S. To do balancing we need to decide what2322 * we will shift to left/right neighbor, or to a new node, where new2323 * item will be etc. To make this analysis simpler we build virtual2324 * node. Virtual node is an array of items, that will replace items of2325 * node S. (For instance if we are going to delete an item, virtual2326 * node does not contain it). Virtual node keeps information about2327 * item sizes and types, mergeability of first and last items, sizes2328 * of all entries in directory item. We use this array of items when2329 * calculating what we can shift to neighbors and how many nodes we2330 * have to have if we do not any shiftings, if we shift to left/right2331 * neighbor or to both.2332 */2333struct virtual_item {2334	int vi_index;		/* index in the array of item operations */2335	unsigned short vi_type;	/* left/right mergeability */2336 2337	/* length of item that it will have after balancing */2338	unsigned short vi_item_len;2339 2340	struct item_head *vi_ih;2341	const char *vi_item;	/* body of item (old or new) */2342	const void *vi_new_data;	/* 0 always but paste mode */2343	void *vi_uarea;		/* item specific area */2344};2345 2346struct virtual_node {2347	/* this is a pointer to the free space in the buffer */2348	char *vn_free_ptr;2349 2350	unsigned short vn_nr_item;	/* number of items in virtual node */2351 2352	/*2353	 * size of node , that node would have if it has2354	 * unlimited size and no balancing is performed2355	 */2356	short vn_size;2357 2358	/* mode of balancing (paste, insert, delete, cut) */2359	short vn_mode;2360 2361	short vn_affected_item_num;2362	short vn_pos_in_item;2363 2364	/* item header of inserted item, 0 for other modes */2365	struct item_head *vn_ins_ih;2366	const void *vn_data;2367 2368	/* array of items (including a new one, excluding item to be deleted) */2369	struct virtual_item *vn_vi;2370};2371 2372/* used by directory items when creating virtual nodes */2373struct direntry_uarea {2374	int flags;2375	__u16 entry_count;2376	__u16 entry_sizes[];2377} __attribute__ ((__packed__));2378 2379/***************************************************************************2380 *                  TREE BALANCE                                           *2381 ***************************************************************************/2382 2383/*2384 * This temporary structure is used in tree balance algorithms, and2385 * constructed as we go to the extent that its various parts are2386 * needed.  It contains arrays of nodes that can potentially be2387 * involved in the balancing of node S, and parameters that define how2388 * each of the nodes must be balanced.  Note that in these algorithms2389 * for balancing the worst case is to need to balance the current node2390 * S and the left and right neighbors and all of their parents plus2391 * create a new node.  We implement S1 balancing for the leaf nodes2392 * and S0 balancing for the internal nodes (S1 and S0 are defined in2393 * our papers.)2394 */2395 2396/* size of the array of buffers to free at end of do_balance */2397#define MAX_FREE_BLOCK 72398 2399/* maximum number of FEB blocknrs on a single level */2400#define MAX_AMOUNT_NEEDED 22401 2402/* someday somebody will prefix every field in this struct with tb_ */2403struct tree_balance {2404	int tb_mode;2405	int need_balance_dirty;2406	struct super_block *tb_sb;2407	struct reiserfs_transaction_handle *transaction_handle;2408	struct treepath *tb_path;2409 2410	/* array of left neighbors of nodes in the path */2411	struct buffer_head *L[MAX_HEIGHT];2412 2413	/* array of right neighbors of nodes in the path */2414	struct buffer_head *R[MAX_HEIGHT];2415 2416	/* array of fathers of the left neighbors */2417	struct buffer_head *FL[MAX_HEIGHT];2418 2419	/* array of fathers of the right neighbors */2420	struct buffer_head *FR[MAX_HEIGHT];2421	/* array of common parents of center node and its left neighbor */2422	struct buffer_head *CFL[MAX_HEIGHT];2423 2424	/* array of common parents of center node and its right neighbor */2425	struct buffer_head *CFR[MAX_HEIGHT];2426 2427	/*2428	 * array of empty buffers. Number of buffers in array equals2429	 * cur_blknum.2430	 */2431	struct buffer_head *FEB[MAX_FEB_SIZE];2432	struct buffer_head *used[MAX_FEB_SIZE];2433	struct buffer_head *thrown[MAX_FEB_SIZE];2434 2435	/*2436	 * array of number of items which must be shifted to the left in2437	 * order to balance the current node; for leaves includes item that2438	 * will be partially shifted; for internal nodes, it is the number2439	 * of child pointers rather than items. It includes the new item2440	 * being created. The code sometimes subtracts one to get the2441	 * number of wholly shifted items for other purposes.2442	 */2443	int lnum[MAX_HEIGHT];2444 2445	/* substitute right for left in comment above */2446	int rnum[MAX_HEIGHT];2447 2448	/*2449	 * array indexed by height h mapping the key delimiting L[h] and2450	 * S[h] to its item number within the node CFL[h]2451	 */2452	int lkey[MAX_HEIGHT];2453 2454	/* substitute r for l in comment above */2455	int rkey[MAX_HEIGHT];2456 2457	/*2458	 * the number of bytes by we are trying to add or remove from2459	 * S[h]. A negative value means removing.2460	 */2461	int insert_size[MAX_HEIGHT];2462 2463	/*2464	 * number of nodes that will replace node S[h] after balancing2465	 * on the level h of the tree.  If 0 then S is being deleted,2466	 * if 1 then S is remaining and no new nodes are being created,2467	 * if 2 or 3 then 1 or 2 new nodes is being created2468	 */2469	int blknum[MAX_HEIGHT];2470 2471	/* fields that are used only for balancing leaves of the tree */2472 2473	/* number of empty blocks having been already allocated */2474	int cur_blknum;2475 2476	/* number of items that fall into left most node when S[0] splits */2477	int s0num;2478 2479	/*2480	 * number of bytes which can flow to the left neighbor from the left2481	 * most liquid item that cannot be shifted from S[0] entirely2482	 * if -1 then nothing will be partially shifted2483	 */2484	int lbytes;2485 2486	/*2487	 * number of bytes which will flow to the right neighbor from the right2488	 * most liquid item that cannot be shifted from S[0] entirely2489	 * if -1 then nothing will be partially shifted2490	 */2491	int rbytes;2492 2493 2494	/*2495	 * index into the array of item headers in2496	 * S[0] of the affected item2497	 */2498	int item_pos;2499 2500	/* new nodes allocated to hold what could not fit into S */2501	struct buffer_head *S_new[2];2502 2503	/*2504	 * number of items that will be placed into nodes in S_new2505	 * when S[0] splits2506	 */2507	int snum[2];2508 2509	/*2510	 * number of bytes which flow to nodes in S_new when S[0] splits2511	 * note: if S[0] splits into 3 nodes, then items do not need to be cut2512	 */2513	int sbytes[2];2514 2515	int pos_in_item;2516	int zeroes_num;2517 2518	/*2519	 * buffers which are to be freed after do_balance finishes2520	 * by unfix_nodes2521	 */2522	struct buffer_head *buf_to_free[MAX_FREE_BLOCK];2523 2524	/*2525	 * kmalloced memory. Used to create virtual node and keep2526	 * map of dirtied bitmap blocks2527	 */2528	char *vn_buf;2529 2530	int vn_buf_size;	/* size of the vn_buf */2531 2532	/* VN starts after bitmap of bitmap blocks */2533	struct virtual_node *tb_vn;2534 2535	/*2536	 * saved value of `reiserfs_generation' counter see2537	 * FILESYSTEM_CHANGED() macro in reiserfs_fs.h2538	 */2539	int fs_gen;2540 2541#ifdef DISPLACE_NEW_PACKING_LOCALITIES2542	/*2543	 * key pointer, to pass to block allocator or2544	 * another low-level subsystem2545	 */2546	struct in_core_key key;2547#endif2548};2549 2550/* These are modes of balancing */2551 2552/* When inserting an item. */2553#define M_INSERT	'i'2554/*2555 * When inserting into (directories only) or appending onto an already2556 * existent item.2557 */2558#define M_PASTE		'p'2559/* When deleting an item. */2560#define M_DELETE	'd'2561/* When truncating an item or removing an entry from a (directory) item. */2562#define M_CUT		'c'2563 2564/* used when balancing on leaf level skipped (in reiserfsck) */2565#define M_INTERNAL	'n'2566 2567/*2568 * When further balancing is not needed, then do_balance does not need2569 * to be called.2570 */2571#define M_SKIP_BALANCING		's'2572#define M_CONVERT	'v'2573 2574/* modes of leaf_move_items */2575#define LEAF_FROM_S_TO_L 02576#define LEAF_FROM_S_TO_R 12577#define LEAF_FROM_R_TO_L 22578#define LEAF_FROM_L_TO_R 32579#define LEAF_FROM_S_TO_SNEW 42580 2581#define FIRST_TO_LAST 02582#define LAST_TO_FIRST 12583 2584/*2585 * used in do_balance for passing parent of node information that has2586 * been gotten from tb struct2587 */2588struct buffer_info {2589	struct tree_balance *tb;2590	struct buffer_head *bi_bh;2591	struct buffer_head *bi_parent;2592	int bi_position;2593};2594 2595static inline struct super_block *sb_from_tb(struct tree_balance *tb)2596{2597	return tb ? tb->tb_sb : NULL;2598}2599 2600static inline struct super_block *sb_from_bi(struct buffer_info *bi)2601{2602	return bi ? sb_from_tb(bi->tb) : NULL;2603}2604 2605/*2606 * there are 4 types of items: stat data, directory item, indirect, direct.2607 * +-------------------+------------+--------------+------------+2608 * |                   |  k_offset  | k_uniqueness | mergeable? |2609 * +-------------------+------------+--------------+------------+2610 * |     stat data     |     0      |      0       |   no       |2611 * +-------------------+------------+--------------+------------+2612 * | 1st directory item| DOT_OFFSET | DIRENTRY_ .. |   no       |2613 * | non 1st directory | hash value | UNIQUENESS   |   yes      |2614 * |     item          |            |              |            |2615 * +-------------------+------------+--------------+------------+2616 * | indirect item     | offset + 1 |TYPE_INDIRECT |    [1]	|2617 * +-------------------+------------+--------------+------------+2618 * | direct item       | offset + 1 |TYPE_DIRECT   |    [2]     |2619 * +-------------------+------------+--------------+------------+2620 *2621 * [1] if this is not the first indirect item of the object2622 * [2] if this is not the first direct item of the object2623*/2624 2625struct item_operations {2626	int (*bytes_number) (struct item_head * ih, int block_size);2627	void (*decrement_key) (struct cpu_key *);2628	int (*is_left_mergeable) (struct reiserfs_key * ih,2629				  unsigned long bsize);2630	void (*print_item) (struct item_head *, char *item);2631	void (*check_item) (struct item_head *, char *item);2632 2633	int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi,2634			  int is_affected, int insert_size);2635	int (*check_left) (struct virtual_item * vi, int free,2636			   int start_skip, int end_skip);2637	int (*check_right) (struct virtual_item * vi, int free);2638	int (*part_size) (struct virtual_item * vi, int from, int to);2639	int (*unit_num) (struct virtual_item * vi);2640	void (*print_vi) (struct virtual_item * vi);2641};2642 2643extern struct item_operations *item_ops[TYPE_ANY + 1];2644 2645#define op_bytes_number(ih,bsize)                    item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize)2646#define op_is_left_mergeable(key,bsize)              item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize)2647#define op_print_item(ih,item)                       item_ops[le_ih_k_type (ih)]->print_item (ih, item)2648#define op_check_item(ih,item)                       item_ops[le_ih_k_type (ih)]->check_item (ih, item)2649#define op_create_vi(vn,vi,is_affected,insert_size)  item_ops[le_ih_k_type ((vi)->vi_ih)]->create_vi (vn,vi,is_affected,insert_size)2650#define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip)2651#define op_check_right(vi,free)                      item_ops[(vi)->vi_index]->check_right (vi, free)2652#define op_part_size(vi,from,to)                     item_ops[(vi)->vi_index]->part_size (vi, from, to)2653#define op_unit_num(vi)				     item_ops[(vi)->vi_index]->unit_num (vi)2654#define op_print_vi(vi)                              item_ops[(vi)->vi_index]->print_vi (vi)2655 2656#define COMP_SHORT_KEYS comp_short_keys2657 2658/* number of blocks pointed to by the indirect item */2659#define I_UNFM_NUM(ih)	(ih_item_len(ih) / UNFM_P_SIZE)2660 2661/*2662 * the used space within the unformatted node corresponding2663 * to pos within the item pointed to by ih2664 */2665#define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size))2666 2667/*2668 * number of bytes contained by the direct item or the2669 * unformatted nodes the indirect item points to2670 */2671 2672/* following defines use reiserfs buffer header and item header */2673 2674/* get stat-data */2675#define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) )2676 2677/* this is 3976 for size==4096 */2678#define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE)2679 2680/*2681 * indirect items consist of entries which contain blocknrs, pos2682 * indicates which entry, and B_I_POS_UNFM_POINTER resolves to the2683 * blocknr contained by the entry pos points to2684 */2685#define B_I_POS_UNFM_POINTER(bh, ih, pos)				\2686	le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos)))2687#define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val)			\2688	(*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val))2689 2690struct reiserfs_iget_args {2691	__u32 objectid;2692	__u32 dirid;2693};2694 2695/***************************************************************************2696 *                    FUNCTION DECLARATIONS                                *2697 ***************************************************************************/2698 2699#define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12)2700 2701#define journal_trans_half(blocksize) \2702	((blocksize - sizeof(struct reiserfs_journal_desc) - 12) / sizeof(__u32))2703 2704/* journal.c see journal.c for all the comments here */2705 2706/* first block written in a commit.  */2707struct reiserfs_journal_desc {2708	__le32 j_trans_id;	/* id of commit */2709 2710	/* length of commit. len +1 is the commit block */2711	__le32 j_len;2712 2713	__le32 j_mount_id;	/* mount id of this trans */2714	__le32 j_realblock[];	/* real locations for each block */2715};2716 2717#define get_desc_trans_id(d)   le32_to_cpu((d)->j_trans_id)2718#define get_desc_trans_len(d)  le32_to_cpu((d)->j_len)2719#define get_desc_mount_id(d)   le32_to_cpu((d)->j_mount_id)2720 2721#define set_desc_trans_id(d,val)       do { (d)->j_trans_id = cpu_to_le32 (val); } while (0)2722#define set_desc_trans_len(d,val)      do { (d)->j_len = cpu_to_le32 (val); } while (0)2723#define set_desc_mount_id(d,val)       do { (d)->j_mount_id = cpu_to_le32 (val); } while (0)2724 2725/* last block written in a commit */2726struct reiserfs_journal_commit {2727	__le32 j_trans_id;	/* must match j_trans_id from the desc block */2728	__le32 j_len;		/* ditto */2729	__le32 j_realblock[];	/* real locations for each block */2730};2731 2732#define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id)2733#define get_commit_trans_len(c)        le32_to_cpu((c)->j_len)2734#define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id)2735 2736#define set_commit_trans_id(c,val)     do { (c)->j_trans_id = cpu_to_le32 (val); } while (0)2737#define set_commit_trans_len(c,val)    do { (c)->j_len = cpu_to_le32 (val); } while (0)2738 2739/*2740 * this header block gets written whenever a transaction is considered2741 * fully flushed, and is more recent than the last fully flushed transaction.2742 * fully flushed means all the log blocks and all the real blocks are on2743 * disk, and this transaction does not need to be replayed.2744 */2745struct reiserfs_journal_header {2746	/* id of last fully flushed transaction */2747	__le32 j_last_flush_trans_id;2748 2749	/* offset in the log of where to start replay after a crash */2750	__le32 j_first_unflushed_offset;2751 2752	__le32 j_mount_id;2753	/* 12 */ struct journal_params jh_journal;2754};2755 2756/* biggest tunable defines are right here */2757#define JOURNAL_BLOCK_COUNT 8192	/* number of blocks in the journal */2758 2759/* biggest possible single transaction, don't change for now (8/3/99) */2760#define JOURNAL_TRANS_MAX_DEFAULT 10242761#define JOURNAL_TRANS_MIN_DEFAULT 2562762 2763/*2764 * max blocks to batch into one transaction,2765 * don't make this any bigger than 9002766 */2767#define JOURNAL_MAX_BATCH_DEFAULT   9002768#define JOURNAL_MIN_RATIO 22769#define JOURNAL_MAX_COMMIT_AGE 302770#define JOURNAL_MAX_TRANS_AGE 302771#define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9)2772#define JOURNAL_BLOCKS_PER_OBJECT(sb)  (JOURNAL_PER_BALANCE_CNT * 3 + \2773					 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \2774					      REISERFS_QUOTA_TRANS_BLOCKS(sb)))2775 2776#ifdef CONFIG_QUOTA2777#define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA))2778/* We need to update data and inode (atime) */2779#define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0)2780/* 1 balancing, 1 bitmap, 1 data per write + stat data update */2781#define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \2782(DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0)2783/* same as with INIT */2784#define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \2785(DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0)2786#else2787#define REISERFS_QUOTA_TRANS_BLOCKS(s) 02788#define REISERFS_QUOTA_INIT_BLOCKS(s) 02789#define REISERFS_QUOTA_DEL_BLOCKS(s) 02790#endif2791 2792/*2793 * both of these can be as low as 1, or as high as you want.  The min is the2794 * number of 4k bitmap nodes preallocated on mount. New nodes are allocated2795 * as needed, and released when transactions are committed.  On release, if2796 * the current number of nodes is > max, the node is freed, otherwise,2797 * it is put on a free list for faster use later.2798*/2799#define REISERFS_MIN_BITMAP_NODES 102800#define REISERFS_MAX_BITMAP_NODES 1002801 2802/* these are based on journal hash size of 8192 */2803#define JBH_HASH_SHIFT 132804#define JBH_HASH_MASK 81912805 2806#define _jhashfn(sb,block)	\2807	(((unsigned long)sb>>L1_CACHE_SHIFT) ^ \2808	 (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12))))2809#define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK])2810 2811/* We need these to make journal.c code more readable */2812#define journal_find_get_block(s, block) __find_get_block(\2813		file_bdev(SB_JOURNAL(s)->j_bdev_file), block, s->s_blocksize)2814#define journal_getblk(s, block) __getblk(file_bdev(SB_JOURNAL(s)->j_bdev_file),\2815		block, s->s_blocksize)2816#define journal_bread(s, block) __bread(file_bdev(SB_JOURNAL(s)->j_bdev_file),\2817		block, s->s_blocksize)2818 2819enum reiserfs_bh_state_bits {2820	BH_JDirty = BH_PrivateStart,	/* buffer is in current transaction */2821	BH_JDirty_wait,2822	/*2823	 * disk block was taken off free list before being in a2824	 * finished transaction, or written to disk. Can be reused immed.2825	 */2826	BH_JNew,2827	BH_JPrepared,2828	BH_JRestore_dirty,2829	BH_JTest,		/* debugging only will go away */2830};2831 2832BUFFER_FNS(JDirty, journaled);2833TAS_BUFFER_FNS(JDirty, journaled);2834BUFFER_FNS(JDirty_wait, journal_dirty);2835TAS_BUFFER_FNS(JDirty_wait, journal_dirty);2836BUFFER_FNS(JNew, journal_new);2837TAS_BUFFER_FNS(JNew, journal_new);2838BUFFER_FNS(JPrepared, journal_prepared);2839TAS_BUFFER_FNS(JPrepared, journal_prepared);2840BUFFER_FNS(JRestore_dirty, journal_restore_dirty);2841TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty);2842BUFFER_FNS(JTest, journal_test);2843TAS_BUFFER_FNS(JTest, journal_test);2844 2845/* transaction handle which is passed around for all journal calls */2846struct reiserfs_transaction_handle {2847	/*2848	 * super for this FS when journal_begin was called. saves calls to2849	 * reiserfs_get_super also used by nested transactions to make2850	 * sure they are nesting on the right FS _must_ be first2851	 * in the handle2852	 */2853	struct super_block *t_super;2854 2855	int t_refcount;2856	int t_blocks_logged;	/* number of blocks this writer has logged */2857	int t_blocks_allocated;	/* number of blocks this writer allocated */2858 2859	/* sanity check, equals the current trans id */2860	unsigned int t_trans_id;2861 2862	void *t_handle_save;	/* save existing current->journal_info */2863 2864	/*2865	 * if new block allocation occurres, that block2866	 * should be displaced from others2867	 */2868	unsigned displace_new_blocks:1;2869 2870	struct list_head t_list;2871};2872 2873/*2874 * used to keep track of ordered and tail writes, attached to the buffer2875 * head through b_journal_head.2876 */2877struct reiserfs_jh {2878	struct reiserfs_journal_list *jl;2879	struct buffer_head *bh;2880	struct list_head list;2881};2882 2883void reiserfs_free_jh(struct buffer_head *bh);2884int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh);2885int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh);2886int journal_mark_dirty(struct reiserfs_transaction_handle *,2887		       struct buffer_head *bh);2888 2889static inline int reiserfs_file_data_log(struct inode *inode)2890{2891	if (reiserfs_data_log(inode->i_sb) ||2892	    (REISERFS_I(inode)->i_flags & i_data_log))2893		return 1;2894	return 0;2895}2896 2897static inline int reiserfs_transaction_running(struct super_block *s)2898{2899	struct reiserfs_transaction_handle *th = current->journal_info;2900	if (th && th->t_super == s)2901		return 1;2902	if (th && th->t_super == NULL)2903		BUG();2904	return 0;2905}2906 2907static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th)2908{2909	return th->t_blocks_allocated - th->t_blocks_logged;2910}2911 2912struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct2913								    super_block2914								    *,2915								    int count);2916int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *);2917void reiserfs_vfs_truncate_file(struct inode *inode);2918int reiserfs_commit_page(struct inode *inode, struct page *page,2919			 unsigned from, unsigned to);2920void reiserfs_flush_old_commits(struct super_block *);2921int reiserfs_commit_for_inode(struct inode *);2922int reiserfs_inode_needs_commit(struct inode *);2923void reiserfs_update_inode_transaction(struct inode *);2924void reiserfs_wait_on_write_block(struct super_block *s);2925void reiserfs_block_writes(struct reiserfs_transaction_handle *th);2926void reiserfs_allow_writes(struct super_block *s);2927void reiserfs_check_lock_depth(struct super_block *s, char *caller);2928int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh,2929				 int wait);2930void reiserfs_restore_prepared_buffer(struct super_block *,2931				      struct buffer_head *bh);2932int journal_init(struct super_block *, const char *j_dev_name, int old_format,2933		 unsigned int);2934int journal_release(struct reiserfs_transaction_handle *, struct super_block *);2935int journal_release_error(struct reiserfs_transaction_handle *,2936			  struct super_block *);2937int journal_end(struct reiserfs_transaction_handle *);2938int journal_end_sync(struct reiserfs_transaction_handle *);2939int journal_mark_freed(struct reiserfs_transaction_handle *,2940		       struct super_block *, b_blocknr_t blocknr);2941int journal_transaction_should_end(struct reiserfs_transaction_handle *, int);2942int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr,2943			 int bit_nr, int searchall, b_blocknr_t *next);2944int journal_begin(struct reiserfs_transaction_handle *,2945		  struct super_block *sb, unsigned long);2946int journal_join_abort(struct reiserfs_transaction_handle *,2947		       struct super_block *sb);2948void reiserfs_abort_journal(struct super_block *sb, int errno);2949void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...);2950int reiserfs_allocate_list_bitmaps(struct super_block *s,2951				   struct reiserfs_list_bitmap *, unsigned int);2952 2953void reiserfs_schedule_old_flush(struct super_block *s);2954void reiserfs_cancel_old_flush(struct super_block *s);2955void add_save_link(struct reiserfs_transaction_handle *th,2956		   struct inode *inode, int truncate);2957int remove_save_link(struct inode *inode, int truncate);2958 2959/* objectid.c */2960__u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th);2961void reiserfs_release_objectid(struct reiserfs_transaction_handle *th,2962			       __u32 objectid_to_release);2963int reiserfs_convert_objectid_map_v1(struct super_block *);2964 2965/* stree.c */2966int B_IS_IN_TREE(const struct buffer_head *);2967extern void copy_item_head(struct item_head *to,2968			   const struct item_head *from);2969 2970/* first key is in cpu form, second - le */2971extern int comp_short_keys(const struct reiserfs_key *le_key,2972			   const struct cpu_key *cpu_key);2973extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from);2974 2975/* both are in le form */2976extern int comp_le_keys(const struct reiserfs_key *,2977			const struct reiserfs_key *);2978extern int comp_short_le_keys(const struct reiserfs_key *,2979			      const struct reiserfs_key *);2980 2981/* * get key version from on disk key - kludge */2982static inline int le_key_version(const struct reiserfs_key *key)2983{2984	int type;2985 2986	type = offset_v2_k_type(&(key->u.k_offset_v2));2987	if (type != TYPE_DIRECT && type != TYPE_INDIRECT2988	    && type != TYPE_DIRENTRY)2989		return KEY_FORMAT_3_5;2990 2991	return KEY_FORMAT_3_6;2992 2993}2994 2995static inline void copy_key(struct reiserfs_key *to,2996			    const struct reiserfs_key *from)2997{2998	memcpy(to, from, KEY_SIZE);2999}3000 3001int comp_items(const struct item_head *stored_ih, const struct treepath *path);3002const struct reiserfs_key *get_rkey(const struct treepath *chk_path,3003				    const struct super_block *sb);3004int search_by_key(struct super_block *, const struct cpu_key *,3005		  struct treepath *, int);3006#define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL)3007int search_for_position_by_key(struct super_block *sb,3008			       const struct cpu_key *cpu_key,3009			       struct treepath *search_path);3010extern void decrement_bcount(struct buffer_head *bh);3011void decrement_counters_in_path(struct treepath *search_path);3012void pathrelse(struct treepath *search_path);3013int reiserfs_check_path(struct treepath *p);3014void pathrelse_and_restore(struct super_block *s, struct treepath *search_path);3015 3016int reiserfs_insert_item(struct reiserfs_transaction_handle *th,3017			 struct treepath *path,3018			 const struct cpu_key *key,3019			 struct item_head *ih,3020			 struct inode *inode, const char *body);3021 3022int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th,3023			     struct treepath *path,3024			     const struct cpu_key *key,3025			     struct inode *inode,3026			     const char *body, int paste_size);3027 3028int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th,3029			   struct treepath *path,3030			   struct cpu_key *key,3031			   struct inode *inode,3032			   struct page *page, loff_t new_file_size);3033 3034int reiserfs_delete_item(struct reiserfs_transaction_handle *th,3035			 struct treepath *path,3036			 const struct cpu_key *key,3037			 struct inode *inode, struct buffer_head *un_bh);3038 3039void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th,3040				struct inode *inode, struct reiserfs_key *key);3041int reiserfs_delete_object(struct reiserfs_transaction_handle *th,3042			   struct inode *inode);3043int reiserfs_do_truncate(struct reiserfs_transaction_handle *th,3044			 struct inode *inode, struct page *,3045			 int update_timestamps);3046 3047#define i_block_size(inode) ((inode)->i_sb->s_blocksize)3048#define file_size(inode) ((inode)->i_size)3049#define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1))3050 3051#define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\3052!STORE_TAIL_IN_UNFM_S1(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):have_small_tails ((inode)->i_sb)?!STORE_TAIL_IN_UNFM_S2(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):0 )3053 3054void padd_item(char *item, int total_length, int length);3055 3056/* inode.c */3057/* args for the create parameter of reiserfs_get_block */3058#define GET_BLOCK_NO_CREATE 0	 /* don't create new blocks or convert tails */3059#define GET_BLOCK_CREATE 1	 /* add anything you need to find block */3060#define GET_BLOCK_NO_HOLE 2	 /* return -ENOENT for file holes */3061#define GET_BLOCK_READ_DIRECT 4	 /* read the tail if indirect item not found */3062#define GET_BLOCK_NO_IMUX     8	 /* i_mutex is not held, don't preallocate */3063#define GET_BLOCK_NO_DANGLE   16 /* don't leave any transactions running */3064 3065void reiserfs_read_locked_inode(struct inode *inode,3066				struct reiserfs_iget_args *args);3067int reiserfs_find_actor(struct inode *inode, void *p);3068int reiserfs_init_locked_inode(struct inode *inode, void *p);3069void reiserfs_evict_inode(struct inode *inode);3070int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc);3071int reiserfs_get_block(struct inode *inode, sector_t block,3072		       struct buffer_head *bh_result, int create);3073struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid,3074				     int fh_len, int fh_type);3075struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid,3076				     int fh_len, int fh_type);3077int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp,3078		       struct inode *parent);3079 3080int reiserfs_truncate_file(struct inode *, int update_timestamps);3081void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset,3082		  int type, int key_length);3083void make_le_item_head(struct item_head *ih, const struct cpu_key *key,3084		       int version,3085		       loff_t offset, int type, int length, int entry_count);3086struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key);3087 3088struct reiserfs_security_handle;3089int reiserfs_new_inode(struct reiserfs_transaction_handle *th,3090		       struct inode *dir, umode_t mode,3091		       const char *symname, loff_t i_size,3092		       struct dentry *dentry, struct inode *inode,3093		       struct reiserfs_security_handle *security);3094 3095void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th,3096			     struct inode *inode, loff_t size);3097 3098static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th,3099				      struct inode *inode)3100{3101	reiserfs_update_sd_size(th, inode, inode->i_size);3102}3103 3104void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode);3105int reiserfs_setattr(struct mnt_idmap *idmap, struct dentry *dentry,3106		     struct iattr *attr);3107 3108int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len);3109 3110/* namei.c */3111void reiserfs_init_priv_inode(struct inode *inode);3112void set_de_name_and_namelen(struct reiserfs_dir_entry *de);3113int search_by_entry_key(struct super_block *sb, const struct cpu_key *key,3114			struct treepath *path, struct reiserfs_dir_entry *de);3115struct dentry *reiserfs_get_parent(struct dentry *);3116 3117#ifdef CONFIG_REISERFS_PROC_INFO3118int reiserfs_proc_info_init(struct super_block *sb);3119int reiserfs_proc_info_done(struct super_block *sb);3120int reiserfs_proc_info_global_init(void);3121int reiserfs_proc_info_global_done(void);3122 3123#define PROC_EXP( e )   e3124 3125#define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data3126#define PROC_INFO_MAX( sb, field, value )								\3127    __PINFO( sb ).field =												\3128        max( REISERFS_SB( sb ) -> s_proc_info_data.field, value )3129#define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) )3130#define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) )3131#define PROC_INFO_BH_STAT( sb, bh, level )							\3132    PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] );						\3133    PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) );	\3134    PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) )3135#else3136static inline int reiserfs_proc_info_init(struct super_block *sb)3137{3138	return 0;3139}3140 3141static inline int reiserfs_proc_info_done(struct super_block *sb)3142{3143	return 0;3144}3145 3146static inline int reiserfs_proc_info_global_init(void)3147{3148	return 0;3149}3150 3151static inline int reiserfs_proc_info_global_done(void)3152{3153	return 0;3154}3155 3156#define PROC_EXP( e )3157#define VOID_V ( ( void ) 0 )3158#define PROC_INFO_MAX( sb, field, value ) VOID_V3159#define PROC_INFO_INC( sb, field ) VOID_V3160#define PROC_INFO_ADD( sb, field, val ) VOID_V3161#define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V3162#endif3163 3164/* dir.c */3165extern const struct inode_operations reiserfs_dir_inode_operations;3166extern const struct inode_operations reiserfs_symlink_inode_operations;3167extern const struct inode_operations reiserfs_special_inode_operations;3168extern const struct file_operations reiserfs_dir_operations;3169int reiserfs_readdir_inode(struct inode *, struct dir_context *);3170 3171/* tail_conversion.c */3172int direct2indirect(struct reiserfs_transaction_handle *, struct inode *,3173		    struct treepath *, struct buffer_head *, loff_t);3174int indirect2direct(struct reiserfs_transaction_handle *, struct inode *,3175		    struct page *, struct treepath *, const struct cpu_key *,3176		    loff_t, char *);3177void reiserfs_unmap_buffer(struct buffer_head *);3178 3179/* file.c */3180extern const struct inode_operations reiserfs_file_inode_operations;3181extern const struct inode_operations reiserfs_priv_file_inode_operations;3182extern const struct file_operations reiserfs_file_operations;3183extern const struct address_space_operations reiserfs_address_space_operations;3184 3185/* fix_nodes.c */3186 3187int fix_nodes(int n_op_mode, struct tree_balance *tb,3188	      struct item_head *ins_ih, const void *);3189void unfix_nodes(struct tree_balance *);3190 3191/* prints.c */3192void __reiserfs_panic(struct super_block *s, const char *id,3193		      const char *function, const char *fmt, ...)3194    __attribute__ ((noreturn));3195#define reiserfs_panic(s, id, fmt, args...) \3196	__reiserfs_panic(s, id, __func__, fmt, ##args)3197void __reiserfs_error(struct super_block *s, const char *id,3198		      const char *function, const char *fmt, ...);3199#define reiserfs_error(s, id, fmt, args...) \3200	 __reiserfs_error(s, id, __func__, fmt, ##args)3201void reiserfs_info(struct super_block *s, const char *fmt, ...);3202void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...);3203void print_indirect_item(struct buffer_head *bh, int item_num);3204void store_print_tb(struct tree_balance *tb);3205void print_cur_tb(char *mes);3206void print_de(struct reiserfs_dir_entry *de);3207void print_bi(struct buffer_info *bi, char *mes);3208#define PRINT_LEAF_ITEMS 1	/* print all items */3209#define PRINT_DIRECTORY_ITEMS 2	/* print directory items */3210#define PRINT_DIRECT_ITEMS 4	/* print contents of direct items */3211void print_block(struct buffer_head *bh, ...);3212void print_bmap(struct super_block *s, int silent);3213void print_bmap_block(int i, char *data, int size, int silent);3214/*void print_super_block (struct super_block * s, char * mes);*/3215void print_objectid_map(struct super_block *s);3216void print_block_head(struct buffer_head *bh, char *mes);3217void check_leaf(struct buffer_head *bh);3218void check_internal(struct buffer_head *bh);3219void print_statistics(struct super_block *s);3220char *reiserfs_hashname(int code);3221 3222/* lbalance.c */3223int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num,3224		    int mov_bytes, struct buffer_head *Snew);3225int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes);3226int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes);3227void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first,3228		       int del_num, int del_bytes);3229void leaf_insert_into_buf(struct buffer_info *bi, int before,3230			  struct item_head * const inserted_item_ih,3231			  const char * const inserted_item_body,3232			  int zeros_number);3233void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num,3234			  int pos_in_item, int paste_size,3235			  const char * const body, int zeros_number);3236void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num,3237			  int pos_in_item, int cut_size);3238void leaf_paste_entries(struct buffer_info *bi, int item_num, int before,3239			int new_entry_count, struct reiserfs_de_head *new_dehs,3240			const char *records, int paste_size);3241/* ibalance.c */3242int balance_internal(struct tree_balance *, int, int, struct item_head *,3243		     struct buffer_head **);3244 3245/* do_balance.c */3246void do_balance_mark_leaf_dirty(struct tree_balance *tb,3247				struct buffer_head *bh, int flag);3248#define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty3249#define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty3250 3251void do_balance(struct tree_balance *tb, struct item_head *ih,3252		const char *body, int flag);3253void reiserfs_invalidate_buffer(struct tree_balance *tb,3254				struct buffer_head *bh);3255 3256int get_left_neighbor_position(struct tree_balance *tb, int h);3257int get_right_neighbor_position(struct tree_balance *tb, int h);3258void replace_key(struct tree_balance *tb, struct buffer_head *, int,3259		 struct buffer_head *, int);3260void make_empty_node(struct buffer_info *);3261struct buffer_head *get_FEB(struct tree_balance *);3262 3263/* bitmap.c */3264 3265/*3266 * structure contains hints for block allocator, and it is a container for3267 * arguments, such as node, search path, transaction_handle, etc.3268 */3269struct __reiserfs_blocknr_hint {3270	/* inode passed to allocator, if we allocate unf. nodes */3271	struct inode *inode;3272 3273	sector_t block;		/* file offset, in blocks */3274	struct in_core_key key;3275 3276	/*3277	 * search path, used by allocator to deternine search_start by3278	 * various ways3279	 */3280	struct treepath *path;3281 3282	/*3283	 * transaction handle is needed to log super blocks3284	 * and bitmap blocks changes3285	 */3286	struct reiserfs_transaction_handle *th;3287 3288	b_blocknr_t beg, end;3289 3290	/*3291	 * a field used to transfer search start value (block number)3292	 * between different block allocator procedures3293	 * (determine_search_start() and others)3294	 */3295	b_blocknr_t search_start;3296 3297	/*3298	 * is set in determine_prealloc_size() function,3299	 * used by underlayed function that do actual allocation3300	 */3301	int prealloc_size;3302 3303	/*3304	 * the allocator uses different polices for getting disk3305	 * space for formatted/unformatted blocks with/without preallocation3306	 */3307	unsigned formatted_node:1;3308	unsigned preallocate:1;3309};3310 3311typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t;3312 3313int reiserfs_parse_alloc_options(struct super_block *, char *);3314void reiserfs_init_alloc_options(struct super_block *s);3315 3316/*3317 * given a directory, this will tell you what packing locality3318 * to use for a new object underneat it.  The locality is returned3319 * in disk byte order (le).3320 */3321__le32 reiserfs_choose_packing(struct inode *dir);3322 3323void show_alloc_options(struct seq_file *seq, struct super_block *s);3324int reiserfs_init_bitmap_cache(struct super_block *sb);3325void reiserfs_free_bitmap_cache(struct super_block *sb);3326void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info);3327struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap);3328int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value);3329void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *,3330			 b_blocknr_t, int for_unformatted);3331int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int,3332			       int);3333static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb,3334					     b_blocknr_t * new_blocknrs,3335					     int amount_needed)3336{3337	reiserfs_blocknr_hint_t hint = {3338		.th = tb->transaction_handle,3339		.path = tb->tb_path,3340		.inode = NULL,3341		.key = tb->key,3342		.block = 0,3343		.formatted_node = 13344	};3345	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed,3346					  0);3347}3348 3349static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle3350					    *th, struct inode *inode,3351					    b_blocknr_t * new_blocknrs,3352					    struct treepath *path,3353					    sector_t block)3354{3355	reiserfs_blocknr_hint_t hint = {3356		.th = th,3357		.path = path,3358		.inode = inode,3359		.block = block,3360		.formatted_node = 0,3361		.preallocate = 03362	};3363	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);3364}3365 3366#ifdef REISERFS_PREALLOCATE3367static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle3368					     *th, struct inode *inode,3369					     b_blocknr_t * new_blocknrs,3370					     struct treepath *path,3371					     sector_t block)3372{3373	reiserfs_blocknr_hint_t hint = {3374		.th = th,3375		.path = path,3376		.inode = inode,3377		.block = block,3378		.formatted_node = 0,3379		.preallocate = 13380	};3381	return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0);3382}3383 3384void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th,3385			       struct inode *inode);3386void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th);3387#endif3388 3389/* hashes.c */3390__u32 keyed_hash(const signed char *msg, int len);3391__u32 yura_hash(const signed char *msg, int len);3392__u32 r5_hash(const signed char *msg, int len);3393 3394#define reiserfs_set_le_bit		__set_bit_le3395#define reiserfs_test_and_set_le_bit	__test_and_set_bit_le3396#define reiserfs_clear_le_bit		__clear_bit_le3397#define reiserfs_test_and_clear_le_bit	__test_and_clear_bit_le3398#define reiserfs_test_le_bit		test_bit_le3399#define reiserfs_find_next_zero_le_bit	find_next_zero_bit_le3400 3401/*3402 * sometimes reiserfs_truncate may require to allocate few new blocks3403 * to perform indirect2direct conversion. People probably used to3404 * think, that truncate should work without problems on a filesystem3405 * without free disk space. They may complain that they can not3406 * truncate due to lack of free disk space. This spare space allows us3407 * to not worry about it. 500 is probably too much, but it should be3408 * absolutely safe3409 */3410#define SPARE_SPACE 5003411 3412/* prototypes from ioctl.c */3413int reiserfs_fileattr_get(struct dentry *dentry, struct fileattr *fa);3414int reiserfs_fileattr_set(struct mnt_idmap *idmap,3415			  struct dentry *dentry, struct fileattr *fa);3416long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);3417long reiserfs_compat_ioctl(struct file *filp,3418		   unsigned int cmd, unsigned long arg);3419int reiserfs_unpack(struct inode *inode);3420