journal.c

Linux Kernel 2.6.9 for OMAP1710

/*
** Write ahead logging implementation copyright Chris Mason 2000
**
** The background commits make this code very interelated, and 
** overly complex.  I need to rethink things a bit....The major players:
**
** journal_begin -- call with the number of blocks you expect to log.  
**                  If the current transaction is too
** 		    old, it will block until the current transaction is 
** 		    finished, and then start a new one.
**		    Usually, your transaction will get joined in with 
**                  previous ones for speed.
**
** journal_join  -- same as journal_begin, but won't block on the current 
**                  transaction regardless of age.  Don't ever call
**                  this.  Ever.  There are only two places it should be 
**                  called from, and they are both inside this file.
**
** journal_mark_dirty -- adds blocks into this transaction.  clears any flags 
**                       that might make them get sent to disk
**                       and then marks them BH_JDirty.  Puts the buffer head 
**                       into the current transaction hash.  
**
** journal_end -- if the current transaction is batchable, it does nothing
**                   otherwise, it could do an async/synchronous commit, or
**                   a full flush of all log and real blocks in the 
**                   transaction.
**
** flush_old_commits -- if the current transaction is too old, it is ended and 
**                      commit blocks are sent to disk.  Forces commit blocks 
**                      to disk for all backgrounded commits that have been 
**                      around too long.
**		     -- Note, if you call this as an immediate flush from 
**		        from within kupdate, it will ignore the immediate flag
*/

#include <linux/config.h>
#include <asm/uaccess.h>
#include <asm/system.h>

#include <linux/time.h>
#include <asm/semaphore.h>

#include <linux/vmalloc.h>
#include <linux/reiserfs_fs.h>

#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/suspend.h>
#include <linux/buffer_head.h>
#include <linux/workqueue.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>


/* gets a struct reiserfs_journal_list * from a list head */
#define JOURNAL_LIST_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
                               j_list))
#define JOURNAL_WORK_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
                               j_working_list))

/* the number of mounted filesystems.  This is used to decide when to
** start and kill the commit workqueue
*/
static int reiserfs_mounted_fs_count;

static struct workqueue_struct *commit_wq;

#define JOURNAL_TRANS_HALF 1018   /* must be correct to keep the desc and commit
				     structs at 4k */
#define BUFNR 64 /*read ahead */

/* cnode stat bits.  Move these into reiserfs_fs.h */

#define BLOCK_FREED 2		/* this block was freed, and can't be written.  */
#define BLOCK_FREED_HOLDER 3    /* this block was freed during this transaction, and can't be written */

#define BLOCK_NEEDS_FLUSH 4	/* used in flush_journal_list */
#define BLOCK_DIRTIED 5


/* journal list state bits */
#define LIST_TOUCHED 1
#define LIST_DIRTY   2
#define LIST_COMMIT_PENDING  4		/* someone will commit this list */

/* flags for do_journal_end */
#define FLUSH_ALL   1		/* flush commit and real blocks */
#define COMMIT_NOW  2		/* end and commit this transaction */
#define WAIT        4		/* wait for the log blocks to hit the disk*/

/* state bits for the journal */
#define WRITERS_BLOCKED 1      /* set when new writers not allowed */
#define WRITERS_QUEUED 2       /* set when log is full due to too many
				* writers
				*/

static int do_journal_end(struct reiserfs_transaction_handle *,struct super_block *,unsigned long nblocks,int flags) ;
static int flush_journal_list(struct super_block *s, struct reiserfs_journal_list *jl, int flushall) ;
static int flush_commit_list(struct super_block *s, struct reiserfs_journal_list *jl, int flushall)  ;
static int can_dirty(struct reiserfs_journal_cnode *cn) ;
static int journal_join(struct reiserfs_transaction_handle *th, struct super_block *p_s_sb, unsigned long nblocks);
static int release_journal_dev( struct super_block *super,
				struct reiserfs_journal *journal );
static int dirty_one_transaction(struct super_block *s,
                                 struct reiserfs_journal_list *jl);
static void flush_async_commits(void *p);

static void init_journal_hash(struct super_block *p_s_sb) {
  memset(SB_JOURNAL(p_s_sb)->j_hash_table, 0, JOURNAL_HASH_SIZE * sizeof(struct reiserfs_journal_cnode *)) ;
}

/*
** clears BH_Dirty and sticks the buffer on the clean list.  Called because I can't allow refile_buffer to
** make schedule happen after I've freed a block.  Look at remove_from_transaction and journal_mark_freed for
** more details.
*/
static int reiserfs_clean_and_file_buffer(struct buffer_head *bh) {
  if (bh) {
    clear_buffer_dirty(bh);
    clear_bit(BH_JTest, &bh->b_state);
  }
  return 0 ;
}

static void disable_barrier(struct super_block *s)
{
    REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_BARRIER_FLUSH);
    printk("reiserfs: disabling flush barriers on %s\n", reiserfs_bdevname(s));
}

static struct reiserfs_bitmap_node *
allocate_bitmap_node(struct super_block *p_s_sb) {
  struct reiserfs_bitmap_node *bn ;
  static int id;

  bn = reiserfs_kmalloc(sizeof(struct reiserfs_bitmap_node), GFP_NOFS, p_s_sb) ;
  if (!bn) {
    return NULL ;
  }
  bn->data = reiserfs_kmalloc(p_s_sb->s_blocksize, GFP_NOFS, p_s_sb) ;
  if (!bn->data) {
    reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb) ;
    return NULL ;
  }
  bn->id = id++ ;
  memset(bn->data, 0, p_s_sb->s_blocksize) ;
  INIT_LIST_HEAD(&bn->list) ;
  return bn ;
}

static struct reiserfs_bitmap_node *
get_bitmap_node(struct super_block *p_s_sb) {
  struct reiserfs_bitmap_node *bn = NULL;
  struct list_head *entry = SB_JOURNAL(p_s_sb)->j_bitmap_nodes.next ;

  SB_JOURNAL(p_s_sb)->j_used_bitmap_nodes++ ;
repeat:

  if(entry != &SB_JOURNAL(p_s_sb)->j_bitmap_nodes) {
    bn = list_entry(entry, struct reiserfs_bitmap_node, list) ;
    list_del(entry) ;
    memset(bn->data, 0, p_s_sb->s_blocksize) ;
    SB_JOURNAL(p_s_sb)->j_free_bitmap_nodes-- ;
    return bn ;
  }
  bn = allocate_bitmap_node(p_s_sb) ;
  if (!bn) {
    yield();
    goto repeat ;
  }
  return bn ;
}
static inline void free_bitmap_node(struct super_block *p_s_sb,
                                    struct reiserfs_bitmap_node *bn) {
  SB_JOURNAL(p_s_sb)->j_used_bitmap_nodes-- ;
  if (SB_JOURNAL(p_s_sb)->j_free_bitmap_nodes > REISERFS_MAX_BITMAP_NODES) {
    reiserfs_kfree(bn->data, p_s_sb->s_blocksize, p_s_sb) ;
    reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb) ;
  } else {
    list_add(&bn->list, &SB_JOURNAL(p_s_sb)->j_bitmap_nodes) ;
    SB_JOURNAL(p_s_sb)->j_free_bitmap_nodes++ ;
  }
}

static void allocate_bitmap_nodes(struct super_block *p_s_sb) {
  int i ;
  struct reiserfs_bitmap_node *bn = NULL ;
  for (i = 0 ; i < REISERFS_MIN_BITMAP_NODES ; i++) {
    bn = allocate_bitmap_node(p_s_sb) ;
    if (bn) {
      list_add(&bn->list, &SB_JOURNAL(p_s_sb)->j_bitmap_nodes) ;
      SB_JOURNAL(p_s_sb)->j_free_bitmap_nodes++ ;
    } else {
      break ; // this is ok, we'll try again when more are needed 
    }
  }
}

static int set_bit_in_list_bitmap(struct super_block *p_s_sb, int block,
                                  struct reiserfs_list_bitmap *jb) {
  int bmap_nr = block / (p_s_sb->s_blocksize << 3) ;
  int bit_nr = block % (p_s_sb->s_blocksize << 3) ;

  if (!jb->bitmaps[bmap_nr]) {
    jb->bitmaps[bmap_nr] = get_bitmap_node(p_s_sb) ;
  }
  set_bit(bit_nr, (unsigned long *)jb->bitmaps[bmap_nr]->data) ;
  return 0 ;
}

static void cleanup_bitmap_list(struct super_block *p_s_sb,
                                struct reiserfs_list_bitmap *jb) {
  int i;
  if (jb->bitmaps == NULL)
    return;

  for (i = 0 ; i < SB_BMAP_NR(p_s_sb) ; i++) {
    if (jb->bitmaps[i]) {
      free_bitmap_node(p_s_sb, jb->bitmaps[i]) ;
      jb->bitmaps[i] = NULL ;
    }
  }
}

/*
** only call this on FS unmount.
*/
static int free_list_bitmaps(struct super_block *p_s_sb,
                             struct reiserfs_list_bitmap *jb_array) {
  int i ;
  struct reiserfs_list_bitmap *jb ;
  for (i = 0 ; i < JOURNAL_NUM_BITMAPS ; i++) {
    jb = jb_array + i ;
    jb->journal_list = NULL ;
    cleanup_bitmap_list(p_s_sb, jb) ;
    vfree(jb->bitmaps) ;
    jb->bitmaps = NULL ;
  }
  return 0;
}

static int free_bitmap_nodes(struct super_block *p_s_sb) {
  struct list_head *next = SB_JOURNAL(p_s_sb)->j_bitmap_nodes.next ;
  struct reiserfs_bitmap_node *bn ;

  while(next != &SB_JOURNAL(p_s_sb)->j_bitmap_nodes) {
    bn = list_entry(next, struct reiserfs_bitmap_node, list) ;
    list_del(next) ;
    reiserfs_kfree(bn->data, p_s_sb->s_blocksize, p_s_sb) ;
    reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb) ;
    next = SB_JOURNAL(p_s_sb)->j_bitmap_nodes.next ;
    SB_JOURNAL(p_s_sb)->j_free_bitmap_nodes-- ;
  }

  return 0 ;
}

/*
** get memory for JOURNAL_NUM_BITMAPS worth of bitmaps. 
** jb_array is the array to be filled in.
*/
int reiserfs_allocate_list_bitmaps(struct super_block *p_s_sb,
                                   struct reiserfs_list_bitmap *jb_array,
				   int bmap_nr) {
  int i ;
  int failed = 0 ;
  struct reiserfs_list_bitmap *jb ;
  int mem = bmap_nr * sizeof(struct reiserfs_bitmap_node *) ;

  for (i = 0 ; i < JOURNAL_NUM_BITMAPS ; i++) {
    jb = jb_array + i ;
    jb->journal_list = NULL ;
    jb->bitmaps = vmalloc( mem ) ;
    if (!jb->bitmaps) {
      reiserfs_warning(p_s_sb, "clm-2000, unable to allocate bitmaps for journal lists") ;
      failed = 1;   
      break ;
    }
    memset(jb->bitmaps, 0, mem) ;
  }
  if (failed) {
    free_list_bitmaps(p_s_sb, jb_array) ;
    return -1 ;
  }
  return 0 ;
}

/*
** find an available list bitmap.  If you can't find one, flush a commit list 
** and try again
*/
static struct reiserfs_list_bitmap *
get_list_bitmap(struct super_block *p_s_sb, struct reiserfs_journal_list *jl) {
  int i,j ; 
  struct reiserfs_list_bitmap *jb = NULL ;

  for (j = 0 ; j < (JOURNAL_NUM_BITMAPS * 3) ; j++) {
    i = SB_JOURNAL(p_s_sb)->j_list_bitmap_index ;
    SB_JOURNAL(p_s_sb)->j_list_bitmap_index = (i + 1) % JOURNAL_NUM_BITMAPS ;
    jb = SB_JOURNAL(p_s_sb)->j_list_bitmap + i ;
    if (SB_JOURNAL(p_s_sb)->j_list_bitmap[i].journal_list) {
      flush_commit_list(p_s_sb, SB_JOURNAL(p_s_sb)->j_list_bitmap[i].journal_list, 1) ;
      if (!SB_JOURNAL(p_s_sb)->j_list_bitmap[i].journal_list) {
	break ;
      }
    } else {
      break ;
    }
  }
  if (jb->journal_list) { /* double check to make sure if flushed correctly */
    return NULL ;
  }
  jb->journal_list = jl ;
  return jb ;
}

/* 
** allocates a new chunk of X nodes, and links them all together as a list.
** Uses the cnode->next and cnode->prev pointers
** returns NULL on failure
*/
static struct reiserfs_journal_cnode *allocate_cnodes(int num_cnodes) {
  struct reiserfs_journal_cnode *head ;
  int i ;
  if (num_cnodes <= 0) {
    return NULL ;
  }
  head = vmalloc(num_cnodes * sizeof(struct reiserfs_journal_cnode)) ;
  if (!head) {
    return NULL ;
  }
  memset(head, 0, num_cnodes * sizeof(struct reiserfs_journal_cnode)) ;
  head[0].prev = NULL ;
  head[0].next = head + 1 ;
  for (i = 1 ; i < num_cnodes; i++) {
    head[i].prev = head + (i - 1) ;
    head[i].next = head + (i + 1) ; /* if last one, overwrite it after the if */
  }
  head[num_cnodes -1].next = NULL ;
  return head ;
}

/*
** pulls a cnode off the free list, or returns NULL on failure 
*/
static struct reiserfs_journal_cnode *get_cnode(struct super_block *p_s_sb) {
  struct reiserfs_journal_cnode *cn ;

  reiserfs_check_lock_depth(p_s_sb, "get_cnode") ;

  if (SB_JOURNAL(p_s_sb)->j_cnode_free <= 0) {
    return NULL ;
  }
  SB_JOURNAL(p_s_sb)->j_cnode_used++ ;
  SB_JOURNAL(p_s_sb)->j_cnode_free-- ;
  cn = SB_JOURNAL(p_s_sb)->j_cnode_free_list ;
  if (!cn) {
    return cn ;
  }
  if (cn->next) {
    cn->next->prev = NULL ;
  }
  SB_JOURNAL(p_s_sb)->j_cnode_free_list = cn->next ;
  memset(cn, 0, sizeof(struct reiserfs_journal_cnode)) ;
  return cn ;
}

/*
** returns a cnode to the free list 
*/
static void free_cnode(struct super_block *p_s_sb, struct reiserfs_journal_cnode *cn) {

  reiserfs_check_lock_depth(p_s_sb, "free_cnode") ;

  SB_JOURNAL(p_s_sb)->j_cnode_used-- ;
  SB_JOURNAL(p_s_sb)->j_cnode_free++ ;
  /* memset(cn, 0, sizeof(struct reiserfs_journal_cnode)) ; */
  cn->next = SB_JOURNAL(p_s_sb)->j_cnode_free_list ;
  if (SB_JOURNAL(p_s_sb)->j_cnode_free_list) {
    SB_JOURNAL(p_s_sb)->j_cnode_free_list->prev = cn ;
  }
  cn->prev = NULL ; /* not needed with the memset, but I might kill the memset, and forget to do this */
  SB_JOURNAL(p_s_sb)->j_cnode_free_list = cn ;
}

static int clear_prepared_bits(struct buffer_head *bh) {
  clear_bit(BH_JPrepared, &bh->b_state) ;
  clear_bit(BH_JRestore_dirty, &bh->b_state) ;
  return 0 ;
}

/* buffer is in current transaction */
inline int buffer_journaled(const struct buffer_head *bh) {
  if (bh)
    return test_bit(BH_JDirty, &bh->b_state) ;
  else
    return 0 ;
}

/* disk block was taken off free list before being in a finished transation, or written to disk
** journal_new blocks can be reused immediately, for any purpose
*/ 
inline int buffer_journal_new(const struct buffer_head *bh) {