/*
- * linux/fs/revoke.c
- *
+ * linux/fs/jbd/revoke.c
+ *
* Written by Stephen C. Tweedie <sct@redhat.com>, 2000
*
* Copyright 2000 Red Hat corp --- All Rights Reserved
* Revoke is the mechanism used to prevent old log records for deleted
* metadata from being replayed on top of newer data using the same
* blocks. The revoke mechanism is used in two separate places:
- *
+ *
* + Commit: during commit we write the entire list of the current
* transaction's revoked blocks to the journal
- *
+ *
* + Recovery: during recovery we record the transaction ID of all
* revoked blocks. If there are multiple revoke records in the log
* for a single block, only the last one counts, and if there is a log
* single transaction:
*
* Block is revoked and then journaled:
- * The desired end result is the journaling of the new block, so we
+ * The desired end result is the journaling of the new block, so we
* cancel the revoke before the transaction commits.
*
* Block is journaled and then revoked:
* transaction must have happened after the block was journaled and so
* the revoke must take precedence.
*
- * Block is revoked and then written as data:
+ * Block is revoked and then written as data:
* The data write is allowed to succeed, but the revoke is _not_
* cancelled. We still need to prevent old log records from
* overwriting the new data. We don't even need to clear the revoke
* buffer has not been revoked, and cancel_revoke
* need do nothing.
* RevokeValid set, Revoked set:
- * buffer has been revoked.
+ * buffer has been revoked.
+ *
+ * Locking rules:
+ * We keep two hash tables of revoke records. One hashtable belongs to the
+ * running transaction (is pointed to by journal->j_revoke), the other one
+ * belongs to the committing transaction. Accesses to the second hash table
+ * happen only from the kjournald and no other thread touches this table. Also
+ * journal_switch_revoke_table() which switches which hashtable belongs to the
+ * running and which to the committing transaction is called only from
+ * kjournald. Therefore we need no locks when accessing the hashtable belonging
+ * to the committing transaction.
+ *
+ * All users operating on the hash table belonging to the running transaction
+ * have a handle to the transaction. Therefore they are safe from kjournald
+ * switching hash tables under them. For operations on the lists of entries in
+ * the hash table j_revoke_lock is used.
+ *
+ * Finally, also replay code uses the hash tables but at this moment noone else
+ * can touch them (filesystem isn't mounted yet) and hence no locking is
+ * needed.
*/
#ifndef __KERNEL__
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/list.h>
-#include <linux/smp_lock.h>
#include <linux/init.h>
+#include <linux/bio.h>
#endif
+#include <linux/log2.h>
-static kmem_cache_t *revoke_record_cache;
-static kmem_cache_t *revoke_table_cache;
+static struct kmem_cache *revoke_record_cache;
+static struct kmem_cache *revoke_table_cache;
/* Each revoke record represents one single revoked block. During
journal replay, this involves recording the transaction ID of the
last transaction to revoke this block. */
-struct jbd_revoke_record_s
+struct jbd_revoke_record_s
{
struct list_head hash;
tid_t sequence; /* Used for recovery only */
- unsigned long blocknr;
+ unsigned int blocknr;
};
{
/* It is conceivable that we might want a larger hash table
* for recovery. Must be a power of two. */
- int hash_size;
- int hash_shift;
+ int hash_size;
+ int hash_shift;
struct list_head *hash_table;
};
#ifdef __KERNEL__
static void write_one_revoke_record(journal_t *, transaction_t *,
struct journal_head **, int *,
- struct jbd_revoke_record_s *);
-static void flush_descriptor(journal_t *, struct journal_head *, int);
+ struct jbd_revoke_record_s *, int);
+static void flush_descriptor(journal_t *, struct journal_head *, int, int);
#endif
/* Utility functions to maintain the revoke table */
/* Borrowed from buffer.c: this is a tried and tested block hash function */
-static inline int hash(journal_t *journal, unsigned long block)
+static inline int hash(journal_t *journal, unsigned int block)
{
struct jbd_revoke_table_s *table = journal->j_revoke;
int hash_shift = table->hash_shift;
(block << (hash_shift - 12))) & (table->hash_size - 1);
}
-int insert_revoke_hash(journal_t *journal, unsigned long blocknr, tid_t seq)
+static int insert_revoke_hash(journal_t *journal, unsigned int blocknr,
+ tid_t seq)
{
struct list_head *hash_list;
struct jbd_revoke_record_s *record;
oom:
if (!journal_oom_retry)
return -ENOMEM;
- jbd_debug(1, "ENOMEM in %s, retrying\n", __FUNCTION__);
+ jbd_debug(1, "ENOMEM in %s, retrying\n", __func__);
yield();
goto repeat;
}
/* Find a revoke record in the journal's hash table. */
static struct jbd_revoke_record_s *find_revoke_record(journal_t *journal,
- unsigned long blocknr)
+ unsigned int blocknr)
{
struct list_head *hash_list;
struct jbd_revoke_record_s *record;
return NULL;
}
+void journal_destroy_revoke_caches(void)
+{
+ if (revoke_record_cache) {
+ kmem_cache_destroy(revoke_record_cache);
+ revoke_record_cache = NULL;
+ }
+ if (revoke_table_cache) {
+ kmem_cache_destroy(revoke_table_cache);
+ revoke_table_cache = NULL;
+ }
+}
+
int __init journal_init_revoke_caches(void)
{
+ J_ASSERT(!revoke_record_cache);
+ J_ASSERT(!revoke_table_cache);
+
revoke_record_cache = kmem_cache_create("revoke_record",
sizeof(struct jbd_revoke_record_s),
- 0, SLAB_HWCACHE_ALIGN, NULL, NULL);
- if (revoke_record_cache == 0)
- return -ENOMEM;
+ 0,
+ SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY,
+ NULL);
+ if (!revoke_record_cache)
+ goto record_cache_failure;
revoke_table_cache = kmem_cache_create("revoke_table",
sizeof(struct jbd_revoke_table_s),
- 0, 0, NULL, NULL);
- if (revoke_table_cache == 0) {
- kmem_cache_destroy(revoke_record_cache);
- revoke_record_cache = NULL;
- return -ENOMEM;
- }
+ 0, SLAB_TEMPORARY, NULL);
+ if (!revoke_table_cache)
+ goto table_cache_failure;
+
return 0;
-}
-void journal_destroy_revoke_caches(void)
-{
- kmem_cache_destroy(revoke_record_cache);
- revoke_record_cache = NULL;
- kmem_cache_destroy(revoke_table_cache);
- revoke_table_cache = NULL;
+table_cache_failure:
+ journal_destroy_revoke_caches();
+record_cache_failure:
+ return -ENOMEM;
}
-/* Initialise the revoke table for a given journal to a given size. */
-
-int journal_init_revoke(journal_t *journal, int hash_size)
+static struct jbd_revoke_table_s *journal_init_revoke_table(int hash_size)
{
- int shift, tmp;
+ int shift = 0;
+ int tmp = hash_size;
+ struct jbd_revoke_table_s *table;
- J_ASSERT (journal->j_revoke_table[0] == NULL);
+ table = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
+ if (!table)
+ goto out;
- shift = 0;
- tmp = hash_size;
while((tmp >>= 1UL) != 0UL)
shift++;
- journal->j_revoke_table[0] = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
- if (!journal->j_revoke_table[0])
- return -ENOMEM;
- journal->j_revoke = journal->j_revoke_table[0];
-
- /* Check that the hash_size is a power of two */
- J_ASSERT ((hash_size & (hash_size-1)) == 0);
-
- journal->j_revoke->hash_size = hash_size;
-
- journal->j_revoke->hash_shift = shift;
-
- journal->j_revoke->hash_table =
+ table->hash_size = hash_size;
+ table->hash_shift = shift;
+ table->hash_table =
kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
- if (!journal->j_revoke->hash_table) {
- kmem_cache_free(revoke_table_cache, journal->j_revoke_table[0]);
- journal->j_revoke = NULL;
- return -ENOMEM;
+ if (!table->hash_table) {
+ kmem_cache_free(revoke_table_cache, table);
+ table = NULL;
+ goto out;
}
for (tmp = 0; tmp < hash_size; tmp++)
- INIT_LIST_HEAD(&journal->j_revoke->hash_table[tmp]);
+ INIT_LIST_HEAD(&table->hash_table[tmp]);
- journal->j_revoke_table[1] = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
- if (!journal->j_revoke_table[1]) {
- kfree(journal->j_revoke_table[0]->hash_table);
- kmem_cache_free(revoke_table_cache, journal->j_revoke_table[0]);
- return -ENOMEM;
- }
+out:
+ return table;
+}
- journal->j_revoke = journal->j_revoke_table[1];
+static void journal_destroy_revoke_table(struct jbd_revoke_table_s *table)
+{
+ int i;
+ struct list_head *hash_list;
+
+ for (i = 0; i < table->hash_size; i++) {
+ hash_list = &table->hash_table[i];
+ J_ASSERT(list_empty(hash_list));
+ }
- /* Check that the hash_size is a power of two */
- J_ASSERT ((hash_size & (hash_size-1)) == 0);
+ kfree(table->hash_table);
+ kmem_cache_free(revoke_table_cache, table);
+}
- journal->j_revoke->hash_size = hash_size;
+/* Initialise the revoke table for a given journal to a given size. */
+int journal_init_revoke(journal_t *journal, int hash_size)
+{
+ J_ASSERT(journal->j_revoke_table[0] == NULL);
+ J_ASSERT(is_power_of_2(hash_size));
- journal->j_revoke->hash_shift = shift;
+ journal->j_revoke_table[0] = journal_init_revoke_table(hash_size);
+ if (!journal->j_revoke_table[0])
+ goto fail0;
- journal->j_revoke->hash_table =
- kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
- if (!journal->j_revoke->hash_table) {
- kfree(journal->j_revoke_table[0]->hash_table);
- kmem_cache_free(revoke_table_cache, journal->j_revoke_table[0]);
- kmem_cache_free(revoke_table_cache, journal->j_revoke_table[1]);
- journal->j_revoke = NULL;
- return -ENOMEM;
- }
+ journal->j_revoke_table[1] = journal_init_revoke_table(hash_size);
+ if (!journal->j_revoke_table[1])
+ goto fail1;
- for (tmp = 0; tmp < hash_size; tmp++)
- INIT_LIST_HEAD(&journal->j_revoke->hash_table[tmp]);
+ journal->j_revoke = journal->j_revoke_table[1];
spin_lock_init(&journal->j_revoke_lock);
return 0;
-}
-/* Destoy a journal's revoke table. The table must already be empty! */
+fail1:
+ journal_destroy_revoke_table(journal->j_revoke_table[0]);
+fail0:
+ return -ENOMEM;
+}
+/* Destroy a journal's revoke table. The table must already be empty! */
void journal_destroy_revoke(journal_t *journal)
{
- struct jbd_revoke_table_s *table;
- struct list_head *hash_list;
- int i;
-
- table = journal->j_revoke_table[0];
- if (!table)
- return;
-
- for (i=0; i<table->hash_size; i++) {
- hash_list = &table->hash_table[i];
- J_ASSERT (list_empty(hash_list));
- }
-
- kfree(table->hash_table);
- kmem_cache_free(revoke_table_cache, table);
- journal->j_revoke = NULL;
-
- table = journal->j_revoke_table[1];
- if (!table)
- return;
-
- for (i=0; i<table->hash_size; i++) {
- hash_list = &table->hash_table[i];
- J_ASSERT (list_empty(hash_list));
- }
-
- kfree(table->hash_table);
- kmem_cache_free(revoke_table_cache, table);
journal->j_revoke = NULL;
+ if (journal->j_revoke_table[0])
+ journal_destroy_revoke_table(journal->j_revoke_table[0]);
+ if (journal->j_revoke_table[1])
+ journal_destroy_revoke_table(journal->j_revoke_table[1]);
}
#ifdef __KERNEL__
-/*
+/*
* journal_revoke: revoke a given buffer_head from the journal. This
* prevents the block from being replayed during recovery if we take a
* crash after this current transaction commits. Any subsequent
* metadata writes of the buffer in this transaction cancel the
- * revoke.
+ * revoke.
*
* Note that this call may block --- it is up to the caller to make
* sure that there are no further calls to journal_write_metadata
* before the revoke is complete. In ext3, this implies calling the
* revoke before clearing the block bitmap when we are deleting
- * metadata.
+ * metadata.
*
* Revoke performs a journal_forget on any buffer_head passed in as a
* parameter, but does _not_ forget the buffer_head if the bh was only
- * found implicitly.
+ * found implicitly.
*
* bh_in may not be a journalled buffer - it may have come off
* the hash tables without an attached journal_head.
* by one.
*/
-int journal_revoke(handle_t *handle, unsigned long blocknr,
+int journal_revoke(handle_t *handle, unsigned int blocknr,
struct buffer_head *bh_in)
{
struct buffer_head *bh = NULL;
}
}
- jbd_debug(2, "insert revoke for block %lu, bh_in=%p\n", blocknr, bh_in);
+ jbd_debug(2, "insert revoke for block %u, bh_in=%p\n", blocknr, bh_in);
err = insert_revoke_hash(journal, blocknr,
handle->h_transaction->t_tid);
BUFFER_TRACE(bh_in, "exit");
* the second time we would still have a pending revoke to cancel. So,
* do not trust the Revoked bit on buffers unless RevokeValid is also
* set.
- *
- * The caller must have the journal locked.
*/
int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
{
else
journal->j_revoke = journal->j_revoke_table[0];
- for (i = 0; i < journal->j_revoke->hash_size; i++)
+ for (i = 0; i < journal->j_revoke->hash_size; i++)
INIT_LIST_HEAD(&journal->j_revoke->hash_table[i]);
}
/*
* Write revoke records to the journal for all entries in the current
* revoke hash, deleting the entries as we go.
- *
- * Called with the journal lock held.
*/
-
-void journal_write_revoke_records(journal_t *journal,
- transaction_t *transaction)
+void journal_write_revoke_records(journal_t *journal,
+ transaction_t *transaction, int write_op)
{
struct journal_head *descriptor;
struct jbd_revoke_record_s *record;
struct list_head *hash_list;
int i, offset, count;
- descriptor = NULL;
+ descriptor = NULL;
offset = 0;
count = 0;
hash_list = &revoke->hash_table[i];
while (!list_empty(hash_list)) {
- record = (struct jbd_revoke_record_s *)
+ record = (struct jbd_revoke_record_s *)
hash_list->next;
write_one_revoke_record(journal, transaction,
- &descriptor, &offset,
- record);
+ &descriptor, &offset,
+ record, write_op);
count++;
list_del(&record->hash);
kmem_cache_free(revoke_record_cache, record);
}
}
if (descriptor)
- flush_descriptor(journal, descriptor, offset);
+ flush_descriptor(journal, descriptor, offset, write_op);
jbd_debug(1, "Wrote %d revoke records\n", count);
}
-/*
+/*
* Write out one revoke record. We need to create a new descriptor
- * block if the old one is full or if we have not already created one.
+ * block if the old one is full or if we have not already created one.
*/
-static void write_one_revoke_record(journal_t *journal,
+static void write_one_revoke_record(journal_t *journal,
transaction_t *transaction,
- struct journal_head **descriptorp,
+ struct journal_head **descriptorp,
int *offsetp,
- struct jbd_revoke_record_s *record)
+ struct jbd_revoke_record_s *record,
+ int write_op)
{
struct journal_head *descriptor;
int offset;
/* Make sure we have a descriptor with space left for the record */
if (descriptor) {
if (offset == journal->j_blocksize) {
- flush_descriptor(journal, descriptor, offset);
+ flush_descriptor(journal, descriptor, offset, write_op);
descriptor = NULL;
}
}
*descriptorp = descriptor;
}
- * ((__be32 *)(&jh2bh(descriptor)->b_data[offset])) =
+ * ((__be32 *)(&jh2bh(descriptor)->b_data[offset])) =
cpu_to_be32(record->blocknr);
offset += 4;
*offsetp = offset;
}
-/*
+/*
* Flush a revoke descriptor out to the journal. If we are aborting,
* this is a noop; otherwise we are generating a buffer which needs to
* be waited for during commit, so it has to go onto the appropriate
* journal buffer list.
*/
-static void flush_descriptor(journal_t *journal,
- struct journal_head *descriptor,
- int offset)
+static void flush_descriptor(journal_t *journal,
+ struct journal_head *descriptor,
+ int offset, int write_op)
{
journal_revoke_header_t *header;
struct buffer_head *bh = jh2bh(descriptor);
set_buffer_jwrite(bh);
BUFFER_TRACE(bh, "write");
set_buffer_dirty(bh);
- ll_rw_block(WRITE, 1, &bh);
+ ll_rw_block((write_op == WRITE) ? SWRITE : SWRITE_SYNC_PLUG, 1, &bh);
}
#endif
-/*
+/*
* Revoke support for recovery.
*
* Recovery needs to be able to:
* check whether a given block in a given transaction should be replayed
* (ie. has not been revoked by a revoke record in that or a subsequent
* transaction)
- *
+ *
* empty the revoke table after recovery.
*/
* First, setting revoke records. We create a new revoke record for
* every block ever revoked in the log as we scan it for recovery, and
* we update the existing records if we find multiple revokes for a
- * single block.
+ * single block.
*/
-int journal_set_revoke(journal_t *journal,
- unsigned long blocknr,
+int journal_set_revoke(journal_t *journal,
+ unsigned int blocknr,
tid_t sequence)
{
struct jbd_revoke_record_s *record;
if (tid_gt(sequence, record->sequence))
record->sequence = sequence;
return 0;
- }
+ }
return insert_revoke_hash(journal, blocknr, sequence);
}
-/*
+/*
* Test revoke records. For a given block referenced in the log, has
* that block been revoked? A revoke record with a given transaction
* sequence number revokes all blocks in that transaction and earlier
* ones, but later transactions still need replayed.
*/
-int journal_test_revoke(journal_t *journal,
- unsigned long blocknr,
+int journal_test_revoke(journal_t *journal,
+ unsigned int blocknr,
tid_t sequence)
{
struct jbd_revoke_record_s *record;