Whamcloud - gitweb
Revert "e2fsck: Change kmem_cache_t to lkmem_cache_t for Solaris"
[tools/e2fsprogs.git] / e2fsck / revoke.c
index 38c265e..fa60878 100644 (file)
@@ -1,14 +1,11 @@
+// SPDX-License-Identifier: GPL-2.0+
 /*
- * linux/fs/revoke.c
+ * linux/fs/jbd2/revoke.c
  *
  * Written by Stephen C. Tweedie <sct@redhat.com>, 2000
  *
  * Copyright 2000 Red Hat corp --- All Rights Reserved
  *
- * This file is part of the Linux kernel and is made available under
- * the terms of the GNU General Public License, version 2, or at your
- * option, any later version, incorporated herein by reference.
- *
  * Journal revoke routines for the generic filesystem journaling code;
  * part of the ext2fs journaling system.
  *
  *   overwriting the new data.  We don't even need to clear the revoke
  *   bit here.
  *
+ * We cache revoke status of a buffer in the current transaction in b_states
+ * bits.  As the name says, revokevalid flag indicates that the cached revoke
+ * status of a buffer is valid and we can rely on the cached status.
+ *
  * Revoke information on buffers is a tri-state value:
  *
  * RevokeValid clear:  no cached revoke status, need to look it up
  *                     need do nothing.
  * RevokeValid set, Revoked set:
  *                     buffer has been revoked.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 no one else
+ * can touch them (filesystem isn't mounted yet) and hence no locking is
+ * needed.
  */
 
 #ifndef __KERNEL__
-#include "config.h"
 #include "jfs_user.h"
 #else
-#include <linux/sched.h>
+#include <linux/time.h>
 #include <linux/fs.h>
-#include <linux/jbd.h>
+#include <linux/jbd2.h>
 #include <linux/errno.h>
 #include <linux/slab.h>
-#include <linux/locks.h>
 #include <linux/list.h>
-#include <linux/smp_lock.h>
 #include <linux/init.h>
+#include <linux/bio.h>
+#include <linux/log2.h>
+#include <linux/hash.h>
 #endif
 
-static lkmem_cache_t *revoke_record_cache;
-static lkmem_cache_t *revoke_table_cache;
+static struct kmem_cache *jbd2_revoke_record_cache;
+static struct kmem_cache *jbd2_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 jbd2_revoke_record_s
 {
        struct list_head  hash;
        tid_t             sequence;     /* Used for recovery only */
-       unsigned long     blocknr;
+       unsigned long long        blocknr;
 };
 
 
 /* The revoke table is just a simple hash table of revoke records. */
-struct jbd_revoke_table_s
+struct jbd2_revoke_table_s
 {
        /* It is conceivable that we might want a larger hash table
         * for recovery.  Must be a power of two. */
@@ -99,167 +119,188 @@ struct jbd_revoke_table_s
 
 
 #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);
+static void write_one_revoke_record(transaction_t *,
+                                   struct list_head *,
+                                   struct buffer_head **, int *,
+                                   struct jbd2_revoke_record_s *);
+static void flush_descriptor(journal_t *, struct buffer_head *, 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 long long block)
 {
-       struct jbd_revoke_table_s *table = journal->j_revoke;
-       int hash_shift = table->hash_shift;
-
-       return ((block << (hash_shift - 6)) ^
-               (block >> 13) ^
-               (block << (hash_shift - 12))) & (table->hash_size - 1);
+       return hash_64(block, journal->j_revoke->hash_shift);
 }
 
-static int insert_revoke_hash(journal_t *journal, unsigned long blocknr,
+static int insert_revoke_hash(journal_t *journal, unsigned long long blocknr,
                              tid_t seq)
 {
        struct list_head *hash_list;
-       struct jbd_revoke_record_s *record;
+       struct jbd2_revoke_record_s *record;
+       gfp_t gfp_mask = GFP_NOFS;
 
-#ifdef __KERNEL__
-repeat:
-#endif
-       record = kmem_cache_alloc(revoke_record_cache, GFP_NOFS);
+       if (journal_oom_retry)
+               gfp_mask |= __GFP_NOFAIL;
+       record = kmem_cache_alloc(jbd2_revoke_record_cache, gfp_mask);
        if (!record)
-               goto oom;
+               return -ENOMEM;
 
        record->sequence = seq;
        record->blocknr = blocknr;
        hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
+       spin_lock(&journal->j_revoke_lock);
        list_add(&record->hash, hash_list);
+       spin_unlock(&journal->j_revoke_lock);
        return 0;
-
-oom:
-#ifdef __KERNEL__
-       if (!journal_oom_retry)
-               return -ENOMEM;
-       jbd_debug(1, "ENOMEM in " __FUNCTION__ ", retrying.\n");
-       current->policy |= SCHED_YIELD;
-       schedule();
-       goto repeat;
-#else
-       return -ENOMEM;
-#endif
 }
 
 /* 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)
+static struct jbd2_revoke_record_s *find_revoke_record(journal_t *journal,
+                                                     unsigned long long blocknr)
 {
        struct list_head *hash_list;
-       struct jbd_revoke_record_s *record;
+       struct jbd2_revoke_record_s *record;
 
        hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
 
-       record = (struct jbd_revoke_record_s *) hash_list->next;
+       spin_lock(&journal->j_revoke_lock);
+       record = (struct jbd2_revoke_record_s *) hash_list->next;
        while (&(record->hash) != hash_list) {
-               if (record->blocknr == blocknr)
+               if (record->blocknr == blocknr) {
+                       spin_unlock(&journal->j_revoke_lock);
                        return record;
-               record = (struct jbd_revoke_record_s *) record->hash.next;
+               }
+               record = (struct jbd2_revoke_record_s *) record->hash.next;
        }
+       spin_unlock(&journal->j_revoke_lock);
        return NULL;
 }
 
-int __init journal_init_revoke_caches(void)
+void jbd2_journal_destroy_revoke_record_cache(void)
 {
-       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;
-
-       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;
-       }
-       return 0;
+       kmem_cache_destroy(jbd2_revoke_record_cache);
+       jbd2_revoke_record_cache = NULL;
 }
 
-void journal_destroy_revoke_caches(void)
+void jbd2_journal_destroy_revoke_table_cache(void)
 {
-       kmem_cache_destroy(revoke_record_cache);
-       revoke_record_cache = 0;
-       kmem_cache_destroy(revoke_table_cache);
-       revoke_table_cache = 0;
+       kmem_cache_destroy(jbd2_revoke_table_cache);
+       jbd2_revoke_table_cache = NULL;
 }
 
-/* Initialise the revoke table for a given journal to a given size. */
-
-int journal_init_revoke(journal_t *journal, int hash_size)
+int __init jbd2_journal_init_revoke_record_cache(void)
 {
-       int shift, tmp;
+       J_ASSERT(!jbd2_revoke_record_cache);
+       jbd2_revoke_record_cache = KMEM_CACHE(jbd2_revoke_record_s,
+                                       SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY);
 
-       J_ASSERT (journal->j_revoke == NULL);
+       if (!jbd2_revoke_record_cache) {
+               pr_emerg("JBD2: failed to create revoke_record cache\n");
+               return -ENOMEM;
+       }
+       return 0;
+}
 
-       journal->j_revoke = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
-       if (!journal->j_revoke)
+int __init jbd2_journal_init_revoke_table_cache(void)
+{
+       J_ASSERT(!jbd2_revoke_table_cache);
+       jbd2_revoke_table_cache = KMEM_CACHE(jbd2_revoke_table_s,
+                                            SLAB_TEMPORARY);
+       if (!jbd2_revoke_table_cache) {
+               pr_emerg("JBD2: failed to create revoke_table cache\n");
                return -ENOMEM;
+       }
+       return 0;
+}
 
-       /* Check that the hash_size is a power of two */
-       J_ASSERT ((hash_size & (hash_size-1)) == 0);
+static struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size)
+{
+       int shift = 0;
+       int tmp = hash_size;
+       struct jbd2_revoke_table_s *table;
 
-       journal->j_revoke->hash_size = hash_size;
+       table = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL);
+       if (!table)
+               goto out;
 
-       shift = 0;
-       tmp = hash_size;
        while((tmp >>= 1UL) != 0UL)
                shift++;
-       journal->j_revoke->hash_shift = shift;
 
-       journal->j_revoke->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);
-               journal->j_revoke = NULL;
-               return -ENOMEM;
+       table->hash_size = hash_size;
+       table->hash_shift = shift;
+       table->hash_table =
+               kmalloc_array(hash_size, sizeof(struct list_head), GFP_KERNEL);
+       if (!table->hash_table) {
+               kmem_cache_free(jbd2_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]);
 
-       return 0;
+out:
+       return table;
 }
 
-/* Destoy a journal's revoke table.  The table must already be empty! */
-
-void journal_destroy_revoke(journal_t *journal)
+static void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table)
 {
-       struct jbd_revoke_table_s *table;
-       struct list_head *hash_list;
        int i;
+       struct list_head *hash_list;
 
-       table = journal->j_revoke;
-       if (!table)
-               return;
-
-       for (i=0; i<table->hash_size; i++) {
+       for (i = 0; i < table->hash_size; i++) {
                hash_list = &table->hash_table[i];
-               J_ASSERT (list_empty(hash_list));
+               J_ASSERT(list_empty(hash_list));
        }
 
        kfree(table->hash_table);
-       kmem_cache_free(revoke_table_cache, table);
+       kmem_cache_free(jbd2_revoke_table_cache, table);
+}
+
+/* Initialise the revoke table for a given journal to a given size. */
+int jbd2_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_table[0] = jbd2_journal_init_revoke_table(hash_size);
+       if (!journal->j_revoke_table[0])
+               goto fail0;
+
+       journal->j_revoke_table[1] = jbd2_journal_init_revoke_table(hash_size);
+       if (!journal->j_revoke_table[1])
+               goto fail1;
+
+       journal->j_revoke = journal->j_revoke_table[1];
+
+       spin_lock_init(&journal->j_revoke_lock);
+
+       return 0;
+
+fail1:
+       jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
+       journal->j_revoke_table[0] = NULL;
+fail0:
+       return -ENOMEM;
+}
+
+/* Destroy a journal's revoke table.  The table must already be empty! */
+void jbd2_journal_destroy_revoke(journal_t *journal)
+{
        journal->j_revoke = NULL;
+       if (journal->j_revoke_table[0])
+               jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
+       if (journal->j_revoke_table[1])
+               jbd2_journal_destroy_revoke_table(journal->j_revoke_table[1]);
 }
 
 
 #ifdef __KERNEL__
 
 /*
- * journal_revoke: revoke a given buffer_head from the journal.  This
+ * jbd2_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
@@ -271,96 +312,104 @@ void journal_destroy_revoke(journal_t *journal)
  * revoke before clearing the block bitmap when we are deleting
  * metadata.
  *
- * Revoke performs a journal_forget on any buffer_head passed in as a
+ * Revoke performs a jbd2_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.
  *
  * bh_in may not be a journalled buffer - it may have come off
  * the hash tables without an attached journal_head.
  *
- * If bh_in is non-zero, journal_revoke() will decrement its b_count
+ * If bh_in is non-zero, jbd2_journal_revoke() will decrement its b_count
  * by one.
  */
 
-int journal_revoke(handle_t *handle, unsigned long blocknr,
+int jbd2_journal_revoke(handle_t *handle, unsigned long long blocknr,
                   struct buffer_head *bh_in)
 {
        struct buffer_head *bh = NULL;
        journal_t *journal;
-       kdev_t dev;
+       struct block_device *bdev;
        int err;
 
+       might_sleep();
        if (bh_in)
                BUFFER_TRACE(bh_in, "enter");
 
        journal = handle->h_transaction->t_journal;
-       if (!journal_set_features(journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)){
+       if (!jbd2_journal_set_features(journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)){
                J_ASSERT (!"Cannot set revoke feature!");
                return -EINVAL;
        }
 
-       dev = journal->j_fs_dev;
+       bdev = journal->j_fs_dev;
        bh = bh_in;
 
        if (!bh) {
-               bh = get_hash_table(dev, blocknr, journal->j_blocksize);
+               bh = __find_get_block(bdev, blocknr, journal->j_blocksize);
                if (bh)
                        BUFFER_TRACE(bh, "found on hash");
        }
-#ifdef JBD_EXPENSIVE_CHECKING
+#ifdef JBD2_EXPENSIVE_CHECKING
        else {
                struct buffer_head *bh2;
 
                /* If there is a different buffer_head lying around in
                 * memory anywhere... */
-               bh2 = get_hash_table(dev, blocknr, journal->j_blocksize);
+               bh2 = __find_get_block(bdev, blocknr, journal->j_blocksize);
                if (bh2) {
                        /* ... and it has RevokeValid status... */
-                       if ((bh2 != bh) &&
-                           test_bit(BH_RevokeValid, &bh2->b_state))
+                       if (bh2 != bh && buffer_revokevalid(bh2))
                                /* ...then it better be revoked too,
                                 * since it's illegal to create a revoke
                                 * record against a buffer_head which is
                                 * not marked revoked --- that would
                                 * risk missing a subsequent revoke
                                 * cancel. */
-                               J_ASSERT_BH(bh2, test_bit(BH_Revoked, &
-                                                         bh2->b_state));
-                       __brelse(bh2);
+                               J_ASSERT_BH(bh2, buffer_revoked(bh2));
+                       put_bh(bh2);
                }
        }
 #endif
 
+       if (WARN_ON_ONCE(handle->h_revoke_credits <= 0)) {
+               if (!bh_in)
+                       brelse(bh);
+               return -EIO;
+       }
        /* We really ought not ever to revoke twice in a row without
            first having the revoke cancelled: it's illegal to free a
            block twice without allocating it in between! */
        if (bh) {
-               J_ASSERT_BH(bh, !test_bit(BH_Revoked, &bh->b_state));
-               set_bit(BH_Revoked, &bh->b_state);
-               set_bit(BH_RevokeValid, &bh->b_state);
+               if (!J_EXPECT_BH(bh, !buffer_revoked(bh),
+                                "inconsistent data on disk")) {
+                       if (!bh_in)
+                               brelse(bh);
+                       return -EIO;
+               }
+               set_buffer_revoked(bh);
+               set_buffer_revokevalid(bh);
                if (bh_in) {
-                       BUFFER_TRACE(bh_in, "call journal_forget");
-                       journal_forget(handle, bh_in);
+                       BUFFER_TRACE(bh_in, "call jbd2_journal_forget");
+                       jbd2_journal_forget(handle, bh_in);
                } else {
                        BUFFER_TRACE(bh, "call brelse");
                        __brelse(bh);
                }
        }
+       handle->h_revoke_credits--;
 
-       lock_journal(journal);
-       jbd_debug(2, "insert revoke for block %lu, bh_in=%p\n", blocknr, bh_in);
+       jbd_debug(2, "insert revoke for block %llu, bh_in=%p\n",blocknr, bh_in);
        err = insert_revoke_hash(journal, blocknr,
                                handle->h_transaction->t_tid);
-       unlock_journal(journal);
        BUFFER_TRACE(bh_in, "exit");
        return err;
 }
 
 /*
  * Cancel an outstanding revoke.  For use only internally by the
- * journaling code (called from journal_get_write_access).
+ * journaling code (called from jbd2_journal_get_write_access).
  *
- * We trust the BH_Revoked bit on the buffer if the buffer is already
+ * We trust buffer_revoked() on the buffer if the buffer is already
  * being journaled: if there is no revoke pending on the buffer, then we
  * don't do anything here.
  *
@@ -370,12 +419,10 @@ int journal_revoke(handle_t *handle, unsigned long blocknr,
  * 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)
+int jbd2_journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
 {
-       struct jbd_revoke_record_s *record;
+       struct jbd2_revoke_record_s *record;
        journal_t *journal = handle->h_transaction->t_journal;
        int need_cancel;
        int did_revoke = 0;     /* akpm: debug */
@@ -387,25 +434,27 @@ int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
         * only perform the full cancel if the revoke bit is set.  If
         * not, we can't trust the revoke bit, and we need to do the
         * full search for a revoke record. */
-       if (test_and_set_bit(BH_RevokeValid, &bh->b_state))
-               need_cancel = (test_and_clear_bit(BH_Revoked, &bh->b_state));
-       else {
+       if (test_set_buffer_revokevalid(bh)) {
+               need_cancel = test_clear_buffer_revoked(bh);
+       else {
                need_cancel = 1;
-               clear_bit(BH_Revoked, &bh->b_state);
+               clear_buffer_revoked(bh);
        }
 
        if (need_cancel) {
                record = find_revoke_record(journal, bh->b_blocknr);
                if (record) {
                        jbd_debug(4, "cancelled existing revoke on "
-                                 "blocknr %lu\n", bh->b_blocknr);
+                                 "blocknr %llu\n", (unsigned long long)bh->b_blocknr);
+                       spin_lock(&journal->j_revoke_lock);
                        list_del(&record->hash);
-                       kmem_cache_free(revoke_record_cache, record);
+                       spin_unlock(&journal->j_revoke_lock);
+                       kmem_cache_free(jbd2_revoke_record_cache, record);
                        did_revoke = 1;
                }
        }
 
-#ifdef JBD_EXPENSIVE_CHECKING
+#ifdef JBD2_EXPENSIVE_CHECKING
        /* There better not be one left behind by now! */
        record = find_revoke_record(journal, bh->b_blocknr);
        J_ASSERT_JH(jh, record == NULL);
@@ -415,52 +464,98 @@ int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
         * buffer_head?  If so, we'd better make sure we clear the
         * revoked status on any hashed alias too, otherwise the revoke
         * state machine will get very upset later on. */
-       if (need_cancel && !bh->b_pprev) {
+       if (need_cancel) {
                struct buffer_head *bh2;
-               bh2 = get_hash_table(bh->b_dev, bh->b_blocknr, bh->b_size);
+               bh2 = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
                if (bh2) {
-                       clear_bit(BH_Revoked, &bh2->b_state);
+                       if (bh2 != bh)
+                               clear_buffer_revoked(bh2);
                        __brelse(bh2);
                }
        }
-
        return did_revoke;
 }
 
+/*
+ * journal_clear_revoked_flag clears revoked flag of buffers in
+ * revoke table to reflect there is no revoked buffers in the next
+ * transaction which is going to be started.
+ */
+void jbd2_clear_buffer_revoked_flags(journal_t *journal)
+{
+       struct jbd2_revoke_table_s *revoke = journal->j_revoke;
+       int i = 0;
+
+       for (i = 0; i < revoke->hash_size; i++) {
+               struct list_head *hash_list;
+               struct list_head *list_entry;
+               hash_list = &revoke->hash_table[i];
+
+               list_for_each(list_entry, hash_list) {
+                       struct jbd2_revoke_record_s *record;
+                       struct buffer_head *bh;
+                       record = (struct jbd2_revoke_record_s *)list_entry;
+                       bh = __find_get_block(journal->j_fs_dev,
+                                             record->blocknr,
+                                             journal->j_blocksize);
+                       if (bh) {
+                               clear_buffer_revoked(bh);
+                               __brelse(bh);
+                       }
+               }
+       }
+}
+
+/* journal_switch_revoke table select j_revoke for next transaction
+ * we do not want to suspend any processing until all revokes are
+ * written -bzzz
+ */
+void jbd2_journal_switch_revoke_table(journal_t *journal)
+{
+       int i;
+
+       if (journal->j_revoke == journal->j_revoke_table[0])
+               journal->j_revoke = journal->j_revoke_table[1];
+       else
+               journal->j_revoke = journal->j_revoke_table[0];
+
+       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 jbd2_journal_write_revoke_records(transaction_t *transaction,
+                                      struct list_head *log_bufs)
 {
-       struct journal_head *descriptor;
-       struct jbd_revoke_record_s *record;
-       struct jbd_revoke_table_s *revoke;
+       journal_t *journal = transaction->t_journal;
+       struct buffer_head *descriptor;
+       struct jbd2_revoke_record_s *record;
+       struct jbd2_revoke_table_s *revoke;
        struct list_head *hash_list;
        int i, offset, count;
 
        descriptor = NULL;
        offset = 0;
        count = 0;
-       revoke = journal->j_revoke;
+
+       /* select revoke table for committing transaction */
+       revoke = journal->j_revoke == journal->j_revoke_table[0] ?
+               journal->j_revoke_table[1] : journal->j_revoke_table[0];
 
        for (i = 0; i < revoke->hash_size; i++) {
                hash_list = &revoke->hash_table[i];
 
                while (!list_empty(hash_list)) {
-                       record = (struct jbd_revoke_record_s *)
+                       record = (struct jbd2_revoke_record_s *)
                                hash_list->next;
-                       write_one_revoke_record(journal, transaction,
-                                               &descriptor, &offset,
-                                               record);
+                       write_one_revoke_record(transaction, log_bufs,
+                                               &descriptor, &offset, record);
                        count++;
                        list_del(&record->hash);
-                       kmem_cache_free(revoke_record_cache, record);
+                       kmem_cache_free(jbd2_revoke_record_cache, record);
                }
        }
        if (descriptor)
@@ -473,19 +568,20 @@ void journal_write_revoke_records(journal_t *journal,
  * block if the old one is full or if we have not already created one.
  */
 
-static void write_one_revoke_record(journal_t *journal,
-                                   transaction_t *transaction,
-                                   struct journal_head **descriptorp,
+static void write_one_revoke_record(transaction_t *transaction,
+                                   struct list_head *log_bufs,
+                                   struct buffer_head **descriptorp,
                                    int *offsetp,
-                                   struct jbd_revoke_record_s *record)
+                                   struct jbd2_revoke_record_s *record)
 {
-       struct journal_head *descriptor;
-       int offset;
-       journal_header_t *header;
+       journal_t *journal = transaction->t_journal;
+       int csum_size = 0;
+       struct buffer_head *descriptor;
+       int sz, offset;
 
        /* If we are already aborting, this all becomes a noop.  We
            still need to go round the loop in
-           journal_write_revoke_records in order to free all of the
+           jbd2_journal_write_revoke_records in order to free all of the
            revoke records: only the IO to the journal is omitted. */
        if (is_journal_aborted(journal))
                return;
@@ -493,34 +589,45 @@ static void write_one_revoke_record(journal_t *journal,
        descriptor = *descriptorp;
        offset = *offsetp;
 
+       /* Do we need to leave space at the end for a checksum? */
+       if (jbd2_journal_has_csum_v2or3(journal))
+               csum_size = sizeof(struct jbd2_journal_block_tail);
+
+       if (jbd2_has_feature_64bit(journal))
+               sz = 8;
+       else
+               sz = 4;
+
        /* Make sure we have a descriptor with space left for the record */
        if (descriptor) {
-               if (offset == journal->j_blocksize) {
+               if (offset + sz > journal->j_blocksize - csum_size) {
                        flush_descriptor(journal, descriptor, offset);
                        descriptor = NULL;
                }
        }
 
        if (!descriptor) {
-               descriptor = journal_get_descriptor_buffer(journal);
+               descriptor = jbd2_journal_get_descriptor_buffer(transaction,
+                                                       JBD2_REVOKE_BLOCK);
                if (!descriptor)
                        return;
-               header = (journal_header_t *) &jh2bh(descriptor)->b_data[0];
-               header->h_magic     = htonl(JFS_MAGIC_NUMBER);
-               header->h_blocktype = htonl(JFS_REVOKE_BLOCK);
-               header->h_sequence  = htonl(transaction->t_tid);
 
                /* Record it so that we can wait for IO completion later */
-               JBUFFER_TRACE(descriptor, "file as BJ_LogCtl");
-               journal_file_buffer(descriptor, transaction, BJ_LogCtl);
+               BUFFER_TRACE(descriptor, "file in log_bufs");
+               jbd2_file_log_bh(log_bufs, descriptor);
 
-               offset = sizeof(journal_revoke_header_t);
+               offset = sizeof(jbd2_journal_revoke_header_t);
                *descriptorp = descriptor;
        }
 
-       * ((unsigned int *)(&jh2bh(descriptor)->b_data[offset])) =
-               htonl(record->blocknr);
-       offset += 4;
+       if (jbd2_has_feature_64bit(journal))
+               * ((__be64 *)(&descriptor->b_data[offset])) =
+                       cpu_to_be64(record->blocknr);
+       else
+               * ((__be32 *)(&descriptor->b_data[offset])) =
+                       cpu_to_be32(record->blocknr);
+       offset += sz;
+
        *offsetp = offset;
 }
 
@@ -532,27 +639,23 @@ static void write_one_revoke_record(journal_t *journal,
  */
 
 static void flush_descriptor(journal_t *journal,
-                            struct journal_head *descriptor,
+                            struct buffer_head *descriptor,
                             int offset)
 {
-       journal_revoke_header_t *header;
+       jbd2_journal_revoke_header_t *header;
 
-       if (is_journal_aborted(journal)) {
-               JBUFFER_TRACE(descriptor, "brelse");
-               __brelse(jh2bh(descriptor));
+       if (is_journal_aborted(journal))
                return;
-       }
 
-       header = (journal_revoke_header_t *) jh2bh(descriptor)->b_data;
-       header->r_count = htonl(offset);
-       set_bit(BH_JWrite, &jh2bh(descriptor)->b_state);
-       {
-               struct buffer_head *bh = jh2bh(descriptor);
-               BUFFER_TRACE(bh, "write");
-               ll_rw_block (WRITE, 1, &bh);
-       }
-}
+       header = (jbd2_journal_revoke_header_t *)descriptor->b_data;
+       header->r_count = cpu_to_be32(offset);
+       jbd2_descriptor_block_csum_set(journal, descriptor);
 
+       set_buffer_jwrite(descriptor);
+       BUFFER_TRACE(descriptor, "write");
+       set_buffer_dirty(descriptor);
+       write_dirty_buffer(descriptor, REQ_SYNC);
+}
 #endif
 
 /*
@@ -577,15 +680,15 @@ static void flush_descriptor(journal_t *journal,
  * single block.
  */
 
-int journal_set_revoke(journal_t *journal,
-                      unsigned long blocknr,
+int jbd2_journal_set_revoke(journal_t *journal,
+                      unsigned long long blocknr,
                       tid_t sequence)
 {
-       struct jbd_revoke_record_s *record;
+       struct jbd2_revoke_record_s *record;
 
        record = find_revoke_record(journal, blocknr);
        if (record) {
-               /* If we have multiple occurences, only record the
+               /* If we have multiple occurrences, only record the
                 * latest sequence number in the hashed record */
                if (tid_gt(sequence, record->sequence))
                        record->sequence = sequence;
@@ -601,11 +704,11 @@ int journal_set_revoke(journal_t *journal,
  * ones, but later transactions still need replayed.
  */
 
-int journal_test_revoke(journal_t *journal,
-                       unsigned long blocknr,
+int jbd2_journal_test_revoke(journal_t *journal,
+                       unsigned long long blocknr,
                        tid_t sequence)
 {
-       struct jbd_revoke_record_s *record;
+       struct jbd2_revoke_record_s *record;
 
        record = find_revoke_record(journal, blocknr);
        if (!record)
@@ -620,22 +723,21 @@ int journal_test_revoke(journal_t *journal,
  * that it can be reused by the running filesystem.
  */
 
-void journal_clear_revoke(journal_t *journal)
+void jbd2_journal_clear_revoke(journal_t *journal)
 {
        int i;
        struct list_head *hash_list;
-       struct jbd_revoke_record_s *record;
-       struct jbd_revoke_table_s *revoke;
+       struct jbd2_revoke_record_s *record;
+       struct jbd2_revoke_table_s *revoke;
 
        revoke = journal->j_revoke;
 
        for (i = 0; i < revoke->hash_size; i++) {
                hash_list = &revoke->hash_table[i];
                while (!list_empty(hash_list)) {
-                       record = (struct jbd_revoke_record_s*) hash_list->next;
+                       record = (struct jbd2_revoke_record_s*) hash_list->next;
                        list_del(&record->hash);
-                       kmem_cache_free(revoke_record_cache, record);
+                       kmem_cache_free(jbd2_revoke_record_cache, record);
                }
        }
 }
-