Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / gfs2 / lops.c
diff --git a/kernel/fs/gfs2/lops.c b/kernel/fs/gfs2/lops.c
new file mode 100644 (file)
index 0000000..2c1ae86
--- /dev/null
@@ -0,0 +1,886 @@
+/*
+ * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
+ * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
+ *
+ * This copyrighted material is made available to anyone wishing to use,
+ * modify, copy, or redistribute it subject to the terms and conditions
+ * of the GNU General Public License version 2.
+ */
+
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/buffer_head.h>
+#include <linux/mempool.h>
+#include <linux/gfs2_ondisk.h>
+#include <linux/bio.h>
+#include <linux/fs.h>
+#include <linux/list_sort.h>
+
+#include "gfs2.h"
+#include "incore.h"
+#include "inode.h"
+#include "glock.h"
+#include "log.h"
+#include "lops.h"
+#include "meta_io.h"
+#include "recovery.h"
+#include "rgrp.h"
+#include "trans.h"
+#include "util.h"
+#include "trace_gfs2.h"
+
+/**
+ * gfs2_pin - Pin a buffer in memory
+ * @sdp: The superblock
+ * @bh: The buffer to be pinned
+ *
+ * The log lock must be held when calling this function
+ */
+void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh)
+{
+       struct gfs2_bufdata *bd;
+
+       BUG_ON(!current->journal_info);
+
+       clear_buffer_dirty(bh);
+       if (test_set_buffer_pinned(bh))
+               gfs2_assert_withdraw(sdp, 0);
+       if (!buffer_uptodate(bh))
+               gfs2_io_error_bh(sdp, bh);
+       bd = bh->b_private;
+       /* If this buffer is in the AIL and it has already been written
+        * to in-place disk block, remove it from the AIL.
+        */
+       spin_lock(&sdp->sd_ail_lock);
+       if (bd->bd_tr)
+               list_move(&bd->bd_ail_st_list, &bd->bd_tr->tr_ail2_list);
+       spin_unlock(&sdp->sd_ail_lock);
+       get_bh(bh);
+       atomic_inc(&sdp->sd_log_pinned);
+       trace_gfs2_pin(bd, 1);
+}
+
+static bool buffer_is_rgrp(const struct gfs2_bufdata *bd)
+{
+       return bd->bd_gl->gl_name.ln_type == LM_TYPE_RGRP;
+}
+
+static void maybe_release_space(struct gfs2_bufdata *bd)
+{
+       struct gfs2_glock *gl = bd->bd_gl;
+       struct gfs2_sbd *sdp = gl->gl_sbd;
+       struct gfs2_rgrpd *rgd = gl->gl_object;
+       unsigned int index = bd->bd_bh->b_blocknr - gl->gl_name.ln_number;
+       struct gfs2_bitmap *bi = rgd->rd_bits + index;
+
+       if (bi->bi_clone == NULL)
+               return;
+       if (sdp->sd_args.ar_discard)
+               gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL);
+       memcpy(bi->bi_clone + bi->bi_offset,
+              bd->bd_bh->b_data + bi->bi_offset, bi->bi_len);
+       clear_bit(GBF_FULL, &bi->bi_flags);
+       rgd->rd_free_clone = rgd->rd_free;
+       rgd->rd_extfail_pt = rgd->rd_free;
+}
+
+/**
+ * gfs2_unpin - Unpin a buffer
+ * @sdp: the filesystem the buffer belongs to
+ * @bh: The buffer to unpin
+ * @ai:
+ * @flags: The inode dirty flags
+ *
+ */
+
+static void gfs2_unpin(struct gfs2_sbd *sdp, struct buffer_head *bh,
+                      struct gfs2_trans *tr)
+{
+       struct gfs2_bufdata *bd = bh->b_private;
+
+       BUG_ON(!buffer_uptodate(bh));
+       BUG_ON(!buffer_pinned(bh));
+
+       lock_buffer(bh);
+       mark_buffer_dirty(bh);
+       clear_buffer_pinned(bh);
+
+       if (buffer_is_rgrp(bd))
+               maybe_release_space(bd);
+
+       spin_lock(&sdp->sd_ail_lock);
+       if (bd->bd_tr) {
+               list_del(&bd->bd_ail_st_list);
+               brelse(bh);
+       } else {
+               struct gfs2_glock *gl = bd->bd_gl;
+               list_add(&bd->bd_ail_gl_list, &gl->gl_ail_list);
+               atomic_inc(&gl->gl_ail_count);
+       }
+       bd->bd_tr = tr;
+       list_add(&bd->bd_ail_st_list, &tr->tr_ail1_list);
+       spin_unlock(&sdp->sd_ail_lock);
+
+       clear_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+       trace_gfs2_pin(bd, 0);
+       unlock_buffer(bh);
+       atomic_dec(&sdp->sd_log_pinned);
+}
+
+static void gfs2_log_incr_head(struct gfs2_sbd *sdp)
+{
+       BUG_ON((sdp->sd_log_flush_head == sdp->sd_log_tail) &&
+              (sdp->sd_log_flush_head != sdp->sd_log_head));
+
+       if (++sdp->sd_log_flush_head == sdp->sd_jdesc->jd_blocks) {
+               sdp->sd_log_flush_head = 0;
+               sdp->sd_log_flush_wrapped = 1;
+       }
+}
+
+static u64 gfs2_log_bmap(struct gfs2_sbd *sdp)
+{
+       unsigned int lbn = sdp->sd_log_flush_head;
+       struct gfs2_journal_extent *je;
+       u64 block;
+
+       list_for_each_entry(je, &sdp->sd_jdesc->extent_list, list) {
+               if ((lbn >= je->lblock) && (lbn < (je->lblock + je->blocks))) {
+                       block = je->dblock + lbn - je->lblock;
+                       gfs2_log_incr_head(sdp);
+                       return block;
+               }
+       }
+
+       return -1;
+}
+
+/**
+ * gfs2_end_log_write_bh - end log write of pagecache data with buffers
+ * @sdp: The superblock
+ * @bvec: The bio_vec
+ * @error: The i/o status
+ *
+ * This finds the relavent buffers and unlocks then and sets the
+ * error flag according to the status of the i/o request. This is
+ * used when the log is writing data which has an in-place version
+ * that is pinned in the pagecache.
+ */
+
+static void gfs2_end_log_write_bh(struct gfs2_sbd *sdp, struct bio_vec *bvec,
+                                 int error)
+{
+       struct buffer_head *bh, *next;
+       struct page *page = bvec->bv_page;
+       unsigned size;
+
+       bh = page_buffers(page);
+       size = bvec->bv_len;
+       while (bh_offset(bh) < bvec->bv_offset)
+               bh = bh->b_this_page;
+       do {
+               if (error)
+                       set_buffer_write_io_error(bh);
+               unlock_buffer(bh);
+               next = bh->b_this_page;
+               size -= bh->b_size;
+               brelse(bh);
+               bh = next;
+       } while(bh && size);
+}
+
+/**
+ * gfs2_end_log_write - end of i/o to the log
+ * @bio: The bio
+ * @error: Status of i/o request
+ *
+ * Each bio_vec contains either data from the pagecache or data
+ * relating to the log itself. Here we iterate over the bio_vec
+ * array, processing both kinds of data.
+ *
+ */
+
+static void gfs2_end_log_write(struct bio *bio, int error)
+{
+       struct gfs2_sbd *sdp = bio->bi_private;
+       struct bio_vec *bvec;
+       struct page *page;
+       int i;
+
+       if (error) {
+               sdp->sd_log_error = error;
+               fs_err(sdp, "Error %d writing to log\n", error);
+       }
+
+       bio_for_each_segment_all(bvec, bio, i) {
+               page = bvec->bv_page;
+               if (page_has_buffers(page))
+                       gfs2_end_log_write_bh(sdp, bvec, error);
+               else
+                       mempool_free(page, gfs2_page_pool);
+       }
+
+       bio_put(bio);
+       if (atomic_dec_and_test(&sdp->sd_log_in_flight))
+               wake_up(&sdp->sd_log_flush_wait);
+}
+
+/**
+ * gfs2_log_flush_bio - Submit any pending log bio
+ * @sdp: The superblock
+ * @rw: The rw flags
+ *
+ * Submit any pending part-built or full bio to the block device. If
+ * there is no pending bio, then this is a no-op.
+ */
+
+void gfs2_log_flush_bio(struct gfs2_sbd *sdp, int rw)
+{
+       if (sdp->sd_log_bio) {
+               atomic_inc(&sdp->sd_log_in_flight);
+               submit_bio(rw, sdp->sd_log_bio);
+               sdp->sd_log_bio = NULL;
+       }
+}
+
+/**
+ * gfs2_log_alloc_bio - Allocate a new bio for log writing
+ * @sdp: The superblock
+ * @blkno: The next device block number we want to write to
+ *
+ * This should never be called when there is a cached bio in the
+ * super block. When it returns, there will be a cached bio in the
+ * super block which will have as many bio_vecs as the device is
+ * happy to handle.
+ *
+ * Returns: Newly allocated bio
+ */
+
+static struct bio *gfs2_log_alloc_bio(struct gfs2_sbd *sdp, u64 blkno)
+{
+       struct super_block *sb = sdp->sd_vfs;
+       unsigned nrvecs = bio_get_nr_vecs(sb->s_bdev);
+       struct bio *bio;
+
+       BUG_ON(sdp->sd_log_bio);
+
+       while (1) {
+               bio = bio_alloc(GFP_NOIO, nrvecs);
+               if (likely(bio))
+                       break;
+               nrvecs = max(nrvecs/2, 1U);
+       }
+
+       bio->bi_iter.bi_sector = blkno * (sb->s_blocksize >> 9);
+       bio->bi_bdev = sb->s_bdev;
+       bio->bi_end_io = gfs2_end_log_write;
+       bio->bi_private = sdp;
+
+       sdp->sd_log_bio = bio;
+
+       return bio;
+}
+
+/**
+ * gfs2_log_get_bio - Get cached log bio, or allocate a new one
+ * @sdp: The superblock
+ * @blkno: The device block number we want to write to
+ *
+ * If there is a cached bio, then if the next block number is sequential
+ * with the previous one, return it, otherwise flush the bio to the
+ * device. If there is not a cached bio, or we just flushed it, then
+ * allocate a new one.
+ *
+ * Returns: The bio to use for log writes
+ */
+
+static struct bio *gfs2_log_get_bio(struct gfs2_sbd *sdp, u64 blkno)
+{
+       struct bio *bio = sdp->sd_log_bio;
+       u64 nblk;
+
+       if (bio) {
+               nblk = bio_end_sector(bio);
+               nblk >>= sdp->sd_fsb2bb_shift;
+               if (blkno == nblk)
+                       return bio;
+               gfs2_log_flush_bio(sdp, WRITE);
+       }
+
+       return gfs2_log_alloc_bio(sdp, blkno);
+}
+
+
+/**
+ * gfs2_log_write - write to log
+ * @sdp: the filesystem
+ * @page: the page to write
+ * @size: the size of the data to write
+ * @offset: the offset within the page 
+ *
+ * Try and add the page segment to the current bio. If that fails,
+ * submit the current bio to the device and create a new one, and
+ * then add the page segment to that.
+ */
+
+static void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page,
+                          unsigned size, unsigned offset)
+{
+       u64 blkno = gfs2_log_bmap(sdp);
+       struct bio *bio;
+       int ret;
+
+       bio = gfs2_log_get_bio(sdp, blkno);
+       ret = bio_add_page(bio, page, size, offset);
+       if (ret == 0) {
+               gfs2_log_flush_bio(sdp, WRITE);
+               bio = gfs2_log_alloc_bio(sdp, blkno);
+               ret = bio_add_page(bio, page, size, offset);
+               WARN_ON(ret == 0);
+       }
+}
+
+/**
+ * gfs2_log_write_bh - write a buffer's content to the log
+ * @sdp: The super block
+ * @bh: The buffer pointing to the in-place location
+ * 
+ * This writes the content of the buffer to the next available location
+ * in the log. The buffer will be unlocked once the i/o to the log has
+ * completed.
+ */
+
+static void gfs2_log_write_bh(struct gfs2_sbd *sdp, struct buffer_head *bh)
+{
+       gfs2_log_write(sdp, bh->b_page, bh->b_size, bh_offset(bh));
+}
+
+/**
+ * gfs2_log_write_page - write one block stored in a page, into the log
+ * @sdp: The superblock
+ * @page: The struct page
+ *
+ * This writes the first block-sized part of the page into the log. Note
+ * that the page must have been allocated from the gfs2_page_pool mempool
+ * and that after this has been called, ownership has been transferred and
+ * the page may be freed at any time.
+ */
+
+void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page)
+{
+       struct super_block *sb = sdp->sd_vfs;
+       gfs2_log_write(sdp, page, sb->s_blocksize, 0);
+}
+
+static struct page *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type,
+                                     u32 ld_length, u32 ld_data1)
+{
+       struct page *page = mempool_alloc(gfs2_page_pool, GFP_NOIO);
+       struct gfs2_log_descriptor *ld = page_address(page);
+       clear_page(ld);
+       ld->ld_header.mh_magic = cpu_to_be32(GFS2_MAGIC);
+       ld->ld_header.mh_type = cpu_to_be32(GFS2_METATYPE_LD);
+       ld->ld_header.mh_format = cpu_to_be32(GFS2_FORMAT_LD);
+       ld->ld_type = cpu_to_be32(ld_type);
+       ld->ld_length = cpu_to_be32(ld_length);
+       ld->ld_data1 = cpu_to_be32(ld_data1);
+       ld->ld_data2 = 0;
+       return page;
+}
+
+static void gfs2_check_magic(struct buffer_head *bh)
+{
+       void *kaddr;
+       __be32 *ptr;
+
+       clear_buffer_escaped(bh);
+       kaddr = kmap_atomic(bh->b_page);
+       ptr = kaddr + bh_offset(bh);
+       if (*ptr == cpu_to_be32(GFS2_MAGIC))
+               set_buffer_escaped(bh);
+       kunmap_atomic(kaddr);
+}
+
+static int blocknr_cmp(void *priv, struct list_head *a, struct list_head *b)
+{
+       struct gfs2_bufdata *bda, *bdb;
+
+       bda = list_entry(a, struct gfs2_bufdata, bd_list);
+       bdb = list_entry(b, struct gfs2_bufdata, bd_list);
+
+       if (bda->bd_bh->b_blocknr < bdb->bd_bh->b_blocknr)
+               return -1;
+       if (bda->bd_bh->b_blocknr > bdb->bd_bh->b_blocknr)
+               return 1;
+       return 0;
+}
+
+static void gfs2_before_commit(struct gfs2_sbd *sdp, unsigned int limit,
+                               unsigned int total, struct list_head *blist,
+                               bool is_databuf)
+{
+       struct gfs2_log_descriptor *ld;
+       struct gfs2_bufdata *bd1 = NULL, *bd2;
+       struct page *page;
+       unsigned int num;
+       unsigned n;
+       __be64 *ptr;
+
+       gfs2_log_lock(sdp);
+       list_sort(NULL, blist, blocknr_cmp);
+       bd1 = bd2 = list_prepare_entry(bd1, blist, bd_list);
+       while(total) {
+               num = total;
+               if (total > limit)
+                       num = limit;
+               gfs2_log_unlock(sdp);
+               page = gfs2_get_log_desc(sdp,
+                                        is_databuf ? GFS2_LOG_DESC_JDATA :
+                                        GFS2_LOG_DESC_METADATA, num + 1, num);
+               ld = page_address(page);
+               gfs2_log_lock(sdp);
+               ptr = (__be64 *)(ld + 1);
+
+               n = 0;
+               list_for_each_entry_continue(bd1, blist, bd_list) {
+                       *ptr++ = cpu_to_be64(bd1->bd_bh->b_blocknr);
+                       if (is_databuf) {
+                               gfs2_check_magic(bd1->bd_bh);
+                               *ptr++ = cpu_to_be64(buffer_escaped(bd1->bd_bh) ? 1 : 0);
+                       }
+                       if (++n >= num)
+                               break;
+               }
+
+               gfs2_log_unlock(sdp);
+               gfs2_log_write_page(sdp, page);
+               gfs2_log_lock(sdp);
+
+               n = 0;
+               list_for_each_entry_continue(bd2, blist, bd_list) {
+                       get_bh(bd2->bd_bh);
+                       gfs2_log_unlock(sdp);
+                       lock_buffer(bd2->bd_bh);
+
+                       if (buffer_escaped(bd2->bd_bh)) {
+                               void *kaddr;
+                               page = mempool_alloc(gfs2_page_pool, GFP_NOIO);
+                               ptr = page_address(page);
+                               kaddr = kmap_atomic(bd2->bd_bh->b_page);
+                               memcpy(ptr, kaddr + bh_offset(bd2->bd_bh),
+                                      bd2->bd_bh->b_size);
+                               kunmap_atomic(kaddr);
+                               *(__be32 *)ptr = 0;
+                               clear_buffer_escaped(bd2->bd_bh);
+                               unlock_buffer(bd2->bd_bh);
+                               brelse(bd2->bd_bh);
+                               gfs2_log_write_page(sdp, page);
+                       } else {
+                               gfs2_log_write_bh(sdp, bd2->bd_bh);
+                       }
+                       gfs2_log_lock(sdp);
+                       if (++n >= num)
+                               break;
+               }
+
+               BUG_ON(total < num);
+               total -= num;
+       }
+       gfs2_log_unlock(sdp);
+}
+
+static void buf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       unsigned int limit = buf_limit(sdp); /* 503 for 4k blocks */
+       unsigned int nbuf;
+       if (tr == NULL)
+               return;
+       nbuf = tr->tr_num_buf_new - tr->tr_num_buf_rm;
+       gfs2_before_commit(sdp, limit, nbuf, &tr->tr_buf, 0);
+}
+
+static void buf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       struct list_head *head;
+       struct gfs2_bufdata *bd;
+
+       if (tr == NULL)
+               return;
+
+       head = &tr->tr_buf;
+       while (!list_empty(head)) {
+               bd = list_entry(head->next, struct gfs2_bufdata, bd_list);
+               list_del_init(&bd->bd_list);
+               gfs2_unpin(sdp, bd->bd_bh, tr);
+       }
+}
+
+static void buf_lo_before_scan(struct gfs2_jdesc *jd,
+                              struct gfs2_log_header_host *head, int pass)
+{
+       if (pass != 0)
+               return;
+
+       jd->jd_found_blocks = 0;
+       jd->jd_replayed_blocks = 0;
+}
+
+static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
+                               struct gfs2_log_descriptor *ld, __be64 *ptr,
+                               int pass)
+{
+       struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+       struct gfs2_glock *gl = ip->i_gl;
+       unsigned int blks = be32_to_cpu(ld->ld_data1);
+       struct buffer_head *bh_log, *bh_ip;
+       u64 blkno;
+       int error = 0;
+
+       if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_METADATA)
+               return 0;
+
+       gfs2_replay_incr_blk(sdp, &start);
+
+       for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
+               blkno = be64_to_cpu(*ptr++);
+
+               jd->jd_found_blocks++;
+
+               if (gfs2_revoke_check(jd, blkno, start))
+                       continue;
+
+               error = gfs2_replay_read_block(jd, start, &bh_log);
+               if (error)
+                       return error;
+
+               bh_ip = gfs2_meta_new(gl, blkno);
+               memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size);
+
+               if (gfs2_meta_check(sdp, bh_ip))
+                       error = -EIO;
+               else
+                       mark_buffer_dirty(bh_ip);
+
+               brelse(bh_log);
+               brelse(bh_ip);
+
+               if (error)
+                       break;
+
+               jd->jd_replayed_blocks++;
+       }
+
+       return error;
+}
+
+/**
+ * gfs2_meta_sync - Sync all buffers associated with a glock
+ * @gl: The glock
+ *
+ */
+
+static void gfs2_meta_sync(struct gfs2_glock *gl)
+{
+       struct address_space *mapping = gfs2_glock2aspace(gl);
+       struct gfs2_sbd *sdp = gl->gl_sbd;
+       int error;
+
+       if (mapping == NULL)
+               mapping = &sdp->sd_aspace;
+
+       filemap_fdatawrite(mapping);
+       error = filemap_fdatawait(mapping);
+
+       if (error)
+               gfs2_io_error(gl->gl_sbd);
+}
+
+static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
+{
+       struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+
+       if (error) {
+               gfs2_meta_sync(ip->i_gl);
+               return;
+       }
+       if (pass != 1)
+               return;
+
+       gfs2_meta_sync(ip->i_gl);
+
+       fs_info(sdp, "jid=%u: Replayed %u of %u blocks\n",
+               jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks);
+}
+
+static void revoke_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       struct gfs2_meta_header *mh;
+       unsigned int offset;
+       struct list_head *head = &sdp->sd_log_le_revoke;
+       struct gfs2_bufdata *bd;
+       struct page *page;
+       unsigned int length;
+
+       gfs2_write_revokes(sdp);
+       if (!sdp->sd_log_num_revoke)
+               return;
+
+       length = gfs2_struct2blk(sdp, sdp->sd_log_num_revoke, sizeof(u64));
+       page = gfs2_get_log_desc(sdp, GFS2_LOG_DESC_REVOKE, length, sdp->sd_log_num_revoke);
+       offset = sizeof(struct gfs2_log_descriptor);
+
+       list_for_each_entry(bd, head, bd_list) {
+               sdp->sd_log_num_revoke--;
+
+               if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) {
+
+                       gfs2_log_write_page(sdp, page);
+                       page = mempool_alloc(gfs2_page_pool, GFP_NOIO);
+                       mh = page_address(page);
+                       clear_page(mh);
+                       mh->mh_magic = cpu_to_be32(GFS2_MAGIC);
+                       mh->mh_type = cpu_to_be32(GFS2_METATYPE_LB);
+                       mh->mh_format = cpu_to_be32(GFS2_FORMAT_LB);
+                       offset = sizeof(struct gfs2_meta_header);
+               }
+
+               *(__be64 *)(page_address(page) + offset) = cpu_to_be64(bd->bd_blkno);
+               offset += sizeof(u64);
+       }
+       gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke);
+
+       gfs2_log_write_page(sdp, page);
+}
+
+static void revoke_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       struct list_head *head = &sdp->sd_log_le_revoke;
+       struct gfs2_bufdata *bd;
+       struct gfs2_glock *gl;
+
+       while (!list_empty(head)) {
+               bd = list_entry(head->next, struct gfs2_bufdata, bd_list);
+               list_del_init(&bd->bd_list);
+               gl = bd->bd_gl;
+               atomic_dec(&gl->gl_revokes);
+               clear_bit(GLF_LFLUSH, &gl->gl_flags);
+               kmem_cache_free(gfs2_bufdata_cachep, bd);
+       }
+}
+
+static void revoke_lo_before_scan(struct gfs2_jdesc *jd,
+                                 struct gfs2_log_header_host *head, int pass)
+{
+       if (pass != 0)
+               return;
+
+       jd->jd_found_revokes = 0;
+       jd->jd_replay_tail = head->lh_tail;
+}
+
+static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
+                                  struct gfs2_log_descriptor *ld, __be64 *ptr,
+                                  int pass)
+{
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+       unsigned int blks = be32_to_cpu(ld->ld_length);
+       unsigned int revokes = be32_to_cpu(ld->ld_data1);
+       struct buffer_head *bh;
+       unsigned int offset;
+       u64 blkno;
+       int first = 1;
+       int error;
+
+       if (pass != 0 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_REVOKE)
+               return 0;
+
+       offset = sizeof(struct gfs2_log_descriptor);
+
+       for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
+               error = gfs2_replay_read_block(jd, start, &bh);
+               if (error)
+                       return error;
+
+               if (!first)
+                       gfs2_metatype_check(sdp, bh, GFS2_METATYPE_LB);
+
+               while (offset + sizeof(u64) <= sdp->sd_sb.sb_bsize) {
+                       blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset));
+
+                       error = gfs2_revoke_add(jd, blkno, start);
+                       if (error < 0) {
+                               brelse(bh);
+                               return error;
+                       }
+                       else if (error)
+                               jd->jd_found_revokes++;
+
+                       if (!--revokes)
+                               break;
+                       offset += sizeof(u64);
+               }
+
+               brelse(bh);
+               offset = sizeof(struct gfs2_meta_header);
+               first = 0;
+       }
+
+       return 0;
+}
+
+static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
+{
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+
+       if (error) {
+               gfs2_revoke_clean(jd);
+               return;
+       }
+       if (pass != 1)
+               return;
+
+       fs_info(sdp, "jid=%u: Found %u revoke tags\n",
+               jd->jd_jid, jd->jd_found_revokes);
+
+       gfs2_revoke_clean(jd);
+}
+
+/**
+ * databuf_lo_before_commit - Scan the data buffers, writing as we go
+ *
+ */
+
+static void databuf_lo_before_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       unsigned int limit = databuf_limit(sdp);
+       unsigned int nbuf;
+       if (tr == NULL)
+               return;
+       nbuf = tr->tr_num_databuf_new - tr->tr_num_databuf_rm;
+       gfs2_before_commit(sdp, limit, nbuf, &tr->tr_databuf, 1);
+}
+
+static int databuf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
+                                   struct gfs2_log_descriptor *ld,
+                                   __be64 *ptr, int pass)
+{
+       struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+       struct gfs2_glock *gl = ip->i_gl;
+       unsigned int blks = be32_to_cpu(ld->ld_data1);
+       struct buffer_head *bh_log, *bh_ip;
+       u64 blkno;
+       u64 esc;
+       int error = 0;
+
+       if (pass != 1 || be32_to_cpu(ld->ld_type) != GFS2_LOG_DESC_JDATA)
+               return 0;
+
+       gfs2_replay_incr_blk(sdp, &start);
+       for (; blks; gfs2_replay_incr_blk(sdp, &start), blks--) {
+               blkno = be64_to_cpu(*ptr++);
+               esc = be64_to_cpu(*ptr++);
+
+               jd->jd_found_blocks++;
+
+               if (gfs2_revoke_check(jd, blkno, start))
+                       continue;
+
+               error = gfs2_replay_read_block(jd, start, &bh_log);
+               if (error)
+                       return error;
+
+               bh_ip = gfs2_meta_new(gl, blkno);
+               memcpy(bh_ip->b_data, bh_log->b_data, bh_log->b_size);
+
+               /* Unescape */
+               if (esc) {
+                       __be32 *eptr = (__be32 *)bh_ip->b_data;
+                       *eptr = cpu_to_be32(GFS2_MAGIC);
+               }
+               mark_buffer_dirty(bh_ip);
+
+               brelse(bh_log);
+               brelse(bh_ip);
+
+               jd->jd_replayed_blocks++;
+       }
+
+       return error;
+}
+
+/* FIXME: sort out accounting for log blocks etc. */
+
+static void databuf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
+{
+       struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
+       struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
+
+       if (error) {
+               gfs2_meta_sync(ip->i_gl);
+               return;
+       }
+       if (pass != 1)
+               return;
+
+       /* data sync? */
+       gfs2_meta_sync(ip->i_gl);
+
+       fs_info(sdp, "jid=%u: Replayed %u of %u data blocks\n",
+               jd->jd_jid, jd->jd_replayed_blocks, jd->jd_found_blocks);
+}
+
+static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+       struct list_head *head;
+       struct gfs2_bufdata *bd;
+
+       if (tr == NULL)
+               return;
+
+       head = &tr->tr_databuf;
+       while (!list_empty(head)) {
+               bd = list_entry(head->next, struct gfs2_bufdata, bd_list);
+               list_del_init(&bd->bd_list);
+               gfs2_unpin(sdp, bd->bd_bh, tr);
+       }
+}
+
+
+const struct gfs2_log_operations gfs2_buf_lops = {
+       .lo_before_commit = buf_lo_before_commit,
+       .lo_after_commit = buf_lo_after_commit,
+       .lo_before_scan = buf_lo_before_scan,
+       .lo_scan_elements = buf_lo_scan_elements,
+       .lo_after_scan = buf_lo_after_scan,
+       .lo_name = "buf",
+};
+
+const struct gfs2_log_operations gfs2_revoke_lops = {
+       .lo_before_commit = revoke_lo_before_commit,
+       .lo_after_commit = revoke_lo_after_commit,
+       .lo_before_scan = revoke_lo_before_scan,
+       .lo_scan_elements = revoke_lo_scan_elements,
+       .lo_after_scan = revoke_lo_after_scan,
+       .lo_name = "revoke",
+};
+
+const struct gfs2_log_operations gfs2_databuf_lops = {
+       .lo_before_commit = databuf_lo_before_commit,
+       .lo_after_commit = databuf_lo_after_commit,
+       .lo_scan_elements = databuf_lo_scan_elements,
+       .lo_after_scan = databuf_lo_after_scan,
+       .lo_name = "databuf",
+};
+
+const struct gfs2_log_operations *gfs2_log_ops[] = {
+       &gfs2_databuf_lops,
+       &gfs2_buf_lops,
+       &gfs2_revoke_lops,
+       NULL,
+};
+