Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / gfs2 / trans.c
diff --git a/kernel/fs/gfs2/trans.c b/kernel/fs/gfs2/trans.c
new file mode 100644 (file)
index 0000000..88bff24
--- /dev/null
@@ -0,0 +1,279 @@
+/*
+ * 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.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/completion.h>
+#include <linux/buffer_head.h>
+#include <linux/kallsyms.h>
+#include <linux/gfs2_ondisk.h>
+
+#include "gfs2.h"
+#include "incore.h"
+#include "glock.h"
+#include "inode.h"
+#include "log.h"
+#include "lops.h"
+#include "meta_io.h"
+#include "trans.h"
+#include "util.h"
+#include "trace_gfs2.h"
+
+int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
+                    unsigned int revokes)
+{
+       struct gfs2_trans *tr;
+       int error;
+
+       BUG_ON(current->journal_info);
+       BUG_ON(blocks == 0 && revokes == 0);
+
+       if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
+               return -EROFS;
+
+       tr = kzalloc(sizeof(struct gfs2_trans), GFP_NOFS);
+       if (!tr)
+               return -ENOMEM;
+
+       tr->tr_ip = _RET_IP_;
+       tr->tr_blocks = blocks;
+       tr->tr_revokes = revokes;
+       tr->tr_reserved = 1;
+       tr->tr_alloced = 1;
+       if (blocks)
+               tr->tr_reserved += 6 + blocks;
+       if (revokes)
+               tr->tr_reserved += gfs2_struct2blk(sdp, revokes,
+                                                  sizeof(u64));
+       INIT_LIST_HEAD(&tr->tr_databuf);
+       INIT_LIST_HEAD(&tr->tr_buf);
+
+       sb_start_intwrite(sdp->sd_vfs);
+
+       error = gfs2_log_reserve(sdp, tr->tr_reserved);
+       if (error)
+               goto fail;
+
+       current->journal_info = tr;
+
+       return 0;
+
+fail:
+       sb_end_intwrite(sdp->sd_vfs);
+       kfree(tr);
+
+       return error;
+}
+
+static void gfs2_print_trans(const struct gfs2_trans *tr)
+{
+       pr_warn("Transaction created at: %pSR\n", (void *)tr->tr_ip);
+       pr_warn("blocks=%u revokes=%u reserved=%u touched=%u\n",
+               tr->tr_blocks, tr->tr_revokes, tr->tr_reserved, tr->tr_touched);
+       pr_warn("Buf %u/%u Databuf %u/%u Revoke %u/%u\n",
+               tr->tr_num_buf_new, tr->tr_num_buf_rm,
+               tr->tr_num_databuf_new, tr->tr_num_databuf_rm,
+               tr->tr_num_revoke, tr->tr_num_revoke_rm);
+}
+
+void gfs2_trans_end(struct gfs2_sbd *sdp)
+{
+       struct gfs2_trans *tr = current->journal_info;
+       s64 nbuf;
+       int alloced = tr->tr_alloced;
+
+       BUG_ON(!tr);
+       current->journal_info = NULL;
+
+       if (!tr->tr_touched) {
+               gfs2_log_release(sdp, tr->tr_reserved);
+               if (alloced) {
+                       kfree(tr);
+                       sb_end_intwrite(sdp->sd_vfs);
+               }
+               return;
+       }
+
+       nbuf = tr->tr_num_buf_new + tr->tr_num_databuf_new;
+       nbuf -= tr->tr_num_buf_rm;
+       nbuf -= tr->tr_num_databuf_rm;
+
+       if (gfs2_assert_withdraw(sdp, (nbuf <= tr->tr_blocks) &&
+                                      (tr->tr_num_revoke <= tr->tr_revokes)))
+               gfs2_print_trans(tr);
+
+       gfs2_log_commit(sdp, tr);
+       if (alloced && !tr->tr_attached)
+                       kfree(tr);
+       up_read(&sdp->sd_log_flush_lock);
+
+       if (sdp->sd_vfs->s_flags & MS_SYNCHRONOUS)
+               gfs2_log_flush(sdp, NULL, NORMAL_FLUSH);
+       if (alloced)
+               sb_end_intwrite(sdp->sd_vfs);
+}
+
+static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
+                                              struct buffer_head *bh,
+                                              const struct gfs2_log_operations *lops)
+{
+       struct gfs2_bufdata *bd;
+
+       bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
+       bd->bd_bh = bh;
+       bd->bd_gl = gl;
+       bd->bd_ops = lops;
+       INIT_LIST_HEAD(&bd->bd_list);
+       bh->b_private = bd;
+       return bd;
+}
+
+/**
+ * gfs2_trans_add_data - Add a databuf to the transaction.
+ * @gl: The inode glock associated with the buffer
+ * @bh: The buffer to add
+ *
+ * This is used in two distinct cases:
+ * i) In ordered write mode
+ *    We put the data buffer on a list so that we can ensure that its
+ *    synced to disk at the right time
+ * ii) In journaled data mode
+ *    We need to journal the data block in the same way as metadata in
+ *    the functions above. The difference is that here we have a tag
+ *    which is two __be64's being the block number (as per meta data)
+ *    and a flag which says whether the data block needs escaping or
+ *    not. This means we need a new log entry for each 251 or so data
+ *    blocks, which isn't an enormous overhead but twice as much as
+ *    for normal metadata blocks.
+ */
+void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
+{
+       struct gfs2_trans *tr = current->journal_info;
+       struct gfs2_sbd *sdp = gl->gl_sbd;
+       struct address_space *mapping = bh->b_page->mapping;
+       struct gfs2_inode *ip = GFS2_I(mapping->host);
+       struct gfs2_bufdata *bd;
+
+       if (!gfs2_is_jdata(ip)) {
+               gfs2_ordered_add_inode(ip);
+               return;
+       }
+
+       lock_buffer(bh);
+       gfs2_log_lock(sdp);
+       bd = bh->b_private;
+       if (bd == NULL) {
+               gfs2_log_unlock(sdp);
+               unlock_buffer(bh);
+               if (bh->b_private == NULL)
+                       bd = gfs2_alloc_bufdata(gl, bh, &gfs2_databuf_lops);
+               lock_buffer(bh);
+               gfs2_log_lock(sdp);
+       }
+       gfs2_assert(sdp, bd->bd_gl == gl);
+       tr->tr_touched = 1;
+       if (list_empty(&bd->bd_list)) {
+               set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+               set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
+               gfs2_pin(sdp, bd->bd_bh);
+               tr->tr_num_databuf_new++;
+               list_add_tail(&bd->bd_list, &tr->tr_databuf);
+       }
+       gfs2_log_unlock(sdp);
+       unlock_buffer(bh);
+}
+
+static void meta_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
+{
+       struct gfs2_meta_header *mh;
+       struct gfs2_trans *tr;
+       enum gfs2_freeze_state state = atomic_read(&sdp->sd_freeze_state);
+
+       tr = current->journal_info;
+       tr->tr_touched = 1;
+       if (!list_empty(&bd->bd_list))
+               return;
+       set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+       set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
+       mh = (struct gfs2_meta_header *)bd->bd_bh->b_data;
+       if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) {
+               pr_err("Attempting to add uninitialised block to journal (inplace block=%lld)\n",
+                      (unsigned long long)bd->bd_bh->b_blocknr);
+               BUG();
+       }
+       if (unlikely(state == SFS_FROZEN)) {
+               printk(KERN_INFO "GFS2:adding buf while frozen\n");
+               gfs2_assert_withdraw(sdp, 0);
+       }
+       gfs2_pin(sdp, bd->bd_bh);
+       mh->__pad0 = cpu_to_be64(0);
+       mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
+       list_add(&bd->bd_list, &tr->tr_buf);
+       tr->tr_num_buf_new++;
+}
+
+void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh)
+{
+
+       struct gfs2_sbd *sdp = gl->gl_sbd;
+       struct gfs2_bufdata *bd;
+
+       lock_buffer(bh);
+       gfs2_log_lock(sdp);
+       bd = bh->b_private;
+       if (bd == NULL) {
+               gfs2_log_unlock(sdp);
+               unlock_buffer(bh);
+               lock_page(bh->b_page);
+               if (bh->b_private == NULL)
+                       bd = gfs2_alloc_bufdata(gl, bh, &gfs2_buf_lops);
+               unlock_page(bh->b_page);
+               lock_buffer(bh);
+               gfs2_log_lock(sdp);
+       }
+       gfs2_assert(sdp, bd->bd_gl == gl);
+       meta_lo_add(sdp, bd);
+       gfs2_log_unlock(sdp);
+       unlock_buffer(bh);
+}
+
+void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
+{
+       struct gfs2_trans *tr = current->journal_info;
+
+       BUG_ON(!list_empty(&bd->bd_list));
+       gfs2_add_revoke(sdp, bd);
+       tr->tr_touched = 1;
+       tr->tr_num_revoke++;
+}
+
+void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
+{
+       struct gfs2_bufdata *bd, *tmp;
+       struct gfs2_trans *tr = current->journal_info;
+       unsigned int n = len;
+
+       gfs2_log_lock(sdp);
+       list_for_each_entry_safe(bd, tmp, &sdp->sd_log_le_revoke, bd_list) {
+               if ((bd->bd_blkno >= blkno) && (bd->bd_blkno < (blkno + len))) {
+                       list_del_init(&bd->bd_list);
+                       gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke);
+                       sdp->sd_log_num_revoke--;
+                       kmem_cache_free(gfs2_bufdata_cachep, bd);
+                       tr->tr_num_revoke_rm++;
+                       if (--n == 0)
+                               break;
+               }
+       }
+       gfs2_log_unlock(sdp);
+}
+