This commit was manufactured by cvs2svn to create branch 'vserver'.
[linux-2.6.git] / fs / gfs2 / ops_super.c
diff --git a/fs/gfs2/ops_super.c b/fs/gfs2/ops_super.c
new file mode 100644 (file)
index 0000000..485ce3d
--- /dev/null
@@ -0,0 +1,512 @@
+/*
+ * 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/statfs.h>
+#include <linux/seq_file.h>
+#include <linux/mount.h>
+#include <linux/kthread.h>
+#include <linux/delay.h>
+#include <linux/gfs2_ondisk.h>
+#include <linux/crc32.h>
+#include <linux/lm_interface.h>
+
+#include "gfs2.h"
+#include "incore.h"
+#include "glock.h"
+#include "inode.h"
+#include "lm.h"
+#include "log.h"
+#include "mount.h"
+#include "ops_super.h"
+#include "quota.h"
+#include "recovery.h"
+#include "rgrp.h"
+#include "super.h"
+#include "sys.h"
+#include "util.h"
+#include "trans.h"
+#include "dir.h"
+#include "eattr.h"
+#include "bmap.h"
+
+/**
+ * gfs2_write_inode - Make sure the inode is stable on the disk
+ * @inode: The inode
+ * @sync: synchronous write flag
+ *
+ * Returns: errno
+ */
+
+static int gfs2_write_inode(struct inode *inode, int sync)
+{
+       struct gfs2_inode *ip = GFS2_I(inode);
+
+       /* Check this is a "normal" inode */
+       if (inode->i_private) {
+               if (current->flags & PF_MEMALLOC)
+                       return 0;
+               if (sync)
+                       gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
+       }
+
+       return 0;
+}
+
+/**
+ * gfs2_put_super - Unmount the filesystem
+ * @sb: The VFS superblock
+ *
+ */
+
+static void gfs2_put_super(struct super_block *sb)
+{
+       struct gfs2_sbd *sdp = sb->s_fs_info;
+       int error;
+
+       if (!sdp)
+               return;
+
+       if (!strncmp(sb->s_type->name, "gfs2meta", 8))
+               return; /* Nothing to do */
+
+       /*  Unfreeze the filesystem, if we need to  */
+
+       mutex_lock(&sdp->sd_freeze_lock);
+       if (sdp->sd_freeze_count)
+               gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
+       mutex_unlock(&sdp->sd_freeze_lock);
+
+       kthread_stop(sdp->sd_quotad_process);
+       kthread_stop(sdp->sd_logd_process);
+       kthread_stop(sdp->sd_recoverd_process);
+       while (sdp->sd_glockd_num--)
+               kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
+       kthread_stop(sdp->sd_scand_process);
+
+       if (!(sb->s_flags & MS_RDONLY)) {
+               error = gfs2_make_fs_ro(sdp);
+               if (error)
+                       gfs2_io_error(sdp);
+       }
+       /*  At this point, we're through modifying the disk  */
+
+       /*  Release stuff  */
+
+       iput(sdp->sd_master_dir);
+       iput(sdp->sd_jindex);
+       iput(sdp->sd_inum_inode);
+       iput(sdp->sd_statfs_inode);
+       iput(sdp->sd_rindex);
+       iput(sdp->sd_quota_inode);
+
+       gfs2_glock_put(sdp->sd_rename_gl);
+       gfs2_glock_put(sdp->sd_trans_gl);
+
+       if (!sdp->sd_args.ar_spectator) {
+               gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
+               gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
+               gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
+               gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
+               gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
+               iput(sdp->sd_ir_inode);
+               iput(sdp->sd_sc_inode);
+               iput(sdp->sd_qc_inode);
+       }
+
+       gfs2_glock_dq_uninit(&sdp->sd_live_gh);
+       gfs2_clear_rgrpd(sdp);
+       gfs2_jindex_free(sdp);
+       /*  Take apart glock structures and buffer lists  */
+       gfs2_gl_hash_clear(sdp, WAIT);
+       /*  Unmount the locking protocol  */
+       gfs2_lm_unmount(sdp);
+
+       /*  At this point, we're through participating in the lockspace  */
+       gfs2_sys_fs_del(sdp);
+       kfree(sdp);
+}
+
+/**
+ * gfs2_write_super
+ * @sb: the superblock
+ *
+ */
+
+static void gfs2_write_super(struct super_block *sb)
+{
+       sb->s_dirt = 0;
+}
+
+/**
+ * gfs2_sync_fs - sync the filesystem
+ * @sb: the superblock
+ *
+ * Flushes the log to disk.
+ */
+static int gfs2_sync_fs(struct super_block *sb, int wait)
+{
+       sb->s_dirt = 0;
+       if (wait)
+               gfs2_log_flush(sb->s_fs_info, NULL);
+       return 0;
+}
+
+/**
+ * gfs2_write_super_lockfs - prevent further writes to the filesystem
+ * @sb: the VFS structure for the filesystem
+ *
+ */
+
+static void gfs2_write_super_lockfs(struct super_block *sb)
+{
+       struct gfs2_sbd *sdp = sb->s_fs_info;
+       int error;
+
+       if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
+               return;
+
+       for (;;) {
+               error = gfs2_freeze_fs(sdp);
+               if (!error)
+                       break;
+
+               switch (error) {
+               case -EBUSY:
+                       fs_err(sdp, "waiting for recovery before freeze\n");
+                       break;
+
+               default:
+                       fs_err(sdp, "error freezing FS: %d\n", error);
+                       break;
+               }
+
+               fs_err(sdp, "retrying...\n");
+               msleep(1000);
+       }
+}
+
+/**
+ * gfs2_unlockfs - reallow writes to the filesystem
+ * @sb: the VFS structure for the filesystem
+ *
+ */
+
+static void gfs2_unlockfs(struct super_block *sb)
+{
+       gfs2_unfreeze_fs(sb->s_fs_info);
+}
+
+/**
+ * gfs2_statfs - Gather and return stats about the filesystem
+ * @sb: The superblock
+ * @statfsbuf: The buffer
+ *
+ * Returns: 0 on success or error code
+ */
+
+static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
+{
+       struct super_block *sb = dentry->d_inode->i_sb;
+       struct gfs2_sbd *sdp = sb->s_fs_info;
+       struct gfs2_statfs_change_host sc;
+       int error;
+
+       if (gfs2_tune_get(sdp, gt_statfs_slow))
+               error = gfs2_statfs_slow(sdp, &sc);
+       else
+               error = gfs2_statfs_i(sdp, &sc);
+
+       if (error)
+               return error;
+
+       buf->f_type = GFS2_MAGIC;
+       buf->f_bsize = sdp->sd_sb.sb_bsize;
+       buf->f_blocks = sc.sc_total;
+       buf->f_bfree = sc.sc_free;
+       buf->f_bavail = sc.sc_free;
+       buf->f_files = sc.sc_dinodes + sc.sc_free;
+       buf->f_ffree = sc.sc_free;
+       buf->f_namelen = GFS2_FNAMESIZE;
+
+       return 0;
+}
+
+/**
+ * gfs2_remount_fs - called when the FS is remounted
+ * @sb:  the filesystem
+ * @flags:  the remount flags
+ * @data:  extra data passed in (not used right now)
+ *
+ * Returns: errno
+ */
+
+static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
+{
+       struct gfs2_sbd *sdp = sb->s_fs_info;
+       int error;
+
+       error = gfs2_mount_args(sdp, data, 1);
+       if (error)
+               return error;
+
+       if (sdp->sd_args.ar_spectator)
+               *flags |= MS_RDONLY;
+       else {
+               if (*flags & MS_RDONLY) {
+                       if (!(sb->s_flags & MS_RDONLY))
+                               error = gfs2_make_fs_ro(sdp);
+               } else if (!(*flags & MS_RDONLY) &&
+                          (sb->s_flags & MS_RDONLY)) {
+                       error = gfs2_make_fs_rw(sdp);
+               }
+       }
+
+       if (*flags & (MS_NOATIME | MS_NODIRATIME))
+               set_bit(SDF_NOATIME, &sdp->sd_flags);
+       else
+               clear_bit(SDF_NOATIME, &sdp->sd_flags);
+
+       /* Don't let the VFS update atimes.  GFS2 handles this itself. */
+       *flags |= MS_NOATIME | MS_NODIRATIME;
+
+       return error;
+}
+
+/**
+ * gfs2_drop_inode - Drop an inode (test for remote unlink)
+ * @inode: The inode to drop
+ *
+ * If we've received a callback on an iopen lock then its because a
+ * remote node tried to deallocate the inode but failed due to this node
+ * still having the inode open. Here we mark the link count zero
+ * since we know that it must have reached zero if the GLF_DEMOTE flag
+ * is set on the iopen glock. If we didn't do a disk read since the
+ * remote node removed the final link then we might otherwise miss
+ * this event. This check ensures that this node will deallocate the
+ * inode's blocks, or alternatively pass the baton on to another
+ * node for later deallocation.
+ */
+static void gfs2_drop_inode(struct inode *inode)
+{
+       if (inode->i_private && inode->i_nlink) {
+               struct gfs2_inode *ip = GFS2_I(inode);
+               struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
+               if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags))
+                       clear_nlink(inode);
+       }
+       generic_drop_inode(inode);
+}
+
+/**
+ * gfs2_clear_inode - Deallocate an inode when VFS is done with it
+ * @inode: The VFS inode
+ *
+ */
+
+static void gfs2_clear_inode(struct inode *inode)
+{
+       /* This tells us its a "real" inode and not one which only
+        * serves to contain an address space (see rgrp.c, meta_io.c)
+        * which therefore doesn't have its own glocks.
+        */
+       if (inode->i_private) {
+               struct gfs2_inode *ip = GFS2_I(inode);
+               ip->i_gl->gl_object = NULL;
+               gfs2_glock_schedule_for_reclaim(ip->i_gl);
+               gfs2_glock_put(ip->i_gl);
+               ip->i_gl = NULL;
+               if (ip->i_iopen_gh.gh_gl)
+                       gfs2_glock_dq_uninit(&ip->i_iopen_gh);
+       }
+}
+
+/**
+ * gfs2_show_options - Show mount options for /proc/mounts
+ * @s: seq_file structure
+ * @mnt: vfsmount
+ *
+ * Returns: 0 on success or error code
+ */
+
+static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
+{
+       struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info;
+       struct gfs2_args *args = &sdp->sd_args;
+
+       if (args->ar_lockproto[0])
+               seq_printf(s, ",lockproto=%s", args->ar_lockproto);
+       if (args->ar_locktable[0])
+               seq_printf(s, ",locktable=%s", args->ar_locktable);
+       if (args->ar_hostdata[0])
+               seq_printf(s, ",hostdata=%s", args->ar_hostdata);
+       if (args->ar_spectator)
+               seq_printf(s, ",spectator");
+       if (args->ar_ignore_local_fs)
+               seq_printf(s, ",ignore_local_fs");
+       if (args->ar_localflocks)
+               seq_printf(s, ",localflocks");
+       if (args->ar_localcaching)
+               seq_printf(s, ",localcaching");
+       if (args->ar_debug)
+               seq_printf(s, ",debug");
+       if (args->ar_upgrade)
+               seq_printf(s, ",upgrade");
+       if (args->ar_num_glockd != GFS2_GLOCKD_DEFAULT)
+               seq_printf(s, ",num_glockd=%u", args->ar_num_glockd);
+       if (args->ar_posix_acl)
+               seq_printf(s, ",acl");
+       if (args->ar_quota != GFS2_QUOTA_DEFAULT) {
+               char *state;
+               switch (args->ar_quota) {
+               case GFS2_QUOTA_OFF:
+                       state = "off";
+                       break;
+               case GFS2_QUOTA_ACCOUNT:
+                       state = "account";
+                       break;
+               case GFS2_QUOTA_ON:
+                       state = "on";
+                       break;
+               default:
+                       state = "unknown";
+                       break;
+               }
+               seq_printf(s, ",quota=%s", state);
+       }
+       if (args->ar_suiddir)
+               seq_printf(s, ",suiddir");
+       if (args->ar_data != GFS2_DATA_DEFAULT) {
+               char *state;
+               switch (args->ar_data) {
+               case GFS2_DATA_WRITEBACK:
+                       state = "writeback";
+                       break;
+               case GFS2_DATA_ORDERED:
+                       state = "ordered";
+                       break;
+               default:
+                       state = "unknown";
+                       break;
+               }
+               seq_printf(s, ",data=%s", state);
+       }
+
+       return 0;
+}
+
+/*
+ * We have to (at the moment) hold the inodes main lock to cover
+ * the gap between unlocking the shared lock on the iopen lock and
+ * taking the exclusive lock. I'd rather do a shared -> exclusive
+ * conversion on the iopen lock, but we can change that later. This
+ * is safe, just less efficient.
+ */
+static void gfs2_delete_inode(struct inode *inode)
+{
+       struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
+       struct gfs2_inode *ip = GFS2_I(inode);
+       struct gfs2_holder gh;
+       int error;
+
+       if (!inode->i_private)
+               goto out;
+
+       error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB, &gh);
+       if (unlikely(error)) {
+               gfs2_glock_dq_uninit(&ip->i_iopen_gh);
+               goto out;
+       }
+
+       gfs2_glock_dq(&ip->i_iopen_gh);
+       gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh);
+       error = gfs2_glock_nq(&ip->i_iopen_gh);
+       if (error)
+               goto out_uninit;
+
+       if (S_ISDIR(inode->i_mode) &&
+           (ip->i_di.di_flags & GFS2_DIF_EXHASH)) {
+               error = gfs2_dir_exhash_dealloc(ip);
+               if (error)
+                       goto out_unlock;
+       }
+
+       if (ip->i_di.di_eattr) {
+               error = gfs2_ea_dealloc(ip);
+               if (error)
+                       goto out_unlock;
+       }
+
+       if (!gfs2_is_stuffed(ip)) {
+               error = gfs2_file_dealloc(ip);
+               if (error)
+                       goto out_unlock;
+       }
+
+       error = gfs2_dinode_dealloc(ip);
+       /*
+        * Must do this before unlock to avoid trying to write back
+        * potentially dirty data now that inode no longer exists
+        * on disk.
+        */
+       truncate_inode_pages(&inode->i_data, 0);
+
+out_unlock:
+       gfs2_glock_dq(&ip->i_iopen_gh);
+out_uninit:
+       gfs2_holder_uninit(&ip->i_iopen_gh);
+       gfs2_glock_dq_uninit(&gh);
+       if (error && error != GLR_TRYFAILED)
+               fs_warn(sdp, "gfs2_delete_inode: %d\n", error);
+out:
+       truncate_inode_pages(&inode->i_data, 0);
+       clear_inode(inode);
+}
+
+
+
+static struct inode *gfs2_alloc_inode(struct super_block *sb)
+{
+       struct gfs2_inode *ip;
+
+       ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL);
+       if (ip) {
+               ip->i_flags = 0;
+               ip->i_gl = NULL;
+               ip->i_last_pfault = jiffies;
+       }
+       return &ip->i_inode;
+}
+
+static void gfs2_destroy_inode(struct inode *inode)
+{
+       kmem_cache_free(gfs2_inode_cachep, inode);
+}
+
+const struct super_operations gfs2_super_ops = {
+       .alloc_inode            = gfs2_alloc_inode,
+       .destroy_inode          = gfs2_destroy_inode,
+       .write_inode            = gfs2_write_inode,
+       .delete_inode           = gfs2_delete_inode,
+       .put_super              = gfs2_put_super,
+       .write_super            = gfs2_write_super,
+       .sync_fs                = gfs2_sync_fs,
+       .write_super_lockfs     = gfs2_write_super_lockfs,
+       .unlockfs               = gfs2_unlockfs,
+       .statfs                 = gfs2_statfs,
+       .remount_fs             = gfs2_remount_fs,
+       .clear_inode            = gfs2_clear_inode,
+       .drop_inode             = gfs2_drop_inode,
+       .show_options           = gfs2_show_options,
+};
+