inode.c

ocfs1.2.7 源码

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * inode.c
 *
 * vfs' aops, fops, dops and iops
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/fs.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <linux/smp_lock.h>

#include <asm/byteorder.h>

#define MLOG_MASK_PREFIX ML_INODE
#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
#include "dlmglue.h"
#include "extent_map.h"
#include "file.h"
#include  "heartbeat.h"
#include "inode.h"
#include "journal.h"
#include "namei.h"
#include "suballoc.h"
#include "super.h"
#include "symlink.h"
#include "sysfile.h"
#include "uptodate.h"
#include "vote.h"

#include "buffer_head_io.h"

#define OCFS2_FI_FLAG_NOWAIT	0x1
#define OCFS2_FI_FLAG_DELETE	0x2
struct ocfs2_find_inode_args
{
	u64		fi_blkno;
	unsigned long	fi_ino;
	unsigned int	fi_flags;
};

static int ocfs2_read_locked_inode(struct inode *inode,
				   struct ocfs2_find_inode_args *args);
static int ocfs2_init_locked_inode(struct inode *inode, void *opaque);
static int ocfs2_find_actor(struct inode *inode, void *opaque);
static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
				    struct inode *inode,
				    struct buffer_head *fe_bh);

struct inode *ocfs2_ilookup_for_vote(struct ocfs2_super *osb,
				     u64 blkno,
				     int delete_vote)
{
	struct ocfs2_find_inode_args args;

	/* ocfs2_ilookup_for_vote should *only* be called from the
	 * vote thread */
	BUG_ON(current != osb->vote_task);

	args.fi_blkno = blkno;
	args.fi_flags = OCFS2_FI_FLAG_NOWAIT;
	if (delete_vote)
		args.fi_flags |= OCFS2_FI_FLAG_DELETE;
	args.fi_ino = ino_from_blkno(osb->sb, blkno);
	return ilookup5(osb->sb, args.fi_ino, ocfs2_find_actor, &args);
}

struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno)
{
	struct inode *inode = NULL;
	struct super_block *sb = osb->sb;
	struct ocfs2_find_inode_args args;

	mlog_entry("(blkno = %"MLFu64")\n", blkno);

	/* Ok. By now we've either got the offsets passed to us by the
	 * caller, or we just pulled them off the bh. Lets do some
	 * sanity checks to make sure they're OK. */
	if (blkno == 0) {
		inode = ERR_PTR(-EINVAL);
		mlog_errno(PTR_ERR(inode));
		goto bail;
	}

	args.fi_blkno = blkno;
	args.fi_flags = 0;
	args.fi_ino = ino_from_blkno(sb, blkno);

	inode = iget5_locked(sb, args.fi_ino, ocfs2_find_actor,
			     ocfs2_init_locked_inode, &args);
	/* inode was *not* in the inode cache. 2.6.x requires
	 * us to do our own read_inode call and unlock it
	 * afterwards. */
	if (inode && inode->i_state & I_NEW) {
		mlog(0, "Inode was not in inode cache, reading it.\n");
		ocfs2_read_locked_inode(inode, &args);
		unlock_new_inode(inode);
	}
	if (inode == NULL) {
		inode = ERR_PTR(-ENOMEM);
		mlog_errno(PTR_ERR(inode));
		goto bail;
	}
	if (is_bad_inode(inode)) {
		iput(inode);
		inode = ERR_PTR(-ESTALE);
		goto bail;
	}

bail:
	if (!IS_ERR(inode)) {
		mlog(0, "returning inode with number %"MLFu64"\n",
		     OCFS2_I(inode)->ip_blkno);
		mlog_exit_ptr(inode);
	}

	return inode;
}


/*
 * here's how inodes get read from disk:
 * iget5_locked -> find_actor -> OCFS2_FIND_ACTOR
 * found? : return the in-memory inode
 * not found? : get_new_inode -> OCFS2_INIT_LOCKED_INODE
 */

static int ocfs2_find_actor(struct inode *inode, void *opaque)
{
	struct ocfs2_find_inode_args *args = NULL;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	int ret = 0;

	mlog_entry("(0x%p, %lu, 0x%p)\n", inode, inode->i_ino, opaque);

	args = opaque;

	mlog_bug_on_msg(!inode, "No inode in find actor!\n");

	if (oi->ip_blkno != args->fi_blkno)
		goto bail;

	/* OCFS2_FI_FLAG_NOWAIT is *only* set from
	 * ocfs2_ilookup_for_vote which won't create an inode for one
	 * that isn't found. The vote thread which doesn't want to get
	 * an inode which is in the process of going away - otherwise
	 * the call to __wait_on_freeing_inode in find_inode_fast will
	 * cause it to deadlock on an inode which may be waiting on a
	 * vote (or lock release) in delete_inode */
	if ((args->fi_flags & OCFS2_FI_FLAG_NOWAIT) &&
	    (inode->i_state & (I_FREEING|I_CLEAR))) {
		/* As stated above, we're not going to return an
		 * inode.  In the case of a delete vote, the voting
		 * code is going to signal the other node to go
		 * ahead. Mark that state here, so this freeing inode
		 * has the state when it gets to delete_inode. */
		if (args->fi_flags & OCFS2_FI_FLAG_DELETE) {
			spin_lock(&oi->ip_lock);
			ocfs2_mark_inode_remotely_deleted(inode,
							O2NM_INVALID_NODE_NUM);
			spin_unlock(&oi->ip_lock);
		}
		goto bail;
	}

	ret = 1;
bail:
	mlog_exit(ret);
	return ret;
}

/*
 * initialize the new inode, but don't do anything that would cause
 * us to sleep.
 * return 0 on success, 1 on failure
 */
static int ocfs2_init_locked_inode(struct inode *inode, void *opaque)
{
	struct ocfs2_find_inode_args *args = opaque;

	mlog_entry("inode = %p, opaque = %p\n", inode, opaque);

	inode->i_ino = args->fi_ino;
	OCFS2_I(inode)->ip_blkno = args->fi_blkno;

	mlog_exit(0);
	return 0;
}

int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
		     	 int create_ino)
{
	struct super_block *sb;
	struct ocfs2_super *osb;
	int status = -EINVAL;

	mlog_entry("(0x%p, size:%"MLFu64")\n", inode, fe->i_size);

	sb = inode->i_sb;
	osb = OCFS2_SB(sb);

	/* this means that read_inode cannot create a superblock inode
	 * today.  change if needed. */
	if (!OCFS2_IS_VALID_DINODE(fe) ||
	    !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
		mlog(0, "Invalid dinode: i_ino=%lu, i_blkno=%"MLFu64", "
		     "signature = %.*s, flags = 0x%x\n",
		     inode->i_ino, le64_to_cpu(fe->i_blkno), 7,
		     fe->i_signature, le32_to_cpu(fe->i_flags));
		goto bail;
	}

	if (le32_to_cpu(fe->i_fs_generation) != osb->fs_generation) {
		mlog(ML_ERROR, "file entry generation does not match "
		     "superblock! osb->fs_generation=%x, "
		     "fe->i_fs_generation=%x\n",
		     osb->fs_generation, le32_to_cpu(fe->i_fs_generation));
		goto bail;
	}

	inode->i_version = 1;
	inode->i_generation = le32_to_cpu(fe->i_generation);
	inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
	inode->i_mode = le16_to_cpu(fe->i_mode);
	inode->i_uid = le32_to_cpu(fe->i_uid);
	inode->i_gid = le32_to_cpu(fe->i_gid);
#ifdef INODE_HAS_BLKSIZE
	inode->i_blksize = (u32)osb->s_clustersize;
#endif

	/* Fast symlinks will have i_size but no allocated clusters. */
	if (S_ISLNK(inode->i_mode) && !fe->i_clusters)
		inode->i_blocks = 0;
	else
		inode->i_blocks =
			ocfs2_align_bytes_to_sectors(le64_to_cpu(fe->i_size));
	inode->i_mapping->a_ops = &ocfs2_aops;
	inode->i_flags |= S_NOATIME;
	inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
	inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
	inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
	inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec);
	inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime);
	inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec);

	if (OCFS2_I(inode)->ip_blkno != le64_to_cpu(fe->i_blkno))
		mlog(ML_ERROR,
		     "ip_blkno %"MLFu64" != i_blkno %"MLFu64"!\n",
		     OCFS2_I(inode)->ip_blkno, fe->i_blkno);

	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
	OCFS2_I(inode)->ip_orphaned_slot = OCFS2_INVALID_SLOT;
	OCFS2_I(inode)->ip_deleting_node = O2NM_INVALID_NODE_NUM;

	if (create_ino)
		inode->i_ino = ino_from_blkno(inode->i_sb,
			       le64_to_cpu(fe->i_blkno));

	mlog(0, "blkno = %"MLFu64", ino = %lu, create_ino = %s\n",
	     fe->i_blkno, inode->i_ino, create_ino ? "true" : "false");

	inode->i_nlink = le16_to_cpu(fe->i_links_count);

	if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) {
		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
		mlog(0, "local alloc inode: i_ino=%lu\n", inode->i_ino);
	} else if (fe->i_flags & cpu_to_le32(OCFS2_BITMAP_FL)) {
		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
	} else if (fe->i_flags & cpu_to_le32(OCFS2_SUPER_BLOCK_FL)) {
		mlog(0, "superblock inode: i_ino=%lu\n", inode->i_ino);
		/* we can't actually hit this as read_inode can't
		 * handle superblocks today ;-) */
		BUG();
	}

	switch (inode->i_mode & S_IFMT) {
	    case S_IFREG:
		    inode->i_fop = &ocfs2_fops;
		    inode->i_op = &ocfs2_file_iops;
		    i_size_write(inode, le64_to_cpu(fe->i_size));
		    OCFS2_I(inode)->ip_mmu_private = inode->i_size;
		    break;
	    case S_IFDIR:
		    inode->i_op = &ocfs2_dir_iops;
		    inode->i_fop = &ocfs2_dops;
		    i_size_write(inode, le64_to_cpu(fe->i_size));
		    break;
	    case S_IFLNK:
		    if (ocfs2_inode_is_fast_symlink(inode))
			inode->i_op = &ocfs2_fast_symlink_inode_operations;
		    else
			inode->i_op = &ocfs2_symlink_inode_operations;
		    i_size_write(inode, le64_to_cpu(fe->i_size));
		    break;
	    default:
		    inode->i_op = &ocfs2_special_file_iops;
		    init_special_inode(inode, inode->i_mode,
				       inode->i_rdev);
		    break;
	}

	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres,
				  OCFS2_LOCK_TYPE_META, inode);
	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_data_lockres,
				  OCFS2_LOCK_TYPE_DATA, inode);

	status = 0;
bail:
	mlog_exit(status);
	return status;
}

static int ocfs2_read_locked_inode(struct inode *inode,
				   struct ocfs2_find_inode_args *args)
{
	struct super_block *sb;
	struct ocfs2_super *osb;
	struct ocfs2_dinode *fe;
	struct buffer_head *bh = NULL;
	int status;
	int sysfile = 0;

	mlog_entry("(0x%p, 0x%p)\n", inode, args);

	status = -EINVAL;
	if (inode == NULL || inode->i_sb == NULL) {
		mlog(ML_ERROR, "bad inode\n");
		goto bail;
	}
	sb = inode->i_sb;
	osb = OCFS2_SB(sb);

	if (!args) {
		mlog(ML_ERROR, "bad inode args\n");
		make_bad_inode(inode);
		goto bail;
	}

	/* Read the FE off disk. This is safe because the kernel only
	 * does one read_inode2 for a new inode, and if it doesn't
	 * exist yet then nobody can be working on it! */
	status = ocfs2_read_block(osb, args->fi_blkno, &bh, 0, NULL);
	if (status < 0) {
		mlog_errno(status);
		make_bad_inode(inode);
		goto bail;
	}

	fe = (struct ocfs2_dinode *) bh->b_data;
	if (!OCFS2_IS_VALID_DINODE(fe)) {
		mlog(ML_ERROR, "Invalid dinode #%"MLFu64": signature = %.*s\n",
		     fe->i_blkno, 7, fe->i_signature);
		make_bad_inode(inode);
		goto bail;
	}

	if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL))
		sysfile = 1;

	if (S_ISCHR(le16_to_cpu(fe->i_mode)) ||
	    S_ISBLK(le16_to_cpu(fe->i_mode)))
    		inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));

	status = -EINVAL;
	if (ocfs2_populate_inode(inode, fe, 0) < 0) {
		make_bad_inode(inode);
		goto bail;
	}

	BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));