xfs_iget.c

来自「Linux Kernel 2.6.9 for OMAP1710」· C语言 代码 · 共 994 行 · 第 1/2 页

/*
 * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 * Further, this software is distributed without any warranty that it is
 * free of the rightful claim of any third person regarding infringement
 * or the like.  Any license provided herein, whether implied or
 * otherwise, applies only to this software file.  Patent licenses, if
 * any, provided herein do not apply to combinations of this program with
 * other software, or any other product whatsoever.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write the Free Software Foundation, Inc., 59
 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
 *
 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
 * Mountain View, CA  94043, or:
 *
 * http://www.sgi.com
 *
 * For further information regarding this notice, see:
 *
 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
 */

#include "xfs.h"

#include "xfs_macros.h"
#include "xfs_types.h"
#include "xfs_inum.h"
#include "xfs_log.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_alloc_btree.h"
#include "xfs_bmap_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_btree.h"
#include "xfs_ialloc.h"
#include "xfs_attr_sf.h"
#include "xfs_dir_sf.h"
#include "xfs_dir2_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_quota.h"
#include "xfs_utils.h"

/*
 * Initialize the inode hash table for the newly mounted file system.
 *
 * mp -- this is the mount point structure for the file system being
 *       initialized
 */
void
xfs_ihash_init(xfs_mount_t *mp)
{
	int	i;

	mp->m_ihsize = XFS_BUCKETS(mp);
	mp->m_ihash = (xfs_ihash_t *)kmem_zalloc(mp->m_ihsize
				      * sizeof(xfs_ihash_t), KM_SLEEP);
	ASSERT(mp->m_ihash != NULL);
	for (i = 0; i < mp->m_ihsize; i++) {
		rwlock_init(&(mp->m_ihash[i].ih_lock));
	}
}

/*
 * Free up structures allocated by xfs_ihash_init, at unmount time.
 */
void
xfs_ihash_free(xfs_mount_t *mp)
{
	kmem_free(mp->m_ihash, mp->m_ihsize*sizeof(xfs_ihash_t));
	mp->m_ihash = NULL;
}

/*
 * Initialize the inode cluster hash table for the newly mounted file system.
 *
 * mp -- this is the mount point structure for the file system being
 *       initialized
 */
void
xfs_chash_init(xfs_mount_t *mp)
{
	int	i;

	/*
	 * m_chash size is based on m_ihash
	 * with a minimum of 37 entries
	 */
	mp->m_chsize = (XFS_BUCKETS(mp)) /
			 (XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog);
	if (mp->m_chsize < 37) {
		mp->m_chsize = 37;
	}
	mp->m_chash = (xfs_chash_t *)kmem_zalloc(mp->m_chsize
						 * sizeof(xfs_chash_t),
						 KM_SLEEP);
	ASSERT(mp->m_chash != NULL);

	for (i = 0; i < mp->m_chsize; i++) {
		spinlock_init(&mp->m_chash[i].ch_lock,"xfshash");
	}
}

/*
 * Free up structures allocated by xfs_chash_init, at unmount time.
 */
void
xfs_chash_free(xfs_mount_t *mp)
{
	int	i;

	for (i = 0; i < mp->m_chsize; i++) {
		spinlock_destroy(&mp->m_chash[i].ch_lock);
	}

	kmem_free(mp->m_chash, mp->m_chsize*sizeof(xfs_chash_t));
	mp->m_chash = NULL;
}

/*
 * Look up an inode by number in the given file system.
 * The inode is looked up in the hash table for the file system
 * represented by the mount point parameter mp.  Each bucket of
 * the hash table is guarded by an individual semaphore.
 *
 * If the inode is found in the hash table, its corresponding vnode
 * is obtained with a call to vn_get().  This call takes care of
 * coordination with the reclamation of the inode and vnode.  Note
 * that the vmap structure is filled in while holding the hash lock.
 * This gives us the state of the inode/vnode when we found it and
 * is used for coordination in vn_get().
 *
 * If it is not in core, read it in from the file system's device and
 * add the inode into the hash table.
 *
 * The inode is locked according to the value of the lock_flags parameter.
 * This flag parameter indicates how and if the inode's IO lock and inode lock
 * should be taken.
 *
 * mp -- the mount point structure for the current file system.  It points
 *       to the inode hash table.
 * tp -- a pointer to the current transaction if there is one.  This is
 *       simply passed through to the xfs_iread() call.
 * ino -- the number of the inode desired.  This is the unique identifier
 *        within the file system for the inode being requested.
 * lock_flags -- flags indicating how to lock the inode.  See the comment
 *		 for xfs_ilock() for a list of valid values.
 * bno -- the block number starting the buffer containing the inode,
 *	  if known (as by bulkstat), else 0.
 */
STATIC int
xfs_iget_core(
	vnode_t		*vp,
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
	xfs_ino_t	ino,
	uint		lock_flags,
	xfs_inode_t	**ipp,
	xfs_daddr_t	bno)
{
	xfs_ihash_t	*ih;
	xfs_inode_t	*ip;
	xfs_inode_t	*iq;
	vnode_t		*inode_vp;
	ulong		version;
	int		error;
	/* REFERENCED */
	int		newnode;
	xfs_chash_t	*ch;
	xfs_chashlist_t	*chl, *chlnew;
	SPLDECL(s);


	ih = XFS_IHASH(mp, ino);

again:
	read_lock(&ih->ih_lock);

	for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) {
		if (ip->i_ino == ino) {

			inode_vp = XFS_ITOV_NULL(ip);

			if (inode_vp == NULL) {
				/* If IRECLAIM is set this inode is
				 * on its way out of the system,
				 * we need to pause and try again.
				 */
				if (ip->i_flags & XFS_IRECLAIM) {
					read_unlock(&ih->ih_lock);
					delay(1);
					XFS_STATS_INC(xs_ig_frecycle);

					goto again;
				}

				vn_trace_exit(vp, "xfs_iget.alloc",
					(inst_t *)__return_address);

				XFS_STATS_INC(xs_ig_found);

				ip->i_flags &= ~XFS_IRECLAIMABLE;
				read_unlock(&ih->ih_lock);

				XFS_MOUNT_ILOCK(mp);
				list_del_init(&ip->i_reclaim);
				XFS_MOUNT_IUNLOCK(mp);

				goto finish_inode;

			} else if (vp != inode_vp) {
				struct inode *inode = LINVFS_GET_IP(inode_vp);

				/* The inode is being torn down, pause and
				 * try again.
				 */
				if (inode->i_state & (I_FREEING | I_CLEAR)) {
					read_unlock(&ih->ih_lock);
					delay(1);
					XFS_STATS_INC(xs_ig_frecycle);

					goto again;
				}
/* Chances are the other vnode (the one in the inode) is being torn
 * down right now, and we landed on top of it. Question is, what do
 * we do? Unhook the old inode and hook up the new one?
 */
				cmn_err(CE_PANIC,
			"xfs_iget_core: ambiguous vns: vp/0x%p, invp/0x%p",
						inode_vp, vp);
			}

			read_unlock(&ih->ih_lock);

			XFS_STATS_INC(xs_ig_found);

finish_inode:
			if (lock_flags != 0) {
				xfs_ilock(ip, lock_flags);
			}

			newnode = (ip->i_d.di_mode == 0);
			if (newnode) {
				xfs_iocore_inode_reinit(ip);
			}
			ip->i_flags &= ~XFS_ISTALE;

			vn_trace_exit(vp, "xfs_iget.found",
						(inst_t *)__return_address);
			goto return_ip;
		}
	}

	/*
	 * Inode cache miss: save the hash chain version stamp and unlock
	 * the chain, so we don't deadlock in vn_alloc.
	 */
	XFS_STATS_INC(xs_ig_missed);

	version = ih->ih_version;

	read_unlock(&ih->ih_lock);

	/*
	 * Read the disk inode attributes into a new inode structure and get
	 * a new vnode for it. This should also initialize i_ino and i_mount.
	 */
	error = xfs_iread(mp, tp, ino, &ip, bno);
	if (error) {
		return error;
	}

	vn_trace_exit(vp, "xfs_iget.alloc", (inst_t *)__return_address);

	xfs_inode_lock_init(ip, vp);
	xfs_iocore_inode_init(ip);

	if (lock_flags != 0) {
		xfs_ilock(ip, lock_flags);
	}

	/*
	 * Put ip on its hash chain, unless someone else hashed a duplicate
	 * after we released the hash lock.
	 */
	write_lock(&ih->ih_lock);

	if (ih->ih_version != version) {
		for (iq = ih->ih_next; iq != NULL; iq = iq->i_next) {
			if (iq->i_ino == ino) {
				write_unlock(&ih->ih_lock);
				xfs_idestroy(ip);

				XFS_STATS_INC(xs_ig_dup);
				goto again;
			}
		}
	}

	/*
	 * These values _must_ be set before releasing ihlock!
	 */
	ip->i_hash = ih;
	if ((iq = ih->ih_next)) {
		iq->i_prevp = &ip->i_next;
	}
	ip->i_next = iq;
	ip->i_prevp = &ih->ih_next;
	ih->ih_next = ip;
	ip->i_udquot = ip->i_gdquot = NULL;
	ih->ih_version++;

	write_unlock(&ih->ih_lock);

	/*
	 * put ip on its cluster's hash chain
	 */
	ASSERT(ip->i_chash == NULL && ip->i_cprev == NULL &&
	       ip->i_cnext == NULL);

	chlnew = NULL;
	ch = XFS_CHASH(mp, ip->i_blkno);
 chlredo:
	s = mutex_spinlock(&ch->ch_lock);
	for (chl = ch->ch_list; chl != NULL; chl = chl->chl_next) {
		if (chl->chl_blkno == ip->i_blkno) {

			/* insert this inode into the doubly-linked list
			 * where chl points */
			if ((iq = chl->chl_ip)) {
				ip->i_cprev = iq->i_cprev;
				iq->i_cprev->i_cnext = ip;
				iq->i_cprev = ip;
				ip->i_cnext = iq;
			} else {
				ip->i_cnext = ip;
				ip->i_cprev = ip;
			}
			chl->chl_ip = ip;
			ip->i_chash = chl;
			break;
		}
	}

	/* no hash list found for this block; add a new hash list */
	if (chl == NULL)  {
		if (chlnew == NULL) {
			mutex_spinunlock(&ch->ch_lock, s);
			ASSERT(xfs_chashlist_zone != NULL);
			chlnew = (xfs_chashlist_t *)
					kmem_zone_alloc(xfs_chashlist_zone,
						KM_SLEEP);
			ASSERT(chlnew != NULL);
			goto chlredo;
		} else {
			ip->i_cnext = ip;
			ip->i_cprev = ip;
			ip->i_chash = chlnew;
			chlnew->chl_ip = ip;
			chlnew->chl_blkno = ip->i_blkno;
			chlnew->chl_next = ch->ch_list;
			ch->ch_list = chlnew;
			chlnew = NULL;
		}
	} else {
		if (chlnew != NULL) {
			kmem_zone_free(xfs_chashlist_zone, chlnew);
		}
	}

	mutex_spinunlock(&ch->ch_lock, s);


	/*
	 * Link ip to its mount and thread it on the mount's inode list.
	 */
	XFS_MOUNT_ILOCK(mp);
	if ((iq = mp->m_inodes)) {
		ASSERT(iq->i_mprev->i_mnext == iq);
		ip->i_mprev = iq->i_mprev;
		iq->i_mprev->i_mnext = ip;
		iq->i_mprev = ip;
		ip->i_mnext = iq;
	} else {
		ip->i_mnext = ip;
		ip->i_mprev = ip;
	}
	mp->m_inodes = ip;

	XFS_MOUNT_IUNLOCK(mp);

	newnode = 1;

 return_ip:
	ASSERT(ip->i_df.if_ext_max ==
	       XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t));

	ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) ==
	       ((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0));

	*ipp = ip;

	/*
	 * If we have a real type for an on-disk inode, we can set ops(&unlock)
	 * now.	 If it's a new inode being created, xfs_ialloc will handle it.
	 */
	VFS_INIT_VNODE(XFS_MTOVFS(mp), vp, XFS_ITOBHV(ip), 1);

	return 0;
}


/*
 * The 'normal' internal xfs_iget, if needed it will
 * 'allocate', or 'get', the vnode.
 */
int
xfs_iget(
	xfs_mount_t	*mp,
	xfs_trans_t	*tp,
	xfs_ino_t	ino,
	uint		lock_flags,
	xfs_inode_t	**ipp,
	xfs_daddr_t	bno)
{
	struct inode	*inode;
	vnode_t		*vp = NULL;
	int		error;

retry:
	XFS_STATS_INC(xs_ig_attempts);

	if ((inode = iget_locked(XFS_MTOVFS(mp)->vfs_super, ino))) {
		bhv_desc_t	*bdp;
		xfs_inode_t	*ip;
		int		newnode;

		vp = LINVFS_GET_VP(inode);
		if (inode->i_state & I_NEW) {
inode_allocate:
			vn_initialize(inode);
			error = xfs_iget_core(vp, mp, tp, ino,
						lock_flags, ipp, bno);
			if (error) {
				vn_mark_bad(vp);
				if (inode->i_state & I_NEW)
					unlock_new_inode(inode);
				iput(inode);
			}
		} else {
			/* These are true if the inode is in inactive or
			 * reclaim. The linux inode is about to go away,
			 * wait for that path to finish, and try again.
			 */
			if (vp->v_flag & (VINACT | VRECLM)) {
				vn_wait(vp);
				iput(inode);
				goto retry;
			}

			bdp = vn_bhv_lookup(VN_BHV_HEAD(vp), &xfs_vnodeops);
			if (bdp == NULL) {
				XFS_STATS_INC(xs_ig_dup);
				goto inode_allocate;
			}
			ip = XFS_BHVTOI(bdp);
			if (lock_flags != 0)
				xfs_ilock(ip, lock_flags);
			newnode = (ip->i_d.di_mode == 0);
			if (newnode)
				xfs_iocore_inode_reinit(ip);
			XFS_STATS_INC(xs_ig_found);
			*ipp = ip;
			error = 0;
		}
	} else
		error = ENOMEM;	/* If we got no inode we are out of memory */

	return error;
}

/*