xfs_filestream.c

linux 内核源代码

/*
 * Copyright (c) 2006-2007 Silicon Graphics, Inc.
 * 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.
 *
 * 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.  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 the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#include "xfs.h"
#include "xfs_bmap_btree.h"
#include "xfs_inum.h"
#include "xfs_dir2.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_ag.h"
#include "xfs_dmapi.h"
#include "xfs_log.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_mount.h"
#include "xfs_bmap.h"
#include "xfs_alloc.h"
#include "xfs_utils.h"
#include "xfs_mru_cache.h"
#include "xfs_filestream.h"

#ifdef XFS_FILESTREAMS_TRACE

ktrace_t *xfs_filestreams_trace_buf;

STATIC void
xfs_filestreams_trace(
	xfs_mount_t	*mp,	/* mount point */
	int		type,	/* type of trace */
	const char	*func,	/* source function */
	int		line,	/* source line number */
	__psunsigned_t	arg0,
	__psunsigned_t	arg1,
	__psunsigned_t	arg2,
	__psunsigned_t	arg3,
	__psunsigned_t	arg4,
	__psunsigned_t	arg5)
{
	ktrace_enter(xfs_filestreams_trace_buf,
		(void *)(__psint_t)(type | (line << 16)),
		(void *)func,
		(void *)(__psunsigned_t)current_pid(),
		(void *)mp,
		(void *)(__psunsigned_t)arg0,
		(void *)(__psunsigned_t)arg1,
		(void *)(__psunsigned_t)arg2,
		(void *)(__psunsigned_t)arg3,
		(void *)(__psunsigned_t)arg4,
		(void *)(__psunsigned_t)arg5,
		NULL, NULL, NULL, NULL, NULL, NULL);
}

#define TRACE0(mp,t)			TRACE6(mp,t,0,0,0,0,0,0)
#define TRACE1(mp,t,a0)			TRACE6(mp,t,a0,0,0,0,0,0)
#define TRACE2(mp,t,a0,a1)		TRACE6(mp,t,a0,a1,0,0,0,0)
#define TRACE3(mp,t,a0,a1,a2)		TRACE6(mp,t,a0,a1,a2,0,0,0)
#define TRACE4(mp,t,a0,a1,a2,a3)	TRACE6(mp,t,a0,a1,a2,a3,0,0)
#define TRACE5(mp,t,a0,a1,a2,a3,a4)	TRACE6(mp,t,a0,a1,a2,a3,a4,0)
#define TRACE6(mp,t,a0,a1,a2,a3,a4,a5) \
	xfs_filestreams_trace(mp, t, __FUNCTION__, __LINE__, \
				(__psunsigned_t)a0, (__psunsigned_t)a1, \
				(__psunsigned_t)a2, (__psunsigned_t)a3, \
				(__psunsigned_t)a4, (__psunsigned_t)a5)

#define TRACE_AG_SCAN(mp, ag, ag2) \
		TRACE2(mp, XFS_FSTRM_KTRACE_AGSCAN, ag, ag2);
#define TRACE_AG_PICK1(mp, max_ag, maxfree) \
		TRACE2(mp, XFS_FSTRM_KTRACE_AGPICK1, max_ag, maxfree);
#define TRACE_AG_PICK2(mp, ag, ag2, cnt, free, scan, flag) \
		TRACE6(mp, XFS_FSTRM_KTRACE_AGPICK2, ag, ag2, \
			 cnt, free, scan, flag)
#define TRACE_UPDATE(mp, ip, ag, cnt, ag2, cnt2) \
		TRACE5(mp, XFS_FSTRM_KTRACE_UPDATE, ip, ag, cnt, ag2, cnt2)
#define TRACE_FREE(mp, ip, pip, ag, cnt) \
		TRACE4(mp, XFS_FSTRM_KTRACE_FREE, ip, pip, ag, cnt)
#define TRACE_LOOKUP(mp, ip, pip, ag, cnt) \
		TRACE4(mp, XFS_FSTRM_KTRACE_ITEM_LOOKUP, ip, pip, ag, cnt)
#define TRACE_ASSOCIATE(mp, ip, pip, ag, cnt) \
		TRACE4(mp, XFS_FSTRM_KTRACE_ASSOCIATE, ip, pip, ag, cnt)
#define TRACE_MOVEAG(mp, ip, pip, oag, ocnt, nag, ncnt) \
		TRACE6(mp, XFS_FSTRM_KTRACE_MOVEAG, ip, pip, oag, ocnt, nag, ncnt)
#define TRACE_ORPHAN(mp, ip, ag) \
		TRACE2(mp, XFS_FSTRM_KTRACE_ORPHAN, ip, ag);


#else
#define TRACE_AG_SCAN(mp, ag, ag2)
#define TRACE_AG_PICK1(mp, max_ag, maxfree)
#define TRACE_AG_PICK2(mp, ag, ag2, cnt, free, scan, flag)
#define TRACE_UPDATE(mp, ip, ag, cnt, ag2, cnt2)
#define TRACE_FREE(mp, ip, pip, ag, cnt)
#define TRACE_LOOKUP(mp, ip, pip, ag, cnt)
#define TRACE_ASSOCIATE(mp, ip, pip, ag, cnt)
#define TRACE_MOVEAG(mp, ip, pip, oag, ocnt, nag, ncnt)
#define TRACE_ORPHAN(mp, ip, ag)
#endif

static kmem_zone_t *item_zone;

/*
 * Structure for associating a file or a directory with an allocation group.
 * The parent directory pointer is only needed for files, but since there will
 * generally be vastly more files than directories in the cache, using the same
 * data structure simplifies the code with very little memory overhead.
 */
typedef struct fstrm_item
{
	xfs_agnumber_t	ag;	/* AG currently in use for the file/directory. */
	xfs_inode_t	*ip;	/* inode self-pointer. */
	xfs_inode_t	*pip;	/* Parent directory inode pointer. */
} fstrm_item_t;


/*
 * Scan the AGs starting at startag looking for an AG that isn't in use and has
 * at least minlen blocks free.
 */
static int
_xfs_filestream_pick_ag(
	xfs_mount_t	*mp,
	xfs_agnumber_t	startag,
	xfs_agnumber_t	*agp,
	int		flags,
	xfs_extlen_t	minlen)
{
	int		err, trylock, nscan;
	xfs_extlen_t	delta, longest, need, free, minfree, maxfree = 0;
	xfs_agnumber_t	ag, max_ag = NULLAGNUMBER;
	struct xfs_perag *pag;

	/* 2% of an AG's blocks must be free for it to be chosen. */
	minfree = mp->m_sb.sb_agblocks / 50;

	ag = startag;
	*agp = NULLAGNUMBER;

	/* For the first pass, don't sleep trying to init the per-AG. */
	trylock = XFS_ALLOC_FLAG_TRYLOCK;

	for (nscan = 0; 1; nscan++) {

		TRACE_AG_SCAN(mp, ag, xfs_filestream_peek_ag(mp, ag));

		pag = mp->m_perag + ag;

		if (!pag->pagf_init) {
			err = xfs_alloc_pagf_init(mp, NULL, ag, trylock);
			if (err && !trylock)
				return err;
		}

		/* Might fail sometimes during the 1st pass with trylock set. */
		if (!pag->pagf_init)
			goto next_ag;

		/* Keep track of the AG with the most free blocks. */
		if (pag->pagf_freeblks > maxfree) {
			maxfree = pag->pagf_freeblks;
			max_ag = ag;
		}

		/*
		 * The AG reference count does two things: it enforces mutual
		 * exclusion when examining the suitability of an AG in this
		 * loop, and it guards against two filestreams being established
		 * in the same AG as each other.
		 */
		if (xfs_filestream_get_ag(mp, ag) > 1) {
			xfs_filestream_put_ag(mp, ag);
			goto next_ag;
		}

		need = XFS_MIN_FREELIST_PAG(pag, mp);
		delta = need > pag->pagf_flcount ? need - pag->pagf_flcount : 0;
		longest = (pag->pagf_longest > delta) ?
		          (pag->pagf_longest - delta) :
		          (pag->pagf_flcount > 0 || pag->pagf_longest > 0);

		if (((minlen && longest >= minlen) ||
		     (!minlen && pag->pagf_freeblks >= minfree)) &&
		    (!pag->pagf_metadata || !(flags & XFS_PICK_USERDATA) ||
		     (flags & XFS_PICK_LOWSPACE))) {

			/* Break out, retaining the reference on the AG. */
			free = pag->pagf_freeblks;
			*agp = ag;
			break;
		}

		/* Drop the reference on this AG, it's not usable. */
		xfs_filestream_put_ag(mp, ag);
next_ag:
		/* Move to the next AG, wrapping to AG 0 if necessary. */
		if (++ag >= mp->m_sb.sb_agcount)
			ag = 0;

		/* If a full pass of the AGs hasn't been done yet, continue. */
		if (ag != startag)
			continue;

		/* Allow sleeping in xfs_alloc_pagf_init() on the 2nd pass. */
		if (trylock != 0) {
			trylock = 0;
			continue;
		}

		/* Finally, if lowspace wasn't set, set it for the 3rd pass. */
		if (!(flags & XFS_PICK_LOWSPACE)) {
			flags |= XFS_PICK_LOWSPACE;
			continue;
		}

		/*
		 * Take the AG with the most free space, regardless of whether
		 * it's already in use by another filestream.
		 */
		if (max_ag != NULLAGNUMBER) {
			xfs_filestream_get_ag(mp, max_ag);
			TRACE_AG_PICK1(mp, max_ag, maxfree);
			free = maxfree;
			*agp = max_ag;
			break;
		}

		/* take AG 0 if none matched */
		TRACE_AG_PICK1(mp, max_ag, maxfree);
		*agp = 0;
		return 0;
	}

	TRACE_AG_PICK2(mp, startag, *agp, xfs_filestream_peek_ag(mp, *agp),
			free, nscan, flags);

	return 0;
}

/*
 * Set the allocation group number for a file or a directory, updating inode
 * references and per-AG references as appropriate.  Must be called with the
 * m_peraglock held in read mode.
 */
static int
_xfs_filestream_update_ag(
	xfs_inode_t	*ip,
	xfs_inode_t	*pip,
	xfs_agnumber_t	ag)
{
	int		err = 0;
	xfs_mount_t	*mp;
	xfs_mru_cache_t	*cache;
	fstrm_item_t	*item;
	xfs_agnumber_t	old_ag;
	xfs_inode_t	*old_pip;

	/*
	 * Either ip is a regular file and pip is a directory, or ip is a
	 * directory and pip is NULL.
	 */
	ASSERT(ip && (((ip->i_d.di_mode & S_IFREG) && pip &&
	               (pip->i_d.di_mode & S_IFDIR)) ||
	              ((ip->i_d.di_mode & S_IFDIR) && !pip)));

	mp = ip->i_mount;
	cache = mp->m_filestream;

	item = xfs_mru_cache_lookup(cache, ip->i_ino);
	if (item) {
		ASSERT(item->ip == ip);
		old_ag = item->ag;
		item->ag = ag;
		old_pip = item->pip;
		item->pip = pip;
		xfs_mru_cache_done(cache);

		/*
		 * If the AG has changed, drop the old ref and take a new one,
		 * effectively transferring the reference from old to new AG.
		 */
		if (ag != old_ag) {
			xfs_filestream_put_ag(mp, old_ag);
			xfs_filestream_get_ag(mp, ag);
		}

		/*
		 * If ip is a file and its pip has changed, drop the old ref and
		 * take a new one.
		 */
		if (pip && pip != old_pip) {
			IRELE(old_pip);
			IHOLD(pip);
		}

		TRACE_UPDATE(mp, ip, old_ag, xfs_filestream_peek_ag(mp, old_ag),
				ag, xfs_filestream_peek_ag(mp, ag));
		return 0;
	}

	item = kmem_zone_zalloc(item_zone, KM_MAYFAIL);
	if (!item)
		return ENOMEM;

	item->ag = ag;
	item->ip = ip;
	item->pip = pip;

	err = xfs_mru_cache_insert(cache, ip->i_ino, item);
	if (err) {
		kmem_zone_free(item_zone, item);
		return err;
	}

	/* Take a reference on the AG. */
	xfs_filestream_get_ag(mp, ag);

	/*
	 * Take a reference on the inode itself regardless of whether it's a
	 * regular file or a directory.
	 */
	IHOLD(ip);

	/*
	 * In the case of a regular file, take a reference on the parent inode
	 * as well to ensure it remains in-core.
	 */
	if (pip)
		IHOLD(pip);

	TRACE_UPDATE(mp, ip, ag, xfs_filestream_peek_ag(mp, ag),
			ag, xfs_filestream_peek_ag(mp, ag));

	return 0;
}

/* xfs_fstrm_free_func(): callback for freeing cached stream items. */
void
xfs_fstrm_free_func(
	unsigned long	ino,
	void		*data)
{
	fstrm_item_t	*item  = (fstrm_item_t *)data;
	xfs_inode_t	*ip = item->ip;
	int ref;

	ASSERT(ip->i_ino == ino);

	xfs_iflags_clear(ip, XFS_IFILESTREAM);

	/* Drop the reference taken on the AG when the item was added. */
	ref = xfs_filestream_put_ag(ip->i_mount, item->ag);

	ASSERT(ref >= 0);
	TRACE_FREE(ip->i_mount, ip, item->pip, item->ag,
		xfs_filestream_peek_ag(ip->i_mount, item->ag));

	/*
	 * _xfs_filestream_update_ag() always takes a reference on the inode
	 * itself, whether it's a file or a directory.  Release it here.
	 * This can result in the inode being freed and so we must
	 * not hold any inode locks when freeing filesstreams objects
	 * otherwise we can deadlock here.
	 */
	IRELE(ip);

	/*
	 * In the case of a regular file, _xfs_filestream_update_ag() also
	 * takes a ref on the parent inode to keep it in-core.  Release that
	 * too.
	 */
	if (item->pip)
		IRELE(item->pip);