Home | History | Annotate | Line # | Download | only in lfs
ulfs_inode.c revision 1.5
      1 /*	$NetBSD: ulfs_inode.c,v 1.5 2013/06/06 00:49:28 dholland Exp $	*/
      2 /*  from NetBSD: ufs_inode.c,v 1.89 2013/01/22 09:39:18 dholland Exp  */
      3 
      4 /*
      5  * Copyright (c) 1991, 1993
      6  *	The Regents of the University of California.  All rights reserved.
      7  * (c) UNIX System Laboratories, Inc.
      8  * All or some portions of this file are derived from material licensed
      9  * to the University of California by American Telephone and Telegraph
     10  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
     11  * the permission of UNIX System Laboratories, Inc.
     12  *
     13  * Redistribution and use in source and binary forms, with or without
     14  * modification, are permitted provided that the following conditions
     15  * are met:
     16  * 1. Redistributions of source code must retain the above copyright
     17  *    notice, this list of conditions and the following disclaimer.
     18  * 2. Redistributions in binary form must reproduce the above copyright
     19  *    notice, this list of conditions and the following disclaimer in the
     20  *    documentation and/or other materials provided with the distribution.
     21  * 3. Neither the name of the University nor the names of its contributors
     22  *    may be used to endorse or promote products derived from this software
     23  *    without specific prior written permission.
     24  *
     25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     35  * SUCH DAMAGE.
     36  *
     37  *	@(#)ufs_inode.c	8.9 (Berkeley) 5/14/95
     38  */
     39 
     40 #include <sys/cdefs.h>
     41 __KERNEL_RCSID(0, "$NetBSD: ulfs_inode.c,v 1.5 2013/06/06 00:49:28 dholland Exp $");
     42 
     43 #if defined(_KERNEL_OPT)
     44 #include "opt_lfs.h"
     45 #include "opt_quota.h"
     46 #include "opt_wapbl.h"
     47 #endif
     48 
     49 #include <sys/param.h>
     50 #include <sys/systm.h>
     51 #include <sys/proc.h>
     52 #include <sys/vnode.h>
     53 #include <sys/mount.h>
     54 #include <sys/kernel.h>
     55 #include <sys/namei.h>
     56 #include <sys/kauth.h>
     57 #include <sys/wapbl.h>
     58 #include <sys/fstrans.h>
     59 #include <sys/kmem.h>
     60 
     61 #include <ufs/lfs/ulfs_inode.h>
     62 #include <ufs/lfs/ulfsmount.h>
     63 #include <ufs/lfs/ulfs_extern.h>
     64 #include <ufs/lfs/ulfs_wapbl.h>
     65 #ifdef LFS_DIRHASH
     66 #include <ufs/lfs/ulfs_dirhash.h>
     67 #endif
     68 #ifdef LFS_EXTATTR
     69 #include <ufs/lfs/ulfs_extattr.h>
     70 #endif
     71 
     72 #include <uvm/uvm.h>
     73 
     74 extern int prtactive;
     75 
     76 /*
     77  * Last reference to an inode.  If necessary, write or delete it.
     78  */
     79 int
     80 ulfs_inactive(void *v)
     81 {
     82 	struct vop_inactive_args /* {
     83 		struct vnode *a_vp;
     84 		struct bool *a_recycle;
     85 	} */ *ap = v;
     86 	struct vnode *vp = ap->a_vp;
     87 	struct inode *ip = VTOI(vp);
     88 	struct mount *transmp;
     89 	mode_t mode;
     90 	int error = 0;
     91 	int logged = 0;
     92 
     93 	ULFS_WAPBL_JUNLOCK_ASSERT(vp->v_mount);
     94 
     95 	transmp = vp->v_mount;
     96 	fstrans_start(transmp, FSTRANS_LAZY);
     97 	/*
     98 	 * Ignore inodes related to stale file handles.
     99 	 */
    100 	if (ip->i_mode == 0)
    101 		goto out;
    102 	if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
    103 #ifdef LFS_EXTATTR
    104 		ulfs_extattr_vnode_inactive(vp, curlwp);
    105 #endif
    106 		error = ULFS_WAPBL_BEGIN(vp->v_mount);
    107 		if (error)
    108 			goto out;
    109 		logged = 1;
    110 		if (ip->i_size != 0) {
    111 			/*
    112 			 * When journaling, only truncate one indirect block
    113 			 * at a time
    114 			 */
    115 			if (vp->v_mount->mnt_wapbl) {
    116 				uint64_t incr = MNINDIR(ip->i_ump) <<
    117 				    vp->v_mount->mnt_fs_bshift; /* Power of 2 */
    118 				uint64_t base = ULFS_NDADDR <<
    119 				    vp->v_mount->mnt_fs_bshift;
    120 				while (!error && ip->i_size > base + incr) {
    121 					/*
    122 					 * round down to next full indirect
    123 					 * block boundary.
    124 					 */
    125 					uint64_t nsize = base +
    126 					    ((ip->i_size - base - 1) &
    127 					    ~(incr - 1));
    128 					error = ULFS_TRUNCATE(vp, nsize, 0,
    129 					    NOCRED);
    130 					if (error)
    131 						break;
    132 					ULFS_WAPBL_END(vp->v_mount);
    133 					error = ULFS_WAPBL_BEGIN(vp->v_mount);
    134 					if (error)
    135 						goto out;
    136 				}
    137 			}
    138 			if (!error)
    139 				error = ULFS_TRUNCATE(vp, (off_t)0, 0, NOCRED);
    140 		}
    141 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
    142 		(void)lfs_chkiq(ip, -1, NOCRED, 0);
    143 #endif
    144 		DIP_ASSIGN(ip, rdev, 0);
    145 		mode = ip->i_mode;
    146 		ip->i_mode = 0;
    147 		ip->i_omode = mode;
    148 		DIP_ASSIGN(ip, mode, 0);
    149 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
    150 		/*
    151 		 * Defer final inode free and update to ulfs_reclaim().
    152 		 */
    153 	}
    154 
    155 	if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
    156 		if (!logged++) {
    157 			int err;
    158 			err = ULFS_WAPBL_BEGIN(vp->v_mount);
    159 			if (err)
    160 				goto out;
    161 		}
    162 		ULFS_UPDATE(vp, NULL, NULL, 0);
    163 	}
    164 	if (logged)
    165 		ULFS_WAPBL_END(vp->v_mount);
    166 out:
    167 	/*
    168 	 * If we are done with the inode, reclaim it
    169 	 * so that it can be reused immediately.
    170 	 */
    171 	*ap->a_recycle = (ip->i_mode == 0);
    172 	VOP_UNLOCK(vp);
    173 	fstrans_done(transmp);
    174 	return (error);
    175 }
    176 
    177 /*
    178  * Reclaim an inode so that it can be used for other purposes.
    179  */
    180 int
    181 ulfs_reclaim(struct vnode *vp)
    182 {
    183 	struct inode *ip = VTOI(vp);
    184 
    185 	if (prtactive && vp->v_usecount > 1)
    186 		vprint("ulfs_reclaim: pushing active", vp);
    187 
    188 	if (!ULFS_WAPBL_BEGIN(vp->v_mount)) {
    189 		ULFS_UPDATE(vp, NULL, NULL, UPDATE_CLOSE);
    190 		ULFS_WAPBL_END(vp->v_mount);
    191 	}
    192 	ULFS_UPDATE(vp, NULL, NULL, UPDATE_CLOSE);
    193 
    194 	/*
    195 	 * Remove the inode from its hash chain.
    196 	 */
    197 	ulfs_ihashrem(ip);
    198 
    199 	if (ip->i_devvp) {
    200 		vrele(ip->i_devvp);
    201 		ip->i_devvp = 0;
    202 	}
    203 #if defined(LFS_QUOTA) || defined(LFS_QUOTA2)
    204 	ulfsquota_free(ip);
    205 #endif
    206 #ifdef LFS_DIRHASH
    207 	if (ip->i_dirhash != NULL)
    208 		ulfsdirhash_free(ip);
    209 #endif
    210 	return (0);
    211 }
    212 
    213 /*
    214  * allocate a range of blocks in a file.
    215  * after this function returns, any page entirely contained within the range
    216  * will map to invalid data and thus must be overwritten before it is made
    217  * accessible to others.
    218  */
    219 
    220 int
    221 ulfs_balloc_range(struct vnode *vp, off_t off, off_t len, kauth_cred_t cred,
    222     int flags)
    223 {
    224 	off_t neweof;	/* file size after the operation */
    225 	off_t neweob;	/* offset next to the last block after the operation */
    226 	off_t pagestart; /* starting offset of range covered by pgs */
    227 	off_t eob;	/* offset next to allocated blocks */
    228 	struct uvm_object *uobj;
    229 	int i, delta, error, npages;
    230 	int bshift = vp->v_mount->mnt_fs_bshift;
    231 	int bsize = 1 << bshift;
    232 	int ppb = MAX(bsize >> PAGE_SHIFT, 1);
    233 	struct vm_page **pgs;
    234 	size_t pgssize;
    235 	UVMHIST_FUNC("ulfs_balloc_range"); UVMHIST_CALLED(ubchist);
    236 	UVMHIST_LOG(ubchist, "vp %p off 0x%x len 0x%x u_size 0x%x",
    237 		    vp, off, len, vp->v_size);
    238 
    239 	neweof = MAX(vp->v_size, off + len);
    240 	GOP_SIZE(vp, neweof, &neweob, 0);
    241 
    242 	error = 0;
    243 	uobj = &vp->v_uobj;
    244 
    245 	/*
    246 	 * read or create pages covering the range of the allocation and
    247 	 * keep them locked until the new block is allocated, so there
    248 	 * will be no window where the old contents of the new block are
    249 	 * visible to racing threads.
    250 	 */
    251 
    252 	pagestart = trunc_page(off) & ~(bsize - 1);
    253 	npages = MIN(ppb, (round_page(neweob) - pagestart) >> PAGE_SHIFT);
    254 	pgssize = npages * sizeof(struct vm_page *);
    255 	pgs = kmem_zalloc(pgssize, KM_SLEEP);
    256 
    257 	/*
    258 	 * adjust off to be block-aligned.
    259 	 */
    260 
    261 	delta = off & (bsize - 1);
    262 	off -= delta;
    263 	len += delta;
    264 
    265 	genfs_node_wrlock(vp);
    266 	mutex_enter(uobj->vmobjlock);
    267 	error = VOP_GETPAGES(vp, pagestart, pgs, &npages, 0,
    268 	    VM_PROT_WRITE, 0, PGO_SYNCIO | PGO_PASTEOF | PGO_NOBLOCKALLOC |
    269 	    PGO_NOTIMESTAMP | PGO_GLOCKHELD);
    270 	if (error) {
    271 		goto out;
    272 	}
    273 
    274 	/*
    275 	 * now allocate the range.
    276 	 */
    277 
    278 	error = GOP_ALLOC(vp, off, len, flags, cred);
    279 	genfs_node_unlock(vp);
    280 
    281 	/*
    282 	 * if the allocation succeeded, clear PG_CLEAN on all the pages
    283 	 * and clear PG_RDONLY on any pages that are now fully backed
    284 	 * by disk blocks.  if the allocation failed, we do not invalidate
    285 	 * the pages since they might have already existed and been dirty,
    286 	 * in which case we need to keep them around.  if we created the pages,
    287 	 * they will be clean and read-only, and leaving such pages
    288 	 * in the cache won't cause any problems.
    289 	 */
    290 
    291 	GOP_SIZE(vp, off + len, &eob, 0);
    292 	mutex_enter(uobj->vmobjlock);
    293 	mutex_enter(&uvm_pageqlock);
    294 	for (i = 0; i < npages; i++) {
    295 		KASSERT((pgs[i]->flags & PG_RELEASED) == 0);
    296 		if (!error) {
    297 			if (off <= pagestart + (i << PAGE_SHIFT) &&
    298 			    pagestart + ((i + 1) << PAGE_SHIFT) <= eob) {
    299 				pgs[i]->flags &= ~PG_RDONLY;
    300 			}
    301 			pgs[i]->flags &= ~PG_CLEAN;
    302 		}
    303 		uvm_pageactivate(pgs[i]);
    304 	}
    305 	mutex_exit(&uvm_pageqlock);
    306 	uvm_page_unbusy(pgs, npages);
    307 	mutex_exit(uobj->vmobjlock);
    308 
    309  out:
    310  	kmem_free(pgs, pgssize);
    311 	return error;
    312 }
    313