vm_swap.c

Version: - FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
    1 /*
    2  * Copyright (c) 1982, 1986, 1989, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 3. All advertising materials mentioning features or use of this software
   14  *    must display the following acknowledgement:
   15  *      This product includes software developed by the University of
   16  *      California, Berkeley and its contributors.
   17  * 4. Neither the name of the University nor the names of its contributors
   18  *    may be used to endorse or promote products derived from this software
   19  *    without specific prior written permission.
   20  *
   21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   31  * SUCH DAMAGE.
   32  *
   33  *      @(#)vm_swap.c   8.5 (Berkeley) 2/17/94
   34  * $FreeBSD$
   35  */
   36 
   37 #include "opt_swap.h"
   38 
   39 #include <sys/param.h>
   40 #include <sys/systm.h>
   41 #include <sys/sysproto.h>
   42 #include <sys/buf.h>
   43 #include <sys/proc.h>
   44 #include <sys/namei.h>
   45 #include <sys/dmap.h>           /* XXX */
   46 #include <sys/vnode.h>
   47 #include <sys/fcntl.h>
   48 #include <sys/blist.h>
   49 #include <sys/kernel.h>
   50 #include <sys/lock.h>
   51 #include <sys/conf.h>
   52 #include <sys/stat.h>
   53 #include <vm/vm.h>
   54 #include <vm/vm_extern.h>
   55 #include <vm/swap_pager.h>
   56 #include <vm/vm_zone.h>
   57 
   58 /*
   59  * Indirect driver for multi-controller paging.
   60  */
   61 
   62 #ifndef NSWAPDEV
   63 #define NSWAPDEV        4
   64 #endif
   65 static struct swdevt should_be_malloced[NSWAPDEV];
   66 static struct swdevt *swdevt = should_be_malloced;
   67 static int nswap;               /* first block after the interleaved devs */
   68 static int nswdev = NSWAPDEV;
   69 int vm_swap_size;
   70 
   71 static int swapdev_strategy __P((struct vop_strategy_args *ap));
   72 struct vnode *swapdev_vp;
   73 
   74 /*
   75  *      swapdev_strategy:
   76  *
   77  *      VOP_STRATEGY() for swapdev_vp.
   78  *      Perform swap strategy interleave device selection.
   79  *
   80  *      The bp is expected to be locked and *not* B_DONE on call.
   81  */
   82 
   83 static int
   84 swapdev_strategy(ap)
   85         struct vop_strategy_args /* {
   86                 struct vnode *a_vp;
   87                 struct buf *a_bp;
   88         } */ *ap;
   89 {
   90         int s, sz, off, seg, index;
   91         register struct swdevt *sp;
   92         struct vnode *vp;
   93         struct buf *bp;
   94 
   95         bp = ap->a_bp;
   96         sz = howmany(bp->b_bcount, PAGE_SIZE);
   97 
   98         /*
   99          * Convert interleaved swap into per-device swap.  Note that
  100          * the block size is left in PAGE_SIZE'd chunks (for the newswap)
  101          * here.
  102          */
  103         if (nswdev > 1) {
  104                 off = bp->b_blkno % dmmax;
  105                 if (off + sz > dmmax) {
  106                         bp->b_error = EINVAL;
  107                         bp->b_flags |= B_ERROR;
  108                         biodone(bp);
  109                         return 0;
  110                 }
  111                 seg = bp->b_blkno / dmmax;
  112                 index = seg % nswdev;
  113                 seg /= nswdev;
  114                 bp->b_blkno = seg * dmmax + off;
  115         } else {
  116                 index = 0;
  117         }
  118         sp = &swdevt[index];
  119         if (bp->b_blkno + sz > sp->sw_nblks) {
  120                 bp->b_error = EINVAL;
  121                 bp->b_flags |= B_ERROR;
  122                 biodone(bp);
  123                 return 0;
  124         }
  125         bp->b_dev = sp->sw_device;
  126         if (sp->sw_vp == NULL) {
  127                 bp->b_error = ENODEV;
  128                 bp->b_flags |= B_ERROR;
  129                 biodone(bp);
  130                 return 0;
  131         }
  132 
  133         /*
  134          * Convert from PAGE_SIZE'd to DEV_BSIZE'd chunks for the actual I/O
  135          */
  136         bp->b_blkno = ctodb(bp->b_blkno);
  137 
  138         vhold(sp->sw_vp);
  139         s = splvm();
  140         if ((bp->b_flags & B_READ) == 0) {
  141                 vp = bp->b_vp;
  142                 if (vp) {
  143                         vp->v_numoutput--;
  144                         if ((vp->v_flag & VBWAIT) && vp->v_numoutput <= 0) {
  145                                 vp->v_flag &= ~VBWAIT;
  146                                 wakeup(&vp->v_numoutput);
  147                         }
  148                 }
  149                 sp->sw_vp->v_numoutput++;
  150         }
  151         pbreassignbuf(bp, sp->sw_vp);
  152         splx(s);
  153         VOP_STRATEGY(bp->b_vp, bp);
  154         return 0;
  155 }
  156 
  157 /*
  158  * Create a special vnode op vector for swapdev_vp - we only use
  159  * VOP_STRATEGY(), everything else returns an error.
  160  */
  161 vop_t **swapdev_vnodeop_p;
  162 static struct vnodeopv_entry_desc swapdev_vnodeop_entries[] = {  
  163         { &vop_default_desc,            (vop_t *) vop_defaultop },
  164         { &vop_strategy_desc,           (vop_t *) swapdev_strategy },
  165         { NULL, NULL }
  166 };
  167 static struct vnodeopv_desc swapdev_vnodeop_opv_desc =
  168         { &swapdev_vnodeop_p, swapdev_vnodeop_entries };
  169 
  170 VNODEOP_SET(swapdev_vnodeop_opv_desc);
  171 
  172 /*
  173  * System call swapon(name) enables swapping on device name,
  174  * which must be in the swdevsw.  Return EBUSY
  175  * if already swapping on this device.
  176  */
  177 #ifndef _SYS_SYSPROTO_H_
  178 struct swapon_args {
  179         char *name;
  180 };
  181 #endif
  182 
  183 /* ARGSUSED */
  184 int
  185 swapon(p, uap)
  186         struct proc *p;
  187         struct swapon_args *uap;
  188 {
  189         struct vattr attr;
  190         register struct vnode *vp;
  191         struct nameidata nd;
  192         int error;
  193 
  194         error = suser(p);
  195         if (error)
  196                 return (error);
  197 
  198         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->name, p);
  199         error = namei(&nd);
  200         if (error)
  201                 return (error);
  202 
  203         NDFREE(&nd, NDF_ONLY_PNBUF);
  204         vp = nd.ni_vp;
  205 
  206         if (vn_isdisk(vp, &error))
  207                 error = swaponvp(p, vp, vp->v_rdev, 0);
  208         else if (vp->v_type == VREG && vp->v_tag == VT_NFS &&
  209             (error = VOP_GETATTR(vp, &attr, p->p_ucred, p)) == 0) {
  210                 /*
  211                  * Allow direct swapping to NFS regular files in the same
  212                  * way that nfs_mountroot() sets up diskless swapping.
  213                  */
  214                 error = swaponvp(p, vp, NODEV, attr.va_size / DEV_BSIZE);
  215         }
  216 
  217         if (error)
  218                 vrele(vp);
  219 
  220         return (error);
  221 }
  222 
  223 /*
  224  * Swfree(index) frees the index'th portion of the swap map.
  225  * Each of the nswdev devices provides 1/nswdev'th of the swap
  226  * space, which is laid out with blocks of dmmax pages circularly
  227  * among the devices.
  228  *
  229  * The new swap code uses page-sized blocks.  The old swap code used
  230  * DEV_BSIZE'd chunks.
  231  *
  232  * XXX locking when multiple swapon's run in parallel
  233  */
  234 int
  235 swaponvp(p, vp, dev, nblks)
  236         struct proc *p;
  237         struct vnode *vp;
  238         dev_t dev;
  239         u_long nblks;
  240 {
  241         int index;
  242         register struct swdevt *sp;
  243         register swblk_t vsbase;
  244         register long blk;
  245         swblk_t dvbase;
  246         int error;
  247         u_long aligned_nblks;
  248 
  249         if (!swapdev_vp) {
  250                 error = getnewvnode(VT_NON, NULL, swapdev_vnodeop_p,
  251                     &swapdev_vp);
  252                 if (error)
  253                         panic("Cannot get vnode for swapdev");
  254                 swapdev_vp->v_type = VNON;      /* Untyped */
  255         }
  256 
  257         ASSERT_VOP_UNLOCKED(vp, "swaponvp");
  258         for (sp = swdevt, index = 0 ; index < nswdev; index++, sp++) {
  259                 if (sp->sw_vp == vp)
  260                         return EBUSY;
  261                 if (!sp->sw_vp)
  262                         goto found;
  263 
  264         }
  265         return EINVAL;
  266     found:
  267         (void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
  268         error = VOP_OPEN(vp, FREAD | FWRITE, p->p_ucred, p);
  269         (void) VOP_UNLOCK(vp, 0, p);
  270         if (error)
  271                 return (error);
  272 
  273         if (nblks == 0 && dev != NODEV && (devsw(dev)->d_psize == 0 ||
  274             (nblks = (*devsw(dev)->d_psize) (dev)) == -1)) {
  275                 (void) VOP_CLOSE(vp, FREAD | FWRITE, p->p_ucred, p);
  276                 return (ENXIO);
  277         }
  278         if (nblks == 0) {
  279                 (void) VOP_CLOSE(vp, FREAD | FWRITE, p->p_ucred, p);
  280                 return (ENXIO);
  281         }
  282 
  283         /*
  284          * If we go beyond this, we get overflows in the radix
  285          * tree bitmap code.
  286          */
  287         if (nblks > 0x40000000 / BLIST_META_RADIX / nswdev) {
  288                 printf("exceeded maximum of %d blocks per swap unit\n",
  289                         0x40000000 / BLIST_META_RADIX / nswdev);
  290                 (void) VOP_CLOSE(vp, FREAD | FWRITE, p->p_ucred, p);
  291                 return (ENXIO);
  292         }
  293         /*
  294          * nblks is in DEV_BSIZE'd chunks, convert to PAGE_SIZE'd chunks.
  295          * First chop nblks off to page-align it, then convert.
  296          * 
  297          * sw->sw_nblks is in page-sized chunks now too.
  298          */
  299         nblks &= ~(ctodb(1) - 1);
  300         nblks = dbtoc(nblks);
  301 
  302         sp->sw_vp = vp;
  303         sp->sw_dev = dev2udev(dev);
  304         sp->sw_device = dev;
  305         sp->sw_flags |= SW_FREED;
  306         sp->sw_nblks = nblks;
  307 
  308         /*
  309          * nblks, nswap, and dmmax are PAGE_SIZE'd parameters now, not
  310          * DEV_BSIZE'd.   aligned_nblks is used to calculate the
  311          * size of the swap bitmap, taking into account the stripe size.
  312          */
  313         aligned_nblks = (nblks + (dmmax - 1)) & ~(u_long)(dmmax - 1);
  314 
  315         if (aligned_nblks * nswdev > nswap)
  316                 nswap = aligned_nblks * nswdev;
  317 
  318         if (swapblist == NULL)
  319                 swapblist = blist_create(nswap);
  320         else
  321                 blist_resize(&swapblist, nswap, 0);
  322 
  323         for (dvbase = dmmax; dvbase < nblks; dvbase += dmmax) {
  324                 blk = min(nblks - dvbase, dmmax);
  325                 vsbase = index * dmmax + dvbase * nswdev;
  326                 blist_free(swapblist, vsbase, blk);
  327                 vm_swap_size += blk;
  328         }
  329 
  330         return (0);
  331 }
Cache object: b6b55e5dee7372154c1ce1d2daf51006
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.
FreeBSD/Linux Kernel Cross Reference sys/vm/vm_swap.c

FreeBSD/Linux Kernel Cross Reference
sys/vm/vm_swap.c
