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: releng/5.0/sys/vm/vm_swap.c 106023 2002-10-27 06:54:06Z rwatson $
   35  */
   36 
   37 #include "opt_mac.h"
   38 #include "opt_swap.h"
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/sysproto.h>
   43 #include <sys/bio.h>
   44 #include <sys/buf.h>
   45 #include <sys/proc.h>
   46 #include <sys/namei.h>
   47 #include <sys/dmap.h>           /* XXX */
   48 #include <sys/vnode.h>
   49 #include <sys/fcntl.h>
   50 #include <sys/blist.h>
   51 #include <sys/kernel.h>
   52 #include <sys/lock.h>
   53 #include <sys/conf.h>
   54 #include <sys/stat.h>
   55 #include <sys/sysctl.h>
   56 #include <sys/mac.h>
   57 #include <sys/mount.h>
   58 #include <vm/vm.h>
   59 #include <vm/vm_extern.h>
   60 #include <vm/vm_param.h>
   61 #include <vm/swap_pager.h>
   62 #include <vm/uma.h>
   63 
   64 /*
   65  * Indirect driver for multi-controller paging.
   66  */
   67 
   68 #ifndef NSWAPDEV
   69 #define NSWAPDEV        4
   70 #endif
   71 static struct swdevt should_be_malloced[NSWAPDEV];
   72 struct swdevt *swdevt = should_be_malloced;
   73 static int nswap;               /* first block after the interleaved devs */
   74 int nswdev = NSWAPDEV;
   75 int vm_swap_size;
   76 
   77 static int swapdev_strategy(struct vop_strategy_args *ap);
   78 struct vnode *swapdev_vp;
   79 
   80 /*
   81  *      swapdev_strategy:
   82  *
   83  *      VOP_STRATEGY() for swapdev_vp.
   84  *      Perform swap strategy interleave device selection.
   85  *
   86  *      The bp is expected to be locked and *not* B_DONE on call.
   87  */
   88 static int
   89 swapdev_strategy(ap)
   90         struct vop_strategy_args /* {
   91                 struct vnode *a_vp;
   92                 struct buf *a_bp;
   93         } */ *ap;
   94 {
   95         int s, sz, off, seg, index;
   96         struct swdevt *sp;
   97         struct vnode *vp;
   98         struct buf *bp;
   99 
  100         bp = ap->a_bp;
  101         sz = howmany(bp->b_bcount, PAGE_SIZE);
  102 
  103         /*
  104          * Convert interleaved swap into per-device swap.  Note that
  105          * the block size is left in PAGE_SIZE'd chunks (for the newswap)
  106          * here.
  107          */
  108         if (nswdev > 1) {
  109                 off = bp->b_blkno % dmmax;
  110                 if (off + sz > dmmax) {
  111                         bp->b_error = EINVAL;
  112                         bp->b_ioflags |= BIO_ERROR;
  113                         bufdone(bp);
  114                         return 0;
  115                 }
  116                 seg = bp->b_blkno / dmmax;
  117                 index = seg % nswdev;
  118                 seg /= nswdev;
  119                 bp->b_blkno = seg * dmmax + off;
  120         } else {
  121                 index = 0;
  122         }
  123         sp = &swdevt[index];
  124         if (bp->b_blkno + sz > sp->sw_nblks) {
  125                 bp->b_error = EINVAL;
  126                 bp->b_ioflags |= BIO_ERROR;
  127                 bufdone(bp);
  128                 return 0;
  129         }
  130         bp->b_dev = sp->sw_device;
  131         if (sp->sw_vp == NULL) {
  132                 bp->b_error = ENODEV;
  133                 bp->b_ioflags |= BIO_ERROR;
  134                 bufdone(bp);
  135                 return 0;
  136         }
  137 
  138         /*
  139          * Convert from PAGE_SIZE'd to DEV_BSIZE'd chunks for the actual I/O
  140          */
  141         bp->b_blkno = ctodb(bp->b_blkno);
  142 
  143         vhold(sp->sw_vp);
  144         s = splvm();
  145         if (bp->b_iocmd == BIO_WRITE) {
  146                 vp = bp->b_vp;
  147                 if (vp) {
  148                         VI_LOCK(vp);
  149                         vp->v_numoutput--;
  150                         if ((vp->v_iflag & VI_BWAIT) && vp->v_numoutput <= 0) {
  151                                 vp->v_iflag &= ~VI_BWAIT;
  152                                 wakeup(&vp->v_numoutput);
  153                         }
  154                         VI_UNLOCK(vp);
  155                 }
  156                 VI_LOCK(sp->sw_vp);
  157                 sp->sw_vp->v_numoutput++;
  158                 VI_UNLOCK(sp->sw_vp);
  159         }
  160         bp->b_vp = sp->sw_vp;
  161         splx(s);
  162         BUF_STRATEGY(bp);
  163         return 0;
  164 }
  165 
  166 /*
  167  * Create a special vnode op vector for swapdev_vp - we only use
  168  * VOP_STRATEGY(), everything else returns an error.
  169  */
  170 vop_t **swapdev_vnodeop_p;
  171 static struct vnodeopv_entry_desc swapdev_vnodeop_entries[] = {  
  172         { &vop_default_desc,            (vop_t *) vop_defaultop },
  173         { &vop_strategy_desc,           (vop_t *) swapdev_strategy },
  174         { NULL, NULL }
  175 };
  176 static struct vnodeopv_desc swapdev_vnodeop_opv_desc =
  177         { &swapdev_vnodeop_p, swapdev_vnodeop_entries };
  178 
  179 VNODEOP_SET(swapdev_vnodeop_opv_desc);
  180 
  181 /*
  182  * System call swapon(name) enables swapping on device name,
  183  * which must be in the swdevsw.  Return EBUSY
  184  * if already swapping on this device.
  185  */
  186 #ifndef _SYS_SYSPROTO_H_
  187 struct swapon_args {
  188         char *name;
  189 };
  190 #endif
  191 
  192 /* 
  193  * MPSAFE
  194  */
  195 /* ARGSUSED */
  196 int
  197 swapon(td, uap)
  198         struct thread *td;
  199         struct swapon_args *uap;
  200 {
  201         struct vattr attr;
  202         struct vnode *vp;
  203         struct nameidata nd;
  204         int error;
  205 
  206         mtx_lock(&Giant);
  207         error = suser(td);
  208         if (error)
  209                 goto done2;
  210 
  211         /*
  212          * Swap metadata may not fit in the KVM if we have physical
  213          * memory of >1GB.
  214          */
  215         if (swap_zone == NULL) {
  216                 error = ENOMEM;
  217                 goto done2;
  218         }
  219 
  220         NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, uap->name, td);
  221         error = namei(&nd);
  222         if (error)
  223                 goto done2;
  224 
  225         NDFREE(&nd, NDF_ONLY_PNBUF);
  226         vp = nd.ni_vp;
  227 
  228         if (vn_isdisk(vp, &error))
  229                 error = swaponvp(td, vp, vp->v_rdev, 0);
  230         else if (vp->v_type == VREG &&
  231             (vp->v_mount->mnt_vfc->vfc_flags & VFCF_NETWORK) != 0 &&
  232             (error = VOP_GETATTR(vp, &attr, td->td_ucred, td)) == 0) {
  233                 /*
  234                  * Allow direct swapping to NFS regular files in the same
  235                  * way that nfs_mountroot() sets up diskless swapping.
  236                  */
  237                 error = swaponvp(td, vp, NODEV, attr.va_size / DEV_BSIZE);
  238         }
  239 
  240         if (error)
  241                 vrele(vp);
  242 done2:
  243         mtx_unlock(&Giant);
  244         return (error);
  245 }
  246 
  247 /*
  248  * Swfree(index) frees the index'th portion of the swap map.
  249  * Each of the nswdev devices provides 1/nswdev'th of the swap
  250  * space, which is laid out with blocks of dmmax pages circularly
  251  * among the devices.
  252  *
  253  * The new swap code uses page-sized blocks.  The old swap code used
  254  * DEV_BSIZE'd chunks.
  255  *
  256  * XXX locking when multiple swapon's run in parallel
  257  */
  258 int
  259 swaponvp(td, vp, dev, nblks)
  260         struct thread *td;
  261         struct vnode *vp;
  262         dev_t dev;
  263         u_long nblks;
  264 {
  265         int index;
  266         struct swdevt *sp;
  267         swblk_t vsbase;
  268         long blk;
  269         swblk_t dvbase;
  270         int error;
  271         u_long aligned_nblks;
  272 
  273         if (!swapdev_vp) {
  274                 error = getnewvnode("none", NULL, swapdev_vnodeop_p,
  275                     &swapdev_vp);
  276                 if (error)
  277                         panic("Cannot get vnode for swapdev");
  278                 swapdev_vp->v_type = VNON;      /* Untyped */
  279         }
  280 
  281         ASSERT_VOP_UNLOCKED(vp, "swaponvp");
  282         for (sp = swdevt, index = 0 ; index < nswdev; index++, sp++) {
  283                 if (sp->sw_vp == vp)
  284                         return EBUSY;
  285                 if (!sp->sw_vp)
  286                         goto found;
  287 
  288         }
  289         return EINVAL;
  290     found:
  291         (void) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
  292 #ifdef MAC
  293         error = mac_check_system_swapon(td->td_ucred, vp);
  294         if (error == 0)
  295 #endif
  296                 error = VOP_OPEN(vp, FREAD | FWRITE, td->td_ucred, td);
  297         (void) VOP_UNLOCK(vp, 0, td);
  298         if (error)
  299                 return (error);
  300 
  301         if (nblks == 0 && dev != NODEV && (devsw(dev)->d_psize == 0 ||
  302             (nblks = (*devsw(dev)->d_psize) (dev)) == -1)) {
  303                 (void) VOP_CLOSE(vp, FREAD | FWRITE, td->td_ucred, td);
  304                 return (ENXIO);
  305         }
  306         if (nblks == 0) {
  307                 (void) VOP_CLOSE(vp, FREAD | FWRITE, td->td_ucred, td);
  308                 return (ENXIO);
  309         }
  310 
  311         /*
  312          * If we go beyond this, we get overflows in the radix
  313          * tree bitmap code.
  314          */
  315         if (nblks > 0x40000000 / BLIST_META_RADIX / nswdev) {
  316                 printf("exceeded maximum of %d blocks per swap unit\n",
  317                         0x40000000 / BLIST_META_RADIX / nswdev);
  318                 (void) VOP_CLOSE(vp, FREAD | FWRITE, td->td_ucred, td);
  319                 return (ENXIO);
  320         }
  321         /*
  322          * nblks is in DEV_BSIZE'd chunks, convert to PAGE_SIZE'd chunks.
  323          * First chop nblks off to page-align it, then convert.
  324          * 
  325          * sw->sw_nblks is in page-sized chunks now too.
  326          */
  327         nblks &= ~(ctodb(1) - 1);
  328         nblks = dbtoc(nblks);
  329 
  330         sp->sw_vp = vp;
  331         sp->sw_dev = dev2udev(dev);
  332         sp->sw_device = dev;
  333         sp->sw_flags |= SW_FREED;
  334         sp->sw_nblks = nblks;
  335         sp->sw_used = 0;
  336 
  337         /*
  338          * nblks, nswap, and dmmax are PAGE_SIZE'd parameters now, not
  339          * DEV_BSIZE'd.   aligned_nblks is used to calculate the
  340          * size of the swap bitmap, taking into account the stripe size.
  341          */
  342         aligned_nblks = (nblks + (dmmax - 1)) & ~(u_long)(dmmax - 1);
  343 
  344         if (aligned_nblks * nswdev > nswap)
  345                 nswap = aligned_nblks * nswdev;
  346 
  347         if (swapblist == NULL)
  348                 swapblist = blist_create(nswap);
  349         else
  350                 blist_resize(&swapblist, nswap, 0);
  351 
  352         for (dvbase = dmmax; dvbase < nblks; dvbase += dmmax) {
  353                 blk = min(nblks - dvbase, dmmax);
  354                 vsbase = index * dmmax + dvbase * nswdev;
  355                 blist_free(swapblist, vsbase, blk);
  356                 vm_swap_size += blk;
  357         }
  358 
  359         return (0);
  360 }
  361 
  362 static int
  363 sysctl_vm_swap_info(SYSCTL_HANDLER_ARGS)
  364 {
  365         int     *name = (int *)arg1;
  366         int     error, i, n;
  367         struct xswdev xs;
  368         struct swdevt *sp;
  369 
  370         if (arg2 != 1) /* name length */
  371                 return (EINVAL);
  372 
  373         for (sp = swdevt, i = 0, n = 0 ; i < nswdev; i++, sp++) {
  374                 if (sp->sw_vp) {
  375                         if (n == *name) {
  376                                 xs.xsw_version = XSWDEV_VERSION;
  377                                 xs.xsw_dev = sp->sw_dev;
  378                                 xs.xsw_flags = sp->sw_flags;
  379                                 xs.xsw_nblks = sp->sw_nblks;
  380                                 xs.xsw_used = sp->sw_used;
  381 
  382                                 error = SYSCTL_OUT(req, &xs, sizeof(xs));
  383                                 return (error);
  384                         }
  385                         n++;
  386                 }
  387 
  388         }
  389         return (ENOENT);
  390 }
  391 
  392 SYSCTL_INT(_vm, OID_AUTO, nswapdev, CTLFLAG_RD, &nswdev, 0,
  393     "Number of swap devices");
  394 SYSCTL_NODE(_vm, OID_AUTO, swap_info, CTLFLAG_RD, sysctl_vm_swap_info,
  395     "Swap statistics by device");
Cache object: a748fbbcdbf6be36e2461bffa27d33a2
[ 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
