freebsd-dev/sys/kern/vfs_default.c
Matthew Dillon 4f467cb8c1 Fix incorrect double-termination of vm_object. When a vm_object is
terminated and flushes pending dirty pages it is possible for the
object to be ref'd (0->1) and then deref'd (1->0) during termination.
We do not terminate the object a second time.

Document vop_stdgetvobject() to explicitly allow it to be called without
the vnode interlock held (for upcoming sync_msync() and ffs_sync()
performance optimizations)

MFC after:	3 days
2001-10-23 01:23:41 +00:00

810 lines
18 KiB
C

/*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from software contributed
* to Berkeley by John Heidemann of the UCLA Ficus project.
*
* Source: * @(#)i405_init.c 2.10 92/04/27 UCLA Ficus project
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/unistd.h>
#include <sys/vnode.h>
#include <sys/poll.h>
#include <machine/limits.h>
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_page.h>
#include <vm/vm_pager.h>
#include <vm/vnode_pager.h>
#include <vm/vm_zone.h>
static int vop_nolookup __P((struct vop_lookup_args *));
static int vop_nostrategy __P((struct vop_strategy_args *));
/*
* This vnode table stores what we want to do if the filesystem doesn't
* implement a particular VOP.
*
* If there is no specific entry here, we will return EOPNOTSUPP.
*
*/
vop_t **default_vnodeop_p;
static struct vnodeopv_entry_desc default_vnodeop_entries[] = {
{ &vop_default_desc, (vop_t *) vop_eopnotsupp },
{ &vop_advlock_desc, (vop_t *) vop_einval },
{ &vop_bmap_desc, (vop_t *) vop_stdbmap },
{ &vop_close_desc, (vop_t *) vop_null },
{ &vop_createvobject_desc, (vop_t *) vop_stdcreatevobject },
{ &vop_destroyvobject_desc, (vop_t *) vop_stddestroyvobject },
{ &vop_fsync_desc, (vop_t *) vop_null },
{ &vop_getpages_desc, (vop_t *) vop_stdgetpages },
{ &vop_getvobject_desc, (vop_t *) vop_stdgetvobject },
{ &vop_inactive_desc, (vop_t *) vop_stdinactive },
{ &vop_ioctl_desc, (vop_t *) vop_enotty },
{ &vop_islocked_desc, (vop_t *) vop_noislocked },
{ &vop_lease_desc, (vop_t *) vop_null },
{ &vop_lock_desc, (vop_t *) vop_nolock },
{ &vop_lookup_desc, (vop_t *) vop_nolookup },
{ &vop_open_desc, (vop_t *) vop_null },
{ &vop_pathconf_desc, (vop_t *) vop_einval },
{ &vop_putpages_desc, (vop_t *) vop_stdputpages },
{ &vop_poll_desc, (vop_t *) vop_nopoll },
{ &vop_readlink_desc, (vop_t *) vop_einval },
{ &vop_revoke_desc, (vop_t *) vop_revoke },
{ &vop_strategy_desc, (vop_t *) vop_nostrategy },
{ &vop_unlock_desc, (vop_t *) vop_nounlock },
{ NULL, NULL }
};
static struct vnodeopv_desc default_vnodeop_opv_desc =
{ &default_vnodeop_p, default_vnodeop_entries };
VNODEOP_SET(default_vnodeop_opv_desc);
int
vop_eopnotsupp(struct vop_generic_args *ap)
{
/*
printf("vop_notsupp[%s]\n", ap->a_desc->vdesc_name);
*/
return (EOPNOTSUPP);
}
int
vop_ebadf(struct vop_generic_args *ap)
{
return (EBADF);
}
int
vop_enotty(struct vop_generic_args *ap)
{
return (ENOTTY);
}
int
vop_einval(struct vop_generic_args *ap)
{
return (EINVAL);
}
int
vop_null(struct vop_generic_args *ap)
{
return (0);
}
int
vop_defaultop(struct vop_generic_args *ap)
{
return (VOCALL(default_vnodeop_p, ap->a_desc->vdesc_offset, ap));
}
int
vop_panic(struct vop_generic_args *ap)
{
panic("filesystem goof: vop_panic[%s]", ap->a_desc->vdesc_name);
}
static int
vop_nolookup(ap)
struct vop_lookup_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
} */ *ap;
{
*ap->a_vpp = NULL;
return (ENOTDIR);
}
/*
* vop_nostrategy:
*
* Strategy routine for VFS devices that have none.
*
* BIO_ERROR and B_INVAL must be cleared prior to calling any strategy
* routine. Typically this is done for a BIO_READ strategy call.
* Typically B_INVAL is assumed to already be clear prior to a write
* and should not be cleared manually unless you just made the buffer
* invalid. BIO_ERROR should be cleared either way.
*/
static int
vop_nostrategy (struct vop_strategy_args *ap)
{
printf("No strategy for buffer at %p\n", ap->a_bp);
vprint("", ap->a_vp);
vprint("", ap->a_bp->b_vp);
ap->a_bp->b_ioflags |= BIO_ERROR;
ap->a_bp->b_error = EOPNOTSUPP;
bufdone(ap->a_bp);
return (EOPNOTSUPP);
}
int
vop_stdpathconf(ap)
struct vop_pathconf_args /* {
struct vnode *a_vp;
int a_name;
int *a_retval;
} */ *ap;
{
switch (ap->a_name) {
case _PC_LINK_MAX:
*ap->a_retval = LINK_MAX;
return (0);
case _PC_MAX_CANON:
*ap->a_retval = MAX_CANON;
return (0);
case _PC_MAX_INPUT:
*ap->a_retval = MAX_INPUT;
return (0);
case _PC_PIPE_BUF:
*ap->a_retval = PIPE_BUF;
return (0);
case _PC_CHOWN_RESTRICTED:
*ap->a_retval = 1;
return (0);
case _PC_VDISABLE:
*ap->a_retval = _POSIX_VDISABLE;
return (0);
default:
return (EINVAL);
}
/* NOTREACHED */
}
/*
* Standard lock, unlock and islocked functions.
*
* These depend on the lock structure being the first element in the
* inode, ie: vp->v_data points to the the lock!
*/
int
vop_stdlock(ap)
struct vop_lock_args /* {
struct vnode *a_vp;
int a_flags;
struct thread *a_td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
#ifndef DEBUG_LOCKS
return (lockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock, ap->a_td));
#else
return (debuglockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock,
ap->a_td, "vop_stdlock", vp->filename, vp->line));
#endif
}
int
vop_stdunlock(ap)
struct vop_unlock_args /* {
struct vnode *a_vp;
int a_flags;
struct thread *a_td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
return (lockmgr(&vp->v_lock, ap->a_flags | LK_RELEASE, &vp->v_interlock,
ap->a_td));
}
int
vop_stdislocked(ap)
struct vop_islocked_args /* {
struct vnode *a_vp;
struct thread *a_td;
} */ *ap;
{
return (lockstatus(&ap->a_vp->v_lock, ap->a_td));
}
int
vop_stdinactive(ap)
struct vop_inactive_args /* {
struct vnode *a_vp;
struct thread *a_td;
} */ *ap;
{
VOP_UNLOCK(ap->a_vp, 0, ap->a_td);
return (0);
}
/*
* Return true for select/poll.
*/
int
vop_nopoll(ap)
struct vop_poll_args /* {
struct vnode *a_vp;
int a_events;
struct ucred *a_cred;
struct thread *a_td;
} */ *ap;
{
/*
* Return true for read/write. If the user asked for something
* special, return POLLNVAL, so that clients have a way of
* determining reliably whether or not the extended
* functionality is present without hard-coding knowledge
* of specific filesystem implementations.
*/
if (ap->a_events & ~POLLSTANDARD)
return (POLLNVAL);
return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
}
/*
* Implement poll for local filesystems that support it.
*/
int
vop_stdpoll(ap)
struct vop_poll_args /* {
struct vnode *a_vp;
int a_events;
struct ucred *a_cred;
struct thread *a_td;
} */ *ap;
{
if (ap->a_events & ~POLLSTANDARD)
return (vn_pollrecord(ap->a_vp, ap->a_td, ap->a_events));
return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
}
/*
* Stubs to use when there is no locking to be done on the underlying object.
* A minimal shared lock is necessary to ensure that the underlying object
* is not revoked while an operation is in progress. So, an active shared
* count is maintained in an auxillary vnode lock structure.
*/
int
vop_sharedlock(ap)
struct vop_lock_args /* {
struct vnode *a_vp;
int a_flags;
struct thread *a_td;
} */ *ap;
{
/*
* This code cannot be used until all the non-locking filesystems
* (notably NFS) are converted to properly lock and release nodes.
* Also, certain vnode operations change the locking state within
* the operation (create, mknod, remove, link, rename, mkdir, rmdir,
* and symlink). Ideally these operations should not change the
* lock state, but should be changed to let the caller of the
* function unlock them. Otherwise all intermediate vnode layers
* (such as union, umapfs, etc) must catch these functions to do
* the necessary locking at their layer. Note that the inactive
* and lookup operations also change their lock state, but this
* cannot be avoided, so these two operations will always need
* to be handled in intermediate layers.
*/
struct vnode *vp = ap->a_vp;
int vnflags, flags = ap->a_flags;
switch (flags & LK_TYPE_MASK) {
case LK_DRAIN:
vnflags = LK_DRAIN;
break;
case LK_EXCLUSIVE:
#ifdef DEBUG_VFS_LOCKS
/*
* Normally, we use shared locks here, but that confuses
* the locking assertions.
*/
vnflags = LK_EXCLUSIVE;
break;
#endif
case LK_SHARED:
vnflags = LK_SHARED;
break;
case LK_UPGRADE:
case LK_EXCLUPGRADE:
case LK_DOWNGRADE:
return (0);
case LK_RELEASE:
default:
panic("vop_sharedlock: bad operation %d", flags & LK_TYPE_MASK);
}
if (flags & LK_INTERLOCK)
vnflags |= LK_INTERLOCK;
#ifndef DEBUG_LOCKS
return (lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_td));
#else
return (debuglockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_td,
"vop_sharedlock", vp->filename, vp->line));
#endif
}
/*
* Stubs to use when there is no locking to be done on the underlying object.
* A minimal shared lock is necessary to ensure that the underlying object
* is not revoked while an operation is in progress. So, an active shared
* count is maintained in an auxillary vnode lock structure.
*/
int
vop_nolock(ap)
struct vop_lock_args /* {
struct vnode *a_vp;
int a_flags;
struct thread *a_td;
} */ *ap;
{
#ifdef notyet
/*
* This code cannot be used until all the non-locking filesystems
* (notably NFS) are converted to properly lock and release nodes.
* Also, certain vnode operations change the locking state within
* the operation (create, mknod, remove, link, rename, mkdir, rmdir,
* and symlink). Ideally these operations should not change the
* lock state, but should be changed to let the caller of the
* function unlock them. Otherwise all intermediate vnode layers
* (such as union, umapfs, etc) must catch these functions to do
* the necessary locking at their layer. Note that the inactive
* and lookup operations also change their lock state, but this
* cannot be avoided, so these two operations will always need
* to be handled in intermediate layers.
*/
struct vnode *vp = ap->a_vp;
int vnflags, flags = ap->a_flags;
switch (flags & LK_TYPE_MASK) {
case LK_DRAIN:
vnflags = LK_DRAIN;
break;
case LK_EXCLUSIVE:
case LK_SHARED:
vnflags = LK_SHARED;
break;
case LK_UPGRADE:
case LK_EXCLUPGRADE:
case LK_DOWNGRADE:
return (0);
case LK_RELEASE:
default:
panic("vop_nolock: bad operation %d", flags & LK_TYPE_MASK);
}
if (flags & LK_INTERLOCK)
vnflags |= LK_INTERLOCK;
return(lockmgr(&vp->v_lock, vnflags, &vp->v_interlock, ap->a_td));
#else /* for now */
/*
* Since we are not using the lock manager, we must clear
* the interlock here.
*/
if (ap->a_flags & LK_INTERLOCK)
mtx_unlock(&ap->a_vp->v_interlock);
return (0);
#endif
}
/*
* Do the inverse of vop_nolock, handling the interlock in a compatible way.
*/
int
vop_nounlock(ap)
struct vop_unlock_args /* {
struct vnode *a_vp;
int a_flags;
struct thread *a_td;
} */ *ap;
{
/*
* Since we are not using the lock manager, we must clear
* the interlock here.
*/
if (ap->a_flags & LK_INTERLOCK)
mtx_unlock(&ap->a_vp->v_interlock);
return (0);
}
/*
* Return whether or not the node is in use.
*/
int
vop_noislocked(ap)
struct vop_islocked_args /* {
struct vnode *a_vp;
struct thread *a_td;
} */ *ap;
{
return (0);
}
/*
* Return our mount point, as we will take charge of the writes.
*/
int
vop_stdgetwritemount(ap)
struct vop_getwritemount_args /* {
struct vnode *a_vp;
struct mount **a_mpp;
} */ *ap;
{
*(ap->a_mpp) = ap->a_vp->v_mount;
return (0);
}
int
vop_stdcreatevobject(ap)
struct vop_createvobject_args /* {
struct vnode *vp;
struct ucred *cred;
struct thread *td;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
struct ucred *cred = ap->a_cred;
struct thread *td = ap->a_td;
struct vattr vat;
vm_object_t object;
int error = 0;
GIANT_REQUIRED;
if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE)
return (0);
retry:
if ((object = vp->v_object) == NULL) {
if (vp->v_type == VREG || vp->v_type == VDIR) {
if ((error = VOP_GETATTR(vp, &vat, cred, td)) != 0)
goto retn;
object = vnode_pager_alloc(vp, vat.va_size, 0, 0);
} else if (devsw(vp->v_rdev) != NULL) {
/*
* This simply allocates the biggest object possible
* for a disk vnode. This should be fixed, but doesn't
* cause any problems (yet).
*/
object = vnode_pager_alloc(vp, IDX_TO_OFF(INT_MAX), 0, 0);
} else {
goto retn;
}
/*
* Dereference the reference we just created. This assumes
* that the object is associated with the vp.
*/
object->ref_count--;
vp->v_usecount--;
} else {
if (object->flags & OBJ_DEAD) {
VOP_UNLOCK(vp, 0, td);
tsleep(object, PVM, "vodead", 0);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
goto retry;
}
}
KASSERT(vp->v_object != NULL, ("vfs_object_create: NULL object"));
vp->v_flag |= VOBJBUF;
retn:
return (error);
}
int
vop_stddestroyvobject(ap)
struct vop_destroyvobject_args /* {
struct vnode *vp;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
vm_object_t obj = vp->v_object;
GIANT_REQUIRED;
if (vp->v_object == NULL)
return (0);
if (obj->ref_count == 0) {
/*
* vclean() may be called twice. The first time
* removes the primary reference to the object,
* the second time goes one further and is a
* special-case to terminate the object.
*
* don't double-terminate the object
*/
if ((obj->flags & OBJ_DEAD) == 0)
vm_object_terminate(obj);
} else {
/*
* Woe to the process that tries to page now :-).
*/
vm_pager_deallocate(obj);
}
return (0);
}
/*
* Return the underlying VM object. This routine may be called with or
* without the vnode interlock held. If called without, the returned
* object is not guarenteed to be valid. The syncer typically gets the
* object without holding the interlock in order to quickly test whether
* it might be dirty before going heavy-weight. vm_object's use zalloc
* and thus stable-storage, so this is safe.
*/
int
vop_stdgetvobject(ap)
struct vop_getvobject_args /* {
struct vnode *vp;
struct vm_object **objpp;
} */ *ap;
{
struct vnode *vp = ap->a_vp;
struct vm_object **objpp = ap->a_objpp;
if (objpp)
*objpp = vp->v_object;
return (vp->v_object ? 0 : EINVAL);
}
int
vop_stdbmap(ap)
struct vop_bmap_args /* {
struct vnode *a_vp;
daddr_t a_bn;
struct vnode **a_vpp;
daddr_t *a_bnp;
int *a_runp;
int *a_runb;
} */ *ap;
{
if (ap->a_vpp != NULL)
*ap->a_vpp = ap->a_vp;
if (ap->a_bnp != NULL)
*ap->a_bnp = ap->a_bn * btodb(ap->a_vp->v_mount->mnt_stat.f_iosize);
if (ap->a_runp != NULL)
*ap->a_runp = 0;
if (ap->a_runb != NULL)
*ap->a_runb = 0;
return (0);
}
int
vop_stdgetpages(ap)
struct vop_getpages_args /* {
struct vnode *a_vp;
vm_page_t *a_m;
int a_count;
int a_reqpage;
vm_ooffset_t a_offset;
} */ *ap;
{
return vnode_pager_generic_getpages(ap->a_vp, ap->a_m,
ap->a_count, ap->a_reqpage);
}
int
vop_stdputpages(ap)
struct vop_putpages_args /* {
struct vnode *a_vp;
vm_page_t *a_m;
int a_count;
int a_sync;
int *a_rtvals;
vm_ooffset_t a_offset;
} */ *ap;
{
return vnode_pager_generic_putpages(ap->a_vp, ap->a_m, ap->a_count,
ap->a_sync, ap->a_rtvals);
}
/*
* vfs default ops
* used to fill the vfs fucntion table to get reasonable default return values.
*/
int
vfs_stdmount (mp, path, data, ndp, td)
struct mount *mp;
char *path;
caddr_t data;
struct nameidata *ndp;
struct thread *td;
{
return (0);
}
int
vfs_stdunmount (mp, mntflags, td)
struct mount *mp;
int mntflags;
struct thread *td;
{
return (0);
}
int
vfs_stdroot (mp, vpp)
struct mount *mp;
struct vnode **vpp;
{
return (EOPNOTSUPP);
}
int
vfs_stdstatfs (mp, sbp, td)
struct mount *mp;
struct statfs *sbp;
struct thread *td;
{
return (EOPNOTSUPP);
}
int
vfs_stdvptofh (vp, fhp)
struct vnode *vp;
struct fid *fhp;
{
return (EOPNOTSUPP);
}
int
vfs_stdstart (mp, flags, td)
struct mount *mp;
int flags;
struct thread *td;
{
return (0);
}
int
vfs_stdquotactl (mp, cmds, uid, arg, td)
struct mount *mp;
int cmds;
uid_t uid;
caddr_t arg;
struct thread *td;
{
return (EOPNOTSUPP);
}
int
vfs_stdsync (mp, waitfor, cred, td)
struct mount *mp;
int waitfor;
struct ucred *cred;
struct thread *td;
{
return (0);
}
int
vfs_stdvget (mp, ino, vpp)
struct mount *mp;
ino_t ino;
struct vnode **vpp;
{
return (EOPNOTSUPP);
}
int
vfs_stdfhtovp (mp, fhp, vpp)
struct mount *mp;
struct fid *fhp;
struct vnode **vpp;
{
return (EOPNOTSUPP);
}
int
vfs_stdinit (vfsp)
struct vfsconf *vfsp;
{
return (0);
}
int
vfs_stduninit (vfsp)
struct vfsconf *vfsp;
{
return(0);
}
int
vfs_stdextattrctl(mp, cmd, filename_vp, attrnamespace, attrname, td)
struct mount *mp;
int cmd;
struct vnode *filename_vp;
int attrnamespace;
const char *attrname;
struct thread *td;
{
return(EOPNOTSUPP);
}
/* end of vfs default ops */