Index: head/sys/fs/devfs/devfs_vnops.c =================================================================== --- head/sys/fs/devfs/devfs_vnops.c (revision 171598) +++ head/sys/fs/devfs/devfs_vnops.c (revision 171599) @@ -1,1403 +1,1404 @@ /*- * Copyright (c) 2000-2004 * Poul-Henning Kamp. All rights reserved. * Copyright (c) 1989, 1992-1993, 1995 * The Regents of the University of California. All rights reserved. * * This code is derived from software donated to Berkeley by * Jan-Simon Pendry. * * 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. 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. * * @(#)kernfs_vnops.c 8.15 (Berkeley) 5/21/95 * From: FreeBSD: src/sys/miscfs/kernfs/kernfs_vnops.c 1.43 * * $FreeBSD$ */ /* * TODO: * remove empty directories * mkdir: want it ? */ #include "opt_mac.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static struct vop_vector devfs_vnodeops; static struct vop_vector devfs_specops; static struct fileops devfs_ops_f; #include #include #include struct mtx devfs_de_interlock; MTX_SYSINIT(devfs_de_interlock, &devfs_de_interlock, "devfs interlock", MTX_DEF); struct sx clone_drain_lock; SX_SYSINIT(clone_drain_lock, &clone_drain_lock, "clone events drain lock"); static int devfs_fp_check(struct file *fp, struct cdev **devp, struct cdevsw **dswp) { *dswp = devvn_refthread(fp->f_vnode, devp); if (*devp != fp->f_data) { if (*dswp != NULL) dev_relthread(*devp); return (ENXIO); } KASSERT((*devp)->si_refcount > 0, ("devfs: un-referenced struct cdev *(%s)", devtoname(*devp))); if (*dswp == NULL) return (ENXIO); return (0); } /* * Construct the fully qualified path name relative to the mountpoint */ static char * devfs_fqpn(char *buf, struct vnode *dvp, struct componentname *cnp) { int i; struct devfs_dirent *de, *dd; struct devfs_mount *dmp; dmp = VFSTODEVFS(dvp->v_mount); dd = dvp->v_data; i = SPECNAMELEN; buf[i] = '\0'; i -= cnp->cn_namelen; if (i < 0) return (NULL); bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen); de = dd; while (de != dmp->dm_rootdir) { i--; if (i < 0) return (NULL); buf[i] = '/'; i -= de->de_dirent->d_namlen; if (i < 0) return (NULL); bcopy(de->de_dirent->d_name, buf + i, de->de_dirent->d_namlen); de = TAILQ_FIRST(&de->de_dlist); /* "." */ de = TAILQ_NEXT(de, de_list); /* ".." */ de = de->de_dir; } return (buf + i); } static int devfs_allocv_drop_refs(int drop_dm_lock, struct devfs_mount *dmp, struct devfs_dirent *de) { int not_found; not_found = 0; if (de->de_flags & DE_DOOMED) not_found = 1; if (DEVFS_DE_DROP(de)) { KASSERT(not_found == 1, ("DEVFS de dropped but not doomed")); devfs_dirent_free(de); } if (DEVFS_DMP_DROP(dmp)) { KASSERT(not_found == 1, ("DEVFS mount struct freed before dirent")); not_found = 2; sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); } if (not_found == 1 || (drop_dm_lock && not_found != 2)) sx_unlock(&dmp->dm_lock); return (not_found); } static void devfs_insmntque_dtr(struct vnode *vp, void *arg) { struct devfs_dirent *de; de = (struct devfs_dirent *)arg; mtx_lock(&devfs_de_interlock); vp->v_data = NULL; de->de_vnode = NULL; mtx_unlock(&devfs_de_interlock); vgone(vp); vput(vp); } /* * devfs_allocv shall be entered with dmp->dm_lock held, and it drops * it on return. */ int devfs_allocv(struct devfs_dirent *de, struct mount *mp, struct vnode **vpp, struct thread *td) { int error; struct vnode *vp; struct cdev *dev; struct devfs_mount *dmp; KASSERT(td == curthread, ("devfs_allocv: td != curthread")); dmp = VFSTODEVFS(mp); if (de->de_flags & DE_DOOMED) { sx_xunlock(&dmp->dm_lock); return (ENOENT); } loop: DEVFS_DE_HOLD(de); DEVFS_DMP_HOLD(dmp); mtx_lock(&devfs_de_interlock); vp = de->de_vnode; if (vp != NULL) { VI_LOCK(vp); mtx_unlock(&devfs_de_interlock); sx_xunlock(&dmp->dm_lock); error = vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td); sx_xlock(&dmp->dm_lock); if (devfs_allocv_drop_refs(0, dmp, de)) { if (error == 0) vput(vp); return (ENOENT); } else if (error) goto loop; sx_xunlock(&dmp->dm_lock); *vpp = vp; return (0); } mtx_unlock(&devfs_de_interlock); if (de->de_dirent->d_type == DT_CHR) { if (!(de->de_cdp->cdp_flags & CDP_ACTIVE)) { devfs_allocv_drop_refs(1, dmp, de); return (ENOENT); } dev = &de->de_cdp->cdp_c; } else { dev = NULL; } error = getnewvnode("devfs", mp, &devfs_vnodeops, &vp); if (error != 0) { devfs_allocv_drop_refs(1, dmp, de); printf("devfs_allocv: failed to allocate new vnode\n"); return (error); } if (de->de_dirent->d_type == DT_CHR) { vp->v_type = VCHR; VI_LOCK(vp); dev_lock(); dev_refl(dev); + /* XXX: v_rdev should be protect by vnode lock */ vp->v_rdev = dev; KASSERT(vp->v_usecount == 1, ("%s %d (%d)\n", __func__, __LINE__, vp->v_usecount)); dev->si_usecount += vp->v_usecount; dev_unlock(); VI_UNLOCK(vp); vp->v_op = &devfs_specops; } else if (de->de_dirent->d_type == DT_DIR) { vp->v_type = VDIR; } else if (de->de_dirent->d_type == DT_LNK) { vp->v_type = VLNK; } else { vp->v_type = VBAD; } vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); mtx_lock(&devfs_de_interlock); vp->v_data = de; de->de_vnode = vp; mtx_unlock(&devfs_de_interlock); error = insmntque1(vp, mp, devfs_insmntque_dtr, de); if (error != 0) { (void) devfs_allocv_drop_refs(1, dmp, de); return (error); } if (devfs_allocv_drop_refs(0, dmp, de)) { vput(vp); return (ENOENT); } #ifdef MAC mac_associate_vnode_devfs(mp, de, vp); #endif sx_xunlock(&dmp->dm_lock); *vpp = vp; return (0); } static int devfs_access(struct vop_access_args *ap) { struct vnode *vp = ap->a_vp; struct devfs_dirent *de; int error; de = vp->v_data; if (vp->v_type == VDIR) de = de->de_dir; error = vaccess(vp->v_type, de->de_mode, de->de_uid, de->de_gid, ap->a_mode, ap->a_cred, NULL); if (!error) return (error); if (error != EACCES) return (error); /* We do, however, allow access to the controlling terminal */ if (!(ap->a_td->td_proc->p_flag & P_CONTROLT)) return (error); if (ap->a_td->td_proc->p_session->s_ttyvp == de->de_vnode) return (0); return (error); } /* ARGSUSED */ static int devfs_advlock(struct vop_advlock_args *ap) { return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL); } /* ARGSUSED */ static int devfs_close(struct vop_close_args *ap) { struct vnode *vp = ap->a_vp, *oldvp; struct thread *td = ap->a_td; struct cdev *dev = vp->v_rdev; struct cdevsw *dsw; int vp_locked, error; /* * Hack: a tty device that is a controlling terminal * has a reference from the session structure. * We cannot easily tell that a character device is * a controlling terminal, unless it is the closing * process' controlling terminal. In that case, * if the reference count is 2 (this last descriptor * plus the session), release the reference from the session. */ oldvp = NULL; sx_xlock(&proctree_lock); if (td && vp == td->td_proc->p_session->s_ttyvp) { SESS_LOCK(td->td_proc->p_session); VI_LOCK(vp); if (count_dev(dev) == 2 && (vp->v_iflag & VI_DOOMED) == 0) { td->td_proc->p_session->s_ttyvp = NULL; oldvp = vp; } VI_UNLOCK(vp); SESS_UNLOCK(td->td_proc->p_session); } sx_xunlock(&proctree_lock); if (oldvp != NULL) vrele(oldvp); /* * We do not want to really close the device if it * is still in use unless we are trying to close it * forcibly. Since every use (buffer, vnode, swap, cmap) * holds a reference to the vnode, and because we mark * any other vnodes that alias this device, when the * sum of the reference counts on all the aliased * vnodes descends to one, we are on last close. */ dsw = dev_refthread(dev); if (dsw == NULL) return (ENXIO); VI_LOCK(vp); if (vp->v_iflag & VI_DOOMED) { /* Forced close. */ } else if (dsw->d_flags & D_TRACKCLOSE) { /* Keep device updated on status. */ } else if (count_dev(dev) > 1) { VI_UNLOCK(vp); dev_relthread(dev); return (0); } vholdl(vp); VI_UNLOCK(vp); vp_locked = VOP_ISLOCKED(vp, td); VOP_UNLOCK(vp, 0, td); KASSERT(dev->si_refcount > 0, ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev))); if (!(dsw->d_flags & D_NEEDGIANT)) { DROP_GIANT(); error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td); PICKUP_GIANT(); } else { error = dsw->d_close(dev, ap->a_fflag, S_IFCHR, td); } dev_relthread(dev); vn_lock(vp, vp_locked | LK_RETRY, td); vdrop(vp); return (error); } static int devfs_close_f(struct file *fp, struct thread *td) { return (vnops.fo_close(fp, td)); } /* ARGSUSED */ static int devfs_fsync(struct vop_fsync_args *ap) { if (!vn_isdisk(ap->a_vp, NULL)) return (0); return (vop_stdfsync(ap)); } static int devfs_getattr(struct vop_getattr_args *ap) { struct vnode *vp = ap->a_vp; struct vattr *vap = ap->a_vap; int error = 0; struct devfs_dirent *de; struct cdev *dev; de = vp->v_data; KASSERT(de != NULL, ("Null dirent in devfs_getattr vp=%p", vp)); if (vp->v_type == VDIR) { de = de->de_dir; KASSERT(de != NULL, ("Null dir dirent in devfs_getattr vp=%p", vp)); } bzero((caddr_t) vap, sizeof(*vap)); vattr_null(vap); vap->va_uid = de->de_uid; vap->va_gid = de->de_gid; vap->va_mode = de->de_mode; if (vp->v_type == VLNK) vap->va_size = strlen(de->de_symlink); else if (vp->v_type == VDIR) vap->va_size = vap->va_bytes = DEV_BSIZE; else vap->va_size = 0; if (vp->v_type != VDIR) vap->va_bytes = 0; vap->va_blocksize = DEV_BSIZE; vap->va_type = vp->v_type; #define fix(aa) \ do { \ if ((aa).tv_sec <= 3600) { \ (aa).tv_sec = boottime.tv_sec; \ (aa).tv_nsec = boottime.tv_usec * 1000; \ } \ } while (0) if (vp->v_type != VCHR) { fix(de->de_atime); vap->va_atime = de->de_atime; fix(de->de_mtime); vap->va_mtime = de->de_mtime; fix(de->de_ctime); vap->va_ctime = de->de_ctime; } else { dev = vp->v_rdev; fix(dev->si_atime); vap->va_atime = dev->si_atime; fix(dev->si_mtime); vap->va_mtime = dev->si_mtime; fix(dev->si_ctime); vap->va_ctime = dev->si_ctime; vap->va_rdev = dev->si_priv->cdp_inode; } vap->va_gen = 0; vap->va_flags = 0; vap->va_nlink = de->de_links; vap->va_fileid = de->de_inode; return (error); } /* ARGSUSED */ static int devfs_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td) { struct cdev *dev; struct cdevsw *dsw; struct vnode *vp; struct vnode *vpold; int error, i; const char *p; struct fiodgname_arg *fgn; error = devfs_fp_check(fp, &dev, &dsw); if (error) return (error); if (com == FIODTYPE) { *(int *)data = dsw->d_flags & D_TYPEMASK; dev_relthread(dev); return (0); } else if (com == FIODGNAME) { fgn = data; p = devtoname(dev); i = strlen(p) + 1; if (i > fgn->len) error = EINVAL; else error = copyout(p, fgn->buf, i); dev_relthread(dev); return (error); } error = dsw->d_ioctl(dev, com, data, fp->f_flag, td); dev_relthread(dev); if (error == ENOIOCTL) error = ENOTTY; if (error == 0 && com == TIOCSCTTY) { vp = fp->f_vnode; /* Do nothing if reassigning same control tty */ sx_slock(&proctree_lock); if (td->td_proc->p_session->s_ttyvp == vp) { sx_sunlock(&proctree_lock); return (0); } mtx_lock(&Giant); /* XXX TTY */ vpold = td->td_proc->p_session->s_ttyvp; VREF(vp); SESS_LOCK(td->td_proc->p_session); td->td_proc->p_session->s_ttyvp = vp; SESS_UNLOCK(td->td_proc->p_session); sx_sunlock(&proctree_lock); /* Get rid of reference to old control tty */ if (vpold) vrele(vpold); mtx_unlock(&Giant); /* XXX TTY */ } return (error); } /* ARGSUSED */ static int devfs_kqfilter_f(struct file *fp, struct knote *kn) { struct cdev *dev; struct cdevsw *dsw; int error; error = devfs_fp_check(fp, &dev, &dsw); if (error) return (error); error = dsw->d_kqfilter(dev, kn); dev_relthread(dev); return (error); } static int devfs_lookupx(struct vop_lookup_args *ap, int *dm_unlock) { struct componentname *cnp; struct vnode *dvp, **vpp; struct thread *td; struct devfs_dirent *de, *dd; struct devfs_dirent **dde; struct devfs_mount *dmp; struct cdev *cdev; int error, flags, nameiop; char specname[SPECNAMELEN + 1], *pname; cnp = ap->a_cnp; vpp = ap->a_vpp; dvp = ap->a_dvp; pname = cnp->cn_nameptr; td = cnp->cn_thread; flags = cnp->cn_flags; nameiop = cnp->cn_nameiop; dmp = VFSTODEVFS(dvp->v_mount); dd = dvp->v_data; *vpp = NULLVP; if ((flags & ISLASTCN) && nameiop == RENAME) return (EOPNOTSUPP); if (dvp->v_type != VDIR) return (ENOTDIR); if ((flags & ISDOTDOT) && (dvp->v_vflag & VV_ROOT)) return (EIO); error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, td); if (error) return (error); if (cnp->cn_namelen == 1 && *pname == '.') { if ((flags & ISLASTCN) && nameiop != LOOKUP) return (EINVAL); *vpp = dvp; VREF(dvp); return (0); } if (flags & ISDOTDOT) { if ((flags & ISLASTCN) && nameiop != LOOKUP) return (EINVAL); VOP_UNLOCK(dvp, 0, td); de = TAILQ_FIRST(&dd->de_dlist); /* "." */ de = TAILQ_NEXT(de, de_list); /* ".." */ de = de->de_dir; error = devfs_allocv(de, dvp->v_mount, vpp, td); *dm_unlock = 0; vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, td); return (error); } DEVFS_DMP_HOLD(dmp); devfs_populate(dmp); if (DEVFS_DMP_DROP(dmp)) { *dm_unlock = 0; sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); return (ENOENT); } dd = dvp->v_data; de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen); while (de == NULL) { /* While(...) so we can use break */ if (nameiop == DELETE) return (ENOENT); /* * OK, we didn't have an entry for the name we were asked for * so we try to see if anybody can create it on demand. */ pname = devfs_fqpn(specname, dvp, cnp); if (pname == NULL) break; cdev = NULL; DEVFS_DMP_HOLD(dmp); sx_xunlock(&dmp->dm_lock); sx_slock(&clone_drain_lock); EVENTHANDLER_INVOKE(dev_clone, td->td_ucred, pname, strlen(pname), &cdev); sx_sunlock(&clone_drain_lock); sx_xlock(&dmp->dm_lock); if (DEVFS_DMP_DROP(dmp)) { *dm_unlock = 0; sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); return (ENOENT); } if (cdev == NULL) break; DEVFS_DMP_HOLD(dmp); devfs_populate(dmp); if (DEVFS_DMP_DROP(dmp)) { *dm_unlock = 0; sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); return (ENOENT); } dev_lock(); dde = &cdev->si_priv->cdp_dirents[dmp->dm_idx]; if (dde != NULL && *dde != NULL) de = *dde; dev_unlock(); dev_rel(cdev); break; } if (de == NULL || de->de_flags & DE_WHITEOUT) { if ((nameiop == CREATE || nameiop == RENAME) && (flags & (LOCKPARENT | WANTPARENT)) && (flags & ISLASTCN)) { cnp->cn_flags |= SAVENAME; return (EJUSTRETURN); } return (ENOENT); } if ((cnp->cn_nameiop == DELETE) && (flags & ISLASTCN)) { error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred, td); if (error) return (error); if (*vpp == dvp) { VREF(dvp); *vpp = dvp; return (0); } } error = devfs_allocv(de, dvp->v_mount, vpp, td); *dm_unlock = 0; return (error); } static int devfs_lookup(struct vop_lookup_args *ap) { int j; struct devfs_mount *dmp; int dm_unlock; dmp = VFSTODEVFS(ap->a_dvp->v_mount); dm_unlock = 1; sx_xlock(&dmp->dm_lock); j = devfs_lookupx(ap, &dm_unlock); if (dm_unlock == 1) sx_xunlock(&dmp->dm_lock); return (j); } static int devfs_mknod(struct vop_mknod_args *ap) { struct componentname *cnp; struct vnode *dvp, **vpp; struct thread *td; struct devfs_dirent *dd, *de; struct devfs_mount *dmp; int error; /* * The only type of node we should be creating here is a * character device, for anything else return EOPNOTSUPP. */ if (ap->a_vap->va_type != VCHR) return (EOPNOTSUPP); dvp = ap->a_dvp; dmp = VFSTODEVFS(dvp->v_mount); cnp = ap->a_cnp; vpp = ap->a_vpp; td = cnp->cn_thread; dd = dvp->v_data; error = ENOENT; sx_xlock(&dmp->dm_lock); TAILQ_FOREACH(de, &dd->de_dlist, de_list) { if (cnp->cn_namelen != de->de_dirent->d_namlen) continue; if (bcmp(cnp->cn_nameptr, de->de_dirent->d_name, de->de_dirent->d_namlen) != 0) continue; if (de->de_flags & DE_WHITEOUT) break; goto notfound; } if (de == NULL) goto notfound; de->de_flags &= ~DE_WHITEOUT; error = devfs_allocv(de, dvp->v_mount, vpp, td); return (error); notfound: sx_xunlock(&dmp->dm_lock); return (error); } /* ARGSUSED */ static int devfs_open(struct vop_open_args *ap) { struct thread *td = ap->a_td; struct vnode *vp = ap->a_vp; struct cdev *dev = vp->v_rdev; struct file *fp = ap->a_fp; int error; struct cdevsw *dsw; if (vp->v_type == VBLK) return (ENXIO); if (dev == NULL) return (ENXIO); /* Make this field valid before any I/O in d_open. */ if (dev->si_iosize_max == 0) dev->si_iosize_max = DFLTPHYS; dsw = dev_refthread(dev); if (dsw == NULL) return (ENXIO); /* XXX: Special casing of ttys for deadfs. Probably redundant. */ if (dsw->d_flags & D_TTY) vp->v_vflag |= VV_ISTTY; VOP_UNLOCK(vp, 0, td); if(!(dsw->d_flags & D_NEEDGIANT)) { DROP_GIANT(); if (dsw->d_fdopen != NULL) error = dsw->d_fdopen(dev, ap->a_mode, td, fp); else error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td); PICKUP_GIANT(); } else { if (dsw->d_fdopen != NULL) error = dsw->d_fdopen(dev, ap->a_mode, td, fp); else error = dsw->d_open(dev, ap->a_mode, S_IFCHR, td); } vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); dev_relthread(dev); if (error) return (error); #if 0 /* /dev/console */ KASSERT(fp != NULL, ("Could not vnode bypass device on NULL fp")); #else if(fp == NULL) return (error); #endif FILE_LOCK(fp); KASSERT(fp->f_ops == &badfileops, ("Could not vnode bypass device on fdops %p", fp->f_ops)); fp->f_data = dev; fp->f_ops = &devfs_ops_f; FILE_UNLOCK(fp); return (error); } static int devfs_pathconf(struct vop_pathconf_args *ap) { switch (ap->a_name) { case _PC_MAC_PRESENT: #ifdef MAC /* * If MAC is enabled, devfs automatically supports * trivial non-persistant label storage. */ *ap->a_retval = 1; #else *ap->a_retval = 0; #endif return (0); default: return (vop_stdpathconf(ap)); } /* NOTREACHED */ } /* ARGSUSED */ static int devfs_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td) { struct cdev *dev; struct cdevsw *dsw; int error; error = devfs_fp_check(fp, &dev, &dsw); if (error) return (error); error = dsw->d_poll(dev, events, td); dev_relthread(dev); return(error); } /* * Print out the contents of a special device vnode. */ static int devfs_print(struct vop_print_args *ap) { printf("\tdev %s\n", devtoname(ap->a_vp->v_rdev)); return (0); } /* ARGSUSED */ static int devfs_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td) { struct cdev *dev; int ioflag, error, resid; struct cdevsw *dsw; error = devfs_fp_check(fp, &dev, &dsw); if (error) return (error); resid = uio->uio_resid; ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT); if (ioflag & O_DIRECT) ioflag |= IO_DIRECT; if ((flags & FOF_OFFSET) == 0) uio->uio_offset = fp->f_offset; error = dsw->d_read(dev, uio, ioflag); if (uio->uio_resid != resid || (error == 0 && resid != 0)) vfs_timestamp(&dev->si_atime); dev_relthread(dev); if ((flags & FOF_OFFSET) == 0) fp->f_offset = uio->uio_offset; fp->f_nextoff = uio->uio_offset; return (error); } static int devfs_readdir(struct vop_readdir_args *ap) { int error; struct uio *uio; struct dirent *dp; struct devfs_dirent *dd; struct devfs_dirent *de; struct devfs_mount *dmp; off_t off, oldoff; int *tmp_ncookies = NULL; if (ap->a_vp->v_type != VDIR) return (ENOTDIR); uio = ap->a_uio; if (uio->uio_offset < 0) return (EINVAL); /* * XXX: This is a temporary hack to get around this filesystem not * supporting cookies. We store the location of the ncookies pointer * in a temporary variable before calling vfs_subr.c:vfs_read_dirent() * and set the number of cookies to 0. We then set the pointer to * NULL so that vfs_read_dirent doesn't try to call realloc() on * ap->a_cookies. Later in this function, we restore the ap->a_ncookies * pointer to its original location before returning to the caller. */ if (ap->a_ncookies != NULL) { tmp_ncookies = ap->a_ncookies; *ap->a_ncookies = 0; ap->a_ncookies = NULL; } dmp = VFSTODEVFS(ap->a_vp->v_mount); sx_xlock(&dmp->dm_lock); DEVFS_DMP_HOLD(dmp); devfs_populate(dmp); if (DEVFS_DMP_DROP(dmp)) { sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); if (tmp_ncookies != NULL) ap->a_ncookies = tmp_ncookies; return (EIO); } error = 0; de = ap->a_vp->v_data; off = 0; oldoff = uio->uio_offset; TAILQ_FOREACH(dd, &de->de_dlist, de_list) { KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__)); if (dd->de_flags & DE_WHITEOUT) continue; if (dd->de_dirent->d_type == DT_DIR) de = dd->de_dir; else de = dd; dp = dd->de_dirent; if (dp->d_reclen > uio->uio_resid) break; dp->d_fileno = de->de_inode; if (off >= uio->uio_offset) { error = vfs_read_dirent(ap, dp, off); if (error) break; } off += dp->d_reclen; } sx_xunlock(&dmp->dm_lock); uio->uio_offset = off; /* * Restore ap->a_ncookies if it wasn't originally NULL in the first * place. */ if (tmp_ncookies != NULL) ap->a_ncookies = tmp_ncookies; return (error); } static int devfs_readlink(struct vop_readlink_args *ap) { struct devfs_dirent *de; de = ap->a_vp->v_data; return (uiomove(de->de_symlink, strlen(de->de_symlink), ap->a_uio)); } static int devfs_reclaim(struct vop_reclaim_args *ap) { struct vnode *vp = ap->a_vp; struct devfs_dirent *de; struct cdev *dev; mtx_lock(&devfs_de_interlock); de = vp->v_data; if (de != NULL) { de->de_vnode = NULL; vp->v_data = NULL; } mtx_unlock(&devfs_de_interlock); vnode_destroy_vobject(vp); dev_lock(); dev = vp->v_rdev; vp->v_rdev = NULL; if (dev == NULL) { dev_unlock(); return (0); } dev->si_usecount -= vp->v_usecount; dev_unlock(); dev_rel(dev); return (0); } static int devfs_remove(struct vop_remove_args *ap) { struct vnode *vp = ap->a_vp; struct devfs_dirent *dd; struct devfs_dirent *de; struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount); sx_xlock(&dmp->dm_lock); dd = ap->a_dvp->v_data; de = vp->v_data; if (de->de_cdp == NULL) { TAILQ_REMOVE(&dd->de_dlist, de, de_list); devfs_delete(dmp, de, 1); } else { de->de_flags |= DE_WHITEOUT; } sx_xunlock(&dmp->dm_lock); return (0); } /* * Revoke is called on a tty when a terminal session ends. The vnode * is orphaned by setting v_op to deadfs so we need to let go of it * as well so that we create a new one next time around. * */ static int devfs_revoke(struct vop_revoke_args *ap) { struct vnode *vp = ap->a_vp, *vp2; struct cdev *dev; struct cdev_priv *cdp; struct devfs_dirent *de; int i; KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL")); dev = vp->v_rdev; cdp = dev->si_priv; dev_lock(); cdp->cdp_inuse++; dev_unlock(); vhold(vp); vgone(vp); vdrop(vp); VOP_UNLOCK(vp,0,curthread); loop: for (;;) { mtx_lock(&devfs_de_interlock); dev_lock(); vp2 = NULL; for (i = 0; i <= cdp->cdp_maxdirent; i++) { de = cdp->cdp_dirents[i]; if (de == NULL) continue; vp2 = de->de_vnode; if (vp2 != NULL) { dev_unlock(); VI_LOCK(vp2); mtx_unlock(&devfs_de_interlock); if (vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK, curthread)) goto loop; vhold(vp2); vgone(vp2); vdrop(vp2); vput(vp2); break; } } if (vp2 != NULL) { continue; } dev_unlock(); mtx_unlock(&devfs_de_interlock); break; } dev_lock(); cdp->cdp_inuse--; if (!(cdp->cdp_flags & CDP_ACTIVE) && cdp->cdp_inuse == 0) { TAILQ_REMOVE(&cdevp_list, cdp, cdp_list); dev_unlock(); dev_rel(&cdp->cdp_c); } else dev_unlock(); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, curthread); return (0); } static int devfs_rioctl(struct vop_ioctl_args *ap) { int error; struct devfs_mount *dmp; dmp = VFSTODEVFS(ap->a_vp->v_mount); sx_xlock(&dmp->dm_lock); DEVFS_DMP_HOLD(dmp); devfs_populate(dmp); if (DEVFS_DMP_DROP(dmp)) { sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); return (ENOENT); } error = devfs_rules_ioctl(dmp, ap->a_command, ap->a_data, ap->a_td); sx_xunlock(&dmp->dm_lock); return (error); } static int devfs_rread(struct vop_read_args *ap) { if (ap->a_vp->v_type != VDIR) return (EINVAL); return (VOP_READDIR(ap->a_vp, ap->a_uio, ap->a_cred, NULL, NULL, NULL)); } static int devfs_setattr(struct vop_setattr_args *ap) { struct devfs_dirent *de; struct vattr *vap; struct vnode *vp; int c, error; uid_t uid; gid_t gid; vap = ap->a_vap; vp = ap->a_vp; if ((vap->va_type != VNON) || (vap->va_nlink != VNOVAL) || (vap->va_fsid != VNOVAL) || (vap->va_fileid != VNOVAL) || (vap->va_blocksize != VNOVAL) || (vap->va_flags != VNOVAL && vap->va_flags != 0) || (vap->va_rdev != VNOVAL) || ((int)vap->va_bytes != VNOVAL) || (vap->va_gen != VNOVAL)) { return (EINVAL); } de = vp->v_data; if (vp->v_type == VDIR) de = de->de_dir; error = c = 0; if (vap->va_uid == (uid_t)VNOVAL) uid = de->de_uid; else uid = vap->va_uid; if (vap->va_gid == (gid_t)VNOVAL) gid = de->de_gid; else gid = vap->va_gid; if (uid != de->de_uid || gid != de->de_gid) { if ((ap->a_cred->cr_uid != de->de_uid) || uid != de->de_uid || (gid != de->de_gid && !groupmember(gid, ap->a_cred))) { error = priv_check(ap->a_td, PRIV_VFS_CHOWN); if (error) return (error); } de->de_uid = uid; de->de_gid = gid; c = 1; } if (vap->va_mode != (mode_t)VNOVAL) { if (ap->a_cred->cr_uid != de->de_uid) { error = priv_check(ap->a_td, PRIV_VFS_ADMIN); if (error) return (error); } de->de_mode = vap->va_mode; c = 1; } if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) { /* See the comment in ufs_vnops::ufs_setattr(). */ if ((error = VOP_ACCESS(vp, VADMIN, ap->a_cred, ap->a_td)) && ((vap->va_vaflags & VA_UTIMES_NULL) == 0 || (error = VOP_ACCESS(vp, VWRITE, ap->a_cred, ap->a_td)))) return (error); if (vap->va_atime.tv_sec != VNOVAL) { if (vp->v_type == VCHR) vp->v_rdev->si_atime = vap->va_atime; else de->de_atime = vap->va_atime; } if (vap->va_mtime.tv_sec != VNOVAL) { if (vp->v_type == VCHR) vp->v_rdev->si_mtime = vap->va_mtime; else de->de_mtime = vap->va_mtime; } c = 1; } if (c) { if (vp->v_type == VCHR) vfs_timestamp(&vp->v_rdev->si_ctime); else vfs_timestamp(&de->de_mtime); } return (0); } #ifdef MAC static int devfs_setlabel(struct vop_setlabel_args *ap) { struct vnode *vp; struct devfs_dirent *de; vp = ap->a_vp; de = vp->v_data; mac_relabel_vnode(ap->a_cred, vp, ap->a_label); mac_update_devfs(vp->v_mount, de, vp); return (0); } #endif static int devfs_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td) { return (vnops.fo_stat(fp, sb, cred, td)); } static int devfs_symlink(struct vop_symlink_args *ap) { int i, error; struct devfs_dirent *dd; struct devfs_dirent *de; struct devfs_mount *dmp; struct thread *td; td = ap->a_cnp->cn_thread; KASSERT(td == curthread, ("devfs_symlink: td != curthread")); error = priv_check(td, PRIV_DEVFS_SYMLINK); if (error) return(error); dmp = VFSTODEVFS(ap->a_dvp->v_mount); dd = ap->a_dvp->v_data; de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen); de->de_uid = 0; de->de_gid = 0; de->de_mode = 0755; de->de_inode = alloc_unr(devfs_inos); de->de_dirent->d_type = DT_LNK; i = strlen(ap->a_target) + 1; de->de_symlink = malloc(i, M_DEVFS, M_WAITOK); bcopy(ap->a_target, de->de_symlink, i); sx_xlock(&dmp->dm_lock); #ifdef MAC mac_create_devfs_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de); #endif TAILQ_INSERT_TAIL(&dd->de_dlist, de, de_list); return (devfs_allocv(de, ap->a_dvp->v_mount, ap->a_vpp, td)); } /* ARGSUSED */ static int devfs_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td) { struct cdev *dev; int error, ioflag, resid; struct cdevsw *dsw; error = devfs_fp_check(fp, &dev, &dsw); if (error) return (error); KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT | O_FSYNC); if (ioflag & O_DIRECT) ioflag |= IO_DIRECT; if ((flags & FOF_OFFSET) == 0) uio->uio_offset = fp->f_offset; resid = uio->uio_resid; error = dsw->d_write(dev, uio, ioflag); if (uio->uio_resid != resid || (error == 0 && resid != 0)) { vfs_timestamp(&dev->si_ctime); dev->si_mtime = dev->si_ctime; } dev_relthread(dev); if ((flags & FOF_OFFSET) == 0) fp->f_offset = uio->uio_offset; fp->f_nextoff = uio->uio_offset; return (error); } dev_t dev2udev(struct cdev *x) { if (x == NULL) return (NODEV); return (x->si_priv->cdp_inode); } static struct fileops devfs_ops_f = { .fo_read = devfs_read_f, .fo_write = devfs_write_f, .fo_ioctl = devfs_ioctl_f, .fo_poll = devfs_poll_f, .fo_kqfilter = devfs_kqfilter_f, .fo_stat = devfs_stat_f, .fo_close = devfs_close_f, .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE }; static struct vop_vector devfs_vnodeops = { .vop_default = &default_vnodeops, .vop_access = devfs_access, .vop_getattr = devfs_getattr, .vop_ioctl = devfs_rioctl, .vop_lookup = devfs_lookup, .vop_mknod = devfs_mknod, .vop_pathconf = devfs_pathconf, .vop_read = devfs_rread, .vop_readdir = devfs_readdir, .vop_readlink = devfs_readlink, .vop_reclaim = devfs_reclaim, .vop_remove = devfs_remove, .vop_revoke = devfs_revoke, .vop_setattr = devfs_setattr, #ifdef MAC .vop_setlabel = devfs_setlabel, #endif .vop_symlink = devfs_symlink, }; static struct vop_vector devfs_specops = { .vop_default = &default_vnodeops, .vop_access = devfs_access, .vop_advlock = devfs_advlock, .vop_bmap = VOP_PANIC, .vop_close = devfs_close, .vop_create = VOP_PANIC, .vop_fsync = devfs_fsync, .vop_getattr = devfs_getattr, .vop_lease = VOP_NULL, .vop_link = VOP_PANIC, .vop_mkdir = VOP_PANIC, .vop_mknod = VOP_PANIC, .vop_open = devfs_open, .vop_pathconf = devfs_pathconf, .vop_print = devfs_print, .vop_read = VOP_PANIC, .vop_readdir = VOP_PANIC, .vop_readlink = VOP_PANIC, .vop_reallocblks = VOP_PANIC, .vop_reclaim = devfs_reclaim, .vop_remove = devfs_remove, .vop_rename = VOP_PANIC, .vop_revoke = devfs_revoke, .vop_rmdir = VOP_PANIC, .vop_setattr = devfs_setattr, #ifdef MAC .vop_setlabel = devfs_setlabel, #endif .vop_strategy = VOP_PANIC, .vop_symlink = VOP_PANIC, .vop_write = VOP_PANIC, }; /* * Our calling convention to the device drivers used to be that we passed * vnode.h IO_* flags to read()/write(), but we're moving to fcntl.h O_ * flags instead since that's what open(), close() and ioctl() takes and * we don't really want vnode.h in device drivers. * We solved the source compatibility by redefining some vnode flags to * be the same as the fcntl ones and by sending down the bitwise OR of * the respective fcntl/vnode flags. These CTASSERTS make sure nobody * pulls the rug out under this. */ CTASSERT(O_NONBLOCK == IO_NDELAY); CTASSERT(O_FSYNC == IO_SYNC); Index: head/sys/fs/fifofs/fifo_vnops.c =================================================================== --- head/sys/fs/fifofs/fifo_vnops.c (revision 171598) +++ head/sys/fs/fifofs/fifo_vnops.c (revision 171599) @@ -1,742 +1,742 @@ /*- * Copyright (c) 1990, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2005 Robert N. M. Watson * All rights reserved. * * 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. * 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. * * @(#)fifo_vnops.c 8.10 (Berkeley) 5/27/95 * $FreeBSD$ */ #include #include #include #include #include #include #include #include #include #include #include #include /* XXXKSE */ #include #include #include #include #include #include #include #include #include static fo_rdwr_t fifo_read_f; static fo_rdwr_t fifo_write_f; static fo_ioctl_t fifo_ioctl_f; static fo_poll_t fifo_poll_f; static fo_kqfilter_t fifo_kqfilter_f; static fo_stat_t fifo_stat_f; static fo_close_t fifo_close_f; struct fileops fifo_ops_f = { .fo_read = fifo_read_f, .fo_write = fifo_write_f, .fo_ioctl = fifo_ioctl_f, .fo_poll = fifo_poll_f, .fo_kqfilter = fifo_kqfilter_f, .fo_stat = fifo_stat_f, .fo_close = fifo_close_f, .fo_flags = DFLAG_PASSABLE }; /* * This structure is associated with the FIFO vnode and stores * the state associated with the FIFO. */ struct fifoinfo { struct socket *fi_readsock; struct socket *fi_writesock; long fi_readers; long fi_writers; }; static vop_print_t fifo_print; static vop_open_t fifo_open; static vop_close_t fifo_close; static vop_ioctl_t fifo_ioctl; static vop_kqfilter_t fifo_kqfilter; static vop_pathconf_t fifo_pathconf; static vop_advlock_t fifo_advlock; static void filt_fifordetach(struct knote *kn); static int filt_fiforead(struct knote *kn, long hint); static void filt_fifowdetach(struct knote *kn); static int filt_fifowrite(struct knote *kn, long hint); static void filt_fifodetach_notsup(struct knote *kn); static int filt_fifo_notsup(struct knote *kn, long hint); static struct filterops fiforead_filtops = { 1, NULL, filt_fifordetach, filt_fiforead }; static struct filterops fifowrite_filtops = { 1, NULL, filt_fifowdetach, filt_fifowrite }; static struct filterops fifo_notsup_filtops = { 1, NULL, filt_fifodetach_notsup, filt_fifo_notsup }; struct vop_vector fifo_specops = { .vop_default = &default_vnodeops, .vop_access = VOP_EBADF, .vop_advlock = fifo_advlock, .vop_close = fifo_close, .vop_create = VOP_PANIC, .vop_getattr = VOP_EBADF, .vop_ioctl = fifo_ioctl, .vop_kqfilter = fifo_kqfilter, .vop_lease = VOP_NULL, .vop_link = VOP_PANIC, .vop_mkdir = VOP_PANIC, .vop_mknod = VOP_PANIC, .vop_open = fifo_open, .vop_pathconf = fifo_pathconf, .vop_print = fifo_print, .vop_read = VOP_PANIC, .vop_readdir = VOP_PANIC, .vop_readlink = VOP_PANIC, .vop_reallocblks = VOP_PANIC, .vop_reclaim = VOP_NULL, .vop_remove = VOP_PANIC, .vop_rename = VOP_PANIC, .vop_rmdir = VOP_PANIC, .vop_setattr = VOP_EBADF, .vop_symlink = VOP_PANIC, .vop_write = VOP_PANIC, }; struct mtx fifo_mtx; MTX_SYSINIT(fifo, &fifo_mtx, "fifo mutex", MTX_DEF); /* * Dispose of fifo resources. */ static void fifo_cleanup(struct vnode *vp) { struct fifoinfo *fip = vp->v_fifoinfo; ASSERT_VOP_LOCKED(vp, "fifo_cleanup"); if (fip->fi_readers == 0 && fip->fi_writers == 0) { vp->v_fifoinfo = NULL; (void)soclose(fip->fi_readsock); (void)soclose(fip->fi_writesock); FREE(fip, M_VNODE); } } /* * Open called to set up a new instance of a fifo or * to find an active instance of a fifo. */ /* ARGSUSED */ static int fifo_open(ap) struct vop_open_args /* { struct vnode *a_vp; int a_mode; struct ucred *a_cred; struct thread *a_td; int a_fdidx; } */ *ap; { struct vnode *vp = ap->a_vp; struct fifoinfo *fip; struct thread *td = ap->a_td; struct ucred *cred = ap->a_cred; struct file *fp = ap->a_fp; struct socket *rso, *wso; int error; - ASSERT_VOP_LOCKED(vp, "fifo_open"); + ASSERT_VOP_ELOCKED(vp, "fifo_open"); if (fp == NULL) return (EINVAL); if ((fip = vp->v_fifoinfo) == NULL) { MALLOC(fip, struct fifoinfo *, sizeof(*fip), M_VNODE, M_WAITOK); error = socreate(AF_LOCAL, &rso, SOCK_STREAM, 0, cred, td); if (error) goto fail1; fip->fi_readsock = rso; error = socreate(AF_LOCAL, &wso, SOCK_STREAM, 0, cred, td); if (error) goto fail2; fip->fi_writesock = wso; error = soconnect2(wso, rso); if (error) { (void)soclose(wso); fail2: (void)soclose(rso); fail1: free(fip, M_VNODE); return (error); } fip->fi_readers = fip->fi_writers = 0; wso->so_snd.sb_lowat = PIPE_BUF; SOCKBUF_LOCK(&rso->so_rcv); rso->so_rcv.sb_state |= SBS_CANTRCVMORE; SOCKBUF_UNLOCK(&rso->so_rcv); KASSERT(vp->v_fifoinfo == NULL, ("fifo_open: v_fifoinfo race")); vp->v_fifoinfo = fip; } /* * General access to fi_readers and fi_writers is protected using * the vnode lock. * * Protect the increment of fi_readers and fi_writers and the * associated calls to wakeup() with the fifo mutex in addition * to the vnode lock. This allows the vnode lock to be dropped * for the msleep() calls below, and using the fifo mutex with * msleep() prevents the wakeup from being missed. */ mtx_lock(&fifo_mtx); if (ap->a_mode & FREAD) { fip->fi_readers++; if (fip->fi_readers == 1) { SOCKBUF_LOCK(&fip->fi_writesock->so_snd); fip->fi_writesock->so_snd.sb_state &= ~SBS_CANTSENDMORE; SOCKBUF_UNLOCK(&fip->fi_writesock->so_snd); if (fip->fi_writers > 0) { wakeup(&fip->fi_writers); sowwakeup(fip->fi_writesock); } } } if (ap->a_mode & FWRITE) { if ((ap->a_mode & O_NONBLOCK) && fip->fi_readers == 0) { mtx_unlock(&fifo_mtx); return (ENXIO); } fip->fi_writers++; if (fip->fi_writers == 1) { SOCKBUF_LOCK(&fip->fi_readsock->so_rcv); fip->fi_readsock->so_rcv.sb_state &= ~SBS_CANTRCVMORE; SOCKBUF_UNLOCK(&fip->fi_readsock->so_rcv); if (fip->fi_readers > 0) { wakeup(&fip->fi_readers); sorwakeup(fip->fi_readsock); } } } if ((ap->a_mode & O_NONBLOCK) == 0) { if ((ap->a_mode & FREAD) && fip->fi_writers == 0) { VOP_UNLOCK(vp, 0, td); error = msleep(&fip->fi_readers, &fifo_mtx, PDROP | PCATCH | PSOCK, "fifoor", 0); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); if (error) { fip->fi_readers--; if (fip->fi_readers == 0) { socantsendmore(fip->fi_writesock); fifo_cleanup(vp); } return (error); } mtx_lock(&fifo_mtx); /* * We must have got woken up because we had a writer. * That (and not still having one) is the condition * that we must wait for. */ } if ((ap->a_mode & FWRITE) && fip->fi_readers == 0) { VOP_UNLOCK(vp, 0, td); error = msleep(&fip->fi_writers, &fifo_mtx, PDROP | PCATCH | PSOCK, "fifoow", 0); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); if (error) { fip->fi_writers--; if (fip->fi_writers == 0) { socantrcvmore(fip->fi_readsock); fifo_cleanup(vp); } return (error); } /* * We must have got woken up because we had * a reader. That (and not still having one) * is the condition that we must wait for. */ mtx_lock(&fifo_mtx); } } mtx_unlock(&fifo_mtx); KASSERT(fp != NULL, ("can't fifo/vnode bypass")); FILE_LOCK(fp); KASSERT(fp->f_ops == &badfileops, ("not badfileops in fifo_open")); fp->f_data = fip; fp->f_ops = &fifo_ops_f; FILE_UNLOCK(fp); return (0); } /* * Now unused vnode ioctl routine. */ /* ARGSUSED */ static int fifo_ioctl(ap) struct vop_ioctl_args /* { struct vnode *a_vp; u_long a_command; caddr_t a_data; int a_fflag; struct ucred *a_cred; struct thread *a_td; } */ *ap; { printf("WARNING: fifo_ioctl called unexpectedly\n"); return (ENOTTY); } /* * Now unused vnode kqfilter routine. */ /* ARGSUSED */ static int fifo_kqfilter(ap) struct vop_kqfilter_args /* { struct vnode *a_vp; struct knote *a_kn; } */ *ap; { printf("WARNING: fifo_kqfilter called unexpectedly\n"); return (EINVAL); } static void filt_fifordetach(struct knote *kn) { struct socket *so = (struct socket *)kn->kn_hook; SOCKBUF_LOCK(&so->so_rcv); knlist_remove(&so->so_rcv.sb_sel.si_note, kn, 1); if (knlist_empty(&so->so_rcv.sb_sel.si_note)) so->so_rcv.sb_flags &= ~SB_KNOTE; SOCKBUF_UNLOCK(&so->so_rcv); } static int filt_fiforead(struct knote *kn, long hint) { struct socket *so = (struct socket *)kn->kn_hook; SOCKBUF_LOCK_ASSERT(&so->so_rcv); kn->kn_data = so->so_rcv.sb_cc; if (so->so_rcv.sb_state & SBS_CANTRCVMORE) { kn->kn_flags |= EV_EOF; return (1); } else { kn->kn_flags &= ~EV_EOF; return (kn->kn_data > 0); } } static void filt_fifowdetach(struct knote *kn) { struct socket *so = (struct socket *)kn->kn_hook; SOCKBUF_LOCK(&so->so_snd); knlist_remove(&so->so_snd.sb_sel.si_note, kn, 1); if (knlist_empty(&so->so_snd.sb_sel.si_note)) so->so_snd.sb_flags &= ~SB_KNOTE; SOCKBUF_UNLOCK(&so->so_snd); } static int filt_fifowrite(struct knote *kn, long hint) { struct socket *so = (struct socket *)kn->kn_hook; SOCKBUF_LOCK_ASSERT(&so->so_snd); kn->kn_data = sbspace(&so->so_snd); if (so->so_snd.sb_state & SBS_CANTSENDMORE) { kn->kn_flags |= EV_EOF; return (1); } else { kn->kn_flags &= ~EV_EOF; return (kn->kn_data >= so->so_snd.sb_lowat); } } static void filt_fifodetach_notsup(struct knote *kn) { } static int filt_fifo_notsup(struct knote *kn, long hint) { return (0); } /* * Device close routine */ /* ARGSUSED */ static int fifo_close(ap) struct vop_close_args /* { struct vnode *a_vp; int a_fflag; struct ucred *a_cred; struct thread *a_td; } */ *ap; { struct vnode *vp = ap->a_vp; struct fifoinfo *fip = vp->v_fifoinfo; ASSERT_VOP_LOCKED(vp, "fifo_close"); KASSERT(fip != NULL, ("fifo_close: no v_fifoinfo")); if (ap->a_fflag & FREAD) { fip->fi_readers--; if (fip->fi_readers == 0) socantsendmore(fip->fi_writesock); } if (ap->a_fflag & FWRITE) { fip->fi_writers--; if (fip->fi_writers == 0) socantrcvmore(fip->fi_readsock); } fifo_cleanup(vp); return (0); } /* * Print out internal contents of a fifo vnode. */ int fifo_printinfo(vp) struct vnode *vp; { register struct fifoinfo *fip = vp->v_fifoinfo; if (fip == NULL){ printf(", NULL v_fifoinfo"); return (0); } printf(", fifo with %ld readers and %ld writers", fip->fi_readers, fip->fi_writers); return (0); } /* * Print out the contents of a fifo vnode. */ static int fifo_print(ap) struct vop_print_args /* { struct vnode *a_vp; } */ *ap; { fifo_printinfo(ap->a_vp); printf("\n"); return (0); } /* * Return POSIX pathconf information applicable to fifo's. */ static int fifo_pathconf(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_PIPE_BUF: *ap->a_retval = PIPE_BUF; return (0); case _PC_CHOWN_RESTRICTED: *ap->a_retval = 1; return (0); default: return (EINVAL); } /* NOTREACHED */ } /* * Fifo advisory byte-level locks. */ /* ARGSUSED */ static int fifo_advlock(ap) struct vop_advlock_args /* { struct vnode *a_vp; caddr_t a_id; int a_op; struct flock *a_fl; int a_flags; } */ *ap; { return (ap->a_flags & F_FLOCK ? EOPNOTSUPP : EINVAL); } static int fifo_close_f(struct file *fp, struct thread *td) { return (vnops.fo_close(fp, td)); } /* * The implementation of ioctl() for named fifos is complicated by the fact * that we permit O_RDWR fifo file descriptors, meaning that the actions of * ioctls may have to be applied to both the underlying sockets rather than * just one. The original implementation simply forward the ioctl to one * or both sockets based on fp->f_flag. We now consider each ioctl * separately, as the composition effect requires careful ordering. * * We do not blindly pass all ioctls through to the socket in order to avoid * providing unnecessary ioctls that might be improperly depended on by * applications (such as socket-specific, routing, and interface ioctls). * * Unlike sys_pipe.c, fifos do not implement the deprecated TIOCSPGRP and * TIOCGPGRP ioctls. Earlier implementations of fifos did forward SIOCSPGRP * and SIOCGPGRP ioctls, so we might need to re-add those here. */ static int fifo_ioctl_f(struct file *fp, u_long com, void *data, struct ucred *cred, struct thread *td) { struct fifoinfo *fi; struct file filetmp; /* Local, so need not be locked. */ int error; error = ENOTTY; fi = fp->f_data; switch (com) { case FIONBIO: /* * Non-blocking I/O is implemented at the fifo layer using * MSG_NBIO, so does not need to be forwarded down the stack. */ return (0); case FIOASYNC: case FIOSETOWN: case FIOGETOWN: /* * These socket ioctls don't have any ordering requirements, * so are called in an arbitrary order, and only on the * sockets indicated by the file descriptor rights. * * XXXRW: If O_RDWR and the read socket accepts an ioctl but * the write socket doesn't, the socketpair is left in an * inconsistent state. */ if (fp->f_flag & FREAD) { filetmp.f_data = fi->fi_readsock; filetmp.f_cred = cred; error = soo_ioctl(&filetmp, com, data, cred, td); if (error) return (error); } if (fp->f_flag & FWRITE) { filetmp.f_data = fi->fi_writesock; filetmp.f_cred = cred; error = soo_ioctl(&filetmp, com, data, cred, td); } return (error); case FIONREAD: /* * FIONREAD will return 0 for non-readable descriptors, and * the results of FIONREAD on the read socket for readable * descriptors. */ if (!(fp->f_flag & FREAD)) { *(int *)data = 0; return (0); } filetmp.f_data = fi->fi_readsock; filetmp.f_cred = cred; return (soo_ioctl(&filetmp, com, data, cred, td)); default: return (ENOTTY); } } /* * Because fifos are now a file descriptor layer object, EVFILT_VNODE is not * implemented. Likely, fifo_kqfilter() should be removed, and * fifo_kqfilter_f() should know how to forward the request to the underling * vnode using f_vnode in the file descriptor here. */ static int fifo_kqfilter_f(struct file *fp, struct knote *kn) { struct fifoinfo *fi; struct socket *so; struct sockbuf *sb; fi = fp->f_data; /* * If a filter is requested that is not supported by this file * descriptor, don't return an error, but also don't ever generate an * event. */ if ((kn->kn_filter == EVFILT_READ) && !(fp->f_flag & FREAD)) { kn->kn_fop = &fifo_notsup_filtops; return (0); } if ((kn->kn_filter == EVFILT_WRITE) && !(fp->f_flag & FWRITE)) { kn->kn_fop = &fifo_notsup_filtops; return (0); } switch (kn->kn_filter) { case EVFILT_READ: kn->kn_fop = &fiforead_filtops; so = fi->fi_readsock; sb = &so->so_rcv; break; case EVFILT_WRITE: kn->kn_fop = &fifowrite_filtops; so = fi->fi_writesock; sb = &so->so_snd; break; default: return (EINVAL); } kn->kn_hook = (caddr_t)so; SOCKBUF_LOCK(sb); knlist_add(&sb->sb_sel.si_note, kn, 1); sb->sb_flags |= SB_KNOTE; SOCKBUF_UNLOCK(sb); return (0); } static int fifo_poll_f(struct file *fp, int events, struct ucred *cred, struct thread *td) { struct fifoinfo *fip; struct file filetmp; int levents, revents = 0; fip = fp->f_data; levents = events & (POLLIN | POLLINIGNEOF | POLLPRI | POLLRDNORM | POLLRDBAND); if ((fp->f_flag & FREAD) && levents) { /* * If POLLIN or POLLRDNORM is requested and POLLINIGNEOF is * not, then convert the first two to the last one. This * tells the socket poll function to ignore EOF so that we * block if there is no writer (and no data). Callers can * set POLLINIGNEOF to get non-blocking behavior. */ if (levents & (POLLIN | POLLRDNORM) && !(levents & POLLINIGNEOF)) { levents &= ~(POLLIN | POLLRDNORM); levents |= POLLINIGNEOF; } filetmp.f_data = fip->fi_readsock; filetmp.f_cred = cred; revents |= soo_poll(&filetmp, levents, cred, td); /* Reverse the above conversion. */ if ((revents & POLLINIGNEOF) && !(events & POLLINIGNEOF)) { revents |= (events & (POLLIN | POLLRDNORM)); revents &= ~POLLINIGNEOF; } } levents = events & (POLLOUT | POLLWRNORM | POLLWRBAND); if ((fp->f_flag & FWRITE) && levents) { filetmp.f_data = fip->fi_writesock; filetmp.f_cred = cred; revents |= soo_poll(&filetmp, levents, cred, td); } return (revents); } static int fifo_read_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td) { struct fifoinfo *fip; int error, sflags; fip = fp->f_data; KASSERT(uio->uio_rw == UIO_READ,("fifo_read mode")); if (uio->uio_resid == 0) return (0); sflags = (fp->f_flag & FNONBLOCK) ? MSG_NBIO : 0; mtx_lock(&Giant); error = soreceive(fip->fi_readsock, NULL, uio, NULL, NULL, &sflags); mtx_unlock(&Giant); return (error); } static int fifo_stat_f(struct file *fp, struct stat *sb, struct ucred *cred, struct thread *td) { return (vnops.fo_stat(fp, sb, cred, td)); } static int fifo_write_f(struct file *fp, struct uio *uio, struct ucred *cred, int flags, struct thread *td) { struct fifoinfo *fip; int error, sflags; fip = fp->f_data; KASSERT(uio->uio_rw == UIO_WRITE,("fifo_write mode")); sflags = (fp->f_flag & FNONBLOCK) ? MSG_NBIO : 0; mtx_lock(&Giant); error = sosend(fip->fi_writesock, NULL, uio, 0, NULL, sflags, td); mtx_unlock(&Giant); return (error); } Index: head/sys/kern/uipc_usrreq.c =================================================================== --- head/sys/kern/uipc_usrreq.c (revision 171598) +++ head/sys/kern/uipc_usrreq.c (revision 171599) @@ -1,2285 +1,2285 @@ /*- * Copyright (c) 1982, 1986, 1989, 1991, 1993 * The Regents of the University of California. * Copyright (c) 2004-2007 Robert N. M. Watson * All rights reserved. * * 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. * 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. * * From: @(#)uipc_usrreq.c 8.3 (Berkeley) 1/4/94 */ /* * UNIX Domain (Local) Sockets * * This is an implementation of UNIX (local) domain sockets. Each socket has * an associated struct unpcb (UNIX protocol control block). Stream sockets * may be connected to 0 or 1 other socket. Datagram sockets may be * connected to 0, 1, or many other sockets. Sockets may be created and * connected in pairs (socketpair(2)), or bound/connected to using the file * system name space. For most purposes, only the receive socket buffer is * used, as sending on one socket delivers directly to the receive socket * buffer of a second socket. * * The implementation is substantially complicated by the fact that * "ancillary data", such as file descriptors or credentials, may be passed * across UNIX domain sockets. The potential for passing UNIX domain sockets * over other UNIX domain sockets requires the implementation of a simple * garbage collector to find and tear down cycles of disconnected sockets. * * TODO: * SEQPACKET, RDM * rethink name space problems * need a proper out-of-band * lock pushdown */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include "opt_mac.h" #include #include #include #include /* XXX must be before */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include static uma_zone_t unp_zone; static unp_gen_t unp_gencnt; static u_int unp_count; /* Count of local sockets. */ static ino_t unp_ino; /* Prototype for fake inode numbers. */ static int unp_rights; /* File descriptors in flight. */ static struct unp_head unp_shead; /* List of local stream sockets. */ static struct unp_head unp_dhead; /* List of local datagram sockets. */ static const struct sockaddr sun_noname = { sizeof(sun_noname), AF_LOCAL }; /* * Garbage collection of cyclic file descriptor/socket references occurs * asynchronously in a taskqueue context in order to avoid recursion and * reentrance in the UNIX domain socket, file descriptor, and socket layer * code. See unp_gc() for a full description. */ static struct task unp_gc_task; /* * Both send and receive buffers are allocated PIPSIZ bytes of buffering for * stream sockets, although the total for sender and receiver is actually * only PIPSIZ. * * Datagram sockets really use the sendspace as the maximum datagram size, * and don't really want to reserve the sendspace. Their recvspace should be * large enough for at least one max-size datagram plus address. */ #ifndef PIPSIZ #define PIPSIZ 8192 #endif static u_long unpst_sendspace = PIPSIZ; static u_long unpst_recvspace = PIPSIZ; static u_long unpdg_sendspace = 2*1024; /* really max datagram size */ static u_long unpdg_recvspace = 4*1024; SYSCTL_NODE(_net, PF_LOCAL, local, CTLFLAG_RW, 0, "Local domain"); SYSCTL_NODE(_net_local, SOCK_STREAM, stream, CTLFLAG_RW, 0, "SOCK_STREAM"); SYSCTL_NODE(_net_local, SOCK_DGRAM, dgram, CTLFLAG_RW, 0, "SOCK_DGRAM"); SYSCTL_ULONG(_net_local_stream, OID_AUTO, sendspace, CTLFLAG_RW, &unpst_sendspace, 0, ""); SYSCTL_ULONG(_net_local_stream, OID_AUTO, recvspace, CTLFLAG_RW, &unpst_recvspace, 0, ""); SYSCTL_ULONG(_net_local_dgram, OID_AUTO, maxdgram, CTLFLAG_RW, &unpdg_sendspace, 0, ""); SYSCTL_ULONG(_net_local_dgram, OID_AUTO, recvspace, CTLFLAG_RW, &unpdg_recvspace, 0, ""); SYSCTL_INT(_net_local, OID_AUTO, inflight, CTLFLAG_RD, &unp_rights, 0, ""); /*- * Locking and synchronization: * * The global UNIX domain socket rwlock (unp_global_rwlock) protects all * global variables, including the linked lists tracking the set of allocated * UNIX domain sockets. The global rwlock also serves to prevent deadlock * when more than one PCB lock is acquired at a time (i.e., during * connect()). Finally, the global rwlock protects uncounted references from * vnodes to sockets bound to those vnodes: to safely dereference the * v_socket pointer, the global rwlock must be held while a full reference is * acquired. * * UNIX domain sockets each have an unpcb hung off of their so_pcb pointer, * allocated in pru_attach() and freed in pru_detach(). The validity of that * pointer is an invariant, so no lock is required to dereference the so_pcb * pointer if a valid socket reference is held by the caller. In practice, * this is always true during operations performed on a socket. Each unpcb * has a back-pointer to its socket, unp_socket, which will be stable under * the same circumstances. * * This pointer may only be safely dereferenced as long as a valid reference * to the unpcb is held. Typically, this reference will be from the socket, * or from another unpcb when the referring unpcb's lock is held (in order * that the reference not be invalidated during use). For example, to follow * unp->unp_conn->unp_socket, you need unlock the lock on unp, not unp_conn, * as unp_socket remains valid as long as the reference to unp_conn is valid. * * Fields of unpcbss are locked using a per-unpcb lock, unp_mtx. Individual * atomic reads without the lock may be performed "lockless", but more * complex reads and read-modify-writes require the mutex to be held. No * lock order is defined between unpcb locks -- multiple unpcb locks may be * acquired at the same time only when holding the global UNIX domain socket * rwlock exclusively, which prevents deadlocks. * * Blocking with UNIX domain sockets is a tricky issue: unlike most network * protocols, bind() is a non-atomic operation, and connect() requires * potential sleeping in the protocol, due to potentially waiting on local or * distributed file systems. We try to separate "lookup" operations, which * may sleep, and the IPC operations themselves, which typically can occur * with relative atomicity as locks can be held over the entire operation. * * Another tricky issue is simultaneous multi-threaded or multi-process * access to a single UNIX domain socket. These are handled by the flags * UNP_CONNECTING and UNP_BINDING, which prevent concurrent connecting or * binding, both of which involve dropping UNIX domain socket locks in order * to perform namei() and other file system operations. */ static struct rwlock unp_global_rwlock; #define UNP_GLOBAL_LOCK_INIT() rw_init(&unp_global_rwlock, \ "unp_global_rwlock") #define UNP_GLOBAL_LOCK_ASSERT() rw_assert(&unp_global_rwlock, \ RA_LOCKED) #define UNP_GLOBAL_UNLOCK_ASSERT() rw_assert(&unp_global_rwlock, \ RA_UNLOCKED) #define UNP_GLOBAL_WLOCK() rw_wlock(&unp_global_rwlock) #define UNP_GLOBAL_WUNLOCK() rw_wunlock(&unp_global_rwlock) #define UNP_GLOBAL_WLOCK_ASSERT() rw_assert(&unp_global_rwlock, \ RA_WLOCKED) #define UNP_GLOBAL_WOWNED() rw_wowned(&unp_global_rwlock) #define UNP_GLOBAL_RLOCK() rw_rlock(&unp_global_rwlock) #define UNP_GLOBAL_RUNLOCK() rw_runlock(&unp_global_rwlock) #define UNP_GLOBAL_RLOCK_ASSERT() rw_assert(&unp_global_rwlock, \ RA_RLOCKED) #define UNP_PCB_LOCK_INIT(unp) mtx_init(&(unp)->unp_mtx, \ "unp_mtx", "unp_mtx", \ MTX_DUPOK|MTX_DEF|MTX_RECURSE) #define UNP_PCB_LOCK_DESTROY(unp) mtx_destroy(&(unp)->unp_mtx) #define UNP_PCB_LOCK(unp) mtx_lock(&(unp)->unp_mtx) #define UNP_PCB_UNLOCK(unp) mtx_unlock(&(unp)->unp_mtx) #define UNP_PCB_LOCK_ASSERT(unp) mtx_assert(&(unp)->unp_mtx, MA_OWNED) static int unp_connect(struct socket *, struct sockaddr *, struct thread *); static int unp_connect2(struct socket *so, struct socket *so2, int); static void unp_disconnect(struct unpcb *unp, struct unpcb *unp2); static void unp_shutdown(struct unpcb *); static void unp_drop(struct unpcb *, int); static void unp_gc(__unused void *, int); static void unp_scan(struct mbuf *, void (*)(struct file *)); static void unp_mark(struct file *); static void unp_discard(struct file *); static void unp_freerights(struct file **, int); static int unp_internalize(struct mbuf **, struct thread *); static struct mbuf *unp_addsockcred(struct thread *, struct mbuf *); /* * Definitions of protocols supported in the LOCAL domain. */ static struct domain localdomain; static struct protosw localsw[] = { { .pr_type = SOCK_STREAM, .pr_domain = &localdomain, .pr_flags = PR_CONNREQUIRED|PR_WANTRCVD|PR_RIGHTS, .pr_ctloutput = &uipc_ctloutput, .pr_usrreqs = &uipc_usrreqs }, { .pr_type = SOCK_DGRAM, .pr_domain = &localdomain, .pr_flags = PR_ATOMIC|PR_ADDR|PR_RIGHTS, .pr_usrreqs = &uipc_usrreqs }, }; static struct domain localdomain = { .dom_family = AF_LOCAL, .dom_name = "local", .dom_init = unp_init, .dom_externalize = unp_externalize, .dom_dispose = unp_dispose, .dom_protosw = localsw, .dom_protoswNPROTOSW = &localsw[sizeof(localsw)/sizeof(localsw[0])] }; DOMAIN_SET(local); static void uipc_abort(struct socket *so) { struct unpcb *unp, *unp2; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_abort: unp == NULL")); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); unp2 = unp->unp_conn; if (unp2 != NULL) { UNP_PCB_LOCK(unp2); unp_drop(unp2, ECONNABORTED); UNP_PCB_UNLOCK(unp2); } UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); } static int uipc_accept(struct socket *so, struct sockaddr **nam) { struct unpcb *unp, *unp2; const struct sockaddr *sa; /* * Pass back name of connected socket, if it was bound and we are * still connected (our peer may have closed already!). */ unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_accept: unp == NULL")); *nam = malloc(sizeof(struct sockaddr_un), M_SONAME, M_WAITOK); UNP_GLOBAL_RLOCK(); unp2 = unp->unp_conn; if (unp2 != NULL && unp2->unp_addr != NULL) { UNP_PCB_LOCK(unp2); sa = (struct sockaddr *) unp2->unp_addr; bcopy(sa, *nam, sa->sa_len); UNP_PCB_UNLOCK(unp2); } else { sa = &sun_noname; bcopy(sa, *nam, sa->sa_len); } UNP_GLOBAL_RUNLOCK(); return (0); } static int uipc_attach(struct socket *so, int proto, struct thread *td) { u_long sendspace, recvspace; struct unpcb *unp; int error, locked; KASSERT(so->so_pcb == NULL, ("uipc_attach: so_pcb != NULL")); if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) { switch (so->so_type) { case SOCK_STREAM: sendspace = unpst_sendspace; recvspace = unpst_recvspace; break; case SOCK_DGRAM: sendspace = unpdg_sendspace; recvspace = unpdg_recvspace; break; default: panic("uipc_attach"); } error = soreserve(so, sendspace, recvspace); if (error) return (error); } unp = uma_zalloc(unp_zone, M_NOWAIT | M_ZERO); if (unp == NULL) return (ENOBUFS); LIST_INIT(&unp->unp_refs); UNP_PCB_LOCK_INIT(unp); unp->unp_socket = so; so->so_pcb = unp; unp->unp_refcount = 1; /* * uipc_attach() may be called indirectly from within the UNIX domain * socket code via sonewconn() in unp_connect(). Since rwlocks can * not be recursed, we do the closest thing. */ locked = 0; if (!UNP_GLOBAL_WOWNED()) { UNP_GLOBAL_WLOCK(); locked = 1; } unp->unp_gencnt = ++unp_gencnt; unp_count++; LIST_INSERT_HEAD(so->so_type == SOCK_DGRAM ? &unp_dhead : &unp_shead, unp, unp_link); if (locked) UNP_GLOBAL_WUNLOCK(); return (0); } static int uipc_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_un *soun = (struct sockaddr_un *)nam; struct vattr vattr; int error, namelen, vfslocked; struct nameidata nd; struct unpcb *unp; struct vnode *vp; struct mount *mp; char *buf; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_bind: unp == NULL")); namelen = soun->sun_len - offsetof(struct sockaddr_un, sun_path); if (namelen <= 0) return (EINVAL); /* * We don't allow simultaneous bind() calls on a single UNIX domain * socket, so flag in-progress operations, and return an error if an * operation is already in progress. * * Historically, we have not allowed a socket to be rebound, so this * also returns an error. Not allowing re-binding simplifies the * implementation and avoids a great many possible failure modes. */ UNP_PCB_LOCK(unp); if (unp->unp_vnode != NULL) { UNP_PCB_UNLOCK(unp); return (EINVAL); } if (unp->unp_flags & UNP_BINDING) { UNP_PCB_UNLOCK(unp); return (EALREADY); } unp->unp_flags |= UNP_BINDING; UNP_PCB_UNLOCK(unp); buf = malloc(namelen + 1, M_TEMP, M_WAITOK); strlcpy(buf, soun->sun_path, namelen + 1); restart: vfslocked = 0; NDINIT(&nd, CREATE, MPSAFE | NOFOLLOW | LOCKPARENT | SAVENAME, UIO_SYSSPACE, buf, td); /* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */ error = namei(&nd); if (error) goto error; vp = nd.ni_vp; vfslocked = NDHASGIANT(&nd); if (vp != NULL || vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) { NDFREE(&nd, NDF_ONLY_PNBUF); if (nd.ni_dvp == vp) vrele(nd.ni_dvp); else vput(nd.ni_dvp); if (vp != NULL) { vrele(vp); error = EADDRINUSE; goto error; } error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH); if (error) goto error; VFS_UNLOCK_GIANT(vfslocked); goto restart; } VATTR_NULL(&vattr); vattr.va_type = VSOCK; vattr.va_mode = (ACCESSPERMS & ~td->td_proc->p_fd->fd_cmask); #ifdef MAC error = mac_check_vnode_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd, &vattr); #endif if (error == 0) { VOP_LEASE(nd.ni_dvp, td, td->td_ucred, LEASE_WRITE); error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr); } NDFREE(&nd, NDF_ONLY_PNBUF); vput(nd.ni_dvp); if (error) { vn_finished_write(mp); goto error; } vp = nd.ni_vp; - ASSERT_VOP_LOCKED(vp, "uipc_bind"); + ASSERT_VOP_ELOCKED(vp, "uipc_bind"); soun = (struct sockaddr_un *)sodupsockaddr(nam, M_WAITOK); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); vp->v_socket = unp->unp_socket; unp->unp_vnode = vp; unp->unp_addr = soun; unp->unp_flags &= ~UNP_BINDING; UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); VOP_UNLOCK(vp, 0, td); vn_finished_write(mp); VFS_UNLOCK_GIANT(vfslocked); free(buf, M_TEMP); return (0); error: VFS_UNLOCK_GIANT(vfslocked); UNP_PCB_LOCK(unp); unp->unp_flags &= ~UNP_BINDING; UNP_PCB_UNLOCK(unp); free(buf, M_TEMP); return (error); } static int uipc_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { int error; KASSERT(td == curthread, ("uipc_connect: td != curthread")); UNP_GLOBAL_WLOCK(); error = unp_connect(so, nam, td); UNP_GLOBAL_WUNLOCK(); return (error); } static void uipc_close(struct socket *so) { struct unpcb *unp, *unp2; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_close: unp == NULL")); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); unp2 = unp->unp_conn; if (unp2 != NULL) { UNP_PCB_LOCK(unp2); unp_disconnect(unp, unp2); UNP_PCB_UNLOCK(unp2); } UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); } int uipc_connect2(struct socket *so1, struct socket *so2) { struct unpcb *unp, *unp2; int error; UNP_GLOBAL_WLOCK(); unp = so1->so_pcb; KASSERT(unp != NULL, ("uipc_connect2: unp == NULL")); UNP_PCB_LOCK(unp); unp2 = so2->so_pcb; KASSERT(unp2 != NULL, ("uipc_connect2: unp2 == NULL")); UNP_PCB_LOCK(unp2); error = unp_connect2(so1, so2, PRU_CONNECT2); UNP_PCB_UNLOCK(unp2); UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); return (error); } /* control is EOPNOTSUPP */ static void uipc_detach(struct socket *so) { struct unpcb *unp, *unp2; struct sockaddr_un *saved_unp_addr; struct vnode *vp; int freeunp, local_unp_rights; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_detach: unp == NULL")); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); LIST_REMOVE(unp, unp_link); unp->unp_gencnt = ++unp_gencnt; --unp_count; /* * XXXRW: Should assert vp->v_socket == so. */ if ((vp = unp->unp_vnode) != NULL) { unp->unp_vnode->v_socket = NULL; unp->unp_vnode = NULL; } unp2 = unp->unp_conn; if (unp2 != NULL) { UNP_PCB_LOCK(unp2); unp_disconnect(unp, unp2); UNP_PCB_UNLOCK(unp2); } /* * We hold the global lock, so it's OK to acquire multiple pcb locks * at a time. */ while (!LIST_EMPTY(&unp->unp_refs)) { struct unpcb *ref = LIST_FIRST(&unp->unp_refs); UNP_PCB_LOCK(ref); unp_drop(ref, ECONNRESET); UNP_PCB_UNLOCK(ref); } UNP_GLOBAL_WUNLOCK(); unp->unp_socket->so_pcb = NULL; local_unp_rights = unp_rights; saved_unp_addr = unp->unp_addr; unp->unp_addr = NULL; unp->unp_refcount--; freeunp = (unp->unp_refcount == 0); if (saved_unp_addr != NULL) FREE(saved_unp_addr, M_SONAME); if (freeunp) { UNP_PCB_LOCK_DESTROY(unp); uma_zfree(unp_zone, unp); } else UNP_PCB_UNLOCK(unp); if (vp) { int vfslocked; vfslocked = VFS_LOCK_GIANT(vp->v_mount); vrele(vp); VFS_UNLOCK_GIANT(vfslocked); } if (local_unp_rights) taskqueue_enqueue(taskqueue_thread, &unp_gc_task); } static int uipc_disconnect(struct socket *so) { struct unpcb *unp, *unp2; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_disconnect: unp == NULL")); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); unp2 = unp->unp_conn; if (unp2 != NULL) { UNP_PCB_LOCK(unp2); unp_disconnect(unp, unp2); UNP_PCB_UNLOCK(unp2); } UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); return (0); } static int uipc_listen(struct socket *so, int backlog, struct thread *td) { struct unpcb *unp; int error; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_listen: unp == NULL")); UNP_PCB_LOCK(unp); if (unp->unp_vnode == NULL) { UNP_PCB_UNLOCK(unp); return (EINVAL); } SOCK_LOCK(so); error = solisten_proto_check(so); if (error == 0) { cru2x(td->td_ucred, &unp->unp_peercred); unp->unp_flags |= UNP_HAVEPCCACHED; solisten_proto(so, backlog); } SOCK_UNLOCK(so); UNP_PCB_UNLOCK(unp); return (error); } static int uipc_peeraddr(struct socket *so, struct sockaddr **nam) { struct unpcb *unp, *unp2; const struct sockaddr *sa; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_peeraddr: unp == NULL")); *nam = malloc(sizeof(struct sockaddr_un), M_SONAME, M_WAITOK); UNP_PCB_LOCK(unp); /* * XXX: It seems that this test always fails even when connection is * established. So, this else clause is added as workaround to * return PF_LOCAL sockaddr. */ unp2 = unp->unp_conn; if (unp2 != NULL) { UNP_PCB_LOCK(unp2); if (unp2->unp_addr != NULL) sa = (struct sockaddr *) unp->unp_conn->unp_addr; else sa = &sun_noname; bcopy(sa, *nam, sa->sa_len); UNP_PCB_UNLOCK(unp2); } else { sa = &sun_noname; bcopy(sa, *nam, sa->sa_len); } UNP_PCB_UNLOCK(unp); return (0); } static int uipc_rcvd(struct socket *so, int flags) { struct unpcb *unp, *unp2; struct socket *so2; u_int mbcnt, sbcc; u_long newhiwat; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_rcvd: unp == NULL")); if (so->so_type == SOCK_DGRAM) panic("uipc_rcvd DGRAM?"); if (so->so_type != SOCK_STREAM) panic("uipc_rcvd unknown socktype"); /* * Adjust backpressure on sender and wakeup any waiting to write. * * The unp lock is acquired to maintain the validity of the unp_conn * pointer; no lock on unp2 is required as unp2->unp_socket will be * static as long as we don't permit unp2 to disconnect from unp, * which is prevented by the lock on unp. We cache values from * so_rcv to avoid holding the so_rcv lock over the entire * transaction on the remote so_snd. */ SOCKBUF_LOCK(&so->so_rcv); mbcnt = so->so_rcv.sb_mbcnt; sbcc = so->so_rcv.sb_cc; SOCKBUF_UNLOCK(&so->so_rcv); UNP_PCB_LOCK(unp); unp2 = unp->unp_conn; if (unp2 == NULL) { UNP_PCB_UNLOCK(unp); return (0); } so2 = unp2->unp_socket; SOCKBUF_LOCK(&so2->so_snd); so2->so_snd.sb_mbmax += unp->unp_mbcnt - mbcnt; newhiwat = so2->so_snd.sb_hiwat + unp->unp_cc - sbcc; (void)chgsbsize(so2->so_cred->cr_uidinfo, &so2->so_snd.sb_hiwat, newhiwat, RLIM_INFINITY); sowwakeup_locked(so2); unp->unp_mbcnt = mbcnt; unp->unp_cc = sbcc; UNP_PCB_UNLOCK(unp); return (0); } /* pru_rcvoob is EOPNOTSUPP */ static int uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { struct unpcb *unp, *unp2; struct socket *so2; u_int mbcnt, sbcc; u_long newhiwat; int error = 0; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_send: unp == NULL")); if (flags & PRUS_OOB) { error = EOPNOTSUPP; goto release; } if (control != NULL && (error = unp_internalize(&control, td))) goto release; if ((nam != NULL) || (flags & PRUS_EOF)) UNP_GLOBAL_WLOCK(); else UNP_GLOBAL_RLOCK(); switch (so->so_type) { case SOCK_DGRAM: { const struct sockaddr *from; unp2 = unp->unp_conn; if (nam != NULL) { UNP_GLOBAL_WLOCK_ASSERT(); if (unp2 != NULL) { error = EISCONN; break; } error = unp_connect(so, nam, td); if (error) break; unp2 = unp->unp_conn; } /* * Because connect() and send() are non-atomic in a sendto() * with a target address, it's possible that the socket will * have disconnected before the send() can run. In that case * return the slightly counter-intuitive but otherwise * correct error that the socket is not connected. */ if (unp2 == NULL) { error = ENOTCONN; break; } /* Lockless read. */ if (unp2->unp_flags & UNP_WANTCRED) control = unp_addsockcred(td, control); UNP_PCB_LOCK(unp); if (unp->unp_addr != NULL) from = (struct sockaddr *)unp->unp_addr; else from = &sun_noname; so2 = unp2->unp_socket; SOCKBUF_LOCK(&so2->so_rcv); if (sbappendaddr_locked(&so2->so_rcv, from, m, control)) { sorwakeup_locked(so2); m = NULL; control = NULL; } else { SOCKBUF_UNLOCK(&so2->so_rcv); error = ENOBUFS; } if (nam != NULL) { UNP_GLOBAL_WLOCK_ASSERT(); UNP_PCB_LOCK(unp2); unp_disconnect(unp, unp2); UNP_PCB_UNLOCK(unp2); } UNP_PCB_UNLOCK(unp); break; } case SOCK_STREAM: /* * Connect if not connected yet. * * Note: A better implementation would complain if not equal * to the peer's address. */ if ((so->so_state & SS_ISCONNECTED) == 0) { if (nam != NULL) { UNP_GLOBAL_WLOCK_ASSERT(); error = unp_connect(so, nam, td); if (error) break; /* XXX */ } else { error = ENOTCONN; break; } } /* Lockless read. */ if (so->so_snd.sb_state & SBS_CANTSENDMORE) { error = EPIPE; break; } /* * Because connect() and send() are non-atomic in a sendto() * with a target address, it's possible that the socket will * have disconnected before the send() can run. In that case * return the slightly counter-intuitive but otherwise * correct error that the socket is not connected. * * Locking here must be done carefully: the global lock * prevents interconnections between unpcbs from changing, so * we can traverse from unp to unp2 without acquiring unp's * lock. Socket buffer locks follow unpcb locks, so we can * acquire both remote and lock socket buffer locks. */ unp2 = unp->unp_conn; if (unp2 == NULL) { error = ENOTCONN; break; } so2 = unp2->unp_socket; UNP_PCB_LOCK(unp2); SOCKBUF_LOCK(&so2->so_rcv); if (unp2->unp_flags & UNP_WANTCRED) { /* * Credentials are passed only once on SOCK_STREAM. */ unp2->unp_flags &= ~UNP_WANTCRED; control = unp_addsockcred(td, control); } /* * Send to paired receive port, and then reduce send buffer * hiwater marks to maintain backpressure. Wake up readers. */ if (control != NULL) { if (sbappendcontrol_locked(&so2->so_rcv, m, control)) control = NULL; } else sbappend_locked(&so2->so_rcv, m); mbcnt = so2->so_rcv.sb_mbcnt - unp2->unp_mbcnt; unp2->unp_mbcnt = so2->so_rcv.sb_mbcnt; sbcc = so2->so_rcv.sb_cc; sorwakeup_locked(so2); SOCKBUF_LOCK(&so->so_snd); newhiwat = so->so_snd.sb_hiwat - (sbcc - unp2->unp_cc); (void)chgsbsize(so->so_cred->cr_uidinfo, &so->so_snd.sb_hiwat, newhiwat, RLIM_INFINITY); so->so_snd.sb_mbmax -= mbcnt; SOCKBUF_UNLOCK(&so->so_snd); unp2->unp_cc = sbcc; UNP_PCB_UNLOCK(unp2); m = NULL; break; default: panic("uipc_send unknown socktype"); } /* * SEND_EOF is equivalent to a SEND followed by a SHUTDOWN. */ if (flags & PRUS_EOF) { UNP_PCB_LOCK(unp); socantsendmore(so); unp_shutdown(unp); UNP_PCB_UNLOCK(unp); } if ((nam != NULL) || (flags & PRUS_EOF)) UNP_GLOBAL_WUNLOCK(); else UNP_GLOBAL_RUNLOCK(); if (control != NULL && error != 0) unp_dispose(control); release: if (control != NULL) m_freem(control); if (m != NULL) m_freem(m); return (error); } static int uipc_sense(struct socket *so, struct stat *sb) { struct unpcb *unp, *unp2; struct socket *so2; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_sense: unp == NULL")); sb->st_blksize = so->so_snd.sb_hiwat; UNP_GLOBAL_RLOCK(); UNP_PCB_LOCK(unp); unp2 = unp->unp_conn; if (so->so_type == SOCK_STREAM && unp2 != NULL) { so2 = unp2->unp_socket; sb->st_blksize += so2->so_rcv.sb_cc; } sb->st_dev = NODEV; if (unp->unp_ino == 0) unp->unp_ino = (++unp_ino == 0) ? ++unp_ino : unp_ino; sb->st_ino = unp->unp_ino; UNP_PCB_UNLOCK(unp); UNP_GLOBAL_RUNLOCK(); return (0); } static int uipc_shutdown(struct socket *so) { struct unpcb *unp; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_shutdown: unp == NULL")); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); socantsendmore(so); unp_shutdown(unp); UNP_PCB_UNLOCK(unp); UNP_GLOBAL_WUNLOCK(); return (0); } static int uipc_sockaddr(struct socket *so, struct sockaddr **nam) { struct unpcb *unp; const struct sockaddr *sa; unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_sockaddr: unp == NULL")); *nam = malloc(sizeof(struct sockaddr_un), M_SONAME, M_WAITOK); UNP_PCB_LOCK(unp); if (unp->unp_addr != NULL) sa = (struct sockaddr *) unp->unp_addr; else sa = &sun_noname; bcopy(sa, *nam, sa->sa_len); UNP_PCB_UNLOCK(unp); return (0); } struct pr_usrreqs uipc_usrreqs = { .pru_abort = uipc_abort, .pru_accept = uipc_accept, .pru_attach = uipc_attach, .pru_bind = uipc_bind, .pru_connect = uipc_connect, .pru_connect2 = uipc_connect2, .pru_detach = uipc_detach, .pru_disconnect = uipc_disconnect, .pru_listen = uipc_listen, .pru_peeraddr = uipc_peeraddr, .pru_rcvd = uipc_rcvd, .pru_send = uipc_send, .pru_sense = uipc_sense, .pru_shutdown = uipc_shutdown, .pru_sockaddr = uipc_sockaddr, .pru_close = uipc_close, }; int uipc_ctloutput(struct socket *so, struct sockopt *sopt) { struct unpcb *unp; struct xucred xu; int error, optval; if (sopt->sopt_level != 0) return (EINVAL); unp = sotounpcb(so); KASSERT(unp != NULL, ("uipc_ctloutput: unp == NULL")); error = 0; switch (sopt->sopt_dir) { case SOPT_GET: switch (sopt->sopt_name) { case LOCAL_PEERCRED: UNP_PCB_LOCK(unp); if (unp->unp_flags & UNP_HAVEPC) xu = unp->unp_peercred; else { if (so->so_type == SOCK_STREAM) error = ENOTCONN; else error = EINVAL; } UNP_PCB_UNLOCK(unp); if (error == 0) error = sooptcopyout(sopt, &xu, sizeof(xu)); break; case LOCAL_CREDS: /* Unocked read. */ optval = unp->unp_flags & UNP_WANTCRED ? 1 : 0; error = sooptcopyout(sopt, &optval, sizeof(optval)); break; case LOCAL_CONNWAIT: /* Unocked read. */ optval = unp->unp_flags & UNP_CONNWAIT ? 1 : 0; error = sooptcopyout(sopt, &optval, sizeof(optval)); break; default: error = EOPNOTSUPP; break; } break; case SOPT_SET: switch (sopt->sopt_name) { case LOCAL_CREDS: case LOCAL_CONNWAIT: error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); if (error) break; #define OPTSET(bit) do { \ UNP_PCB_LOCK(unp); \ if (optval) \ unp->unp_flags |= bit; \ else \ unp->unp_flags &= ~bit; \ UNP_PCB_UNLOCK(unp); \ } while (0) switch (sopt->sopt_name) { case LOCAL_CREDS: OPTSET(UNP_WANTCRED); break; case LOCAL_CONNWAIT: OPTSET(UNP_CONNWAIT); break; default: break; } break; #undef OPTSET default: error = ENOPROTOOPT; break; } break; default: error = EOPNOTSUPP; break; } return (error); } static int unp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct sockaddr_un *soun = (struct sockaddr_un *)nam; struct vnode *vp; struct socket *so2, *so3; struct unpcb *unp, *unp2, *unp3; int error, len, vfslocked; struct nameidata nd; char buf[SOCK_MAXADDRLEN]; struct sockaddr *sa; UNP_GLOBAL_WLOCK_ASSERT(); UNP_GLOBAL_WUNLOCK(); unp = sotounpcb(so); KASSERT(unp != NULL, ("unp_connect: unp == NULL")); len = nam->sa_len - offsetof(struct sockaddr_un, sun_path); if (len <= 0) return (EINVAL); strlcpy(buf, soun->sun_path, len + 1); UNP_PCB_LOCK(unp); if (unp->unp_flags & UNP_CONNECTING) { UNP_PCB_UNLOCK(unp); return (EALREADY); } unp->unp_flags |= UNP_CONNECTING; UNP_PCB_UNLOCK(unp); sa = malloc(sizeof(struct sockaddr_un), M_SONAME, M_WAITOK); NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW | LOCKLEAF, UIO_SYSSPACE, buf, td); error = namei(&nd); if (error) vp = NULL; else vp = nd.ni_vp; ASSERT_VOP_LOCKED(vp, "unp_connect"); vfslocked = NDHASGIANT(&nd); NDFREE(&nd, NDF_ONLY_PNBUF); if (error) goto bad; if (vp->v_type != VSOCK) { error = ENOTSOCK; goto bad; } #ifdef MAC error = mac_check_vnode_open(td->td_ucred, vp, VWRITE | VREAD); if (error) goto bad; #endif error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td); if (error) goto bad; VFS_UNLOCK_GIANT(vfslocked); unp = sotounpcb(so); KASSERT(unp != NULL, ("unp_connect: unp == NULL")); /* * Lock global lock for two reasons: make sure v_socket is stable, * and to protect simultaneous locking of multiple pcbs. */ UNP_GLOBAL_WLOCK(); so2 = vp->v_socket; if (so2 == NULL) { error = ECONNREFUSED; goto bad2; } if (so->so_type != so2->so_type) { error = EPROTOTYPE; goto bad2; } if (so->so_proto->pr_flags & PR_CONNREQUIRED) { if (so2->so_options & SO_ACCEPTCONN) { /* * We can't drop the global lock here or 'so2' may * become invalid. As a result, we need to handle * possibly lock recursion in uipc_attach. */ so3 = sonewconn(so2, 0); } else so3 = NULL; if (so3 == NULL) { error = ECONNREFUSED; goto bad2; } unp = sotounpcb(so); unp2 = sotounpcb(so2); unp3 = sotounpcb(so3); UNP_PCB_LOCK(unp); UNP_PCB_LOCK(unp2); UNP_PCB_LOCK(unp3); if (unp2->unp_addr != NULL) { bcopy(unp2->unp_addr, sa, unp2->unp_addr->sun_len); unp3->unp_addr = (struct sockaddr_un *) sa; sa = NULL; } /* * unp_peercred management: * * The connecter's (client's) credentials are copied from its * process structure at the time of connect() (which is now). */ cru2x(td->td_ucred, &unp3->unp_peercred); unp3->unp_flags |= UNP_HAVEPC; /* * The receiver's (server's) credentials are copied from the * unp_peercred member of socket on which the former called * listen(); uipc_listen() cached that process's credentials * at that time so we can use them now. */ KASSERT(unp2->unp_flags & UNP_HAVEPCCACHED, ("unp_connect: listener without cached peercred")); memcpy(&unp->unp_peercred, &unp2->unp_peercred, sizeof(unp->unp_peercred)); unp->unp_flags |= UNP_HAVEPC; if (unp2->unp_flags & UNP_WANTCRED) unp3->unp_flags |= UNP_WANTCRED; UNP_PCB_UNLOCK(unp3); UNP_PCB_UNLOCK(unp2); UNP_PCB_UNLOCK(unp); #ifdef MAC SOCK_LOCK(so); mac_set_socket_peer_from_socket(so, so3); mac_set_socket_peer_from_socket(so3, so); SOCK_UNLOCK(so); #endif so2 = so3; } unp = sotounpcb(so); KASSERT(unp != NULL, ("unp_connect: unp == NULL")); unp2 = sotounpcb(so2); KASSERT(unp2 != NULL, ("unp_connect: unp2 == NULL")); UNP_PCB_LOCK(unp); UNP_PCB_LOCK(unp2); error = unp_connect2(so, so2, PRU_CONNECT); UNP_PCB_UNLOCK(unp2); UNP_PCB_UNLOCK(unp); bad2: UNP_GLOBAL_WUNLOCK(); if (vfslocked) /* * Giant has been previously acquired. This means filesystem * isn't MPSAFE. Do it once again. */ mtx_lock(&Giant); bad: if (vp != NULL) vput(vp); VFS_UNLOCK_GIANT(vfslocked); free(sa, M_SONAME); UNP_GLOBAL_WLOCK(); UNP_PCB_LOCK(unp); unp->unp_flags &= ~UNP_CONNECTING; UNP_PCB_UNLOCK(unp); return (error); } static int unp_connect2(struct socket *so, struct socket *so2, int req) { struct unpcb *unp; struct unpcb *unp2; unp = sotounpcb(so); KASSERT(unp != NULL, ("unp_connect2: unp == NULL")); unp2 = sotounpcb(so2); KASSERT(unp2 != NULL, ("unp_connect2: unp2 == NULL")); UNP_GLOBAL_WLOCK_ASSERT(); UNP_PCB_LOCK_ASSERT(unp); UNP_PCB_LOCK_ASSERT(unp2); if (so2->so_type != so->so_type) return (EPROTOTYPE); unp->unp_conn = unp2; switch (so->so_type) { case SOCK_DGRAM: LIST_INSERT_HEAD(&unp2->unp_refs, unp, unp_reflink); soisconnected(so); break; case SOCK_STREAM: unp2->unp_conn = unp; if (req == PRU_CONNECT && ((unp->unp_flags | unp2->unp_flags) & UNP_CONNWAIT)) soisconnecting(so); else soisconnected(so); soisconnected(so2); break; default: panic("unp_connect2"); } return (0); } static void unp_disconnect(struct unpcb *unp, struct unpcb *unp2) { struct socket *so; KASSERT(unp2 != NULL, ("unp_disconnect: unp2 == NULL")); UNP_GLOBAL_WLOCK_ASSERT(); UNP_PCB_LOCK_ASSERT(unp); UNP_PCB_LOCK_ASSERT(unp2); unp->unp_conn = NULL; switch (unp->unp_socket->so_type) { case SOCK_DGRAM: LIST_REMOVE(unp, unp_reflink); so = unp->unp_socket; SOCK_LOCK(so); so->so_state &= ~SS_ISCONNECTED; SOCK_UNLOCK(so); break; case SOCK_STREAM: soisdisconnected(unp->unp_socket); unp2->unp_conn = NULL; soisdisconnected(unp2->unp_socket); break; } } /* * unp_pcblist() walks the global list of struct unpcb's to generate a * pointer list, bumping the refcount on each unpcb. It then copies them out * sequentially, validating the generation number on each to see if it has * been detached. All of this is necessary because copyout() may sleep on * disk I/O. */ static int unp_pcblist(SYSCTL_HANDLER_ARGS) { int error, i, n; int freeunp; struct unpcb *unp, **unp_list; unp_gen_t gencnt; struct xunpgen *xug; struct unp_head *head; struct xunpcb *xu; head = ((intptr_t)arg1 == SOCK_DGRAM ? &unp_dhead : &unp_shead); /* * The process of preparing the PCB list is too time-consuming and * resource-intensive to repeat twice on every request. */ if (req->oldptr == NULL) { n = unp_count; req->oldidx = 2 * (sizeof *xug) + (n + n/8) * sizeof(struct xunpcb); return (0); } if (req->newptr != NULL) return (EPERM); /* * OK, now we're committed to doing something. */ xug = malloc(sizeof(*xug), M_TEMP, M_WAITOK); UNP_GLOBAL_RLOCK(); gencnt = unp_gencnt; n = unp_count; UNP_GLOBAL_RUNLOCK(); xug->xug_len = sizeof *xug; xug->xug_count = n; xug->xug_gen = gencnt; xug->xug_sogen = so_gencnt; error = SYSCTL_OUT(req, xug, sizeof *xug); if (error) { free(xug, M_TEMP); return (error); } unp_list = malloc(n * sizeof *unp_list, M_TEMP, M_WAITOK); UNP_GLOBAL_RLOCK(); for (unp = LIST_FIRST(head), i = 0; unp && i < n; unp = LIST_NEXT(unp, unp_link)) { UNP_PCB_LOCK(unp); if (unp->unp_gencnt <= gencnt) { if (cr_cansee(req->td->td_ucred, unp->unp_socket->so_cred)) { UNP_PCB_UNLOCK(unp); continue; } unp_list[i++] = unp; unp->unp_refcount++; } UNP_PCB_UNLOCK(unp); } UNP_GLOBAL_RUNLOCK(); n = i; /* In case we lost some during malloc. */ error = 0; xu = malloc(sizeof(*xu), M_TEMP, M_WAITOK | M_ZERO); for (i = 0; i < n; i++) { unp = unp_list[i]; UNP_PCB_LOCK(unp); unp->unp_refcount--; if (unp->unp_refcount != 0 && unp->unp_gencnt <= gencnt) { xu->xu_len = sizeof *xu; xu->xu_unpp = unp; /* * XXX - need more locking here to protect against * connect/disconnect races for SMP. */ if (unp->unp_addr != NULL) bcopy(unp->unp_addr, &xu->xu_addr, unp->unp_addr->sun_len); if (unp->unp_conn != NULL && unp->unp_conn->unp_addr != NULL) bcopy(unp->unp_conn->unp_addr, &xu->xu_caddr, unp->unp_conn->unp_addr->sun_len); bcopy(unp, &xu->xu_unp, sizeof *unp); sotoxsocket(unp->unp_socket, &xu->xu_socket); UNP_PCB_UNLOCK(unp); error = SYSCTL_OUT(req, xu, sizeof *xu); } else { freeunp = (unp->unp_refcount == 0); UNP_PCB_UNLOCK(unp); if (freeunp) { UNP_PCB_LOCK_DESTROY(unp); uma_zfree(unp_zone, unp); } } } free(xu, M_TEMP); if (!error) { /* * Give the user an updated idea of our state. If the * generation differs from what we told her before, she knows * that something happened while we were processing this * request, and it might be necessary to retry. */ xug->xug_gen = unp_gencnt; xug->xug_sogen = so_gencnt; xug->xug_count = unp_count; error = SYSCTL_OUT(req, xug, sizeof *xug); } free(unp_list, M_TEMP); free(xug, M_TEMP); return (error); } SYSCTL_PROC(_net_local_dgram, OID_AUTO, pcblist, CTLFLAG_RD, (caddr_t)(long)SOCK_DGRAM, 0, unp_pcblist, "S,xunpcb", "List of active local datagram sockets"); SYSCTL_PROC(_net_local_stream, OID_AUTO, pcblist, CTLFLAG_RD, (caddr_t)(long)SOCK_STREAM, 0, unp_pcblist, "S,xunpcb", "List of active local stream sockets"); static void unp_shutdown(struct unpcb *unp) { struct unpcb *unp2; struct socket *so; UNP_GLOBAL_WLOCK_ASSERT(); UNP_PCB_LOCK_ASSERT(unp); unp2 = unp->unp_conn; if (unp->unp_socket->so_type == SOCK_STREAM && unp2 != NULL) { so = unp2->unp_socket; if (so != NULL) socantrcvmore(so); } } static void unp_drop(struct unpcb *unp, int errno) { struct socket *so = unp->unp_socket; struct unpcb *unp2; UNP_GLOBAL_WLOCK_ASSERT(); UNP_PCB_LOCK_ASSERT(unp); so->so_error = errno; unp2 = unp->unp_conn; if (unp2 == NULL) return; UNP_PCB_LOCK(unp2); unp_disconnect(unp, unp2); UNP_PCB_UNLOCK(unp2); } static void unp_freerights(struct file **rp, int fdcount) { int i; struct file *fp; for (i = 0; i < fdcount; i++) { /* * Zero the pointer before calling unp_discard since it may * end up in unp_gc().. * * XXXRW: This is less true than it used to be. */ fp = *rp; *rp++ = NULL; unp_discard(fp); } } int unp_externalize(struct mbuf *control, struct mbuf **controlp) { struct thread *td = curthread; /* XXX */ struct cmsghdr *cm = mtod(control, struct cmsghdr *); int i; int *fdp; struct file **rp; struct file *fp; void *data; socklen_t clen = control->m_len, datalen; int error, newfds; int f; u_int newlen; UNP_GLOBAL_UNLOCK_ASSERT(); error = 0; if (controlp != NULL) /* controlp == NULL => free control messages */ *controlp = NULL; while (cm != NULL) { if (sizeof(*cm) > clen || cm->cmsg_len > clen) { error = EINVAL; break; } data = CMSG_DATA(cm); datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; if (cm->cmsg_level == SOL_SOCKET && cm->cmsg_type == SCM_RIGHTS) { newfds = datalen / sizeof(struct file *); rp = data; /* If we're not outputting the descriptors free them. */ if (error || controlp == NULL) { unp_freerights(rp, newfds); goto next; } FILEDESC_XLOCK(td->td_proc->p_fd); /* if the new FD's will not fit free them. */ if (!fdavail(td, newfds)) { FILEDESC_XUNLOCK(td->td_proc->p_fd); error = EMSGSIZE; unp_freerights(rp, newfds); goto next; } /* * Now change each pointer to an fd in the global * table to an integer that is the index to the local * fd table entry that we set up to point to the * global one we are transferring. */ newlen = newfds * sizeof(int); *controlp = sbcreatecontrol(NULL, newlen, SCM_RIGHTS, SOL_SOCKET); if (*controlp == NULL) { FILEDESC_XUNLOCK(td->td_proc->p_fd); error = E2BIG; unp_freerights(rp, newfds); goto next; } fdp = (int *) CMSG_DATA(mtod(*controlp, struct cmsghdr *)); for (i = 0; i < newfds; i++) { if (fdalloc(td, 0, &f)) panic("unp_externalize fdalloc failed"); fp = *rp++; td->td_proc->p_fd->fd_ofiles[f] = fp; FILE_LOCK(fp); fp->f_msgcount--; FILE_UNLOCK(fp); unp_rights--; *fdp++ = f; } FILEDESC_XUNLOCK(td->td_proc->p_fd); } else { /* We can just copy anything else across. */ if (error || controlp == NULL) goto next; *controlp = sbcreatecontrol(NULL, datalen, cm->cmsg_type, cm->cmsg_level); if (*controlp == NULL) { error = ENOBUFS; goto next; } bcopy(data, CMSG_DATA(mtod(*controlp, struct cmsghdr *)), datalen); } controlp = &(*controlp)->m_next; next: if (CMSG_SPACE(datalen) < clen) { clen -= CMSG_SPACE(datalen); cm = (struct cmsghdr *) ((caddr_t)cm + CMSG_SPACE(datalen)); } else { clen = 0; cm = NULL; } } m_freem(control); return (error); } static void unp_zone_change(void *tag) { uma_zone_set_max(unp_zone, maxsockets); } void unp_init(void) { unp_zone = uma_zcreate("unpcb", sizeof(struct unpcb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); if (unp_zone == NULL) panic("unp_init"); uma_zone_set_max(unp_zone, maxsockets); EVENTHANDLER_REGISTER(maxsockets_change, unp_zone_change, NULL, EVENTHANDLER_PRI_ANY); LIST_INIT(&unp_dhead); LIST_INIT(&unp_shead); TASK_INIT(&unp_gc_task, 0, unp_gc, NULL); UNP_GLOBAL_LOCK_INIT(); } static int unp_internalize(struct mbuf **controlp, struct thread *td) { struct mbuf *control = *controlp; struct proc *p = td->td_proc; struct filedesc *fdescp = p->p_fd; struct cmsghdr *cm = mtod(control, struct cmsghdr *); struct cmsgcred *cmcred; struct file **rp; struct file *fp; struct timeval *tv; int i, fd, *fdp; void *data; socklen_t clen = control->m_len, datalen; int error, oldfds; u_int newlen; UNP_GLOBAL_UNLOCK_ASSERT(); error = 0; *controlp = NULL; while (cm != NULL) { if (sizeof(*cm) > clen || cm->cmsg_level != SOL_SOCKET || cm->cmsg_len > clen) { error = EINVAL; goto out; } data = CMSG_DATA(cm); datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; switch (cm->cmsg_type) { /* * Fill in credential information. */ case SCM_CREDS: *controlp = sbcreatecontrol(NULL, sizeof(*cmcred), SCM_CREDS, SOL_SOCKET); if (*controlp == NULL) { error = ENOBUFS; goto out; } cmcred = (struct cmsgcred *) CMSG_DATA(mtod(*controlp, struct cmsghdr *)); cmcred->cmcred_pid = p->p_pid; cmcred->cmcred_uid = td->td_ucred->cr_ruid; cmcred->cmcred_gid = td->td_ucred->cr_rgid; cmcred->cmcred_euid = td->td_ucred->cr_uid; cmcred->cmcred_ngroups = MIN(td->td_ucred->cr_ngroups, CMGROUP_MAX); for (i = 0; i < cmcred->cmcred_ngroups; i++) cmcred->cmcred_groups[i] = td->td_ucred->cr_groups[i]; break; case SCM_RIGHTS: oldfds = datalen / sizeof (int); /* * Check that all the FDs passed in refer to legal * files. If not, reject the entire operation. */ fdp = data; FILEDESC_SLOCK(fdescp); for (i = 0; i < oldfds; i++) { fd = *fdp++; if ((unsigned)fd >= fdescp->fd_nfiles || fdescp->fd_ofiles[fd] == NULL) { FILEDESC_SUNLOCK(fdescp); error = EBADF; goto out; } fp = fdescp->fd_ofiles[fd]; if (!(fp->f_ops->fo_flags & DFLAG_PASSABLE)) { FILEDESC_SUNLOCK(fdescp); error = EOPNOTSUPP; goto out; } } /* * Now replace the integer FDs with pointers to * the associated global file table entry.. */ newlen = oldfds * sizeof(struct file *); *controlp = sbcreatecontrol(NULL, newlen, SCM_RIGHTS, SOL_SOCKET); if (*controlp == NULL) { FILEDESC_SUNLOCK(fdescp); error = E2BIG; goto out; } fdp = data; rp = (struct file **) CMSG_DATA(mtod(*controlp, struct cmsghdr *)); for (i = 0; i < oldfds; i++) { fp = fdescp->fd_ofiles[*fdp++]; *rp++ = fp; FILE_LOCK(fp); fp->f_count++; fp->f_msgcount++; FILE_UNLOCK(fp); unp_rights++; } FILEDESC_SUNLOCK(fdescp); break; case SCM_TIMESTAMP: *controlp = sbcreatecontrol(NULL, sizeof(*tv), SCM_TIMESTAMP, SOL_SOCKET); if (*controlp == NULL) { error = ENOBUFS; goto out; } tv = (struct timeval *) CMSG_DATA(mtod(*controlp, struct cmsghdr *)); microtime(tv); break; default: error = EINVAL; goto out; } controlp = &(*controlp)->m_next; if (CMSG_SPACE(datalen) < clen) { clen -= CMSG_SPACE(datalen); cm = (struct cmsghdr *) ((caddr_t)cm + CMSG_SPACE(datalen)); } else { clen = 0; cm = NULL; } } out: m_freem(control); return (error); } static struct mbuf * unp_addsockcred(struct thread *td, struct mbuf *control) { struct mbuf *m, *n, *n_prev; struct sockcred *sc; const struct cmsghdr *cm; int ngroups; int i; ngroups = MIN(td->td_ucred->cr_ngroups, CMGROUP_MAX); m = sbcreatecontrol(NULL, SOCKCREDSIZE(ngroups), SCM_CREDS, SOL_SOCKET); if (m == NULL) return (control); sc = (struct sockcred *) CMSG_DATA(mtod(m, struct cmsghdr *)); sc->sc_uid = td->td_ucred->cr_ruid; sc->sc_euid = td->td_ucred->cr_uid; sc->sc_gid = td->td_ucred->cr_rgid; sc->sc_egid = td->td_ucred->cr_gid; sc->sc_ngroups = ngroups; for (i = 0; i < sc->sc_ngroups; i++) sc->sc_groups[i] = td->td_ucred->cr_groups[i]; /* * Unlink SCM_CREDS control messages (struct cmsgcred), since just * created SCM_CREDS control message (struct sockcred) has another * format. */ if (control != NULL) for (n = control, n_prev = NULL; n != NULL;) { cm = mtod(n, struct cmsghdr *); if (cm->cmsg_level == SOL_SOCKET && cm->cmsg_type == SCM_CREDS) { if (n_prev == NULL) control = n->m_next; else n_prev->m_next = n->m_next; n = m_free(n); } else { n_prev = n; n = n->m_next; } } /* Prepend it to the head. */ m->m_next = control; return (m); } /* * unp_defer indicates whether additional work has been defered for a future * pass through unp_gc(). It is thread local and does not require explicit * synchronization. */ static int unp_defer; static int unp_taskcount; SYSCTL_INT(_net_local, OID_AUTO, taskcount, CTLFLAG_RD, &unp_taskcount, 0, ""); static int unp_recycled; SYSCTL_INT(_net_local, OID_AUTO, recycled, CTLFLAG_RD, &unp_recycled, 0, ""); static void unp_gc(__unused void *arg, int pending) { struct file *fp, *nextfp; struct socket *so; struct file **extra_ref, **fpp; int nunref, i; int nfiles_snap; int nfiles_slack = 20; unp_taskcount++; unp_defer = 0; /* * Before going through all this, set all FDs to be NOT deferred and * NOT externally accessible. */ sx_slock(&filelist_lock); LIST_FOREACH(fp, &filehead, f_list) fp->f_gcflag &= ~(FMARK|FDEFER); do { KASSERT(unp_defer >= 0, ("unp_gc: unp_defer %d", unp_defer)); LIST_FOREACH(fp, &filehead, f_list) { FILE_LOCK(fp); /* * If the file is not open, skip it -- could be a * file in the process of being opened, or in the * process of being closed. If the file is * "closing", it may have been marked for deferred * consideration. Clear the flag now if so. */ if (fp->f_count == 0) { if (fp->f_gcflag & FDEFER) unp_defer--; fp->f_gcflag &= ~(FMARK|FDEFER); FILE_UNLOCK(fp); continue; } /* * If we already marked it as 'defer' in a * previous pass, then try to process it this * time and un-mark it. */ if (fp->f_gcflag & FDEFER) { fp->f_gcflag &= ~FDEFER; unp_defer--; } else { /* * If it's not deferred, then check if it's * already marked.. if so skip it */ if (fp->f_gcflag & FMARK) { FILE_UNLOCK(fp); continue; } /* * If all references are from messages in * transit, then skip it. it's not externally * accessible. */ if (fp->f_count == fp->f_msgcount) { FILE_UNLOCK(fp); continue; } /* * If it got this far then it must be * externally accessible. */ fp->f_gcflag |= FMARK; } /* * Either it was deferred, or it is externally * accessible and not already marked so. Now check * if it is possibly one of OUR sockets. */ if (fp->f_type != DTYPE_SOCKET || (so = fp->f_data) == NULL) { FILE_UNLOCK(fp); continue; } if (so->so_proto->pr_domain != &localdomain || (so->so_proto->pr_flags & PR_RIGHTS) == 0) { FILE_UNLOCK(fp); continue; } /* * Tell any other threads that do a subsequent * fdrop() that we are scanning the message * buffers. */ fp->f_gcflag |= FWAIT; FILE_UNLOCK(fp); /* * So, Ok, it's one of our sockets and it IS * externally accessible (or was deferred). Now we * look to see if we hold any file descriptors in its * message buffers. Follow those links and mark them * as accessible too. */ SOCKBUF_LOCK(&so->so_rcv); unp_scan(so->so_rcv.sb_mb, unp_mark); SOCKBUF_UNLOCK(&so->so_rcv); /* * Wake up any threads waiting in fdrop(). */ FILE_LOCK(fp); fp->f_gcflag &= ~FWAIT; wakeup(&fp->f_gcflag); FILE_UNLOCK(fp); } } while (unp_defer); sx_sunlock(&filelist_lock); /* * XXXRW: The following comments need updating for a post-SMPng and * deferred unp_gc() world, but are still generally accurate. * * We grab an extra reference to each of the file table entries that * are not otherwise accessible and then free the rights that are * stored in messages on them. * * The bug in the orginal code is a little tricky, so I'll describe * what's wrong with it here. * * It is incorrect to simply unp_discard each entry for f_msgcount * times -- consider the case of sockets A and B that contain * references to each other. On a last close of some other socket, * we trigger a gc since the number of outstanding rights (unp_rights) * is non-zero. If during the sweep phase the gc code unp_discards, * we end up doing a (full) closef on the descriptor. A closef on A * results in the following chain. Closef calls soo_close, which * calls soclose. Soclose calls first (through the switch * uipc_usrreq) unp_detach, which re-invokes unp_gc. Unp_gc simply * returns because the previous instance had set unp_gcing, and we * return all the way back to soclose, which marks the socket with * SS_NOFDREF, and then calls sofree. Sofree calls sorflush to free * up the rights that are queued in messages on the socket A, i.e., * the reference on B. The sorflush calls via the dom_dispose switch * unp_dispose, which unp_scans with unp_discard. This second * instance of unp_discard just calls closef on B. * * Well, a similar chain occurs on B, resulting in a sorflush on B, * which results in another closef on A. Unfortunately, A is already * being closed, and the descriptor has already been marked with * SS_NOFDREF, and soclose panics at this point. * * Here, we first take an extra reference to each inaccessible * descriptor. Then, we call sorflush ourself, since we know it is a * Unix domain socket anyhow. After we destroy all the rights * carried in messages, we do a last closef to get rid of our extra * reference. This is the last close, and the unp_detach etc will * shut down the socket. * * 91/09/19, bsy@cs.cmu.edu */ again: nfiles_snap = openfiles + nfiles_slack; /* some slack */ extra_ref = malloc(nfiles_snap * sizeof(struct file *), M_TEMP, M_WAITOK); sx_slock(&filelist_lock); if (nfiles_snap < openfiles) { sx_sunlock(&filelist_lock); free(extra_ref, M_TEMP); nfiles_slack += 20; goto again; } for (nunref = 0, fp = LIST_FIRST(&filehead), fpp = extra_ref; fp != NULL; fp = nextfp) { nextfp = LIST_NEXT(fp, f_list); FILE_LOCK(fp); /* * If it's not open, skip it */ if (fp->f_count == 0) { FILE_UNLOCK(fp); continue; } /* * If all refs are from msgs, and it's not marked accessible * then it must be referenced from some unreachable cycle of * (shut-down) FDs, so include it in our list of FDs to * remove. */ if (fp->f_count == fp->f_msgcount && !(fp->f_gcflag & FMARK)) { *fpp++ = fp; nunref++; fp->f_count++; } FILE_UNLOCK(fp); } sx_sunlock(&filelist_lock); /* * For each FD on our hit list, do the following two things: */ for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp) { struct file *tfp = *fpp; FILE_LOCK(tfp); if (tfp->f_type == DTYPE_SOCKET && tfp->f_data != NULL) { FILE_UNLOCK(tfp); sorflush(tfp->f_data); } else { FILE_UNLOCK(tfp); } } for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp) { closef(*fpp, (struct thread *) NULL); unp_recycled++; } free(extra_ref, M_TEMP); } void unp_dispose(struct mbuf *m) { if (m) unp_scan(m, unp_discard); } static void unp_scan(struct mbuf *m0, void (*op)(struct file *)) { struct mbuf *m; struct file **rp; struct cmsghdr *cm; void *data; int i; socklen_t clen, datalen; int qfds; while (m0 != NULL) { for (m = m0; m; m = m->m_next) { if (m->m_type != MT_CONTROL) continue; cm = mtod(m, struct cmsghdr *); clen = m->m_len; while (cm != NULL) { if (sizeof(*cm) > clen || cm->cmsg_len > clen) break; data = CMSG_DATA(cm); datalen = (caddr_t)cm + cm->cmsg_len - (caddr_t)data; if (cm->cmsg_level == SOL_SOCKET && cm->cmsg_type == SCM_RIGHTS) { qfds = datalen / sizeof (struct file *); rp = data; for (i = 0; i < qfds; i++) (*op)(*rp++); } if (CMSG_SPACE(datalen) < clen) { clen -= CMSG_SPACE(datalen); cm = (struct cmsghdr *) ((caddr_t)cm + CMSG_SPACE(datalen)); } else { clen = 0; cm = NULL; } } } m0 = m0->m_act; } } static void unp_mark(struct file *fp) { /* XXXRW: Should probably assert file list lock here. */ if (fp->f_gcflag & FMARK) return; unp_defer++; fp->f_gcflag |= (FMARK|FDEFER); } static void unp_discard(struct file *fp) { UNP_GLOBAL_WLOCK(); FILE_LOCK(fp); fp->f_msgcount--; unp_rights--; FILE_UNLOCK(fp); UNP_GLOBAL_WUNLOCK(); (void) closef(fp, (struct thread *)NULL); } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_unpflags(int unp_flags) { int comma; comma = 0; if (unp_flags & UNP_HAVEPC) { db_printf("%sUNP_HAVEPC", comma ? ", " : ""); comma = 1; } if (unp_flags & UNP_HAVEPCCACHED) { db_printf("%sUNP_HAVEPCCACHED", comma ? ", " : ""); comma = 1; } if (unp_flags & UNP_WANTCRED) { db_printf("%sUNP_WANTCRED", comma ? ", " : ""); comma = 1; } if (unp_flags & UNP_CONNWAIT) { db_printf("%sUNP_CONNWAIT", comma ? ", " : ""); comma = 1; } if (unp_flags & UNP_CONNECTING) { db_printf("%sUNP_CONNECTING", comma ? ", " : ""); comma = 1; } if (unp_flags & UNP_BINDING) { db_printf("%sUNP_BINDING", comma ? ", " : ""); comma = 1; } } static void db_print_xucred(int indent, struct xucred *xu) { int comma, i; db_print_indent(indent); db_printf("cr_version: %u cr_uid: %u cr_ngroups: %d\n", xu->cr_version, xu->cr_uid, xu->cr_ngroups); db_print_indent(indent); db_printf("cr_groups: "); comma = 0; for (i = 0; i < xu->cr_ngroups; i++) { db_printf("%s%u", comma ? ", " : "", xu->cr_groups[i]); comma = 1; } db_printf("\n"); } static void db_print_unprefs(int indent, struct unp_head *uh) { struct unpcb *unp; int counter; counter = 0; LIST_FOREACH(unp, uh, unp_reflink) { if (counter % 4 == 0) db_print_indent(indent); db_printf("%p ", unp); if (counter % 4 == 3) db_printf("\n"); counter++; } if (counter != 0 && counter % 4 != 0) db_printf("\n"); } DB_SHOW_COMMAND(unpcb, db_show_unpcb) { struct unpcb *unp; if (!have_addr) { db_printf("usage: show unpcb \n"); return; } unp = (struct unpcb *)addr; db_printf("unp_socket: %p unp_vnode: %p\n", unp->unp_socket, unp->unp_vnode); db_printf("unp_ino: %d unp_conn: %p\n", unp->unp_ino, unp->unp_conn); db_printf("unp_refs:\n"); db_print_unprefs(2, &unp->unp_refs); /* XXXRW: Would be nice to print the full address, if any. */ db_printf("unp_addr: %p\n", unp->unp_addr); db_printf("unp_cc: %d unp_mbcnt: %d unp_gencnt: %llu\n", unp->unp_cc, unp->unp_mbcnt, (unsigned long long)unp->unp_gencnt); db_printf("unp_flags: %x (", unp->unp_flags); db_print_unpflags(unp->unp_flags); db_printf(")\n"); db_printf("unp_peercred:\n"); db_print_xucred(2, &unp->unp_peercred); db_printf("unp_refcount: %u\n", unp->unp_refcount); } #endif Index: head/sys/kern/vfs_vnops.c =================================================================== --- head/sys/kern/vfs_vnops.c (revision 171598) +++ head/sys/kern/vfs_vnops.c (revision 171599) @@ -1,1252 +1,1252 @@ /*- * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * 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. * 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. * * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94 */ #include __FBSDID("$FreeBSD$"); #include "opt_mac.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static fo_rdwr_t vn_read; static fo_rdwr_t vn_write; static fo_ioctl_t vn_ioctl; static fo_poll_t vn_poll; static fo_kqfilter_t vn_kqfilter; static fo_stat_t vn_statfile; static fo_close_t vn_closefile; struct fileops vnops = { .fo_read = vn_read, .fo_write = vn_write, .fo_ioctl = vn_ioctl, .fo_poll = vn_poll, .fo_kqfilter = vn_kqfilter, .fo_stat = vn_statfile, .fo_close = vn_closefile, .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE }; int vn_open(ndp, flagp, cmode, fp) struct nameidata *ndp; int *flagp, cmode; struct file *fp; { struct thread *td = ndp->ni_cnd.cn_thread; return (vn_open_cred(ndp, flagp, cmode, td->td_ucred, fp)); } /* * Common code for vnode open operations. * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. * * Note that this does NOT free nameidata for the successful case, * due to the NDINIT being done elsewhere. */ int vn_open_cred(ndp, flagp, cmode, cred, fp) struct nameidata *ndp; int *flagp, cmode; struct ucred *cred; struct file *fp; { struct vnode *vp; struct mount *mp; struct thread *td = ndp->ni_cnd.cn_thread; struct vattr vat; struct vattr *vap = &vat; int mode, fmode, error; int vfslocked, mpsafe; mpsafe = ndp->ni_cnd.cn_flags & MPSAFE; restart: vfslocked = 0; fmode = *flagp; if (fmode & O_CREAT) { ndp->ni_cnd.cn_nameiop = CREATE; ndp->ni_cnd.cn_flags = ISOPEN | LOCKPARENT | LOCKLEAF | MPSAFE | AUDITVNODE1; if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0) ndp->ni_cnd.cn_flags |= FOLLOW; bwillwrite(); if ((error = namei(ndp)) != 0) return (error); vfslocked = NDHASGIANT(ndp); if (!mpsafe) ndp->ni_cnd.cn_flags &= ~MPSAFE; if (ndp->ni_vp == NULL) { VATTR_NULL(vap); vap->va_type = VREG; vap->va_mode = cmode; if (fmode & O_EXCL) vap->va_vaflags |= VA_EXCLUSIVE; if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { NDFREE(ndp, NDF_ONLY_PNBUF); vput(ndp->ni_dvp); VFS_UNLOCK_GIANT(vfslocked); if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0) return (error); goto restart; } #ifdef MAC error = mac_check_vnode_create(cred, ndp->ni_dvp, &ndp->ni_cnd, vap); if (error == 0) { #endif VOP_LEASE(ndp->ni_dvp, td, cred, LEASE_WRITE); error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, &ndp->ni_cnd, vap); #ifdef MAC } #endif vput(ndp->ni_dvp); vn_finished_write(mp); if (error) { VFS_UNLOCK_GIANT(vfslocked); NDFREE(ndp, NDF_ONLY_PNBUF); return (error); } fmode &= ~O_TRUNC; vp = ndp->ni_vp; } else { if (ndp->ni_dvp == ndp->ni_vp) vrele(ndp->ni_dvp); else vput(ndp->ni_dvp); ndp->ni_dvp = NULL; vp = ndp->ni_vp; if (fmode & O_EXCL) { error = EEXIST; goto bad; } fmode &= ~O_CREAT; } } else { ndp->ni_cnd.cn_nameiop = LOOKUP; ndp->ni_cnd.cn_flags = ISOPEN | ((fmode & O_NOFOLLOW) ? NOFOLLOW : FOLLOW) | - LOCKSHARED | LOCKLEAF | MPSAFE | AUDITVNODE1; + LOCKLEAF | MPSAFE | AUDITVNODE1; if ((error = namei(ndp)) != 0) return (error); if (!mpsafe) ndp->ni_cnd.cn_flags &= ~MPSAFE; vfslocked = NDHASGIANT(ndp); vp = ndp->ni_vp; } if (vp->v_type == VLNK) { error = EMLINK; goto bad; } if (vp->v_type == VSOCK) { error = EOPNOTSUPP; goto bad; } mode = 0; if (fmode & (FWRITE | O_TRUNC)) { if (vp->v_type == VDIR) { error = EISDIR; goto bad; } mode |= VWRITE; } if (fmode & FREAD) mode |= VREAD; if (fmode & O_APPEND) mode |= VAPPEND; #ifdef MAC error = mac_check_vnode_open(cred, vp, mode); if (error) goto bad; #endif if ((fmode & O_CREAT) == 0) { if (mode & VWRITE) { error = vn_writechk(vp); if (error) goto bad; } if (mode) { error = VOP_ACCESS(vp, mode, cred, td); if (error) goto bad; } } if ((error = VOP_OPEN(vp, fmode, cred, td, fp)) != 0) goto bad; if (fmode & FWRITE) vp->v_writecount++; *flagp = fmode; - ASSERT_VOP_LOCKED(vp, "vn_open_cred"); + ASSERT_VOP_ELOCKED(vp, "vn_open_cred"); if (!mpsafe) VFS_UNLOCK_GIANT(vfslocked); return (0); bad: NDFREE(ndp, NDF_ONLY_PNBUF); vput(vp); VFS_UNLOCK_GIANT(vfslocked); *flagp = fmode; ndp->ni_vp = NULL; return (error); } /* * Check for write permissions on the specified vnode. * Prototype text segments cannot be written. */ int vn_writechk(vp) register struct vnode *vp; { ASSERT_VOP_LOCKED(vp, "vn_writechk"); /* * If there's shared text associated with * the vnode, try to free it up once. If * we fail, we can't allow writing. */ if (vp->v_vflag & VV_TEXT) return (ETXTBSY); return (0); } /* * Vnode close call */ int vn_close(vp, flags, file_cred, td) register struct vnode *vp; int flags; struct ucred *file_cred; struct thread *td; { struct mount *mp; int error; VFS_ASSERT_GIANT(vp->v_mount); vn_start_write(vp, &mp, V_WAIT); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); if (flags & FWRITE) { VNASSERT(vp->v_writecount > 0, vp, ("vn_close: negative writecount")); vp->v_writecount--; } error = VOP_CLOSE(vp, flags, file_cred, td); vput(vp); vn_finished_write(mp); return (error); } /* * Sequential heuristic - detect sequential operation */ static __inline int sequential_heuristic(struct uio *uio, struct file *fp) { if ((uio->uio_offset == 0 && fp->f_seqcount > 0) || uio->uio_offset == fp->f_nextoff) { /* * XXX we assume that the filesystem block size is * the default. Not true, but still gives us a pretty * good indicator of how sequential the read operations * are. */ fp->f_seqcount += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE; if (fp->f_seqcount > IO_SEQMAX) fp->f_seqcount = IO_SEQMAX; return(fp->f_seqcount << IO_SEQSHIFT); } /* * Not sequential, quick draw-down of seqcount */ if (fp->f_seqcount > 1) fp->f_seqcount = 1; else fp->f_seqcount = 0; return(0); } /* * Package up an I/O request on a vnode into a uio and do it. */ int vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred, aresid, td) enum uio_rw rw; struct vnode *vp; void *base; int len; off_t offset; enum uio_seg segflg; int ioflg; struct ucred *active_cred; struct ucred *file_cred; int *aresid; struct thread *td; { struct uio auio; struct iovec aiov; struct mount *mp; struct ucred *cred; int error; VFS_ASSERT_GIANT(vp->v_mount); if ((ioflg & IO_NODELOCKED) == 0) { mp = NULL; if (rw == UIO_WRITE) { if (vp->v_type != VCHR && (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) return (error); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); } else { /* * XXX This should be LK_SHARED but I don't trust VFS * enough to leave it like that until it has been * reviewed further. */ vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); } } ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held"); auio.uio_iov = &aiov; auio.uio_iovcnt = 1; aiov.iov_base = base; aiov.iov_len = len; auio.uio_resid = len; auio.uio_offset = offset; auio.uio_segflg = segflg; auio.uio_rw = rw; auio.uio_td = td; error = 0; #ifdef MAC if ((ioflg & IO_NOMACCHECK) == 0) { if (rw == UIO_READ) error = mac_check_vnode_read(active_cred, file_cred, vp); else error = mac_check_vnode_write(active_cred, file_cred, vp); } #endif if (error == 0) { if (file_cred) cred = file_cred; else cred = active_cred; if (rw == UIO_READ) error = VOP_READ(vp, &auio, ioflg, cred); else error = VOP_WRITE(vp, &auio, ioflg, cred); } if (aresid) *aresid = auio.uio_resid; else if (auio.uio_resid && error == 0) error = EIO; if ((ioflg & IO_NODELOCKED) == 0) { if (rw == UIO_WRITE && vp->v_type != VCHR) vn_finished_write(mp); VOP_UNLOCK(vp, 0, td); } return (error); } /* * Package up an I/O request on a vnode into a uio and do it. The I/O * request is split up into smaller chunks and we try to avoid saturating * the buffer cache while potentially holding a vnode locked, so we * check bwillwrite() before calling vn_rdwr(). We also call uio_yield() * to give other processes a chance to lock the vnode (either other processes * core'ing the same binary, or unrelated processes scanning the directory). */ int vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, active_cred, file_cred, aresid, td) enum uio_rw rw; struct vnode *vp; void *base; size_t len; off_t offset; enum uio_seg segflg; int ioflg; struct ucred *active_cred; struct ucred *file_cred; size_t *aresid; struct thread *td; { int error = 0; int iaresid; VFS_ASSERT_GIANT(vp->v_mount); do { int chunk; /* * Force `offset' to a multiple of MAXBSIZE except possibly * for the first chunk, so that filesystems only need to * write full blocks except possibly for the first and last * chunks. */ chunk = MAXBSIZE - (uoff_t)offset % MAXBSIZE; if (chunk > len) chunk = len; if (rw != UIO_READ && vp->v_type == VREG) bwillwrite(); iaresid = 0; error = vn_rdwr(rw, vp, base, chunk, offset, segflg, ioflg, active_cred, file_cred, &iaresid, td); len -= chunk; /* aresid calc already includes length */ if (error) break; offset += chunk; base = (char *)base + chunk; uio_yield(); } while (len); if (aresid) *aresid = len + iaresid; return (error); } /* * File table vnode read routine. */ static int vn_read(fp, uio, active_cred, flags, td) struct file *fp; struct uio *uio; struct ucred *active_cred; struct thread *td; int flags; { struct vnode *vp; int error, ioflag; int vfslocked; KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); vp = fp->f_vnode; ioflag = 0; if (fp->f_flag & FNONBLOCK) ioflag |= IO_NDELAY; if (fp->f_flag & O_DIRECT) ioflag |= IO_DIRECT; vfslocked = VFS_LOCK_GIANT(vp->v_mount); VOP_LEASE(vp, td, fp->f_cred, LEASE_READ); /* * According to McKusick the vn lock was protecting f_offset here. * It is now protected by the FOFFSET_LOCKED flag. */ if ((flags & FOF_OFFSET) == 0) { FILE_LOCK(fp); while(fp->f_vnread_flags & FOFFSET_LOCKED) { fp->f_vnread_flags |= FOFFSET_LOCK_WAITING; msleep(&fp->f_vnread_flags,fp->f_mtxp,PUSER -1,"vnread offlock",0); } fp->f_vnread_flags |= FOFFSET_LOCKED; FILE_UNLOCK(fp); vn_lock(vp, LK_SHARED | LK_RETRY, td); uio->uio_offset = fp->f_offset; } else vn_lock(vp, LK_SHARED | LK_RETRY, td); ioflag |= sequential_heuristic(uio, fp); #ifdef MAC error = mac_check_vnode_read(active_cred, fp->f_cred, vp); if (error == 0) #endif error = VOP_READ(vp, uio, ioflag, fp->f_cred); if ((flags & FOF_OFFSET) == 0) { fp->f_offset = uio->uio_offset; FILE_LOCK(fp); if (fp->f_vnread_flags & FOFFSET_LOCK_WAITING) wakeup(&fp->f_vnread_flags); fp->f_vnread_flags = 0; FILE_UNLOCK(fp); } fp->f_nextoff = uio->uio_offset; VOP_UNLOCK(vp, 0, td); VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * File table vnode write routine. */ static int vn_write(fp, uio, active_cred, flags, td) struct file *fp; struct uio *uio; struct ucred *active_cred; struct thread *td; int flags; { struct vnode *vp; struct mount *mp; int error, ioflag; int vfslocked; KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); vp = fp->f_vnode; vfslocked = VFS_LOCK_GIANT(vp->v_mount); if (vp->v_type == VREG) bwillwrite(); ioflag = IO_UNIT; if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) ioflag |= IO_APPEND; if (fp->f_flag & FNONBLOCK) ioflag |= IO_NDELAY; if (fp->f_flag & O_DIRECT) ioflag |= IO_DIRECT; if ((fp->f_flag & O_FSYNC) || (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) ioflag |= IO_SYNC; mp = NULL; if (vp->v_type != VCHR && (error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) goto unlock; VOP_LEASE(vp, td, fp->f_cred, LEASE_WRITE); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); if ((flags & FOF_OFFSET) == 0) uio->uio_offset = fp->f_offset; ioflag |= sequential_heuristic(uio, fp); #ifdef MAC error = mac_check_vnode_write(active_cred, fp->f_cred, vp); if (error == 0) #endif error = VOP_WRITE(vp, uio, ioflag, fp->f_cred); if ((flags & FOF_OFFSET) == 0) fp->f_offset = uio->uio_offset; fp->f_nextoff = uio->uio_offset; VOP_UNLOCK(vp, 0, td); if (vp->v_type != VCHR) vn_finished_write(mp); unlock: VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * File table vnode stat routine. */ static int vn_statfile(fp, sb, active_cred, td) struct file *fp; struct stat *sb; struct ucred *active_cred; struct thread *td; { struct vnode *vp = fp->f_vnode; int vfslocked; int error; vfslocked = VFS_LOCK_GIANT(vp->v_mount); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); error = vn_stat(vp, sb, active_cred, fp->f_cred, td); VOP_UNLOCK(vp, 0, td); VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * Stat a vnode; implementation for the stat syscall */ int vn_stat(vp, sb, active_cred, file_cred, td) struct vnode *vp; register struct stat *sb; struct ucred *active_cred; struct ucred *file_cred; struct thread *td; { struct vattr vattr; register struct vattr *vap; int error; u_short mode; #ifdef MAC error = mac_check_vnode_stat(active_cred, file_cred, vp); if (error) return (error); #endif vap = &vattr; error = VOP_GETATTR(vp, vap, active_cred, td); if (error) return (error); /* * Zero the spare stat fields */ bzero(sb, sizeof *sb); /* * Copy from vattr table */ if (vap->va_fsid != VNOVAL) sb->st_dev = vap->va_fsid; else sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; sb->st_ino = vap->va_fileid; mode = vap->va_mode; switch (vap->va_type) { case VREG: mode |= S_IFREG; break; case VDIR: mode |= S_IFDIR; break; case VBLK: mode |= S_IFBLK; break; case VCHR: mode |= S_IFCHR; break; case VLNK: mode |= S_IFLNK; /* This is a cosmetic change, symlinks do not have a mode. */ if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW) sb->st_mode &= ~ACCESSPERMS; /* 0000 */ else sb->st_mode |= ACCESSPERMS; /* 0777 */ break; case VSOCK: mode |= S_IFSOCK; break; case VFIFO: mode |= S_IFIFO; break; default: return (EBADF); }; sb->st_mode = mode; sb->st_nlink = vap->va_nlink; sb->st_uid = vap->va_uid; sb->st_gid = vap->va_gid; sb->st_rdev = vap->va_rdev; if (vap->va_size > OFF_MAX) return (EOVERFLOW); sb->st_size = vap->va_size; sb->st_atimespec = vap->va_atime; sb->st_mtimespec = vap->va_mtime; sb->st_ctimespec = vap->va_ctime; sb->st_birthtimespec = vap->va_birthtime; /* * According to www.opengroup.org, the meaning of st_blksize is * "a filesystem-specific preferred I/O block size for this * object. In some filesystem types, this may vary from file * to file" * Default to PAGE_SIZE after much discussion. * XXX: min(PAGE_SIZE, vp->v_bufobj.bo_bsize) may be more correct. */ sb->st_blksize = PAGE_SIZE; sb->st_flags = vap->va_flags; if (priv_check(td, PRIV_VFS_GENERATION)) sb->st_gen = 0; else sb->st_gen = vap->va_gen; sb->st_blocks = vap->va_bytes / S_BLKSIZE; return (0); } /* * File table vnode ioctl routine. */ static int vn_ioctl(fp, com, data, active_cred, td) struct file *fp; u_long com; void *data; struct ucred *active_cred; struct thread *td; { struct vnode *vp = fp->f_vnode; struct vattr vattr; int vfslocked; int error; vfslocked = VFS_LOCK_GIANT(vp->v_mount); error = ENOTTY; switch (vp->v_type) { case VREG: case VDIR: if (com == FIONREAD) { vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); error = VOP_GETATTR(vp, &vattr, active_cred, td); VOP_UNLOCK(vp, 0, td); if (!error) *(int *)data = vattr.va_size - fp->f_offset; } if (com == FIONBIO || com == FIOASYNC) /* XXX */ error = 0; else error = VOP_IOCTL(vp, com, data, fp->f_flag, active_cred, td); break; default: break; } VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * File table vnode poll routine. */ static int vn_poll(fp, events, active_cred, td) struct file *fp; int events; struct ucred *active_cred; struct thread *td; { struct vnode *vp; int vfslocked; int error; vp = fp->f_vnode; vfslocked = VFS_LOCK_GIANT(vp->v_mount); #ifdef MAC vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); error = mac_check_vnode_poll(active_cred, fp->f_cred, vp); VOP_UNLOCK(vp, 0, td); if (!error) #endif error = VOP_POLL(vp, events, fp->f_cred, td); VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * Check that the vnode is still valid, and if so * acquire requested lock. */ int _vn_lock(struct vnode *vp, int flags, struct thread *td, char *file, int line) { int error; do { if ((flags & LK_INTERLOCK) == 0) VI_LOCK(vp); if ((flags & LK_NOWAIT || (flags & LK_TYPE_MASK) == 0) && vp->v_iflag & VI_DOOMED) { VI_UNLOCK(vp); return (ENOENT); } /* * Just polling to check validity. */ if ((flags & LK_TYPE_MASK) == 0) { VI_UNLOCK(vp); return (0); } /* * lockmgr drops interlock before it will return for * any reason. So force the code above to relock it. */ error = VOP_LOCK1(vp, flags | LK_INTERLOCK, td, file, line); flags &= ~LK_INTERLOCK; KASSERT((flags & LK_RETRY) == 0 || error == 0, ("LK_RETRY set with incompatible flags %d\n", flags)); /* * Callers specify LK_RETRY if they wish to get dead vnodes. * If RETRY is not set, we return ENOENT instead. */ if (error == 0 && vp->v_iflag & VI_DOOMED && (flags & LK_RETRY) == 0) { VOP_UNLOCK(vp, 0, td); error = ENOENT; break; } } while (flags & LK_RETRY && error != 0); return (error); } /* * File table vnode close routine. */ static int vn_closefile(fp, td) struct file *fp; struct thread *td; { struct vnode *vp; struct flock lf; int vfslocked; int error; vp = fp->f_vnode; vfslocked = VFS_LOCK_GIANT(vp->v_mount); if (fp->f_type == DTYPE_VNODE && fp->f_flag & FHASLOCK) { lf.l_whence = SEEK_SET; lf.l_start = 0; lf.l_len = 0; lf.l_type = F_UNLCK; (void) VOP_ADVLOCK(vp, fp, F_UNLCK, &lf, F_FLOCK); } fp->f_ops = &badfileops; error = vn_close(vp, fp->f_flag, fp->f_cred, td); VFS_UNLOCK_GIANT(vfslocked); return (error); } /* * Preparing to start a filesystem write operation. If the operation is * permitted, then we bump the count of operations in progress and * proceed. If a suspend request is in progress, we wait until the * suspension is over, and then proceed. */ int vn_start_write(vp, mpp, flags) struct vnode *vp; struct mount **mpp; int flags; { struct mount *mp; int error; error = 0; /* * If a vnode is provided, get and return the mount point that * to which it will write. */ if (vp != NULL) { if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) { *mpp = NULL; if (error != EOPNOTSUPP) return (error); return (0); } } if ((mp = *mpp) == NULL) return (0); MNT_ILOCK(mp); if (vp == NULL) MNT_REF(mp); /* * Check on status of suspension. */ while ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) { if (flags & V_NOWAIT) { error = EWOULDBLOCK; goto unlock; } error = msleep(&mp->mnt_flag, MNT_MTX(mp), (PUSER - 1) | (flags & PCATCH), "suspfs", 0); if (error) goto unlock; } if (flags & V_XSLEEP) goto unlock; mp->mnt_writeopcount++; unlock: MNT_REL(mp); MNT_IUNLOCK(mp); return (error); } /* * Secondary suspension. Used by operations such as vop_inactive * routines that are needed by the higher level functions. These * are allowed to proceed until all the higher level functions have * completed (indicated by mnt_writeopcount dropping to zero). At that * time, these operations are halted until the suspension is over. */ int vn_write_suspend_wait(vp, mp, flags) struct vnode *vp; struct mount *mp; int flags; { int error; if (vp != NULL) { if ((error = VOP_GETWRITEMOUNT(vp, &mp)) != 0) { if (error != EOPNOTSUPP) return (error); return (0); } } /* * If we are not suspended or have not yet reached suspended * mode, then let the operation proceed. */ if (mp == NULL) return (0); MNT_ILOCK(mp); if (vp == NULL) MNT_REF(mp); if ((mp->mnt_kern_flag & MNTK_SUSPENDED) == 0) { MNT_REL(mp); MNT_IUNLOCK(mp); return (0); } if (flags & V_NOWAIT) { MNT_REL(mp); MNT_IUNLOCK(mp); return (EWOULDBLOCK); } /* * Wait for the suspension to finish. */ error = msleep(&mp->mnt_flag, MNT_MTX(mp), (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0); vfs_rel(mp); return (error); } /* * Secondary suspension. Used by operations such as vop_inactive * routines that are needed by the higher level functions. These * are allowed to proceed until all the higher level functions have * completed (indicated by mnt_writeopcount dropping to zero). At that * time, these operations are halted until the suspension is over. */ int vn_start_secondary_write(vp, mpp, flags) struct vnode *vp; struct mount **mpp; int flags; { struct mount *mp; int error; retry: if (vp != NULL) { if ((error = VOP_GETWRITEMOUNT(vp, mpp)) != 0) { *mpp = NULL; if (error != EOPNOTSUPP) return (error); return (0); } } /* * If we are not suspended or have not yet reached suspended * mode, then let the operation proceed. */ if ((mp = *mpp) == NULL) return (0); MNT_ILOCK(mp); if (vp == NULL) MNT_REF(mp); if ((mp->mnt_kern_flag & (MNTK_SUSPENDED | MNTK_SUSPEND2)) == 0) { mp->mnt_secondary_writes++; mp->mnt_secondary_accwrites++; MNT_REL(mp); MNT_IUNLOCK(mp); return (0); } if (flags & V_NOWAIT) { MNT_REL(mp); MNT_IUNLOCK(mp); return (EWOULDBLOCK); } /* * Wait for the suspension to finish. */ error = msleep(&mp->mnt_flag, MNT_MTX(mp), (PUSER - 1) | (flags & PCATCH) | PDROP, "suspfs", 0); vfs_rel(mp); if (error == 0) goto retry; return (error); } /* * Filesystem write operation has completed. If we are suspending and this * operation is the last one, notify the suspender that the suspension is * now in effect. */ void vn_finished_write(mp) struct mount *mp; { if (mp == NULL) return; MNT_ILOCK(mp); mp->mnt_writeopcount--; if (mp->mnt_writeopcount < 0) panic("vn_finished_write: neg cnt"); if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 && mp->mnt_writeopcount <= 0) wakeup(&mp->mnt_writeopcount); MNT_IUNLOCK(mp); } /* * Filesystem secondary write operation has completed. If we are * suspending and this operation is the last one, notify the suspender * that the suspension is now in effect. */ void vn_finished_secondary_write(mp) struct mount *mp; { if (mp == NULL) return; MNT_ILOCK(mp); mp->mnt_secondary_writes--; if (mp->mnt_secondary_writes < 0) panic("vn_finished_secondary_write: neg cnt"); if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0 && mp->mnt_secondary_writes <= 0) wakeup(&mp->mnt_secondary_writes); MNT_IUNLOCK(mp); } /* * Request a filesystem to suspend write operations. */ int vfs_write_suspend(mp) struct mount *mp; { struct thread *td = curthread; int error; MNT_ILOCK(mp); if (mp->mnt_kern_flag & MNTK_SUSPEND) { MNT_IUNLOCK(mp); return (0); } mp->mnt_kern_flag |= MNTK_SUSPEND; if (mp->mnt_writeopcount > 0) (void) msleep(&mp->mnt_writeopcount, MNT_MTX(mp), (PUSER - 1)|PDROP, "suspwt", 0); else MNT_IUNLOCK(mp); if ((error = VFS_SYNC(mp, MNT_SUSPEND, td)) != 0) vfs_write_resume(mp); return (error); } /* * Request a filesystem to resume write operations. */ void vfs_write_resume(mp) struct mount *mp; { MNT_ILOCK(mp); if ((mp->mnt_kern_flag & MNTK_SUSPEND) != 0) { mp->mnt_kern_flag &= ~(MNTK_SUSPEND | MNTK_SUSPEND2 | MNTK_SUSPENDED); wakeup(&mp->mnt_writeopcount); wakeup(&mp->mnt_flag); } MNT_IUNLOCK(mp); } /* * Implement kqueues for files by translating it to vnode operation. */ static int vn_kqfilter(struct file *fp, struct knote *kn) { int vfslocked; int error; vfslocked = VFS_LOCK_GIANT(fp->f_vnode->v_mount); error = VOP_KQFILTER(fp->f_vnode, kn); VFS_UNLOCK_GIANT(vfslocked); return error; } /* * Simplified in-kernel wrapper calls for extended attribute access. * Both calls pass in a NULL credential, authorizing as "kernel" access. * Set IO_NODELOCKED in ioflg if the vnode is already locked. */ int vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, const char *attrname, int *buflen, char *buf, struct thread *td) { struct uio auio; struct iovec iov; int error; iov.iov_len = *buflen; iov.iov_base = buf; auio.uio_iov = &iov; auio.uio_iovcnt = 1; auio.uio_rw = UIO_READ; auio.uio_segflg = UIO_SYSSPACE; auio.uio_td = td; auio.uio_offset = 0; auio.uio_resid = *buflen; if ((ioflg & IO_NODELOCKED) == 0) vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held"); /* authorize attribute retrieval as kernel */ error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, td); if ((ioflg & IO_NODELOCKED) == 0) VOP_UNLOCK(vp, 0, td); if (error == 0) { *buflen = *buflen - auio.uio_resid; } return (error); } /* * XXX failure mode if partially written? */ int vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, const char *attrname, int buflen, char *buf, struct thread *td) { struct uio auio; struct iovec iov; struct mount *mp; int error; iov.iov_len = buflen; iov.iov_base = buf; auio.uio_iov = &iov; auio.uio_iovcnt = 1; auio.uio_rw = UIO_WRITE; auio.uio_segflg = UIO_SYSSPACE; auio.uio_td = td; auio.uio_offset = 0; auio.uio_resid = buflen; if ((ioflg & IO_NODELOCKED) == 0) { if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) return (error); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); } ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held"); /* authorize attribute setting as kernel */ error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, td); if ((ioflg & IO_NODELOCKED) == 0) { vn_finished_write(mp); VOP_UNLOCK(vp, 0, td); } return (error); } int vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, const char *attrname, struct thread *td) { struct mount *mp; int error; if ((ioflg & IO_NODELOCKED) == 0) { if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) return (error); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); } ASSERT_VOP_LOCKED(vp, "IO_NODELOCKED with no vp lock held"); /* authorize attribute removal as kernel */ error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, td); if (error == EOPNOTSUPP) error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, NULL, td); if ((ioflg & IO_NODELOCKED) == 0) { vn_finished_write(mp); VOP_UNLOCK(vp, 0, td); } return (error); } Index: head/sys/vm/vnode_pager.c =================================================================== --- head/sys/vm/vnode_pager.c (revision 171598) +++ head/sys/vm/vnode_pager.c (revision 171599) @@ -1,1222 +1,1222 @@ /*- * Copyright (c) 1990 University of Utah. * Copyright (c) 1991 The Regents of the University of California. * All rights reserved. * Copyright (c) 1993, 1994 John S. Dyson * Copyright (c) 1995, David Greenman * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * 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. * * from: @(#)vnode_pager.c 7.5 (Berkeley) 4/20/91 */ /* * Page to/from files (vnodes). */ /* * TODO: * Implement VOP_GETPAGES/PUTPAGES interface for filesystems. Will * greatly re-simplify the vnode_pager. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static int vnode_pager_addr(struct vnode *vp, vm_ooffset_t address, daddr_t *rtaddress, int *run); static int vnode_pager_input_smlfs(vm_object_t object, vm_page_t m); static int vnode_pager_input_old(vm_object_t object, vm_page_t m); static void vnode_pager_dealloc(vm_object_t); static int vnode_pager_getpages(vm_object_t, vm_page_t *, int, int); static void vnode_pager_putpages(vm_object_t, vm_page_t *, int, boolean_t, int *); static boolean_t vnode_pager_haspage(vm_object_t, vm_pindex_t, int *, int *); static vm_object_t vnode_pager_alloc(void *, vm_ooffset_t, vm_prot_t, vm_ooffset_t); struct pagerops vnodepagerops = { .pgo_alloc = vnode_pager_alloc, .pgo_dealloc = vnode_pager_dealloc, .pgo_getpages = vnode_pager_getpages, .pgo_putpages = vnode_pager_putpages, .pgo_haspage = vnode_pager_haspage, }; int vnode_pbuf_freecnt; /* Create the VM system backing object for this vnode */ int vnode_create_vobject(struct vnode *vp, off_t isize, struct thread *td) { vm_object_t object; vm_ooffset_t size = isize; struct vattr va; if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE) return (0); while ((object = vp->v_object) != NULL) { VM_OBJECT_LOCK(object); if (!(object->flags & OBJ_DEAD)) { VM_OBJECT_UNLOCK(object); return (0); } VOP_UNLOCK(vp, 0, td); vm_object_set_flag(object, OBJ_DISCONNECTWNT); msleep(object, VM_OBJECT_MTX(object), PDROP | PVM, "vodead", 0); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); } if (size == 0) { if (vn_isdisk(vp, NULL)) { size = IDX_TO_OFF(INT_MAX); } else { if (VOP_GETATTR(vp, &va, td->td_ucred, td) != 0) return (0); size = va.va_size; } } object = vnode_pager_alloc(vp, size, 0, 0); /* * Dereference the reference we just created. This assumes * that the object is associated with the vp. */ VM_OBJECT_LOCK(object); object->ref_count--; VM_OBJECT_UNLOCK(object); vrele(vp); KASSERT(vp->v_object != NULL, ("vnode_create_vobject: NULL object")); return (0); } void vnode_destroy_vobject(struct vnode *vp) { struct vm_object *obj; obj = vp->v_object; if (obj == NULL) return; - ASSERT_VOP_LOCKED(vp, "vnode_destroy_vobject"); + ASSERT_VOP_ELOCKED(vp, "vnode_destroy_vobject"); VM_OBJECT_LOCK(obj); 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 VM_OBJECT_UNLOCK(obj); } else { /* * Woe to the process that tries to page now :-). */ vm_pager_deallocate(obj); VM_OBJECT_UNLOCK(obj); } vp->v_object = NULL; } /* * Allocate (or lookup) pager for a vnode. * Handle is a vnode pointer. * * MPSAFE */ vm_object_t vnode_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot, vm_ooffset_t offset) { vm_object_t object; struct vnode *vp; /* * Pageout to vnode, no can do yet. */ if (handle == NULL) return (NULL); vp = (struct vnode *) handle; - ASSERT_VOP_LOCKED(vp, "vnode_pager_alloc"); + ASSERT_VOP_ELOCKED(vp, "vnode_pager_alloc"); /* * If the object is being terminated, wait for it to * go away. */ while ((object = vp->v_object) != NULL) { VM_OBJECT_LOCK(object); if ((object->flags & OBJ_DEAD) == 0) break; vm_object_set_flag(object, OBJ_DISCONNECTWNT); msleep(object, VM_OBJECT_MTX(object), PDROP | PVM, "vadead", 0); } if (vp->v_usecount == 0) panic("vnode_pager_alloc: no vnode reference"); if (object == NULL) { /* * And an object of the appropriate size */ object = vm_object_allocate(OBJT_VNODE, OFF_TO_IDX(round_page(size))); object->un_pager.vnp.vnp_size = size; object->handle = handle; if (VFS_NEEDSGIANT(vp->v_mount)) vm_object_set_flag(object, OBJ_NEEDGIANT); vp->v_object = object; } else { object->ref_count++; VM_OBJECT_UNLOCK(object); } vref(vp); return (object); } /* * The object must be locked. */ static void vnode_pager_dealloc(object) vm_object_t object; { struct vnode *vp = object->handle; if (vp == NULL) panic("vnode_pager_dealloc: pager already dealloced"); VM_OBJECT_LOCK_ASSERT(object, MA_OWNED); vm_object_pip_wait(object, "vnpdea"); object->handle = NULL; object->type = OBJT_DEAD; if (object->flags & OBJ_DISCONNECTWNT) { vm_object_clear_flag(object, OBJ_DISCONNECTWNT); wakeup(object); } - ASSERT_VOP_LOCKED(vp, "vnode_pager_dealloc"); + ASSERT_VOP_ELOCKED(vp, "vnode_pager_dealloc"); vp->v_object = NULL; vp->v_vflag &= ~VV_TEXT; } static boolean_t vnode_pager_haspage(object, pindex, before, after) vm_object_t object; vm_pindex_t pindex; int *before; int *after; { struct vnode *vp = object->handle; daddr_t bn; int err; daddr_t reqblock; int poff; int bsize; int pagesperblock, blocksperpage; int vfslocked; VM_OBJECT_LOCK_ASSERT(object, MA_OWNED); /* * If no vp or vp is doomed or marked transparent to VM, we do not * have the page. */ if (vp == NULL || vp->v_iflag & VI_DOOMED) return FALSE; /* * If the offset is beyond end of file we do * not have the page. */ if (IDX_TO_OFF(pindex) >= object->un_pager.vnp.vnp_size) return FALSE; bsize = vp->v_mount->mnt_stat.f_iosize; pagesperblock = bsize / PAGE_SIZE; blocksperpage = 0; if (pagesperblock > 0) { reqblock = pindex / pagesperblock; } else { blocksperpage = (PAGE_SIZE / bsize); reqblock = pindex * blocksperpage; } VM_OBJECT_UNLOCK(object); vfslocked = VFS_LOCK_GIANT(vp->v_mount); err = VOP_BMAP(vp, reqblock, NULL, &bn, after, before); VFS_UNLOCK_GIANT(vfslocked); VM_OBJECT_LOCK(object); if (err) return TRUE; if (bn == -1) return FALSE; if (pagesperblock > 0) { poff = pindex - (reqblock * pagesperblock); if (before) { *before *= pagesperblock; *before += poff; } if (after) { int numafter; *after *= pagesperblock; numafter = pagesperblock - (poff + 1); if (IDX_TO_OFF(pindex + numafter) > object->un_pager.vnp.vnp_size) { numafter = OFF_TO_IDX(object->un_pager.vnp.vnp_size) - pindex; } *after += numafter; } } else { if (before) { *before /= blocksperpage; } if (after) { *after /= blocksperpage; } } return TRUE; } /* * Lets the VM system know about a change in size for a file. * We adjust our own internal size and flush any cached pages in * the associated object that are affected by the size change. * * Note: this routine may be invoked as a result of a pager put * operation (possibly at object termination time), so we must be careful. */ void vnode_pager_setsize(vp, nsize) struct vnode *vp; vm_ooffset_t nsize; { vm_object_t object; vm_page_t m; vm_pindex_t nobjsize; if ((object = vp->v_object) == NULL) return; VM_OBJECT_LOCK(object); if (nsize == object->un_pager.vnp.vnp_size) { /* * Hasn't changed size */ VM_OBJECT_UNLOCK(object); return; } nobjsize = OFF_TO_IDX(nsize + PAGE_MASK); if (nsize < object->un_pager.vnp.vnp_size) { /* * File has shrunk. Toss any cached pages beyond the new EOF. */ if (nobjsize < object->size) vm_object_page_remove(object, nobjsize, object->size, FALSE); /* * this gets rid of garbage at the end of a page that is now * only partially backed by the vnode. * * XXX for some reason (I don't know yet), if we take a * completely invalid page and mark it partially valid * it can screw up NFS reads, so we don't allow the case. */ if ((nsize & PAGE_MASK) && (m = vm_page_lookup(object, OFF_TO_IDX(nsize))) != NULL && m->valid != 0) { int base = (int)nsize & PAGE_MASK; int size = PAGE_SIZE - base; /* * Clear out partial-page garbage in case * the page has been mapped. */ pmap_zero_page_area(m, base, size); /* * XXX work around SMP data integrity race * by unmapping the page from user processes. * The garbage we just cleared may be mapped * to a user process running on another cpu * and this code is not running through normal * I/O channels which handle SMP issues for * us, so unmap page to synchronize all cpus. * * XXX should vm_pager_unmap_page() have * dealt with this? */ vm_page_lock_queues(); pmap_remove_all(m); /* * Clear out partial-page dirty bits. This * has the side effect of setting the valid * bits, but that is ok. There are a bunch * of places in the VM system where we expected * m->dirty == VM_PAGE_BITS_ALL. The file EOF * case is one of them. If the page is still * partially dirty, make it fully dirty. * * note that we do not clear out the valid * bits. This would prevent bogus_page * replacement from working properly. */ vm_page_set_validclean(m, base, size); if (m->dirty != 0) m->dirty = VM_PAGE_BITS_ALL; vm_page_unlock_queues(); } } object->un_pager.vnp.vnp_size = nsize; object->size = nobjsize; VM_OBJECT_UNLOCK(object); } /* * calculate the linear (byte) disk address of specified virtual * file address */ static int vnode_pager_addr(struct vnode *vp, vm_ooffset_t address, daddr_t *rtaddress, int *run) { int bsize; int err; daddr_t vblock; daddr_t voffset; if (address < 0) return -1; if (vp->v_iflag & VI_DOOMED) return -1; bsize = vp->v_mount->mnt_stat.f_iosize; vblock = address / bsize; voffset = address % bsize; err = VOP_BMAP(vp, vblock, NULL, rtaddress, run, NULL); if (err == 0) { if (*rtaddress != -1) *rtaddress += voffset / DEV_BSIZE; if (run) { *run += 1; *run *= bsize/PAGE_SIZE; *run -= voffset/PAGE_SIZE; } } return (err); } /* * small block filesystem vnode pager input */ static int vnode_pager_input_smlfs(object, m) vm_object_t object; vm_page_t m; { int i; struct vnode *vp; struct bufobj *bo; struct buf *bp; struct sf_buf *sf; daddr_t fileaddr; vm_offset_t bsize; int error = 0; vp = object->handle; if (vp->v_iflag & VI_DOOMED) return VM_PAGER_BAD; bsize = vp->v_mount->mnt_stat.f_iosize; VOP_BMAP(vp, 0, &bo, 0, NULL, NULL); sf = sf_buf_alloc(m, 0); for (i = 0; i < PAGE_SIZE / bsize; i++) { vm_ooffset_t address; if (vm_page_bits(i * bsize, bsize) & m->valid) continue; address = IDX_TO_OFF(m->pindex) + i * bsize; if (address >= object->un_pager.vnp.vnp_size) { fileaddr = -1; } else { error = vnode_pager_addr(vp, address, &fileaddr, NULL); if (error) break; } if (fileaddr != -1) { bp = getpbuf(&vnode_pbuf_freecnt); /* build a minimal buffer header */ bp->b_iocmd = BIO_READ; bp->b_iodone = bdone; KASSERT(bp->b_rcred == NOCRED, ("leaking read ucred")); KASSERT(bp->b_wcred == NOCRED, ("leaking write ucred")); bp->b_rcred = crhold(curthread->td_ucred); bp->b_wcred = crhold(curthread->td_ucred); bp->b_data = (caddr_t)sf_buf_kva(sf) + i * bsize; bp->b_blkno = fileaddr; pbgetbo(bo, bp); bp->b_bcount = bsize; bp->b_bufsize = bsize; bp->b_runningbufspace = bp->b_bufsize; atomic_add_int(&runningbufspace, bp->b_runningbufspace); /* do the input */ bp->b_iooffset = dbtob(bp->b_blkno); bstrategy(bp); bwait(bp, PVM, "vnsrd"); if ((bp->b_ioflags & BIO_ERROR) != 0) error = EIO; /* * free the buffer header back to the swap buffer pool */ pbrelbo(bp); relpbuf(bp, &vnode_pbuf_freecnt); if (error) break; VM_OBJECT_LOCK(object); vm_page_lock_queues(); vm_page_set_validclean(m, (i * bsize) & PAGE_MASK, bsize); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); } else { VM_OBJECT_LOCK(object); vm_page_lock_queues(); vm_page_set_validclean(m, (i * bsize) & PAGE_MASK, bsize); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); bzero((caddr_t)sf_buf_kva(sf) + i * bsize, bsize); } } sf_buf_free(sf); vm_page_lock_queues(); pmap_clear_modify(m); vm_page_unlock_queues(); if (error) { return VM_PAGER_ERROR; } return VM_PAGER_OK; } /* * old style vnode pager input routine */ static int vnode_pager_input_old(object, m) vm_object_t object; vm_page_t m; { struct uio auio; struct iovec aiov; int error; int size; struct sf_buf *sf; struct vnode *vp; VM_OBJECT_LOCK_ASSERT(object, MA_OWNED); error = 0; /* * Return failure if beyond current EOF */ if (IDX_TO_OFF(m->pindex) >= object->un_pager.vnp.vnp_size) { return VM_PAGER_BAD; } else { size = PAGE_SIZE; if (IDX_TO_OFF(m->pindex) + size > object->un_pager.vnp.vnp_size) size = object->un_pager.vnp.vnp_size - IDX_TO_OFF(m->pindex); vp = object->handle; VM_OBJECT_UNLOCK(object); /* * Allocate a kernel virtual address and initialize so that * we can use VOP_READ/WRITE routines. */ sf = sf_buf_alloc(m, 0); aiov.iov_base = (caddr_t)sf_buf_kva(sf); aiov.iov_len = size; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_offset = IDX_TO_OFF(m->pindex); auio.uio_segflg = UIO_SYSSPACE; auio.uio_rw = UIO_READ; auio.uio_resid = size; auio.uio_td = curthread; error = VOP_READ(vp, &auio, 0, curthread->td_ucred); if (!error) { int count = size - auio.uio_resid; if (count == 0) error = EINVAL; else if (count != PAGE_SIZE) bzero((caddr_t)sf_buf_kva(sf) + count, PAGE_SIZE - count); } sf_buf_free(sf); VM_OBJECT_LOCK(object); } vm_page_lock_queues(); pmap_clear_modify(m); vm_page_undirty(m); vm_page_unlock_queues(); if (!error) m->valid = VM_PAGE_BITS_ALL; return error ? VM_PAGER_ERROR : VM_PAGER_OK; } /* * generic vnode pager input routine */ /* * Local media VFS's that do not implement their own VOP_GETPAGES * should have their VOP_GETPAGES call to vnode_pager_generic_getpages() * to implement the previous behaviour. * * All other FS's should use the bypass to get to the local media * backing vp's VOP_GETPAGES. */ static int vnode_pager_getpages(object, m, count, reqpage) vm_object_t object; vm_page_t *m; int count; int reqpage; { int rtval; struct vnode *vp; int bytes = count * PAGE_SIZE; int vfslocked; vp = object->handle; VM_OBJECT_UNLOCK(object); vfslocked = VFS_LOCK_GIANT(vp->v_mount); rtval = VOP_GETPAGES(vp, m, bytes, reqpage, 0); KASSERT(rtval != EOPNOTSUPP, ("vnode_pager: FS getpages not implemented\n")); VFS_UNLOCK_GIANT(vfslocked); VM_OBJECT_LOCK(object); return rtval; } /* * This is now called from local media FS's to operate against their * own vnodes if they fail to implement VOP_GETPAGES. */ int vnode_pager_generic_getpages(vp, m, bytecount, reqpage) struct vnode *vp; vm_page_t *m; int bytecount; int reqpage; { vm_object_t object; vm_offset_t kva; off_t foff, tfoff, nextoff; int i, j, size, bsize, first; daddr_t firstaddr, reqblock; struct bufobj *bo; int runpg; int runend; struct buf *bp; int count; int error; object = vp->v_object; count = bytecount / PAGE_SIZE; KASSERT(vp->v_type != VCHR && vp->v_type != VBLK, ("vnode_pager_generic_getpages does not support devices")); if (vp->v_iflag & VI_DOOMED) return VM_PAGER_BAD; bsize = vp->v_mount->mnt_stat.f_iosize; /* get the UNDERLYING device for the file with VOP_BMAP() */ /* * originally, we did not check for an error return value -- assuming * an fs always has a bmap entry point -- that assumption is wrong!!! */ foff = IDX_TO_OFF(m[reqpage]->pindex); /* * if we can't bmap, use old VOP code */ error = VOP_BMAP(vp, foff / bsize, &bo, &reqblock, NULL, NULL); if (error == EOPNOTSUPP) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (i = 0; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); PCPU_INC(cnt.v_vnodein); PCPU_INC(cnt.v_vnodepgsin); error = vnode_pager_input_old(object, m[reqpage]); VM_OBJECT_UNLOCK(object); return (error); } else if (error != 0) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (i = 0; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); return (VM_PAGER_ERROR); /* * if the blocksize is smaller than a page size, then use * special small filesystem code. NFS sometimes has a small * blocksize, but it can handle large reads itself. */ } else if ((PAGE_SIZE / bsize) > 1 && (vp->v_mount->mnt_stat.f_type != nfs_mount_type)) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (i = 0; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); PCPU_INC(cnt.v_vnodein); PCPU_INC(cnt.v_vnodepgsin); return vnode_pager_input_smlfs(object, m[reqpage]); } /* * If we have a completely valid page available to us, we can * clean up and return. Otherwise we have to re-read the * media. */ VM_OBJECT_LOCK(object); if (m[reqpage]->valid == VM_PAGE_BITS_ALL) { vm_page_lock_queues(); for (i = 0; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); return VM_PAGER_OK; } else if (reqblock == -1) { pmap_zero_page(m[reqpage]); vm_page_undirty(m[reqpage]); m[reqpage]->valid = VM_PAGE_BITS_ALL; vm_page_lock_queues(); for (i = 0; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); return (VM_PAGER_OK); } m[reqpage]->valid = 0; VM_OBJECT_UNLOCK(object); /* * here on direct device I/O */ firstaddr = -1; /* * calculate the run that includes the required page */ for (first = 0, i = 0; i < count; i = runend) { if (vnode_pager_addr(vp, IDX_TO_OFF(m[i]->pindex), &firstaddr, &runpg) != 0) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (; i < count; i++) if (i != reqpage) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); return (VM_PAGER_ERROR); } if (firstaddr == -1) { VM_OBJECT_LOCK(object); if (i == reqpage && foff < object->un_pager.vnp.vnp_size) { panic("vnode_pager_getpages: unexpected missing page: firstaddr: %jd, foff: 0x%jx%08jx, vnp_size: 0x%jx%08jx", (intmax_t)firstaddr, (uintmax_t)(foff >> 32), (uintmax_t)foff, (uintmax_t) (object->un_pager.vnp.vnp_size >> 32), (uintmax_t)object->un_pager.vnp.vnp_size); } vm_page_lock_queues(); vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); runend = i + 1; first = runend; continue; } runend = i + runpg; if (runend <= reqpage) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (j = i; j < runend; j++) vm_page_free(m[j]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); } else { if (runpg < (count - first)) { VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (i = first + runpg; i < count; i++) vm_page_free(m[i]); vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); count = first + runpg; } break; } first = runend; } /* * the first and last page have been calculated now, move input pages * to be zero based... */ if (first != 0) { m += first; count -= first; reqpage -= first; } /* * calculate the file virtual address for the transfer */ foff = IDX_TO_OFF(m[0]->pindex); /* * calculate the size of the transfer */ size = count * PAGE_SIZE; KASSERT(count > 0, ("zero count")); if ((foff + size) > object->un_pager.vnp.vnp_size) size = object->un_pager.vnp.vnp_size - foff; KASSERT(size > 0, ("zero size")); /* * round up physical size for real devices. */ if (1) { int secmask = bo->bo_bsize - 1; KASSERT(secmask < PAGE_SIZE && secmask > 0, ("vnode_pager_generic_getpages: sector size %d too large", secmask + 1)); size = (size + secmask) & ~secmask; } bp = getpbuf(&vnode_pbuf_freecnt); kva = (vm_offset_t) bp->b_data; /* * and map the pages to be read into the kva */ pmap_qenter(kva, m, count); /* build a minimal buffer header */ bp->b_iocmd = BIO_READ; bp->b_iodone = bdone; KASSERT(bp->b_rcred == NOCRED, ("leaking read ucred")); KASSERT(bp->b_wcred == NOCRED, ("leaking write ucred")); bp->b_rcred = crhold(curthread->td_ucred); bp->b_wcred = crhold(curthread->td_ucred); bp->b_blkno = firstaddr; pbgetbo(bo, bp); bp->b_bcount = size; bp->b_bufsize = size; bp->b_runningbufspace = bp->b_bufsize; atomic_add_int(&runningbufspace, bp->b_runningbufspace); PCPU_INC(cnt.v_vnodein); PCPU_ADD(cnt.v_vnodepgsin, count); /* do the input */ bp->b_iooffset = dbtob(bp->b_blkno); bstrategy(bp); bwait(bp, PVM, "vnread"); if ((bp->b_ioflags & BIO_ERROR) != 0) error = EIO; if (!error) { if (size != count * PAGE_SIZE) bzero((caddr_t) kva + size, PAGE_SIZE * count - size); } pmap_qremove(kva, count); /* * free the buffer header back to the swap buffer pool */ pbrelbo(bp); relpbuf(bp, &vnode_pbuf_freecnt); VM_OBJECT_LOCK(object); vm_page_lock_queues(); for (i = 0, tfoff = foff; i < count; i++, tfoff = nextoff) { vm_page_t mt; nextoff = tfoff + PAGE_SIZE; mt = m[i]; if (nextoff <= object->un_pager.vnp.vnp_size) { /* * Read filled up entire page. */ mt->valid = VM_PAGE_BITS_ALL; vm_page_undirty(mt); /* should be an assert? XXX */ pmap_clear_modify(mt); } else { /* * Read did not fill up entire page. Since this * is getpages, the page may be mapped, so we have * to zero the invalid portions of the page even * though we aren't setting them valid. * * Currently we do not set the entire page valid, * we just try to clear the piece that we couldn't * read. */ vm_page_set_validclean(mt, 0, object->un_pager.vnp.vnp_size - tfoff); /* handled by vm_fault now */ /* vm_page_zero_invalid(mt, FALSE); */ } if (i != reqpage) { /* * whether or not to leave the page activated is up in * the air, but we should put the page on a page queue * somewhere. (it already is in the object). Result: * It appears that empirical results show that * deactivating pages is best. */ /* * just in case someone was asking for this page we * now tell them that it is ok to use */ if (!error) { if (mt->oflags & VPO_WANTED) vm_page_activate(mt); else vm_page_deactivate(mt); vm_page_wakeup(mt); } else { vm_page_free(mt); } } } vm_page_unlock_queues(); VM_OBJECT_UNLOCK(object); if (error) { printf("vnode_pager_getpages: I/O read error\n"); } return (error ? VM_PAGER_ERROR : VM_PAGER_OK); } /* * EOPNOTSUPP is no longer legal. For local media VFS's that do not * implement their own VOP_PUTPAGES, their VOP_PUTPAGES should call to * vnode_pager_generic_putpages() to implement the previous behaviour. * * All other FS's should use the bypass to get to the local media * backing vp's VOP_PUTPAGES. */ static void vnode_pager_putpages(object, m, count, sync, rtvals) vm_object_t object; vm_page_t *m; int count; boolean_t sync; int *rtvals; { int rtval; struct vnode *vp; struct mount *mp; int bytes = count * PAGE_SIZE; /* * Force synchronous operation if we are extremely low on memory * to prevent a low-memory deadlock. VOP operations often need to * allocate more memory to initiate the I/O ( i.e. do a BMAP * operation ). The swapper handles the case by limiting the amount * of asynchronous I/O, but that sort of solution doesn't scale well * for the vnode pager without a lot of work. * * Also, the backing vnode's iodone routine may not wake the pageout * daemon up. This should be probably be addressed XXX. */ if ((cnt.v_free_count + cnt.v_cache_count) < cnt.v_pageout_free_min) sync |= OBJPC_SYNC; /* * Call device-specific putpages function */ vp = object->handle; VM_OBJECT_UNLOCK(object); if (vp->v_type != VREG) mp = NULL; rtval = VOP_PUTPAGES(vp, m, bytes, sync, rtvals, 0); KASSERT(rtval != EOPNOTSUPP, ("vnode_pager: stale FS putpages\n")); VM_OBJECT_LOCK(object); } /* * This is now called from local media FS's to operate against their * own vnodes if they fail to implement VOP_PUTPAGES. * * This is typically called indirectly via the pageout daemon and * clustering has already typically occured, so in general we ask the * underlying filesystem to write the data out asynchronously rather * then delayed. */ int vnode_pager_generic_putpages(vp, m, bytecount, flags, rtvals) struct vnode *vp; vm_page_t *m; int bytecount; int flags; int *rtvals; { int i; vm_object_t object; int count; int maxsize, ncount; vm_ooffset_t poffset; struct uio auio; struct iovec aiov; int error; int ioflags; int ppscheck = 0; static struct timeval lastfail; static int curfail; object = vp->v_object; count = bytecount / PAGE_SIZE; for (i = 0; i < count; i++) rtvals[i] = VM_PAGER_AGAIN; if ((int64_t)m[0]->pindex < 0) { printf("vnode_pager_putpages: attempt to write meta-data!!! -- 0x%lx(%lx)\n", (long)m[0]->pindex, (u_long)m[0]->dirty); rtvals[0] = VM_PAGER_BAD; return VM_PAGER_BAD; } maxsize = count * PAGE_SIZE; ncount = count; poffset = IDX_TO_OFF(m[0]->pindex); /* * If the page-aligned write is larger then the actual file we * have to invalidate pages occuring beyond the file EOF. However, * there is an edge case where a file may not be page-aligned where * the last page is partially invalid. In this case the filesystem * may not properly clear the dirty bits for the entire page (which * could be VM_PAGE_BITS_ALL due to the page having been mmap()d). * With the page locked we are free to fix-up the dirty bits here. * * We do not under any circumstances truncate the valid bits, as * this will screw up bogus page replacement. */ if (maxsize + poffset > object->un_pager.vnp.vnp_size) { if (object->un_pager.vnp.vnp_size > poffset) { int pgoff; maxsize = object->un_pager.vnp.vnp_size - poffset; ncount = btoc(maxsize); if ((pgoff = (int)maxsize & PAGE_MASK) != 0) { vm_page_lock_queues(); vm_page_clear_dirty(m[ncount - 1], pgoff, PAGE_SIZE - pgoff); vm_page_unlock_queues(); } } else { maxsize = 0; ncount = 0; } if (ncount < count) { for (i = ncount; i < count; i++) { rtvals[i] = VM_PAGER_BAD; } } } /* * pageouts are already clustered, use IO_ASYNC t o force a bawrite() * rather then a bdwrite() to prevent paging I/O from saturating * the buffer cache. Dummy-up the sequential heuristic to cause * large ranges to cluster. If neither IO_SYNC or IO_ASYNC is set, * the system decides how to cluster. */ ioflags = IO_VMIO; if (flags & (VM_PAGER_PUT_SYNC | VM_PAGER_PUT_INVAL)) ioflags |= IO_SYNC; else if ((flags & VM_PAGER_CLUSTER_OK) == 0) ioflags |= IO_ASYNC; ioflags |= (flags & VM_PAGER_PUT_INVAL) ? IO_INVAL: 0; ioflags |= IO_SEQMAX << IO_SEQSHIFT; aiov.iov_base = (caddr_t) 0; aiov.iov_len = maxsize; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_offset = poffset; auio.uio_segflg = UIO_NOCOPY; auio.uio_rw = UIO_WRITE; auio.uio_resid = maxsize; auio.uio_td = (struct thread *) 0; error = VOP_WRITE(vp, &auio, ioflags, curthread->td_ucred); PCPU_INC(cnt.v_vnodeout); PCPU_ADD(cnt.v_vnodepgsout, ncount); if (error) { if ((ppscheck = ppsratecheck(&lastfail, &curfail, 1))) printf("vnode_pager_putpages: I/O error %d\n", error); } if (auio.uio_resid) { if (ppscheck || ppsratecheck(&lastfail, &curfail, 1)) printf("vnode_pager_putpages: residual I/O %d at %lu\n", auio.uio_resid, (u_long)m[0]->pindex); } for (i = 0; i < ncount; i++) { rtvals[i] = VM_PAGER_OK; } return rtvals[0]; } struct vnode * vnode_pager_lock(vm_object_t first_object) { struct vnode *vp; vm_object_t backing_object, object; VM_OBJECT_LOCK_ASSERT(first_object, MA_OWNED); for (object = first_object; object != NULL; object = backing_object) { if (object->type != OBJT_VNODE) { if ((backing_object = object->backing_object) != NULL) VM_OBJECT_LOCK(backing_object); if (object != first_object) VM_OBJECT_UNLOCK(object); continue; } retry: if (object->flags & OBJ_DEAD) { if (object != first_object) VM_OBJECT_UNLOCK(object); return NULL; } vp = object->handle; VI_LOCK(vp); VM_OBJECT_UNLOCK(object); if (first_object != object) VM_OBJECT_UNLOCK(first_object); VFS_ASSERT_GIANT(vp->v_mount); if (vget(vp, LK_CANRECURSE | LK_INTERLOCK | LK_RETRY | LK_SHARED, curthread)) { VM_OBJECT_LOCK(first_object); if (object != first_object) VM_OBJECT_LOCK(object); if (object->type != OBJT_VNODE) { if (object != first_object) VM_OBJECT_UNLOCK(object); return NULL; } printf("vnode_pager_lock: retrying\n"); goto retry; } VM_OBJECT_LOCK(first_object); return (vp); } return NULL; }