diff --git a/sys/fs/devfs/devfs_vnops.c b/sys/fs/devfs/devfs_vnops.c index 043cee74fde2..a1a06eb6c3e6 100644 --- a/sys/fs/devfs/devfs_vnops.c +++ b/sys/fs/devfs/devfs_vnops.c @@ -1,2123 +1,2123 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * 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: * mkdir: want it ? */ #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 #include static struct vop_vector devfs_vnodeops; static struct vop_vector devfs_specops; static struct fileops devfs_ops_f; #include #include #include #include #include #include static MALLOC_DEFINE(M_CDEVPDATA, "DEVFSP", "Metainfo for cdev-fp data"); 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"); struct mtx cdevpriv_mtx; MTX_SYSINIT(cdevpriv_mtx, &cdevpriv_mtx, "cdevpriv lock", MTX_DEF); SYSCTL_DECL(_vfs_devfs); static int devfs_dotimes; SYSCTL_INT(_vfs_devfs, OID_AUTO, dotimes, CTLFLAG_RW, &devfs_dotimes, 0, "Update timestamps on DEVFS with default precision"); /* * Update devfs node timestamp. Note that updates are unlocked and * stat(2) could see partially updated times. */ static void devfs_timestamp(struct timespec *tsp) { time_t ts; if (devfs_dotimes) { vfs_timestamp(tsp); } else { ts = time_second; if (tsp->tv_sec != ts) { tsp->tv_sec = ts; tsp->tv_nsec = 0; } } } static int devfs_fp_check(struct file *fp, struct cdev **devp, struct cdevsw **dswp, int *ref) { *dswp = devvn_refthread(fp->f_vnode, devp, ref); if (*devp != fp->f_data) { if (*dswp != NULL) dev_relthread(*devp, *ref); return (ENXIO); } KASSERT((*devp)->si_refcount > 0, ("devfs: un-referenced struct cdev *(%s)", devtoname(*devp))); if (*dswp == NULL) return (ENXIO); curthread->td_fpop = fp; return (0); } int devfs_get_cdevpriv(void **datap) { struct file *fp; struct cdev_privdata *p; int error; fp = curthread->td_fpop; if (fp == NULL) return (EBADF); p = fp->f_cdevpriv; if (p != NULL) { error = 0; *datap = p->cdpd_data; } else error = ENOENT; return (error); } int devfs_set_cdevpriv(void *priv, d_priv_dtor_t *priv_dtr) { struct file *fp; struct cdev_priv *cdp; struct cdev_privdata *p; int error; fp = curthread->td_fpop; if (fp == NULL) return (ENOENT); cdp = cdev2priv((struct cdev *)fp->f_data); p = malloc(sizeof(struct cdev_privdata), M_CDEVPDATA, M_WAITOK); p->cdpd_data = priv; p->cdpd_dtr = priv_dtr; p->cdpd_fp = fp; mtx_lock(&cdevpriv_mtx); if (fp->f_cdevpriv == NULL) { LIST_INSERT_HEAD(&cdp->cdp_fdpriv, p, cdpd_list); fp->f_cdevpriv = p; mtx_unlock(&cdevpriv_mtx); error = 0; } else { mtx_unlock(&cdevpriv_mtx); free(p, M_CDEVPDATA); error = EBUSY; } return (error); } void devfs_destroy_cdevpriv(struct cdev_privdata *p) { mtx_assert(&cdevpriv_mtx, MA_OWNED); KASSERT(p->cdpd_fp->f_cdevpriv == p, ("devfs_destoy_cdevpriv %p != %p", p->cdpd_fp->f_cdevpriv, p)); p->cdpd_fp->f_cdevpriv = NULL; LIST_REMOVE(p, cdpd_list); mtx_unlock(&cdevpriv_mtx); (p->cdpd_dtr)(p->cdpd_data); free(p, M_CDEVPDATA); } static void devfs_fpdrop(struct file *fp) { struct cdev_privdata *p; mtx_lock(&cdevpriv_mtx); if ((p = fp->f_cdevpriv) == NULL) { mtx_unlock(&cdevpriv_mtx); return; } devfs_destroy_cdevpriv(p); } void devfs_clear_cdevpriv(void) { struct file *fp; fp = curthread->td_fpop; if (fp == NULL) return; devfs_fpdrop(fp); } static void devfs_usecount_add(struct vnode *vp) { struct devfs_dirent *de; struct cdev *dev; mtx_lock(&devfs_de_interlock); VI_LOCK(vp); VNPASS(vp->v_type == VCHR || vp->v_type == VBAD, vp); if (VN_IS_DOOMED(vp)) { goto out_unlock; } de = vp->v_data; dev = vp->v_rdev; MPASS(de != NULL); MPASS(dev != NULL); dev->si_usecount++; de->de_usecount++; out_unlock: VI_UNLOCK(vp); mtx_unlock(&devfs_de_interlock); } static void devfs_usecount_subl(struct vnode *vp) { struct devfs_dirent *de; struct cdev *dev; mtx_assert(&devfs_de_interlock, MA_OWNED); ASSERT_VI_LOCKED(vp, __func__); VNPASS(vp->v_type == VCHR || vp->v_type == VBAD, vp); de = vp->v_data; dev = vp->v_rdev; if (de == NULL) return; if (dev == NULL) { MPASS(de->de_usecount == 0); return; } if (dev->si_usecount < de->de_usecount) panic("%s: si_usecount underflow for dev %p " "(has %ld, dirent has %d)\n", __func__, dev, dev->si_usecount, de->de_usecount); if (VN_IS_DOOMED(vp)) { dev->si_usecount -= de->de_usecount; de->de_usecount = 0; } else { if (de->de_usecount == 0) panic("%s: de_usecount underflow for dev %p\n", __func__, dev); dev->si_usecount--; de->de_usecount--; } } static void devfs_usecount_sub(struct vnode *vp) { mtx_lock(&devfs_de_interlock); VI_LOCK(vp); devfs_usecount_subl(vp); VI_UNLOCK(vp); mtx_unlock(&devfs_de_interlock); } static int devfs_usecountl(struct vnode *vp) { VNPASS(vp->v_type == VCHR, vp); mtx_assert(&devfs_de_interlock, MA_OWNED); ASSERT_VI_LOCKED(vp, __func__); return (vp->v_rdev->si_usecount); } int devfs_usecount(struct vnode *vp) { int count; VNPASS(vp->v_type == VCHR, vp); mtx_lock(&devfs_de_interlock); VI_LOCK(vp); count = devfs_usecountl(vp); VI_UNLOCK(vp); mtx_unlock(&devfs_de_interlock); return (count); } void devfs_ctty_ref(struct vnode *vp) { vrefact(vp); devfs_usecount_add(vp); } void devfs_ctty_unref(struct vnode *vp) { devfs_usecount_sub(vp); vrele(vp); } /* * On success devfs_populate_vp() returns with dmp->dm_lock held. */ static int devfs_populate_vp(struct vnode *vp) { struct devfs_dirent *de; struct devfs_mount *dmp; int locked; ASSERT_VOP_LOCKED(vp, "devfs_populate_vp"); dmp = VFSTODEVFS(vp->v_mount); if (!devfs_populate_needed(dmp)) { sx_xlock(&dmp->dm_lock); goto out_nopopulate; } locked = VOP_ISLOCKED(vp); sx_xlock(&dmp->dm_lock); DEVFS_DMP_HOLD(dmp); /* Can't call devfs_populate() with the vnode lock held. */ VOP_UNLOCK(vp); devfs_populate(dmp); sx_xunlock(&dmp->dm_lock); vn_lock(vp, locked | LK_RETRY); sx_xlock(&dmp->dm_lock); if (DEVFS_DMP_DROP(dmp)) { sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); return (ERESTART); } out_nopopulate: if (VN_IS_DOOMED(vp)) { sx_xunlock(&dmp->dm_lock); return (ERESTART); } de = vp->v_data; KASSERT(de != NULL, ("devfs_populate_vp: vp->v_data == NULL but vnode not doomed")); if ((de->de_flags & DE_DOOMED) != 0) { sx_xunlock(&dmp->dm_lock); return (ERESTART); } return (0); } static int devfs_vptocnp(struct vop_vptocnp_args *ap) { struct vnode *vp = ap->a_vp; struct vnode **dvp = ap->a_vpp; struct devfs_mount *dmp; char *buf = ap->a_buf; size_t *buflen = ap->a_buflen; struct devfs_dirent *dd, *de; int i, error; dmp = VFSTODEVFS(vp->v_mount); error = devfs_populate_vp(vp); if (error != 0) return (error); if (vp->v_type != VCHR && vp->v_type != VDIR) { error = ENOENT; goto finished; } dd = vp->v_data; if (vp->v_type == VDIR && dd == dmp->dm_rootdir) { *dvp = vp; vref(*dvp); goto finished; } i = *buflen; i -= dd->de_dirent->d_namlen; if (i < 0) { error = ENOMEM; goto finished; } bcopy(dd->de_dirent->d_name, buf + i, dd->de_dirent->d_namlen); *buflen = i; de = devfs_parent_dirent(dd); if (de == NULL) { error = ENOENT; goto finished; } mtx_lock(&devfs_de_interlock); *dvp = de->de_vnode; if (*dvp != NULL) { VI_LOCK(*dvp); mtx_unlock(&devfs_de_interlock); vholdl(*dvp); VI_UNLOCK(*dvp); vref(*dvp); vdrop(*dvp); } else { mtx_unlock(&devfs_de_interlock); error = ENOENT; } finished: sx_xunlock(&dmp->dm_lock); return (error); } /* * Construct the fully qualified path name relative to the mountpoint. * If a NULL cnp is provided, no '/' is appended to the resulting path. */ char * devfs_fqpn(char *buf, struct devfs_mount *dmp, struct devfs_dirent *dd, struct componentname *cnp) { int i; struct devfs_dirent *de; sx_assert(&dmp->dm_lock, SA_LOCKED); i = SPECNAMELEN; buf[i] = '\0'; if (cnp != NULL) i -= cnp->cn_namelen; if (i < 0) return (NULL); if (cnp != NULL) bcopy(cnp->cn_nameptr, buf + i, cnp->cn_namelen); de = dd; while (de != dmp->dm_rootdir) { if (cnp != NULL || i < SPECNAMELEN) { 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 = devfs_parent_dirent(de); if (de == NULL) return (NULL); } 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, int lockmode, struct vnode **vpp) { int error; struct vnode *vp; struct cdev *dev; struct devfs_mount *dmp; struct cdevsw *dsw; enum vgetstate vs; 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) { vs = vget_prep(vp); mtx_unlock(&devfs_de_interlock); sx_xunlock(&dmp->dm_lock); vget_finish(vp, lockmode | LK_RETRY, vs); sx_xlock(&dmp->dm_lock); if (devfs_allocv_drop_refs(0, dmp, de)) { vput(vp); return (ENOENT); } else if (VN_IS_DOOMED(vp)) { mtx_lock(&devfs_de_interlock); if (de->de_vnode == vp) { de->de_vnode = NULL; vp->v_data = NULL; } mtx_unlock(&devfs_de_interlock); vput(vp); 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; VNPASS(vp->v_usecount == 1, vp); /* Special casing of ttys for deadfs. Probably redundant. */ dsw = dev->si_devsw; if (dsw != NULL && (dsw->d_flags & D_TTY) != 0) vp->v_vflag |= VV_ISTTY; dev_unlock(); VI_UNLOCK(vp); if ((dev->si_flags & SI_ETERNAL) != 0) vp->v_vflag |= VV_ETERNALDEV; 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 | LK_NOWITNESS); VN_LOCK_ASHARE(vp); 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_devfs_vnode_associate(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; struct proc *p; 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_accmode, ap->a_cred); if (error == 0) return (0); if (error != EACCES) return (error); p = ap->a_td->td_proc; /* We do, however, allow access to the controlling terminal */ PROC_LOCK(p); if (!(p->p_flag & P_CONTROLT)) { PROC_UNLOCK(p); return (error); } if (p->p_session->s_ttydp == de->de_cdp) error = 0; PROC_UNLOCK(p); return (error); } _Static_assert(((FMASK | FCNTLFLAGS) & (FLASTCLOSE | FREVOKE)) == 0, "devfs-only flag reuse failed"); static int devfs_close(struct vop_close_args *ap) { struct vnode *vp = ap->a_vp, *oldvp; struct thread *td = ap->a_td; struct proc *p; struct cdev *dev = vp->v_rdev; struct cdevsw *dsw; struct devfs_dirent *de = vp->v_data; int dflags, error, ref, vp_locked; /* * XXX: Don't call d_close() if we were called because of * XXX: insmntque1() failure. */ if (vp->v_data == NULL) return (0); /* * 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. */ if (de->de_usecount == 2 && td != NULL) { p = td->td_proc; PROC_LOCK(p); if (vp == p->p_session->s_ttyvp) { PROC_UNLOCK(p); oldvp = NULL; sx_xlock(&proctree_lock); if (vp == p->p_session->s_ttyvp) { SESS_LOCK(p->p_session); mtx_lock(&devfs_de_interlock); VI_LOCK(vp); if (devfs_usecountl(vp) == 2 && !VN_IS_DOOMED(vp)) { p->p_session->s_ttyvp = NULL; p->p_session->s_ttydp = NULL; oldvp = vp; } VI_UNLOCK(vp); mtx_unlock(&devfs_de_interlock); SESS_UNLOCK(p->p_session); } sx_xunlock(&proctree_lock); if (oldvp != NULL) devfs_ctty_unref(oldvp); } else PROC_UNLOCK(p); } /* * 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, &ref); if (dsw == NULL) return (ENXIO); dflags = 0; mtx_lock(&devfs_de_interlock); VI_LOCK(vp); if (devfs_usecountl(vp) == 1) dflags |= FLASTCLOSE; devfs_usecount_subl(vp); mtx_unlock(&devfs_de_interlock); if (VN_IS_DOOMED(vp)) { /* Forced close. */ dflags |= FREVOKE | FNONBLOCK; } else if (dsw->d_flags & D_TRACKCLOSE) { /* Keep device updated on status. */ } else if ((dflags & FLASTCLOSE) == 0) { VI_UNLOCK(vp); dev_relthread(dev, ref); return (0); } vholdnz(vp); VI_UNLOCK(vp); vp_locked = VOP_ISLOCKED(vp); VOP_UNLOCK(vp); KASSERT(dev->si_refcount > 0, ("devfs_close() on un-referenced struct cdev *(%s)", devtoname(dev))); error = dsw->d_close(dev, ap->a_fflag | dflags, S_IFCHR, td); dev_relthread(dev, ref); vn_lock(vp, vp_locked | LK_RETRY); vdrop(vp); return (error); } static int devfs_close_f(struct file *fp, struct thread *td) { int error; struct file *fpop; /* * NB: td may be NULL if this descriptor is closed due to * garbage collection from a closed UNIX domain socket. */ fpop = curthread->td_fpop; curthread->td_fpop = fp; error = vnops.fo_close(fp, td); curthread->td_fpop = fpop; /* * The f_cdevpriv cannot be assigned non-NULL value while we * are destroying the file. */ if (fp->f_cdevpriv != NULL) devfs_fpdrop(fp); return (error); } static int devfs_getattr(struct vop_getattr_args *ap) { struct vnode *vp = ap->a_vp; struct vattr *vap = ap->a_vap; struct devfs_dirent *de; struct devfs_mount *dmp; struct cdev *dev; struct timeval boottime; int error; error = devfs_populate_vp(vp); if (error != 0) return (error); dmp = VFSTODEVFS(vp->v_mount); sx_xunlock(&dmp->dm_lock); 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)); } 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; getboottime(&boottime); #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 = cdev2priv(dev)->cdp_inode; } vap->va_gen = 0; vap->va_flags = 0; vap->va_filerev = 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 file *fpop; int error; fpop = td->td_fpop; td->td_fpop = fp; error = vnops.fo_ioctl(fp, com, data, cred, td); td->td_fpop = fpop; return (error); } void * fiodgname_buf_get_ptr(void *fgnp, u_long com) { union { struct fiodgname_arg fgn; #ifdef COMPAT_FREEBSD32 struct fiodgname_arg32 fgn32; #endif } *fgnup; fgnup = fgnp; switch (com) { case FIODGNAME: return (fgnup->fgn.buf); #ifdef COMPAT_FREEBSD32 case FIODGNAME_32: return ((void *)(uintptr_t)fgnup->fgn32.buf); #endif default: panic("Unhandled ioctl command %ld", com); } } static int devfs_ioctl(struct vop_ioctl_args *ap) { struct fiodgname_arg *fgn; struct vnode *vpold, *vp; struct cdevsw *dsw; struct thread *td; struct session *sess; struct cdev *dev; int error, ref, i; const char *p; u_long com; vp = ap->a_vp; com = ap->a_command; td = ap->a_td; dsw = devvn_refthread(vp, &dev, &ref); if (dsw == NULL) return (ENXIO); KASSERT(dev->si_refcount > 0, ("devfs: un-referenced struct cdev *(%s)", devtoname(dev))); switch (com) { case FIODTYPE: *(int *)ap->a_data = dsw->d_flags & D_TYPEMASK; error = 0; break; case FIODGNAME: #ifdef COMPAT_FREEBSD32 case FIODGNAME_32: #endif fgn = ap->a_data; p = devtoname(dev); i = strlen(p) + 1; if (i > fgn->len) error = EINVAL; else error = copyout(p, fiodgname_buf_get_ptr(fgn, com), i); break; default: error = dsw->d_ioctl(dev, com, ap->a_data, ap->a_fflag, td); } dev_relthread(dev, ref); if (error == ENOIOCTL) error = ENOTTY; if (error == 0 && com == TIOCSCTTY) { /* * Do nothing if reassigning same control tty, or if the * control tty has already disappeared. If it disappeared, * it's because we were racing with TIOCNOTTY. TIOCNOTTY * already took care of releasing the old vnode and we have * nothing left to do. */ sx_slock(&proctree_lock); sess = td->td_proc->p_session; if (sess->s_ttyvp == vp || sess->s_ttyp == NULL) { sx_sunlock(&proctree_lock); return (0); } devfs_ctty_ref(vp); SESS_LOCK(sess); vpold = sess->s_ttyvp; sess->s_ttyvp = vp; sess->s_ttydp = cdev2priv(dev); SESS_UNLOCK(sess); sx_sunlock(&proctree_lock); /* Get rid of reference to old control tty */ if (vpold) devfs_ctty_unref(vpold); } return (error); } /* ARGSUSED */ static int devfs_kqfilter_f(struct file *fp, struct knote *kn) { struct cdev *dev; struct cdevsw *dsw; int error, ref; struct file *fpop; struct thread *td; td = curthread; fpop = td->td_fpop; error = devfs_fp_check(fp, &dev, &dsw, &ref); if (error) return (error); error = dsw->d_kqfilter(dev, kn); td->td_fpop = fpop; dev_relthread(dev, ref); return (error); } static inline int devfs_prison_check(struct devfs_dirent *de, struct thread *td) { struct cdev_priv *cdp; struct ucred *dcr; struct proc *p; int error; cdp = de->de_cdp; if (cdp == NULL) return (0); dcr = cdp->cdp_c.si_cred; if (dcr == NULL) return (0); error = prison_check(td->td_ucred, dcr); if (error == 0) return (0); /* We do, however, allow access to the controlling terminal */ p = td->td_proc; PROC_LOCK(p); if (!(p->p_flag & P_CONTROLT)) { PROC_UNLOCK(p); return (error); } if (p->p_session->s_ttydp == cdp) error = 0; PROC_UNLOCK(p); 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 mount *mp; struct cdev *cdev; int error, flags, nameiop, dvplocked; 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; mp = dvp->v_mount; dmp = VFSTODEVFS(mp); 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 = vn_dir_check_exec(dvp, cnp); if (error != 0) 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); de = devfs_parent_dirent(dd); if (de == NULL) return (ENOENT); dvplocked = VOP_ISLOCKED(dvp); VOP_UNLOCK(dvp); error = devfs_allocv(de, mp, cnp->cn_lkflags & LK_TYPE_MASK, vpp); *dm_unlock = 0; vn_lock(dvp, dvplocked | LK_RETRY); return (error); } dd = dvp->v_data; de = devfs_find(dd, cnp->cn_nameptr, cnp->cn_namelen, 0); 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, dmp, dd, 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); if (cdev == NULL) sx_xlock(&dmp->dm_lock); else if (devfs_populate_vp(dvp) != 0) { *dm_unlock = 0; sx_xlock(&dmp->dm_lock); if (DEVFS_DMP_DROP(dmp)) { sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); } else sx_xunlock(&dmp->dm_lock); dev_rel(cdev); return (ENOENT); } if (DEVFS_DMP_DROP(dmp)) { *dm_unlock = 0; sx_xunlock(&dmp->dm_lock); devfs_unmount_final(dmp); if (cdev != NULL) dev_rel(cdev); return (ENOENT); } if (cdev == NULL) break; dev_lock(); dde = &cdev2priv(cdev)->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 (devfs_prison_check(de, td)) 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, mp, cnp->cn_lkflags & LK_TYPE_MASK, vpp); *dm_unlock = 0; return (error); } static int devfs_lookup(struct vop_lookup_args *ap) { int j; struct devfs_mount *dmp; int dm_unlock; if (devfs_populate_vp(ap->a_dvp) != 0) return (ENOTDIR); dmp = VFSTODEVFS(ap->a_dvp->v_mount); dm_unlock = 1; 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 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; 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 (de->de_dirent->d_type == DT_CHR && (de->de_cdp->cdp_flags & CDP_ACTIVE) == 0) 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, LK_EXCLUSIVE, vpp); 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, ref, vlocked; struct cdevsw *dsw; struct file *fpop; 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, &ref); if (dsw == NULL) return (ENXIO); if (fp == NULL && dsw->d_fdopen != NULL) { dev_relthread(dev, ref); return (ENXIO); } if (vp->v_type == VCHR) devfs_usecount_add(vp); vlocked = VOP_ISLOCKED(vp); VOP_UNLOCK(vp); fpop = td->td_fpop; td->td_fpop = fp; if (fp != NULL) { fp->f_data = dev; fp->f_vnode = vp; } 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); /* Clean up any cdevpriv upon error. */ if (error != 0) devfs_clear_cdevpriv(); td->td_fpop = fpop; vn_lock(vp, vlocked | LK_RETRY); if (error != 0 && vp->v_type == VCHR) devfs_usecount_sub(vp); dev_relthread(dev, ref); if (error != 0) { if (error == ERESTART) error = EINTR; return (error); } #if 0 /* /dev/console */ KASSERT(fp != NULL, ("Could not vnode bypass device on NULL fp")); #else if (fp == NULL) return (error); #endif if (fp->f_ops == &badfileops) finit(fp, fp->f_flag, DTYPE_VNODE, dev, &devfs_ops_f); return (error); } static int devfs_pathconf(struct vop_pathconf_args *ap) { switch (ap->a_name) { case _PC_FILESIZEBITS: *ap->a_retval = 64; return (0); case _PC_NAME_MAX: *ap->a_retval = NAME_MAX; return (0); case _PC_LINK_MAX: *ap->a_retval = INT_MAX; return (0); case _PC_SYMLINK_MAX: *ap->a_retval = MAXPATHLEN; return (0); case _PC_MAX_CANON: if (ap->a_vp->v_vflag & VV_ISTTY) { *ap->a_retval = MAX_CANON; return (0); } return (EINVAL); case _PC_MAX_INPUT: if (ap->a_vp->v_vflag & VV_ISTTY) { *ap->a_retval = MAX_INPUT; return (0); } return (EINVAL); case _PC_VDISABLE: if (ap->a_vp->v_vflag & VV_ISTTY) { *ap->a_retval = _POSIX_VDISABLE; return (0); } return (EINVAL); case _PC_MAC_PRESENT: #ifdef MAC /* * If MAC is enabled, devfs automatically supports - * trivial non-persistant label storage. + * trivial non-persistent label storage. */ *ap->a_retval = 1; #else *ap->a_retval = 0; #endif return (0); case _PC_CHOWN_RESTRICTED: *ap->a_retval = 1; 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, ref; struct file *fpop; fpop = td->td_fpop; error = devfs_fp_check(fp, &dev, &dsw, &ref); if (error != 0) { error = vnops.fo_poll(fp, events, cred, td); return (error); } error = dsw->d_poll(dev, events, td); td->td_fpop = fpop; dev_relthread(dev, ref); 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); } 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, ref; ssize_t resid; struct cdevsw *dsw; struct file *fpop; if (uio->uio_resid > DEVFS_IOSIZE_MAX) return (EINVAL); fpop = td->td_fpop; error = devfs_fp_check(fp, &dev, &dsw, &ref); if (error != 0) { error = vnops.fo_read(fp, uio, cred, flags, td); return (error); } resid = uio->uio_resid; ioflag = fp->f_flag & (O_NONBLOCK | O_DIRECT); if (ioflag & O_DIRECT) ioflag |= IO_DIRECT; foffset_lock_uio(fp, uio, flags | FOF_NOLOCK); error = dsw->d_read(dev, uio, ioflag); if (uio->uio_resid != resid || (error == 0 && resid != 0)) devfs_timestamp(&dev->si_atime); td->td_fpop = fpop; dev_relthread(dev, ref); foffset_unlock_uio(fp, uio, flags | FOF_NOLOCK | FOF_NEXTOFF_R); 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; 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); if (devfs_populate_vp(ap->a_vp) != 0) { if (tmp_ncookies != NULL) ap->a_ncookies = tmp_ncookies; return (EIO); } error = 0; de = ap->a_vp->v_data; off = 0; TAILQ_FOREACH(dd, &de->de_dlist, de_list) { KASSERT(dd->de_cdp != (void *)0xdeadc0de, ("%s %d\n", __func__, __LINE__)); if (dd->de_flags & (DE_COVERED | DE_WHITEOUT)) continue; if (devfs_prison_check(dd, uio->uio_td)) continue; if (dd->de_dirent->d_type == DT_DIR) de = dd->de_dir; else de = dd; dp = dd->de_dirent; MPASS(dp->d_reclen == GENERIC_DIRSIZ(dp)); if (dp->d_reclen > uio->uio_resid) break; dp->d_fileno = de->de_inode; /* NOTE: d_off is the offset for the *next* entry. */ dp->d_off = off + dp->d_reclen; 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 void devfs_reclaiml(struct vnode *vp) { struct devfs_dirent *de; mtx_assert(&devfs_de_interlock, MA_OWNED); de = vp->v_data; if (de != NULL) { MPASS(de->de_usecount == 0); de->de_vnode = NULL; vp->v_data = NULL; } } static int devfs_reclaim(struct vop_reclaim_args *ap) { struct vnode *vp; vp = ap->a_vp; mtx_lock(&devfs_de_interlock); devfs_reclaiml(vp); mtx_unlock(&devfs_de_interlock); return (0); } static int devfs_reclaim_vchr(struct vop_reclaim_args *ap) { struct vnode *vp; struct cdev *dev; vp = ap->a_vp; MPASS(vp->v_type == VCHR); mtx_lock(&devfs_de_interlock); VI_LOCK(vp); devfs_usecount_subl(vp); devfs_reclaiml(vp); mtx_unlock(&devfs_de_interlock); dev_lock(); dev = vp->v_rdev; vp->v_rdev = NULL; dev_unlock(); VI_UNLOCK(vp); if (dev != NULL) dev_rel(dev); return (0); } static int devfs_remove(struct vop_remove_args *ap) { struct vnode *dvp = ap->a_dvp; struct vnode *vp = ap->a_vp; struct devfs_dirent *dd; struct devfs_dirent *de, *de_covered; struct devfs_mount *dmp = VFSTODEVFS(vp->v_mount); ASSERT_VOP_ELOCKED(dvp, "devfs_remove"); ASSERT_VOP_ELOCKED(vp, "devfs_remove"); 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); if (de->de_dirent->d_type == DT_LNK) { de_covered = devfs_find(dd, de->de_dirent->d_name, de->de_dirent->d_namlen, 0); if (de_covered != NULL) de_covered->de_flags &= ~DE_COVERED; } /* We need to unlock dvp because devfs_delete() may lock it. */ VOP_UNLOCK(vp); if (dvp != vp) VOP_UNLOCK(dvp); devfs_delete(dmp, de, 0); sx_xunlock(&dmp->dm_lock); if (dvp != vp) vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); } 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; enum vgetstate vs; u_int i; KASSERT((ap->a_flags & REVOKEALL) != 0, ("devfs_revoke !REVOKEALL")); dev = vp->v_rdev; cdp = cdev2priv(dev); dev_lock(); cdp->cdp_inuse++; dev_unlock(); vhold(vp); vgone(vp); vdrop(vp); VOP_UNLOCK(vp); 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(); vs = vget_prep(vp2); mtx_unlock(&devfs_de_interlock); if (vget_finish(vp2, LK_EXCLUSIVE, vs) != 0) 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); return (0); } static int devfs_rioctl(struct vop_ioctl_args *ap) { struct vnode *vp; struct devfs_mount *dmp; int error; vp = ap->a_vp; vn_lock(vp, LK_SHARED | LK_RETRY); if (VN_IS_DOOMED(vp)) { VOP_UNLOCK(vp); return (EBADF); } dmp = VFSTODEVFS(vp->v_mount); sx_xlock(&dmp->dm_lock); VOP_UNLOCK(vp); 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; struct thread *td; int c, error; uid_t uid; gid_t gid; vap = ap->a_vap; vp = ap->a_vp; td = curthread; 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); } error = devfs_populate_vp(vp); if (error != 0) return (error); de = vp->v_data; if (vp->v_type == VDIR) de = de->de_dir; 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(td, PRIV_VFS_CHOWN); if (error != 0) goto ret; } 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(td, PRIV_VFS_ADMIN); if (error != 0) goto ret; } de->de_mode = vap->va_mode; c = 1; } if (vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL) { error = vn_utimes_perm(vp, vap, ap->a_cred, td); if (error != 0) goto ret; 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); } ret: sx_xunlock(&VFSTODEVFS(vp->v_mount)->dm_lock); return (error); } #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_vnode_relabel(ap->a_cred, vp, ap->a_label); mac_devfs_update(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, *de_covered, *de_dotdot; struct devfs_mount *dmp; error = priv_check(curthread, PRIV_DEVFS_SYMLINK); if (error) return(error); dmp = VFSTODEVFS(ap->a_dvp->v_mount); if (devfs_populate_vp(ap->a_dvp) != 0) return (ENOENT); dd = ap->a_dvp->v_data; de = devfs_newdirent(ap->a_cnp->cn_nameptr, ap->a_cnp->cn_namelen); de->de_flags = DE_USER; de->de_uid = 0; de->de_gid = 0; de->de_mode = 0755; de->de_inode = alloc_unr(devfs_inos); de->de_dir = dd; 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); #ifdef MAC mac_devfs_create_symlink(ap->a_cnp->cn_cred, dmp->dm_mount, dd, de); #endif de_covered = devfs_find(dd, de->de_dirent->d_name, de->de_dirent->d_namlen, 0); if (de_covered != NULL) { if ((de_covered->de_flags & DE_USER) != 0) { devfs_delete(dmp, de, DEVFS_DEL_NORECURSE); sx_xunlock(&dmp->dm_lock); return (EEXIST); } KASSERT((de_covered->de_flags & DE_COVERED) == 0, ("devfs_symlink: entry %p already covered", de_covered)); de_covered->de_flags |= DE_COVERED; } de_dotdot = TAILQ_FIRST(&dd->de_dlist); /* "." */ de_dotdot = TAILQ_NEXT(de_dotdot, de_list); /* ".." */ TAILQ_INSERT_AFTER(&dd->de_dlist, de_dotdot, de, de_list); devfs_dir_ref_de(dmp, dd); devfs_rules_apply(dmp, de); return (devfs_allocv(de, ap->a_dvp->v_mount, LK_EXCLUSIVE, ap->a_vpp)); } static int devfs_truncate_f(struct file *fp, off_t length, struct ucred *cred, struct thread *td) { return (vnops.fo_truncate(fp, length, cred, td)); } 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, ref; ssize_t resid; struct cdevsw *dsw; struct file *fpop; if (uio->uio_resid > DEVFS_IOSIZE_MAX) return (EINVAL); fpop = td->td_fpop; error = devfs_fp_check(fp, &dev, &dsw, &ref); if (error != 0) { error = vnops.fo_write(fp, uio, cred, flags, td); 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; foffset_lock_uio(fp, uio, flags | FOF_NOLOCK); resid = uio->uio_resid; error = dsw->d_write(dev, uio, ioflag); if (uio->uio_resid != resid || (error == 0 && resid != 0)) { devfs_timestamp(&dev->si_ctime); dev->si_mtime = dev->si_ctime; } td->td_fpop = fpop; dev_relthread(dev, ref); foffset_unlock_uio(fp, uio, flags | FOF_NOLOCK | FOF_NEXTOFF_W); return (error); } static int devfs_mmap_f(struct file *fp, vm_map_t map, vm_offset_t *addr, vm_size_t size, vm_prot_t prot, vm_prot_t cap_maxprot, int flags, vm_ooffset_t foff, struct thread *td) { struct cdev *dev; struct cdevsw *dsw; struct mount *mp; struct vnode *vp; struct file *fpop; vm_object_t object; vm_prot_t maxprot; int error, ref; vp = fp->f_vnode; /* * Ensure that file and memory protections are * compatible. */ mp = vp->v_mount; if (mp != NULL && (mp->mnt_flag & MNT_NOEXEC) != 0) { maxprot = VM_PROT_NONE; if ((prot & VM_PROT_EXECUTE) != 0) return (EACCES); } else maxprot = VM_PROT_EXECUTE; if ((fp->f_flag & FREAD) != 0) maxprot |= VM_PROT_READ; else if ((prot & VM_PROT_READ) != 0) return (EACCES); /* * If we are sharing potential changes via MAP_SHARED and we * are trying to get write permission although we opened it * without asking for it, bail out. * * Note that most character devices always share mappings. * The one exception is that D_MMAP_ANON devices * (i.e. /dev/zero) permit private writable mappings. * * Rely on vm_mmap_cdev() to fail invalid MAP_PRIVATE requests * as well as updating maxprot to permit writing for * D_MMAP_ANON devices rather than doing that here. */ if ((flags & MAP_SHARED) != 0) { if ((fp->f_flag & FWRITE) != 0) maxprot |= VM_PROT_WRITE; else if ((prot & VM_PROT_WRITE) != 0) return (EACCES); } maxprot &= cap_maxprot; fpop = td->td_fpop; error = devfs_fp_check(fp, &dev, &dsw, &ref); if (error != 0) return (error); error = vm_mmap_cdev(td, size, prot, &maxprot, &flags, dev, dsw, &foff, &object); td->td_fpop = fpop; dev_relthread(dev, ref); if (error != 0) return (error); error = vm_mmap_object(map, addr, size, prot, maxprot, flags, object, foff, FALSE, td); if (error != 0) vm_object_deallocate(object); return (error); } dev_t dev2udev(struct cdev *x) { if (x == NULL) return (NODEV); return (cdev2priv(x)->cdp_inode); } static struct fileops devfs_ops_f = { .fo_read = devfs_read_f, .fo_write = devfs_write_f, .fo_truncate = devfs_truncate_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_chmod = vn_chmod, .fo_chown = vn_chown, .fo_sendfile = vn_sendfile, .fo_seek = vn_seek, .fo_fill_kinfo = vn_fill_kinfo, .fo_mmap = devfs_mmap_f, .fo_flags = DFLAG_PASSABLE | DFLAG_SEEKABLE }; /* Vops for non-CHR vnodes in /dev. */ 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, .vop_vptocnp = devfs_vptocnp, .vop_lock1 = vop_lock, .vop_unlock = vop_unlock, .vop_islocked = vop_islocked, }; VFS_VOP_VECTOR_REGISTER(devfs_vnodeops); /* Vops for VCHR vnodes in /dev. */ static struct vop_vector devfs_specops = { .vop_default = &default_vnodeops, .vop_access = devfs_access, .vop_bmap = VOP_PANIC, .vop_close = devfs_close, .vop_create = VOP_PANIC, .vop_fsync = vop_stdfsync, .vop_getattr = devfs_getattr, .vop_ioctl = devfs_ioctl, .vop_link = VOP_PANIC, .vop_mkdir = VOP_PANIC, .vop_mknod = VOP_PANIC, .vop_open = devfs_open, .vop_pathconf = devfs_pathconf, .vop_poll = dead_poll, .vop_print = devfs_print, .vop_read = dead_read, .vop_readdir = VOP_PANIC, .vop_readlink = VOP_PANIC, .vop_reallocblks = VOP_PANIC, .vop_reclaim = devfs_reclaim_vchr, .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_vptocnp = devfs_vptocnp, .vop_write = dead_write, .vop_lock1 = vop_lock, .vop_unlock = vop_unlock, .vop_islocked = vop_islocked, }; VFS_VOP_VECTOR_REGISTER(devfs_specops); /* * 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); diff --git a/sys/fs/ext2fs/ext2_alloc.c b/sys/fs/ext2fs/ext2_alloc.c index 4c265a1a8bcc..79df804aceb8 100644 --- a/sys/fs/ext2fs/ext2_alloc.c +++ b/sys/fs/ext2fs/ext2_alloc.c @@ -1,1577 +1,1577 @@ /*- * modified for Lites 1.1 * * Aug 1995, Godmar Back (gback@cs.utah.edu) * University of Utah, Department of Computer Science */ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1989, 1993 * The Regents of the University of California. 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. * 3. 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. * * @(#)ffs_alloc.c 8.8 (Berkeley) 2/21/94 * $FreeBSD$ */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include SDT_PROVIDER_DEFINE(ext2fs); /* * ext2fs trace probe: * arg0: verbosity. Higher numbers give more verbose messages * arg1: Textual message */ SDT_PROBE_DEFINE2(ext2fs, , alloc, trace, "int", "char*"); SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_reallocblks_realloc, "ino_t", "e2fs_lbn_t", "e2fs_lbn_t"); SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_bap, "uint32_t"); SDT_PROBE_DEFINE1(ext2fs, , alloc, ext2_reallocblks_blkno, "e2fs_daddr_t"); SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "char*", "int"); SDT_PROBE_DEFINE3(ext2fs, , alloc, ext2_nodealloccg_bmap_corrupted, "int", "daddr_t", "char*"); SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_blkfree_bad_block, "ino_t", "e4fs_daddr_t"); SDT_PROBE_DEFINE2(ext2fs, , alloc, ext2_vfree_doublefree, "char*", "ino_t"); static daddr_t ext2_alloccg(struct inode *, int, daddr_t, int); static daddr_t ext2_clusteralloc(struct inode *, int, daddr_t, int); static u_long ext2_dirpref(struct inode *); static e4fs_daddr_t ext2_hashalloc(struct inode *, int, long, int, daddr_t (*)(struct inode *, int, daddr_t, int)); static daddr_t ext2_nodealloccg(struct inode *, int, daddr_t, int); static daddr_t ext2_mapsearch(struct m_ext2fs *, char *, daddr_t); /* * Allocate a block in the filesystem. * * A preference may be optionally specified. If a preference is given * the following hierarchy is used to allocate a block: * 1) allocate the requested block. * 2) allocate a rotationally optimal block in the same cylinder. * 3) allocate a block in the same cylinder group. - * 4) quadradically rehash into other cylinder groups, until an + * 4) quadratically rehash into other cylinder groups, until an * available block is located. * If no block preference is given the following hierarchy is used * to allocate a block: * 1) allocate a block in the cylinder group that contains the * inode for the file. - * 2) quadradically rehash into other cylinder groups, until an + * 2) quadratically rehash into other cylinder groups, until an * available block is located. */ int ext2_alloc(struct inode *ip, daddr_t lbn, e4fs_daddr_t bpref, int size, struct ucred *cred, e4fs_daddr_t *bnp) { struct m_ext2fs *fs; struct ext2mount *ump; e4fs_daddr_t bno; int cg; *bnp = 0; fs = ip->i_e2fs; ump = ip->i_ump; mtx_assert(EXT2_MTX(ump), MA_OWNED); #ifdef INVARIANTS if ((u_int)size > fs->e2fs_bsize || blkoff(fs, size) != 0) { vn_printf(ip->i_devvp, "bsize = %lu, size = %d, fs = %s\n", (long unsigned int)fs->e2fs_bsize, size, fs->e2fs_fsmnt); panic("ext2_alloc: bad size"); } if (cred == NOCRED) panic("ext2_alloc: missing credential"); #endif /* INVARIANTS */ if (size == fs->e2fs_bsize && fs->e2fs_fbcount == 0) goto nospace; if (cred->cr_uid != 0 && fs->e2fs_fbcount < fs->e2fs_rbcount) goto nospace; if (bpref >= fs->e2fs_bcount) bpref = 0; if (bpref == 0) cg = ino_to_cg(fs, ip->i_number); else cg = dtog(fs, bpref); bno = (daddr_t)ext2_hashalloc(ip, cg, bpref, fs->e2fs_bsize, ext2_alloccg); if (bno > 0) { /* set next_alloc fields as done in block_getblk */ ip->i_next_alloc_block = lbn; ip->i_next_alloc_goal = bno; ip->i_blocks += btodb(fs->e2fs_bsize); ip->i_flag |= IN_CHANGE | IN_UPDATE; *bnp = bno; return (0); } nospace: EXT2_UNLOCK(ump); SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate data block"); return (ENOSPC); } /* * Allocate EA's block for inode. */ e4fs_daddr_t ext2_alloc_meta(struct inode *ip) { struct m_ext2fs *fs; daddr_t blk; fs = ip->i_e2fs; EXT2_LOCK(ip->i_ump); blk = ext2_hashalloc(ip, ino_to_cg(fs, ip->i_number), 0, fs->e2fs_bsize, ext2_alloccg); if (0 == blk) { EXT2_UNLOCK(ip->i_ump); SDT_PROBE2(ext2fs, , alloc, trace, 1, "cannot allocate meta block"); } return (blk); } /* * Reallocate a sequence of blocks into a contiguous sequence of blocks. * * The vnode and an array of buffer pointers for a range of sequential * logical blocks to be made contiguous is given. The allocator attempts * to find a range of sequential blocks starting as close as possible to * an fs_rotdelay offset from the end of the allocation for the logical * block immediately preceding the current range. If successful, the * physical block numbers in the buffer pointers and in the inode are * changed to reflect the new allocation. If unsuccessful, the allocation * is left unchanged. The success in doing the reallocation is returned. * Note that the error return is not reflected back to the user. Rather * the previous block allocation will be used. */ static SYSCTL_NODE(_vfs, OID_AUTO, ext2fs, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "EXT2FS filesystem"); static int doasyncfree = 1; SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doasyncfree, CTLFLAG_RW, &doasyncfree, 0, "Use asynchronous writes to update block pointers when freeing blocks"); static int doreallocblks = 0; SYSCTL_INT(_vfs_ext2fs, OID_AUTO, doreallocblks, CTLFLAG_RW, &doreallocblks, 0, ""); int ext2_reallocblks(struct vop_reallocblks_args *ap) { struct m_ext2fs *fs; struct inode *ip; struct vnode *vp; struct buf *sbp, *ebp; uint32_t *bap, *sbap, *ebap; struct ext2mount *ump; struct cluster_save *buflist; struct indir start_ap[EXT2_NIADDR + 1], end_ap[EXT2_NIADDR + 1], *idp; e2fs_lbn_t start_lbn, end_lbn; int soff; e2fs_daddr_t newblk, blkno; int i, len, start_lvl, end_lvl, pref, ssize; if (doreallocblks == 0) return (ENOSPC); vp = ap->a_vp; ip = VTOI(vp); fs = ip->i_e2fs; ump = ip->i_ump; if (fs->e2fs_contigsumsize <= 0 || ip->i_flag & IN_E4EXTENTS) return (ENOSPC); buflist = ap->a_buflist; len = buflist->bs_nchildren; start_lbn = buflist->bs_children[0]->b_lblkno; end_lbn = start_lbn + len - 1; #ifdef INVARIANTS for (i = 1; i < len; i++) if (buflist->bs_children[i]->b_lblkno != start_lbn + i) panic("ext2_reallocblks: non-cluster"); #endif /* * If the cluster crosses the boundary for the first indirect * block, leave space for the indirect block. Indirect blocks * are initially laid out in a position after the last direct * block. Block reallocation would usually destroy locality by * moving the indirect block out of the way to make room for * data blocks if we didn't compensate here. We should also do * this for other indirect block boundaries, but it is only * important for the first one. */ if (start_lbn < EXT2_NDADDR && end_lbn >= EXT2_NDADDR) return (ENOSPC); /* * If the latest allocation is in a new cylinder group, assume that * the filesystem has decided to move and do not force it back to * the previous cylinder group. */ if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) != dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno))) return (ENOSPC); if (ext2_getlbns(vp, start_lbn, start_ap, &start_lvl) || ext2_getlbns(vp, end_lbn, end_ap, &end_lvl)) return (ENOSPC); /* * Get the starting offset and block map for the first block. */ if (start_lvl == 0) { sbap = &ip->i_db[0]; soff = start_lbn; } else { idp = &start_ap[start_lvl - 1]; if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &sbp)) { brelse(sbp); return (ENOSPC); } sbap = (u_int *)sbp->b_data; soff = idp->in_off; } /* * If the block range spans two block maps, get the second map. */ ebap = NULL; if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) { ssize = len; } else { #ifdef INVARIANTS if (start_ap[start_lvl - 1].in_lbn == idp->in_lbn) panic("ext2_reallocblks: start == end"); #endif ssize = len - (idp->in_off + 1); if (bread(vp, idp->in_lbn, (int)fs->e2fs_bsize, NOCRED, &ebp)) goto fail; ebap = (u_int *)ebp->b_data; } /* * Find the preferred location for the cluster. */ EXT2_LOCK(ump); pref = ext2_blkpref(ip, start_lbn, soff, sbap, 0); /* * Search the block map looking for an allocation of the desired size. */ if ((newblk = (e2fs_daddr_t)ext2_hashalloc(ip, dtog(fs, pref), pref, len, ext2_clusteralloc)) == 0) { EXT2_UNLOCK(ump); goto fail; } /* * We have found a new contiguous block. * * First we have to replace the old block pointers with the new * block pointers in the inode and indirect blocks associated * with the file. */ SDT_PROBE3(ext2fs, , alloc, ext2_reallocblks_realloc, ip->i_number, start_lbn, end_lbn); blkno = newblk; for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->e2fs_fpb) { if (i == ssize) { bap = ebap; soff = -i; } #ifdef INVARIANTS if (buflist->bs_children[i]->b_blkno != fsbtodb(fs, *bap)) panic("ext2_reallocblks: alloc mismatch"); #endif SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_bap, *bap); *bap++ = blkno; } /* * Next we must write out the modified inode and indirect blocks. * For strict correctness, the writes should be synchronous since * the old block values may have been written to disk. In practise * they are almost never written, but if we are concerned about * strict correctness, the `doasyncfree' flag should be set to zero. * * The test on `doasyncfree' should be changed to test a flag * that shows whether the associated buffers and inodes have * been written. The flag should be set when the cluster is * started and cleared whenever the buffer or inode is flushed. * We can then check below to see if it is set, and do the * synchronous write only when it has been cleared. */ if (sbap != &ip->i_db[0]) { if (doasyncfree) bdwrite(sbp); else bwrite(sbp); } else { ip->i_flag |= IN_CHANGE | IN_UPDATE; if (!doasyncfree) ext2_update(vp, 1); } if (ssize < len) { if (doasyncfree) bdwrite(ebp); else bwrite(ebp); } /* * Last, free the old blocks and assign the new blocks to the buffers. */ for (blkno = newblk, i = 0; i < len; i++, blkno += fs->e2fs_fpb) { ext2_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->e2fs_bsize); buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno); SDT_PROBE1(ext2fs, , alloc, ext2_reallocblks_blkno, blkno); } return (0); fail: if (ssize < len) brelse(ebp); if (sbap != &ip->i_db[0]) brelse(sbp); return (ENOSPC); } /* * Allocate an inode in the filesystem. * */ int ext2_valloc(struct vnode *pvp, int mode, struct ucred *cred, struct vnode **vpp) { struct timespec ts; struct m_ext2fs *fs; struct ext2mount *ump; struct inode *pip; struct inode *ip; struct vnode *vp; struct thread *td; ino_t ino, ipref; int error, cg; *vpp = NULL; pip = VTOI(pvp); fs = pip->i_e2fs; ump = pip->i_ump; EXT2_LOCK(ump); if (fs->e2fs_ficount == 0) goto noinodes; /* * If it is a directory then obtain a cylinder group based on * ext2_dirpref else obtain it using ino_to_cg. The preferred inode is * always the next inode. */ if ((mode & IFMT) == IFDIR) { cg = ext2_dirpref(pip); if (fs->e2fs_contigdirs[cg] < 255) fs->e2fs_contigdirs[cg]++; } else { cg = ino_to_cg(fs, pip->i_number); if (fs->e2fs_contigdirs[cg] > 0) fs->e2fs_contigdirs[cg]--; } ipref = cg * fs->e2fs_ipg + 1; ino = (ino_t)ext2_hashalloc(pip, cg, (long)ipref, mode, ext2_nodealloccg); if (ino == 0) goto noinodes; td = curthread; error = vfs_hash_get(ump->um_mountp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL); if (error || *vpp != NULL) { return (error); } ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO); /* Allocate a new vnode/inode. */ if ((error = getnewvnode("ext2fs", ump->um_mountp, &ext2_vnodeops, &vp)) != 0) { free(ip, M_EXT2NODE); return (error); } lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL); vp->v_data = ip; ip->i_vnode = vp; ip->i_e2fs = fs = ump->um_e2fs; ip->i_ump = ump; ip->i_number = ino; ip->i_block_group = ino_to_cg(fs, ino); ip->i_next_alloc_block = 0; ip->i_next_alloc_goal = 0; error = insmntque(vp, ump->um_mountp); if (error) { free(ip, M_EXT2NODE); return (error); } error = vfs_hash_insert(vp, ino, LK_EXCLUSIVE, td, vpp, NULL, NULL); if (error || *vpp != NULL) { *vpp = NULL; free(ip, M_EXT2NODE); return (error); } if ((error = ext2_vinit(ump->um_mountp, &ext2_fifoops, &vp)) != 0) { vput(vp); *vpp = NULL; free(ip, M_EXT2NODE); return (error); } if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_EXTENTS) && (S_ISREG(mode) || S_ISDIR(mode))) ext4_ext_tree_init(ip); else memset(ip->i_data, 0, sizeof(ip->i_data)); /* * Set up a new generation number for this inode. * Avoid zero values. */ do { ip->i_gen = arc4random(); } while (ip->i_gen == 0); vfs_timestamp(&ts); ip->i_birthtime = ts.tv_sec; ip->i_birthnsec = ts.tv_nsec; *vpp = vp; return (0); noinodes: EXT2_UNLOCK(ump); SDT_PROBE2(ext2fs, , alloc, trace, 1, "out of inodes"); return (ENOSPC); } /* * 64-bit compatible getters and setters for struct ext2_gd from ext2fs.h */ uint64_t e2fs_gd_get_b_bitmap(struct ext2_gd *gd) { return (((uint64_t)(le32toh(gd->ext4bgd_b_bitmap_hi)) << 32) | le32toh(gd->ext2bgd_b_bitmap)); } uint64_t e2fs_gd_get_i_bitmap(struct ext2_gd *gd) { return (((uint64_t)(le32toh(gd->ext4bgd_i_bitmap_hi)) << 32) | le32toh(gd->ext2bgd_i_bitmap)); } uint64_t e2fs_gd_get_i_tables(struct ext2_gd *gd) { return (((uint64_t)(le32toh(gd->ext4bgd_i_tables_hi)) << 32) | le32toh(gd->ext2bgd_i_tables)); } static uint32_t e2fs_gd_get_nbfree(struct ext2_gd *gd) { return (((uint32_t)(le16toh(gd->ext4bgd_nbfree_hi)) << 16) | le16toh(gd->ext2bgd_nbfree)); } static void e2fs_gd_set_nbfree(struct ext2_gd *gd, uint32_t val) { gd->ext2bgd_nbfree = htole16(val & 0xffff); gd->ext4bgd_nbfree_hi = htole16(val >> 16); } static uint32_t e2fs_gd_get_nifree(struct ext2_gd *gd) { return (((uint32_t)(le16toh(gd->ext4bgd_nifree_hi)) << 16) | le16toh(gd->ext2bgd_nifree)); } static void e2fs_gd_set_nifree(struct ext2_gd *gd, uint32_t val) { gd->ext2bgd_nifree = htole16(val & 0xffff); gd->ext4bgd_nifree_hi = htole16(val >> 16); } uint32_t e2fs_gd_get_ndirs(struct ext2_gd *gd) { return (((uint32_t)(le16toh(gd->ext4bgd_ndirs_hi)) << 16) | le16toh(gd->ext2bgd_ndirs)); } static void e2fs_gd_set_ndirs(struct ext2_gd *gd, uint32_t val) { gd->ext2bgd_ndirs = htole16(val & 0xffff); gd->ext4bgd_ndirs_hi = htole16(val >> 16); } static uint32_t e2fs_gd_get_i_unused(struct ext2_gd *gd) { return ((uint32_t)(le16toh(gd->ext4bgd_i_unused_hi) << 16) | le16toh(gd->ext4bgd_i_unused)); } static void e2fs_gd_set_i_unused(struct ext2_gd *gd, uint32_t val) { gd->ext4bgd_i_unused = htole16(val & 0xffff); gd->ext4bgd_i_unused_hi = htole16(val >> 16); } /* * Find a cylinder to place a directory. * * The policy implemented by this algorithm is to allocate a * directory inode in the same cylinder group as its parent * directory, but also to reserve space for its files inodes * and data. Restrict the number of directories which may be * allocated one after another in the same cylinder group * without intervening allocation of files. * * If we allocate a first level directory then force allocation * in another cylinder group. * */ static u_long ext2_dirpref(struct inode *pip) { struct m_ext2fs *fs; int cg, prefcg, cgsize; uint64_t avgbfree, minbfree; u_int avgifree, avgndir, curdirsize; u_int minifree, maxndir; u_int mincg, minndir; u_int dirsize, maxcontigdirs; mtx_assert(EXT2_MTX(pip->i_ump), MA_OWNED); fs = pip->i_e2fs; avgifree = fs->e2fs_ficount / fs->e2fs_gcount; avgbfree = fs->e2fs_fbcount / fs->e2fs_gcount; avgndir = fs->e2fs_total_dir / fs->e2fs_gcount; /* * Force allocation in another cg if creating a first level dir. */ ASSERT_VOP_LOCKED(ITOV(pip), "ext2fs_dirpref"); if (ITOV(pip)->v_vflag & VV_ROOT) { prefcg = arc4random() % fs->e2fs_gcount; mincg = prefcg; minndir = fs->e2fs_ipg; for (cg = prefcg; cg < fs->e2fs_gcount; cg++) if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir && e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree && e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) { mincg = cg; minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]); } for (cg = 0; cg < prefcg; cg++) if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < minndir && e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree && e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= avgbfree) { mincg = cg; minndir = e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]); } return (mincg); } /* * Count various limits which used for * optimal allocation of a directory inode. */ maxndir = min(avgndir + fs->e2fs_ipg / 16, fs->e2fs_ipg); minifree = avgifree - avgifree / 4; if (minifree < 1) minifree = 1; minbfree = avgbfree - avgbfree / 4; if (minbfree < 1) minbfree = 1; cgsize = fs->e2fs_fsize * fs->e2fs_fpg; dirsize = AVGDIRSIZE; curdirsize = avgndir ? (cgsize - avgbfree * fs->e2fs_bsize) / avgndir : 0; if (dirsize < curdirsize) dirsize = curdirsize; maxcontigdirs = min((avgbfree * fs->e2fs_bsize) / dirsize, 255); maxcontigdirs = min(maxcontigdirs, fs->e2fs_ipg / AFPDIR); if (maxcontigdirs == 0) maxcontigdirs = 1; /* * Limit number of dirs in one cg and reserve space for * regular files, but only if we have no deficit in * inodes or space. */ prefcg = ino_to_cg(fs, pip->i_number); for (cg = prefcg; cg < fs->e2fs_gcount; cg++) if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir && e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree && e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) { if (fs->e2fs_contigdirs[cg] < maxcontigdirs) return (cg); } for (cg = 0; cg < prefcg; cg++) if (e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) < maxndir && e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= minifree && e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) >= minbfree) { if (fs->e2fs_contigdirs[cg] < maxcontigdirs) return (cg); } /* * This is a backstop when we have deficit in space. */ for (cg = prefcg; cg < fs->e2fs_gcount; cg++) if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree) return (cg); for (cg = 0; cg < prefcg; cg++) if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) >= avgifree) break; return (cg); } /* * Select the desired position for the next block in a file. * * we try to mimic what Remy does in inode_getblk/block_getblk * * we note: blocknr == 0 means that we're about to allocate either * a direct block or a pointer block at the first level of indirection * (In other words, stuff that will go in i_db[] or i_ib[]) * * blocknr != 0 means that we're allocating a block that is none * of the above. Then, blocknr tells us the number of the block * that will hold the pointer */ e4fs_daddr_t ext2_blkpref(struct inode *ip, e2fs_lbn_t lbn, int indx, e2fs_daddr_t *bap, e2fs_daddr_t blocknr) { struct m_ext2fs *fs; int tmp; fs = ip->i_e2fs; mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED); /* * If the next block is actually what we thought it is, then set the * goal to what we thought it should be. */ if (ip->i_next_alloc_block == lbn && ip->i_next_alloc_goal != 0) return ip->i_next_alloc_goal; /* * Now check whether we were provided with an array that basically * tells us previous blocks to which we want to stay close. */ if (bap) for (tmp = indx - 1; tmp >= 0; tmp--) if (bap[tmp]) return (le32toh(bap[tmp])); /* * Else lets fall back to the blocknr or, if there is none, follow * the rule that a block should be allocated near its inode. */ return (blocknr ? blocknr : (e2fs_daddr_t)(ip->i_block_group * EXT2_BLOCKS_PER_GROUP(fs)) + le32toh(fs->e2fs->e2fs_first_dblock)); } /* * Implement the cylinder overflow algorithm. * * The policy implemented by this algorithm is: * 1) allocate the block in its requested cylinder group. - * 2) quadradically rehash on the cylinder group number. + * 2) quadratically rehash on the cylinder group number. * 3) brute force search for a free block. */ static e4fs_daddr_t ext2_hashalloc(struct inode *ip, int cg, long pref, int size, daddr_t (*allocator) (struct inode *, int, daddr_t, int)) { struct m_ext2fs *fs; e4fs_daddr_t result; int i, icg = cg; mtx_assert(EXT2_MTX(ip->i_ump), MA_OWNED); fs = ip->i_e2fs; /* * 1: preferred cylinder group */ result = (*allocator)(ip, cg, pref, size); if (result) return (result); /* * 2: quadratic rehash */ for (i = 1; i < fs->e2fs_gcount; i *= 2) { cg += i; if (cg >= fs->e2fs_gcount) cg -= fs->e2fs_gcount; result = (*allocator)(ip, cg, 0, size); if (result) return (result); } /* * 3: brute force search * Note that we start at i == 2, since 0 was checked initially, * and 1 is always checked in the quadratic rehash. */ cg = (icg + 2) % fs->e2fs_gcount; for (i = 2; i < fs->e2fs_gcount; i++) { result = (*allocator)(ip, cg, 0, size); if (result) return (result); cg++; if (cg == fs->e2fs_gcount) cg = 0; } return (0); } static uint64_t ext2_cg_number_gdb_nometa(struct m_ext2fs *fs, int cg) { if (!ext2_cg_has_sb(fs, cg)) return (0); if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG)) return (le32toh(fs->e2fs->e3fs_first_meta_bg)); return ((fs->e2fs_gcount + EXT2_DESCS_PER_BLOCK(fs) - 1) / EXT2_DESCS_PER_BLOCK(fs)); } static uint64_t ext2_cg_number_gdb_meta(struct m_ext2fs *fs, int cg) { unsigned long metagroup; int first, last; metagroup = cg / EXT2_DESCS_PER_BLOCK(fs); first = metagroup * EXT2_DESCS_PER_BLOCK(fs); last = first + EXT2_DESCS_PER_BLOCK(fs) - 1; if (cg == first || cg == first + 1 || cg == last) return (1); return (0); } uint64_t ext2_cg_number_gdb(struct m_ext2fs *fs, int cg) { unsigned long first_meta_bg, metagroup; first_meta_bg = le32toh(fs->e2fs->e3fs_first_meta_bg); metagroup = cg / EXT2_DESCS_PER_BLOCK(fs); if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) || metagroup < first_meta_bg) return (ext2_cg_number_gdb_nometa(fs, cg)); return ext2_cg_number_gdb_meta(fs, cg); } static int ext2_number_base_meta_blocks(struct m_ext2fs *fs, int cg) { int number; number = ext2_cg_has_sb(fs, cg); if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) || cg < le32toh(fs->e2fs->e3fs_first_meta_bg) * EXT2_DESCS_PER_BLOCK(fs)) { if (number) { number += ext2_cg_number_gdb(fs, cg); number += le16toh(fs->e2fs->e2fs_reserved_ngdb); } } else { number += ext2_cg_number_gdb(fs, cg); } return (number); } static void ext2_mark_bitmap_end(int start_bit, int end_bit, char *bitmap) { int i; if (start_bit >= end_bit) return; for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++) setbit(bitmap, i); if (i < end_bit) memset(bitmap + (i >> 3), 0xff, (end_bit - i) >> 3); } static int ext2_get_group_number(struct m_ext2fs *fs, e4fs_daddr_t block) { return ((block - le32toh(fs->e2fs->e2fs_first_dblock)) / fs->e2fs_bsize); } static int ext2_block_in_group(struct m_ext2fs *fs, e4fs_daddr_t block, int cg) { return ((ext2_get_group_number(fs, block) == cg) ? 1 : 0); } static int ext2_cg_block_bitmap_init(struct m_ext2fs *fs, int cg, struct buf *bp) { int bit, bit_max, inodes_per_block; uint64_t start, tmp; if (!(le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_BLOCK_UNINIT)) return (0); memset(bp->b_data, 0, fs->e2fs_bsize); bit_max = ext2_number_base_meta_blocks(fs, cg); if ((bit_max >> 3) >= fs->e2fs_bsize) return (EINVAL); for (bit = 0; bit < bit_max; bit++) setbit(bp->b_data, bit); start = (uint64_t)cg * fs->e2fs_bpg + le32toh(fs->e2fs->e2fs_first_dblock); /* Set bits for block and inode bitmaps, and inode table. */ tmp = e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg]); if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) || ext2_block_in_group(fs, tmp, cg)) setbit(bp->b_data, tmp - start); tmp = e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg]); if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) || ext2_block_in_group(fs, tmp, cg)) setbit(bp->b_data, tmp - start); tmp = e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]); inodes_per_block = fs->e2fs_bsize/EXT2_INODE_SIZE(fs); while( tmp < e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) + fs->e2fs_ipg / inodes_per_block ) { if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) || ext2_block_in_group(fs, tmp, cg)) setbit(bp->b_data, tmp - start); tmp++; } /* * Also if the number of blocks within the group is less than * the blocksize * 8 ( which is the size of bitmap ), set rest * of the block bitmap to 1 */ ext2_mark_bitmap_end(fs->e2fs_bpg, fs->e2fs_bsize * 8, bp->b_data); /* Clean the flag */ fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh( fs->e2fs_gd[cg].ext4bgd_flags) & ~EXT2_BG_BLOCK_UNINIT); return (0); } static int ext2_b_bitmap_validate(struct m_ext2fs *fs, struct buf *bp, int cg) { struct ext2_gd *gd; uint64_t group_first_block; unsigned int offset, max_bit; if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG)) { /* * It is not possible to check block bitmap in case of this * feature, because the inode and block bitmaps and inode table * blocks may not be in the group at all. * So, skip check in this case. */ return (0); } gd = &fs->e2fs_gd[cg]; max_bit = fs->e2fs_fpg; group_first_block = ((uint64_t)cg) * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock); /* Check block bitmap block number */ offset = e2fs_gd_get_b_bitmap(gd) - group_first_block; if (offset >= max_bit || !isset(bp->b_data, offset)) { SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "bad block bitmap, group", cg); return (EINVAL); } /* Check inode bitmap block number */ offset = e2fs_gd_get_i_bitmap(gd) - group_first_block; if (offset >= max_bit || !isset(bp->b_data, offset)) { SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "bad inode bitmap", cg); return (EINVAL); } /* Check inode table */ offset = e2fs_gd_get_i_tables(gd) - group_first_block; if (offset >= max_bit || offset + fs->e2fs_itpg >= max_bit) { SDT_PROBE2(ext2fs, , alloc, ext2_b_bitmap_validate_error, "bad inode table, group", cg); return (EINVAL); } return (0); } /* * Determine whether a block can be allocated. * * Check to see if a block of the appropriate size is available, * and if it is, allocate it. */ static daddr_t ext2_alloccg(struct inode *ip, int cg, daddr_t bpref, int size) { struct m_ext2fs *fs; struct buf *bp; struct ext2mount *ump; daddr_t bno, runstart, runlen; int bit, loc, end, error, start; char *bbp; /* XXX ondisk32 */ fs = ip->i_e2fs; ump = ip->i_ump; if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0) return (0); EXT2_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])), (int)fs->e2fs_bsize, NOCRED, &bp); if (error) goto fail; if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) || EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) { error = ext2_cg_block_bitmap_init(fs, cg, bp); if (error) goto fail; ext2_gd_b_bitmap_csum_set(fs, cg, bp); } error = ext2_gd_b_bitmap_csum_verify(fs, cg, bp); if (error) goto fail; error = ext2_b_bitmap_validate(fs,bp, cg); if (error) goto fail; /* * Check, that another thread did not not allocate the last block in * this group while we were waiting for the buffer. */ if (e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) == 0) goto fail; bbp = (char *)bp->b_data; if (dtog(fs, bpref) != cg) bpref = 0; if (bpref != 0) { bpref = dtogd(fs, bpref); /* * if the requested block is available, use it */ if (isclr(bbp, bpref)) { bno = bpref; goto gotit; } } /* * no blocks in the requested cylinder, so take next * available one in this cylinder group. * first try to get 8 contigous blocks, then fall back to a single * block. */ if (bpref) start = dtogd(fs, bpref) / NBBY; else start = 0; end = howmany(fs->e2fs_fpg, NBBY) - start; retry: runlen = 0; runstart = 0; for (loc = start; loc < end; loc++) { if (bbp[loc] == (char)0xff) { runlen = 0; continue; } /* Start of a run, find the number of high clear bits. */ if (runlen == 0) { bit = fls(bbp[loc]); runlen = NBBY - bit; runstart = loc * NBBY + bit; } else if (bbp[loc] == 0) { /* Continue a run. */ runlen += NBBY; } else { /* * Finish the current run. If it isn't long * enough, start a new one. */ bit = ffs(bbp[loc]) - 1; runlen += bit; if (runlen >= 8) { bno = runstart; goto gotit; } /* Run was too short, start a new one. */ bit = fls(bbp[loc]); runlen = NBBY - bit; runstart = loc * NBBY + bit; } /* If the current run is long enough, use it. */ if (runlen >= 8) { bno = runstart; goto gotit; } } if (start != 0) { end = start; start = 0; goto retry; } bno = ext2_mapsearch(fs, bbp, bpref); if (bno < 0) goto fail; gotit: #ifdef INVARIANTS if (isset(bbp, bno)) { printf("ext2fs_alloccgblk: cg=%d bno=%jd fs=%s\n", cg, (intmax_t)bno, fs->e2fs_fsmnt); panic("ext2fs_alloccg: dup alloc"); } #endif setbit(bbp, bno); EXT2_LOCK(ump); ext2_clusteracct(fs, bbp, cg, bno, -1); fs->e2fs_fbcount--; e2fs_gd_set_nbfree(&fs->e2fs_gd[cg], e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1); fs->e2fs_fmod = 1; EXT2_UNLOCK(ump); ext2_gd_b_bitmap_csum_set(fs, cg, bp); bdwrite(bp); return (((uint64_t)cg) * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock) + bno); fail: brelse(bp); EXT2_LOCK(ump); return (0); } /* * Determine whether a cluster can be allocated. */ static daddr_t ext2_clusteralloc(struct inode *ip, int cg, daddr_t bpref, int len) { struct m_ext2fs *fs; struct ext2mount *ump; struct buf *bp; char *bbp; int bit, error, got, i, loc, run; int32_t *lp; daddr_t bno; fs = ip->i_e2fs; ump = ip->i_ump; if (fs->e2fs_maxcluster[cg] < len) return (0); EXT2_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])), (int)fs->e2fs_bsize, NOCRED, &bp); if (error) goto fail_lock; bbp = (char *)bp->b_data; EXT2_LOCK(ump); /* * Check to see if a cluster of the needed size (or bigger) is * available in this cylinder group. */ lp = &fs->e2fs_clustersum[cg].cs_sum[len]; for (i = len; i <= fs->e2fs_contigsumsize; i++) if (*lp++ > 0) break; if (i > fs->e2fs_contigsumsize) { /* * Update the cluster summary information to reflect * the true maximum-sized cluster so that future cluster * allocation requests can avoid reading the bitmap only * to find no cluster. */ lp = &fs->e2fs_clustersum[cg].cs_sum[len - 1]; for (i = len - 1; i > 0; i--) if (*lp-- > 0) break; fs->e2fs_maxcluster[cg] = i; goto fail; } EXT2_UNLOCK(ump); /* Search the bitmap to find a big enough cluster like in FFS. */ if (dtog(fs, bpref) != cg) bpref = 0; if (bpref != 0) bpref = dtogd(fs, bpref); loc = bpref / NBBY; bit = 1 << (bpref % NBBY); for (run = 0, got = bpref; got < fs->e2fs_fpg; got++) { if ((bbp[loc] & bit) != 0) run = 0; else { run++; if (run == len) break; } if ((got & (NBBY - 1)) != (NBBY - 1)) bit <<= 1; else { loc++; bit = 1; } } if (got >= fs->e2fs_fpg) goto fail_lock; /* Allocate the cluster that we found. */ for (i = 1; i < len; i++) if (!isclr(bbp, got - run + i)) panic("ext2_clusteralloc: map mismatch"); bno = got - run + 1; if (bno >= fs->e2fs_fpg) panic("ext2_clusteralloc: allocated out of group"); EXT2_LOCK(ump); for (i = 0; i < len; i += fs->e2fs_fpb) { setbit(bbp, bno + i); ext2_clusteracct(fs, bbp, cg, bno + i, -1); fs->e2fs_fbcount--; e2fs_gd_set_nbfree(&fs->e2fs_gd[cg], e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) - 1); } fs->e2fs_fmod = 1; EXT2_UNLOCK(ump); bdwrite(bp); return (cg * fs->e2fs_fpg + le32toh(fs->e2fs->e2fs_first_dblock) + bno); fail_lock: EXT2_LOCK(ump); fail: brelse(bp); return (0); } static int ext2_zero_inode_table(struct inode *ip, int cg) { struct m_ext2fs *fs; struct buf *bp; int i, all_blks, used_blks; fs = ip->i_e2fs; if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_INODE_ZEROED) return (0); all_blks = le16toh(fs->e2fs->e2fs_inode_size) * fs->e2fs_ipg / fs->e2fs_bsize; used_blks = howmany(fs->e2fs_ipg - e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]), fs->e2fs_bsize / EXT2_INODE_SIZE(fs)); for (i = 0; i < all_blks - used_blks; i++) { bp = getblk(ip->i_devvp, fsbtodb(fs, e2fs_gd_get_i_tables(&fs->e2fs_gd[cg]) + used_blks + i), fs->e2fs_bsize, 0, 0, 0); if (!bp) return (EIO); vfs_bio_bzero_buf(bp, 0, fs->e2fs_bsize); bawrite(bp); } fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh( fs->e2fs_gd[cg].ext4bgd_flags) | EXT2_BG_INODE_ZEROED); return (0); } static void ext2_fix_bitmap_tail(unsigned char *bitmap, int first, int last) { int i; for (i = first; i <= last; i++) bitmap[i] = 0xff; } /* * Determine whether an inode can be allocated. * * Check to see if an inode is available, and if it is, * allocate it using tode in the specified cylinder group. */ static daddr_t ext2_nodealloccg(struct inode *ip, int cg, daddr_t ipref, int mode) { struct m_ext2fs *fs; struct buf *bp; struct ext2mount *ump; int error, start, len, ifree, ibytes; char *ibp, *loc; ipref--; /* to avoid a lot of (ipref -1) */ if (ipref == -1) ipref = 0; fs = ip->i_e2fs; ump = ip->i_ump; if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0) return (0); EXT2_UNLOCK(ump); error = bread(ip->i_devvp, fsbtodb(fs, e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])), (int)fs->e2fs_bsize, NOCRED, &bp); if (error) { EXT2_LOCK(ump); return (0); } if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) || EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) { if (le16toh(fs->e2fs_gd[cg].ext4bgd_flags) & EXT2_BG_INODE_UNINIT) { ibytes = fs->e2fs_ipg / 8; memset(bp->b_data, 0, ibytes - 1); ext2_fix_bitmap_tail(bp->b_data, ibytes, fs->e2fs_bsize - 1); fs->e2fs_gd[cg].ext4bgd_flags = htole16(le16toh( fs->e2fs_gd[cg].ext4bgd_flags) & ~EXT2_BG_INODE_UNINIT); } ext2_gd_i_bitmap_csum_set(fs, cg, bp); error = ext2_zero_inode_table(ip, cg); if (error) { brelse(bp); EXT2_LOCK(ump); return (0); } } error = ext2_gd_i_bitmap_csum_verify(fs, cg, bp); if (error) { brelse(bp); EXT2_LOCK(ump); return (0); } if (e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) == 0) { /* * Another thread allocated the last i-node in this * group while we were waiting for the buffer. */ brelse(bp); EXT2_LOCK(ump); return (0); } ibp = (char *)bp->b_data; if (ipref) { ipref %= fs->e2fs_ipg; if (isclr(ibp, ipref)) goto gotit; } start = ipref / NBBY; len = howmany(fs->e2fs_ipg - ipref, NBBY); loc = memcchr(&ibp[start], 0xff, len); if (loc == NULL) { len = start + 1; start = 0; loc = memcchr(&ibp[start], 0xff, len); if (loc == NULL) { SDT_PROBE3(ext2fs, , alloc, ext2_nodealloccg_bmap_corrupted, cg, ipref, fs->e2fs_fsmnt); brelse(bp); EXT2_LOCK(ump); return (0); } } ipref = (loc - ibp) * NBBY + ffs(~*loc) - 1; gotit: setbit(ibp, ipref); EXT2_LOCK(ump); e2fs_gd_set_nifree(&fs->e2fs_gd[cg], e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) - 1); if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) || EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) { ifree = fs->e2fs_ipg - e2fs_gd_get_i_unused(&fs->e2fs_gd[cg]); if (ipref + 1 > ifree) e2fs_gd_set_i_unused(&fs->e2fs_gd[cg], fs->e2fs_ipg - (ipref + 1)); } fs->e2fs_ficount--; fs->e2fs_fmod = 1; if ((mode & IFMT) == IFDIR) { e2fs_gd_set_ndirs(&fs->e2fs_gd[cg], e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) + 1); fs->e2fs_total_dir++; } EXT2_UNLOCK(ump); ext2_gd_i_bitmap_csum_set(fs, cg, bp); bdwrite(bp); return ((uint64_t)cg * fs->e2fs_ipg + ipref + 1); } /* * Free a block or fragment. * */ void ext2_blkfree(struct inode *ip, e4fs_daddr_t bno, long size) { struct m_ext2fs *fs; struct buf *bp; struct ext2mount *ump; int cg, error; char *bbp; fs = ip->i_e2fs; ump = ip->i_ump; cg = dtog(fs, bno); if (bno >= fs->e2fs_bcount) { SDT_PROBE2(ext2fs, , alloc, ext2_blkfree_bad_block, ip->i_number, bno); return; } error = bread(ip->i_devvp, fsbtodb(fs, e2fs_gd_get_b_bitmap(&fs->e2fs_gd[cg])), (int)fs->e2fs_bsize, NOCRED, &bp); if (error) { return; } bbp = (char *)bp->b_data; bno = dtogd(fs, bno); if (isclr(bbp, bno)) { panic("ext2_blkfree: freeing free block %lld, fs=%s", (long long)bno, fs->e2fs_fsmnt); } clrbit(bbp, bno); EXT2_LOCK(ump); ext2_clusteracct(fs, bbp, cg, bno, 1); fs->e2fs_fbcount++; e2fs_gd_set_nbfree(&fs->e2fs_gd[cg], e2fs_gd_get_nbfree(&fs->e2fs_gd[cg]) + 1); fs->e2fs_fmod = 1; EXT2_UNLOCK(ump); ext2_gd_b_bitmap_csum_set(fs, cg, bp); bdwrite(bp); } /* * Free an inode. * */ int ext2_vfree(struct vnode *pvp, ino_t ino, int mode) { struct m_ext2fs *fs; struct inode *pip; struct buf *bp; struct ext2mount *ump; int error, cg; char *ibp; pip = VTOI(pvp); fs = pip->i_e2fs; ump = pip->i_ump; if ((u_int)ino > fs->e2fs_ipg * fs->e2fs_gcount) panic("ext2_vfree: range: devvp = %p, ino = %ju, fs = %s", pip->i_devvp, (uintmax_t)ino, fs->e2fs_fsmnt); cg = ino_to_cg(fs, ino); error = bread(pip->i_devvp, fsbtodb(fs, e2fs_gd_get_i_bitmap(&fs->e2fs_gd[cg])), (int)fs->e2fs_bsize, NOCRED, &bp); if (error) { return (0); } ibp = (char *)bp->b_data; ino = (ino - 1) % fs->e2fs_ipg; if (isclr(ibp, ino)) { SDT_PROBE2(ext2fs, , alloc, ext2_vfree_doublefree, fs->e2fs_fsmnt, ino); if (fs->e2fs_ronly == 0) panic("ext2_vfree: freeing free inode"); } clrbit(ibp, ino); EXT2_LOCK(ump); fs->e2fs_ficount++; e2fs_gd_set_nifree(&fs->e2fs_gd[cg], e2fs_gd_get_nifree(&fs->e2fs_gd[cg]) + 1); if ((mode & IFMT) == IFDIR) { e2fs_gd_set_ndirs(&fs->e2fs_gd[cg], e2fs_gd_get_ndirs(&fs->e2fs_gd[cg]) - 1); fs->e2fs_total_dir--; } fs->e2fs_fmod = 1; EXT2_UNLOCK(ump); ext2_gd_i_bitmap_csum_set(fs, cg, bp); bdwrite(bp); return (0); } /* * Find a block in the specified cylinder group. * * It is a panic if a request is made to find a block if none are * available. */ static daddr_t ext2_mapsearch(struct m_ext2fs *fs, char *bbp, daddr_t bpref) { char *loc; int start, len; /* * find the fragment by searching through the free block * map for an appropriate bit pattern */ if (bpref) start = dtogd(fs, bpref) / NBBY; else start = 0; len = howmany(fs->e2fs_fpg, NBBY) - start; loc = memcchr(&bbp[start], 0xff, len); if (loc == NULL) { len = start + 1; start = 0; loc = memcchr(&bbp[start], 0xff, len); if (loc == NULL) { panic("ext2_mapsearch: map corrupted: start=%d, len=%d," "fs=%s", start, len, fs->e2fs_fsmnt); /* NOTREACHED */ } } return ((loc - bbp) * NBBY + ffs(~*loc) - 1); } int ext2_cg_has_sb(struct m_ext2fs *fs, int cg) { int a3, a5, a7; if (cg == 0) return (1); if (EXT2_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) { if (cg == le32toh(fs->e2fs->e4fs_backup_bgs[0]) || cg == le32toh(fs->e2fs->e4fs_backup_bgs[1])) return (1); return (0); } if ((cg <= 1) || !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER)) return (1); if (!(cg & 1)) return (0); for (a3 = 3, a5 = 5, a7 = 7; a3 <= cg || a5 <= cg || a7 <= cg; a3 *= 3, a5 *= 5, a7 *= 7) if (cg == a3 || cg == a5 || cg == a7) return (1); return (0); }