freebsd-dev/sys/kern/vfs_syscalls.c

4765 lines
99 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +00:00
* Copyright (c) 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_syscalls.c 8.13 (Berkeley) 4/15/94
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_capsicum.h"
#include "opt_compat.h"
#include "opt_ktrace.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/capsicum.h>
#include <sys/disk.h>
#include <sys/sysent.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/sysproto.h>
1994-05-24 10:09:53 +00:00
#include <sys/namei.h>
#include <sys/filedesc.h>
#include <sys/kernel.h>
#include <sys/fcntl.h>
1994-05-24 10:09:53 +00:00
#include <sys/file.h>
#include <sys/filio.h>
#include <sys/limits.h>
#include <sys/linker.h>
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
#include <sys/rwlock.h>
#include <sys/sdt.h>
1994-05-24 10:09:53 +00:00
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/unistd.h>
1994-05-24 10:09:53 +00:00
#include <sys/vnode.h>
#include <sys/priv.h>
1994-05-24 10:09:53 +00:00
#include <sys/proc.h>
#include <sys/dirent.h>
#include <sys/jail.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
1994-05-24 10:09:53 +00:00
#include <machine/stdarg.h>
#include <security/audit/audit.h>
#include <security/mac/mac_framework.h>
1994-05-24 10:09:53 +00:00
#include <vm/vm.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/uma.h>
1994-05-24 10:09:53 +00:00
#include <ufs/ufs/quota.h>
MALLOC_DEFINE(M_FADVISE, "fadvise", "posix_fadvise(2) information");
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
SDT_PROVIDER_DEFINE(vfs);
SDT_PROBE_DEFINE2(vfs, , stat, mode, "char *", "int");
SDT_PROBE_DEFINE2(vfs, , stat, reg, "char *", "int");
2002-03-19 21:25:46 +00:00
static int chroot_refuse_vdir_fds(struct filedesc *fdp);
static int kern_chflagsat(struct thread *td, int fd, const char *path,
enum uio_seg pathseg, u_long flags, int atflag);
static int setfflags(struct thread *td, struct vnode *, u_long);
static int getutimes(const struct timeval *, enum uio_seg, struct timespec *);
static int getutimens(const struct timespec *, enum uio_seg,
struct timespec *, int *);
2002-03-19 21:25:46 +00:00
static int setutimes(struct thread *td, struct vnode *,
const struct timespec *, int, int);
2002-03-19 21:25:46 +00:00
static int vn_access(struct vnode *vp, int user_flags, struct ucred *cred,
struct thread *td);
/*
* The module initialization routine for POSIX asynchronous I/O will
* set this to the version of AIO that it implements. (Zero means
* that it is not implemented.) This value is used here by pathconf()
* and in kern_descrip.c by fpathconf().
*/
int async_io_version;
1994-05-24 10:09:53 +00:00
/*
* Sync each mounted filesystem.
*/
#ifndef _SYS_SYSPROTO_H_
struct sync_args {
int dummy;
};
#endif
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
sys_sync(td, uap)
struct thread *td;
1994-05-24 10:09:53 +00:00
struct sync_args *uap;
{
struct mount *mp, *nmp;
int save;
1994-05-24 10:09:53 +00:00
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
if ((mp->mnt_flag & MNT_RDONLY) == 0 &&
vn_start_write(NULL, &mp, V_NOWAIT) == 0) {
save = curthread_pflags_set(TDP_SYNCIO);
vfs_msync(mp, MNT_NOWAIT);
VFS_SYNC(mp, MNT_NOWAIT);
curthread_pflags_restore(save);
vn_finished_write(mp);
1994-05-24 10:09:53 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
nmp = TAILQ_NEXT(mp, mnt_list);
vfs_unbusy(mp);
1994-05-24 10:09:53 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Change filesystem quotas.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct quotactl_args {
char *path;
int cmd;
int uid;
caddr_t arg;
};
#endif
int
sys_quotactl(td, uap)
struct thread *td;
register struct quotactl_args /* {
char *path;
int cmd;
int uid;
caddr_t arg;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_CMD(uap->cmd);
AUDIT_ARG_UID(uap->uid);
if (!prison_allow(td->td_ucred, PR_ALLOW_QUOTAS))
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
return (EPERM);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
uap->path, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
mp = nd.ni_vp->v_mount;
vfs_ref(mp);
vput(nd.ni_vp);
error = vfs_busy(mp, 0);
vfs_rel(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
error = VFS_QUOTACTL(mp, uap->cmd, uap->uid, uap->arg);
/*
* Since quota on operation typically needs to open quota
* file, the Q_QUOTAON handler needs to unbusy the mount point
* before calling into namei. Otherwise, unmount might be
* started between two vfs_busy() invocations (first is our,
* second is from mount point cross-walk code in lookup()),
* causing deadlock.
*
* Require that Q_QUOTAON handles the vfs_busy() reference on
* its own, always returning with ubusied mount point.
*/
if ((uap->cmd >> SUBCMDSHIFT) != Q_QUOTAON)
vfs_unbusy(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Used by statfs conversion routines to scale the block size up if
* necessary so that all of the block counts are <= 'max_size'. Note
* that 'max_size' should be a bitmask, i.e. 2^n - 1 for some non-zero
* value of 'n'.
*/
void
statfs_scale_blocks(struct statfs *sf, long max_size)
{
uint64_t count;
int shift;
KASSERT(powerof2(max_size + 1), ("%s: invalid max_size", __func__));
/*
* Attempt to scale the block counts to give a more accurate
* overview to userland of the ratio of free space to used
* space. To do this, find the largest block count and compute
* a divisor that lets it fit into a signed integer <= max_size.
*/
if (sf->f_bavail < 0)
count = -sf->f_bavail;
else
count = sf->f_bavail;
count = MAX(sf->f_blocks, MAX(sf->f_bfree, count));
if (count <= max_size)
return;
count >>= flsl(max_size);
shift = 0;
while (count > 0) {
shift++;
count >>=1;
}
sf->f_bsize <<= shift;
sf->f_blocks >>= shift;
sf->f_bfree >>= shift;
sf->f_bavail >>= shift;
}
1994-05-24 10:09:53 +00:00
/*
* Get filesystem statistics.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct statfs_args {
char *path;
struct statfs *buf;
};
#endif
int
sys_statfs(td, uap)
struct thread *td;
register struct statfs_args /* {
char *path;
struct statfs *buf;
} */ *uap;
{
struct statfs sf;
int error;
error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
if (error == 0)
error = copyout(&sf, uap->buf, sizeof(sf));
return (error);
}
int
kern_statfs(struct thread *td, char *path, enum uio_seg pathseg,
struct statfs *buf)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
struct statfs *sp, sb;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
pathseg, path, td);
error = namei(&nd);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
mp = nd.ni_vp->v_mount;
vfs_ref(mp);
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_vp);
error = vfs_busy(mp, 0);
vfs_rel(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
#ifdef MAC
error = mac_mount_check_stat(td->td_ucred, mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
/*
* Set these in case the underlying filesystem fails to do so.
*/
sp = &mp->mnt_stat;
sp->f_version = STATFS_VERSION;
sp->f_namemax = NAME_MAX;
sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
error = VFS_STATFS(mp, sp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
if (priv_check(td, PRIV_VFS_GENERATION)) {
2002-06-28 23:02:38 +00:00
bcopy(sp, &sb, sizeof(sb));
sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
prison_enforce_statfs(td->td_ucred, mp, &sb);
sp = &sb;
}
*buf = *sp;
out:
vfs_unbusy(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Get filesystem statistics.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fstatfs_args {
int fd;
struct statfs *buf;
};
#endif
int
sys_fstatfs(td, uap)
struct thread *td;
register struct fstatfs_args /* {
int fd;
struct statfs *buf;
} */ *uap;
{
struct statfs sf;
int error;
error = kern_fstatfs(td, uap->fd, &sf);
if (error == 0)
error = copyout(&sf, uap->buf, sizeof(sf));
return (error);
}
int
kern_fstatfs(struct thread *td, int fd, struct statfs *buf)
1994-05-24 10:09:53 +00:00
{
struct file *fp;
struct mount *mp;
struct statfs *sp, sb;
struct vnode *vp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(fd);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, fd,
cap_rights_init(&rights, CAP_FSTATFS), &fp);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = fp->f_vnode;
vn_lock(vp, LK_SHARED | LK_RETRY);
#ifdef AUDIT
AUDIT_ARG_VNODE1(vp);
#endif
mp = vp->v_mount;
if (mp)
vfs_ref(mp);
VOP_UNLOCK(vp, 0);
fdrop(fp, td);
if (mp == NULL) {
error = EBADF;
goto out;
}
error = vfs_busy(mp, 0);
vfs_rel(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
#ifdef MAC
error = mac_mount_check_stat(td->td_ucred, mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
/*
* Set these in case the underlying filesystem fails to do so.
*/
sp = &mp->mnt_stat;
sp->f_version = STATFS_VERSION;
sp->f_namemax = NAME_MAX;
sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
error = VFS_STATFS(mp, sp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
if (priv_check(td, PRIV_VFS_GENERATION)) {
2002-06-28 23:02:38 +00:00
bcopy(sp, &sb, sizeof(sb));
sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
prison_enforce_statfs(td->td_ucred, mp, &sb);
sp = &sb;
}
*buf = *sp;
out:
if (mp)
vfs_unbusy(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Get statistics on all filesystems.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct getfsstat_args {
struct statfs *buf;
long bufsize;
int flags;
};
#endif
int
sys_getfsstat(td, uap)
struct thread *td;
register struct getfsstat_args /* {
struct statfs *buf;
long bufsize;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
size_t count;
int error;
error = kern_getfsstat(td, &uap->buf, uap->bufsize, &count,
UIO_USERSPACE, uap->flags);
if (error == 0)
td->td_retval[0] = count;
return (error);
}
/*
* If (bufsize > 0 && bufseg == UIO_SYSSPACE)
2013-03-16 22:37:30 +00:00
* The caller is responsible for freeing memory which will be allocated
* in '*buf'.
*/
int
kern_getfsstat(struct thread *td, struct statfs **buf, size_t bufsize,
size_t *countp, enum uio_seg bufseg, int flags)
{
struct mount *mp, *nmp;
struct statfs *sfsp, *sp, sb;
size_t count, maxcount;
int error;
1994-05-24 10:09:53 +00:00
maxcount = bufsize / sizeof(struct statfs);
if (bufsize == 0)
sfsp = NULL;
else if (bufseg == UIO_USERSPACE)
sfsp = *buf;
else /* if (bufseg == UIO_SYSSPACE) */ {
count = 0;
mtx_lock(&mountlist_mtx);
TAILQ_FOREACH(mp, &mountlist, mnt_list) {
count++;
}
mtx_unlock(&mountlist_mtx);
if (maxcount > count)
maxcount = count;
sfsp = *buf = malloc(maxcount * sizeof(struct statfs), M_TEMP,
M_WAITOK);
}
count = 0;
mtx_lock(&mountlist_mtx);
for (mp = TAILQ_FIRST(&mountlist); mp != NULL; mp = nmp) {
if (prison_canseemount(td->td_ucred, mp) != 0) {
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
#ifdef MAC
if (mac_mount_check_stat(td->td_ucred, mp) != 0) {
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
#endif
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, MBF_NOWAIT | MBF_MNTLSTLOCK)) {
nmp = TAILQ_NEXT(mp, mnt_list);
continue;
}
if (sfsp && count < maxcount) {
1994-05-24 10:09:53 +00:00
sp = &mp->mnt_stat;
/*
* Set these in case the underlying filesystem
* fails to do so.
*/
sp->f_version = STATFS_VERSION;
sp->f_namemax = NAME_MAX;
sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
1994-05-24 10:09:53 +00:00
/*
* If MNT_NOWAIT or MNT_LAZY is specified, do not
* refresh the fsstat cache. MNT_NOWAIT or MNT_LAZY
* overrides MNT_WAIT.
1994-05-24 10:09:53 +00:00
*/
if (((flags & (MNT_LAZY|MNT_NOWAIT)) == 0 ||
(flags & MNT_WAIT)) &&
(error = VFS_STATFS(mp, sp))) {
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
nmp = TAILQ_NEXT(mp, mnt_list);
vfs_unbusy(mp);
1994-05-24 10:09:53 +00:00
continue;
}
if (priv_check(td, PRIV_VFS_GENERATION)) {
bcopy(sp, &sb, sizeof(sb));
sb.f_fsid.val[0] = sb.f_fsid.val[1] = 0;
prison_enforce_statfs(td->td_ucred, mp, &sb);
sp = &sb;
}
if (bufseg == UIO_SYSSPACE)
bcopy(sp, sfsp, sizeof(*sp));
else /* if (bufseg == UIO_USERSPACE) */ {
error = copyout(sp, sfsp, sizeof(*sp));
2013-09-05 00:19:30 +00:00
if (error != 0) {
vfs_unbusy(mp);
return (error);
}
}
sfsp++;
1994-05-24 10:09:53 +00:00
}
count++;
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_lock(&mountlist_mtx);
nmp = TAILQ_NEXT(mp, mnt_list);
vfs_unbusy(mp);
1994-05-24 10:09:53 +00:00
}
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
mtx_unlock(&mountlist_mtx);
1994-05-24 10:09:53 +00:00
if (sfsp && count > maxcount)
*countp = maxcount;
1994-05-24 10:09:53 +00:00
else
*countp = count;
1994-05-24 10:09:53 +00:00
return (0);
}
#ifdef COMPAT_FREEBSD4
/*
* Get old format filesystem statistics.
*/
static void cvtstatfs(struct statfs *, struct ostatfs *);
#ifndef _SYS_SYSPROTO_H_
struct freebsd4_statfs_args {
char *path;
struct ostatfs *buf;
};
#endif
int
freebsd4_statfs(td, uap)
struct thread *td;
struct freebsd4_statfs_args /* {
char *path;
struct ostatfs *buf;
} */ *uap;
{
struct ostatfs osb;
struct statfs sf;
int error;
error = kern_statfs(td, uap->path, UIO_USERSPACE, &sf);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
cvtstatfs(&sf, &osb);
return (copyout(&osb, uap->buf, sizeof(osb)));
}
/*
* Get filesystem statistics.
*/
#ifndef _SYS_SYSPROTO_H_
struct freebsd4_fstatfs_args {
int fd;
struct ostatfs *buf;
};
#endif
int
freebsd4_fstatfs(td, uap)
struct thread *td;
struct freebsd4_fstatfs_args /* {
int fd;
struct ostatfs *buf;
} */ *uap;
{
struct ostatfs osb;
struct statfs sf;
int error;
error = kern_fstatfs(td, uap->fd, &sf);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
cvtstatfs(&sf, &osb);
return (copyout(&osb, uap->buf, sizeof(osb)));
}
/*
* Get statistics on all filesystems.
*/
#ifndef _SYS_SYSPROTO_H_
struct freebsd4_getfsstat_args {
struct ostatfs *buf;
long bufsize;
int flags;
};
#endif
int
freebsd4_getfsstat(td, uap)
struct thread *td;
register struct freebsd4_getfsstat_args /* {
struct ostatfs *buf;
long bufsize;
int flags;
} */ *uap;
{
struct statfs *buf, *sp;
struct ostatfs osb;
size_t count, size;
int error;
count = uap->bufsize / sizeof(struct ostatfs);
size = count * sizeof(struct statfs);
error = kern_getfsstat(td, &buf, size, &count, UIO_SYSSPACE,
uap->flags);
if (size > 0) {
sp = buf;
while (count > 0 && error == 0) {
cvtstatfs(sp, &osb);
error = copyout(&osb, uap->buf, sizeof(osb));
sp++;
uap->buf++;
count--;
}
free(buf, M_TEMP);
}
if (error == 0)
td->td_retval[0] = count;
return (error);
}
/*
* Implement fstatfs() for (NFS) file handles.
*/
#ifndef _SYS_SYSPROTO_H_
struct freebsd4_fhstatfs_args {
struct fhandle *u_fhp;
struct ostatfs *buf;
};
#endif
int
freebsd4_fhstatfs(td, uap)
struct thread *td;
struct freebsd4_fhstatfs_args /* {
struct fhandle *u_fhp;
struct ostatfs *buf;
} */ *uap;
{
struct ostatfs osb;
struct statfs sf;
fhandle_t fh;
int error;
error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
error = kern_fhstatfs(td, fh, &sf);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
cvtstatfs(&sf, &osb);
return (copyout(&osb, uap->buf, sizeof(osb)));
}
/*
* Convert a new format statfs structure to an old format statfs structure.
*/
static void
cvtstatfs(nsp, osp)
struct statfs *nsp;
struct ostatfs *osp;
{
statfs_scale_blocks(nsp, LONG_MAX);
bzero(osp, sizeof(*osp));
osp->f_bsize = nsp->f_bsize;
osp->f_iosize = MIN(nsp->f_iosize, LONG_MAX);
osp->f_blocks = nsp->f_blocks;
osp->f_bfree = nsp->f_bfree;
osp->f_bavail = nsp->f_bavail;
osp->f_files = MIN(nsp->f_files, LONG_MAX);
osp->f_ffree = MIN(nsp->f_ffree, LONG_MAX);
osp->f_owner = nsp->f_owner;
osp->f_type = nsp->f_type;
osp->f_flags = nsp->f_flags;
osp->f_syncwrites = MIN(nsp->f_syncwrites, LONG_MAX);
osp->f_asyncwrites = MIN(nsp->f_asyncwrites, LONG_MAX);
osp->f_syncreads = MIN(nsp->f_syncreads, LONG_MAX);
osp->f_asyncreads = MIN(nsp->f_asyncreads, LONG_MAX);
strlcpy(osp->f_fstypename, nsp->f_fstypename,
MIN(MFSNAMELEN, OMFSNAMELEN));
strlcpy(osp->f_mntonname, nsp->f_mntonname,
MIN(MNAMELEN, OMNAMELEN));
strlcpy(osp->f_mntfromname, nsp->f_mntfromname,
MIN(MNAMELEN, OMNAMELEN));
osp->f_fsid = nsp->f_fsid;
}
#endif /* COMPAT_FREEBSD4 */
1994-05-24 10:09:53 +00:00
/*
* Change current working directory to a given file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fchdir_args {
int fd;
};
#endif
int
sys_fchdir(td, uap)
struct thread *td;
struct fchdir_args /* {
int fd;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
register struct filedesc *fdp = td->td_proc->p_fd;
struct vnode *vp, *tdp, *vpold;
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->fd);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(fdp, uap->fd, cap_rights_init(&rights, CAP_FCHDIR),
&fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = fp->f_vnode;
VREF(vp);
fdrop(fp, td);
vn_lock(vp, LK_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(vp);
error = change_dir(vp, td);
while (!error && (mp = vp->v_mountedhere) != NULL) {
2008-11-02 10:15:42 +00:00
if (vfs_busy(mp, 0))
continue;
error = VFS_ROOT(mp, LK_SHARED, &tdp);
vfs_unbusy(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
break;
vput(vp);
vp = tdp;
}
2013-09-05 00:19:30 +00:00
if (error != 0) {
vput(vp);
1994-05-24 10:09:53 +00:00
return (error);
}
VOP_UNLOCK(vp, 0);
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XLOCK(fdp);
vpold = fdp->fd_cdir;
1994-05-24 10:09:53 +00:00
fdp->fd_cdir = vp;
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XUNLOCK(fdp);
vrele(vpold);
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Change current working directory (``.'').
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct chdir_args {
char *path;
};
#endif
int
sys_chdir(td, uap)
struct thread *td;
struct chdir_args /* {
char *path;
} */ *uap;
{
return (kern_chdir(td, uap->path, UIO_USERSPACE));
}
int
kern_chdir(struct thread *td, char *path, enum uio_seg pathseg)
1994-05-24 10:09:53 +00:00
{
register struct filedesc *fdp = td->td_proc->p_fd;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
struct vnode *vp;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
pathseg, path, td);
if ((error = namei(&nd)) != 0)
return (error);
if ((error = change_dir(nd.ni_vp, td)) != 0) {
vput(nd.ni_vp);
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
return (error);
}
VOP_UNLOCK(nd.ni_vp, 0);
NDFREE(&nd, NDF_ONLY_PNBUF);
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XLOCK(fdp);
vp = fdp->fd_cdir;
1994-05-24 10:09:53 +00:00
fdp->fd_cdir = nd.ni_vp;
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XUNLOCK(fdp);
vrele(vp);
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Helper function for raised chroot(2) security function: Refuse if
* any filedescriptors are open directories.
*/
static int
chroot_refuse_vdir_fds(fdp)
struct filedesc *fdp;
{
struct vnode *vp;
struct file *fp;
int fd;
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_LOCK_ASSERT(fdp);
for (fd = 0; fd <= fdp->fd_lastfile; fd++) {
fp = fget_locked(fdp, fd);
if (fp == NULL)
continue;
if (fp->f_type == DTYPE_VNODE) {
vp = fp->f_vnode;
if (vp->v_type == VDIR)
return (EPERM);
}
}
return (0);
}
/*
* This sysctl determines if we will allow a process to chroot(2) if it
* has a directory open:
* 0: disallowed for all processes.
* 1: allowed for processes that were not already chroot(2)'ed.
* 2: allowed for all processes.
*/
static int chroot_allow_open_directories = 1;
SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
&chroot_allow_open_directories, 0,
"Allow a process to chroot(2) if it has a directory open");
1994-05-24 10:09:53 +00:00
/*
* Change notion of root (``/'') directory.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct chroot_args {
char *path;
};
#endif
int
sys_chroot(td, uap)
struct thread *td;
struct chroot_args /* {
char *path;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
error = priv_check(td, PRIV_VFS_CHROOT);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
UIO_USERSPACE, uap->path, td);
error = namei(&nd);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto error;
2013-09-05 00:19:30 +00:00
error = change_dir(nd.ni_vp, td);
if (error != 0)
goto e_vunlock;
#ifdef MAC
2013-09-05 00:19:30 +00:00
error = mac_vnode_check_chroot(td->td_ucred, nd.ni_vp);
if (error != 0)
goto e_vunlock;
#endif
VOP_UNLOCK(nd.ni_vp, 0);
error = change_root(nd.ni_vp, td);
vrele(nd.ni_vp);
NDFREE(&nd, NDF_ONLY_PNBUF);
return (error);
e_vunlock:
vput(nd.ni_vp);
error:
NDFREE(&nd, NDF_ONLY_PNBUF);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Common routine for chroot and chdir. Callers must provide a locked vnode
* instance.
1994-05-24 10:09:53 +00:00
*/
int
change_dir(vp, td)
struct vnode *vp;
struct thread *td;
1994-05-24 10:09:53 +00:00
{
2013-09-05 00:19:30 +00:00
#ifdef MAC
1994-05-24 10:09:53 +00:00
int error;
2013-09-05 00:19:30 +00:00
#endif
1994-05-24 10:09:53 +00:00
ASSERT_VOP_LOCKED(vp, "change_dir(): vp not locked");
1994-05-24 10:09:53 +00:00
if (vp->v_type != VDIR)
return (ENOTDIR);
#ifdef MAC
error = mac_vnode_check_chdir(td->td_ucred, vp);
if (error != 0)
return (error);
#endif
2013-09-05 00:19:30 +00:00
return (VOP_ACCESS(vp, VEXEC, td->td_ucred, td));
1994-05-24 10:09:53 +00:00
}
/*
* Common routine for kern_chroot() and jail_attach(). The caller is
* responsible for invoking priv_check() and mac_vnode_check_chroot() to
* authorize this operation.
*/
int
change_root(vp, td)
struct vnode *vp;
struct thread *td;
{
struct filedesc *fdp;
struct vnode *oldvp;
int error;
fdp = td->td_proc->p_fd;
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XLOCK(fdp);
if (chroot_allow_open_directories == 0 ||
(chroot_allow_open_directories == 1 && fdp->fd_rdir != rootvnode)) {
error = chroot_refuse_vdir_fds(fdp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XUNLOCK(fdp);
return (error);
}
}
oldvp = fdp->fd_rdir;
fdp->fd_rdir = vp;
VREF(fdp->fd_rdir);
if (!fdp->fd_jdir) {
fdp->fd_jdir = vp;
VREF(fdp->fd_jdir);
}
Replace custom file descriptor array sleep lock constructed using a mutex and flags with an sxlock. This leads to a significant and measurable performance improvement as a result of access to shared locking for frequent lookup operations, reduced general overhead, and reduced overhead in the event of contention. All of these are imported for threaded applications where simultaneous access to a shared file descriptor array occurs frequently. Kris has reported 2x-4x transaction rate improvements on 8-core MySQL benchmarks; smaller improvements can be expected for many workloads as a result of reduced overhead. - Generally eliminate the distinction between "fast" and regular acquisisition of the filedesc lock; the plan is that they will now all be fast. Change all locking instances to either shared or exclusive locks. - Correct a bug (pointed out by kib) in fdfree() where previously msleep() was called without the mutex held; sx_sleep() is now always called with the sxlock held exclusively. - Universally hold the struct file lock over changes to struct file, rather than the filedesc lock or no lock. Always update the f_ops field last. A further memory barrier is required here in the future (discussed with jhb). - Improve locking and reference management in linux_at(), which fails to properly acquire vnode references before using vnode pointers. Annotate improper use of vn_fullpath(), which will be replaced at a future date. In fcntl(), we conservatively acquire an exclusive lock, even though in some cases a shared lock may be sufficient, which should be revisited. The dropping of the filedesc lock in fdgrowtable() is no longer required as the sxlock can be held over the sleep operation; we should consider removing that (pointed out by attilio). Tested by: kris Discussed with: jhb, kris, attilio, jeff
2007-04-04 09:11:34 +00:00
FILEDESC_XUNLOCK(fdp);
vrele(oldvp);
return (0);
}
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
static __inline void
flags_to_rights(int flags, cap_rights_t *rightsp)
{
if (flags & O_EXEC) {
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_FEXECVE);
} else {
switch ((flags & O_ACCMODE)) {
case O_RDONLY:
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_READ);
break;
case O_RDWR:
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_READ);
/* FALLTHROUGH */
case O_WRONLY:
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_WRITE);
if (!(flags & (O_APPEND | O_TRUNC)))
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_SEEK);
break;
}
}
if (flags & O_CREAT)
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_CREATE);
if (flags & O_TRUNC)
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_FTRUNCATE);
if (flags & (O_SYNC | O_FSYNC))
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_FSYNC);
if (flags & (O_EXLOCK | O_SHLOCK))
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(rightsp, CAP_FLOCK);
}
1994-05-24 10:09:53 +00:00
/*
* Check permissions, allocate an open file structure, and call the device
* open routine if any.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct open_args {
char *path;
int flags;
int mode;
};
#endif
int
sys_open(td, uap)
struct thread *td;
register struct open_args /* {
char *path;
int flags;
int mode;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->flags, uap->mode));
}
#ifndef _SYS_SYSPROTO_H_
struct openat_args {
int fd;
char *path;
int flag;
int mode;
};
#endif
int
sys_openat(struct thread *td, struct openat_args *uap)
{
return (kern_openat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
uap->mode));
}
int
kern_openat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
int flags, int mode)
{
struct proc *p = td->td_proc;
struct filedesc *fdp = p->p_fd;
struct file *fp;
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
2013-09-05 00:19:30 +00:00
int cmode, error, indx;
indx = -1;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FFLAGS(flags);
AUDIT_ARG_MODE(mode);
/* XXX: audit dirfd */
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_LOOKUP);
flags_to_rights(flags, &rights);
/*
* Only one of the O_EXEC, O_RDONLY, O_WRONLY and O_RDWR flags
* may be specified.
*/
if (flags & O_EXEC) {
if (flags & O_ACCMODE)
return (EINVAL);
2012-06-13 21:11:58 +00:00
} else if ((flags & O_ACCMODE) == O_ACCMODE) {
return (EINVAL);
2012-06-13 21:13:18 +00:00
} else {
flags = FFLAGS(flags);
2012-06-13 21:11:58 +00:00
}
/*
2012-06-13 21:11:58 +00:00
* Allocate the file descriptor, but don't install a descriptor yet.
*/
error = falloc_noinstall(td, &fp);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
/*
* An extra reference on `fp' has been held for us by
* falloc_noinstall().
*/
/* Set the flags early so the finit in devfs can pick them up. */
fp->f_flag = flags & FMASK;
2013-09-05 00:19:30 +00:00
cmode = ((mode & ~fdp->fd_cmask) & ALLPERMS) & ~S_ISTXT;
NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
&rights, td);
td->td_dupfd = -1; /* XXX check for fdopen */
error = vn_open(&nd, &flags, cmode, fp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
/*
* If the vn_open replaced the method vector, something
* wonderous happened deep below and we just pass it up
* pretending we know what we do.
*/
if (error == ENXIO && fp->f_ops != &badfileops)
goto success;
/*
* Handle special fdopen() case. bleh.
*
* Don't do this for relative (capability) lookups; we don't
* understand exactly what would happen, and we don't think
* that it ever should.
*/
2012-06-13 21:11:58 +00:00
if (nd.ni_strictrelative == 0 &&
(error == ENODEV || error == ENXIO) &&
2012-06-13 21:11:58 +00:00
td->td_dupfd >= 0) {
error = dupfdopen(td, fdp, td->td_dupfd, flags, error,
&indx);
if (error == 0)
goto success;
1994-05-24 10:09:53 +00:00
}
goto bad;
1994-05-24 10:09:53 +00:00
}
td->td_dupfd = 0;
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vp = nd.ni_vp;
/*
* Store the vnode, for any f_type. Typically, the vnode use
* count is decremented by direct call to vn_closefile() for
* files that switched type in the cdevsw fdopen() method.
*/
fp->f_vnode = vp;
/*
* If the file wasn't claimed by devfs bind it to the normal
* vnode operations here.
*/
if (fp->f_ops == &badfileops) {
KASSERT(vp->v_type != VFIFO, ("Unexpected fifo."));
fp->f_seqcount = 1;
2013-09-05 00:19:30 +00:00
finit(fp, (flags & FMASK) | (fp->f_flag & FHASLOCK),
DTYPE_VNODE, vp, &vnops);
}
VOP_UNLOCK(vp, 0);
if (flags & O_TRUNC) {
error = fo_truncate(fp, 0, td->td_ucred, td);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto bad;
}
success:
/*
* If we haven't already installed the FD (for dupfdopen), do so now.
*/
if (indx == -1) {
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
struct filecaps *fcaps;
#ifdef CAPABILITIES
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
if (nd.ni_strictrelative == 1)
fcaps = &nd.ni_filecaps;
else
#endif
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
fcaps = NULL;
error = finstall(td, fp, &indx, flags, fcaps);
/* On success finstall() consumes fcaps. */
if (error != 0) {
filecaps_free(&nd.ni_filecaps);
goto bad;
}
} else {
filecaps_free(&nd.ni_filecaps);
}
/*
* Release our private reference, leaving the one associated with
* the descriptor table intact.
*/
fdrop(fp, td);
td->td_retval[0] = indx;
1994-05-24 10:09:53 +00:00
return (0);
bad:
KASSERT(indx == -1, ("indx=%d, should be -1", indx));
fdrop(fp, td);
return (error);
1994-05-24 10:09:53 +00:00
}
#ifdef COMPAT_43
/*
* Create a file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct ocreat_args {
char *path;
int mode;
};
#endif
int
ocreat(td, uap)
struct thread *td;
register struct ocreat_args /* {
char *path;
int mode;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_openat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
O_WRONLY | O_CREAT | O_TRUNC, uap->mode));
1994-05-24 10:09:53 +00:00
}
#endif /* COMPAT_43 */
/*
* Create a special file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct mknod_args {
char *path;
int mode;
int dev;
};
#endif
int
sys_mknod(td, uap)
struct thread *td;
register struct mknod_args /* {
char *path;
int mode;
int dev;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_mknodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->mode, uap->dev));
}
#ifndef _SYS_SYSPROTO_H_
struct mknodat_args {
int fd;
char *path;
mode_t mode;
dev_t dev;
};
#endif
int
sys_mknodat(struct thread *td, struct mknodat_args *uap)
{
return (kern_mknodat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode,
uap->dev));
}
int
kern_mknodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
int mode, int dev)
{
struct vnode *vp;
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
2013-09-05 00:19:30 +00:00
int error, whiteout = 0;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_MODE(mode);
AUDIT_ARG_DEV(dev);
switch (mode & S_IFMT) {
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
case S_IFCHR:
case S_IFBLK:
error = priv_check(td, PRIV_VFS_MKNOD_DEV);
break;
case S_IFMT:
error = priv_check(td, PRIV_VFS_MKNOD_BAD);
break;
case S_IFWHT:
error = priv_check(td, PRIV_VFS_MKNOD_WHT);
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
break;
case S_IFIFO:
if (dev == 0)
return (kern_mkfifoat(td, fd, path, pathseg, mode));
/* FALLTHROUGH */
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
default:
error = EINVAL;
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
break;
}
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKNODAT),
td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = nd.ni_vp;
if (vp != NULL) {
NDFREE(&nd, NDF_ONLY_PNBUF);
if (vp == nd.ni_dvp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
vrele(vp);
return (EEXIST);
} else {
1994-05-24 10:09:53 +00:00
VATTR_NULL(&vattr);
vattr.va_mode = (mode & ALLPERMS) &
~td->td_proc->p_fd->fd_cmask;
vattr.va_rdev = dev;
whiteout = 0;
1994-05-24 10:09:53 +00:00
switch (mode & S_IFMT) {
1994-05-24 10:09:53 +00:00
case S_IFMT: /* used by badsect to flag bad sectors */
vattr.va_type = VBAD;
break;
case S_IFCHR:
vattr.va_type = VCHR;
break;
case S_IFBLK:
vattr.va_type = VBLK;
break;
case S_IFWHT:
whiteout = 1;
break;
1994-05-24 10:09:53 +00:00
default:
panic("kern_mknod: invalid mode");
1994-05-24 10:09:53 +00:00
}
}
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
#ifdef MAC
if (error == 0 && !whiteout)
error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp,
&nd.ni_cnd, &vattr);
#endif
2013-09-05 00:19:30 +00:00
if (error == 0) {
if (whiteout)
error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, CREATE);
else {
error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp,
&nd.ni_cnd, &vattr);
1999-11-13 14:35:50 +00:00
if (error == 0)
vput(nd.ni_vp);
}
1994-05-24 10:09:53 +00:00
}
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
vn_finished_write(mp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Create a named pipe.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct mkfifo_args {
char *path;
int mode;
};
#endif
int
sys_mkfifo(td, uap)
struct thread *td;
register struct mkfifo_args /* {
char *path;
int mode;
} */ *uap;
{
return (kern_mkfifoat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->mode));
}
#ifndef _SYS_SYSPROTO_H_
struct mkfifoat_args {
int fd;
char *path;
mode_t mode;
};
#endif
int
sys_mkfifoat(struct thread *td, struct mkfifoat_args *uap)
{
return (kern_mkfifoat(td, uap->fd, uap->path, UIO_USERSPACE,
uap->mode));
}
int
kern_mkfifoat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
int mode)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
2013-09-05 00:19:30 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_MODE(mode);
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
NOCACHE, pathseg, path, fd, cap_rights_init(&rights, CAP_MKFIFOAT),
td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
if (nd.ni_vp != NULL) {
NDFREE(&nd, NDF_ONLY_PNBUF);
if (nd.ni_vp == nd.ni_dvp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
vrele(nd.ni_vp);
1994-05-24 10:09:53 +00:00
return (EEXIST);
}
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
1994-05-24 10:09:53 +00:00
VATTR_NULL(&vattr);
vattr.va_type = VFIFO;
vattr.va_mode = (mode & ALLPERMS) & ~td->td_proc->p_fd->fd_cmask;
#ifdef MAC
error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
&vattr);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
1999-11-13 14:35:50 +00:00
if (error == 0)
vput(nd.ni_vp);
#ifdef MAC
out:
#endif
vput(nd.ni_dvp);
vn_finished_write(mp);
NDFREE(&nd, NDF_ONLY_PNBUF);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Make a hard file link.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct link_args {
char *path;
char *link;
};
#endif
int
sys_link(td, uap)
struct thread *td;
register struct link_args /* {
char *path;
char *link;
} */ *uap;
{
return (kern_linkat(td, AT_FDCWD, AT_FDCWD, uap->path, uap->link,
UIO_USERSPACE, FOLLOW));
}
#ifndef _SYS_SYSPROTO_H_
struct linkat_args {
int fd1;
char *path1;
int fd2;
char *path2;
int flag;
};
#endif
int
sys_linkat(struct thread *td, struct linkat_args *uap)
{
int flag;
flag = uap->flag;
if (flag & ~AT_SYMLINK_FOLLOW)
return (EINVAL);
return (kern_linkat(td, uap->fd1, uap->fd2, uap->path1, uap->path2,
UIO_USERSPACE, (flag & AT_SYMLINK_FOLLOW) ? FOLLOW : NOFOLLOW));
}
int hardlink_check_uid = 0;
SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_uid, CTLFLAG_RW,
&hardlink_check_uid, 0,
"Unprivileged processes cannot create hard links to files owned by other "
"users");
static int hardlink_check_gid = 0;
SYSCTL_INT(_security_bsd, OID_AUTO, hardlink_check_gid, CTLFLAG_RW,
&hardlink_check_gid, 0,
"Unprivileged processes cannot create hard links to files owned by other "
"groups");
static int
can_hardlink(struct vnode *vp, struct ucred *cred)
{
struct vattr va;
int error;
if (!hardlink_check_uid && !hardlink_check_gid)
return (0);
error = VOP_GETATTR(vp, &va, cred);
if (error != 0)
return (error);
if (hardlink_check_uid && cred->cr_uid != va.va_uid) {
error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
}
if (hardlink_check_gid && !groupmember(va.va_gid, cred)) {
error = priv_check_cred(cred, PRIV_VFS_LINK, 0);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
}
return (0);
}
int
kern_linkat(struct thread *td, int fd1, int fd2, char *path1, char *path2,
enum uio_seg segflg, int follow)
1994-05-24 10:09:53 +00:00
{
struct vnode *vp;
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
again:
bwillwrite();
NDINIT_AT(&nd, LOOKUP, follow | AUDITVNODE1, segflg, path1, fd1, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vp = nd.ni_vp;
if (vp->v_type == VDIR) {
vrele(vp);
return (EPERM); /* POSIX */
}
NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE2 |
NOCACHE, segflg, path2, fd2, cap_rights_init(&rights, CAP_LINKAT),
td);
if ((error = namei(&nd)) == 0) {
if (nd.ni_vp != NULL) {
NDFREE(&nd, NDF_ONLY_PNBUF);
if (nd.ni_dvp == nd.ni_vp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
vrele(nd.ni_vp);
vrele(vp);
return (EEXIST);
} else if (nd.ni_dvp->v_mount != vp->v_mount) {
/*
* Cross-device link. No need to recheck
* vp->v_type, since it cannot change, except
* to VBAD.
*/
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
vrele(vp);
return (EXDEV);
} else if ((error = vn_lock(vp, LK_EXCLUSIVE)) == 0) {
error = can_hardlink(vp, td->td_ucred);
#ifdef MAC
if (error == 0)
error = mac_vnode_check_link(td->td_ucred,
nd.ni_dvp, vp, &nd.ni_cnd);
#endif
if (error != 0) {
vput(vp);
vput(nd.ni_dvp);
NDFREE(&nd, NDF_ONLY_PNBUF);
return (error);
}
error = vn_start_write(vp, &mp, V_NOWAIT);
if (error != 0) {
vput(vp);
vput(nd.ni_dvp);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = vn_start_write(NULL, &mp,
V_XSLEEP | PCATCH);
if (error != 0)
return (error);
goto again;
}
error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
VOP_UNLOCK(vp, 0);
vput(nd.ni_dvp);
vn_finished_write(mp);
NDFREE(&nd, NDF_ONLY_PNBUF);
} else {
vput(nd.ni_dvp);
NDFREE(&nd, NDF_ONLY_PNBUF);
vrele(vp);
goto again;
1994-05-24 10:09:53 +00:00
}
}
vrele(vp);
return (error);
}
/*
* Make a symbolic link.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct symlink_args {
char *path;
char *link;
};
#endif
int
sys_symlink(td, uap)
struct thread *td;
register struct symlink_args /* {
char *path;
char *link;
} */ *uap;
{
return (kern_symlinkat(td, uap->path, AT_FDCWD, uap->link,
UIO_USERSPACE));
}
#ifndef _SYS_SYSPROTO_H_
struct symlinkat_args {
char *path;
int fd;
char *path2;
};
#endif
int
sys_symlinkat(struct thread *td, struct symlinkat_args *uap)
{
return (kern_symlinkat(td, uap->path1, uap->fd, uap->path2,
UIO_USERSPACE));
}
int
kern_symlinkat(struct thread *td, char *path1, int fd, char *path2,
enum uio_seg segflg)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
char *syspath;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
if (segflg == UIO_SYSSPACE) {
syspath = path1;
} else {
syspath = uma_zalloc(namei_zone, M_WAITOK);
if ((error = copyinstr(path1, syspath, MAXPATHLEN, NULL)) != 0)
goto out;
}
AUDIT_ARG_TEXT(syspath);
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
NOCACHE, segflg, path2, fd, cap_rights_init(&rights, CAP_SYMLINKAT),
td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
goto out;
if (nd.ni_vp) {
NDFREE(&nd, NDF_ONLY_PNBUF);
if (nd.ni_vp == nd.ni_dvp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
vrele(nd.ni_vp);
1994-05-24 10:09:53 +00:00
error = EEXIST;
goto out;
}
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
goto out;
goto restart;
}
1994-05-24 10:09:53 +00:00
VATTR_NULL(&vattr);
vattr.va_mode = ACCESSPERMS &~ td->td_proc->p_fd->fd_cmask;
#ifdef MAC
vattr.va_type = VLNK;
error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
&vattr);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out2;
#endif
error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr, syspath);
if (error == 0)
vput(nd.ni_vp);
#ifdef MAC
out2:
#endif
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
vn_finished_write(mp);
1994-05-24 10:09:53 +00:00
out:
if (segflg != UIO_SYSSPACE)
uma_zfree(namei_zone, syspath);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Delete a whiteout from the filesystem.
*/
int
sys_undelete(td, uap)
struct thread *td;
register struct undelete_args /* {
char *path;
} */ *uap;
{
struct mount *mp;
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
restart:
bwillwrite();
NDINIT(&nd, DELETE, LOCKPARENT | DOWHITEOUT | AUDITVNODE1,
UIO_USERSPACE, uap->path, td);
error = namei(&nd);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
if (nd.ni_vp != NULLVP || !(nd.ni_cnd.cn_flags & ISWHITEOUT)) {
NDFREE(&nd, NDF_ONLY_PNBUF);
if (nd.ni_vp == nd.ni_dvp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
if (nd.ni_vp)
vrele(nd.ni_vp);
return (EEXIST);
}
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
error = VOP_WHITEOUT(nd.ni_dvp, &nd.ni_cnd, DELETE);
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
vn_finished_write(mp);
return (error);
}
1994-05-24 10:09:53 +00:00
/*
* Delete a name from the filesystem.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct unlink_args {
char *path;
};
#endif
int
sys_unlink(td, uap)
struct thread *td;
struct unlink_args /* {
char *path;
} */ *uap;
{
return (kern_unlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE, 0));
}
#ifndef _SYS_SYSPROTO_H_
struct unlinkat_args {
int fd;
char *path;
int flag;
};
#endif
int
sys_unlinkat(struct thread *td, struct unlinkat_args *uap)
{
int flag = uap->flag;
int fd = uap->fd;
char *path = uap->path;
if (flag & ~AT_REMOVEDIR)
return (EINVAL);
if (flag & AT_REMOVEDIR)
return (kern_rmdirat(td, fd, path, UIO_USERSPACE));
else
Background: When renaming a directory it passes through several intermediate states. First its new name will be created causing it to have two names (from possibly different parents). Next, if it has different parents, its value of ".." will be changed from pointing to the old parent to pointing to the new parent. Concurrently, its old name will be removed bringing it back into a consistent state. When fsck encounters an extra name for a directory, it offers to remove the "extraneous hard link"; when it finds that the names have been changed but the update to ".." has not happened, it offers to rewrite ".." to point at the correct parent. Both of these changes were considered unexpected so would cause fsck in preen mode or fsck in background mode to fail with the need to run fsck manually to fix these problems. Fsck running in preen mode or background mode now corrects these expected inconsistencies that arise during directory rename. The functionality added with this update is used by fsck running in background mode to make these fixes. Solution: This update adds three new fsck sysctl commands to support background fsck in correcting expected inconsistencies that arise from incomplete directory rename operations. They are: setcwd(dirinode) - set the current directory to dirinode in the filesystem associated with the snapshot. setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." in the current directory is oldvalue then change it to newvalue. unlink(nameptr, oldvalue) - Verify that the inode number associated with nameptr in the current directory is oldvalue then unlink it. As with all other fsck sysctls, these new ones may only be used by processes with appropriate priviledge. Reported by: jeff Security issues: rwatson
2010-01-11 20:44:05 +00:00
return (kern_unlinkat(td, fd, path, UIO_USERSPACE, 0));
}
int
Background: When renaming a directory it passes through several intermediate states. First its new name will be created causing it to have two names (from possibly different parents). Next, if it has different parents, its value of ".." will be changed from pointing to the old parent to pointing to the new parent. Concurrently, its old name will be removed bringing it back into a consistent state. When fsck encounters an extra name for a directory, it offers to remove the "extraneous hard link"; when it finds that the names have been changed but the update to ".." has not happened, it offers to rewrite ".." to point at the correct parent. Both of these changes were considered unexpected so would cause fsck in preen mode or fsck in background mode to fail with the need to run fsck manually to fix these problems. Fsck running in preen mode or background mode now corrects these expected inconsistencies that arise during directory rename. The functionality added with this update is used by fsck running in background mode to make these fixes. Solution: This update adds three new fsck sysctl commands to support background fsck in correcting expected inconsistencies that arise from incomplete directory rename operations. They are: setcwd(dirinode) - set the current directory to dirinode in the filesystem associated with the snapshot. setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." in the current directory is oldvalue then change it to newvalue. unlink(nameptr, oldvalue) - Verify that the inode number associated with nameptr in the current directory is oldvalue then unlink it. As with all other fsck sysctls, these new ones may only be used by processes with appropriate priviledge. Reported by: jeff Security issues: rwatson
2010-01-11 20:44:05 +00:00
kern_unlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
ino_t oldinum)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
Background: When renaming a directory it passes through several intermediate states. First its new name will be created causing it to have two names (from possibly different parents). Next, if it has different parents, its value of ".." will be changed from pointing to the old parent to pointing to the new parent. Concurrently, its old name will be removed bringing it back into a consistent state. When fsck encounters an extra name for a directory, it offers to remove the "extraneous hard link"; when it finds that the names have been changed but the update to ".." has not happened, it offers to rewrite ".." to point at the correct parent. Both of these changes were considered unexpected so would cause fsck in preen mode or fsck in background mode to fail with the need to run fsck manually to fix these problems. Fsck running in preen mode or background mode now corrects these expected inconsistencies that arise during directory rename. The functionality added with this update is used by fsck running in background mode to make these fixes. Solution: This update adds three new fsck sysctl commands to support background fsck in correcting expected inconsistencies that arise from incomplete directory rename operations. They are: setcwd(dirinode) - set the current directory to dirinode in the filesystem associated with the snapshot. setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." in the current directory is oldvalue then change it to newvalue. unlink(nameptr, oldvalue) - Verify that the inode number associated with nameptr in the current directory is oldvalue then unlink it. As with all other fsck sysctls, these new ones may only be used by processes with appropriate priviledge. Reported by: jeff Security issues: rwatson
2010-01-11 20:44:05 +00:00
struct stat sb;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
if ((error = namei(&nd)) != 0)
return (error == EINVAL ? EPERM : error);
1994-05-24 10:09:53 +00:00
vp = nd.ni_vp;
Background: When renaming a directory it passes through several intermediate states. First its new name will be created causing it to have two names (from possibly different parents). Next, if it has different parents, its value of ".." will be changed from pointing to the old parent to pointing to the new parent. Concurrently, its old name will be removed bringing it back into a consistent state. When fsck encounters an extra name for a directory, it offers to remove the "extraneous hard link"; when it finds that the names have been changed but the update to ".." has not happened, it offers to rewrite ".." to point at the correct parent. Both of these changes were considered unexpected so would cause fsck in preen mode or fsck in background mode to fail with the need to run fsck manually to fix these problems. Fsck running in preen mode or background mode now corrects these expected inconsistencies that arise during directory rename. The functionality added with this update is used by fsck running in background mode to make these fixes. Solution: This update adds three new fsck sysctl commands to support background fsck in correcting expected inconsistencies that arise from incomplete directory rename operations. They are: setcwd(dirinode) - set the current directory to dirinode in the filesystem associated with the snapshot. setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." in the current directory is oldvalue then change it to newvalue. unlink(nameptr, oldvalue) - Verify that the inode number associated with nameptr in the current directory is oldvalue then unlink it. As with all other fsck sysctls, these new ones may only be used by processes with appropriate priviledge. Reported by: jeff Security issues: rwatson
2010-01-11 20:44:05 +00:00
if (vp->v_type == VDIR && oldinum == 0) {
error = EPERM; /* POSIX */
Background: When renaming a directory it passes through several intermediate states. First its new name will be created causing it to have two names (from possibly different parents). Next, if it has different parents, its value of ".." will be changed from pointing to the old parent to pointing to the new parent. Concurrently, its old name will be removed bringing it back into a consistent state. When fsck encounters an extra name for a directory, it offers to remove the "extraneous hard link"; when it finds that the names have been changed but the update to ".." has not happened, it offers to rewrite ".." to point at the correct parent. Both of these changes were considered unexpected so would cause fsck in preen mode or fsck in background mode to fail with the need to run fsck manually to fix these problems. Fsck running in preen mode or background mode now corrects these expected inconsistencies that arise during directory rename. The functionality added with this update is used by fsck running in background mode to make these fixes. Solution: This update adds three new fsck sysctl commands to support background fsck in correcting expected inconsistencies that arise from incomplete directory rename operations. They are: setcwd(dirinode) - set the current directory to dirinode in the filesystem associated with the snapshot. setdotdot(oldvalue, newvalue) - Verify that the inode number for ".." in the current directory is oldvalue then change it to newvalue. unlink(nameptr, oldvalue) - Verify that the inode number associated with nameptr in the current directory is oldvalue then unlink it. As with all other fsck sysctls, these new ones may only be used by processes with appropriate priviledge. Reported by: jeff Security issues: rwatson
2010-01-11 20:44:05 +00:00
} else if (oldinum != 0 &&
((error = vn_stat(vp, &sb, td->td_ucred, NOCRED, td)) == 0) &&
sb.st_ino != oldinum) {
error = EIDRM; /* Identifier removed */
} else {
1994-05-24 10:09:53 +00:00
/*
* The root of a mounted filesystem cannot be deleted.
*
* XXX: can this only be a VDIR case?
1994-05-24 10:09:53 +00:00
*/
if (vp->v_vflag & VV_ROOT)
1994-05-24 10:09:53 +00:00
error = EBUSY;
}
if (error == 0) {
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if (vp == nd.ni_dvp)
vrele(vp);
else
vput(vp);
if ((error = vn_start_write(NULL, &mp,
V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
#ifdef MAC
error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
&nd.ni_cnd);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd);
#ifdef MAC
out:
#endif
vn_finished_write(mp);
1994-05-24 10:09:53 +00:00
}
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if (vp == nd.ni_dvp)
vrele(vp);
else
vput(vp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Reposition read/write file offset.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct lseek_args {
int fd;
int pad;
off_t offset;
int whence;
};
#endif
int
sys_lseek(td, uap)
struct thread *td;
register struct lseek_args /* {
int fd;
int pad;
off_t offset;
int whence;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(uap->fd);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, uap->fd, cap_rights_init(&rights, CAP_SEEK), &fp);
if (error != 0)
return (error);
error = (fp->f_ops->fo_flags & DFLAG_SEEKABLE) != 0 ?
fo_seek(fp, uap->offset, uap->whence, td) : ESPIPE;
fdrop(fp, td);
return (error);
1994-05-24 10:09:53 +00:00
}
#if defined(COMPAT_43)
1994-05-24 10:09:53 +00:00
/*
* Reposition read/write file offset.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct olseek_args {
int fd;
long offset;
int whence;
};
#endif
int
olseek(td, uap)
struct thread *td;
register struct olseek_args /* {
int fd;
long offset;
int whence;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct lseek_args /* {
int fd;
int pad;
off_t offset;
int whence;
} */ nuap;
1994-05-24 10:09:53 +00:00
2002-12-14 01:56:26 +00:00
nuap.fd = uap->fd;
nuap.offset = uap->offset;
nuap.whence = uap->whence;
return (sys_lseek(td, &nuap));
1994-05-24 10:09:53 +00:00
}
#endif /* COMPAT_43 */
#if defined(COMPAT_FREEBSD6)
/* Version with the 'pad' argument */
int
freebsd6_lseek(td, uap)
struct thread *td;
register struct freebsd6_lseek_args *uap;
{
struct lseek_args ouap;
ouap.fd = uap->fd;
ouap.offset = uap->offset;
ouap.whence = uap->whence;
return (sys_lseek(td, &ouap));
}
#endif
1994-05-24 10:09:53 +00:00
/*
* Check access permissions using passed credentials.
*/
static int
vn_access(vp, user_flags, cred, td)
struct vnode *vp;
int user_flags;
struct ucred *cred;
struct thread *td;
{
accmode_t accmode;
2013-09-05 00:19:30 +00:00
int error;
/* Flags == 0 means only check for existence. */
if (user_flags == 0)
return (0);
accmode = 0;
if (user_flags & R_OK)
accmode |= VREAD;
if (user_flags & W_OK)
accmode |= VWRITE;
if (user_flags & X_OK)
accmode |= VEXEC;
#ifdef MAC
error = mac_vnode_check_access(cred, vp, accmode);
if (error != 0)
return (error);
#endif
if ((accmode & VWRITE) == 0 || (error = vn_writechk(vp)) == 0)
error = VOP_ACCESS(vp, accmode, cred, td);
return (error);
}
/*
* Check access permissions using "real" credentials.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct access_args {
char *path;
int amode;
1994-05-24 10:09:53 +00:00
};
#endif
int
sys_access(td, uap)
struct thread *td;
register struct access_args /* {
char *path;
int amode;
} */ *uap;
{
return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
0, uap->amode));
}
#ifndef _SYS_SYSPROTO_H_
struct faccessat_args {
int dirfd;
char *path;
int amode;
int flag;
}
#endif
2008-10-23 21:50:16 +00:00
int
sys_faccessat(struct thread *td, struct faccessat_args *uap)
{
return (kern_accessat(td, uap->fd, uap->path, UIO_USERSPACE, uap->flag,
uap->amode));
}
int
kern_accessat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
int flag, int amode)
1994-05-24 10:09:53 +00:00
{
struct ucred *cred, *usecred;
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
1994-05-24 10:09:53 +00:00
if (flag & ~AT_EACCESS)
return (EINVAL);
if (amode != F_OK && (amode & ~(R_OK | W_OK | X_OK)) != 0)
return (EINVAL);
/*
* Create and modify a temporary credential instead of one that
* is potentially shared (if we need one).
*/
cred = td->td_ucred;
if ((flag & AT_EACCESS) == 0 &&
((cred->cr_uid != cred->cr_ruid ||
cred->cr_rgid != cred->cr_groups[0]))) {
usecred = crdup(cred);
usecred->cr_uid = cred->cr_ruid;
usecred->cr_groups[0] = cred->cr_rgid;
td->td_ucred = usecred;
} else
usecred = cred;
AUDIT_ARG_VALUE(amode);
NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | LOCKSHARED | LOCKLEAF |
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
AUDITVNODE1, pathseg, path, fd, cap_rights_init(&rights, CAP_FSTAT),
td);
if ((error = namei(&nd)) != 0)
goto out;
1994-05-24 10:09:53 +00:00
vp = nd.ni_vp;
error = vn_access(vp, amode, usecred, td);
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vput(vp);
out:
if (usecred != cred) {
td->td_ucred = cred;
crfree(usecred);
}
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Check access permissions using "effective" credentials.
*/
#ifndef _SYS_SYSPROTO_H_
struct eaccess_args {
char *path;
int amode;
};
#endif
int
sys_eaccess(td, uap)
struct thread *td;
register struct eaccess_args /* {
char *path;
int amode;
} */ *uap;
{
return (kern_accessat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
AT_EACCESS, uap->amode));
}
#if defined(COMPAT_43)
1994-05-24 10:09:53 +00:00
/*
* Get file status; this version follows links.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct ostat_args {
char *path;
struct ostat *ub;
};
#endif
int
ostat(td, uap)
struct thread *td;
register struct ostat_args /* {
char *path;
struct ostat *ub;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct stat sb;
struct ostat osb;
int error;
error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
&sb, NULL);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
cvtstat(&sb, &osb);
2013-09-05 00:19:30 +00:00
return (copyout(&osb, uap->ub, sizeof (osb)));
1994-05-24 10:09:53 +00:00
}
/*
* Get file status; this version does not follow links.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct olstat_args {
char *path;
struct ostat *ub;
};
#endif
int
olstat(td, uap)
struct thread *td;
register struct olstat_args /* {
char *path;
struct ostat *ub;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct stat sb;
1994-05-24 10:09:53 +00:00
struct ostat osb;
int error;
error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
UIO_USERSPACE, &sb, NULL);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
1994-05-24 10:09:53 +00:00
cvtstat(&sb, &osb);
2013-09-05 00:19:30 +00:00
return (copyout(&osb, uap->ub, sizeof (osb)));
1994-05-24 10:09:53 +00:00
}
/*
* Convert from an old to a new stat structure.
*/
void
1994-05-24 10:09:53 +00:00
cvtstat(st, ost)
struct stat *st;
struct ostat *ost;
{
ost->st_dev = st->st_dev;
ost->st_ino = st->st_ino;
ost->st_mode = st->st_mode;
ost->st_nlink = st->st_nlink;
ost->st_uid = st->st_uid;
ost->st_gid = st->st_gid;
ost->st_rdev = st->st_rdev;
if (st->st_size < (quad_t)1 << 32)
ost->st_size = st->st_size;
else
ost->st_size = -2;
ost->st_atim = st->st_atim;
ost->st_mtim = st->st_mtim;
ost->st_ctim = st->st_ctim;
1994-05-24 10:09:53 +00:00
ost->st_blksize = st->st_blksize;
ost->st_blocks = st->st_blocks;
ost->st_flags = st->st_flags;
ost->st_gen = st->st_gen;
}
#endif /* COMPAT_43 */
1994-05-24 10:09:53 +00:00
/*
* Get file status; this version follows links.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct stat_args {
char *path;
struct stat *ub;
};
#endif
int
sys_stat(td, uap)
struct thread *td;
register struct stat_args /* {
char *path;
struct stat *ub;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct stat sb;
int error;
error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
&sb, NULL);
if (error == 0)
error = copyout(&sb, uap->ub, sizeof (sb));
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct fstatat_args {
int fd;
char *path;
struct stat *buf;
int flag;
}
#endif
int
sys_fstatat(struct thread *td, struct fstatat_args *uap)
{
struct stat sb;
int error;
error = kern_statat(td, uap->flag, uap->fd, uap->path,
UIO_USERSPACE, &sb, NULL);
if (error == 0)
error = copyout(&sb, uap->buf, sizeof (sb));
return (error);
}
int
kern_statat(struct thread *td, int flag, int fd, char *path,
enum uio_seg pathseg, struct stat *sbp,
void (*hook)(struct vnode *vp, struct stat *sbp))
{
1994-05-24 10:09:53 +00:00
struct nameidata nd;
struct stat sb;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
1994-05-24 10:09:53 +00:00
if (flag & ~AT_SYMLINK_NOFOLLOW)
return (EINVAL);
NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
FOLLOW) | LOCKSHARED | LOCKLEAF | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_FSTAT), td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
Make similar changes to fo_stat() and fo_poll() as made earlier to fo_read() and fo_write(): explicitly use the cred argument to fo_poll() as "active_cred" using the passed file descriptor's f_cred reference to provide access to the file credential. Add an active_cred argument to fo_stat() so that implementers have access to the active credential as well as the file credential. Generally modify callers of fo_stat() to pass in td->td_ucred rather than fp->f_cred, which was redundantly provided via the fp argument. This set of modifications also permits threads to perform these operations on behalf of another thread without modifying their credential. Trickle this change down into fo_stat/poll() implementations: - badfo_poll(), badfo_stat(): modify/add arguments. - kqueue_poll(), kqueue_stat(): modify arguments. - pipe_poll(), pipe_stat(): modify/add arguments, pass active_cred to MAC checks rather than td->td_ucred. - soo_poll(), soo_stat(): modify/add arguments, pass fp->f_cred rather than cred to pru_sopoll() to maintain current semantics. - sopoll(): moidfy arguments. - vn_poll(), vn_statfile(): modify/add arguments, pass new arguments to vn_stat(). Pass active_cred to MAC and fp->f_cred to VOP_POLL() to maintian current semantics. - vn_close(): rename cred to file_cred to reflect reality while I'm here. - vn_stat(): Add active_cred and file_cred arguments to vn_stat() and consumers so that this distinction is maintained at the VFS as well as 'struct file' layer. Pass active_cred instead of td->td_ucred to MAC and to VOP_GETATTR() to maintain current semantics. - fifofs: modify the creation of a "filetemp" so that the file credential is properly initialized and can be used in the socket code if desired. Pass ap->a_td->td_ucred as the active credential to soo_poll(). If we teach the vnop interface about the distinction between file and active credentials, we would use the active credential here. Note that current inconsistent passing of active_cred vs. file_cred to VOP's is maintained. It's not clear why GETATTR would be authorized using active_cred while POLL would be authorized using file_cred at the file system level. Obtained from: TrustedBSD Project Sponsored by: DARPA, NAI Labs
2002-08-16 12:52:03 +00:00
error = vn_stat(nd.ni_vp, &sb, td->td_ucred, NOCRED, td);
2013-09-05 00:19:30 +00:00
if (error == 0) {
SDT_PROBE(vfs, , stat, mode, path, sb.st_mode, 0, 0, 0);
if (S_ISREG(sb.st_mode))
SDT_PROBE(vfs, , stat, reg, path, pathseg, 0, 0, 0);
if (__predict_false(hook != NULL))
hook(nd.ni_vp, &sb);
}
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vput(nd.ni_vp);
2013-09-05 00:19:30 +00:00
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
*sbp = sb;
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrstat(&sb);
#endif
return (0);
1994-05-24 10:09:53 +00:00
}
/*
* Get file status; this version does not follow links.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct lstat_args {
char *path;
struct stat *ub;
};
#endif
int
sys_lstat(td, uap)
struct thread *td;
register struct lstat_args /* {
char *path;
struct stat *ub;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct stat sb;
1994-05-24 10:09:53 +00:00
int error;
error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
UIO_USERSPACE, &sb, NULL);
if (error == 0)
error = copyout(&sb, uap->ub, sizeof (sb));
return (error);
}
/*
* Implementation of the NetBSD [l]stat() functions.
*/
void
cvtnstat(sb, nsb)
struct stat *sb;
struct nstat *nsb;
{
2013-09-05 00:19:30 +00:00
bzero(nsb, sizeof *nsb);
nsb->st_dev = sb->st_dev;
nsb->st_ino = sb->st_ino;
nsb->st_mode = sb->st_mode;
nsb->st_nlink = sb->st_nlink;
nsb->st_uid = sb->st_uid;
nsb->st_gid = sb->st_gid;
nsb->st_rdev = sb->st_rdev;
nsb->st_atim = sb->st_atim;
nsb->st_mtim = sb->st_mtim;
nsb->st_ctim = sb->st_ctim;
nsb->st_size = sb->st_size;
nsb->st_blocks = sb->st_blocks;
nsb->st_blksize = sb->st_blksize;
nsb->st_flags = sb->st_flags;
nsb->st_gen = sb->st_gen;
nsb->st_birthtim = sb->st_birthtim;
}
#ifndef _SYS_SYSPROTO_H_
struct nstat_args {
char *path;
struct nstat *ub;
};
#endif
int
sys_nstat(td, uap)
struct thread *td;
register struct nstat_args /* {
char *path;
struct nstat *ub;
} */ *uap;
{
struct stat sb;
struct nstat nsb;
int error;
error = kern_statat(td, 0, AT_FDCWD, uap->path, UIO_USERSPACE,
&sb, NULL);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
cvtnstat(&sb, &nsb);
2013-09-05 00:19:30 +00:00
return (copyout(&nsb, uap->ub, sizeof (nsb)));
}
/*
* NetBSD lstat. Get file status; this version does not follow links.
*/
#ifndef _SYS_SYSPROTO_H_
struct lstat_args {
char *path;
struct stat *ub;
};
#endif
int
sys_nlstat(td, uap)
struct thread *td;
register struct nlstat_args /* {
char *path;
struct nstat *ub;
} */ *uap;
{
struct stat sb;
struct nstat nsb;
int error;
error = kern_statat(td, AT_SYMLINK_NOFOLLOW, AT_FDCWD, uap->path,
UIO_USERSPACE, &sb, NULL);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
cvtnstat(&sb, &nsb);
2013-09-05 00:19:30 +00:00
return (copyout(&nsb, uap->ub, sizeof (nsb)));
}
1994-05-24 10:09:53 +00:00
/*
* Get configurable pathname variables.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct pathconf_args {
char *path;
int name;
};
#endif
int
sys_pathconf(td, uap)
struct thread *td;
register struct pathconf_args /* {
char *path;
int name;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name, FOLLOW));
}
#ifndef _SYS_SYSPROTO_H_
struct lpathconf_args {
char *path;
int name;
};
#endif
int
sys_lpathconf(td, uap)
struct thread *td;
register struct lpathconf_args /* {
char *path;
int name;
} */ *uap;
{
return (kern_pathconf(td, uap->path, UIO_USERSPACE, uap->name,
NOFOLLOW));
}
int
kern_pathconf(struct thread *td, char *path, enum uio_seg pathseg, int name,
u_long flags)
{
1994-05-24 10:09:53 +00:00
struct nameidata nd;
int error;
1994-05-24 10:09:53 +00:00
NDINIT(&nd, LOOKUP, LOCKSHARED | LOCKLEAF | AUDITVNODE1 | flags,
pathseg, path, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
/* If asynchronous I/O is available, it works for all files. */
if (name == _PC_ASYNC_IO)
td->td_retval[0] = async_io_version;
else
error = VOP_PATHCONF(nd.ni_vp, name, td->td_retval);
1994-05-24 10:09:53 +00:00
vput(nd.ni_vp);
return (error);
}
/*
* Return target name of a symbolic link.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct readlink_args {
char *path;
char *buf;
size_t count;
1994-05-24 10:09:53 +00:00
};
#endif
int
sys_readlink(td, uap)
struct thread *td;
register struct readlink_args /* {
char *path;
char *buf;
size_t count;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_readlinkat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->buf, UIO_USERSPACE, uap->count));
}
#ifndef _SYS_SYSPROTO_H_
struct readlinkat_args {
int fd;
char *path;
char *buf;
size_t bufsize;
};
#endif
int
sys_readlinkat(struct thread *td, struct readlinkat_args *uap)
{
return (kern_readlinkat(td, uap->fd, uap->path, UIO_USERSPACE,
uap->buf, UIO_USERSPACE, uap->bufsize));
}
int
kern_readlinkat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
char *buf, enum uio_seg bufseg, size_t count)
{
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct iovec aiov;
struct uio auio;
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
if (count > IOSIZE_MAX)
return (EINVAL);
NDINIT_AT(&nd, LOOKUP, NOFOLLOW | LOCKSHARED | LOCKLEAF | AUDITVNODE1,
pathseg, path, fd, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vp = nd.ni_vp;
#ifdef MAC
error = mac_vnode_check_readlink(td->td_ucred, vp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
vput(vp);
return (error);
}
#endif
1994-05-24 10:09:53 +00:00
if (vp->v_type != VLNK)
error = EINVAL;
else {
aiov.iov_base = buf;
aiov.iov_len = count;
1994-05-24 10:09:53 +00:00
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = 0;
auio.uio_rw = UIO_READ;
auio.uio_segflg = bufseg;
auio.uio_td = td;
auio.uio_resid = count;
error = VOP_READLINK(vp, &auio, td->td_ucred);
td->td_retval[0] = count - auio.uio_resid;
1994-05-24 10:09:53 +00:00
}
vput(vp);
return (error);
}
/*
* Common implementation code for chflags() and fchflags().
*/
static int
setfflags(td, vp, flags)
struct thread *td;
struct vnode *vp;
u_long flags;
{
struct mount *mp;
struct vattr vattr;
2013-09-05 00:19:30 +00:00
int error;
/* We can't support the value matching VNOVAL. */
if (flags == VNOVAL)
return (EOPNOTSUPP);
/*
* Prevent non-root users from setting flags on devices. When
* a device is reused, users can retain ownership of the device
* if they are allowed to set flags and programs assume that
* chown can't fail when done as root.
*/
if (vp->v_type == VCHR || vp->v_type == VBLK) {
error = priv_check(td, PRIV_VFS_CHFLAGS_DEV);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
}
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
return (error);
VATTR_NULL(&vattr);
vattr.va_flags = flags;
2013-03-01 21:57:02 +00:00
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
#ifdef MAC
error = mac_vnode_check_setflags(td->td_ucred, vp, vattr.va_flags);
if (error == 0)
#endif
error = VOP_SETATTR(vp, &vattr, td->td_ucred);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
return (error);
}
1994-05-24 10:09:53 +00:00
/*
* Change flags of a file given a path name.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct chflags_args {
const char *path;
u_long flags;
1994-05-24 10:09:53 +00:00
};
#endif
int
sys_chflags(td, uap)
struct thread *td;
register struct chflags_args /* {
const char *path;
u_long flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->flags, 0));
}
#ifndef _SYS_SYSPROTO_H_
struct chflagsat_args {
int fd;
const char *path;
u_long flags;
int atflag;
}
#endif
int
sys_chflagsat(struct thread *td, struct chflagsat_args *uap)
{
int fd = uap->fd;
const char *path = uap->path;
u_long flags = uap->flags;
int atflag = uap->atflag;
if (atflag & ~AT_SYMLINK_NOFOLLOW)
return (EINVAL);
return (kern_chflagsat(td, fd, path, UIO_USERSPACE, flags, atflag));
}
/*
* Same as chflags() but doesn't follow symlinks.
*/
int
sys_lchflags(td, uap)
struct thread *td;
register struct lchflags_args /* {
const char *path;
u_long flags;
} */ *uap;
{
return (kern_chflagsat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->flags, AT_SYMLINK_NOFOLLOW));
}
static int
kern_chflagsat(struct thread *td, int fd, const char *path,
enum uio_seg pathseg, u_long flags, int atflag)
{
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error, follow;
AUDIT_ARG_FFLAGS(flags);
follow = (atflag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_FCHFLAGS), td);
if ((error = namei(&nd)) != 0)
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = setfflags(td, nd.ni_vp, flags);
vrele(nd.ni_vp);
return (error);
}
1994-05-24 10:09:53 +00:00
/*
* Change flags of a file given a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fchflags_args {
int fd;
u_long flags;
1994-05-24 10:09:53 +00:00
};
#endif
int
sys_fchflags(td, uap)
struct thread *td;
register struct fchflags_args /* {
int fd;
u_long flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->fd);
AUDIT_ARG_FFLAGS(uap->flags);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, uap->fd,
cap_rights_init(&rights, CAP_FCHFLAGS), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
#ifdef AUDIT
vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(fp->f_vnode);
VOP_UNLOCK(fp->f_vnode, 0);
#endif
error = setfflags(td, fp->f_vnode, uap->flags);
fdrop(fp, td);
return (error);
}
/*
* Common implementation code for chmod(), lchmod() and fchmod().
*/
int
setfmode(td, cred, vp, mode)
struct thread *td;
struct ucred *cred;
struct vnode *vp;
int mode;
{
struct mount *mp;
struct vattr vattr;
2013-09-05 00:19:30 +00:00
int error;
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
return (error);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
VATTR_NULL(&vattr);
vattr.va_mode = mode & ALLPERMS;
#ifdef MAC
error = mac_vnode_check_setmode(cred, vp, vattr.va_mode);
if (error == 0)
#endif
error = VOP_SETATTR(vp, &vattr, cred);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Change mode of a file given path name.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct chmod_args {
char *path;
int mode;
};
#endif
int
sys_chmod(td, uap)
struct thread *td;
register struct chmod_args /* {
char *path;
int mode;
} */ *uap;
{
return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->mode, 0));
}
#ifndef _SYS_SYSPROTO_H_
struct fchmodat_args {
int dirfd;
char *path;
mode_t mode;
int flag;
}
#endif
int
sys_fchmodat(struct thread *td, struct fchmodat_args *uap)
{
int flag = uap->flag;
int fd = uap->fd;
char *path = uap->path;
mode_t mode = uap->mode;
if (flag & ~AT_SYMLINK_NOFOLLOW)
return (EINVAL);
return (kern_fchmodat(td, fd, path, UIO_USERSPACE, mode, flag));
}
/*
* Change mode of a file given path name (don't follow links.)
*/
#ifndef _SYS_SYSPROTO_H_
struct lchmod_args {
char *path;
int mode;
};
#endif
int
sys_lchmod(td, uap)
struct thread *td;
register struct lchmod_args /* {
char *path;
int mode;
} */ *uap;
{
return (kern_fchmodat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->mode, AT_SYMLINK_NOFOLLOW));
}
int
kern_fchmodat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
mode_t mode, int flag)
{
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
2013-09-05 00:19:30 +00:00
int error, follow;
AUDIT_ARG_MODE(mode);
follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
2013-03-16 22:36:24 +00:00
NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_FCHMOD), td);
if ((error = namei(&nd)) != 0)
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = setfmode(td, td->td_ucred, nd.ni_vp, mode);
vrele(nd.ni_vp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Change mode of a file given a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fchmod_args {
int fd;
int mode;
};
#endif
int
sys_fchmod(struct thread *td, struct fchmod_args *uap)
1994-05-24 10:09:53 +00:00
{
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->fd);
AUDIT_ARG_MODE(uap->mode);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHMOD), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
error = fo_chmod(fp, uap->mode, td->td_ucred, td);
fdrop(fp, td);
return (error);
}
/*
* Common implementation for chown(), lchown(), and fchown()
*/
int
setfown(td, cred, vp, uid, gid)
struct thread *td;
struct ucred *cred;
struct vnode *vp;
uid_t uid;
gid_t gid;
{
struct mount *mp;
struct vattr vattr;
2013-09-05 00:19:30 +00:00
int error;
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
return (error);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
VATTR_NULL(&vattr);
vattr.va_uid = uid;
vattr.va_gid = gid;
#ifdef MAC
error = mac_vnode_check_setowner(cred, vp, vattr.va_uid,
vattr.va_gid);
if (error == 0)
#endif
error = VOP_SETATTR(vp, &vattr, cred);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Set ownership given a path name.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct chown_args {
char *path;
int uid;
int gid;
};
#endif
int
sys_chown(td, uap)
struct thread *td;
register struct chown_args /* {
char *path;
int uid;
int gid;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE, uap->uid,
uap->gid, 0));
}
#ifndef _SYS_SYSPROTO_H_
struct fchownat_args {
int fd;
const char * path;
uid_t uid;
gid_t gid;
int flag;
};
#endif
int
sys_fchownat(struct thread *td, struct fchownat_args *uap)
{
int flag;
flag = uap->flag;
if (flag & ~AT_SYMLINK_NOFOLLOW)
return (EINVAL);
return (kern_fchownat(td, uap->fd, uap->path, UIO_USERSPACE, uap->uid,
uap->gid, uap->flag));
}
int
kern_fchownat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
int uid, int gid, int flag)
{
1994-05-24 10:09:53 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error, follow;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_OWNER(uid, gid);
follow = (flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW : FOLLOW;
NDINIT_ATRIGHTS(&nd, LOOKUP, follow | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_FCHOWN), td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = setfown(td, td->td_ucred, nd.ni_vp, uid, gid);
vrele(nd.ni_vp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Set ownership given a path name, do not cross symlinks.
*/
#ifndef _SYS_SYSPROTO_H_
struct lchown_args {
char *path;
int uid;
int gid;
};
#endif
int
sys_lchown(td, uap)
struct thread *td;
register struct lchown_args /* {
char *path;
int uid;
int gid;
} */ *uap;
{
return (kern_fchownat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->uid, uap->gid, AT_SYMLINK_NOFOLLOW));
}
1994-05-24 10:09:53 +00:00
/*
* Set ownership given a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fchown_args {
int fd;
int uid;
int gid;
};
#endif
int
sys_fchown(td, uap)
struct thread *td;
register struct fchown_args /* {
int fd;
int uid;
int gid;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->fd);
AUDIT_ARG_OWNER(uap->uid, uap->gid);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, uap->fd, cap_rights_init(&rights, CAP_FCHOWN), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
error = fo_chown(fp, uap->uid, uap->gid, td->td_ucred, td);
fdrop(fp, td);
return (error);
}
/*
* Common implementation code for utimes(), lutimes(), and futimes().
*/
static int
getutimes(usrtvp, tvpseg, tsp)
const struct timeval *usrtvp;
enum uio_seg tvpseg;
struct timespec *tsp;
{
struct timeval tv[2];
const struct timeval *tvp;
int error;
if (usrtvp == NULL) {
vfs_timestamp(&tsp[0]);
tsp[1] = tsp[0];
} else {
if (tvpseg == UIO_SYSSPACE) {
tvp = usrtvp;
} else {
if ((error = copyin(usrtvp, tv, sizeof(tv))) != 0)
return (error);
tvp = tv;
}
2006-01-04 07:50:54 +00:00
if (tvp[0].tv_usec < 0 || tvp[0].tv_usec >= 1000000 ||
tvp[1].tv_usec < 0 || tvp[1].tv_usec >= 1000000)
return (EINVAL);
TIMEVAL_TO_TIMESPEC(&tvp[0], &tsp[0]);
TIMEVAL_TO_TIMESPEC(&tvp[1], &tsp[1]);
}
return (0);
}
/*
* Common implementation code for futimens(), utimensat().
*/
#define UTIMENS_NULL 0x1
#define UTIMENS_EXIT 0x2
static int
getutimens(const struct timespec *usrtsp, enum uio_seg tspseg,
struct timespec *tsp, int *retflags)
{
struct timespec tsnow;
int error;
vfs_timestamp(&tsnow);
*retflags = 0;
if (usrtsp == NULL) {
tsp[0] = tsnow;
tsp[1] = tsnow;
*retflags |= UTIMENS_NULL;
return (0);
}
if (tspseg == UIO_SYSSPACE) {
tsp[0] = usrtsp[0];
tsp[1] = usrtsp[1];
} else if ((error = copyin(usrtsp, tsp, sizeof(*tsp) * 2)) != 0)
return (error);
if (tsp[0].tv_nsec == UTIME_OMIT && tsp[1].tv_nsec == UTIME_OMIT)
*retflags |= UTIMENS_EXIT;
if (tsp[0].tv_nsec == UTIME_NOW && tsp[1].tv_nsec == UTIME_NOW)
*retflags |= UTIMENS_NULL;
if (tsp[0].tv_nsec == UTIME_OMIT)
tsp[0].tv_sec = VNOVAL;
else if (tsp[0].tv_nsec == UTIME_NOW)
tsp[0] = tsnow;
else if (tsp[0].tv_nsec < 0 || tsp[0].tv_nsec >= 1000000000L)
return (EINVAL);
if (tsp[1].tv_nsec == UTIME_OMIT)
tsp[1].tv_sec = VNOVAL;
else if (tsp[1].tv_nsec == UTIME_NOW)
tsp[1] = tsnow;
else if (tsp[1].tv_nsec < 0 || tsp[1].tv_nsec >= 1000000000L)
return (EINVAL);
return (0);
}
/*
* Common implementation code for utimes(), lutimes(), futimes(), futimens(),
* and utimensat().
*/
static int
setutimes(td, vp, ts, numtimes, nullflag)
struct thread *td;
struct vnode *vp;
const struct timespec *ts;
int numtimes;
int nullflag;
{
struct mount *mp;
struct vattr vattr;
2013-09-05 00:19:30 +00:00
int error, setbirthtime;
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0)
return (error);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
setbirthtime = 0;
if (numtimes < 3 && !VOP_GETATTR(vp, &vattr, td->td_ucred) &&
timespeccmp(&ts[1], &vattr.va_birthtime, < ))
setbirthtime = 1;
VATTR_NULL(&vattr);
vattr.va_atime = ts[0];
vattr.va_mtime = ts[1];
if (setbirthtime)
vattr.va_birthtime = ts[1];
if (numtimes > 2)
vattr.va_birthtime = ts[2];
if (nullflag)
vattr.va_vaflags |= VA_UTIMES_NULL;
#ifdef MAC
error = mac_vnode_check_setutimes(td->td_ucred, vp, vattr.va_atime,
vattr.va_mtime);
#endif
if (error == 0)
error = VOP_SETATTR(vp, &vattr, td->td_ucred);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
return (error);
1994-05-24 10:09:53 +00:00
}
/*
* Set the access and modification times of a file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct utimes_args {
char *path;
struct timeval *tptr;
};
#endif
int
sys_utimes(td, uap)
struct thread *td;
register struct utimes_args /* {
char *path;
struct timeval *tptr;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_utimesat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->tptr, UIO_USERSPACE));
}
#ifndef _SYS_SYSPROTO_H_
struct futimesat_args {
int fd;
const char * path;
const struct timeval * times;
};
#endif
int
sys_futimesat(struct thread *td, struct futimesat_args *uap)
{
return (kern_utimesat(td, uap->fd, uap->path, UIO_USERSPACE,
uap->times, UIO_USERSPACE));
}
int
kern_utimesat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
struct timeval *tptr, enum uio_seg tptrseg)
{
1994-05-24 10:09:53 +00:00
struct nameidata nd;
struct timespec ts[2];
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
1994-05-24 10:09:53 +00:00
if ((error = getutimes(tptr, tptrseg, ts)) != 0)
return (error);
NDINIT_ATRIGHTS(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, fd,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_FUTIMES), td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
vrele(nd.ni_vp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Set the access and modification times of a file.
*/
#ifndef _SYS_SYSPROTO_H_
struct lutimes_args {
char *path;
struct timeval *tptr;
};
#endif
int
sys_lutimes(td, uap)
struct thread *td;
register struct lutimes_args /* {
char *path;
struct timeval *tptr;
} */ *uap;
{
return (kern_lutimes(td, uap->path, UIO_USERSPACE, uap->tptr,
UIO_USERSPACE));
}
int
kern_lutimes(struct thread *td, char *path, enum uio_seg pathseg,
struct timeval *tptr, enum uio_seg tptrseg)
{
struct timespec ts[2];
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
if ((error = getutimes(tptr, tptrseg, ts)) != 0)
return (error);
NDINIT(&nd, LOOKUP, NOFOLLOW | AUDITVNODE1, pathseg, path, td);
if ((error = namei(&nd)) != 0)
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
error = setutimes(td, nd.ni_vp, ts, 2, tptr == NULL);
vrele(nd.ni_vp);
return (error);
}
/*
* Set the access and modification times of a file.
*/
#ifndef _SYS_SYSPROTO_H_
struct futimes_args {
int fd;
struct timeval *tptr;
};
#endif
int
sys_futimes(td, uap)
struct thread *td;
register struct futimes_args /* {
int fd;
struct timeval *tptr;
} */ *uap;
{
return (kern_futimes(td, uap->fd, uap->tptr, UIO_USERSPACE));
}
int
kern_futimes(struct thread *td, int fd, struct timeval *tptr,
enum uio_seg tptrseg)
{
struct timespec ts[2];
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
AUDIT_ARG_FD(fd);
2013-09-05 00:19:30 +00:00
error = getutimes(tptr, tptrseg, ts);
if (error != 0)
return (error);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, fd,
cap_rights_init(&rights, CAP_FUTIMES), &fp);
if (error != 0)
return (error);
#ifdef AUDIT
vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(fp->f_vnode);
VOP_UNLOCK(fp->f_vnode, 0);
#endif
error = setutimes(td, fp->f_vnode, ts, 2, tptr == NULL);
fdrop(fp, td);
return (error);
}
int
sys_futimens(struct thread *td, struct futimens_args *uap)
{
return (kern_futimens(td, uap->fd, uap->times, UIO_USERSPACE));
}
int
kern_futimens(struct thread *td, int fd, struct timespec *tptr,
enum uio_seg tptrseg)
{
struct timespec ts[2];
struct file *fp;
cap_rights_t rights;
int error, flags;
AUDIT_ARG_FD(fd);
error = getutimens(tptr, tptrseg, ts, &flags);
if (error != 0)
return (error);
if (flags & UTIMENS_EXIT)
return (0);
error = getvnode(td->td_proc->p_fd, fd,
cap_rights_init(&rights, CAP_FUTIMES), &fp);
if (error != 0)
return (error);
#ifdef AUDIT
vn_lock(fp->f_vnode, LK_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(fp->f_vnode);
VOP_UNLOCK(fp->f_vnode, 0);
#endif
error = setutimes(td, fp->f_vnode, ts, 2, flags & UTIMENS_NULL);
fdrop(fp, td);
return (error);
}
int
sys_utimensat(struct thread *td, struct utimensat_args *uap)
{
return (kern_utimensat(td, uap->fd, uap->path, UIO_USERSPACE,
uap->times, UIO_USERSPACE, uap->flag));
}
int
kern_utimensat(struct thread *td, int fd, char *path, enum uio_seg pathseg,
struct timespec *tptr, enum uio_seg tptrseg, int flag)
{
struct nameidata nd;
struct timespec ts[2];
cap_rights_t rights;
int error, flags;
if (flag & ~AT_SYMLINK_NOFOLLOW)
return (EINVAL);
if ((error = getutimens(tptr, tptrseg, ts, &flags)) != 0)
return (error);
NDINIT_ATRIGHTS(&nd, LOOKUP, ((flag & AT_SYMLINK_NOFOLLOW) ? NOFOLLOW :
FOLLOW) | AUDITVNODE1, pathseg, path, fd,
cap_rights_init(&rights, CAP_FUTIMES), td);
if ((error = namei(&nd)) != 0)
return (error);
/*
* We are allowed to call namei() regardless of 2xUTIME_OMIT.
* POSIX states:
* "If both tv_nsec fields are UTIME_OMIT... EACCESS may be detected."
* "Search permission is denied by a component of the path prefix."
*/
NDFREE(&nd, NDF_ONLY_PNBUF);
if ((flags & UTIMENS_EXIT) == 0)
error = setutimes(td, nd.ni_vp, ts, 2, flags & UTIMENS_NULL);
vrele(nd.ni_vp);
return (error);
}
1994-05-24 10:09:53 +00:00
/*
* Truncate a file given its path name.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct truncate_args {
char *path;
int pad;
off_t length;
};
#endif
int
sys_truncate(td, uap)
struct thread *td;
register struct truncate_args /* {
char *path;
int pad;
off_t length;
} */ *uap;
{
return (kern_truncate(td, uap->path, UIO_USERSPACE, uap->length));
}
int
kern_truncate(struct thread *td, char *path, enum uio_seg pathseg, off_t length)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
struct vnode *vp;
void *rl_cookie;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
struct nameidata nd;
int error;
1994-05-24 10:09:53 +00:00
if (length < 0)
return(EINVAL);
NDINIT(&nd, LOOKUP, FOLLOW | AUDITVNODE1, pathseg, path, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = nd.ni_vp;
rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX);
if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) {
vn_rangelock_unlock(vp, rl_cookie);
vrele(vp);
return (error);
}
NDFREE(&nd, NDF_ONLY_PNBUF);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1994-05-24 10:09:53 +00:00
if (vp->v_type == VDIR)
error = EISDIR;
#ifdef MAC
else if ((error = mac_vnode_check_write(td->td_ucred, NOCRED, vp))) {
}
#endif
1994-05-24 10:09:53 +00:00
else if ((error = vn_writechk(vp)) == 0 &&
(error = VOP_ACCESS(vp, VWRITE, td->td_ucred, td)) == 0) {
1994-05-24 10:09:53 +00:00
VATTR_NULL(&vattr);
vattr.va_size = length;
error = VOP_SETATTR(vp, &vattr, td->td_ucred);
1994-05-24 10:09:53 +00:00
}
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
vn_rangelock_unlock(vp, rl_cookie);
vrele(vp);
1994-05-24 10:09:53 +00:00
return (error);
}
#if defined(COMPAT_43)
1994-05-24 10:09:53 +00:00
/*
* Truncate a file given its path name.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct otruncate_args {
char *path;
long length;
};
#endif
int
otruncate(td, uap)
struct thread *td;
register struct otruncate_args /* {
char *path;
long length;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct truncate_args /* {
char *path;
int pad;
off_t length;
} */ nuap;
2002-12-14 01:56:26 +00:00
nuap.path = uap->path;
nuap.length = uap->length;
return (sys_truncate(td, &nuap));
1994-05-24 10:09:53 +00:00
}
#endif /* COMPAT_43 */
1994-05-24 10:09:53 +00:00
#if defined(COMPAT_FREEBSD6)
/* Versions with the pad argument */
int
freebsd6_truncate(struct thread *td, struct freebsd6_truncate_args *uap)
{
struct truncate_args ouap;
ouap.path = uap->path;
ouap.length = uap->length;
return (sys_truncate(td, &ouap));
}
int
freebsd6_ftruncate(struct thread *td, struct freebsd6_ftruncate_args *uap)
{
struct ftruncate_args ouap;
ouap.fd = uap->fd;
ouap.length = uap->length;
return (sys_ftruncate(td, &ouap));
}
#endif
1994-05-24 10:09:53 +00:00
/*
* Sync an open file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct fsync_args {
int fd;
};
#endif
int
sys_fsync(td, uap)
struct thread *td;
struct fsync_args /* {
int fd;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct vnode *vp;
struct mount *mp;
1994-05-24 10:09:53 +00:00
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error, lock_flags;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(uap->fd);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, uap->fd,
cap_rights_init(&rights, CAP_FSYNC), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = fp->f_vnode;
2013-09-05 00:19:30 +00:00
error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
if (error != 0)
goto drop;
if (MNT_SHARED_WRITES(mp) ||
((mp == NULL) && MNT_SHARED_WRITES(vp->v_mount))) {
lock_flags = LK_SHARED;
} else {
lock_flags = LK_EXCLUSIVE;
}
vn_lock(vp, lock_flags | LK_RETRY);
AUDIT_ARG_VNODE1(vp);
if (vp->v_object != NULL) {
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WLOCK(vp->v_object);
vm_object_page_clean(vp->v_object, 0, 0, 0);
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
VM_OBJECT_WUNLOCK(vp->v_object);
}
error = VOP_FSYNC(vp, MNT_WAIT, td);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
drop:
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Rename files. Source and destination must either both be directories, or
* both not be directories. If target is a directory, it must be empty.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct rename_args {
char *from;
char *to;
};
#endif
int
sys_rename(td, uap)
struct thread *td;
register struct rename_args /* {
char *from;
char *to;
} */ *uap;
{
return (kern_renameat(td, AT_FDCWD, uap->from, AT_FDCWD,
uap->to, UIO_USERSPACE));
}
#ifndef _SYS_SYSPROTO_H_
struct renameat_args {
int oldfd;
char *old;
int newfd;
char *new;
};
#endif
int
sys_renameat(struct thread *td, struct renameat_args *uap)
{
return (kern_renameat(td, uap->oldfd, uap->old, uap->newfd, uap->new,
UIO_USERSPACE));
}
int
kern_renameat(struct thread *td, int oldfd, char *old, int newfd, char *new,
enum uio_seg pathseg)
1994-05-24 10:09:53 +00:00
{
struct mount *mp = NULL;
struct vnode *tvp, *fvp, *tdvp;
1994-05-24 10:09:53 +00:00
struct nameidata fromnd, tond;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
again:
bwillwrite();
#ifdef MAC
NDINIT_ATRIGHTS(&fromnd, DELETE, LOCKPARENT | LOCKLEAF | SAVESTART |
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
AUDITVNODE1, pathseg, old, oldfd,
cap_rights_init(&rights, CAP_RENAMEAT), td);
#else
NDINIT_ATRIGHTS(&fromnd, DELETE, WANTPARENT | SAVESTART | AUDITVNODE1,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
pathseg, old, oldfd, cap_rights_init(&rights, CAP_RENAMEAT), td);
#endif
if ((error = namei(&fromnd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
#ifdef MAC
error = mac_vnode_check_rename_from(td->td_ucred, fromnd.ni_dvp,
fromnd.ni_vp, &fromnd.ni_cnd);
VOP_UNLOCK(fromnd.ni_dvp, 0);
if (fromnd.ni_dvp != fromnd.ni_vp)
VOP_UNLOCK(fromnd.ni_vp, 0);
#endif
1994-05-24 10:09:53 +00:00
fvp = fromnd.ni_vp;
NDINIT_ATRIGHTS(&tond, RENAME, LOCKPARENT | LOCKLEAF | NOCACHE |
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
SAVESTART | AUDITVNODE2, pathseg, new, newfd,
cap_rights_init(&rights, CAP_LINKAT), td);
if (fromnd.ni_vp->v_type == VDIR)
tond.ni_cnd.cn_flags |= WILLBEDIR;
if ((error = namei(&tond)) != 0) {
/* Translate error code for rename("dir1", "dir2/."). */
if (error == EISDIR && fvp->v_type == VDIR)
error = EINVAL;
NDFREE(&fromnd, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
vrele(fromnd.ni_dvp);
vrele(fvp);
goto out1;
}
tdvp = tond.ni_dvp;
tvp = tond.ni_vp;
error = vn_start_write(fvp, &mp, V_NOWAIT);
if (error != 0) {
NDFREE(&fromnd, NDF_ONLY_PNBUF);
NDFREE(&tond, NDF_ONLY_PNBUF);
if (tvp != NULL)
vput(tvp);
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
vrele(fromnd.ni_dvp);
vrele(fvp);
vrele(tond.ni_startdir);
if (fromnd.ni_startdir != NULL)
vrele(fromnd.ni_startdir);
error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH);
if (error != 0)
return (error);
goto again;
}
1994-05-24 10:09:53 +00:00
if (tvp != NULL) {
if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
error = ENOTDIR;
goto out;
} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
error = EISDIR;
goto out;
}
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
#ifdef CAPABILITIES
if (newfd != AT_FDCWD) {
/*
* If the target already exists we require CAP_UNLINKAT
* from 'newfd'.
*/
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = cap_check(&tond.ni_filecaps.fc_rights,
cap_rights_init(&rights, CAP_UNLINKAT));
if (error != 0)
goto out;
}
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
#endif
1994-05-24 10:09:53 +00:00
}
if (fvp == tdvp) {
1994-05-24 10:09:53 +00:00
error = EINVAL;
goto out;
}
1994-05-24 10:09:53 +00:00
/*
* If the source is the same as the destination (that is, if they
* are links to the same vnode), then there is nothing to do.
1994-05-24 10:09:53 +00:00
*/
if (fvp == tvp)
1994-05-24 10:09:53 +00:00
error = -1;
#ifdef MAC
else
error = mac_vnode_check_rename_to(td->td_ucred, tdvp,
tond.ni_vp, fromnd.ni_dvp == tdvp, &tond.ni_cnd);
#endif
1994-05-24 10:09:53 +00:00
out:
2013-09-05 00:19:30 +00:00
if (error == 0) {
1994-05-24 10:09:53 +00:00
error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2013-09-05 00:19:30 +00:00
tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
NDFREE(&fromnd, NDF_ONLY_PNBUF);
NDFREE(&tond, NDF_ONLY_PNBUF);
1994-05-24 10:09:53 +00:00
} else {
NDFREE(&fromnd, NDF_ONLY_PNBUF);
NDFREE(&tond, NDF_ONLY_PNBUF);
2013-09-05 00:19:30 +00:00
if (tvp != NULL)
vput(tvp);
1994-05-24 10:09:53 +00:00
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
vrele(fromnd.ni_dvp);
vrele(fvp);
}
vrele(tond.ni_startdir);
vn_finished_write(mp);
out1:
1994-05-24 10:09:53 +00:00
if (fromnd.ni_startdir)
vrele(fromnd.ni_startdir);
if (error == -1)
return (0);
return (error);
}
/*
* Make a directory file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct mkdir_args {
char *path;
int mode;
};
#endif
int
sys_mkdir(td, uap)
struct thread *td;
register struct mkdir_args /* {
char *path;
int mode;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_mkdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE,
uap->mode));
}
#ifndef _SYS_SYSPROTO_H_
struct mkdirat_args {
int fd;
char *path;
mode_t mode;
};
#endif
int
sys_mkdirat(struct thread *td, struct mkdirat_args *uap)
{
return (kern_mkdirat(td, uap->fd, uap->path, UIO_USERSPACE, uap->mode));
}
int
kern_mkdirat(struct thread *td, int fd, char *path, enum uio_seg segflg,
int mode)
{
struct mount *mp;
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
2013-09-05 00:19:30 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_MODE(mode);
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, CREATE, LOCKPARENT | SAVENAME | AUDITVNODE1 |
NOCACHE, segflg, path, fd, cap_rights_init(&rights, CAP_MKDIRAT),
td);
nd.ni_cnd.cn_flags |= WILLBEDIR;
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = nd.ni_vp;
if (vp != NULL) {
NDFREE(&nd, NDF_ONLY_PNBUF);
/*
* XXX namei called with LOCKPARENT but not LOCKLEAF has
* the strange behaviour of leaving the vnode unlocked
* if the target is the same vnode as the parent.
*/
if (vp == nd.ni_dvp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
vrele(vp);
1994-05-24 10:09:53 +00:00
return (EEXIST);
}
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
1994-05-24 10:09:53 +00:00
VATTR_NULL(&vattr);
vattr.va_type = VDIR;
vattr.va_mode = (mode & ACCESSPERMS) &~ td->td_proc->p_fd->fd_cmask;
#ifdef MAC
error = mac_vnode_check_create(td->td_ucred, nd.ni_dvp, &nd.ni_cnd,
&vattr);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
1994-05-24 10:09:53 +00:00
error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
#ifdef MAC
out:
#endif
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(nd.ni_dvp);
2013-09-05 00:19:30 +00:00
if (error == 0)
1994-05-24 10:09:53 +00:00
vput(nd.ni_vp);
vn_finished_write(mp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Remove a directory file.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct rmdir_args {
char *path;
};
#endif
int
sys_rmdir(td, uap)
struct thread *td;
struct rmdir_args /* {
char *path;
} */ *uap;
{
return (kern_rmdirat(td, AT_FDCWD, uap->path, UIO_USERSPACE));
}
int
kern_rmdirat(struct thread *td, int fd, char *path, enum uio_seg pathseg)
1994-05-24 10:09:53 +00:00
{
struct mount *mp;
struct vnode *vp;
2013-09-05 00:19:30 +00:00
struct nameidata nd;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
restart:
bwillwrite();
NDINIT_ATRIGHTS(&nd, DELETE, LOCKPARENT | LOCKLEAF | AUDITVNODE1,
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
pathseg, path, fd, cap_rights_init(&rights, CAP_UNLINKAT), td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = nd.ni_vp;
if (vp->v_type != VDIR) {
error = ENOTDIR;
goto out;
}
/*
* No rmdir "." please.
*/
if (nd.ni_dvp == vp) {
error = EINVAL;
goto out;
}
/*
* The root of a mounted filesystem cannot be deleted.
*/
if (vp->v_vflag & VV_ROOT) {
1994-05-24 10:09:53 +00:00
error = EBUSY;
goto out;
1994-05-24 10:09:53 +00:00
}
#ifdef MAC
error = mac_vnode_check_unlink(td->td_ucred, nd.ni_dvp, vp,
&nd.ni_cnd);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
if (vn_start_write(nd.ni_dvp, &mp, V_NOWAIT) != 0) {
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(vp);
if (nd.ni_dvp == vp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
if ((error = vn_start_write(NULL, &mp, V_XSLEEP | PCATCH)) != 0)
return (error);
goto restart;
}
vfs_notify_upper(vp, VFS_NOTIFY_UPPER_UNLINK);
error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
vn_finished_write(mp);
out:
NDFREE(&nd, NDF_ONLY_PNBUF);
vput(vp);
if (nd.ni_dvp == vp)
vrele(nd.ni_dvp);
else
vput(nd.ni_dvp);
1994-05-24 10:09:53 +00:00
return (error);
}
#ifdef COMPAT_43
/*
2002-05-16 21:28:32 +00:00
* Read a block of directory entries in a filesystem independent format.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct ogetdirentries_args {
int fd;
char *buf;
u_int count;
long *basep;
};
#endif
int
ogetdirentries(struct thread *td, struct ogetdirentries_args *uap)
{
long loff;
int error;
error = kern_ogetdirentries(td, uap, &loff);
if (error == 0)
error = copyout(&loff, uap->basep, sizeof(long));
return (error);
}
int
kern_ogetdirentries(struct thread *td, struct ogetdirentries_args *uap,
long *ploff)
1994-05-24 10:09:53 +00:00
{
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct file *fp;
struct uio auio, kuio;
struct iovec aiov, kiov;
struct dirent *dp, *edp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
caddr_t dirbuf;
int error, eofflag, readcnt;
1994-05-24 10:09:53 +00:00
long loff;
off_t foffset;
1994-05-24 10:09:53 +00:00
/* XXX arbitrary sanity limit on `count'. */
2002-12-14 01:56:26 +00:00
if (uap->count > 64 * 1024)
return (EINVAL);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, uap->fd,
cap_rights_init(&rights, CAP_READ), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
if ((fp->f_flag & FREAD) == 0) {
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (EBADF);
}
vp = fp->f_vnode;
foffset = foffset_lock(fp, 0);
unionread:
if (vp->v_type != VDIR) {
foffset_unlock(fp, foffset, 0);
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (EINVAL);
}
2002-12-14 01:56:26 +00:00
aiov.iov_base = uap->buf;
aiov.iov_len = uap->count;
1994-05-24 10:09:53 +00:00
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_rw = UIO_READ;
auio.uio_segflg = UIO_USERSPACE;
auio.uio_td = td;
2002-12-14 01:56:26 +00:00
auio.uio_resid = uap->count;
vn_lock(vp, LK_SHARED | LK_RETRY);
loff = auio.uio_offset = foffset;
#ifdef MAC
error = mac_vnode_check_readdir(td->td_ucred, vp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
VOP_UNLOCK(vp, 0);
foffset_unlock(fp, foffset, FOF_NOUPDATE);
fdrop(fp, td);
return (error);
}
#endif
1994-05-24 10:09:53 +00:00
# if (BYTE_ORDER != LITTLE_ENDIAN)
if (vp->v_mount->mnt_maxsymlinklen <= 0) {
error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag,
NULL, NULL);
foffset = auio.uio_offset;
1994-05-24 10:09:53 +00:00
} else
# endif
{
kuio = auio;
kuio.uio_iov = &kiov;
kuio.uio_segflg = UIO_SYSSPACE;
2002-12-14 01:56:26 +00:00
kiov.iov_len = uap->count;
dirbuf = malloc(uap->count, M_TEMP, M_WAITOK);
1994-05-24 10:09:53 +00:00
kiov.iov_base = dirbuf;
error = VOP_READDIR(vp, &kuio, fp->f_cred, &eofflag,
NULL, NULL);
foffset = kuio.uio_offset;
1994-05-24 10:09:53 +00:00
if (error == 0) {
2002-12-14 01:56:26 +00:00
readcnt = uap->count - kuio.uio_resid;
1994-05-24 10:09:53 +00:00
edp = (struct dirent *)&dirbuf[readcnt];
for (dp = (struct dirent *)dirbuf; dp < edp; ) {
# if (BYTE_ORDER == LITTLE_ENDIAN)
/*
* The expected low byte of
* dp->d_namlen is our dp->d_type.
* The high MBZ byte of dp->d_namlen
* is our dp->d_namlen.
*/
dp->d_type = dp->d_namlen;
dp->d_namlen = 0;
# else
/*
* The dp->d_type is the high byte
* of the expected dp->d_namlen,
* so must be zero'ed.
*/
dp->d_type = 0;
# endif
if (dp->d_reclen > 0) {
dp = (struct dirent *)
((char *)dp + dp->d_reclen);
} else {
error = EIO;
break;
}
}
if (dp >= edp)
error = uiomove(dirbuf, readcnt, &auio);
}
free(dirbuf, M_TEMP);
1994-05-24 10:09:53 +00:00
}
2013-09-05 00:19:30 +00:00
if (error != 0) {
VOP_UNLOCK(vp, 0);
foffset_unlock(fp, foffset, 0);
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
if (uap->count == auio.uio_resid &&
(vp->v_vflag & VV_ROOT) &&
(vp->v_mount->mnt_flag & MNT_UNION)) {
struct vnode *tvp = vp;
vp = vp->v_mount->mnt_vnodecovered;
VREF(vp);
fp->f_vnode = vp;
fp->f_data = vp;
foffset = 0;
vput(tvp);
goto unionread;
}
VOP_UNLOCK(vp, 0);
foffset_unlock(fp, foffset, 0);
fdrop(fp, td);
2002-12-14 01:56:26 +00:00
td->td_retval[0] = uap->count - auio.uio_resid;
if (error == 0)
*ploff = loff;
1994-05-24 10:09:53 +00:00
return (error);
}
#endif /* COMPAT_43 */
1994-05-24 10:09:53 +00:00
/*
2002-05-16 21:28:32 +00:00
* Read a block of directory entries in a filesystem independent format.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct getdirentries_args {
int fd;
char *buf;
u_int count;
long *basep;
};
#endif
int
sys_getdirentries(td, uap)
struct thread *td;
register struct getdirentries_args /* {
int fd;
char *buf;
u_int count;
long *basep;
} */ *uap;
{
long base;
int error;
error = kern_getdirentries(td, uap->fd, uap->buf, uap->count, &base,
NULL, UIO_USERSPACE);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
if (uap->basep != NULL)
error = copyout(&base, uap->basep, sizeof(long));
return (error);
}
int
kern_getdirentries(struct thread *td, int fd, char *buf, u_int count,
long *basep, ssize_t *residp, enum uio_seg bufseg)
1994-05-24 10:09:53 +00:00
{
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct file *fp;
struct uio auio;
struct iovec aiov;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
long loff;
int error, eofflag;
off_t foffset;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(fd);
if (count > IOSIZE_MAX)
return (EINVAL);
auio.uio_resid = count;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = getvnode(td->td_proc->p_fd, fd,
cap_rights_init(&rights, CAP_READ), &fp);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
if ((fp->f_flag & FREAD) == 0) {
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (EBADF);
}
vp = fp->f_vnode;
foffset = foffset_lock(fp, 0);
1994-05-24 10:09:53 +00:00
unionread:
if (vp->v_type != VDIR) {
error = EINVAL;
goto fail;
}
aiov.iov_base = buf;
aiov.iov_len = count;
1994-05-24 10:09:53 +00:00
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_rw = UIO_READ;
auio.uio_segflg = bufseg;
auio.uio_td = td;
vn_lock(vp, LK_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(vp);
loff = auio.uio_offset = foffset;
#ifdef MAC
error = mac_vnode_check_readdir(td->td_ucred, vp);
if (error == 0)
#endif
error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, NULL,
NULL);
foffset = auio.uio_offset;
2013-09-05 00:19:30 +00:00
if (error != 0) {
VOP_UNLOCK(vp, 0);
goto fail;
}
if (count == auio.uio_resid &&
(vp->v_vflag & VV_ROOT) &&
(vp->v_mount->mnt_flag & MNT_UNION)) {
struct vnode *tvp = vp;
2013-09-05 00:19:30 +00:00
vp = vp->v_mount->mnt_vnodecovered;
VREF(vp);
fp->f_vnode = vp;
fp->f_data = vp;
foffset = 0;
vput(tvp);
goto unionread;
}
VOP_UNLOCK(vp, 0);
*basep = loff;
if (residp != NULL)
*residp = auio.uio_resid;
td->td_retval[0] = count - auio.uio_resid;
fail:
foffset_unlock(fp, foffset, 0);
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct getdents_args {
int fd;
char *buf;
size_t count;
};
#endif
int
sys_getdents(td, uap)
struct thread *td;
register struct getdents_args /* {
int fd;
char *buf;
u_int count;
} */ *uap;
{
struct getdirentries_args ap;
2013-09-05 00:19:30 +00:00
ap.fd = uap->fd;
ap.buf = uap->buf;
ap.count = uap->count;
ap.basep = NULL;
return (sys_getdirentries(td, &ap));
}
1994-05-24 10:09:53 +00:00
/*
* Set the mode mask for creation of filesystem nodes.
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct umask_args {
int newmask;
};
#endif
int
sys_umask(td, uap)
struct thread *td;
struct umask_args /* {
int newmask;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct filedesc *fdp;
1994-05-24 10:09:53 +00:00
fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
td->td_retval[0] = fdp->fd_cmask;
2002-12-14 01:56:26 +00:00
fdp->fd_cmask = uap->newmask & ALLPERMS;
FILEDESC_XUNLOCK(fdp);
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Void all references to file by ripping underlying filesystem away from
* vnode.
1994-05-24 10:09:53 +00:00
*/
#ifndef _SYS_SYSPROTO_H_
1994-05-24 10:09:53 +00:00
struct revoke_args {
char *path;
};
#endif
int
sys_revoke(td, uap)
struct thread *td;
register struct revoke_args /* {
char *path;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct vnode *vp;
1994-05-24 10:09:53 +00:00
struct vattr vattr;
struct nameidata nd;
2013-09-05 00:19:30 +00:00
int error;
1994-05-24 10:09:53 +00:00
NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
uap->path, td);
if ((error = namei(&nd)) != 0)
1994-05-24 10:09:53 +00:00
return (error);
vp = nd.ni_vp;
NDFREE(&nd, NDF_ONLY_PNBUF);
if (vp->v_type != VCHR || vp->v_rdev == NULL) {
error = EINVAL;
goto out;
}
#ifdef MAC
error = mac_vnode_check_revoke(td->td_ucred, vp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
error = VOP_GETATTR(vp, &vattr, td->td_ucred);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
if (td->td_ucred->cr_uid != vattr.va_uid) {
error = priv_check(td, PRIV_VFS_ADMIN);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
}
if (vcount(vp) > 1)
VOP_REVOKE(vp, REVOKEALL);
out:
vput(vp);
1994-05-24 10:09:53 +00:00
return (error);
}
/*
* Convert a user file descriptor to a kernel file entry and check that, if it
* is a capability, the correct rights are present. A reference on the file
* entry is held upon returning.
1994-05-24 10:09:53 +00:00
*/
int
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
getvnode(struct filedesc *fdp, int fd, cap_rights_t *rightsp, struct file **fpp)
1994-05-24 10:09:53 +00:00
{
struct file *fp;
int error;
1994-05-24 10:09:53 +00:00
error = fget_unlocked(fdp, fd, rightsp, &fp, NULL);
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
if (error != 0)
return (error);
/*
* The file could be not of the vnode type, or it may be not
* yet fully initialized, in which case the f_vnode pointer
* may be set, but f_ops is still badfileops. E.g.,
* devfs_open() transiently create such situation to
* facilitate csw d_fdopen().
*
* Dupfdopen() handling in kern_openat() installs the
* half-baked file into the process descriptor table, allowing
* other thread to dereference it. Guard against the race by
* checking f_ops.
*/
if (fp->f_vnode == NULL || fp->f_ops == &badfileops) {
fdrop(fp, curthread);
return (EINVAL);
}
1994-05-24 10:09:53 +00:00
*fpp = fp;
return (0);
1994-05-24 10:09:53 +00:00
}
/*
* Get an (NFS) file handle.
*/
#ifndef _SYS_SYSPROTO_H_
struct lgetfh_args {
char *fname;
fhandle_t *fhp;
};
#endif
int
sys_lgetfh(td, uap)
struct thread *td;
register struct lgetfh_args *uap;
{
struct nameidata nd;
fhandle_t fh;
register struct vnode *vp;
int error;
error = priv_check(td, PRIV_VFS_GETFH);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
NDINIT(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
uap->fname, td);
error = namei(&nd);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
vp = nd.ni_vp;
bzero(&fh, sizeof(fh));
fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
error = VOP_VPTOFH(vp, &fh.fh_fid);
vput(vp);
2013-09-05 00:19:30 +00:00
if (error == 0)
error = copyout(&fh, uap->fhp, sizeof (fh));
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct getfh_args {
char *fname;
fhandle_t *fhp;
};
#endif
int
sys_getfh(td, uap)
struct thread *td;
register struct getfh_args *uap;
{
struct nameidata nd;
fhandle_t fh;
register struct vnode *vp;
int error;
error = priv_check(td, PRIV_VFS_GETFH);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | AUDITVNODE1, UIO_USERSPACE,
uap->fname, td);
error = namei(&nd);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
NDFREE(&nd, NDF_ONLY_PNBUF);
vp = nd.ni_vp;
bzero(&fh, sizeof(fh));
fh.fh_fsid = vp->v_mount->mnt_stat.f_fsid;
error = VOP_VPTOFH(vp, &fh.fh_fid);
vput(vp);
2013-09-05 00:19:30 +00:00
if (error == 0)
error = copyout(&fh, uap->fhp, sizeof (fh));
return (error);
}
/*
* syscall for the rpc.lockd to use to translate a NFS file handle into an
* open descriptor.
*
* warning: do not remove the priv_check() call or this becomes one giant
* security hole.
*/
#ifndef _SYS_SYSPROTO_H_
struct fhopen_args {
const struct fhandle *u_fhp;
int flags;
};
#endif
int
sys_fhopen(td, uap)
struct thread *td;
struct fhopen_args /* {
const struct fhandle *u_fhp;
int flags;
} */ *uap;
{
struct mount *mp;
struct vnode *vp;
struct fhandle fhp;
struct file *fp;
int fmode, error;
int indx;
error = priv_check(td, PRIV_VFS_FHOPEN);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
indx = -1;
2002-12-14 01:56:26 +00:00
fmode = FFLAGS(uap->flags);
/* why not allow a non-read/write open for our lockd? */
if (((fmode & (FREAD | FWRITE)) == 0) || (fmode & O_CREAT))
return (EINVAL);
2002-12-14 01:56:26 +00:00
error = copyin(uap->u_fhp, &fhp, sizeof(fhp));
2013-09-05 00:19:30 +00:00
if (error != 0)
return(error);
/* find the mount point */
mp = vfs_busyfs(&fhp.fh_fsid);
if (mp == NULL)
return (ESTALE);
/* now give me my vnode, it gets returned to me locked */
error = VFS_FHTOVP(mp, &fhp.fh_fid, LK_EXCLUSIVE, &vp);
vfs_unbusy(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
error = falloc_noinstall(td, &fp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
vput(vp);
return (error);
}
/*
* An extra reference on `fp' has been held for us by
* falloc_noinstall().
*/
#ifdef INVARIANTS
td->td_dupfd = -1;
#endif
error = vn_open_vnode(vp, fmode, td->td_ucred, td, fp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
KASSERT(fp->f_ops == &badfileops,
("VOP_OPEN in fhopen() set f_ops"));
KASSERT(td->td_dupfd < 0,
("fhopen() encountered fdopen()"));
vput(vp);
goto bad;
}
#ifdef INVARIANTS
td->td_dupfd = 0;
#endif
fp->f_vnode = vp;
fp->f_seqcount = 1;
finit(fp, (fmode & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE, vp,
&vnops);
VOP_UNLOCK(vp, 0);
2013-09-05 00:19:30 +00:00
if ((fmode & O_TRUNC) != 0) {
error = fo_truncate(fp, 0, td->td_ucred, td);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto bad;
}
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
error = finstall(td, fp, &indx, fmode, NULL);
bad:
fdrop(fp, td);
td->td_retval[0] = indx;
return (error);
}
/*
* Stat an (NFS) file handle.
*/
#ifndef _SYS_SYSPROTO_H_
struct fhstat_args {
struct fhandle *u_fhp;
struct stat *sb;
};
#endif
int
sys_fhstat(td, uap)
struct thread *td;
register struct fhstat_args /* {
struct fhandle *u_fhp;
struct stat *sb;
} */ *uap;
{
struct stat sb;
struct fhandle fh;
int error;
error = copyin(uap->u_fhp, &fh, sizeof(fh));
if (error != 0)
return (error);
error = kern_fhstat(td, fh, &sb);
2013-09-05 00:19:30 +00:00
if (error == 0)
error = copyout(&sb, uap->sb, sizeof(sb));
return (error);
}
int
kern_fhstat(struct thread *td, struct fhandle fh, struct stat *sb)
{
struct mount *mp;
struct vnode *vp;
int error;
error = priv_check(td, PRIV_VFS_FHSTAT);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
return (ESTALE);
error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
vfs_unbusy(mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
vput(vp);
return (error);
}
/*
* Implement fstatfs() for (NFS) file handles.
*/
#ifndef _SYS_SYSPROTO_H_
struct fhstatfs_args {
struct fhandle *u_fhp;
struct statfs *buf;
};
#endif
int
sys_fhstatfs(td, uap)
struct thread *td;
struct fhstatfs_args /* {
struct fhandle *u_fhp;
struct statfs *buf;
} */ *uap;
{
struct statfs sf;
fhandle_t fh;
int error;
error = copyin(uap->u_fhp, &fh, sizeof(fhandle_t));
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
error = kern_fhstatfs(td, fh, &sf);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
return (copyout(&sf, uap->buf, sizeof(sf)));
}
int
kern_fhstatfs(struct thread *td, fhandle_t fh, struct statfs *buf)
{
struct statfs *sp;
struct mount *mp;
struct vnode *vp;
int error;
error = priv_check(td, PRIV_VFS_FHSTATFS);
2013-09-05 00:19:30 +00:00
if (error != 0)
return (error);
if ((mp = vfs_busyfs(&fh.fh_fsid)) == NULL)
return (ESTALE);
error = VFS_FHTOVP(mp, &fh.fh_fid, LK_EXCLUSIVE, &vp);
2013-09-05 00:19:30 +00:00
if (error != 0) {
vfs_unbusy(mp);
return (error);
}
vput(vp);
error = prison_canseemount(td->td_ucred, mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#ifdef MAC
error = mac_mount_check_stat(td->td_ucred, mp);
2013-09-05 00:19:30 +00:00
if (error != 0)
goto out;
#endif
/*
* Set these in case the underlying filesystem fails to do so.
*/
sp = &mp->mnt_stat;
sp->f_version = STATFS_VERSION;
sp->f_namemax = NAME_MAX;
sp->f_flags = mp->mnt_flag & MNT_VISFLAGMASK;
error = VFS_STATFS(mp, sp);
if (error == 0)
*buf = *sp;
out:
vfs_unbusy(mp);
return (error);
}
int
kern_posix_fallocate(struct thread *td, int fd, off_t offset, off_t len)
{
struct file *fp;
struct mount *mp;
struct vnode *vp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
off_t olen, ooffset;
int error;
if (offset < 0 || len <= 0)
return (EINVAL);
/* Check for wrap. */
if (offset > OFF_MAX - len)
return (EFBIG);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, fd, cap_rights_init(&rights, CAP_WRITE), &fp);
if (error != 0)
return (error);
if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
error = ESPIPE;
goto out;
}
2011-04-18 16:40:47 +00:00
if ((fp->f_flag & FWRITE) == 0) {
error = EBADF;
goto out;
2011-04-18 16:40:47 +00:00
}
if (fp->f_type != DTYPE_VNODE) {
error = ENODEV;
goto out;
}
vp = fp->f_vnode;
if (vp->v_type != VREG) {
error = ENODEV;
goto out;
}
/* Allocating blocks may take a long time, so iterate. */
for (;;) {
olen = len;
ooffset = offset;
bwillwrite();
mp = NULL;
error = vn_start_write(vp, &mp, V_WAIT | PCATCH);
if (error != 0)
break;
error = vn_lock(vp, LK_EXCLUSIVE);
if (error != 0) {
vn_finished_write(mp);
break;
}
#ifdef MAC
error = mac_vnode_check_write(td->td_ucred, fp->f_cred, vp);
if (error == 0)
#endif
error = VOP_ALLOCATE(vp, &offset, &len);
VOP_UNLOCK(vp, 0);
vn_finished_write(mp);
if (olen + ooffset != offset + len) {
panic("offset + len changed from %jx/%jx to %jx/%jx",
ooffset, olen, offset, len);
}
if (error != 0 || len == 0)
break;
KASSERT(olen > len, ("Iteration did not make progress?"));
maybe_yield();
}
out:
fdrop(fp, td);
return (error);
}
int
sys_posix_fallocate(struct thread *td, struct posix_fallocate_args *uap)
{
td->td_retval[0] = kern_posix_fallocate(td, uap->fd, uap->offset,
uap->len);
return (0);
}
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
/*
* Unlike madvise(2), we do not make a best effort to remember every
* possible caching hint. Instead, we remember the last setting with
* the exception that we will allow POSIX_FADV_NORMAL to adjust the
* region of any current setting.
*/
int
kern_posix_fadvise(struct thread *td, int fd, off_t offset, off_t len,
int advice)
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
{
struct fadvise_info *fa, *new;
struct file *fp;
struct vnode *vp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
off_t end;
int error;
if (offset < 0 || len < 0 || offset > OFF_MAX - len)
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
return (EINVAL);
switch (advice) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
case POSIX_FADV_SEQUENTIAL:
case POSIX_FADV_RANDOM:
case POSIX_FADV_NOREUSE:
new = malloc(sizeof(*fa), M_FADVISE, M_WAITOK);
break;
case POSIX_FADV_NORMAL:
case POSIX_FADV_WILLNEED:
case POSIX_FADV_DONTNEED:
new = NULL;
break;
default:
return (EINVAL);
}
/* XXX: CAP_POSIX_FADVISE? */
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
error = fget(td, fd, cap_rights_init(&rights), &fp);
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
if (error != 0)
goto out;
if ((fp->f_ops->fo_flags & DFLAG_SEEKABLE) == 0) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
error = ESPIPE;
goto out;
}
if (fp->f_type != DTYPE_VNODE) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
error = ENODEV;
goto out;
}
vp = fp->f_vnode;
if (vp->v_type != VREG) {
error = ENODEV;
goto out;
}
if (len == 0)
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
end = OFF_MAX;
else
end = offset + len - 1;
switch (advice) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
case POSIX_FADV_SEQUENTIAL:
case POSIX_FADV_RANDOM:
case POSIX_FADV_NOREUSE:
/*
* Try to merge any existing non-standard region with
* this new region if possible, otherwise create a new
* non-standard region for this request.
*/
mtx_pool_lock(mtxpool_sleep, fp);
fa = fp->f_advice;
if (fa != NULL && fa->fa_advice == advice &&
((fa->fa_start <= end && fa->fa_end >= offset) ||
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
(end != OFF_MAX && fa->fa_start == end + 1) ||
(fa->fa_end != OFF_MAX && fa->fa_end + 1 == offset))) {
if (offset < fa->fa_start)
fa->fa_start = offset;
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
if (end > fa->fa_end)
fa->fa_end = end;
} else {
new->fa_advice = advice;
new->fa_start = offset;
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
new->fa_end = end;
new->fa_prevstart = 0;
new->fa_prevend = 0;
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
fp->f_advice = new;
new = fa;
}
mtx_pool_unlock(mtxpool_sleep, fp);
break;
case POSIX_FADV_NORMAL:
/*
* If a the "normal" region overlaps with an existing
* non-standard region, trim or remove the
* non-standard region.
*/
mtx_pool_lock(mtxpool_sleep, fp);
fa = fp->f_advice;
if (fa != NULL) {
if (offset <= fa->fa_start && end >= fa->fa_end) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
new = fa;
fp->f_advice = NULL;
} else if (offset <= fa->fa_start &&
2013-03-16 22:37:30 +00:00
end >= fa->fa_start)
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
fa->fa_start = end + 1;
else if (offset <= fa->fa_end && end >= fa->fa_end)
fa->fa_end = offset - 1;
else if (offset >= fa->fa_start && end <= fa->fa_end) {
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
/*
* If the "normal" region is a middle
* portion of the existing
* non-standard region, just remove
* the whole thing rather than picking
* one side or the other to
* preserve.
*/
new = fa;
fp->f_advice = NULL;
}
}
mtx_pool_unlock(mtxpool_sleep, fp);
break;
case POSIX_FADV_WILLNEED:
case POSIX_FADV_DONTNEED:
error = VOP_ADVISE(vp, offset, end, advice);
Add the posix_fadvise(2) system call. It is somewhat similar to madvise(2) except that it operates on a file descriptor instead of a memory region. It is currently only supported on regular files. Just as with madvise(2), the advice given to posix_fadvise(2) can be divided into two types. The first type provide hints about data access patterns and are used in the file read and write routines to modify the I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are thus filesystem independent. Note that to ease implementation (and since this API is only advisory anyway), only a single non-normal range is allowed per file descriptor. The second type of hints are used to hint to the OS that data will or will not be used. These hints are implemented via a new VOP_ADVISE(). A default implementation is provided which does nothing for the WILLNEED request and attempts to move any clean pages to the cache page queue for the DONTNEED request. This latter case required two other changes. First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests vinvalbuf() to only flush clean buffers for the vnode from the buffer cache and to not remove any backing pages from the vnode. This is used to ensure clean pages are not wired into the buffer cache before attempting to move them to the cache page queue. The second change adds a new vm_object_page_cache() method. This method is somewhat similar to vm_object_page_remove() except that instead of freeing each page in the specified range, it attempts to move clean pages to the cache queue if possible. To preserve the ABI of struct file, the f_cdevpriv pointer is now reused in a union to point to the currently active advice region if one is present for regular files. Reviewed by: jilles, kib, arch@ Approved by: re (kib) MFC after: 1 month
2011-11-04 04:02:50 +00:00
break;
}
out:
if (fp != NULL)
fdrop(fp, td);
free(new, M_FADVISE);
return (error);
}
int
sys_posix_fadvise(struct thread *td, struct posix_fadvise_args *uap)
{
td->td_retval[0] = kern_posix_fadvise(td, uap->fd, uap->offset,
uap->len, uap->advice);
return (0);
}