OpenBSD manual page server

Manual Page Search Parameters
PLEDGE(2) System Calls Manual PLEDGE(2)

pledge — restrict system operations

#include <unistd.h>

int
pledge(const char *promises, const char *execpromises);

The () system call forces the current process into a restricted-service operating mode. A few subsets are available, roughly described as computation, memory management, read-write operations on file descriptors, opening of files, networking (and notably separate, DNS resolution). In general, these modes were selected by studying the operation of many programs using libc and other such interfaces, and setting promises or execpromises.

Use of () in an application will require at least some study and understanding of the interfaces called. Subsequent calls to pledge() can reduce the abilities further, but abilities can never be regained.

A process which attempts a restricted operation is killed with an uncatchable SIGABRT, delivering a core file if possible. A process currently running with pledge has state ‘p’ in ps(1) output; a process that was terminated due to a pledge violation is accounted by lastcomm(1) with the ‘P’ flag.

A promises value of "" restricts the process to the _exit(2) system call. This can be used for pure computation operating on memory shared with another process.

Passing NULL to promises or execpromises specifies to not change the current value.

Some system calls, when allowed, have restrictions applied to them:

access(2):
May check for existence of /etc/localtime.
adjtime(2):
Read-only, for ntpd(8).
chmod(2), fchmod(2), fchmodat(2), chown(2), lchown(2), fchown(2), fchownat(2), mkfifo(2), and mknod(2):
Setuid/setgid/sticky bits are ignored. The user or group cannot be changed on a file.
ioctl(2):
Only the FIONREAD, FIONBIO, FIOCLEX, and FIONCLEX operations are allowed by default. Various ioctl requests are allowed against specific file descriptors based upon the requests audio, bpf, disklabel, drm, inet, pf, route, wroute, tape, tty, video, and vmm.
mmap(2) and mprotect(2):
isn't allowed.
open(2):
May open /etc/localtime and any files below /usr/share/zoneinfo.
profil(2):
Can only disable profiling.
():
Can only reduce permissions for promises and execpromises.
sysctl(2):
A small set of read-only operations are allowed, sufficient to support: getdomainname(3), gethostname(3), getifaddrs(3), uname(3), and system sensor readings.

The promises argument is specified as a string, with space separated keywords:

The following system calls are permitted. sendto(2) is only permitted if its destination socket address is NULL. As a result, all the expected functionalities of libc stdio work.

clock_getres(2), clock_gettime(2), close(2), closefrom(2), dup(2), dup2(2), dup3(2), fchdir(2), fcntl(2), fstat(2), fsync(2), ftruncate(2), getdtablecount(2), getegid(2), getentropy(2), geteuid(2), getgid(2), getgroups(2), getitimer(2), getlogin(2), getpgid(2), getpgrp(2), getpid(2), getppid(2), getresgid(2), getresuid(2), getrlimit(2), getrtable(2), getsid(2), getthrid(2), gettimeofday(2), getuid(2), issetugid(2), kevent(2), kqueue(2), kqueue1(2), lseek(2), madvise(2), minherit(2), mmap(2), mprotect(2), mquery(2), munmap(2), nanosleep(2), pipe(2), pipe2(2), poll(2), pread(2), preadv(2), profil(2), pwrite(2), pwritev(2), read(2), readv(2), recvfrom(2), recvmsg(2), select(2), sendmsg(2), sendsyslog(2), sendto(2), setitimer(2), shutdown(2), sigaction(2), sigprocmask(2), sigreturn(2), socketpair(2), umask(2), wait4(2), waitid(2), write(2), writev(2)

A number of system calls are allowed if they only cause read-only effects on the filesystem, or expose filenames to programs:

chdir(2), getcwd(3), getdents(2), openat(2), fstatat(2), faccessat(2), readlinkat(2), lstat(2), chmod(2), fchmod(2), fchmodat(2), chflags(2), chflagsat(2), chown(2), fchown(2), fchownat(2), fstat(2), getfsstat(2)

A number of system calls are allowed and may cause write-effects on the filesystem:

getcwd(3), openat(2), fstatat(2), faccessat(2), readlinkat(2), lstat(2), chmod(2), fchmod(2), fchmodat(2), chflags(2), chflagsat(2), chown(2), fchown(2), fchownat(2), fstat(2)

A number of system calls and sub-modes are allowed, which may create new files or directories in the filesystem:

rename(2), renameat(2), link(2), linkat(2), symlink(2), symlinkat(2), unlink(2), unlinkat(2), mkdir(2), mkdirat(2), rmdir(2)

A number of system calls are allowed to create special files:

mkfifo(2), mknod(2)

A number of system calls are allowed to do operations in the /tmp directory, including create, read, or write:

lstat(2), chmod(2), chflags(2), chown(2), unlink(2), fstat(2)

The following system calls are allowed to operate in the AF_INET and AF_INET6 domains (though setsockopt(2) has been substantially reduced in functionality):

socket(2), listen(2), bind(2), connect(2), accept4(2), accept(2), getpeername(2), getsockname(2), setsockopt(2), getsockopt(2)

In combination with inet give back functionality to setsockopt(2) for operating on multicast sockets.
The following system calls are allowed to make explicit changes to fields in struct stat relating to a file:

utimes(2), futimes(2), utimensat(2), futimens(2), chmod(2), fchmod(2), fchmodat(2), chflags(2), chflagsat(2), chown(2), fchownat(2), lchown(2), fchown(2), utimes(2)

The chown(2) family is allowed to change the user or group on a file.
File locking via fcntl(2), flock(2), lockf(3), and open(2) is allowed. No distinction is made between shared and exclusive locks. This promise is required for unlock as well as lock.
The following system calls are allowed to operate in the AF_UNIX domain:

socket(2), listen(2), bind(2), connect(2), accept4(2), accept(2), getpeername(2), getsockname(2), setsockopt(2), getsockopt(2)

Subsequent to a successful open(2) of /etc/resolv.conf, a few system calls become able to allow DNS network transactions:

sendto(2), recvfrom(2), socket(2), connect(2)

This allows read-only opening of files in /etc for the getpwnam(3), getgrnam(3), getgrouplist(3), and initgroups(3) family of functions, including lookups via the yp(8) protocol for YP and LDAP databases.
Allows sending of file descriptors using sendmsg(2). File descriptors referring to directories may not be passed.
Allows receiving of file descriptors using recvmsg(2). File descriptors referring to directories may not be passed.
Allow MTIOCGET and MTIOCTOP operations against tape drives.
In addition to allowing read-write operations on /dev/tty, this opens up a variety of ioctl(2) requests used by tty devices. If tty is accompanied with rpath, revoke(2) is permitted. Otherwise only the following ioctl(2) requests are permitted:

TIOCSPGRP, TIOCGETA, TIOCGPGRP, TIOCGWINSZ, TIOCSWINSZ, TIOCSBRK, TIOCCDTR, TIOCSETA, TIOCSETAW, TIOCSETAF, TIOCUCNTL

Allows the following process relationship operations:

fork(2), vfork(2), kill(2), getpriority(2), setpriority(2), setrlimit(2), setpgid(2), setsid(2)

Allows a process to call execve(2). Coupled with the proc promise, this allows a process to fork and execute another program. If execpromises has been previously set the new program begins with those promises, unless setuid/setgid bits are set in which case execution is blocked with EACCES. Otherwise the new program starts running without pledge active, and hopefully makes a new pledge soon.
Allows the use of PROT_EXEC with mmap(2) and mprotect(2).
Allows the setting of system time, via the settimeofday(2), adjtime(2), and adjfreq(2) system calls.
Allows enough sysctl(2) interfaces to allow inspection of processes operating on the system using programs like ps(1).
Allows enough sysctl(2) interfaces to allow inspection of the system's virtual memory by programs like top(1) and vmstat(8).
Allows the following system calls which can change the rights of a process:

setuid(2), seteuid(2), setreuid(2), setresuid(2), setgid(2), setegid(2), setregid(2), setresgid(2), setgroups(2), setlogin(2), setrlimit(2), getpriority(2), setpriority(2), setrtable(2)

Allows a subset of ioctl(2) operations on the pf(4) device:

DIOCADDRULE, DIOCGETSTATUS, DIOCNATLOOK, DIOCRADDTABLES, DIOCRCLRADDRS, DIOCRCLRTABLES, DIOCRCLRTSTATS, DIOCRGETTSTATS, DIOCRSETADDRS, DIOCXBEGIN, DIOCXCOMMIT

Allow inspection of the routing table.
Allow changes to the routing table.
Allows a subset of ioctl(2) operations on audio(4) devices (see sio_open(3) for more information):

AUDIO_GETPOS, AUDIO_GETPAR, AUDIO_SETPAR, AUDIO_START, AUDIO_STOP, AUDIO_MIXER_DEVINFO, AUDIO_MIXER_READ, AUDIO_MIXER_WRITE

Allows a subset of ioctl(2) operations on video(4) devices:

VIDIOC_DQBUF, VIDIOC_ENUM_FMT, VIDIOC_ENUM_FRAMEINTERVALS, VIDIOC_ENUM_FRAMESIZES, VIDIOC_G_CTRL, VIDIOC_G_PARM, VIDIOC_QBUF, VIDIOC_QUERYBUF, VIDIOC_QUERYCAP, VIDIOC_QUERYCTRL, VIDIOC_S_CTRL, VIDIOC_S_FMT, VIDIOC_S_PARM, VIDIOC_STREAMOFF, VIDIOC_STREAMON, VIDIOC_TRY_FMT, VIDIOC_REQBUFS

Allow BIOCGSTATS operation for statistics collection from a bpf(4) device.
Allow unveil(2) to be called.
Rather than killing the process upon violation, indicate error with ENOSYS.

Also when () is called with higher promises or execpromises, those changes will be ignored and return success. This is useful when a parent enforces execpromises but an execve'd child has a different idea.

Upon successful completion, the value 0 is returned; otherwise the value -1 is returned and the global variable errno is set to indicate the error.

pledge() will fail if:

[]
promises or execpromises points outside the process's allocated address space.
[]
promises is malformed or contains invalid keywords.
[]
This process is attempting to increase permissions.

The pledge() system call first appeared in OpenBSD 5.9.

September 17, 2024 OpenBSD-current