FUTEX(2) Linux Programmer's Manual FUTEX(2)
NAME
futex(2,4) - Fast Userspace Locking system call
SYNOPSIS
#include <linux/futex.h>
#include <sys/time.h>
int futex(2,4) (int *uaddr, int op, int val, const struct timespec *timeout(1,3x,3x cbreak),
int *uaddr2, int val3);
DESCRIPTION
The futex(2,4) system call provides a method for a program to wait for a
value at a given address to change, and a method to wake up anyone
waiting on a particular address (while the addresses for the same mem-
ory in(1,8) separate processes may not be equal, the kernel maps them inter-
nally so the same memory mapped in(1,8) different locations will correspond
for futex(2,4) calls). It is typically used to implement the contended case
of a lock in(1,8) shared memory, as described in(1,8) futex(2,4)(4).
When a futex(2,4)(4) operation did not finish uncontended in(1,8) userspace, a
call needs to be made to the kernel to arbitrate. Arbitration can
either mean putting the calling process to sleep(1,3) or, conversely, waking
a waiting process.
Callers of this function are expected to adhere to the semantics as set(7,n,1 builtins)
out in(1,8) futex(2,4)(4). As these semantics involve writing non-portable
assembly instructions, this in(1,8) turn probably means that most users(1,5) will
in(1,8) fact be library authors and not general application developers.
The uaddr argument needs to point to an aligned integer which stores
the counter. The operation to execute is passed via the op parameter,
along with a value val.
Five operations are currently defined:
FUTEX_WAIT
This operation atomically verifies that the futex(2,4) address uaddr
still contains the value val, and sleeps awaiting FUTEX_WAKE on
this futex(2,4) address. If the timeout(1,3x,3x cbreak) argument is non-NULL, its
contents describe the maximum duration of the wait, which is
infinite otherwise. The arguments uaddr2 and val3 are ignored.
For futex(2,4)(4), this call is executed if(3,n) decrementing the count
gave a negative value (indicating contention), and will sleep(1,3)
until another process releases the futex(2,4) and executes the
FUTEX_WAKE operation.
FUTEX_WAKE
This operation wakes at most val processes waiting on this futex(2,4)
address (ie. inside FUTEX_WAIT). The arguments timeout(1,3x,3x cbreak), uaddr2
and val3 are ignored.
For futex(2,4)(4), this is executed if(3,n) incrementing the count showed
that there were waiters, once the futex(2,4) value has been set(7,n,1 builtins) to 1
(indicating that it is available).
FUTEX_FD
To support asynchronous wakeups, this operation associates a
file(1,n) descriptor with a futex. If another process executes a
FUTEX_WAKE, the process will receive the signal(2,7) number that was
passed in(1,8) val. The calling process must close(2,7,n) the returned file(1,n)
descriptor after use. The arguments timeout(1,3x,3x cbreak), uaddr2 and val3
are ignored.
To prevent race conditions, the caller should test if(3,n) the futex(2,4)
has been upped after FUTEX_FD returns.
FUTEX_REQUEUE (since Linux 2.5.70)
This operation was introduced in(1,8) order to avoid a "thundering
herd" effect when FUTEX_WAKE is used and all processes woken up
need to acquire another futex. This call wakes up val processes,
and requeues all other waiters on the futex(2,4) at address uaddr2.
The arguments timeout(1,3x,3x cbreak) and val3 are ignored.
FUTEX_CMP_REQUEUE (since Linux 2.6.7)
There was a race in(1,8) the intended use of FUTEX_REQUEUE, so
FUTEX_CMP_REQUEUE was introduced. This is similar to
FUTEX_REQUEUE, but first checks whether the location uaddr still
contains the value val3. If not, an error(8,n) EAGAIN is returned.
The argument timeout(1,3x,3x cbreak) is ignored.
RETURN VALUE
Depending on which operation was executed, the returned value can have
differing meanings.
FUTEX_WAIT
Returns 0 if(3,n) the process was woken by a FUTEX_WAKE call. In case
of timeout(1,3x,3x cbreak), ETIMEDOUT is returned. If the futex(2,4) was not equal to
the expected value, the operation returns EWOULDBLOCK. Signals
(or other spurious wakeups) cause FUTEX_WAIT to return EINTR.
FUTEX_WAKE
Returns the number of processes woken up.
FUTEX_FD
Returns the new file(1,n) descriptor associated with the futex.
FUTEX_REQUEUE
Returns the number of processes woken up.
FUTEX_CMP_REQUEUE
Returns the number of processes woken up.
ERRORS
EACCES No read(2,n,1 builtins) access(2,5) to futex(2,4) memory.
EAGAIN FUTEX_CMP_REQUEUE found an unexpected futex(2,4) value. (This proba-
bly indicates a race; use the safe FUTEX_WAKE now.)
EFAULT Error in(1,8) getting timeout(1,3x,3x cbreak) information from userspace.
EINVAL An operation was not defined or error(8,n) in(1,8) page alignment.
ENFILE The system limit on the total number of open(2,3,n) files has been
reached.
NOTES
To reiterate, bare futexes are not intended as an easy to use abstrac-
tion for end-users. Implementors are expected to be assembly literate
and to have read(2,n,1 builtins) the sources of the futex(2,4) userspace library referenced
below.
VERSIONS
Initial futex(2,4) support was merged in(1,8) Linux 2.5.7 but with different
semantics from what was described above. A 4-parameter system call with
the semantics given here was introduced in(1,8) Linux 2.5.40. In Linux
2.5.70 one parameter was added. In Linux 2.6.7 a sixth parameter was
added - messy, especially on the s390 architecture.
SEE ALSO
futex(2,4)(4), `Fuss, Futexes and Furwocks: Fast Userlevel Locking in(1,8) Linux'
(proceedings of the Ottawa Linux Symposium 2002), futex(2,4) example
library, futex-*.tar.bz2 <URL:ftp://ftp.nl.kernel.org:/pub/linux/ker-
nel/people/rusty/>.
Linux 2.6.7 2004-10-07 FUTEX(2)