Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
futex: Add sys_futex_wait()
To complement sys_futex_waitv()/wake(), add sys_futex_wait(). This
syscall implements what was previously known as FUTEX_WAIT_BITSET
except it uses 'unsigned long' for the value and bitmask arguments,
takes timespec and clockid_t arguments for the absolute timeout and
uses FUTEX2 flags.
The 'unsigned long' allows FUTEX2_SIZE_U64 on 64bit platforms.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/20230921105248.164324363@noisy.programming.kicks-ass.net
+156
-57
+1
arch/alpha/kernel/syscalls/syscall.tbl
+1
arch/alpha/kernel/syscalls/syscall.tbl
+1
arch/arm/tools/syscall.tbl
+1
arch/arm/tools/syscall.tbl
+1
-1
arch/arm64/include/asm/unistd.h
+1
-1
arch/arm64/include/asm/unistd.h
+2
arch/arm64/include/asm/unistd32.h
+2
arch/arm64/include/asm/unistd32.h
···
913
913
__SYSCALL(__NR_fchmodat2, sys_fchmodat2)
914
914
#define __NR_futex_wake 454
915
915
__SYSCALL(__NR_futex_wake, sys_futex_wake)
916
+
#define __NR_futex_wait 455
917
+
__SYSCALL(__NR_futex_wait, sys_futex_wait)
916
918
917
919
/*
918
920
* Please add new compat syscalls above this comment and update
+1
arch/ia64/kernel/syscalls/syscall.tbl
+1
arch/ia64/kernel/syscalls/syscall.tbl
+1
arch/m68k/kernel/syscalls/syscall.tbl
+1
arch/m68k/kernel/syscalls/syscall.tbl
+1
arch/microblaze/kernel/syscalls/syscall.tbl
+1
arch/microblaze/kernel/syscalls/syscall.tbl
+1
arch/mips/kernel/syscalls/syscall_n32.tbl
+1
arch/mips/kernel/syscalls/syscall_n32.tbl
+1
arch/mips/kernel/syscalls/syscall_n64.tbl
+1
arch/mips/kernel/syscalls/syscall_n64.tbl
+1
arch/mips/kernel/syscalls/syscall_o32.tbl
+1
arch/mips/kernel/syscalls/syscall_o32.tbl
+1
arch/parisc/kernel/syscalls/syscall.tbl
+1
arch/parisc/kernel/syscalls/syscall.tbl
+1
arch/powerpc/kernel/syscalls/syscall.tbl
+1
arch/powerpc/kernel/syscalls/syscall.tbl
+1
arch/s390/kernel/syscalls/syscall.tbl
+1
arch/s390/kernel/syscalls/syscall.tbl
+1
arch/sh/kernel/syscalls/syscall.tbl
+1
arch/sh/kernel/syscalls/syscall.tbl
+1
arch/sparc/kernel/syscalls/syscall.tbl
+1
arch/sparc/kernel/syscalls/syscall.tbl
+1
arch/x86/entry/syscalls/syscall_32.tbl
+1
arch/x86/entry/syscalls/syscall_32.tbl
+1
arch/x86/entry/syscalls/syscall_64.tbl
+1
arch/x86/entry/syscalls/syscall_64.tbl
+1
arch/xtensa/kernel/syscalls/syscall.tbl
+1
arch/xtensa/kernel/syscalls/syscall.tbl
+4
include/linux/syscalls.h
+4
include/linux/syscalls.h
···
552
552
553
553
asmlinkage long sys_futex_wake(void __user *uaddr, unsigned long mask, int nr, unsigned int flags);
554
554
555
+
asmlinkage long sys_futex_wait(void __user *uaddr, unsigned long val, unsigned long mask,
556
+
unsigned int flags, struct __kernel_timespec __user *timespec,
557
+
clockid_t clockid);
558
+
555
559
asmlinkage long sys_nanosleep(struct __kernel_timespec __user *rqtp,
556
560
struct __kernel_timespec __user *rmtp);
557
561
asmlinkage long sys_nanosleep_time32(struct old_timespec32 __user *rqtp,
+3
-1
include/uapi/asm-generic/unistd.h
+3
-1
include/uapi/asm-generic/unistd.h
···
824
824
__SYSCALL(__NR_fchmodat2, sys_fchmodat2)
825
825
#define __NR_futex_wake 454
826
826
__SYSCALL(__NR_futex_wake, sys_futex_wake)
827
+
#define __NR_futex_wait 455
828
+
__SYSCALL(__NR_futex_wait, sys_futex_wait)
827
829
828
830
#undef __NR_syscalls
829
-
#define __NR_syscalls 455
831
+
#define __NR_syscalls 456
830
832
831
833
/*
832
834
* 32 bit systems traditionally used different
+3
kernel/futex/futex.h
+3
kernel/futex/futex.h
···
332
332
u32 __user *uaddr2, int nr_wake, int nr_requeue,
333
333
u32 *cmpval, int requeue_pi);
334
334
335
+
extern int __futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
336
+
struct hrtimer_sleeper *to, u32 bitset);
337
+
335
338
extern int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
336
339
ktime_t *abs_time, u32 bitset);
337
340
+90
-30
kernel/futex/syscalls.c
+90
-30
kernel/futex/syscalls.c
···
221
221
return 0;
222
222
}
223
223
224
+
static int futex2_setup_timeout(struct __kernel_timespec __user *timeout,
225
+
clockid_t clockid, struct hrtimer_sleeper *to)
226
+
{
227
+
int flag_clkid = 0, flag_init = 0;
228
+
struct timespec64 ts;
229
+
ktime_t time;
230
+
int ret;
231
+
232
+
if (!timeout)
233
+
return 0;
234
+
235
+
if (clockid == CLOCK_REALTIME) {
236
+
flag_clkid = FLAGS_CLOCKRT;
237
+
flag_init = FUTEX_CLOCK_REALTIME;
238
+
}
239
+
240
+
if (clockid != CLOCK_REALTIME && clockid != CLOCK_MONOTONIC)
241
+
return -EINVAL;
242
+
243
+
if (get_timespec64(&ts, timeout))
244
+
return -EFAULT;
245
+
246
+
/*
247
+
* Since there's no opcode for futex_waitv, use
248
+
* FUTEX_WAIT_BITSET that uses absolute timeout as well
249
+
*/
250
+
ret = futex_init_timeout(FUTEX_WAIT_BITSET, flag_init, &ts, &time);
251
+
if (ret)
252
+
return ret;
253
+
254
+
futex_setup_timer(&time, to, flag_clkid, 0);
255
+
return 0;
256
+
}
257
+
258
+
static inline void futex2_destroy_timeout(struct hrtimer_sleeper *to)
259
+
{
260
+
hrtimer_cancel(&to->timer);
261
+
destroy_hrtimer_on_stack(&to->timer);
262
+
}
263
+
224
264
/**
225
265
* sys_futex_waitv - Wait on a list of futexes
226
266
* @waiters: List of futexes to wait on
···
290
250
{
291
251
struct hrtimer_sleeper to;
292
252
struct futex_vector *futexv;
293
-
struct timespec64 ts;
294
-
ktime_t time;
295
253
int ret;
296
254
297
255
/* This syscall supports no flags for now */
···
299
261
if (!nr_futexes || nr_futexes > FUTEX_WAITV_MAX || !waiters)
300
262
return -EINVAL;
301
263
302
-
if (timeout) {
303
-
int flag_clkid = 0, flag_init = 0;
304
-
305
-
if (clockid == CLOCK_REALTIME) {
306
-
flag_clkid = FLAGS_CLOCKRT;
307
-
flag_init = FUTEX_CLOCK_REALTIME;
308
-
}
309
-
310
-
if (clockid != CLOCK_REALTIME && clockid != CLOCK_MONOTONIC)
311
-
return -EINVAL;
312
-
313
-
if (get_timespec64(&ts, timeout))
314
-
return -EFAULT;
315
-
316
-
/*
317
-
* Since there's no opcode for futex_waitv, use
318
-
* FUTEX_WAIT_BITSET that uses absolute timeout as well
319
-
*/
320
-
ret = futex_init_timeout(FUTEX_WAIT_BITSET, flag_init, &ts, &time);
321
-
if (ret)
322
-
return ret;
323
-
324
-
futex_setup_timer(&time, &to, flag_clkid, 0);
325
-
}
264
+
if (timeout && (ret = futex2_setup_timeout(timeout, clockid, &to)))
265
+
return ret;
326
266
327
267
futexv = kcalloc(nr_futexes, sizeof(*futexv), GFP_KERNEL);
328
268
if (!futexv) {
···
315
299
kfree(futexv);
316
300
317
301
destroy_timer:
318
-
if (timeout) {
319
-
hrtimer_cancel(&to.timer);
320
-
destroy_hrtimer_on_stack(&to.timer);
321
-
}
302
+
if (timeout)
303
+
futex2_destroy_timeout(&to);
322
304
return ret;
323
305
}
324
306
···
348
334
return -EINVAL;
349
335
350
336
return futex_wake(uaddr, FLAGS_STRICT | flags, nr, mask);
337
+
}
338
+
339
+
/*
340
+
* sys_futex_wait - Wait on a futex
341
+
* @uaddr: Address of the futex to wait on
342
+
* @val: Value of @uaddr
343
+
* @mask: bitmask
344
+
* @flags: FUTEX2 flags
345
+
* @timeout: Optional absolute timeout
346
+
* @clockid: Clock to be used for the timeout, realtime or monotonic
347
+
*
348
+
* Identical to the traditional FUTEX_WAIT_BITSET op, except it is part of the
349
+
* futex2 familiy of calls.
350
+
*/
351
+
352
+
SYSCALL_DEFINE6(futex_wait,
353
+
void __user *, uaddr,
354
+
unsigned long, val,
355
+
unsigned long, mask,
356
+
unsigned int, flags,
357
+
struct __kernel_timespec __user *, timeout,
358
+
clockid_t, clockid)
359
+
{
360
+
struct hrtimer_sleeper to;
361
+
int ret;
362
+
363
+
if (flags & ~FUTEX2_VALID_MASK)
364
+
return -EINVAL;
365
+
366
+
flags = futex2_to_flags(flags);
367
+
if (!futex_flags_valid(flags))
368
+
return -EINVAL;
369
+
370
+
if (!futex_validate_input(flags, val) ||
371
+
!futex_validate_input(flags, mask))
372
+
return -EINVAL;
373
+
374
+
if (timeout && (ret = futex2_setup_timeout(timeout, clockid, &to)))
375
+
return ret;
376
+
377
+
ret = __futex_wait(uaddr, flags, val, timeout ? &to : NULL, mask);
378
+
379
+
if (timeout)
380
+
futex2_destroy_timeout(&to);
381
+
382
+
return ret;
351
383
}
352
384
353
385
#ifdef CONFIG_COMPAT
+36
-25
kernel/futex/waitwake.c
+36
-25
kernel/futex/waitwake.c
···
632
632
return ret;
633
633
}
634
634
635
-
int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val, ktime_t *abs_time, u32 bitset)
635
+
int __futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
636
+
struct hrtimer_sleeper *to, u32 bitset)
636
637
{
637
-
struct hrtimer_sleeper timeout, *to;
638
-
struct restart_block *restart;
639
-
struct futex_hash_bucket *hb;
640
638
struct futex_q q = futex_q_init;
639
+
struct futex_hash_bucket *hb;
641
640
int ret;
642
641
643
642
if (!bitset)
644
643
return -EINVAL;
644
+
645
645
q.bitset = bitset;
646
646
647
-
to = futex_setup_timer(abs_time, &timeout, flags,
648
-
current->timer_slack_ns);
649
647
retry:
650
648
/*
651
649
* Prepare to wait on uaddr. On success, it holds hb->lock and q
···
651
653
*/
652
654
ret = futex_wait_setup(uaddr, val, flags, &q, &hb);
653
655
if (ret)
654
-
goto out;
656
+
return ret;
655
657
656
658
/* futex_queue and wait for wakeup, timeout, or a signal. */
657
659
futex_wait_queue(hb, &q, to);
658
660
659
661
/* If we were woken (and unqueued), we succeeded, whatever. */
660
-
ret = 0;
661
662
if (!futex_unqueue(&q))
662
-
goto out;
663
-
ret = -ETIMEDOUT;
663
+
return 0;
664
+
664
665
if (to && !to->task)
665
-
goto out;
666
+
return -ETIMEDOUT;
666
667
667
668
/*
668
669
* We expect signal_pending(current), but we might be the
···
670
673
if (!signal_pending(current))
671
674
goto retry;
672
675
673
-
ret = -ERESTARTSYS;
674
-
if (!abs_time)
675
-
goto out;
676
+
return -ERESTARTSYS;
677
+
}
676
678
677
-
restart = ¤t->restart_block;
678
-
restart->futex.uaddr = uaddr;
679
-
restart->futex.val = val;
680
-
restart->futex.time = *abs_time;
681
-
restart->futex.bitset = bitset;
682
-
restart->futex.flags = flags | FLAGS_HAS_TIMEOUT;
679
+
int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val, ktime_t *abs_time, u32 bitset)
680
+
{
681
+
struct hrtimer_sleeper timeout, *to;
682
+
struct restart_block *restart;
683
+
int ret;
683
684
684
-
ret = set_restart_fn(restart, futex_wait_restart);
685
+
to = futex_setup_timer(abs_time, &timeout, flags,
686
+
current->timer_slack_ns);
685
687
686
-
out:
687
-
if (to) {
688
-
hrtimer_cancel(&to->timer);
689
-
destroy_hrtimer_on_stack(&to->timer);
688
+
ret = __futex_wait(uaddr, flags, val, to, bitset);
689
+
690
+
/* No timeout, nothing to clean up. */
691
+
if (!to)
692
+
return ret;
693
+
694
+
hrtimer_cancel(&to->timer);
695
+
destroy_hrtimer_on_stack(&to->timer);
696
+
697
+
if (ret == -ERESTARTSYS) {
698
+
restart = ¤t->restart_block;
699
+
restart->futex.uaddr = uaddr;
700
+
restart->futex.val = val;
701
+
restart->futex.time = *abs_time;
702
+
restart->futex.bitset = bitset;
703
+
restart->futex.flags = flags | FLAGS_HAS_TIMEOUT;
704
+
705
+
return set_restart_fn(restart, futex_wait_restart);
690
706
}
707
+
691
708
return ret;
692
709
}
693
710