]>
git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/select.c
2 * This file contains the procedures for the handling of select and poll
4 * Created for Linux based loosely upon Mathius Lattner's minix
5 * patches by Peter MacDonald. Heavily edited by Linus.
8 * COFF/ELF binary emulation. If the process has the STICKY_TIMEOUTS
9 * flag set in its personality we do *not* modify the given timeout
10 * parameter to reflect time remaining.
13 * Changed sys_poll()/do_poll() to use PAGE_SIZE chunk-based allocation
14 * of fds to overcome nfds < 16390 descriptors limit (Tigran Aivazian).
17 #include <linux/kernel.h>
18 #include <linux/sched.h>
19 #include <linux/syscalls.h>
20 #include <linux/export.h>
21 #include <linux/slab.h>
22 #include <linux/poll.h>
23 #include <linux/personality.h> /* for STICKY_TIMEOUTS */
24 #include <linux/file.h>
25 #include <linux/fdtable.h>
27 #include <linux/rcupdate.h>
28 #include <linux/hrtimer.h>
30 #include <asm/uaccess.h>
34 * Estimate expected accuracy in ns from a timeval.
36 * After quite a bit of churning around, we've settled on
37 * a simple thing of taking 0.1% of the timeout as the
38 * slack, with a cap of 100 msec.
39 * "nice" tasks get a 0.5% slack instead.
41 * Consider this comment an open invitation to come up with even
45 #define MAX_SLACK (100 * NSEC_PER_MSEC)
47 static long __estimate_accuracy(struct timespec
*tv
)
55 if (task_nice(current
) > 0)
56 divfactor
= divfactor
/ 5;
58 if (tv
->tv_sec
> MAX_SLACK
/ (NSEC_PER_SEC
/divfactor
))
61 slack
= tv
->tv_nsec
/ divfactor
;
62 slack
+= tv
->tv_sec
* (NSEC_PER_SEC
/divfactor
);
64 if (slack
> MAX_SLACK
)
70 long select_estimate_accuracy(struct timespec
*tv
)
76 * Realtime tasks get a slack of 0 for obvious reasons.
83 now
= timespec_sub(*tv
, now
);
84 ret
= __estimate_accuracy(&now
);
85 if (ret
< current
->timer_slack_ns
)
86 return current
->timer_slack_ns
;
92 struct poll_table_page
{
93 struct poll_table_page
* next
;
94 struct poll_table_entry
* entry
;
95 struct poll_table_entry entries
[0];
98 #define POLL_TABLE_FULL(table) \
99 ((unsigned long)((table)->entry+1) > PAGE_SIZE + (unsigned long)(table))
102 * Ok, Peter made a complicated, but straightforward multiple_wait() function.
103 * I have rewritten this, taking some shortcuts: This code may not be easy to
104 * follow, but it should be free of race-conditions, and it's practical. If you
105 * understand what I'm doing here, then you understand how the linux
106 * sleep/wakeup mechanism works.
108 * Two very simple procedures, poll_wait() and poll_freewait() make all the
109 * work. poll_wait() is an inline-function defined in <linux/poll.h>,
110 * as all select/poll functions have to call it to add an entry to the
113 static void __pollwait(struct file
*filp
, wait_queue_head_t
*wait_address
,
116 void poll_initwait(struct poll_wqueues
*pwq
)
118 init_poll_funcptr(&pwq
->pt
, __pollwait
);
119 pwq
->polling_task
= current
;
123 pwq
->inline_index
= 0;
125 EXPORT_SYMBOL(poll_initwait
);
127 static void free_poll_entry(struct poll_table_entry
*entry
)
129 remove_wait_queue(entry
->wait_address
, &entry
->wait
);
133 void poll_freewait(struct poll_wqueues
*pwq
)
135 struct poll_table_page
* p
= pwq
->table
;
137 for (i
= 0; i
< pwq
->inline_index
; i
++)
138 free_poll_entry(pwq
->inline_entries
+ i
);
140 struct poll_table_entry
* entry
;
141 struct poll_table_page
*old
;
146 free_poll_entry(entry
);
147 } while (entry
> p
->entries
);
150 free_page((unsigned long) old
);
153 EXPORT_SYMBOL(poll_freewait
);
155 static struct poll_table_entry
*poll_get_entry(struct poll_wqueues
*p
)
157 struct poll_table_page
*table
= p
->table
;
159 if (p
->inline_index
< N_INLINE_POLL_ENTRIES
)
160 return p
->inline_entries
+ p
->inline_index
++;
162 if (!table
|| POLL_TABLE_FULL(table
)) {
163 struct poll_table_page
*new_table
;
165 new_table
= (struct poll_table_page
*) __get_free_page(GFP_KERNEL
);
170 new_table
->entry
= new_table
->entries
;
171 new_table
->next
= table
;
172 p
->table
= new_table
;
176 return table
->entry
++;
179 static int __pollwake(wait_queue_t
*wait
, unsigned mode
, int sync
, void *key
)
181 struct poll_wqueues
*pwq
= wait
->private;
182 DECLARE_WAITQUEUE(dummy_wait
, pwq
->polling_task
);
185 * Although this function is called under waitqueue lock, LOCK
186 * doesn't imply write barrier and the users expect write
187 * barrier semantics on wakeup functions. The following
188 * smp_wmb() is equivalent to smp_wmb() in try_to_wake_up()
189 * and is paired with set_mb() in poll_schedule_timeout.
195 * Perform the default wake up operation using a dummy
198 * TODO: This is hacky but there currently is no interface to
199 * pass in @sync. @sync is scheduled to be removed and once
200 * that happens, wake_up_process() can be used directly.
202 return default_wake_function(&dummy_wait
, mode
, sync
, key
);
205 static int pollwake(wait_queue_t
*wait
, unsigned mode
, int sync
, void *key
)
207 struct poll_table_entry
*entry
;
209 entry
= container_of(wait
, struct poll_table_entry
, wait
);
210 if (key
&& !((unsigned long)key
& entry
->key
))
212 return __pollwake(wait
, mode
, sync
, key
);
215 /* Add a new entry */
216 static void __pollwait(struct file
*filp
, wait_queue_head_t
*wait_address
,
219 struct poll_wqueues
*pwq
= container_of(p
, struct poll_wqueues
, pt
);
220 struct poll_table_entry
*entry
= poll_get_entry(pwq
);
225 entry
->wait_address
= wait_address
;
226 entry
->key
= p
->_key
;
227 init_waitqueue_func_entry(&entry
->wait
, pollwake
);
228 entry
->wait
.private = pwq
;
229 add_wait_queue(wait_address
, &entry
->wait
);
232 int poll_schedule_timeout(struct poll_wqueues
*pwq
, int state
,
233 ktime_t
*expires
, unsigned long slack
)
237 set_current_state(state
);
239 rc
= schedule_hrtimeout_range(expires
, slack
, HRTIMER_MODE_ABS
);
240 __set_current_state(TASK_RUNNING
);
243 * Prepare for the next iteration.
245 * The following set_mb() serves two purposes. First, it's
246 * the counterpart rmb of the wmb in pollwake() such that data
247 * written before wake up is always visible after wake up.
248 * Second, the full barrier guarantees that triggered clearing
249 * doesn't pass event check of the next iteration. Note that
250 * this problem doesn't exist for the first iteration as
251 * add_wait_queue() has full barrier semantics.
253 set_mb(pwq
->triggered
, 0);
257 EXPORT_SYMBOL(poll_schedule_timeout
);
260 * poll_select_set_timeout - helper function to setup the timeout value
261 * @to: pointer to timespec variable for the final timeout
262 * @sec: seconds (from user space)
263 * @nsec: nanoseconds (from user space)
265 * Note, we do not use a timespec for the user space value here, That
266 * way we can use the function for timeval and compat interfaces as well.
268 * Returns -EINVAL if sec/nsec are not normalized. Otherwise 0.
270 int poll_select_set_timeout(struct timespec
*to
, long sec
, long nsec
)
272 struct timespec ts
= {.tv_sec
= sec
, .tv_nsec
= nsec
};
274 if (!timespec_valid(&ts
))
277 /* Optimize for the zero timeout value here */
279 to
->tv_sec
= to
->tv_nsec
= 0;
282 *to
= timespec_add_safe(*to
, ts
);
287 static int poll_select_copy_remaining(struct timespec
*end_time
, void __user
*p
,
288 int timeval
, int ret
)
296 if (current
->personality
& STICKY_TIMEOUTS
)
299 /* No update for zero timeout */
300 if (!end_time
->tv_sec
&& !end_time
->tv_nsec
)
304 rts
= timespec_sub(*end_time
, rts
);
306 rts
.tv_sec
= rts
.tv_nsec
= 0;
309 if (sizeof(rtv
) > sizeof(rtv
.tv_sec
) + sizeof(rtv
.tv_usec
))
310 memset(&rtv
, 0, sizeof(rtv
));
311 rtv
.tv_sec
= rts
.tv_sec
;
312 rtv
.tv_usec
= rts
.tv_nsec
/ NSEC_PER_USEC
;
314 if (!copy_to_user(p
, &rtv
, sizeof(rtv
)))
317 } else if (!copy_to_user(p
, &rts
, sizeof(rts
)))
321 * If an application puts its timeval in read-only memory, we
322 * don't want the Linux-specific update to the timeval to
323 * cause a fault after the select has completed
324 * successfully. However, because we're not updating the
325 * timeval, we can't restart the system call.
329 if (ret
== -ERESTARTNOHAND
)
334 #define FDS_IN(fds, n) (fds->in + n)
335 #define FDS_OUT(fds, n) (fds->out + n)
336 #define FDS_EX(fds, n) (fds->ex + n)
338 #define BITS(fds, n) (*FDS_IN(fds, n)|*FDS_OUT(fds, n)|*FDS_EX(fds, n))
340 static int max_select_fd(unsigned long n
, fd_set_bits
*fds
)
342 unsigned long *open_fds
;
347 /* handle last in-complete long-word first */
348 set
= ~(~0UL << (n
& (BITS_PER_LONG
-1)));
350 fdt
= files_fdtable(current
->files
);
351 open_fds
= fdt
->open_fds
+ n
;
356 if (!(set
& ~*open_fds
))
367 if (set
& ~*open_fds
)
376 max
+= n
* BITS_PER_LONG
;
382 #define POLLIN_SET (POLLRDNORM | POLLRDBAND | POLLIN | POLLHUP | POLLERR)
383 #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
384 #define POLLEX_SET (POLLPRI)
386 static inline void wait_key_set(poll_table
*wait
, unsigned long in
,
387 unsigned long out
, unsigned long bit
)
389 wait
->_key
= POLLEX_SET
;
391 wait
->_key
|= POLLIN_SET
;
393 wait
->_key
|= POLLOUT_SET
;
396 int do_select(int n
, fd_set_bits
*fds
, struct timespec
*end_time
)
398 ktime_t expire
, *to
= NULL
;
399 struct poll_wqueues table
;
401 int retval
, i
, timed_out
= 0;
402 unsigned long slack
= 0;
405 retval
= max_select_fd(n
, fds
);
412 poll_initwait(&table
);
414 if (end_time
&& !end_time
->tv_sec
&& !end_time
->tv_nsec
) {
419 if (end_time
&& !timed_out
)
420 slack
= select_estimate_accuracy(end_time
);
424 unsigned long *rinp
, *routp
, *rexp
, *inp
, *outp
, *exp
;
426 inp
= fds
->in
; outp
= fds
->out
; exp
= fds
->ex
;
427 rinp
= fds
->res_in
; routp
= fds
->res_out
; rexp
= fds
->res_ex
;
429 for (i
= 0; i
< n
; ++rinp
, ++routp
, ++rexp
) {
430 unsigned long in
, out
, ex
, all_bits
, bit
= 1, mask
, j
;
431 unsigned long res_in
= 0, res_out
= 0, res_ex
= 0;
432 const struct file_operations
*f_op
= NULL
;
433 struct file
*file
= NULL
;
435 in
= *inp
++; out
= *outp
++; ex
= *exp
++;
436 all_bits
= in
| out
| ex
;
442 for (j
= 0; j
< BITS_PER_LONG
; ++j
, ++i
, bit
<<= 1) {
446 if (!(bit
& all_bits
))
448 file
= fget_light(i
, &fput_needed
);
451 mask
= DEFAULT_POLLMASK
;
452 if (f_op
&& f_op
->poll
) {
453 wait_key_set(wait
, in
, out
, bit
);
454 mask
= (*f_op
->poll
)(file
, wait
);
456 fput_light(file
, fput_needed
);
457 if ((mask
& POLLIN_SET
) && (in
& bit
)) {
462 if ((mask
& POLLOUT_SET
) && (out
& bit
)) {
467 if ((mask
& POLLEX_SET
) && (ex
& bit
)) {
483 if (retval
|| timed_out
|| signal_pending(current
))
486 retval
= table
.error
;
491 * If this is the first loop and we have a timeout
492 * given, then we convert to ktime_t and set the to
493 * pointer to the expiry value.
495 if (end_time
&& !to
) {
496 expire
= timespec_to_ktime(*end_time
);
500 if (!poll_schedule_timeout(&table
, TASK_INTERRUPTIBLE
,
505 poll_freewait(&table
);
511 * We can actually return ERESTARTSYS instead of EINTR, but I'd
512 * like to be certain this leads to no problems. So I return
513 * EINTR just for safety.
515 * Update: ERESTARTSYS breaks at least the xview clock binary, so
516 * I'm trying ERESTARTNOHAND which restart only when you want to.
518 int core_sys_select(int n
, fd_set __user
*inp
, fd_set __user
*outp
,
519 fd_set __user
*exp
, struct timespec
*end_time
)
526 /* Allocate small arguments on the stack to save memory and be faster */
527 long stack_fds
[SELECT_STACK_ALLOC
/sizeof(long)];
533 /* max_fds can increase, so grab it once to avoid race */
535 fdt
= files_fdtable(current
->files
);
536 max_fds
= fdt
->max_fds
;
542 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
543 * since we used fdset we need to allocate memory in units of
548 if (size
> sizeof(stack_fds
) / 6) {
549 /* Not enough space in on-stack array; must use kmalloc */
551 bits
= kmalloc(6 * size
, GFP_KERNEL
);
556 fds
.out
= bits
+ size
;
557 fds
.ex
= bits
+ 2*size
;
558 fds
.res_in
= bits
+ 3*size
;
559 fds
.res_out
= bits
+ 4*size
;
560 fds
.res_ex
= bits
+ 5*size
;
562 if ((ret
= get_fd_set(n
, inp
, fds
.in
)) ||
563 (ret
= get_fd_set(n
, outp
, fds
.out
)) ||
564 (ret
= get_fd_set(n
, exp
, fds
.ex
)))
566 zero_fd_set(n
, fds
.res_in
);
567 zero_fd_set(n
, fds
.res_out
);
568 zero_fd_set(n
, fds
.res_ex
);
570 ret
= do_select(n
, &fds
, end_time
);
575 ret
= -ERESTARTNOHAND
;
576 if (signal_pending(current
))
581 if (set_fd_set(n
, inp
, fds
.res_in
) ||
582 set_fd_set(n
, outp
, fds
.res_out
) ||
583 set_fd_set(n
, exp
, fds
.res_ex
))
587 if (bits
!= stack_fds
)
593 SYSCALL_DEFINE5(select
, int, n
, fd_set __user
*, inp
, fd_set __user
*, outp
,
594 fd_set __user
*, exp
, struct timeval __user
*, tvp
)
596 struct timespec end_time
, *to
= NULL
;
601 if (copy_from_user(&tv
, tvp
, sizeof(tv
)))
605 if (poll_select_set_timeout(to
,
606 tv
.tv_sec
+ (tv
.tv_usec
/ USEC_PER_SEC
),
607 (tv
.tv_usec
% USEC_PER_SEC
) * NSEC_PER_USEC
))
611 ret
= core_sys_select(n
, inp
, outp
, exp
, to
);
612 ret
= poll_select_copy_remaining(&end_time
, tvp
, 1, ret
);
617 static long do_pselect(int n
, fd_set __user
*inp
, fd_set __user
*outp
,
618 fd_set __user
*exp
, struct timespec __user
*tsp
,
619 const sigset_t __user
*sigmask
, size_t sigsetsize
)
621 sigset_t ksigmask
, sigsaved
;
622 struct timespec ts
, end_time
, *to
= NULL
;
626 if (copy_from_user(&ts
, tsp
, sizeof(ts
)))
630 if (poll_select_set_timeout(to
, ts
.tv_sec
, ts
.tv_nsec
))
635 /* XXX: Don't preclude handling different sized sigset_t's. */
636 if (sigsetsize
!= sizeof(sigset_t
))
638 if (copy_from_user(&ksigmask
, sigmask
, sizeof(ksigmask
)))
641 sigdelsetmask(&ksigmask
, sigmask(SIGKILL
)|sigmask(SIGSTOP
));
642 sigprocmask(SIG_SETMASK
, &ksigmask
, &sigsaved
);
645 ret
= core_sys_select(n
, inp
, outp
, exp
, to
);
646 ret
= poll_select_copy_remaining(&end_time
, tsp
, 0, ret
);
648 if (ret
== -ERESTARTNOHAND
) {
650 * Don't restore the signal mask yet. Let do_signal() deliver
651 * the signal on the way back to userspace, before the signal
655 memcpy(¤t
->saved_sigmask
, &sigsaved
,
657 set_restore_sigmask();
660 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
666 * Most architectures can't handle 7-argument syscalls. So we provide a
667 * 6-argument version where the sixth argument is a pointer to a structure
668 * which has a pointer to the sigset_t itself followed by a size_t containing
671 SYSCALL_DEFINE6(pselect6
, int, n
, fd_set __user
*, inp
, fd_set __user
*, outp
,
672 fd_set __user
*, exp
, struct timespec __user
*, tsp
,
675 size_t sigsetsize
= 0;
676 sigset_t __user
*up
= NULL
;
679 if (!access_ok(VERIFY_READ
, sig
, sizeof(void *)+sizeof(size_t))
680 || __get_user(up
, (sigset_t __user
* __user
*)sig
)
681 || __get_user(sigsetsize
,
682 (size_t __user
*)(sig
+sizeof(void *))))
686 return do_pselect(n
, inp
, outp
, exp
, tsp
, up
, sigsetsize
);
689 #ifdef __ARCH_WANT_SYS_OLD_SELECT
690 struct sel_arg_struct
{
692 fd_set __user
*inp
, *outp
, *exp
;
693 struct timeval __user
*tvp
;
696 SYSCALL_DEFINE1(old_select
, struct sel_arg_struct __user
*, arg
)
698 struct sel_arg_struct a
;
700 if (copy_from_user(&a
, arg
, sizeof(a
)))
702 return sys_select(a
.n
, a
.inp
, a
.outp
, a
.exp
, a
.tvp
);
707 struct poll_list
*next
;
709 struct pollfd entries
[0];
712 #define POLLFD_PER_PAGE ((PAGE_SIZE-sizeof(struct poll_list)) / sizeof(struct pollfd))
715 * Fish for pollable events on the pollfd->fd file descriptor. We're only
716 * interested in events matching the pollfd->events mask, and the result
717 * matching that mask is both recorded in pollfd->revents and returned. The
718 * pwait poll_table will be used by the fd-provided poll handler for waiting,
719 * if pwait->_qproc is non-NULL.
721 static inline unsigned int do_pollfd(struct pollfd
*pollfd
, poll_table
*pwait
)
732 file
= fget_light(fd
, &fput_needed
);
735 mask
= DEFAULT_POLLMASK
;
736 if (file
->f_op
&& file
->f_op
->poll
) {
737 pwait
->_key
= pollfd
->events
|POLLERR
|POLLHUP
;
738 mask
= file
->f_op
->poll(file
, pwait
);
740 /* Mask out unneeded events. */
741 mask
&= pollfd
->events
| POLLERR
| POLLHUP
;
742 fput_light(file
, fput_needed
);
745 pollfd
->revents
= mask
;
750 static int do_poll(unsigned int nfds
, struct poll_list
*list
,
751 struct poll_wqueues
*wait
, struct timespec
*end_time
)
753 poll_table
* pt
= &wait
->pt
;
754 ktime_t expire
, *to
= NULL
;
755 int timed_out
= 0, count
= 0;
756 unsigned long slack
= 0;
758 /* Optimise the no-wait case */
759 if (end_time
&& !end_time
->tv_sec
&& !end_time
->tv_nsec
) {
764 if (end_time
&& !timed_out
)
765 slack
= select_estimate_accuracy(end_time
);
768 struct poll_list
*walk
;
770 for (walk
= list
; walk
!= NULL
; walk
= walk
->next
) {
771 struct pollfd
* pfd
, * pfd_end
;
774 pfd_end
= pfd
+ walk
->len
;
775 for (; pfd
!= pfd_end
; pfd
++) {
777 * Fish for events. If we found one, record it
778 * and kill poll_table->_qproc, so we don't
779 * needlessly register any other waiters after
780 * this. They'll get immediately deregistered
781 * when we break out and return.
783 if (do_pollfd(pfd
, pt
)) {
790 * All waiters have already been registered, so don't provide
791 * a poll_table->_qproc to them on the next loop iteration.
796 if (signal_pending(current
))
799 if (count
|| timed_out
)
803 * If this is the first loop and we have a timeout
804 * given, then we convert to ktime_t and set the to
805 * pointer to the expiry value.
807 if (end_time
&& !to
) {
808 expire
= timespec_to_ktime(*end_time
);
812 if (!poll_schedule_timeout(wait
, TASK_INTERRUPTIBLE
, to
, slack
))
818 #define N_STACK_PPS ((sizeof(stack_pps) - sizeof(struct poll_list)) / \
819 sizeof(struct pollfd))
821 int do_sys_poll(struct pollfd __user
*ufds
, unsigned int nfds
,
822 struct timespec
*end_time
)
824 struct poll_wqueues table
;
825 int err
= -EFAULT
, fdcount
, len
, size
;
826 /* Allocate small arguments on the stack to save memory and be
827 faster - use long to make sure the buffer is aligned properly
828 on 64 bit archs to avoid unaligned access */
829 long stack_pps
[POLL_STACK_ALLOC
/sizeof(long)];
830 struct poll_list
*const head
= (struct poll_list
*)stack_pps
;
831 struct poll_list
*walk
= head
;
832 unsigned long todo
= nfds
;
834 if (nfds
> rlimit(RLIMIT_NOFILE
))
837 len
= min_t(unsigned int, nfds
, N_STACK_PPS
);
844 if (copy_from_user(walk
->entries
, ufds
+ nfds
-todo
,
845 sizeof(struct pollfd
) * walk
->len
))
852 len
= min(todo
, POLLFD_PER_PAGE
);
853 size
= sizeof(struct poll_list
) + sizeof(struct pollfd
) * len
;
854 walk
= walk
->next
= kmalloc(size
, GFP_KERNEL
);
861 poll_initwait(&table
);
862 fdcount
= do_poll(nfds
, head
, &table
, end_time
);
863 poll_freewait(&table
);
865 for (walk
= head
; walk
; walk
= walk
->next
) {
866 struct pollfd
*fds
= walk
->entries
;
869 for (j
= 0; j
< walk
->len
; j
++, ufds
++)
870 if (__put_user(fds
[j
].revents
, &ufds
->revents
))
878 struct poll_list
*pos
= walk
;
886 static long do_restart_poll(struct restart_block
*restart_block
)
888 struct pollfd __user
*ufds
= restart_block
->poll
.ufds
;
889 int nfds
= restart_block
->poll
.nfds
;
890 struct timespec
*to
= NULL
, end_time
;
893 if (restart_block
->poll
.has_timeout
) {
894 end_time
.tv_sec
= restart_block
->poll
.tv_sec
;
895 end_time
.tv_nsec
= restart_block
->poll
.tv_nsec
;
899 ret
= do_sys_poll(ufds
, nfds
, to
);
902 restart_block
->fn
= do_restart_poll
;
903 ret
= -ERESTART_RESTARTBLOCK
;
908 SYSCALL_DEFINE3(poll
, struct pollfd __user
*, ufds
, unsigned int, nfds
,
911 struct timespec end_time
, *to
= NULL
;
914 if (timeout_msecs
>= 0) {
916 poll_select_set_timeout(to
, timeout_msecs
/ MSEC_PER_SEC
,
917 NSEC_PER_MSEC
* (timeout_msecs
% MSEC_PER_SEC
));
920 ret
= do_sys_poll(ufds
, nfds
, to
);
923 struct restart_block
*restart_block
;
925 restart_block
= ¤t_thread_info()->restart_block
;
926 restart_block
->fn
= do_restart_poll
;
927 restart_block
->poll
.ufds
= ufds
;
928 restart_block
->poll
.nfds
= nfds
;
930 if (timeout_msecs
>= 0) {
931 restart_block
->poll
.tv_sec
= end_time
.tv_sec
;
932 restart_block
->poll
.tv_nsec
= end_time
.tv_nsec
;
933 restart_block
->poll
.has_timeout
= 1;
935 restart_block
->poll
.has_timeout
= 0;
937 ret
= -ERESTART_RESTARTBLOCK
;
942 SYSCALL_DEFINE5(ppoll
, struct pollfd __user
*, ufds
, unsigned int, nfds
,
943 struct timespec __user
*, tsp
, const sigset_t __user
*, sigmask
,
946 sigset_t ksigmask
, sigsaved
;
947 struct timespec ts
, end_time
, *to
= NULL
;
951 if (copy_from_user(&ts
, tsp
, sizeof(ts
)))
955 if (poll_select_set_timeout(to
, ts
.tv_sec
, ts
.tv_nsec
))
960 /* XXX: Don't preclude handling different sized sigset_t's. */
961 if (sigsetsize
!= sizeof(sigset_t
))
963 if (copy_from_user(&ksigmask
, sigmask
, sizeof(ksigmask
)))
966 sigdelsetmask(&ksigmask
, sigmask(SIGKILL
)|sigmask(SIGSTOP
));
967 sigprocmask(SIG_SETMASK
, &ksigmask
, &sigsaved
);
970 ret
= do_sys_poll(ufds
, nfds
, to
);
972 /* We can restart this syscall, usually */
975 * Don't restore the signal mask yet. Let do_signal() deliver
976 * the signal on the way back to userspace, before the signal
980 memcpy(¤t
->saved_sigmask
, &sigsaved
,
982 set_restore_sigmask();
984 ret
= -ERESTARTNOHAND
;
986 sigprocmask(SIG_SETMASK
, &sigsaved
, NULL
);
988 ret
= poll_select_copy_remaining(&end_time
, tsp
, 0, ret
);