2 * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
28 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
30 * Copyright (c) 1982, 1986, 1989, 1993
31 * The Regents of the University of California. All rights reserved.
32 * (c) UNIX System Laboratories, Inc.
33 * All or some portions of this file are derived from material licensed
34 * to the University of California by American Telephone and Telegraph
35 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
36 * the permission of UNIX System Laboratories, Inc.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)sys_generic.c 8.9 (Berkeley) 2/14/95
69 * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce
70 * support for mandatory and extensible security protections. This notice
71 * is included in support of clause 2.2 (b) of the Apple Public License,
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/filedesc.h>
78 #include <sys/ioctl.h>
79 #include <sys/file_internal.h>
80 #include <sys/proc_internal.h>
81 #include <sys/socketvar.h>
82 #include <sys/uio_internal.h>
83 #include <sys/kernel.h>
85 #include <sys/malloc.h>
86 #include <sys/sysproto.h>
88 #include <sys/mount_internal.h>
89 #include <sys/protosw.h>
92 #include <sys/kdebug.h>
94 #include <sys/event.h>
95 #include <sys/eventvar.h>
97 #include <mach/mach_types.h>
98 #include <kern/kern_types.h>
99 #include <kern/assert.h>
100 #include <kern/kalloc.h>
101 #include <kern/thread.h>
102 #include <kern/clock.h>
104 #include <sys/mbuf.h>
105 #include <sys/socket.h>
106 #include <sys/socketvar.h>
107 #include <sys/errno.h>
108 #include <sys/syscall.h>
109 #include <sys/pipe.h>
111 #include <bsm/audit_kernel.h>
114 #include <net/route.h>
116 #include <netinet/in.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/ip.h>
119 #include <netinet/in_pcb.h>
120 #include <netinet/ip_var.h>
121 #include <netinet/ip6.h>
122 #include <netinet/tcp.h>
123 #include <netinet/tcp_fsm.h>
124 #include <netinet/tcp_seq.h>
125 #include <netinet/tcp_timer.h>
126 #include <netinet/tcp_var.h>
127 #include <netinet/tcpip.h>
128 #include <netinet/tcp_debug.h>
129 /* for wait queue based select */
130 #include <kern/wait_queue.h>
131 #include <kern/kalloc.h>
132 #include <sys/vnode_internal.h>
134 /* XXX should be in a header file somewhere */
135 void evsofree(struct socket
*);
136 void evpipefree(struct pipe
*);
137 void postpipeevent(struct pipe
*, int);
138 void postevent(struct socket
*, struct sockbuf
*, int);
139 extern kern_return_t
IOBSDGetPlatformUUID(__darwin_uuid_t uuid
, mach_timespec_t timeoutp
);
141 int rd_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
);
142 int wr_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
);
143 extern void *get_bsduthreadarg(thread_t
);
144 extern int *get_bsduthreadrval(thread_t
);
146 __private_extern__
int dofileread(vfs_context_t ctx
, struct fileproc
*fp
,
147 user_addr_t bufp
, user_size_t nbyte
,
148 off_t offset
, int flags
, user_ssize_t
*retval
);
149 __private_extern__
int dofilewrite(vfs_context_t ctx
, struct fileproc
*fp
,
150 user_addr_t bufp
, user_size_t nbyte
,
151 off_t offset
, int flags
, user_ssize_t
*retval
);
152 __private_extern__
int preparefileread(struct proc
*p
, struct fileproc
**fp_ret
, int fd
, int check_for_vnode
);
153 __private_extern__
void donefileread(struct proc
*p
, struct fileproc
*fp_ret
, int fd
);
156 extern int appletalk_inited
;
159 #define f_flag f_fglob->fg_flag
160 #define f_type f_fglob->fg_type
161 #define f_msgcount f_fglob->fg_msgcount
162 #define f_cred f_fglob->fg_cred
163 #define f_ops f_fglob->fg_ops
164 #define f_offset f_fglob->fg_offset
165 #define f_data f_fglob->fg_data
171 * preparefileread:EBADF
172 * preparefileread:ESPIPE
173 * preparefileread:ENXIO
174 * preparefileread:EBADF
178 read(struct proc
*p
, struct read_args
*uap
, user_ssize_t
*retval
)
180 __pthread_testcancel(1);
181 return(read_nocancel(p
, (struct read_nocancel_args
*)uap
, retval
));
185 read_nocancel(struct proc
*p
, struct read_nocancel_args
*uap
, user_ssize_t
*retval
)
191 if ( (error
= preparefileread(p
, &fp
, fd
, 0)) )
194 error
= dofileread(vfs_context_current(), fp
, uap
->cbuf
, uap
->nbyte
,
195 (off_t
)-1, 0, retval
);
197 donefileread(p
, fp
, fd
);
206 * preparefileread:EBADF
207 * preparefileread:ESPIPE
208 * preparefileread:ENXIO
209 * preparefileread:EBADF
213 pread(struct proc
*p
, struct pread_args
*uap
, user_ssize_t
*retval
)
215 __pthread_testcancel(1);
216 return(pread_nocancel(p
, (struct pread_nocancel_args
*)uap
, retval
));
220 pread_nocancel(struct proc
*p
, struct pread_nocancel_args
*uap
, user_ssize_t
*retval
)
222 struct fileproc
*fp
= NULL
; /* fp set by preparefileread() */
226 if ( (error
= preparefileread(p
, &fp
, fd
, 1)) )
229 error
= dofileread(vfs_context_current(), fp
, uap
->buf
, uap
->nbyte
,
230 uap
->offset
, FOF_OFFSET
, retval
);
232 donefileread(p
, fp
, fd
);
235 KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO
, SYS_pread
) | DBG_FUNC_NONE
),
236 uap
->fd
, uap
->nbyte
, (unsigned int)((uap
->offset
>> 32)), (unsigned int)(uap
->offset
), 0);
242 * Code common for read and pread
246 donefileread(struct proc
*p
, struct fileproc
*fp
, int fd
)
250 fp
->f_flags
&= ~FP_INCHRREAD
;
252 fp_drop(p
, fd
, fp
, 1);
265 preparefileread(struct proc
*p
, struct fileproc
**fp_ret
, int fd
, int check_for_pread
)
273 error
= fp_lookup(p
, fd
, &fp
, 1);
279 if ((fp
->f_flag
& FREAD
) == 0) {
283 if (check_for_pread
&& (fp
->f_type
!= DTYPE_VNODE
)) {
287 if (fp
->f_type
== DTYPE_VNODE
) {
288 vp
= (struct vnode
*)fp
->f_fglob
->fg_data
;
290 if (check_for_pread
&& (vnode_isfifo(vp
))) {
294 if (check_for_pread
&& (vp
->v_flag
& VISTTY
)) {
298 if (vp
->v_type
== VCHR
)
299 fp
->f_flags
|= FP_INCHRREAD
;
308 fp_drop(p
, fd
, fp
, 1);
319 __private_extern__
int
320 dofileread(vfs_context_t ctx
, struct fileproc
*fp
,
321 user_addr_t bufp
, user_size_t nbyte
, off_t offset
, int flags
,
322 user_ssize_t
*retval
)
325 user_ssize_t bytecnt
;
327 char uio_buf
[ UIO_SIZEOF(1) ];
329 // LP64todo - do we want to raise this?
333 if (IS_64BIT_PROCESS(vfs_context_proc(ctx
))) {
334 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE64
, UIO_READ
,
335 &uio_buf
[0], sizeof(uio_buf
));
337 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE32
, UIO_READ
,
338 &uio_buf
[0], sizeof(uio_buf
));
340 uio_addiov(auio
, bufp
, nbyte
);
344 if ((error
= fo_read(fp
, auio
, flags
, ctx
))) {
345 if (uio_resid(auio
) != bytecnt
&& (error
== ERESTART
||
346 error
== EINTR
|| error
== EWOULDBLOCK
))
349 bytecnt
-= uio_resid(auio
);
357 * Scatter read system call.
366 readv(struct proc
*p
, struct readv_args
*uap
, user_ssize_t
*retval
)
368 __pthread_testcancel(1);
369 return(readv_nocancel(p
, (struct readv_nocancel_args
*)uap
, retval
));
373 readv_nocancel(struct proc
*p
, struct readv_nocancel_args
*uap
, user_ssize_t
*retval
)
378 struct user_iovec
*iovp
;
380 /* Verify range bedfore calling uio_create() */
381 if (uap
->iovcnt
<= 0 || uap
->iovcnt
> UIO_MAXIOV
)
384 /* allocate a uio large enough to hold the number of iovecs passed */
385 auio
= uio_create(uap
->iovcnt
, 0,
386 (IS_64BIT_PROCESS(p
) ? UIO_USERSPACE64
: UIO_USERSPACE32
),
389 /* get location of iovecs within the uio. then copyin the iovecs from
392 iovp
= uio_iovsaddr(auio
);
395 goto ExitThisRoutine
;
397 size_of_iovec
= (IS_64BIT_PROCESS(p
) ? sizeof(struct user_iovec
) : sizeof(struct iovec
));
398 error
= copyin(uap
->iovp
, (caddr_t
)iovp
, (uap
->iovcnt
* size_of_iovec
));
400 goto ExitThisRoutine
;
403 /* finalize uio_t for use and do the IO
405 uio_calculateresid(auio
);
406 error
= rd_uio(p
, uap
->fd
, auio
, retval
);
424 write(struct proc
*p
, struct write_args
*uap
, user_ssize_t
*retval
)
426 __pthread_testcancel(1);
427 return(write_nocancel(p
, (struct write_nocancel_args
*)uap
, retval
));
432 write_nocancel(struct proc
*p
, struct write_nocancel_args
*uap
, user_ssize_t
*retval
)
438 error
= fp_lookup(p
,fd
,&fp
,0);
441 if ((fp
->f_flag
& FWRITE
) == 0) {
444 struct vfs_context context
= *(vfs_context_current());
445 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
447 error
= dofilewrite(&context
, fp
, uap
->cbuf
, uap
->nbyte
,
448 (off_t
)-1, 0, retval
);
451 fp_drop_written(p
, fd
, fp
);
453 fp_drop(p
, fd
, fp
, 0);
469 pwrite(struct proc
*p
, struct pwrite_args
*uap
, user_ssize_t
*retval
)
471 __pthread_testcancel(1);
472 return(pwrite_nocancel(p
, (struct pwrite_nocancel_args
*)uap
, retval
));
476 pwrite_nocancel(struct proc
*p
, struct pwrite_nocancel_args
*uap
, user_ssize_t
*retval
)
481 vnode_t vp
= (vnode_t
)0;
483 error
= fp_lookup(p
,fd
,&fp
,0);
487 if ((fp
->f_flag
& FWRITE
) == 0) {
490 struct vfs_context context
= *vfs_context_current();
491 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
493 if (fp
->f_type
!= DTYPE_VNODE
) {
497 vp
= (vnode_t
)fp
->f_fglob
->fg_data
;
498 if (vnode_isfifo(vp
)) {
502 if ((vp
->v_flag
& VISTTY
)) {
506 if (uap
->offset
== (off_t
)-1) {
511 error
= dofilewrite(&context
, fp
, uap
->buf
, uap
->nbyte
,
512 uap
->offset
, FOF_OFFSET
, retval
);
516 fp_drop_written(p
, fd
, fp
);
518 fp_drop(p
, fd
, fp
, 0);
521 KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO
, SYS_pwrite
) | DBG_FUNC_NONE
),
522 uap
->fd
, uap
->nbyte
, (unsigned int)((uap
->offset
>> 32)), (unsigned int)(uap
->offset
), 0);
531 * <fo_write>:??? [indirect through struct fileops]
533 __private_extern__
int
534 dofilewrite(vfs_context_t ctx
, struct fileproc
*fp
,
535 user_addr_t bufp
, user_size_t nbyte
, off_t offset
, int flags
,
536 user_ssize_t
*retval
)
540 user_ssize_t bytecnt
;
541 char uio_buf
[ UIO_SIZEOF(1) ];
543 // LP64todo - do we want to raise this?
547 if (IS_64BIT_PROCESS(vfs_context_proc(ctx
))) {
548 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE64
, UIO_WRITE
,
549 &uio_buf
[0], sizeof(uio_buf
));
551 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE32
, UIO_WRITE
,
552 &uio_buf
[0], sizeof(uio_buf
));
554 uio_addiov(auio
, bufp
, nbyte
);
557 if ((error
= fo_write(fp
, auio
, flags
, ctx
))) {
558 if (uio_resid(auio
) != bytecnt
&& (error
== ERESTART
||
559 error
== EINTR
|| error
== EWOULDBLOCK
))
561 /* The socket layer handles SIGPIPE */
562 if (error
== EPIPE
&& fp
->f_type
!= DTYPE_SOCKET
) {
563 /* XXX Raise the signal on the thread? */
564 psignal(vfs_context_proc(ctx
), SIGPIPE
);
567 bytecnt
-= uio_resid(auio
);
574 * Gather write system call
577 writev(struct proc
*p
, struct writev_args
*uap
, user_ssize_t
*retval
)
579 __pthread_testcancel(1);
580 return(writev_nocancel(p
, (struct writev_nocancel_args
*)uap
, retval
));
584 writev_nocancel(struct proc
*p
, struct writev_nocancel_args
*uap
, user_ssize_t
*retval
)
589 struct user_iovec
*iovp
;
591 /* Verify range bedfore calling uio_create() */
592 if (uap
->iovcnt
<= 0 || uap
->iovcnt
> UIO_MAXIOV
)
595 /* allocate a uio large enough to hold the number of iovecs passed */
596 auio
= uio_create(uap
->iovcnt
, 0,
597 (IS_64BIT_PROCESS(p
) ? UIO_USERSPACE64
: UIO_USERSPACE32
),
600 /* get location of iovecs within the uio. then copyin the iovecs from
603 iovp
= uio_iovsaddr(auio
);
606 goto ExitThisRoutine
;
608 size_of_iovec
= (IS_64BIT_PROCESS(p
) ? sizeof(struct user_iovec
) : sizeof(struct iovec
));
609 error
= copyin(uap
->iovp
, (caddr_t
)iovp
, (uap
->iovcnt
* size_of_iovec
));
611 goto ExitThisRoutine
;
614 /* finalize uio_t for use and do the IO
616 uio_calculateresid(auio
);
617 error
= wr_uio(p
, uap
->fd
, auio
, retval
);
628 wr_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
)
633 struct vfs_context context
= *vfs_context_current();
635 error
= fp_lookup(p
,fdes
,&fp
,0);
639 if ((fp
->f_flag
& FWRITE
) == 0) {
643 count
= uio_resid(uio
);
645 context
.vc_ucred
= fp
->f_cred
;
646 error
= fo_write(fp
, uio
, 0, &context
);
648 if (uio_resid(uio
) != count
&& (error
== ERESTART
||
649 error
== EINTR
|| error
== EWOULDBLOCK
))
651 /* The socket layer handles SIGPIPE */
652 if (error
== EPIPE
&& fp
->f_type
!= DTYPE_SOCKET
)
655 *retval
= count
- uio_resid(uio
);
659 fp_drop_written(p
, fdes
, fp
);
661 fp_drop(p
, fdes
, fp
, 0);
667 rd_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
)
672 struct vfs_context context
= *vfs_context_current();
674 if ( (error
= preparefileread(p
, &fp
, fdes
, 0)) )
677 count
= uio_resid(uio
);
679 context
.vc_ucred
= fp
->f_cred
;
681 error
= fo_read(fp
, uio
, 0, &context
);
684 if (uio_resid(uio
) != count
&& (error
== ERESTART
||
685 error
== EINTR
|| error
== EWOULDBLOCK
))
688 *retval
= count
- uio_resid(uio
);
690 donefileread(p
, fp
, fdes
);
705 * fp_lookup:EBADF Bad file descriptor
709 ioctl(struct proc
*p
, struct ioctl_args
*uap
, __unused register_t
*retval
)
718 #define STK_PARAMS 128
719 char stkbuf
[STK_PARAMS
];
721 struct vfs_context context
= *vfs_context_current();
723 AUDIT_ARG(fd
, uap
->fd
);
724 AUDIT_ARG(cmd
, CAST_DOWN(int, uap
->com
)); /* LP64todo: uap->com is a user-land long */
725 AUDIT_ARG(addr
, uap
->data
);
727 is64bit
= proc_is64bit(p
);
730 error
= fp_lookup(p
,fd
,&fp
,1);
736 AUDIT_ARG(file
, p
, fp
);
738 if ((fp
->f_flag
& (FREAD
| FWRITE
)) == 0) {
743 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
746 error
= mac_file_check_ioctl(context
.vc_ucred
, fp
->f_fglob
, uap
->com
);
753 * ### LD 6/11/97 Hack Alert: this is to get AppleTalk to work
754 * while implementing an ATioctl system call
757 if (appletalk_inited
&& ((uap
->com
& 0x0000FFFF) == 0xff99)) {
758 u_long fixed_command
;
760 #ifdef APPLETALK_DEBUG
761 kprintf("ioctl: special AppleTalk \n");
764 *(user_addr_t
*)datap
= uap
->data
;
765 fixed_command
= _IOW(0, 0xff99, uap
->data
);
766 error
= fo_ioctl(fp
, fixed_command
, datap
, &context
);
774 switch (com
= uap
->com
) {
776 *fdflags(p
, uap
->fd
) &= ~UF_EXCLOSE
;
780 *fdflags(p
, uap
->fd
) |= UF_EXCLOSE
;
786 * Interpret high order word to find amount of data to be
787 * copied to/from the user's address space.
789 size
= IOCPARM_LEN(com
);
790 if (size
> IOCPARM_MAX
) {
795 if (size
> sizeof (stkbuf
)) {
797 if ((memp
= (caddr_t
)kalloc(size
)) == 0) {
809 error
= copyin(uap
->data
, datap
, size
);
818 /* XXX - IOC_IN and no size? we should proably return an error here!! */
820 *(user_addr_t
*)datap
= uap
->data
;
823 *(uint32_t *)datap
= (uint32_t)uap
->data
;
826 } else if ((com
&IOC_OUT
) && size
)
828 * Zero the buffer so the user always
829 * gets back something deterministic.
832 else if (com
&IOC_VOID
) {
833 /* XXX - this is odd since IOC_VOID means no parameters */
835 *(user_addr_t
*)datap
= uap
->data
;
838 *(uint32_t *)datap
= (uint32_t)uap
->data
;
845 if ( (tmp
= *(int *)datap
) )
846 fp
->f_flag
|= FNONBLOCK
;
848 fp
->f_flag
&= ~FNONBLOCK
;
849 error
= fo_ioctl(fp
, FIONBIO
, (caddr_t
)&tmp
, &context
);
853 if ( (tmp
= *(int *)datap
) )
854 fp
->f_flag
|= FASYNC
;
856 fp
->f_flag
&= ~FASYNC
;
857 error
= fo_ioctl(fp
, FIOASYNC
, (caddr_t
)&tmp
, &context
);
862 if (fp
->f_type
== DTYPE_SOCKET
) {
863 ((struct socket
*)fp
->f_data
)->so_pgid
= tmp
;
867 if (fp
->f_type
== DTYPE_PIPE
) {
868 error
= fo_ioctl(fp
, (int)TIOCSPGRP
, (caddr_t
)&tmp
, &context
);
874 struct proc
*p1
= proc_find(tmp
);
882 error
= fo_ioctl(fp
, (int)TIOCSPGRP
, (caddr_t
)&tmp
, &context
);
886 if (fp
->f_type
== DTYPE_SOCKET
) {
888 *(int *)datap
= ((struct socket
*)fp
->f_data
)->so_pgid
;
891 error
= fo_ioctl(fp
, TIOCGPGRP
, datap
, &context
);
892 *(int *)datap
= -*(int *)datap
;
896 error
= fo_ioctl(fp
, com
, datap
, &context
);
898 * Copy any data to user, size was
899 * already set and checked above.
901 if (error
== 0 && (com
&IOC_OUT
) && size
)
902 error
= copyout(datap
, uap
->data
, (u_int
)size
);
910 fp_drop(p
, fd
, fp
, 1);
915 int selwait
, nselcoll
;
916 #define SEL_FIRSTPASS 1
917 #define SEL_SECONDPASS 2
918 extern int selcontinue(int error
);
919 extern int selprocess(int error
, int sel_pass
);
920 static int selscan(struct proc
*p
, struct _select
* sel
,
921 int nfd
, register_t
*retval
, int sel_pass
, wait_queue_sub_t wqsub
);
922 static int selcount(struct proc
*p
, u_int32_t
*ibits
, u_int32_t
*obits
,
923 int nfd
, int * count
, int *kfcount
);
924 static int seldrop(struct proc
*p
, u_int32_t
*ibits
, int nfd
);
925 extern uint64_t tvtoabstime(struct timeval
*tvp
);
928 * Select system call.
931 * EINVAL Invalid argument
932 * EAGAIN Nonconformant error if allocation fails
936 select(struct proc
*p
, struct select_args
*uap
, register_t
*retval
)
938 __pthread_testcancel(1);
939 return(select_nocancel(p
, (struct select_nocancel_args
*)uap
, retval
));
943 select_nocancel(struct proc
*p
, struct select_nocancel_args
*uap
, register_t
*retval
)
950 int needzerofill
= 1;
954 th_act
= current_thread();
955 uth
= get_bsdthread_info(th_act
);
956 sel
= &uth
->uu_select
;
957 retval
= (int *)get_bsduthreadrval(th_act
);
964 /* select on thread of process that already called proc_exit() */
965 if (p
->p_fd
== NULL
) {
969 if (uap
->nd
> p
->p_fd
->fd_nfiles
)
970 uap
->nd
= p
->p_fd
->fd_nfiles
; /* forgiving; slightly wrong */
972 nw
= howmany(uap
->nd
, NFDBITS
);
973 ni
= nw
* sizeof(fd_mask
);
976 * if the previously allocated space for the bits is smaller than
977 * what is requested or no space has yet been allocated for this
978 * thread, allocate enough space now.
980 * Note: If this process fails, select() will return EAGAIN; this
981 * is the same thing pool() returns in a no-memory situation, but
982 * it is not a POSIX compliant error code for select().
984 if (sel
->nbytes
< (3 * ni
)) {
987 /* Free previous allocation, if any */
988 if (sel
->ibits
!= NULL
)
989 FREE(sel
->ibits
, M_TEMP
);
990 if (sel
->obits
!= NULL
) {
991 FREE(sel
->obits
, M_TEMP
);
992 /* NULL out; subsequent ibits allocation may fail */
996 MALLOC(sel
->ibits
, u_int32_t
*, nbytes
, M_TEMP
, M_WAITOK
| M_ZERO
);
997 if (sel
->ibits
== NULL
)
999 MALLOC(sel
->obits
, u_int32_t
*, nbytes
, M_TEMP
, M_WAITOK
| M_ZERO
);
1000 if (sel
->obits
== NULL
) {
1001 FREE(sel
->ibits
, M_TEMP
);
1005 sel
->nbytes
= nbytes
;
1010 bzero((caddr_t
)sel
->ibits
, sel
->nbytes
);
1011 bzero((caddr_t
)sel
->obits
, sel
->nbytes
);
1015 * get the bits from the user address space
1017 #define getbits(name, x) \
1019 if (uap->name && (error = copyin(uap->name, \
1020 (caddr_t)&sel->ibits[(x) * nw], ni))) \
1021 goto continuation; \
1031 if (IS_64BIT_PROCESS(p
)) {
1032 struct user_timeval atv64
;
1033 error
= copyin(uap
->tv
, (caddr_t
)&atv64
, sizeof(atv64
));
1034 /* Loses resolution - assume timeout < 68 years */
1035 atv
.tv_sec
= atv64
.tv_sec
;
1036 atv
.tv_usec
= atv64
.tv_usec
;
1038 error
= copyin(uap
->tv
, (caddr_t
)&atv
, sizeof(atv
));
1042 if (itimerfix(&atv
)) {
1047 clock_absolutetime_interval_to_deadline(
1048 tvtoabstime(&atv
), &sel
->abstime
);
1054 if ( (error
= selcount(p
, sel
->ibits
, sel
->obits
, uap
->nd
, &count
, &kfcount
)) ) {
1058 sel
->kfcount
= kfcount
;
1059 size
= SIZEOF_WAITQUEUE_SET
+ (count
* SIZEOF_WAITQUEUE_LINK
);
1060 if (uth
->uu_allocsize
) {
1061 if (uth
->uu_wqset
== 0)
1062 panic("select: wql memory smashed");
1063 /* needed for the select now */
1064 if (size
> uth
->uu_allocsize
) {
1065 kfree(uth
->uu_wqset
, uth
->uu_allocsize
);
1066 uth
->uu_allocsize
= size
;
1067 uth
->uu_wqset
= (wait_queue_set_t
)kalloc(size
);
1068 if (uth
->uu_wqset
== (wait_queue_set_t
)NULL
)
1069 panic("failed to allocate memory for waitqueue\n");
1073 uth
->uu_allocsize
= size
;
1074 uth
->uu_wqset
= (wait_queue_set_t
)kalloc(uth
->uu_allocsize
);
1075 if (uth
->uu_wqset
== (wait_queue_set_t
)NULL
)
1076 panic("failed to allocate memory for waitqueue\n");
1078 bzero(uth
->uu_wqset
, size
);
1079 sel
->wql
= (char *)uth
->uu_wqset
+ SIZEOF_WAITQUEUE_SET
;
1080 wait_queue_set_init(uth
->uu_wqset
, (SYNC_POLICY_FIFO
| SYNC_POLICY_PREPOST
));
1083 return selprocess(error
, SEL_FIRSTPASS
);
1087 selcontinue(int error
)
1089 return selprocess(error
, SEL_SECONDPASS
);
1093 selprocess(int error
, int sel_pass
)
1098 struct uthread
*uth
;
1100 struct select_args
*uap
;
1102 struct _select
*sel
;
1107 wait_result_t wait_result
;
1110 th_act
= current_thread();
1111 uap
= (struct select_args
*)get_bsduthreadarg(th_act
);
1112 retval
= (int *)get_bsduthreadrval(th_act
);
1113 uth
= get_bsdthread_info(th_act
);
1114 sel
= &uth
->uu_select
;
1116 /* if it is first pass wait queue is not setup yet */
1117 if ((error
!= 0) && (sel_pass
== SEL_FIRSTPASS
))
1119 if (sel
->count
== 0)
1127 OSBitOrAtomic(P_SELECT
, (UInt32
*)&p
->p_flag
);
1128 /* skip scans if the select is just for timeouts */
1130 if (sel_pass
== SEL_FIRSTPASS
)
1131 wait_queue_sub_clearrefs(uth
->uu_wqset
);
1133 error
= selscan(p
, sel
, uap
->nd
, retval
, sel_pass
, (wait_queue_sub_t
)uth
->uu_wqset
);
1134 if (error
|| *retval
) {
1138 /* if the select of log, then we canwakeup and discover some one
1139 * else already read the data; go toselct again if time permits
1153 clock_get_uptime(&now
);
1154 if (now
>= sel
->abstime
)
1159 /* cleanup obits and try again */
1161 sel_pass
= SEL_FIRSTPASS
;
1166 * To effect a poll, the timeout argument should be
1167 * non-nil, pointing to a zero-valued timeval structure.
1169 if (uap
->tv
&& sel
->abstime
== 0) {
1173 /* No spurious wakeups due to colls,no need to check for them */
1174 if ((sel_pass
== SEL_SECONDPASS
) || ((p
->p_flag
& P_SELECT
) == 0)) {
1175 sel_pass
= SEL_FIRSTPASS
;
1179 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1181 /* if the select is just for timeout skip check */
1182 if (sel
->count
&&(sel_pass
== SEL_SECONDPASS
))
1183 panic("selprocess: 2nd pass assertwaiting");
1185 /* Wait Queue Subordinate has waitqueue as first element */
1186 wait_result
= wait_queue_assert_wait((wait_queue_t
)uth
->uu_wqset
,
1187 &selwait
, THREAD_ABORTSAFE
, sel
->abstime
);
1188 if (wait_result
!= THREAD_AWAKENED
) {
1189 /* there are no preposted events */
1190 error
= tsleep1(NULL
, PSOCK
| PCATCH
,
1191 "select", 0, selcontinue
);
1197 sel_pass
= SEL_SECONDPASS
;
1205 wait_subqueue_unlink_all(uth
->uu_wqset
);
1206 seldrop(p
, sel
->ibits
, uap
->nd
);
1208 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1209 /* select is not restarted after signals... */
1210 if (error
== ERESTART
)
1212 if (error
== EWOULDBLOCK
)
1214 nw
= howmany(uap
->nd
, NFDBITS
);
1215 ni
= nw
* sizeof(fd_mask
);
1217 #define putbits(name, x) \
1219 if (uap->name && (error2 = \
1220 copyout((caddr_t)&sel->obits[(x) * nw], uap->name, ni))) \
1236 selscan(struct proc
*p
, struct _select
*sel
, int nfd
, register_t
*retval
,
1237 int sel_pass
, wait_queue_sub_t wqsub
)
1239 struct filedesc
*fdp
= p
->p_fd
;
1242 struct fileproc
*fp
;
1245 static int flag
[3] = { FREAD
, FWRITE
, 0 };
1246 u_int32_t
*iptr
, *optr
;
1248 u_int32_t
*ibits
, *obits
;
1252 boolean_t funnel_state
;
1254 struct vfs_context context
= *vfs_context_current();
1257 * Problems when reboot; due to MacOSX signal probs
1258 * in Beaker1C ; verify that the p->p_fd is valid
1268 nw
= howmany(nfd
, NFDBITS
);
1271 kfcount
= sel
->kfcount
;
1273 if (kfcount
> count
)
1274 panic("selscan: count < kfcount");
1277 funnel_state
= thread_funnel_set(kernel_flock
, TRUE
);
1280 for (msk
= 0; msk
< 3; msk
++) {
1281 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1282 optr
= (u_int32_t
*)&obits
[msk
* nw
];
1284 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1285 bits
= iptr
[i
/NFDBITS
];
1287 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1289 fp
= fdp
->fd_ofiles
[fd
];
1292 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1294 thread_funnel_set(kernel_flock
, funnel_state
);
1297 if (sel_pass
== SEL_SECONDPASS
) {
1298 wql_ptr
= (char *)0;
1299 fp
->f_flags
&= ~FP_INSELECT
;
1300 fp
->f_waddr
= (void *)0;
1302 wql_ptr
= (wql
+ nc
* SIZEOF_WAITQUEUE_LINK
);
1303 fp
->f_flags
|= FP_INSELECT
;
1304 fp
->f_waddr
= (void *)wqsub
;
1307 context
.vc_ucred
= fp
->f_cred
;
1309 if (fp
->f_ops
&& (fp
->f_type
== DTYPE_VNODE
)
1310 && ((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1311 && (vp
->v_type
== VCHR
)
1312 && fo_select(fp
, flag
[msk
], wql_ptr
, &context
)) {
1313 optr
[fd
/NFDBITS
] |= (1 << (fd
% NFDBITS
));
1321 thread_funnel_set(kernel_flock
, funnel_state
);
1325 if (kfcount
!= count
) {
1327 for (msk
= 0; msk
< 3; msk
++) {
1328 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1329 optr
= (u_int32_t
*)&obits
[msk
* nw
];
1331 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1332 bits
= iptr
[i
/NFDBITS
];
1334 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1336 fp
= fdp
->fd_ofiles
[fd
];
1339 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1343 if (sel_pass
== SEL_SECONDPASS
) {
1344 wql_ptr
= (char *)0;
1345 fp
->f_flags
&= ~FP_INSELECT
;
1346 fp
->f_waddr
= (void *)0;
1348 wql_ptr
= (wql
+ nc
* SIZEOF_WAITQUEUE_LINK
);
1349 fp
->f_flags
|= FP_INSELECT
;
1350 fp
->f_waddr
= (void *)wqsub
;
1353 context
.vc_ucred
= fp
->f_cred
;
1356 ((fp
->f_type
!= DTYPE_VNODE
)
1357 || (((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1358 && (vp
->v_type
!= VCHR
))
1360 && fo_select(fp
, flag
[msk
], wql_ptr
, &context
))) {
1361 optr
[fd
/NFDBITS
] |= (1 << (fd
% NFDBITS
));
1374 int poll_callback(struct kqueue
*, struct kevent
*, void *);
1376 struct poll_continue_args
{
1377 user_addr_t pca_fds
;
1383 poll(struct proc
*p
, struct poll_args
*uap
, register_t
*retval
)
1385 __pthread_testcancel(1);
1386 return(poll_nocancel(p
, (struct poll_nocancel_args
*)uap
, retval
));
1391 poll_nocancel(struct proc
*p
, struct poll_nocancel_args
*uap
, register_t
*retval
)
1393 struct poll_continue_args
*cont
;
1397 int ncoll
, error
= 0;
1398 u_int nfds
= uap
->nfds
;
1404 * This is kinda bogus. We have fd limits, but that is not
1405 * really related to the size of the pollfd array. Make sure
1406 * we let the process use at least FD_SETSIZE entries and at
1407 * least enough for the current limits. We want to be reasonably
1408 * safe, but not overly restrictive.
1410 if (nfds
> OPEN_MAX
||
1411 (nfds
> p
->p_rlimit
[RLIMIT_NOFILE
].rlim_cur
&& (proc_suser(p
) || nfds
> FD_SETSIZE
)))
1414 kq
= kqueue_alloc(p
);
1418 ni
= nfds
* sizeof(struct pollfd
) + sizeof(struct poll_continue_args
);
1419 MALLOC(cont
, struct poll_continue_args
*, ni
, M_TEMP
, M_WAITOK
);
1425 fds
= (struct pollfd
*)&cont
[1];
1426 error
= copyin(uap
->fds
, fds
, nfds
* sizeof(struct pollfd
));
1430 if (uap
->timeout
!= -1) {
1433 atv
.tv_sec
= uap
->timeout
/ 1000;
1434 atv
.tv_usec
= (uap
->timeout
% 1000) * 1000;
1435 if (itimerfix(&atv
)) {
1439 getmicrouptime(&rtv
);
1440 timevaladd(&atv
, &rtv
);
1446 /* JMM - all this P_SELECT stuff is bogus */
1448 OSBitOrAtomic(P_SELECT
, (UInt32
*)&p
->p_flag
);
1449 for (i
= 0; i
< nfds
; i
++) {
1450 short events
= fds
[i
].events
;
1454 /* per spec, ignore fd values below zero */
1455 if (fds
[i
].fd
< 0) {
1460 /* convert the poll event into a kqueue kevent */
1461 kev
.ident
= fds
[i
].fd
;
1462 kev
.flags
= EV_ADD
| EV_ONESHOT
| EV_POLL
;
1463 kev
.fflags
= NOTE_LOWAT
;
1464 kev
.data
= 1; /* efficiency be damned: any data should trigger */
1465 kev
.udata
= CAST_USER_ADDR_T(&fds
[i
]);
1467 /* Handle input events */
1468 if (events
& ( POLLIN
| POLLRDNORM
| POLLPRI
| POLLRDBAND
| POLLHUP
)) {
1469 kev
.filter
= EVFILT_READ
;
1470 if (!(events
& ( POLLIN
| POLLRDNORM
)))
1471 kev
.flags
|= EV_OOBAND
;
1472 kerror
= kevent_register(kq
, &kev
, p
);
1475 /* Handle output events */
1477 events
& ( POLLOUT
| POLLWRNORM
| POLLWRBAND
)) {
1478 kev
.filter
= EVFILT_WRITE
;
1479 kerror
= kevent_register(kq
, &kev
, p
);
1482 /* Handle BSD extension vnode events */
1484 events
& ( POLLEXTEND
| POLLATTRIB
| POLLNLINK
| POLLWRITE
)) {
1485 kev
.filter
= EVFILT_VNODE
;
1487 if (events
& POLLEXTEND
)
1488 kev
.fflags
|= NOTE_EXTEND
;
1489 if (events
& POLLATTRIB
)
1490 kev
.fflags
|= NOTE_ATTRIB
;
1491 if (events
& POLLNLINK
)
1492 kev
.fflags
|= NOTE_LINK
;
1493 if (events
& POLLWRITE
)
1494 kev
.fflags
|= NOTE_WRITE
;
1495 kerror
= kevent_register(kq
, &kev
, p
);
1499 fds
[i
].revents
= POLLNVAL
;
1505 /* Did we have any trouble registering? */
1509 /* scan for, and possibly wait for, the kevents to trigger */
1510 cont
->pca_fds
= uap
->fds
;
1511 cont
->pca_nfds
= nfds
;
1512 cont
->pca_rfds
= rfds
;
1513 error
= kevent_scan(kq
, poll_callback
, NULL
, cont
, &atv
, p
);
1514 rfds
= cont
->pca_rfds
;
1517 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1518 /* poll is not restarted after signals... */
1519 if (error
== ERESTART
)
1521 if (error
== EWOULDBLOCK
)
1524 error
= copyout(fds
, uap
->fds
, nfds
* sizeof(struct pollfd
));
1537 poll_callback(__unused
struct kqueue
*kq
, struct kevent
*kevp
, void *data
)
1539 struct poll_continue_args
*cont
= (struct poll_continue_args
*)data
;
1540 struct pollfd
*fds
= CAST_DOWN(struct pollfd
*, kevp
->udata
);
1543 /* convert the results back into revents */
1544 if (kevp
->flags
& EV_EOF
)
1545 fds
->revents
|= POLLHUP
;
1546 if (kevp
->flags
& EV_ERROR
)
1547 fds
->revents
|= POLLERR
;
1549 switch (kevp
->filter
) {
1551 if (fds
->revents
& POLLHUP
)
1552 mask
= (POLLIN
| POLLRDNORM
| POLLPRI
| POLLRDBAND
);
1555 if (kevp
->data
!= 0)
1556 mask
|= (POLLIN
| POLLRDNORM
);
1557 if (kevp
->flags
& EV_OOBAND
)
1558 mask
|= ( POLLPRI
| POLLRDBAND
);
1560 fds
->revents
|= (fds
->events
& mask
);
1564 if (!(fds
->revents
& POLLHUP
))
1565 fds
->revents
|= (fds
->events
& ( POLLOUT
| POLLWRNORM
| POLLWRBAND
));
1569 if (kevp
->fflags
& NOTE_EXTEND
)
1570 fds
->revents
|= (fds
->events
& POLLEXTEND
);
1571 if (kevp
->fflags
& NOTE_ATTRIB
)
1572 fds
->revents
|= (fds
->events
& POLLATTRIB
);
1573 if (kevp
->fflags
& NOTE_LINK
)
1574 fds
->revents
|= (fds
->events
& POLLNLINK
);
1575 if (kevp
->fflags
& NOTE_WRITE
)
1576 fds
->revents
|= (fds
->events
& POLLWRITE
);
1587 seltrue(__unused dev_t dev
, __unused
int flag
, __unused
struct proc
*p
)
1594 selcount(struct proc
*p
, u_int32_t
*ibits
, __unused u_int32_t
*obits
,
1595 int nfd
, int *countp
, int * kfcountp
)
1597 struct filedesc
*fdp
= p
->p_fd
;
1600 struct fileproc
*fp
;
1610 * Problems when reboot; due to MacOSX signal probs
1611 * in Beaker1C ; verify that the p->p_fd is valid
1618 nw
= howmany(nfd
, NFDBITS
);
1621 for (msk
= 0; msk
< 3; msk
++) {
1622 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1623 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1624 bits
= iptr
[i
/NFDBITS
];
1625 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1627 fp
= fdp
->fd_ofiles
[fd
];
1629 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1636 if ((fp
->f_type
== DTYPE_VNODE
)
1637 && ((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1638 && (vp
->v_type
== VCHR
) )
1655 /* undo the iocounts */
1656 for (msk
= 0; msk
< 3; msk
++) {
1657 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1658 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1659 bits
= iptr
[i
/NFDBITS
];
1660 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1662 fp
= fdp
->fd_ofiles
[fd
];
1667 if (p
->p_fpdrainwait
&& fp
->f_iocount
== 0) {
1668 p
->p_fpdrainwait
= 0;
1669 wakeup(&p
->p_fpdrainwait
);
1681 seldrop(struct proc
*p
, u_int32_t
*ibits
, int nfd
)
1683 struct filedesc
*fdp
= p
->p_fd
;
1686 struct fileproc
*fp
;
1692 * Problems when reboot; due to MacOSX signal probs
1693 * in Beaker1C ; verify that the p->p_fd is valid
1699 nw
= howmany(nfd
, NFDBITS
);
1703 for (msk
= 0; msk
< 3; msk
++) {
1704 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1705 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1706 bits
= iptr
[i
/NFDBITS
];
1707 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1709 fp
= fdp
->fd_ofiles
[fd
];
1712 /* if you are here then it is being closed */
1713 || (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)
1721 fp
->f_flags
&= ~FP_INSELECT
;
1723 if (p
->p_fpdrainwait
&& fp
->f_iocount
== 0) {
1724 p
->p_fpdrainwait
= 0;
1725 wakeup(&p
->p_fpdrainwait
);
1735 * Record a select request.
1738 selrecord(__unused
struct proc
*selector
, struct selinfo
*sip
, void * p_wql
)
1740 thread_t cur_act
= current_thread();
1741 struct uthread
* ut
= get_bsdthread_info(cur_act
);
1743 /* need to look at collisions */
1745 if ((p_wql
== (void *)0) && ((sip
->si_flags
& SI_INITED
) == 0)) {
1749 /*do not record if this is second pass of select */
1750 if((p_wql
== (void *)0)) {
1754 if ((sip
->si_flags
& SI_INITED
) == 0) {
1755 wait_queue_init(&sip
->si_wait_queue
, SYNC_POLICY_FIFO
);
1756 sip
->si_flags
|= SI_INITED
;
1757 sip
->si_flags
&= ~SI_CLEAR
;
1760 if (sip
->si_flags
& SI_RECORDED
) {
1761 sip
->si_flags
|= SI_COLL
;
1763 sip
->si_flags
&= ~SI_COLL
;
1765 sip
->si_flags
|= SI_RECORDED
;
1766 if (!wait_queue_member(&sip
->si_wait_queue
, ut
->uu_wqset
))
1767 wait_queue_link_noalloc(&sip
->si_wait_queue
, ut
->uu_wqset
,
1768 (wait_queue_link_t
)p_wql
);
1774 selwakeup(struct selinfo
*sip
)
1777 if ((sip
->si_flags
& SI_INITED
) == 0) {
1781 if (sip
->si_flags
& SI_COLL
) {
1783 sip
->si_flags
&= ~SI_COLL
;
1785 /* will not support */
1786 //wakeup((caddr_t)&selwait);
1790 if (sip
->si_flags
& SI_RECORDED
) {
1791 wait_queue_wakeup_all(&sip
->si_wait_queue
, &selwait
, THREAD_AWAKENED
);
1792 sip
->si_flags
&= ~SI_RECORDED
;
1798 selthreadclear(struct selinfo
*sip
)
1801 if ((sip
->si_flags
& SI_INITED
) == 0) {
1804 if (sip
->si_flags
& SI_RECORDED
) {
1806 sip
->si_flags
&= ~(SI_RECORDED
| SI_COLL
);
1808 sip
->si_flags
|= SI_CLEAR
;
1809 wait_queue_unlinkall_nofree(&sip
->si_wait_queue
);
1815 #define DBG_POST 0x10
1816 #define DBG_WATCH 0x11
1817 #define DBG_WAIT 0x12
1818 #define DBG_MOD 0x13
1819 #define DBG_EWAKEUP 0x14
1820 #define DBG_ENQUEUE 0x15
1821 #define DBG_DEQUEUE 0x16
1823 #define DBG_MISC_POST MISCDBG_CODE(DBG_EVENT,DBG_POST)
1824 #define DBG_MISC_WATCH MISCDBG_CODE(DBG_EVENT,DBG_WATCH)
1825 #define DBG_MISC_WAIT MISCDBG_CODE(DBG_EVENT,DBG_WAIT)
1826 #define DBG_MISC_MOD MISCDBG_CODE(DBG_EVENT,DBG_MOD)
1827 #define DBG_MISC_EWAKEUP MISCDBG_CODE(DBG_EVENT,DBG_EWAKEUP)
1828 #define DBG_MISC_ENQUEUE MISCDBG_CODE(DBG_EVENT,DBG_ENQUEUE)
1829 #define DBG_MISC_DEQUEUE MISCDBG_CODE(DBG_EVENT,DBG_DEQUEUE)
1832 #define EVPROCDEQUE(p, evq) do { \
1834 if (evq->ee_flags & EV_QUEUED) { \
1835 TAILQ_REMOVE(&p->p_evlist, evq, ee_plist); \
1836 evq->ee_flags &= ~EV_QUEUED; \
1843 * called upon socket close. deque and free all events for
1844 * the socket... socket must be locked by caller.
1847 evsofree(struct socket
*sp
)
1849 struct eventqelt
*evq
, *next
;
1855 for (evq
= sp
->so_evlist
.tqh_first
; evq
!= NULL
; evq
= next
) {
1856 next
= evq
->ee_slist
.tqe_next
;
1859 if (evq
->ee_flags
& EV_QUEUED
) {
1860 EVPROCDEQUE(p
, evq
);
1862 TAILQ_REMOVE(&sp
->so_evlist
, evq
, ee_slist
); // remove from socket q
1869 * called upon pipe close. deque and free all events for
1870 * the pipe... pipe must be locked by caller
1873 evpipefree(struct pipe
*cpipe
)
1875 struct eventqelt
*evq
, *next
;
1878 for (evq
= cpipe
->pipe_evlist
.tqh_first
; evq
!= NULL
; evq
= next
) {
1879 next
= evq
->ee_slist
.tqe_next
;
1882 EVPROCDEQUE(p
, evq
);
1884 TAILQ_REMOVE(&cpipe
->pipe_evlist
, evq
, ee_slist
); // remove from pipe q
1891 * enqueue this event if it's not already queued. wakeup
1892 * the proc if we do queue this event to it...
1893 * entered with proc lock held... we drop it before
1894 * doing the wakeup and return in that state
1897 evprocenque(struct eventqelt
*evq
)
1904 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_START
, (uint32_t)evq
, evq
->ee_flags
, evq
->ee_eventmask
,0,0);
1908 if (evq
->ee_flags
& EV_QUEUED
) {
1911 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_END
, 0,0,0,0,0);
1914 evq
->ee_flags
|= EV_QUEUED
;
1916 TAILQ_INSERT_TAIL(&p
->p_evlist
, evq
, ee_plist
);
1920 wakeup(&p
->p_evlist
);
1922 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_END
, 0,0,0,0,0);
1927 * pipe lock must be taken by the caller
1930 postpipeevent(struct pipe
*pipep
, int event
)
1933 struct eventqelt
*evq
;
1937 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_START
, event
,0,0,1,0);
1939 for (evq
= pipep
->pipe_evlist
.tqh_first
;
1940 evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
1942 if (evq
->ee_eventmask
== 0)
1946 switch (event
& (EV_RWBYTES
| EV_RCLOSED
| EV_WCLOSED
)) {
1949 if ((evq
->ee_eventmask
& EV_RE
) && pipep
->pipe_buffer
.cnt
) {
1951 evq
->ee_req
.er_rcnt
= pipep
->pipe_buffer
.cnt
;
1953 if ((evq
->ee_eventmask
& EV_WR
) &&
1954 (pipep
->pipe_buffer
.size
- pipep
->pipe_buffer
.cnt
) >= PIPE_BUF
) {
1956 if (pipep
->pipe_state
& PIPE_EOF
) {
1957 mask
|= EV_WR
|EV_RESET
;
1961 evq
->ee_req
.er_wcnt
= pipep
->pipe_buffer
.size
- pipep
->pipe_buffer
.cnt
;
1967 if ((evq
->ee_eventmask
& EV_RE
)) {
1968 mask
|= EV_RE
|EV_RCLOSED
;
1970 if ((evq
->ee_eventmask
& EV_WR
)) {
1971 mask
|= EV_WR
|EV_WCLOSED
;
1980 * disarm... postevents are nops until this event is 'read' via
1981 * waitevent and then re-armed via modwatch
1983 evq
->ee_eventmask
= 0;
1986 * since events are disarmed until after the waitevent
1987 * the ee_req.er_xxxx fields can't change once we've
1988 * inserted this event into the proc queue...
1989 * therefore, the waitevent will see a 'consistent'
1990 * snapshot of the event, even though it won't hold
1991 * the pipe lock, and we're updating the event outside
1992 * of the proc lock, which it will hold
1994 evq
->ee_req
.er_eventbits
|= mask
;
1996 KERNEL_DEBUG(DBG_MISC_POST
, (uint32_t)evq
, evq
->ee_req
.er_eventbits
, mask
, 1,0);
2001 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, 0,0,0,1,0);
2006 * given either a sockbuf or a socket run down the
2007 * event list and queue ready events found...
2008 * the socket must be locked by the caller
2011 postevent(struct socket
*sp
, struct sockbuf
*sb
, int event
)
2014 struct eventqelt
*evq
;
2022 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_START
, (int)sp
, event
, 0, 0, 0);
2024 for (evq
= sp
->so_evlist
.tqh_first
;
2025 evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2027 if (evq
->ee_eventmask
== 0)
2031 /* ready for reading:
2032 - byte cnt >= receive low water mark
2033 - read-half of conn closed
2034 - conn pending for listening sock
2035 - socket error pending
2038 - byte cnt avail >= send low water mark
2039 - write half of conn closed
2040 - socket error pending
2041 - non-blocking conn completed successfully
2045 - sock at out of band mark
2048 switch (event
& EV_DMASK
) {
2051 if ((evq
->ee_eventmask
& EV_EX
)) {
2052 if (sp
->so_oobmark
|| ((sp
->so_state
& SS_RCVATMARK
)))
2053 mask
|= EV_EX
|EV_OOB
;
2057 case EV_RWBYTES
|EV_OOB
:
2058 if ((evq
->ee_eventmask
& EV_EX
)) {
2059 if (sp
->so_oobmark
|| ((sp
->so_state
& SS_RCVATMARK
)))
2060 mask
|= EV_EX
|EV_OOB
;
2063 * fall into the next case
2066 if ((evq
->ee_eventmask
& EV_RE
) && soreadable(sp
)) {
2068 if ((sp
->so_type
== SOCK_STREAM
) && ((sp
->so_error
== ECONNREFUSED
) || (sp
->so_error
== ECONNRESET
))) {
2069 if ((sp
->so_pcb
== 0) || (((struct inpcb
*)sp
->so_pcb
)->inp_state
== INPCB_STATE_DEAD
) || !(tp
= sototcpcb(sp
)) ||
2070 (tp
->t_state
== TCPS_CLOSED
)) {
2071 mask
|= EV_RE
|EV_RESET
;
2077 evq
->ee_req
.er_rcnt
= sp
->so_rcv
.sb_cc
;
2079 if (sp
->so_state
& SS_CANTRCVMORE
) {
2084 if ((evq
->ee_eventmask
& EV_WR
) && sowriteable(sp
)) {
2086 if ((sp
->so_type
== SOCK_STREAM
) && ((sp
->so_error
== ECONNREFUSED
) || (sp
->so_error
== ECONNRESET
))) {
2087 if ((sp
->so_pcb
== 0) || (((struct inpcb
*)sp
->so_pcb
)->inp_state
== INPCB_STATE_DEAD
) || !(tp
= sototcpcb(sp
)) ||
2088 (tp
->t_state
== TCPS_CLOSED
)) {
2089 mask
|= EV_WR
|EV_RESET
;
2095 evq
->ee_req
.er_wcnt
= sbspace(&sp
->so_snd
);
2100 if ((evq
->ee_eventmask
& EV_RE
)) {
2101 mask
|= EV_RE
|EV_RCONN
;
2102 evq
->ee_req
.er_rcnt
= sp
->so_qlen
+ 1; // incl this one
2107 if ((evq
->ee_eventmask
& EV_WR
)) {
2108 mask
|= EV_WR
|EV_WCONN
;
2113 if ((evq
->ee_eventmask
& EV_RE
)) {
2114 mask
|= EV_RE
|EV_RCLOSED
;
2119 if ((evq
->ee_eventmask
& EV_WR
)) {
2120 mask
|= EV_WR
|EV_WCLOSED
;
2125 if (evq
->ee_eventmask
& EV_RE
) {
2126 mask
|= EV_RE
|EV_FIN
;
2132 if (evq
->ee_eventmask
& EV_RE
) {
2133 mask
|= EV_RE
| event
;
2135 if (evq
->ee_eventmask
& EV_WR
) {
2136 mask
|= EV_WR
| event
;
2141 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, (int)sp
, -1, 0, 0, 0);
2145 KERNEL_DEBUG(DBG_MISC_POST
, (int)evq
, evq
->ee_eventmask
, evq
->ee_req
.er_eventbits
, mask
, 0);
2149 * disarm... postevents are nops until this event is 'read' via
2150 * waitevent and then re-armed via modwatch
2152 evq
->ee_eventmask
= 0;
2155 * since events are disarmed until after the waitevent
2156 * the ee_req.er_xxxx fields can't change once we've
2157 * inserted this event into the proc queue...
2158 * since waitevent can't see this event until we
2159 * enqueue it, waitevent will see a 'consistent'
2160 * snapshot of the event, even though it won't hold
2161 * the socket lock, and we're updating the event outside
2162 * of the proc lock, which it will hold
2164 evq
->ee_req
.er_eventbits
|= mask
;
2169 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, (int)sp
, 0, 0, 0, 0);
2171 #endif /* SOCKETS */
2175 * watchevent system call. user passes us an event to watch
2176 * for. we malloc an event object, initialize it, and queue
2177 * it to the open socket. when the event occurs, postevent()
2178 * will enque it back to our proc where we can retrieve it
2181 * should this prevent duplicate events on same socket?
2184 * ENOMEM No memory for operation
2188 watchevent(proc_t p
, struct watchevent_args
*uap
, __unused
int *retval
)
2190 struct eventqelt
*evq
= (struct eventqelt
*)0;
2191 struct eventqelt
*np
= NULL
;
2192 struct eventreq64
*erp
;
2193 struct fileproc
*fp
= NULL
;
2196 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_START
, 0,0,0,0,0);
2198 // get a qelt and fill with users req
2199 MALLOC(evq
, struct eventqelt
*, sizeof(struct eventqelt
), M_TEMP
, M_WAITOK
);
2205 // get users request pkt
2207 if (IS_64BIT_PROCESS(p
)) {
2208 error
= copyin(uap
->u_req
, (caddr_t
)erp
, sizeof(struct eventreq64
));
2210 struct eventreq32 er32
;
2212 error
= copyin(uap
->u_req
, (caddr_t
)&er32
, sizeof(struct eventreq32
));
2215 * the user only passes in the
2216 * er_type, er_handle and er_data...
2217 * the other fields are initialized
2218 * below, so don't bother to copy
2220 erp
->er_type
= er32
.er_type
;
2221 erp
->er_handle
= er32
.er_handle
;
2222 erp
->er_data
= (user_addr_t
)er32
.er_data
;
2227 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, error
,0,0,0,0);
2231 KERNEL_DEBUG(DBG_MISC_WATCH
, erp
->er_handle
,uap
->u_eventmask
,(uint32_t)evq
,0,0);
2233 // validate, freeing qelt if errors
2237 if (erp
->er_type
!= EV_FD
) {
2239 } else if ((error
= fp_lookup(p
, erp
->er_handle
, &fp
, 1)) != 0) {
2242 } else if (fp
->f_type
== DTYPE_SOCKET
) {
2243 socket_lock((struct socket
*)fp
->f_data
, 1);
2244 np
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2245 #endif /* SOCKETS */
2246 } else if (fp
->f_type
== DTYPE_PIPE
) {
2247 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2248 np
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2250 fp_drop(p
, erp
->er_handle
, fp
, 1);
2258 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, error
,0,0,0,0);
2263 * only allow one watch per file per proc
2265 for ( ; np
!= NULL
; np
= np
->ee_slist
.tqe_next
) {
2266 if (np
->ee_proc
== p
) {
2268 if (fp
->f_type
== DTYPE_SOCKET
)
2269 socket_unlock((struct socket
*)fp
->f_data
, 1);
2271 #endif /* SOCKETS */
2272 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2273 fp_drop(p
, erp
->er_handle
, fp
, 0);
2276 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2280 erp
->er_ecnt
= erp
->er_rcnt
= erp
->er_wcnt
= erp
->er_eventbits
= 0;
2282 evq
->ee_eventmask
= uap
->u_eventmask
& EV_MASK
;
2286 if (fp
->f_type
== DTYPE_SOCKET
) {
2287 TAILQ_INSERT_TAIL(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2288 postevent((struct socket
*)fp
->f_data
, 0, EV_RWBYTES
); // catch existing events
2290 socket_unlock((struct socket
*)fp
->f_data
, 1);
2292 #endif /* SOCKETS */
2294 TAILQ_INSERT_TAIL(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2295 postpipeevent((struct pipe
*)fp
->f_data
, EV_RWBYTES
);
2297 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2299 fp_drop_event(p
, erp
->er_handle
, fp
);
2301 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, 0,0,0,0,0);
2308 * waitevent system call.
2309 * grabs the next waiting event for this proc and returns
2310 * it. if no events, user can request to sleep with timeout
2311 * or without or poll mode
2312 * ((tv != NULL && interval == 0) || tv == -1)
2315 waitevent(proc_t p
, struct waitevent_args
*uap
, int *retval
)
2318 struct eventqelt
*evq
;
2319 struct eventreq64
*erp
;
2320 uint64_t abstime
, interval
;
2321 boolean_t fast_poll
= FALSE
;
2323 struct eventreq64 er64
;
2324 struct eventreq32 er32
;
2332 * check for fast poll method
2334 if (IS_64BIT_PROCESS(p
)) {
2335 if (uap
->tv
== (user_addr_t
)-1)
2337 } else if (uap
->tv
== (user_addr_t
)((uint32_t)-1))
2340 if (fast_poll
== TRUE
) {
2341 if (p
->p_evlist
.tqh_first
== NULL
) {
2342 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_NONE
, -1,0,0,0,0);
2352 error
= copyin(uap
->tv
, (caddr_t
)&atv
, sizeof (atv
));
2356 if (itimerfix(&atv
)) {
2360 interval
= tvtoabstime(&atv
);
2362 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_START
, 0,0,0,0,0);
2366 if ((evq
= p
->p_evlist
.tqh_first
) != NULL
) {
2368 * found one... make a local copy while it's still on the queue
2369 * to prevent it from changing while in the midst of copying
2370 * don't want to hold the proc lock across a copyout because
2371 * it might block on a page fault at the target in user space
2375 if (IS_64BIT_PROCESS(p
))
2376 bcopy((caddr_t
)erp
, (caddr_t
)&uer
.er64
, sizeof (struct eventreq64
));
2378 uer
.er32
.er_type
= erp
->er_type
;
2379 uer
.er32
.er_handle
= erp
->er_handle
;
2380 uer
.er32
.er_data
= (uint32_t)erp
->er_data
;
2381 uer
.er32
.er_ecnt
= erp
->er_ecnt
;
2382 uer
.er32
.er_rcnt
= erp
->er_rcnt
;
2383 uer
.er32
.er_wcnt
= erp
->er_wcnt
;
2384 uer
.er32
.er_eventbits
= erp
->er_eventbits
;
2386 TAILQ_REMOVE(&p
->p_evlist
, evq
, ee_plist
);
2388 evq
->ee_flags
&= ~EV_QUEUED
;
2392 if (IS_64BIT_PROCESS(p
))
2393 error
= copyout((caddr_t
)&uer
.er64
, uap
->u_req
, sizeof(struct eventreq64
));
2395 error
= copyout((caddr_t
)&uer
.er32
, uap
->u_req
, sizeof(struct eventreq32
));
2397 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, error
,
2398 evq
->ee_req
.er_handle
,evq
->ee_req
.er_eventbits
,(uint32_t)evq
,0);
2402 if (uap
->tv
&& interval
== 0) {
2404 *retval
= 1; // poll failed
2406 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, error
,0,0,0,0);
2410 clock_absolutetime_interval_to_deadline(interval
, &abstime
);
2414 KERNEL_DEBUG(DBG_MISC_WAIT
, 1,(uint32_t)&p
->p_evlist
,0,0,0);
2416 error
= msleep1(&p
->p_evlist
, &p
->p_mlock
, (PSOCK
| PCATCH
), "waitevent", abstime
);
2418 KERNEL_DEBUG(DBG_MISC_WAIT
, 2,(uint32_t)&p
->p_evlist
,0,0,0);
2422 if (error
== ERESTART
)
2424 if (error
== EWOULDBLOCK
) {
2431 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, 0,0,0,0,0);
2437 * modwatch system call. user passes in event to modify.
2438 * if we find it we reset the event bits and que/deque event
2442 modwatch(proc_t p
, struct modwatch_args
*uap
, __unused
int *retval
)
2444 struct eventreq64 er
;
2445 struct eventreq64
*erp
= &er
;
2446 struct eventqelt
*evq
= NULL
; /* protected by error return */
2448 struct fileproc
*fp
;
2451 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_START
, 0,0,0,0,0);
2454 * get user's request pkt
2455 * just need the er_type and er_handle which sit above the
2456 * problematic er_data (32/64 issue)... so only copy in
2459 if ((error
= copyin(uap
->u_req
, (caddr_t
)erp
, sizeof(er
.er_type
) + sizeof(er
.er_handle
)))) {
2460 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, error
,0,0,0,0);
2465 if (erp
->er_type
!= EV_FD
) {
2467 } else if ((error
= fp_lookup(p
, erp
->er_handle
, &fp
, 1)) != 0) {
2470 } else if (fp
->f_type
== DTYPE_SOCKET
) {
2471 socket_lock((struct socket
*)fp
->f_data
, 1);
2472 evq
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2473 #endif /* SOCKETS */
2474 } else if (fp
->f_type
== DTYPE_PIPE
) {
2475 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2476 evq
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2478 fp_drop(p
, erp
->er_handle
, fp
, 1);
2484 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, error
,0,0,0,0);
2488 if ((uap
->u_eventmask
== EV_RM
) && (fp
->f_flags
& FP_WAITEVENT
)) {
2489 fp
->f_flags
&= ~FP_WAITEVENT
;
2493 // locate event if possible
2494 for ( ; evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2495 if (evq
->ee_proc
== p
)
2500 if (fp
->f_type
== DTYPE_SOCKET
)
2501 socket_unlock((struct socket
*)fp
->f_data
, 1);
2503 #endif /* SOCKETS */
2504 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2505 fp_drop(p
, erp
->er_handle
, fp
, 0);
2506 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2509 KERNEL_DEBUG(DBG_MISC_MOD
, erp
->er_handle
,uap
->u_eventmask
,(uint32_t)evq
,0,0);
2511 if (uap
->u_eventmask
== EV_RM
) {
2512 EVPROCDEQUE(p
, evq
);
2515 if (fp
->f_type
== DTYPE_SOCKET
) {
2516 TAILQ_REMOVE(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2517 socket_unlock((struct socket
*)fp
->f_data
, 1);
2519 #endif /* SOCKETS */
2521 TAILQ_REMOVE(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2522 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2524 fp_drop(p
, erp
->er_handle
, fp
, 0);
2526 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, 0,0,0,0,0);
2529 switch (uap
->u_eventmask
& EV_MASK
) {
2547 case EV_EX
|EV_RE
|EV_WR
:
2548 flag
= EV_OOB
|EV_RWBYTES
;
2553 if (fp
->f_type
== DTYPE_SOCKET
)
2554 socket_unlock((struct socket
*)fp
->f_data
, 1);
2556 #endif /* SOCKETS */
2557 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2558 fp_drop(p
, erp
->er_handle
, fp
, 0);
2559 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2563 * since we're holding the socket/pipe lock, the event
2564 * cannot go from the unqueued state to the queued state
2565 * however, it can go from the queued state to the unqueued state
2566 * since that direction is protected by the proc_lock...
2567 * so do a quick check for EV_QUEUED w/o holding the proc lock
2568 * since by far the common case will be NOT EV_QUEUED, this saves
2569 * us taking the proc_lock the majority of the time
2571 if (evq
->ee_flags
& EV_QUEUED
) {
2573 * EVPROCDEQUE will recheck the state after it grabs the proc_lock
2575 EVPROCDEQUE(p
, evq
);
2578 * while the event is off the proc queue and
2579 * we're holding the socket/pipe lock
2580 * it's safe to update these fields...
2582 evq
->ee_req
.er_eventbits
= 0;
2583 evq
->ee_eventmask
= uap
->u_eventmask
& EV_MASK
;
2586 if (fp
->f_type
== DTYPE_SOCKET
) {
2587 postevent((struct socket
*)fp
->f_data
, 0, flag
);
2588 socket_unlock((struct socket
*)fp
->f_data
, 1);
2590 #endif /* SOCKETS */
2592 postpipeevent((struct pipe
*)fp
->f_data
, flag
);
2593 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2595 fp_drop(p
, erp
->er_handle
, fp
, 0);
2596 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, evq
->ee_req
.er_handle
,evq
->ee_eventmask
,(uint32_t)fp
->f_data
,flag
,0);
2600 /* this routine is called from the close of fd with proc_fdlock held */
2602 waitevent_close(struct proc
*p
, struct fileproc
*fp
)
2604 struct eventqelt
*evq
;
2607 fp
->f_flags
&= ~FP_WAITEVENT
;
2610 if (fp
->f_type
== DTYPE_SOCKET
) {
2611 socket_lock((struct socket
*)fp
->f_data
, 1);
2612 evq
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2614 #endif /* SOCKETS */
2615 if (fp
->f_type
== DTYPE_PIPE
) {
2616 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2617 evq
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2625 // locate event if possible
2626 for ( ; evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2627 if (evq
->ee_proc
== p
)
2632 if (fp
->f_type
== DTYPE_SOCKET
)
2633 socket_unlock((struct socket
*)fp
->f_data
, 1);
2635 #endif /* SOCKETS */
2636 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2642 EVPROCDEQUE(p
, evq
);
2645 if (fp
->f_type
== DTYPE_SOCKET
) {
2646 TAILQ_REMOVE(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2647 socket_unlock((struct socket
*)fp
->f_data
, 1);
2649 #endif /* SOCKETS */
2651 TAILQ_REMOVE(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2652 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2665 * Description: Get the host UUID from IOKit and return it to user space.
2667 * Parameters: uuid_buf Pointer to buffer to receive UUID
2668 * timeout Timespec for timout
2670 * Returns: 0 Success
2671 * EWOULDBLOCK Timeout is too short
2672 * copyout:EFAULT Bad user buffer
2674 * Notes: A timeout seems redundant, since if it's tolerable to not
2675 * have a system UUID in hand, then why ask for one?
2678 gethostuuid(struct proc
*p
, struct gethostuuid_args
*uap
, __unused register_t
*retval
)
2682 mach_timespec_t mach_ts
; /* for IOKit call */
2683 __darwin_uuid_t uuid_kern
; /* for IOKit call */
2685 /* Convert the 32/64 bit timespec into a mach_timespec_t */
2686 if ( proc_is64bit(p
) ) {
2687 struct user_timespec ts
;
2688 error
= copyin(uap
->timeoutp
, &ts
, sizeof(ts
));
2691 mach_ts
.tv_sec
= ts
.tv_sec
;
2692 mach_ts
.tv_nsec
= ts
.tv_nsec
;
2695 error
= copyin(uap
->timeoutp
, &ts
, sizeof(ts
) );
2698 mach_ts
.tv_sec
= ts
.tv_sec
;
2699 mach_ts
.tv_nsec
= ts
.tv_nsec
;
2702 /* Call IOKit with the stack buffer to get the UUID */
2703 kret
= IOBSDGetPlatformUUID(uuid_kern
, mach_ts
);
2706 * If we get it, copy out the data to the user buffer; note that a
2707 * uuid_t is an array of characters, so this is size invariant for
2710 if (kret
== KERN_SUCCESS
) {
2711 error
= copyout(uuid_kern
, uap
->uuid_buf
, sizeof(uuid_kern
));
2713 error
= EWOULDBLOCK
;