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);
243 * Code common for read and pread
247 donefileread(struct proc
*p
, struct fileproc
*fp
, int fd
)
251 fp
->f_flags
&= ~FP_INCHRREAD
;
253 fp_drop(p
, fd
, fp
, 1);
266 preparefileread(struct proc
*p
, struct fileproc
**fp_ret
, int fd
, int check_for_pread
)
274 error
= fp_lookup(p
, fd
, &fp
, 1);
280 if ((fp
->f_flag
& FREAD
) == 0) {
284 if (check_for_pread
&& (fp
->f_type
!= DTYPE_VNODE
)) {
288 if (fp
->f_type
== DTYPE_VNODE
) {
289 vp
= (struct vnode
*)fp
->f_fglob
->fg_data
;
291 if (check_for_pread
&& (vnode_isfifo(vp
))) {
295 if (check_for_pread
&& (vp
->v_flag
& VISTTY
)) {
299 if (vp
->v_type
== VCHR
)
300 fp
->f_flags
|= FP_INCHRREAD
;
309 fp_drop(p
, fd
, fp
, 1);
320 __private_extern__
int
321 dofileread(vfs_context_t ctx
, struct fileproc
*fp
,
322 user_addr_t bufp
, user_size_t nbyte
, off_t offset
, int flags
,
323 user_ssize_t
*retval
)
326 user_ssize_t bytecnt
;
328 char uio_buf
[ UIO_SIZEOF(1) ];
330 // LP64todo - do we want to raise this?
334 if (IS_64BIT_PROCESS(vfs_context_proc(ctx
))) {
335 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE64
, UIO_READ
,
336 &uio_buf
[0], sizeof(uio_buf
));
338 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE32
, UIO_READ
,
339 &uio_buf
[0], sizeof(uio_buf
));
341 uio_addiov(auio
, bufp
, nbyte
);
345 if ((error
= fo_read(fp
, auio
, flags
, ctx
))) {
346 if (uio_resid(auio
) != bytecnt
&& (error
== ERESTART
||
347 error
== EINTR
|| error
== EWOULDBLOCK
))
350 bytecnt
-= uio_resid(auio
);
358 * Scatter read system call.
367 readv(struct proc
*p
, struct readv_args
*uap
, user_ssize_t
*retval
)
369 __pthread_testcancel(1);
370 return(readv_nocancel(p
, (struct readv_nocancel_args
*)uap
, retval
));
374 readv_nocancel(struct proc
*p
, struct readv_nocancel_args
*uap
, user_ssize_t
*retval
)
379 struct user_iovec
*iovp
;
381 /* Verify range bedfore calling uio_create() */
382 if (uap
->iovcnt
<= 0 || uap
->iovcnt
> UIO_MAXIOV
)
385 /* allocate a uio large enough to hold the number of iovecs passed */
386 auio
= uio_create(uap
->iovcnt
, 0,
387 (IS_64BIT_PROCESS(p
) ? UIO_USERSPACE64
: UIO_USERSPACE32
),
390 /* get location of iovecs within the uio. then copyin the iovecs from
393 iovp
= uio_iovsaddr(auio
);
396 goto ExitThisRoutine
;
398 size_of_iovec
= (IS_64BIT_PROCESS(p
) ? sizeof(struct user_iovec
) : sizeof(struct iovec
));
399 error
= copyin(uap
->iovp
, (caddr_t
)iovp
, (uap
->iovcnt
* size_of_iovec
));
401 goto ExitThisRoutine
;
404 /* finalize uio_t for use and do the IO
406 uio_calculateresid(auio
);
407 error
= rd_uio(p
, uap
->fd
, auio
, retval
);
425 write(struct proc
*p
, struct write_args
*uap
, user_ssize_t
*retval
)
427 __pthread_testcancel(1);
428 return(write_nocancel(p
, (struct write_nocancel_args
*)uap
, retval
));
433 write_nocancel(struct proc
*p
, struct write_nocancel_args
*uap
, user_ssize_t
*retval
)
439 error
= fp_lookup(p
,fd
,&fp
,0);
442 if ((fp
->f_flag
& FWRITE
) == 0) {
445 struct vfs_context context
= *(vfs_context_current());
446 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
448 error
= dofilewrite(&context
, fp
, uap
->cbuf
, uap
->nbyte
,
449 (off_t
)-1, 0, retval
);
452 fp_drop_written(p
, fd
, fp
);
454 fp_drop(p
, fd
, fp
, 0);
470 pwrite(struct proc
*p
, struct pwrite_args
*uap
, user_ssize_t
*retval
)
472 __pthread_testcancel(1);
473 return(pwrite_nocancel(p
, (struct pwrite_nocancel_args
*)uap
, retval
));
477 pwrite_nocancel(struct proc
*p
, struct pwrite_nocancel_args
*uap
, user_ssize_t
*retval
)
482 vnode_t vp
= (vnode_t
)0;
484 error
= fp_lookup(p
,fd
,&fp
,0);
488 if ((fp
->f_flag
& FWRITE
) == 0) {
491 struct vfs_context context
= *vfs_context_current();
492 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
494 if (fp
->f_type
!= DTYPE_VNODE
) {
498 vp
= (vnode_t
)fp
->f_fglob
->fg_data
;
499 if (vnode_isfifo(vp
)) {
503 if ((vp
->v_flag
& VISTTY
)) {
507 if (uap
->offset
== (off_t
)-1) {
512 error
= dofilewrite(&context
, fp
, uap
->buf
, uap
->nbyte
,
513 uap
->offset
, FOF_OFFSET
, retval
);
517 fp_drop_written(p
, fd
, fp
);
519 fp_drop(p
, fd
, fp
, 0);
522 KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO
, SYS_pwrite
) | DBG_FUNC_NONE
),
523 uap
->fd
, uap
->nbyte
, (unsigned int)((uap
->offset
>> 32)), (unsigned int)(uap
->offset
), 0);
532 * <fo_write>:??? [indirect through struct fileops]
534 __private_extern__
int
535 dofilewrite(vfs_context_t ctx
, struct fileproc
*fp
,
536 user_addr_t bufp
, user_size_t nbyte
, off_t offset
, int flags
,
537 user_ssize_t
*retval
)
541 user_ssize_t bytecnt
;
542 char uio_buf
[ UIO_SIZEOF(1) ];
544 // LP64todo - do we want to raise this?
548 if (IS_64BIT_PROCESS(vfs_context_proc(ctx
))) {
549 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE64
, UIO_WRITE
,
550 &uio_buf
[0], sizeof(uio_buf
));
552 auio
= uio_createwithbuffer(1, offset
, UIO_USERSPACE32
, UIO_WRITE
,
553 &uio_buf
[0], sizeof(uio_buf
));
555 uio_addiov(auio
, bufp
, nbyte
);
558 if ((error
= fo_write(fp
, auio
, flags
, ctx
))) {
559 if (uio_resid(auio
) != bytecnt
&& (error
== ERESTART
||
560 error
== EINTR
|| error
== EWOULDBLOCK
))
562 /* The socket layer handles SIGPIPE */
563 if (error
== EPIPE
&& fp
->f_type
!= DTYPE_SOCKET
) {
564 /* XXX Raise the signal on the thread? */
565 psignal(vfs_context_proc(ctx
), SIGPIPE
);
568 bytecnt
-= uio_resid(auio
);
575 * Gather write system call
578 writev(struct proc
*p
, struct writev_args
*uap
, user_ssize_t
*retval
)
580 __pthread_testcancel(1);
581 return(writev_nocancel(p
, (struct writev_nocancel_args
*)uap
, retval
));
585 writev_nocancel(struct proc
*p
, struct writev_nocancel_args
*uap
, user_ssize_t
*retval
)
590 struct user_iovec
*iovp
;
592 /* Verify range bedfore calling uio_create() */
593 if (uap
->iovcnt
<= 0 || uap
->iovcnt
> UIO_MAXIOV
)
596 /* allocate a uio large enough to hold the number of iovecs passed */
597 auio
= uio_create(uap
->iovcnt
, 0,
598 (IS_64BIT_PROCESS(p
) ? UIO_USERSPACE64
: UIO_USERSPACE32
),
601 /* get location of iovecs within the uio. then copyin the iovecs from
604 iovp
= uio_iovsaddr(auio
);
607 goto ExitThisRoutine
;
609 size_of_iovec
= (IS_64BIT_PROCESS(p
) ? sizeof(struct user_iovec
) : sizeof(struct iovec
));
610 error
= copyin(uap
->iovp
, (caddr_t
)iovp
, (uap
->iovcnt
* size_of_iovec
));
612 goto ExitThisRoutine
;
615 /* finalize uio_t for use and do the IO
617 uio_calculateresid(auio
);
618 error
= wr_uio(p
, uap
->fd
, auio
, retval
);
629 wr_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
)
634 struct vfs_context context
= *vfs_context_current();
636 error
= fp_lookup(p
,fdes
,&fp
,0);
640 if ((fp
->f_flag
& FWRITE
) == 0) {
644 count
= uio_resid(uio
);
646 context
.vc_ucred
= fp
->f_cred
;
647 error
= fo_write(fp
, uio
, 0, &context
);
649 if (uio_resid(uio
) != count
&& (error
== ERESTART
||
650 error
== EINTR
|| error
== EWOULDBLOCK
))
652 /* The socket layer handles SIGPIPE */
653 if (error
== EPIPE
&& fp
->f_type
!= DTYPE_SOCKET
)
656 *retval
= count
- uio_resid(uio
);
660 fp_drop_written(p
, fdes
, fp
);
662 fp_drop(p
, fdes
, fp
, 0);
668 rd_uio(struct proc
*p
, int fdes
, uio_t uio
, user_ssize_t
*retval
)
673 struct vfs_context context
= *vfs_context_current();
675 if ( (error
= preparefileread(p
, &fp
, fdes
, 0)) )
678 count
= uio_resid(uio
);
680 context
.vc_ucred
= fp
->f_cred
;
682 error
= fo_read(fp
, uio
, 0, &context
);
685 if (uio_resid(uio
) != count
&& (error
== ERESTART
||
686 error
== EINTR
|| error
== EWOULDBLOCK
))
689 *retval
= count
- uio_resid(uio
);
691 donefileread(p
, fp
, fdes
);
706 * fp_lookup:EBADF Bad file descriptor
710 ioctl(struct proc
*p
, struct ioctl_args
*uap
, __unused register_t
*retval
)
719 #define STK_PARAMS 128
720 char stkbuf
[STK_PARAMS
];
722 struct vfs_context context
= *vfs_context_current();
724 AUDIT_ARG(fd
, uap
->fd
);
725 AUDIT_ARG(cmd
, CAST_DOWN(int, uap
->com
)); /* LP64todo: uap->com is a user-land long */
726 AUDIT_ARG(addr
, uap
->data
);
728 is64bit
= proc_is64bit(p
);
731 error
= fp_lookup(p
,fd
,&fp
,1);
737 AUDIT_ARG(file
, p
, fp
);
739 if ((fp
->f_flag
& (FREAD
| FWRITE
)) == 0) {
744 context
.vc_ucred
= fp
->f_fglob
->fg_cred
;
747 error
= mac_file_check_ioctl(context
.vc_ucred
, fp
->f_fglob
, uap
->com
);
754 * ### LD 6/11/97 Hack Alert: this is to get AppleTalk to work
755 * while implementing an ATioctl system call
758 if (appletalk_inited
&& ((uap
->com
& 0x0000FFFF) == 0xff99)) {
759 u_long fixed_command
;
761 #ifdef APPLETALK_DEBUG
762 kprintf("ioctl: special AppleTalk \n");
765 *(user_addr_t
*)datap
= uap
->data
;
766 fixed_command
= _IOW(0, 0xff99, uap
->data
);
767 error
= fo_ioctl(fp
, fixed_command
, datap
, &context
);
775 switch (com
= uap
->com
) {
777 *fdflags(p
, uap
->fd
) &= ~UF_EXCLOSE
;
781 *fdflags(p
, uap
->fd
) |= UF_EXCLOSE
;
787 * Interpret high order word to find amount of data to be
788 * copied to/from the user's address space.
790 size
= IOCPARM_LEN(com
);
791 if (size
> IOCPARM_MAX
) {
796 if (size
> sizeof (stkbuf
)) {
798 if ((memp
= (caddr_t
)kalloc(size
)) == 0) {
810 error
= copyin(uap
->data
, datap
, size
);
819 /* XXX - IOC_IN and no size? we should proably return an error here!! */
821 *(user_addr_t
*)datap
= uap
->data
;
824 *(uint32_t *)datap
= (uint32_t)uap
->data
;
827 } else if ((com
&IOC_OUT
) && size
)
829 * Zero the buffer so the user always
830 * gets back something deterministic.
833 else if (com
&IOC_VOID
) {
834 /* XXX - this is odd since IOC_VOID means no parameters */
836 *(user_addr_t
*)datap
= uap
->data
;
839 *(uint32_t *)datap
= (uint32_t)uap
->data
;
846 if ( (tmp
= *(int *)datap
) )
847 fp
->f_flag
|= FNONBLOCK
;
849 fp
->f_flag
&= ~FNONBLOCK
;
850 error
= fo_ioctl(fp
, FIONBIO
, (caddr_t
)&tmp
, &context
);
854 if ( (tmp
= *(int *)datap
) )
855 fp
->f_flag
|= FASYNC
;
857 fp
->f_flag
&= ~FASYNC
;
858 error
= fo_ioctl(fp
, FIOASYNC
, (caddr_t
)&tmp
, &context
);
863 if (fp
->f_type
== DTYPE_SOCKET
) {
864 ((struct socket
*)fp
->f_data
)->so_pgid
= tmp
;
868 if (fp
->f_type
== DTYPE_PIPE
) {
869 error
= fo_ioctl(fp
, (int)TIOCSPGRP
, (caddr_t
)&tmp
, &context
);
875 struct proc
*p1
= proc_find(tmp
);
883 error
= fo_ioctl(fp
, (int)TIOCSPGRP
, (caddr_t
)&tmp
, &context
);
887 if (fp
->f_type
== DTYPE_SOCKET
) {
889 *(int *)datap
= ((struct socket
*)fp
->f_data
)->so_pgid
;
892 error
= fo_ioctl(fp
, TIOCGPGRP
, datap
, &context
);
893 *(int *)datap
= -*(int *)datap
;
897 error
= fo_ioctl(fp
, com
, datap
, &context
);
899 * Copy any data to user, size was
900 * already set and checked above.
902 if (error
== 0 && (com
&IOC_OUT
) && size
)
903 error
= copyout(datap
, uap
->data
, (u_int
)size
);
911 fp_drop(p
, fd
, fp
, 1);
916 int selwait
, nselcoll
;
917 #define SEL_FIRSTPASS 1
918 #define SEL_SECONDPASS 2
919 extern int selcontinue(int error
);
920 extern int selprocess(int error
, int sel_pass
);
921 static int selscan(struct proc
*p
, struct _select
* sel
,
922 int nfd
, register_t
*retval
, int sel_pass
, wait_queue_sub_t wqsub
);
923 static int selcount(struct proc
*p
, u_int32_t
*ibits
, u_int32_t
*obits
,
924 int nfd
, int * count
, int *kfcount
);
925 static int seldrop(struct proc
*p
, u_int32_t
*ibits
, int nfd
);
926 extern uint64_t tvtoabstime(struct timeval
*tvp
);
929 * Select system call.
932 * EINVAL Invalid argument
933 * EAGAIN Nonconformant error if allocation fails
937 select(struct proc
*p
, struct select_args
*uap
, register_t
*retval
)
939 __pthread_testcancel(1);
940 return(select_nocancel(p
, (struct select_nocancel_args
*)uap
, retval
));
944 select_nocancel(struct proc
*p
, struct select_nocancel_args
*uap
, register_t
*retval
)
951 int needzerofill
= 1;
955 th_act
= current_thread();
956 uth
= get_bsdthread_info(th_act
);
957 sel
= &uth
->uu_select
;
958 retval
= (int *)get_bsduthreadrval(th_act
);
965 /* select on thread of process that already called proc_exit() */
966 if (p
->p_fd
== NULL
) {
970 if (uap
->nd
> p
->p_fd
->fd_nfiles
)
971 uap
->nd
= p
->p_fd
->fd_nfiles
; /* forgiving; slightly wrong */
973 nw
= howmany(uap
->nd
, NFDBITS
);
974 ni
= nw
* sizeof(fd_mask
);
977 * if the previously allocated space for the bits is smaller than
978 * what is requested or no space has yet been allocated for this
979 * thread, allocate enough space now.
981 * Note: If this process fails, select() will return EAGAIN; this
982 * is the same thing pool() returns in a no-memory situation, but
983 * it is not a POSIX compliant error code for select().
985 if (sel
->nbytes
< (3 * ni
)) {
988 /* Free previous allocation, if any */
989 if (sel
->ibits
!= NULL
)
990 FREE(sel
->ibits
, M_TEMP
);
991 if (sel
->obits
!= NULL
) {
992 FREE(sel
->obits
, M_TEMP
);
993 /* NULL out; subsequent ibits allocation may fail */
997 MALLOC(sel
->ibits
, u_int32_t
*, nbytes
, M_TEMP
, M_WAITOK
| M_ZERO
);
998 if (sel
->ibits
== NULL
)
1000 MALLOC(sel
->obits
, u_int32_t
*, nbytes
, M_TEMP
, M_WAITOK
| M_ZERO
);
1001 if (sel
->obits
== NULL
) {
1002 FREE(sel
->ibits
, M_TEMP
);
1006 sel
->nbytes
= nbytes
;
1011 bzero((caddr_t
)sel
->ibits
, sel
->nbytes
);
1012 bzero((caddr_t
)sel
->obits
, sel
->nbytes
);
1016 * get the bits from the user address space
1018 #define getbits(name, x) \
1020 if (uap->name && (error = copyin(uap->name, \
1021 (caddr_t)&sel->ibits[(x) * nw], ni))) \
1022 goto continuation; \
1032 if (IS_64BIT_PROCESS(p
)) {
1033 struct user_timeval atv64
;
1034 error
= copyin(uap
->tv
, (caddr_t
)&atv64
, sizeof(atv64
));
1035 /* Loses resolution - assume timeout < 68 years */
1036 atv
.tv_sec
= atv64
.tv_sec
;
1037 atv
.tv_usec
= atv64
.tv_usec
;
1039 error
= copyin(uap
->tv
, (caddr_t
)&atv
, sizeof(atv
));
1043 if (itimerfix(&atv
)) {
1048 clock_absolutetime_interval_to_deadline(
1049 tvtoabstime(&atv
), &sel
->abstime
);
1055 if ( (error
= selcount(p
, sel
->ibits
, sel
->obits
, uap
->nd
, &count
, &kfcount
)) ) {
1059 sel
->kfcount
= kfcount
;
1060 size
= SIZEOF_WAITQUEUE_SET
+ (count
* SIZEOF_WAITQUEUE_LINK
);
1061 if (uth
->uu_allocsize
) {
1062 if (uth
->uu_wqset
== 0)
1063 panic("select: wql memory smashed");
1064 /* needed for the select now */
1065 if (size
> uth
->uu_allocsize
) {
1066 kfree(uth
->uu_wqset
, uth
->uu_allocsize
);
1067 uth
->uu_allocsize
= size
;
1068 uth
->uu_wqset
= (wait_queue_set_t
)kalloc(size
);
1069 if (uth
->uu_wqset
== (wait_queue_set_t
)NULL
)
1070 panic("failed to allocate memory for waitqueue\n");
1074 uth
->uu_allocsize
= size
;
1075 uth
->uu_wqset
= (wait_queue_set_t
)kalloc(uth
->uu_allocsize
);
1076 if (uth
->uu_wqset
== (wait_queue_set_t
)NULL
)
1077 panic("failed to allocate memory for waitqueue\n");
1079 bzero(uth
->uu_wqset
, size
);
1080 sel
->wql
= (char *)uth
->uu_wqset
+ SIZEOF_WAITQUEUE_SET
;
1081 wait_queue_set_init(uth
->uu_wqset
, (SYNC_POLICY_FIFO
| SYNC_POLICY_PREPOST
));
1084 return selprocess(error
, SEL_FIRSTPASS
);
1088 selcontinue(int error
)
1090 return selprocess(error
, SEL_SECONDPASS
);
1094 selprocess(int error
, int sel_pass
)
1099 struct uthread
*uth
;
1101 struct select_args
*uap
;
1103 struct _select
*sel
;
1108 wait_result_t wait_result
;
1111 th_act
= current_thread();
1112 uap
= (struct select_args
*)get_bsduthreadarg(th_act
);
1113 retval
= (int *)get_bsduthreadrval(th_act
);
1114 uth
= get_bsdthread_info(th_act
);
1115 sel
= &uth
->uu_select
;
1117 /* if it is first pass wait queue is not setup yet */
1118 if ((error
!= 0) && (sel_pass
== SEL_FIRSTPASS
))
1120 if (sel
->count
== 0)
1128 OSBitOrAtomic(P_SELECT
, (UInt32
*)&p
->p_flag
);
1129 /* skip scans if the select is just for timeouts */
1131 if (sel_pass
== SEL_FIRSTPASS
)
1132 wait_queue_sub_clearrefs(uth
->uu_wqset
);
1134 error
= selscan(p
, sel
, uap
->nd
, retval
, sel_pass
, (wait_queue_sub_t
)uth
->uu_wqset
);
1135 if (error
|| *retval
) {
1139 /* if the select of log, then we canwakeup and discover some one
1140 * else already read the data; go toselct again if time permits
1154 clock_get_uptime(&now
);
1155 if (now
>= sel
->abstime
)
1160 /* cleanup obits and try again */
1162 sel_pass
= SEL_FIRSTPASS
;
1167 * To effect a poll, the timeout argument should be
1168 * non-nil, pointing to a zero-valued timeval structure.
1170 if (uap
->tv
&& sel
->abstime
== 0) {
1174 /* No spurious wakeups due to colls,no need to check for them */
1175 if ((sel_pass
== SEL_SECONDPASS
) || ((p
->p_flag
& P_SELECT
) == 0)) {
1176 sel_pass
= SEL_FIRSTPASS
;
1180 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1182 /* if the select is just for timeout skip check */
1183 if (sel
->count
&&(sel_pass
== SEL_SECONDPASS
))
1184 panic("selprocess: 2nd pass assertwaiting");
1186 /* Wait Queue Subordinate has waitqueue as first element */
1187 wait_result
= wait_queue_assert_wait((wait_queue_t
)uth
->uu_wqset
,
1188 &selwait
, THREAD_ABORTSAFE
, sel
->abstime
);
1189 if (wait_result
!= THREAD_AWAKENED
) {
1190 /* there are no preposted events */
1191 error
= tsleep1(NULL
, PSOCK
| PCATCH
,
1192 "select", 0, selcontinue
);
1198 sel_pass
= SEL_SECONDPASS
;
1206 wait_subqueue_unlink_all(uth
->uu_wqset
);
1207 seldrop(p
, sel
->ibits
, uap
->nd
);
1209 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1210 /* select is not restarted after signals... */
1211 if (error
== ERESTART
)
1213 if (error
== EWOULDBLOCK
)
1215 nw
= howmany(uap
->nd
, NFDBITS
);
1216 ni
= nw
* sizeof(fd_mask
);
1218 #define putbits(name, x) \
1220 if (uap->name && (error2 = \
1221 copyout((caddr_t)&sel->obits[(x) * nw], uap->name, ni))) \
1237 selscan(struct proc
*p
, struct _select
*sel
, int nfd
, register_t
*retval
,
1238 int sel_pass
, wait_queue_sub_t wqsub
)
1240 struct filedesc
*fdp
= p
->p_fd
;
1243 struct fileproc
*fp
;
1246 static int flag
[3] = { FREAD
, FWRITE
, 0 };
1247 u_int32_t
*iptr
, *optr
;
1249 u_int32_t
*ibits
, *obits
;
1253 boolean_t funnel_state
;
1255 struct vfs_context context
= *vfs_context_current();
1258 * Problems when reboot; due to MacOSX signal probs
1259 * in Beaker1C ; verify that the p->p_fd is valid
1269 nw
= howmany(nfd
, NFDBITS
);
1272 kfcount
= sel
->kfcount
;
1274 if (kfcount
> count
)
1275 panic("selscan: count < kfcount");
1278 funnel_state
= thread_funnel_set(kernel_flock
, TRUE
);
1281 for (msk
= 0; msk
< 3; msk
++) {
1282 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1283 optr
= (u_int32_t
*)&obits
[msk
* nw
];
1285 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1286 bits
= iptr
[i
/NFDBITS
];
1288 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1290 fp
= fdp
->fd_ofiles
[fd
];
1293 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1295 thread_funnel_set(kernel_flock
, funnel_state
);
1298 if (sel_pass
== SEL_SECONDPASS
) {
1299 wql_ptr
= (char *)0;
1300 fp
->f_flags
&= ~FP_INSELECT
;
1301 fp
->f_waddr
= (void *)0;
1303 wql_ptr
= (wql
+ nc
* SIZEOF_WAITQUEUE_LINK
);
1304 fp
->f_flags
|= FP_INSELECT
;
1305 fp
->f_waddr
= (void *)wqsub
;
1308 context
.vc_ucred
= fp
->f_cred
;
1310 if (fp
->f_ops
&& (fp
->f_type
== DTYPE_VNODE
)
1311 && ((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1312 && (vp
->v_type
== VCHR
)
1313 && fo_select(fp
, flag
[msk
], wql_ptr
, &context
)) {
1314 optr
[fd
/NFDBITS
] |= (1 << (fd
% NFDBITS
));
1322 thread_funnel_set(kernel_flock
, funnel_state
);
1326 if (kfcount
!= count
) {
1328 for (msk
= 0; msk
< 3; msk
++) {
1329 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1330 optr
= (u_int32_t
*)&obits
[msk
* nw
];
1332 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1333 bits
= iptr
[i
/NFDBITS
];
1335 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1337 fp
= fdp
->fd_ofiles
[fd
];
1340 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1344 if (sel_pass
== SEL_SECONDPASS
) {
1345 wql_ptr
= (char *)0;
1346 fp
->f_flags
&= ~FP_INSELECT
;
1347 fp
->f_waddr
= (void *)0;
1349 wql_ptr
= (wql
+ nc
* SIZEOF_WAITQUEUE_LINK
);
1350 fp
->f_flags
|= FP_INSELECT
;
1351 fp
->f_waddr
= (void *)wqsub
;
1354 context
.vc_ucred
= fp
->f_cred
;
1357 ((fp
->f_type
!= DTYPE_VNODE
)
1358 || (((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1359 && (vp
->v_type
!= VCHR
))
1361 && fo_select(fp
, flag
[msk
], wql_ptr
, &context
))) {
1362 optr
[fd
/NFDBITS
] |= (1 << (fd
% NFDBITS
));
1375 int poll_callback(struct kqueue
*, struct kevent
*, void *);
1377 struct poll_continue_args
{
1378 user_addr_t pca_fds
;
1384 poll(struct proc
*p
, struct poll_args
*uap
, register_t
*retval
)
1386 __pthread_testcancel(1);
1387 return(poll_nocancel(p
, (struct poll_nocancel_args
*)uap
, retval
));
1392 poll_nocancel(struct proc
*p
, struct poll_nocancel_args
*uap
, register_t
*retval
)
1394 struct poll_continue_args
*cont
;
1398 int ncoll
, error
= 0;
1399 u_int nfds
= uap
->nfds
;
1405 * This is kinda bogus. We have fd limits, but that is not
1406 * really related to the size of the pollfd array. Make sure
1407 * we let the process use at least FD_SETSIZE entries and at
1408 * least enough for the current limits. We want to be reasonably
1409 * safe, but not overly restrictive.
1411 if (nfds
> OPEN_MAX
||
1412 (nfds
> p
->p_rlimit
[RLIMIT_NOFILE
].rlim_cur
&& (proc_suser(p
) || nfds
> FD_SETSIZE
)))
1415 kq
= kqueue_alloc(p
);
1419 ni
= nfds
* sizeof(struct pollfd
) + sizeof(struct poll_continue_args
);
1420 MALLOC(cont
, struct poll_continue_args
*, ni
, M_TEMP
, M_WAITOK
);
1426 fds
= (struct pollfd
*)&cont
[1];
1427 error
= copyin(uap
->fds
, fds
, nfds
* sizeof(struct pollfd
));
1431 if (uap
->timeout
!= -1) {
1434 atv
.tv_sec
= uap
->timeout
/ 1000;
1435 atv
.tv_usec
= (uap
->timeout
% 1000) * 1000;
1436 if (itimerfix(&atv
)) {
1440 getmicrouptime(&rtv
);
1441 timevaladd(&atv
, &rtv
);
1447 /* JMM - all this P_SELECT stuff is bogus */
1449 OSBitOrAtomic(P_SELECT
, (UInt32
*)&p
->p_flag
);
1450 for (i
= 0; i
< nfds
; i
++) {
1451 short events
= fds
[i
].events
;
1455 /* per spec, ignore fd values below zero */
1456 if (fds
[i
].fd
< 0) {
1461 /* convert the poll event into a kqueue kevent */
1462 kev
.ident
= fds
[i
].fd
;
1463 kev
.flags
= EV_ADD
| EV_ONESHOT
| EV_POLL
;
1464 kev
.fflags
= NOTE_LOWAT
;
1465 kev
.data
= 1; /* efficiency be damned: any data should trigger */
1466 kev
.udata
= CAST_USER_ADDR_T(&fds
[i
]);
1468 /* Handle input events */
1469 if (events
& ( POLLIN
| POLLRDNORM
| POLLPRI
| POLLRDBAND
| POLLHUP
)) {
1470 kev
.filter
= EVFILT_READ
;
1471 if (!(events
& ( POLLIN
| POLLRDNORM
)))
1472 kev
.flags
|= EV_OOBAND
;
1473 kerror
= kevent_register(kq
, &kev
, p
);
1476 /* Handle output events */
1478 events
& ( POLLOUT
| POLLWRNORM
| POLLWRBAND
)) {
1479 kev
.filter
= EVFILT_WRITE
;
1480 kerror
= kevent_register(kq
, &kev
, p
);
1483 /* Handle BSD extension vnode events */
1485 events
& ( POLLEXTEND
| POLLATTRIB
| POLLNLINK
| POLLWRITE
)) {
1486 kev
.filter
= EVFILT_VNODE
;
1488 if (events
& POLLEXTEND
)
1489 kev
.fflags
|= NOTE_EXTEND
;
1490 if (events
& POLLATTRIB
)
1491 kev
.fflags
|= NOTE_ATTRIB
;
1492 if (events
& POLLNLINK
)
1493 kev
.fflags
|= NOTE_LINK
;
1494 if (events
& POLLWRITE
)
1495 kev
.fflags
|= NOTE_WRITE
;
1496 kerror
= kevent_register(kq
, &kev
, p
);
1500 fds
[i
].revents
= POLLNVAL
;
1506 /* Did we have any trouble registering? */
1510 /* scan for, and possibly wait for, the kevents to trigger */
1511 cont
->pca_fds
= uap
->fds
;
1512 cont
->pca_nfds
= nfds
;
1513 cont
->pca_rfds
= rfds
;
1514 error
= kevent_scan(kq
, poll_callback
, NULL
, cont
, &atv
, p
);
1515 rfds
= cont
->pca_rfds
;
1518 OSBitAndAtomic(~((uint32_t)P_SELECT
), (UInt32
*)&p
->p_flag
);
1519 /* poll is not restarted after signals... */
1520 if (error
== ERESTART
)
1522 if (error
== EWOULDBLOCK
)
1525 error
= copyout(fds
, uap
->fds
, nfds
* sizeof(struct pollfd
));
1538 poll_callback(__unused
struct kqueue
*kq
, struct kevent
*kevp
, void *data
)
1540 struct poll_continue_args
*cont
= (struct poll_continue_args
*)data
;
1541 struct pollfd
*fds
= CAST_DOWN(struct pollfd
*, kevp
->udata
);
1544 /* convert the results back into revents */
1545 if (kevp
->flags
& EV_EOF
)
1546 fds
->revents
|= POLLHUP
;
1547 if (kevp
->flags
& EV_ERROR
)
1548 fds
->revents
|= POLLERR
;
1550 switch (kevp
->filter
) {
1552 if (fds
->revents
& POLLHUP
)
1553 mask
= (POLLIN
| POLLRDNORM
| POLLPRI
| POLLRDBAND
);
1556 if (kevp
->data
!= 0)
1557 mask
|= (POLLIN
| POLLRDNORM
);
1558 if (kevp
->flags
& EV_OOBAND
)
1559 mask
|= ( POLLPRI
| POLLRDBAND
);
1561 fds
->revents
|= (fds
->events
& mask
);
1565 if (!(fds
->revents
& POLLHUP
))
1566 fds
->revents
|= (fds
->events
& ( POLLOUT
| POLLWRNORM
| POLLWRBAND
));
1570 if (kevp
->fflags
& NOTE_EXTEND
)
1571 fds
->revents
|= (fds
->events
& POLLEXTEND
);
1572 if (kevp
->fflags
& NOTE_ATTRIB
)
1573 fds
->revents
|= (fds
->events
& POLLATTRIB
);
1574 if (kevp
->fflags
& NOTE_LINK
)
1575 fds
->revents
|= (fds
->events
& POLLNLINK
);
1576 if (kevp
->fflags
& NOTE_WRITE
)
1577 fds
->revents
|= (fds
->events
& POLLWRITE
);
1588 seltrue(__unused dev_t dev
, __unused
int flag
, __unused
struct proc
*p
)
1595 selcount(struct proc
*p
, u_int32_t
*ibits
, __unused u_int32_t
*obits
,
1596 int nfd
, int *countp
, int * kfcountp
)
1598 struct filedesc
*fdp
= p
->p_fd
;
1601 struct fileproc
*fp
;
1611 * Problems when reboot; due to MacOSX signal probs
1612 * in Beaker1C ; verify that the p->p_fd is valid
1619 nw
= howmany(nfd
, NFDBITS
);
1622 for (msk
= 0; msk
< 3; msk
++) {
1623 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1624 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1625 bits
= iptr
[i
/NFDBITS
];
1626 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1628 fp
= fdp
->fd_ofiles
[fd
];
1630 (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)) {
1637 if ((fp
->f_type
== DTYPE_VNODE
)
1638 && ((vp
= (struct vnode
*)fp
->f_data
) != NULLVP
)
1639 && (vp
->v_type
== VCHR
) )
1656 /* undo the iocounts */
1657 for (msk
= 0; msk
< 3; msk
++) {
1658 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1659 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1660 bits
= iptr
[i
/NFDBITS
];
1661 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1663 fp
= fdp
->fd_ofiles
[fd
];
1668 if (p
->p_fpdrainwait
&& fp
->f_iocount
== 0) {
1669 p
->p_fpdrainwait
= 0;
1670 wakeup(&p
->p_fpdrainwait
);
1682 seldrop(struct proc
*p
, u_int32_t
*ibits
, int nfd
)
1684 struct filedesc
*fdp
= p
->p_fd
;
1687 struct fileproc
*fp
;
1693 * Problems when reboot; due to MacOSX signal probs
1694 * in Beaker1C ; verify that the p->p_fd is valid
1700 nw
= howmany(nfd
, NFDBITS
);
1704 for (msk
= 0; msk
< 3; msk
++) {
1705 iptr
= (u_int32_t
*)&ibits
[msk
* nw
];
1706 for (i
= 0; i
< nfd
; i
+= NFDBITS
) {
1707 bits
= iptr
[i
/NFDBITS
];
1708 while ((j
= ffs(bits
)) && (fd
= i
+ --j
) < nfd
) {
1710 fp
= fdp
->fd_ofiles
[fd
];
1713 /* if you are here then it is being closed */
1714 || (fdp
->fd_ofileflags
[fd
] & UF_RESERVED
)
1722 fp
->f_flags
&= ~FP_INSELECT
;
1724 if (p
->p_fpdrainwait
&& fp
->f_iocount
== 0) {
1725 p
->p_fpdrainwait
= 0;
1726 wakeup(&p
->p_fpdrainwait
);
1736 * Record a select request.
1739 selrecord(__unused
struct proc
*selector
, struct selinfo
*sip
, void * p_wql
)
1741 thread_t cur_act
= current_thread();
1742 struct uthread
* ut
= get_bsdthread_info(cur_act
);
1744 /* need to look at collisions */
1746 if ((p_wql
== (void *)0) && ((sip
->si_flags
& SI_INITED
) == 0)) {
1750 /*do not record if this is second pass of select */
1751 if((p_wql
== (void *)0)) {
1755 if ((sip
->si_flags
& SI_INITED
) == 0) {
1756 wait_queue_init(&sip
->si_wait_queue
, SYNC_POLICY_FIFO
);
1757 sip
->si_flags
|= SI_INITED
;
1758 sip
->si_flags
&= ~SI_CLEAR
;
1761 if (sip
->si_flags
& SI_RECORDED
) {
1762 sip
->si_flags
|= SI_COLL
;
1764 sip
->si_flags
&= ~SI_COLL
;
1766 sip
->si_flags
|= SI_RECORDED
;
1767 if (!wait_queue_member(&sip
->si_wait_queue
, ut
->uu_wqset
))
1768 wait_queue_link_noalloc(&sip
->si_wait_queue
, ut
->uu_wqset
,
1769 (wait_queue_link_t
)p_wql
);
1775 selwakeup(struct selinfo
*sip
)
1778 if ((sip
->si_flags
& SI_INITED
) == 0) {
1782 if (sip
->si_flags
& SI_COLL
) {
1784 sip
->si_flags
&= ~SI_COLL
;
1786 /* will not support */
1787 //wakeup((caddr_t)&selwait);
1791 if (sip
->si_flags
& SI_RECORDED
) {
1792 wait_queue_wakeup_all(&sip
->si_wait_queue
, &selwait
, THREAD_AWAKENED
);
1793 sip
->si_flags
&= ~SI_RECORDED
;
1799 selthreadclear(struct selinfo
*sip
)
1802 if ((sip
->si_flags
& SI_INITED
) == 0) {
1805 if (sip
->si_flags
& SI_RECORDED
) {
1807 sip
->si_flags
&= ~(SI_RECORDED
| SI_COLL
);
1809 sip
->si_flags
|= SI_CLEAR
;
1810 wait_queue_unlinkall_nofree(&sip
->si_wait_queue
);
1816 #define DBG_POST 0x10
1817 #define DBG_WATCH 0x11
1818 #define DBG_WAIT 0x12
1819 #define DBG_MOD 0x13
1820 #define DBG_EWAKEUP 0x14
1821 #define DBG_ENQUEUE 0x15
1822 #define DBG_DEQUEUE 0x16
1824 #define DBG_MISC_POST MISCDBG_CODE(DBG_EVENT,DBG_POST)
1825 #define DBG_MISC_WATCH MISCDBG_CODE(DBG_EVENT,DBG_WATCH)
1826 #define DBG_MISC_WAIT MISCDBG_CODE(DBG_EVENT,DBG_WAIT)
1827 #define DBG_MISC_MOD MISCDBG_CODE(DBG_EVENT,DBG_MOD)
1828 #define DBG_MISC_EWAKEUP MISCDBG_CODE(DBG_EVENT,DBG_EWAKEUP)
1829 #define DBG_MISC_ENQUEUE MISCDBG_CODE(DBG_EVENT,DBG_ENQUEUE)
1830 #define DBG_MISC_DEQUEUE MISCDBG_CODE(DBG_EVENT,DBG_DEQUEUE)
1833 #define EVPROCDEQUE(p, evq) do { \
1835 if (evq->ee_flags & EV_QUEUED) { \
1836 TAILQ_REMOVE(&p->p_evlist, evq, ee_plist); \
1837 evq->ee_flags &= ~EV_QUEUED; \
1844 * called upon socket close. deque and free all events for
1845 * the socket... socket must be locked by caller.
1848 evsofree(struct socket
*sp
)
1850 struct eventqelt
*evq
, *next
;
1856 for (evq
= sp
->so_evlist
.tqh_first
; evq
!= NULL
; evq
= next
) {
1857 next
= evq
->ee_slist
.tqe_next
;
1860 if (evq
->ee_flags
& EV_QUEUED
) {
1861 EVPROCDEQUE(p
, evq
);
1863 TAILQ_REMOVE(&sp
->so_evlist
, evq
, ee_slist
); // remove from socket q
1870 * called upon pipe close. deque and free all events for
1871 * the pipe... pipe must be locked by caller
1874 evpipefree(struct pipe
*cpipe
)
1876 struct eventqelt
*evq
, *next
;
1879 for (evq
= cpipe
->pipe_evlist
.tqh_first
; evq
!= NULL
; evq
= next
) {
1880 next
= evq
->ee_slist
.tqe_next
;
1883 EVPROCDEQUE(p
, evq
);
1885 TAILQ_REMOVE(&cpipe
->pipe_evlist
, evq
, ee_slist
); // remove from pipe q
1892 * enqueue this event if it's not already queued. wakeup
1893 * the proc if we do queue this event to it...
1894 * entered with proc lock held... we drop it before
1895 * doing the wakeup and return in that state
1898 evprocenque(struct eventqelt
*evq
)
1905 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_START
, (uint32_t)evq
, evq
->ee_flags
, evq
->ee_eventmask
,0,0);
1909 if (evq
->ee_flags
& EV_QUEUED
) {
1912 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_END
, 0,0,0,0,0);
1915 evq
->ee_flags
|= EV_QUEUED
;
1917 TAILQ_INSERT_TAIL(&p
->p_evlist
, evq
, ee_plist
);
1921 wakeup(&p
->p_evlist
);
1923 KERNEL_DEBUG(DBG_MISC_ENQUEUE
|DBG_FUNC_END
, 0,0,0,0,0);
1928 * pipe lock must be taken by the caller
1931 postpipeevent(struct pipe
*pipep
, int event
)
1934 struct eventqelt
*evq
;
1938 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_START
, event
,0,0,1,0);
1940 for (evq
= pipep
->pipe_evlist
.tqh_first
;
1941 evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
1943 if (evq
->ee_eventmask
== 0)
1947 switch (event
& (EV_RWBYTES
| EV_RCLOSED
| EV_WCLOSED
)) {
1950 if ((evq
->ee_eventmask
& EV_RE
) && pipep
->pipe_buffer
.cnt
) {
1952 evq
->ee_req
.er_rcnt
= pipep
->pipe_buffer
.cnt
;
1954 if ((evq
->ee_eventmask
& EV_WR
) &&
1955 (pipep
->pipe_buffer
.size
- pipep
->pipe_buffer
.cnt
) >= PIPE_BUF
) {
1957 if (pipep
->pipe_state
& PIPE_EOF
) {
1958 mask
|= EV_WR
|EV_RESET
;
1962 evq
->ee_req
.er_wcnt
= pipep
->pipe_buffer
.size
- pipep
->pipe_buffer
.cnt
;
1968 if ((evq
->ee_eventmask
& EV_RE
)) {
1969 mask
|= EV_RE
|EV_RCLOSED
;
1971 if ((evq
->ee_eventmask
& EV_WR
)) {
1972 mask
|= EV_WR
|EV_WCLOSED
;
1981 * disarm... postevents are nops until this event is 'read' via
1982 * waitevent and then re-armed via modwatch
1984 evq
->ee_eventmask
= 0;
1987 * since events are disarmed until after the waitevent
1988 * the ee_req.er_xxxx fields can't change once we've
1989 * inserted this event into the proc queue...
1990 * therefore, the waitevent will see a 'consistent'
1991 * snapshot of the event, even though it won't hold
1992 * the pipe lock, and we're updating the event outside
1993 * of the proc lock, which it will hold
1995 evq
->ee_req
.er_eventbits
|= mask
;
1997 KERNEL_DEBUG(DBG_MISC_POST
, (uint32_t)evq
, evq
->ee_req
.er_eventbits
, mask
, 1,0);
2002 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, 0,0,0,1,0);
2007 * given either a sockbuf or a socket run down the
2008 * event list and queue ready events found...
2009 * the socket must be locked by the caller
2012 postevent(struct socket
*sp
, struct sockbuf
*sb
, int event
)
2015 struct eventqelt
*evq
;
2023 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_START
, (int)sp
, event
, 0, 0, 0);
2025 for (evq
= sp
->so_evlist
.tqh_first
;
2026 evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2028 if (evq
->ee_eventmask
== 0)
2032 /* ready for reading:
2033 - byte cnt >= receive low water mark
2034 - read-half of conn closed
2035 - conn pending for listening sock
2036 - socket error pending
2039 - byte cnt avail >= send low water mark
2040 - write half of conn closed
2041 - socket error pending
2042 - non-blocking conn completed successfully
2046 - sock at out of band mark
2049 switch (event
& EV_DMASK
) {
2052 if ((evq
->ee_eventmask
& EV_EX
)) {
2053 if (sp
->so_oobmark
|| ((sp
->so_state
& SS_RCVATMARK
)))
2054 mask
|= EV_EX
|EV_OOB
;
2058 case EV_RWBYTES
|EV_OOB
:
2059 if ((evq
->ee_eventmask
& EV_EX
)) {
2060 if (sp
->so_oobmark
|| ((sp
->so_state
& SS_RCVATMARK
)))
2061 mask
|= EV_EX
|EV_OOB
;
2064 * fall into the next case
2067 if ((evq
->ee_eventmask
& EV_RE
) && soreadable(sp
)) {
2069 if ((sp
->so_type
== SOCK_STREAM
) && ((sp
->so_error
== ECONNREFUSED
) || (sp
->so_error
== ECONNRESET
))) {
2070 if ((sp
->so_pcb
== 0) || (((struct inpcb
*)sp
->so_pcb
)->inp_state
== INPCB_STATE_DEAD
) || !(tp
= sototcpcb(sp
)) ||
2071 (tp
->t_state
== TCPS_CLOSED
)) {
2072 mask
|= EV_RE
|EV_RESET
;
2078 evq
->ee_req
.er_rcnt
= sp
->so_rcv
.sb_cc
;
2080 if (sp
->so_state
& SS_CANTRCVMORE
) {
2085 if ((evq
->ee_eventmask
& EV_WR
) && sowriteable(sp
)) {
2087 if ((sp
->so_type
== SOCK_STREAM
) && ((sp
->so_error
== ECONNREFUSED
) || (sp
->so_error
== ECONNRESET
))) {
2088 if ((sp
->so_pcb
== 0) || (((struct inpcb
*)sp
->so_pcb
)->inp_state
== INPCB_STATE_DEAD
) || !(tp
= sototcpcb(sp
)) ||
2089 (tp
->t_state
== TCPS_CLOSED
)) {
2090 mask
|= EV_WR
|EV_RESET
;
2096 evq
->ee_req
.er_wcnt
= sbspace(&sp
->so_snd
);
2101 if ((evq
->ee_eventmask
& EV_RE
)) {
2102 mask
|= EV_RE
|EV_RCONN
;
2103 evq
->ee_req
.er_rcnt
= sp
->so_qlen
+ 1; // incl this one
2108 if ((evq
->ee_eventmask
& EV_WR
)) {
2109 mask
|= EV_WR
|EV_WCONN
;
2114 if ((evq
->ee_eventmask
& EV_RE
)) {
2115 mask
|= EV_RE
|EV_RCLOSED
;
2120 if ((evq
->ee_eventmask
& EV_WR
)) {
2121 mask
|= EV_WR
|EV_WCLOSED
;
2126 if (evq
->ee_eventmask
& EV_RE
) {
2127 mask
|= EV_RE
|EV_FIN
;
2133 if (evq
->ee_eventmask
& EV_RE
) {
2134 mask
|= EV_RE
| event
;
2136 if (evq
->ee_eventmask
& EV_WR
) {
2137 mask
|= EV_WR
| event
;
2142 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, (int)sp
, -1, 0, 0, 0);
2146 KERNEL_DEBUG(DBG_MISC_POST
, (int)evq
, evq
->ee_eventmask
, evq
->ee_req
.er_eventbits
, mask
, 0);
2150 * disarm... postevents are nops until this event is 'read' via
2151 * waitevent and then re-armed via modwatch
2153 evq
->ee_eventmask
= 0;
2156 * since events are disarmed until after the waitevent
2157 * the ee_req.er_xxxx fields can't change once we've
2158 * inserted this event into the proc queue...
2159 * since waitevent can't see this event until we
2160 * enqueue it, waitevent will see a 'consistent'
2161 * snapshot of the event, even though it won't hold
2162 * the socket lock, and we're updating the event outside
2163 * of the proc lock, which it will hold
2165 evq
->ee_req
.er_eventbits
|= mask
;
2170 KERNEL_DEBUG(DBG_MISC_POST
|DBG_FUNC_END
, (int)sp
, 0, 0, 0, 0);
2172 #endif /* SOCKETS */
2176 * watchevent system call. user passes us an event to watch
2177 * for. we malloc an event object, initialize it, and queue
2178 * it to the open socket. when the event occurs, postevent()
2179 * will enque it back to our proc where we can retrieve it
2182 * should this prevent duplicate events on same socket?
2185 * ENOMEM No memory for operation
2189 watchevent(proc_t p
, struct watchevent_args
*uap
, __unused
int *retval
)
2191 struct eventqelt
*evq
= (struct eventqelt
*)0;
2192 struct eventqelt
*np
= NULL
;
2193 struct eventreq64
*erp
;
2194 struct fileproc
*fp
= NULL
;
2197 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_START
, 0,0,0,0,0);
2199 // get a qelt and fill with users req
2200 MALLOC(evq
, struct eventqelt
*, sizeof(struct eventqelt
), M_TEMP
, M_WAITOK
);
2206 // get users request pkt
2208 if (IS_64BIT_PROCESS(p
)) {
2209 error
= copyin(uap
->u_req
, (caddr_t
)erp
, sizeof(struct eventreq64
));
2211 struct eventreq32 er32
;
2213 error
= copyin(uap
->u_req
, (caddr_t
)&er32
, sizeof(struct eventreq32
));
2216 * the user only passes in the
2217 * er_type, er_handle and er_data...
2218 * the other fields are initialized
2219 * below, so don't bother to copy
2221 erp
->er_type
= er32
.er_type
;
2222 erp
->er_handle
= er32
.er_handle
;
2223 erp
->er_data
= (user_addr_t
)er32
.er_data
;
2228 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, error
,0,0,0,0);
2232 KERNEL_DEBUG(DBG_MISC_WATCH
, erp
->er_handle
,uap
->u_eventmask
,(uint32_t)evq
,0,0);
2234 // validate, freeing qelt if errors
2238 if (erp
->er_type
!= EV_FD
) {
2240 } else if ((error
= fp_lookup(p
, erp
->er_handle
, &fp
, 1)) != 0) {
2243 } else if (fp
->f_type
== DTYPE_SOCKET
) {
2244 socket_lock((struct socket
*)fp
->f_data
, 1);
2245 np
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2246 #endif /* SOCKETS */
2247 } else if (fp
->f_type
== DTYPE_PIPE
) {
2248 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2249 np
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2251 fp_drop(p
, erp
->er_handle
, fp
, 1);
2259 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, error
,0,0,0,0);
2264 * only allow one watch per file per proc
2266 for ( ; np
!= NULL
; np
= np
->ee_slist
.tqe_next
) {
2267 if (np
->ee_proc
== p
) {
2269 if (fp
->f_type
== DTYPE_SOCKET
)
2270 socket_unlock((struct socket
*)fp
->f_data
, 1);
2272 #endif /* SOCKETS */
2273 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2274 fp_drop(p
, erp
->er_handle
, fp
, 0);
2277 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2281 erp
->er_ecnt
= erp
->er_rcnt
= erp
->er_wcnt
= erp
->er_eventbits
= 0;
2283 evq
->ee_eventmask
= uap
->u_eventmask
& EV_MASK
;
2287 if (fp
->f_type
== DTYPE_SOCKET
) {
2288 TAILQ_INSERT_TAIL(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2289 postevent((struct socket
*)fp
->f_data
, 0, EV_RWBYTES
); // catch existing events
2291 socket_unlock((struct socket
*)fp
->f_data
, 1);
2293 #endif /* SOCKETS */
2295 TAILQ_INSERT_TAIL(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2296 postpipeevent((struct pipe
*)fp
->f_data
, EV_RWBYTES
);
2298 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2300 fp_drop_event(p
, erp
->er_handle
, fp
);
2302 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, 0,0,0,0,0);
2309 * waitevent system call.
2310 * grabs the next waiting event for this proc and returns
2311 * it. if no events, user can request to sleep with timeout
2312 * or without or poll mode
2313 * ((tv != NULL && interval == 0) || tv == -1)
2316 waitevent(proc_t p
, struct waitevent_args
*uap
, int *retval
)
2319 struct eventqelt
*evq
;
2320 struct eventreq64
*erp
;
2321 uint64_t abstime
, interval
;
2322 boolean_t fast_poll
= FALSE
;
2324 struct eventreq64 er64
;
2325 struct eventreq32 er32
;
2333 * check for fast poll method
2335 if (IS_64BIT_PROCESS(p
)) {
2336 if (uap
->tv
== (user_addr_t
)-1)
2338 } else if (uap
->tv
== (user_addr_t
)((uint32_t)-1))
2341 if (fast_poll
== TRUE
) {
2342 if (p
->p_evlist
.tqh_first
== NULL
) {
2343 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_NONE
, -1,0,0,0,0);
2353 error
= copyin(uap
->tv
, (caddr_t
)&atv
, sizeof (atv
));
2357 if (itimerfix(&atv
)) {
2361 interval
= tvtoabstime(&atv
);
2363 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_START
, 0,0,0,0,0);
2367 if ((evq
= p
->p_evlist
.tqh_first
) != NULL
) {
2369 * found one... make a local copy while it's still on the queue
2370 * to prevent it from changing while in the midst of copying
2371 * don't want to hold the proc lock across a copyout because
2372 * it might block on a page fault at the target in user space
2376 if (IS_64BIT_PROCESS(p
))
2377 bcopy((caddr_t
)erp
, (caddr_t
)&uer
.er64
, sizeof (struct eventreq64
));
2379 uer
.er32
.er_type
= erp
->er_type
;
2380 uer
.er32
.er_handle
= erp
->er_handle
;
2381 uer
.er32
.er_data
= (uint32_t)erp
->er_data
;
2382 uer
.er32
.er_ecnt
= erp
->er_ecnt
;
2383 uer
.er32
.er_rcnt
= erp
->er_rcnt
;
2384 uer
.er32
.er_wcnt
= erp
->er_wcnt
;
2385 uer
.er32
.er_eventbits
= erp
->er_eventbits
;
2387 TAILQ_REMOVE(&p
->p_evlist
, evq
, ee_plist
);
2389 evq
->ee_flags
&= ~EV_QUEUED
;
2393 if (IS_64BIT_PROCESS(p
))
2394 error
= copyout((caddr_t
)&uer
.er64
, uap
->u_req
, sizeof(struct eventreq64
));
2396 error
= copyout((caddr_t
)&uer
.er32
, uap
->u_req
, sizeof(struct eventreq32
));
2398 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, error
,
2399 evq
->ee_req
.er_handle
,evq
->ee_req
.er_eventbits
,(uint32_t)evq
,0);
2403 if (uap
->tv
&& interval
== 0) {
2405 *retval
= 1; // poll failed
2407 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, error
,0,0,0,0);
2411 clock_absolutetime_interval_to_deadline(interval
, &abstime
);
2415 KERNEL_DEBUG(DBG_MISC_WAIT
, 1,(uint32_t)&p
->p_evlist
,0,0,0);
2417 error
= msleep1(&p
->p_evlist
, &p
->p_mlock
, (PSOCK
| PCATCH
), "waitevent", abstime
);
2419 KERNEL_DEBUG(DBG_MISC_WAIT
, 2,(uint32_t)&p
->p_evlist
,0,0,0);
2423 if (error
== ERESTART
)
2425 if (error
== EWOULDBLOCK
) {
2432 KERNEL_DEBUG(DBG_MISC_WAIT
|DBG_FUNC_END
, 0,0,0,0,0);
2438 * modwatch system call. user passes in event to modify.
2439 * if we find it we reset the event bits and que/deque event
2443 modwatch(proc_t p
, struct modwatch_args
*uap
, __unused
int *retval
)
2445 struct eventreq64 er
;
2446 struct eventreq64
*erp
= &er
;
2447 struct eventqelt
*evq
= NULL
; /* protected by error return */
2449 struct fileproc
*fp
;
2452 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_START
, 0,0,0,0,0);
2455 * get user's request pkt
2456 * just need the er_type and er_handle which sit above the
2457 * problematic er_data (32/64 issue)... so only copy in
2460 if ((error
= copyin(uap
->u_req
, (caddr_t
)erp
, sizeof(er
.er_type
) + sizeof(er
.er_handle
)))) {
2461 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, error
,0,0,0,0);
2466 if (erp
->er_type
!= EV_FD
) {
2468 } else if ((error
= fp_lookup(p
, erp
->er_handle
, &fp
, 1)) != 0) {
2471 } else if (fp
->f_type
== DTYPE_SOCKET
) {
2472 socket_lock((struct socket
*)fp
->f_data
, 1);
2473 evq
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2474 #endif /* SOCKETS */
2475 } else if (fp
->f_type
== DTYPE_PIPE
) {
2476 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2477 evq
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2479 fp_drop(p
, erp
->er_handle
, fp
, 1);
2485 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, error
,0,0,0,0);
2489 if ((uap
->u_eventmask
== EV_RM
) && (fp
->f_flags
& FP_WAITEVENT
)) {
2490 fp
->f_flags
&= ~FP_WAITEVENT
;
2494 // locate event if possible
2495 for ( ; evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2496 if (evq
->ee_proc
== p
)
2501 if (fp
->f_type
== DTYPE_SOCKET
)
2502 socket_unlock((struct socket
*)fp
->f_data
, 1);
2504 #endif /* SOCKETS */
2505 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2506 fp_drop(p
, erp
->er_handle
, fp
, 0);
2507 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2510 KERNEL_DEBUG(DBG_MISC_MOD
, erp
->er_handle
,uap
->u_eventmask
,(uint32_t)evq
,0,0);
2512 if (uap
->u_eventmask
== EV_RM
) {
2513 EVPROCDEQUE(p
, evq
);
2516 if (fp
->f_type
== DTYPE_SOCKET
) {
2517 TAILQ_REMOVE(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2518 socket_unlock((struct socket
*)fp
->f_data
, 1);
2520 #endif /* SOCKETS */
2522 TAILQ_REMOVE(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2523 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2525 fp_drop(p
, erp
->er_handle
, fp
, 0);
2527 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, 0,0,0,0,0);
2530 switch (uap
->u_eventmask
& EV_MASK
) {
2548 case EV_EX
|EV_RE
|EV_WR
:
2549 flag
= EV_OOB
|EV_RWBYTES
;
2554 if (fp
->f_type
== DTYPE_SOCKET
)
2555 socket_unlock((struct socket
*)fp
->f_data
, 1);
2557 #endif /* SOCKETS */
2558 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2559 fp_drop(p
, erp
->er_handle
, fp
, 0);
2560 KERNEL_DEBUG(DBG_MISC_WATCH
|DBG_FUNC_END
, EINVAL
,0,0,0,0);
2564 * since we're holding the socket/pipe lock, the event
2565 * cannot go from the unqueued state to the queued state
2566 * however, it can go from the queued state to the unqueued state
2567 * since that direction is protected by the proc_lock...
2568 * so do a quick check for EV_QUEUED w/o holding the proc lock
2569 * since by far the common case will be NOT EV_QUEUED, this saves
2570 * us taking the proc_lock the majority of the time
2572 if (evq
->ee_flags
& EV_QUEUED
) {
2574 * EVPROCDEQUE will recheck the state after it grabs the proc_lock
2576 EVPROCDEQUE(p
, evq
);
2579 * while the event is off the proc queue and
2580 * we're holding the socket/pipe lock
2581 * it's safe to update these fields...
2583 evq
->ee_req
.er_eventbits
= 0;
2584 evq
->ee_eventmask
= uap
->u_eventmask
& EV_MASK
;
2587 if (fp
->f_type
== DTYPE_SOCKET
) {
2588 postevent((struct socket
*)fp
->f_data
, 0, flag
);
2589 socket_unlock((struct socket
*)fp
->f_data
, 1);
2591 #endif /* SOCKETS */
2593 postpipeevent((struct pipe
*)fp
->f_data
, flag
);
2594 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2596 fp_drop(p
, erp
->er_handle
, fp
, 0);
2597 KERNEL_DEBUG(DBG_MISC_MOD
|DBG_FUNC_END
, evq
->ee_req
.er_handle
,evq
->ee_eventmask
,(uint32_t)fp
->f_data
,flag
,0);
2601 /* this routine is called from the close of fd with proc_fdlock held */
2603 waitevent_close(struct proc
*p
, struct fileproc
*fp
)
2605 struct eventqelt
*evq
;
2608 fp
->f_flags
&= ~FP_WAITEVENT
;
2611 if (fp
->f_type
== DTYPE_SOCKET
) {
2612 socket_lock((struct socket
*)fp
->f_data
, 1);
2613 evq
= ((struct socket
*)fp
->f_data
)->so_evlist
.tqh_first
;
2615 #endif /* SOCKETS */
2616 if (fp
->f_type
== DTYPE_PIPE
) {
2617 PIPE_LOCK((struct pipe
*)fp
->f_data
);
2618 evq
= ((struct pipe
*)fp
->f_data
)->pipe_evlist
.tqh_first
;
2626 // locate event if possible
2627 for ( ; evq
!= NULL
; evq
= evq
->ee_slist
.tqe_next
) {
2628 if (evq
->ee_proc
== p
)
2633 if (fp
->f_type
== DTYPE_SOCKET
)
2634 socket_unlock((struct socket
*)fp
->f_data
, 1);
2636 #endif /* SOCKETS */
2637 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2643 EVPROCDEQUE(p
, evq
);
2646 if (fp
->f_type
== DTYPE_SOCKET
) {
2647 TAILQ_REMOVE(&((struct socket
*)fp
->f_data
)->so_evlist
, evq
, ee_slist
);
2648 socket_unlock((struct socket
*)fp
->f_data
, 1);
2650 #endif /* SOCKETS */
2652 TAILQ_REMOVE(&((struct pipe
*)fp
->f_data
)->pipe_evlist
, evq
, ee_slist
);
2653 PIPE_UNLOCK((struct pipe
*)fp
->f_data
);
2666 * Description: Get the host UUID from IOKit and return it to user space.
2668 * Parameters: uuid_buf Pointer to buffer to receive UUID
2669 * timeout Timespec for timout
2671 * Returns: 0 Success
2672 * EWOULDBLOCK Timeout is too short
2673 * copyout:EFAULT Bad user buffer
2675 * Notes: A timeout seems redundant, since if it's tolerable to not
2676 * have a system UUID in hand, then why ask for one?
2679 gethostuuid(struct proc
*p
, struct gethostuuid_args
*uap
, __unused register_t
*retval
)
2683 mach_timespec_t mach_ts
; /* for IOKit call */
2684 __darwin_uuid_t uuid_kern
; /* for IOKit call */
2686 /* Convert the 32/64 bit timespec into a mach_timespec_t */
2687 if ( proc_is64bit(p
) ) {
2688 struct user_timespec ts
;
2689 error
= copyin(uap
->timeoutp
, &ts
, sizeof(ts
));
2692 mach_ts
.tv_sec
= ts
.tv_sec
;
2693 mach_ts
.tv_nsec
= ts
.tv_nsec
;
2696 error
= copyin(uap
->timeoutp
, &ts
, sizeof(ts
) );
2699 mach_ts
.tv_sec
= ts
.tv_sec
;
2700 mach_ts
.tv_nsec
= ts
.tv_nsec
;
2703 /* Call IOKit with the stack buffer to get the UUID */
2704 kret
= IOBSDGetPlatformUUID(uuid_kern
, mach_ts
);
2707 * If we get it, copy out the data to the user buffer; note that a
2708 * uuid_t is an array of characters, so this is size invariant for
2711 if (kret
== KERN_SUCCESS
) {
2712 error
= copyout(uuid_kern
, uap
->uuid_buf
, sizeof(uuid_kern
));
2714 error
= EWOULDBLOCK
;