diff --git a/lib/libc/sys/aio_return.2 b/lib/libc/sys/aio_return.2 index 2ef6c25953f7..df558734ed41 100644 --- a/lib/libc/sys/aio_return.2 +++ b/lib/libc/sys/aio_return.2 @@ -1,102 +1,102 @@ .\" Copyright (c) 1999 Softweyr LLC. .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY Softweyr LLC AND CONTRIBUTORS ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL Softweyr LLC OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" -.Dd October 7, 2006 +.Dd March 21, 2016 .Dt AIO_RETURN 2 .Os .Sh NAME .Nm aio_return .Nd retrieve return status of asynchronous I/O operation (REALTIME) .Sh LIBRARY .Lb libc .Sh SYNOPSIS .In aio.h -.Ft int +.Ft ssize_t .Fn aio_return "struct aiocb *iocb" .Sh DESCRIPTION The .Fn aio_return system call returns the final status of the asynchronous I/O request associated with the structure pointed to by .Fa iocb . .Pp The .Fn aio_return system call should only be called once, to obtain the final status of an asynchronous I/O operation once it has completed .Xr ( aio_error 2 returns something other than .Er EINPROGRESS ) . .Sh RETURN VALUES If the asynchronous I/O request has completed, the status is returned as described in .Xr read 2 , .Xr write 2 , or .Xr fsync 2 . Otherwise, .Fn aio_return returns \-1 and sets .Va errno to indicate the error condition. .Sh ERRORS The .Fn aio_return system call will fail if: .Bl -tag -width Er .It Bq Er EINVAL The .Fa iocb argument does not reference a completed asynchronous I/O request. .El .Sh SEE ALSO .Xr aio_cancel 2 , .Xr aio_error 2 , .Xr aio_suspend 2 , .Xr aio_waitcomplete 2 , .Xr aio_write 2 , .Xr fsync 2 , .Xr read 2 , .Xr write 2 , .Xr aio 4 .Sh STANDARDS The .Fn aio_return system call is expected to conform to the .St -p1003.1 standard. .Sh HISTORY The .Fn aio_return system call first appeared in .Fx 3.0 . .Sh AUTHORS This manual page was written by .An Wes Peters Aq Mt wes@softweyr.com . diff --git a/lib/libc/sys/aio_waitcomplete.2 b/lib/libc/sys/aio_waitcomplete.2 index 885b3d7fa481..66e7208d7b48 100644 --- a/lib/libc/sys/aio_waitcomplete.2 +++ b/lib/libc/sys/aio_waitcomplete.2 @@ -1,137 +1,137 @@ .\" Copyright (c) 1999 Christopher M Sedore. .\" All rights reserved. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE .\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF .\" SUCH DAMAGE. .\" .\" $FreeBSD$ .\" -.Dd January 19, 2000 +.Dd March 21, 2016 .Dt AIO_WAITCOMPLETE 2 .Os .Sh NAME .Nm aio_waitcomplete .Nd wait for the next completion of an aio request .Sh LIBRARY .Lb libc .Sh SYNOPSIS .In aio.h -.Ft int +.Ft ssize_t .Fn aio_waitcomplete "struct aiocb **iocbp" "struct timespec *timeout" .Sh DESCRIPTION The .Fn aio_waitcomplete system call waits for completion of an asynchronous I/O request. Upon completion, .Fn aio_waitcomplete returns the result of the function and sets .Fa iocbp to point to the structure associated with the original request. If an asynchronous I/O request is completed before .Fn aio_waitcomplete is called, it returns immediately with the completed request. .Pp If .Fa timeout is a non-NULL pointer, it specifies a maximum interval to wait for a asynchronous I/O request to complete. If .Fa timeout is a NULL pointer, .Fn aio_waitcomplete waits indefinitely. To effect a poll, the .Fa timeout argument should be non-NULL, pointing to a zero-valued timeval structure. .Pp The .Fn aio_waitcomplete system call also serves the function of .Fn aio_return , thus .Fn aio_return should not be called for the control block returned in .Fa iocbp . .Sh RETURN VALUES If an asynchronous I/O request has completed, .Fa iocbp is set to point to the control block passed with the original request, and the status is returned as described in .Xr read 2 , .Xr write 2 , or .Xr fsync 2 . On failure, .Fn aio_waitcomplete returns .Dv -1 , sets iocbp to .Dv NULL and sets .Va errno to indicate the error condition. .Sh ERRORS The .Fn aio_waitcomplete system call fails if: .Bl -tag -width Er .It Bq Er EINVAL The specified time limit is invalid. .It Bq Er EAGAIN The process has not yet called .Fn aio_read or .Fn aio_write . .It Bq Er EINTR A signal was delivered before the timeout expired and before any asynchronous I/O requests completed. .It Bq Er EWOULDBLOCK .It Bq Er EINPROGRESS The specified time limit expired before any asynchronous I/O requests completed. .El .Sh SEE ALSO .Xr aio_cancel 2 , .Xr aio_error 2 , .Xr aio_read 2 , .Xr aio_return 2 , .Xr aio_suspend 2 , .Xr aio_write 2 , .Xr fsync 2 , .Xr read 2 , .Xr write 2 , .Xr aio 4 .Sh STANDARDS The .Fn aio_waitcomplete system call is a .Fx Ns -specific extension. .Sh HISTORY The .Fn aio_waitcomplete system call first appeared in .Fx 4.0 . .Sh AUTHORS The .Fn aio_waitcomplete system call and this manual page were written by .An Christopher M Sedore Aq Mt cmsedore@maxwell.syr.edu . diff --git a/sys/kern/syscalls.master b/sys/kern/syscalls.master index f08f5e3b0b43..56754256b410 100644 --- a/sys/kern/syscalls.master +++ b/sys/kern/syscalls.master @@ -1,999 +1,999 @@ $FreeBSD$ ; from: @(#)syscalls.master 8.2 (Berkeley) 1/13/94 ; ; System call name/number master file. ; Processed to created init_sysent.c, syscalls.c and syscall.h. ; Columns: number audit type name alt{name,tag,rtyp}/comments ; number system call number, must be in order ; audit the audit event associated with the system call ; A value of AUE_NULL means no auditing, but it also means that ; there is no audit event for the call at this time. For the ; case where the event exists, but we don't want auditing, the ; event should be #defined to AUE_NULL in audit_kevents.h. ; type one of STD, OBSOL, UNIMPL, COMPAT, COMPAT4, COMPAT6, ; COMPAT7, NODEF, NOARGS, NOPROTO, NOSTD ; The COMPAT* options may be combined with one or more NO* ; options separated by '|' with no spaces (e.g. COMPAT|NOARGS) ; name psuedo-prototype of syscall routine ; If one of the following alts is different, then all appear: ; altname name of system call if different ; alttag name of args struct tag if different from [o]`name'"_args" ; altrtyp return type if not int (bogus - syscalls always return int) ; for UNIMPL/OBSOL, name continues with comments ; types: ; STD always included ; COMPAT included on COMPAT #ifdef ; COMPAT4 included on COMPAT4 #ifdef (FreeBSD 4 compat) ; COMPAT6 included on COMPAT6 #ifdef (FreeBSD 6 compat) ; COMPAT7 included on COMPAT7 #ifdef (FreeBSD 7 compat) ; OBSOL obsolete, not included in system, only specifies name ; UNIMPL not implemented, placeholder only ; NOSTD implemented but as a lkm that can be statically ; compiled in; sysent entry will be filled with lkmressys ; so the SYSCALL_MODULE macro works ; NOARGS same as STD except do not create structure in sys/sysproto.h ; NODEF same as STD except only have the entry in the syscall table ; added. Meaning - do not create structure or function ; prototype in sys/sysproto.h ; NOPROTO same as STD except do not create structure or ; function prototype in sys/sysproto.h. Does add a ; definition to syscall.h besides adding a sysent. ; NOTSTATIC syscall is loadable ; ; Please copy any additions and changes to the following compatability tables: ; sys/compat/freebsd32/syscalls.master ; #ifdef's, etc. may be included, and are copied to the output files. #include #include #include ; Reserved/unimplemented system calls in the range 0-150 inclusive ; are reserved for use in future Berkeley releases. ; Additional system calls implemented in vendor and other ; redistributions should be placed in the reserved range at the end ; of the current calls. 0 AUE_NULL STD { int nosys(void); } syscall nosys_args int 1 AUE_EXIT STD { void sys_exit(int rval); } exit \ sys_exit_args void 2 AUE_FORK STD { int fork(void); } 3 AUE_NULL STD { ssize_t read(int fd, void *buf, \ size_t nbyte); } 4 AUE_NULL STD { ssize_t write(int fd, const void *buf, \ size_t nbyte); } 5 AUE_OPEN_RWTC STD { int open(char *path, int flags, int mode); } ; XXX should be { int open(const char *path, int flags, ...); } ; but we're not ready for `const' or varargs. ; XXX man page says `mode_t mode'. 6 AUE_CLOSE STD { int close(int fd); } 7 AUE_WAIT4 STD { int wait4(int pid, int *status, \ int options, struct rusage *rusage); } 8 AUE_CREAT COMPAT { int creat(char *path, int mode); } 9 AUE_LINK STD { int link(char *path, char *link); } 10 AUE_UNLINK STD { int unlink(char *path); } 11 AUE_NULL OBSOL execv 12 AUE_CHDIR STD { int chdir(char *path); } 13 AUE_FCHDIR STD { int fchdir(int fd); } 14 AUE_MKNOD STD { int mknod(char *path, int mode, int dev); } 15 AUE_CHMOD STD { int chmod(char *path, int mode); } 16 AUE_CHOWN STD { int chown(char *path, int uid, int gid); } 17 AUE_NULL STD { int obreak(char *nsize); } break \ obreak_args int 18 AUE_GETFSSTAT COMPAT4 { int getfsstat(struct ostatfs *buf, \ long bufsize, int flags); } 19 AUE_LSEEK COMPAT { long lseek(int fd, long offset, \ int whence); } 20 AUE_GETPID STD { pid_t getpid(void); } 21 AUE_MOUNT STD { int mount(char *type, char *path, \ int flags, caddr_t data); } ; XXX `path' should have type `const char *' but we're not ready for that. 22 AUE_UMOUNT STD { int unmount(char *path, int flags); } 23 AUE_SETUID STD { int setuid(uid_t uid); } 24 AUE_GETUID STD { uid_t getuid(void); } 25 AUE_GETEUID STD { uid_t geteuid(void); } 26 AUE_PTRACE STD { int ptrace(int req, pid_t pid, \ caddr_t addr, int data); } 27 AUE_RECVMSG STD { int recvmsg(int s, struct msghdr *msg, \ int flags); } 28 AUE_SENDMSG STD { int sendmsg(int s, struct msghdr *msg, \ int flags); } 29 AUE_RECVFROM STD { int recvfrom(int s, caddr_t buf, \ size_t len, int flags, \ struct sockaddr * __restrict from, \ __socklen_t * __restrict fromlenaddr); } 30 AUE_ACCEPT STD { int accept(int s, \ struct sockaddr * __restrict name, \ __socklen_t * __restrict anamelen); } 31 AUE_GETPEERNAME STD { int getpeername(int fdes, \ struct sockaddr * __restrict asa, \ __socklen_t * __restrict alen); } 32 AUE_GETSOCKNAME STD { int getsockname(int fdes, \ struct sockaddr * __restrict asa, \ __socklen_t * __restrict alen); } 33 AUE_ACCESS STD { int access(char *path, int amode); } 34 AUE_CHFLAGS STD { int chflags(const char *path, u_long flags); } 35 AUE_FCHFLAGS STD { int fchflags(int fd, u_long flags); } 36 AUE_SYNC STD { int sync(void); } 37 AUE_KILL STD { int kill(int pid, int signum); } 38 AUE_STAT COMPAT { int stat(char *path, struct ostat *ub); } 39 AUE_GETPPID STD { pid_t getppid(void); } 40 AUE_LSTAT COMPAT { int lstat(char *path, struct ostat *ub); } 41 AUE_DUP STD { int dup(u_int fd); } 42 AUE_PIPE STD { int pipe(void); } 43 AUE_GETEGID STD { gid_t getegid(void); } 44 AUE_PROFILE STD { int profil(caddr_t samples, size_t size, \ size_t offset, u_int scale); } 45 AUE_KTRACE STD { int ktrace(const char *fname, int ops, \ int facs, int pid); } 46 AUE_SIGACTION COMPAT { int sigaction(int signum, \ struct osigaction *nsa, \ struct osigaction *osa); } 47 AUE_GETGID STD { gid_t getgid(void); } 48 AUE_SIGPROCMASK COMPAT { int sigprocmask(int how, osigset_t mask); } ; XXX note nonstandard (bogus) calling convention - the libc stub passes ; us the mask, not a pointer to it, and we return the old mask as the ; (int) return value. 49 AUE_GETLOGIN STD { int getlogin(char *namebuf, u_int \ namelen); } 50 AUE_SETLOGIN STD { int setlogin(char *namebuf); } 51 AUE_ACCT STD { int acct(char *path); } 52 AUE_SIGPENDING COMPAT { int sigpending(void); } 53 AUE_SIGALTSTACK STD { int sigaltstack(stack_t *ss, \ stack_t *oss); } 54 AUE_IOCTL STD { int ioctl(int fd, u_long com, \ caddr_t data); } 55 AUE_REBOOT STD { int reboot(int opt); } 56 AUE_REVOKE STD { int revoke(char *path); } 57 AUE_SYMLINK STD { int symlink(char *path, char *link); } 58 AUE_READLINK STD { ssize_t readlink(char *path, char *buf, \ size_t count); } 59 AUE_EXECVE STD { int execve(char *fname, char **argv, \ char **envv); } 60 AUE_UMASK STD { int umask(int newmask); } umask umask_args \ int 61 AUE_CHROOT STD { int chroot(char *path); } 62 AUE_FSTAT COMPAT { int fstat(int fd, struct ostat *sb); } 63 AUE_NULL COMPAT { int getkerninfo(int op, char *where, \ size_t *size, int arg); } getkerninfo \ getkerninfo_args int 64 AUE_NULL COMPAT { int getpagesize(void); } getpagesize \ getpagesize_args int 65 AUE_MSYNC STD { int msync(void *addr, size_t len, \ int flags); } 66 AUE_VFORK STD { int vfork(void); } 67 AUE_NULL OBSOL vread 68 AUE_NULL OBSOL vwrite 69 AUE_SBRK STD { int sbrk(int incr); } 70 AUE_SSTK STD { int sstk(int incr); } 71 AUE_MMAP COMPAT { int mmap(void *addr, int len, int prot, \ int flags, int fd, long pos); } 72 AUE_O_VADVISE STD { int ovadvise(int anom); } vadvise \ ovadvise_args int 73 AUE_MUNMAP STD { int munmap(void *addr, size_t len); } 74 AUE_MPROTECT STD { int mprotect(const void *addr, size_t len, \ int prot); } 75 AUE_MADVISE STD { int madvise(void *addr, size_t len, \ int behav); } 76 AUE_NULL OBSOL vhangup 77 AUE_NULL OBSOL vlimit 78 AUE_MINCORE STD { int mincore(const void *addr, size_t len, \ char *vec); } 79 AUE_GETGROUPS STD { int getgroups(u_int gidsetsize, \ gid_t *gidset); } 80 AUE_SETGROUPS STD { int setgroups(u_int gidsetsize, \ gid_t *gidset); } 81 AUE_GETPGRP STD { int getpgrp(void); } 82 AUE_SETPGRP STD { int setpgid(int pid, int pgid); } 83 AUE_SETITIMER STD { int setitimer(u_int which, struct \ itimerval *itv, struct itimerval *oitv); } 84 AUE_WAIT4 COMPAT { int wait(void); } 85 AUE_SWAPON STD { int swapon(char *name); } 86 AUE_GETITIMER STD { int getitimer(u_int which, \ struct itimerval *itv); } 87 AUE_SYSCTL COMPAT { int gethostname(char *hostname, \ u_int len); } gethostname \ gethostname_args int 88 AUE_SYSCTL COMPAT { int sethostname(char *hostname, \ u_int len); } sethostname \ sethostname_args int 89 AUE_GETDTABLESIZE STD { int getdtablesize(void); } 90 AUE_DUP2 STD { int dup2(u_int from, u_int to); } 91 AUE_NULL UNIMPL getdopt 92 AUE_FCNTL STD { int fcntl(int fd, int cmd, long arg); } ; XXX should be { int fcntl(int fd, int cmd, ...); } ; but we're not ready for varargs. 93 AUE_SELECT STD { int select(int nd, fd_set *in, fd_set *ou, \ fd_set *ex, struct timeval *tv); } 94 AUE_NULL UNIMPL setdopt 95 AUE_FSYNC STD { int fsync(int fd); } 96 AUE_SETPRIORITY STD { int setpriority(int which, int who, \ int prio); } 97 AUE_SOCKET STD { int socket(int domain, int type, \ int protocol); } 98 AUE_CONNECT STD { int connect(int s, caddr_t name, \ int namelen); } 99 AUE_ACCEPT COMPAT|NOARGS { int accept(int s, caddr_t name, \ int *anamelen); } accept accept_args int 100 AUE_GETPRIORITY STD { int getpriority(int which, int who); } 101 AUE_SEND COMPAT { int send(int s, caddr_t buf, int len, \ int flags); } 102 AUE_RECV COMPAT { int recv(int s, caddr_t buf, int len, \ int flags); } 103 AUE_SIGRETURN COMPAT { int sigreturn( \ struct osigcontext *sigcntxp); } 104 AUE_BIND STD { int bind(int s, caddr_t name, \ int namelen); } 105 AUE_SETSOCKOPT STD { int setsockopt(int s, int level, int name, \ caddr_t val, int valsize); } 106 AUE_LISTEN STD { int listen(int s, int backlog); } 107 AUE_NULL OBSOL vtimes 108 AUE_NULL COMPAT { int sigvec(int signum, struct sigvec *nsv, \ struct sigvec *osv); } 109 AUE_NULL COMPAT { int sigblock(int mask); } 110 AUE_NULL COMPAT { int sigsetmask(int mask); } 111 AUE_NULL COMPAT { int sigsuspend(osigset_t mask); } ; XXX note nonstandard (bogus) calling convention - the libc stub passes ; us the mask, not a pointer to it. 112 AUE_NULL COMPAT { int sigstack(struct sigstack *nss, \ struct sigstack *oss); } 113 AUE_RECVMSG COMPAT { int recvmsg(int s, struct omsghdr *msg, \ int flags); } 114 AUE_SENDMSG COMPAT { int sendmsg(int s, caddr_t msg, \ int flags); } 115 AUE_NULL OBSOL vtrace 116 AUE_GETTIMEOFDAY STD { int gettimeofday(struct timeval *tp, \ struct timezone *tzp); } 117 AUE_GETRUSAGE STD { int getrusage(int who, \ struct rusage *rusage); } 118 AUE_GETSOCKOPT STD { int getsockopt(int s, int level, int name, \ caddr_t val, int *avalsize); } 119 AUE_NULL UNIMPL resuba (BSD/OS 2.x) 120 AUE_READV STD { int readv(int fd, struct iovec *iovp, \ u_int iovcnt); } 121 AUE_WRITEV STD { int writev(int fd, struct iovec *iovp, \ u_int iovcnt); } 122 AUE_SETTIMEOFDAY STD { int settimeofday(struct timeval *tv, \ struct timezone *tzp); } 123 AUE_FCHOWN STD { int fchown(int fd, int uid, int gid); } 124 AUE_FCHMOD STD { int fchmod(int fd, int mode); } 125 AUE_RECVFROM COMPAT|NOARGS { int recvfrom(int s, caddr_t buf, \ size_t len, int flags, caddr_t from, int \ *fromlenaddr); } recvfrom recvfrom_args \ int 126 AUE_SETREUID STD { int setreuid(int ruid, int euid); } 127 AUE_SETREGID STD { int setregid(int rgid, int egid); } 128 AUE_RENAME STD { int rename(char *from, char *to); } 129 AUE_TRUNCATE COMPAT { int truncate(char *path, long length); } 130 AUE_FTRUNCATE COMPAT { int ftruncate(int fd, long length); } 131 AUE_FLOCK STD { int flock(int fd, int how); } 132 AUE_MKFIFO STD { int mkfifo(char *path, int mode); } 133 AUE_SENDTO STD { int sendto(int s, caddr_t buf, size_t len, \ int flags, caddr_t to, int tolen); } 134 AUE_SHUTDOWN STD { int shutdown(int s, int how); } 135 AUE_SOCKETPAIR STD { int socketpair(int domain, int type, \ int protocol, int *rsv); } 136 AUE_MKDIR STD { int mkdir(char *path, int mode); } 137 AUE_RMDIR STD { int rmdir(char *path); } 138 AUE_UTIMES STD { int utimes(char *path, \ struct timeval *tptr); } 139 AUE_NULL OBSOL 4.2 sigreturn 140 AUE_ADJTIME STD { int adjtime(struct timeval *delta, \ struct timeval *olddelta); } 141 AUE_GETPEERNAME COMPAT { int getpeername(int fdes, caddr_t asa, \ int *alen); } 142 AUE_SYSCTL COMPAT { long gethostid(void); } 143 AUE_SYSCTL COMPAT { int sethostid(long hostid); } 144 AUE_GETRLIMIT COMPAT { int getrlimit(u_int which, struct \ orlimit *rlp); } 145 AUE_SETRLIMIT COMPAT { int setrlimit(u_int which, \ struct orlimit *rlp); } 146 AUE_KILLPG COMPAT { int killpg(int pgid, int signum); } 147 AUE_SETSID STD { int setsid(void); } 148 AUE_QUOTACTL STD { int quotactl(char *path, int cmd, int uid, \ caddr_t arg); } 149 AUE_O_QUOTA COMPAT { int quota(void); } 150 AUE_GETSOCKNAME COMPAT|NOARGS { int getsockname(int fdec, \ caddr_t asa, int *alen); } getsockname \ getsockname_args int ; Syscalls 151-180 inclusive are reserved for vendor-specific ; system calls. (This includes various calls added for compatibity ; with other Unix variants.) ; Some of these calls are now supported by BSD... 151 AUE_NULL UNIMPL sem_lock (BSD/OS 2.x) 152 AUE_NULL UNIMPL sem_wakeup (BSD/OS 2.x) 153 AUE_NULL UNIMPL asyncdaemon (BSD/OS 2.x) ; 154 is initialised by the NLM code, if present. 154 AUE_NULL NOSTD { int nlm_syscall(int debug_level, int grace_period, int addr_count, char **addrs); } ; 155 is initialized by the NFS code, if present. 155 AUE_NFS_SVC NOSTD { int nfssvc(int flag, caddr_t argp); } 156 AUE_GETDIRENTRIES COMPAT { int getdirentries(int fd, char *buf, \ u_int count, long *basep); } 157 AUE_STATFS COMPAT4 { int statfs(char *path, \ struct ostatfs *buf); } 158 AUE_FSTATFS COMPAT4 { int fstatfs(int fd, \ struct ostatfs *buf); } 159 AUE_NULL UNIMPL nosys 160 AUE_LGETFH STD { int lgetfh(char *fname, \ struct fhandle *fhp); } 161 AUE_NFS_GETFH STD { int getfh(char *fname, \ struct fhandle *fhp); } 162 AUE_SYSCTL COMPAT4 { int getdomainname(char *domainname, \ int len); } 163 AUE_SYSCTL COMPAT4 { int setdomainname(char *domainname, \ int len); } 164 AUE_NULL COMPAT4 { int uname(struct utsname *name); } 165 AUE_SYSARCH STD { int sysarch(int op, char *parms); } 166 AUE_RTPRIO STD { int rtprio(int function, pid_t pid, \ struct rtprio *rtp); } 167 AUE_NULL UNIMPL nosys 168 AUE_NULL UNIMPL nosys 169 AUE_SEMSYS NOSTD { int semsys(int which, int a2, int a3, \ int a4, int a5); } ; XXX should be { int semsys(int which, ...); } 170 AUE_MSGSYS NOSTD { int msgsys(int which, int a2, int a3, \ int a4, int a5, int a6); } ; XXX should be { int msgsys(int which, ...); } 171 AUE_SHMSYS NOSTD { int shmsys(int which, int a2, int a3, \ int a4); } ; XXX should be { int shmsys(int which, ...); } 172 AUE_NULL UNIMPL nosys 173 AUE_PREAD COMPAT6 { ssize_t pread(int fd, void *buf, \ size_t nbyte, int pad, off_t offset); } 174 AUE_PWRITE COMPAT6 { ssize_t pwrite(int fd, \ const void *buf, \ size_t nbyte, int pad, off_t offset); } 175 AUE_NULL STD { int setfib(int fibnum); } 176 AUE_NTP_ADJTIME STD { int ntp_adjtime(struct timex *tp); } 177 AUE_NULL UNIMPL sfork (BSD/OS 2.x) 178 AUE_NULL UNIMPL getdescriptor (BSD/OS 2.x) 179 AUE_NULL UNIMPL setdescriptor (BSD/OS 2.x) 180 AUE_NULL UNIMPL nosys ; Syscalls 181-199 are used by/reserved for BSD 181 AUE_SETGID STD { int setgid(gid_t gid); } 182 AUE_SETEGID STD { int setegid(gid_t egid); } 183 AUE_SETEUID STD { int seteuid(uid_t euid); } 184 AUE_NULL UNIMPL lfs_bmapv 185 AUE_NULL UNIMPL lfs_markv 186 AUE_NULL UNIMPL lfs_segclean 187 AUE_NULL UNIMPL lfs_segwait 188 AUE_STAT STD { int stat(char *path, struct stat *ub); } 189 AUE_FSTAT STD { int fstat(int fd, struct stat *sb); } 190 AUE_LSTAT STD { int lstat(char *path, struct stat *ub); } 191 AUE_PATHCONF STD { int pathconf(char *path, int name); } 192 AUE_FPATHCONF STD { int fpathconf(int fd, int name); } 193 AUE_NULL UNIMPL nosys 194 AUE_GETRLIMIT STD { int getrlimit(u_int which, \ struct rlimit *rlp); } getrlimit \ __getrlimit_args int 195 AUE_SETRLIMIT STD { int setrlimit(u_int which, \ struct rlimit *rlp); } setrlimit \ __setrlimit_args int 196 AUE_GETDIRENTRIES STD { int getdirentries(int fd, char *buf, \ u_int count, long *basep); } 197 AUE_MMAP COMPAT6 { caddr_t mmap(caddr_t addr, \ size_t len, int prot, int flags, int fd, \ int pad, off_t pos); } 198 AUE_NULL NOPROTO { int nosys(void); } __syscall \ __syscall_args int 199 AUE_LSEEK COMPAT6 { off_t lseek(int fd, int pad, \ off_t offset, int whence); } 200 AUE_TRUNCATE COMPAT6 { int truncate(char *path, int pad, \ off_t length); } 201 AUE_FTRUNCATE COMPAT6 { int ftruncate(int fd, int pad, \ off_t length); } 202 AUE_SYSCTL STD { int __sysctl(int *name, u_int namelen, \ void *old, size_t *oldlenp, void *new, \ size_t newlen); } __sysctl sysctl_args int 203 AUE_MLOCK STD { int mlock(const void *addr, size_t len); } 204 AUE_MUNLOCK STD { int munlock(const void *addr, size_t len); } 205 AUE_UNDELETE STD { int undelete(char *path); } 206 AUE_FUTIMES STD { int futimes(int fd, struct timeval *tptr); } 207 AUE_GETPGID STD { int getpgid(pid_t pid); } 208 AUE_NULL UNIMPL newreboot (NetBSD) 209 AUE_POLL STD { int poll(struct pollfd *fds, u_int nfds, \ int timeout); } ; ; The following are reserved for loadable syscalls ; 210 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 211 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 212 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 213 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 214 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 215 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 216 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 217 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 218 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int 219 AUE_NULL NODEF|NOTSTATIC lkmnosys lkmnosys nosys_args int ; ; The following were introduced with NetBSD/4.4Lite-2 220 AUE_SEMCTL COMPAT7|NOSTD { int __semctl(int semid, int semnum, \ int cmd, union semun_old *arg); } 221 AUE_SEMGET NOSTD { int semget(key_t key, int nsems, \ int semflg); } 222 AUE_SEMOP NOSTD { int semop(int semid, struct sembuf *sops, \ size_t nsops); } 223 AUE_NULL UNIMPL semconfig 224 AUE_MSGCTL COMPAT7|NOSTD { int msgctl(int msqid, int cmd, \ struct msqid_ds_old *buf); } 225 AUE_MSGGET NOSTD { int msgget(key_t key, int msgflg); } 226 AUE_MSGSND NOSTD { int msgsnd(int msqid, const void *msgp, \ size_t msgsz, int msgflg); } 227 AUE_MSGRCV NOSTD { int msgrcv(int msqid, void *msgp, \ size_t msgsz, long msgtyp, int msgflg); } 228 AUE_SHMAT NOSTD { int shmat(int shmid, const void *shmaddr, \ int shmflg); } 229 AUE_SHMCTL COMPAT7|NOSTD { int shmctl(int shmid, int cmd, \ struct shmid_ds_old *buf); } 230 AUE_SHMDT NOSTD { int shmdt(const void *shmaddr); } 231 AUE_SHMGET NOSTD { int shmget(key_t key, size_t size, \ int shmflg); } ; 232 AUE_NULL STD { int clock_gettime(clockid_t clock_id, \ struct timespec *tp); } 233 AUE_CLOCK_SETTIME STD { int clock_settime( \ clockid_t clock_id, \ const struct timespec *tp); } 234 AUE_NULL STD { int clock_getres(clockid_t clock_id, \ struct timespec *tp); } 235 AUE_NULL STD { int ktimer_create(clockid_t clock_id, \ struct sigevent *evp, int *timerid); } 236 AUE_NULL STD { int ktimer_delete(int timerid); } 237 AUE_NULL STD { int ktimer_settime(int timerid, int flags, \ const struct itimerspec *value, \ struct itimerspec *ovalue); } 238 AUE_NULL STD { int ktimer_gettime(int timerid, struct \ itimerspec *value); } 239 AUE_NULL STD { int ktimer_getoverrun(int timerid); } 240 AUE_NULL STD { int nanosleep(const struct timespec *rqtp, \ struct timespec *rmtp); } 241 AUE_NULL STD { int ffclock_getcounter(ffcounter *ffcount); } 242 AUE_NULL STD { int ffclock_setestimate( \ struct ffclock_estimate *cest); } 243 AUE_NULL STD { int ffclock_getestimate( \ struct ffclock_estimate *cest); } 244 AUE_NULL UNIMPL nosys 245 AUE_NULL UNIMPL nosys 246 AUE_NULL UNIMPL nosys 247 AUE_NULL STD { int clock_getcpuclockid2(id_t id,\ int which, clockid_t *clock_id); } 248 AUE_NULL STD { int ntp_gettime(struct ntptimeval *ntvp); } 249 AUE_NULL UNIMPL nosys ; syscall numbers initially used in OpenBSD 250 AUE_MINHERIT STD { int minherit(void *addr, size_t len, \ int inherit); } 251 AUE_RFORK STD { int rfork(int flags); } 252 AUE_POLL STD { int openbsd_poll(struct pollfd *fds, \ u_int nfds, int timeout); } 253 AUE_ISSETUGID STD { int issetugid(void); } 254 AUE_LCHOWN STD { int lchown(char *path, int uid, int gid); } 255 AUE_NULL STD { int aio_read(struct aiocb *aiocbp); } 256 AUE_NULL STD { int aio_write(struct aiocb *aiocbp); } 257 AUE_NULL STD { int lio_listio(int mode, \ struct aiocb * const *acb_list, \ int nent, struct sigevent *sig); } 258 AUE_NULL UNIMPL nosys 259 AUE_NULL UNIMPL nosys 260 AUE_NULL UNIMPL nosys 261 AUE_NULL UNIMPL nosys 262 AUE_NULL UNIMPL nosys 263 AUE_NULL UNIMPL nosys 264 AUE_NULL UNIMPL nosys 265 AUE_NULL UNIMPL nosys 266 AUE_NULL UNIMPL nosys 267 AUE_NULL UNIMPL nosys 268 AUE_NULL UNIMPL nosys 269 AUE_NULL UNIMPL nosys 270 AUE_NULL UNIMPL nosys 271 AUE_NULL UNIMPL nosys 272 AUE_O_GETDENTS STD { int getdents(int fd, char *buf, \ size_t count); } 273 AUE_NULL UNIMPL nosys 274 AUE_LCHMOD STD { int lchmod(char *path, mode_t mode); } 275 AUE_LCHOWN NOPROTO { int lchown(char *path, uid_t uid, \ gid_t gid); } netbsd_lchown lchown_args \ int 276 AUE_LUTIMES STD { int lutimes(char *path, \ struct timeval *tptr); } 277 AUE_MSYNC NOPROTO { int msync(void *addr, size_t len, \ int flags); } netbsd_msync msync_args int 278 AUE_STAT STD { int nstat(char *path, struct nstat *ub); } 279 AUE_FSTAT STD { int nfstat(int fd, struct nstat *sb); } 280 AUE_LSTAT STD { int nlstat(char *path, struct nstat *ub); } 281 AUE_NULL UNIMPL nosys 282 AUE_NULL UNIMPL nosys 283 AUE_NULL UNIMPL nosys 284 AUE_NULL UNIMPL nosys 285 AUE_NULL UNIMPL nosys 286 AUE_NULL UNIMPL nosys 287 AUE_NULL UNIMPL nosys 288 AUE_NULL UNIMPL nosys ; 289 and 290 from NetBSD (OpenBSD: 267 and 268) 289 AUE_PREADV STD { ssize_t preadv(int fd, struct iovec *iovp, \ u_int iovcnt, off_t offset); } 290 AUE_PWRITEV STD { ssize_t pwritev(int fd, struct iovec *iovp, \ u_int iovcnt, off_t offset); } 291 AUE_NULL UNIMPL nosys 292 AUE_NULL UNIMPL nosys 293 AUE_NULL UNIMPL nosys 294 AUE_NULL UNIMPL nosys 295 AUE_NULL UNIMPL nosys 296 AUE_NULL UNIMPL nosys ; XXX 297 is 300 in NetBSD 297 AUE_FHSTATFS COMPAT4 { int fhstatfs( \ const struct fhandle *u_fhp, \ struct ostatfs *buf); } 298 AUE_FHOPEN STD { int fhopen(const struct fhandle *u_fhp, \ int flags); } 299 AUE_FHSTAT STD { int fhstat(const struct fhandle *u_fhp, \ struct stat *sb); } ; syscall numbers for FreeBSD 300 AUE_NULL STD { int modnext(int modid); } 301 AUE_NULL STD { int modstat(int modid, \ struct module_stat *stat); } 302 AUE_NULL STD { int modfnext(int modid); } 303 AUE_NULL STD { int modfind(const char *name); } 304 AUE_MODLOAD STD { int kldload(const char *file); } 305 AUE_MODUNLOAD STD { int kldunload(int fileid); } 306 AUE_NULL STD { int kldfind(const char *file); } 307 AUE_NULL STD { int kldnext(int fileid); } 308 AUE_NULL STD { int kldstat(int fileid, struct \ kld_file_stat* stat); } 309 AUE_NULL STD { int kldfirstmod(int fileid); } 310 AUE_GETSID STD { int getsid(pid_t pid); } 311 AUE_SETRESUID STD { int setresuid(uid_t ruid, uid_t euid, \ uid_t suid); } 312 AUE_SETRESGID STD { int setresgid(gid_t rgid, gid_t egid, \ gid_t sgid); } 313 AUE_NULL OBSOL signanosleep -314 AUE_NULL STD { int aio_return(struct aiocb *aiocbp); } +314 AUE_NULL STD { ssize_t aio_return(struct aiocb *aiocbp); } 315 AUE_NULL STD { int aio_suspend( \ struct aiocb * const * aiocbp, int nent, \ const struct timespec *timeout); } 316 AUE_NULL STD { int aio_cancel(int fd, \ struct aiocb *aiocbp); } 317 AUE_NULL STD { int aio_error(struct aiocb *aiocbp); } 318 AUE_NULL COMPAT6 { int aio_read(struct oaiocb *aiocbp); } 319 AUE_NULL COMPAT6 { int aio_write(struct oaiocb *aiocbp); } 320 AUE_NULL COMPAT6 { int lio_listio(int mode, \ struct oaiocb * const *acb_list, \ int nent, struct osigevent *sig); } 321 AUE_NULL STD { int yield(void); } 322 AUE_NULL OBSOL thr_sleep 323 AUE_NULL OBSOL thr_wakeup 324 AUE_MLOCKALL STD { int mlockall(int how); } 325 AUE_MUNLOCKALL STD { int munlockall(void); } 326 AUE_GETCWD STD { int __getcwd(char *buf, u_int buflen); } 327 AUE_NULL STD { int sched_setparam (pid_t pid, \ const struct sched_param *param); } 328 AUE_NULL STD { int sched_getparam (pid_t pid, struct \ sched_param *param); } 329 AUE_NULL STD { int sched_setscheduler (pid_t pid, int \ policy, const struct sched_param \ *param); } 330 AUE_NULL STD { int sched_getscheduler (pid_t pid); } 331 AUE_NULL STD { int sched_yield (void); } 332 AUE_NULL STD { int sched_get_priority_max (int policy); } 333 AUE_NULL STD { int sched_get_priority_min (int policy); } 334 AUE_NULL STD { int sched_rr_get_interval (pid_t pid, \ struct timespec *interval); } 335 AUE_NULL STD { int utrace(const void *addr, size_t len); } 336 AUE_SENDFILE COMPAT4 { int sendfile(int fd, int s, \ off_t offset, size_t nbytes, \ struct sf_hdtr *hdtr, off_t *sbytes, \ int flags); } 337 AUE_NULL STD { int kldsym(int fileid, int cmd, \ void *data); } 338 AUE_JAIL STD { int jail(struct jail *jail); } 339 AUE_NULL NOSTD|NOTSTATIC { int nnpfs_syscall(int operation, \ char *a_pathP, int a_opcode, \ void *a_paramsP, int a_followSymlinks); } 340 AUE_SIGPROCMASK STD { int sigprocmask(int how, \ const sigset_t *set, sigset_t *oset); } 341 AUE_SIGSUSPEND STD { int sigsuspend(const sigset_t *sigmask); } 342 AUE_SIGACTION COMPAT4 { int sigaction(int sig, const \ struct sigaction *act, \ struct sigaction *oact); } 343 AUE_SIGPENDING STD { int sigpending(sigset_t *set); } 344 AUE_SIGRETURN COMPAT4 { int sigreturn( \ const struct ucontext4 *sigcntxp); } 345 AUE_SIGWAIT STD { int sigtimedwait(const sigset_t *set, \ siginfo_t *info, \ const struct timespec *timeout); } 346 AUE_NULL STD { int sigwaitinfo(const sigset_t *set, \ siginfo_t *info); } 347 AUE_NULL STD { int __acl_get_file(const char *path, \ acl_type_t type, struct acl *aclp); } 348 AUE_NULL STD { int __acl_set_file(const char *path, \ acl_type_t type, struct acl *aclp); } 349 AUE_NULL STD { int __acl_get_fd(int filedes, \ acl_type_t type, struct acl *aclp); } 350 AUE_NULL STD { int __acl_set_fd(int filedes, \ acl_type_t type, struct acl *aclp); } 351 AUE_NULL STD { int __acl_delete_file(const char *path, \ acl_type_t type); } 352 AUE_NULL STD { int __acl_delete_fd(int filedes, \ acl_type_t type); } 353 AUE_NULL STD { int __acl_aclcheck_file(const char *path, \ acl_type_t type, struct acl *aclp); } 354 AUE_NULL STD { int __acl_aclcheck_fd(int filedes, \ acl_type_t type, struct acl *aclp); } 355 AUE_EXTATTRCTL STD { int extattrctl(const char *path, int cmd, \ const char *filename, int attrnamespace, \ const char *attrname); } 356 AUE_EXTATTR_SET_FILE STD { ssize_t extattr_set_file( \ const char *path, int attrnamespace, \ const char *attrname, void *data, \ size_t nbytes); } 357 AUE_EXTATTR_GET_FILE STD { ssize_t extattr_get_file( \ const char *path, int attrnamespace, \ const char *attrname, void *data, \ size_t nbytes); } 358 AUE_EXTATTR_DELETE_FILE STD { int extattr_delete_file(const char *path, \ int attrnamespace, \ const char *attrname); } -359 AUE_NULL STD { int aio_waitcomplete( \ +359 AUE_NULL STD { ssize_t aio_waitcomplete( \ struct aiocb **aiocbp, \ struct timespec *timeout); } 360 AUE_GETRESUID STD { int getresuid(uid_t *ruid, uid_t *euid, \ uid_t *suid); } 361 AUE_GETRESGID STD { int getresgid(gid_t *rgid, gid_t *egid, \ gid_t *sgid); } 362 AUE_KQUEUE STD { int kqueue(void); } 363 AUE_NULL STD { int kevent(int fd, \ struct kevent *changelist, int nchanges, \ struct kevent *eventlist, int nevents, \ const struct timespec *timeout); } 364 AUE_NULL UNIMPL __cap_get_proc 365 AUE_NULL UNIMPL __cap_set_proc 366 AUE_NULL UNIMPL __cap_get_fd 367 AUE_NULL UNIMPL __cap_get_file 368 AUE_NULL UNIMPL __cap_set_fd 369 AUE_NULL UNIMPL __cap_set_file 370 AUE_NULL UNIMPL nosys 371 AUE_EXTATTR_SET_FD STD { ssize_t extattr_set_fd(int fd, \ int attrnamespace, const char *attrname, \ void *data, size_t nbytes); } 372 AUE_EXTATTR_GET_FD STD { ssize_t extattr_get_fd(int fd, \ int attrnamespace, const char *attrname, \ void *data, size_t nbytes); } 373 AUE_EXTATTR_DELETE_FD STD { int extattr_delete_fd(int fd, \ int attrnamespace, \ const char *attrname); } 374 AUE_NULL STD { int __setugid(int flag); } 375 AUE_NULL UNIMPL nfsclnt 376 AUE_EACCESS STD { int eaccess(char *path, int amode); } 377 AUE_NULL NOSTD|NOTSTATIC { int afs3_syscall(long syscall, \ long parm1, long parm2, long parm3, \ long parm4, long parm5, long parm6); } 378 AUE_NMOUNT STD { int nmount(struct iovec *iovp, \ unsigned int iovcnt, int flags); } 379 AUE_NULL UNIMPL kse_exit 380 AUE_NULL UNIMPL kse_wakeup 381 AUE_NULL UNIMPL kse_create 382 AUE_NULL UNIMPL kse_thr_interrupt 383 AUE_NULL UNIMPL kse_release 384 AUE_NULL STD { int __mac_get_proc(struct mac *mac_p); } 385 AUE_NULL STD { int __mac_set_proc(struct mac *mac_p); } 386 AUE_NULL STD { int __mac_get_fd(int fd, \ struct mac *mac_p); } 387 AUE_NULL STD { int __mac_get_file(const char *path_p, \ struct mac *mac_p); } 388 AUE_NULL STD { int __mac_set_fd(int fd, \ struct mac *mac_p); } 389 AUE_NULL STD { int __mac_set_file(const char *path_p, \ struct mac *mac_p); } 390 AUE_NULL STD { int kenv(int what, const char *name, \ char *value, int len); } 391 AUE_LCHFLAGS STD { int lchflags(const char *path, \ u_long flags); } 392 AUE_NULL STD { int uuidgen(struct uuid *store, \ int count); } 393 AUE_SENDFILE STD { int sendfile(int fd, int s, off_t offset, \ size_t nbytes, struct sf_hdtr *hdtr, \ off_t *sbytes, int flags); } 394 AUE_NULL STD { int mac_syscall(const char *policy, \ int call, void *arg); } 395 AUE_GETFSSTAT STD { int getfsstat(struct statfs *buf, \ long bufsize, int flags); } 396 AUE_STATFS STD { int statfs(char *path, \ struct statfs *buf); } 397 AUE_FSTATFS STD { int fstatfs(int fd, struct statfs *buf); } 398 AUE_FHSTATFS STD { int fhstatfs(const struct fhandle *u_fhp, \ struct statfs *buf); } 399 AUE_NULL UNIMPL nosys 400 AUE_NULL NOSTD { int ksem_close(semid_t id); } 401 AUE_NULL NOSTD { int ksem_post(semid_t id); } 402 AUE_NULL NOSTD { int ksem_wait(semid_t id); } 403 AUE_NULL NOSTD { int ksem_trywait(semid_t id); } 404 AUE_NULL NOSTD { int ksem_init(semid_t *idp, \ unsigned int value); } 405 AUE_NULL NOSTD { int ksem_open(semid_t *idp, \ const char *name, int oflag, \ mode_t mode, unsigned int value); } 406 AUE_NULL NOSTD { int ksem_unlink(const char *name); } 407 AUE_NULL NOSTD { int ksem_getvalue(semid_t id, int *val); } 408 AUE_NULL NOSTD { int ksem_destroy(semid_t id); } 409 AUE_NULL STD { int __mac_get_pid(pid_t pid, \ struct mac *mac_p); } 410 AUE_NULL STD { int __mac_get_link(const char *path_p, \ struct mac *mac_p); } 411 AUE_NULL STD { int __mac_set_link(const char *path_p, \ struct mac *mac_p); } 412 AUE_EXTATTR_SET_LINK STD { ssize_t extattr_set_link( \ const char *path, int attrnamespace, \ const char *attrname, void *data, \ size_t nbytes); } 413 AUE_EXTATTR_GET_LINK STD { ssize_t extattr_get_link( \ const char *path, int attrnamespace, \ const char *attrname, void *data, \ size_t nbytes); } 414 AUE_EXTATTR_DELETE_LINK STD { int extattr_delete_link( \ const char *path, int attrnamespace, \ const char *attrname); } 415 AUE_NULL STD { int __mac_execve(char *fname, char **argv, \ char **envv, struct mac *mac_p); } 416 AUE_SIGACTION STD { int sigaction(int sig, \ const struct sigaction *act, \ struct sigaction *oact); } 417 AUE_SIGRETURN STD { int sigreturn( \ const struct __ucontext *sigcntxp); } 418 AUE_NULL UNIMPL __xstat 419 AUE_NULL UNIMPL __xfstat 420 AUE_NULL UNIMPL __xlstat 421 AUE_NULL STD { int getcontext(struct __ucontext *ucp); } 422 AUE_NULL STD { int setcontext( \ const struct __ucontext *ucp); } 423 AUE_NULL STD { int swapcontext(struct __ucontext *oucp, \ const struct __ucontext *ucp); } 424 AUE_SWAPOFF STD { int swapoff(const char *name); } 425 AUE_NULL STD { int __acl_get_link(const char *path, \ acl_type_t type, struct acl *aclp); } 426 AUE_NULL STD { int __acl_set_link(const char *path, \ acl_type_t type, struct acl *aclp); } 427 AUE_NULL STD { int __acl_delete_link(const char *path, \ acl_type_t type); } 428 AUE_NULL STD { int __acl_aclcheck_link(const char *path, \ acl_type_t type, struct acl *aclp); } 429 AUE_SIGWAIT STD { int sigwait(const sigset_t *set, \ int *sig); } 430 AUE_NULL STD { int thr_create(ucontext_t *ctx, long *id, \ int flags); } 431 AUE_NULL STD { void thr_exit(long *state); } 432 AUE_NULL STD { int thr_self(long *id); } 433 AUE_NULL STD { int thr_kill(long id, int sig); } 434 AUE_NULL UNIMPL nosys 435 AUE_NULL UNIMPL nosys 436 AUE_NULL STD { int jail_attach(int jid); } 437 AUE_EXTATTR_LIST_FD STD { ssize_t extattr_list_fd(int fd, \ int attrnamespace, void *data, \ size_t nbytes); } 438 AUE_EXTATTR_LIST_FILE STD { ssize_t extattr_list_file( \ const char *path, int attrnamespace, \ void *data, size_t nbytes); } 439 AUE_EXTATTR_LIST_LINK STD { ssize_t extattr_list_link( \ const char *path, int attrnamespace, \ void *data, size_t nbytes); } 440 AUE_NULL UNIMPL kse_switchin 441 AUE_NULL NOSTD { int ksem_timedwait(semid_t id, \ const struct timespec *abstime); } 442 AUE_NULL STD { int thr_suspend( \ const struct timespec *timeout); } 443 AUE_NULL STD { int thr_wake(long id); } 444 AUE_MODUNLOAD STD { int kldunloadf(int fileid, int flags); } 445 AUE_AUDIT STD { int audit(const void *record, \ u_int length); } 446 AUE_AUDITON STD { int auditon(int cmd, void *data, \ u_int length); } 447 AUE_GETAUID STD { int getauid(uid_t *auid); } 448 AUE_SETAUID STD { int setauid(uid_t *auid); } 449 AUE_GETAUDIT STD { int getaudit(struct auditinfo *auditinfo); } 450 AUE_SETAUDIT STD { int setaudit(struct auditinfo *auditinfo); } 451 AUE_GETAUDIT_ADDR STD { int getaudit_addr( \ struct auditinfo_addr *auditinfo_addr, \ u_int length); } 452 AUE_SETAUDIT_ADDR STD { int setaudit_addr( \ struct auditinfo_addr *auditinfo_addr, \ u_int length); } 453 AUE_AUDITCTL STD { int auditctl(char *path); } 454 AUE_NULL STD { int _umtx_op(void *obj, int op, \ u_long val, void *uaddr1, void *uaddr2); } 455 AUE_NULL STD { int thr_new(struct thr_param *param, \ int param_size); } 456 AUE_NULL STD { int sigqueue(pid_t pid, int signum, void *value); } 457 AUE_NULL NOSTD { int kmq_open(const char *path, int flags, \ mode_t mode, const struct mq_attr *attr); } 458 AUE_NULL NOSTD { int kmq_setattr(int mqd, \ const struct mq_attr *attr, \ struct mq_attr *oattr); } 459 AUE_NULL NOSTD { int kmq_timedreceive(int mqd, \ char *msg_ptr, size_t msg_len, \ unsigned *msg_prio, \ const struct timespec *abs_timeout); } 460 AUE_NULL NOSTD { int kmq_timedsend(int mqd, \ const char *msg_ptr, size_t msg_len,\ unsigned msg_prio, \ const struct timespec *abs_timeout);} 461 AUE_NULL NOSTD { int kmq_notify(int mqd, \ const struct sigevent *sigev); } 462 AUE_NULL NOSTD { int kmq_unlink(const char *path); } 463 AUE_NULL STD { int abort2(const char *why, int nargs, void **args); } 464 AUE_NULL STD { int thr_set_name(long id, const char *name); } 465 AUE_NULL STD { int aio_fsync(int op, struct aiocb *aiocbp); } 466 AUE_RTPRIO STD { int rtprio_thread(int function, \ lwpid_t lwpid, struct rtprio *rtp); } 467 AUE_NULL UNIMPL nosys 468 AUE_NULL UNIMPL nosys 469 AUE_NULL UNIMPL __getpath_fromfd 470 AUE_NULL UNIMPL __getpath_fromaddr 471 AUE_NULL NOSTD { int sctp_peeloff(int sd, uint32_t name); } 472 AUE_NULL NOSTD { int sctp_generic_sendmsg(int sd, caddr_t msg, int mlen, \ caddr_t to, __socklen_t tolen, \ struct sctp_sndrcvinfo *sinfo, int flags); } 473 AUE_NULL NOSTD { int sctp_generic_sendmsg_iov(int sd, struct iovec *iov, int iovlen, \ caddr_t to, __socklen_t tolen, \ struct sctp_sndrcvinfo *sinfo, int flags); } 474 AUE_NULL NOSTD { int sctp_generic_recvmsg(int sd, struct iovec *iov, int iovlen, \ struct sockaddr * from, __socklen_t *fromlenaddr, \ struct sctp_sndrcvinfo *sinfo, int *msg_flags); } 475 AUE_PREAD STD { ssize_t pread(int fd, void *buf, \ size_t nbyte, off_t offset); } 476 AUE_PWRITE STD { ssize_t pwrite(int fd, const void *buf, \ size_t nbyte, off_t offset); } 477 AUE_MMAP STD { caddr_t mmap(caddr_t addr, size_t len, \ int prot, int flags, int fd, off_t pos); } 478 AUE_LSEEK STD { off_t lseek(int fd, off_t offset, \ int whence); } 479 AUE_TRUNCATE STD { int truncate(char *path, off_t length); } 480 AUE_FTRUNCATE STD { int ftruncate(int fd, off_t length); } 481 AUE_KILL STD { int thr_kill2(pid_t pid, long id, int sig); } 482 AUE_SHMOPEN STD { int shm_open(const char *path, int flags, \ mode_t mode); } 483 AUE_SHMUNLINK STD { int shm_unlink(const char *path); } 484 AUE_NULL STD { int cpuset(cpusetid_t *setid); } 485 AUE_NULL STD { int cpuset_setid(cpuwhich_t which, id_t id, \ cpusetid_t setid); } 486 AUE_NULL STD { int cpuset_getid(cpulevel_t level, \ cpuwhich_t which, id_t id, \ cpusetid_t *setid); } 487 AUE_NULL STD { int cpuset_getaffinity(cpulevel_t level, \ cpuwhich_t which, id_t id, size_t cpusetsize, \ cpuset_t *mask); } 488 AUE_NULL STD { int cpuset_setaffinity(cpulevel_t level, \ cpuwhich_t which, id_t id, size_t cpusetsize, \ const cpuset_t *mask); } 489 AUE_FACCESSAT STD { int faccessat(int fd, char *path, int amode, \ int flag); } 490 AUE_FCHMODAT STD { int fchmodat(int fd, char *path, mode_t mode, \ int flag); } 491 AUE_FCHOWNAT STD { int fchownat(int fd, char *path, uid_t uid, \ gid_t gid, int flag); } 492 AUE_FEXECVE STD { int fexecve(int fd, char **argv, \ char **envv); } 493 AUE_FSTATAT STD { int fstatat(int fd, char *path, \ struct stat *buf, int flag); } 494 AUE_FUTIMESAT STD { int futimesat(int fd, char *path, \ struct timeval *times); } 495 AUE_LINKAT STD { int linkat(int fd1, char *path1, int fd2, \ char *path2, int flag); } 496 AUE_MKDIRAT STD { int mkdirat(int fd, char *path, mode_t mode); } 497 AUE_MKFIFOAT STD { int mkfifoat(int fd, char *path, mode_t mode); } 498 AUE_MKNODAT STD { int mknodat(int fd, char *path, mode_t mode, \ dev_t dev); } ; XXX: see the comment for open 499 AUE_OPENAT_RWTC STD { int openat(int fd, char *path, int flag, \ mode_t mode); } 500 AUE_READLINKAT STD { int readlinkat(int fd, char *path, char *buf, \ size_t bufsize); } 501 AUE_RENAMEAT STD { int renameat(int oldfd, char *old, int newfd, \ char *new); } 502 AUE_SYMLINKAT STD { int symlinkat(char *path1, int fd, \ char *path2); } 503 AUE_UNLINKAT STD { int unlinkat(int fd, char *path, int flag); } 504 AUE_POSIX_OPENPT STD { int posix_openpt(int flags); } ; 505 is initialised by the kgssapi code, if present. 505 AUE_NULL NOSTD { int gssd_syscall(char *path); } 506 AUE_NULL STD { int jail_get(struct iovec *iovp, \ unsigned int iovcnt, int flags); } 507 AUE_NULL STD { int jail_set(struct iovec *iovp, \ unsigned int iovcnt, int flags); } 508 AUE_NULL STD { int jail_remove(int jid); } 509 AUE_CLOSEFROM STD { int closefrom(int lowfd); } 510 AUE_SEMCTL NOSTD { int __semctl(int semid, int semnum, \ int cmd, union semun *arg); } 511 AUE_MSGCTL NOSTD { int msgctl(int msqid, int cmd, \ struct msqid_ds *buf); } 512 AUE_SHMCTL NOSTD { int shmctl(int shmid, int cmd, \ struct shmid_ds *buf); } 513 AUE_LPATHCONF STD { int lpathconf(char *path, int name); } 514 AUE_NULL OBSOL cap_new 515 AUE_CAP_RIGHTS_GET STD { int __cap_rights_get(int version, \ int fd, cap_rights_t *rightsp); } 516 AUE_CAP_ENTER STD { int cap_enter(void); } 517 AUE_CAP_GETMODE STD { int cap_getmode(u_int *modep); } 518 AUE_PDFORK STD { int pdfork(int *fdp, int flags); } 519 AUE_PDKILL STD { int pdkill(int fd, int signum); } 520 AUE_PDGETPID STD { int pdgetpid(int fd, pid_t *pidp); } 521 AUE_PDWAIT UNIMPL pdwait4 522 AUE_SELECT STD { int pselect(int nd, fd_set *in, \ fd_set *ou, fd_set *ex, \ const struct timespec *ts, \ const sigset_t *sm); } 523 AUE_NULL STD { int getloginclass(char *namebuf, \ size_t namelen); } 524 AUE_NULL STD { int setloginclass(const char *namebuf); } 525 AUE_NULL STD { int rctl_get_racct(const void *inbufp, \ size_t inbuflen, void *outbufp, \ size_t outbuflen); } 526 AUE_NULL STD { int rctl_get_rules(const void *inbufp, \ size_t inbuflen, void *outbufp, \ size_t outbuflen); } 527 AUE_NULL STD { int rctl_get_limits(const void *inbufp, \ size_t inbuflen, void *outbufp, \ size_t outbuflen); } 528 AUE_NULL STD { int rctl_add_rule(const void *inbufp, \ size_t inbuflen, void *outbufp, \ size_t outbuflen); } 529 AUE_NULL STD { int rctl_remove_rule(const void *inbufp, \ size_t inbuflen, void *outbufp, \ size_t outbuflen); } 530 AUE_NULL STD { int posix_fallocate(int fd, \ off_t offset, off_t len); } 531 AUE_NULL STD { int posix_fadvise(int fd, off_t offset, \ off_t len, int advice); } 532 AUE_WAIT6 STD { int wait6(idtype_t idtype, id_t id, \ int *status, int options, \ struct __wrusage *wrusage, \ siginfo_t *info); } 533 AUE_CAP_RIGHTS_LIMIT STD { int cap_rights_limit(int fd, \ cap_rights_t *rightsp); } 534 AUE_CAP_IOCTLS_LIMIT STD { int cap_ioctls_limit(int fd, \ const u_long *cmds, size_t ncmds); } 535 AUE_CAP_IOCTLS_GET STD { ssize_t cap_ioctls_get(int fd, \ u_long *cmds, size_t maxcmds); } 536 AUE_CAP_FCNTLS_LIMIT STD { int cap_fcntls_limit(int fd, \ uint32_t fcntlrights); } 537 AUE_CAP_FCNTLS_GET STD { int cap_fcntls_get(int fd, \ uint32_t *fcntlrightsp); } 538 AUE_BINDAT STD { int bindat(int fd, int s, caddr_t name, \ int namelen); } 539 AUE_CONNECTAT STD { int connectat(int fd, int s, caddr_t name, \ int namelen); } 540 AUE_CHFLAGSAT STD { int chflagsat(int fd, const char *path, \ u_long flags, int atflag); } 541 AUE_ACCEPT STD { int accept4(int s, \ struct sockaddr * __restrict name, \ __socklen_t * __restrict anamelen, \ int flags); } 542 AUE_PIPE STD { int pipe2(int *fildes, int flags); } 543 AUE_NULL STD { int aio_mlock(struct aiocb *aiocbp); } 544 AUE_NULL STD { int procctl(idtype_t idtype, id_t id, \ int com, void *data); } 545 AUE_POLL STD { int ppoll(struct pollfd *fds, u_int nfds, \ const struct timespec *ts, \ const sigset_t *set); } 546 AUE_FUTIMES STD { int futimens(int fd, \ struct timespec *times); } 547 AUE_FUTIMESAT STD { int utimensat(int fd, \ char *path, \ struct timespec *times, int flag); } 548 AUE_NULL STD { int numa_getaffinity(cpuwhich_t which, \ id_t id, \ struct vm_domain_policy_entry *policy); } 549 AUE_NULL STD { int numa_setaffinity(cpuwhich_t which, \ id_t id, \ const struct vm_domain_policy_entry *policy); } ; Please copy any additions and changes to the following compatability tables: ; sys/compat/freebsd32/syscalls.master diff --git a/sys/kern/vfs_aio.c b/sys/kern/vfs_aio.c index 95e4c53ea0ab..c2c26f8328d4 100644 --- a/sys/kern/vfs_aio.c +++ b/sys/kern/vfs_aio.c @@ -1,2951 +1,2951 @@ /*- * Copyright (c) 1997 John S. Dyson. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. John S. Dyson's name may not be used to endorse or promote products * derived from this software without specific prior written permission. * * DISCLAIMER: This code isn't warranted to do anything useful. Anything * bad that happens because of using this software isn't the responsibility * of the author. This software is distributed AS-IS. */ /* * This file contains support for the POSIX 1003.1B AIO/LIO facility. */ #include __FBSDID("$FreeBSD$"); #include "opt_compat.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Counter for allocating reference ids to new jobs. Wrapped to 1 on * overflow. (XXX will be removed soon.) */ static u_long jobrefid; /* * Counter for aio_fsync. */ static uint64_t jobseqno; #ifndef MAX_AIO_PER_PROC #define MAX_AIO_PER_PROC 32 #endif #ifndef MAX_AIO_QUEUE_PER_PROC #define MAX_AIO_QUEUE_PER_PROC 256 /* Bigger than AIO_LISTIO_MAX */ #endif #ifndef MAX_AIO_QUEUE #define MAX_AIO_QUEUE 1024 /* Bigger than AIO_LISTIO_MAX */ #endif #ifndef MAX_BUF_AIO #define MAX_BUF_AIO 16 #endif FEATURE(aio, "Asynchronous I/O"); static MALLOC_DEFINE(M_LIO, "lio", "listio aio control block list"); static SYSCTL_NODE(_vfs, OID_AUTO, aio, CTLFLAG_RW, 0, "Async IO management"); static int enable_aio_unsafe = 0; SYSCTL_INT(_vfs_aio, OID_AUTO, enable_unsafe, CTLFLAG_RW, &enable_aio_unsafe, 0, "Permit asynchronous IO on all file types, not just known-safe types"); static int max_aio_procs = MAX_AIO_PROCS; SYSCTL_INT(_vfs_aio, OID_AUTO, max_aio_procs, CTLFLAG_RW, &max_aio_procs, 0, "Maximum number of kernel processes to use for handling async IO "); static int num_aio_procs = 0; SYSCTL_INT(_vfs_aio, OID_AUTO, num_aio_procs, CTLFLAG_RD, &num_aio_procs, 0, "Number of presently active kernel processes for async IO"); /* * The code will adjust the actual number of AIO processes towards this * number when it gets a chance. */ static int target_aio_procs = TARGET_AIO_PROCS; SYSCTL_INT(_vfs_aio, OID_AUTO, target_aio_procs, CTLFLAG_RW, &target_aio_procs, 0, "Preferred number of ready kernel processes for async IO"); static int max_queue_count = MAX_AIO_QUEUE; SYSCTL_INT(_vfs_aio, OID_AUTO, max_aio_queue, CTLFLAG_RW, &max_queue_count, 0, "Maximum number of aio requests to queue, globally"); static int num_queue_count = 0; SYSCTL_INT(_vfs_aio, OID_AUTO, num_queue_count, CTLFLAG_RD, &num_queue_count, 0, "Number of queued aio requests"); static int num_buf_aio = 0; SYSCTL_INT(_vfs_aio, OID_AUTO, num_buf_aio, CTLFLAG_RD, &num_buf_aio, 0, "Number of aio requests presently handled by the buf subsystem"); /* Number of async I/O processes in the process of being started */ /* XXX This should be local to aio_aqueue() */ static int num_aio_resv_start = 0; static int aiod_lifetime; SYSCTL_INT(_vfs_aio, OID_AUTO, aiod_lifetime, CTLFLAG_RW, &aiod_lifetime, 0, "Maximum lifetime for idle aiod"); static int max_aio_per_proc = MAX_AIO_PER_PROC; SYSCTL_INT(_vfs_aio, OID_AUTO, max_aio_per_proc, CTLFLAG_RW, &max_aio_per_proc, 0, "Maximum active aio requests per process (stored in the process)"); static int max_aio_queue_per_proc = MAX_AIO_QUEUE_PER_PROC; SYSCTL_INT(_vfs_aio, OID_AUTO, max_aio_queue_per_proc, CTLFLAG_RW, &max_aio_queue_per_proc, 0, "Maximum queued aio requests per process (stored in the process)"); static int max_buf_aio = MAX_BUF_AIO; SYSCTL_INT(_vfs_aio, OID_AUTO, max_buf_aio, CTLFLAG_RW, &max_buf_aio, 0, "Maximum buf aio requests per process (stored in the process)"); #ifdef COMPAT_FREEBSD6 typedef struct oaiocb { int aio_fildes; /* File descriptor */ off_t aio_offset; /* File offset for I/O */ volatile void *aio_buf; /* I/O buffer in process space */ size_t aio_nbytes; /* Number of bytes for I/O */ struct osigevent aio_sigevent; /* Signal to deliver */ int aio_lio_opcode; /* LIO opcode */ int aio_reqprio; /* Request priority -- ignored */ struct __aiocb_private _aiocb_private; } oaiocb_t; #endif /* * Below is a key of locks used to protect each member of struct kaiocb * aioliojob and kaioinfo and any backends. * * * - need not protected * a - locked by kaioinfo lock * b - locked by backend lock, the backend lock can be null in some cases, * for example, BIO belongs to this type, in this case, proc lock is * reused. * c - locked by aio_job_mtx, the lock for the generic file I/O backend. */ /* * If the routine that services an AIO request blocks while running in an * AIO kernel process it can starve other I/O requests. BIO requests * queued via aio_qphysio() complete in GEOM and do not use AIO kernel * processes at all. Socket I/O requests use a separate pool of * kprocs and also force non-blocking I/O. Other file I/O requests * use the generic fo_read/fo_write operations which can block. The * fsync and mlock operations can also block while executing. Ideally * none of these requests would block while executing. * * Note that the service routines cannot toggle O_NONBLOCK in the file * structure directly while handling a request due to races with * userland threads. */ /* jobflags */ #define KAIOCB_QUEUEING 0x01 #define KAIOCB_CANCELLED 0x02 #define KAIOCB_CANCELLING 0x04 #define KAIOCB_CHECKSYNC 0x08 #define KAIOCB_CLEARED 0x10 #define KAIOCB_FINISHED 0x20 /* * AIO process info */ #define AIOP_FREE 0x1 /* proc on free queue */ struct aioproc { int aioprocflags; /* (c) AIO proc flags */ TAILQ_ENTRY(aioproc) list; /* (c) list of processes */ struct proc *aioproc; /* (*) the AIO proc */ }; /* * data-structure for lio signal management */ struct aioliojob { int lioj_flags; /* (a) listio flags */ int lioj_count; /* (a) listio flags */ int lioj_finished_count; /* (a) listio flags */ struct sigevent lioj_signal; /* (a) signal on all I/O done */ TAILQ_ENTRY(aioliojob) lioj_list; /* (a) lio list */ struct knlist klist; /* (a) list of knotes */ ksiginfo_t lioj_ksi; /* (a) Realtime signal info */ }; #define LIOJ_SIGNAL 0x1 /* signal on all done (lio) */ #define LIOJ_SIGNAL_POSTED 0x2 /* signal has been posted */ #define LIOJ_KEVENT_POSTED 0x4 /* kevent triggered */ /* * per process aio data structure */ struct kaioinfo { struct mtx kaio_mtx; /* the lock to protect this struct */ int kaio_flags; /* (a) per process kaio flags */ int kaio_maxactive_count; /* (*) maximum number of AIOs */ int kaio_active_count; /* (c) number of currently used AIOs */ int kaio_qallowed_count; /* (*) maxiumu size of AIO queue */ int kaio_count; /* (a) size of AIO queue */ int kaio_ballowed_count; /* (*) maximum number of buffers */ int kaio_buffer_count; /* (a) number of physio buffers */ TAILQ_HEAD(,kaiocb) kaio_all; /* (a) all AIOs in a process */ TAILQ_HEAD(,kaiocb) kaio_done; /* (a) done queue for process */ TAILQ_HEAD(,aioliojob) kaio_liojoblist; /* (a) list of lio jobs */ TAILQ_HEAD(,kaiocb) kaio_jobqueue; /* (a) job queue for process */ TAILQ_HEAD(,kaiocb) kaio_syncqueue; /* (a) queue for aio_fsync */ TAILQ_HEAD(,kaiocb) kaio_syncready; /* (a) second q for aio_fsync */ struct task kaio_task; /* (*) task to kick aio processes */ struct task kaio_sync_task; /* (*) task to schedule fsync jobs */ }; #define AIO_LOCK(ki) mtx_lock(&(ki)->kaio_mtx) #define AIO_UNLOCK(ki) mtx_unlock(&(ki)->kaio_mtx) #define AIO_LOCK_ASSERT(ki, f) mtx_assert(&(ki)->kaio_mtx, (f)) #define AIO_MTX(ki) (&(ki)->kaio_mtx) #define KAIO_RUNDOWN 0x1 /* process is being run down */ #define KAIO_WAKEUP 0x2 /* wakeup process when AIO completes */ /* * Operations used to interact with userland aio control blocks. * Different ABIs provide their own operations. */ struct aiocb_ops { int (*copyin)(struct aiocb *ujob, struct aiocb *kjob); long (*fetch_status)(struct aiocb *ujob); long (*fetch_error)(struct aiocb *ujob); int (*store_status)(struct aiocb *ujob, long status); int (*store_error)(struct aiocb *ujob, long error); int (*store_kernelinfo)(struct aiocb *ujob, long jobref); int (*store_aiocb)(struct aiocb **ujobp, struct aiocb *ujob); }; static TAILQ_HEAD(,aioproc) aio_freeproc; /* (c) Idle daemons */ static struct sema aio_newproc_sem; static struct mtx aio_job_mtx; static TAILQ_HEAD(,kaiocb) aio_jobs; /* (c) Async job list */ static struct unrhdr *aiod_unr; void aio_init_aioinfo(struct proc *p); static int aio_onceonly(void); static int aio_free_entry(struct kaiocb *job); static void aio_process_rw(struct kaiocb *job); static void aio_process_sync(struct kaiocb *job); static void aio_process_mlock(struct kaiocb *job); static void aio_schedule_fsync(void *context, int pending); static int aio_newproc(int *); int aio_aqueue(struct thread *td, struct aiocb *ujob, struct aioliojob *lio, int type, struct aiocb_ops *ops); static int aio_queue_file(struct file *fp, struct kaiocb *job); static void aio_physwakeup(struct bio *bp); static void aio_proc_rundown(void *arg, struct proc *p); static void aio_proc_rundown_exec(void *arg, struct proc *p, struct image_params *imgp); static int aio_qphysio(struct proc *p, struct kaiocb *job); static void aio_daemon(void *param); static void aio_bio_done_notify(struct proc *userp, struct kaiocb *job); static int aio_kick(struct proc *userp); static void aio_kick_nowait(struct proc *userp); static void aio_kick_helper(void *context, int pending); static int filt_aioattach(struct knote *kn); static void filt_aiodetach(struct knote *kn); static int filt_aio(struct knote *kn, long hint); static int filt_lioattach(struct knote *kn); static void filt_liodetach(struct knote *kn); static int filt_lio(struct knote *kn, long hint); /* * Zones for: * kaio Per process async io info * aiop async io process data * aiocb async io jobs * aiol list io job pointer - internal to aio_suspend XXX * aiolio list io jobs */ static uma_zone_t kaio_zone, aiop_zone, aiocb_zone, aiol_zone, aiolio_zone; /* kqueue filters for aio */ static struct filterops aio_filtops = { .f_isfd = 0, .f_attach = filt_aioattach, .f_detach = filt_aiodetach, .f_event = filt_aio, }; static struct filterops lio_filtops = { .f_isfd = 0, .f_attach = filt_lioattach, .f_detach = filt_liodetach, .f_event = filt_lio }; static eventhandler_tag exit_tag, exec_tag; TASKQUEUE_DEFINE_THREAD(aiod_kick); /* * Main operations function for use as a kernel module. */ static int aio_modload(struct module *module, int cmd, void *arg) { int error = 0; switch (cmd) { case MOD_LOAD: aio_onceonly(); break; case MOD_SHUTDOWN: break; default: error = EOPNOTSUPP; break; } return (error); } static moduledata_t aio_mod = { "aio", &aio_modload, NULL }; DECLARE_MODULE(aio, aio_mod, SI_SUB_VFS, SI_ORDER_ANY); MODULE_VERSION(aio, 1); /* * Startup initialization */ static int aio_onceonly(void) { exit_tag = EVENTHANDLER_REGISTER(process_exit, aio_proc_rundown, NULL, EVENTHANDLER_PRI_ANY); exec_tag = EVENTHANDLER_REGISTER(process_exec, aio_proc_rundown_exec, NULL, EVENTHANDLER_PRI_ANY); kqueue_add_filteropts(EVFILT_AIO, &aio_filtops); kqueue_add_filteropts(EVFILT_LIO, &lio_filtops); TAILQ_INIT(&aio_freeproc); sema_init(&aio_newproc_sem, 0, "aio_new_proc"); mtx_init(&aio_job_mtx, "aio_job", NULL, MTX_DEF); TAILQ_INIT(&aio_jobs); aiod_unr = new_unrhdr(1, INT_MAX, NULL); kaio_zone = uma_zcreate("AIO", sizeof(struct kaioinfo), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); aiop_zone = uma_zcreate("AIOP", sizeof(struct aioproc), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); aiocb_zone = uma_zcreate("AIOCB", sizeof(struct kaiocb), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); aiol_zone = uma_zcreate("AIOL", AIO_LISTIO_MAX*sizeof(intptr_t) , NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); aiolio_zone = uma_zcreate("AIOLIO", sizeof(struct aioliojob), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); aiod_lifetime = AIOD_LIFETIME_DEFAULT; jobrefid = 1; p31b_setcfg(CTL_P1003_1B_ASYNCHRONOUS_IO, _POSIX_ASYNCHRONOUS_IO); p31b_setcfg(CTL_P1003_1B_AIO_LISTIO_MAX, AIO_LISTIO_MAX); p31b_setcfg(CTL_P1003_1B_AIO_MAX, MAX_AIO_QUEUE); p31b_setcfg(CTL_P1003_1B_AIO_PRIO_DELTA_MAX, 0); return (0); } /* * Init the per-process aioinfo structure. The aioinfo limits are set * per-process for user limit (resource) management. */ void aio_init_aioinfo(struct proc *p) { struct kaioinfo *ki; ki = uma_zalloc(kaio_zone, M_WAITOK); mtx_init(&ki->kaio_mtx, "aiomtx", NULL, MTX_DEF | MTX_NEW); ki->kaio_flags = 0; ki->kaio_maxactive_count = max_aio_per_proc; ki->kaio_active_count = 0; ki->kaio_qallowed_count = max_aio_queue_per_proc; ki->kaio_count = 0; ki->kaio_ballowed_count = max_buf_aio; ki->kaio_buffer_count = 0; TAILQ_INIT(&ki->kaio_all); TAILQ_INIT(&ki->kaio_done); TAILQ_INIT(&ki->kaio_jobqueue); TAILQ_INIT(&ki->kaio_liojoblist); TAILQ_INIT(&ki->kaio_syncqueue); TAILQ_INIT(&ki->kaio_syncready); TASK_INIT(&ki->kaio_task, 0, aio_kick_helper, p); TASK_INIT(&ki->kaio_sync_task, 0, aio_schedule_fsync, ki); PROC_LOCK(p); if (p->p_aioinfo == NULL) { p->p_aioinfo = ki; PROC_UNLOCK(p); } else { PROC_UNLOCK(p); mtx_destroy(&ki->kaio_mtx); uma_zfree(kaio_zone, ki); } while (num_aio_procs < MIN(target_aio_procs, max_aio_procs)) aio_newproc(NULL); } static int aio_sendsig(struct proc *p, struct sigevent *sigev, ksiginfo_t *ksi) { struct thread *td; int error; error = sigev_findtd(p, sigev, &td); if (error) return (error); if (!KSI_ONQ(ksi)) { ksiginfo_set_sigev(ksi, sigev); ksi->ksi_code = SI_ASYNCIO; ksi->ksi_flags |= KSI_EXT | KSI_INS; tdsendsignal(p, td, ksi->ksi_signo, ksi); } PROC_UNLOCK(p); return (error); } /* * Free a job entry. Wait for completion if it is currently active, but don't * delay forever. If we delay, we return a flag that says that we have to * restart the queue scan. */ static int aio_free_entry(struct kaiocb *job) { struct kaioinfo *ki; struct aioliojob *lj; struct proc *p; p = job->userproc; MPASS(curproc == p); ki = p->p_aioinfo; MPASS(ki != NULL); AIO_LOCK_ASSERT(ki, MA_OWNED); MPASS(job->jobflags & KAIOCB_FINISHED); atomic_subtract_int(&num_queue_count, 1); ki->kaio_count--; MPASS(ki->kaio_count >= 0); TAILQ_REMOVE(&ki->kaio_done, job, plist); TAILQ_REMOVE(&ki->kaio_all, job, allist); lj = job->lio; if (lj) { lj->lioj_count--; lj->lioj_finished_count--; if (lj->lioj_count == 0) { TAILQ_REMOVE(&ki->kaio_liojoblist, lj, lioj_list); /* lio is going away, we need to destroy any knotes */ knlist_delete(&lj->klist, curthread, 1); PROC_LOCK(p); sigqueue_take(&lj->lioj_ksi); PROC_UNLOCK(p); uma_zfree(aiolio_zone, lj); } } /* job is going away, we need to destroy any knotes */ knlist_delete(&job->klist, curthread, 1); PROC_LOCK(p); sigqueue_take(&job->ksi); PROC_UNLOCK(p); MPASS(job->bp == NULL); AIO_UNLOCK(ki); /* * The thread argument here is used to find the owning process * and is also passed to fo_close() which may pass it to various * places such as devsw close() routines. Because of that, we * need a thread pointer from the process owning the job that is * persistent and won't disappear out from under us or move to * another process. * * Currently, all the callers of this function call it to remove * a kaiocb from the current process' job list either via a * syscall or due to the current process calling exit() or * execve(). Thus, we know that p == curproc. We also know that * curthread can't exit since we are curthread. * * Therefore, we use curthread as the thread to pass to * knlist_delete(). This does mean that it is possible for the * thread pointer at close time to differ from the thread pointer * at open time, but this is already true of file descriptors in * a multithreaded process. */ if (job->fd_file) fdrop(job->fd_file, curthread); crfree(job->cred); uma_zfree(aiocb_zone, job); AIO_LOCK(ki); return (0); } static void aio_proc_rundown_exec(void *arg, struct proc *p, struct image_params *imgp __unused) { aio_proc_rundown(arg, p); } static int aio_cancel_job(struct proc *p, struct kaioinfo *ki, struct kaiocb *job) { aio_cancel_fn_t *func; int cancelled; AIO_LOCK_ASSERT(ki, MA_OWNED); if (job->jobflags & (KAIOCB_CANCELLED | KAIOCB_FINISHED)) return (0); MPASS((job->jobflags & KAIOCB_CANCELLING) == 0); job->jobflags |= KAIOCB_CANCELLED; func = job->cancel_fn; /* * If there is no cancel routine, just leave the job marked as * cancelled. The job should be in active use by a caller who * should complete it normally or when it fails to install a * cancel routine. */ if (func == NULL) return (0); /* * Set the CANCELLING flag so that aio_complete() will defer * completions of this job. This prevents the job from being * freed out from under the cancel callback. After the * callback any deferred completion (whether from the callback * or any other source) will be completed. */ job->jobflags |= KAIOCB_CANCELLING; AIO_UNLOCK(ki); func(job); AIO_LOCK(ki); job->jobflags &= ~KAIOCB_CANCELLING; if (job->jobflags & KAIOCB_FINISHED) { cancelled = job->uaiocb._aiocb_private.error == ECANCELED; TAILQ_REMOVE(&ki->kaio_jobqueue, job, plist); aio_bio_done_notify(p, job); } else { /* * The cancel callback might have scheduled an * operation to cancel this request, but it is * only counted as cancelled if the request is * cancelled when the callback returns. */ cancelled = 0; } return (cancelled); } /* * Rundown the jobs for a given process. */ static void aio_proc_rundown(void *arg, struct proc *p) { struct kaioinfo *ki; struct aioliojob *lj; struct kaiocb *job, *jobn; KASSERT(curthread->td_proc == p, ("%s: called on non-curproc", __func__)); ki = p->p_aioinfo; if (ki == NULL) return; AIO_LOCK(ki); ki->kaio_flags |= KAIO_RUNDOWN; restart: /* * Try to cancel all pending requests. This code simulates * aio_cancel on all pending I/O requests. */ TAILQ_FOREACH_SAFE(job, &ki->kaio_jobqueue, plist, jobn) { aio_cancel_job(p, ki, job); } /* Wait for all running I/O to be finished */ if (TAILQ_FIRST(&ki->kaio_jobqueue) || ki->kaio_active_count != 0) { ki->kaio_flags |= KAIO_WAKEUP; msleep(&p->p_aioinfo, AIO_MTX(ki), PRIBIO, "aioprn", hz); goto restart; } /* Free all completed I/O requests. */ while ((job = TAILQ_FIRST(&ki->kaio_done)) != NULL) aio_free_entry(job); while ((lj = TAILQ_FIRST(&ki->kaio_liojoblist)) != NULL) { if (lj->lioj_count == 0) { TAILQ_REMOVE(&ki->kaio_liojoblist, lj, lioj_list); knlist_delete(&lj->klist, curthread, 1); PROC_LOCK(p); sigqueue_take(&lj->lioj_ksi); PROC_UNLOCK(p); uma_zfree(aiolio_zone, lj); } else { panic("LIO job not cleaned up: C:%d, FC:%d\n", lj->lioj_count, lj->lioj_finished_count); } } AIO_UNLOCK(ki); taskqueue_drain(taskqueue_aiod_kick, &ki->kaio_task); taskqueue_drain(taskqueue_aiod_kick, &ki->kaio_sync_task); mtx_destroy(&ki->kaio_mtx); uma_zfree(kaio_zone, ki); p->p_aioinfo = NULL; } /* * Select a job to run (called by an AIO daemon). */ static struct kaiocb * aio_selectjob(struct aioproc *aiop) { struct kaiocb *job; struct kaioinfo *ki; struct proc *userp; mtx_assert(&aio_job_mtx, MA_OWNED); restart: TAILQ_FOREACH(job, &aio_jobs, list) { userp = job->userproc; ki = userp->p_aioinfo; if (ki->kaio_active_count < ki->kaio_maxactive_count) { TAILQ_REMOVE(&aio_jobs, job, list); if (!aio_clear_cancel_function(job)) goto restart; /* Account for currently active jobs. */ ki->kaio_active_count++; break; } } return (job); } /* * Move all data to a permanent storage device. This code * simulates the fsync syscall. */ static int aio_fsync_vnode(struct thread *td, struct vnode *vp) { struct mount *mp; int error; if ((error = vn_start_write(vp, &mp, V_WAIT | PCATCH)) != 0) goto drop; vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); if (vp->v_object != NULL) { VM_OBJECT_WLOCK(vp->v_object); vm_object_page_clean(vp->v_object, 0, 0, 0); VM_OBJECT_WUNLOCK(vp->v_object); } error = VOP_FSYNC(vp, MNT_WAIT, td); VOP_UNLOCK(vp, 0); vn_finished_write(mp); drop: return (error); } /* * The AIO processing activity for LIO_READ/LIO_WRITE. This is the code that * does the I/O request for the non-physio version of the operations. The * normal vn operations are used, and this code should work in all instances * for every type of file, including pipes, sockets, fifos, and regular files. * * XXX I don't think it works well for socket, pipe, and fifo. */ static void aio_process_rw(struct kaiocb *job) { struct ucred *td_savedcred; struct thread *td; struct aiocb *cb; struct file *fp; struct uio auio; struct iovec aiov; - int cnt; + ssize_t cnt; int error; int oublock_st, oublock_end; int inblock_st, inblock_end; KASSERT(job->uaiocb.aio_lio_opcode == LIO_READ || job->uaiocb.aio_lio_opcode == LIO_WRITE, ("%s: opcode %d", __func__, job->uaiocb.aio_lio_opcode)); aio_switch_vmspace(job); td = curthread; td_savedcred = td->td_ucred; td->td_ucred = job->cred; cb = &job->uaiocb; fp = job->fd_file; aiov.iov_base = (void *)(uintptr_t)cb->aio_buf; aiov.iov_len = cb->aio_nbytes; auio.uio_iov = &aiov; auio.uio_iovcnt = 1; auio.uio_offset = cb->aio_offset; auio.uio_resid = cb->aio_nbytes; cnt = cb->aio_nbytes; auio.uio_segflg = UIO_USERSPACE; auio.uio_td = td; inblock_st = td->td_ru.ru_inblock; oublock_st = td->td_ru.ru_oublock; /* * aio_aqueue() acquires a reference to the file that is * released in aio_free_entry(). */ if (cb->aio_lio_opcode == LIO_READ) { auio.uio_rw = UIO_READ; if (auio.uio_resid == 0) error = 0; else error = fo_read(fp, &auio, fp->f_cred, FOF_OFFSET, td); } else { if (fp->f_type == DTYPE_VNODE) bwillwrite(); auio.uio_rw = UIO_WRITE; error = fo_write(fp, &auio, fp->f_cred, FOF_OFFSET, td); } inblock_end = td->td_ru.ru_inblock; oublock_end = td->td_ru.ru_oublock; job->inputcharge = inblock_end - inblock_st; job->outputcharge = oublock_end - oublock_st; if ((error) && (auio.uio_resid != cnt)) { if (error == ERESTART || error == EINTR || error == EWOULDBLOCK) error = 0; if ((error == EPIPE) && (cb->aio_lio_opcode == LIO_WRITE)) { PROC_LOCK(job->userproc); kern_psignal(job->userproc, SIGPIPE); PROC_UNLOCK(job->userproc); } } cnt -= auio.uio_resid; td->td_ucred = td_savedcred; aio_complete(job, cnt, error); } static void aio_process_sync(struct kaiocb *job) { struct thread *td = curthread; struct ucred *td_savedcred = td->td_ucred; struct file *fp = job->fd_file; int error = 0; KASSERT(job->uaiocb.aio_lio_opcode == LIO_SYNC, ("%s: opcode %d", __func__, job->uaiocb.aio_lio_opcode)); td->td_ucred = job->cred; if (fp->f_vnode != NULL) error = aio_fsync_vnode(td, fp->f_vnode); td->td_ucred = td_savedcred; aio_complete(job, 0, error); } static void aio_process_mlock(struct kaiocb *job) { struct aiocb *cb = &job->uaiocb; int error; KASSERT(job->uaiocb.aio_lio_opcode == LIO_MLOCK, ("%s: opcode %d", __func__, job->uaiocb.aio_lio_opcode)); aio_switch_vmspace(job); error = vm_mlock(job->userproc, job->cred, __DEVOLATILE(void *, cb->aio_buf), cb->aio_nbytes); aio_complete(job, 0, error); } static void aio_bio_done_notify(struct proc *userp, struct kaiocb *job) { struct aioliojob *lj; struct kaioinfo *ki; struct kaiocb *sjob, *sjobn; int lj_done; bool schedule_fsync; ki = userp->p_aioinfo; AIO_LOCK_ASSERT(ki, MA_OWNED); lj = job->lio; lj_done = 0; if (lj) { lj->lioj_finished_count++; if (lj->lioj_count == lj->lioj_finished_count) lj_done = 1; } TAILQ_INSERT_TAIL(&ki->kaio_done, job, plist); MPASS(job->jobflags & KAIOCB_FINISHED); if (ki->kaio_flags & KAIO_RUNDOWN) goto notification_done; if (job->uaiocb.aio_sigevent.sigev_notify == SIGEV_SIGNAL || job->uaiocb.aio_sigevent.sigev_notify == SIGEV_THREAD_ID) aio_sendsig(userp, &job->uaiocb.aio_sigevent, &job->ksi); KNOTE_LOCKED(&job->klist, 1); if (lj_done) { if (lj->lioj_signal.sigev_notify == SIGEV_KEVENT) { lj->lioj_flags |= LIOJ_KEVENT_POSTED; KNOTE_LOCKED(&lj->klist, 1); } if ((lj->lioj_flags & (LIOJ_SIGNAL|LIOJ_SIGNAL_POSTED)) == LIOJ_SIGNAL && (lj->lioj_signal.sigev_notify == SIGEV_SIGNAL || lj->lioj_signal.sigev_notify == SIGEV_THREAD_ID)) { aio_sendsig(userp, &lj->lioj_signal, &lj->lioj_ksi); lj->lioj_flags |= LIOJ_SIGNAL_POSTED; } } notification_done: if (job->jobflags & KAIOCB_CHECKSYNC) { schedule_fsync = false; TAILQ_FOREACH_SAFE(sjob, &ki->kaio_syncqueue, list, sjobn) { if (job->fd_file == sjob->fd_file && job->seqno < sjob->seqno) { if (--sjob->pending == 0) { TAILQ_REMOVE(&ki->kaio_syncqueue, sjob, list); if (!aio_clear_cancel_function(sjob)) continue; TAILQ_INSERT_TAIL(&ki->kaio_syncready, sjob, list); schedule_fsync = true; } } } if (schedule_fsync) taskqueue_enqueue(taskqueue_aiod_kick, &ki->kaio_sync_task); } if (ki->kaio_flags & KAIO_WAKEUP) { ki->kaio_flags &= ~KAIO_WAKEUP; wakeup(&userp->p_aioinfo); } } static void aio_schedule_fsync(void *context, int pending) { struct kaioinfo *ki; struct kaiocb *job; ki = context; AIO_LOCK(ki); while (!TAILQ_EMPTY(&ki->kaio_syncready)) { job = TAILQ_FIRST(&ki->kaio_syncready); TAILQ_REMOVE(&ki->kaio_syncready, job, list); AIO_UNLOCK(ki); aio_schedule(job, aio_process_sync); AIO_LOCK(ki); } AIO_UNLOCK(ki); } bool aio_cancel_cleared(struct kaiocb *job) { struct kaioinfo *ki; /* * The caller should hold the same queue lock held when * aio_clear_cancel_function() was called and set this flag * ensuring this check sees an up-to-date value. However, * there is no way to assert that. */ ki = job->userproc->p_aioinfo; return ((job->jobflags & KAIOCB_CLEARED) != 0); } bool aio_clear_cancel_function(struct kaiocb *job) { struct kaioinfo *ki; ki = job->userproc->p_aioinfo; AIO_LOCK(ki); MPASS(job->cancel_fn != NULL); if (job->jobflags & KAIOCB_CANCELLING) { job->jobflags |= KAIOCB_CLEARED; AIO_UNLOCK(ki); return (false); } job->cancel_fn = NULL; AIO_UNLOCK(ki); return (true); } bool aio_set_cancel_function(struct kaiocb *job, aio_cancel_fn_t *func) { struct kaioinfo *ki; ki = job->userproc->p_aioinfo; AIO_LOCK(ki); if (job->jobflags & KAIOCB_CANCELLED) { AIO_UNLOCK(ki); return (false); } job->cancel_fn = func; AIO_UNLOCK(ki); return (true); } void aio_complete(struct kaiocb *job, long status, int error) { struct kaioinfo *ki; struct proc *userp; job->uaiocb._aiocb_private.error = error; job->uaiocb._aiocb_private.status = status; userp = job->userproc; ki = userp->p_aioinfo; AIO_LOCK(ki); KASSERT(!(job->jobflags & KAIOCB_FINISHED), ("duplicate aio_complete")); job->jobflags |= KAIOCB_FINISHED; if ((job->jobflags & (KAIOCB_QUEUEING | KAIOCB_CANCELLING)) == 0) { TAILQ_REMOVE(&ki->kaio_jobqueue, job, plist); aio_bio_done_notify(userp, job); } AIO_UNLOCK(ki); } void aio_cancel(struct kaiocb *job) { aio_complete(job, -1, ECANCELED); } void aio_switch_vmspace(struct kaiocb *job) { vmspace_switch_aio(job->userproc->p_vmspace); } /* * The AIO daemon, most of the actual work is done in aio_process_*, * but the setup (and address space mgmt) is done in this routine. */ static void aio_daemon(void *_id) { struct kaiocb *job; struct aioproc *aiop; struct kaioinfo *ki; struct proc *p; struct vmspace *myvm; struct thread *td = curthread; int id = (intptr_t)_id; /* * Grab an extra reference on the daemon's vmspace so that it * doesn't get freed by jobs that switch to a different * vmspace. */ p = td->td_proc; myvm = vmspace_acquire_ref(p); KASSERT(p->p_textvp == NULL, ("kthread has a textvp")); /* * Allocate and ready the aio control info. There is one aiop structure * per daemon. */ aiop = uma_zalloc(aiop_zone, M_WAITOK); aiop->aioproc = p; aiop->aioprocflags = 0; /* * Wakeup parent process. (Parent sleeps to keep from blasting away * and creating too many daemons.) */ sema_post(&aio_newproc_sem); mtx_lock(&aio_job_mtx); for (;;) { /* * Take daemon off of free queue */ if (aiop->aioprocflags & AIOP_FREE) { TAILQ_REMOVE(&aio_freeproc, aiop, list); aiop->aioprocflags &= ~AIOP_FREE; } /* * Check for jobs. */ while ((job = aio_selectjob(aiop)) != NULL) { mtx_unlock(&aio_job_mtx); ki = job->userproc->p_aioinfo; job->handle_fn(job); mtx_lock(&aio_job_mtx); /* Decrement the active job count. */ ki->kaio_active_count--; } /* * Disconnect from user address space. */ if (p->p_vmspace != myvm) { mtx_unlock(&aio_job_mtx); vmspace_switch_aio(myvm); mtx_lock(&aio_job_mtx); /* * We have to restart to avoid race, we only sleep if * no job can be selected. */ continue; } mtx_assert(&aio_job_mtx, MA_OWNED); TAILQ_INSERT_HEAD(&aio_freeproc, aiop, list); aiop->aioprocflags |= AIOP_FREE; /* * If daemon is inactive for a long time, allow it to exit, * thereby freeing resources. */ if (msleep(p, &aio_job_mtx, PRIBIO, "aiordy", aiod_lifetime) == EWOULDBLOCK && TAILQ_EMPTY(&aio_jobs) && (aiop->aioprocflags & AIOP_FREE) && num_aio_procs > target_aio_procs) break; } TAILQ_REMOVE(&aio_freeproc, aiop, list); num_aio_procs--; mtx_unlock(&aio_job_mtx); uma_zfree(aiop_zone, aiop); free_unr(aiod_unr, id); vmspace_free(myvm); KASSERT(p->p_vmspace == myvm, ("AIOD: bad vmspace for exiting daemon")); KASSERT(myvm->vm_refcnt > 1, ("AIOD: bad vm refcnt for exiting daemon: %d", myvm->vm_refcnt)); kproc_exit(0); } /* * Create a new AIO daemon. This is mostly a kernel-thread fork routine. The * AIO daemon modifies its environment itself. */ static int aio_newproc(int *start) { int error; struct proc *p; int id; id = alloc_unr(aiod_unr); error = kproc_create(aio_daemon, (void *)(intptr_t)id, &p, RFNOWAIT, 0, "aiod%d", id); if (error == 0) { /* * Wait until daemon is started. */ sema_wait(&aio_newproc_sem); mtx_lock(&aio_job_mtx); num_aio_procs++; if (start != NULL) (*start)--; mtx_unlock(&aio_job_mtx); } else { free_unr(aiod_unr, id); } return (error); } /* * Try the high-performance, low-overhead physio method for eligible * VCHR devices. This method doesn't use an aio helper thread, and * thus has very low overhead. * * Assumes that the caller, aio_aqueue(), has incremented the file * structure's reference count, preventing its deallocation for the * duration of this call. */ static int aio_qphysio(struct proc *p, struct kaiocb *job) { struct aiocb *cb; struct file *fp; struct bio *bp; struct buf *pbuf; struct vnode *vp; struct cdevsw *csw; struct cdev *dev; struct kaioinfo *ki; int error, ref, unmap, poff; vm_prot_t prot; cb = &job->uaiocb; fp = job->fd_file; if (fp == NULL || fp->f_type != DTYPE_VNODE) return (-1); vp = fp->f_vnode; if (vp->v_type != VCHR) return (-1); if (vp->v_bufobj.bo_bsize == 0) return (-1); if (cb->aio_nbytes % vp->v_bufobj.bo_bsize) return (-1); ref = 0; csw = devvn_refthread(vp, &dev, &ref); if (csw == NULL) return (ENXIO); if ((csw->d_flags & D_DISK) == 0) { error = -1; goto unref; } if (cb->aio_nbytes > dev->si_iosize_max) { error = -1; goto unref; } ki = p->p_aioinfo; poff = (vm_offset_t)cb->aio_buf & PAGE_MASK; unmap = ((dev->si_flags & SI_UNMAPPED) && unmapped_buf_allowed); if (unmap) { if (cb->aio_nbytes > MAXPHYS) { error = -1; goto unref; } } else { if (cb->aio_nbytes > MAXPHYS - poff) { error = -1; goto unref; } if (ki->kaio_buffer_count >= ki->kaio_ballowed_count) { error = -1; goto unref; } } job->bp = bp = g_alloc_bio(); if (!unmap) { job->pbuf = pbuf = (struct buf *)getpbuf(NULL); BUF_KERNPROC(pbuf); } else pbuf = NULL; AIO_LOCK(ki); if (!unmap) ki->kaio_buffer_count++; AIO_UNLOCK(ki); bp->bio_length = cb->aio_nbytes; bp->bio_bcount = cb->aio_nbytes; bp->bio_done = aio_physwakeup; bp->bio_data = (void *)(uintptr_t)cb->aio_buf; bp->bio_offset = cb->aio_offset; bp->bio_cmd = cb->aio_lio_opcode == LIO_WRITE ? BIO_WRITE : BIO_READ; bp->bio_dev = dev; bp->bio_caller1 = (void *)job; prot = VM_PROT_READ; if (cb->aio_lio_opcode == LIO_READ) prot |= VM_PROT_WRITE; /* Less backwards than it looks */ if ((job->npages = vm_fault_quick_hold_pages( &curproc->p_vmspace->vm_map, (vm_offset_t)bp->bio_data, bp->bio_length, prot, job->pages, sizeof(job->pages)/sizeof(job->pages[0]))) < 0) { error = EFAULT; goto doerror; } if (!unmap) { pmap_qenter((vm_offset_t)pbuf->b_data, job->pages, job->npages); bp->bio_data = pbuf->b_data + poff; } else { bp->bio_ma = job->pages; bp->bio_ma_n = job->npages; bp->bio_ma_offset = poff; bp->bio_data = unmapped_buf; bp->bio_flags |= BIO_UNMAPPED; } if (!unmap) atomic_add_int(&num_buf_aio, 1); /* Perform transfer. */ csw->d_strategy(bp); dev_relthread(dev, ref); return (0); doerror: AIO_LOCK(ki); if (!unmap) ki->kaio_buffer_count--; AIO_UNLOCK(ki); if (pbuf) { relpbuf(pbuf, NULL); job->pbuf = NULL; } g_destroy_bio(bp); job->bp = NULL; unref: dev_relthread(dev, ref); return (error); } #ifdef COMPAT_FREEBSD6 static int convert_old_sigevent(struct osigevent *osig, struct sigevent *nsig) { /* * Only SIGEV_NONE, SIGEV_SIGNAL, and SIGEV_KEVENT are * supported by AIO with the old sigevent structure. */ nsig->sigev_notify = osig->sigev_notify; switch (nsig->sigev_notify) { case SIGEV_NONE: break; case SIGEV_SIGNAL: nsig->sigev_signo = osig->__sigev_u.__sigev_signo; break; case SIGEV_KEVENT: nsig->sigev_notify_kqueue = osig->__sigev_u.__sigev_notify_kqueue; nsig->sigev_value.sival_ptr = osig->sigev_value.sival_ptr; break; default: return (EINVAL); } return (0); } static int aiocb_copyin_old_sigevent(struct aiocb *ujob, struct aiocb *kjob) { struct oaiocb *ojob; int error; bzero(kjob, sizeof(struct aiocb)); error = copyin(ujob, kjob, sizeof(struct oaiocb)); if (error) return (error); ojob = (struct oaiocb *)kjob; return (convert_old_sigevent(&ojob->aio_sigevent, &kjob->aio_sigevent)); } #endif static int aiocb_copyin(struct aiocb *ujob, struct aiocb *kjob) { return (copyin(ujob, kjob, sizeof(struct aiocb))); } static long aiocb_fetch_status(struct aiocb *ujob) { return (fuword(&ujob->_aiocb_private.status)); } static long aiocb_fetch_error(struct aiocb *ujob) { return (fuword(&ujob->_aiocb_private.error)); } static int aiocb_store_status(struct aiocb *ujob, long status) { return (suword(&ujob->_aiocb_private.status, status)); } static int aiocb_store_error(struct aiocb *ujob, long error) { return (suword(&ujob->_aiocb_private.error, error)); } static int aiocb_store_kernelinfo(struct aiocb *ujob, long jobref) { return (suword(&ujob->_aiocb_private.kernelinfo, jobref)); } static int aiocb_store_aiocb(struct aiocb **ujobp, struct aiocb *ujob) { return (suword(ujobp, (long)ujob)); } static struct aiocb_ops aiocb_ops = { .copyin = aiocb_copyin, .fetch_status = aiocb_fetch_status, .fetch_error = aiocb_fetch_error, .store_status = aiocb_store_status, .store_error = aiocb_store_error, .store_kernelinfo = aiocb_store_kernelinfo, .store_aiocb = aiocb_store_aiocb, }; #ifdef COMPAT_FREEBSD6 static struct aiocb_ops aiocb_ops_osigevent = { .copyin = aiocb_copyin_old_sigevent, .fetch_status = aiocb_fetch_status, .fetch_error = aiocb_fetch_error, .store_status = aiocb_store_status, .store_error = aiocb_store_error, .store_kernelinfo = aiocb_store_kernelinfo, .store_aiocb = aiocb_store_aiocb, }; #endif /* * Queue a new AIO request. Choosing either the threaded or direct physio VCHR * technique is done in this code. */ int aio_aqueue(struct thread *td, struct aiocb *ujob, struct aioliojob *lj, int type, struct aiocb_ops *ops) { struct proc *p = td->td_proc; cap_rights_t rights; struct file *fp; struct kaiocb *job; struct kaioinfo *ki; struct kevent kev; int opcode; int error; int fd, kqfd; int jid; u_short evflags; if (p->p_aioinfo == NULL) aio_init_aioinfo(p); ki = p->p_aioinfo; ops->store_status(ujob, -1); ops->store_error(ujob, 0); ops->store_kernelinfo(ujob, -1); if (num_queue_count >= max_queue_count || ki->kaio_count >= ki->kaio_qallowed_count) { ops->store_error(ujob, EAGAIN); return (EAGAIN); } job = uma_zalloc(aiocb_zone, M_WAITOK | M_ZERO); knlist_init_mtx(&job->klist, AIO_MTX(ki)); error = ops->copyin(ujob, &job->uaiocb); if (error) { ops->store_error(ujob, error); uma_zfree(aiocb_zone, job); return (error); } - /* XXX: aio_nbytes is later casted to signed types. */ - if (job->uaiocb.aio_nbytes > INT_MAX) { + if (job->uaiocb.aio_nbytes > IOSIZE_MAX) { uma_zfree(aiocb_zone, job); return (EINVAL); } if (job->uaiocb.aio_sigevent.sigev_notify != SIGEV_KEVENT && job->uaiocb.aio_sigevent.sigev_notify != SIGEV_SIGNAL && job->uaiocb.aio_sigevent.sigev_notify != SIGEV_THREAD_ID && job->uaiocb.aio_sigevent.sigev_notify != SIGEV_NONE) { ops->store_error(ujob, EINVAL); uma_zfree(aiocb_zone, job); return (EINVAL); } if ((job->uaiocb.aio_sigevent.sigev_notify == SIGEV_SIGNAL || job->uaiocb.aio_sigevent.sigev_notify == SIGEV_THREAD_ID) && !_SIG_VALID(job->uaiocb.aio_sigevent.sigev_signo)) { uma_zfree(aiocb_zone, job); return (EINVAL); } ksiginfo_init(&job->ksi); /* Save userspace address of the job info. */ job->ujob = ujob; /* Get the opcode. */ if (type != LIO_NOP) job->uaiocb.aio_lio_opcode = type; opcode = job->uaiocb.aio_lio_opcode; /* * Validate the opcode and fetch the file object for the specified * file descriptor. * * XXXRW: Moved the opcode validation up here so that we don't * retrieve a file descriptor without knowing what the capabiltity * should be. */ fd = job->uaiocb.aio_fildes; switch (opcode) { case LIO_WRITE: error = fget_write(td, fd, cap_rights_init(&rights, CAP_PWRITE), &fp); break; case LIO_READ: error = fget_read(td, fd, cap_rights_init(&rights, CAP_PREAD), &fp); break; case LIO_SYNC: error = fget(td, fd, cap_rights_init(&rights, CAP_FSYNC), &fp); break; case LIO_MLOCK: fp = NULL; break; case LIO_NOP: error = fget(td, fd, cap_rights_init(&rights), &fp); break; default: error = EINVAL; } if (error) { uma_zfree(aiocb_zone, job); ops->store_error(ujob, error); return (error); } if (opcode == LIO_SYNC && fp->f_vnode == NULL) { error = EINVAL; goto aqueue_fail; } if (opcode != LIO_SYNC && job->uaiocb.aio_offset == -1LL) { error = EINVAL; goto aqueue_fail; } job->fd_file = fp; mtx_lock(&aio_job_mtx); jid = jobrefid++; job->seqno = jobseqno++; mtx_unlock(&aio_job_mtx); error = ops->store_kernelinfo(ujob, jid); if (error) { error = EINVAL; goto aqueue_fail; } job->uaiocb._aiocb_private.kernelinfo = (void *)(intptr_t)jid; if (opcode == LIO_NOP) { fdrop(fp, td); uma_zfree(aiocb_zone, job); return (0); } if (job->uaiocb.aio_sigevent.sigev_notify != SIGEV_KEVENT) goto no_kqueue; evflags = job->uaiocb.aio_sigevent.sigev_notify_kevent_flags; if ((evflags & ~(EV_CLEAR | EV_DISPATCH | EV_ONESHOT)) != 0) { error = EINVAL; goto aqueue_fail; } kqfd = job->uaiocb.aio_sigevent.sigev_notify_kqueue; kev.ident = (uintptr_t)job->ujob; kev.filter = EVFILT_AIO; kev.flags = EV_ADD | EV_ENABLE | EV_FLAG1 | evflags; kev.data = (intptr_t)job; kev.udata = job->uaiocb.aio_sigevent.sigev_value.sival_ptr; error = kqfd_register(kqfd, &kev, td, 1); if (error) goto aqueue_fail; no_kqueue: ops->store_error(ujob, EINPROGRESS); job->uaiocb._aiocb_private.error = EINPROGRESS; job->userproc = p; job->cred = crhold(td->td_ucred); job->jobflags = KAIOCB_QUEUEING; job->lio = lj; if (opcode == LIO_MLOCK) { aio_schedule(job, aio_process_mlock); error = 0; } else if (fp->f_ops->fo_aio_queue == NULL) error = aio_queue_file(fp, job); else error = fo_aio_queue(fp, job); if (error) goto aqueue_fail; AIO_LOCK(ki); job->jobflags &= ~KAIOCB_QUEUEING; TAILQ_INSERT_TAIL(&ki->kaio_all, job, allist); ki->kaio_count++; if (lj) lj->lioj_count++; atomic_add_int(&num_queue_count, 1); if (job->jobflags & KAIOCB_FINISHED) { /* * The queue callback completed the request synchronously. * The bulk of the completion is deferred in that case * until this point. */ aio_bio_done_notify(p, job); } else TAILQ_INSERT_TAIL(&ki->kaio_jobqueue, job, plist); AIO_UNLOCK(ki); return (0); aqueue_fail: knlist_delete(&job->klist, curthread, 0); if (fp) fdrop(fp, td); uma_zfree(aiocb_zone, job); ops->store_error(ujob, error); return (error); } static void aio_cancel_daemon_job(struct kaiocb *job) { mtx_lock(&aio_job_mtx); if (!aio_cancel_cleared(job)) TAILQ_REMOVE(&aio_jobs, job, list); mtx_unlock(&aio_job_mtx); aio_cancel(job); } void aio_schedule(struct kaiocb *job, aio_handle_fn_t *func) { mtx_lock(&aio_job_mtx); if (!aio_set_cancel_function(job, aio_cancel_daemon_job)) { mtx_unlock(&aio_job_mtx); aio_cancel(job); return; } job->handle_fn = func; TAILQ_INSERT_TAIL(&aio_jobs, job, list); aio_kick_nowait(job->userproc); mtx_unlock(&aio_job_mtx); } static void aio_cancel_sync(struct kaiocb *job) { struct kaioinfo *ki; ki = job->userproc->p_aioinfo; mtx_lock(&aio_job_mtx); if (!aio_cancel_cleared(job)) TAILQ_REMOVE(&ki->kaio_syncqueue, job, list); mtx_unlock(&aio_job_mtx); aio_cancel(job); } int aio_queue_file(struct file *fp, struct kaiocb *job) { struct aioliojob *lj; struct kaioinfo *ki; struct kaiocb *job2; int error, opcode; lj = job->lio; ki = job->userproc->p_aioinfo; opcode = job->uaiocb.aio_lio_opcode; if (opcode == LIO_SYNC) goto queueit; if ((error = aio_qphysio(job->userproc, job)) == 0) goto done; #if 0 /* * XXX: This means qphysio() failed with EFAULT. The current * behavior is to retry the operation via fo_read/fo_write. * Wouldn't it be better to just complete the request with an * error here? */ if (error > 0) goto done; #endif queueit: if (!enable_aio_unsafe) return (EOPNOTSUPP); if (opcode == LIO_SYNC) { AIO_LOCK(ki); TAILQ_FOREACH(job2, &ki->kaio_jobqueue, plist) { if (job2->fd_file == job->fd_file && job2->uaiocb.aio_lio_opcode != LIO_SYNC && job2->seqno < job->seqno) { job2->jobflags |= KAIOCB_CHECKSYNC; job->pending++; } } if (job->pending != 0) { if (!aio_set_cancel_function(job, aio_cancel_sync)) { AIO_UNLOCK(ki); aio_cancel(job); return (0); } TAILQ_INSERT_TAIL(&ki->kaio_syncqueue, job, list); AIO_UNLOCK(ki); return (0); } AIO_UNLOCK(ki); } switch (opcode) { case LIO_READ: case LIO_WRITE: aio_schedule(job, aio_process_rw); error = 0; break; case LIO_SYNC: aio_schedule(job, aio_process_sync); error = 0; break; default: error = EINVAL; } done: return (error); } static void aio_kick_nowait(struct proc *userp) { struct kaioinfo *ki = userp->p_aioinfo; struct aioproc *aiop; mtx_assert(&aio_job_mtx, MA_OWNED); if ((aiop = TAILQ_FIRST(&aio_freeproc)) != NULL) { TAILQ_REMOVE(&aio_freeproc, aiop, list); aiop->aioprocflags &= ~AIOP_FREE; wakeup(aiop->aioproc); } else if (num_aio_resv_start + num_aio_procs < max_aio_procs && ki->kaio_active_count + num_aio_resv_start < ki->kaio_maxactive_count) { taskqueue_enqueue(taskqueue_aiod_kick, &ki->kaio_task); } } static int aio_kick(struct proc *userp) { struct kaioinfo *ki = userp->p_aioinfo; struct aioproc *aiop; int error, ret = 0; mtx_assert(&aio_job_mtx, MA_OWNED); retryproc: if ((aiop = TAILQ_FIRST(&aio_freeproc)) != NULL) { TAILQ_REMOVE(&aio_freeproc, aiop, list); aiop->aioprocflags &= ~AIOP_FREE; wakeup(aiop->aioproc); } else if (num_aio_resv_start + num_aio_procs < max_aio_procs && ki->kaio_active_count + num_aio_resv_start < ki->kaio_maxactive_count) { num_aio_resv_start++; mtx_unlock(&aio_job_mtx); error = aio_newproc(&num_aio_resv_start); mtx_lock(&aio_job_mtx); if (error) { num_aio_resv_start--; goto retryproc; } } else { ret = -1; } return (ret); } static void aio_kick_helper(void *context, int pending) { struct proc *userp = context; mtx_lock(&aio_job_mtx); while (--pending >= 0) { if (aio_kick(userp)) break; } mtx_unlock(&aio_job_mtx); } /* * Support the aio_return system call, as a side-effect, kernel resources are * released. */ static int kern_aio_return(struct thread *td, struct aiocb *ujob, struct aiocb_ops *ops) { struct proc *p = td->td_proc; struct kaiocb *job; struct kaioinfo *ki; - int status, error; + long status, error; ki = p->p_aioinfo; if (ki == NULL) return (EINVAL); AIO_LOCK(ki); TAILQ_FOREACH(job, &ki->kaio_done, plist) { if (job->ujob == ujob) break; } if (job != NULL) { MPASS(job->jobflags & KAIOCB_FINISHED); status = job->uaiocb._aiocb_private.status; error = job->uaiocb._aiocb_private.error; td->td_retval[0] = status; if (job->uaiocb.aio_lio_opcode == LIO_WRITE) { td->td_ru.ru_oublock += job->outputcharge; job->outputcharge = 0; } else if (job->uaiocb.aio_lio_opcode == LIO_READ) { td->td_ru.ru_inblock += job->inputcharge; job->inputcharge = 0; } aio_free_entry(job); AIO_UNLOCK(ki); ops->store_error(ujob, error); ops->store_status(ujob, status); } else { error = EINVAL; AIO_UNLOCK(ki); } return (error); } int sys_aio_return(struct thread *td, struct aio_return_args *uap) { return (kern_aio_return(td, uap->aiocbp, &aiocb_ops)); } /* * Allow a process to wakeup when any of the I/O requests are completed. */ static int kern_aio_suspend(struct thread *td, int njoblist, struct aiocb **ujoblist, struct timespec *ts) { struct proc *p = td->td_proc; struct timeval atv; struct kaioinfo *ki; struct kaiocb *firstjob, *job; int error, i, timo; timo = 0; if (ts) { if (ts->tv_nsec < 0 || ts->tv_nsec >= 1000000000) return (EINVAL); TIMESPEC_TO_TIMEVAL(&atv, ts); if (itimerfix(&atv)) return (EINVAL); timo = tvtohz(&atv); } ki = p->p_aioinfo; if (ki == NULL) return (EAGAIN); if (njoblist == 0) return (0); AIO_LOCK(ki); for (;;) { firstjob = NULL; error = 0; TAILQ_FOREACH(job, &ki->kaio_all, allist) { for (i = 0; i < njoblist; i++) { if (job->ujob == ujoblist[i]) { if (firstjob == NULL) firstjob = job; if (job->jobflags & KAIOCB_FINISHED) goto RETURN; } } } /* All tasks were finished. */ if (firstjob == NULL) break; ki->kaio_flags |= KAIO_WAKEUP; error = msleep(&p->p_aioinfo, AIO_MTX(ki), PRIBIO | PCATCH, "aiospn", timo); if (error == ERESTART) error = EINTR; if (error) break; } RETURN: AIO_UNLOCK(ki); return (error); } int sys_aio_suspend(struct thread *td, struct aio_suspend_args *uap) { struct timespec ts, *tsp; struct aiocb **ujoblist; int error; if (uap->nent < 0 || uap->nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->timeout) { /* Get timespec struct. */ if ((error = copyin(uap->timeout, &ts, sizeof(ts))) != 0) return (error); tsp = &ts; } else tsp = NULL; ujoblist = uma_zalloc(aiol_zone, M_WAITOK); error = copyin(uap->aiocbp, ujoblist, uap->nent * sizeof(ujoblist[0])); if (error == 0) error = kern_aio_suspend(td, uap->nent, ujoblist, tsp); uma_zfree(aiol_zone, ujoblist); return (error); } /* * aio_cancel cancels any non-physio aio operations not currently in * progress. */ int sys_aio_cancel(struct thread *td, struct aio_cancel_args *uap) { struct proc *p = td->td_proc; struct kaioinfo *ki; struct kaiocb *job, *jobn; struct file *fp; cap_rights_t rights; int error; int cancelled = 0; int notcancelled = 0; struct vnode *vp; /* Lookup file object. */ error = fget(td, uap->fd, cap_rights_init(&rights), &fp); if (error) return (error); ki = p->p_aioinfo; if (ki == NULL) goto done; if (fp->f_type == DTYPE_VNODE) { vp = fp->f_vnode; if (vn_isdisk(vp, &error)) { fdrop(fp, td); td->td_retval[0] = AIO_NOTCANCELED; return (0); } } AIO_LOCK(ki); TAILQ_FOREACH_SAFE(job, &ki->kaio_jobqueue, plist, jobn) { if ((uap->fd == job->uaiocb.aio_fildes) && ((uap->aiocbp == NULL) || (uap->aiocbp == job->ujob))) { if (aio_cancel_job(p, ki, job)) { cancelled++; } else { notcancelled++; } if (uap->aiocbp != NULL) break; } } AIO_UNLOCK(ki); done: fdrop(fp, td); if (uap->aiocbp != NULL) { if (cancelled) { td->td_retval[0] = AIO_CANCELED; return (0); } } if (notcancelled) { td->td_retval[0] = AIO_NOTCANCELED; return (0); } if (cancelled) { td->td_retval[0] = AIO_CANCELED; return (0); } td->td_retval[0] = AIO_ALLDONE; return (0); } /* * aio_error is implemented in the kernel level for compatibility purposes * only. For a user mode async implementation, it would be best to do it in * a userland subroutine. */ static int kern_aio_error(struct thread *td, struct aiocb *ujob, struct aiocb_ops *ops) { struct proc *p = td->td_proc; struct kaiocb *job; struct kaioinfo *ki; int status; ki = p->p_aioinfo; if (ki == NULL) { td->td_retval[0] = EINVAL; return (0); } AIO_LOCK(ki); TAILQ_FOREACH(job, &ki->kaio_all, allist) { if (job->ujob == ujob) { if (job->jobflags & KAIOCB_FINISHED) td->td_retval[0] = job->uaiocb._aiocb_private.error; else td->td_retval[0] = EINPROGRESS; AIO_UNLOCK(ki); return (0); } } AIO_UNLOCK(ki); /* * Hack for failure of aio_aqueue. */ status = ops->fetch_status(ujob); if (status == -1) { td->td_retval[0] = ops->fetch_error(ujob); return (0); } td->td_retval[0] = EINVAL; return (0); } int sys_aio_error(struct thread *td, struct aio_error_args *uap) { return (kern_aio_error(td, uap->aiocbp, &aiocb_ops)); } /* syscall - asynchronous read from a file (REALTIME) */ #ifdef COMPAT_FREEBSD6 int freebsd6_aio_read(struct thread *td, struct freebsd6_aio_read_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_READ, &aiocb_ops_osigevent)); } #endif int sys_aio_read(struct thread *td, struct aio_read_args *uap) { return (aio_aqueue(td, uap->aiocbp, NULL, LIO_READ, &aiocb_ops)); } /* syscall - asynchronous write to a file (REALTIME) */ #ifdef COMPAT_FREEBSD6 int freebsd6_aio_write(struct thread *td, struct freebsd6_aio_write_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_WRITE, &aiocb_ops_osigevent)); } #endif int sys_aio_write(struct thread *td, struct aio_write_args *uap) { return (aio_aqueue(td, uap->aiocbp, NULL, LIO_WRITE, &aiocb_ops)); } int sys_aio_mlock(struct thread *td, struct aio_mlock_args *uap) { return (aio_aqueue(td, uap->aiocbp, NULL, LIO_MLOCK, &aiocb_ops)); } static int kern_lio_listio(struct thread *td, int mode, struct aiocb * const *uacb_list, struct aiocb **acb_list, int nent, struct sigevent *sig, struct aiocb_ops *ops) { struct proc *p = td->td_proc; struct aiocb *job; struct kaioinfo *ki; struct aioliojob *lj; struct kevent kev; int error; int nerror; int i; if ((mode != LIO_NOWAIT) && (mode != LIO_WAIT)) return (EINVAL); if (nent < 0 || nent > AIO_LISTIO_MAX) return (EINVAL); if (p->p_aioinfo == NULL) aio_init_aioinfo(p); ki = p->p_aioinfo; lj = uma_zalloc(aiolio_zone, M_WAITOK); lj->lioj_flags = 0; lj->lioj_count = 0; lj->lioj_finished_count = 0; knlist_init_mtx(&lj->klist, AIO_MTX(ki)); ksiginfo_init(&lj->lioj_ksi); /* * Setup signal. */ if (sig && (mode == LIO_NOWAIT)) { bcopy(sig, &lj->lioj_signal, sizeof(lj->lioj_signal)); if (lj->lioj_signal.sigev_notify == SIGEV_KEVENT) { /* Assume only new style KEVENT */ kev.filter = EVFILT_LIO; kev.flags = EV_ADD | EV_ENABLE | EV_FLAG1; kev.ident = (uintptr_t)uacb_list; /* something unique */ kev.data = (intptr_t)lj; /* pass user defined sigval data */ kev.udata = lj->lioj_signal.sigev_value.sival_ptr; error = kqfd_register( lj->lioj_signal.sigev_notify_kqueue, &kev, td, 1); if (error) { uma_zfree(aiolio_zone, lj); return (error); } } else if (lj->lioj_signal.sigev_notify == SIGEV_NONE) { ; } else if (lj->lioj_signal.sigev_notify == SIGEV_SIGNAL || lj->lioj_signal.sigev_notify == SIGEV_THREAD_ID) { if (!_SIG_VALID(lj->lioj_signal.sigev_signo)) { uma_zfree(aiolio_zone, lj); return EINVAL; } lj->lioj_flags |= LIOJ_SIGNAL; } else { uma_zfree(aiolio_zone, lj); return EINVAL; } } AIO_LOCK(ki); TAILQ_INSERT_TAIL(&ki->kaio_liojoblist, lj, lioj_list); /* * Add extra aiocb count to avoid the lio to be freed * by other threads doing aio_waitcomplete or aio_return, * and prevent event from being sent until we have queued * all tasks. */ lj->lioj_count = 1; AIO_UNLOCK(ki); /* * Get pointers to the list of I/O requests. */ nerror = 0; for (i = 0; i < nent; i++) { job = acb_list[i]; if (job != NULL) { error = aio_aqueue(td, job, lj, LIO_NOP, ops); if (error != 0) nerror++; } } error = 0; AIO_LOCK(ki); if (mode == LIO_WAIT) { while (lj->lioj_count - 1 != lj->lioj_finished_count) { ki->kaio_flags |= KAIO_WAKEUP; error = msleep(&p->p_aioinfo, AIO_MTX(ki), PRIBIO | PCATCH, "aiospn", 0); if (error == ERESTART) error = EINTR; if (error) break; } } else { if (lj->lioj_count - 1 == lj->lioj_finished_count) { if (lj->lioj_signal.sigev_notify == SIGEV_KEVENT) { lj->lioj_flags |= LIOJ_KEVENT_POSTED; KNOTE_LOCKED(&lj->klist, 1); } if ((lj->lioj_flags & (LIOJ_SIGNAL|LIOJ_SIGNAL_POSTED)) == LIOJ_SIGNAL && (lj->lioj_signal.sigev_notify == SIGEV_SIGNAL || lj->lioj_signal.sigev_notify == SIGEV_THREAD_ID)) { aio_sendsig(p, &lj->lioj_signal, &lj->lioj_ksi); lj->lioj_flags |= LIOJ_SIGNAL_POSTED; } } } lj->lioj_count--; if (lj->lioj_count == 0) { TAILQ_REMOVE(&ki->kaio_liojoblist, lj, lioj_list); knlist_delete(&lj->klist, curthread, 1); PROC_LOCK(p); sigqueue_take(&lj->lioj_ksi); PROC_UNLOCK(p); AIO_UNLOCK(ki); uma_zfree(aiolio_zone, lj); } else AIO_UNLOCK(ki); if (nerror) return (EIO); return (error); } /* syscall - list directed I/O (REALTIME) */ #ifdef COMPAT_FREEBSD6 int freebsd6_lio_listio(struct thread *td, struct freebsd6_lio_listio_args *uap) { struct aiocb **acb_list; struct sigevent *sigp, sig; struct osigevent osig; int error, nent; if ((uap->mode != LIO_NOWAIT) && (uap->mode != LIO_WAIT)) return (EINVAL); nent = uap->nent; if (nent < 0 || nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->sig && (uap->mode == LIO_NOWAIT)) { error = copyin(uap->sig, &osig, sizeof(osig)); if (error) return (error); error = convert_old_sigevent(&osig, &sig); if (error) return (error); sigp = &sig; } else sigp = NULL; acb_list = malloc(sizeof(struct aiocb *) * nent, M_LIO, M_WAITOK); error = copyin(uap->acb_list, acb_list, nent * sizeof(acb_list[0])); if (error == 0) error = kern_lio_listio(td, uap->mode, (struct aiocb * const *)uap->acb_list, acb_list, nent, sigp, &aiocb_ops_osigevent); free(acb_list, M_LIO); return (error); } #endif /* syscall - list directed I/O (REALTIME) */ int sys_lio_listio(struct thread *td, struct lio_listio_args *uap) { struct aiocb **acb_list; struct sigevent *sigp, sig; int error, nent; if ((uap->mode != LIO_NOWAIT) && (uap->mode != LIO_WAIT)) return (EINVAL); nent = uap->nent; if (nent < 0 || nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->sig && (uap->mode == LIO_NOWAIT)) { error = copyin(uap->sig, &sig, sizeof(sig)); if (error) return (error); sigp = &sig; } else sigp = NULL; acb_list = malloc(sizeof(struct aiocb *) * nent, M_LIO, M_WAITOK); error = copyin(uap->acb_list, acb_list, nent * sizeof(acb_list[0])); if (error == 0) error = kern_lio_listio(td, uap->mode, uap->acb_list, acb_list, nent, sigp, &aiocb_ops); free(acb_list, M_LIO); return (error); } static void aio_physwakeup(struct bio *bp) { struct kaiocb *job = (struct kaiocb *)bp->bio_caller1; struct proc *userp; struct kaioinfo *ki; size_t nbytes; int error, nblks; /* Release mapping into kernel space. */ userp = job->userproc; ki = userp->p_aioinfo; if (job->pbuf) { pmap_qremove((vm_offset_t)job->pbuf->b_data, job->npages); relpbuf(job->pbuf, NULL); job->pbuf = NULL; atomic_subtract_int(&num_buf_aio, 1); AIO_LOCK(ki); ki->kaio_buffer_count--; AIO_UNLOCK(ki); } vm_page_unhold_pages(job->pages, job->npages); bp = job->bp; job->bp = NULL; nbytes = job->uaiocb.aio_nbytes - bp->bio_resid; error = 0; if (bp->bio_flags & BIO_ERROR) error = bp->bio_error; nblks = btodb(nbytes); if (job->uaiocb.aio_lio_opcode == LIO_WRITE) job->outputcharge += nblks; else job->inputcharge += nblks; aio_complete(job, nbytes, error); g_destroy_bio(bp); } /* syscall - wait for the next completion of an aio request */ static int kern_aio_waitcomplete(struct thread *td, struct aiocb **ujobp, struct timespec *ts, struct aiocb_ops *ops) { struct proc *p = td->td_proc; struct timeval atv; struct kaioinfo *ki; struct kaiocb *job; struct aiocb *ujob; - int error, status, timo; + long error, status; + int timo; ops->store_aiocb(ujobp, NULL); if (ts == NULL) { timo = 0; } else if (ts->tv_sec == 0 && ts->tv_nsec == 0) { timo = -1; } else { if ((ts->tv_nsec < 0) || (ts->tv_nsec >= 1000000000)) return (EINVAL); TIMESPEC_TO_TIMEVAL(&atv, ts); if (itimerfix(&atv)) return (EINVAL); timo = tvtohz(&atv); } if (p->p_aioinfo == NULL) aio_init_aioinfo(p); ki = p->p_aioinfo; error = 0; job = NULL; AIO_LOCK(ki); while ((job = TAILQ_FIRST(&ki->kaio_done)) == NULL) { if (timo == -1) { error = EWOULDBLOCK; break; } ki->kaio_flags |= KAIO_WAKEUP; error = msleep(&p->p_aioinfo, AIO_MTX(ki), PRIBIO | PCATCH, "aiowc", timo); if (timo && error == ERESTART) error = EINTR; if (error) break; } if (job != NULL) { MPASS(job->jobflags & KAIOCB_FINISHED); ujob = job->ujob; status = job->uaiocb._aiocb_private.status; error = job->uaiocb._aiocb_private.error; td->td_retval[0] = status; if (job->uaiocb.aio_lio_opcode == LIO_WRITE) { td->td_ru.ru_oublock += job->outputcharge; job->outputcharge = 0; } else if (job->uaiocb.aio_lio_opcode == LIO_READ) { td->td_ru.ru_inblock += job->inputcharge; job->inputcharge = 0; } aio_free_entry(job); AIO_UNLOCK(ki); ops->store_aiocb(ujobp, ujob); ops->store_error(ujob, error); ops->store_status(ujob, status); } else AIO_UNLOCK(ki); return (error); } int sys_aio_waitcomplete(struct thread *td, struct aio_waitcomplete_args *uap) { struct timespec ts, *tsp; int error; if (uap->timeout) { /* Get timespec struct. */ error = copyin(uap->timeout, &ts, sizeof(ts)); if (error) return (error); tsp = &ts; } else tsp = NULL; return (kern_aio_waitcomplete(td, uap->aiocbp, tsp, &aiocb_ops)); } static int kern_aio_fsync(struct thread *td, int op, struct aiocb *ujob, struct aiocb_ops *ops) { struct proc *p = td->td_proc; struct kaioinfo *ki; if (op != O_SYNC) /* XXX lack of O_DSYNC */ return (EINVAL); ki = p->p_aioinfo; if (ki == NULL) aio_init_aioinfo(p); return (aio_aqueue(td, ujob, NULL, LIO_SYNC, ops)); } int sys_aio_fsync(struct thread *td, struct aio_fsync_args *uap) { return (kern_aio_fsync(td, uap->op, uap->aiocbp, &aiocb_ops)); } /* kqueue attach function */ static int filt_aioattach(struct knote *kn) { struct kaiocb *job = (struct kaiocb *)kn->kn_sdata; /* * The job pointer must be validated before using it, so * registration is restricted to the kernel; the user cannot * set EV_FLAG1. */ if ((kn->kn_flags & EV_FLAG1) == 0) return (EPERM); kn->kn_ptr.p_aio = job; kn->kn_flags &= ~EV_FLAG1; knlist_add(&job->klist, kn, 0); return (0); } /* kqueue detach function */ static void filt_aiodetach(struct knote *kn) { struct knlist *knl; knl = &kn->kn_ptr.p_aio->klist; knl->kl_lock(knl->kl_lockarg); if (!knlist_empty(knl)) knlist_remove(knl, kn, 1); knl->kl_unlock(knl->kl_lockarg); } /* kqueue filter function */ /*ARGSUSED*/ static int filt_aio(struct knote *kn, long hint) { struct kaiocb *job = kn->kn_ptr.p_aio; kn->kn_data = job->uaiocb._aiocb_private.error; if (!(job->jobflags & KAIOCB_FINISHED)) return (0); kn->kn_flags |= EV_EOF; return (1); } /* kqueue attach function */ static int filt_lioattach(struct knote *kn) { struct aioliojob * lj = (struct aioliojob *)kn->kn_sdata; /* * The aioliojob pointer must be validated before using it, so * registration is restricted to the kernel; the user cannot * set EV_FLAG1. */ if ((kn->kn_flags & EV_FLAG1) == 0) return (EPERM); kn->kn_ptr.p_lio = lj; kn->kn_flags &= ~EV_FLAG1; knlist_add(&lj->klist, kn, 0); return (0); } /* kqueue detach function */ static void filt_liodetach(struct knote *kn) { struct knlist *knl; knl = &kn->kn_ptr.p_lio->klist; knl->kl_lock(knl->kl_lockarg); if (!knlist_empty(knl)) knlist_remove(knl, kn, 1); knl->kl_unlock(knl->kl_lockarg); } /* kqueue filter function */ /*ARGSUSED*/ static int filt_lio(struct knote *kn, long hint) { struct aioliojob * lj = kn->kn_ptr.p_lio; return (lj->lioj_flags & LIOJ_KEVENT_POSTED); } #ifdef COMPAT_FREEBSD32 #include #include #include #include #include #include #include struct __aiocb_private32 { int32_t status; int32_t error; uint32_t kernelinfo; }; #ifdef COMPAT_FREEBSD6 typedef struct oaiocb32 { int aio_fildes; /* File descriptor */ uint64_t aio_offset __packed; /* File offset for I/O */ uint32_t aio_buf; /* I/O buffer in process space */ uint32_t aio_nbytes; /* Number of bytes for I/O */ struct osigevent32 aio_sigevent; /* Signal to deliver */ int aio_lio_opcode; /* LIO opcode */ int aio_reqprio; /* Request priority -- ignored */ struct __aiocb_private32 _aiocb_private; } oaiocb32_t; #endif typedef struct aiocb32 { int32_t aio_fildes; /* File descriptor */ uint64_t aio_offset __packed; /* File offset for I/O */ uint32_t aio_buf; /* I/O buffer in process space */ uint32_t aio_nbytes; /* Number of bytes for I/O */ int __spare__[2]; uint32_t __spare2__; int aio_lio_opcode; /* LIO opcode */ int aio_reqprio; /* Request priority -- ignored */ struct __aiocb_private32 _aiocb_private; struct sigevent32 aio_sigevent; /* Signal to deliver */ } aiocb32_t; #ifdef COMPAT_FREEBSD6 static int convert_old_sigevent32(struct osigevent32 *osig, struct sigevent *nsig) { /* * Only SIGEV_NONE, SIGEV_SIGNAL, and SIGEV_KEVENT are * supported by AIO with the old sigevent structure. */ CP(*osig, *nsig, sigev_notify); switch (nsig->sigev_notify) { case SIGEV_NONE: break; case SIGEV_SIGNAL: nsig->sigev_signo = osig->__sigev_u.__sigev_signo; break; case SIGEV_KEVENT: nsig->sigev_notify_kqueue = osig->__sigev_u.__sigev_notify_kqueue; PTRIN_CP(*osig, *nsig, sigev_value.sival_ptr); break; default: return (EINVAL); } return (0); } static int aiocb32_copyin_old_sigevent(struct aiocb *ujob, struct aiocb *kjob) { struct oaiocb32 job32; int error; bzero(kjob, sizeof(struct aiocb)); error = copyin(ujob, &job32, sizeof(job32)); if (error) return (error); CP(job32, *kjob, aio_fildes); CP(job32, *kjob, aio_offset); PTRIN_CP(job32, *kjob, aio_buf); CP(job32, *kjob, aio_nbytes); CP(job32, *kjob, aio_lio_opcode); CP(job32, *kjob, aio_reqprio); CP(job32, *kjob, _aiocb_private.status); CP(job32, *kjob, _aiocb_private.error); PTRIN_CP(job32, *kjob, _aiocb_private.kernelinfo); return (convert_old_sigevent32(&job32.aio_sigevent, &kjob->aio_sigevent)); } #endif static int aiocb32_copyin(struct aiocb *ujob, struct aiocb *kjob) { struct aiocb32 job32; int error; error = copyin(ujob, &job32, sizeof(job32)); if (error) return (error); CP(job32, *kjob, aio_fildes); CP(job32, *kjob, aio_offset); PTRIN_CP(job32, *kjob, aio_buf); CP(job32, *kjob, aio_nbytes); CP(job32, *kjob, aio_lio_opcode); CP(job32, *kjob, aio_reqprio); CP(job32, *kjob, _aiocb_private.status); CP(job32, *kjob, _aiocb_private.error); PTRIN_CP(job32, *kjob, _aiocb_private.kernelinfo); return (convert_sigevent32(&job32.aio_sigevent, &kjob->aio_sigevent)); } static long aiocb32_fetch_status(struct aiocb *ujob) { struct aiocb32 *ujob32; ujob32 = (struct aiocb32 *)ujob; return (fuword32(&ujob32->_aiocb_private.status)); } static long aiocb32_fetch_error(struct aiocb *ujob) { struct aiocb32 *ujob32; ujob32 = (struct aiocb32 *)ujob; return (fuword32(&ujob32->_aiocb_private.error)); } static int aiocb32_store_status(struct aiocb *ujob, long status) { struct aiocb32 *ujob32; ujob32 = (struct aiocb32 *)ujob; return (suword32(&ujob32->_aiocb_private.status, status)); } static int aiocb32_store_error(struct aiocb *ujob, long error) { struct aiocb32 *ujob32; ujob32 = (struct aiocb32 *)ujob; return (suword32(&ujob32->_aiocb_private.error, error)); } static int aiocb32_store_kernelinfo(struct aiocb *ujob, long jobref) { struct aiocb32 *ujob32; ujob32 = (struct aiocb32 *)ujob; return (suword32(&ujob32->_aiocb_private.kernelinfo, jobref)); } static int aiocb32_store_aiocb(struct aiocb **ujobp, struct aiocb *ujob) { return (suword32(ujobp, (long)ujob)); } static struct aiocb_ops aiocb32_ops = { .copyin = aiocb32_copyin, .fetch_status = aiocb32_fetch_status, .fetch_error = aiocb32_fetch_error, .store_status = aiocb32_store_status, .store_error = aiocb32_store_error, .store_kernelinfo = aiocb32_store_kernelinfo, .store_aiocb = aiocb32_store_aiocb, }; #ifdef COMPAT_FREEBSD6 static struct aiocb_ops aiocb32_ops_osigevent = { .copyin = aiocb32_copyin_old_sigevent, .fetch_status = aiocb32_fetch_status, .fetch_error = aiocb32_fetch_error, .store_status = aiocb32_store_status, .store_error = aiocb32_store_error, .store_kernelinfo = aiocb32_store_kernelinfo, .store_aiocb = aiocb32_store_aiocb, }; #endif int freebsd32_aio_return(struct thread *td, struct freebsd32_aio_return_args *uap) { return (kern_aio_return(td, (struct aiocb *)uap->aiocbp, &aiocb32_ops)); } int freebsd32_aio_suspend(struct thread *td, struct freebsd32_aio_suspend_args *uap) { struct timespec32 ts32; struct timespec ts, *tsp; struct aiocb **ujoblist; uint32_t *ujoblist32; int error, i; if (uap->nent < 0 || uap->nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->timeout) { /* Get timespec struct. */ if ((error = copyin(uap->timeout, &ts32, sizeof(ts32))) != 0) return (error); CP(ts32, ts, tv_sec); CP(ts32, ts, tv_nsec); tsp = &ts; } else tsp = NULL; ujoblist = uma_zalloc(aiol_zone, M_WAITOK); ujoblist32 = (uint32_t *)ujoblist; error = copyin(uap->aiocbp, ujoblist32, uap->nent * sizeof(ujoblist32[0])); if (error == 0) { for (i = uap->nent; i > 0; i--) ujoblist[i] = PTRIN(ujoblist32[i]); error = kern_aio_suspend(td, uap->nent, ujoblist, tsp); } uma_zfree(aiol_zone, ujoblist); return (error); } int freebsd32_aio_error(struct thread *td, struct freebsd32_aio_error_args *uap) { return (kern_aio_error(td, (struct aiocb *)uap->aiocbp, &aiocb32_ops)); } #ifdef COMPAT_FREEBSD6 int freebsd6_freebsd32_aio_read(struct thread *td, struct freebsd6_freebsd32_aio_read_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_READ, &aiocb32_ops_osigevent)); } #endif int freebsd32_aio_read(struct thread *td, struct freebsd32_aio_read_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_READ, &aiocb32_ops)); } #ifdef COMPAT_FREEBSD6 int freebsd6_freebsd32_aio_write(struct thread *td, struct freebsd6_freebsd32_aio_write_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_WRITE, &aiocb32_ops_osigevent)); } #endif int freebsd32_aio_write(struct thread *td, struct freebsd32_aio_write_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_WRITE, &aiocb32_ops)); } int freebsd32_aio_mlock(struct thread *td, struct freebsd32_aio_mlock_args *uap) { return (aio_aqueue(td, (struct aiocb *)uap->aiocbp, NULL, LIO_MLOCK, &aiocb32_ops)); } int freebsd32_aio_waitcomplete(struct thread *td, struct freebsd32_aio_waitcomplete_args *uap) { struct timespec32 ts32; struct timespec ts, *tsp; int error; if (uap->timeout) { /* Get timespec struct. */ error = copyin(uap->timeout, &ts32, sizeof(ts32)); if (error) return (error); CP(ts32, ts, tv_sec); CP(ts32, ts, tv_nsec); tsp = &ts; } else tsp = NULL; return (kern_aio_waitcomplete(td, (struct aiocb **)uap->aiocbp, tsp, &aiocb32_ops)); } int freebsd32_aio_fsync(struct thread *td, struct freebsd32_aio_fsync_args *uap) { return (kern_aio_fsync(td, uap->op, (struct aiocb *)uap->aiocbp, &aiocb32_ops)); } #ifdef COMPAT_FREEBSD6 int freebsd6_freebsd32_lio_listio(struct thread *td, struct freebsd6_freebsd32_lio_listio_args *uap) { struct aiocb **acb_list; struct sigevent *sigp, sig; struct osigevent32 osig; uint32_t *acb_list32; int error, i, nent; if ((uap->mode != LIO_NOWAIT) && (uap->mode != LIO_WAIT)) return (EINVAL); nent = uap->nent; if (nent < 0 || nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->sig && (uap->mode == LIO_NOWAIT)) { error = copyin(uap->sig, &osig, sizeof(osig)); if (error) return (error); error = convert_old_sigevent32(&osig, &sig); if (error) return (error); sigp = &sig; } else sigp = NULL; acb_list32 = malloc(sizeof(uint32_t) * nent, M_LIO, M_WAITOK); error = copyin(uap->acb_list, acb_list32, nent * sizeof(uint32_t)); if (error) { free(acb_list32, M_LIO); return (error); } acb_list = malloc(sizeof(struct aiocb *) * nent, M_LIO, M_WAITOK); for (i = 0; i < nent; i++) acb_list[i] = PTRIN(acb_list32[i]); free(acb_list32, M_LIO); error = kern_lio_listio(td, uap->mode, (struct aiocb * const *)uap->acb_list, acb_list, nent, sigp, &aiocb32_ops_osigevent); free(acb_list, M_LIO); return (error); } #endif int freebsd32_lio_listio(struct thread *td, struct freebsd32_lio_listio_args *uap) { struct aiocb **acb_list; struct sigevent *sigp, sig; struct sigevent32 sig32; uint32_t *acb_list32; int error, i, nent; if ((uap->mode != LIO_NOWAIT) && (uap->mode != LIO_WAIT)) return (EINVAL); nent = uap->nent; if (nent < 0 || nent > AIO_LISTIO_MAX) return (EINVAL); if (uap->sig && (uap->mode == LIO_NOWAIT)) { error = copyin(uap->sig, &sig32, sizeof(sig32)); if (error) return (error); error = convert_sigevent32(&sig32, &sig); if (error) return (error); sigp = &sig; } else sigp = NULL; acb_list32 = malloc(sizeof(uint32_t) * nent, M_LIO, M_WAITOK); error = copyin(uap->acb_list, acb_list32, nent * sizeof(uint32_t)); if (error) { free(acb_list32, M_LIO); return (error); } acb_list = malloc(sizeof(struct aiocb *) * nent, M_LIO, M_WAITOK); for (i = 0; i < nent; i++) acb_list[i] = PTRIN(acb_list32[i]); free(acb_list32, M_LIO); error = kern_lio_listio(td, uap->mode, (struct aiocb * const *)uap->acb_list, acb_list, nent, sigp, &aiocb32_ops); free(acb_list, M_LIO); return (error); } #endif diff --git a/sys/sys/aio.h b/sys/sys/aio.h index 25ffb280e9e3..0abee2d5d8a9 100644 --- a/sys/sys/aio.h +++ b/sys/sys/aio.h @@ -1,249 +1,249 @@ /*- * Copyright (c) 1997 John S. Dyson. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. John S. Dyson's name may not be used to endorse or promote products * derived from this software without specific prior written permission. * * DISCLAIMER: This code isn't warranted to do anything useful. Anything * bad that happens because of using this software isn't the responsibility * of the author. This software is distributed AS-IS. * * $FreeBSD$ */ #ifndef _SYS_AIO_H_ #define _SYS_AIO_H_ #include #include #ifdef _KERNEL #include #include #include #endif /* * Returned by aio_cancel: */ #define AIO_CANCELED 0x1 #define AIO_NOTCANCELED 0x2 #define AIO_ALLDONE 0x3 /* * LIO opcodes */ #define LIO_NOP 0x0 #define LIO_WRITE 0x1 #define LIO_READ 0x2 #ifdef _KERNEL #define LIO_SYNC 0x3 #define LIO_MLOCK 0x4 #endif /* * LIO modes */ #define LIO_NOWAIT 0x0 #define LIO_WAIT 0x1 /* * Maximum number of allowed LIO operations */ #define AIO_LISTIO_MAX 16 #ifdef _KERNEL /* Default values of tunables for the AIO worker pool. */ #ifndef MAX_AIO_PROCS #define MAX_AIO_PROCS 32 #endif #ifndef TARGET_AIO_PROCS #define TARGET_AIO_PROCS 4 #endif #ifndef AIOD_LIFETIME_DEFAULT #define AIOD_LIFETIME_DEFAULT (30 * hz) #endif #endif /* * Private members for aiocb -- don't access * directly. */ struct __aiocb_private { long status; long error; void *kernelinfo; }; /* * I/O control block */ typedef struct aiocb { int aio_fildes; /* File descriptor */ off_t aio_offset; /* File offset for I/O */ volatile void *aio_buf; /* I/O buffer in process space */ size_t aio_nbytes; /* Number of bytes for I/O */ int __spare__[2]; void *__spare2__; int aio_lio_opcode; /* LIO opcode */ int aio_reqprio; /* Request priority -- ignored */ struct __aiocb_private _aiocb_private; struct sigevent aio_sigevent; /* Signal to deliver */ } aiocb_t; #ifdef _KERNEL typedef void aio_cancel_fn_t(struct kaiocb *); typedef void aio_handle_fn_t(struct kaiocb *); /* * Kernel version of an I/O control block. * * Locking key: * * - need not protected * a - locked by kaioinfo lock * b - locked by backend lock * c - locked by aio_job_mtx */ struct kaiocb { TAILQ_ENTRY(kaiocb) list; /* (b) backend-specific list of jobs */ TAILQ_ENTRY(kaiocb) plist; /* (a) lists of pending / done jobs */ TAILQ_ENTRY(kaiocb) allist; /* (a) list of all jobs in proc */ int jobflags; /* (a) job flags */ int inputcharge; /* (*) input blocks */ int outputcharge; /* (*) output blocks */ struct bio *bp; /* (*) BIO backend BIO pointer */ struct buf *pbuf; /* (*) BIO backend buffer pointer */ struct vm_page *pages[btoc(MAXPHYS)+1]; /* BIO backend pages */ int npages; /* BIO backend number of pages */ struct proc *userproc; /* (*) user process */ struct ucred *cred; /* (*) active credential when created */ struct file *fd_file; /* (*) pointer to file structure */ struct aioliojob *lio; /* (*) optional lio job */ struct aiocb *ujob; /* (*) pointer in userspace of aiocb */ struct knlist klist; /* (a) list of knotes */ struct aiocb uaiocb; /* (*) copy of user I/O control block */ ksiginfo_t ksi; /* (a) realtime signal info */ uint64_t seqno; /* (*) job number */ int pending; /* (a) number of pending I/O, aio_fsync only */ aio_cancel_fn_t *cancel_fn; /* (a) backend cancel function */ aio_handle_fn_t *handle_fn; /* (c) backend handle function */ }; struct socket; struct sockbuf; /* * AIO backends should permit cancellation of queued requests waiting to * be serviced by installing a cancel routine while the request is * queued. The cancellation routine should dequeue the request if * necessary and cancel it. Care must be used to handle races between * queueing and dequeueing requests and cancellation. * * When queueing a request somewhere such that it can be cancelled, the * caller should: * * 1) Acquire lock that protects the associated queue. * 2) Call aio_set_cancel_function() to install the cancel routine. * 3) If that fails, the request has a pending cancel and should be * cancelled via aio_cancel(). * 4) Queue the request. * * When dequeueing a request to service it or hand it off to somewhere else, * the caller should: * * 1) Acquire the lock that protects the associated queue. * 2) Dequeue the request. * 3) Call aio_clear_cancel_function() to clear the cancel routine. * 4) If that fails, the cancel routine is about to be called. The * caller should ignore the request. * * The cancel routine should: * * 1) Acquire the lock that protects the associated queue. * 2) Call aio_cancel_cleared() to determine if the request is already * dequeued due to a race with dequeueing thread. * 3) If that fails, dequeue the request. * 4) Cancel the request via aio_cancel(). */ bool aio_cancel_cleared(struct kaiocb *job); void aio_cancel(struct kaiocb *job); bool aio_clear_cancel_function(struct kaiocb *job); void aio_complete(struct kaiocb *job, long status, int error); void aio_schedule(struct kaiocb *job, aio_handle_fn_t *func); bool aio_set_cancel_function(struct kaiocb *job, aio_cancel_fn_t *func); void aio_switch_vmspace(struct kaiocb *job); #else /* !_KERNEL */ struct timespec; __BEGIN_DECLS /* * Asynchronously read from a file */ int aio_read(struct aiocb *); /* * Asynchronously write to file */ int aio_write(struct aiocb *); /* * List I/O Asynchronously/synchronously read/write to/from file * "lio_mode" specifies whether or not the I/O is synchronous. * "acb_list" is an array of "nacb_listent" I/O control blocks. * when all I/Os are complete, the optional signal "sig" is sent. */ int lio_listio(int, struct aiocb * const [], int, struct sigevent *); /* * Get completion status * returns EINPROGRESS until I/O is complete. * this routine does not block. */ int aio_error(const struct aiocb *); /* * Finish up I/O, releasing I/O resources and returns the value * that would have been associated with a synchronous I/O request. * This routine must be called once and only once for each * I/O control block who has had I/O associated with it. */ ssize_t aio_return(struct aiocb *); /* * Cancel I/O */ int aio_cancel(int, struct aiocb *); /* * Suspend until all specified I/O or timeout is complete. */ int aio_suspend(const struct aiocb * const[], int, const struct timespec *); /* * Asynchronous mlock */ int aio_mlock(struct aiocb *); #ifdef __BSD_VISIBLE -int aio_waitcomplete(struct aiocb **, struct timespec *); +ssize_t aio_waitcomplete(struct aiocb **, struct timespec *); #endif int aio_fsync(int op, struct aiocb *aiocbp); __END_DECLS #endif /* !_KERNEL */ #endif /* !_SYS_AIO_H_ */ diff --git a/tests/sys/aio/aio_test.c b/tests/sys/aio/aio_test.c index 607f46db5acb..99ecc283f27e 100644 --- a/tests/sys/aio/aio_test.c +++ b/tests/sys/aio/aio_test.c @@ -1,663 +1,739 @@ /*- * Copyright (c) 2004 Robert N. M. Watson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ /* * Regression test to do some very basic AIO exercising on several types of * file descriptors. Currently, the tests consist of initializing a fixed * size buffer with pseudo-random data, writing it to one fd using AIO, then * reading it from a second descriptor using AIO. For some targets, the same * fd is used for write and read (i.e., file, md device), but for others the * operation is performed on a peer (pty, socket, fifo, etc). A timeout is * initiated to detect undo blocking. This test does not attempt to exercise * error cases or more subtle asynchronous behavior, just make sure that the * basic operations work on some basic object types. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "freebsd_test_suite/macros.h" #include "local.h" #define PATH_TEMPLATE "aio.XXXXXXXXXX" /* * GLOBAL_MAX sets the largest usable buffer size to be read and written, as * it sizes ac_buffer in the aio_context structure. It is also the default * size for file I/O. For other types, we use smaller blocks or we risk * blocking (and we run in a single process/thread so that would be bad). */ #define GLOBAL_MAX 16384 #define BUFFER_MAX GLOBAL_MAX struct aio_context { int ac_read_fd, ac_write_fd; long ac_seed; char ac_buffer[GLOBAL_MAX]; int ac_buflen; int ac_seconds; void (*ac_cleanup)(void *arg); void *ac_cleanup_arg; }; static int aio_timedout; /* * Each test run specifies a timeout in seconds. Use the somewhat obsoleted * signal(3) and alarm(3) APIs to set this up. */ static void aio_timeout_signal(int sig __unused) { aio_timedout = 1; } static void aio_timeout_start(int seconds) { aio_timedout = 0; ATF_REQUIRE_MSG(signal(SIGALRM, aio_timeout_signal) != SIG_ERR, "failed to set SIGALRM handler: %s", strerror(errno)); alarm(seconds); } static void aio_timeout_stop(void) { ATF_REQUIRE_MSG(signal(SIGALRM, NULL) != SIG_ERR, "failed to reset SIGALRM handler to default: %s", strerror(errno)); alarm(0); } /* * Fill a buffer given a seed that can be fed into srandom() to initialize * the PRNG in a repeatable manner. */ static void aio_fill_buffer(char *buffer, int len, long seed) { char ch; int i; srandom(seed); for (i = 0; i < len; i++) { ch = random() & 0xff; buffer[i] = ch; } } /* * Test that a buffer matches a given seed. See aio_fill_buffer(). Return * (1) on a match, (0) on a mismatch. */ static int aio_test_buffer(char *buffer, int len, long seed) { char ch; int i; srandom(seed); for (i = 0; i < len; i++) { ch = random() & 0xff; if (buffer[i] != ch) return (0); } return (1); } /* * Initialize a testing context given the file descriptors provided by the * test setup. */ static void aio_context_init(struct aio_context *ac, int read_fd, int write_fd, int buflen, int seconds, void (*cleanup)(void *), void *cleanup_arg) { ATF_REQUIRE_MSG(buflen <= BUFFER_MAX, "aio_context_init: buffer too large (%d > %d)", buflen, BUFFER_MAX); bzero(ac, sizeof(*ac)); ac->ac_read_fd = read_fd; ac->ac_write_fd = write_fd; ac->ac_buflen = buflen; srandomdev(); ac->ac_seed = random(); aio_fill_buffer(ac->ac_buffer, buflen, ac->ac_seed); ATF_REQUIRE_MSG(aio_test_buffer(ac->ac_buffer, buflen, ac->ac_seed) != 0, "aio_test_buffer: internal error"); ac->ac_seconds = seconds; ac->ac_cleanup = cleanup; ac->ac_cleanup_arg = cleanup_arg; } /* * Each tester can register a callback to clean up in the event the test * fails. Preserve the value of errno so that subsequent calls to errx() * work properly. */ static void aio_cleanup(struct aio_context *ac) { int error; if (ac->ac_cleanup == NULL) return; error = errno; (ac->ac_cleanup)(ac->ac_cleanup_arg); errno = error; } /* * Perform a simple write test of our initialized data buffer to the provided * file descriptor. */ static void aio_write_test(struct aio_context *ac) { struct aiocb aio, *aiop; ssize_t len; ATF_REQUIRE_KERNEL_MODULE("aio"); bzero(&aio, sizeof(aio)); aio.aio_buf = ac->ac_buffer; aio.aio_nbytes = ac->ac_buflen; aio.aio_fildes = ac->ac_write_fd; aio.aio_offset = 0; aio_timeout_start(ac->ac_seconds); if (aio_write(&aio) < 0) { if (errno == EINTR) { if (aio_timedout) { aio_cleanup(ac); atf_tc_fail("aio_write timed out"); } } aio_cleanup(ac); atf_tc_fail("aio_write failed: %s", strerror(errno)); } len = aio_waitcomplete(&aiop, NULL); if (len < 0) { if (errno == EINTR) { if (aio_timedout) { aio_cleanup(ac); atf_tc_fail("aio_waitcomplete timed out"); } } aio_cleanup(ac); atf_tc_fail("aio_waitcomplete failed: %s", strerror(errno)); } aio_timeout_stop(); if (len != ac->ac_buflen) { aio_cleanup(ac); atf_tc_fail("aio_waitcomplete short write (%jd)", (intmax_t)len); } } /* * Perform a simple read test of our initialized data buffer from the * provided file descriptor. */ static void aio_read_test(struct aio_context *ac) { struct aiocb aio, *aiop; ssize_t len; ATF_REQUIRE_KERNEL_MODULE("aio"); bzero(ac->ac_buffer, ac->ac_buflen); bzero(&aio, sizeof(aio)); aio.aio_buf = ac->ac_buffer; aio.aio_nbytes = ac->ac_buflen; aio.aio_fildes = ac->ac_read_fd; aio.aio_offset = 0; aio_timeout_start(ac->ac_seconds); if (aio_read(&aio) < 0) { if (errno == EINTR) { if (aio_timedout) { aio_cleanup(ac); atf_tc_fail("aio_write timed out"); } } aio_cleanup(ac); atf_tc_fail("aio_read failed: %s", strerror(errno)); } len = aio_waitcomplete(&aiop, NULL); if (len < 0) { if (errno == EINTR) { if (aio_timedout) { aio_cleanup(ac); atf_tc_fail("aio_waitcomplete timed out"); } } aio_cleanup(ac); atf_tc_fail("aio_waitcomplete failed: %s", strerror(errno)); } aio_timeout_stop(); if (len != ac->ac_buflen) { aio_cleanup(ac); atf_tc_fail("aio_waitcomplete short read (%jd)", (intmax_t)len); } if (aio_test_buffer(ac->ac_buffer, ac->ac_buflen, ac->ac_seed) == 0) { aio_cleanup(ac); atf_tc_fail("buffer mismatched"); } } /* * Series of type-specific tests for AIO. For now, we just make sure we can * issue a write and then a read to each type. We assume that once a write * is issued, a read can follow. */ /* * Test with a classic file. Assumes we can create a moderate size temporary * file. */ struct aio_file_arg { int afa_fd; char *afa_pathname; }; static void aio_file_cleanup(void *arg) { struct aio_file_arg *afa; afa = arg; close(afa->afa_fd); unlink(afa->afa_pathname); } #define FILE_LEN GLOBAL_MAX #define FILE_TIMEOUT 30 ATF_TC_WITHOUT_HEAD(aio_file_test); ATF_TC_BODY(aio_file_test, tc) { char pathname[PATH_MAX]; struct aio_file_arg arg; struct aio_context ac; int fd; ATF_REQUIRE_KERNEL_MODULE("aio"); ATF_REQUIRE_UNSAFE_AIO(); strcpy(pathname, PATH_TEMPLATE); fd = mkstemp(pathname); ATF_REQUIRE_MSG(fd != -1, "mkstemp failed: %s", strerror(errno)); arg.afa_fd = fd; arg.afa_pathname = pathname; aio_context_init(&ac, fd, fd, FILE_LEN, FILE_TIMEOUT, aio_file_cleanup, &arg); aio_write_test(&ac); aio_read_test(&ac); aio_file_cleanup(&arg); } struct aio_fifo_arg { int afa_read_fd; int afa_write_fd; char *afa_pathname; }; static void aio_fifo_cleanup(void *arg) { struct aio_fifo_arg *afa; afa = arg; if (afa->afa_read_fd != -1) close(afa->afa_read_fd); if (afa->afa_write_fd != -1) close(afa->afa_write_fd); unlink(afa->afa_pathname); } #define FIFO_LEN 256 #define FIFO_TIMEOUT 30 ATF_TC_WITHOUT_HEAD(aio_fifo_test); ATF_TC_BODY(aio_fifo_test, tc) { int error, read_fd = -1, write_fd = -1; struct aio_fifo_arg arg; char pathname[PATH_MAX]; struct aio_context ac; ATF_REQUIRE_KERNEL_MODULE("aio"); ATF_REQUIRE_UNSAFE_AIO(); /* * In theory, mkstemp() can return a name that is then collided with. * Because this is a regression test, we treat that as a test failure * rather than retrying. */ strcpy(pathname, PATH_TEMPLATE); ATF_REQUIRE_MSG(mkstemp(pathname) != -1, "mkstemp failed: %s", strerror(errno)); ATF_REQUIRE_MSG(unlink(pathname) == 0, "unlink failed: %s", strerror(errno)); ATF_REQUIRE_MSG(mkfifo(pathname, 0600) != -1, "mkfifo failed: %s", strerror(errno)); arg.afa_pathname = pathname; arg.afa_read_fd = -1; arg.afa_write_fd = -1; read_fd = open(pathname, O_RDONLY | O_NONBLOCK); if (read_fd == -1) { error = errno; aio_fifo_cleanup(&arg); errno = error; atf_tc_fail("read_fd open failed: %s", strerror(errno)); } arg.afa_read_fd = read_fd; write_fd = open(pathname, O_WRONLY); if (write_fd == -1) { error = errno; aio_fifo_cleanup(&arg); errno = error; atf_tc_fail("write_fd open failed: %s", strerror(errno)); } arg.afa_write_fd = write_fd; aio_context_init(&ac, read_fd, write_fd, FIFO_LEN, FIFO_TIMEOUT, aio_fifo_cleanup, &arg); aio_write_test(&ac); aio_read_test(&ac); aio_fifo_cleanup(&arg); } struct aio_unix_socketpair_arg { int asa_sockets[2]; }; static void aio_unix_socketpair_cleanup(void *arg) { struct aio_unix_socketpair_arg *asa; asa = arg; close(asa->asa_sockets[0]); close(asa->asa_sockets[1]); } #define UNIX_SOCKETPAIR_LEN 256 #define UNIX_SOCKETPAIR_TIMEOUT 30 ATF_TC_WITHOUT_HEAD(aio_unix_socketpair_test); ATF_TC_BODY(aio_unix_socketpair_test, tc) { struct aio_unix_socketpair_arg arg; struct aio_context ac; int sockets[2]; ATF_REQUIRE_KERNEL_MODULE("aio"); ATF_REQUIRE_MSG(socketpair(PF_UNIX, SOCK_STREAM, 0, sockets) != -1, "socketpair failed: %s", strerror(errno)); arg.asa_sockets[0] = sockets[0]; arg.asa_sockets[1] = sockets[1]; aio_context_init(&ac, sockets[0], sockets[1], UNIX_SOCKETPAIR_LEN, UNIX_SOCKETPAIR_TIMEOUT, aio_unix_socketpair_cleanup, &arg); aio_write_test(&ac); aio_read_test(&ac); aio_unix_socketpair_cleanup(&arg); } struct aio_pty_arg { int apa_read_fd; int apa_write_fd; }; static void aio_pty_cleanup(void *arg) { struct aio_pty_arg *apa; apa = arg; close(apa->apa_read_fd); close(apa->apa_write_fd); }; #define PTY_LEN 256 #define PTY_TIMEOUT 30 ATF_TC_WITHOUT_HEAD(aio_pty_test); ATF_TC_BODY(aio_pty_test, tc) { struct aio_pty_arg arg; struct aio_context ac; int read_fd, write_fd; struct termios ts; int error; ATF_REQUIRE_KERNEL_MODULE("aio"); ATF_REQUIRE_UNSAFE_AIO(); ATF_REQUIRE_MSG(openpty(&read_fd, &write_fd, NULL, NULL, NULL) == 0, "openpty failed: %s", strerror(errno)); arg.apa_read_fd = read_fd; arg.apa_write_fd = write_fd; if (tcgetattr(write_fd, &ts) < 0) { error = errno; aio_pty_cleanup(&arg); errno = error; atf_tc_fail("tcgetattr failed: %s", strerror(errno)); } cfmakeraw(&ts); if (tcsetattr(write_fd, TCSANOW, &ts) < 0) { error = errno; aio_pty_cleanup(&arg); errno = error; atf_tc_fail("tcsetattr failed: %s", strerror(errno)); } aio_context_init(&ac, read_fd, write_fd, PTY_LEN, PTY_TIMEOUT, aio_pty_cleanup, &arg); aio_write_test(&ac); aio_read_test(&ac); aio_pty_cleanup(&arg); } static void aio_pipe_cleanup(void *arg) { int *pipes = arg; close(pipes[0]); close(pipes[1]); } #define PIPE_LEN 256 #define PIPE_TIMEOUT 30 ATF_TC_WITHOUT_HEAD(aio_pipe_test); ATF_TC_BODY(aio_pipe_test, tc) { struct aio_context ac; int pipes[2]; ATF_REQUIRE_KERNEL_MODULE("aio"); ATF_REQUIRE_UNSAFE_AIO(); ATF_REQUIRE_MSG(pipe(pipes) != -1, "pipe failed: %s", strerror(errno)); aio_context_init(&ac, pipes[0], pipes[1], PIPE_LEN, PIPE_TIMEOUT, aio_pipe_cleanup, pipes); aio_write_test(&ac); aio_read_test(&ac); aio_pipe_cleanup(pipes); } struct aio_md_arg { int ama_mdctl_fd; int ama_unit; int ama_fd; }; static void aio_md_cleanup(void *arg) { struct aio_md_arg *ama; struct md_ioctl mdio; int error; ama = arg; if (ama->ama_fd != -1) close(ama->ama_fd); if (ama->ama_unit != -1) { bzero(&mdio, sizeof(mdio)); mdio.md_version = MDIOVERSION; mdio.md_unit = ama->ama_unit; if (ioctl(ama->ama_mdctl_fd, MDIOCDETACH, &mdio) == -1) { error = errno; close(ama->ama_mdctl_fd); errno = error; atf_tc_fail("ioctl MDIOCDETACH failed: %s", strerror(errno)); } } close(ama->ama_mdctl_fd); } #define MD_LEN GLOBAL_MAX #define MD_TIMEOUT 30 ATF_TC(aio_md_test); ATF_TC_HEAD(aio_md_test, tc) { atf_tc_set_md_var(tc, "require.user", "root"); } ATF_TC_BODY(aio_md_test, tc) { int error, fd, mdctl_fd, unit; char pathname[PATH_MAX]; struct aio_md_arg arg; struct aio_context ac; struct md_ioctl mdio; ATF_REQUIRE_KERNEL_MODULE("aio"); mdctl_fd = open("/dev/" MDCTL_NAME, O_RDWR, 0); ATF_REQUIRE_MSG(mdctl_fd != -1, "opening /dev/%s failed: %s", MDCTL_NAME, strerror(errno)); bzero(&mdio, sizeof(mdio)); mdio.md_version = MDIOVERSION; mdio.md_type = MD_MALLOC; mdio.md_options = MD_AUTOUNIT | MD_COMPRESS; mdio.md_mediasize = GLOBAL_MAX; mdio.md_sectorsize = 512; arg.ama_mdctl_fd = mdctl_fd; arg.ama_unit = -1; arg.ama_fd = -1; if (ioctl(mdctl_fd, MDIOCATTACH, &mdio) < 0) { error = errno; aio_md_cleanup(&arg); errno = error; atf_tc_fail("ioctl MDIOCATTACH failed: %s", strerror(errno)); } arg.ama_unit = unit = mdio.md_unit; snprintf(pathname, PATH_MAX, "/dev/md%d", unit); fd = open(pathname, O_RDWR); ATF_REQUIRE_MSG(fd != -1, "opening %s failed: %s", pathname, strerror(errno)); arg.ama_fd = fd; aio_context_init(&ac, fd, fd, MD_LEN, MD_TIMEOUT, aio_md_cleanup, &arg); aio_write_test(&ac); aio_read_test(&ac); aio_md_cleanup(&arg); } +ATF_TC_WITHOUT_HEAD(aio_large_read_test); +ATF_TC_BODY(aio_large_read_test, tc) +{ + char pathname[PATH_MAX]; + struct aiocb cb, *cbp; + ssize_t nread; + size_t len; + int fd; +#ifdef __LP64__ + int clamped; +#endif + + ATF_REQUIRE_KERNEL_MODULE("aio"); + ATF_REQUIRE_UNSAFE_AIO(); + +#ifdef __LP64__ + len = sizeof(clamped); + if (sysctlbyname("debug.iosize_max_clamp", &clamped, &len, NULL, 0) == + -1) + atf_libc_error(errno, "Failed to read debug.iosize_max_clamp"); +#endif + + /* Determine the maximum supported read(2) size. */ + len = SSIZE_MAX; +#ifdef __LP64__ + if (clamped) + len = INT_MAX; +#endif + + strcpy(pathname, PATH_TEMPLATE); + fd = mkstemp(pathname); + ATF_REQUIRE_MSG(fd != -1, "mkstemp failed: %s", strerror(errno)); + + unlink(pathname); + + memset(&cb, 0, sizeof(cb)); + cb.aio_nbytes = len; + cb.aio_fildes = fd; + cb.aio_buf = NULL; + if (aio_read(&cb) == -1) + atf_tc_fail("aio_read() of maximum read size failed: %s", + strerror(errno)); + + nread = aio_waitcomplete(&cbp, NULL); + if (nread == -1) + atf_tc_fail("aio_waitcomplete() failed: %s", strerror(errno)); + if (nread != 0) + atf_tc_fail("aio_read() from /dev/null returned data: %zd", + nread); + + memset(&cb, 0, sizeof(cb)); + cb.aio_nbytes = len + 1; + cb.aio_fildes = fd; + cb.aio_buf = NULL; + if (aio_read(&cb) == -1) { + if (errno == EINVAL) + goto finished; + atf_tc_fail("aio_read() of too large read size failed: %s", + strerror(errno)); + } + + nread = aio_waitcomplete(&cbp, NULL); + if (nread == -1) { + if (errno == EINVAL) + goto finished; + atf_tc_fail("aio_waitcomplete() failed: %s", strerror(errno)); + } + atf_tc_fail( + "aio_read() of too large read size from /dev/null returned: %zd", + nread); + +finished: + close(fd); +} + ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, aio_file_test); ATF_TP_ADD_TC(tp, aio_fifo_test); ATF_TP_ADD_TC(tp, aio_unix_socketpair_test); ATF_TP_ADD_TC(tp, aio_pty_test); ATF_TP_ADD_TC(tp, aio_pipe_test); ATF_TP_ADD_TC(tp, aio_md_test); + ATF_TP_ADD_TC(tp, aio_large_read_test); return (atf_no_error()); }