d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add new system call - aio_mlock(). The name speaks for itself. It allows

Browse Source 
to perform the mlock(2) operation, which can consume a lot of time, under
control of aio(4).

Reviewed by:	kib, jilles
Sponsored by:	Nginx, Inc.
This commit is contained in:
glebius 2013-06-08 13:27:57 +00:00
parent bf96f17d6f
commit 9a02f3097d
7 changed files with 189 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
lib/libc/sys/Makefile.inc
View File

@ -85,6 +85,7 @@ MAN+= abort2.2 \
adjtime.2 \
aio_cancel.2 \
aio_error.2 \
aio_mlock.2 \
aio_read.2 \
aio_return.2 \
aio_suspend.2 \

1
lib/libc/sys/Symbol.map
View File

@ -379,6 +379,7 @@ FBSD_1.2 {
FBSD_1.3 {
accept4;
aio_mlock;
bindat;
cap_fcntls_get;
cap_fcntls_limit;

133
lib/libc/sys/aio_mlock.2 Normal file
View File

@ -0,0 +1,133 @@
.\" Copyright (c) 2013 Gleb Smirnoff <glebius@FreeBSD.org>
.\" 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.
.\"
.\" $FreeBSD$
.\"
.Dd June 3, 2013
.Dt AIO_MLOCK 2
.Os
.Sh NAME
.Nm aio_mlock
.Nd asynchronous
.Xr mlock 2
operation
.Sh LIBRARY
.Lb libc
.Sh SYNOPSIS
.In aio.h
.Ft int
.Fn aio_mlock "struct aiocb *iocb"
.Sh DESCRIPTION
The
.Fn aio_mlock
system call allows the calling process to lock into memory the
physical pages associated with the virtual address range starting at
.Fa iocb->aio_buf
for
.Fa iocb->aio_nbytes
bytes.
The call returns immediately after the locking request has
been enqueued; the operation may or may not have completed at the time
the call returns.
.Pp
The
.Fa iocb
pointer may be subsequently used as an argument to
.Fn aio_return
and
.Fn aio_error
in order to determine return or error status for the enqueued operation
while it is in progress.
.Pp
If the request could not be enqueued (generally due to
.Xr aio 4
limits),
then the call returns without having enqueued the request.
.Sh RESTRICTIONS
The Asynchronous I/O Control Block structure pointed to by
.Fa iocb
and the buffer that the
.Fa iocb->aio_buf
member of that structure references must remain valid until the
operation has completed.
For this reason, use of auto (stack) variables
for these objects is discouraged.
.Pp
The asynchronous I/O control buffer
.Fa iocb
should be zeroed before the
.Fn aio_mlock
call to avoid passing bogus context information to the kernel.
.Pp
Modifications of the Asynchronous I/O Control Block structure or the
buffer contents after the request has been enqueued, but before the
request has completed, are not allowed.
.Sh RETURN VALUES
.Rv -std aio_mlock
.Sh ERRORS
The
.Fn aio_read
system call will fail if:
.Bl -tag -width Er
.It Bq Er EAGAIN
The request was not queued because of system resource limitations.
.It Bq Er ENOSYS
The
.Fn aio_mlock
system call is not supported.
.El
.Pp
If the request is successfully enqueued, but subsequently cancelled
or an error occurs, the value returned by the
.Fn aio_return
system call is per the
.Xr mlock 2
system call, and the value returned by the
.Fn aio_error
system call is one of the error returns from the
.Xr mlock 2
system call, or
.Er ECANCELED
if the request was explicitly cancelled via a call to
.Fn aio_cancel .
.Sh SEE ALSO
.Xr aio_cancel 2 ,
.Xr aio_error 2 ,
.Xr aio_return 2 ,
.Xr aio 4 ,
.Xr mlock 2
.Sh PORTABILITY
The
.Fn aio_mlock
system call is a
.Fx
extension, and should not be used in portable code.
.Sh HISTORY
The
.Fn aio_mlock
system call first appeared in
.Fx 10.0 .
.Sh AUTHORS
The system call was introduced by
.An Gleb Smirnoff Aq glebius@FreeBSD.org .

2
sys/compat/freebsd32/syscalls.master
View File

@ -1044,3 +1044,5 @@
__socklen_t * __restrict anamelen, \
int flags); }
542 AUE_PIPE NOPROTO { int pipe2(int *fildes, int flags); }
543 AUE_NULL NOSTD { int freebsd32_aio_mlock( \
struct aiocb32 *aiocbp); }

1
sys/kern/syscalls.master
View File

@ -977,5 +977,6 @@
__socklen_t * __restrict anamelen, \
int flags); }
542 AUE_PIPE STD { int pipe2(int *fildes, int flags); }
543 AUE_NULL NOSTD { int aio_mlock(struct aiocb *aiocbp); }
; Please copy any additions and changes to the following compatability tables:
; sys/compat/freebsd32/syscalls.master

49
sys/kern/vfs_aio.c
View File

@ -340,6 +340,7 @@ static int aio_onceonly(void);
static int aio_free_entry(struct aiocblist *aiocbe);
static void aio_process_rw(struct aiocblist *aiocbe);
static void aio_process_sync(struct aiocblist *aiocbe);
static void aio_process_mlock(struct aiocblist *aiocbe);
static int aio_newproc(int *);
int aio_aqueue(struct thread *td, struct aiocb *job,
struct aioliojob *lio, int type, struct aiocb_ops *ops);
@ -426,6 +427,7 @@ static struct syscall_helper_data aio_syscalls[] = {
SYSCALL_INIT_HELPER(aio_cancel),
SYSCALL_INIT_HELPER(aio_error),
SYSCALL_INIT_HELPER(aio_fsync),
SYSCALL_INIT_HELPER(aio_mlock),
SYSCALL_INIT_HELPER(aio_read),
SYSCALL_INIT_HELPER(aio_return),
SYSCALL_INIT_HELPER(aio_suspend),
@ -453,6 +455,7 @@ static struct syscall_helper_data aio32_syscalls[] = {
SYSCALL32_INIT_HELPER(freebsd32_aio_cancel),
SYSCALL32_INIT_HELPER(freebsd32_aio_error),
SYSCALL32_INIT_HELPER(freebsd32_aio_fsync),
SYSCALL32_INIT_HELPER(freebsd32_aio_mlock),
SYSCALL32_INIT_HELPER(freebsd32_aio_read),
SYSCALL32_INIT_HELPER(freebsd32_aio_write),
SYSCALL32_INIT_HELPER(freebsd32_aio_waitcomplete),
@ -702,7 +705,8 @@ aio_free_entry(struct aiocblist *aiocbe)
* at open time, but this is already true of file descriptors in
* a multithreaded process.
*/
fdrop(aiocbe->fd_file, curthread);
if (aiocbe->fd_file)
fdrop(aiocbe->fd_file, curthread);
crfree(aiocbe->cred);
uma_zfree(aiocb_zone, aiocbe);
AIO_LOCK(ki);
@ -967,6 +971,21 @@ aio_process_sync(struct aiocblist *aiocbe)
td->td_ucred = td_savedcred;
}
static void
aio_process_mlock(struct aiocblist *aiocbe)
{
struct aiocb *cb = &aiocbe->uaiocb;
int error;
KASSERT(aiocbe->uaiocb.aio_lio_opcode == LIO_MLOCK,
("%s: opcode %d", __func__, aiocbe->uaiocb.aio_lio_opcode));
error = vm_mlock(aiocbe->userproc, aiocbe->cred,
__DEVOLATILE(void *, cb->aio_buf), cb->aio_nbytes);
cb->_aiocb_private.error = error;
cb->_aiocb_private.status = 0;
}
static void
aio_bio_done_notify(struct proc *userp, struct aiocblist *aiocbe, int type)
{
@ -1143,6 +1162,9 @@ aio_daemon(void *_id)
case LIO_SYNC:
aio_process_sync(aiocbe);
break;
case LIO_MLOCK:
aio_process_mlock(aiocbe);
break;
}
mtx_lock(&aio_job_mtx);
@ -1283,7 +1305,7 @@ aio_qphysio(struct proc *p, struct aiocblist *aiocbe)
cb = &aiocbe->uaiocb;
fp = aiocbe->fd_file;
if (fp->f_type != DTYPE_VNODE)
if (fp == NULL || fp->f_type != DTYPE_VNODE)
return (-1);
vp = fp->f_vnode;
@ -1635,6 +1657,9 @@ aio_aqueue(struct thread *td, struct aiocb *job, struct aioliojob *lj,
case LIO_SYNC:
error = fget(td, fd, CAP_FSYNC, &fp);
break;
case LIO_MLOCK:
fp = NULL;
break;
case LIO_NOP:
error = fget(td, fd, CAP_NONE, &fp);
break;
@ -1692,7 +1717,8 @@ aio_aqueue(struct thread *td, struct aiocb *job, struct aioliojob *lj,
error = kqfd_register(kqfd, &kev, td, 1);
aqueue_fail:
if (error) {
fdrop(fp, td);
if (fp)
fdrop(fp, td);
uma_zfree(aiocb_zone, aiocbe);
ops->store_error(job, error);
goto done;
@ -1709,7 +1735,7 @@ aio_aqueue(struct thread *td, struct aiocb *job, struct aioliojob *lj,
if (opcode == LIO_SYNC)
goto queueit;
if (fp->f_type == DTYPE_SOCKET) {
if (fp && fp->f_type == DTYPE_SOCKET) {
/*
* Alternate queueing for socket ops: Reach down into the
* descriptor to get the socket data. Then check to see if the
@ -2187,6 +2213,13 @@ 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,
@ -2928,6 +2961,14 @@ freebsd32_aio_write(struct thread *td, struct freebsd32_aio_write_args *uap)
&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)

6
sys/sys/aio.h
View File

@ -37,6 +37,7 @@
#define LIO_READ 0x2
#ifdef _KERNEL
#define LIO_SYNC 0x3
#define LIO_MLOCK 0x4
#endif
/*
@ -124,6 +125,11 @@ int aio_cancel(int, struct aiocb *);
*/
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 *);
#endif