freebsd-dev/share/man/man9/taskqueue.9

298 lines
9.0 KiB
Groff
Raw Normal View History

2000-05-28 16:53:50 +00:00
.\" -*- nroff -*-
.\"
.\" Copyright (c) 2000 Doug Rabson
.\"
.\" All rights reserved.
.\"
.\" This program is free software.
.\"
.\" 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 DEVELOPERS ``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 DEVELOPERS 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.
.\"
.\" $FreeBSD$
.\"
2005-05-19 18:31:42 +00:00
.Dd May 19, 2005
2000-05-28 16:53:50 +00:00
.Dt TASKQUEUE 9
.Os
.Sh NAME
.Nm taskqueue
.Nd asynchronous task execution
.Sh SYNOPSIS
.In sys/param.h
.In sys/kernel.h
.In sys/malloc.h
.In sys/queue.h
.In sys/taskqueue.h
2000-05-28 16:53:50 +00:00
.Bd -literal
typedef void (*task_fn_t)(void *context, int pending);
2000-05-28 16:53:50 +00:00
typedef void (*taskqueue_enqueue_fn)(void *context);
struct task {
STAILQ_ENTRY(task) ta_link; /* link for queue */
2005-05-01 02:12:44 +00:00
u_short ta_pending; /* count times queued */
u_short ta_priority; /* priority of task in queue */
task_fn_t ta_func; /* task handler */
2000-05-28 16:53:50 +00:00
void *ta_context; /* argument for handler */
};
.Ed
.Ft struct taskqueue *
2005-05-01 02:12:44 +00:00
.Fn taskqueue_create "const char *name" "int mflags" "taskqueue_enqueue_fn enqueue" "void *context" "struct proc **"
2000-05-28 16:53:50 +00:00
.Ft void
.Fn taskqueue_free "struct taskqueue *queue"
.Ft struct taskqueue *
.Fn taskqueue_find "const char *name"
.Ft int
.Fn taskqueue_enqueue "struct taskqueue *queue" "struct task *task"
2004-01-02 07:23:40 +00:00
.Ft int
.Fn taskqueue_enqueue_fast "struct taskqueue *queue" "struct task *task"
2000-05-28 16:53:50 +00:00
.Ft void
.Fn taskqueue_run "struct taskqueue *queue"
.Ft void
.Fn taskqueue_run_fast "struct taskqueue *queue"
2005-05-19 18:31:42 +00:00
.Ft void
.Fn taskqueue_drain "struct taskqueue *queue" "struct task *task"
.Fn TASK_INIT "struct task *task" "int priority" "task_fn_t *func" "void *context"
2000-05-28 16:53:50 +00:00
.Fn TASKQUEUE_DECLARE "name"
.Fn TASKQUEUE_DEFINE "name" "taskqueue_enqueue_fn enqueue" "void *context" "init"
.Fn TASKQUEUE_DEFINE_THREAD "name"
2000-05-28 16:53:50 +00:00
.Sh DESCRIPTION
These functions provide a simple interface for asynchronous execution
of code.
.Pp
The function
.Fn taskqueue_create
is used to create new queues.
The arguments to
.Fn taskqueue_create
2005-05-01 02:12:44 +00:00
include a name that should be unique,
2000-05-28 16:53:50 +00:00
a set of
.Xr malloc 9
2005-05-01 02:12:44 +00:00
flags that specify whether the call to
2000-05-28 16:53:50 +00:00
.Fn malloc
2005-05-01 02:12:44 +00:00
is allowed to sleep,
a function that is called from
2000-05-28 16:53:50 +00:00
.Fn taskqueue_enqueue
2005-05-01 02:12:44 +00:00
when a task is added to the queue,
and a pointer to the memory location where the identity of the
thread that services the queue is recorded.
2000-05-28 16:53:50 +00:00
.\" XXX The rest of the sentence gets lots in relation to the first part.
2005-05-01 02:12:44 +00:00
The function called from
.Fn taskqueue_enqueue
must arrange for the queue to be processed
2000-05-28 16:53:50 +00:00
(for instance by scheduling a software interrupt or waking a kernel
thread).
2005-05-01 02:12:44 +00:00
The memory location where the thread identity is recorded is used
to signal the service thread(s) to terminate--when this value is set to
zero and the thread is signaled it will terminate.
2000-05-28 16:53:50 +00:00
.Pp
The function
.Fn taskqueue_free
should be used to remove the queue from the global list of queues
and free the memory used by the queue.
2005-05-01 02:12:44 +00:00
Any tasks that are on the queue will be executed at this time after
which the thread servicing the queue will be signaled that it should exit.
2000-05-28 16:53:50 +00:00
.Pp
The system maintains a list of all queues which can be searched using
.Fn taskqueue_find .
The first queue whose name matches is returned, otherwise
.Dv NULL .
.Pp
To add a task to the list of tasks queued on a taskqueue, call
.Fn taskqueue_enqueue
with pointers to the queue and task.
If the task's
.Va ta_pending
field is non-zero,
then it is simply incremented to reflect the number of times the task
was enqueued.
Otherwise,
the task is added to the list before the first task which has a lower
.Va ta_priority
value or at the end of the list if no tasks have a lower priority.
Enqueueing a task does not perform any memory allocation which makes
it suitable for calling from an interrupt handler.
This function will return
.Er EPIPE
2000-05-28 16:53:50 +00:00
if the queue is being freed.
.Pp
2004-01-02 07:23:40 +00:00
The function
.Fn taskqueue_enqueue_fast
should be used in place of
.Fn taskqueue_enqueue
when the enqueuing must happen from a fast interrupt handler.
This method uses spin locks to avoid the possibility of sleeping in the fast
interrupt context.
.Pp
2000-05-28 16:53:50 +00:00
To execute all the tasks on a queue,
call
.Fn taskqueue_run
or
.Fn taskqueue_run_fast
depending on the flavour of the queue.
2000-05-28 16:53:50 +00:00
When a task is executed,
first it is removed from the queue,
the value of
.Va ta_pending
is recorded and then the field is zeroed.
The function
.Va ta_func
from the task structure is called with the value of the field
.Va ta_context
as its first argument
and the value of
.Va ta_pending
as its second argument.
.Pp
The
2005-05-19 18:31:42 +00:00
.Fn taskqueue_drain
function is used to wait for the task to finish.
There is no guarantee that the task will not be
enqueued after call to
.Fn taskqueue_drain .
.Pp
2000-05-28 16:53:50 +00:00
A convenience macro,
.Fn TASK_INIT "task" "priority" "func" "context"
is provided to initialise a
.Va task
structure.
The values of
.Va priority ,
.Va func ,
and
.Va context
are simply copied into the task structure fields and the
.Va ta_pending
field is cleared.
.Pp
Three macros
.Fn TASKQUEUE_DECLARE "name" ,
.Fn TASKQUEUE_DEFINE "name" "enqueue" "context" "init" ,
2000-05-28 16:53:50 +00:00
and
.Fn TASKQUEUE_DEFINE_THREAD "name"
are used to declare a reference to a global queue, to define the
2005-01-12 21:48:25 +00:00
implementation of the queue, and declare a queue that uses its own thread.
2000-05-28 16:53:50 +00:00
The
.Fn TASKQUEUE_DEFINE
macro arranges to call
.Fn taskqueue_create
with the values of its
.Va name ,
.Va enqueue
and
.Va context
arguments during system initialisation.
After calling
.Fn taskqueue_create ,
the
.Va init
argument to the macro is executed as a C statement,
allowing any further initialisation to be performed
(such as registering an interrupt handler etc.)
.Pp
The
.Fn TASKQUEUE_DEFINE_THREAD
2005-01-12 21:48:25 +00:00
macro defines a new taskqueue with its own kernel thread to serve tasks.
The variable
.Vt struct proc *taskqueue_name_proc
is defined which contains the kernel thread serving the tasks.
The variable
.Vt struct taskqueue *taskqueue_name
is used to enqueue tasks onto the queue.
.Ss Predefined Task Queues
The system provides four global taskqueues,
.Va taskqueue_fast ,
2000-05-28 16:53:50 +00:00
.Va taskqueue_swi ,
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
.Va taskqueue_swi_giant ,
2004-07-02 19:07:33 +00:00
and
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
.Va taskqueue_thread .
The
.Va taskqueue_fast
queue is for swi handlers dispatched from fast interrupt handlers,
where sleep mutexes cannot be used.
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
The swi taskqueues are run via a software interrupt mechanism.
The
.Va taskqueue_swi
queue runs without the protection of the
.Va Giant
kernel lock, and the
.Va taskqueue_swi_giant
queue runs with the protection of the
.Va Giant
kernel lock.
The thread taskqueue
.Va taskqueue_thread
runs in a kernel thread context, and tasks run from this thread do
not run under the
.Va Giant
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
kernel lock.
If the caller wants to run under
.Va Giant ,
he should explicitly acquire and release
.Va Giant
in his taskqueue handler routine.
.Pp
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
To use these queues,
2000-05-28 16:53:50 +00:00
call
.Fn taskqueue_enqueue
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
with the value of the global taskqueue variable for the queue you wish to
2005-01-12 21:48:25 +00:00
use
.Va ( taskqueue_swi ,
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
.Va taskqueue_swi_giant ,
or
2005-01-12 21:48:25 +00:00
.Va taskqueue_thread ) .
Use
.Fn taskqueue_enqueue_fast
for the global taskqueue variable
.Va taskqueue_fast .
2000-05-28 16:53:50 +00:00
.Pp
The software interrupt queues can be used,
2000-05-28 16:53:50 +00:00
for instance, for implementing interrupt handlers which must perform a
significant amount of processing in the handler.
The hardware interrupt handler would perform minimal processing of the
interrupt and then enqueue a task to finish the work.
This reduces to a minimum
the amount of time spent with interrupts disabled.
Move dynamic sysctl(8) variable creation for the cd(4) and da(4) drivers out of cdregister() and daregister(), which are run from interrupt context. The sysctl code does blocking mallocs (M_WAITOK), which causes problems if malloc(9) actually needs to sleep. The eventual fix for this issue will involve moving the CAM probe process inside a kernel thread. For now, though, I have fixed the issue by moving dynamic sysctl variable creation for these two drivers to a task queue running in a kernel thread. The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in software interrupt handlers, which wouldn't fix the problem at hand. So I have created a new task queue, taskqueue_thread, that runs inside a kernel thread. (It also runs outside of Giant -- clients must explicitly acquire and release Giant in their taskqueue functions.) scsi_cd.c: Remove sysctl variable creation code from cdregister(), and move it to a new function, cdsysctlinit(). Queue cdsysctlinit() to the taskqueue_thread taskqueue once we have fully registered the cd(4) driver instance. scsi_da.c: Remove sysctl variable creation code from daregister(), and move it to move it to a new function, dasysctlinit(). Queue dasysctlinit() to the taskqueue_thread taskqueue once we have fully registered the da(4) instance. taskqueue.h: Declare the new taskqueue_thread taskqueue, update some comments. subr_taskqueue.c: Create the new kernel thread taskqueue. This taskqueue runs outside of Giant, so any functions queued to it would need to explicitly acquire/release Giant if they need it. cd.4: Update the cd(4) man page to talk about the minimum command size sysctl/loader tunable. Also note that the changer variables are available as loader tunables as well. da.4: Update the da(4) man page to cover the retry_count, default_timeout and minimum_cmd_size sysctl variables/loader tunables. Remove references to /dev/r???, they aren't used any longer. cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY quirk. taskqueue.9: Update the taskqueue(9) man page to describe the new thread task queue, and the taskqueue_swi_giant queue. MFC after: 3 days
2003-09-03 04:46:28 +00:00
.Pp
The thread queue can be used, for instance, by interrupt level routines
that need to call kernel functions that do things that can only be done
from a thread context.
(e.g., call malloc with the M_WAITOK flag.)
2005-05-01 02:12:44 +00:00
.Pp
Note that tasks queued on shared taskqueues such as
.Va taskqueue_swi
may be delayed an indeterminate amount of time before execution.
If queueing delays cannot be tolerated then a private taskqueue should
be created with a dedicated processing thread.
2005-04-15 14:46:59 +00:00
.Sh SEE ALSO
.Xr ithread 9 ,
.Xr kthread 9 ,
.Xr swi 9
2000-05-28 16:53:50 +00:00
.Sh HISTORY
This interface first appeared in
.Fx 5.0 .
There is a similar facility called tqueue in the Linux kernel.
.Sh AUTHORS
This man page was written by
.An Doug Rabson .
.Sh BUGS
There is no
.Fn taskqueue_create_fast .