Add prior release version 1.04 to repo.

python3 rel_1.04
Alex Tsariounov 15 years ago
commit 7c9c4134b3

@ -0,0 +1,5 @@
set nocompatible
set hlsearch
set ai
set sw=4
set expandtab

@ -0,0 +1,2 @@
Alex Tsariounov <alext@novell.com>
http://www.novell.com/realtime

@ -0,0 +1,340 @@
GNU GENERAL PUBLIC LICENSE
Version 2, June 1991
Copyright (C) 1989, 1991 Free Software Foundation, Inc.
59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.
Preamble
The licenses for most software are designed to take away your
freedom to share and change it. By contrast, the GNU General Public
License is intended to guarantee your freedom to share and change free
software--to make sure the software is free for all its users. This
General Public License applies to most of the Free Software
Foundation's software and to any other program whose authors commit to
using it. (Some other Free Software Foundation software is covered by
the GNU Library General Public License instead.) You can apply it to
your programs, too.
When we speak of free software, we are referring to freedom, not
price. Our General Public Licenses are designed to make sure that you
have the freedom to distribute copies of free software (and charge for
this service if you wish), that you receive source code or can get it
if you want it, that you can change the software or use pieces of it
in new free programs; and that you know you can do these things.
To protect your rights, we need to make restrictions that forbid
anyone to deny you these rights or to ask you to surrender the rights.
These restrictions translate to certain responsibilities for you if you
distribute copies of the software, or if you modify it.
For example, if you distribute copies of such a program, whether
gratis or for a fee, you must give the recipients all the rights that
you have. You must make sure that they, too, receive or can get the
source code. And you must show them these terms so they know their
rights.
We protect your rights with two steps: (1) copyright the software, and
(2) offer you this license which gives you legal permission to copy,
distribute and/or modify the software.
Also, for each author's protection and ours, we want to make certain
that everyone understands that there is no warranty for this free
software. If the software is modified by someone else and passed on, we
want its recipients to know that what they have is not the original, so
that any problems introduced by others will not reflect on the original
authors' reputations.
Finally, any free program is threatened constantly by software
patents. We wish to avoid the danger that redistributors of a free
program will individually obtain patent licenses, in effect making the
program proprietary. To prevent this, we have made it clear that any
patent must be licensed for everyone's free use or not licensed at all.
The precise terms and conditions for copying, distribution and
modification follow.
GNU GENERAL PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
0. This License applies to any program or other work which contains
a notice placed by the copyright holder saying it may be distributed
under the terms of this General Public License. The "Program", below,
refers to any such program or work, and a "work based on the Program"
means either the Program or any derivative work under copyright law:
that is to say, a work containing the Program or a portion of it,
either verbatim or with modifications and/or translated into another
language. (Hereinafter, translation is included without limitation in
the term "modification".) Each licensee is addressed as "you".
Activities other than copying, distribution and modification are not
covered by this License; they are outside its scope. The act of
running the Program is not restricted, and the output from the Program
is covered only if its contents constitute a work based on the
Program (independent of having been made by running the Program).
Whether that is true depends on what the Program does.
1. You may copy and distribute verbatim copies of the Program's
source code as you receive it, in any medium, provided that you
conspicuously and appropriately publish on each copy an appropriate
copyright notice and disclaimer of warranty; keep intact all the
notices that refer to this License and to the absence of any warranty;
and give any other recipients of the Program a copy of this License
along with the Program.
You may charge a fee for the physical act of transferring a copy, and
you may at your option offer warranty protection in exchange for a fee.
2. You may modify your copy or copies of the Program or any portion
of it, thus forming a work based on the Program, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:
a) You must cause the modified files to carry prominent notices
stating that you changed the files and the date of any change.
b) You must cause any work that you distribute or publish, that in
whole or in part contains or is derived from the Program or any
part thereof, to be licensed as a whole at no charge to all third
parties under the terms of this License.
c) If the modified program normally reads commands interactively
when run, you must cause it, when started running for such
interactive use in the most ordinary way, to print or display an
announcement including an appropriate copyright notice and a
notice that there is no warranty (or else, saying that you provide
a warranty) and that users may redistribute the program under
these conditions, and telling the user how to view a copy of this
License. (Exception: if the Program itself is interactive but
does not normally print such an announcement, your work based on
the Program is not required to print an announcement.)
These requirements apply to the modified work as a whole. If
identifiable sections of that work are not derived from the Program,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works. But when you
distribute the same sections as part of a whole which is a work based
on the Program, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote it.
Thus, it is not the intent of this section to claim rights or contest
your rights to work written entirely by you; rather, the intent is to
exercise the right to control the distribution of derivative or
collective works based on the Program.
In addition, mere aggregation of another work not based on the Program
with the Program (or with a work based on the Program) on a volume of
a storage or distribution medium does not bring the other work under
the scope of this License.
3. You may copy and distribute the Program (or a work based on it,
under Section 2) in object code or executable form under the terms of
Sections 1 and 2 above provided that you also do one of the following:
a) Accompany it with the complete corresponding machine-readable
source code, which must be distributed under the terms of Sections
1 and 2 above on a medium customarily used for software interchange; or,
b) Accompany it with a written offer, valid for at least three
years, to give any third party, for a charge no more than your
cost of physically performing source distribution, a complete
machine-readable copy of the corresponding source code, to be
distributed under the terms of Sections 1 and 2 above on a medium
customarily used for software interchange; or,
c) Accompany it with the information you received as to the offer
to distribute corresponding source code. (This alternative is
allowed only for noncommercial distribution and only if you
received the program in object code or executable form with such
an offer, in accord with Subsection b above.)
The source code for a work means the preferred form of the work for
making modifications to it. For an executable work, complete source
code means all the source code for all modules it contains, plus any
associated interface definition files, plus the scripts used to
control compilation and installation of the executable. However, as a
special exception, the source code distributed need not include
anything that is normally distributed (in either source or binary
form) with the major components (compiler, kernel, and so on) of the
operating system on which the executable runs, unless that component
itself accompanies the executable.
If distribution of executable or object code is made by offering
access to copy from a designated place, then offering equivalent
access to copy the source code from the same place counts as
distribution of the source code, even though third parties are not
compelled to copy the source along with the object code.
4. You may not copy, modify, sublicense, or distribute the Program
except as expressly provided under this License. Any attempt
otherwise to copy, modify, sublicense or distribute the Program is
void, and will automatically terminate your rights under this License.
However, parties who have received copies, or rights, from you under
this License will not have their licenses terminated so long as such
parties remain in full compliance.
5. You are not required to accept this License, since you have not
signed it. However, nothing else grants you permission to modify or
distribute the Program or its derivative works. These actions are
prohibited by law if you do not accept this License. Therefore, by
modifying or distributing the Program (or any work based on the
Program), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Program or works based on it.
6. Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions. You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties to
this License.
7. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License. If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Program at all. For example, if a patent
license would not permit royalty-free redistribution of the Program by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Program.
If any portion of this section is held invalid or unenforceable under
any particular circumstance, the balance of the section is intended to
apply and the section as a whole is intended to apply in other
circumstances.
It is not the purpose of this section to induce you to infringe any
patents or other property right claims or to contest validity of any
such claims; this section has the sole purpose of protecting the
integrity of the free software distribution system, which is
implemented by public license practices. Many people have made
generous contributions to the wide range of software distributed
through that system in reliance on consistent application of that
system; it is up to the author/donor to decide if he or she is willing
to distribute software through any other system and a licensee cannot
impose that choice.
This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.
8. If the distribution and/or use of the Program is restricted in
certain countries either by patents or by copyrighted interfaces, the
original copyright holder who places the Program under this License
may add an explicit geographical distribution limitation excluding
those countries, so that distribution is permitted only in or among
countries not thus excluded. In such case, this License incorporates
the limitation as if written in the body of this License.
9. The Free Software Foundation may publish revised and/or new versions
of the General Public License from time to time. Such new versions will
be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.
Each version is given a distinguishing version number. If the Program
specifies a version number of this License which applies to it and "any
later version", you have the option of following the terms and conditions
either of that version or of any later version published by the Free
Software Foundation. If the Program does not specify a version number of
this License, you may choose any version ever published by the Free Software
Foundation.
10. If you wish to incorporate parts of the Program into other free
programs whose distribution conditions are different, write to the author
to ask for permission. For software which is copyrighted by the Free
Software Foundation, write to the Free Software Foundation; we sometimes
make exceptions for this. Our decision will be guided by the two goals
of preserving the free status of all derivatives of our free software and
of promoting the sharing and reuse of software generally.
NO WARRANTY
11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS
TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
REPAIR OR CORRECTION.
12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
END OF TERMS AND CONDITIONS
How to Apply These Terms to Your New Programs
If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest
to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the program's name and a brief idea of what it does.>
Copyright (C) <year> <name of author>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
Also add information on how to contact you by electronic and paper mail.
If the program is interactive, make it output a short notice like this
when it starts in an interactive mode:
Gnomovision version 69, Copyright (C) year name of author
Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
This is free software, and you are welcome to redistribute it
under certain conditions; type `show c' for details.
The hypothetical commands `show w' and `show c' should show the appropriate
parts of the General Public License. Of course, the commands you use may
be called something other than `show w' and `show c'; they could even be
mouse-clicks or menu items--whatever suits your program.
You should also get your employer (if you work as a programmer) or your
school, if any, to sign a "copyright disclaimer" for the program, if
necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the program
`Gnomovision' (which makes passes at compilers) written by James Hacker.
<signature of Ty Coon>, 1 April 1989
Ty Coon, President of Vice
This General Public License does not permit incorporating your program into
proprietary programs. If your program is a subroutine library, you may
consider it more useful to permit linking proprietary applications with the
library. If this is what you want to do, use the GNU Library General
Public License instead of this License.

@ -0,0 +1,22 @@
This is a distutils-based package and as such, it contains
a number of differnet installation methods. By far the most
convenient is to create an rpm. This rpm can then be later
updated or removed by the rpm packaging utility (rpm).
To create an rpm, use the following command:
python setup.py bdist_rpm
For basic installation:
python setup.py install
By default, the above command installs cpuset in the
$HOME/{bin,lib,share} directories. For a different location,
use the --prefix option:
python setup.py install --prefix=/usr
For more information:
http://docs.python.org/inst/inst.html

@ -0,0 +1,41 @@
AUTHORS
COPYING
INSTALL
MANIFEST
MANIFEST.in
Makefile
README
TODO
cset
setup.cfg
setup.py
cpuset/__init__.py
cpuset/cset.py
cpuset/main.py
cpuset/task.py
cpuset/util.py
cpuset/version.py
cpuset/commands/__init__.py
cpuset/commands/common.py
cpuset/commands/mem.py
cpuset/commands/proc.py
cpuset/commands/set.py
cpuset/commands/shield.py
doc/Makefile
doc/asciidoc.conf
doc/callouts.xsl
doc/cset-proc.1
doc/cset-proc.html
doc/cset-proc.txt
doc/cset-set.1
doc/cset-set.html
doc/cset-set.txt
doc/cset-shield.1
doc/cset-shield.html
doc/cset-shield.txt
doc/cset.1
doc/cset.html
doc/cset.txt
doc/tutorial.conf
doc/tutorial.txt
t/README

@ -0,0 +1,3 @@
include README Makefile MANIFEST MANIFEST.in AUTHORS COPYING INSTALL ChangeLog TODO
include t/README
include doc/*.txt doc/Makefile doc/*.conf doc/callouts.xsl doc/*.1 doc/*.html

@ -0,0 +1,32 @@
PREFIX ?= $(HOME)
DESTDIR ?= /
PYTHON ?= python
all:
$(PYTHON) setup.py build
install:
$(PYTHON) setup.py install --prefix=$(PREFIX) --root=$(DESTDIR)
test:
cd t && $(MAKE) all
doc:
cd doc && $(MAKE) all
clean:
for dir in doc t; do \
(cd $$dir && $(MAKE) clean); \
done
rm -rf build
rm -f cpuset/*.pyc
rm -f cpuset/commands/*.pyc
rm -f TAGS
clobber: clean
rm -rf dist
tags:
ctags -e -R cpuset/*
.PHONY: all install doc test clean

@ -0,0 +1,25 @@
Cpuset is a Python application that forms a wrapper around the
standard Linux filesystem calls to make using the cpusets facilities
in the Linux kernel easier.
For the latest version see http://opensuse.org/build_service/cpuset
Bugs or feature requests should be sent to http://bugs.opensuse.org
-----
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 only as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

@ -0,0 +1,4 @@
The TODO list before 1.0:
- implement test suite in tests/

@ -0,0 +1,5 @@
set nocompatible
set hlsearch
set ai
set sw=4
set expandtab

@ -0,0 +1,23 @@
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
---------------------------------------------------------------------
Substrate code and ideas taken from the excellent stgit 0.13, see
https://gna.org/projects/stgit and http://www.procode.org/stgit/
Stacked GIT is under GPL V2 or later.
---------------------------------------------------------------------
"""

@ -0,0 +1,5 @@
set nocompatible
set hlsearch
set ai
set sw=4
set expandtab

@ -0,0 +1,23 @@
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
---------------------------------------------------------------------
Substrate code and ideas taken from the excellent stgit 0.13, see
https://gna.org/projects/stgit and http://www.procode.org/stgit/
Stacked GIT is under GPL V2 or later.
---------------------------------------------------------------------
"""

@ -0,0 +1,25 @@
"""Common functions and variables for all commands
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
# Command exception class
class CmdException(Exception):
pass

@ -0,0 +1,63 @@
"""Memory node manipulation command
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, os, logging
from optparse import OptionParser, make_option
from cpuset.commands.common import *
from cpuset import cset
from cpuset.util import *
global log
log = logging.getLogger('mem')
help = 'create and destroy memory nodes within cpusets'
usage = """%prog [options] [cpuset name]
Create and manage memory node assignments to cpusets. Note that for
non-NUMA machines, the memory node assignment will always be 0 (zero)
and is so set by default. Thus this command only needs to be used
for NUMA machines.
"""
options = [make_option('-l', '--list',
help = 'list memory nodes in specified cpuset',
action = 'store_true'),
make_option('-c', '--create',
metavar = 'NODESPEC',
help = 'create a memory node in specified cpuset'),
make_option('-d', '--destroy',
help = 'destroy specified memory node in specified cpuset',
action = 'store_true'),
make_option('-m', '--move',
help = 'move specified memory node to specified cpuset',
action = 'store_true'),
make_option('-s', '--set',
metavar = 'CPUSET',
help = 'specify immediate cpuset'),
make_option('-t', '--toset',
help = 'specify destination cpuset'),
make_option('-f', '--fromset',
help = 'specify origination cpuset')
]
def func(parser, options, args):
log.debug("entering mem, options=%s, args=%s", options, args)

@ -0,0 +1,572 @@
"""Process manipulation command
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, os, re, logging, pwd, grp
from optparse import OptionParser, make_option
from cpuset import cset
from cpuset.util import *
from cpuset.commands.common import *
try: from cpuset.commands import set
except: pass
global log
log = logging.getLogger('proc')
help = 'create and manage processes within cpusets'
usage = """%prog [options] [path/program [args]]
This command is used to run and manage arbitrary processes on
specified cpusets. It is also used to move pre-existing processes
and threads to specified cpusets. You may note there is no
"kill" or "destroy" option -- use the standard OS ^C or kill
commands for that.
To list which tasks are running in a particular cpuset, use the
--list command.
For example:
# cset proc --list --set priset
This command will list all the tasks running in the
cpuset called "priset".
Processes are created by specifying the path to the executable
and specifying the cpuset that the process is to be created in.
For example:
# cset proc --set=blazing_cpuset --exec /usr/bin/fast_code
This command will execute the /usr/bin/fast_code program
on the "blazing_cpuset" cpuset.
The PIDSPEC argument taken for the move command is a comma
separated list of PIDs or TIDs. The list can also include
brackets of PIDs or TIDs (i.e. tasks) that are inclusive of the
endpoints.
For example:
1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700
Note that the range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.
To move a PIDSPEC to a specific cpuset, you can either specify
the PIDSPEC with --pid and the destination cpuset with --toset,
or use the short hand and list the cpuset name after the PIDSPEC
for the --move arguments.
The move command accepts multiple common calling methods.
For example, the following commands are equivalent:
# cset proc --move 2442,3000-3200 reserved_set
# cset proc --move --pid=2442,3000-3200 --toset=reserved_set
These commands move the tasks defined as 2442 and any
running task between 3000 and 3200 inclusive of the ends
to the cpuset called "reserved_set".
Specifying the --fromset is not necesary since the tasks will be
moved to the destination cpuset no matter which cpuset they are
currently running on.
Note however that if you do specify a cpuset with the --fromset
option, then only those tasks that are both in the PIDSPEC *and*
are running in the cpuset specified by --fromset will be moved.
I.e., if there is a task running on the system but not in
--fromset that is in PIDSPEC, it will not be moved.
To move all userspace tasks from one cpuset to another, you need
to specify the source and destination cpuset by name.
For example:
# cset proc --move --fromset=comp1 --toset=comp42
This command specifies that all processes and threads
running on cpuset "comp1" be moved to cpuset "comp42".
Note that the move command will not move kernel threads unless
the -k/--kthread switch is specified. If it is, then all unbound
kernel threads will be added to the move. Unbound kernel threads
are those that can run on any CPU. If you also specify the
-a/--all switch, then all tasks, kernel or not, bound or not,
will be moved.
CAUTION: Please be cautious with the --all switch, since moving a
kernel thread that is bound to a specific CPU to a cpuset that
does not include that CPU can cause a system hang.
You must specify unique cpuset names for the both exec and move
commands. If a simple name passed to the --fromset, --toset and
--set parameters is unique on the system then that command
executes. However, if there are multiple cpusets by that name,
then you will need to specify which one you mean with a full path
rooted at the base cpuset tree.
For example, suppose you have the following cpuset tree:
/group1
/myset
/yourset
/group2
/myset
/yourset
Then, to move a process from myset in group1 to yourset in
group2, you would have to issue the following command:
# cset proc --move --pid=50 --fromset=/group1/myset \\
--toset=/group2/yourset
"""
verbose = 0
options = [make_option('-l', '--list',
help = 'list processes in the specified cpuset',
action = 'store_true'),
make_option('-e', '--exec',
help = 'execute arguments in the specified cpuset',
dest = 'exc',
action = 'store_true'),
make_option('-u', '--user',
help = 'use this USER to --exec (id or name)'),
make_option('-g', '--group',
help = 'use this GROUP to --exec (id or name)'),
make_option('-m', '--move',
help = 'move specified tasks to specified cpuset; '
'to move a PIDSPEC to a cpuset, use -m PIDSPEC cpuset; '
'to move all tasks specify --fromset and --toset',
action = 'store_true'),
make_option('-p', '--pid',
metavar = 'PIDSPEC',
help = 'specify pid or tid specification'),
make_option('-s', '--set',
metavar = 'CPUSET',
help = 'specify name of immediate cpuset'),
make_option('-t', '--toset',
help = 'specify name of destination cpuset'),
make_option('-f', '--fromset',
help = 'specify name of origination cpuset'),
make_option('-k', '--kthread',
help = 'move, or include moving, unbound kernel threads',
action = 'store_true'),
make_option('-a', '--all',
help = 'force all processes and threads to be moved',
action = 'store_true'),
make_option('-v', '--verbose',
help = 'prints more detailed output, additive',
action = 'count')
]
def func(parser, options, args):
log.debug("entering func, options=%s, args=%s", options, args)
global verbose
if options.verbose: verbose = options.verbose
cset.rescan()
tset = None
if options.list or options.exc:
if options.set:
tset = cset.unique_set(options.set)
elif options.toset:
tset = cset.unique_set(options.toset)
elif len(args) > 0:
tset = cset.unique_set(args[0])
if options.exc: del args[0]
else: tset = args
else:
raise CpusetException("cpuset not specified")
try:
log.debug("operating on set %s", tset.path)
except:
log.debug("operating on sets %s", tset)
if options.exc: run(tset, args, options.user, options.group)
if options.list:
list_sets(tset)
return
if options.move or options.kthread:
# first, we need to know the destination
tset = None
if options.toset:
tset = cset.unique_set(options.toset)
elif options.set:
tset = cset.unique_set(options.set)
elif len(args) > 0:
if len(args) > 1 and options.pid == None:
options.pid = args[0]
tset = cset.unique_set(args[1])
else:
tset = cset.unique_set(args[0])
else:
raise CpusetException("destination cpuset not specified")
set.active(tset)
# next, if there is a pidspec, move just that
if options.pid:
fset = None
if options.fromset and not options.all:
fset = cset.unique_set(options.fromset)
pids = pidspec_to_list(options.pid, fset)
if len(pids):
l = []
l.append('--> moving following pidspec: %s' % options.pid)
l.extend(task_detail_header(' '))
l.extend(task_detail_table(pids, ' ', 76))
log.info("\n".join(l))
selective_move(None, tset, pids, options.kthread, options.all)
log.info('done')
else:
# here we assume move everything from fromset to toset
if options.fromset == None:
raise CpusetException("origination cpuset not specified")
fset = cset.unique_set(options.fromset)
nt = len(fset.tasks)
if nt == 0:
raise CpusetException('no tasks to move from cpuset "%s"'
% fset.path)
if options.move:
log.info('--> moving all tasks from "%s" to "%s"...',
fset.path, tset.path)
selective_move(fset, tset, None, options.kthread, options.all)
else:
log.info('--> moving all kernel threads from "%s" to "%s"...',
fset.path, tset.path)
# this is a -k "move", so only move kernel threads
pids = []
for task in fset.tasks:
try: os.readlink('/proc/'+task+'/exe')
except: pids.append(task)
selective_move(fset, tset, pids, options.kthread, options.all)
log.info('done')
return
# default no options is list
list_sets(args)
def list_sets(args):
l = []
if isinstance(args, list):
for s in args: l.extend(cset.find_sets(s))
else:
l.extend(cset.find_sets(args))
for s in l:
if len(s.tasks) > 0:
if verbose:
log_detailed_task_table(s, ' ')
else:
log_detailed_task_table(s, ' ', 76)
else:
log.info(cset.summary(s))
def move(fromset, toset, plist=None):
log.debug('entering move, fromset=%s toset=%s list=%s', fromset, toset, plist)
if isinstance(fromset, str):
fset = cset.unique_set(fromset)
elif not isinstance(fromset, cset.CpuSet) and plist == None:
raise CpusetException(
"passed fromset=%s, which is not a string or CpuSet" % fromset)
else:
fset = fromset
if isinstance(toset, str):
tset = cset.unique_set(toset)
elif not isinstance(toset, cset.CpuSet):
raise CpusetException(
"passed toset=%s, which is not a string or CpuSet" % toset)
else:
tset = toset
if plist == None:
log.debug('moving default of all processes')
tset.tasks = fset.tasks
else:
tset.tasks = plist
def selective_move(fset, tset, plist=None, kthread=None, force=None):
log.debug('entering selective_move, fset=%s tset=%s plist=%s kthread=%s force=%s',
fset, tset, plist, kthread, force)
target = cset.unique_set(tset)
tasks = []
task_heap = []
task_check = []
utsk = 0
ktsk = 0
autsk = 0
aktsk = 0
utsknr = 0
ktsknr = 0
ktskb = 0
if fset:
task_check = cset.unique_set(fset).tasks
if plist:
task_heap = plist
else:
task_heap = cset.unique_set(fset).tasks
for task in task_heap:
try:
# kernel threads do not have an excutable image
os.readlink('/proc/'+task+'/exe')
autsk += 1
if fset and not force:
try:
task_check.index(task)
tasks.append(task)
utsk += 1
except:
log.debug(' task %s not running in %s, skipped',
task, fset.name)
utsknr += 1
else:
tasks.append(task)
utsk += 1
except:
aktsk += 1
try:
# this is in try because the task may not exist by the
# time we do this, in that case, just ignore it
if kthread:
if force:
tasks.append(task)
ktsk += 1
else:
if is_unbound(task):
tasks.append(task)
ktsk += 1
else:
log.debug(' kernel thread %s is bound, not adding',
task)
ktskb += 1
except:
log.debug(' kernel thread %s not found , perhaps it went away',
task)
ktsknr += 1
# ok, move 'em
log.debug('moving %d tasks to "%s"...', len(tasks), tset.name)
if len(tasks) == 0:
log.info('** no task matched move criteria')
if autsk > 0:
l = []
l.append('moving')
l.append(str(utsk))
l.append('userspace tasks')
if utsknr > 0:
l.append('- not moving')
l.append(str(utsknr))
l.append('tasks (not in fromset)')
log.info(' '.join(l))
if ktsk > 0 or kthread:
l = []
l.append('moving')
l.append(str(ktsk))
l.append('kernel threads')
if ktskb > 0:
l.append('- not moving')
l.append(str(ktskb))
l.append('threads (not unbound)')
log.info(' '.join(l))
if aktsk > 0 and force and not kthread and autsk == 0:
log.info('** not moving kernel threads since both --all and --kthread needed')
if ktsknr > 0:
l = []
l.append('--> not moving')
l.append(str(ktsknr))
l.append('tasks because they are missing (race)')
move(None, target, tasks)
def run(tset, args, usr_par=None, grp_par=None):
if isinstance(tset, str):
s = cset.unique_set(tset)
elif not isinstance(tset, cset.CpuSet):
raise CpusetException(
"passed set=%s, which is not a string or CpuSet" % tset)
else:
s = tset
log.debug('entering run, set=%s args=%s ', s.path, args)
set.active(s)
# check user
if usr_par:
try:
user = pwd.getpwnam(usr_par)[2]
except KeyError:
try:
user = pwd.getpwuid(int(usr_par))[2]
except:
raise CpusetException('unknown user: "%s"' % usr_par)
if grp_par:
try:
group = grp.getgrnam(grp_par)[2]
except KeyError:
try:
group = grp.getgrgid(int(grp_par))[2]
except:
raise CpusetException('unknown group: "%s"' % grp_par)
elif usr_par:
# if user is specified but group is not, and user is not root,
# then use the users group
if user != 0:
try:
group = grp.getgrnam('users')[2]
grp_par = True
except:
pass # just forget it
# move myself into target cpuset and exec child
move_pidspec(str(os.getpid()), s)
log.info('--> last message, executed args into cpuset "%s", new pid is: %s',
s.path, os.getpid())
# change user and group before exec
if grp_par: os.setgid(group)
if usr_par: os.setuid(user)
os.execvp(args[0], args)
def is_unbound(proc):
# FIXME: popen is slow... need to use sched_getaffinity() directly,
# but python doesn't have it... maybe use pyrex to wrap....
line = os.popen('/usr/bin/taskset -p ' + str(proc), 'r').readline()
aff = line.split()[-1]
log.debug('is_unbound, proc=%s aff=%s allcpumask=%s',
proc, aff, cset.allcpumask)
if aff == cset.allcpumask: return True
return False
def pidspec_to_list(pidspec, fset=None):
log.debug('entering pidspecToList, pidspec=%s', pidspec)
if fset:
if isinstance(fset, str): fset = cset.unique_set(fset)
elif not isinstance(fset, cset.CpuSet):
raise CpusetException("passed fset=%s, which is not a string or CpuSet" % fset)
log.debug('from-set specified as: %s', fset.path)
if not isinstance(pidspec, str):
raise CpusetException('pidspec=%s is not a string' % pidspec)
groups = pidspec.split(',')
plist = []
if fset: chktsk = fset.tasks
log.debug('parsing groups: %s', groups)
for sub in groups:
items = sub.split('-')
if len(items) == 1:
if not len(items[0]):
# two consecutive commas in pidspec, just continue processing
continue
# one pid in this group
if fset:
try:
chktsk.index(items[0])
plist.append(items[0])
log.debug(' added single pid: %s', items[0])
except:
log.debug(' task %s not running in %s, skipped', items[0], fset.name)
else:
plist.append(items[0])
log.debug(' added single pid: %s', items[0])
elif len(items) == 2:
# a range of pids, only include those that exist
rng = [str(x) for x in range(int(items[0]), int(items[1])+1)
if os.access('/proc/'+str(x), os.F_OK)]
if fset:
for tsk in rng:
try:
chktsk.index(tsk)
plist.append(tsk)
log.debug(' added task from range: %s', tsk)
except:
log.debug(' task %s not running in %s, skipped', tsk, fset.name)
else:
plist.extend(rng)
log.debug(' added range of pids from %s-%s: %s', items[0], items[1], rng)
else:
raise CpusetException('pidspec=%s has bad group=%s' % (pidspec, items))
log.debug('returning parsed pid list: %s', plist)
log.info('%s tasks match criteria', len(plist))
return plist
def move_pidspec(pidspec, toset, fset=None):
log.debug('entering move_pidspec, pidspec=%s toset=%s', pidspec, toset)
if not fset:
pids = pidspec_to_list(pidspec)
else:
# if fromset is specified, only move tasks that are in pidspec
# and are running in fromset
log.debug('specified fset=%s', fset)
pids = pidspec_to_list(pidspec, fset)
if len(pids) == 0:
raise CpusetException('tasks do not match all criteria, none moved')
move(None, toset, pids)
def task_detail(pid, width=65):
# get task details from /proc
pid = str(pid)
if not os.access('/proc/'+pid, os.F_OK):
raise CpusetException('task "%s" does not exist' % pid)
status = file('/proc/'+pid+'/status', 'r').readlines()
stdict = {}
for line in status:
try:
stdict[line.split()[0][:-1]] = line.split(':')[1].strip()
except:
pass # sometimes, we get an extra \n out of this file...
cmdline = file('/proc/'+pid+'/cmdline').readline()
# assume that a zero delimits the cmdline (it does now...)
cmdline = cmdline.replace('\0', ' ')
used = 0
out = pwd.getpwuid(int(stdict['Uid'].split()[0]))[0][:8].ljust(8)
used += 8
out += stdict['Pid'].rjust(6)
used += 6
out += stdict['PPid'].rjust(6)
used += 6
out += stdict['State'].split()[0].center(3)
used += 3
try:
os.readlink('/proc/'+pid+'/exe')
#prog = stdict['Name'] + ' '.join(cmdline.split()[1:])
prog = cmdline
except:
prog = '['+stdict['Name']+']'
if width == 0:
out += prog
else:
out += prog[:(width-used)]
return out
def task_detail_header(indent=None):
if indent == None: istr = ""
else: istr = indent
l = []
l.append(istr + 'USER PID PPID S TASK NAME')
l.append(istr + '-------- ----- ----- - ---------')
return l
def task_detail_table(pids, indent=None, width=None):
l = []
if indent == None: istr = ""
else: istr = indent
for task in pids:
if width: l.append(istr + task_detail(task, width))
else: l.append(istr + task_detail(task, 0))
return l
def log_detailed_task_table(set, indent=None, width=None):
log.debug("entering print_detailed_task_table, set=%s indent=%s width=%s",
set.path, indent, width)
l = []
l.append(cset.summary(set))
l.extend(task_detail_header(indent))
l.extend(task_detail_table(set.tasks, indent, width))
log.info("\n".join(l))

@ -0,0 +1,393 @@
"""Cpuset manipulation command
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, os, logging
from optparse import OptionParser, make_option
from cpuset import cset
from cpuset.util import *
from cpuset.commands.common import *
try: from cpuset.commands import proc
except: pass
global log
log = logging.getLogger('set')
help = 'create, modify and destroy cpusets'
usage = """%prog [options] [cpuset name]
This command is used to create, modify, and destroy cpusets.
Cpusets form a tree-like structure rooted at the root cpuset
which always includes all system CPUs and all system memory
nodes.
A cpuset is an organizational unit that defines a group of CPUs
and a group of memory nodes where a process or thread (i.e. task)
is allowed to run on. For non-NUMA machines, the memory node is
always 0 (zero) and cannot be set to anything else. For NUMA
machines, the memory node can be set to a similar specifcation
as the CPU definition and will tie those memory nodes to that
cpuset. You will usually want the memory nodes that belong to
the CPUs defined to be in the same cpuset.
A cpuset can have exclusive right to the CPUs defined in it.
This means that only this cpuset can own these CPUs. Similarly,
a cpuset can have exclusive right to the memory nodes defined in
it. This means that only this cpuset can own these memory
nodes.
Cpusets can be specified by name or by path; however, care
should be taken when specifying by name if the name is not
unique. This tool will generally not let you do destructive
things to non-unique cpuset names.
Cpusets are uniquely specified by path. The path starts at where
the cpusets filesystem is mounted so you generally do not have to
know where that is. For example, so specify a cpuset that is
called "two" which is a subset of "one" which in turn is a subset
of the root cpuset, use the path "/one/two" regardless of where
the cpusets filesystem is mounted.
When specifying CPUs, a so-called CPUSPEC is used. The CPUSPEC
will accept a comma-separated list of CPUs and inclusive range
specifications. For example, --cpu=1,3,5-7 will assign CPU1,
CPU3, CPU5, CPU6, and CPU7 to the specified cpuset.
Note that cpusets follow certain rules. For example, children
can only include CPUs that the parents already have. If you do
not follow those rules, the kernel cpuset subsystem will not let
you create that cpuset. For example, if you create a cpuset that
contains CPU3, and then attempt to create a child of that cpuset
with a CPU other than 3, you will get an error, and the cpuset
will not be active. The error is somewhat cryptic in that it is
usually a "Permission denied" error.
Memory nodes are specified with a MEMSPEC in a similar way to
the CPUSPEC. For example, --mem=1,3-6 will assign MEM1, MEM3,
MEM4, MEM5, and MEM6 to the specified cpuset.
Note that if you attempt to create or modify a cpuset with a
memory node specification that is not valid, you may get a
cryptic error message, "No space left on device", and the
modification will not be allowed.
When you destroy a cpuset, then the tasks running in that set are
moved to the parent of that cpuset. If this is not what you
want, then manually move those tasks to the cpuset of your choice
with the 'cset proc' command (see 'cset proc --help' for more
information).
EXAMPLES
Create a cpuset with the default memory specification:
# cset set --cpu=2,4,6-8 --set=new_set
This command creates a cpuset called "new_set" located
off the root cpuset which holds CPUS 2,4,6,7,8 and node 0
(interleaved) memory. Note that --set is optional, and
you can just specify the name for the new cpuset after
all arguments.
Create a cpuset that specifies both CPUs and memory nodes:
# cset set --cpu=3 --mem=3 /rad/set_one
Note that this command uses the full path method to
specify the name of the new cpuset "/rad/set_one". It
also names the new cpuset implicitily (i.e. no --set
option, although you can use that if you want to). If
the "set_one" name is unique, you can subsequently refer
to is just by that. Memory node 3 is assigned to this
cpuset as well as CPU 3.
The above commands will create the new cpusets, or if they
already exist, they will modify them to the new specifications."""
verbose = 0
options = [make_option('-l', '--list',
help = 'list the named cpuset(s); recursive list if also -r; '
'members if also -a',
action = 'store_true'),
make_option('-c', '--cpu',
help = 'create or modify cpuset in the specified '
'cpuset with CPUSPEC specification',
metavar = 'CPUSPEC'),
make_option('-m', '--mem',
help = 'specify which memory nodes to assign '
'to the created or modified cpuset',
metavar = 'MEMSPEC'),
make_option('-d', '--destroy',
help = 'destroy specified cpuset',
action = 'store_true'),
make_option('-s', '--set',
metavar = 'CPUSET',
help = 'specify cpuset'),
make_option('-a', '--all',
help = 'also do listing of member cpusets',
action = 'store_true'),
make_option('-r', '--recurse',
help = 'do recursive listing, for use with --list',
action = 'store_true'),
make_option('-v', '--verbose',
help = 'prints more detailed output, additive',
action = 'count'),
make_option('--cpu_exclusive',
help = 'mark this cpuset as owning its CPUs exclusively',
action = 'store_true'),
make_option('--mem_exclusive',
help = 'mark this cpuset as owning its MEMs exclusively',
action = 'store_true'),
]
def func(parser, options, args):
log.debug("entering func, options=%s, args=%s", options, args)
global verbose
if options.verbose: verbose = options.verbose
cset.rescan()
if options.list:
if options.set:
list_sets(options.set, options.recurse, options.all)
return
if len(args): list_sets(args, options.recurse, options.all)
else: list_sets('root', options.recurse, options.all)
return
if options.cpu or options.mem:
# create or modify cpuset
create_from_options(options, args)
return
if options.destroy:
if options.set: destroy_sets(options.set)
else: destroy_sets(args)
return
if options.cpu_exclusive or options.mem_exclusive:
# modification of existing cpusets for exclusivity
return
# default behavior if no options specified is list
log.debug('no options set, default is listing cpusets')
if len(args): list_sets(args, options.recurse, options.all)
else: list_sets('root', options.recurse, options.all)
def list_sets(tset, recurse=None, members=None):
log.debug('entering list_sets, tset=%s recurse=%s', tset, recurse)
sl = []
if isinstance(tset, list):
for s in tset: sl.extend(cset.find_sets(s))
else:
sl.extend(cset.find_sets(tset))
log.debug('total unique sets in passed tset: %d', len(sl))
if recurse: members = True
if members:
sl2 = []
for s in sl:
sl2.append(s)
if len(s.subsets) > 0:
sl2.extend(s.subsets)
if recurse:
for node in s.subsets:
for nd in cset.walk_set(node):
sl2.append(nd)
sl = sl2
pl = ['']
pl.extend(set_header(' '))
for s in sl:
if verbose:
pl.append(set_details(s,' ', 0))
else:
pl.append(set_details(s,' '))
log.info("\n".join(pl))
def destroy_sets(sets):
log.debug('enter destroy_sets, sets=%s', sets)
nl = []
try:
nl.extend(sets)
except:
nl.append(sets)
# check that sets passed are ok, will raise if one is bad
for s in nl:
st = cset.unique_set(s)
if len(st.subsets) > 0:
raise CpusetException('cpuset "%s" has subsets, delete them first'
% st.path)
# ok, good to go
for s in nl:
s = cset.unique_set(s)
log.info('--> processing cpuset "%s", moving %s tasks to parent "%s"...',
s.name, len(s.tasks), s.parent.path)
proc.move(s, s.parent)
log.info('deleting cpuset "%s"', s.path)
destroy(s)
log.info('done')
def destroy(name):
log.debug('entering destroy, name=%s', name)
if isinstance(name, str):
set = cset.unique_set(name)
elif not isinstance(name, cset.CpuSet):
raise CpusetException(
"passed name=%s, which is not a string or CpuSet" % name)
else:
set = name
if len(set.tasks) > 0:
log.debug('%i tasks still running in set %s', len(set.tasks), name)
raise CpusetException(
"trying to destroy cpuset %s with tasks running" % name)
os.rmdir(cset.CpuSet.basepath+set.path)
# fixme: perhaps reparsing the all the sets is not so efficient...
cset.rescan()
def create_from_options(options, args):
log.debug('entering create_from_options, options=%s args=%s', options, args)
# figure out target cpuset name, if --set not used, use first arg
if options.set:
tset = options.set
elif len(args) > 0:
tset = args[0]
else:
raise CpusetException('cpuset not specified')
cspec = None
mspec = None
cx = None
mx = None
if options.cpu:
cset.cpuspec_check(options.cpu)
cspec = options.cpu
if options.mem:
cset.memspec_check(options.mem)
mspec = options.mem
if options.cpu_exclusive: cx = options.cpu_exclusive
if options.mem_exclusive: mx = options.mem_exclusive
try:
create(tset, cspec, mspec, cx, mx)
if not mspec: modify(tset, memspec='0') # always need at least this
log.info('--> created cpuset "%s"', tset)
except CpusetExists:
modify(tset, cspec, mspec, cx, mx)
log.info('--> modified cpuset "%s"', tset)
active(tset)
def create(name, cpuspec, memspec, cx, mx):
log.debug('entering create, name=%s cpuspec=%s memspec=%s cx=%s mx=%s',
name, cpuspec, memspec, cx, mx)
try:
cset.unique_set(name)
except CpusetNotFound:
pass
except:
raise CpusetException('cpuset "%s" not unique, please specify by path' % name)
else:
raise CpusetExists('attempt to create already existing set: "%s"' % name)
# FIXME: check if name is a path here
os.mkdir(cset.CpuSet.basepath+'/'+name)
# fixme: perhaps reparsing the all the sets is not so efficient...
cset.rescan()
log.debug('created new cpuset "%s"', name)
modify(name, cpuspec, memspec, cx, mx)
def modify(name, cpuspec=None, memspec=None, cx=None, mx=None):
log.debug('entering modify, name=%s cpuspec=%s memspec=%s cx=%s mx=%s',
name, cpuspec, memspec, cx, mx)
if isinstance(name, str):
nset = cset.unique_set(name)
elif not isinstance(name, cset.CpuSet):
raise CpusetException(
"passed name=%s, which is not a string or CpuSet" % name)
else:
nset = name
log.debug('modifying cpuset "%s"', nset.name)
if cpuspec: nset.cpus = cpuspec
if memspec: nset.mems = memspec
if cx: nset.cpu_exclusive = cx
if mx: nset.mem_exclusive = mx
def active(name):
log.debug("entering active, name=%s", name)
if isinstance(name, str):
set = cset.unique_set(name)
elif not isinstance(name, cset.CpuSet):
raise CpusetException("passing bogus name=%s" % name)
else:
set = name
if set.cpus == '':
raise CpusetException('"%s" cpuset not active, no cpus defined' % set.path)
if set.mems == '':
raise CpusetException('"%s" cpuset not active, no mems defined' % set.path)
def set_header(indent=None):
if indent: istr = indent
else: istr = ''
l = []
# '1234567890-1234567890-1234567890-1234567890-1234567890'
l.append(istr + ' Name CPUs-X MEMs-X Tasks Subs Path')
l.append(istr + '------------ ---------- - ------- - ----- ---- ----------')
return l
def set_details(name, indent=None, width=75):
if isinstance(name, str):
set = cset.unique_set(name)
elif not isinstance(name, cset.CpuSet):
raise CpusetException("passing bogus set=%s" % name)
else:
set = name
if indent: istr = indent
else: istr = ''
l = []
used = 0
l.append(istr)
used += len(istr)
l.append(set.name.rjust(12))
used += 12
cs = set.cpus
if cs == '': cs = '*****'
l.append(cs.rjust(11))
used += 11
if set.cpu_exclusive:
l.append(' y')
else:
l.append(' n')
used += 2
cs = set.mems
if cs == '': cs = '*****'
l.append(cs.rjust(8))
used += 8
if set.mem_exclusive:
l.append(' y')
else:
l.append(' n')
used += 2
l.append(str(len(set.tasks)).rjust(6))
used += 6
l.append(str(len(set.subsets)).rjust(5))
used += 5
l.append(' ')
used += 1
if width == 0:
l.append(set.path)
else:
l.append(set.path[:(width-used)])
return ''.join(l)

@ -0,0 +1,472 @@
"""Shield supercommand
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, os, logging
from optparse import OptionParser, make_option
from cpuset.commands.common import *
from cpuset.commands import proc
from cpuset.commands import set
from cpuset import cset
from cpuset.util import *
global log
log = logging.getLogger('shield')
help = 'supercommand to set up and manage basic shielding'
usage = """%prog [options] [path/program]
This is a supercommand that creates basic cpu shielding. The
normal cset commands can of course be used to create this basic
shield, but the shield command combines many such commands to
create and manage a common type of cpu shielding setup.
The concept of shielding implies at minimum three cpusets, for
example: root, user and system. The root cpuset always exists in
all implementations of cpusets and contains all available CPUs on
the machine. The system cpuset is so named because normal system
tasks are made to run on it. The user cpuset is so named because
that is the "shielded" cpuset on which you would run your tasks
of interest.
Usually, CPU zero would be in the system set and the rest of the
CPUs would be in the user set. After creation of the cpusets,
all processes running in the root cpuset are moved to the system
cpuset. Thus any new processes or threads spawned from these
processes will also run the system cpuset.
If the optional --kthread=on option is given to the shield
command, then all kernel threads (with exception of the per-CPU
bound interrupt kernel threads) are also moved to the system set.
One executes processes on the shielded user cpuset with the
--exec subcommand or moves processes or threads to the shielded
cpuset with the --shield subcommand. Note that you do not need to
specify which cpuset a process or thread is running in initially
when using the --shield subcommand.
To create a shield, you would execute the shield command with the
--cpu option that specifies CPUSPEC argument that assigns CPUs to
be under the shield (this means assigned to the user cpuset, all
other cpus will be assigned to the system set).
For example:
# cset shield --cpu=3
On a 4-way machine, this command will dedicate the first
3 processors, CPU0-CPU2, for the system set (unshielded)
and only the last processor, CPU3, for the user set
(shielded).
The CPUSPEC will accept a comma separated list of CPUs and
inclusive range specifications. For example, --cpu=1,3,5-7 will
assign CPU1, CPU3, CPU5, CPU6, and CPU7 to the user (or shielded)
cpuset.
If you do not like the names "system" and "user" for the
unshielded and shielded sets respectively, or if those names are
used already, then use the --sysset and --userset options.
For example:
# cset shield --sysset=free --userset=cage --cpu=2,3 --kthread=on
The above command will use the name "free" for the unshielded
system cpuset, the name "cage" for the shielded user cpuset,
initialize these cpusets and dedicate CPU0 and CPU1 to the "free"
set and (on a 4-way machine) dedicate CPU2 and CPU3 to the "cage"
set. Further, the command moves all processes and threads,
including kernel threads from the root cpuset to the "free"
cpuset. Note however that if you do use the --syset/--userset
options, then you must continue to use those for every invocation
of the shield supercommand.
After initialization, you can run the process of interest on the
shielded cpuset with the --exec subcommand, or move processes or
threads already running to the shielded cpuset with the --shield
subcommand.
The PIDSPEC argument taken for the --shield (or -s) subcommand is
a comma separated list of PIDs or TIDs. The list can also
include brackets of PIDs or TIDs that are inclusive of the
endpoints.
For example:
1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700
# cset shield --shield=50-65
This command moves all processes and threads with PID or
TID in the range 50-65 inclusive, from any cpuset they may
be running in into the shielded user cpuset.
Note that the range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.
DANGER: Please note that there is no checking of processes you
request to move into the shield with the --shield command. This
means that the tool will happily move, for example, kernel
threads that are bound to specific CPUs with this command. You
can hang your system by indiscriminately specifying arbitrary
PIDs to the --shield command so please be careful.
The --unshield (or -u) subcommand will remove the specified
processes or threads from the shielded cpuset and move them into
the unshielded (or system) cpuset. This option also takes a
PIDSPEC argument, the same as for the --shield subcommand.
Both the --shield and the --unshield commands will also finally
output the number of tasks running in the shield and out of the
shield. If you do not specify a PIDSPEC to these commands, then
just that status is output. By specifying also a --verbose in
addition, then you will get a listing of every task that is
running in either the shield or out of the shield.
Using no subcommand, ie. only "cset shield", will output the
status of both shield and non-shield. Tasks will be listed if
--verbose is used.
You can adjust which CPUs are in the shielded cpuset by issuing
the --cpu subcommand again anytime after the shield has been
initialized.
For example if the original shield contained CPU0 and CPU1 in the
system set and CPU2 and CPU3 in the user set, if you then issue
the following command:
# cset shield --cpu=1,2,3
then that command will move CPU1 into the shielded "user" cpuset.
Any processes or threads that were running on CPU1 that belonged
to the unshielded "system" cpuset are migrated to CPU0 by the
system.
The --reset subcommand will in essence destroy the shield. For
example, if there was a shield on a 4-way machine with CPU0 in
system and CPUs 1-3 in user with processes running on the user
cpuset (i.e. in the shield), and a --reset subcommand was issued,
then all processes running in both system and user cpusets would
be migrated to the root cpuset (which has access to all CPUs and
never goes away), after which both system and user cpusets would
be destroyed.
Note that even though you can mix general usage of cpusets with
the shielding concepts described here, you generally will not
want to. For more complex shielding or usage scenarios, one
would generally use the normal cpuset commands (i.e. cset set
and proc) directly."""
USR_SET = '/user'
SYS_SET = '/system'
verbose = 0
# callback for --shield/--unshield, allows optional argument
# if no arg, then status of either shield or unshield is displayed
def shield_cb(option, opt_str, value, parser):
if value == None:
try:
arg = parser.rargs[0]
if arg[0] != '-':
# assign and consume argument
value = arg
del parser.rargs[0]
else:
value = True
except IndexError, err:
# no further arguments
value = True
except:
raise
setattr(parser.values, option.dest, value)
options = [make_option('-c', '--cpu',
metavar = 'CPUSPEC',
help = 'modifies or initializes the shield cpusets'),
make_option('-r', '--reset',
help = 'destroys the shield',
action = 'store_true'),
make_option('-e', '--exec',
help = 'executes args in the shield',
dest = 'exc',
action = 'store_true'),
make_option('--user',
help = 'use this USER for --exec (id or name)'),
make_option('--group',
help = 'use this GROUP for --exec (id or name)'),
make_option('-s', '--shield',
action = 'callback',
callback=shield_cb,
dest = 'shield',
metavar = 'PIDSPEC',
help = 'shield specified PIDSPEC of processes or threads'),
make_option('-u', '--unshield',
action = 'callback',
callback=shield_cb,
dest = 'unshield',
metavar = 'PIDSPEC',
help = 'remove specified PIDSPEC of processes or threads from shield'),
make_option('-k', '--kthread',
metavar = 'on|off',
choices = ['on', 'off'],
help = 'shield from unbound interrupt threads as well'),
make_option('-f', '--force',
help = 'force operation, use with care',
action = 'store_true'),
make_option('-v', '--verbose',
help = 'prints more detailed output, additive',
action = 'count'),
make_option('--sysset',
help = 'optionally specify system cpuset name'),
make_option('--userset',
help = 'optionally specify user cpuset name')
]
def func(parser, options, args):
log.debug("entering shield, options=%s, args=%s", options, args)
global verbose
if options.verbose: verbose = options.verbose
cset.rescan()
if options.sysset:
global SYS_SET
SYS_SET = options.sysset
if options.userset:
global USR_SET
USR_SET = options.userset
if (not options.cpu and not options.reset and not options.exc and
not options.shield and not options.unshield and not options.kthread):
shield_exists()
if len(args) == 0:
log.info("--> shielding system active with")
print_all_stats()
else:
exec_args(args, options.user, options.group)
return
if options.reset:
reset_shield()
return
# note that the following options fall through to others
# to allow for multiple options/commands on one cmdline
if options.cpu: make_shield(options.cpu, options.kthread)
elif options.kthread: make_kthread(options.kthread)
if options.exc: exec_args(args, options.user, options.group)
if options.shield:
shield_exists()
try:
log.info('--> shielding following pidspec: %s', options.shield)
if options.force:
proc.move_pidspec(options.shield, USR_SET)
else:
proc.move_pidspec(options.shield, USR_SET, SYS_SET)
log.info('done')
except Exception, err:
if str(err).find('Permission denied') != -1:
raise
if options.shield != True:
log.info(err)
log.info('--> bad pidspec: %s, shield state is:', options.shield)
else:
log.info('--> shielded state is:')
print_usr_stats()
if options.unshield:
shield_exists()
try:
log.info('--> unshielding following pidspec: %s', options.unshield)
if options.force:
proc.move_pidspec(options.unshield, SYS_SET)
else:
proc.move_pidspec(options.unshield, SYS_SET, USR_SET)
log.info('done')
except Exception, err:
if str(err).find('Permission denied') != -1:
raise
if options.unshield != True:
log.info('--> bad pidspec: %s, unshielded state is:', options.unshield)
else:
log.info('--> not shielded state is:')
print_sys_stats()
def print_all_stats():
print_sys_stats()
print_usr_stats()
def print_sys_stats():
if verbose and len(cset.unique_set(SYS_SET).tasks) > 0:
if verbose == 1:
proc.log_detailed_task_table(cset.unique_set(SYS_SET), ' ', 76)
else:
proc.log_detailed_task_table(cset.unique_set(SYS_SET), ' ')
else:
log.info(cset.summary(cset.unique_set(SYS_SET)))
def print_usr_stats():
if verbose and len(cset.unique_set(USR_SET).tasks) > 0:
if verbose == 1:
proc.log_detailed_task_table(cset.unique_set(USR_SET), ' ', 76)
else:
proc.log_detailed_task_table(cset.unique_set(USR_SET), ' ')
else:
log.info(cset.summary(cset.unique_set(USR_SET)))
def shield_exists():
try:
cset.unique_set(USR_SET)
cset.unique_set(SYS_SET)
return True
except CpusetNotFound:
log.info('can\'t find "%s" and "%s" cpusets on system...', SYS_SET, USR_SET)
raise CpusetException('--> shielding not active on system')
def reset_shield():
log.info("--> deactivating/reseting shielding")
shield_exists()
tasks = cset.unique_set(USR_SET).tasks
log.info('moving %s tasks from "%s" user set to root set...',
len(tasks), USR_SET)
proc.move(USR_SET, 'root')
tasks = cset.unique_set(SYS_SET).tasks
log.info('moving %s tasks from "%s" system set to root set...',
len(tasks), SYS_SET)
proc.move(SYS_SET, 'root')
log.info('deleting "%s" and "%s" sets', USR_SET, SYS_SET)
set.destroy(USR_SET)
set.destroy(SYS_SET)
log.info('done')
def make_shield(cpuspec, kthread):
memspec = '0' # FIXME: for numa, we probably want a more intelligent scheme
log.debug("entering make_shield, cpuspec=%s kthread=%s", cpuspec, kthread)
# create base cpusets for shield
cset.cpuspec_check(cpuspec)
cpuspec_inv = cset.cpuspec_inverse(cpuspec)
try:
shield_exists()
except:
log.debug("shielding does not exist, creating")
try:
set.create(USR_SET, cpuspec, memspec, True, False)
set.create(SYS_SET, cpuspec_inv, memspec, True, False)
except Exception, instance:
# unroll
try: set.destroy(USR_SET)
except: pass
try: set.destroy(SYS_SET)
except: pass
log.critical('--> failed to create shield, hint: do other cpusets exist?')
raise instance
log.info('--> activating shielding:')
else:
log.debug("shielding exists, modifying cpuspec")
# note, since we're going to modify the cpu assigments to these sets,
# they cannot be exclusive, the following modify() calls will make
# them exclusive again
cset.unique_set(USR_SET).cpu_exclusive = False
cset.unique_set(SYS_SET).cpu_exclusive = False
set.modify(USR_SET, cpuspec, memspec, False, False)
set.modify(SYS_SET, cpuspec_inv, memspec, False, False)
# reset cpu exlusivity
cset.unique_set(USR_SET).cpu_exclusive = True
cset.unique_set(SYS_SET).cpu_exclusive = True
log.info('--> shielding modified with:')
# move root tasks into system set
root_tasks = cset.unique_set('/').tasks
log.debug("number of root tasks are: %s", len(root_tasks))
# figure out what in root set is not a kernel thread
tasks = []
for task in root_tasks:
try:
os.readlink('/proc/'+task+'/exe')
tasks.append(task)
except:
pass
if len(tasks) != 0:
log.info("moving %s tasks from root into system cpuset...", len(tasks))
proc.move('root', SYS_SET, tasks)
# move kernel theads into system set if asked for
if kthread == 'on':
root_tasks = cset.unique_set('/').tasks
tasks = []
for task in root_tasks:
try:
if proc.is_unbound(task): tasks.append(task)
except:
pass
if len(tasks) != 0:
log.info("kthread shield activated, moving %s tasks into system cpuset...",
len(tasks))
proc.move('root', SYS_SET, tasks)
# print out stats
print_all_stats()
def make_kthread(state):
log.debug("entering make_kthread, state=%s", state)
shield_exists()
if state == 'on':
log.info('--> activating kthread shielding')
root_tasks = cset.unique_set('/').tasks
log.debug('root set has %d tasks, checking for unbound',
len(root_tasks))
tasks = []
for task in root_tasks:
try:
if proc.is_unbound(task): tasks.append(task)
except:
pass
if len(tasks) != 0:
log.debug("total root tasks %s", len(root_tasks))
log.info("kthread shield activated, moving %s tasks into system cpuset...",
len(tasks))
proc.move('root', SYS_SET, tasks)
if verbose == 1:
proc.log_detailed_task_table(tasks, ' ', 76)
elif verbose > 1:
proc.log_detailed_task_table(tasks, ' ')
else:
log.info('--> deactivating kthread shielding')
usr_tasks = cset.unique_set(SYS_SET).tasks
tasks = []
for task in usr_tasks:
try:
os.readlink('/proc/'+task+'/exe')
except:
tasks.append(task)
if len(tasks) != 0:
log.info("moving %s tasks into root cpuset...", len(tasks))
proc.move(SYS_SET, '/', tasks)
if verbose == 1:
proc.log_detailed_task_table(tasks, ' ', 76)
elif verbose > 1:
proc.log_detailed_task_table(tasks, ' ')
log.info('done')
def exec_args(args, upar, gpar):
log.debug("entering exec_args, args=%s", args)
shield_exists()
proc.run(USR_SET, args, upar, gpar)

@ -0,0 +1,432 @@
"""Cpuset class and cpuset graph, importing module will create model
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import os, re, sys, logging
if __name__ == '__main__':
sys.path.insert(0, "..")
logging.basicConfig()
from cpuset.util import *
global log
log = logging.getLogger('cset')
class CpuSet(object):
# sets is a class variable dict that keeps track of all
# cpusets discovered such that we can link them in properly.
# The basepath is it's base path, the sets are indexed via
# a relative path from this basepath.
sets = {}
basepath = ''
def __init__(self, path=None):
log.debug("initializing CpuSet")
if (path == None):
# recursively find all cpusets and link together
# note: a breadth-first search could do this in one
# pass, but there are never many cpusets, so
# that optimization is left for the future
log.debug("finding all cpusets")
path = self.locate_cpusets()
CpuSet.basepath = path
log.debug("creating root node at %s", path)
self.__root = True
self.name = 'root'
self.path = '/'
self.parent = self
if (CpuSet.sets):
del CpuSet.sets
CpuSet.sets = {}
CpuSet.sets[self.path] = self
# bottom-up search otherwise links will not exist
log.debug("starting bottom-up discovery walk...")
for dir, dirs, files in os.walk(path, topdown=False):
log.debug("*** walking %s", dir)
if dir != CpuSet.basepath:
node = CpuSet(dir)
else:
node = self
node.subsets = []
for sub in dirs:
if len(sub) > 0:
relpath = os.path.join(dir,sub).replace(CpuSet.basepath, '')
else:
relpath = '/'
node.subsets.append(CpuSet.sets[relpath])
log.debug("%s has %i subsets: [%s]", dir,
len(node.subsets), '|'.join(dirs))
log.debug("staring top-down parenting walk...")
for dir, dirs, files in os.walk(path):
dir = dir.replace(CpuSet.basepath, '')
if len(dir) == 0: dir = '/'
node = CpuSet.sets[dir]
log.debug("~~~ walking %s", node.path)
if dir == '/':
log.debug("parent is self (root cpuset), skipping")
else:
parpath = dir[0:dir.rfind('/')]
log.debug('parpath decodes to: %s from dir of: %s', parpath, dir)
if CpuSet.sets.has_key(parpath):
log.debug("parent is %s", parpath)
node.parent = CpuSet.sets[parpath]
else:
log.debug("parent is root cpuset")
node.parent = CpuSet.sets['/']
log.debug("found %i cpusets", len(CpuSet.sets))
else:
# one new cpuset node
log.debug("new cpuset node absolute: %s", path)
if len(path) > len(CpuSet.basepath):
path = path.replace(CpuSet.basepath, '')
else:
path = '/'
log.debug(" relative: %s", path)
if CpuSet.sets.has_key(path):
log.debug("the cpuset %s already exists, skipping", path)
self = CpuSet.sets[path] # questionable....
return
cpus = CpuSet.basepath + path + "/cpus"
if not os.access(cpus, os.F_OK):
# not a cpuset directory
str = '%s is not a cpuset directory' % (CpuSet.basepath + path)
log.error(str)
raise CpusetException(str)
self.__root = False
self.read_cpuset(path)
CpuSet.sets[path] = self
def locate_cpusets(self):
log.debug("locating cpuset filesystem...")
cpuset = re.compile(r"none (/.+) cpuset .+")
cgroup = re.compile(r"none (/.+) cgroup .+")
path = None
f = file("/proc/mounts")
for line in f:
res = cpuset.search(line)
if res:
path = res.group(1)
break
else:
if cgroup.search(line):
groups = line.split()
if re.search("cpuset", groups[3]):
path = groups[1]
break
f.close()
if not path:
# mounted cpusets not found, so mount them
# FIXME: provide default mount directory from config file
if not os.access("/cpusets", os.F_OK):
os.mkdir("/cpusets")
ret = os.system("mount -t cpuset none /cpusets")
if ret:
raise CpusetException(
'mount of cpuset filesystem failed, do you have permission?')
path = "/cpusets"
log.debug("cpusets mounted at: " + path)
return path
def read_cpuset(self, path):
log.debug("reading cpuset passed relpath: %s", path)
self.path = path
log.debug("...path=%s", path)
self.name = path[path.rfind('/')+1:]
log.debug("...name=%s", self.name)
# Properties of cpuset node
def delprop(self):
raise AttributeError, "deletion of properties not allowed"
def getcpus(self):
f = file(CpuSet.basepath+self.path+"/cpus")
return f.readline()[:-1]
def setcpus(self, newval):
cpuspec_check(newval)
f = file(CpuSet.basepath+self.path+"/cpus",'w')
f.write(str(newval))
f.close()
log.debug("-> prop_set %s.cpus = %s", self.path, newval)
cpus = property(fget=getcpus, fset=setcpus, fdel=delprop, doc="CPU specifier")
def getmems(self):
f = file(CpuSet.basepath+self.path+"/mems")
return f.readline()[:-1]
def setmems(self, newval):
# FIXME: check format for correctness
f = file(CpuSet.basepath+self.path+"/mems",'w')
f.write(str(newval))
f.close()
log.debug("-> prop_set %s.mems = %s", self.path, newval)
mems = property(getmems, setmems, delprop, "Mem node specifier")
def getcpuxlsv(self):
f = file(CpuSet.basepath+self.path+"/cpu_exclusive")
if f.readline()[:-1] == '1':
return True
else:
return False
def setcpuxlsv(self, newval):
log.debug("-> prop_set %s.cpu_exclusive = %s", self.path, newval)
f = file(CpuSet.basepath+self.path+"/cpu_exclusive",'w')
if newval:
f.write('1')
else:
f.write('0')
f.close()
cpu_exclusive = property(getcpuxlsv, setcpuxlsv, delprop,
"CPU exclusive flag")
def getmemxlsv(self):
f = file(CpuSet.basepath+self.path+"/mem_exclusive")
if f.readline()[:-1] == '1':
return True
else:
return False
def setmemxlsv(self, newval):
log.debug("-> prop_set %s.mem_exclusive = %s", self.path, newval)
f = file(CpuSet.basepath+self.path+"/mem_exclusive",'w')
if newval:
f.write('1')
else:
f.write('0')
f.close()
mem_exclusive = property(getmemxlsv, setmemxlsv, delprop,
"Memory exclusive flag")
def gettasks(self):
f = file(CpuSet.basepath+self.path+"/tasks")
lst = []
for task in f: lst.append(task[:-1])
return lst
def settasks(self, tasklist):
if len(tasklist) > 3:
pb = ProgressBar(len(tasklist), '=')
tick = 0
prog = True
else:
prog = False
for task in tasklist:
try:
f = file(CpuSet.basepath+self.path+"/tasks",'w')
f.write(task)
f.close()
except Exception, err:
if str(err).find('Permission denied') != -1:
raise
# if here, means process is already gone, racy stuff...
if prog:
tick += 1
pb(tick)
log.debug("-> prop_set %s.tasks set with %s tasks", self.path,
len(tasklist))
tasks = property(gettasks, settasks, delprop, "Task list")
#
# Helper functions
#
def unique_set(name):
"""find a unique cpuset by name or path, raise if multiple sets found"""
log.debug("entering unique_set, name=%s", name)
if isinstance(name, CpuSet): return name
nl = find_sets(name)
if len(nl) > 1:
raise CpusetNotUnique('cpuset name "%s" not unique: %s' % (name,
[x.path for x in nl]) )
return nl[0]
def find_sets(name):
"""find cpusets by name or path, return None if not found"""
log = logging.getLogger("cset.find_sets")
log.debug('finding "%s" in cpusets', name)
nodelist = []
if name.find('/') == -1:
log.debug("find by name")
name = name.lower()
if name == 'root':
log.debug("returning root set")
nodelist.append(RootSet)
else:
log.debug("walking from: %s", RootSet.name)
for node in walk_set(RootSet):
if node.name == name:
log.debug('... found node "%s"', name)
nodelist.append(node)
else:
log.debug("find by path")
if name in CpuSet.sets:
log.debug('... found node "%s"', CpuSet.sets[name].name)
nodelist.append(CpuSet.sets[name])
if len(nodelist) == 0:
raise CpusetNotFound('cpuset "%s" not found in cpusets' % name)
return nodelist
def walk_set(set):
""" generator for walking cpuset graph, breadth-first, more or less... """
log = logging.getLogger("cset.walk_set")
for node in set.subsets:
log.debug("+++ yield %s", node.name)
yield node
for node in set.subsets:
for result in walk_set(node):
log.debug("++++++ yield %s", node.name)
yield result
def rescan():
""" re-read the cpuset directory to sync system with data structs """
global RootSet, maxcpu, allcpumask
RootSet = CpuSet()
# figure out system properties
maxcpu = int(RootSet.cpus[-1])
allcpumask = calc_cpumask(maxcpu)
def cpuspec_check(cpuspec):
""" check format of cpuspec for validity """
log.debug("cpuspec_check(%s), maxcpu=%s", cpuspec, maxcpu)
groups = cpuspec.split(',')
if int(groups[-1].split('-')[-1]) > int(maxcpu):
str = 'CPUSPEC "%s" specifies higher max(%s) than available(%s)' % \
(cpuspec, groups[-1].split('-')[-1], maxcpu)
log.debug(str)
raise CpusetException(str)
mo = re.search("[^0-9,\-]", cpuspec)
if mo:
str = 'CPUSPEC "%s" contains invalid charaters: %s' % (cpuspec, mo.group())
log.debug(str)
raise CpusetException(str)
def memspec_check(memspec):
"""check format of memspec for validity """
# FIXME: look under /sys/devices/system/node for numa memory node
# information and check the memspec that way, currently we only do
# a basic check
log.debug("memspec_check(%s)", memspec)
mo = re.search("[^0-9,\-]", memspec)
if mo:
str = 'MEMSPEC "%s" contains invalid charaters: %s' % (memspec, mo.group())
log.debug(str)
raise CpusetException(str)
def cpuspec_inverse(cpuspec):
""" calculate inverse of cpu specification """
cpus = [0 for x in range(maxcpu+1)]
groups = cpuspec.split(',')
log.debug("cpuspec_inverse(%s) maxcpu=%d groups=%d",
cpuspec, maxcpu, len(groups))
for set in groups:
items = set.split('-')
if len(items) == 1:
if not len(items[0]):
# common error of two consecutive commas in cpuspec,
# just ignore it and keep going
continue
cpus[int(items[0])] = 1
elif len(items) == 2:
for x in range(int(items[0]), int(items[1])+1):
cpus[x] = 1
else:
raise CpusetException("cpuspec(%s) has bad group %s" % (cpuspec, set))
log.debug("cpuspec array: %s", cpus)
# calculate inverse of array
for x in range(0, len(cpus)):
if cpus[x] == 0:
cpus[x] = 1
else:
cpus[x] = 0
log.debug(" inverse: %s", cpus)
# build cpuspec expression
nspec = ""
ingrp = False
for x in range(0, len(cpus)):
if cpus[x] == 0 and ingrp:
nspec += str(begin)
if x > begin+1:
if cpus[x] == 1:
nspec += '-' + str(x)
else:
nspec += '-' + str(x-1)
ingrp = False
if cpus[x] == 1:
if not ingrp:
if len(nspec): nspec += ','
begin = x
ingrp = True
if x == len(cpus)-1:
nspec += str(begin)
if x > begin:
nspec += '-' + str(x)
log.debug("inverse cpuspec: %s", nspec)
return nspec
def summary(set):
"""return summary of cpuset with number of tasks running"""
log.debug("entering summary, set=%s", set.path)
return ('"%s" cpuset of: %+10s cpu, with: %+5s tasks running' %
(set.name, set.cpus, len(set.tasks)) )
def calc_cpumask(max):
all = 1
ii = 1
while ii < max+1:
all |= 1 << ii
ii += 1
return "%x" % all
# Test if stand-alone execution
if __name__ == '__main__':
rescan()
# first create them, then find them
try:
os.makedirs(CpuSet.basepath+'/csettest/one/x')
os.mkdir(CpuSet.basepath+'/csettest/one/y')
os.makedirs(CpuSet.basepath+'/csettest/two/x')
os.mkdir(CpuSet.basepath+'/csettest/two/y')
except:
pass
print 'Max cpu on system:', maxcpu
print 'All cpu mask: 0x%s' % allcpumask
print '------- find_sets tests --------'
print 'Find by root of "root" -> ', find_sets("root")
print 'Find by path of "/" -> ', find_sets("/")
print 'Find by path of "/csettest/one" -> ', find_sets("/csettest/one")
print 'Find by name of "one" -> ', find_sets("one")
print 'Find by path of "/csettest/two" -> ', find_sets("/csettest/two")
print 'Find by name of "two" -> ', find_sets("two")
print 'Find by path of "/csettest/one/x" -> ', find_sets("/csettest/one/x")
print 'Find by name of "x" -> ', find_sets("x")
print 'Find by path of "/csettest/two/y" -> ', find_sets("/csettest/two/y")
print 'Find by name of "y" -> ', find_sets("y")
try:
node = find_sets("cantfindmenoway")
print 'Found "cantfindmenoway??!? -> ', node
except CpusetException, err:
print 'Caught exeption for non-existant set (correctly)-> ', err

@ -0,0 +1,212 @@
"""Front end command line tool for Linux cpusets
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, os
from optparse import OptionParser
import cpuset.commands
from cpuset.commands.common import CmdException
from cpuset.util import CpusetException
#
# The commands map
#
class Commands(dict):
"""Commands class. It performs on-demand module loading
"""
def canonical_cmd(self, key):
"""Return the canonical name for a possibly-shortenned
command name.
"""
candidates = [cmd for cmd in self.keys() if cmd.startswith(key)]
if not candidates:
log.error('Unknown command: %s', key)
log.error('Try "%s help" for a list of supported commands', prog)
sys.exit(1)
elif len(candidates) > 1:
log.error('Ambiguous command: %s', key)
log.error('Candidates are: %s', ', '.join(candidates))
sys.exit(1)
return candidates[0]
def __getitem__(self, key):
"""Return the command python module name based.
"""
global prog
cmd_mod = self.get(key) or self.get(self.canonical_cmd(key))
__import__('cpuset.commands.' + cmd_mod)
return getattr(cpuset.commands, cmd_mod)
commands = Commands({
'shield': 'shield',
'set': 'set',
# 'mem': 'mem',
'proc': 'proc',
})
supercommands = (
'shield',
)
def _print_helpstring(cmd):
print ' ' + cmd + ' ' * (12 - len(cmd)) + commands[cmd].help
def print_help():
print 'Usage: %s <command> [options]' % os.path.basename(sys.argv[0])
print
print 'Generic commands:'
print ' help print the detailed command usage'
print ' version display version information'
print ' copyright display copyright information'
cmds = commands.keys()
cmds.sort()
print
print 'Super commands (high-level and multi-function):'
for cmd in supercommands:
_print_helpstring(cmd)
print
print 'Regular commands:'
for cmd in cmds:
if not cmd in supercommands:
_print_helpstring(cmd)
def main():
global prog
prog = os.path.basename(sys.argv[0])
global logfile
logfile = None
if len(sys.argv) < 2:
print >> sys.stderr, 'usage: %s <command>' % prog
print >> sys.stderr, \
' Try "%s --help" for a list of supported commands' % prog
sys.exit(1)
# configure logging
import logging
console = logging.StreamHandler(sys.stdout)
console.setLevel(logging.INFO)
formatter = logging.Formatter(prog + ': %(message)s')
console.setFormatter(formatter)
logging.getLogger('').addHandler(console)
global log
log = logging.getLogger('')
log.setLevel(logging.DEBUG)
cmd = sys.argv[1]
if cmd in ['-l', '--log']:
# FIXME: very fragile
logfile = sys.argv[2]
#trace = logging.FileHandler('/var/log/cset.log', 'w')
trace = logging.FileHandler(logfile, 'a')
trace.setLevel(logging.DEBUG)
formatter = logging.Formatter('%(asctime)s %(name)-6s %(levelname)-8s %(message)s',
'%y%m%d-%H:%M:%S')
trace.setFormatter(formatter)
logging.getLogger('').addHandler(trace)
log.debug("---------- STARTING ----------")
from cpuset.version import version
log.debug('Cpuset (cset) %s' % version)
del(sys.argv[2])
del(sys.argv[1])
if len(sys.argv) > 1:
cmd = sys.argv[1]
else:
log.debug("no arguments to process, exit")
print >> sys.stderr, 'usage: %s <command>' % prog
print >> sys.stderr, \
' Try "%s --help" for a list of supported commands' % prog
sys.exit(1)
if cmd in ['-h', '--help']:
if len(sys.argv) >= 3:
cmd = commands.canonical_cmd(sys.argv[2])
sys.argv[2] = '--help'
else:
print_help()
sys.exit(0)
if cmd == 'help':
if len(sys.argv) == 3 and not sys.argv[2] in ['-h', '--help']:
cmd = commands.canonical_cmd(sys.argv[2])
if not cmd in commands:
log.error('help: "%s" command unknown' % cmd)
sys.exit(1)
sys.argv[0] += ' %s' % cmd
command = commands[cmd]
parser = OptionParser(usage = command.usage,
option_list = command.options)
from pydoc import pager
pager(parser.format_help())
else:
print_help()
sys.exit(0)
if cmd in ['-v', '--version', 'version']:
from cpuset.version import version
log.info('Cpuset (cset) %s' % version)
sys.exit(0)
if cmd in ['copyright']:
log.info(__copyright__)
sys.exit(0)
# re-build the command line arguments
cmd = commands.canonical_cmd(cmd)
sys.argv[0] += ' %s' % cmd
del(sys.argv[1])
log.debug('cmdline: ' + ' '.join(sys.argv))
try:
debug_level = int(os.environ['CSET_DEBUG_LEVEL'])
except KeyError:
debug_level = 0
except ValueError:
log.error('Invalid CSET_DEBUG_LEVEL environment variable')
sys.exit(1)
try:
# importing the cset class creates the model
log.debug("creating cpuset model")
import cpuset.cset
command = commands[cmd]
usage = command.usage.split('\n')[0].strip()
parser = OptionParser(usage = usage, option_list = command.options)
parser.disable_interspersed_args()
options, args = parser.parse_args()
command.func(parser, options, args)
except (ValueError, OSError, IOError, CpusetException, CmdException), err:
log.critical('%s: %s' % (cmd, err))
if str(err).find('Permission denied') != -1:
log.critical('insufficient permissions, you probably need to be root')
if str(err).find('invalid literal') != -1:
log.critical('option not understood')
if debug_level:
raise
else:
sys.exit(2)
except KeyboardInterrupt:
sys.exit(1)
sys.exit(0)

@ -0,0 +1,21 @@
"""Task class
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""

@ -0,0 +1,92 @@
"""Utility functions
"""
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
import sys, time
class CpusetException(Exception):
pass
class CpusetAbort(CpusetException):
pass
class CpusetNotFound(CpusetException):
pass
class CpusetNotUnique(CpusetException):
pass
class CpusetExists(CpusetException):
pass
# a twirling bar progress indicator
class TwirlyBar:
def __init__(self):
import sys
self.__dict__['__call__'] = self.tick
self.__state = 0
self.__bar = ('|', '/', '-', '\\')
def tick(self):
print '\b' + self.__bar[self.__state] + '\b',
self.__state = self.__state + 1
if self.__state > 3: self.__state = 0
def fastick(self):
for x in range(10):
self.tick()
time.sleep(0.04)
# a progress bar indicator
class ProgressBar:
def __init__(self, finalcount, progresschar=None):
self.__dict__['__call__'] = self.progress
self.finalcount=finalcount
self.blockcount=0
# Use ascii block char for progress if none passed
if not progresschar:
self.block=chr(178)
else:
self.block=progresschar
self.f=sys.stdout
if not self.finalcount: return
self.f.write('[')
for i in range(50): self.f.write(' ')
self.f.write(']%')
for i in range(52): self.f.write('\b')
def progress(self, count):
count=min(count, self.finalcount)
if self.finalcount:
percentcomplete=int(round(100*count/self.finalcount))
if percentcomplete < 1: percentcomplete=1
else:
percentcomplete=100
blockcount=int(percentcomplete/2)
if blockcount > self.blockcount:
for i in range(self.blockcount,blockcount):
self.f.write(self.block)
self.f.flush()
if percentcomplete == 100: self.f.write("]\n")
self.blockcount=blockcount

@ -0,0 +1,19 @@
__copyright__ = """
Copyright (C) 2008 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
"""
version = '1.04'

47
cset

@ -0,0 +1,47 @@
#!/usr/bin/env python
__copyright__ = """
Copyright (C) 2007 Novell Inc.
Author: Alex Tsariounov <alext@novell.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License version 2 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
---------------------------------------------------------------------
Substrate code and ideas taken from the excellent stgit 0.13, see
https://gna.org/projects/stgit and http://www.procode.org/stgit/
Stacked GIT is under GPL V2 or later.
---------------------------------------------------------------------
"""
import sys, os
# Try to detect where it is run from and set prefix and the search path.
# It is assumed that the user installed cpuset using the --prefix= option
prefix, bin = os.path.split(sys.path[0])
if bin == 'bin' and prefix != sys.prefix:
sys.prefix = prefix
sys.exec_prefix = prefix
major, minor = sys.version_info[0:2]
local_path = [os.path.join(prefix, 'lib', 'python'),
os.path.join(prefix, 'lib', 'python%s.%s' % (major, minor)),
os.path.join(prefix, 'lib', 'python%s.%s' % (major, minor),
'site-packages')]
sys.path = local_path + sys.path
from cpuset.main import main
if __name__ == '__main__':
main()

@ -0,0 +1,70 @@
MAN1_TXT=$(wildcard cset*.txt)
DOC_HTML=$(patsubst %.txt,%.html,$(MAN1_TXT) $(MAN7_TXT))
ARTICLES =
# with their own formatting rules.
SP_ARTICLES = tutorial
DOC_HTML += $(patsubst %,%.html,$(ARTICLES) $(SP_ARTICLES))
DOC_PDF += $(patsubst %,%.pdf,$(ARTICLES) $(SP_ARTICLES))
DOC_MAN1=$(patsubst %.txt,%.1,$(MAN1_TXT))
DOC_MAN7=$(patsubst %.txt,%.7,$(MAN7_TXT))
prefix?=$(HOME)/man
mandir?=$(prefix)
man1dir=$(mandir)/man1
# DESTDIR=
ASCIIDOC=asciidoc --unsafe
ASCIIDOC_EXTRA =
INSTALL?=install
#
# Please note that there is a minor bug in asciidoc.
# The version after 6.0.3 _will_ include the patch found here:
# http://marc.theaimsgroup.com/?l=git&m=111558757202243&w=2
#
# Until that version is released you may have to apply the patch
# yourself - yes, all 6 characters of it!
#
all: html man
html: $(DOC_HTML)
pdf: $(DOC_PDF)
$(DOC_HTML) $(DOC_MAN1) $(DOC_MAN7): asciidoc.conf
man: man1
man1: $(DOC_MAN1)
install: man
$(INSTALL) -d -m755 $(DESTDIR)$(man1dir)
$(INSTALL) -m644 $(DOC_MAN1) $(DESTDIR)$(man1dir)
cd $(DESTDIR)$(man1dir); gzip *.1
-include doc.dep
clean:
#rm -f *.xml *.html *.pdf *.1 *.7 doc.dep
rm -f *.xml *.pdf doc.dep
%.html : %.txt
$(ASCIIDOC) -b xhtml11 -d manpage -f asciidoc.conf $(ASCIIDOC_EXTRA) $<
%.1 %.7 : %.xml
xmlto -m callouts.xsl man $<
%.xml : %.txt
$(ASCIIDOC) -b docbook -d manpage -f asciidoc.conf $<
%.pdf: %.xml
xmlto pdf $<
# special formatting rules
tutorial.html : %.html : %.txt
$(ASCIIDOC) -b xhtml11 -d article -f tutorial.conf $(ASCIIDOC_EXTRA) $<
tutorial.xml : %.xml : %.txt
$(ASCIIDOC) -b docbook -d article -f tutorial.conf $<

@ -0,0 +1,59 @@
## csetlink: macro
#
# Usage: csetlink:command[manpage-section]
#
# Note, {0} is the manpage section, while {target} is the command.
#
# Show cset link as: <command>(<section>); if section is defined, else just show
# the command.
[attributes]
caret=^
startsb=&#91;
endsb=&#93;
tilde=&#126;
ifdef::backend-docbook[]
[manlink-inlinemacro]
{0%{target}}
{0#<citerefentry>}
{0#<refentrytitle>{target}</refentrytitle><manvolnum>{0}</manvolnum>}
{0#</citerefentry>}
endif::backend-docbook[]
ifdef::backend-docbook[]
# "unbreak" docbook-xsl v1.68 for manpages. v1.69 works with or without this.
[listingblock]
<example><title>{title}</title>
<literallayout>
|
</literallayout>
{title#}</example>
endif::backend-docbook[]
ifdef::backend-xhtml11[]
[manlink-inlinemacro]
<a href="{target}.html">{target}{0?({0})}</a>
endif::backend-xhtml11[]
# csetlink
ifdef::backend-docbook[]
[csetlink-inlinemacro]
<citerefentry>
<refentrytitle>cset-{target}</refentrytitle><manvolnum>1</manvolnum>
</citerefentry>
endif::backend-docbook[]
ifdef::backend-xhtml11[]
[csetlink-inlinemacro]
<a href="cset-{target}.html">cset-{target}(1)</a>
endif::backend-xhtml11[]
# csetdesc
[csetdesc-inlinemacro]
{sys:../cset help|grep " {target}" | tr -s ' '| cut -d' ' -f3-}
[csetdesc-blockmacro]
csetlink:{target}[]::
csetdesc:{target}[]

@ -0,0 +1,30 @@
<!-- callout.xsl: converts asciidoc callouts to man page format -->
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0">
<xsl:template match="co">
<xsl:value-of select="concat('\fB(',substring-after(@id,'-'),')\fR')"/>
</xsl:template>
<xsl:template match="calloutlist">
<xsl:text>.sp&#10;</xsl:text>
<xsl:apply-templates/>
<xsl:text>&#10;</xsl:text>
</xsl:template>
<xsl:template match="callout">
<xsl:value-of select="concat('\fB',substring-after(@arearefs,'-'),'. \fR')"/>
<xsl:apply-templates/>
<xsl:text>.br&#10;</xsl:text>
</xsl:template>
<!-- sorry, this is not about callouts, but attempts to work around
spurious .sp at the tail of the line docbook stylesheets seem to add -->
<xsl:template match="simpara">
<xsl:variable name="content">
<xsl:apply-templates/>
</xsl:variable>
<xsl:value-of select="normalize-space($content)"/>
<xsl:if test="not(ancestor::authorblurb) and
not(ancestor::personblurb)">
<xsl:text>&#10;&#10;</xsl:text>
</xsl:if>
</xsl:template>
</xsl:stylesheet>

@ -0,0 +1,173 @@
.\" ** You probably do not want to edit this file directly **
.\" It was generated using the DocBook XSL Stylesheets (version 1.69.1).
.\" Instead of manually editing it, you probably should edit the DocBook XML
.\" source for it and then use the DocBook XSL Stylesheets to regenerate it.
.TH "CSET\-PROC" "1" "05/30/2008" "" ""
.\" disable hyphenation
.nh
.\" disable justification (adjust text to left margin only)
.ad l
.SH "NAME"
cset\-proc \- manage processes running in cpusets
.SH "SYNOPSIS"
.sp
.nf
\fIcset\fR [cset options] \fIproc\fR [proc options] [args]
\fIcset\fR proc \-\-help
\fIcset\fR proc
\fIcset\fR proc my_set my_other_set
\fIcset\fR proc \-\-list \-\-set my_set
\fIcset\fR proc \-\-exec my_set /opt/software/my_code \-\-my_opt_1
\fIcset\fR proc \-\-set my_set \-\-exec /opt/software/my_code \-\-my_opt_1
\fIcset\fR proc \-\-move 2442,3000\-3200 my_set
\fIcset\fR proc \-\-move \-\-pid=2442,3000\-3200 \-\-toset=my_set
\fIcset\fR proc \-\-move \-\-fromset=my_set_1 \-\-toset=my_set_2
\fIcset\fR proc \-\-move \-\-pid=42 \-\-fromset=/group1/myset \-\-toset=/group2/yourset
.fi
.SH "OPTIONS"
.TP
\-h, \-\-help
prints the list of options for this command
.TP
\-l, \-\-list
list processes in the specified cpuset
.TP
\-e, \-\-exec
execute arguments in the specified cpuset
.TP
\-u USER, \-\-user=USER
use this USER to \-\-exec (id or name)
.TP
\-g GROUP, \-\-group=GROUP
use this GROUP to \-\-exec (id or name)
.TP
\-m, \-\-move
move specified tasks to specified cpuset; to move a PIDSPEC to a cpuset, use \-m PIDSPEC cpuset; to move all tasks specify \-\-fromset and \-\-toset
.TP
\-p PIDSPEC, \-\-pid=PIDSPEC
specify pid or tid specification
.TP
\-s CPUSET, \-\-set=CPUSET
specify name of immediate cpuset
.TP
\-t TOSET, \-\-toset=TOSET
specify name of destination cpuset
.TP
\-f FROMSET, \-\-fromset=FROMSET
specify name of origination cpuset
.TP
\-k, \-\-kthread
move, or include moving, unbound kernel threads
.TP
\-a, \-\-all
force all processes and threads to be moved
.TP
\-v, \-\-verbose
prints more detailed output, additive
.SH "DESCRIPTION"
This command is used to run and manage arbitrary processes on specified cpusets. It is also used to move pre\-existing processes and threads to specified cpusets. You may note there is no "kill" or "destroy" option \(em use the standard OS ^C or kill commands for that.
To list which tasks are running in a particular cpuset, use the \-\-list command.
For example:
\fB# cset proc \-\-list \-\-set priset\fR
This command will list all the tasks running in the cpuset called "priset".
Processes are created by specifying the path to the executable and specifying the cpuset that the process is to be created in.
For example:
\fB# cset proc \-\-set=blazing_cpuset \-\-exec /usr/bin/fast_code\fR
This command will execute the /usr/bin/fast_code program on the "blazing_cpuset" cpuset.
The PIDSPEC argument taken for the move command is a comma separated list of PIDs or TIDs. The list can also include brackets of PIDs or TIDs (i.e. tasks) that are inclusive of the endpoints.
For example:
.sp
.nf
1,2,5 Means processes 1, 2 and 5
1,2,600\-700 Means processes 1, 2 and from 600 to 700
.fi
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
The range of PIDs or TIDs does not need to have every position populated. In other words, for the example above, if there is only one process, say PID 57, in the range of 50\-65, then only that process will be moved.
To move a PIDSPEC to a specific cpuset, you can either specify the PIDSPEC with \-\-pid and the destination cpuset with \-\-toset, or use the short hand and list the cpuset name after the PIDSPEC for the \-\-move arguments.
The move command accepts multiple common calling methods. For example, the following commands are equivalent:
\fB# cset proc \-\-move 2442,3000\-3200 reserved_set\fR
\fB# cset proc \-\-move \-\-pid=2442,3000\-3200 \-\-toset=reserved_set\fR
These commands move the tasks defined as 2442 and any running task between 3000 and 3200 inclusive of the ends to the cpuset called "reserved_set".
Specifying the \-\-fromset is not necesary since the tasks will be moved to the destination cpuset no matter which cpuset they are currently running on.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
However, if you do specify a cpuset with the \-\-fromset option, then only those tasks that are both in the PIDSPEC \fBand\fR are running in the cpuset specified by \-\-fromset will be moved. I.e., if there is a task running on the system but not in \-\-fromset that is in PIDSPEC, it will not be moved.
To move all userspace tasks from one cpuset to another, you need to specify the source and destination cpuset by name.
For example:
\fB# cset proc \-\-move \-\-fromset=comp1 \-\-toset=comp42\fR
This command specifies that all processes and threads running on cpuset "comp1" be moved to cpuset "comp42".
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
This move command will not move kernel threads unless the \-k/\-\-kthread switch is specified. If it is, then all unbound kernel threads will be added to the move. Unbound kernel threads are those that can run on any CPU. If you \fBalso specify\fR the \-a/\-\-all switch, then all tasks, kernel or not, bound or not, will be moved.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBCaution\fR
Please be cautious with the \-\-all switch, since moving a kernel thread that is bound to a specific CPU to a cpuset that does not include that CPU can cause a system hang.
You must specify unique cpuset names for the both exec and move commands. If a simple name passed to the \-\-fromset, \-\-toset and \-\-set parameters is unique on the system then that command executes. However, if there are multiple cpusets by that name, then you will need to specify which one you mean with a full path rooted at the base cpuset tree.
For example, suppose you have the following cpuset tree:
.sp
.nf
/cpusets
/group1
/myset
/yourset
/group2
/myset
/yourset
.fi
Then, to move a process from myset in group1 to yourset in group2, you would have to issue the following command:
# cset proc \-\-move \-\-pid=50 \-\-fromset=/group1/myset \-\-toset=/group2/yourset
You do not have to worry about where in the Linux filesystem the cpuset filesystem is mounted. The cset command takes care of that. Any cpusets that are specifed by path (such as above), are done with respect to the root of the cpuset filesystem.
.SH "LICENSE"
Cpuset is licensed under the GNU GPL V2 only.
.SH "COPYRIGHT"
Copyright (c) 2008 Novell Inc.
.SH "AUTHOR"
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked GIT (stgit) v0.13 (see http://gna.org/projects/stgit and http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
.SH "SEE ALSO"
\fBcset\fR(1), \fBcset\-set\fR(1), \fBcset\-shield\fR(1), \fBtaskset\fR(1), \fBchrt\fR(1)
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,577 @@
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<meta name="generator" content="AsciiDoc 8.1.0" />
<style type="text/css">
/* Debug borders */
p, li, dt, dd, div, pre, h1, h2, h3, h4, h5, h6 {
/*
border: 1px solid red;
*/
}
body {
margin: 1em 5% 1em 5%;
}
a {
color: blue;
text-decoration: underline;
}
a:visited {
color: fuchsia;
}
em {
font-style: italic;
}
strong {
font-weight: bold;
}
tt {
color: navy;
}
h1, h2, h3, h4, h5, h6 {
color: #527bbd;
font-family: sans-serif;
margin-top: 1.2em;
margin-bottom: 0.5em;
line-height: 1.3;
}
h1 {
border-bottom: 2px solid silver;
}
h2 {
border-bottom: 2px solid silver;
padding-top: 0.5em;
}
div.sectionbody {
font-family: serif;
margin-left: 0;
}
hr {
border: 1px solid silver;
}
p {
margin-top: 0.5em;
margin-bottom: 0.5em;
}
pre {
padding: 0;
margin: 0;
}
span#author {
color: #527bbd;
font-family: sans-serif;
font-weight: bold;
font-size: 1.1em;
}
span#email {
}
span#revision {
font-family: sans-serif;
}
div#footer {
font-family: sans-serif;
font-size: small;
border-top: 2px solid silver;
padding-top: 0.5em;
margin-top: 4.0em;
}
div#footer-text {
float: left;
padding-bottom: 0.5em;
}
div#footer-badges {
float: right;
padding-bottom: 0.5em;
}
div#preamble,
div.tableblock, div.imageblock, div.exampleblock, div.verseblock,
div.quoteblock, div.literalblock, div.listingblock, div.sidebarblock,
div.admonitionblock {
margin-right: 10%;
margin-top: 1.5em;
margin-bottom: 1.5em;
}
div.admonitionblock {
margin-top: 2.5em;
margin-bottom: 2.5em;
}
div.content { /* Block element content. */
padding: 0;
}
/* Block element titles. */
div.title, caption.title {
font-family: sans-serif;
font-weight: bold;
text-align: left;
margin-top: 1.0em;
margin-bottom: 0.5em;
}
div.title + * {
margin-top: 0;
}
td div.title:first-child {
margin-top: 0.0em;
}
div.content div.title:first-child {
margin-top: 0.0em;
}
div.content + div.title {
margin-top: 0.0em;
}
div.sidebarblock > div.content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.listingblock {
margin-right: 0%;
}
div.listingblock > div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock > div.content {
padding-left: 2.0em;
}
div.attribution {
text-align: right;
}
div.verseblock + div.attribution {
text-align: left;
}
div.admonitionblock .icon {
vertical-align: top;
font-size: 1.1em;
font-weight: bold;
text-decoration: underline;
color: #527bbd;
padding-right: 0.5em;
}
div.admonitionblock td.content {
padding-left: 0.5em;
border-left: 2px solid silver;
}
div.exampleblock > div.content {
border-left: 2px solid silver;
padding: 0.5em;
}
div.verseblock div.content {
white-space: pre;
}
div.imageblock div.content { padding-left: 0; }
div.imageblock img { border: 1px solid silver; }
span.image img { border-style: none; }
dl {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
dt {
margin-top: 0.5em;
margin-bottom: 0;
font-style: italic;
}
dd > *:first-child {
margin-top: 0;
}
ul, ol {
list-style-position: outside;
}
ol.olist2 {
list-style-type: lower-alpha;
}
div.tableblock > table {
border: 3px solid #527bbd;
}
thead {
font-family: sans-serif;
font-weight: bold;
}
tfoot {
font-weight: bold;
}
div.hlist {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
td.hlist1 {
vertical-align: top;
font-style: italic;
padding-right: 0.8em;
}
td.hlist2 {
vertical-align: top;
}
@media print {
div#footer-badges { display: none; }
}
div#toctitle {
color: #527bbd;
font-family: sans-serif;
font-size: 1.1em;
font-weight: bold;
margin-top: 1.0em;
margin-bottom: 0.1em;
}
div.toclevel1, div.toclevel2 {
margin-top: 0;
margin-bottom: 0;
}
div.toclevel2 {
margin-left: 2em;
}
include1::./stylesheets/xhtml11-manpage.css[]
/* Workarounds for IE6's broken and incomplete CSS2. */
div.sidebar-content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.sidebar-title, div.image-title {
font-family: sans-serif;
font-weight: bold;
margin-top: 0.0em;
margin-bottom: 0.5em;
}
div.listingblock div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock-content {
padding-left: 2.0em;
}
div.exampleblock-content {
border-left: 2px solid silver;
padding-left: 0.5em;
}
/* IE6 sets dynamically generated links as visited. */
div#toc a:visited { color: blue; }
</style>
<title>cset-proc(1)</title>
</head>
<body>
<div id="header">
<h1>
cset-proc(1) Manual Page
</h1>
<h2>NAME</h2>
<div class="sectionbody">
<p>cset-proc -
manage processes running in cpusets
</p>
</div>
</div>
<h2>SYNOPSIS</h2>
<div class="sectionbody">
<div class="verseblock">
<div class="content"><em>cset</em> [cset options] <em>proc</em> [proc options] [args]
<em>cset</em> proc --help
<em>cset</em> proc
<em>cset</em> proc my_set my_other_set
<em>cset</em> proc --list --set my_set
<em>cset</em> proc --exec my_set /opt/software/my_code --my_opt_1
<em>cset</em> proc --set my_set --exec /opt/software/my_code --my_opt_1
<em>cset</em> proc --move 2442,3000-3200 my_set
<em>cset</em> proc --move --pid=2442,3000-3200 --toset=my_set
<em>cset</em> proc --move --fromset=my_set_1 --toset=my_set_2
<em>cset</em> proc --move --pid=42 --fromset=/group1/myset --toset=/group2/yourset</div></div>
</div>
<h2>OPTIONS</h2>
<div class="sectionbody">
<dl>
<dt>
-h, --help
</dt>
<dd>
<p>
prints the list of options for this command
</p>
</dd>
<dt>
-l, --list
</dt>
<dd>
<p>
list processes in the specified cpuset
</p>
</dd>
<dt>
-e, --exec
</dt>
<dd>
<p>
execute arguments in the specified cpuset
</p>
</dd>
<dt>
-u USER, --user=USER
</dt>
<dd>
<p>
use this USER to --exec (id or name)
</p>
</dd>
<dt>
-g GROUP, --group=GROUP
</dt>
<dd>
<p>
use this GROUP to --exec (id or name)
</p>
</dd>
<dt>
-m, --move
</dt>
<dd>
<p>
move specified tasks to specified cpuset; to move a PIDSPEC to a
cpuset, use -m PIDSPEC cpuset; to move all tasks specify --fromset
and --toset
</p>
</dd>
<dt>
-p PIDSPEC, --pid=PIDSPEC
</dt>
<dd>
<p>
specify pid or tid specification
</p>
</dd>
<dt>
-s CPUSET, --set=CPUSET
</dt>
<dd>
<p>
specify name of immediate cpuset
</p>
</dd>
<dt>
-t TOSET, --toset=TOSET
</dt>
<dd>
<p>
specify name of destination cpuset
</p>
</dd>
<dt>
-f FROMSET, --fromset=FROMSET
</dt>
<dd>
<p>
specify name of origination cpuset
</p>
</dd>
<dt>
-k, --kthread
</dt>
<dd>
<p>
move, or include moving, unbound kernel threads
</p>
</dd>
<dt>
-a, --all
</dt>
<dd>
<p>
force all processes and threads to be moved
</p>
</dd>
<dt>
-v, --verbose
</dt>
<dd>
<p>
prints more detailed output, additive
</p>
</dd>
</dl>
</div>
<h2>DESCRIPTION</h2>
<div class="sectionbody">
<p>This command is used to run and manage arbitrary processes on
specified cpusets. It is also used to move pre-existing processes and
threads to specified cpusets. You may note there is no "kill" or
"destroy" option &#8212; use the standard OS ^C or kill commands for that.</p>
<p>To list which tasks are running in a particular cpuset, use the --list
command.</p>
<p>For example:</p>
<p><strong><tt># cset proc --list --set priset</tt></strong></p>
<p>This command will list all the tasks running in the cpuset called
"priset".</p>
<p>Processes are created by specifying the path to the executable
and specifying the cpuset that the process is to be created in.</p>
<p>For example:</p>
<p><strong><tt># cset proc --set=blazing_cpuset --exec /usr/bin/fast_code</tt></strong></p>
<p>This command will execute the /usr/bin/fast_code program on the
"blazing_cpuset" cpuset.</p>
<p>The PIDSPEC argument taken for the move command is a comma
separated list of PIDs or TIDs. The list can also include
brackets of PIDs or TIDs (i.e. tasks) that are inclusive of the
endpoints.</p>
<p>For example:</p>
<div class="literalblock">
<div class="content">
<pre><tt>1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700</tt></pre>
</div></div>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">The range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.</td>
</tr></table>
</div>
<p>To move a PIDSPEC to a specific cpuset, you can either specify
the PIDSPEC with --pid and the destination cpuset with --toset,
or use the short hand and list the cpuset name after the PIDSPEC
for the --move arguments.</p>
<p>The move command accepts multiple common calling methods.
For example, the following commands are equivalent:</p>
<p><strong><tt># cset proc --move 2442,3000-3200 reserved_set</tt></strong></p>
<p><strong><tt># cset proc --move --pid=2442,3000-3200 --toset=reserved_set</tt></strong></p>
<p>These commands move the tasks defined as 2442 and any running task
between 3000 and 3200 inclusive of the ends to the cpuset called
"reserved_set".</p>
<p>Specifying the --fromset is not necesary since the tasks will be
moved to the destination cpuset no matter which cpuset they are
currently running on.</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">However, if you do specify a cpuset with the --fromset
option, then only those tasks that are both in the PIDSPEC <strong>and</strong>
are running in the cpuset specified by --fromset will be moved.
I.e., if there is a task running on the system but not in
--fromset that is in PIDSPEC, it will not be moved.</td>
</tr></table>
</div>
<p>To move all userspace tasks from one cpuset to another, you need
to specify the source and destination cpuset by name.</p>
<p>For example:</p>
<p><strong><tt># cset proc --move --fromset=comp1 --toset=comp42</tt></strong></p>
<p>This command specifies that all processes and threads
running on cpuset "comp1" be moved to cpuset "comp42".</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">This move command will not move kernel threads unless
the -k/--kthread switch is specified. If it is, then all unbound
kernel threads will be added to the move. Unbound kernel threads
are those that can run on any CPU. If you <strong>also specify</strong> the
-a/--all switch, then all tasks, kernel or not, bound or not,
will be moved.</td>
</tr></table>
</div>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Caution</div>
</td>
<td class="content">Please be cautious with the --all switch, since moving a
kernel thread that is bound to a specific CPU to a cpuset that
does not include that CPU can cause a system hang.</td>
</tr></table>
</div>
<p>You must specify unique cpuset names for the both exec and move
commands. If a simple name passed to the --fromset, --toset and
--set parameters is unique on the system then that command
executes. However, if there are multiple cpusets by that name,
then you will need to specify which one you mean with a full path
rooted at the base cpuset tree.</p>
<p>For example, suppose you have the following cpuset tree:</p>
<div class="literalblock">
<div class="content">
<pre><tt>/cpusets
/group1
/myset
/yourset
/group2
/myset
/yourset</tt></pre>
</div></div>
<p>Then, to move a process from myset in group1 to yourset in
group2, you would have to issue the following command:</p>
<p><tt># cset proc --move --pid=50 --fromset=/group1/myset --toset=/group2/yourset</tt></p>
<p>You do not have to worry about where in the Linux filesystem the
cpuset filesystem is mounted. The cset command takes care of that.
Any cpusets that are specifed by path (such as above), are done with
respect to the root of the cpuset filesystem.</p>
</div>
<h2>LICENSE</h2>
<div class="sectionbody">
<p>Cpuset is licensed under the GNU GPL V2 only.</p>
</div>
<h2>COPYRIGHT</h2>
<div class="sectionbody">
<p>Copyright (c) 2008 Novell Inc.</p>
</div>
<h2>AUTHOR</h2>
<div class="sectionbody">
<p>Written by Alex Tsariounov &lt;alext@novell.com&gt;.</p>
<p>Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.</p>
</div>
<h2>SEE ALSO</h2>
<div class="sectionbody">
<p><a href="cset.html">cset(1)</a>, <a href="cset-set.html">cset-set(1)</a>, <a href="cset-shield.html">cset-shield(1)</a>, <a href="taskset.html">taskset(1)</a>,
<a href="chrt.html">chrt(1)</a></p>
<p>/usr/src/linux/Documentation/cpusets.txt</p>
</div>
<div id="footer">
<div id="footer-text">
Version 1.0<br />
Last updated 30-May-2008 14:33:23 MDT
</div>
</div>
</body>
</html>

@ -0,0 +1,205 @@
cset-proc(1)
============
Alex Tsariounov <alext@novell.com>
v1.0, May 2008
NAME
----
cset-proc - manage processes running in cpusets
SYNOPSIS
--------
[verse]
'cset' [cset options] 'proc' [proc options] [args]
'cset' proc --help
'cset' proc
'cset' proc my_set my_other_set
'cset' proc --list --set my_set
'cset' proc --exec my_set /opt/software/my_code --my_opt_1
'cset' proc --set my_set --exec /opt/software/my_code --my_opt_1
'cset' proc --move 2442,3000-3200 my_set
'cset' proc --move --pid=2442,3000-3200 --toset=my_set
'cset' proc --move --fromset=my_set_1 --toset=my_set_2
'cset' proc --move --pid=42 --fromset=/group1/myset --toset=/group2/yourset
OPTIONS
-------
-h, --help::
prints the list of options for this command
-l, --list::
list processes in the specified cpuset
-e, --exec::
execute arguments in the specified cpuset
-u USER, --user=USER::
use this USER to --exec (id or name)
-g GROUP, --group=GROUP::
use this GROUP to --exec (id or name)
-m, --move::
move specified tasks to specified cpuset; to move a PIDSPEC to a
cpuset, use -m PIDSPEC cpuset; to move all tasks specify --fromset
and --toset
-p PIDSPEC, --pid=PIDSPEC::
specify pid or tid specification
-s CPUSET, --set=CPUSET::
specify name of immediate cpuset
-t TOSET, --toset=TOSET::
specify name of destination cpuset
-f FROMSET, --fromset=FROMSET::
specify name of origination cpuset
-k, --kthread::
move, or include moving, unbound kernel threads
-a, --all::
force all processes and threads to be moved
-v, --verbose::
prints more detailed output, additive
DESCRIPTION
-----------
This command is used to run and manage arbitrary processes on
specified cpusets. It is also used to move pre-existing processes and
threads to specified cpusets. You may note there is no "kill" or
"destroy" option -- use the standard OS ^C or kill commands for that.
To list which tasks are running in a particular cpuset, use the --list
command.
For example:
*+# cset proc --list --set priset+*
This command will list all the tasks running in the cpuset called
"priset".
Processes are created by specifying the path to the executable
and specifying the cpuset that the process is to be created in.
For example:
*+# cset proc --set=blazing_cpuset --exec /usr/bin/fast_code+*
This command will execute the /usr/bin/fast_code program on the
"blazing_cpuset" cpuset.
The PIDSPEC argument taken for the move command is a comma
separated list of PIDs or TIDs. The list can also include
brackets of PIDs or TIDs (i.e. tasks) that are inclusive of the
endpoints.
For example:
1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700
NOTE: The range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.
To move a PIDSPEC to a specific cpuset, you can either specify
the PIDSPEC with --pid and the destination cpuset with --toset,
or use the short hand and list the cpuset name after the PIDSPEC
for the --move arguments.
The move command accepts multiple common calling methods.
For example, the following commands are equivalent:
*+# cset proc --move 2442,3000-3200 reserved_set+*
*+# cset proc --move --pid=2442,3000-3200 --toset=reserved_set+*
These commands move the tasks defined as 2442 and any running task
between 3000 and 3200 inclusive of the ends to the cpuset called
"reserved_set".
Specifying the --fromset is not necesary since the tasks will be
moved to the destination cpuset no matter which cpuset they are
currently running on.
NOTE: However, if you do specify a cpuset with the --fromset
option, then only those tasks that are both in the PIDSPEC *and*
are running in the cpuset specified by --fromset will be moved.
I.e., if there is a task running on the system but not in
--fromset that is in PIDSPEC, it will not be moved.
To move all userspace tasks from one cpuset to another, you need
to specify the source and destination cpuset by name.
For example:
*+# cset proc --move --fromset=comp1 --toset=comp42+*
This command specifies that all processes and threads
running on cpuset "comp1" be moved to cpuset "comp42".
NOTE: This move command will not move kernel threads unless
the -k/--kthread switch is specified. If it is, then all unbound
kernel threads will be added to the move. Unbound kernel threads
are those that can run on any CPU. If you *also specify* the
-a/--all switch, then all tasks, kernel or not, bound or not,
will be moved.
CAUTION: Please be cautious with the --all switch, since moving a
kernel thread that is bound to a specific CPU to a cpuset that
does not include that CPU can cause a system hang.
You must specify unique cpuset names for the both exec and move
commands. If a simple name passed to the --fromset, --toset and
--set parameters is unique on the system then that command
executes. However, if there are multiple cpusets by that name,
then you will need to specify which one you mean with a full path
rooted at the base cpuset tree.
For example, suppose you have the following cpuset tree:
/cpusets
/group1
/myset
/yourset
/group2
/myset
/yourset
Then, to move a process from myset in group1 to yourset in
group2, you would have to issue the following command:
+# cset proc --move --pid=50 --fromset=/group1/myset --toset=/group2/yourset+
You do not have to worry about where in the Linux filesystem the
cpuset filesystem is mounted. The cset command takes care of that.
Any cpusets that are specifed by path (such as above), are done with
respect to the root of the cpuset filesystem.
LICENSE
-------
Cpuset is licensed under the GNU GPL V2 only.
COPYRIGHT
---------
Copyright (c) 2008 Novell Inc.
AUTHOR
------
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
SEE ALSO
--------
manlink:cset[1], csetlink:set[], csetlink:shield[], manlink:taskset[1],
manlink:chrt[1]
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,109 @@
.\" ** You probably do not want to edit this file directly **
.\" It was generated using the DocBook XSL Stylesheets (version 1.69.1).
.\" Instead of manually editing it, you probably should edit the DocBook XML
.\" source for it and then use the DocBook XSL Stylesheets to regenerate it.
.TH "CSET\-SET" "1" "05/30/2008" "" ""
.\" disable hyphenation
.nh
.\" disable justification (adjust text to left margin only)
.ad l
.SH "NAME"
cset\-set \- manage sets of cpus
.SH "SYNOPSIS"
.sp
.nf
\fIcset\fR [cset options] \fIset\fR [set options] [args]
\fIcset\fR set \-\-help
\fIcset\fR set
\fIcset\fR set \-\-all
\fIcset\fR set \-\-recurse
\fIcset\fR set \-\-list myset
\fIcset\fR set myset
\fIcset\fR set \-\-all myset
\fIcset\fR set \-\-recurse \-\-list myset
\fIcset\fR set \-\-cpu 2\-5 \-\-mem 0 \-\-set newset
\fIcset\fR set \-\-cpu 2\-5 newset
\fIcset\fR set \-\-cpu 1,2,5\-7 another_set
\fIcset\fR set \-\-destroy newset
\fIcset\fR set \-\-destroy /mygroup_sets/my_set
.fi
.SH "OPTIONS"
.TP
\-h, \-\-help
prints the list of options for this command
.TP
\-l, \-\-list
list the named cpuset(s); if \-a is used, will list members of named cpuset; if \-r is used, will list recursively
.TP
\-c CPUSPEC, \-\-cpu=CPUSPEC
create or modify cpuset in the specified cpuset with CPUSPEC specification
.TP
\-m MEMSPEC, \-\-mem=MEMSPEC
specify which memory nodes to assign to the created or modified cpuset
.TP
\-d, \-\-destroy
destroy specified cpuset
.TP
\-s CPUSET, \-\-set=CPUSET
specify cpuset name to be acted on
.TP
\-a, \-\-all
also do listing of members of specified cpuset, for use with \-\-list
.TP
\-r, \-\-recurse
do recursive listing, for use with \-\-list
.TP
\-v, \-\-verbose
prints more detailed output, for the set command, using this flag will not chop listing to fit in 80 columns
.TP
\-\-cpu_exclusive
mark this cpuset as owning its CPUs exclusively
.TP
\-\-mem_exclusive
mark this cpuset as owning its MEMs exclusively
.SH "DESCRIPTION"
This command is used to create, modify, and destroy cpusets. Cpusets form a tree\-like structure rooted at the root cpuset which always includes all system CPUs and all system memory nodes.
A cpuset is an organizational unit that defines a group of CPUs and a group of memory nodes where a process or thread (i.e. task) is allowed to run on. For non\-NUMA machines, the memory node is always 0 (zero) and cannot be set to anything else. For NUMA machines, the memory node can be set to a similar specifcation as the CPU definition and will tie those memory nodes to that cpuset. You will usually want the memory nodes that belong to the CPUs defined to be in the same cpuset.
A cpuset can have exclusive right to the CPUs defined in it. This means that only this cpuset can own these CPUs. Similarly, a cpuset can have exclusive right to the memory nodes defined in it. This means that only this cpuset can own these memory nodes.
Cpusets can be specified by name or by path; however, care should be taken when specifying by name if the name is not unique. This tool will generally not let you do destructive things to non\-unique cpuset names.
Cpusets are uniquely specified by path. The path starts at where the cpusets filesystem is mounted so you generally do not have to know where that is. For example, so specify a cpuset that is called "two" which is a subset of "one" which in turn is a subset of the root cpuset, use the path "/one/two" regardless of where the cpusets filesystem is mounted.
When specifying CPUs, a so\-called CPUSPEC is used. The CPUSPEC will accept a comma\-separated list of CPUs and inclusive range specifications. For example, \-\-cpu=1,3,5\-7 will assign CPU1, CPU3, CPU5, CPU6, and CPU7 to the specified cpuset.
Note that cpusets follow certain rules. For example, children can only include CPUs that the parents already have. If you do not follow those rules, the kernel cpuset subsystem will not let you create that cpuset. For example, if you create a cpuset that contains CPU3, and then attempt to create a child of that cpuset with a CPU other than 3, you will get an error, and the cpuset will not be active. The error is somewhat cryptic in that it is usually a "Permission denied" error.
Memory nodes are specified with a MEMSPEC in a similar way to the CPUSPEC. For example, \-\-mem=1,3\-6 will assign MEM1, MEM3, MEM4, MEM5, and MEM6 to the specified cpuset.
Note that if you attempt to create or modify a cpuset with a memory node specification that is not valid, you may get a cryptic error message, "No space left on device", and the modification will not be allowed.
When you destroy a cpuset, then the tasks running in that set are moved to the parent of that cpuset. If this is not what you want, then manually move those tasks to the cpuset of your choice with the \fIcset proc\fR command (see \fIcset proc \-\-help\fR for more information).
.SS "EXAMPLES"
.TP
Create a cpuset with the default memory specification
\fB# cset set \-\-cpu=2,4,6\-8 \-\-set=new_set\fR
This command creates a cpuset called "new_set" located off the root cpuset which holds CPUS 2,4,6,7,8 and node 0 (interleaved) memory. Note that \-\-set is optional, and you can just specify the name for the new cpuset after all arguments.
.TP
Create a cpuset that specifies both CPUs and memory nodes
\fB# cset set \-\-cpu=3 \-\-mem=3 /rad/set_one\fR
Note that this command uses the full path method to specify the name of the new cpuset "/rad/set_one". It also names the new cpuset implicitily (i.e. no \-\-set option, although you can use that if you want to). If the "set_one" name is unique, you can subsequently refer to is just by that. Memory node 3 is assigned to this cpuset as well as CPU 3.
The above commands will create the new cpusets, or if they already exist, they will modify them to the new specifications.
.SH "LICENSE"
Cpuset is licensed under the GNU GPL V2 only.
.SH "COPYRIGHT"
Copyright (c) 2008 Novell Inc.
.SH "AUTHOR"
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked GIT (stgit) v0.13 (see http://gna.org/projects/stgit and http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
.SH "SEE ALSO"
\fBcset\fR(1), \fBcset\-proc\fR(1), \fBcset\-shield\fR(1), \fBtaskset\fR(1), \fBchrt\fR(1)
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,536 @@
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<meta name="generator" content="AsciiDoc 8.1.0" />
<style type="text/css">
/* Debug borders */
p, li, dt, dd, div, pre, h1, h2, h3, h4, h5, h6 {
/*
border: 1px solid red;
*/
}
body {
margin: 1em 5% 1em 5%;
}
a {
color: blue;
text-decoration: underline;
}
a:visited {
color: fuchsia;
}
em {
font-style: italic;
}
strong {
font-weight: bold;
}
tt {
color: navy;
}
h1, h2, h3, h4, h5, h6 {
color: #527bbd;
font-family: sans-serif;
margin-top: 1.2em;
margin-bottom: 0.5em;
line-height: 1.3;
}
h1 {
border-bottom: 2px solid silver;
}
h2 {
border-bottom: 2px solid silver;
padding-top: 0.5em;
}
div.sectionbody {
font-family: serif;
margin-left: 0;
}
hr {
border: 1px solid silver;
}
p {
margin-top: 0.5em;
margin-bottom: 0.5em;
}
pre {
padding: 0;
margin: 0;
}
span#author {
color: #527bbd;
font-family: sans-serif;
font-weight: bold;
font-size: 1.1em;
}
span#email {
}
span#revision {
font-family: sans-serif;
}
div#footer {
font-family: sans-serif;
font-size: small;
border-top: 2px solid silver;
padding-top: 0.5em;
margin-top: 4.0em;
}
div#footer-text {
float: left;
padding-bottom: 0.5em;
}
div#footer-badges {
float: right;
padding-bottom: 0.5em;
}
div#preamble,
div.tableblock, div.imageblock, div.exampleblock, div.verseblock,
div.quoteblock, div.literalblock, div.listingblock, div.sidebarblock,
div.admonitionblock {
margin-right: 10%;
margin-top: 1.5em;
margin-bottom: 1.5em;
}
div.admonitionblock {
margin-top: 2.5em;
margin-bottom: 2.5em;
}
div.content { /* Block element content. */
padding: 0;
}
/* Block element titles. */
div.title, caption.title {
font-family: sans-serif;
font-weight: bold;
text-align: left;
margin-top: 1.0em;
margin-bottom: 0.5em;
}
div.title + * {
margin-top: 0;
}
td div.title:first-child {
margin-top: 0.0em;
}
div.content div.title:first-child {
margin-top: 0.0em;
}
div.content + div.title {
margin-top: 0.0em;
}
div.sidebarblock > div.content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.listingblock {
margin-right: 0%;
}
div.listingblock > div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock > div.content {
padding-left: 2.0em;
}
div.attribution {
text-align: right;
}
div.verseblock + div.attribution {
text-align: left;
}
div.admonitionblock .icon {
vertical-align: top;
font-size: 1.1em;
font-weight: bold;
text-decoration: underline;
color: #527bbd;
padding-right: 0.5em;
}
div.admonitionblock td.content {
padding-left: 0.5em;
border-left: 2px solid silver;
}
div.exampleblock > div.content {
border-left: 2px solid silver;
padding: 0.5em;
}
div.verseblock div.content {
white-space: pre;
}
div.imageblock div.content { padding-left: 0; }
div.imageblock img { border: 1px solid silver; }
span.image img { border-style: none; }
dl {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
dt {
margin-top: 0.5em;
margin-bottom: 0;
font-style: italic;
}
dd > *:first-child {
margin-top: 0;
}
ul, ol {
list-style-position: outside;
}
ol.olist2 {
list-style-type: lower-alpha;
}
div.tableblock > table {
border: 3px solid #527bbd;
}
thead {
font-family: sans-serif;
font-weight: bold;
}
tfoot {
font-weight: bold;
}
div.hlist {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
td.hlist1 {
vertical-align: top;
font-style: italic;
padding-right: 0.8em;
}
td.hlist2 {
vertical-align: top;
}
@media print {
div#footer-badges { display: none; }
}
div#toctitle {
color: #527bbd;
font-family: sans-serif;
font-size: 1.1em;
font-weight: bold;
margin-top: 1.0em;
margin-bottom: 0.1em;
}
div.toclevel1, div.toclevel2 {
margin-top: 0;
margin-bottom: 0;
}
div.toclevel2 {
margin-left: 2em;
}
include1::./stylesheets/xhtml11-manpage.css[]
/* Workarounds for IE6's broken and incomplete CSS2. */
div.sidebar-content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.sidebar-title, div.image-title {
font-family: sans-serif;
font-weight: bold;
margin-top: 0.0em;
margin-bottom: 0.5em;
}
div.listingblock div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock-content {
padding-left: 2.0em;
}
div.exampleblock-content {
border-left: 2px solid silver;
padding-left: 0.5em;
}
/* IE6 sets dynamically generated links as visited. */
div#toc a:visited { color: blue; }
</style>
<title>cset-set(1)</title>
</head>
<body>
<div id="header">
<h1>
cset-set(1) Manual Page
</h1>
<h2>NAME</h2>
<div class="sectionbody">
<p>cset-set -
manage sets of cpus
</p>
</div>
</div>
<h2>SYNOPSIS</h2>
<div class="sectionbody">
<div class="verseblock">
<div class="content"><em>cset</em> [cset options] <em>set</em> [set options] [args]
<em>cset</em> set --help
<em>cset</em> set
<em>cset</em> set --all
<em>cset</em> set --recurse
<em>cset</em> set --list myset
<em>cset</em> set myset
<em>cset</em> set --all myset
<em>cset</em> set --recurse --list myset
<em>cset</em> set --cpu 2-5 --mem 0 --set newset
<em>cset</em> set --cpu 2-5 newset
<em>cset</em> set --cpu 1,2,5-7 another_set
<em>cset</em> set --destroy newset
<em>cset</em> set --destroy /mygroup_sets/my_set</div></div>
</div>
<h2>OPTIONS</h2>
<div class="sectionbody">
<dl>
<dt>
-h, --help
</dt>
<dd>
<p>
prints the list of options for this command
</p>
</dd>
<dt>
-l, --list
</dt>
<dd>
<p>
list the named cpuset(s); if -a is used, will list members of named
cpuset; if -r is used, will list recursively
</p>
</dd>
<dt>
-c CPUSPEC, --cpu=CPUSPEC
</dt>
<dd>
<p>
create or modify cpuset in the specified cpuset with CPUSPEC
specification
</p>
</dd>
<dt>
-m MEMSPEC, --mem=MEMSPEC
</dt>
<dd>
<p>
specify which memory nodes to assign to the created or modified
cpuset
</p>
</dd>
<dt>
-d, --destroy
</dt>
<dd>
<p>
destroy specified cpuset
</p>
</dd>
<dt>
-s CPUSET, --set=CPUSET
</dt>
<dd>
<p>
specify cpuset name to be acted on
</p>
</dd>
<dt>
-a, --all
</dt>
<dd>
<p>
also do listing of members of specified cpuset, for use with --list
</p>
</dd>
<dt>
-r, --recurse
</dt>
<dd>
<p>
do recursive listing, for use with --list
</p>
</dd>
<dt>
-v, --verbose
</dt>
<dd>
<p>
prints more detailed output, for the set command, using this flag
will not chop listing to fit in 80 columns
</p>
</dd>
<dt>
--cpu_exclusive
</dt>
<dd>
<p>
mark this cpuset as owning its CPUs exclusively
</p>
</dd>
<dt>
--mem_exclusive
</dt>
<dd>
<p>
mark this cpuset as owning its MEMs exclusively
</p>
</dd>
</dl>
</div>
<h2>DESCRIPTION</h2>
<div class="sectionbody">
<p>This command is used to create, modify, and destroy cpusets.
Cpusets form a tree-like structure rooted at the root cpuset
which always includes all system CPUs and all system memory
nodes.</p>
<p>A cpuset is an organizational unit that defines a group of CPUs
and a group of memory nodes where a process or thread (i.e. task)
is allowed to run on. For non-NUMA machines, the memory node is
always 0 (zero) and cannot be set to anything else. For NUMA
machines, the memory node can be set to a similar specifcation
as the CPU definition and will tie those memory nodes to that
cpuset. You will usually want the memory nodes that belong to
the CPUs defined to be in the same cpuset.</p>
<p>A cpuset can have exclusive right to the CPUs defined in it.
This means that only this cpuset can own these CPUs. Similarly,
a cpuset can have exclusive right to the memory nodes defined in
it. This means that only this cpuset can own these memory
nodes.</p>
<p>Cpusets can be specified by name or by path; however, care
should be taken when specifying by name if the name is not
unique. This tool will generally not let you do destructive
things to non-unique cpuset names.</p>
<p>Cpusets are uniquely specified by path. The path starts at where
the cpusets filesystem is mounted so you generally do not have to
know where that is. For example, so specify a cpuset that is
called "two" which is a subset of "one" which in turn is a subset
of the root cpuset, use the path "/one/two" regardless of where
the cpusets filesystem is mounted.</p>
<p>When specifying CPUs, a so-called CPUSPEC is used. The CPUSPEC
will accept a comma-separated list of CPUs and inclusive range
specifications. For example, --cpu=1,3,5-7 will assign CPU1,
CPU3, CPU5, CPU6, and CPU7 to the specified cpuset.</p>
<p>Note that cpusets follow certain rules. For example, children
can only include CPUs that the parents already have. If you do
not follow those rules, the kernel cpuset subsystem will not let
you create that cpuset. For example, if you create a cpuset that
contains CPU3, and then attempt to create a child of that cpuset
with a CPU other than 3, you will get an error, and the cpuset
will not be active. The error is somewhat cryptic in that it is
usually a "Permission denied" error.</p>
<p>Memory nodes are specified with a MEMSPEC in a similar way to
the CPUSPEC. For example, --mem=1,3-6 will assign MEM1, MEM3,
MEM4, MEM5, and MEM6 to the specified cpuset.</p>
<p>Note that if you attempt to create or modify a cpuset with a
memory node specification that is not valid, you may get a
cryptic error message, "No space left on device", and the
modification will not be allowed.</p>
<p>When you destroy a cpuset, then the tasks running in that set are
moved to the parent of that cpuset. If this is not what you
want, then manually move those tasks to the cpuset of your choice
with the <em>cset proc</em> command (see <em>cset proc --help</em> for more
information).</p>
<h3>EXAMPLES</h3>
<dl>
<dt>
Create a cpuset with the default memory specification
</dt>
<dd>
<p>
<strong><tt># cset set --cpu=2,4,6-8 --set=new_set</tt></strong>
</p>
</dd>
</dl>
<p>This command creates a cpuset called "new_set" located
off the root cpuset which holds CPUS 2,4,6,7,8 and node 0
(interleaved) memory. Note that --set is optional, and
you can just specify the name for the new cpuset after
all arguments.</p>
<dl>
<dt>
Create a cpuset that specifies both CPUs and memory nodes
</dt>
<dd>
<p>
<strong><tt># cset set --cpu=3 --mem=3 /rad/set_one</tt></strong>
</p>
</dd>
</dl>
<p>Note that this command uses the full path method to
specify the name of the new cpuset "/rad/set_one". It
also names the new cpuset implicitily (i.e. no --set
option, although you can use that if you want to). If
the "set_one" name is unique, you can subsequently refer
to is just by that. Memory node 3 is assigned to this
cpuset as well as CPU 3.</p>
<p>The above commands will create the new cpusets, or if they
already exist, they will modify them to the new specifications.</p>
</div>
<h2>LICENSE</h2>
<div class="sectionbody">
<p>Cpuset is licensed under the GNU GPL V2 only.</p>
</div>
<h2>COPYRIGHT</h2>
<div class="sectionbody">
<p>Copyright (c) 2008 Novell Inc.</p>
</div>
<h2>AUTHOR</h2>
<div class="sectionbody">
<p>Written by Alex Tsariounov &lt;alext@novell.com&gt;.</p>
<p>Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.</p>
</div>
<h2>SEE ALSO</h2>
<div class="sectionbody">
<p><a href="cset.html">cset(1)</a>, <a href="cset-proc.html">cset-proc(1)</a>, <a href="cset-shield.html">cset-shield(1)</a>, <a href="taskset.html">taskset(1)</a>,
<a href="chrt.html">chrt(1)</a></p>
<p>/usr/src/linux/Documentation/cpusets.txt</p>
</div>
<div id="footer">
<div id="footer-text">
Version 1.0<br />
Last updated 30-May-2008 14:33:23 MDT
</div>
</div>
</body>
</html>

@ -0,0 +1,179 @@
cset-set(1)
===========
Alex Tsariounov <alext@novell.com>
v1.0, May 2008
NAME
----
cset-set - manage sets of cpus
SYNOPSIS
--------
[verse]
'cset' [cset options] 'set' [set options] [args]
'cset' set --help
'cset' set
'cset' set --all
'cset' set --recurse
'cset' set --list myset
'cset' set myset
'cset' set --all myset
'cset' set --recurse --list myset
'cset' set --cpu 2-5 --mem 0 --set newset
'cset' set --cpu 2-5 newset
'cset' set --cpu 1,2,5-7 another_set
'cset' set --destroy newset
'cset' set --destroy /mygroup_sets/my_set
OPTIONS
-------
-h, --help::
prints the list of options for this command
-l, --list::
list the named cpuset(s); if -a is used, will list members of named
cpuset; if -r is used, will list recursively
-c CPUSPEC, --cpu=CPUSPEC::
create or modify cpuset in the specified cpuset with CPUSPEC
specification
-m MEMSPEC, --mem=MEMSPEC::
specify which memory nodes to assign to the created or modified
cpuset
-d, --destroy::
destroy specified cpuset
-s CPUSET, --set=CPUSET::
specify cpuset name to be acted on
-a, --all::
also do listing of members of specified cpuset, for use with --list
-r, --recurse::
do recursive listing, for use with --list
-v, --verbose::
prints more detailed output, for the set command, using this flag
will not chop listing to fit in 80 columns
--cpu_exclusive::
mark this cpuset as owning its CPUs exclusively
--mem_exclusive::
mark this cpuset as owning its MEMs exclusively
DESCRIPTION
-----------
This command is used to create, modify, and destroy cpusets.
Cpusets form a tree-like structure rooted at the root cpuset
which always includes all system CPUs and all system memory
nodes.
A cpuset is an organizational unit that defines a group of CPUs
and a group of memory nodes where a process or thread (i.e. task)
is allowed to run on. For non-NUMA machines, the memory node is
always 0 (zero) and cannot be set to anything else. For NUMA
machines, the memory node can be set to a similar specifcation
as the CPU definition and will tie those memory nodes to that
cpuset. You will usually want the memory nodes that belong to
the CPUs defined to be in the same cpuset.
A cpuset can have exclusive right to the CPUs defined in it.
This means that only this cpuset can own these CPUs. Similarly,
a cpuset can have exclusive right to the memory nodes defined in
it. This means that only this cpuset can own these memory
nodes.
Cpusets can be specified by name or by path; however, care
should be taken when specifying by name if the name is not
unique. This tool will generally not let you do destructive
things to non-unique cpuset names.
Cpusets are uniquely specified by path. The path starts at where
the cpusets filesystem is mounted so you generally do not have to
know where that is. For example, so specify a cpuset that is
called "two" which is a subset of "one" which in turn is a subset
of the root cpuset, use the path "/one/two" regardless of where
the cpusets filesystem is mounted.
When specifying CPUs, a so-called CPUSPEC is used. The CPUSPEC
will accept a comma-separated list of CPUs and inclusive range
specifications. For example, --cpu=1,3,5-7 will assign CPU1,
CPU3, CPU5, CPU6, and CPU7 to the specified cpuset.
Note that cpusets follow certain rules. For example, children
can only include CPUs that the parents already have. If you do
not follow those rules, the kernel cpuset subsystem will not let
you create that cpuset. For example, if you create a cpuset that
contains CPU3, and then attempt to create a child of that cpuset
with a CPU other than 3, you will get an error, and the cpuset
will not be active. The error is somewhat cryptic in that it is
usually a "Permission denied" error.
Memory nodes are specified with a MEMSPEC in a similar way to
the CPUSPEC. For example, --mem=1,3-6 will assign MEM1, MEM3,
MEM4, MEM5, and MEM6 to the specified cpuset.
Note that if you attempt to create or modify a cpuset with a
memory node specification that is not valid, you may get a
cryptic error message, "No space left on device", and the
modification will not be allowed.
When you destroy a cpuset, then the tasks running in that set are
moved to the parent of that cpuset. If this is not what you
want, then manually move those tasks to the cpuset of your choice
with the 'cset proc' command (see 'cset proc --help' for more
information).
EXAMPLES
~~~~~~~~
Create a cpuset with the default memory specification::
*+# cset set --cpu=2,4,6-8 --set=new_set+*
This command creates a cpuset called "new_set" located
off the root cpuset which holds CPUS 2,4,6,7,8 and node 0
(interleaved) memory. Note that --set is optional, and
you can just specify the name for the new cpuset after
all arguments.
Create a cpuset that specifies both CPUs and memory nodes::
*+# cset set --cpu=3 --mem=3 /rad/set_one+*
Note that this command uses the full path method to
specify the name of the new cpuset "/rad/set_one". It
also names the new cpuset implicitily (i.e. no --set
option, although you can use that if you want to). If
the "set_one" name is unique, you can subsequently refer
to is just by that. Memory node 3 is assigned to this
cpuset as well as CPU 3.
The above commands will create the new cpusets, or if they
already exist, they will modify them to the new specifications.
LICENSE
-------
Cpuset is licensed under the GNU GPL V2 only.
COPYRIGHT
---------
Copyright (c) 2008 Novell Inc.
AUTHOR
------
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
SEE ALSO
--------
manlink:cset[1], csetlink:proc[], csetlink:shield[], manlink:taskset[1],
manlink:chrt[1]
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,173 @@
.\" ** You probably do not want to edit this file directly **
.\" It was generated using the DocBook XSL Stylesheets (version 1.69.1).
.\" Instead of manually editing it, you probably should edit the DocBook XML
.\" source for it and then use the DocBook XSL Stylesheets to regenerate it.
.TH "CSET\-SHIELD" "1" "05/30/2008" "" ""
.\" disable hyphenation
.nh
.\" disable justification (adjust text to left margin only)
.ad l
.SH "NAME"
cset\-shield \- cpuset supercommand which implements cpu shielding
.SH "SYNOPSIS"
.sp
.nf
\fIcset\fR [cset options] \fIshield\fR [shield options] [args]
\fIcset\fR shield \-\-help
\fIcset\fR shield
\fIcset\fR shield \-\-cpu 1\-7
\fIcset\fR shield \-\-cpu 1\-7 \-\-kthread=on
\fIcset\fR shield \-\-exec /opt/software/myapp/doit \-\-my_opt1 \-\-my_opt2
\fIcset\fR shield \-\-user appuser \-\-exec run_benchmark.sh
\fIcset\fR shield \-\-shield 1024,2048,5000\-1000
\fIcset\fR shield \-\-unshield 6000\-8500
\fIcset\fR shield \-\-kthread=off
\fIcset\fR shield \-\-kthread=on
\fIcset\fR shield \-\-shield bash
.fi
.SH "OPTIONS"
.TP
\-h, \-\-help
prints the list of options for this command
.TP
\-c CPUSPEC, \-\-cpu=CPUSPEC
modifies or initializes the shield cpusets
.TP
\-r, \-\-reset
destroys the shield
.TP
\-e, \-\-exec
executes args in the shield
.TP
\-\-user=USER
use this USER for \-\-exec (id or name)
.TP
\-\-group=GROUP
use this GROUP for \-\-exec (id or name)
.TP
\-s PIDSPEC, \-\-shield=PIDSPEC
shield specified PIDSPEC of processes or threads
.TP
\-u PIDSPEC, \-\-unshield=PIDSPEC
remove specified PIDSPEC of processes or threads from the shield, the task keep running in the unshielded cpuset
.TP
\-k on|off, \-\-kthread=on|off
shield from unbound interrupt threads as well
.TP
\-f, \-\-force
force operation, use with care
.TP
\-v, \-\-verbose
prints more detailed output, additive
.TP
\-\-sysset=SYSSET
optionally specify system cpuset name
.TP
\-\-userset=USERSET
optionally specify user cpuset name
.SH "DESCRIPTION"
This is a supercommand that creates basic cpu shielding. The normal cset commands can of course be used to create this basic shield, but the shield command combines many such commands to create and manage a common type of cpu shielding setup.
The concept of shielding implies at minimum three cpusets, for example: root, user and system. The root cpuset always exists in all implementations of cpusets and contains all available CPUs on the machine. The system cpuset is so named because normal system tasks are made to run on it. The user cpuset is so named because that is the "shielded" cpuset on which you would run your tasks of interest.
Usually, CPU zero would be in the system set and the rest of the CPUs would be in the user set. After creation of the cpusets, all processes running in the root cpuset are moved to the system cpuset. Thus any new processes or threads spawned from these processes will also run the system cpuset.
If the optional \-\-kthread=on option is given to the shield command, then all kernel threads (with exception of the per\-CPU bound interrupt kernel threads) are also moved to the system set.
One executes processes on the shielded user cpuset with the \-\-exec subcommand or moves processes or threads to the shielded cpuset with the \-\-shield subcommand.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
You do not need to specify which cpuset a process or thread is running in initially when using the \-\-shield subcommand.
To create a shield, you would execute the shield command with the \-\-cpu option that specifies CPUSPEC argument that assigns CPUs to be under the shield (this means assigned to the user cpuset, all other cpus will be assigned to the system set).
For example:
\fB# cset shield \-\-cpu=3\fR
On a 4\-way machine, this command will dedicate the first 3 processors, CPU0\-CPU2, for the system set (unshielded) and only the last processor, CPU3, for the user set (shielded).
The CPUSPEC will accept a comma separated list of CPUs and inclusive range specifications. For example, \-\-cpu=1,3,5\-7 will assign CPU1, CPU3, CPU5, CPU6, and CPU7 to the user (or shielded) cpuset.
If you do not like the names "system" and "user" for the unshielded and shielded sets respectively, or if those names are used already, then use the \-\-sysset and \-\-userset options.
For example:
\fB# cset shield \-\-sysset=free \-\-userset=cage \-\-cpu=2,3 \-\-kthread=on\fR
The above command will use the name "free" for the unshielded system cpuset, the name "cage" for the shielded user cpuset, initialize these cpusets and dedicate CPU0 and CPU1 to the "free" set and (on a 4\-way machine) dedicate CPU2 and CPU3 to the "cage" set. Further, the command moves all processes and threads, including kernel threads from the root cpuset to the "free" cpuset.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
If you do use the \-\-syset/\-\-userset options, then you must continue to use those for every invocation of the shield supercommand.
After initialization, you can run the process of interest on the shielded cpuset with the \-\-exec subcommand, or move processes or threads already running to the shielded cpuset with the \-\-shield subcommand.
The PIDSPEC argument taken for the \-\-shield (or \-s) subcommand is a comma separated list of PIDs or TIDs. The list can also include brackets of PIDs or TIDs that are inclusive of the endpoints.
For example:
.sp
.nf
1,2,5 Means processes 1, 2 and 5
1,2,600\-700 Means processes 1, 2 and from 600 to 700
.fi
\fB# cset shield \-\-shield=50\-65\fR
The above command moves all processes and threads with PID or TID in the range 50\-65 inclusive, from any cpuset they may be running in into the shielded user cpuset.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
The range of PIDs or TIDs does not need to have every position populated. In other words, for the example above, if there is only one process, say PID 57, in the range of 50\-65, then only that process will be moved.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBCaution\fR
Please note that there is no checking of processes you request to move into the shield with the \-\-shield command. This means that the tool will happily move, for example, kernel threads that are bound to specific CPUs with this command. You can hang your system by indiscriminately specifying arbitrary PIDs to the \-\-shield command so please be careful.
The \-\-unshield (or \-u) subcommand will remove the specified processes or threads from the shielded cpuset and move them into the unshielded (or system) cpuset. This option also takes a PIDSPEC argument, the same as for the \-\-shield subcommand.
Both the \-\-shield and the \-\-unshield commands will also finally output the number of tasks running in the shield and out of the shield. If you do not specify a PIDSPEC to these commands, then just that status is output. By specifying also a \-\-verbose in addition, then you will get a listing of every task that is running in either the shield or out of the shield.
Using no subcommand, ie. only "cset shield", will output the status of both shield and non\-shield. Tasks will be listed if \-\-verbose is used.
You can adjust which CPUs are in the shielded cpuset by issuing the \-\-cpu subcommand again anytime after the shield has been initialized.
For example if the original shield contained CPU0 and CPU1 in the system set and CPU2 and CPU3 in the user set, if you then issue the following command:
\fB# cset shield \-\-cpu=1,2,3\fR
then that command will move CPU1 into the shielded "user" cpuset. Any processes or threads that were running on CPU1 that belonged to the unshielded "system" cpuset are migrated to CPU0 by the system.
The \-\-reset subcommand will in essence destroy the shield. For example, if there was a shield on a 4\-way machine with CPU0 in system and CPUs 1\-3 in user with processes running on the user cpuset (i.e. in the shield), and a \-\-reset subcommand was issued, then all processes running in both system and user cpusets would be migrated to the root cpuset (which has access to all CPUs and never goes away), after which both system and user cpusets would be destroyed.
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
Even though you can mix general usage of cpusets with the shielding concepts described here, you generally will not want to. For more complex shielding or usage scenarios, one would generally use the normal cpuset commands (i.e. cset set and proc) directly.
.SH "LICENSE"
Cpuset is licensed under the GNU GPL V2 only.
.SH "COPYRIGHT"
Copyright (c) 2008 Novell Inc.
.SH "AUTHOR"
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked GIT (stgit) v0.13 (see http://gna.org/projects/stgit and http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
.SH "SEE ALSO"
\fBcset\fR(1), \fBcset\-set\fR(1), \fBcset\-proc\fR(1), \fBtaskset\fR(1), \fBchrt\fR(1)
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,613 @@
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<meta name="generator" content="AsciiDoc 8.1.0" />
<style type="text/css">
/* Debug borders */
p, li, dt, dd, div, pre, h1, h2, h3, h4, h5, h6 {
/*
border: 1px solid red;
*/
}
body {
margin: 1em 5% 1em 5%;
}
a {
color: blue;
text-decoration: underline;
}
a:visited {
color: fuchsia;
}
em {
font-style: italic;
}
strong {
font-weight: bold;
}
tt {
color: navy;
}
h1, h2, h3, h4, h5, h6 {
color: #527bbd;
font-family: sans-serif;
margin-top: 1.2em;
margin-bottom: 0.5em;
line-height: 1.3;
}
h1 {
border-bottom: 2px solid silver;
}
h2 {
border-bottom: 2px solid silver;
padding-top: 0.5em;
}
div.sectionbody {
font-family: serif;
margin-left: 0;
}
hr {
border: 1px solid silver;
}
p {
margin-top: 0.5em;
margin-bottom: 0.5em;
}
pre {
padding: 0;
margin: 0;
}
span#author {
color: #527bbd;
font-family: sans-serif;
font-weight: bold;
font-size: 1.1em;
}
span#email {
}
span#revision {
font-family: sans-serif;
}
div#footer {
font-family: sans-serif;
font-size: small;
border-top: 2px solid silver;
padding-top: 0.5em;
margin-top: 4.0em;
}
div#footer-text {
float: left;
padding-bottom: 0.5em;
}
div#footer-badges {
float: right;
padding-bottom: 0.5em;
}
div#preamble,
div.tableblock, div.imageblock, div.exampleblock, div.verseblock,
div.quoteblock, div.literalblock, div.listingblock, div.sidebarblock,
div.admonitionblock {
margin-right: 10%;
margin-top: 1.5em;
margin-bottom: 1.5em;
}
div.admonitionblock {
margin-top: 2.5em;
margin-bottom: 2.5em;
}
div.content { /* Block element content. */
padding: 0;
}
/* Block element titles. */
div.title, caption.title {
font-family: sans-serif;
font-weight: bold;
text-align: left;
margin-top: 1.0em;
margin-bottom: 0.5em;
}
div.title + * {
margin-top: 0;
}
td div.title:first-child {
margin-top: 0.0em;
}
div.content div.title:first-child {
margin-top: 0.0em;
}
div.content + div.title {
margin-top: 0.0em;
}
div.sidebarblock > div.content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.listingblock {
margin-right: 0%;
}
div.listingblock > div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock > div.content {
padding-left: 2.0em;
}
div.attribution {
text-align: right;
}
div.verseblock + div.attribution {
text-align: left;
}
div.admonitionblock .icon {
vertical-align: top;
font-size: 1.1em;
font-weight: bold;
text-decoration: underline;
color: #527bbd;
padding-right: 0.5em;
}
div.admonitionblock td.content {
padding-left: 0.5em;
border-left: 2px solid silver;
}
div.exampleblock > div.content {
border-left: 2px solid silver;
padding: 0.5em;
}
div.verseblock div.content {
white-space: pre;
}
div.imageblock div.content { padding-left: 0; }
div.imageblock img { border: 1px solid silver; }
span.image img { border-style: none; }
dl {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
dt {
margin-top: 0.5em;
margin-bottom: 0;
font-style: italic;
}
dd > *:first-child {
margin-top: 0;
}
ul, ol {
list-style-position: outside;
}
ol.olist2 {
list-style-type: lower-alpha;
}
div.tableblock > table {
border: 3px solid #527bbd;
}
thead {
font-family: sans-serif;
font-weight: bold;
}
tfoot {
font-weight: bold;
}
div.hlist {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
td.hlist1 {
vertical-align: top;
font-style: italic;
padding-right: 0.8em;
}
td.hlist2 {
vertical-align: top;
}
@media print {
div#footer-badges { display: none; }
}
div#toctitle {
color: #527bbd;
font-family: sans-serif;
font-size: 1.1em;
font-weight: bold;
margin-top: 1.0em;
margin-bottom: 0.1em;
}
div.toclevel1, div.toclevel2 {
margin-top: 0;
margin-bottom: 0;
}
div.toclevel2 {
margin-left: 2em;
}
include1::./stylesheets/xhtml11-manpage.css[]
/* Workarounds for IE6's broken and incomplete CSS2. */
div.sidebar-content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.sidebar-title, div.image-title {
font-family: sans-serif;
font-weight: bold;
margin-top: 0.0em;
margin-bottom: 0.5em;
}
div.listingblock div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock-content {
padding-left: 2.0em;
}
div.exampleblock-content {
border-left: 2px solid silver;
padding-left: 0.5em;
}
/* IE6 sets dynamically generated links as visited. */
div#toc a:visited { color: blue; }
</style>
<title>cset-shield(1)</title>
</head>
<body>
<div id="header">
<h1>
cset-shield(1) Manual Page
</h1>
<h2>NAME</h2>
<div class="sectionbody">
<p>cset-shield -
cpuset supercommand which implements cpu shielding
</p>
</div>
</div>
<h2>SYNOPSIS</h2>
<div class="sectionbody">
<div class="verseblock">
<div class="content"><em>cset</em> [cset options] <em>shield</em> [shield options] [args]
<em>cset</em> shield --help
<em>cset</em> shield
<em>cset</em> shield --cpu 1-7
<em>cset</em> shield --cpu 1-7 --kthread=on
<em>cset</em> shield --exec /opt/software/myapp/doit --my_opt1 --my_opt2
<em>cset</em> shield --user appuser --exec run_benchmark.sh
<em>cset</em> shield --shield 1024,2048,5000-1000
<em>cset</em> shield --unshield 6000-8500
<em>cset</em> shield --kthread=off
<em>cset</em> shield --kthread=on
<em>cset</em> shield --shield bash</div></div>
</div>
<h2>OPTIONS</h2>
<div class="sectionbody">
<dl>
<dt>
-h, --help
</dt>
<dd>
<p>
prints the list of options for this command
</p>
</dd>
<dt>
-c CPUSPEC, --cpu=CPUSPEC
</dt>
<dd>
<p>
modifies or initializes the shield cpusets
</p>
</dd>
<dt>
-r, --reset
</dt>
<dd>
<p>
destroys the shield
</p>
</dd>
<dt>
-e, --exec
</dt>
<dd>
<p>
executes args in the shield
</p>
</dd>
<dt>
--user=USER
</dt>
<dd>
<p>
use this USER for --exec (id or name)
</p>
</dd>
<dt>
--group=GROUP
</dt>
<dd>
<p>
use this GROUP for --exec (id or name)
</p>
</dd>
<dt>
-s PIDSPEC, --shield=PIDSPEC
</dt>
<dd>
<p>
shield specified PIDSPEC of processes or threads
</p>
</dd>
<dt>
-u PIDSPEC, --unshield=PIDSPEC
</dt>
<dd>
<p>
remove specified PIDSPEC of processes or threads from the shield,
the task keep running in the unshielded cpuset
</p>
</dd>
<dt>
-k on|off, --kthread=on|off
</dt>
<dd>
<p>
shield from unbound interrupt threads as well
</p>
</dd>
<dt>
-f, --force
</dt>
<dd>
<p>
force operation, use with care
</p>
</dd>
<dt>
-v, --verbose
</dt>
<dd>
<p>
prints more detailed output, additive
</p>
</dd>
<dt>
--sysset=SYSSET
</dt>
<dd>
<p>
optionally specify system cpuset name
</p>
</dd>
<dt>
--userset=USERSET
</dt>
<dd>
<p>
optionally specify user cpuset name
</p>
</dd>
</dl>
</div>
<h2>DESCRIPTION</h2>
<div class="sectionbody">
<p>This is a supercommand that creates basic cpu shielding. The
normal cset commands can of course be used to create this basic
shield, but the shield command combines many such commands to
create and manage a common type of cpu shielding setup.</p>
<p>The concept of shielding implies at minimum three cpusets, for
example: root, user and system. The root cpuset always exists in
all implementations of cpusets and contains all available CPUs on
the machine. The system cpuset is so named because normal system
tasks are made to run on it. The user cpuset is so named because
that is the "shielded" cpuset on which you would run your tasks
of interest.</p>
<p>Usually, CPU zero would be in the system set and the rest of the
CPUs would be in the user set. After creation of the cpusets,
all processes running in the root cpuset are moved to the system
cpuset. Thus any new processes or threads spawned from these
processes will also run the system cpuset.</p>
<p>If the optional --kthread=on option is given to the shield
command, then all kernel threads (with exception of the per-CPU
bound interrupt kernel threads) are also moved to the system set.</p>
<p>One executes processes on the shielded user cpuset with the
--exec subcommand or moves processes or threads to the shielded
cpuset with the --shield subcommand.</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">You do not need to specify which cpuset a process or thread
is running in initially when using the --shield subcommand.</td>
</tr></table>
</div>
<p>To create a shield, you would execute the shield command with the
--cpu option that specifies CPUSPEC argument that assigns CPUs to
be under the shield (this means assigned to the user cpuset, all
other cpus will be assigned to the system set).</p>
<p>For example:</p>
<p><strong><tt># cset shield --cpu=3</tt></strong></p>
<p>On a 4-way machine, this command will dedicate the first
3 processors, CPU0-CPU2, for the system set (unshielded)
and only the last processor, CPU3, for the user set
(shielded).</p>
<p>The CPUSPEC will accept a comma separated list of CPUs and
inclusive range specifications. For example, --cpu=1,3,5-7 will
assign CPU1, CPU3, CPU5, CPU6, and CPU7 to the user (or shielded)
cpuset.</p>
<p>If you do not like the names "system" and "user" for the
unshielded and shielded sets respectively, or if those names are
used already, then use the --sysset and --userset options.</p>
<p>For example:</p>
<p><strong><tt># cset shield --sysset=free --userset=cage --cpu=2,3 --kthread=on</tt></strong></p>
<p>The above command will use the name "free" for the unshielded
system cpuset, the name "cage" for the shielded user cpuset,
initialize these cpusets and dedicate CPU0 and CPU1 to the "free"
set and (on a 4-way machine) dedicate CPU2 and CPU3 to the "cage"
set. Further, the command moves all processes and threads,
including kernel threads from the root cpuset to the "free"
cpuset.</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">If you do use the --syset/--userset options, then you must
continue to use those for every invocation of the shield supercommand.</td>
</tr></table>
</div>
<p>After initialization, you can run the process of interest on the
shielded cpuset with the --exec subcommand, or move processes or
threads already running to the shielded cpuset with the --shield
subcommand.</p>
<p>The PIDSPEC argument taken for the --shield (or -s) subcommand is
a comma separated list of PIDs or TIDs. The list can also
include brackets of PIDs or TIDs that are inclusive of the
endpoints.</p>
<p>For example:</p>
<div class="literalblock">
<div class="content">
<pre><tt>1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700</tt></pre>
</div></div>
<p><strong><tt># cset shield --shield=50-65</tt></strong></p>
<p>The above command moves all processes and threads with PID or
TID in the range 50-65 inclusive, from any cpuset they may
be running in into the shielded user cpuset.</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">The range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.</td>
</tr></table>
</div>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Caution</div>
</td>
<td class="content">Please note that there is no checking of processes you
request to move into the shield with the --shield command. This
means that the tool will happily move, for example, kernel
threads that are bound to specific CPUs with this command. You
can hang your system by indiscriminately specifying arbitrary
PIDs to the --shield command so please be careful.</td>
</tr></table>
</div>
<p>The --unshield (or -u) subcommand will remove the specified
processes or threads from the shielded cpuset and move them into
the unshielded (or system) cpuset. This option also takes a
PIDSPEC argument, the same as for the --shield subcommand.</p>
<p>Both the --shield and the --unshield commands will also finally
output the number of tasks running in the shield and out of the
shield. If you do not specify a PIDSPEC to these commands, then
just that status is output. By specifying also a --verbose in
addition, then you will get a listing of every task that is
running in either the shield or out of the shield.</p>
<p>Using no subcommand, ie. only "cset shield", will output the
status of both shield and non-shield. Tasks will be listed if
--verbose is used.</p>
<p>You can adjust which CPUs are in the shielded cpuset by issuing
the --cpu subcommand again anytime after the shield has been
initialized.</p>
<p>For example if the original shield contained CPU0 and CPU1 in the
system set and CPU2 and CPU3 in the user set, if you then issue
the following command:</p>
<p><strong><tt># cset shield --cpu=1,2,3</tt></strong></p>
<p>then that command will move CPU1 into the shielded "user" cpuset.
Any processes or threads that were running on CPU1 that belonged
to the unshielded "system" cpuset are migrated to CPU0 by the
system.</p>
<p>The --reset subcommand will in essence destroy the shield. For
example, if there was a shield on a 4-way machine with CPU0 in
system and CPUs 1-3 in user with processes running on the user
cpuset (i.e. in the shield), and a --reset subcommand was issued,
then all processes running in both system and user cpusets would
be migrated to the root cpuset (which has access to all CPUs and
never goes away), after which both system and user cpusets would
be destroyed.</p>
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">Even though you can mix general usage of cpusets with
the shielding concepts described here, you generally will not
want to. For more complex shielding or usage scenarios, one
would generally use the normal cpuset commands (i.e. cset set
and proc) directly.</td>
</tr></table>
</div>
</div>
<h2>LICENSE</h2>
<div class="sectionbody">
<p>Cpuset is licensed under the GNU GPL V2 only.</p>
</div>
<h2>COPYRIGHT</h2>
<div class="sectionbody">
<p>Copyright (c) 2008 Novell Inc.</p>
</div>
<h2>AUTHOR</h2>
<div class="sectionbody">
<p>Written by Alex Tsariounov &lt;alext@novell.com&gt;.</p>
<p>Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.</p>
</div>
<h2>SEE ALSO</h2>
<div class="sectionbody">
<p><a href="cset.html">cset(1)</a>, <a href="cset-set.html">cset-set(1)</a>, <a href="cset-proc.html">cset-proc(1)</a>, <a href="taskset.html">taskset(1)</a>,
<a href="chrt.html">chrt(1)</a></p>
<p>/usr/src/linux/Documentation/cpusets.txt</p>
</div>
<div id="footer">
<div id="footer-text">
Version 1.0<br />
Last updated 30-May-2008 14:33:24 MDT
</div>
</div>
</body>
</html>

@ -0,0 +1,238 @@
cset-shield(1)
==============
Alex Tsariounov <alext@novell.com>
v1.0, May 2008
NAME
----
cset-shield - cpuset supercommand which implements cpu shielding
SYNOPSIS
--------
[verse]
'cset' [cset options] 'shield' [shield options] [args]
'cset' shield --help
'cset' shield
'cset' shield --cpu 1-7
'cset' shield --cpu 1-7 --kthread=on
'cset' shield --exec /opt/software/myapp/doit --my_opt1 --my_opt2
'cset' shield --user appuser --exec run_benchmark.sh
'cset' shield --shield 1024,2048,5000-1000
'cset' shield --unshield 6000-8500
'cset' shield --kthread=off
'cset' shield --kthread=on
'cset' shield --shield bash
OPTIONS
-------
-h, --help::
prints the list of options for this command
-c CPUSPEC, --cpu=CPUSPEC::
modifies or initializes the shield cpusets
-r, --reset::
destroys the shield
-e, --exec::
executes args in the shield
--user=USER::
use this USER for --exec (id or name)
--group=GROUP::
use this GROUP for --exec (id or name)
-s PIDSPEC, --shield=PIDSPEC::
shield specified PIDSPEC of processes or threads
-u PIDSPEC, --unshield=PIDSPEC::
remove specified PIDSPEC of processes or threads from the shield,
the task keep running in the unshielded cpuset
-k on|off, --kthread=on|off::
shield from unbound interrupt threads as well
-f, --force::
force operation, use with care
-v, --verbose::
prints more detailed output, additive
--sysset=SYSSET::
optionally specify system cpuset name
--userset=USERSET::
optionally specify user cpuset name
DESCRIPTION
-----------
This is a supercommand that creates basic cpu shielding. The
normal cset commands can of course be used to create this basic
shield, but the shield command combines many such commands to
create and manage a common type of cpu shielding setup.
The concept of shielding implies at minimum three cpusets, for
example: root, user and system. The root cpuset always exists in
all implementations of cpusets and contains all available CPUs on
the machine. The system cpuset is so named because normal system
tasks are made to run on it. The user cpuset is so named because
that is the "shielded" cpuset on which you would run your tasks
of interest.
Usually, CPU zero would be in the system set and the rest of the
CPUs would be in the user set. After creation of the cpusets,
all processes running in the root cpuset are moved to the system
cpuset. Thus any new processes or threads spawned from these
processes will also run the system cpuset.
If the optional --kthread=on option is given to the shield
command, then all kernel threads (with exception of the per-CPU
bound interrupt kernel threads) are also moved to the system set.
One executes processes on the shielded user cpuset with the
--exec subcommand or moves processes or threads to the shielded
cpuset with the --shield subcommand.
NOTE: You do not need to specify which cpuset a process or thread
is running in initially when using the --shield subcommand.
To create a shield, you would execute the shield command with the
--cpu option that specifies CPUSPEC argument that assigns CPUs to
be under the shield (this means assigned to the user cpuset, all
other cpus will be assigned to the system set).
For example:
*+# cset shield --cpu=3+*
On a 4-way machine, this command will dedicate the first
3 processors, CPU0-CPU2, for the system set (unshielded)
and only the last processor, CPU3, for the user set
(shielded).
The CPUSPEC will accept a comma separated list of CPUs and
inclusive range specifications. For example, --cpu=1,3,5-7 will
assign CPU1, CPU3, CPU5, CPU6, and CPU7 to the user (or shielded)
cpuset.
If you do not like the names "system" and "user" for the
unshielded and shielded sets respectively, or if those names are
used already, then use the --sysset and --userset options.
For example:
*+# cset shield --sysset=free --userset=cage --cpu=2,3 --kthread=on+*
The above command will use the name "free" for the unshielded
system cpuset, the name "cage" for the shielded user cpuset,
initialize these cpusets and dedicate CPU0 and CPU1 to the "free"
set and (on a 4-way machine) dedicate CPU2 and CPU3 to the "cage"
set. Further, the command moves all processes and threads,
including kernel threads from the root cpuset to the "free"
cpuset.
NOTE: If you do use the --syset/--userset options, then you must
continue to use those for every invocation of the shield supercommand.
After initialization, you can run the process of interest on the
shielded cpuset with the --exec subcommand, or move processes or
threads already running to the shielded cpuset with the --shield
subcommand.
The PIDSPEC argument taken for the --shield (or -s) subcommand is
a comma separated list of PIDs or TIDs. The list can also
include brackets of PIDs or TIDs that are inclusive of the
endpoints.
For example:
1,2,5 Means processes 1, 2 and 5
1,2,600-700 Means processes 1, 2 and from 600 to 700
*+# cset shield --shield=50-65+*
The above command moves all processes and threads with PID or
TID in the range 50-65 inclusive, from any cpuset they may
be running in into the shielded user cpuset.
NOTE: The range of PIDs or TIDs does not need to have every
position populated. In other words, for the example above, if
there is only one process, say PID 57, in the range of 50-65,
then only that process will be moved.
CAUTION: Please note that there is no checking of processes you
request to move into the shield with the --shield command. This
means that the tool will happily move, for example, kernel
threads that are bound to specific CPUs with this command. You
can hang your system by indiscriminately specifying arbitrary
PIDs to the --shield command so please be careful.
The --unshield (or -u) subcommand will remove the specified
processes or threads from the shielded cpuset and move them into
the unshielded (or system) cpuset. This option also takes a
PIDSPEC argument, the same as for the --shield subcommand.
Both the --shield and the --unshield commands will also finally
output the number of tasks running in the shield and out of the
shield. If you do not specify a PIDSPEC to these commands, then
just that status is output. By specifying also a --verbose in
addition, then you will get a listing of every task that is
running in either the shield or out of the shield.
Using no subcommand, ie. only "cset shield", will output the
status of both shield and non-shield. Tasks will be listed if
--verbose is used.
You can adjust which CPUs are in the shielded cpuset by issuing
the --cpu subcommand again anytime after the shield has been
initialized.
For example if the original shield contained CPU0 and CPU1 in the
system set and CPU2 and CPU3 in the user set, if you then issue
the following command:
*+# cset shield --cpu=1,2,3+*
then that command will move CPU1 into the shielded "user" cpuset.
Any processes or threads that were running on CPU1 that belonged
to the unshielded "system" cpuset are migrated to CPU0 by the
system.
The --reset subcommand will in essence destroy the shield. For
example, if there was a shield on a 4-way machine with CPU0 in
system and CPUs 1-3 in user with processes running on the user
cpuset (i.e. in the shield), and a --reset subcommand was issued,
then all processes running in both system and user cpusets would
be migrated to the root cpuset (which has access to all CPUs and
never goes away), after which both system and user cpusets would
be destroyed.
NOTE: Even though you can mix general usage of cpusets with
the shielding concepts described here, you generally will not
want to. For more complex shielding or usage scenarios, one
would generally use the normal cpuset commands (i.e. cset set
and proc) directly.
LICENSE
-------
Cpuset is licensed under the GNU GPL V2 only.
COPYRIGHT
---------
Copyright (c) 2008 Novell Inc.
AUTHOR
------
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
SEE ALSO
--------
manlink:cset[1], csetlink:set[], csetlink:proc[], manlink:taskset[1],
manlink:chrt[1]
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,90 @@
.\" ** You probably do not want to edit this file directly **
.\" It was generated using the DocBook XSL Stylesheets (version 1.69.1).
.\" Instead of manually editing it, you probably should edit the DocBook XML
.\" source for it and then use the DocBook XSL Stylesheets to regenerate it.
.TH "CSET" "1" "05/30/2008" "" ""
.\" disable hyphenation
.nh
.\" disable justification (adjust text to left margin only)
.ad l
.SH "NAME"
cset \- manage cpusets functions in the Linux kernel
.SH "SYNOPSIS"
.sp
.nf
\fIcset\fR [\-\-version | \-\-help | \-\-log]
\fIcset\fR [help <command> | <command> \-\-help]
\fIcset\fR [cset options] <command> [command options] [args]
.fi
.SH "DESCRIPTION"
.sp
.it 1 an-trap
.nr an-no-space-flag 1
.nr an-break-flag 1
.br
\fBNote\fR
In general, you need to have root permissions to run cset. The tool mounts the cpusets filesystem and manipulates it. Non\-root users do not have permission for these actions.
Cpuset is a Python application to make using the cpusets facilities in the Linux kernel easier. The actual included command is called \fIcset\fR and it allows manipulation of cpusets on the system and provides higher level functions such as implementation and control of a basic cpu shielding setup.
.SS "Typical uses of cpuset include"
.TP
Setting up and managing a simple shielded CPU environment
The concept of \fIshielded\fR cpus is that a certain number of cpus are partitioned off on the system and only processes that are of interest are run on these cpus (i.e., inside the shield).
For a simple shielded configuration, one typically uses three cpusets: the root set, a system set and a user set. \fICset\fR includes a super command that implements this strategy and lets you easily manage it. See \fBcset\-shield\fR(1) for more details.
.TP
Setting up and managing a comples shielding environment
Shielding can be more complex of course where concepts such as priority cpusets and intersecting cpuset can be used. You can use \fIcset\fR to help manage this type of shielding as well. You will need to use the \fBcset\-set\fR(1) and \fBcset\-proc\fR(1) subcommands directly to do that.
.TP
Managing cpusets on the system
The cset subcommand \fBcset\-set\fR(1) allows you to create and destroy arbitrary cpusets on the system and assign arbitrary cpus and memory nodes to them. The cpusets so created have to follow the Linux kernel cpuset rules. See the \fBcset\-set\fR(1) subcommand for more details.
.TP
Managing processes that run on various system cpusets
The cset subcommand \fBcset\-proc\fR(1) allows you to manage processes running on various cpusets created on the system. You can exec new processes in specific cpusets and move tasks around existing cpusets. See the \fBcset\-proc\fR(1) subcommand for more details.
.SH "OPTIONS"
The following generic option flags are available. Additional options are available per\-command, and documented in the command\-specific documentation.
.TP
\fIcset\fR \-\-version
Prints the cpuset suite version that the \fIcset\fR program came from, as well as version of other components used, such as Python and the kernel.
.TP
\fIcset\fR \-\-help
Prints the synopsis and a list of all commands.
.TP
\fIcset\fR \-\-log <filename>
Creates a log file for the current run. All manner of useful information is stored in this file. This is usually used to debug cpuset when things don't go as planned.
.TP
\fIcset\fR help \fIcommand\fR
Although not strictly an "option", this is a very helpful way to execute help with cpuset. The command prints out a lengthy summary of how the specified subcommand works.
.SH "CPUSET COMMANDS"
The cpuset commands are divided into groups, according to the primary purpose of those commands. Following is a short description of each command. A more detailed description is available in individual command manpages. Those manpages are named \fIcset\-<command>(1)\fR. The first command, \fIhelp\fR, is especially useful as it prints out a long summary of what a particular command does.
.TP
\fIcset help\fR
print the detailed command usage
.TP
\fIcset version\fR
display version information
.TP
\fIcset copyright\fR
display copyright information
.TP
\fIcset shield\fR
supercommand to set up and manage basic shielding (see \fBcset\-shield\fR(1))
.TP
\fIcset set\fR
create, modify and destroy cpusets (see \fBcset\-set\fR(1))
.TP
\fIcset proc\fR
create and manage processes within cpusets (see \fBcset\-proc\fR(1))
.SH "LICENSE"
Cpuset is licensed under the GNU GPL V2 only.
.SH "COPYRIGHT"
Copyright (c) 2008 Novell Inc.
.SH "AUTHOR"
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked GIT (stgit) v0.13 (see http://gna.org/projects/stgit and http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
.SH "SEE ALSO"
\fBcset\-set\fR(1), \fBcset\-proc\fR(1), \fBcset\-shield\fR(1), \fBtaskset\fR(1), \fBchrt\fR(1)
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,517 @@
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<meta name="generator" content="AsciiDoc 8.1.0" />
<style type="text/css">
/* Debug borders */
p, li, dt, dd, div, pre, h1, h2, h3, h4, h5, h6 {
/*
border: 1px solid red;
*/
}
body {
margin: 1em 5% 1em 5%;
}
a {
color: blue;
text-decoration: underline;
}
a:visited {
color: fuchsia;
}
em {
font-style: italic;
}
strong {
font-weight: bold;
}
tt {
color: navy;
}
h1, h2, h3, h4, h5, h6 {
color: #527bbd;
font-family: sans-serif;
margin-top: 1.2em;
margin-bottom: 0.5em;
line-height: 1.3;
}
h1 {
border-bottom: 2px solid silver;
}
h2 {
border-bottom: 2px solid silver;
padding-top: 0.5em;
}
div.sectionbody {
font-family: serif;
margin-left: 0;
}
hr {
border: 1px solid silver;
}
p {
margin-top: 0.5em;
margin-bottom: 0.5em;
}
pre {
padding: 0;
margin: 0;
}
span#author {
color: #527bbd;
font-family: sans-serif;
font-weight: bold;
font-size: 1.1em;
}
span#email {
}
span#revision {
font-family: sans-serif;
}
div#footer {
font-family: sans-serif;
font-size: small;
border-top: 2px solid silver;
padding-top: 0.5em;
margin-top: 4.0em;
}
div#footer-text {
float: left;
padding-bottom: 0.5em;
}
div#footer-badges {
float: right;
padding-bottom: 0.5em;
}
div#preamble,
div.tableblock, div.imageblock, div.exampleblock, div.verseblock,
div.quoteblock, div.literalblock, div.listingblock, div.sidebarblock,
div.admonitionblock {
margin-right: 10%;
margin-top: 1.5em;
margin-bottom: 1.5em;
}
div.admonitionblock {
margin-top: 2.5em;
margin-bottom: 2.5em;
}
div.content { /* Block element content. */
padding: 0;
}
/* Block element titles. */
div.title, caption.title {
font-family: sans-serif;
font-weight: bold;
text-align: left;
margin-top: 1.0em;
margin-bottom: 0.5em;
}
div.title + * {
margin-top: 0;
}
td div.title:first-child {
margin-top: 0.0em;
}
div.content div.title:first-child {
margin-top: 0.0em;
}
div.content + div.title {
margin-top: 0.0em;
}
div.sidebarblock > div.content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.listingblock {
margin-right: 0%;
}
div.listingblock > div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock > div.content {
padding-left: 2.0em;
}
div.attribution {
text-align: right;
}
div.verseblock + div.attribution {
text-align: left;
}
div.admonitionblock .icon {
vertical-align: top;
font-size: 1.1em;
font-weight: bold;
text-decoration: underline;
color: #527bbd;
padding-right: 0.5em;
}
div.admonitionblock td.content {
padding-left: 0.5em;
border-left: 2px solid silver;
}
div.exampleblock > div.content {
border-left: 2px solid silver;
padding: 0.5em;
}
div.verseblock div.content {
white-space: pre;
}
div.imageblock div.content { padding-left: 0; }
div.imageblock img { border: 1px solid silver; }
span.image img { border-style: none; }
dl {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
dt {
margin-top: 0.5em;
margin-bottom: 0;
font-style: italic;
}
dd > *:first-child {
margin-top: 0;
}
ul, ol {
list-style-position: outside;
}
ol.olist2 {
list-style-type: lower-alpha;
}
div.tableblock > table {
border: 3px solid #527bbd;
}
thead {
font-family: sans-serif;
font-weight: bold;
}
tfoot {
font-weight: bold;
}
div.hlist {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
td.hlist1 {
vertical-align: top;
font-style: italic;
padding-right: 0.8em;
}
td.hlist2 {
vertical-align: top;
}
@media print {
div#footer-badges { display: none; }
}
div#toctitle {
color: #527bbd;
font-family: sans-serif;
font-size: 1.1em;
font-weight: bold;
margin-top: 1.0em;
margin-bottom: 0.1em;
}
div.toclevel1, div.toclevel2 {
margin-top: 0;
margin-bottom: 0;
}
div.toclevel2 {
margin-left: 2em;
}
include1::./stylesheets/xhtml11-manpage.css[]
/* Workarounds for IE6's broken and incomplete CSS2. */
div.sidebar-content {
background: #ffffee;
border: 1px solid silver;
padding: 0.5em;
}
div.sidebar-title, div.image-title {
font-family: sans-serif;
font-weight: bold;
margin-top: 0.0em;
margin-bottom: 0.5em;
}
div.listingblock div.content {
border: 1px solid silver;
background: #f4f4f4;
padding: 0.5em;
}
div.quoteblock-content {
padding-left: 2.0em;
}
div.exampleblock-content {
border-left: 2px solid silver;
padding-left: 0.5em;
}
/* IE6 sets dynamically generated links as visited. */
div#toc a:visited { color: blue; }
</style>
<title>cset(1)</title>
</head>
<body>
<div id="header">
<h1>
cset(1) Manual Page
</h1>
<h2>NAME</h2>
<div class="sectionbody">
<p>cset -
manage cpusets functions in the Linux kernel
</p>
</div>
</div>
<h2>SYNOPSIS</h2>
<div class="sectionbody">
<div class="verseblock">
<div class="content"><em>cset</em> [--version | --help | --log]
<em>cset</em> [help &lt;command&gt; | &lt;command&gt; --help]
<em>cset</em> [cset options] &lt;command&gt; [command options] [args]</div></div>
</div>
<h2>DESCRIPTION</h2>
<div class="sectionbody">
<div class="admonitionblock">
<table><tr>
<td class="icon">
<div class="title">Note</div>
</td>
<td class="content">In general, you need to have root permissions to run cset. The
tool mounts the cpusets filesystem and manipulates it. Non-root users
do not have permission for these actions.</td>
</tr></table>
</div>
<p>Cpuset is a Python application to make using the cpusets facilities in
the Linux kernel easier. The actual included command is called <em>cset</em>
and it allows manipulation of cpusets on the system and provides
higher level functions such as implementation and control of a basic
cpu shielding setup.</p>
<h3>Typical uses of cpuset include</h3>
<dl>
<dt>
Setting up and managing a simple shielded CPU environment
</dt>
<dd>
<p>
The concept of <em>shielded</em> cpus is that a certain number of
cpus are partitioned off on the system and only processes that
are of interest are run on these cpus (i.e., inside the shield).
</p>
<p>For a simple shielded configuration, one typically uses three cpusets:
the root set, a system set and a user set. <em>Cset</em> includes a super
command that implements this strategy and lets you easily manage it.
See <a href="cset-shield.html">cset-shield(1)</a> for more details.</p>
</dd>
<dt>
Setting up and managing a comples shielding environment
</dt>
<dd>
<p>
Shielding can be more complex of course where concepts such as
priority cpusets and intersecting cpuset can be used. You can
use <em>cset</em> to help manage this type of shielding as well. You
will need to use the <a href="cset-set.html">cset-set(1)</a> and <a href="cset-proc.html">cset-proc(1)</a>
subcommands directly to do that.
</p>
</dd>
<dt>
Managing cpusets on the system
</dt>
<dd>
<p>
The cset subcommand <a href="cset-set.html">cset-set(1)</a> allows you to create and
destroy arbitrary cpusets on the system and assign arbitrary
cpus and memory nodes to them. The cpusets so created have to
follow the Linux kernel cpuset rules. See the <a href="cset-set.html">cset-set(1)</a>
subcommand for more details.
</p>
</dd>
<dt>
Managing processes that run on various system cpusets
</dt>
<dd>
<p>
The cset subcommand <a href="cset-proc.html">cset-proc(1)</a> allows you to manage
processes running on various cpusets created on the system.
You can exec new processes in specific cpusets and move tasks
around existing cpusets. See the <a href="cset-proc.html">cset-proc(1)</a> subcommand
for more details.
</p>
</dd>
</dl>
</div>
<h2>OPTIONS</h2>
<div class="sectionbody">
<p>The following generic option flags are available. Additional options
are available per-command, and documented in the command-specific
documentation.</p>
<dl>
<dt>
<em>cset</em> --version
</dt>
<dd>
<p>
Prints the cpuset suite version that the <em>cset</em> program came
from, as well as version of other components used, such as
Python and the kernel.
</p>
</dd>
<dt>
<em>cset</em> --help
</dt>
<dd>
<p>
Prints the synopsis and a list of all commands.
</p>
</dd>
<dt>
<em>cset</em> --log &lt;filename&gt;
</dt>
<dd>
<p>
Creates a log file for the current run. All manner of useful
information is stored in this file. This is usually used to
debug cpuset when things don't go as planned.
</p>
</dd>
<dt>
<em>cset</em> help <em>command</em>
</dt>
<dd>
<p>
Although not strictly an "option", this is a very helpful way
to execute help with cpuset. The command prints out a lengthy
summary of how the specified subcommand works.
</p>
</dd>
</dl>
</div>
<h2>CPUSET COMMANDS</h2>
<div class="sectionbody">
<p>The cpuset commands are divided into groups, according to the primary
purpose of those commands. Following is a short description of each
command. A more detailed description is available in individual
command manpages. Those manpages are named <em>cset-&lt;command&gt;(1)</em>. The
first command, <em>help</em>, is especially useful as it prints out a long
summary of what a particular command does.</p>
<dl>
<dt>
<em>cset help</em>
</dt>
<dd>
<p>
print the detailed command usage
</p>
</dd>
<dt>
<em>cset version</em>
</dt>
<dd>
<p>
display version information
</p>
</dd>
<dt>
<em>cset copyright</em>
</dt>
<dd>
<p>
display copyright information
</p>
</dd>
<dt>
<em>cset shield</em>
</dt>
<dd>
<p>
supercommand to set up and manage basic shielding (see
<a href="cset-shield.html">cset-shield(1)</a>)
</p>
</dd>
<dt>
<em>cset set</em>
</dt>
<dd>
<p>
create, modify and destroy cpusets (see <a href="cset-set.html">cset-set(1)</a>)
</p>
</dd>
<dt>
<em>cset proc</em>
</dt>
<dd>
<p>
create and manage processes within cpusets (see
<a href="cset-proc.html">cset-proc(1)</a>)
</p>
</dd>
</dl>
</div>
<h2>LICENSE</h2>
<div class="sectionbody">
<p>Cpuset is licensed under the GNU GPL V2 only.</p>
</div>
<h2>COPYRIGHT</h2>
<div class="sectionbody">
<p>Copyright (c) 2008 Novell Inc.</p>
</div>
<h2>AUTHOR</h2>
<div class="sectionbody">
<p>Written by Alex Tsariounov &lt;alext@novell.com&gt;.</p>
<p>Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.</p>
</div>
<h2>SEE ALSO</h2>
<div class="sectionbody">
<p><a href="cset-set.html">cset-set(1)</a>, <a href="cset-proc.html">cset-proc(1)</a>, <a href="cset-shield.html">cset-shield(1)</a>, <a href="taskset.html">taskset(1)</a>,
<a href="chrt.html">chrt(1)</a></p>
<p>/usr/src/linux/Documentation/cpusets.txt</p>
</div>
<div id="footer">
<div id="footer-text">
Version 1.0<br />
Last updated 30-May-2008 14:33:24 MDT
</div>
</div>
</body>
</html>

@ -0,0 +1,132 @@
cset(1)
======
Alex Tsariounov <alext@novell.com>
v1.0, May 2008
NAME
----
cset - manage cpusets functions in the Linux kernel
SYNOPSIS
--------
[verse]
'cset' [--version | --help | --log]
'cset' [help <command> | <command> --help]
'cset' [cset options] <command> [command options] [args]
DESCRIPTION
-----------
NOTE: In general, you need to have root permissions to run cset. The
tool mounts the cpusets filesystem and manipulates it. Non-root users
do not have permission for these actions.
Cpuset is a Python application to make using the cpusets facilities in
the Linux kernel easier. The actual included command is called 'cset'
and it allows manipulation of cpusets on the system and provides
higher level functions such as implementation and control of a basic
cpu shielding setup.
Typical uses of cpuset include
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Setting up and managing a simple shielded CPU environment::
The concept of 'shielded' cpus is that a certain number of
cpus are partitioned off on the system and only processes that
are of interest are run on these cpus (i.e., inside the shield).
+
For a simple shielded configuration, one typically uses three cpusets:
the root set, a system set and a user set. 'Cset' includes a super
command that implements this strategy and lets you easily manage it.
See csetlink:shield[] for more details.
Setting up and managing a comples shielding environment::
Shielding can be more complex of course where concepts such as
priority cpusets and intersecting cpuset can be used. You can
use 'cset' to help manage this type of shielding as well. You
will need to use the csetlink:set[] and csetlink:proc[]
subcommands directly to do that.
Managing cpusets on the system::
The cset subcommand csetlink:set[] allows you to create and
destroy arbitrary cpusets on the system and assign arbitrary
cpus and memory nodes to them. The cpusets so created have to
follow the Linux kernel cpuset rules. See the csetlink:set[]
subcommand for more details.
Managing processes that run on various system cpusets::
The cset subcommand csetlink:proc[] allows you to manage
processes running on various cpusets created on the system.
You can exec new processes in specific cpusets and move tasks
around existing cpusets. See the csetlink:proc[] subcommand
for more details.
OPTIONS
-------
The following generic option flags are available. Additional options
are available per-command, and documented in the command-specific
documentation.
'cset' --version::
Prints the cpuset suite version that the 'cset' program came
from, as well as version of other components used, such as
Python and the kernel.
'cset' --help::
Prints the synopsis and a list of all commands.
'cset' --log <filename>::
Creates a log file for the current run. All manner of useful
information is stored in this file. This is usually used to
debug cpuset when things don't go as planned.
'cset' help 'command'::
Although not strictly an "option", this is a very helpful way
to execute help with cpuset. The command prints out a lengthy
summary of how the specified subcommand works.
CPUSET COMMANDS
---------------
The cpuset commands are divided into groups, according to the primary
purpose of those commands. Following is a short description of each
command. A more detailed description is available in individual
command manpages. Those manpages are named 'cset-<command>(1)'. The
first command, 'help', is especially useful as it prints out a long
summary of what a particular command does.
'cset help'::
print the detailed command usage
'cset version'::
display version information
'cset copyright'::
display copyright information
'cset shield'::
supercommand to set up and manage basic shielding (see
csetlink:shield[])
'cset set'::
create, modify and destroy cpusets (see csetlink:set[])
'cset proc'::
create and manage processes within cpusets (see
csetlink:proc[])
LICENSE
-------
Cpuset is licensed under the GNU GPL V2 only.
COPYRIGHT
---------
Copyright (c) 2008 Novell Inc.
AUTHOR
------
Written by Alex Tsariounov <alext@novell.com>.
Some substrate code and ideas were taken from the excellent Stacked
GIT (stgit) v0.13 (see http://gna.org/projects/stgit and
http://www.procode.org/stgit). Stacked GIT is under GPL V2 or later.
SEE ALSO
--------
csetlink:set[], csetlink:proc[], csetlink:shield[], manlink:taskset[1],
manlink:chrt[1]
/usr/src/linux/Documentation/cpusets.txt

@ -0,0 +1,3 @@
[titles]
underlines="##","==","--","~~","^^"

@ -0,0 +1,51 @@
Cpuset Tutorial
###############
Cpuset (cset) is a Python application that provides a command line
application front end for the cpusets functionality found in the Linux
kernel.
Basic Operation
===============
Help
----
For a full list of cset commands:
cset help
For help on individual subcommands:
cset help <command>
For options of individual subcommands:
cset <command> (-h | --help)
Built-in Super Commands
-----------------------
Cpuset has some built-in "super commands" that encapsulate common use
cases of cpusets and make their interface simple to use.
cset shield
~~~~~~~~~~~
This super command sets up a shielded set of cpus environment.
Basic cpuset Manipulation Commands
----------------------------------
Advanced Usage
==============
This describes more advanced usage of cpusets.
Priority Realtime CPU Sets
---------------------------

@ -0,0 +1,7 @@
[install]
prefix: /usr
[bdist_rpm]
group=System/Management
build_requires=python python-devel asciidoc xmlto
requires=python

@ -0,0 +1,27 @@
#!/usr/bin/env python
import glob
from distutils.core import setup
from cpuset.version import version
setup(name = 'cpuset',
version = version,
license = 'GPLv2',
author = 'Alex Tsariounov',
author_email = 'alext@novell.com',
url = 'http://developer.novell.com/wiki/index.php/Cpuset',
description = 'Allows manipluation of cpusets and provides higher level functions.',
long_description = \
'Cpuset is a Python application to make using the cpusets facilities in the Linux\n'
'kernel easier. The actual included command is called cset and it allows\n'
'manipulation of cpusets on the system and provides higher level functions such as\n'
'implementation and control of a basic cpu shielding setup.',
scripts = ['cset'],
packages = ['cpuset', 'cpuset.commands'],
data_files = [
('share/doc/packages/cpuset', ['README', 'COPYING']),
('share/doc/packages/cpuset', glob.glob('doc/*.html')),
('share/doc/packages/cpuset', glob.glob('doc/*.txt')),
]
)

@ -0,0 +1,27 @@
# Run cpuset tests
# Suite design taken from git which is Copyright (c) 2005 Junio C Hamano
#
# Copyright (C) 2007 Novell Inc.
# Author: Alex Tsariounov <alext@novell.com>
#
#CSET_TEST_OPTS=--verbose --debug
SHELL_PATH ?= $(SHELL)
TAR ?= $(TAR)
# Shell quote;
SHELL_PATH_SQ = $(subst ','\'',$(SHELL_PATH))
T = $(wildcard t[0-9][0-9][0-9][0-9]-*.sh)
all: $(T) clean
$(T):
@echo "*** $@ ***"; '$(SHELL_PATH_SQ)' $@ $(GIT_TEST_OPTS)
clean:
rm -fr trash
.PHONY: $(T) clean
.NOPARALLEL:

@ -0,0 +1,4 @@
CPUSET Tests
============
Loading…
Cancel
Save