summaryrefslogtreecommitdiffstats
path: root/contrib/top/INSTALL
blob: c435c06142c28c5c8ee47735774dbf6f02050424 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
                             TOP
                         Version 3.5

                       William LeFebvre
		      and a cast of many

INSTALLATION

Configuration and installation of top is very straightforward.  After
unpacking the sources, run the script "Configure".  It will present you
with a series of questions, all of which should be explained in the
presentation.  After you have answered all the questions, "Configure" will
perform all the necessary configuration.  Once this is finished, type
"make install".  Make will compile the sources then install the resulting
executable and manual page in the appropriate places.

The most difficult step in the configuration is the choice of an
appropriate machine-specific module.  The Configure script gives you a
list of choices complete with brief descriptions of when each choice is
appropriate.  Each module is contained in a separate c file in the
directory "machine".  The module contains all of the machine-specific code
that makes top work correctly on the architecture in question.  All of the
code in the top-level directory is machine-independent (or at least
strives to be).  Hints for some module choices that are not obvious are
given at the end of this file.

The first comment in each c file in that directory contains the synopsis
AND a detailed description of the machines for which that module is
appropriate.  It also contains a list of authors for that module.  If you
are really stumped in this choice, use grep to find your machine
manufacturer's name or operating system name in machine/*.c.  If you still
can't find one that is appropriate, then chances are very good that one
hasn't been written yet.  If that is the case, then you are out of luck.

HANDLING MULTIPLE ARCHITECTURES

If you need to recompile top for a different architecture (that is, using
a different module) you need to reconfigure top.  A short cut is available
to make this a little easier.  If all of your previous answers to the
configuration questions (except for the module name of course) are
adequate for the new architecture, then you can just use the command
"Configure <modulename>".  The configuration script will reconfigure top
using the new module and all the answers you gave last time.  It will
finish with a "make clean".  Once that completes, type "make install"
and make will compile the sources and do the installation.

HANDLING MULTIPLE OS VERSIONS

By far the most frequently received bug report for top is something like
this: "We just upgraded our operating system to version 99.9.9.9 and top
broke.  What should we do?"  The simple answer is "recompile".

Top is very sensitive to changes in internal kernel data structures
(especially the proc and user structures).  Some operating systems
(especially SunOS) are notorious for changing these structure in every
minor release of the OS.  This means that a top executable made under one
version of the OS will not always work correctly (if even at all) under
another version.  This is just one of those tough facts of life.  There is
really no way around it.

To make life even worse, some operating systems (SunOS again) will use
slightly different proc and user structures on different models.  For
example, "top" built on a SparcStation 2 will not run correctly on a
SparcStation 10, even if they are both running SunOS 4.1.3.  These
unfortunate circumstances make maintaining top very difficult, especially
in an environment that runs several different versions of the same
operating system.

But there is hope.  If your operating system has a properly functioning
"uname" command then you can handle this problem rather gracefully.
Included in the distribution is a shell file called "metatop".  All this
shell file does is:

	exec top-`uname -m`-`uname -r` "$@"

So when you run this script, it execs a filename that is unique to your
specific machine architecture and your OS revision number.

To use "metatop", do the following:

	. on any machine, run Configure and choose the module that is
	  appropriate for the machine
	. for all machines which use the same module:
	    . group machines according to machine architecture AND OS
	      revision number (i.e.: sun4-4.1.1, sun4c-4.1.1, sun4c-4.1.2,
	      sun4-4.1.3, sun4c-4.1.3, sun4m-4.1.3, ...)
	    . for each group, choose one machine from that group and on it
	      run "make clean; make installmeta".


The "installmeta" rule in the makefile will insure that top is compiled,
install the shell file "metatop" as "top", then install the executable
"top" with a name appropriate to the machine architecture and OS revision.


HINTS FOR CHOOSING THE CORRECT MODULE:

SOLARIS 2.x

All versions of Solaris will now work with the module sunos5.  Version
specific modules (such as sunos54) no longer exist.


SUNOS 4.x AND MULTIPROCESSOR ARCHITECTURES

First, we need to be speaking the same language:

sun4	a regular sparc sun 4 architecture machine (sparc station 1,
	sparc station 2, IPC, SLC, etc.)

sun4m	a multiprocessor sparc (Sparc 10, 4/670, 4/690)

I intended to write the sunos4 module so that an executable compiled on a
sun4m machine would work correctly on a sun4 machine.  Unfortunately my
experiments indicate that this cannot be done.  It turns out that the user
structure is so different between these two architectures that nothing
short of a serious hack will make the same executable work correctly on
both machines.  I recommend that you use the separate module "sunos4mp"
when making an executable for a sun4m architecture, and use "sunos4" when
making an executable for sun4 or sun4c architectures.

DIGITAL UNIX V4.0

This is the successor to DECOSF/1.  Use the module decosf1.

SOLBOURNE OPERATING SYSTEM (OS/MP)

If you are running OS/MP version 4.1A, then use the module "osmp4.1a".

If you are running a version of OS/MP OLDER than 4.1A (that is, one
of its predecessors), use the module "sunos4".

If you are running OS/MP 4.1B or LATER, use the module "sunos4mp".

HP/UX OPERATING SYSTEM

The module hpux8 works on all version 8 systems.  Some say that it works
with version 9 as well, but one user did send me a separate module for
version 9.  This module has only been tested on series 800 machines.  I
would recommend the following for those running version 9: try hpux9 and
if it doesn't work then try hpux8.  If neither work, then send mail to me
and/or the modules' authors.  Another note:  we have a model 730 supposedly
running version 9.01.  The module hpux9 did not compile successfully, but
the module hpux8 worked fine.  The module hpux10 works on all revisions of
HP/UX 10 except 10.10, where HP removed the definition of the proc structure
from the system include files.

NET/2 386BSD SYSTEMS

If your version of the operating system has patchkit 2.4 installed,
then you will need to modify machine/m_386bsd.c and uncomment the
definition of PATCHED_KVM.  This patchkit makes what more than a few
people believe to be a wholly unnecessary patch to the way the kvm
routines work.

A/UX SYSTEMS

There is a module for A/UX 3.0 and 3.1.  Whether or not it works for
any other version is not known.  Proceed at your own risk.

Although AUX does not generally have a renice systemcall, it can be
implemented by tweeking kernel memory.  The flag IMPLEMENT_SETPRIORITY
controls the inclusion of this code.  It is off be default.  While
such a simple hack should not be difficult to get right, USE THIS
FEATURE AT YOUR OWN RISK!

OpenPOWER on IntegriCloud