openssl/INSTALL
Richard Levitte 3c65577f1a Rename INSTALL_PREFIX to DESTDIR, remove option --install_prefix
INSTALL_PREFIX is a confusing name, as there's also --prefix.
Instead, tag along with the rest of the open source world and adopt
the Makefile variable DESTDIR to designate the desired staging
directory.

The Configure option --install_prefix is removed, the only way to
designate a staging directory is with the Makefile variable (this is
also implemented for VMS' descrip.mms et al).

Reviewed-by: Rich Salz <rsalz@openssl.org>
2016-02-12 21:54:07 +01:00

396 lines
16 KiB
Plaintext

INSTALLATION ON THE UNIX PLATFORM
---------------------------------
[Installation on DOS (with djgpp), Windows, OpenVMS, MacOS (before MacOS X)
and NetWare is described in INSTALL.DJGPP, INSTALL.WIN, INSTALL.VMS,
INSTALL.MacOS and INSTALL.NW.
This document describes installation on operating systems in the Unix
family.]
To install OpenSSL, you will need:
* make
* Perl 5 with core modules (please read README.PERL)
* The perl module Text::Template (please read README.PERL)
* an ANSI C compiler
* a development environment in form of development libraries and C
header files
* a supported Unix operating system
Quick Start
-----------
If you want to just get on with it, do:
$ ./config
$ make
$ make test
$ make install
[If any of these steps fails, see section Installation in Detail below.]
This will build and install OpenSSL in the default location, which is (for
historical reasons) /usr/local/ssl. If you want to install it anywhere else,
run config like this:
$ ./config --prefix=/usr/local --openssldir=/usr/local/openssl
Configuration Options
---------------------
There are several options to ./config (or ./Configure) to customize
the build:
--prefix=DIR Install in DIR/bin, DIR/lib, DIR/include/openssl.
Configuration files used by OpenSSL will be in DIR/ssl
or the directory specified by --openssldir.
--openssldir=DIR Directory for OpenSSL files. If no prefix is specified,
the library files and binaries are also installed there.
no-autoalginit Don't automatically load all supported ciphers and digests.
Typically OpenSSL will make available all of its supported
ciphers and digests. For a statically linked application this
may be undesirable if small executable size is an objective.
This only affects libcrypto. Ciphers and digests will have to be
loaded manually using EVP_add_cipher() and EVP_add_digest() if
this option is used.
no-autoerrinit Don't automatically load all libcrypto/libssl error strings.
Typically OpenSSL will automatically load human readable error
strings. For a statically linked application this may be
undesirable if small executable size is an objective.
no-threads Don't try to build with support for multi-threaded
applications.
threads Build with support for multi-threaded applications.
This will usually require additional system-dependent options!
See "Note on multi-threading" below.
no-zlib Don't try to build with support for zlib compression and
decompression.
zlib Build with support for zlib compression/decompression.
zlib-dynamic Like "zlib", but has OpenSSL load the zlib library dynamically
when needed. This is only supported on systems where loading
of shared libraries is supported. This is the default choice.
no-shared Don't try to create shared libraries.
shared In addition to the usual static libraries, create shared
libraries on platforms where it's supported. See "Note on
shared libraries" below.
no-asm Do not use assembler code.
386 Use the 80386 instruction set only (the default x86 code is
more efficient, but requires at least a 486). Note: Use
compiler flags for any other CPU specific configuration,
e.g. "-m32" to build x86 code on an x64 system.
no-sse2 Exclude SSE2 code pathes. Normally SSE2 extension is
detected at run-time, but the decision whether or not the
machine code will be executed is taken solely on CPU
capability vector. This means that if you happen to run OS
kernel which does not support SSE2 extension on Intel P4
processor, then your application might be exposed to
"illegal instruction" exception. There might be a way
to enable support in kernel, e.g. FreeBSD kernel can be
compiled with CPU_ENABLE_SSE, and there is a way to
disengage SSE2 code pathes upon application start-up,
but if you aim for wider "audience" running such kernel,
consider no-sse2. Both 386 and no-asm options above imply
no-sse2.
no-<cipher> Build without the specified cipher (bf, cast, des, dh, dsa,
hmac, md2, md5, mdc2, rc2, rc4, rc5, rsa, sha).
The crypto/<cipher> directory can be removed after running
"make depend".
-Dxxx, -lxxx, -Lxxx, -fxxx, -mXXX, -Kxxx These system specific options will
be passed through to the compiler to allow you to
define preprocessor symbols, specify additional libraries,
library directories or other compiler options.
Installation in Detail
----------------------
1a. Configure OpenSSL for your operation system automatically:
$ ./config [options]
This guesses at your operating system (and compiler, if necessary) and
configures OpenSSL based on this guess. Run ./config -t to see
if it guessed correctly. If you want to use a different compiler, you
are cross-compiling for another platform, or the ./config guess was
wrong for other reasons, go to step 1b. Otherwise go to step 2.
On some systems, you can include debugging information as follows:
$ ./config -d [options]
1b. Configure OpenSSL for your operating system manually
OpenSSL knows about a range of different operating system, hardware and
compiler combinations. To see the ones it knows about, run
$ ./Configure
Pick a suitable name from the list that matches your system. For most
operating systems there is a choice between using "cc" or "gcc". When
you have identified your system (and if necessary compiler) use this name
as the argument to ./Configure. For example, a "linux-elf" user would
run:
$ ./Configure linux-elf [options]
If your system is not available, you will have to edit the Configure
program and add the correct configuration for your system. The
generic configurations "cc" or "gcc" should usually work on 32 bit
systems.
Configure creates the file Makefile.ssl from Makefile.in and
defines various macros in crypto/opensslconf.h (generated from
crypto/opensslconf.h.in).
2. Build OpenSSL by running:
$ make
This will build the OpenSSL libraries (libcrypto.a and libssl.a) and the
OpenSSL binary ("openssl"). The libraries will be built in the top-level
directory, and the binary will be in the "apps" directory.
If "make" fails, look at the output. There may be reasons for
the failure that aren't problems in OpenSSL itself (like missing
standard headers). If it is a problem with OpenSSL itself, please
report the problem to <openssl-bugs@openssl.org> (note that your
message will be recorded in the request tracker publicly readable
at https://www.openssl.org/community/index.html#bugs and will be
forwarded to a public mailing list). Include the output of "make
report" in your message. Please check out the request tracker. Maybe
the bug was already reported or has already been fixed.
[If you encounter assembler error messages, try the "no-asm"
configuration option as an immediate fix.]
Compiling parts of OpenSSL with gcc and others with the system
compiler will result in unresolved symbols on some systems.
3. After a successful build, the libraries should be tested. Run:
$ make test
If some tests fail, look at the output. There may be reasons for
the failure that isn't a problem in OpenSSL itself (like a
malfunction with Perl). You may want increased verbosity, that
can be accomplished like this:
$ HARNESS_VERBOSE=yes make test
If you want to run just one or a few specific tests, you can use
the make variable TESTS to specify them, like this:
$ make TESTS='test_rsa test_dsa' test
And of course, you can combine:
$ HARNESS_VERBOSE=yes make TESTS='test_rsa test_dsa' test
You can find the list of available tests like this:
$ make list-tests
Have a look at the manual for the perl module Test::Harness to
see what other HARNESS_* variables there are.
If you find a problem with OpenSSL itself, try removing any
compiler optimization flags from the CFLAG line in Makefile and
run "make clean; make".
Please send a bug report to <openssl-bugs@openssl.org>, and when
you do, please run the following and include the output in your
report:
$ make report
4. If everything tests ok, install OpenSSL with
$ make install
This will create the installation directory (if it does not exist) and
then the following subdirectories:
certs Initially empty, this is the default location
for certificate files.
man/man1 Manual pages for the 'openssl' command line tool
man/man3 Manual pages for the libraries (very incomplete)
misc Various scripts.
private Initially empty, this is the default location
for private key files.
If you didn't choose a different installation prefix, the
following additional subdirectories will be created:
bin Contains the openssl binary and a few other
utility programs.
include/openssl Contains the header files needed if you want to
compile programs with libcrypto or libssl.
lib Contains the OpenSSL library files themselves.
Use "make install_sw" to install the software without documentation,
and "install_docs_html" to install HTML renditions of the manual
pages.
Package builders who want to configure the library for standard
locations, but have the package installed somewhere else so that
it can easily be packaged, can use
$ make DESTDIR=/tmp/package-root install
The specified destination directory will be prepended to all
installation target filenames.
NOTE: The header files used to reside directly in the include
directory, but have now been moved to include/openssl so that
OpenSSL can co-exist with other libraries which use some of the
same filenames. This means that applications that use OpenSSL
should now use C preprocessor directives of the form
#include <openssl/ssl.h>
instead of "#include <ssl.h>", which was used with library versions
up to OpenSSL 0.9.2b.
If you install a new version of OpenSSL over an old library version,
you should delete the old header files in the include directory.
Compatibility issues:
* COMPILING existing applications
To compile an application that uses old filenames -- e.g.
"#include <ssl.h>" --, it will usually be enough to find
the CFLAGS definition in the application's Makefile and
add a C option such as
-I/usr/local/ssl/include/openssl
to it.
But don't delete the existing -I option that points to
the ..../include directory! Otherwise, OpenSSL header files
could not #include each other.
* WRITING applications
To write an application that is able to handle both the new
and the old directory layout, so that it can still be compiled
with library versions up to OpenSSL 0.9.2b without bothering
the user, you can proceed as follows:
- Always use the new filename of OpenSSL header files,
e.g. #include <openssl/ssl.h>.
- Create a directory "incl" that contains only a symbolic
link named "openssl", which points to the "include" directory
of OpenSSL.
For example, your application's Makefile might contain the
following rule, if OPENSSLDIR is a pathname (absolute or
relative) of the directory where OpenSSL resides:
incl/openssl:
-mkdir incl
cd $(OPENSSLDIR) # Check whether the directory really exists
-ln -s `cd $(OPENSSLDIR); pwd`/include incl/openssl
You will have to add "incl/openssl" to the dependencies
of those C files that include some OpenSSL header file.
- Add "-Iincl" to your CFLAGS.
With these additions, the OpenSSL header files will be available
under both name variants if an old library version is used:
Your application can reach them under names like <openssl/foo.h>,
while the header files still are able to #include each other
with names of the form <foo.h>.
Note on multi-threading
-----------------------
For some systems, the OpenSSL Configure script knows what compiler options
are needed to generate a library that is suitable for multi-threaded
applications. On these systems, support for multi-threading is enabled
by default; use the "no-threads" option to disable (this should never be
necessary).
On other systems, to enable support for multi-threading, you will have
to specify at least two options: "threads", and a system-dependent option.
(The latter is "-D_REENTRANT" on various systems.) The default in this
case, obviously, is not to include support for multi-threading (but
you can still use "no-threads" to suppress an annoying warning message
from the Configure script.)
OpenSSL provides built-in support for two threading models: pthreads (found on
most UNIX/Linux systems), and Windows threads. No other threading models are
supported. If your platform does not provide pthreads or Windows threads then
you should Configure with the "no-threads" option.
Note on shared libraries
------------------------
Shared libraries have certain caveats. Binary backward compatibility
can't be guaranteed before OpenSSL version 1.0. The only reason to
use them would be to conserve memory on systems where several programs
are using OpenSSL.
For some systems, the OpenSSL Configure script knows what is needed to
build shared libraries for libcrypto and libssl. On these systems,
the shared libraries are currently not created by default, but giving
the option "shared" will get them created. This method supports Makefile
targets for shared library creation, like linux-shared. Those targets
can currently be used on their own just as well, but this is expected
to change in future versions of OpenSSL.
Note on random number generation
--------------------------------
Availability of cryptographically secure random numbers is required for
secret key generation. OpenSSL provides several options to seed the
internal PRNG. If not properly seeded, the internal PRNG will refuse
to deliver random bytes and a "PRNG not seeded error" will occur.
On systems without /dev/urandom (or similar) device, it may be necessary
to install additional support software to obtain random seed.
Please check out the manual pages for RAND_add(), RAND_bytes(), RAND_egd(),
and the FAQ for more information.
Note on support for multiple builds
-----------------------------------
OpenSSL is usually built in its source tree. Unfortunately, this doesn't
support building for multiple platforms from the same source tree very well.
It is however possible to build in a separate tree through the use of lots
of symbolic links, which should be prepared like this:
mkdir -p objtree/"`uname -s`-`uname -r`-`uname -m`"
cd objtree/"`uname -s`-`uname -r`-`uname -m`"
(cd $OPENSSL_SOURCE; find . -type f) | while read F; do
mkdir -p `dirname $F`
rm -f $F; ln -s $OPENSSL_SOURCE/$F $F
echo $F '->' $OPENSSL_SOURCE/$F
done
make -f Makefile.in clean
OPENSSL_SOURCE is an environment variable that contains the absolute (this
is important!) path to the OpenSSL source tree.
Also, operations like 'make update' should still be made in the source tree.