configure.ac is not detecting -Wextra in compilers that implement the
option. The problem is that -Wextra implies -Wunused-parameter, and the
C excerpt used by aclocal.m4 does not use argv. Patch from pedro at
ambientworks.net, ok djm@
This is a frankenstein monster of AMD64 instructions/calling conventions
but with a 4GB address space. Allegedly deprecated but people still run
into it causing weird sandbox failures, e.g. bz#3085
When the test's child signals its parent and it exits the result of
getppid changes. On Ubuntu 20.04 this results in the ppid being that
of the GDM session, causing it to exit. Analysis and testing from pedro
at ambientworks.net
If we don't have LLONG_{MIN,MAX} but do have LONG_LONG_{MIN,MAX}
then use those instead. We do calculate these values in configure,
but it turns out that at least one compiler (old HP ANSI C) can't
parse "-9223372036854775808LL" without mangling it. (It can parse
"-9223372036854775807LL" which is presumably why its limits.h defines
LONG_LONG_MIN as the latter minus 1.)
Fixes rekey test when compiled with the aforementioned compiler.
On some platforms (at least older HP-UXes such as 11.11, possibly others)
setting SA_RESTART on signal handers will cause it to not interrupt
select(), at least for calls that do not specify a timeout. Try to
detect this and if found, don't use SA_RESTART.
POSIX says "If SA_RESTART has been set for the interrupting signal, it
is implementation-dependent whether select() restarts or returns with
[EINTR]" so this behaviour is within spec.
... unless we are actually going to use it. Fixes build on HP-UX
without the potential impact to other platforms of a header change
shortly before release.
If the system (or one of the dependencies) implements memmem but does
not define the header, we would not declare it either resulting in
compiler warnings. Check for declaration explicitly. bz#3102.
linking against the (previously external) USB HID middleware. The dlopen()
capability still exists for alternate middlewares, e.g. for Bluetooth, NFC
and test/debugging.
OpenBSD-Commit-ID: 14446cf170ac0351f0d4792ba0bca53024930069
Including a function call in the test programs for the gcc stack
protector flag tests exercises more of the compiler and makes it more
likely it'll detect problems.
Enable -Wextra if compiler supports it
Set -Wno-error=format-truncation if available to prevent expected
string truncations in openbsd-compat from breaking -Werror builds
Rather than attempt to apply 14 years' worth of changes to OpenBSD's sha2
I imported the current versions directly then re-applied the portability
changes. This also allowed re-syncing digest-libc.c against upstream.
We shipped a BSD implementation of realpath() because sftp-server
depended on its behaviour.
OpenBSD is now moving to a more strictly POSIX-compliant realpath(2),
so sftp-server now unconditionally requires its own BSD-style realpath
implementation. As such, there is no need to carry another independant
implementation in openbsd-compat.
ok dtucker@
Previously configure would not select the "doc" man page format if
mandoc was present but nroff was not. This checks for mandoc first
and removes a now-superflous AC_PATH_PROG. Based on a patch from
vehk at vehk.de and feedback from schwarze at usta.de.
Don't call OpenSSL_add_all_algorithms() unless OpenSSL actually
supports it.
Move all libcrypto initialisation to a single function, and call that
from seed_rng() that is called early in each tool's main().
Prompted by patch from Rosen Penev
If configure could not find a working OpenSSL installation it would
fall back to checking in /usr/local/ssl. This made sense back when
systems did not ship with OpenSSL, but most do and OpenSSL 1.1 doesn't
use that as a default any more. The fallback behaviour also meant
that if you pointed --with-ssl-dir at a specific directory and it
didn't work, it would silently use either the system libs or the ones
in /usr/local/ssl. If you want to use /usr/local/ssl you'll need to
pass configure --with-ssl-dir=/usr/local/ssl. ok djm@
Check for the existence of openssl version functions and use the ones
detected instead of trying to guess based on the int32 version
identifier. Fixes builds with LibreSSL.
Current impementions of the gcc spectre mitigation flags cause
miscompilations when combined with other flags and do not provide much
protection. Found by fweimer at redhat.com, ok djm@
Cygwin's latest 7.x GCC allows to specify -mfunction-return=thunk
as well as -mindirect-branch=thunk on the command line, albeit
producing invalid code, leading to an error at link stage.
The check in configure.ac only checks if the option is present,
but not if it produces valid code.
This patch fixes it by special-casing Cygwin. Another solution
may be to change these to linker checks.
Signed-off-by: Corinna Vinschen <vinschen@redhat.com>
Most people will 1) be using modern multi-factor authentication methods
like TOTP/OATH etc and 2) be getting support for multi-factor
authentication via PAM or BSD Auth.
Add getline for the benefit of platforms that don't have it. Sourced
from NetBSD (OpenBSD's implementation is a little too chummy with the
internals of FILE).
Since autoconf always uses $CC to link C programs, allowing users to
override LD caused mismatches between what LD_LINK_IFELSE thought worked
and what ld thought worked. If you do need to do this kind of thing you
need to set a compiler flag such as gcc's -fuse-ld in LDFLAGS.
On at least some revisions of AIX, strndup returns unterminated strings
under some conditions, apparently because strnlen returns incorrect
values in those cases. Disable both on AIX and use the replacements
from openbsd-compat. Fixes problem with ECDSA keys there, ok djm.
Currently seccomp_audit_arch is set to AUDIT_ARCH_MIPS64 or
AUDIT_ARCH_MIPSEL64 (depending on the endinness) when openssh is built
for MIPS64. However, that's only valid for n64 ABI. The right macros for
n32 ABI defined in seccomp.h are AUDIT_ARCH_MIPS64N32 and
AUDIT_ARCH_MIPSEL64N32, for big and little endian respectively.
Because of that an sshd built for MIPS64 n32 rejects connection attempts
and the output of strace reveals that the problem is related to seccomp
audit:
[pid 194] prctl(PR_SET_SECCOMP, SECCOMP_MODE_FILTER, {len=57,
filter=0x555d5da0}) = 0
[pid 194] write(7, "\0\0\0]\0\0\0\5\0\0\0Ulist_hostkey_types: "..., 97) = ?
[pid 193] <... poll resumed> ) = 2 ([{fd=5, revents=POLLIN|POLLHUP},
{fd=6, revents=POLLHUP}])
[pid 194] +++ killed by SIGSYS +++
This patch fixes that problem by setting the right value to
seccomp_audit_arch taking into account the MIPS64 ABI.
Signed-off-by: Vicente Olivert Riera <Vincent.Riera@imgtec.com>
Some compilers (gcc 2.9.53, 3.0 and probably others, see gcc bug #3481)
do not accept __attribute__ on function pointer prototype args. Check for
this and hide them if they're not accepted.
Configure assumes that if malloc(0) returns null then calloc(0,n)
also does. On some old platforms (SunOS4) malloc behaves as expected
(as determined by AC_FUNC_MALLOC) but calloc doesn't. Test for this
at configure time and activate the replacement function if found, plus
handle this case in rpl_calloc.
BindInterface required getifaddr and friends so disable if not available
(eg Solaris 10). We should be able to add support for some systems with
a bit more work but this gets the building again.
The code required to support it is quite invasive to the mainline
code that is synced with upstream and is an ongoing maintenance burden.
Both the hardware and software are literal museum pieces these days and
we could not find anyone still running OpenSSH on one.
This adds checks for gcc and clang flags for mitigations for Spectre
variant 2, ie "retpoline". It'll automatically enabled if the compiler
supports it as part of toolchain hardening flag. ok djm@
Adds a .depend file containing dependency information generated by
makedepend, which is appended to the generated Makefile by configure.
You can regen the file with "make -f Makefile.in depend" if necessary,
but we'll be looking at some way to automatically keep this up to date.
"no objection" djm@
Check for MIKDIR_P and use it instead of mkinstalldirs. Should fix "mkdir:
cannot create directory:... File exists" during "make install".
Patch from eb at emlix.com.
We don't support openssl-1.1.x yet (see multiple threads on the
openssh-unix-dev@ mailing list for the reason), but previously
./configure would accept it and the compilation would subsequently
fail. This makes ./configure display an explicit error message and
abort.
ok dtucker@
On some platforms (AIX, maybe others) allocating zero bytes of memory
via the various *alloc functions returns NULL, which is permitted
by the standards. Autoconf has some macros for detecting this (with
the exception of calloc for some reason) so use these and if necessary
activate shims for them. ok djm@
It's still on by default, but now it's possible to turn it off using
--without-hardening. This is useful since it's known to cause problems
with some -fsanitize options. ok dtucker@
Some CFLAGS/LDFLAGS may disrupt the configure script's operation,
in particular santization and fuzzer options that break assumptions
about memory and file descriptor dispositions.
This adds two flags to configure --with-cflags-after and
--with-ldflags-after that allow specifying additional compiler and
linker options that are added to the resultant Makefiles but not
used in the configure run itself.
E.g.
env CC=clang-3.9 ./configure \
--with-cflags-after=-fsantize=address \
--with-ldflags-after="-g -fsanitize=address"
FreeBSD's <sys/capability.h> was renamed to <sys/capsicum.h> in 2014 to
avoid future conflicts with POSIX capabilities (the last release that
didn't have it was 9.3) so switch to that. Patch from des at des.no.
Include replacement timespeccmp() for systems that lack it.
Support time_t struct stat->st_mtime in addition to
timespec stat->st_mtim, as well as unsorted fallback.
Should fix bz#2603 - "Build with ldns and without kerberos support
fails if ldns compiled with kerberos support" by including correct
cflags/libs
ok dtucker@
Having _XOPEN_SOURCE unconditionally causes problems on some platforms
and configurations, notably Solaris 64-bit binaries. It was there for
the benefit of Linux put the required bits in the *-*linux* section.
Patch from yvoinov at gmail.com.
pedro martelletto
Damien Miller
Darren Tucker
djm@openbsd.org
Andreas Schwab
Ruben Kerkhof
Dag-Erling Smørgrav
Corinna Vinschen
Tim Rice
Vicente Olivert Riera
Darren Tucker
Mike Frysinger