Since QEMU 7.1 we don't support Ubuntu 18.04 anymore, so the last big
important Linux distro that did not have a pre-packaged libslirp has
been dismissed. All other major distros seem to have a libslirp package
in their distribution already - according to repology.org:

          Fedora 35: 4.6.1
  CentOS 8 (RHEL-8): 4.4.0
          Debian 11: 4.4.0
 OpenSUSE Leap 15.3: 4.3.1
   Ubuntu LTS 20.04: 4.1.0
      FreeBSD Ports: 4.7.0
      NetBSD pkgsrc: 4.7.0
           Homebrew: 4.7.0
        MSYS2 mingw: 4.7.0

The only one that was still missing a libslirp package is OpenBSD - but
the next version (OpenBSD 7.2 which will be shipped in October) is going
to include a libslirp package. Since QEMU 7.2 will be published after
OpenBSD 7.2, we should be fine there, too.

So there is no real urgent need for keeping the slirp submodule in
the QEMU tree anymore. Thus let's drop the slirp submodule now and
rely on the libslirp packages from the distributions instead.

Message-Id: <20220824151122.704946-7-thuth@redhat.com>
Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>

A Linux headers update to v6.0-rc switches some definitions from GNU
'zero-length-array' extension to the C-standard-defined flexible array
member. e.g.

struct kvm_msrs {
        __u32 nmsrs; /* number of msrs in entries */
        __u32 pad;

-       struct kvm_msr_entry entries[0];
+       struct kvm_msr_entry entries[];
};

Those (unlike the GNU zero-length-array) have some extra restrictions like
'this must be put at the end of a struct', which clang build would complain
about. e.g. the current code

struct {
        struct kvm_msrs info;
        struct kvm_msr_entry entries[1];
} msr_data = { }

generates the warning like:

target/i386/kvm/kvm.c:2868:25: error: field 'info' with variable sized
type 'struct kvm_msrs' not at the end of a struct or class is a GNU
extension [-Werror,-Wgnu-variable-sized-type-not-at-end]
        struct kvm_msrs info;
                        ^
In fact, the variable length 'entries[]' field in 'info' is zero-sized in
GNU defined semantics, which can give predictable offset for 'entries[1]'
in local msr_data. The local defined struct is just there to force a stack
allocation large enough for 1 kvm_msr_entry, a clever trick but requires to
turn off this clang warning.

Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Chenyi Qiang <chenyi.qiang@intel.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20220915091035.3897-2-chenyi.qiang@intel.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

We use the non-POSIX 'local' keyword in just two places in configure;
rewrite to avoid it.

In do_compiler(), just drop the 'local' keyword.  The variable
'compiler' is only used elsewhere in the do_compiler_werror()
function, which already uses the variable as a normal non-local one.

In probe_target_compiler(), $try and $t are both local; make them
normal variables and use a more obviously distinct variable name
for $t.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-8-peter.maydell@linaro.org

Shellcheck warns that we have one place where we run a command and
then check if it failed using $?; this is better written to simply
check the command in the 'if' statement directly.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-7-peter.maydell@linaro.org

There's only one place in configure where we use `...` to execute a
command and capture the result.  Switch to $() to match the rest of
the script. This silences a shellcheck warning.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-6-peter.maydell@linaro.org

Shellcheck warns that in
 rm -f */config-devices.mak.d
the glob might expand to something with a '-' in it, which would
then be misinterpreted as an option to rm. Fix this by adding './'.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-5-peter.maydell@linaro.org

This commit adds quotes in some places which:
 * are spotted by shellcheck
 * are obviously incorrect
 * are easy to fix just by adding the quotes

It doesn't attempt fix all of the places shellcheck finds errors,
or even all the ones which are easy to fix. It's just a random
sampling which is hopefully easy to review and which cuts
down the size of the problem for next time somebody wants to
try to look at shellcheck errors.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-4-peter.maydell@linaro.org

The meson_args variable was added in commit 3b4da132, but
was not used in that commit and isn't used today.  Delete the
unnecessary assignment.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-3-peter.maydell@linaro.org

Shellcheck correctly reports that we set python_version and never use
it.  This is a leftover from commit f9332757: we used to use
python_version purely to as part of the summary information printed
at the end of a configure run, and that commit changed to printing
the information from meson (which looks up the python version
itself). Remove the unused variable.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220825150703.4074125-2-peter.maydell@linaro.org

This is working around current limitation of Meson's handling of
--disable-pie.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220914155950.804707-19-alex.bennee@linaro.org>

Instead of using feature_not_found(), which is not a good match because
there is no "remedy" to fix the lack of makecontext(), just print a
custom error.

This happens to remove the last use of feature_not_found(), so remove
the definition and the documentation.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

probe_target_compiler returns nonempty $target_cc for installed toolchains
and $container_cross_cc for container-based toolchains.  In both cases
however the flags (coming from $cross_cc_cflags_${target_arch}) must be
in $target_cflags.

Therefore, do not clear them prior to returning from probe_target_compiler.

Reported-by: Taylor Simpson <tsimpson@quicinc.com>
Fixes: 92e288fc ("build: try both native and cross compilers", 2022-07-08)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

When moving this code out of probe_target_compiler(), we failed to adjust
the variable in which the target is located, resulting in e.g.
powerpc64-linux-user-linux-gnu-gcc-10

Fixes: cd362def ("tests/tcg: merge configure.sh back into main configure script")
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220728183901.1290113-1-richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

The '==' operator to test is a bashism; the standard way to copmare
strings is '='. This causes dash to complain:

../../configure: 681: test: linux: unexpected operator

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-id: 20220720152631.450903-6-peter.maydell@linaro.org

In commit 823eb013 we moved the setting of ARCH from configure
to meson.build, but we accidentally left behind one attempt to use
$ARCH in configure, which was trying to add -msmall-data to the
compiler flags on Alpha hosts.  Since ARCH is now never set, the test
always fails and we never add the flag.

There isn't actually any need to use this compiler flag on Alpha:
the original intent was that it would allow us to simplify our TCG
codegen on that platform, but we never actually made the TCG changes
that would rely on -msmall-data.

Drop the effectively-dead code from configure, as we don't need it.

This was spotted by shellcheck:

In ./configure line 2254:
case "$ARCH" in
      ^---^ SC2153: Possible misspelling: ARCH may not be assigned, but arch is.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-id: 20220720152631.450903-5-peter.maydell@linaro.org

The variable string-replacement syntax ${var/old/new} is a bashism
(though it is also supported by some other shells), and for instance
does not work with the NetBSD /bin/sh, which complains:
 ../src/configure: 687: Syntax error: Bad substitution

Replace it with a more portable sed-based approach, similar to
what we already do in quote_sh().

Note that shellcheck also diagnoses this:

In ./configure line 687:
    e=${e/'\'/'\\'}
      ^-----------^ SC2039: In POSIX sh, string replacement is undefined.
           ^-- SC1003: Want to escape a single quote? echo 'This is how it'\''s done'.
                ^-- SC1003: Want to escape a single quote? echo 'This is how it'\''s done'.

In ./configure line 688:
    e=${e/\"/'\"'}
      ^----------^ SC2039: In POSIX sh, string replacement is undefined.

Fixes: 8154f5e6 ("meson: Prefix each element of firmware path")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-id: 20220720152631.450903-4-peter.maydell@linaro.org

In shell script syntax, $var[something] is not special for variable
expansion: $var is expanded.  However, as it can look as if it were
intended to be an array element access (the correct syntax for which
is ${var[something]}), shellcheck recommends using explicit braces
around ${var} to clarify the intended expansion.

This fixes the warning:

In ./configure line 2346:
        if "$target_ld" -verbose 2>&1 | grep -q "^[[:space:]]*$emu[[:space:]]*$"; then
                                                              ^-- SC1087: Use braces when expanding arrays, e.g. ${array[idx]} (or ${var}[.. to quiet).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-id: 20220720152631.450903-3-peter.maydell@linaro.org

In commit 7d7dbf9d we added a line to the configure script
which is not valid POSIX shell syntax, because it is missing a space
after a '!' character. shellcheck diagnoses this:

if !(GIT="$git" "$source_path/scripts/git-submodule.sh" "$git_submodules_action" "$git_submodules"); then
    ^-- SC1035: You are missing a required space after the !.

and the OpenBSD shell will not correctly handle this without the space.

Fixes: 7d7dbf9d ("configure: replace --enable/disable-git-update with --with-git-submodules")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-id: 20220720152631.450903-2-peter.maydell@linaro.org

Use the pre-packaged toolchain provided by Loongson via github.

Tested-by: Song Gao <gaosong@loongson.cn>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220704070824.965429-1-richard.henderson@linaro.org>

Signed-off-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Message-Id: <20220624154042.51512-1-akihiko.odaki@gmail.com>
[Rewrite shell function without using Bash extensions. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

If we don't need to emulate any target, we certainly don't need TCG.

This should also help to compile again with
 ".../configure --enable-tools --disable-system --disable-user"
on systems that do not have a TCG backend.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
[thuth: Re-arranged the code, remove check-softfloat from buildtest.yml]
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20220706153816.768143-1-thuth@redhat.com>

Configure is trying to fall back on cross compilers for targets that
can have bi-arch or bi-endian toolchains, but there are many corner
cases where just checking the name can go wrong.  For example, the RHEL
ppc64le compiler is bi-arch and bi-endian, but multilibs are disabled.
Therefore it cannot be used to build 32-bit hosted binaries like the
linux-user TCG tests.

Trying the cross compiler first also does not work, and an example for
this is also ppc64le.  The powerpc64-linux-gnu-gcc binary from the
cross-gcc package is theoretically multilib-friendly, but it cannot
find the CRT files on a ppc64le host, because they are not in the .../le
multilib subdirectory.

This can be fixed by testing both the native compiler and the cross
compiler, and proceeding with the first one that works.  To do this,
move the compiler usability check from the tests/tcg snippet to inside
probe_target_compiler and, while at it, restrict the softmmu emulation
target to basically a test for the presence of libgcc.

Tested-by: Matheus Kowalczuk Ferst <matheus.ferst@eldorado.org.br>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Let probe_target_compiler know if it is looking for a compiler for a
softmmu (freestanding) or a linux-user (hosted) environment.  The
detection for the compiler has to be done differently in the two
cases.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Add more pairs of bi-arch compilers, so that it is not necessary to have
e.g. both little-endian and big-endian ARM compilers.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Use the flags passed to the configure script for the ppc cross compiler,
which in fact default to those that are needed to get the 32-bit ISA.
Add the endianness flag so that it remains possible to use a ppc64le
compiler to compile VOF.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

QEMU_CFLAGS is not available in pc-bios/s390-ccw/netboot.mak, but the Makefile
needs to access the flags passed to the configure script for the s390x
cross compiler.  Fix everything and rename QEMU_CFLAGS to EXTRA_CFLAGS for
consistency with tests/tcg.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

The optionrom build is disregarding the flags passed to the configure
script via --cross-cflags-i386.  Pass it down and add it to the Makefile.

This will make it possible to get the -m32 flag from $target_cflags to
force a 32-bit build on 64-bit hosts, instead of supplying manually the
arcane -Wa,-32 and linker emulation options.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

add the libvfio-user library as a submodule. build it as a meson
subproject.

libvfio-user is distributed with BSD 3-Clause license and
json-c with MIT (Expat) license

Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Signed-off-by: John G Johnson <john.g.johnson@oracle.com>
Signed-off-by: Jagannathan Raman <jag.raman@oracle.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: c2adec87958b081d1dc8775d4aa05c897912f025.1655151679.git.jag.raman@oracle.com

[Changed submodule URL to QEMU's libvfio-user mirror on GitLab. The QEMU
project mirrors its dependencies so that it can provide full source code
even in the event that its dependencies become unavailable. Note that
the mirror repo is manually updated, so please contact me to make newer
libvfio-user commits available. If I become a bottleneck we can set up a
cronjob.

Updated scripts/meson-buildoptions.sh to match the meson_options.txt
change. Failure to do so can result in scripts/meson-buildoptions.sh
being modified by the build system later on and you end up with a dirty
working tree.
--Stefan]

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Include the full path in TARGET_DIR, so that messages from sub-Makefiles
are clearer.  Also, prepare for possibly building firmware outside
pc-bios/ from the Makefile,

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Place it only inside the 'if test "$pie" = "no"' conditional.  Since
commit 43924d1e ("pc-bios/optionrom: detect -fno-pie", 2022-05-12),
the PIE options are detected independently by pc-bios/optionrom/Makefile,
and the CFLAGS_NOPIE/LDFLAGS_NOPIE variables are not used anymore.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

INSTALL and LIBTOOL are not used anymore, but OBJCFLAGS is new and
was not listed.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

The move from tests/tcg/configure.sh started enabling the container image
for xtensa-linux-user, which fails because the compiler does not have
the full set of headers.  The cause is the "xtensa*-softmmu)" case
in tests/tcg/configure.sh which became just "xtensa*)" in the new
probe_target_compiler shell function.  Look out for xtensa*-linux-user
and do not configure it.

Reported-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220608135727.1341946-1-pbonzini@redhat.com>
Fixes: cd362def ("tests/tcg: merge configure.sh back into main configure script")
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220613171258.1905715-6-alex.bennee@linaro.org>

Signed-off-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: Xiaojuan Yang <yangxiaojuan@loongson.cn>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Acked-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220606124333.2060567-44-yangxiaojuan@loongson.cn>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

The only compiler variable that is still needed is $(CC), for
contrib/plugins/Makefile.  All firmware builds have their own
config-host.mak file.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220517092616.1272238-17-pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220527153603.887929-28-alex.bennee@linaro.org>

While container-based cross compilers are not supported, this already
makes it possible to build vof on any machine that has an installation
of GCC and binutils for 32- or 64-bit PowerPC.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220517092616.1272238-16-pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220527153603.887929-27-alex.bennee@linaro.org>

While container-based cross compilers are not supported, this already makes
it possible to build x86 optionroms on any machine that has an installation
of GCC and binutils for 32- or 64-bit x86.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220517092616.1272238-15-pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220527153603.887929-26-alex.bennee@linaro.org>

While container-based cross compilers are not supported, this already makes
it possible to build s390-ccw on any machine that has s390x GCC and binutils
installed.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220517092616.1272238-14-pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Acked-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20220527153603.887929-25-alex.bennee@linaro.org>

Ensure that the pc-bios/optionrom and pc-bios/s390-ccw directory
exist at the time when we'll write out the compiler configuration
for them.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20220517092616.1272238-13-pbonzini@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20220527153603.887929-24-alex.bennee@linaro.org>