qemu/meson.build

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

4619 lines
165 KiB
Meson
Raw Normal View History

meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
project('qemu', ['c'], meson_version: '>=1.1.0',
default_options: ['warning_level=1', 'c_std=gnu11', 'cpp_std=gnu++11', 'b_colorout=auto',
'b_staticpic=false', 'stdsplit=false', 'optimization=2', 'b_pie=true'],
version: files('VERSION'))
add_test_setup('quick', exclude_suites: ['slow', 'thorough'], is_default: true)
add_test_setup('slow', exclude_suites: ['thorough'], env: ['G_TEST_SLOW=1', 'SPEED=slow'])
add_test_setup('thorough', env: ['G_TEST_SLOW=1', 'SPEED=thorough'])
meson.add_postconf_script(find_program('scripts/symlink-install-tree.py'))
####################
# Global variables #
####################
not_found = dependency('', required: false)
keyval = import('keyval')
ss = import('sourceset')
fs = import('fs')
host_os = host_machine.system()
config_host = keyval.load(meson.current_build_dir() / 'config-host.mak')
# Temporary directory used for files created while
# configure runs. Since it is in the build directory
# we can safely blow away any previous version of it
# (and we need not jump through hoops to try to delete
# it when configure exits.)
tmpdir = meson.current_build_dir() / 'meson-private/temp'
if get_option('qemu_suffix').startswith('/')
error('qemu_suffix cannot start with a /')
endif
qemu_confdir = get_option('sysconfdir') / get_option('qemu_suffix')
qemu_datadir = get_option('datadir') / get_option('qemu_suffix')
qemu_docdir = get_option('docdir') / get_option('qemu_suffix')
qemu_moddir = get_option('libdir') / get_option('qemu_suffix')
qemu_desktopdir = get_option('datadir') / 'applications'
qemu_icondir = get_option('datadir') / 'icons'
genh = []
qapi_trace_events = []
bsd_oses = ['gnu/kfreebsd', 'freebsd', 'netbsd', 'openbsd', 'dragonfly', 'darwin']
supported_oses = ['windows', 'freebsd', 'netbsd', 'openbsd', 'darwin', 'sunos', 'linux']
supported_cpus = ['ppc', 'ppc64', 's390x', 'riscv32', 'riscv64', 'x86', 'x86_64',
'arm', 'aarch64', 'loongarch64', 'mips', 'mips64', 'sparc64']
cpu = host_machine.cpu_family()
target_dirs = config_host['TARGET_DIRS'].split()
############
# Programs #
############
sh = find_program('sh')
python = import('python').find_installation()
cc = meson.get_compiler('c')
all_languages = ['c']
if host_os == 'windows' and add_languages('cpp', required: false, native: false)
all_languages += ['cpp']
cxx = meson.get_compiler('cpp')
endif
if host_os == 'darwin' and \
add_languages('objc', required: true, native: false)
all_languages += ['objc']
objc = meson.get_compiler('objc')
endif
dtrace = not_found
stap = not_found
if 'dtrace' in get_option('trace_backends')
dtrace = find_program('dtrace', required: true)
stap = find_program('stap', required: false)
if stap.found()
# Workaround to avoid dtrace(1) producing a file with 'hidden' symbol
# visibility. Define STAP_SDT_V2 to produce 'default' symbol visibility
# instead. QEMU --enable-modules depends on this because the SystemTap
# semaphores are linked into the main binary and not the module's shared
# object.
add_global_arguments('-DSTAP_SDT_V2',
native: false, language: all_languages)
endif
endif
if get_option('iasl') == ''
iasl = find_program('iasl', required: false)
else
iasl = find_program(get_option('iasl'), required: true)
endif
edk2_targets = [ 'arm-softmmu', 'aarch64-softmmu', 'i386-softmmu', 'x86_64-softmmu', 'riscv64-softmmu', 'loongarch64-softmmu' ]
unpack_edk2_blobs = false
foreach target : edk2_targets
if target in target_dirs
bzip2 = find_program('bzip2', required: get_option('install_blobs'))
unpack_edk2_blobs = bzip2.found()
break
endif
endforeach
#####################
# Option validation #
#####################
# Fuzzing
if get_option('fuzzing') and get_option('fuzzing_engine') == '' and \
not cc.links('''
#include <stdint.h>
#include <sys/types.h>
int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size);
int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) { return 0; }
''',
args: ['-Werror', '-fsanitize=fuzzer'])
error('Your compiler does not support -fsanitize=fuzzer')
endif
# Tracing backends
if 'ftrace' in get_option('trace_backends') and host_os != 'linux'
error('ftrace is supported only on Linux')
endif
if 'syslog' in get_option('trace_backends') and not cc.compiles('''
#include <syslog.h>
int main(void) {
openlog("qemu", LOG_PID, LOG_DAEMON);
syslog(LOG_INFO, "configure");
return 0;
}''')
error('syslog is not supported on this system')
endif
# Miscellaneous Linux-only features
get_option('mpath') \
.require(host_os == 'linux', error_message: 'Multipath is supported only on Linux')
multiprocess_allowed = get_option('multiprocess') \
.require(host_os == 'linux', error_message: 'Multiprocess QEMU is supported only on Linux') \
.allowed()
vfio_user_server_allowed = get_option('vfio_user_server') \
.require(host_os == 'linux', error_message: 'vfio-user server is supported only on Linux') \
.allowed()
have_tpm = get_option('tpm') \
.require(host_os != 'windows', error_message: 'TPM emulation only available on POSIX systems') \
.allowed()
# vhost
have_vhost_user = get_option('vhost_user') \
.disable_auto_if(host_os != 'linux') \
.require(host_os != 'windows',
error_message: 'vhost-user is not available on Windows').allowed()
have_vhost_vdpa = get_option('vhost_vdpa') \
.require(host_os == 'linux',
error_message: 'vhost-vdpa is only available on Linux').allowed()
have_vhost_kernel = get_option('vhost_kernel') \
.require(host_os == 'linux',
error_message: 'vhost-kernel is only available on Linux').allowed()
have_vhost_user_crypto = get_option('vhost_crypto') \
.require(have_vhost_user,
error_message: 'vhost-crypto requires vhost-user to be enabled').allowed()
have_vhost = have_vhost_user or have_vhost_vdpa or have_vhost_kernel
have_vhost_net_user = have_vhost_user and get_option('vhost_net').allowed()
have_vhost_net_vdpa = have_vhost_vdpa and get_option('vhost_net').allowed()
have_vhost_net_kernel = have_vhost_kernel and get_option('vhost_net').allowed()
have_vhost_net = have_vhost_net_kernel or have_vhost_net_user or have_vhost_net_vdpa
# type of binaries to build
have_linux_user = false
have_bsd_user = false
have_system = false
foreach target : target_dirs
have_linux_user = have_linux_user or target.endswith('linux-user')
have_bsd_user = have_bsd_user or target.endswith('bsd-user')
have_system = have_system or target.endswith('-softmmu')
endforeach
have_user = have_linux_user or have_bsd_user
have_tools = get_option('tools') \
.disable_auto_if(not have_system) \
.allowed()
have_ga = get_option('guest_agent') \
.disable_auto_if(not have_system and not have_tools) \
.require(host_os in ['sunos', 'linux', 'windows', 'freebsd', 'netbsd', 'openbsd'],
error_message: 'unsupported OS for QEMU guest agent') \
.allowed()
have_block = have_system or have_tools
enable_modules = get_option('modules') \
.require(host_os != 'windows',
error_message: 'Modules are not available for Windows') \
.require(not get_option('prefer_static'),
error_message: 'Modules are incompatible with static linking') \
.allowed()
#######################################
# Variables for host and accelerators #
#######################################
if cpu not in supported_cpus
host_arch = 'unknown'
elif cpu == 'x86'
host_arch = 'i386'
elif cpu == 'mips64'
host_arch = 'mips'
elif cpu in ['riscv32', 'riscv64']
host_arch = 'riscv'
else
host_arch = cpu
endif
if cpu in ['x86', 'x86_64']
kvm_targets = ['i386-softmmu', 'x86_64-softmmu']
elif cpu == 'aarch64'
kvm_targets = ['aarch64-softmmu']
elif cpu == 's390x'
kvm_targets = ['s390x-softmmu']
elif cpu in ['ppc', 'ppc64']
kvm_targets = ['ppc-softmmu', 'ppc64-softmmu']
elif cpu in ['mips', 'mips64']
kvm_targets = ['mips-softmmu', 'mipsel-softmmu', 'mips64-softmmu', 'mips64el-softmmu']
elif cpu in ['riscv32']
kvm_targets = ['riscv32-softmmu']
elif cpu in ['riscv64']
kvm_targets = ['riscv64-softmmu']
elif cpu in ['loongarch64']
kvm_targets = ['loongarch64-softmmu']
else
kvm_targets = []
endif
accelerator_targets = { 'CONFIG_KVM': kvm_targets }
if cpu in ['x86', 'x86_64']
xen_targets = ['i386-softmmu', 'x86_64-softmmu']
elif cpu in ['arm', 'aarch64']
# i386 emulator provides xenpv machine type for multiple architectures
xen_targets = ['i386-softmmu', 'x86_64-softmmu', 'aarch64-softmmu']
else
xen_targets = []
endif
accelerator_targets += { 'CONFIG_XEN': xen_targets }
if cpu in ['aarch64']
accelerator_targets += {
'CONFIG_HVF': ['aarch64-softmmu']
}
endif
if cpu in ['x86', 'x86_64']
accelerator_targets += {
'CONFIG_HVF': ['x86_64-softmmu'],
'CONFIG_NVMM': ['i386-softmmu', 'x86_64-softmmu'],
'CONFIG_WHPX': ['i386-softmmu', 'x86_64-softmmu'],
}
endif
modular_tcg = []
# Darwin does not support references to thread-local variables in modules
if host_os != 'darwin'
modular_tcg = ['i386-softmmu', 'x86_64-softmmu']
endif
##################
# Compiler flags #
##################
foreach lang : all_languages
compiler = meson.get_compiler(lang)
if compiler.get_id() == 'gcc' and compiler.version().version_compare('>=7.4')
# ok
elif compiler.get_id() == 'clang' and compiler.compiles('''
#ifdef __apple_build_version__
# if __clang_major__ < 12 || (__clang_major__ == 12 && __clang_minor__ < 0)
# error You need at least XCode Clang v12.0 to compile QEMU
# endif
#else
# if __clang_major__ < 10 || (__clang_major__ == 10 && __clang_minor__ < 0)
# error You need at least Clang v10.0 to compile QEMU
# endif
#endif''')
# ok
else
error('You either need GCC v7.4 or Clang v10.0 (or XCode Clang v12.0) to compile QEMU')
endif
endforeach
# default flags for all hosts
# We use -fwrapv to tell the compiler that we require a C dialect where
# left shift of signed integers is well defined and has the expected
# 2s-complement style results. (Both clang and gcc agree that it
# provides these semantics.)
qemu_common_flags = [
'-D_GNU_SOURCE', '-D_FILE_OFFSET_BITS=64', '-D_LARGEFILE_SOURCE',
'-fno-strict-aliasing', '-fno-common', '-fwrapv' ]
qemu_cflags = []
qemu_ldflags = []
if host_os == 'darwin'
# Disable attempts to use ObjectiveC features in os/object.h since they
# won't work when we're compiling with gcc as a C compiler.
if compiler.get_id() == 'gcc'
qemu_common_flags += '-DOS_OBJECT_USE_OBJC=0'
endif
elif host_os == 'sunos'
# needed for CMSG_ macros in sys/socket.h
qemu_common_flags += '-D_XOPEN_SOURCE=600'
# needed for TIOCWIN* defines in termios.h
qemu_common_flags += '-D__EXTENSIONS__'
elif host_os == 'haiku'
qemu_common_flags += ['-DB_USE_POSITIVE_POSIX_ERRORS', '-D_BSD_SOURCE', '-fPIC']
elif host_os == 'windows'
if not compiler.compiles('struct x { int y; } __attribute__((gcc_struct));',
args: '-Werror')
error('Your compiler does not support __attribute__((gcc_struct)) - please use GCC instead of Clang')
endif
endif
# __sync_fetch_and_and requires at least -march=i486. Many toolchains
# use i686 as default anyway, but for those that don't, an explicit
# specification is necessary
if host_arch == 'i386' and not cc.links('''
static int sfaa(int *ptr)
{
return __sync_fetch_and_and(ptr, 0);
}
int main(void)
{
int val = 42;
val = __sync_val_compare_and_swap(&val, 0, 1);
sfaa(&val);
return val;
}''')
qemu_common_flags = ['-march=i486'] + qemu_common_flags
endif
# Pick x86-64 baseline version
if host_arch in ['i386', 'x86_64']
if get_option('x86_version') == '0' and host_arch == 'x86_64'
error('x86_64-v1 required for x86-64 hosts')
endif
# add flags for individual instruction set extensions
if get_option('x86_version') >= '1'
if host_arch == 'i386'
qemu_common_flags = ['-mfpmath=sse'] + qemu_common_flags
else
# present on basically all processors but technically not part of
# x86-64-v1, so only include -mneeded for x86-64 version 2 and above
qemu_common_flags = ['-mcx16'] + qemu_common_flags
endif
endif
if get_option('x86_version') >= '2'
qemu_common_flags = ['-mpopcnt'] + qemu_common_flags
qemu_common_flags = cc.get_supported_arguments('-mneeded') + qemu_common_flags
endif
if get_option('x86_version') >= '3'
qemu_common_flags = ['-mmovbe', '-mabm', '-mbmi1', '-mbmi2', '-mfma', '-mf16c'] + qemu_common_flags
endif
# add required vector instruction set (each level implies those below)
if get_option('x86_version') == '1'
qemu_common_flags = ['-msse2'] + qemu_common_flags
elif get_option('x86_version') == '2'
qemu_common_flags = ['-msse4.2'] + qemu_common_flags
elif get_option('x86_version') == '3'
qemu_common_flags = ['-mavx2'] + qemu_common_flags
elif get_option('x86_version') == '4'
qemu_common_flags = ['-mavx512f', '-mavx512bw', '-mavx512cd', '-mavx512dq', '-mavx512vl'] + qemu_common_flags
endif
endif
if get_option('prefer_static')
qemu_ldflags += get_option('b_pie') ? '-static-pie' : '-static'
endif
# Meson currently only handles pie as a boolean for now, so if the user
# has explicitly disabled PIE we need to extend our cflags.
#
# -no-pie is supposedly a linker flag that has no effect on the compiler
# command line, but some distros, that didn't quite know what they were
# doing, made local changes to gcc's specs file that turned it into
# a compiler command-line flag.
#
# What about linker flags? For a static build, no PIE is implied by -static
# which we added above (and if it's not because of the same specs patching,
# there's nothing we can do: compilation will fail, report a bug to your
# distro and do not use --disable-pie in the meanwhile). For dynamic linking,
# instead, we can't add -no-pie because it overrides -shared: the linker then
# tries to build an executable instead of a shared library and fails. So
# don't add -no-pie anywhere and cross fingers. :(
if not get_option('b_pie')
qemu_common_flags += cc.get_supported_arguments('-fno-pie', '-no-pie')
endif
if not get_option('stack_protector').disabled()
stack_protector_probe = '''
int main(int argc, char *argv[])
{
char arr[64], *p = arr, *c = argv[argc - 1];
while (*c) {
*p++ = *c++;
}
return 0;
}'''
have_stack_protector = false
foreach arg : ['-fstack-protector-strong', '-fstack-protector-all']
# We need to check both a compile and a link, since some compiler
# setups fail only on a .c->.o compile and some only at link time
if cc.compiles(stack_protector_probe, args: ['-Werror', arg]) and \
cc.links(stack_protector_probe, args: ['-Werror', arg])
have_stack_protector = true
qemu_cflags += arg
qemu_ldflags += arg
break
endif
endforeach
get_option('stack_protector') \
.require(have_stack_protector, error_message: 'Stack protector not supported')
endif
coroutine_backend = get_option('coroutine_backend')
ucontext_probe = '''
#include <ucontext.h>
#ifdef __stub_makecontext
#error Ignoring glibc stub makecontext which will always fail
#endif
int main(void) { makecontext(0, 0, 0); return 0; }'''
# On Windows the only valid backend is the Windows specific one.
# For POSIX prefer ucontext, but it's not always possible. The fallback
# is sigcontext.
supported_backends = []
if host_os == 'windows'
supported_backends += ['windows']
else
if host_os != 'darwin' and cc.links(ucontext_probe)
supported_backends += ['ucontext']
endif
supported_backends += ['sigaltstack']
endif
if coroutine_backend == 'auto'
coroutine_backend = supported_backends[0]
elif coroutine_backend not in supported_backends
error('"@0@" backend requested but not available. Available backends: @1@' \
.format(coroutine_backend, ', '.join(supported_backends)))
endif
# Compiles if SafeStack *not* enabled
safe_stack_probe = '''
int main(void)
{
#if defined(__has_feature)
#if __has_feature(safe_stack)
#error SafeStack Enabled
#endif
#endif
return 0;
}'''
if get_option('safe_stack') != not cc.compiles(safe_stack_probe)
safe_stack_arg = get_option('safe_stack') ? '-fsanitize=safe-stack' : '-fno-sanitize=safe-stack'
if get_option('safe_stack') != not cc.compiles(safe_stack_probe, args: safe_stack_arg)
error(get_option('safe_stack') \
? 'SafeStack not supported by your compiler' \
: 'Cannot disable SafeStack')
endif
qemu_cflags += safe_stack_arg
qemu_ldflags += safe_stack_arg
endif
if get_option('safe_stack') and coroutine_backend != 'ucontext'
error('SafeStack is only supported with the ucontext coroutine backend')
endif
if get_option('asan')
if cc.has_argument('-fsanitize=address')
qemu_cflags = ['-fsanitize=address'] + qemu_cflags
qemu_ldflags = ['-fsanitize=address'] + qemu_ldflags
else
error('Your compiler does not support -fsanitize=address')
endif
endif
if get_option('ubsan')
# Detect static linking issue with ubsan:
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84285
if cc.links('int main(int argc, char **argv) { return argc + 1; }',
args: [qemu_ldflags, '-fsanitize=undefined'])
qemu_cflags += ['-fsanitize=undefined']
qemu_ldflags += ['-fsanitize=undefined']
# Suppress undefined behaviour from function call to mismatched type.
# In addition, tcg prologue does not emit function type prefix
# required by function call sanitizer.
if cc.has_argument('-fno-sanitize=function')
qemu_cflags += ['-fno-sanitize=function']
endif
else
error('Your compiler does not support -fsanitize=undefined')
endif
endif
# Thread sanitizer is, for now, much noisier than the other sanitizers;
# keep it separate until that is not the case.
if get_option('tsan')
if get_option('asan') or get_option('ubsan')
error('TSAN is not supported with other sanitizers')
endif
if not cc.has_function('__tsan_create_fiber',
args: '-fsanitize=thread',
prefix: '#include <sanitizer/tsan_interface.h>')
error('Cannot enable TSAN due to missing fiber annotation interface')
endif
qemu_cflags = ['-fsanitize=thread'] + qemu_cflags
qemu_ldflags = ['-fsanitize=thread'] + qemu_ldflags
endif
# Detect support for PT_GNU_RELRO + DT_BIND_NOW.
# The combination is known as "full relro", because .got.plt is read-only too.
qemu_ldflags += cc.get_supported_link_arguments('-Wl,-z,relro', '-Wl,-z,now')
if host_os == 'windows'
qemu_ldflags += cc.get_supported_link_arguments('-Wl,--no-seh', '-Wl,--nxcompat')
qemu_ldflags += cc.get_supported_link_arguments('-Wl,--dynamicbase', '-Wl,--high-entropy-va')
endif
if get_option('fuzzing')
# Specify a filter to only instrument code that is directly related to
# virtual-devices.
configure_file(output: 'instrumentation-filter',
input: 'scripts/oss-fuzz/instrumentation-filter-template',
copy: true)
if cc.compiles('int main () { return 0; }',
name: '-fsanitize-coverage-allowlist=/dev/null',
args: ['-fsanitize-coverage-allowlist=/dev/null',
'-fsanitize-coverage=trace-pc'] )
qemu_common_flags += ['-fsanitize-coverage-allowlist=instrumentation-filter']
endif
if get_option('fuzzing_engine') == ''
# Add CFLAGS to tell clang to add fuzzer-related instrumentation to all the
# compiled code. To build non-fuzzer binaries with --enable-fuzzing, link
# everything with fsanitize=fuzzer-no-link. Otherwise, the linker will be
# unable to bind the fuzzer-related callbacks added by instrumentation.
qemu_common_flags += ['-fsanitize=fuzzer-no-link']
qemu_ldflags += ['-fsanitize=fuzzer-no-link']
# For the actual fuzzer binaries, we need to link against the libfuzzer
# library. They need to be configurable, to support OSS-Fuzz
fuzz_exe_ldflags = ['-fsanitize=fuzzer']
else
# LIB_FUZZING_ENGINE was set; assume we are running on OSS-Fuzz, and
# the needed CFLAGS have already been provided
fuzz_exe_ldflags = get_option('fuzzing_engine').split()
endif
endif
if get_option('cfi')
cfi_flags=[]
# Check for dependency on LTO
if not get_option('b_lto')
error('Selected Control-Flow Integrity but LTO is disabled')
endif
if enable_modules
error('Selected Control-Flow Integrity is not compatible with modules')
endif
# Check for cfi flags. CFI requires LTO so we can't use
# get_supported_arguments, but need a more complex "compiles" which allows
# custom arguments
if cc.compiles('int main () { return 0; }', name: '-fsanitize=cfi-icall',
args: ['-flto', '-fsanitize=cfi-icall'] )
cfi_flags += '-fsanitize=cfi-icall'
else
error('-fsanitize=cfi-icall is not supported by the compiler')
endif
if cc.compiles('int main () { return 0; }',
name: '-fsanitize-cfi-icall-generalize-pointers',
args: ['-flto', '-fsanitize=cfi-icall',
'-fsanitize-cfi-icall-generalize-pointers'] )
cfi_flags += '-fsanitize-cfi-icall-generalize-pointers'
else
error('-fsanitize-cfi-icall-generalize-pointers is not supported by the compiler')
endif
if get_option('cfi_debug')
if cc.compiles('int main () { return 0; }',
name: '-fno-sanitize-trap=cfi-icall',
args: ['-flto', '-fsanitize=cfi-icall',
'-fno-sanitize-trap=cfi-icall'] )
cfi_flags += '-fno-sanitize-trap=cfi-icall'
else
error('-fno-sanitize-trap=cfi-icall is not supported by the compiler')
endif
endif
add_global_arguments(cfi_flags, native: false, language: all_languages)
add_global_link_arguments(cfi_flags, native: false, language: all_languages)
endif
meson: mitigate against ROP exploits with -fzero-call-used-regs To quote wikipedia: "Return-oriented programming (ROP) is a computer security exploit technique that allows an attacker to execute code in the presence of security defenses such as executable space protection and code signing. In this technique, an attacker gains control of the call stack to hijack program control flow and then executes carefully chosen machine instruction sequences that are already present in the machine's memory, called "gadgets". Each gadget typically ends in a return instruction and is located in a subroutine within the existing program and/or shared library code. Chained together, these gadgets allow an attacker to perform arbitrary operations on a machine employing defenses that thwart simpler attacks." QEMU is by no means perfect with an ever growing set of CVEs from flawed hardware device emulation, which could potentially be exploited using ROP techniques. Since GCC 11 there has been a compiler option that can mitigate against this exploit technique: -fzero-call-user-regs To understand it refer to these two resources: https://www.jerkeby.se/newsletter/posts/rop-reduction-zero-call-user-regs/ https://gcc.gnu.org/pipermail/gcc-patches/2020-August/552262.html I used two programs to scan qemu-system-x86_64 for ROP gadgets: https://github.com/0vercl0k/rp https://github.com/JonathanSalwan/ROPgadget When asked to find 8 byte gadgets, the 'rp' tool reports: A total of 440278 gadgets found. You decided to keep only the unique ones, 156143 unique gadgets found. While the ROPgadget tool reports: Unique gadgets found: 353122 With the --ropchain argument, the latter attempts to use the found gadgets to product a chain that can execute arbitrary syscalls. With current QEMU it succeeds in this task, which is an undesirable situation. With QEMU modified to use -fzero-call-user-regs=used-gpr the 'rp' tool reports A total of 528991 gadgets found. You decided to keep only the unique ones, 121128 unique gadgets found. This is 22% fewer unique gadgets While the ROPgadget tool reports: Unique gadgets found: 328605 This is 7% fewer unique gadgets. Crucially though, despite this more modest reduction, the ROPgadget tool is no longer able to identify a chain of gadgets for executing arbitrary syscalls. It fails at the very first step, unable to find gadgets for populating registers for a future syscall. Having said that, more advanced tools do still manage to put together a viable ROP chain. Also this only takes into account QEMU code. QEMU links to many 3rd party shared libraries and ideally all of them would be compiled with this same hardening. That becomes a distro policy question though. In terms of performance impact, TCG was used as an evaluation test case. We're not interested in protecting TCG since it isn't designed to provide a security barrier, but it is performance sensitive code, so useful as a guide to how other areas of QEMU might be impacted. With the -fzero-call-user-regs=used-gpr argument present, using the real world test of booting a linux kernel and having init immediately poweroff, there is a ~1% slow down in performance under TCG. The QEMU binary size also grows by approximately 1%. By comparison, using the more aggressive -fzero-call-user-regs=all, results in a slowdown of over 25% in TCG, which is clearly not an acceptable impact, and a binary size increase of 5%. Considering that 'used-gpr' successfully stopped ROPgadget assembling a chain, this more targeted protection is a justifiable hardening / performance tradeoff. Reviewed-by: Thomas Huth <thuth@redhat.com> Signed-off-by: "Daniel P. Berrangé" <berrange@redhat.com> Message-ID: <20240103123414.2401208-2-berrange@redhat.com> Signed-off-by: Thomas Huth <thuth@redhat.com>
2024-01-03 15:34:13 +03:00
# Check further flags that make QEMU more robust against malicious parties
hardening_flags = [
meson: mitigate against use of uninitialize stack for exploits When variables are used without being initialized, there is potential to take advantage of data that was pre-existing on the stack from an earlier call, to drive an exploit. It is good practice to always initialize variables, and the compiler can warn about flaws when -Wuninitialized is present. This warning, however, is by no means foolproof with its output varying depending on compiler version and which optimizations are enabled. The -ftrivial-auto-var-init option can be used to tell the compiler to always initialize all variables. This increases the security and predictability of the program, closing off certain attack vectors, reducing the risk of unsafe memory disclosure. While the option takes several possible values, using 'zero' is considered to be the option that is likely to lead to semantically correct or safe behaviour[1]. eg sizes/indexes are not likely to lead to out-of-bounds accesses when initialized to zero. Pointers are less likely to point something useful if initialized to zero. Even with -ftrivial-auto-var-init=zero set, GCC will still issue warnings with -Wuninitialized if it discovers a problem, so we are not loosing diagnostics for developers, just hardening runtime behaviour and making QEMU behave more predictably in case of hitting bad codepaths. [1] https://lists.llvm.org/pipermail/cfe-dev/2020-April/065221.html Signed-off-by: "Daniel P. Berrangé" <berrange@redhat.com> Message-ID: <20240103123414.2401208-3-berrange@redhat.com> Signed-off-by: Thomas Huth <thuth@redhat.com>
2024-01-03 15:34:14 +03:00
# Initialize all stack variables to zero. This makes
# it harder to take advantage of uninitialized stack
# data to drive exploits
'-ftrivial-auto-var-init=zero',
meson: mitigate against ROP exploits with -fzero-call-used-regs To quote wikipedia: "Return-oriented programming (ROP) is a computer security exploit technique that allows an attacker to execute code in the presence of security defenses such as executable space protection and code signing. In this technique, an attacker gains control of the call stack to hijack program control flow and then executes carefully chosen machine instruction sequences that are already present in the machine's memory, called "gadgets". Each gadget typically ends in a return instruction and is located in a subroutine within the existing program and/or shared library code. Chained together, these gadgets allow an attacker to perform arbitrary operations on a machine employing defenses that thwart simpler attacks." QEMU is by no means perfect with an ever growing set of CVEs from flawed hardware device emulation, which could potentially be exploited using ROP techniques. Since GCC 11 there has been a compiler option that can mitigate against this exploit technique: -fzero-call-user-regs To understand it refer to these two resources: https://www.jerkeby.se/newsletter/posts/rop-reduction-zero-call-user-regs/ https://gcc.gnu.org/pipermail/gcc-patches/2020-August/552262.html I used two programs to scan qemu-system-x86_64 for ROP gadgets: https://github.com/0vercl0k/rp https://github.com/JonathanSalwan/ROPgadget When asked to find 8 byte gadgets, the 'rp' tool reports: A total of 440278 gadgets found. You decided to keep only the unique ones, 156143 unique gadgets found. While the ROPgadget tool reports: Unique gadgets found: 353122 With the --ropchain argument, the latter attempts to use the found gadgets to product a chain that can execute arbitrary syscalls. With current QEMU it succeeds in this task, which is an undesirable situation. With QEMU modified to use -fzero-call-user-regs=used-gpr the 'rp' tool reports A total of 528991 gadgets found. You decided to keep only the unique ones, 121128 unique gadgets found. This is 22% fewer unique gadgets While the ROPgadget tool reports: Unique gadgets found: 328605 This is 7% fewer unique gadgets. Crucially though, despite this more modest reduction, the ROPgadget tool is no longer able to identify a chain of gadgets for executing arbitrary syscalls. It fails at the very first step, unable to find gadgets for populating registers for a future syscall. Having said that, more advanced tools do still manage to put together a viable ROP chain. Also this only takes into account QEMU code. QEMU links to many 3rd party shared libraries and ideally all of them would be compiled with this same hardening. That becomes a distro policy question though. In terms of performance impact, TCG was used as an evaluation test case. We're not interested in protecting TCG since it isn't designed to provide a security barrier, but it is performance sensitive code, so useful as a guide to how other areas of QEMU might be impacted. With the -fzero-call-user-regs=used-gpr argument present, using the real world test of booting a linux kernel and having init immediately poweroff, there is a ~1% slow down in performance under TCG. The QEMU binary size also grows by approximately 1%. By comparison, using the more aggressive -fzero-call-user-regs=all, results in a slowdown of over 25% in TCG, which is clearly not an acceptable impact, and a binary size increase of 5%. Considering that 'used-gpr' successfully stopped ROPgadget assembling a chain, this more targeted protection is a justifiable hardening / performance tradeoff. Reviewed-by: Thomas Huth <thuth@redhat.com> Signed-off-by: "Daniel P. Berrangé" <berrange@redhat.com> Message-ID: <20240103123414.2401208-2-berrange@redhat.com> Signed-off-by: Thomas Huth <thuth@redhat.com>
2024-01-03 15:34:13 +03:00
]
# Zero out registers used during a function call
# upon its return. This makes it harder to assemble
# ROP gadgets into something usable
#
# NB: Clang 17 is broken and SEGVs
# https://github.com/llvm/llvm-project/issues/75168
#
# NB2: This clashes with the "retguard" extension of OpenBSD's Clang
# https://gitlab.com/qemu-project/qemu/-/issues/2278
if host_os != 'openbsd' and \
cc.compiles('extern struct { void (*cb)(void); } s; void f(void) { s.cb(); }',
name: '-fzero-call-used-regs=used-gpr',
args: ['-O2', '-fzero-call-used-regs=used-gpr'])
hardening_flags += '-fzero-call-used-regs=used-gpr'
endif
meson: mitigate against ROP exploits with -fzero-call-used-regs To quote wikipedia: "Return-oriented programming (ROP) is a computer security exploit technique that allows an attacker to execute code in the presence of security defenses such as executable space protection and code signing. In this technique, an attacker gains control of the call stack to hijack program control flow and then executes carefully chosen machine instruction sequences that are already present in the machine's memory, called "gadgets". Each gadget typically ends in a return instruction and is located in a subroutine within the existing program and/or shared library code. Chained together, these gadgets allow an attacker to perform arbitrary operations on a machine employing defenses that thwart simpler attacks." QEMU is by no means perfect with an ever growing set of CVEs from flawed hardware device emulation, which could potentially be exploited using ROP techniques. Since GCC 11 there has been a compiler option that can mitigate against this exploit technique: -fzero-call-user-regs To understand it refer to these two resources: https://www.jerkeby.se/newsletter/posts/rop-reduction-zero-call-user-regs/ https://gcc.gnu.org/pipermail/gcc-patches/2020-August/552262.html I used two programs to scan qemu-system-x86_64 for ROP gadgets: https://github.com/0vercl0k/rp https://github.com/JonathanSalwan/ROPgadget When asked to find 8 byte gadgets, the 'rp' tool reports: A total of 440278 gadgets found. You decided to keep only the unique ones, 156143 unique gadgets found. While the ROPgadget tool reports: Unique gadgets found: 353122 With the --ropchain argument, the latter attempts to use the found gadgets to product a chain that can execute arbitrary syscalls. With current QEMU it succeeds in this task, which is an undesirable situation. With QEMU modified to use -fzero-call-user-regs=used-gpr the 'rp' tool reports A total of 528991 gadgets found. You decided to keep only the unique ones, 121128 unique gadgets found. This is 22% fewer unique gadgets While the ROPgadget tool reports: Unique gadgets found: 328605 This is 7% fewer unique gadgets. Crucially though, despite this more modest reduction, the ROPgadget tool is no longer able to identify a chain of gadgets for executing arbitrary syscalls. It fails at the very first step, unable to find gadgets for populating registers for a future syscall. Having said that, more advanced tools do still manage to put together a viable ROP chain. Also this only takes into account QEMU code. QEMU links to many 3rd party shared libraries and ideally all of them would be compiled with this same hardening. That becomes a distro policy question though. In terms of performance impact, TCG was used as an evaluation test case. We're not interested in protecting TCG since it isn't designed to provide a security barrier, but it is performance sensitive code, so useful as a guide to how other areas of QEMU might be impacted. With the -fzero-call-user-regs=used-gpr argument present, using the real world test of booting a linux kernel and having init immediately poweroff, there is a ~1% slow down in performance under TCG. The QEMU binary size also grows by approximately 1%. By comparison, using the more aggressive -fzero-call-user-regs=all, results in a slowdown of over 25% in TCG, which is clearly not an acceptable impact, and a binary size increase of 5%. Considering that 'used-gpr' successfully stopped ROPgadget assembling a chain, this more targeted protection is a justifiable hardening / performance tradeoff. Reviewed-by: Thomas Huth <thuth@redhat.com> Signed-off-by: "Daniel P. Berrangé" <berrange@redhat.com> Message-ID: <20240103123414.2401208-2-berrange@redhat.com> Signed-off-by: Thomas Huth <thuth@redhat.com>
2024-01-03 15:34:13 +03:00
qemu_common_flags += cc.get_supported_arguments(hardening_flags)
add_global_arguments(qemu_common_flags, native: false, language: all_languages)
add_global_link_arguments(qemu_ldflags, native: false, language: all_languages)
# Collect warning flags we want to set, sorted alphabetically
warn_flags = [
# First enable interesting warnings
'-Wempty-body',
'-Wendif-labels',
'-Wexpansion-to-defined',
'-Wformat-security',
'-Wformat-y2k',
'-Wignored-qualifiers',
'-Wimplicit-fallthrough=2',
'-Winit-self',
'-Wmissing-format-attribute',
'-Wmissing-prototypes',
'-Wnested-externs',
'-Wold-style-declaration',
'-Wold-style-definition',
'-Wredundant-decls',
'-Wshadow=local',
'-Wstrict-prototypes',
'-Wtype-limits',
'-Wundef',
'-Wvla',
'-Wwrite-strings',
# Then disable some undesirable warnings
'-Wno-gnu-variable-sized-type-not-at-end',
'-Wno-initializer-overrides',
'-Wno-missing-include-dirs',
'-Wno-psabi',
'-Wno-shift-negative-value',
'-Wno-string-plus-int',
'-Wno-tautological-type-limit-compare',
'-Wno-typedef-redefinition',
]
if host_os != 'darwin'
tsa_has_cleanup = cc.compiles('''
struct __attribute__((capability("mutex"))) mutex {};
void lock(struct mutex *m) __attribute__((acquire_capability(m)));
void unlock(struct mutex *m) __attribute__((release_capability(m)));
void test(void) {
struct mutex __attribute__((cleanup(unlock))) m;
lock(&m);
}
''', args: ['-Wthread-safety', '-Werror'])
if tsa_has_cleanup
warn_flags += ['-Wthread-safety']
endif
endif
# Set up C++ compiler flags
qemu_cxxflags = []
if 'cpp' in all_languages
qemu_cxxflags = ['-D__STDC_LIMIT_MACROS', '-D__STDC_CONSTANT_MACROS', '-D__STDC_FORMAT_MACROS'] + qemu_cflags
endif
add_project_arguments(qemu_cflags, native: false, language: 'c')
add_project_arguments(cc.get_supported_arguments(warn_flags), native: false, language: 'c')
if 'cpp' in all_languages
add_project_arguments(qemu_cxxflags, native: false, language: 'cpp')
add_project_arguments(cxx.get_supported_arguments(warn_flags), native: false, language: 'cpp')
endif
if 'objc' in all_languages
# Note sanitizer flags are not applied to Objective-C sources!
add_project_arguments(objc.get_supported_arguments(warn_flags), native: false, language: 'objc')
endif
if host_os == 'linux'
add_project_arguments('-isystem', meson.current_source_dir() / 'linux-headers',
'-isystem', 'linux-headers',
language: all_languages)
endif
add_project_arguments('-iquote', '.',
'-iquote', meson.current_source_dir(),
'-iquote', meson.current_source_dir() / 'include',
language: all_languages)
# If a host-specific include directory exists, list that first...
host_include = meson.current_source_dir() / 'host/include/'
if fs.is_dir(host_include / host_arch)
add_project_arguments('-iquote', host_include / host_arch,
language: all_languages)
endif
# ... followed by the generic fallback.
add_project_arguments('-iquote', host_include / 'generic',
language: all_languages)
sparse = find_program('cgcc', required: get_option('sparse'))
if sparse.found()
run_target('sparse',
command: [find_program('scripts/check_sparse.py'),
'compile_commands.json', sparse.full_path(), '-Wbitwise',
'-Wno-transparent-union', '-Wno-old-initializer',
'-Wno-non-pointer-null'])
endif
#####################################
# Host-specific libraries and flags #
#####################################
libm = cc.find_library('m', required: false)
threads = dependency('threads')
util = cc.find_library('util', required: false)
winmm = []
socket = []
version_res = []
coref = []
iokit = []
emulator_link_args = []
midl = not_found
widl = not_found
pathcch = not_found
host_dsosuf = '.so'
if host_os == 'windows'
midl = find_program('midl', required: false)
widl = find_program('widl', required: false)
pathcch = cc.find_library('pathcch')
socket = cc.find_library('ws2_32')
winmm = cc.find_library('winmm')
win = import('windows')
version_res = win.compile_resources('version.rc',
depend_files: files('pc-bios/qemu-nsis.ico'),
include_directories: include_directories('.'))
host_dsosuf = '.dll'
elif host_os == 'darwin'
coref = dependency('appleframeworks', modules: 'CoreFoundation')
iokit = dependency('appleframeworks', modules: 'IOKit', required: false)
host_dsosuf = '.dylib'
elif host_os == 'sunos'
socket = [cc.find_library('socket'),
cc.find_library('nsl'),
cc.find_library('resolv')]
elif host_os == 'haiku'
socket = [cc.find_library('posix_error_mapper'),
cc.find_library('network'),
cc.find_library('bsd')]
elif host_os == 'openbsd'
if get_option('tcg').allowed() and target_dirs.length() > 0
# Disable OpenBSD W^X if available
emulator_link_args = cc.get_supported_link_arguments('-Wl,-z,wxneeded')
endif
endif
###############################################
# Host-specific configuration of accelerators #
###############################################
accelerators = []
if get_option('kvm').allowed() and host_os == 'linux'
accelerators += 'CONFIG_KVM'
endif
if get_option('whpx').allowed() and host_os == 'windows'
if get_option('whpx').enabled() and host_machine.cpu() != 'x86_64'
error('WHPX requires 64-bit host')
elif cc.has_header('winhvplatform.h', required: get_option('whpx')) and \
cc.has_header('winhvemulation.h', required: get_option('whpx'))
accelerators += 'CONFIG_WHPX'
endif
endif
hvf = not_found
if get_option('hvf').allowed()
hvf = dependency('appleframeworks', modules: 'Hypervisor',
required: get_option('hvf'))
if hvf.found()
accelerators += 'CONFIG_HVF'
endif
endif
nvmm = not_found
if host_os == 'netbsd'
nvmm = cc.find_library('nvmm', required: get_option('nvmm'))
if nvmm.found()
accelerators += 'CONFIG_NVMM'
endif
endif
tcg_arch = host_arch
if get_option('tcg').allowed()
if host_arch == 'unknown'
if not get_option('tcg_interpreter')
error('Unsupported CPU @0@, try --enable-tcg-interpreter'.format(cpu))
endif
elif get_option('tcg_interpreter')
warning('Use of the TCG interpreter is not recommended on this host')
warning('architecture. There is a native TCG execution backend available')
warning('which provides substantially better performance and reliability.')
warning('It is strongly recommended to remove the --enable-tcg-interpreter')
warning('configuration option on this architecture to use the native')
warning('backend.')
endif
if get_option('tcg_interpreter')
tcg_arch = 'tci'
elif host_arch == 'x86_64'
tcg_arch = 'i386'
elif host_arch == 'ppc64'
tcg_arch = 'ppc'
endif
add_project_arguments('-iquote', meson.current_source_dir() / 'tcg' / tcg_arch,
language: all_languages)
accelerators += 'CONFIG_TCG'
endif
if 'CONFIG_KVM' not in accelerators and get_option('kvm').enabled()
error('KVM not available on this platform')
endif
if 'CONFIG_HVF' not in accelerators and get_option('hvf').enabled()
error('HVF not available on this platform')
endif
if 'CONFIG_NVMM' not in accelerators and get_option('nvmm').enabled()
error('NVMM not available on this platform')
endif
if 'CONFIG_WHPX' not in accelerators and get_option('whpx').enabled()
error('WHPX not available on this platform')
endif
xen = not_found
if get_option('xen').enabled() or (get_option('xen').auto() and have_system)
xencontrol = dependency('xencontrol', required: false,
method: 'pkg-config')
if xencontrol.found()
xen_pc = declare_dependency(version: xencontrol.version(),
dependencies: [
xencontrol,
# disabler: true makes xen_pc.found() return false if any is not found
dependency('xenstore', required: false,
method: 'pkg-config',
disabler: true),
dependency('xenforeignmemory', required: false,
method: 'pkg-config',
disabler: true),
dependency('xengnttab', required: false,
method: 'pkg-config',
disabler: true),
dependency('xenevtchn', required: false,
method: 'pkg-config',
disabler: true),
dependency('xendevicemodel', required: false,
method: 'pkg-config',
disabler: true),
# optional, no "disabler: true"
dependency('xentoolcore', required: false,
method: 'pkg-config')])
if xen_pc.found()
xen = xen_pc
endif
endif
if not xen.found()
xen_tests = [ '4.11.0', '4.10.0', '4.9.0', '4.8.0', '4.7.1' ]
xen_libs = {
'4.11.0': [ 'xenstore', 'xenctrl', 'xendevicemodel', 'xenforeignmemory', 'xengnttab', 'xenevtchn', 'xentoolcore' ],
'4.10.0': [ 'xenstore', 'xenctrl', 'xendevicemodel', 'xenforeignmemory', 'xengnttab', 'xenevtchn', 'xentoolcore' ],
'4.9.0': [ 'xenstore', 'xenctrl', 'xendevicemodel', 'xenforeignmemory', 'xengnttab', 'xenevtchn' ],
'4.8.0': [ 'xenstore', 'xenctrl', 'xenforeignmemory', 'xengnttab', 'xenevtchn' ],
'4.7.1': [ 'xenstore', 'xenctrl', 'xenforeignmemory', 'xengnttab', 'xenevtchn' ],
}
xen_deps = {}
foreach ver: xen_tests
# cache the various library tests to avoid polluting the logs
xen_test_deps = []
foreach l: xen_libs[ver]
if l not in xen_deps
xen_deps += { l: cc.find_library(l, required: false) }
endif
xen_test_deps += xen_deps[l]
endforeach
# Use -D to pick just one of the test programs in scripts/xen-detect.c
xen_version = ver.split('.')
xen_ctrl_version = xen_version[0] + \
('0' + xen_version[1]).substring(-2) + \
('0' + xen_version[2]).substring(-2)
if cc.links(files('scripts/xen-detect.c'),
args: '-DCONFIG_XEN_CTRL_INTERFACE_VERSION=' + xen_ctrl_version,
dependencies: xen_test_deps)
xen = declare_dependency(version: ver, dependencies: xen_test_deps)
break
endif
endforeach
endif
if xen.found()
accelerators += 'CONFIG_XEN'
elif get_option('xen').enabled()
error('could not compile and link Xen test program')
endif
endif
have_xen_pci_passthrough = get_option('xen_pci_passthrough') \
.require(xen.found(),
error_message: 'Xen PCI passthrough requested but Xen not enabled') \
.require(host_os == 'linux',
error_message: 'Xen PCI passthrough not available on this platform') \
.require(cpu == 'x86' or cpu == 'x86_64',
error_message: 'Xen PCI passthrough not available on this platform') \
.allowed()
################
# Dependencies #
################
# When bumping glib minimum version, please check also whether to increase
# the _WIN32_WINNT setting in osdep.h according to the value from glib
glib_req_ver = '>=2.66.0'
glib_pc = dependency('glib-2.0', version: glib_req_ver, required: true,
method: 'pkg-config')
glib_cflags = []
if enable_modules
gmodule = dependency('gmodule-export-2.0', version: glib_req_ver, required: true,
method: 'pkg-config')
elif get_option('plugins')
gmodule = dependency('gmodule-no-export-2.0', version: glib_req_ver, required: true,
method: 'pkg-config')
else
gmodule = not_found
endif
# This workaround is required due to a bug in pkg-config file for glib as it
# doesn't define GLIB_STATIC_COMPILATION for pkg-config --static
if host_os == 'windows' and get_option('prefer_static')
glib_cflags += ['-DGLIB_STATIC_COMPILATION']
endif
# Sanity check that the current size_t matches the
# size that glib thinks it should be. This catches
# problems on multi-arch where people try to build
# 32-bit QEMU while pointing at 64-bit glib headers
if not cc.compiles('''
#include <glib.h>
#include <unistd.h>
#define QEMU_BUILD_BUG_ON(x) \
typedef char qemu_build_bug_on[(x)?-1:1] __attribute__((unused));
int main(void) {
QEMU_BUILD_BUG_ON(sizeof(size_t) != GLIB_SIZEOF_SIZE_T);
return 0;
}''', dependencies: glib_pc, args: glib_cflags)
error('''sizeof(size_t) doesn't match GLIB_SIZEOF_SIZE_T.
You probably need to set PKG_CONFIG_LIBDIR" to point
to the right pkg-config files for your build target.''')
endif
glib = declare_dependency(dependencies: [glib_pc, gmodule],
compile_args: glib_cflags,
version: glib_pc.version())
# Check whether glib has gslice, which we have to avoid for correctness.
# TODO: remove this check and the corresponding workaround (qtree) when
# the minimum supported glib is >= 2.75.3
glib_has_gslice = glib.version().version_compare('<2.75.3')
# override glib dep to include the above refinements
meson.override_dependency('glib-2.0', glib)
# The path to glib.h is added to all compilation commands.
add_project_dependencies(glib.partial_dependency(compile_args: true, includes: true),
native: false, language: all_languages)
gio = not_found
gdbus_codegen = not_found
gdbus_codegen_error = '@0@ requires gdbus-codegen, please install libgio'
if not get_option('gio').auto() or have_system
gio = dependency('gio-2.0', required: get_option('gio'),
method: 'pkg-config')
if gio.found() and not cc.links('''
#include <gio/gio.h>
int main(void)
{
g_dbus_proxy_new_sync(0, 0, 0, 0, 0, 0, 0, 0);
return 0;
}''', dependencies: [glib, gio])
if get_option('gio').enabled()
error('The installed libgio is broken for static linking')
endif
gio = not_found
endif
if gio.found()
gdbus_codegen = find_program(gio.get_variable('gdbus_codegen'),
required: get_option('gio'))
gio_unix = dependency('gio-unix-2.0', required: get_option('gio'),
method: 'pkg-config')
gio = declare_dependency(dependencies: [gio, gio_unix],
version: gio.version())
endif
endif
if gdbus_codegen.found() and get_option('cfi')
gdbus_codegen = not_found
gdbus_codegen_error = '@0@ uses gdbus-codegen, which does not support control flow integrity'
endif
xml_pp = find_program('scripts/xml-preprocess.py')
lttng = not_found
if 'ust' in get_option('trace_backends')
lttng = dependency('lttng-ust', required: true, version: '>= 2.1',
method: 'pkg-config')
endif
pixman = not_found
if not get_option('pixman').auto() or have_system or have_tools
pixman = dependency('pixman-1', required: get_option('pixman'), version:'>=0.21.8',
method: 'pkg-config')
endif
zlib = dependency('zlib', required: true)
libaio = not_found
if not get_option('linux_aio').auto() or have_block
libaio = cc.find_library('aio', has_headers: ['libaio.h'],
required: get_option('linux_aio'))
endif
linux_io_uring_test = '''
#include <liburing.h>
#include <linux/errqueue.h>
int main(void) { return 0; }'''
linux_io_uring = not_found
if not get_option('linux_io_uring').auto() or have_block
linux_io_uring = dependency('liburing', version: '>=0.3',
required: get_option('linux_io_uring'),
method: 'pkg-config')
if not cc.links(linux_io_uring_test)
linux_io_uring = not_found
endif
endif
libnfs = not_found
if not get_option('libnfs').auto() or have_block
libnfs = dependency('libnfs', version: '>=1.9.3',
required: get_option('libnfs'),
method: 'pkg-config')
endif
libattr_test = '''
#include <stddef.h>
#include <sys/types.h>
#ifdef CONFIG_LIBATTR
#include <attr/xattr.h>
#else
#include <sys/xattr.h>
#endif
int main(void) { getxattr(NULL, NULL, NULL, 0); setxattr(NULL, NULL, NULL, 0, 0); return 0; }'''
libattr = not_found
have_old_libattr = false
if get_option('attr').allowed()
if cc.links(libattr_test)
libattr = declare_dependency()
else
libattr = cc.find_library('attr', has_headers: ['attr/xattr.h'],
required: get_option('attr'))
if libattr.found() and not \
cc.links(libattr_test, dependencies: libattr, args: '-DCONFIG_LIBATTR')
libattr = not_found
if get_option('attr').enabled()
error('could not link libattr')
else
warning('could not link libattr, disabling')
endif
else
have_old_libattr = libattr.found()
endif
endif
endif
cocoa = dependency('appleframeworks',
modules: ['Cocoa', 'CoreVideo', 'QuartzCore'],
required: get_option('cocoa'))
vmnet = dependency('appleframeworks', modules: 'vmnet', required: get_option('vmnet'))
if vmnet.found() and not cc.has_header_symbol('vmnet/vmnet.h',
'VMNET_BRIDGED_MODE',
dependencies: vmnet)
vmnet = not_found
if get_option('vmnet').enabled()
error('vmnet.framework API is outdated')
else
warning('vmnet.framework API is outdated, disabling')
endif
endif
seccomp = not_found
seccomp_has_sysrawrc = false
if not get_option('seccomp').auto() or have_system or have_tools
seccomp = dependency('libseccomp', version: '>=2.3.0',
required: get_option('seccomp'),
method: 'pkg-config')
if seccomp.found()
seccomp_has_sysrawrc = cc.has_header_symbol('seccomp.h',
'SCMP_FLTATR_API_SYSRAWRC',
dependencies: seccomp)
endif
endif
libcap_ng = not_found
if not get_option('cap_ng').auto() or have_system or have_tools
libcap_ng = cc.find_library('cap-ng', has_headers: ['cap-ng.h'],
required: get_option('cap_ng'))
endif
if libcap_ng.found() and not cc.links('''
#include <cap-ng.h>
int main(void)
{
capng_capability_to_name(CAPNG_EFFECTIVE);
return 0;
}''', dependencies: libcap_ng)
libcap_ng = not_found
if get_option('cap_ng').enabled()
error('could not link libcap-ng')
else
warning('could not link libcap-ng, disabling')
endif
endif
if get_option('xkbcommon').auto() and not have_system and not have_tools
xkbcommon = not_found
else
xkbcommon = dependency('xkbcommon', required: get_option('xkbcommon'),
method: 'pkg-config')
endif
slirp = not_found
if not get_option('slirp').auto() or have_system
slirp = dependency('slirp', required: get_option('slirp'),
method: 'pkg-config')
# slirp < 4.7 is incompatible with CFI support in QEMU. This is because
# it passes function pointers within libslirp as callbacks for timers.
# When using a system-wide shared libslirp, the type information for the
# callback is missing and the timer call produces a false positive with CFI.
# Do not use the "version" keyword argument to produce a better error.
# with control-flow integrity.
if get_option('cfi') and slirp.found() and slirp.version().version_compare('<4.7')
if get_option('slirp').enabled()
error('Control-Flow Integrity requires libslirp 4.7.')
else
warning('Cannot use libslirp since Control-Flow Integrity requires libslirp >= 4.7.')
slirp = not_found
endif
endif
endif
vde = not_found
if not get_option('vde').auto() or have_system or have_tools
vde = cc.find_library('vdeplug', has_headers: ['libvdeplug.h'],
required: get_option('vde'))
endif
if vde.found() and not cc.links('''
#include <libvdeplug.h>
int main(void)
{
struct vde_open_args a = {0, 0, 0};
char s[] = "";
vde_open(s, s, &a);
return 0;
}''', dependencies: vde)
vde = not_found
if get_option('cap_ng').enabled()
error('could not link libvdeplug')
else
warning('could not link libvdeplug, disabling')
endif
endif
pulse = not_found
if not get_option('pa').auto() or (host_os == 'linux' and have_system)
pulse = dependency('libpulse', required: get_option('pa'),
method: 'pkg-config')
endif
alsa = not_found
if not get_option('alsa').auto() or (host_os == 'linux' and have_system)
alsa = dependency('alsa', required: get_option('alsa'),
method: 'pkg-config')
endif
jack = not_found
if not get_option('jack').auto() or have_system
jack = dependency('jack', required: get_option('jack'),
method: 'pkg-config')
endif
pipewire = not_found
if not get_option('pipewire').auto() or (host_os == 'linux' and have_system)
pipewire = dependency('libpipewire-0.3', version: '>=0.3.60',
required: get_option('pipewire'),
method: 'pkg-config')
endif
sndio = not_found
if not get_option('sndio').auto() or have_system
sndio = dependency('sndio', required: get_option('sndio'),
method: 'pkg-config')
endif
spice_protocol = not_found
if not get_option('spice_protocol').auto() or have_system
spice_protocol = dependency('spice-protocol', version: '>=0.14.0',
required: get_option('spice_protocol'),
method: 'pkg-config')
endif
spice = not_found
if get_option('spice') \
.disable_auto_if(not have_system) \
.require(pixman.found(),
error_message: 'cannot enable SPICE if pixman is not available') \
.allowed()
spice = dependency('spice-server', version: '>=0.14.0',
required: get_option('spice'),
method: 'pkg-config')
endif
spice_headers = spice.partial_dependency(compile_args: true, includes: true)
rt = cc.find_library('rt', required: false)
libiscsi = not_found
if not get_option('libiscsi').auto() or have_block
libiscsi = dependency('libiscsi', version: '>=1.9.0',
required: get_option('libiscsi'),
method: 'pkg-config')
endif
zstd = not_found
if not get_option('zstd').auto() or have_block
zstd = dependency('libzstd', version: '>=1.4.0',
required: get_option('zstd'),
method: 'pkg-config')
endif
qpl = not_found
if not get_option('qpl').auto() or have_system
qpl = dependency('qpl', version: '>=1.5.0',
required: get_option('qpl'),
method: 'pkg-config')
endif
uadk = not_found
if not get_option('uadk').auto() or have_system
libwd = dependency('libwd', version: '>=2.6',
required: get_option('uadk'),
method: 'pkg-config')
libwd_comp = dependency('libwd_comp', version: '>=2.6',
required: get_option('uadk'),
method: 'pkg-config')
if libwd.found() and libwd_comp.found()
uadk = declare_dependency(dependencies: [libwd, libwd_comp])
endif
endif
qatzip = not_found
if not get_option('qatzip').auto() or have_system
qatzip = dependency('qatzip', version: '>=1.1.2',
required: get_option('qatzip'),
method: 'pkg-config')
endif
virgl = not_found
have_vhost_user_gpu = have_tools and host_os == 'linux' and pixman.found()
if not get_option('virglrenderer').auto() or have_system or have_vhost_user_gpu
virgl = dependency('virglrenderer',
method: 'pkg-config',
required: get_option('virglrenderer'))
endif
rutabaga = not_found
if not get_option('rutabaga_gfx').auto() or have_system or have_vhost_user_gpu
rutabaga = dependency('rutabaga_gfx_ffi',
method: 'pkg-config',
required: get_option('rutabaga_gfx'))
endif
blkio: add libblkio block driver libblkio (https://gitlab.com/libblkio/libblkio/) is a library for high-performance disk I/O. It currently supports io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa with additional drivers under development. One of the reasons for developing libblkio is that other applications besides QEMU can use it. This will be particularly useful for virtio-blk-vhost-user which applications may wish to use for connecting to qemu-storage-daemon. libblkio also gives us an opportunity to develop in Rust behind a C API that is easy to consume from QEMU. This commit adds io_uring, nvme-io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa BlockDrivers to QEMU using libblkio. It will be easy to add other libblkio drivers since they will share the majority of code. For now I/O buffers are copied through bounce buffers if the libblkio driver requires it. Later commits add an optimization for pre-registering guest RAM to avoid bounce buffers. The syntax is: --blockdev io_uring,node-name=drive0,filename=test.img,readonly=on|off,cache.direct=on|off --blockdev nvme-io_uring,node-name=drive0,filename=/dev/ng0n1,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-vdpa,node-name=drive0,path=/dev/vdpa...,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-user,node-name=drive0,path=vhost-user-blk.sock,readonly=on|off,cache.direct=on Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Message-id: 20221013185908.1297568-3-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-10-13 21:58:57 +03:00
blkio = not_found
if not get_option('blkio').auto() or have_block
blkio = dependency('blkio',
method: 'pkg-config',
required: get_option('blkio'))
blkio: add libblkio block driver libblkio (https://gitlab.com/libblkio/libblkio/) is a library for high-performance disk I/O. It currently supports io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa with additional drivers under development. One of the reasons for developing libblkio is that other applications besides QEMU can use it. This will be particularly useful for virtio-blk-vhost-user which applications may wish to use for connecting to qemu-storage-daemon. libblkio also gives us an opportunity to develop in Rust behind a C API that is easy to consume from QEMU. This commit adds io_uring, nvme-io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa BlockDrivers to QEMU using libblkio. It will be easy to add other libblkio drivers since they will share the majority of code. For now I/O buffers are copied through bounce buffers if the libblkio driver requires it. Later commits add an optimization for pre-registering guest RAM to avoid bounce buffers. The syntax is: --blockdev io_uring,node-name=drive0,filename=test.img,readonly=on|off,cache.direct=on|off --blockdev nvme-io_uring,node-name=drive0,filename=/dev/ng0n1,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-vdpa,node-name=drive0,path=/dev/vdpa...,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-user,node-name=drive0,path=vhost-user-blk.sock,readonly=on|off,cache.direct=on Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Message-id: 20221013185908.1297568-3-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-10-13 21:58:57 +03:00
endif
curl = not_found
if not get_option('curl').auto() or have_block
curl = dependency('libcurl', version: '>=7.29.0',
method: 'pkg-config',
required: get_option('curl'))
endif
libudev = not_found
if host_os == 'linux' and (have_system or have_tools)
libudev = dependency('libudev',
method: 'pkg-config',
required: get_option('libudev'))
endif
mpathlibs = [libudev]
mpathpersist = not_found
if host_os == 'linux' and have_tools and get_option('mpath').allowed()
mpath_test_source = '''
#include <libudev.h>
#include <mpath_persist.h>
unsigned mpath_mx_alloc_len = 1024;
int logsink;
static struct config *multipath_conf;
extern struct udev *udev;
extern struct config *get_multipath_config(void);
extern void put_multipath_config(struct config *conf);
struct udev *udev;
struct config *get_multipath_config(void) { return multipath_conf; }
void put_multipath_config(struct config *conf) { }
int main(void) {
udev = udev_new();
multipath_conf = mpath_lib_init();
return 0;
}'''
libmpathpersist = cc.find_library('mpathpersist',
required: get_option('mpath'))
if libmpathpersist.found()
mpathlibs += libmpathpersist
if get_option('prefer_static')
mpathlibs += cc.find_library('devmapper',
required: get_option('mpath'))
endif
mpathlibs += cc.find_library('multipath',
required: get_option('mpath'))
foreach lib: mpathlibs
if not lib.found()
mpathlibs = []
break
endif
endforeach
if mpathlibs.length() == 0
msg = 'Dependencies missing for libmpathpersist'
elif cc.links(mpath_test_source, dependencies: mpathlibs)
mpathpersist = declare_dependency(dependencies: mpathlibs)
else
msg = 'Cannot detect libmpathpersist API'
endif
if not mpathpersist.found()
if get_option('mpath').enabled()
error(msg)
else
warning(msg + ', disabling')
endif
endif
endif
endif
iconv = not_found
curses = not_found
if have_system and get_option('curses').allowed()
curses_test = '''
#if defined(__APPLE__) || defined(__OpenBSD__)
#define _XOPEN_SOURCE_EXTENDED 1
#endif
#include <locale.h>
#include <curses.h>
#include <wchar.h>
int main(void) {
wchar_t wch = L'w';
setlocale(LC_ALL, "");
resize_term(0, 0);
addwstr(L"wide chars\n");
addnwstr(&wch, 1);
add_wch(WACS_DEGREE);
return 0;
}'''
curses_dep_list = host_os == 'windows' ? ['ncurses', 'ncursesw'] : ['ncursesw']
curses = dependency(curses_dep_list,
required: false,
method: 'pkg-config')
msg = get_option('curses').enabled() ? 'curses library not found' : ''
curses_compile_args = ['-DNCURSES_WIDECHAR=1']
if curses.found()
if cc.links(curses_test, args: curses_compile_args, dependencies: [curses])
curses = declare_dependency(compile_args: curses_compile_args, dependencies: [curses],
version: curses.version())
else
msg = 'curses package not usable'
curses = not_found
endif
endif
if not curses.found()
has_curses_h = cc.has_header('curses.h', args: curses_compile_args)
if host_os != 'windows' and not has_curses_h
message('Trying with /usr/include/ncursesw')
curses_compile_args += ['-I/usr/include/ncursesw']
has_curses_h = cc.has_header('curses.h', args: curses_compile_args)
endif
if has_curses_h
curses_libname_list = (host_os == 'windows' ? ['pdcurses'] : ['ncursesw', 'cursesw'])
foreach curses_libname : curses_libname_list
libcurses = cc.find_library(curses_libname,
required: false)
if libcurses.found()
if cc.links(curses_test, args: curses_compile_args, dependencies: libcurses)
curses = declare_dependency(compile_args: curses_compile_args,
dependencies: [libcurses])
break
else
msg = 'curses library not usable'
endif
endif
endforeach
endif
endif
if get_option('iconv').allowed()
foreach link_args : [ ['-liconv'], [] ]
# Programs will be linked with glib and this will bring in libiconv on FreeBSD.
# We need to use libiconv if available because mixing libiconv's headers with
# the system libc does not work.
# However, without adding glib to the dependencies -L/usr/local/lib will not be
# included in the command line and libiconv will not be found.
if cc.links('''
#include <iconv.h>
int main(void) {
iconv_t conv = iconv_open("WCHAR_T", "UCS-2");
return conv != (iconv_t) -1;
}''', args: link_args, dependencies: glib)
iconv = declare_dependency(link_args: link_args, dependencies: glib)
break
endif
endforeach
endif
if curses.found() and not iconv.found()
if get_option('iconv').enabled()
error('iconv not available')
endif
msg = 'iconv required for curses UI but not available'
curses = not_found
endif
if not curses.found() and msg != ''
if get_option('curses').enabled()
error(msg)
else
warning(msg + ', disabling')
endif
endif
endif
brlapi = not_found
if not get_option('brlapi').auto() or have_system
brlapi = cc.find_library('brlapi', has_headers: ['brlapi.h'],
required: get_option('brlapi'))
if brlapi.found() and not cc.links('''
#include <brlapi.h>
#include <stddef.h>
int main(void) { return brlapi__openConnection (NULL, NULL, NULL); }''', dependencies: brlapi)
brlapi = not_found
if get_option('brlapi').enabled()
error('could not link brlapi')
else
warning('could not link brlapi, disabling')
endif
endif
endif
sdl = not_found
if not get_option('sdl').auto() or have_system
sdl = dependency('sdl2', required: get_option('sdl'))
sdl_image = not_found
endif
if sdl.found()
# Some versions of SDL have problems with -Wundef
if not cc.compiles('''
#include <SDL.h>
#include <SDL_syswm.h>
int main(int argc, char *argv[]) { return 0; }
''', dependencies: sdl, args: '-Werror=undef')
sdl = declare_dependency(compile_args: '-Wno-undef',
dependencies: sdl,
version: sdl.version())
endif
sdl_image = dependency('SDL2_image', required: get_option('sdl_image'),
method: 'pkg-config')
else
if get_option('sdl_image').enabled()
error('sdl-image required, but SDL was @0@'.format(
get_option('sdl').disabled() ? 'disabled' : 'not found'))
endif
sdl_image = not_found
endif
rbd = not_found
if not get_option('rbd').auto() or have_block
librados = cc.find_library('rados', required: get_option('rbd'))
librbd = cc.find_library('rbd', has_headers: ['rbd/librbd.h'],
required: get_option('rbd'))
if librados.found() and librbd.found()
if cc.links('''
#include <stdio.h>
#include <rbd/librbd.h>
int main(void) {
rados_t cluster;
rados_create(&cluster, NULL);
#if LIBRBD_VERSION_CODE < LIBRBD_VERSION(1, 12, 0)
#error
#endif
return 0;
}''', dependencies: [librbd, librados])
rbd = declare_dependency(dependencies: [librbd, librados])
elif get_option('rbd').enabled()
error('librbd >= 1.12.0 required')
else
warning('librbd >= 1.12.0 not found, disabling')
endif
endif
endif
glusterfs = not_found
glusterfs_ftruncate_has_stat = false
glusterfs_iocb_has_stat = false
if not get_option('glusterfs').auto() or have_block
glusterfs = dependency('glusterfs-api', version: '>=3',
required: get_option('glusterfs'),
method: 'pkg-config')
if glusterfs.found()
glusterfs_ftruncate_has_stat = cc.links('''
#include <glusterfs/api/glfs.h>
int
main(void)
{
/* new glfs_ftruncate() passes two additional args */
return glfs_ftruncate(NULL, 0, NULL, NULL);
}
''', dependencies: glusterfs)
glusterfs_iocb_has_stat = cc.links('''
#include <glusterfs/api/glfs.h>
/* new glfs_io_cbk() passes two additional glfs_stat structs */
static void
glusterfs_iocb(glfs_fd_t *fd, ssize_t ret, struct glfs_stat *prestat, struct glfs_stat *poststat, void *data)
{}
int
main(void)
{
glfs_io_cbk iocb = &glusterfs_iocb;
iocb(NULL, 0 , NULL, NULL, NULL);
return 0;
}
''', dependencies: glusterfs)
endif
endif
Add Hyper-V Dynamic Memory Protocol driver (hv-balloon) base This driver is like virtio-balloon on steroids: it allows both changing the guest memory allocation via ballooning and (in the next patch) inserting pieces of extra RAM into it on demand from a provided memory backend. The actual resizing is done via ballooning interface (for example, via the "balloon" HMP command). This includes resizing the guest past its boot size - that is, hot-adding additional memory in granularity limited only by the guest alignment requirements, as provided by the next patch. In contrast with ACPI DIMM hotplug where one can only request to unplug a whole DIMM stick this driver allows removing memory from guest in single page (4k) units via ballooning. After a VM reboot the guest is back to its original (boot) size. In the future, the guest boot memory size might be changed on reboot instead, taking into account the effective size that VM had before that reboot (much like Hyper-V does). For performance reasons, the guest-released memory is tracked in a few range trees, as a series of (start, count) ranges. Each time a new page range is inserted into such tree its neighbors are checked as candidates for possible merging with it. Besides performance reasons, the Dynamic Memory protocol itself uses page ranges as the data structure in its messages, so relevant pages need to be merged into such ranges anyway. One has to be careful when tracking the guest-released pages, since the guest can maliciously report returning pages outside its current address space, which later clash with the address range of newly added memory. Similarly, the guest can report freeing the same page twice. The above design results in much better ballooning performance than when using virtio-balloon with the same guest: 230 GB / minute with this driver versus 70 GB / minute with virtio-balloon. During a ballooning operation most of time is spent waiting for the guest to come up with newly freed page ranges, processing the received ranges on the host side (in QEMU and KVM) is nearly instantaneous. The unballoon operation is also pretty much instantaneous: thanks to the merging of the ballooned out page ranges 200 GB of memory can be returned to the guest in about 1 second. With virtio-balloon this operation takes about 2.5 minutes. These tests were done against a Windows Server 2019 guest running on a Xeon E5-2699, after dirtying the whole memory inside guest before each balloon operation. Using a range tree instead of a bitmap to track the removed memory also means that the solution scales well with the guest size: even a 1 TB range takes just a few bytes of such metadata. Since the required GTree operations aren't present in every Glib version a check for them was added to the meson build script, together with new "--enable-hv-balloon" and "--disable-hv-balloon" configure arguments. If these GTree operations are missing in the system's Glib version this driver will be skipped during QEMU build. An optional "status-report=on" device parameter requests memory status events from the guest (typically sent every second), which allow the host to learn both the guest memory available and the guest memory in use counts. Following commits will add support for their external emission as "HV_BALLOON_STATUS_REPORT" QMP events. The driver is named hv-balloon since the Linux kernel client driver for the Dynamic Memory Protocol is named as such and to follow the naming pattern established by the virtio-balloon driver. The whole protocol runs over Hyper-V VMBus. The driver was tested against Windows Server 2012 R2, Windows Server 2016 and Windows Server 2019 guests and obeys the guest alignment requirements reported to the host via DM_CAPABILITIES_REPORT message. Acked-by: David Hildenbrand <david@redhat.com> Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
2023-06-12 17:00:54 +03:00
hv_balloon = false
if get_option('hv_balloon').allowed() and have_system
if cc.links('''
#include <string.h>
#include <gmodule.h>
int main(void) {
GTree *tree;
tree = g_tree_new((GCompareFunc)strcmp);
(void)g_tree_node_first(tree);
g_tree_destroy(tree);
return 0;
}
''', dependencies: glib)
hv_balloon = true
else
if get_option('hv_balloon').enabled()
error('could not enable hv-balloon, update your glib')
else
warning('could not find glib support for hv-balloon, disabling')
endif
endif
endif
libssh = not_found
if not get_option('libssh').auto() or have_block
libssh = dependency('libssh', version: '>=0.8.7',
method: 'pkg-config',
required: get_option('libssh'))
endif
libbzip2 = not_found
if not get_option('bzip2').auto() or have_block
libbzip2 = cc.find_library('bz2', has_headers: ['bzlib.h'],
required: get_option('bzip2'))
if libbzip2.found() and not cc.links('''
#include <bzlib.h>
int main(void) { BZ2_bzlibVersion(); return 0; }''', dependencies: libbzip2)
libbzip2 = not_found
if get_option('bzip2').enabled()
error('could not link libbzip2')
else
warning('could not link libbzip2, disabling')
endif
endif
endif
liblzfse = not_found
if not get_option('lzfse').auto() or have_block
liblzfse = cc.find_library('lzfse', has_headers: ['lzfse.h'],
required: get_option('lzfse'))
endif
if liblzfse.found() and not cc.links('''
#include <lzfse.h>
int main(void) { lzfse_decode_scratch_size(); return 0; }''', dependencies: liblzfse)
liblzfse = not_found
if get_option('lzfse').enabled()
error('could not link liblzfse')
else
warning('could not link liblzfse, disabling')
endif
endif
oss = not_found
if get_option('oss').allowed() and have_system
if not cc.has_header('sys/soundcard.h')
# not found
elif host_os == 'netbsd'
oss = cc.find_library('ossaudio', required: get_option('oss'))
else
oss = declare_dependency()
endif
if not oss.found()
if get_option('oss').enabled()
error('OSS not found')
endif
endif
endif
dsound = not_found
if not get_option('dsound').auto() or (host_os == 'windows' and have_system)
if cc.has_header('dsound.h')
dsound = declare_dependency(link_args: ['-lole32', '-ldxguid'])
endif
if not dsound.found()
if get_option('dsound').enabled()
error('DirectSound not found')
endif
endif
endif
coreaudio = not_found
if not get_option('coreaudio').auto() or (host_os == 'darwin' and have_system)
coreaudio = dependency('appleframeworks', modules: 'CoreAudio',
required: get_option('coreaudio'))
endif
opengl = not_found
if not get_option('opengl').auto() or have_system or have_vhost_user_gpu
epoxy = dependency('epoxy', method: 'pkg-config',
required: get_option('opengl'))
if cc.has_header('epoxy/egl.h', dependencies: epoxy)
opengl = epoxy
elif get_option('opengl').enabled()
error('epoxy/egl.h not found')
endif
endif
gbm = not_found
if (have_system or have_tools) and (virgl.found() or opengl.found())
gbm = dependency('gbm', method: 'pkg-config', required: false)
endif
have_vhost_user_gpu = have_vhost_user_gpu and virgl.found() and opengl.found() and gbm.found()
gnutls = not_found
gnutls_crypto = not_found
if get_option('gnutls').enabled() or (get_option('gnutls').auto() and have_system)
# For general TLS support our min gnutls matches
# that implied by our platform support matrix
#
# For the crypto backends, we look for a newer
# gnutls:
#
# Version 3.6.8 is needed to get XTS
# Version 3.6.13 is needed to get PBKDF
# Version 3.6.14 is needed to get HW accelerated XTS
#
# If newer enough gnutls isn't available, we can
# still use a different crypto backend to satisfy
# the platform support requirements
gnutls_crypto = dependency('gnutls', version: '>=3.6.14',
method: 'pkg-config',
required: false)
if gnutls_crypto.found()
gnutls = gnutls_crypto
else
# Our min version if all we need is TLS
gnutls = dependency('gnutls', version: '>=3.5.18',
method: 'pkg-config',
required: get_option('gnutls'))
endif
endif
# We prefer use of gnutls for crypto, unless the options
# explicitly asked for nettle or gcrypt.
#
# If gnutls isn't available for crypto, then we'll prefer
# gcrypt over nettle for performance reasons.
gcrypt = not_found
nettle = not_found
hogweed = not_found
crypto_sm4 = not_found
xts = 'none'
if get_option('nettle').enabled() and get_option('gcrypt').enabled()
error('Only one of gcrypt & nettle can be enabled')
endif
# Explicit nettle/gcrypt request, so ignore gnutls for crypto
if get_option('nettle').enabled() or get_option('gcrypt').enabled()
gnutls_crypto = not_found
endif
if not gnutls_crypto.found()
if (not get_option('gcrypt').auto() or have_system) and not get_option('nettle').enabled()
gcrypt = dependency('libgcrypt', version: '>=1.8',
required: get_option('gcrypt'))
# Debian has removed -lgpg-error from libgcrypt-config
# as it "spreads unnecessary dependencies" which in
# turn breaks static builds...
if gcrypt.found() and get_option('prefer_static')
gcrypt = declare_dependency(dependencies:
[gcrypt,
cc.find_library('gpg-error', required: true)],
version: gcrypt.version())
endif
crypto_sm4 = gcrypt
# SM4 ALG is available in libgcrypt >= 1.9
if gcrypt.found() and not cc.links('''
#include <gcrypt.h>
int main(void) {
gcry_cipher_hd_t handler;
gcry_cipher_open(&handler, GCRY_CIPHER_SM4, GCRY_CIPHER_MODE_ECB, 0);
return 0;
}''', dependencies: gcrypt)
crypto_sm4 = not_found
endif
endif
if (not get_option('nettle').auto() or have_system) and not gcrypt.found()
nettle = dependency('nettle', version: '>=3.4',
method: 'pkg-config',
required: get_option('nettle'))
if nettle.found() and not cc.has_header('nettle/xts.h', dependencies: nettle)
xts = 'private'
endif
crypto_sm4 = nettle
# SM4 ALG is available in nettle >= 3.9
if nettle.found() and not cc.links('''
#include <nettle/sm4.h>
int main(void) {
struct sm4_ctx ctx;
unsigned char key[16] = {0};
sm4_set_encrypt_key(&ctx, key);
return 0;
}''', dependencies: nettle)
crypto_sm4 = not_found
endif
endif
endif
capstone = not_found
if not get_option('capstone').auto() or have_system or have_user
capstone = dependency('capstone', version: '>=3.0.5',
method: 'pkg-config',
required: get_option('capstone'))
# Some versions of capstone have broken pkg-config file
# that reports a wrong -I path, causing the #include to
# fail later. If the system has such a broken version
# do not use it.
if capstone.found() and not cc.compiles('#include <capstone.h>',
dependencies: [capstone])
capstone = not_found
if get_option('capstone').enabled()
error('capstone requested, but it does not appear to work')
endif
endif
endif
gmp = dependency('gmp', required: false, method: 'pkg-config')
if nettle.found() and gmp.found()
hogweed = dependency('hogweed', version: '>=3.4',
method: 'pkg-config',
required: get_option('nettle'))
endif
gtk = not_found
gtkx11 = not_found
vte = not_found
have_gtk_clipboard = get_option('gtk_clipboard').enabled()
if get_option('gtk') \
.disable_auto_if(not have_system) \
.require(pixman.found(),
error_message: 'cannot enable GTK if pixman is not available') \
.allowed()
gtk = dependency('gtk+-3.0', version: '>=3.22.0',
method: 'pkg-config',
required: get_option('gtk'))
if gtk.found()
gtkx11 = dependency('gtk+-x11-3.0', version: '>=3.22.0',
method: 'pkg-config',
required: false)
gtk = declare_dependency(dependencies: [gtk, gtkx11],
version: gtk.version())
if not get_option('vte').auto() or have_system
vte = dependency('vte-2.91',
method: 'pkg-config',
required: get_option('vte'))
endif
elif have_gtk_clipboard
error('GTK clipboard requested, but GTK not found')
endif
endif
x11 = not_found
if gtkx11.found()
x11 = dependency('x11', method: 'pkg-config', required: gtkx11.found())
endif
png = not_found
if get_option('png').allowed() and have_system
png = dependency('libpng', version: '>=1.6.34', required: get_option('png'),
method: 'pkg-config')
endif
vnc = not_found
jpeg = not_found
sasl = not_found
if get_option('vnc') \
.disable_auto_if(not have_system) \
.require(pixman.found(),
error_message: 'cannot enable VNC if pixman is not available') \
.allowed()
vnc = declare_dependency() # dummy dependency
jpeg = dependency('libjpeg', required: get_option('vnc_jpeg'),
method: 'pkg-config')
sasl = cc.find_library('sasl2', has_headers: ['sasl/sasl.h'],
required: get_option('vnc_sasl'))
if sasl.found()
sasl = declare_dependency(dependencies: sasl,
compile_args: '-DSTRUCT_IOVEC_DEFINED')
endif
endif
pam = not_found
if not get_option('auth_pam').auto() or have_system
pam = cc.find_library('pam', has_headers: ['security/pam_appl.h'],
required: get_option('auth_pam'))
endif
if pam.found() and not cc.links('''
#include <stddef.h>
#include <security/pam_appl.h>
int main(void) {
const char *service_name = "qemu";
const char *user = "frank";
const struct pam_conv pam_conv = { 0 };
pam_handle_t *pamh = NULL;
pam_start(service_name, user, &pam_conv, &pamh);
return 0;
}''', dependencies: pam)
pam = not_found
if get_option('auth_pam').enabled()
error('could not link libpam')
else
warning('could not link libpam, disabling')
endif
endif
snappy = not_found
if not get_option('snappy').auto() or have_system
snappy = cc.find_library('snappy', has_headers: ['snappy-c.h'],
required: get_option('snappy'))
endif
if snappy.found() and not cc.links('''
#include <snappy-c.h>
int main(void) { snappy_max_compressed_length(4096); return 0; }''', dependencies: snappy)
snappy = not_found
if get_option('snappy').enabled()
error('could not link libsnappy')
else
warning('could not link libsnappy, disabling')
endif
endif
lzo = not_found
if not get_option('lzo').auto() or have_system
lzo = cc.find_library('lzo2', has_headers: ['lzo/lzo1x.h'],
required: get_option('lzo'))
endif
if lzo.found() and not cc.links('''
#include <lzo/lzo1x.h>
int main(void) { lzo_version(); return 0; }''', dependencies: lzo)
lzo = not_found
if get_option('lzo').enabled()
error('could not link liblzo2')
else
warning('could not link liblzo2, disabling')
endif
endif
numa = not_found
if not get_option('numa').auto() or have_system or have_tools
numa = cc.find_library('numa', has_headers: ['numa.h'],
required: get_option('numa'))
endif
if numa.found() and not cc.links('''
#include <numa.h>
int main(void) { return numa_available(); }
''', dependencies: numa)
numa = not_found
if get_option('numa').enabled()
error('could not link numa')
else
warning('could not link numa, disabling')
endif
endif
fdt = not_found
fdt_opt = get_option('fdt')
if fdt_opt == 'enabled' and get_option('wrap_mode') == 'nodownload'
fdt_opt = 'system'
endif
if fdt_opt in ['enabled', 'system'] or (fdt_opt == 'auto' and have_system)
fdt = cc.find_library('fdt', required: fdt_opt == 'system')
if fdt.found() and cc.links('''
#include <libfdt.h>
#include <libfdt_env.h>
int main(void) { fdt_find_max_phandle(NULL, NULL); return 0; }''',
dependencies: fdt)
fdt_opt = 'system'
elif fdt_opt != 'system'
fdt_opt = get_option('wrap_mode') == 'nodownload' ? 'disabled' : 'internal'
fdt = not_found
else
error('system libfdt is too old (1.5.1 or newer required)')
endif
endif
if fdt_opt == 'internal'
assert(not fdt.found())
libfdt_proj = subproject('dtc', required: true,
default_options: ['tools=false', 'yaml=disabled',
'python=disabled', 'default_library=static'])
fdt = libfdt_proj.get_variable('libfdt_dep')
endif
rdma = not_found
if not get_option('rdma').auto() or have_system
rdma_libs = [cc.find_library('rdmacm', has_headers: ['rdma/rdma_cma.h'],
required: get_option('rdma')),
cc.find_library('ibverbs', required: get_option('rdma'))]
rdma = declare_dependency(dependencies: rdma_libs)
foreach lib: rdma_libs
if not lib.found()
rdma = not_found
endif
endforeach
endif
cacard = not_found
if not get_option('smartcard').auto() or have_system
cacard = dependency('libcacard', required: get_option('smartcard'),
version: '>=2.5.1', method: 'pkg-config')
endif
u2f = not_found
if not get_option('u2f').auto() or have_system
u2f = dependency('u2f-emu', required: get_option('u2f'),
method: 'pkg-config')
endif
canokey = not_found
if not get_option('canokey').auto() or have_system
canokey = dependency('canokey-qemu', required: get_option('canokey'),
method: 'pkg-config')
endif
usbredir = not_found
if not get_option('usb_redir').auto() or have_system
usbredir = dependency('libusbredirparser-0.5', required: get_option('usb_redir'),
version: '>=0.6', method: 'pkg-config')
endif
libusb = not_found
if not get_option('libusb').auto() or have_system
libusb = dependency('libusb-1.0', required: get_option('libusb'),
version: '>=1.0.13', method: 'pkg-config')
endif
libpmem = not_found
if not get_option('libpmem').auto() or have_system
libpmem = dependency('libpmem', required: get_option('libpmem'),
method: 'pkg-config')
endif
libdaxctl = not_found
if not get_option('libdaxctl').auto() or have_system
libdaxctl = dependency('libdaxctl', required: get_option('libdaxctl'),
version: '>=57', method: 'pkg-config')
endif
tasn1 = not_found
if gnutls.found()
tasn1 = dependency('libtasn1',
required: false,
method: 'pkg-config')
endif
keyutils = not_found
if not get_option('libkeyutils').auto() or have_block
keyutils = dependency('libkeyutils', required: get_option('libkeyutils'),
method: 'pkg-config')
endif
has_gettid = cc.has_function('gettid')
nbd/server: Add --selinux-label option Under SELinux, Unix domain sockets have two labels. One is on the disk and can be set with commands such as chcon(1). There is a different label stored in memory (called the process label). This can only be set by the process creating the socket. When using SELinux + SVirt and wanting qemu to be able to connect to a qemu-nbd instance, you must set both labels correctly first. For qemu-nbd the options to set the second label are awkward. You can create the socket in a wrapper program and then exec into qemu-nbd. Or you could try something with LD_PRELOAD. This commit adds the ability to set the label straightforwardly on the command line, via the new --selinux-label flag. (The name of the flag is the same as the equivalent nbdkit option.) A worked example showing how to use the new option can be found in this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Signed-off-by: Richard W.M. Jones <rjones@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> [eblake: rebase to configure changes, reject --selinux-label if it is not compiled in or not used on a Unix socket] Note that we may relax some of these restrictions at a later date, such as making it possible to label a TCP socket, although it may be smarter to do so as a generic QMP action rather than more one-off command lines in qemu-nbd. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20211115202944.615966-1-eblake@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> [eblake: adjust meson output as suggested by thuth] Signed-off-by: Eric Blake <eblake@redhat.com>
2021-11-15 23:29:43 +03:00
# libselinux
selinux = dependency('libselinux',
required: get_option('selinux'),
method: 'pkg-config')
nbd/server: Add --selinux-label option Under SELinux, Unix domain sockets have two labels. One is on the disk and can be set with commands such as chcon(1). There is a different label stored in memory (called the process label). This can only be set by the process creating the socket. When using SELinux + SVirt and wanting qemu to be able to connect to a qemu-nbd instance, you must set both labels correctly first. For qemu-nbd the options to set the second label are awkward. You can create the socket in a wrapper program and then exec into qemu-nbd. Or you could try something with LD_PRELOAD. This commit adds the ability to set the label straightforwardly on the command line, via the new --selinux-label flag. (The name of the flag is the same as the equivalent nbdkit option.) A worked example showing how to use the new option can be found in this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Signed-off-by: Richard W.M. Jones <rjones@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> [eblake: rebase to configure changes, reject --selinux-label if it is not compiled in or not used on a Unix socket] Note that we may relax some of these restrictions at a later date, such as making it possible to label a TCP socket, although it may be smarter to do so as a generic QMP action rather than more one-off command lines in qemu-nbd. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20211115202944.615966-1-eblake@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> [eblake: adjust meson output as suggested by thuth] Signed-off-by: Eric Blake <eblake@redhat.com>
2021-11-15 23:29:43 +03:00
# Malloc tests
malloc = []
if get_option('malloc') == 'system'
has_malloc_trim = \
get_option('malloc_trim').allowed() and \
cc.has_function('malloc_trim', prefix: '#include <malloc.h>')
else
has_malloc_trim = false
malloc = cc.find_library(get_option('malloc'), required: true)
endif
if not has_malloc_trim and get_option('malloc_trim').enabled()
if get_option('malloc') == 'system'
error('malloc_trim not available on this platform.')
else
error('malloc_trim not available with non-libc memory allocator')
endif
endif
gnu_source_prefix = '''
#ifndef _GNU_SOURCE
#define _GNU_SOURCE
#endif
'''
# Check whether the glibc provides STATX_BASIC_STATS
has_statx = cc.has_header_symbol('sys/stat.h', 'STATX_BASIC_STATS', prefix: gnu_source_prefix)
# Check whether statx() provides mount ID information
has_statx_mnt_id = cc.has_header_symbol('sys/stat.h', 'STATX_MNT_ID', prefix: gnu_source_prefix)
have_vhost_user_blk_server = get_option('vhost_user_blk_server') \
.require(host_os == 'linux',
error_message: 'vhost_user_blk_server requires linux') \
.require(have_vhost_user,
error_message: 'vhost_user_blk_server requires vhost-user support') \
.disable_auto_if(not have_tools and not have_system) \
.allowed()
if get_option('fuse').disabled() and get_option('fuse_lseek').enabled()
error('Cannot enable fuse-lseek while fuse is disabled')
endif
fuse = dependency('fuse3', required: get_option('fuse'),
version: '>=3.1', method: 'pkg-config')
fuse_lseek = not_found
if get_option('fuse_lseek').allowed()
if fuse.version().version_compare('>=3.8')
# Dummy dependency
fuse_lseek = declare_dependency()
elif get_option('fuse_lseek').enabled()
if fuse.found()
error('fuse-lseek requires libfuse >=3.8, found ' + fuse.version())
else
error('fuse-lseek requires libfuse, which was not found')
endif
endif
endif
have_libvduse = (host_os == 'linux')
if get_option('libvduse').enabled()
if host_os != 'linux'
error('libvduse requires linux')
endif
elif get_option('libvduse').disabled()
have_libvduse = false
endif
have_vduse_blk_export = (have_libvduse and host_os == 'linux')
if get_option('vduse_blk_export').enabled()
if host_os != 'linux'
error('vduse_blk_export requires linux')
elif not have_libvduse
error('vduse_blk_export requires libvduse support')
endif
elif get_option('vduse_blk_export').disabled()
have_vduse_blk_export = false
endif
# libbpf
bpf_version = '1.1.0'
libbpf = dependency('libbpf', version: '>=' + bpf_version, required: get_option('bpf'), method: 'pkg-config')
if libbpf.found() and not cc.links('''
#include <bpf/libbpf.h>
#include <linux/bpf.h>
int main(void)
{
// check flag availability
int flag = BPF_F_MMAPABLE;
bpf_object__destroy_skeleton(NULL);
return 0;
}''', dependencies: libbpf)
libbpf = not_found
if get_option('bpf').enabled()
error('libbpf skeleton/mmaping test failed')
else
warning('libbpf skeleton/mmaping test failed, disabling')
endif
endif
net: add initial support for AF_XDP network backend AF_XDP is a network socket family that allows communication directly with the network device driver in the kernel, bypassing most or all of the kernel networking stack. In the essence, the technology is pretty similar to netmap. But, unlike netmap, AF_XDP is Linux-native and works with any network interfaces without driver modifications. Unlike vhost-based backends (kernel, user, vdpa), AF_XDP doesn't require access to character devices or unix sockets. Only access to the network interface itself is necessary. This patch implements a network backend that communicates with the kernel by creating an AF_XDP socket. A chunk of userspace memory is shared between QEMU and the host kernel. 4 ring buffers (Tx, Rx, Fill and Completion) are placed in that memory along with a pool of memory buffers for the packet data. Data transmission is done by allocating one of the buffers, copying packet data into it and placing the pointer into Tx ring. After transmission, device will return the buffer via Completion ring. On Rx, device will take a buffer form a pre-populated Fill ring, write the packet data into it and place the buffer into Rx ring. AF_XDP network backend takes on the communication with the host kernel and the network interface and forwards packets to/from the peer device in QEMU. Usage example: -device virtio-net-pci,netdev=guest1,mac=00:16:35:AF:AA:5C -netdev af-xdp,ifname=ens6f1np1,id=guest1,mode=native,queues=1 XDP program bridges the socket with a network interface. It can be attached to the interface in 2 different modes: 1. skb - this mode should work for any interface and doesn't require driver support. With a caveat of lower performance. 2. native - this does require support from the driver and allows to bypass skb allocation in the kernel and potentially use zero-copy while getting packets in/out userspace. By default, QEMU will try to use native mode and fall back to skb. Mode can be forced via 'mode' option. To force 'copy' even in native mode, use 'force-copy=on' option. This might be useful if there is some issue with the driver. Option 'queues=N' allows to specify how many device queues should be open. Note that all the queues that are not open are still functional and can receive traffic, but it will not be delivered to QEMU. So, the number of device queues should generally match the QEMU configuration, unless the device is shared with something else and the traffic re-direction to appropriate queues is correctly configured on a device level (e.g. with ethtool -N). 'start-queue=M' option can be used to specify from which queue id QEMU should start configuring 'N' queues. It might also be necessary to use this option with certain NICs, e.g. MLX5 NICs. See the docs for examples. In a general case QEMU will need CAP_NET_ADMIN and CAP_SYS_ADMIN or CAP_BPF capabilities in order to load default XSK/XDP programs to the network interface and configure BPF maps. It is possible, however, to run with no capabilities. For that to work, an external process with enough capabilities will need to pre-load default XSK program, create AF_XDP sockets and pass their file descriptors to QEMU process on startup via 'sock-fds' option. Network backend will need to be configured with 'inhibit=on' to avoid loading of the program. QEMU will need 32 MB of locked memory (RLIMIT_MEMLOCK) per queue or CAP_IPC_LOCK. There are few performance challenges with the current network backends. First is that they do not support IO threads. This means that data path is handled by the main thread in QEMU and may slow down other work or may be slowed down by some other work. This also means that taking advantage of multi-queue is generally not possible today. Another thing is that data path is going through the device emulation code, which is not really optimized for performance. The fastest "frontend" device is virtio-net. But it's not optimized for heavy traffic either, because it expects such use-cases to be handled via some implementation of vhost (user, kernel, vdpa). In practice, we have virtio notifications and rcu lock/unlock on a per-packet basis and not very efficient accesses to the guest memory. Communication channels between backend and frontend devices do not allow passing more than one packet at a time as well. Some of these challenges can be avoided in the future by adding better batching into device emulation or by implementing vhost-af-xdp variant. There are also a few kernel limitations. AF_XDP sockets do not support any kinds of checksum or segmentation offloading. Buffers are limited to a page size (4K), i.e. MTU is limited. Multi-buffer support implementation for AF_XDP is in progress, but not ready yet. Also, transmission in all non-zero-copy modes is synchronous, i.e. done in a syscall. That doesn't allow high packet rates on virtual interfaces. However, keeping in mind all of these challenges, current implementation of the AF_XDP backend shows a decent performance while running on top of a physical NIC with zero-copy support. Test setup: 2 VMs running on 2 physical hosts connected via ConnectX6-Dx card. Network backend is configured to open the NIC directly in native mode. The driver supports zero-copy. NIC is configured to use 1 queue. Inside a VM - iperf3 for basic TCP performance testing and dpdk-testpmd for PPS testing. iperf3 result: TCP stream : 19.1 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 3.4 Mpps Rx only : 2.0 Mpps L2 FWD Loopback : 1.5 Mpps In skb mode the same setup shows much lower performance, similar to the setup where pair of physical NICs is replaced with veth pair: iperf3 result: TCP stream : 9 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 1.2 Mpps Rx only : 1.0 Mpps L2 FWD Loopback : 0.7 Mpps Results in skb mode or over the veth are close to results of a tap backend with vhost=on and disabled segmentation offloading bridged with a NIC. Signed-off-by: Ilya Maximets <i.maximets@ovn.org> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> (docker/lcitool) Signed-off-by: Jason Wang <jasowang@redhat.com>
2023-09-13 21:34:37 +03:00
# libxdp
libxdp = not_found
if not get_option('af_xdp').auto() or have_system
libxdp = dependency('libxdp', required: get_option('af_xdp'),
version: '>=1.4.0', method: 'pkg-config')
endif
# libdw
libdw = not_found
if not get_option('libdw').auto() or \
(not get_option('prefer_static') and (have_system or have_user))
libdw = dependency('libdw',
method: 'pkg-config',
required: get_option('libdw'))
endif
#################
# config-host.h #
#################
config_host_data = configuration_data()
audio_drivers_selected = []
if have_system
audio_drivers_available = {
'alsa': alsa.found(),
'coreaudio': coreaudio.found(),
'dsound': dsound.found(),
'jack': jack.found(),
'oss': oss.found(),
'pa': pulse.found(),
'pipewire': pipewire.found(),
'sdl': sdl.found(),
'sndio': sndio.found(),
}
foreach k, v: audio_drivers_available
config_host_data.set('CONFIG_AUDIO_' + k.to_upper(), v)
endforeach
# Default to native drivers first, OSS second, SDL third
audio_drivers_priority = \
[ 'pa', 'coreaudio', 'dsound', 'sndio', 'oss' ] + \
(host_os == 'linux' ? [] : [ 'sdl' ])
audio_drivers_default = []
foreach k: audio_drivers_priority
if audio_drivers_available[k]
audio_drivers_default += k
endif
endforeach
foreach k: get_option('audio_drv_list')
if k == 'default'
audio_drivers_selected += audio_drivers_default
elif not audio_drivers_available[k]
error('Audio driver "@0@" not available.'.format(k))
else
audio_drivers_selected += k
endif
endforeach
endif
config_host_data.set('CONFIG_AUDIO_DRIVERS',
'"' + '", "'.join(audio_drivers_selected) + '", ')
have_host_block_device = (host_os != 'darwin' or
cc.has_header('IOKit/storage/IOMedia.h'))
dbus_display = get_option('dbus_display') \
.require(gio.version().version_compare('>=2.64'),
error_message: '-display dbus requires glib>=2.64') \
.require(gdbus_codegen.found(),
error_message: gdbus_codegen_error.format('-display dbus')) \
.allowed()
have_virtfs = get_option('virtfs') \
.require(host_os == 'linux' or host_os == 'darwin',
error_message: 'virtio-9p (virtfs) requires Linux or macOS') \
.require(host_os == 'linux' or cc.has_function('pthread_fchdir_np'),
error_message: 'virtio-9p (virtfs) on macOS requires the presence of pthread_fchdir_np') \
.require(host_os == 'darwin' or libattr.found(),
error_message: 'virtio-9p (virtfs) on Linux requires libattr-devel') \
.disable_auto_if(not have_tools and not have_system) \
.allowed()
qga_fsfreeze = false
qga_fstrim = false
if host_os == 'linux'
if cc.has_header_symbol('linux/fs.h', 'FIFREEZE')
qga_fsfreeze = true
endif
if cc.has_header_symbol('linux/fs.h', 'FITRIM')
qga_fstrim = true
endif
elif host_os == 'freebsd' and cc.has_header_symbol('ufs/ffs/fs.h', 'UFSSUSPEND')
qga_fsfreeze = true
endif
if get_option('block_drv_ro_whitelist') == ''
config_host_data.set('CONFIG_BDRV_RO_WHITELIST', '')
else
config_host_data.set('CONFIG_BDRV_RO_WHITELIST',
'"' + get_option('block_drv_ro_whitelist').replace(',', '", "') + '", ')
endif
if get_option('block_drv_rw_whitelist') == ''
config_host_data.set('CONFIG_BDRV_RW_WHITELIST', '')
else
config_host_data.set('CONFIG_BDRV_RW_WHITELIST',
'"' + get_option('block_drv_rw_whitelist').replace(',', '", "') + '", ')
endif
foreach k : get_option('trace_backends')
config_host_data.set('CONFIG_TRACE_' + k.to_upper(), true)
endforeach
config_host_data.set_quoted('CONFIG_TRACE_FILE', get_option('trace_file'))
config_host_data.set_quoted('CONFIG_TLS_PRIORITY', get_option('tls_priority'))
if iasl.found()
config_host_data.set_quoted('CONFIG_IASL', iasl.full_path())
endif
config_host_data.set_quoted('CONFIG_BINDIR', get_option('prefix') / get_option('bindir'))
config_host_data.set_quoted('CONFIG_PREFIX', get_option('prefix'))
config_host_data.set_quoted('CONFIG_QEMU_CONFDIR', get_option('prefix') / qemu_confdir)
config_host_data.set_quoted('CONFIG_QEMU_DATADIR', get_option('prefix') / qemu_datadir)
config_host_data.set_quoted('CONFIG_QEMU_DESKTOPDIR', get_option('prefix') / qemu_desktopdir)
qemu_firmwarepath = ''
foreach k : get_option('qemu_firmwarepath')
qemu_firmwarepath += '"' + get_option('prefix') / k + '", '
endforeach
config_host_data.set('CONFIG_QEMU_FIRMWAREPATH', qemu_firmwarepath)
config_host_data.set_quoted('CONFIG_QEMU_HELPERDIR', get_option('prefix') / get_option('libexecdir'))
config_host_data.set_quoted('CONFIG_QEMU_ICONDIR', get_option('prefix') / qemu_icondir)
config_host_data.set_quoted('CONFIG_QEMU_LOCALEDIR', get_option('prefix') / get_option('localedir'))
config_host_data.set_quoted('CONFIG_QEMU_LOCALSTATEDIR', get_option('prefix') / get_option('localstatedir'))
config_host_data.set_quoted('CONFIG_QEMU_MODDIR', get_option('prefix') / qemu_moddir)
config_host_data.set_quoted('CONFIG_SYSCONFDIR', get_option('prefix') / get_option('sysconfdir'))
if enable_modules
config_host_data.set('CONFIG_STAMP', run_command(
meson.current_source_dir() / 'scripts/qemu-stamp.py',
meson.project_version(), get_option('pkgversion'), '--',
meson.current_source_dir() / 'configure',
capture: true, check: true).stdout().strip())
endif
have_slirp_smbd = get_option('slirp_smbd') \
.require(host_os != 'windows', error_message: 'Host smbd not supported on this platform.') \
.allowed()
if have_slirp_smbd
smbd_path = get_option('smbd')
if smbd_path == ''
smbd_path = (host_os == 'sunos' ? '/usr/sfw/sbin/smbd' : '/usr/sbin/smbd')
endif
config_host_data.set_quoted('CONFIG_SMBD_COMMAND', smbd_path)
endif
config_host_data.set('HOST_' + host_arch.to_upper(), 1)
kvm_targets_c = '""'
if get_option('kvm').allowed() and host_os == 'linux'
kvm_targets_c = '"' + '" ,"'.join(kvm_targets) + '"'
endif
config_host_data.set('CONFIG_KVM_TARGETS', kvm_targets_c)
if get_option('module_upgrades') and not enable_modules
error('Cannot enable module-upgrades as modules are not enabled')
endif
config_host_data.set('CONFIG_MODULE_UPGRADES', get_option('module_upgrades'))
config_host_data.set('CONFIG_ATTR', libattr.found())
config_host_data.set('CONFIG_BDRV_WHITELIST_TOOLS', get_option('block_drv_whitelist_in_tools'))
config_host_data.set('CONFIG_BRLAPI', brlapi.found())
config_host_data.set('CONFIG_BSD', host_os in bsd_oses)
config_host_data.set('CONFIG_FREEBSD', host_os == 'freebsd')
config_host_data.set('CONFIG_CAPSTONE', capstone.found())
config_host_data.set('CONFIG_COCOA', cocoa.found())
config_host_data.set('CONFIG_DARWIN', host_os == 'darwin')
config_host_data.set('CONFIG_FDT', fdt.found())
config_host_data.set('CONFIG_FUZZ', get_option('fuzzing'))
config_host_data.set('CONFIG_GCOV', get_option('b_coverage'))
config_host_data.set('CONFIG_LIBUDEV', libudev.found())
config_host_data.set('CONFIG_LINUX', host_os == 'linux')
config_host_data.set('CONFIG_POSIX', host_os != 'windows')
config_host_data.set('CONFIG_WIN32', host_os == 'windows')
config_host_data.set('CONFIG_LZO', lzo.found())
config_host_data.set('CONFIG_MPATH', mpathpersist.found())
blkio: add libblkio block driver libblkio (https://gitlab.com/libblkio/libblkio/) is a library for high-performance disk I/O. It currently supports io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa with additional drivers under development. One of the reasons for developing libblkio is that other applications besides QEMU can use it. This will be particularly useful for virtio-blk-vhost-user which applications may wish to use for connecting to qemu-storage-daemon. libblkio also gives us an opportunity to develop in Rust behind a C API that is easy to consume from QEMU. This commit adds io_uring, nvme-io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa BlockDrivers to QEMU using libblkio. It will be easy to add other libblkio drivers since they will share the majority of code. For now I/O buffers are copied through bounce buffers if the libblkio driver requires it. Later commits add an optimization for pre-registering guest RAM to avoid bounce buffers. The syntax is: --blockdev io_uring,node-name=drive0,filename=test.img,readonly=on|off,cache.direct=on|off --blockdev nvme-io_uring,node-name=drive0,filename=/dev/ng0n1,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-vdpa,node-name=drive0,path=/dev/vdpa...,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-user,node-name=drive0,path=vhost-user-blk.sock,readonly=on|off,cache.direct=on Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Message-id: 20221013185908.1297568-3-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-10-13 21:58:57 +03:00
config_host_data.set('CONFIG_BLKIO', blkio.found())
if blkio.found()
config_host_data.set('CONFIG_BLKIO_VHOST_VDPA_FD',
blkio.version().version_compare('>=1.3.0'))
config_host_data.set('CONFIG_BLKIO_WRITE_ZEROS_FUA',
blkio.version().version_compare('>=1.4.0'))
endif
config_host_data.set('CONFIG_CURL', curl.found())
config_host_data.set('CONFIG_CURSES', curses.found())
config_host_data.set('CONFIG_GBM', gbm.found())
config_host_data.set('CONFIG_GIO', gio.found())
config_host_data.set('CONFIG_GLUSTERFS', glusterfs.found())
if glusterfs.found()
config_host_data.set('CONFIG_GLUSTERFS_XLATOR_OPT', glusterfs.version().version_compare('>=4'))
config_host_data.set('CONFIG_GLUSTERFS_DISCARD', glusterfs.version().version_compare('>=5'))
config_host_data.set('CONFIG_GLUSTERFS_FALLOCATE', glusterfs.version().version_compare('>=6'))
config_host_data.set('CONFIG_GLUSTERFS_ZEROFILL', glusterfs.version().version_compare('>=6'))
config_host_data.set('CONFIG_GLUSTERFS_FTRUNCATE_HAS_STAT', glusterfs_ftruncate_has_stat)
config_host_data.set('CONFIG_GLUSTERFS_IOCB_HAS_STAT', glusterfs_iocb_has_stat)
endif
config_host_data.set('CONFIG_GTK', gtk.found())
config_host_data.set('CONFIG_VTE', vte.found())
config_host_data.set('CONFIG_GTK_CLIPBOARD', have_gtk_clipboard)
config_host_data.set('CONFIG_HEXAGON_IDEF_PARSER', get_option('hexagon_idef_parser'))
config_host_data.set('CONFIG_LIBATTR', have_old_libattr)
config_host_data.set('CONFIG_LIBCAP_NG', libcap_ng.found())
config_host_data.set('CONFIG_EBPF', libbpf.found())
net: add initial support for AF_XDP network backend AF_XDP is a network socket family that allows communication directly with the network device driver in the kernel, bypassing most or all of the kernel networking stack. In the essence, the technology is pretty similar to netmap. But, unlike netmap, AF_XDP is Linux-native and works with any network interfaces without driver modifications. Unlike vhost-based backends (kernel, user, vdpa), AF_XDP doesn't require access to character devices or unix sockets. Only access to the network interface itself is necessary. This patch implements a network backend that communicates with the kernel by creating an AF_XDP socket. A chunk of userspace memory is shared between QEMU and the host kernel. 4 ring buffers (Tx, Rx, Fill and Completion) are placed in that memory along with a pool of memory buffers for the packet data. Data transmission is done by allocating one of the buffers, copying packet data into it and placing the pointer into Tx ring. After transmission, device will return the buffer via Completion ring. On Rx, device will take a buffer form a pre-populated Fill ring, write the packet data into it and place the buffer into Rx ring. AF_XDP network backend takes on the communication with the host kernel and the network interface and forwards packets to/from the peer device in QEMU. Usage example: -device virtio-net-pci,netdev=guest1,mac=00:16:35:AF:AA:5C -netdev af-xdp,ifname=ens6f1np1,id=guest1,mode=native,queues=1 XDP program bridges the socket with a network interface. It can be attached to the interface in 2 different modes: 1. skb - this mode should work for any interface and doesn't require driver support. With a caveat of lower performance. 2. native - this does require support from the driver and allows to bypass skb allocation in the kernel and potentially use zero-copy while getting packets in/out userspace. By default, QEMU will try to use native mode and fall back to skb. Mode can be forced via 'mode' option. To force 'copy' even in native mode, use 'force-copy=on' option. This might be useful if there is some issue with the driver. Option 'queues=N' allows to specify how many device queues should be open. Note that all the queues that are not open are still functional and can receive traffic, but it will not be delivered to QEMU. So, the number of device queues should generally match the QEMU configuration, unless the device is shared with something else and the traffic re-direction to appropriate queues is correctly configured on a device level (e.g. with ethtool -N). 'start-queue=M' option can be used to specify from which queue id QEMU should start configuring 'N' queues. It might also be necessary to use this option with certain NICs, e.g. MLX5 NICs. See the docs for examples. In a general case QEMU will need CAP_NET_ADMIN and CAP_SYS_ADMIN or CAP_BPF capabilities in order to load default XSK/XDP programs to the network interface and configure BPF maps. It is possible, however, to run with no capabilities. For that to work, an external process with enough capabilities will need to pre-load default XSK program, create AF_XDP sockets and pass their file descriptors to QEMU process on startup via 'sock-fds' option. Network backend will need to be configured with 'inhibit=on' to avoid loading of the program. QEMU will need 32 MB of locked memory (RLIMIT_MEMLOCK) per queue or CAP_IPC_LOCK. There are few performance challenges with the current network backends. First is that they do not support IO threads. This means that data path is handled by the main thread in QEMU and may slow down other work or may be slowed down by some other work. This also means that taking advantage of multi-queue is generally not possible today. Another thing is that data path is going through the device emulation code, which is not really optimized for performance. The fastest "frontend" device is virtio-net. But it's not optimized for heavy traffic either, because it expects such use-cases to be handled via some implementation of vhost (user, kernel, vdpa). In practice, we have virtio notifications and rcu lock/unlock on a per-packet basis and not very efficient accesses to the guest memory. Communication channels between backend and frontend devices do not allow passing more than one packet at a time as well. Some of these challenges can be avoided in the future by adding better batching into device emulation or by implementing vhost-af-xdp variant. There are also a few kernel limitations. AF_XDP sockets do not support any kinds of checksum or segmentation offloading. Buffers are limited to a page size (4K), i.e. MTU is limited. Multi-buffer support implementation for AF_XDP is in progress, but not ready yet. Also, transmission in all non-zero-copy modes is synchronous, i.e. done in a syscall. That doesn't allow high packet rates on virtual interfaces. However, keeping in mind all of these challenges, current implementation of the AF_XDP backend shows a decent performance while running on top of a physical NIC with zero-copy support. Test setup: 2 VMs running on 2 physical hosts connected via ConnectX6-Dx card. Network backend is configured to open the NIC directly in native mode. The driver supports zero-copy. NIC is configured to use 1 queue. Inside a VM - iperf3 for basic TCP performance testing and dpdk-testpmd for PPS testing. iperf3 result: TCP stream : 19.1 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 3.4 Mpps Rx only : 2.0 Mpps L2 FWD Loopback : 1.5 Mpps In skb mode the same setup shows much lower performance, similar to the setup where pair of physical NICs is replaced with veth pair: iperf3 result: TCP stream : 9 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 1.2 Mpps Rx only : 1.0 Mpps L2 FWD Loopback : 0.7 Mpps Results in skb mode or over the veth are close to results of a tap backend with vhost=on and disabled segmentation offloading bridged with a NIC. Signed-off-by: Ilya Maximets <i.maximets@ovn.org> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> (docker/lcitool) Signed-off-by: Jason Wang <jasowang@redhat.com>
2023-09-13 21:34:37 +03:00
config_host_data.set('CONFIG_AF_XDP', libxdp.found())
config_host_data.set('CONFIG_LIBDAXCTL', libdaxctl.found())
config_host_data.set('CONFIG_LIBISCSI', libiscsi.found())
config_host_data.set('CONFIG_LIBNFS', libnfs.found())
config_host_data.set('CONFIG_LIBSSH', libssh.found())
config_host_data.set('CONFIG_LINUX_AIO', libaio.found())
config_host_data.set('CONFIG_LINUX_IO_URING', linux_io_uring.found())
config_host_data.set('CONFIG_LIBPMEM', libpmem.found())
config_host_data.set('CONFIG_MODULES', enable_modules)
config_host_data.set('CONFIG_NUMA', numa.found())
if numa.found()
config_host_data.set('HAVE_NUMA_HAS_PREFERRED_MANY',
cc.has_function('numa_has_preferred_many',
dependencies: numa))
endif
config_host_data.set('CONFIG_OPENGL', opengl.found())
config_host_data.set('CONFIG_PLUGIN', get_option('plugins'))
config_host_data.set('CONFIG_RBD', rbd.found())
config_host_data.set('CONFIG_RDMA', rdma.found())
meson, cutils: allow non-relocatable installs Say QEMU is configured with bindir = "/usr/bin" and a firmware path that starts with "/usr/share/qemu". Ever since QEMU 5.2, QEMU's install has been relocatable: if you move qemu-system-x86_64 from /usr/bin to /home/username/bin, it will start looking for firmware in /home/username/share/qemu. Previously, you would get a non-relocatable install where the moved QEMU will keep looking for firmware in /usr/share/qemu. Windows almost always wants relocatable installs, and in fact that is why QEMU 5.2 introduced relocatability in the first place. However, newfangled distribution mechanisms such as AppImage (https://docs.appimage.org/reference/best-practices.html), and possibly NixOS, also dislike using at runtime the absolute paths that were established at build time. On POSIX systems you almost never care; if you do, your usecase dictates which one is desirable, so there's no single answer. Obviously relocatability works fine most of the time, because not many people have complained about QEMU's switch to relocatable install, and that's why until now there was no way to disable relocatability. But a non-relocatable, non-modular binary can help if you want to do experiments with old firmware and new QEMU or vice versa (because you can just upgrade/downgrade the firmware package, and use rpm2cpio or similar to extract the QEMU binaries outside /usr), so allow both. This patch allows one to build a non-relocatable install using a new option to configure. Why? Because it's not too hard, and because it helps the user double check the relocatability of their install. Note that the same code that handles relocation also lets you run QEMU from the build tree and pick e.g. firmware files from the source tree transparently. Therefore that part remains active with this patch, even if you configure with --disable-relocatable. Suggested-by: Michael Tokarev <mjt@tls.msk.ru> Reviewed-by: Emmanouil Pitsidianakis <manos.pitsidianakis@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-10-05 15:19:34 +03:00
config_host_data.set('CONFIG_RELOCATABLE', get_option('relocatable'))
config_host_data.set('CONFIG_SAFESTACK', get_option('safe_stack'))
config_host_data.set('CONFIG_SDL', sdl.found())
config_host_data.set('CONFIG_SDL_IMAGE', sdl_image.found())
config_host_data.set('CONFIG_SECCOMP', seccomp.found())
if seccomp.found()
config_host_data.set('CONFIG_SECCOMP_SYSRAWRC', seccomp_has_sysrawrc)
endif
config_host_data.set('CONFIG_PIXMAN', pixman.found())
config_host_data.set('CONFIG_SLIRP', slirp.found())
config_host_data.set('CONFIG_SNAPPY', snappy.found())
config_host_data.set('CONFIG_SOLARIS', host_os == 'sunos')
if get_option('tcg').allowed()
config_host_data.set('CONFIG_TCG', 1)
config_host_data.set('CONFIG_TCG_INTERPRETER', tcg_arch == 'tci')
endif
config_host_data.set('CONFIG_TPM', have_tpm)
config_host_data.set('CONFIG_TSAN', get_option('tsan'))
config_host_data.set('CONFIG_USB_LIBUSB', libusb.found())
config_host_data.set('CONFIG_VDE', vde.found())
config_host_data.set('CONFIG_VHOST', have_vhost)
config_host_data.set('CONFIG_VHOST_NET', have_vhost_net)
config_host_data.set('CONFIG_VHOST_NET_USER', have_vhost_net_user)
config_host_data.set('CONFIG_VHOST_NET_VDPA', have_vhost_net_vdpa)
config_host_data.set('CONFIG_VHOST_KERNEL', have_vhost_kernel)
config_host_data.set('CONFIG_VHOST_USER', have_vhost_user)
config_host_data.set('CONFIG_VHOST_CRYPTO', have_vhost_user_crypto)
config_host_data.set('CONFIG_VHOST_VDPA', have_vhost_vdpa)
config_host_data.set('CONFIG_VMNET', vmnet.found())
config_host_data.set('CONFIG_VHOST_USER_BLK_SERVER', have_vhost_user_blk_server)
config_host_data.set('CONFIG_VDUSE_BLK_EXPORT', have_vduse_blk_export)
config_host_data.set('CONFIG_PNG', png.found())
config_host_data.set('CONFIG_VNC', vnc.found())
config_host_data.set('CONFIG_VNC_JPEG', jpeg.found())
config_host_data.set('CONFIG_VNC_SASL', sasl.found())
if virgl.found()
config_host_data.set('HAVE_VIRGL_D3D_INFO_EXT',
cc.has_member('struct virgl_renderer_resource_info_ext', 'd3d_tex2d',
prefix: '#include <virglrenderer.h>',
dependencies: virgl))
endif
config_host_data.set('CONFIG_VIRTFS', have_virtfs)
config_host_data.set('CONFIG_VTE', vte.found())
config_host_data.set('CONFIG_XKBCOMMON', xkbcommon.found())
config_host_data.set('CONFIG_KEYUTILS', keyutils.found())
config_host_data.set('CONFIG_GETTID', has_gettid)
config_host_data.set('CONFIG_GNUTLS', gnutls.found())
config_host_data.set('CONFIG_GNUTLS_CRYPTO', gnutls_crypto.found())
config_host_data.set('CONFIG_TASN1', tasn1.found())
config_host_data.set('CONFIG_GCRYPT', gcrypt.found())
config_host_data.set('CONFIG_NETTLE', nettle.found())
config_host_data.set('CONFIG_CRYPTO_SM4', crypto_sm4.found())
config_host_data.set('CONFIG_HOGWEED', hogweed.found())
config_host_data.set('CONFIG_QEMU_PRIVATE_XTS', xts == 'private')
config_host_data.set('CONFIG_MALLOC_TRIM', has_malloc_trim)
config_host_data.set('CONFIG_STATX', has_statx)
config_host_data.set('CONFIG_STATX_MNT_ID', has_statx_mnt_id)
config_host_data.set('CONFIG_ZSTD', zstd.found())
config_host_data.set('CONFIG_QPL', qpl.found())
config_host_data.set('CONFIG_UADK', uadk.found())
config_host_data.set('CONFIG_QATZIP', qatzip.found())
config_host_data.set('CONFIG_FUSE', fuse.found())
config_host_data.set('CONFIG_FUSE_LSEEK', fuse_lseek.found())
config_host_data.set('CONFIG_SPICE_PROTOCOL', spice_protocol.found())
if spice_protocol.found()
config_host_data.set('CONFIG_SPICE_PROTOCOL_MAJOR', spice_protocol.version().split('.')[0])
config_host_data.set('CONFIG_SPICE_PROTOCOL_MINOR', spice_protocol.version().split('.')[1])
config_host_data.set('CONFIG_SPICE_PROTOCOL_MICRO', spice_protocol.version().split('.')[2])
endif
config_host_data.set('CONFIG_SPICE', spice.found())
config_host_data.set('CONFIG_X11', x11.found())
config_host_data.set('CONFIG_DBUS_DISPLAY', dbus_display)
config_host_data.set('CONFIG_CFI', get_option('cfi'))
nbd/server: Add --selinux-label option Under SELinux, Unix domain sockets have two labels. One is on the disk and can be set with commands such as chcon(1). There is a different label stored in memory (called the process label). This can only be set by the process creating the socket. When using SELinux + SVirt and wanting qemu to be able to connect to a qemu-nbd instance, you must set both labels correctly first. For qemu-nbd the options to set the second label are awkward. You can create the socket in a wrapper program and then exec into qemu-nbd. Or you could try something with LD_PRELOAD. This commit adds the ability to set the label straightforwardly on the command line, via the new --selinux-label flag. (The name of the flag is the same as the equivalent nbdkit option.) A worked example showing how to use the new option can be found in this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Signed-off-by: Richard W.M. Jones <rjones@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> [eblake: rebase to configure changes, reject --selinux-label if it is not compiled in or not used on a Unix socket] Note that we may relax some of these restrictions at a later date, such as making it possible to label a TCP socket, although it may be smarter to do so as a generic QMP action rather than more one-off command lines in qemu-nbd. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20211115202944.615966-1-eblake@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> [eblake: adjust meson output as suggested by thuth] Signed-off-by: Eric Blake <eblake@redhat.com>
2021-11-15 23:29:43 +03:00
config_host_data.set('CONFIG_SELINUX', selinux.found())
config_host_data.set('CONFIG_XEN_BACKEND', xen.found())
config_host_data.set('CONFIG_LIBDW', libdw.found())
if xen.found()
# protect from xen.version() having less than three components
xen_version = xen.version().split('.') + ['0', '0']
xen_ctrl_version = xen_version[0] + \
('0' + xen_version[1]).substring(-2) + \
('0' + xen_version[2]).substring(-2)
config_host_data.set('CONFIG_XEN_CTRL_INTERFACE_VERSION', xen_ctrl_version)
endif
config_host_data.set('QEMU_VERSION', '"@0@"'.format(meson.project_version()))
config_host_data.set('QEMU_VERSION_MAJOR', meson.project_version().split('.')[0])
config_host_data.set('QEMU_VERSION_MINOR', meson.project_version().split('.')[1])
config_host_data.set('QEMU_VERSION_MICRO', meson.project_version().split('.')[2])
config_host_data.set_quoted('CONFIG_HOST_DSOSUF', host_dsosuf)
config_host_data.set('HAVE_HOST_BLOCK_DEVICE', have_host_block_device)
have_coroutine_pool = get_option('coroutine_pool')
if get_option('debug_stack_usage') and have_coroutine_pool
message('Disabling coroutine pool to measure stack usage')
have_coroutine_pool = false
endif
config_host_data.set('CONFIG_COROUTINE_POOL', have_coroutine_pool)
config_host_data.set('CONFIG_DEBUG_GRAPH_LOCK', get_option('debug_graph_lock'))
config_host_data.set('CONFIG_DEBUG_MUTEX', get_option('debug_mutex'))
config_host_data.set('CONFIG_DEBUG_STACK_USAGE', get_option('debug_stack_usage'))
config_host_data.set('CONFIG_DEBUG_TCG', get_option('debug_tcg'))
config_host_data.set('CONFIG_DEBUG_REMAP', get_option('debug_remap'))
config_host_data.set('CONFIG_QOM_CAST_DEBUG', get_option('qom_cast_debug'))
config_host_data.set('CONFIG_REPLICATION', get_option('replication').allowed())
config_host_data.set('CONFIG_FSFREEZE', qga_fsfreeze)
config_host_data.set('CONFIG_FSTRIM', qga_fstrim)
# has_header
config_host_data.set('CONFIG_EPOLL', cc.has_header('sys/epoll.h'))
config_host_data.set('CONFIG_LINUX_MAGIC_H', cc.has_header('linux/magic.h'))
config_host_data.set('CONFIG_VALGRIND_H', cc.has_header('valgrind/valgrind.h'))
config_host_data.set('HAVE_BTRFS_H', cc.has_header('linux/btrfs.h'))
config_host_data.set('HAVE_DRM_H', cc.has_header('libdrm/drm.h'))
config_host_data.set('HAVE_PTY_H', cc.has_header('pty.h'))
config_host_data.set('HAVE_SYS_DISK_H', cc.has_header('sys/disk.h'))
config_host_data.set('HAVE_SYS_IOCCOM_H', cc.has_header('sys/ioccom.h'))
config_host_data.set('HAVE_SYS_KCOV_H', cc.has_header('sys/kcov.h'))
if host_os == 'windows'
config_host_data.set('HAVE_AFUNIX_H', cc.has_header('afunix.h'))
endif
# has_function
os-posix: asynchronous teardown for shutdown on Linux This patch adds support for asynchronously tearing down a VM on Linux. When qemu terminates, either naturally or because of a fatal signal, the VM is torn down. If the VM is huge, it can take a considerable amount of time for it to be cleaned up. In case of a protected VM, it might take even longer than a non-protected VM (this is the case on s390x, for example). Some users might want to shut down a VM and restart it immediately, without having to wait. This is especially true if management infrastructure like libvirt is used. This patch implements a simple trick on Linux to allow qemu to return immediately, with the teardown of the VM being performed asynchronously. If the new commandline option -async-teardown is used, a new process is spawned from qemu at startup, using the clone syscall, in such way that it will share its address space with qemu.The new process will have the name "cleanup/<QEMU_PID>". It will wait until qemu terminates completely, and then it will exit itself. This allows qemu to terminate quickly, without having to wait for the whole address space to be torn down. The cleanup process will exit after qemu, so it will be the last user of the address space, and therefore it will take care of the actual teardown. The cleanup process will share the same cgroups as qemu, so both memory usage and cpu time will be accounted properly. If possible, close_range will be used in the cleanup process to close all open file descriptors. If it is not available or if it fails, /proc will be used to determine which file descriptors to close. If the cleanup process is forcefully killed with SIGKILL before the main qemu process has terminated completely, the mechanism is defeated and the teardown will not be asynchronous. This feature can already be used with libvirt by adding the following to the XML domain definition to pass the parameter to qemu directly: <commandline xmlns="http://libvirt.org/schemas/domain/qemu/1.0"> <arg value='-async-teardown'/> </commandline> Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com> Reviewed-by: Murilo Opsfelder Araujo <muriloo@linux.ibm.com> Tested-by: Murilo Opsfelder Araujo <muriloo@linux.ibm.com> Message-Id: <20220812133453.82671-1-imbrenda@linux.ibm.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2022-08-12 16:34:53 +03:00
config_host_data.set('CONFIG_CLOSE_RANGE', cc.has_function('close_range'))
config_host_data.set('CONFIG_ACCEPT4', cc.has_function('accept4'))
config_host_data.set('CONFIG_CLOCK_ADJTIME', cc.has_function('clock_adjtime'))
config_host_data.set('CONFIG_DUP3', cc.has_function('dup3'))
config_host_data.set('CONFIG_FALLOCATE', cc.has_function('fallocate'))
config_host_data.set('CONFIG_POSIX_FALLOCATE', cc.has_function('posix_fallocate'))
config_host_data.set('CONFIG_GETCPU', cc.has_function('getcpu', prefix: gnu_source_prefix))
config_host_data.set('CONFIG_SCHED_GETCPU', cc.has_function('sched_getcpu', prefix: '#include <sched.h>'))
# Note that we need to specify prefix: here to avoid incorrectly
# thinking that Windows has posix_memalign()
config_host_data.set('CONFIG_POSIX_MEMALIGN', cc.has_function('posix_memalign', prefix: '#include <stdlib.h>'))
config_host_data.set('CONFIG_ALIGNED_MALLOC', cc.has_function('_aligned_malloc'))
config_host_data.set('CONFIG_VALLOC', cc.has_function('valloc'))
config_host_data.set('CONFIG_MEMALIGN', cc.has_function('memalign'))
config_host_data.set('CONFIG_PPOLL', cc.has_function('ppoll'))
configure: Move preadv check to meson.build Move the preadv availability check to meson.build. This is what we want to be doing for host-OS-feature-checks anyway, but it also fixes a problem with building for macOS with the most recent XCode SDK on a Catalina host. On that configuration, 'preadv()' is provided as a weak symbol, so that programs can be built with optional support for it and make a runtime availability check to see whether the preadv() they have is a working one or one which they must not call because it will runtime-assert. QEMU's configure test passes (unless you're building with --enable-werror) because the test program using preadv() compiles, but then QEMU crashes at runtime when preadv() is called, with errors like: dyld: lazy symbol binding failed: Symbol not found: _preadv Referenced from: /Users/pm215/src/qemu/./build/x86/tests/test-replication Expected in: /usr/lib/libSystem.B.dylib dyld: Symbol not found: _preadv Referenced from: /Users/pm215/src/qemu/./build/x86/tests/test-replication Expected in: /usr/lib/libSystem.B.dylib Meson's own function availability check has a special case for macOS which adds '-Wl,-no_weak_imports' to the compiler flags, which forces the test to require the real function, not the macOS-version-too-old stub. So this commit fixes the bug where macOS builds on Catalina currently require --disable-werror. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Acked-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Message-id: 20210126155846.17109-1-peter.maydell@linaro.org
2021-01-26 18:58:46 +03:00
config_host_data.set('CONFIG_PREADV', cc.has_function('preadv', prefix: '#include <sys/uio.h>'))
config_host_data.set('CONFIG_PTHREAD_FCHDIR_NP', cc.has_function('pthread_fchdir_np'))
config_host_data.set('CONFIG_SENDFILE', cc.has_function('sendfile'))
config_host_data.set('CONFIG_SETNS', cc.has_function('setns') and cc.has_function('unshare'))
config_host_data.set('CONFIG_SYNCFS', cc.has_function('syncfs'))
config_host_data.set('CONFIG_SYNC_FILE_RANGE', cc.has_function('sync_file_range'))
config_host_data.set('CONFIG_TIMERFD', cc.has_function('timerfd_create'))
config_host_data.set('HAVE_COPY_FILE_RANGE', cc.has_function('copy_file_range'))
config_host_data.set('HAVE_GETIFADDRS', cc.has_function('getifaddrs'))
config_host_data.set('HAVE_GLIB_WITH_SLICE_ALLOCATOR', glib_has_gslice)
config_host_data.set('HAVE_OPENPTY', cc.has_function('openpty', dependencies: util))
config_host_data.set('HAVE_STRCHRNUL', cc.has_function('strchrnul'))
config_host_data.set('HAVE_SYSTEM_FUNCTION', cc.has_function('system', prefix: '#include <stdlib.h>'))
if rbd.found()
config_host_data.set('HAVE_RBD_NAMESPACE_EXISTS',
cc.has_function('rbd_namespace_exists',
dependencies: rbd,
prefix: '#include <rbd/librbd.h>'))
endif
if rdma.found()
config_host_data.set('HAVE_IBV_ADVISE_MR',
cc.has_function('ibv_advise_mr',
dependencies: rdma,
prefix: '#include <infiniband/verbs.h>'))
endif
configure: Move preadv check to meson.build Move the preadv availability check to meson.build. This is what we want to be doing for host-OS-feature-checks anyway, but it also fixes a problem with building for macOS with the most recent XCode SDK on a Catalina host. On that configuration, 'preadv()' is provided as a weak symbol, so that programs can be built with optional support for it and make a runtime availability check to see whether the preadv() they have is a working one or one which they must not call because it will runtime-assert. QEMU's configure test passes (unless you're building with --enable-werror) because the test program using preadv() compiles, but then QEMU crashes at runtime when preadv() is called, with errors like: dyld: lazy symbol binding failed: Symbol not found: _preadv Referenced from: /Users/pm215/src/qemu/./build/x86/tests/test-replication Expected in: /usr/lib/libSystem.B.dylib dyld: Symbol not found: _preadv Referenced from: /Users/pm215/src/qemu/./build/x86/tests/test-replication Expected in: /usr/lib/libSystem.B.dylib Meson's own function availability check has a special case for macOS which adds '-Wl,-no_weak_imports' to the compiler flags, which forces the test to require the real function, not the macOS-version-too-old stub. So this commit fixes the bug where macOS builds on Catalina currently require --disable-werror. Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Acked-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> Message-id: 20210126155846.17109-1-peter.maydell@linaro.org
2021-01-26 18:58:46 +03:00
have_asan_fiber = false
if get_option('asan') and \
not cc.has_function('__sanitizer_start_switch_fiber',
args: '-fsanitize=address',
prefix: '#include <sanitizer/asan_interface.h>')
warning('Missing ASAN due to missing fiber annotation interface')
warning('Without code annotation, the report may be inferior.')
else
have_asan_fiber = true
endif
config_host_data.set('CONFIG_ASAN_IFACE_FIBER', have_asan_fiber)
have_inotify_init = cc.has_header_symbol('sys/inotify.h', 'inotify_init')
have_inotify_init1 = cc.has_header_symbol('sys/inotify.h', 'inotify_init1')
inotify = not_found
if (have_inotify_init or have_inotify_init1) and host_os == 'freebsd'
# libinotify-kqueue
inotify = cc.find_library('inotify')
if have_inotify_init
have_inotify_init = inotify.found()
endif
if have_inotify_init1
have_inotify_init1 = inotify.found()
endif
endif
config_host_data.set('CONFIG_INOTIFY', have_inotify_init)
config_host_data.set('CONFIG_INOTIFY1', have_inotify_init1)
# has_header_symbol
config_host_data.set('CONFIG_BLKZONED',
cc.has_header_symbol('linux/blkzoned.h', 'BLKOPENZONE'))
config_host_data.set('CONFIG_EPOLL_CREATE1',
cc.has_header_symbol('sys/epoll.h', 'epoll_create1'))
config_host_data.set('CONFIG_FALLOCATE_PUNCH_HOLE',
cc.has_header_symbol('linux/falloc.h', 'FALLOC_FL_PUNCH_HOLE') and
cc.has_header_symbol('linux/falloc.h', 'FALLOC_FL_KEEP_SIZE'))
config_host_data.set('CONFIG_FALLOCATE_ZERO_RANGE',
cc.has_header_symbol('linux/falloc.h', 'FALLOC_FL_ZERO_RANGE'))
config_host_data.set('CONFIG_FIEMAP',
cc.has_header('linux/fiemap.h') and
cc.has_header_symbol('linux/fs.h', 'FS_IOC_FIEMAP'))
config_host_data.set('CONFIG_GETRANDOM',
cc.has_function('getrandom') and
cc.has_header_symbol('sys/random.h', 'GRND_NONBLOCK'))
config_host_data.set('CONFIG_PRCTL_PR_SET_TIMERSLACK',
cc.has_header_symbol('sys/prctl.h', 'PR_SET_TIMERSLACK'))
config_host_data.set('CONFIG_RTNETLINK',
cc.has_header_symbol('linux/rtnetlink.h', 'IFLA_PROTO_DOWN'))
config_host_data.set('CONFIG_SYSMACROS',
cc.has_header_symbol('sys/sysmacros.h', 'makedev'))
config_host_data.set('HAVE_OPTRESET',
cc.has_header_symbol('getopt.h', 'optreset'))
config_host_data.set('HAVE_IPPROTO_MPTCP',
cc.has_header_symbol('netinet/in.h', 'IPPROTO_MPTCP'))
# has_member
config_host_data.set('HAVE_SIGEV_NOTIFY_THREAD_ID',
cc.has_member('struct sigevent', 'sigev_notify_thread_id',
prefix: '#include <signal.h>'))
config_host_data.set('HAVE_STRUCT_STAT_ST_ATIM',
cc.has_member('struct stat', 'st_atim',
prefix: '#include <sys/stat.h>'))
config_host_data.set('HAVE_BLK_ZONE_REP_CAPACITY',
cc.has_member('struct blk_zone', 'capacity',
prefix: '#include <linux/blkzoned.h>'))
# has_type
config_host_data.set('CONFIG_IOVEC',
cc.has_type('struct iovec',
prefix: '#include <sys/uio.h>'))
config_host_data.set('HAVE_UTMPX',
cc.has_type('struct utmpx',
prefix: '#include <utmpx.h>'))
config_host_data.set('CONFIG_EVENTFD', cc.links('''
#include <sys/eventfd.h>
int main(void) { return eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC); }'''))
config_host_data.set('CONFIG_FDATASYNC', cc.links(gnu_source_prefix + '''
#include <unistd.h>
int main(void) {
#if defined(_POSIX_SYNCHRONIZED_IO) && _POSIX_SYNCHRONIZED_IO > 0
return fdatasync(0);
#else
#error Not supported
#endif
}'''))
has_madvise = cc.links(gnu_source_prefix + '''
#include <sys/types.h>
#include <sys/mman.h>
#include <stddef.h>
int main(void) { return madvise(NULL, 0, MADV_DONTNEED); }''')
missing_madvise_proto = false
if has_madvise
# Some platforms (illumos and Solaris before Solaris 11) provide madvise()
# but forget to prototype it. In this case, has_madvise will be true (the
# test program links despite a compile warning). To detect the
# missing-prototype case, we try again with a definitely-bogus prototype.
# This will only compile if the system headers don't provide the prototype;
# otherwise the conflicting prototypes will cause a compiler error.
missing_madvise_proto = cc.links(gnu_source_prefix + '''
#include <sys/types.h>
#include <sys/mman.h>
#include <stddef.h>
extern int madvise(int);
int main(void) { return madvise(0); }''')
endif
config_host_data.set('CONFIG_MADVISE', has_madvise)
config_host_data.set('HAVE_MADVISE_WITHOUT_PROTOTYPE', missing_madvise_proto)
config_host_data.set('CONFIG_MEMFD', cc.links(gnu_source_prefix + '''
#include <sys/mman.h>
int main(void) { return memfd_create("foo", MFD_ALLOW_SEALING); }'''))
config_host_data.set('CONFIG_OPEN_BY_HANDLE', cc.links(gnu_source_prefix + '''
#include <fcntl.h>
#if !defined(AT_EMPTY_PATH)
# error missing definition
#else
int main(void) { struct file_handle fh; return open_by_handle_at(0, &fh, 0); }
#endif'''))
# On Darwin posix_madvise() has the same return semantics as plain madvise(),
# i.e. errno is set and -1 is returned. That's not really how POSIX defines the
# function. On the flip side, it has madvise() which is preferred anyways.
if host_os != 'darwin'
config_host_data.set('CONFIG_POSIX_MADVISE', cc.links(gnu_source_prefix + '''
#include <sys/mman.h>
#include <stddef.h>
int main(void) { return posix_madvise(NULL, 0, POSIX_MADV_DONTNEED); }'''))
endif
config_host_data.set('CONFIG_PTHREAD_SETNAME_NP_W_TID', cc.links(gnu_source_prefix + '''
#include <pthread.h>
static void *f(void *p) { return NULL; }
int main(void)
{
pthread_t thread;
pthread_create(&thread, 0, f, 0);
pthread_setname_np(thread, "QEMU");
return 0;
}''', dependencies: threads))
config_host_data.set('CONFIG_PTHREAD_SETNAME_NP_WO_TID', cc.links(gnu_source_prefix + '''
#include <pthread.h>
static void *f(void *p) { pthread_setname_np("QEMU"); return NULL; }
int main(void)
{
pthread_t thread;
pthread_create(&thread, 0, f, 0);
return 0;
}''', dependencies: threads))
config_host_data.set('CONFIG_PTHREAD_SET_NAME_NP', cc.links(gnu_source_prefix + '''
#include <pthread.h>
#include <pthread_np.h>
static void *f(void *p) { return NULL; }
int main(void)
{
pthread_t thread;
pthread_create(&thread, 0, f, 0);
pthread_set_name_np(thread, "QEMU");
return 0;
}''', dependencies: threads))
config_host_data.set('CONFIG_PTHREAD_CONDATTR_SETCLOCK', cc.links(gnu_source_prefix + '''
#include <pthread.h>
#include <time.h>
int main(void)
{
pthread_condattr_t attr
pthread_condattr_init(&attr);
pthread_condattr_setclock(&attr, CLOCK_MONOTONIC);
return 0;
}''', dependencies: threads))
config_host_data.set('CONFIG_PTHREAD_AFFINITY_NP', cc.links(gnu_source_prefix + '''
#include <pthread.h>
static void *f(void *p) { return NULL; }
int main(void)
{
int setsize = CPU_ALLOC_SIZE(64);
pthread_t thread;
cpu_set_t *cpuset;
pthread_create(&thread, 0, f, 0);
cpuset = CPU_ALLOC(64);
CPU_ZERO_S(setsize, cpuset);
pthread_setaffinity_np(thread, setsize, cpuset);
pthread_getaffinity_np(thread, setsize, cpuset);
CPU_FREE(cpuset);
return 0;
}''', dependencies: threads))
config_host_data.set('CONFIG_SIGNALFD', cc.links(gnu_source_prefix + '''
#include <sys/signalfd.h>
#include <stddef.h>
int main(void) { return signalfd(-1, NULL, SFD_CLOEXEC); }'''))
config_host_data.set('CONFIG_SPLICE', cc.links(gnu_source_prefix + '''
#include <unistd.h>
#include <fcntl.h>
#include <limits.h>
int main(void)
{
int len, fd = 0;
len = tee(STDIN_FILENO, STDOUT_FILENO, INT_MAX, SPLICE_F_NONBLOCK);
splice(STDIN_FILENO, NULL, fd, NULL, len, SPLICE_F_MOVE);
return 0;
}'''))
config_host_data.set('HAVE_MLOCKALL', cc.links(gnu_source_prefix + '''
#include <sys/mman.h>
int main(void) {
return mlockall(MCL_FUTURE);
}'''))
have_l2tpv3 = false
if get_option('l2tpv3').allowed() and have_system
have_l2tpv3 = cc.has_type('struct mmsghdr',
prefix: gnu_source_prefix + '''
#include <sys/socket.h>
#include <linux/ip.h>''')
endif
config_host_data.set('CONFIG_L2TPV3', have_l2tpv3)
have_netmap = false
if get_option('netmap').allowed() and have_system
have_netmap = cc.compiles('''
#include <inttypes.h>
#include <net/if.h>
#include <net/netmap.h>
#include <net/netmap_user.h>
#if (NETMAP_API < 11) || (NETMAP_API > 15)
#error
#endif
int main(void) { return 0; }''')
if not have_netmap and get_option('netmap').enabled()
error('Netmap headers not available')
endif
endif
config_host_data.set('CONFIG_NETMAP', have_netmap)
# Work around a system header bug with some kernel/XFS header
# versions where they both try to define 'struct fsxattr':
# xfs headers will not try to redefine structs from linux headers
# if this macro is set.
config_host_data.set('HAVE_FSXATTR', cc.links('''
#include <linux/fs.h>
struct fsxattr foo;
int main(void) {
return 0;
}'''))
# Some versions of Mac OS X incorrectly define SIZE_MAX
config_host_data.set('HAVE_BROKEN_SIZE_MAX', not cc.compiles('''
#include <stdint.h>
#include <stdio.h>
int main(void) {
return printf("%zu", SIZE_MAX);
}''', args: ['-Werror']))
# See if 64-bit atomic operations are supported.
# Note that without __atomic builtins, we can only
# assume atomic loads/stores max at pointer size.
config_host_data.set('CONFIG_ATOMIC64', cc.links('''
#include <stdint.h>
int main(void)
{
uint64_t x = 0, y = 0;
y = __atomic_load_n(&x, __ATOMIC_RELAXED);
__atomic_store_n(&x, y, __ATOMIC_RELAXED);
__atomic_compare_exchange_n(&x, &y, x, 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
__atomic_exchange_n(&x, y, __ATOMIC_RELAXED);
__atomic_fetch_add(&x, y, __ATOMIC_RELAXED);
return 0;
}'''))
has_int128_type = cc.compiles('''
__int128_t a;
__uint128_t b;
int main(void) { b = a; }''')
config_host_data.set('CONFIG_INT128_TYPE', has_int128_type)
has_int128 = has_int128_type and cc.links('''
__int128_t a;
__uint128_t b;
int main (void) {
a = a + b;
b = a * b;
a = a * a;
return 0;
}''')
config_host_data.set('CONFIG_INT128', has_int128)
if has_int128_type
# "do we have 128-bit atomics which are handled inline and specifically not
# via libatomic". The reason we can't use libatomic is documented in the
# comment starting "GCC is a house divided" in include/qemu/atomic128.h.
# We only care about these operations on 16-byte aligned pointers, so
# force 16-byte alignment of the pointer, which may be greater than
# __alignof(unsigned __int128) for the host.
atomic_test_128 = '''
int main(int ac, char **av) {
__uint128_t *p = __builtin_assume_aligned(av[ac - 1], 16);
p[1] = __atomic_load_n(&p[0], __ATOMIC_RELAXED);
__atomic_store_n(&p[2], p[3], __ATOMIC_RELAXED);
__atomic_compare_exchange_n(&p[4], &p[5], p[6], 0, __ATOMIC_RELAXED, __ATOMIC_RELAXED);
return 0;
}'''
has_atomic128 = cc.links(atomic_test_128)
config_host_data.set('CONFIG_ATOMIC128', has_atomic128)
if not has_atomic128
# Even with __builtin_assume_aligned, the above test may have failed
# without optimization enabled. Try again with optimizations locally
# enabled for the function. See
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107389
has_atomic128_opt = cc.links('__attribute__((optimize("O1")))' + atomic_test_128)
config_host_data.set('CONFIG_ATOMIC128_OPT', has_atomic128_opt)
if not has_atomic128_opt
config_host_data.set('CONFIG_CMPXCHG128', cc.links('''
int main(void)
{
__uint128_t x = 0, y = 0;
__sync_val_compare_and_swap_16(&x, y, x);
return 0;
}
'''))
endif
endif
endif
config_host_data.set('CONFIG_GETAUXVAL', cc.links(gnu_source_prefix + '''
#include <sys/auxv.h>
int main(void) {
return getauxval(AT_HWCAP) == 0;
}'''))
config_host_data.set('CONFIG_ELF_AUX_INFO', cc.links(gnu_source_prefix + '''
#include <sys/auxv.h>
int main(void) {
unsigned long hwcap = 0;
elf_aux_info(AT_HWCAP, &hwcap, sizeof(hwcap));
return hwcap;
}'''))
config_host_data.set('CONFIG_USBFS', have_linux_user and cc.compiles('''
#include <linux/usbdevice_fs.h>
#ifndef USBDEVFS_GET_CAPABILITIES
#error "USBDEVFS_GET_CAPABILITIES undefined"
#endif
#ifndef USBDEVFS_DISCONNECT_CLAIM
#error "USBDEVFS_DISCONNECT_CLAIM undefined"
#endif
int main(void) { return 0; }'''))
have_keyring = get_option('keyring') \
.require(host_os == 'linux', error_message: 'keyring is only available on Linux') \
.require(cc.compiles('''
#include <errno.h>
#include <asm/unistd.h>
#include <linux/keyctl.h>
#include <sys/syscall.h>
#include <unistd.h>
int main(void) {
return syscall(__NR_keyctl, KEYCTL_READ, 0, NULL, NULL, 0);
}'''), error_message: 'keyctl syscall not available on this system').allowed()
config_host_data.set('CONFIG_SECRET_KEYRING', have_keyring)
have_cpuid_h = cc.links('''
#include <cpuid.h>
int main(void) {
unsigned a, b, c, d;
unsigned max = __get_cpuid_max(0, 0);
if (max >= 1) {
__cpuid(1, a, b, c, d);
}
if (max >= 7) {
__cpuid_count(7, 0, a, b, c, d);
}
return 0;
}''')
config_host_data.set('CONFIG_CPUID_H', have_cpuid_h)
# Don't bother to advertise asm/hwprobe.h for old versions that do
# not contain RISCV_HWPROBE_EXT_ZBA.
config_host_data.set('CONFIG_ASM_HWPROBE_H',
cc.has_header_symbol('asm/hwprobe.h',
'RISCV_HWPROBE_EXT_ZBA'))
config_host_data.set('CONFIG_AVX2_OPT', get_option('avx2') \
.require(have_cpuid_h, error_message: 'cpuid.h not available, cannot enable AVX2') \
.require(cc.links('''
#include <cpuid.h>
#include <immintrin.h>
static int __attribute__((target("avx2"))) bar(void *a) {
__m256i x = *(__m256i *)a;
return _mm256_testz_si256(x, x);
}
int main(int argc, char *argv[]) { return bar(argv[argc - 1]); }
'''), error_message: 'AVX2 not available').allowed())
config_host_data.set('CONFIG_AVX512BW_OPT', get_option('avx512bw') \
.require(have_cpuid_h, error_message: 'cpuid.h not available, cannot enable AVX512BW') \
.require(cc.links('''
#include <cpuid.h>
#include <immintrin.h>
static int __attribute__((target("avx512bw"))) bar(void *a) {
__m512i *x = a;
__m512i res= _mm512_abs_epi8(*x);
return res[1];
}
int main(int argc, char *argv[]) { return bar(argv[0]); }
'''), error_message: 'AVX512BW not available').allowed())
# For both AArch64 and AArch32, detect if builtins are available.
config_host_data.set('CONFIG_ARM_AES_BUILTIN', cc.compiles('''
#include <arm_neon.h>
#ifndef __ARM_FEATURE_AES
__attribute__((target("+crypto")))
#endif
void foo(uint8x16_t *p) { *p = vaesmcq_u8(*p); }
'''))
if get_option('membarrier').disabled()
have_membarrier = false
elif host_os == 'windows'
have_membarrier = true
elif host_os == 'linux'
have_membarrier = cc.compiles('''
#include <linux/membarrier.h>
#include <sys/syscall.h>
#include <unistd.h>
#include <stdlib.h>
int main(void) {
syscall(__NR_membarrier, MEMBARRIER_CMD_QUERY, 0);
syscall(__NR_membarrier, MEMBARRIER_CMD_SHARED, 0);
exit(0);
}''')
endif
config_host_data.set('CONFIG_MEMBARRIER', get_option('membarrier') \
.require(have_membarrier, error_message: 'membarrier system call not available') \
.allowed())
have_afalg = get_option('crypto_afalg') \
.require(cc.compiles(gnu_source_prefix + '''
#include <errno.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <linux/if_alg.h>
int main(void) {
int sock;
sock = socket(AF_ALG, SOCK_SEQPACKET, 0);
return sock;
}
'''), error_message: 'AF_ALG requested but could not be detected').allowed()
config_host_data.set('CONFIG_AF_ALG', have_afalg)
config_host_data.set('CONFIG_AF_VSOCK', cc.has_header_symbol(
'linux/vm_sockets.h', 'AF_VSOCK',
prefix: '#include <sys/socket.h>',
))
have_vss = false
have_vss_sdk = false # old xp/2003 SDK
if host_os == 'windows' and 'cpp' in all_languages
have_vss = cxx.compiles('''
#define __MIDL_user_allocate_free_DEFINED__
#include <vss.h>
int main(void) { return VSS_CTX_BACKUP; }''')
have_vss_sdk = cxx.has_header('vscoordint.h')
endif
config_host_data.set('HAVE_VSS_SDK', have_vss_sdk)
# Older versions of MinGW do not import _lock_file and _unlock_file properly.
# This was fixed for v6.0.0 with commit b48e3ac8969d.
if host_os == 'windows'
config_host_data.set('HAVE__LOCK_FILE', cc.links('''
#include <stdio.h>
int main(void) {
_lock_file(NULL);
_unlock_file(NULL);
return 0;
}''', name: '_lock_file and _unlock_file'))
endif
if host_os == 'windows'
mingw_has_setjmp_longjmp = cc.links('''
#include <setjmp.h>
int main(void) {
/*
* These functions are not available in setjmp header, but may be
* available at link time, from libmingwex.a.
*/
extern int __mingw_setjmp(jmp_buf);
extern void __attribute__((noreturn)) __mingw_longjmp(jmp_buf, int);
jmp_buf env;
__mingw_setjmp(env);
__mingw_longjmp(env, 0);
}
''', name: 'mingw setjmp and longjmp')
if cpu == 'aarch64' and not mingw_has_setjmp_longjmp
error('mingw must provide setjmp/longjmp for windows-arm64')
endif
endif
########################
# Target configuration #
########################
minikconf = find_program('scripts/minikconf.py')
config_all_accel = {}
config_all_devices = {}
config_devices_mak_list = []
config_devices_h = {}
config_target_h = {}
config_target_mak = {}
disassemblers = {
'alpha' : ['CONFIG_ALPHA_DIS'],
'avr' : ['CONFIG_AVR_DIS'],
'hexagon' : ['CONFIG_HEXAGON_DIS'],
'hppa' : ['CONFIG_HPPA_DIS'],
'i386' : ['CONFIG_I386_DIS'],
'x86_64' : ['CONFIG_I386_DIS'],
'm68k' : ['CONFIG_M68K_DIS'],
'microblaze' : ['CONFIG_MICROBLAZE_DIS'],
'mips' : ['CONFIG_MIPS_DIS'],
'or1k' : ['CONFIG_OPENRISC_DIS'],
'ppc' : ['CONFIG_PPC_DIS'],
'riscv' : ['CONFIG_RISCV_DIS'],
'rx' : ['CONFIG_RX_DIS'],
's390' : ['CONFIG_S390_DIS'],
'sh4' : ['CONFIG_SH4_DIS'],
'sparc' : ['CONFIG_SPARC_DIS'],
'xtensa' : ['CONFIG_XTENSA_DIS'],
'loongarch' : ['CONFIG_LOONGARCH_DIS'],
}
have_ivshmem = config_host_data.get('CONFIG_EVENTFD')
host_kconfig = \
(get_option('fuzzing') ? ['CONFIG_FUZZ=y'] : []) + \
(have_tpm ? ['CONFIG_TPM=y'] : []) + \
(pixman.found() ? ['CONFIG_PIXMAN=y'] : []) + \
(spice.found() ? ['CONFIG_SPICE=y'] : []) + \
(have_ivshmem ? ['CONFIG_IVSHMEM=y'] : []) + \
(opengl.found() ? ['CONFIG_OPENGL=y'] : []) + \
(x11.found() ? ['CONFIG_X11=y'] : []) + \
(fdt.found() ? ['CONFIG_FDT=y'] : []) + \
(have_vhost_user ? ['CONFIG_VHOST_USER=y'] : []) + \
(have_vhost_vdpa ? ['CONFIG_VHOST_VDPA=y'] : []) + \
(have_vhost_kernel ? ['CONFIG_VHOST_KERNEL=y'] : []) + \
(have_virtfs ? ['CONFIG_VIRTFS=y'] : []) + \
(host_os == 'linux' ? ['CONFIG_LINUX=y'] : []) + \
(multiprocess_allowed ? ['CONFIG_MULTIPROCESS_ALLOWED=y'] : []) + \
Add Hyper-V Dynamic Memory Protocol driver (hv-balloon) base This driver is like virtio-balloon on steroids: it allows both changing the guest memory allocation via ballooning and (in the next patch) inserting pieces of extra RAM into it on demand from a provided memory backend. The actual resizing is done via ballooning interface (for example, via the "balloon" HMP command). This includes resizing the guest past its boot size - that is, hot-adding additional memory in granularity limited only by the guest alignment requirements, as provided by the next patch. In contrast with ACPI DIMM hotplug where one can only request to unplug a whole DIMM stick this driver allows removing memory from guest in single page (4k) units via ballooning. After a VM reboot the guest is back to its original (boot) size. In the future, the guest boot memory size might be changed on reboot instead, taking into account the effective size that VM had before that reboot (much like Hyper-V does). For performance reasons, the guest-released memory is tracked in a few range trees, as a series of (start, count) ranges. Each time a new page range is inserted into such tree its neighbors are checked as candidates for possible merging with it. Besides performance reasons, the Dynamic Memory protocol itself uses page ranges as the data structure in its messages, so relevant pages need to be merged into such ranges anyway. One has to be careful when tracking the guest-released pages, since the guest can maliciously report returning pages outside its current address space, which later clash with the address range of newly added memory. Similarly, the guest can report freeing the same page twice. The above design results in much better ballooning performance than when using virtio-balloon with the same guest: 230 GB / minute with this driver versus 70 GB / minute with virtio-balloon. During a ballooning operation most of time is spent waiting for the guest to come up with newly freed page ranges, processing the received ranges on the host side (in QEMU and KVM) is nearly instantaneous. The unballoon operation is also pretty much instantaneous: thanks to the merging of the ballooned out page ranges 200 GB of memory can be returned to the guest in about 1 second. With virtio-balloon this operation takes about 2.5 minutes. These tests were done against a Windows Server 2019 guest running on a Xeon E5-2699, after dirtying the whole memory inside guest before each balloon operation. Using a range tree instead of a bitmap to track the removed memory also means that the solution scales well with the guest size: even a 1 TB range takes just a few bytes of such metadata. Since the required GTree operations aren't present in every Glib version a check for them was added to the meson build script, together with new "--enable-hv-balloon" and "--disable-hv-balloon" configure arguments. If these GTree operations are missing in the system's Glib version this driver will be skipped during QEMU build. An optional "status-report=on" device parameter requests memory status events from the guest (typically sent every second), which allow the host to learn both the guest memory available and the guest memory in use counts. Following commits will add support for their external emission as "HV_BALLOON_STATUS_REPORT" QMP events. The driver is named hv-balloon since the Linux kernel client driver for the Dynamic Memory Protocol is named as such and to follow the naming pattern established by the virtio-balloon driver. The whole protocol runs over Hyper-V VMBus. The driver was tested against Windows Server 2012 R2, Windows Server 2016 and Windows Server 2019 guests and obeys the guest alignment requirements reported to the host via DM_CAPABILITIES_REPORT message. Acked-by: David Hildenbrand <david@redhat.com> Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
2023-06-12 17:00:54 +03:00
(vfio_user_server_allowed ? ['CONFIG_VFIO_USER_SERVER_ALLOWED=y'] : []) + \
(hv_balloon ? ['CONFIG_HV_BALLOON_POSSIBLE=y'] : [])
ignored = [ 'TARGET_XML_FILES', 'TARGET_ABI_DIR', 'TARGET_ARCH' ]
default_targets = 'CONFIG_DEFAULT_TARGETS' in config_host
actual_target_dirs = []
fdt_required = []
foreach target : target_dirs
config_target = { 'TARGET_NAME': target.split('-')[0] }
if target.endswith('linux-user')
if host_os != 'linux'
if default_targets
continue
endif
error('Target @0@ is only available on a Linux host'.format(target))
endif
config_target += { 'CONFIG_LINUX_USER': 'y' }
elif target.endswith('bsd-user')
if host_os not in bsd_oses
if default_targets
continue
endif
error('Target @0@ is only available on a BSD host'.format(target))
endif
config_target += { 'CONFIG_BSD_USER': 'y' }
elif target.endswith('softmmu')
config_target += { 'CONFIG_SYSTEM_ONLY': 'y' }
config_target += { 'CONFIG_SOFTMMU': 'y' }
endif
if target.endswith('-user')
config_target += {
'CONFIG_USER_ONLY': 'y',
'CONFIG_QEMU_INTERP_PREFIX':
get_option('interp_prefix').replace('%M', config_target['TARGET_NAME'])
}
endif
target_kconfig = []
foreach sym: accelerators
if sym == 'CONFIG_TCG' or target in accelerator_targets.get(sym, [])
config_target += { sym: 'y' }
config_all_accel += { sym: 'y' }
if target in modular_tcg
config_target += { 'CONFIG_TCG_MODULAR': 'y' }
else
config_target += { 'CONFIG_TCG_BUILTIN': 'y' }
endif
target_kconfig += [ sym + '=y' ]
endif
endforeach
if target_kconfig.length() == 0
if default_targets
continue
endif
error('No accelerator available for target @0@'.format(target))
endif
config_target += keyval.load('configs/targets' / target + '.mak')
config_target += { 'TARGET_' + config_target['TARGET_ARCH'].to_upper(): 'y' }
if 'TARGET_NEED_FDT' in config_target and not fdt.found()
if default_targets
warning('Disabling ' + target + ' due to missing libfdt')
else
fdt_required += target
endif
continue
endif
actual_target_dirs += target
# Add default keys
if 'TARGET_BASE_ARCH' not in config_target
config_target += {'TARGET_BASE_ARCH': config_target['TARGET_ARCH']}
endif
if 'TARGET_ABI_DIR' not in config_target
config_target += {'TARGET_ABI_DIR': config_target['TARGET_ARCH']}
endif
if 'TARGET_BIG_ENDIAN' not in config_target
config_target += {'TARGET_BIG_ENDIAN': 'n'}
endif
foreach k, v: disassemblers
if host_arch.startswith(k) or config_target['TARGET_BASE_ARCH'].startswith(k)
foreach sym: v
config_target += { sym: 'y' }
endforeach
endif
endforeach
config_target_data = configuration_data()
foreach k, v: config_target
if not k.startswith('TARGET_') and not k.startswith('CONFIG_')
# do nothing
elif ignored.contains(k)
# do nothing
elif k == 'TARGET_BASE_ARCH'
# Note that TARGET_BASE_ARCH ends up in config-target.h but it is
# not used to select files from sourcesets.
config_target_data.set('TARGET_' + v.to_upper(), 1)
elif k == 'TARGET_NAME' or k == 'CONFIG_QEMU_INTERP_PREFIX'
config_target_data.set_quoted(k, v)
elif v == 'y'
config_target_data.set(k, 1)
elif v == 'n'
config_target_data.set(k, 0)
else
config_target_data.set(k, v)
endif
endforeach
config_target_data.set('QEMU_ARCH',
'QEMU_ARCH_' + config_target['TARGET_BASE_ARCH'].to_upper())
config_target_h += {target: configure_file(output: target + '-config-target.h',
configuration: config_target_data)}
if target.endswith('-softmmu')
target_kconfig += 'CONFIG_' + config_target['TARGET_ARCH'].to_upper() + '=y'
target_kconfig += 'CONFIG_TARGET_BIG_ENDIAN=' + config_target['TARGET_BIG_ENDIAN']
config_input = meson.get_external_property(target, 'default')
config_devices_mak = target + '-config-devices.mak'
config_devices_mak = configure_file(
input: ['configs/devices' / target / config_input + '.mak', 'Kconfig'],
output: config_devices_mak,
depfile: config_devices_mak + '.d',
capture: true,
command: [minikconf,
get_option('default_devices') ? '--defconfig' : '--allnoconfig',
config_devices_mak, '@DEPFILE@', '@INPUT@',
host_kconfig, target_kconfig])
config_devices_data = configuration_data()
config_devices = keyval.load(config_devices_mak)
foreach k, v: config_devices
config_devices_data.set(k, 1)
endforeach
config_devices_mak_list += config_devices_mak
config_devices_h += {target: configure_file(output: target + '-config-devices.h',
configuration: config_devices_data)}
config_target += config_devices
config_all_devices += config_devices
endif
config_target_mak += {target: config_target}
endforeach
target_dirs = actual_target_dirs
target_configs_h = []
foreach target: target_dirs
target_configs_h += config_target_h[target]
target_configs_h += config_devices_h.get(target, [])
endforeach
genh += custom_target('config-poison.h',
input: [target_configs_h],
output: 'config-poison.h',
capture: true,
command: [find_program('scripts/make-config-poison.sh'),
target_configs_h])
if fdt_required.length() > 0
error('fdt disabled but required by targets ' + ', '.join(fdt_required))
endif
###############
# Subprojects #
###############
libvfio_user_dep = not_found
if have_system and vfio_user_server_allowed
libvfio_user_proj = subproject('libvfio-user', required: true)
libvfio_user_dep = libvfio_user_proj.get_variable('libvfio_user_dep')
endif
vhost_user = not_found
if host_os == 'linux' and have_vhost_user
libvhost_user = subproject('libvhost-user')
vhost_user = libvhost_user.get_variable('vhost_user_dep')
endif
libvduse = not_found
if have_libvduse
libvduse_proj = subproject('libvduse')
libvduse = libvduse_proj.get_variable('libvduse_dep')
endif
#####################
# Generated sources #
#####################
genh += configure_file(output: 'config-host.h', configuration: config_host_data)
hxtool = find_program('scripts/hxtool')
shaderinclude = find_program('scripts/shaderinclude.py')
qapi_gen = find_program('scripts/qapi-gen.py')
qapi_gen_depends = [ meson.current_source_dir() / 'scripts/qapi/__init__.py',
meson.current_source_dir() / 'scripts/qapi/commands.py',
meson.current_source_dir() / 'scripts/qapi/common.py',
meson.current_source_dir() / 'scripts/qapi/error.py',
meson.current_source_dir() / 'scripts/qapi/events.py',
meson.current_source_dir() / 'scripts/qapi/expr.py',
meson.current_source_dir() / 'scripts/qapi/gen.py',
meson.current_source_dir() / 'scripts/qapi/introspect.py',
meson.current_source_dir() / 'scripts/qapi/main.py',
meson.current_source_dir() / 'scripts/qapi/parser.py',
meson.current_source_dir() / 'scripts/qapi/schema.py',
meson.current_source_dir() / 'scripts/qapi/source.py',
meson.current_source_dir() / 'scripts/qapi/types.py',
meson.current_source_dir() / 'scripts/qapi/visit.py',
meson.current_source_dir() / 'scripts/qapi-gen.py'
]
tracetool = [
python, files('scripts/tracetool.py'),
'--backend=' + ','.join(get_option('trace_backends'))
]
tracetool_depends = files(
'scripts/tracetool/backend/log.py',
'scripts/tracetool/backend/__init__.py',
'scripts/tracetool/backend/dtrace.py',
'scripts/tracetool/backend/ftrace.py',
'scripts/tracetool/backend/simple.py',
'scripts/tracetool/backend/syslog.py',
'scripts/tracetool/backend/ust.py',
'scripts/tracetool/format/ust_events_c.py',
'scripts/tracetool/format/ust_events_h.py',
'scripts/tracetool/format/__init__.py',
'scripts/tracetool/format/d.py',
'scripts/tracetool/format/simpletrace_stap.py',
'scripts/tracetool/format/c.py',
'scripts/tracetool/format/h.py',
'scripts/tracetool/format/log_stap.py',
'scripts/tracetool/format/stap.py',
'scripts/tracetool/__init__.py',
)
qemu_version_cmd = [find_program('scripts/qemu-version.sh'),
meson.current_source_dir(),
get_option('pkgversion'), meson.project_version()]
qemu_version = custom_target('qemu-version.h',
output: 'qemu-version.h',
command: qemu_version_cmd,
capture: true,
build_by_default: true,
build_always_stale: true)
genh += qemu_version
hxdep = []
hx_headers = [
['qemu-options.hx', 'qemu-options.def'],
['qemu-img-cmds.hx', 'qemu-img-cmds.h'],
]
if have_system
hx_headers += [
['hmp-commands.hx', 'hmp-commands.h'],
['hmp-commands-info.hx', 'hmp-commands-info.h'],
]
endif
foreach d : hx_headers
hxdep += custom_target(d[1],
input: files(d[0]),
output: d[1],
capture: true,
command: [hxtool, '-h', '@INPUT0@'])
endforeach
genh += hxdep
###############
# Trace files #
###############
# TODO: add each directory to the subdirs from its own meson.build, once
# we have those
trace_events_subdirs = [
'crypto',
'qapi',
'qom',
'monitor',
'util',
'gdbstub',
]
if have_linux_user
trace_events_subdirs += [ 'linux-user' ]
endif
if have_bsd_user
trace_events_subdirs += [ 'bsd-user' ]
endif
if have_block
trace_events_subdirs += [
'authz',
'block',
'chardev',
'io',
'nbd',
'scsi',
]
endif
if have_system
trace_events_subdirs += [
'accel/kvm',
'audio',
'backends',
'backends/tpm',
'ebpf',
'hw/9pfs',
'hw/acpi',
'hw/adc',
'hw/alpha',
'hw/arm',
'hw/audio',
'hw/block',
'hw/char',
'hw/display',
'hw/dma',
'hw/fsi',
'hw/hyperv',
'hw/i2c',
'hw/i386',
'hw/i386/xen',
'hw/i386/kvm',
'hw/ide',
'hw/input',
'hw/intc',
'hw/isa',
'hw/mem',
'hw/mips',
'hw/misc',
'hw/misc/macio',
'hw/net',
'hw/net/can',
'hw/nubus',
'hw/nvme',
'hw/nvram',
'hw/pci',
'hw/pci-host',
'hw/ppc',
'hw/rtc',
'hw/s390x',
'hw/scsi',
'hw/sd',
'hw/sh4',
'hw/sparc',
'hw/sparc64',
'hw/ssi',
'hw/timer',
'hw/tpm',
'hw/ufs',
'hw/usb',
'hw/vfio',
'hw/virtio',
'hw/watchdog',
'hw/xen',
'hw/gpio',
'migration',
'net',
'system',
'ui',
'hw/remote',
]
endif
if have_system or have_user
trace_events_subdirs += [
'accel/tcg',
'hw/core',
'target/arm',
'target/arm/hvf',
'target/hppa',
'target/i386',
'target/i386/kvm',
'target/loongarch',
'target/mips/tcg',
'target/ppc',
'target/riscv',
'target/s390x',
'target/s390x/kvm',
'target/sparc',
]
endif
###################
# Collect sources #
###################
authz_ss = ss.source_set()
blockdev_ss = ss.source_set()
block_ss = ss.source_set()
chardev_ss = ss.source_set()
common_ss = ss.source_set()
crypto_ss = ss.source_set()
hwcore_ss = ss.source_set()
io_ss = ss.source_set()
qmp_ss = ss.source_set()
qom_ss = ss.source_set()
system_ss = ss.source_set()
specific_fuzz_ss = ss.source_set()
specific_ss = ss.source_set()
stub_ss = ss.source_set()
trace_ss = ss.source_set()
user_ss = ss.source_set()
util_ss = ss.source_set()
# accel modules
qtest_module_ss = ss.source_set()
tcg_module_ss = ss.source_set()
modules = {}
target_modules = {}
hw_arch = {}
target_arch = {}
target_system_arch = {}
target_user_arch = {}
# NOTE: the trace/ subdirectory needs the qapi_trace_events variable
# that is filled in by qapi/.
subdir('qapi')
subdir('qobject')
subdir('stubs')
subdir('trace')
subdir('util')
subdir('qom')
subdir('authz')
subdir('crypto')
subdir('ui')
subdir('gdbstub')
if have_system
subdir('hw')
else
subdir('hw/core')
endif
if enable_modules
libmodulecommon = static_library('module-common', files('module-common.c') + genh, pic: true, c_args: '-DBUILD_DSO')
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
modulecommon = declare_dependency(objects: libmodulecommon.extract_all_objects(recursive: false), compile_args: '-DBUILD_DSO')
endif
qom_ss = qom_ss.apply({})
Introduce event-loop-base abstract class Introduce the 'event-loop-base' abstract class, it'll hold the properties common to all event loops and provide the necessary hooks for their creation and maintenance. Then have iothread inherit from it. EventLoopBaseClass is defined as user creatable and provides a hook for its children to attach themselves to the user creatable class 'complete' function. It also provides an update_params() callback to propagate property changes onto its children. The new 'event-loop-base' class will live in the root directory. It is built on its own using the 'link_whole' option (there are no direct function dependencies between the class and its children, it all happens trough 'constructor' magic). And also imposes new compilation dependencies: qom <- event-loop-base <- blockdev (iothread.c) And in subsequent patches: qom <- event-loop-base <- qemuutil (util/main-loop.c) All this forced some amount of reordering in meson.build: - Moved qom build definition before qemuutil. Doing it the other way around (i.e. moving qemuutil after qom) isn't possible as a lot of core libraries that live in between the two depend on it. - Process the 'hw' subdir earlier, as it introduces files into the 'qom' source set. No functional changes intended. Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Message-id: 20220425075723.20019-2-nsaenzju@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-04-25 10:57:21 +03:00
libqom = static_library('qom', qom_ss.sources() + genh,
dependencies: [qom_ss.dependencies()],
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
qom = declare_dependency(objects: libqom.extract_all_objects(recursive: false),
dependencies: qom_ss.dependencies())
Introduce event-loop-base abstract class Introduce the 'event-loop-base' abstract class, it'll hold the properties common to all event loops and provide the necessary hooks for their creation and maintenance. Then have iothread inherit from it. EventLoopBaseClass is defined as user creatable and provides a hook for its children to attach themselves to the user creatable class 'complete' function. It also provides an update_params() callback to propagate property changes onto its children. The new 'event-loop-base' class will live in the root directory. It is built on its own using the 'link_whole' option (there are no direct function dependencies between the class and its children, it all happens trough 'constructor' magic). And also imposes new compilation dependencies: qom <- event-loop-base <- blockdev (iothread.c) And in subsequent patches: qom <- event-loop-base <- qemuutil (util/main-loop.c) All this forced some amount of reordering in meson.build: - Moved qom build definition before qemuutil. Doing it the other way around (i.e. moving qemuutil after qom) isn't possible as a lot of core libraries that live in between the two depend on it. - Process the 'hw' subdir earlier, as it introduces files into the 'qom' source set. No functional changes intended. Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Message-id: 20220425075723.20019-2-nsaenzju@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-04-25 10:57:21 +03:00
event_loop_base = files('event-loop-base.c')
event_loop_base = static_library('event-loop-base',
sources: event_loop_base + genh,
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
event_loop_base = declare_dependency(objects: event_loop_base.extract_all_objects(recursive: false),
Introduce event-loop-base abstract class Introduce the 'event-loop-base' abstract class, it'll hold the properties common to all event loops and provide the necessary hooks for their creation and maintenance. Then have iothread inherit from it. EventLoopBaseClass is defined as user creatable and provides a hook for its children to attach themselves to the user creatable class 'complete' function. It also provides an update_params() callback to propagate property changes onto its children. The new 'event-loop-base' class will live in the root directory. It is built on its own using the 'link_whole' option (there are no direct function dependencies between the class and its children, it all happens trough 'constructor' magic). And also imposes new compilation dependencies: qom <- event-loop-base <- blockdev (iothread.c) And in subsequent patches: qom <- event-loop-base <- qemuutil (util/main-loop.c) All this forced some amount of reordering in meson.build: - Moved qom build definition before qemuutil. Doing it the other way around (i.e. moving qemuutil after qom) isn't possible as a lot of core libraries that live in between the two depend on it. - Process the 'hw' subdir earlier, as it introduces files into the 'qom' source set. No functional changes intended. Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Message-id: 20220425075723.20019-2-nsaenzju@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-04-25 10:57:21 +03:00
dependencies: [qom])
stub_ss = stub_ss.apply({})
util_ss.add_all(trace_ss)
util_ss = util_ss.apply({})
libqemuutil = static_library('qemuutil',
build_by_default: false,
sources: util_ss.sources() + stub_ss.sources() + genh,
dependencies: [util_ss.dependencies(), libm, threads, glib, socket, malloc])
qemuutil = declare_dependency(link_with: libqemuutil,
sources: genh + version_res,
dependencies: [event_loop_base])
if have_system or have_user
decodetree = generator(find_program('scripts/decodetree.py'),
output: 'decode-@BASENAME@.c.inc',
arguments: ['@INPUT@', '@EXTRA_ARGS@', '-o', '@OUTPUT@'])
subdir('libdecnumber')
subdir('target')
endif
subdir('audio')
subdir('io')
subdir('chardev')
subdir('fsdev')
subdir('dump')
if have_block
block_ss.add(files(
'block.c',
'blockjob.c',
'job.c',
'qemu-io-cmds.c',
))
if config_host_data.get('CONFIG_REPLICATION')
block_ss.add(files('replication.c'))
endif
subdir('nbd')
subdir('scsi')
subdir('block')
blockdev_ss.add(files(
'blockdev.c',
'blockdev-nbd.c',
'iothread.c',
'job-qmp.c',
))
# os-posix.c contains POSIX-specific functions used by qemu-storage-daemon,
# os-win32.c does not
if host_os == 'windows'
system_ss.add(files('os-win32.c'))
else
blockdev_ss.add(files('os-posix.c'))
endif
endif
common_ss.add(files('cpu-common.c'))
specific_ss.add(files('cpu-target.c'))
subdir('system')
exec: Build page-vary-common.c with -fno-lto In bbc17caf81f, we used an alias attribute to allow target_page to be declared const, and yet be initialized late. This fails when using LTO with several versions of gcc. The compiler looks through the alias and decides that the const variable is statically initialized to zero, then propagates that zero to many uses of the variable. This can be avoided by compiling one object file with -fno-lto. In this way, any initializer cannot be seen, and the constant propagation does not occur. Since we are certain to have this separate compilation unit, we can drop the alias attribute as well. We simply have differing declarations for target_page in different compilation units. Drop the use of init_target_page, and drop the configure detection for CONFIG_ATTRIBUTE_ALIAS. In order to change the compilation flags for a file with meson, we must use a static_library. This runs into specific_ss, where we would need to create many static_library instances. Fix this by splitting page-vary.c: the page-vary-common.c part is compiled once as a static_library, while the page-vary.c part is left in specific_ss in order to handle the target-specific value of TARGET_PAGE_BITS_MIN. Reported-by: Gavin Shan <gshan@redhat.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20210321211534.2101231-1-richard.henderson@linaro.org> [PMD: Fix typo in subject, split original patch in 3] Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Tested-by: Gavin Shan <gshan@redhat.com> Message-Id: <20210322112427.4045204-4-f4bug@amsat.org> [rth: Update MAINTAINERS] Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2021-03-22 14:24:26 +03:00
# Work around a gcc bug/misfeature wherein constant propagation looks
# through an alias:
# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99696
# to guess that a const variable is always zero. Without lto, this is
# impossible, as the alias is restricted to page-vary-common.c. Indeed,
# without lto, not even the alias is required -- we simply use different
# declarations in different compilation units.
pagevary = files('page-vary-common.c')
if get_option('b_lto')
pagevary_flags = ['-fno-lto']
if get_option('cfi')
pagevary_flags += '-fno-sanitize=cfi-icall'
endif
pagevary = static_library('page-vary-common', sources: pagevary + genh,
exec: Build page-vary-common.c with -fno-lto In bbc17caf81f, we used an alias attribute to allow target_page to be declared const, and yet be initialized late. This fails when using LTO with several versions of gcc. The compiler looks through the alias and decides that the const variable is statically initialized to zero, then propagates that zero to many uses of the variable. This can be avoided by compiling one object file with -fno-lto. In this way, any initializer cannot be seen, and the constant propagation does not occur. Since we are certain to have this separate compilation unit, we can drop the alias attribute as well. We simply have differing declarations for target_page in different compilation units. Drop the use of init_target_page, and drop the configure detection for CONFIG_ATTRIBUTE_ALIAS. In order to change the compilation flags for a file with meson, we must use a static_library. This runs into specific_ss, where we would need to create many static_library instances. Fix this by splitting page-vary.c: the page-vary-common.c part is compiled once as a static_library, while the page-vary.c part is left in specific_ss in order to handle the target-specific value of TARGET_PAGE_BITS_MIN. Reported-by: Gavin Shan <gshan@redhat.com> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20210321211534.2101231-1-richard.henderson@linaro.org> [PMD: Fix typo in subject, split original patch in 3] Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Tested-by: Gavin Shan <gshan@redhat.com> Message-Id: <20210322112427.4045204-4-f4bug@amsat.org> [rth: Update MAINTAINERS] Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
2021-03-22 14:24:26 +03:00
c_args: pagevary_flags)
pagevary = declare_dependency(link_with: pagevary)
endif
common_ss.add(pagevary)
specific_ss.add(files('page-target.c', 'page-vary-target.c'))
subdir('backends')
subdir('disas')
subdir('migration')
subdir('monitor')
subdir('net')
subdir('replay')
subdir('semihosting')
subdir('stats')
subdir('tcg')
subdir('fpu')
subdir('accel')
subdir('plugins')
subdir('ebpf')
common_user_inc = []
subdir('common-user')
subdir('bsd-user')
subdir('linux-user')
# needed for fuzzing binaries
subdir('tests/qtest/libqos')
subdir('tests/qtest/fuzz')
# accel modules
tcg_real_module_ss = ss.source_set()
tcg_real_module_ss.add_all(when: 'CONFIG_TCG_MODULAR', if_true: tcg_module_ss)
specific_ss.add_all(when: 'CONFIG_TCG_BUILTIN', if_true: tcg_module_ss)
target_modules += { 'accel' : { 'qtest': qtest_module_ss,
'tcg': tcg_real_module_ss }}
##############################################
# Internal static_libraries and dependencies #
##############################################
modinfo_collect = find_program('scripts/modinfo-collect.py')
modinfo_generate = find_program('scripts/modinfo-generate.py')
modinfo_files = []
block_mods = []
system_mods = []
emulator_modules = []
foreach d, list : modules
if not (d == 'block' ? have_block : have_system)
continue
endif
foreach m, module_ss : list
if enable_modules
module_ss.add(modulecommon)
module_ss = module_ss.apply(config_all_devices, strict: false)
sl = static_library(d + '-' + m, [genh, module_ss.sources()],
dependencies: module_ss.dependencies(), pic: true)
if d == 'block'
block_mods += sl
else
system_mods += sl
endif
emulator_modules += shared_module(sl.name(),
name_prefix: '',
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
objects: sl.extract_all_objects(recursive: false),
dependencies: module_ss.dependencies(),
install: true,
install_dir: qemu_moddir)
if module_ss.sources() != []
# FIXME: Should use sl.extract_all_objects(recursive: true) as
# input. Sources can be used multiple times but objects are
# unique when it comes to lookup in compile_commands.json.
# Depnds on a mesion version with
# https://github.com/mesonbuild/meson/pull/8900
modinfo_files += custom_target(d + '-' + m + '.modinfo',
output: d + '-' + m + '.modinfo',
input: module_ss.sources() + genh,
capture: true,
command: [modinfo_collect, module_ss.sources()])
endif
else
if d == 'block'
block_ss.add_all(module_ss)
else
system_ss.add_all(module_ss)
endif
endif
endforeach
endforeach
foreach d, list : target_modules
foreach m, module_ss : list
if enable_modules
module_ss.add(modulecommon)
foreach target : target_dirs
if target.endswith('-softmmu')
config_target = config_target_mak[target]
target_inc = [include_directories('target' / config_target['TARGET_BASE_ARCH'])]
c_args = ['-DCOMPILING_PER_TARGET',
'-DCONFIG_TARGET="@0@-config-target.h"'.format(target),
'-DCONFIG_DEVICES="@0@-config-devices.h"'.format(target)]
target_module_ss = module_ss.apply(config_target, strict: false)
if target_module_ss.sources() != []
module_name = d + '-' + m + '-' + config_target['TARGET_NAME']
sl = static_library(module_name,
[genh, target_module_ss.sources()],
dependencies: target_module_ss.dependencies(),
include_directories: target_inc,
c_args: c_args,
pic: true)
system_mods += sl
emulator_modules += shared_module(sl.name(),
name_prefix: '',
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
objects: sl.extract_all_objects(recursive: false),
dependencies: target_module_ss.dependencies(),
install: true,
install_dir: qemu_moddir)
# FIXME: Should use sl.extract_all_objects(recursive: true) too.
modinfo_files += custom_target(module_name + '.modinfo',
output: module_name + '.modinfo',
input: target_module_ss.sources() + genh,
capture: true,
command: [modinfo_collect, '--target', target, target_module_ss.sources()])
endif
endif
endforeach
else
specific_ss.add_all(module_ss)
endif
endforeach
endforeach
if enable_modules
foreach target : target_dirs
if target.endswith('-softmmu')
config_target = config_target_mak[target]
config_devices_mak = target + '-config-devices.mak'
modinfo_src = custom_target('modinfo-' + target + '.c',
output: 'modinfo-' + target + '.c',
input: modinfo_files,
command: [modinfo_generate, '--devices', config_devices_mak, '@INPUT@'],
capture: true)
modinfo_lib = static_library('modinfo-' + target + '.c', modinfo_src)
modinfo_dep = declare_dependency(link_with: modinfo_lib)
arch = config_target['TARGET_NAME'] == 'sparc64' ? 'sparc64' : config_target['TARGET_BASE_ARCH']
hw_arch[arch].add(modinfo_dep)
endif
endforeach
if emulator_modules.length() > 0
alias_target('modules', emulator_modules)
endif
endif
nm = find_program('nm')
undefsym = find_program('scripts/undefsym.py')
block_syms = custom_target('block.syms', output: 'block.syms',
input: [libqemuutil, block_mods],
capture: true,
command: [undefsym, nm, '@INPUT@'])
qemu_syms = custom_target('qemu.syms', output: 'qemu.syms',
input: [libqemuutil, system_mods],
capture: true,
command: [undefsym, nm, '@INPUT@'])
authz_ss = authz_ss.apply({})
libauthz = static_library('authz', authz_ss.sources() + genh,
dependencies: [authz_ss.dependencies()],
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
authz = declare_dependency(objects: libauthz.extract_all_objects(recursive: false),
dependencies: [authz_ss.dependencies(), qom])
crypto_ss = crypto_ss.apply({})
libcrypto = static_library('crypto', crypto_ss.sources() + genh,
dependencies: [crypto_ss.dependencies()],
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
crypto = declare_dependency(objects: libcrypto.extract_all_objects(recursive: false),
dependencies: [crypto_ss.dependencies(), authz, qom])
io_ss = io_ss.apply({})
libio = static_library('io', io_ss.sources() + genh,
dependencies: [io_ss.dependencies()],
link_with: libqemuutil,
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
io = declare_dependency(objects: libio.extract_all_objects(recursive: false),
dependencies: [io_ss.dependencies(), crypto, qom])
libmigration = static_library('migration', sources: migration_files + genh,
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
migration = declare_dependency(objects: libmigration.extract_all_objects(recursive: false),
dependencies: [qom, io])
system_ss.add(migration)
block_ss = block_ss.apply({})
libblock = static_library('block', block_ss.sources() + genh,
dependencies: block_ss.dependencies(),
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
block = declare_dependency(objects: libblock.extract_all_objects(recursive: false),
dependencies: [block_ss.dependencies(), crypto, io])
blockdev_ss = blockdev_ss.apply({})
libblockdev = static_library('blockdev', blockdev_ss.sources() + genh,
dependencies: blockdev_ss.dependencies(),
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
blockdev = declare_dependency(objects: libblockdev.extract_all_objects(recursive: false),
dependencies: [blockdev_ss.dependencies(), block, event_loop_base])
qmp_ss = qmp_ss.apply({})
libqmp = static_library('qmp', qmp_ss.sources() + genh,
dependencies: qmp_ss.dependencies(),
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
qmp = declare_dependency(objects: libqmp.extract_all_objects(recursive: false),
dependencies: qmp_ss.dependencies())
libchardev = static_library('chardev', chardev_ss.sources() + genh,
dependencies: chardev_ss.dependencies(),
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
chardev = declare_dependency(objects: libchardev.extract_all_objects(recursive: false),
dependencies: chardev_ss.dependencies())
hwcore_ss = hwcore_ss.apply({})
libhwcore = static_library('hwcore', sources: hwcore_ss.sources() + genh,
build_by_default: false)
meson: Pass objects and dependencies to declare_dependency() We used to request declare_dependency() to link_whole static libraries. If a static library is a thin archive, GNU ld keeps all object files referenced by the archive open, and sometimes exceeds the open file limit. Another problem with link_whole is that suboptimal handling of nested dependencies. link_whole by itself does not propagate dependencies. In particular, gnutls, a dependency of crypto, is not propagated to its users, and we currently workaround the issue by declaring gnutls as a dependency for each crypto user. On the other hand, if you write something like libfoo = static_library('foo', 'foo.c', dependencies: gnutls) foo = declare_dependency(link_whole: libfoo) libbar = static_library('bar', 'bar.c', dependencies: foo) bar = declare_dependency(link_whole: libbar, dependencies: foo) executable('prog', sources: files('prog.c'), dependencies: [foo, bar]) hoping to propagate the gnutls dependency into bar.c, you'll see a linking failure for "prog", because the foo.c.o object file is included in libbar.a and therefore it is linked twice into "prog": once from libfoo.a and once from libbar.a. Here Meson does not see the duplication, it just asks the linker to link all of libfoo.a and libbar.a into "prog". Instead of using link_whole, extract objects included in static libraries and pass them to declare_dependency(); and then the dependencies can be added as well so that they are propagated, because object files on the linker command line are always deduplicated. This requires Meson 1.1.0 or later. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Message-ID: <20240524-objects-v1-1-07cbbe96166b@daynix.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2024-05-24 11:00:22 +03:00
hwcore = declare_dependency(objects: libhwcore.extract_all_objects(recursive: false))
common_ss.add(hwcore)
###########
# Targets #
###########
system_ss.add(authz, blockdev, chardev, crypto, io, qmp)
common_ss.add(qom, qemuutil)
common_ss.add_all(when: 'CONFIG_SYSTEM_ONLY', if_true: [system_ss])
common_ss.add_all(when: 'CONFIG_USER_ONLY', if_true: user_ss)
# Note that this library is never used directly (only through extract_objects)
# and is not built by default; therefore, source files not used by the build
# configuration will be in build.ninja, but are never built by default.
common_all = static_library('common',
build_by_default: false,
sources: common_ss.all_sources() + genh,
include_directories: common_user_inc,
implicit_include_directories: false,
dependencies: common_ss.all_dependencies())
feature_to_c = find_program('scripts/feature_to_c.py')
if host_os == 'darwin'
entitlement = find_program('scripts/entitlement.sh')
endif
traceable = []
emulators = {}
foreach target : target_dirs
config_target = config_target_mak[target]
target_name = config_target['TARGET_NAME']
target_base_arch = config_target['TARGET_BASE_ARCH']
arch_srcs = [config_target_h[target]]
arch_deps = []
c_args = ['-DCOMPILING_PER_TARGET',
'-DCONFIG_TARGET="@0@-config-target.h"'.format(target),
'-DCONFIG_DEVICES="@0@-config-devices.h"'.format(target)]
link_args = emulator_link_args
target_inc = [include_directories('target' / config_target['TARGET_BASE_ARCH'])]
if host_os == 'linux'
target_inc += include_directories('linux-headers', is_system: true)
endif
if target.endswith('-softmmu')
target_type='system'
t = target_system_arch[target_base_arch].apply(config_target, strict: false)
arch_srcs += t.sources()
arch_deps += t.dependencies()
hw_dir = target_name == 'sparc64' ? 'sparc64' : target_base_arch
if hw_arch.has_key(hw_dir)
hw = hw_arch[hw_dir].apply(config_target, strict: false)
arch_srcs += hw.sources()
arch_deps += hw.dependencies()
endif
arch_srcs += config_devices_h[target]
link_args += ['@block.syms', '@qemu.syms']
else
abi = config_target['TARGET_ABI_DIR']
target_type='user'
target_inc += common_user_inc
if target_base_arch in target_user_arch
t = target_user_arch[target_base_arch].apply(config_target, strict: false)
arch_srcs += t.sources()
arch_deps += t.dependencies()
endif
if 'CONFIG_LINUX_USER' in config_target
base_dir = 'linux-user'
endif
if 'CONFIG_BSD_USER' in config_target
base_dir = 'bsd-user'
target_inc += include_directories('bsd-user/' / host_os)
target_inc += include_directories('bsd-user/host/' / host_arch)
dir = base_dir / abi
arch_srcs += files(dir / 'signal.c', dir / 'target_arch_cpu.c')
endif
target_inc += include_directories(
base_dir,
base_dir / abi,
)
if 'CONFIG_LINUX_USER' in config_target
dir = base_dir / abi
arch_srcs += files(dir / 'signal.c', dir / 'cpu_loop.c')
if config_target.has_key('TARGET_SYSTBL_ABI')
arch_srcs += \
syscall_nr_generators[abi].process(base_dir / abi / config_target['TARGET_SYSTBL'],
extra_args : config_target['TARGET_SYSTBL_ABI'])
endif
endif
endif
if 'TARGET_XML_FILES' in config_target
gdbstub_xml = custom_target(target + '-gdbstub-xml.c',
output: target + '-gdbstub-xml.c',
input: files(config_target['TARGET_XML_FILES'].split()),
command: [feature_to_c, '@INPUT@'],
capture: true)
arch_srcs += gdbstub_xml
endif
t = target_arch[target_base_arch].apply(config_target, strict: false)
arch_srcs += t.sources()
arch_deps += t.dependencies()
target_common = common_ss.apply(config_target, strict: false)
objects = common_all.extract_objects(target_common.sources())
arch_deps += target_common.dependencies()
target_specific = specific_ss.apply(config_target, strict: false)
arch_srcs += target_specific.sources()
arch_deps += target_specific.dependencies()
# allow using headers from the dependencies but do not include the sources,
# because this emulator only needs those in "objects". For external
# dependencies, the full dependency is included below in the executable.
lib_deps = []
foreach dep : arch_deps
lib_deps += dep.partial_dependency(compile_args: true, includes: true)
endforeach
lib = static_library('qemu-' + target,
sources: arch_srcs + genh,
dependencies: lib_deps,
objects: objects,
include_directories: target_inc,
c_args: c_args,
build_by_default: false)
if target.endswith('-softmmu')
execs = [{
'name': 'qemu-system-' + target_name,
'win_subsystem': 'console',
'sources': files('system/main.c'),
'dependencies': []
}]
if host_os == 'windows' and (sdl.found() or gtk.found())
execs += [{
'name': 'qemu-system-' + target_name + 'w',
'win_subsystem': 'windows',
'sources': files('system/main.c'),
'dependencies': []
}]
endif
if get_option('fuzzing')
specific_fuzz = specific_fuzz_ss.apply(config_target, strict: false)
execs += [{
'name': 'qemu-fuzz-' + target_name,
'win_subsystem': 'console',
'sources': specific_fuzz.sources(),
'dependencies': specific_fuzz.dependencies(),
}]
endif
else
execs = [{
'name': 'qemu-' + target_name,
'win_subsystem': 'console',
'sources': [],
'dependencies': []
}]
endif
foreach exe: execs
exe_name = exe['name']
if host_os == 'darwin'
exe_name += '-unsigned'
endif
emulator = executable(exe_name, exe['sources'],
install: true,
c_args: c_args,
dependencies: arch_deps + exe['dependencies'],
objects: lib.extract_all_objects(recursive: true),
link_depends: [block_syms, qemu_syms],
link_args: link_args,
win_subsystem: exe['win_subsystem'])
if host_os == 'darwin'
icon = 'pc-bios/qemu.rsrc'
build_input = [emulator, files(icon)]
install_input = [
get_option('bindir') / exe_name,
meson.current_source_dir() / icon
]
if 'CONFIG_HVF' in config_target
entitlements = 'accel/hvf/entitlements.plist'
build_input += files(entitlements)
install_input += meson.current_source_dir() / entitlements
endif
emulators += {exe['name'] : custom_target(exe['name'],
input: build_input,
output: exe['name'],
meson: Use find_program() to resolve the entitlement.sh script Using ../configure without any particular option generates 31 targets on Darwin, and meson search for the entitlement.sh script 31 times: Program nm found: YES Program scripts/undefsym.py found: YES (/opt/homebrew/opt/python@3.9/bin/python3.9 /Code/qemu/scripts/undefsym.py) Program scripts/feature_to_c.sh found: YES (/bin/sh /Code/qemu/scripts/feature_to_c.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Configuring 50-edk2-i386-secure.json using configuration Configuring 50-edk2-x86_64-secure.json using configuration Use find_program() which seems to cache the script path once found. Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Akihiko Odaki <akihiko.odaki@gmail.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20220122002052.83745-1-f4bug@amsat.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2022-01-22 03:20:52 +03:00
command: [entitlement, '@OUTPUT@', '@INPUT@'])
}
meson: Use find_program() to resolve the entitlement.sh script Using ../configure without any particular option generates 31 targets on Darwin, and meson search for the entitlement.sh script 31 times: Program nm found: YES Program scripts/undefsym.py found: YES (/opt/homebrew/opt/python@3.9/bin/python3.9 /Code/qemu/scripts/undefsym.py) Program scripts/feature_to_c.sh found: YES (/bin/sh /Code/qemu/scripts/feature_to_c.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Program scripts/entitlement.sh found: YES (/Code/qemu/scripts/entitlement.sh) Configuring 50-edk2-i386-secure.json using configuration Configuring 50-edk2-x86_64-secure.json using configuration Use find_program() which seems to cache the script path once found. Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Reviewed-by: Akihiko Odaki <akihiko.odaki@gmail.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Message-Id: <20220122002052.83745-1-f4bug@amsat.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2022-01-22 03:20:52 +03:00
meson.add_install_script(entitlement, '--install',
get_option('bindir') / exe['name'],
install_input)
else
emulators += {exe['name']: emulator}
endif
traceable += [{
'exe': exe['name'],
'probe-prefix': 'qemu.' + target_type + '.' + target_name,
}]
endforeach
endforeach
# Other build targets
if get_option('plugins')
install_headers('include/qemu/qemu-plugin.h')
if host_os == 'windows'
# On windows, we want to deliver the qemu_plugin_api.lib file in the qemu installer,
# so that plugin authors can compile against it.
install_data(win32_qemu_plugin_api_lib, install_dir: 'lib')
endif
endif
subdir('qga')
# Don't build qemu-keymap if xkbcommon is not explicitly enabled
# when we don't build tools or system
if xkbcommon.found()
# used for the update-keymaps target, so include rules even if !have_tools
qemu_keymap = executable('qemu-keymap', files('qemu-keymap.c', 'ui/input-keymap.c') + genh,
dependencies: [qemuutil, xkbcommon], install: have_tools)
endif
if have_tools
qemu_img = executable('qemu-img', [files('qemu-img.c'), hxdep],
link_args: '@block.syms', link_depends: block_syms,
dependencies: [authz, block, crypto, io, qom, qemuutil], install: true)
qemu_io = executable('qemu-io', files('qemu-io.c'),
link_args: '@block.syms', link_depends: block_syms,
dependencies: [block, qemuutil], install: true)
qemu_nbd = executable('qemu-nbd', files('qemu-nbd.c'),
link_args: '@block.syms', link_depends: block_syms,
dependencies: [blockdev, qemuutil, selinux],
nbd/server: Add --selinux-label option Under SELinux, Unix domain sockets have two labels. One is on the disk and can be set with commands such as chcon(1). There is a different label stored in memory (called the process label). This can only be set by the process creating the socket. When using SELinux + SVirt and wanting qemu to be able to connect to a qemu-nbd instance, you must set both labels correctly first. For qemu-nbd the options to set the second label are awkward. You can create the socket in a wrapper program and then exec into qemu-nbd. Or you could try something with LD_PRELOAD. This commit adds the ability to set the label straightforwardly on the command line, via the new --selinux-label flag. (The name of the flag is the same as the equivalent nbdkit option.) A worked example showing how to use the new option can be found in this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Signed-off-by: Richard W.M. Jones <rjones@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> [eblake: rebase to configure changes, reject --selinux-label if it is not compiled in or not used on a Unix socket] Note that we may relax some of these restrictions at a later date, such as making it possible to label a TCP socket, although it may be smarter to do so as a generic QMP action rather than more one-off command lines in qemu-nbd. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20211115202944.615966-1-eblake@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> [eblake: adjust meson output as suggested by thuth] Signed-off-by: Eric Blake <eblake@redhat.com>
2021-11-15 23:29:43 +03:00
install: true)
subdir('storage-daemon')
foreach exe: [ 'qemu-img', 'qemu-io', 'qemu-nbd', 'qemu-storage-daemon']
traceable += [{
'exe': exe,
'probe-prefix': 'qemu.' + exe.substring(5).replace('-', '_')
}]
endforeach
subdir('contrib/elf2dmp')
executable('qemu-edid', files('qemu-edid.c', 'hw/display/edid-generate.c'),
dependencies: qemuutil,
install: true)
if have_vhost_user
subdir('contrib/vhost-user-blk')
subdir('contrib/vhost-user-gpu')
subdir('contrib/vhost-user-input')
subdir('contrib/vhost-user-scsi')
endif
if host_os == 'linux'
executable('qemu-bridge-helper', files('qemu-bridge-helper.c'),
dependencies: [qemuutil, libcap_ng],
install: true,
install_dir: get_option('libexecdir'))
executable('qemu-pr-helper', files('scsi/qemu-pr-helper.c', 'scsi/utils.c'),
dependencies: [authz, crypto, io, qom, qemuutil,
libcap_ng, mpathpersist],
install: true)
if cpu in ['x86', 'x86_64']
executable('qemu-vmsr-helper', files('tools/i386/qemu-vmsr-helper.c'),
dependencies: [authz, crypto, io, qom, qemuutil,
libcap_ng, mpathpersist],
install: true)
endif
endif
if have_ivshmem
subdir('contrib/ivshmem-client')
subdir('contrib/ivshmem-server')
endif
endif
if stap.found()
foreach t: traceable
foreach stp: [
{'ext': '.stp-build', 'fmt': 'stap', 'bin': meson.current_build_dir() / t['exe'], 'install': false},
{'ext': '.stp', 'fmt': 'stap', 'bin': get_option('prefix') / get_option('bindir') / t['exe'], 'install': true},
{'ext': '-simpletrace.stp', 'fmt': 'simpletrace-stap', 'bin': '', 'install': true},
{'ext': '-log.stp', 'fmt': 'log-stap', 'bin': '', 'install': true},
]
cmd = [
tracetool, '--group=all', '--format=' + stp['fmt'],
'--binary=' + stp['bin'],
'--probe-prefix=' + t['probe-prefix'],
'@INPUT@', '@OUTPUT@'
]
custom_target(t['exe'] + stp['ext'],
input: trace_events_all,
output: t['exe'] + stp['ext'],
install: stp['install'],
install_dir: get_option('datadir') / 'systemtap/tapset',
command: cmd,
depend_files: tracetool_depends)
endforeach
endforeach
endif
subdir('scripts')
subdir('tools')
subdir('pc-bios')
subdir('docs')
subdir('tests')
if gtk.found()
subdir('po')
endif
if host_machine.system() == 'windows'
nsis_cmd = [
find_program('scripts/nsis.py'),
'@OUTPUT@',
get_option('prefix'),
meson.current_source_dir(),
glib_pc.get_variable('bindir'),
host_machine.cpu(),
'--',
'-DDISPLAYVERSION=' + meson.project_version(),
]
if build_docs
nsis_cmd += '-DCONFIG_DOCUMENTATION=y'
endif
if gtk.found()
nsis_cmd += '-DCONFIG_GTK=y'
endif
nsis = custom_target('nsis',
output: 'qemu-setup-' + meson.project_version() + '.exe',
input: files('qemu.nsi'),
build_always_stale: true,
command: nsis_cmd + ['@INPUT@'])
alias_target('installer', nsis)
endif
#########################
# Configuration summary #
#########################
# Build environment
summary_info = {}
summary_info += {'Build directory': meson.current_build_dir()}
summary_info += {'Source path': meson.current_source_dir()}
summary_info += {'Download dependencies': get_option('wrap_mode') != 'nodownload'}
summary(summary_info, bool_yn: true, section: 'Build environment')
# Directories
summary_info += {'Install prefix': get_option('prefix')}
summary_info += {'BIOS directory': qemu_datadir}
pathsep = host_os == 'windows' ? ';' : ':'
summary_info += {'firmware path': pathsep.join(get_option('qemu_firmwarepath'))}
summary_info += {'binary directory': get_option('prefix') / get_option('bindir')}
summary_info += {'library directory': get_option('prefix') / get_option('libdir')}
summary_info += {'module directory': qemu_moddir}
summary_info += {'libexec directory': get_option('prefix') / get_option('libexecdir')}
summary_info += {'include directory': get_option('prefix') / get_option('includedir')}
summary_info += {'config directory': get_option('prefix') / get_option('sysconfdir')}
if host_os != 'windows'
summary_info += {'local state directory': get_option('prefix') / get_option('localstatedir')}
summary_info += {'Manual directory': get_option('prefix') / get_option('mandir')}
else
summary_info += {'local state directory': 'queried at runtime'}
endif
summary_info += {'Doc directory': get_option('prefix') / get_option('docdir')}
summary(summary_info, bool_yn: true, section: 'Directories')
# Host binaries
summary_info = {}
summary_info += {'python': '@0@ (version: @1@)'.format(python.full_path(), python.language_version())}
summary_info += {'sphinx-build': sphinx_build}
# FIXME: the [binaries] section of machine files, which can be probed
# with find_program(), would be great for passing gdb and genisoimage
# paths from configure to Meson. However, there seems to be no way to
# hide a program (for example if gdb is too old).
if config_host.has_key('GDB')
summary_info += {'gdb': config_host['GDB']}
endif
summary_info += {'iasl': iasl}
summary_info += {'genisoimage': config_host['GENISOIMAGE']}
if host_os == 'windows' and have_ga
summary_info += {'wixl': wixl}
endif
if slirp.found() and have_system
summary_info += {'smbd': have_slirp_smbd ? smbd_path : false}
endif
summary(summary_info, bool_yn: true, section: 'Host binaries')
# Configurable features
summary_info = {}
summary_info += {'Documentation': build_docs}
summary_info += {'system-mode emulation': have_system}
summary_info += {'user-mode emulation': have_user}
summary_info += {'block layer': have_block}
summary_info += {'Install blobs': get_option('install_blobs')}
summary_info += {'module support': enable_modules}
if enable_modules
summary_info += {'alternative module path': get_option('module_upgrades')}
endif
summary_info += {'fuzzing support': get_option('fuzzing')}
if have_system
summary_info += {'Audio drivers': ' '.join(audio_drivers_selected)}
endif
summary_info += {'Trace backends': ','.join(get_option('trace_backends'))}
if 'simple' in get_option('trace_backends')
summary_info += {'Trace output file': get_option('trace_file') + '-<pid>'}
endif
summary_info += {'D-Bus display': dbus_display}
summary_info += {'QOM debugging': get_option('qom_cast_debug')}
meson, cutils: allow non-relocatable installs Say QEMU is configured with bindir = "/usr/bin" and a firmware path that starts with "/usr/share/qemu". Ever since QEMU 5.2, QEMU's install has been relocatable: if you move qemu-system-x86_64 from /usr/bin to /home/username/bin, it will start looking for firmware in /home/username/share/qemu. Previously, you would get a non-relocatable install where the moved QEMU will keep looking for firmware in /usr/share/qemu. Windows almost always wants relocatable installs, and in fact that is why QEMU 5.2 introduced relocatability in the first place. However, newfangled distribution mechanisms such as AppImage (https://docs.appimage.org/reference/best-practices.html), and possibly NixOS, also dislike using at runtime the absolute paths that were established at build time. On POSIX systems you almost never care; if you do, your usecase dictates which one is desirable, so there's no single answer. Obviously relocatability works fine most of the time, because not many people have complained about QEMU's switch to relocatable install, and that's why until now there was no way to disable relocatability. But a non-relocatable, non-modular binary can help if you want to do experiments with old firmware and new QEMU or vice versa (because you can just upgrade/downgrade the firmware package, and use rpm2cpio or similar to extract the QEMU binaries outside /usr), so allow both. This patch allows one to build a non-relocatable install using a new option to configure. Why? Because it's not too hard, and because it helps the user double check the relocatability of their install. Note that the same code that handles relocation also lets you run QEMU from the build tree and pick e.g. firmware files from the source tree transparently. Therefore that part remains active with this patch, even if you configure with --disable-relocatable. Suggested-by: Michael Tokarev <mjt@tls.msk.ru> Reviewed-by: Emmanouil Pitsidianakis <manos.pitsidianakis@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-10-05 15:19:34 +03:00
summary_info += {'Relocatable install': get_option('relocatable')}
summary_info += {'vhost-kernel support': have_vhost_kernel}
summary_info += {'vhost-net support': have_vhost_net}
summary_info += {'vhost-user support': have_vhost_user}
summary_info += {'vhost-user-crypto support': have_vhost_user_crypto}
summary_info += {'vhost-user-blk server support': have_vhost_user_blk_server}
summary_info += {'vhost-vdpa support': have_vhost_vdpa}
summary_info += {'build guest agent': have_ga}
summary(summary_info, bool_yn: true, section: 'Configurable features')
# Compilation information
summary_info = {}
summary_info += {'host CPU': cpu}
summary_info += {'host endianness': build_machine.endian()}
summary_info += {'C compiler': ' '.join(meson.get_compiler('c').cmd_array())}
summary_info += {'Host C compiler': ' '.join(meson.get_compiler('c', native: true).cmd_array())}
if 'cpp' in all_languages
summary_info += {'C++ compiler': ' '.join(meson.get_compiler('cpp').cmd_array())}
else
summary_info += {'C++ compiler': false}
endif
if 'objc' in all_languages
summary_info += {'Objective-C compiler': ' '.join(meson.get_compiler('objc').cmd_array())}
else
summary_info += {'Objective-C compiler': false}
endif
option_cflags = (get_option('debug') ? ['-g'] : [])
if get_option('optimization') != 'plain'
option_cflags += ['-O' + get_option('optimization')]
endif
summary_info += {'CFLAGS': ' '.join(get_option('c_args') + option_cflags)}
if 'cpp' in all_languages
summary_info += {'CXXFLAGS': ' '.join(get_option('cpp_args') + option_cflags)}
endif
if 'objc' in all_languages
summary_info += {'OBJCFLAGS': ' '.join(get_option('objc_args') + option_cflags)}
endif
link_args = get_option('c_link_args')
if link_args.length() > 0
summary_info += {'LDFLAGS': ' '.join(link_args)}
endif
summary_info += {'QEMU_CFLAGS': ' '.join(qemu_common_flags + qemu_cflags)}
if 'cpp' in all_languages
summary_info += {'QEMU_CXXFLAGS': ' '.join(qemu_common_flags + qemu_cxxflags)}
endif
if 'objc' in all_languages
summary_info += {'QEMU_OBJCFLAGS': ' '.join(qemu_common_flags)}
endif
summary_info += {'QEMU_LDFLAGS': ' '.join(qemu_ldflags)}
summary_info += {'link-time optimization (LTO)': get_option('b_lto')}
summary_info += {'PIE': get_option('b_pie')}
summary_info += {'static build': get_option('prefer_static')}
summary_info += {'malloc trim support': has_malloc_trim}
summary_info += {'membarrier': have_membarrier}
summary_info += {'debug graph lock': get_option('debug_graph_lock')}
summary_info += {'debug stack usage': get_option('debug_stack_usage')}
summary_info += {'mutex debugging': get_option('debug_mutex')}
summary_info += {'memory allocator': get_option('malloc')}
summary_info += {'avx2 optimization': config_host_data.get('CONFIG_AVX2_OPT')}
summary_info += {'avx512bw optimization': config_host_data.get('CONFIG_AVX512BW_OPT')}
summary_info += {'gcov': get_option('b_coverage')}
summary_info += {'thread sanitizer': get_option('tsan')}
summary_info += {'CFI support': get_option('cfi')}
if get_option('cfi')
summary_info += {'CFI debug support': get_option('cfi_debug')}
endif
summary_info += {'strip binaries': get_option('strip')}
summary_info += {'sparse': sparse}
summary_info += {'mingw32 support': host_os == 'windows'}
summary(summary_info, bool_yn: true, section: 'Compilation')
# snarf the cross-compilation information for tests
summary_info = {}
have_cross = false
foreach target: target_dirs
tcg_mak = meson.current_build_dir() / 'tests/tcg' / target / 'config-target.mak'
if fs.exists(tcg_mak)
config_cross_tcg = keyval.load(tcg_mak)
if 'CC' in config_cross_tcg
summary_info += {config_cross_tcg['TARGET_NAME']: config_cross_tcg['CC']}
have_cross = true
endif
endif
endforeach
if have_cross
summary(summary_info, bool_yn: true, section: 'Cross compilers')
endif
# Targets and accelerators
summary_info = {}
if have_system
summary_info += {'KVM support': config_all_accel.has_key('CONFIG_KVM')}
summary_info += {'HVF support': config_all_accel.has_key('CONFIG_HVF')}
summary_info += {'WHPX support': config_all_accel.has_key('CONFIG_WHPX')}
summary_info += {'NVMM support': config_all_accel.has_key('CONFIG_NVMM')}
summary_info += {'Xen support': xen.found()}
if xen.found()
summary_info += {'xen ctrl version': xen.version()}
endif
summary_info += {'Xen emulation': config_all_devices.has_key('CONFIG_XEN_EMU')}
endif
summary_info += {'TCG support': config_all_accel.has_key('CONFIG_TCG')}
if config_all_accel.has_key('CONFIG_TCG')
if get_option('tcg_interpreter')
summary_info += {'TCG backend': 'TCI (TCG with bytecode interpreter, slow)'}
else
summary_info += {'TCG backend': 'native (@0@)'.format(cpu)}
endif
summary_info += {'TCG plugins': get_option('plugins')}
summary_info += {'TCG debug enabled': get_option('debug_tcg')}
if have_linux_user or have_bsd_user
summary_info += {'syscall buffer debugging support': get_option('debug_remap')}
endif
endif
summary_info += {'target list': ' '.join(target_dirs)}
if have_system
summary_info += {'default devices': get_option('default_devices')}
summary_info += {'out of process emulation': multiprocess_allowed}
summary_info += {'vfio-user server': vfio_user_server_allowed}
endif
summary(summary_info, bool_yn: true, section: 'Targets and accelerators')
# Block layer
summary_info = {}
summary_info += {'coroutine backend': coroutine_backend}
summary_info += {'coroutine pool': have_coroutine_pool}
if have_block
summary_info += {'Block whitelist (rw)': get_option('block_drv_rw_whitelist')}
summary_info += {'Block whitelist (ro)': get_option('block_drv_ro_whitelist')}
summary_info += {'Use block whitelist in tools': get_option('block_drv_whitelist_in_tools')}
summary_info += {'VirtFS (9P) support': have_virtfs}
summary_info += {'replication support': config_host_data.get('CONFIG_REPLICATION')}
summary_info += {'bochs support': get_option('bochs').allowed()}
summary_info += {'cloop support': get_option('cloop').allowed()}
summary_info += {'dmg support': get_option('dmg').allowed()}
summary_info += {'qcow v1 support': get_option('qcow1').allowed()}
summary_info += {'vdi support': get_option('vdi').allowed()}
summary_info += {'vhdx support': get_option('vhdx').allowed()}
summary_info += {'vmdk support': get_option('vmdk').allowed()}
summary_info += {'vpc support': get_option('vpc').allowed()}
summary_info += {'vvfat support': get_option('vvfat').allowed()}
summary_info += {'qed support': get_option('qed').allowed()}
summary_info += {'parallels support': get_option('parallels').allowed()}
summary_info += {'FUSE exports': fuse}
summary_info += {'VDUSE block exports': have_vduse_blk_export}
endif
summary(summary_info, bool_yn: true, section: 'Block layer support')
# Crypto
summary_info = {}
summary_info += {'TLS priority': get_option('tls_priority')}
summary_info += {'GNUTLS support': gnutls}
if gnutls.found()
summary_info += {' GNUTLS crypto': gnutls_crypto.found()}
endif
summary_info += {'libgcrypt': gcrypt}
summary_info += {'nettle': nettle}
if nettle.found()
summary_info += {' XTS': xts != 'private'}
endif
summary_info += {'SM4 ALG support': crypto_sm4}
summary_info += {'AF_ALG support': have_afalg}
summary_info += {'rng-none': get_option('rng_none')}
summary_info += {'Linux keyring': have_keyring}
summary_info += {'Linux keyutils': keyutils}
summary(summary_info, bool_yn: true, section: 'Crypto')
# UI
summary_info = {}
if host_os == 'darwin'
summary_info += {'Cocoa support': cocoa}
endif
summary_info += {'SDL support': sdl}
summary_info += {'SDL image support': sdl_image}
summary_info += {'GTK support': gtk}
summary_info += {'pixman': pixman}
summary_info += {'VTE support': vte}
summary_info += {'PNG support': png}
summary_info += {'VNC support': vnc}
if vnc.found()
summary_info += {'VNC SASL support': sasl}
summary_info += {'VNC JPEG support': jpeg}
endif
summary_info += {'spice protocol support': spice_protocol}
if spice_protocol.found()
summary_info += {' spice server support': spice}
endif
summary_info += {'curses support': curses}
summary_info += {'brlapi support': brlapi}
summary(summary_info, bool_yn: true, section: 'User interface')
# Graphics backends
summary_info = {}
summary_info += {'VirGL support': virgl}
summary_info += {'Rutabaga support': rutabaga}
summary(summary_info, bool_yn: true, section: 'Graphics backends')
# Audio backends
summary_info = {}
if host_os not in ['darwin', 'haiku', 'windows']
summary_info += {'OSS support': oss}
summary_info += {'sndio support': sndio}
elif host_os == 'darwin'
summary_info += {'CoreAudio support': coreaudio}
elif host_os == 'windows'
summary_info += {'DirectSound support': dsound}
endif
if host_os == 'linux'
summary_info += {'ALSA support': alsa}
summary_info += {'PulseAudio support': pulse}
endif
summary_info += {'PipeWire support': pipewire}
summary_info += {'JACK support': jack}
summary(summary_info, bool_yn: true, section: 'Audio backends')
# Network backends
summary_info = {}
if host_os == 'darwin'
summary_info += {'vmnet.framework support': vmnet}
endif
net: add initial support for AF_XDP network backend AF_XDP is a network socket family that allows communication directly with the network device driver in the kernel, bypassing most or all of the kernel networking stack. In the essence, the technology is pretty similar to netmap. But, unlike netmap, AF_XDP is Linux-native and works with any network interfaces without driver modifications. Unlike vhost-based backends (kernel, user, vdpa), AF_XDP doesn't require access to character devices or unix sockets. Only access to the network interface itself is necessary. This patch implements a network backend that communicates with the kernel by creating an AF_XDP socket. A chunk of userspace memory is shared between QEMU and the host kernel. 4 ring buffers (Tx, Rx, Fill and Completion) are placed in that memory along with a pool of memory buffers for the packet data. Data transmission is done by allocating one of the buffers, copying packet data into it and placing the pointer into Tx ring. After transmission, device will return the buffer via Completion ring. On Rx, device will take a buffer form a pre-populated Fill ring, write the packet data into it and place the buffer into Rx ring. AF_XDP network backend takes on the communication with the host kernel and the network interface and forwards packets to/from the peer device in QEMU. Usage example: -device virtio-net-pci,netdev=guest1,mac=00:16:35:AF:AA:5C -netdev af-xdp,ifname=ens6f1np1,id=guest1,mode=native,queues=1 XDP program bridges the socket with a network interface. It can be attached to the interface in 2 different modes: 1. skb - this mode should work for any interface and doesn't require driver support. With a caveat of lower performance. 2. native - this does require support from the driver and allows to bypass skb allocation in the kernel and potentially use zero-copy while getting packets in/out userspace. By default, QEMU will try to use native mode and fall back to skb. Mode can be forced via 'mode' option. To force 'copy' even in native mode, use 'force-copy=on' option. This might be useful if there is some issue with the driver. Option 'queues=N' allows to specify how many device queues should be open. Note that all the queues that are not open are still functional and can receive traffic, but it will not be delivered to QEMU. So, the number of device queues should generally match the QEMU configuration, unless the device is shared with something else and the traffic re-direction to appropriate queues is correctly configured on a device level (e.g. with ethtool -N). 'start-queue=M' option can be used to specify from which queue id QEMU should start configuring 'N' queues. It might also be necessary to use this option with certain NICs, e.g. MLX5 NICs. See the docs for examples. In a general case QEMU will need CAP_NET_ADMIN and CAP_SYS_ADMIN or CAP_BPF capabilities in order to load default XSK/XDP programs to the network interface and configure BPF maps. It is possible, however, to run with no capabilities. For that to work, an external process with enough capabilities will need to pre-load default XSK program, create AF_XDP sockets and pass their file descriptors to QEMU process on startup via 'sock-fds' option. Network backend will need to be configured with 'inhibit=on' to avoid loading of the program. QEMU will need 32 MB of locked memory (RLIMIT_MEMLOCK) per queue or CAP_IPC_LOCK. There are few performance challenges with the current network backends. First is that they do not support IO threads. This means that data path is handled by the main thread in QEMU and may slow down other work or may be slowed down by some other work. This also means that taking advantage of multi-queue is generally not possible today. Another thing is that data path is going through the device emulation code, which is not really optimized for performance. The fastest "frontend" device is virtio-net. But it's not optimized for heavy traffic either, because it expects such use-cases to be handled via some implementation of vhost (user, kernel, vdpa). In practice, we have virtio notifications and rcu lock/unlock on a per-packet basis and not very efficient accesses to the guest memory. Communication channels between backend and frontend devices do not allow passing more than one packet at a time as well. Some of these challenges can be avoided in the future by adding better batching into device emulation or by implementing vhost-af-xdp variant. There are also a few kernel limitations. AF_XDP sockets do not support any kinds of checksum or segmentation offloading. Buffers are limited to a page size (4K), i.e. MTU is limited. Multi-buffer support implementation for AF_XDP is in progress, but not ready yet. Also, transmission in all non-zero-copy modes is synchronous, i.e. done in a syscall. That doesn't allow high packet rates on virtual interfaces. However, keeping in mind all of these challenges, current implementation of the AF_XDP backend shows a decent performance while running on top of a physical NIC with zero-copy support. Test setup: 2 VMs running on 2 physical hosts connected via ConnectX6-Dx card. Network backend is configured to open the NIC directly in native mode. The driver supports zero-copy. NIC is configured to use 1 queue. Inside a VM - iperf3 for basic TCP performance testing and dpdk-testpmd for PPS testing. iperf3 result: TCP stream : 19.1 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 3.4 Mpps Rx only : 2.0 Mpps L2 FWD Loopback : 1.5 Mpps In skb mode the same setup shows much lower performance, similar to the setup where pair of physical NICs is replaced with veth pair: iperf3 result: TCP stream : 9 Gbps dpdk-testpmd (single queue, single CPU core, 64 B packets) results: Tx only : 1.2 Mpps Rx only : 1.0 Mpps L2 FWD Loopback : 0.7 Mpps Results in skb mode or over the veth are close to results of a tap backend with vhost=on and disabled segmentation offloading bridged with a NIC. Signed-off-by: Ilya Maximets <i.maximets@ovn.org> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> (docker/lcitool) Signed-off-by: Jason Wang <jasowang@redhat.com>
2023-09-13 21:34:37 +03:00
summary_info += {'AF_XDP support': libxdp}
summary_info += {'slirp support': slirp}
summary_info += {'vde support': vde}
summary_info += {'netmap support': have_netmap}
summary_info += {'l2tpv3 support': have_l2tpv3}
summary(summary_info, bool_yn: true, section: 'Network backends')
# Libraries
summary_info = {}
summary_info += {'libtasn1': tasn1}
summary_info += {'PAM': pam}
summary_info += {'iconv support': iconv}
blkio: add libblkio block driver libblkio (https://gitlab.com/libblkio/libblkio/) is a library for high-performance disk I/O. It currently supports io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa with additional drivers under development. One of the reasons for developing libblkio is that other applications besides QEMU can use it. This will be particularly useful for virtio-blk-vhost-user which applications may wish to use for connecting to qemu-storage-daemon. libblkio also gives us an opportunity to develop in Rust behind a C API that is easy to consume from QEMU. This commit adds io_uring, nvme-io_uring, virtio-blk-vhost-user, and virtio-blk-vhost-vdpa BlockDrivers to QEMU using libblkio. It will be easy to add other libblkio drivers since they will share the majority of code. For now I/O buffers are copied through bounce buffers if the libblkio driver requires it. Later commits add an optimization for pre-registering guest RAM to avoid bounce buffers. The syntax is: --blockdev io_uring,node-name=drive0,filename=test.img,readonly=on|off,cache.direct=on|off --blockdev nvme-io_uring,node-name=drive0,filename=/dev/ng0n1,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-vdpa,node-name=drive0,path=/dev/vdpa...,readonly=on|off,cache.direct=on --blockdev virtio-blk-vhost-user,node-name=drive0,path=vhost-user-blk.sock,readonly=on|off,cache.direct=on Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Acked-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Message-id: 20221013185908.1297568-3-stefanha@redhat.com Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
2022-10-13 21:58:57 +03:00
summary_info += {'blkio support': blkio}
summary_info += {'curl support': curl}
summary_info += {'Multipath support': mpathpersist}
summary_info += {'Linux AIO support': libaio}
summary_info += {'Linux io_uring support': linux_io_uring}
summary_info += {'ATTR/XATTR support': libattr}
summary_info += {'RDMA support': rdma}
summary_info += {'fdt support': fdt_opt == 'internal' ? 'internal' : fdt}
summary_info += {'libcap-ng support': libcap_ng}
summary_info += {'bpf support': libbpf}
summary_info += {'rbd support': rbd}
summary_info += {'smartcard support': cacard}
summary_info += {'U2F support': u2f}
summary_info += {'libusb': libusb}
summary_info += {'usb net redir': usbredir}
summary_info += {'OpenGL support (epoxy)': opengl}
summary_info += {'GBM': gbm}
summary_info += {'libiscsi support': libiscsi}
summary_info += {'libnfs support': libnfs}
if host_os == 'windows'
if have_ga
summary_info += {'QGA VSS support': have_qga_vss}
endif
endif
summary_info += {'seccomp support': seccomp}
summary_info += {'GlusterFS support': glusterfs}
Add Hyper-V Dynamic Memory Protocol driver (hv-balloon) base This driver is like virtio-balloon on steroids: it allows both changing the guest memory allocation via ballooning and (in the next patch) inserting pieces of extra RAM into it on demand from a provided memory backend. The actual resizing is done via ballooning interface (for example, via the "balloon" HMP command). This includes resizing the guest past its boot size - that is, hot-adding additional memory in granularity limited only by the guest alignment requirements, as provided by the next patch. In contrast with ACPI DIMM hotplug where one can only request to unplug a whole DIMM stick this driver allows removing memory from guest in single page (4k) units via ballooning. After a VM reboot the guest is back to its original (boot) size. In the future, the guest boot memory size might be changed on reboot instead, taking into account the effective size that VM had before that reboot (much like Hyper-V does). For performance reasons, the guest-released memory is tracked in a few range trees, as a series of (start, count) ranges. Each time a new page range is inserted into such tree its neighbors are checked as candidates for possible merging with it. Besides performance reasons, the Dynamic Memory protocol itself uses page ranges as the data structure in its messages, so relevant pages need to be merged into such ranges anyway. One has to be careful when tracking the guest-released pages, since the guest can maliciously report returning pages outside its current address space, which later clash with the address range of newly added memory. Similarly, the guest can report freeing the same page twice. The above design results in much better ballooning performance than when using virtio-balloon with the same guest: 230 GB / minute with this driver versus 70 GB / minute with virtio-balloon. During a ballooning operation most of time is spent waiting for the guest to come up with newly freed page ranges, processing the received ranges on the host side (in QEMU and KVM) is nearly instantaneous. The unballoon operation is also pretty much instantaneous: thanks to the merging of the ballooned out page ranges 200 GB of memory can be returned to the guest in about 1 second. With virtio-balloon this operation takes about 2.5 minutes. These tests were done against a Windows Server 2019 guest running on a Xeon E5-2699, after dirtying the whole memory inside guest before each balloon operation. Using a range tree instead of a bitmap to track the removed memory also means that the solution scales well with the guest size: even a 1 TB range takes just a few bytes of such metadata. Since the required GTree operations aren't present in every Glib version a check for them was added to the meson build script, together with new "--enable-hv-balloon" and "--disable-hv-balloon" configure arguments. If these GTree operations are missing in the system's Glib version this driver will be skipped during QEMU build. An optional "status-report=on" device parameter requests memory status events from the guest (typically sent every second), which allow the host to learn both the guest memory available and the guest memory in use counts. Following commits will add support for their external emission as "HV_BALLOON_STATUS_REPORT" QMP events. The driver is named hv-balloon since the Linux kernel client driver for the Dynamic Memory Protocol is named as such and to follow the naming pattern established by the virtio-balloon driver. The whole protocol runs over Hyper-V VMBus. The driver was tested against Windows Server 2012 R2, Windows Server 2016 and Windows Server 2019 guests and obeys the guest alignment requirements reported to the host via DM_CAPABILITIES_REPORT message. Acked-by: David Hildenbrand <david@redhat.com> Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
2023-06-12 17:00:54 +03:00
summary_info += {'hv-balloon support': hv_balloon}
summary_info += {'TPM support': have_tpm}
summary_info += {'libssh support': libssh}
summary_info += {'lzo support': lzo}
summary_info += {'snappy support': snappy}
summary_info += {'bzip2 support': libbzip2}
summary_info += {'lzfse support': liblzfse}
summary_info += {'zstd support': zstd}
summary_info += {'Query Processing Library support': qpl}
summary_info += {'UADK Library support': uadk}
summary_info += {'qatzip support': qatzip}
summary_info += {'NUMA host support': numa}
summary_info += {'capstone': capstone}
summary_info += {'libpmem support': libpmem}
summary_info += {'libdaxctl support': libdaxctl}
summary_info += {'libudev': libudev}
# Dummy dependency, keep .found()
summary_info += {'FUSE lseek': fuse_lseek.found()}
nbd/server: Add --selinux-label option Under SELinux, Unix domain sockets have two labels. One is on the disk and can be set with commands such as chcon(1). There is a different label stored in memory (called the process label). This can only be set by the process creating the socket. When using SELinux + SVirt and wanting qemu to be able to connect to a qemu-nbd instance, you must set both labels correctly first. For qemu-nbd the options to set the second label are awkward. You can create the socket in a wrapper program and then exec into qemu-nbd. Or you could try something with LD_PRELOAD. This commit adds the ability to set the label straightforwardly on the command line, via the new --selinux-label flag. (The name of the flag is the same as the equivalent nbdkit option.) A worked example showing how to use the new option can be found in this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1984938 Signed-off-by: Richard W.M. Jones <rjones@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> [eblake: rebase to configure changes, reject --selinux-label if it is not compiled in or not used on a Unix socket] Note that we may relax some of these restrictions at a later date, such as making it possible to label a TCP socket, although it may be smarter to do so as a generic QMP action rather than more one-off command lines in qemu-nbd. Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20211115202944.615966-1-eblake@redhat.com> Reviewed-by: Thomas Huth <thuth@redhat.com> [eblake: adjust meson output as suggested by thuth] Signed-off-by: Eric Blake <eblake@redhat.com>
2021-11-15 23:29:43 +03:00
summary_info += {'selinux': selinux}
summary_info += {'libdw': libdw}
if host_os == 'freebsd'
summary_info += {'libinotify-kqueue': inotify}
endif
summary(summary_info, bool_yn: true, section: 'Dependencies')
if host_arch == 'unknown'
message()
warning('UNSUPPORTED HOST CPU')
message()
message('Support for CPU host architecture ' + cpu + ' is not currently')
message('maintained. The QEMU project does not guarantee that QEMU will')
message('compile or work on this host CPU. You can help by volunteering')
message('to maintain it and providing a build host for our continuous')
message('integration setup.')
if get_option('tcg').allowed() and target_dirs.length() > 0
message()
message('configure has succeeded and you can continue to build, but')
message('QEMU will use a slow interpreter to emulate the target CPU.')
endif
endif
if not supported_oses.contains(host_os)
message()
warning('UNSUPPORTED HOST OS')
message()
message('Support for host OS ' + host_os + 'is not currently maintained.')
message('configure has succeeded and you can continue to build, but')
message('the QEMU project does not guarantee that QEMU will compile or')
message('work on this operating system. You can help by volunteering')
message('to maintain it and providing a build host for our continuous')
message('integration setup. This will ensure that future versions of QEMU')
message('will keep working on ' + host_os + '.')
endif
if host_arch == 'unknown' or not supported_oses.contains(host_os)
message()
message('If you want to help supporting QEMU on this platform, please')
message('contact the developers at qemu-devel@nongnu.org.')
endif
meson, cutils: allow non-relocatable installs Say QEMU is configured with bindir = "/usr/bin" and a firmware path that starts with "/usr/share/qemu". Ever since QEMU 5.2, QEMU's install has been relocatable: if you move qemu-system-x86_64 from /usr/bin to /home/username/bin, it will start looking for firmware in /home/username/share/qemu. Previously, you would get a non-relocatable install where the moved QEMU will keep looking for firmware in /usr/share/qemu. Windows almost always wants relocatable installs, and in fact that is why QEMU 5.2 introduced relocatability in the first place. However, newfangled distribution mechanisms such as AppImage (https://docs.appimage.org/reference/best-practices.html), and possibly NixOS, also dislike using at runtime the absolute paths that were established at build time. On POSIX systems you almost never care; if you do, your usecase dictates which one is desirable, so there's no single answer. Obviously relocatability works fine most of the time, because not many people have complained about QEMU's switch to relocatable install, and that's why until now there was no way to disable relocatability. But a non-relocatable, non-modular binary can help if you want to do experiments with old firmware and new QEMU or vice versa (because you can just upgrade/downgrade the firmware package, and use rpm2cpio or similar to extract the QEMU binaries outside /usr), so allow both. This patch allows one to build a non-relocatable install using a new option to configure. Why? Because it's not too hard, and because it helps the user double check the relocatability of their install. Note that the same code that handles relocation also lets you run QEMU from the build tree and pick e.g. firmware files from the source tree transparently. Therefore that part remains active with this patch, even if you configure with --disable-relocatable. Suggested-by: Michael Tokarev <mjt@tls.msk.ru> Reviewed-by: Emmanouil Pitsidianakis <manos.pitsidianakis@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-10-05 15:19:34 +03:00
actually_reloc = get_option('relocatable')
# check if get_relocated_path() is actually able to relocate paths
if get_option('relocatable') and \
not (get_option('prefix') / get_option('bindir')).startswith(get_option('prefix') / '')
message()
warning('bindir not included within prefix, the installation will not be relocatable.')
actually_reloc = false
endif
if not actually_reloc and (host_os == 'windows' or get_option('relocatable'))
if host_os == 'windows'
meson, cutils: allow non-relocatable installs Say QEMU is configured with bindir = "/usr/bin" and a firmware path that starts with "/usr/share/qemu". Ever since QEMU 5.2, QEMU's install has been relocatable: if you move qemu-system-x86_64 from /usr/bin to /home/username/bin, it will start looking for firmware in /home/username/share/qemu. Previously, you would get a non-relocatable install where the moved QEMU will keep looking for firmware in /usr/share/qemu. Windows almost always wants relocatable installs, and in fact that is why QEMU 5.2 introduced relocatability in the first place. However, newfangled distribution mechanisms such as AppImage (https://docs.appimage.org/reference/best-practices.html), and possibly NixOS, also dislike using at runtime the absolute paths that were established at build time. On POSIX systems you almost never care; if you do, your usecase dictates which one is desirable, so there's no single answer. Obviously relocatability works fine most of the time, because not many people have complained about QEMU's switch to relocatable install, and that's why until now there was no way to disable relocatability. But a non-relocatable, non-modular binary can help if you want to do experiments with old firmware and new QEMU or vice versa (because you can just upgrade/downgrade the firmware package, and use rpm2cpio or similar to extract the QEMU binaries outside /usr), so allow both. This patch allows one to build a non-relocatable install using a new option to configure. Why? Because it's not too hard, and because it helps the user double check the relocatability of their install. Note that the same code that handles relocation also lets you run QEMU from the build tree and pick e.g. firmware files from the source tree transparently. Therefore that part remains active with this patch, even if you configure with --disable-relocatable. Suggested-by: Michael Tokarev <mjt@tls.msk.ru> Reviewed-by: Emmanouil Pitsidianakis <manos.pitsidianakis@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-10-05 15:19:34 +03:00
message()
warning('Windows installs should usually be relocatable.')
endif
message()
message('QEMU will have to be installed under ' + get_option('prefix') + '.')
message('Use --disable-relocatable to remove this warning.')
endif