* remotes/huth-gitlab/tags/pull-request-2019-03-08:
cirrus.yml: Add macOS continuous integration task
tests/bios-tables: Improve portability by searching bash in the $PATH
vhost-user-test: fix leaks
tests: Do not use "\n" in g_test_message() strings
hw/devices: Remove unused TC6393XB_RAM definition
hw: Remove unused 'hw/devices.h' include
tests: Move qdict-test-data.txt to tests/data/qobject/
Signed-off-by: Peter Maydell <[email protected]>
# Conflicts:
# tests/vhost-user-test.c
* remotes/bonzini/tags/for-upstream-qgraph: (57 commits)
qos-test: megasas test node
qos-test: e1000 test node
qos-test: eepro100 test node
qos-test: es1370 test node
qos-test: vmxnet3 test node
qos-test: usb-hcd-ohci test node
qos-test: spapr-phb test node
qos-test: pcnet test node
qos-test: nvme test node
qos-test: ne2k_pci test node
qos-test: ipoctal232 test node
qos-test: tpci200 test node
qos-test: ac97 test node
tests: move virtio entirely to qos-test
tests/libqos: remove pre-qgraph QVirtioPCIDevice API
qos-test: virtio-scsi test node
tests/libqos: virtio-scsi driver and interface nodes
qos-test: vhost-user test node
vhost-user-test: always use 256 MiB of guest memory
tests/libqos: support multiqueue for virtio-net
...
Peter Maydell [Fri, 8 Mar 2019 11:55:53 +0000 (11:55 +0000)]
Merge remote-tracking branch 'remotes/thibault/tags/samuel-thibault' into staging
Slirp updates
Greg Kurz (1):
slirp: Fix build with gcc 9
Marc-André Lureau (7):
slirp: adapt a subset of QEMU vmstate code
slirp: use libslirp migration code
slirp: use "slirp_" prefix for inet_aton() win32 implementation
slirp: move sources to src/ subdirectory
slirp: add a standalone Makefile
build-sys: link with slirp as an external project
slirp: remove QEMU Makefile.objs
Samuel Thibault (2):
slirp: fix big/little endian conversion in ident protocol
slirp: Mark pieces missing IPv6 support
Vic Lee (1):
slirp: check for ioctlsocket error and 0-length udp payload.
William Bowling (1):
slirp: check sscanf result when emulating ident
# gpg: Signature made Thu 07 Mar 2019 11:51:20 GMT
# gpg: using RSA key E61DBB15D4172BDEC97E92D9DB550E89F0FA54F3
# gpg: Good signature from "Samuel Thibault <[email protected]>" [unknown]
# gpg: aka "Samuel Thibault <[email protected]>" [marginal]
# gpg: aka "Samuel Thibault <[email protected]>" [unknown]
# gpg: aka "Samuel Thibault <[email protected]>" [marginal]
# gpg: aka "Samuel Thibault <[email protected]>" [marginal]
# gpg: aka "Samuel Thibault <[email protected]>" [marginal]
# gpg: aka "Samuel Thibault <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 900C B024 B679 31D4 0F82 304B D017 8C76 7D06 9EE6
# Subkey fingerprint: E61D BB15 D417 2BDE C97E 92D9 DB55 0E89 F0FA 54F3
* remotes/thibault/tags/samuel-thibault:
slirp: remove QEMU Makefile.objs
build-sys: link with slirp as an external project
slirp: add a standalone Makefile
slirp: move sources to src/ subdirectory
slirp: use "slirp_" prefix for inet_aton() win32 implementation
slirp: use libslirp migration code
slirp: adapt a subset of QEMU vmstate code
slirp: Mark pieces missing IPv6 support
slirp: fix big/little endian conversion in ident protocol
slirp: check sscanf result when emulating ident
slirp: check for ioctlsocket error and 0-length udp payload.
slirp: Fix build with gcc 9
* remotes/bonzini/tags/for-upstream-kconfig: (54 commits)
kconfig: add documentation
.travis.yml: test that no-default-device builds do not regress
xtensa-softmmu.mak: express dependencies with Kconfig
unicore32-softmmu.mak: express dependencies with Kconfig
sparc64-softmmu.mak: express dependencies with Kconfig
sparc-softmmu.mak: express dependencies with Kconfig
sh4-softmmu.mak: express dependencies with Kconfig
s390x: express dependencies with Kconfig
riscv-softmmu.mak: replace CONFIG_* with Kconfig "select" directives
or1k-softmmu.mak: express dependencies with Kconfig
nios2-softmmu.mak: express dependencies with Kconfig
moxie-softmmu.mak: express dependencies with Kconfig
microblaze-softmmu.mak: express dependencies with Kconfig
m68k-softmmu.mak: express dependencies with Kconfig
lm32-softmmu.mak: express dependencies with Kconfig
hppa-softmmu.mak: express dependencies with Kconfig
cris-softmmu.mak: express dependencies with Kconfig
alpha-softmmu.mak: express dependencies with Kconfig
ppc: Express dependencies of the embedded machines with kconfig
ppc: Express dependencies of the Sam460EX machines with kconfig
...
Thomas Huth [Sun, 3 Mar 2019 14:42:59 +0000 (15:42 +0100)]
cirrus.yml: Add macOS continuous integration task
cirrus-ci.com also has the possibility to run CI tasks on macOS.
Since most of the QEMU developers do not have access to macOS yet,
let's add a CI pipeline for this operating system here, too.
tests/bios-tables: Improve portability by searching bash in the $PATH
Bash is not always installed as /bin/bash. In particular on OpenBSD,
the package installs it in /usr/local/bin.
Use the 'env' shebang to search bash in the $PATH.
tests: Move qdict-test-data.txt to tests/data/qobject/
Move qdict-test-data.txt to the tests/data/qobject/ subdirectory,
and remove the unnecessary symlinking.
(See 4b2ff65a1f3 for similar test-data cleanup).
Thomas Huth [Tue, 29 Jan 2019 09:42:14 +0000 (10:42 +0100)]
s390x: express dependencies with Kconfig
Instead of hard-coding all config switches in the config file
default-configs/s390x-softmmu.mak, let's use the new Kconfig files
to express the necessary dependencies: The S390_CCW_VIRTIO config switch
for the "s390-ccw-virtio" machine now selects all non-optional devices.
And since we already have the VIRTIO_PCI and VIRTIO_MMIO config switches
for the other two virtio transports, this patch also introduces a new
config switch VIRTIO_CCW for the third, s390x-specific virtio transport,
so that all three virtio transports are now handled in the same way.
Thomas Huth [Thu, 31 Jan 2019 13:53:29 +0000 (14:53 +0100)]
ppc: Express dependencies of the Sam460EX machines with kconfig
Most of the dependencies are now directly selected by the SAM460EX
switch. We can drop CONFIG_VGA_CIRRUS since this device is already
selected automatically when CONFIG_PCI_DEVICES is set.
Thomas Huth [Thu, 31 Jan 2019 13:53:27 +0000 (14:53 +0100)]
ppc: Express dependencies of the 'prep' and '40p' machines with kconfig
Select the required devices in hw/ppc/Kconfig instead, so that
ppc-softmmu.mak only contains the user-selectable PREP switch.
Plug-in devices like NE2000_ISA are pulled in automatically by the
Kconfig build system now.
Thomas Huth [Thu, 31 Jan 2019 13:53:26 +0000 (14:53 +0100)]
ppc64: Express dependencies of 'pseries' and 'powernv' machines with kconfig
The POWERNV switch should always select ISA_IPMI_BT, then the other
IPMI options are turned on automatically now.
CONFIG_DIMM should always be selected by the pseries machine,
which in turn depends on CONFIG_MEM_DEVICE since DIMM implements
this interface.
CONFIG_VIRTIO_VGA can be dropped from default-configs/ppc64-softmmu.mak
completely since this device is already automatically enabled via
hw/display/Kconfig now.
Paolo Bonzini [Wed, 23 Jan 2019 06:56:06 +0000 (14:56 +0800)]
build: convert usb.mak to Kconfig
Instead of including the same list of devices for each target,
let the host controllers select CONFIG_USB and make the devices
default to present whenever USB is available.
Done with the following script:
while read i; do
i=${i%=y}; i=${i#CONFIG_}
sed -i -e'/^config '$i'$/!b' -en \
-e'a\' -e' default y\' -e' depends on USB' \
`grep -lw $i hw/*/Kconfig`
done < default-configs/usb.mak
followed by adding "select USB" on the host controllers.
Paolo Bonzini [Wed, 23 Jan 2019 06:56:05 +0000 (14:56 +0800)]
build: convert sound.mak to Kconfig
There is really nothing special in these devices; they are just
ISA devices. Instead of including them for each target,
set CONFIG_ISA_BUS to true, and make the devices default to present
whenever ISA is available. More conversion of ISA devices will
follow.
Done with the following script:
while read i; do
i=${i%=y}; i=${i#CONFIG_}
sed -i -e'/^config '$i'$/!b' -en \
-e'a\' -e' default y\' -e' depends on ISA_BUS' \
`grep -lw $i hw/*/Kconfig`
done < default-configs/sound.mak
Paolo Bonzini [Wed, 23 Jan 2019 06:56:04 +0000 (14:56 +0800)]
build: convert pci.mak to Kconfig
Instead of including the same list of devices for each target,
set CONFIG_PCI to true, and make the devices default to present
whenever PCI is available. However, s390x does not want all the
PCI devices, so there is a separate symbol to enable them.
Done mostly with the following script:
while read i; do
i=${i%=y}; i=${i#CONFIG_}
sed -i -e'/^config '$i'$/!b' -en \
-e'a\' -e' default y if PCI_DEVICES\' -e' depends on PCI' \
`grep -lw $i hw/*/Kconfig`
done < default-configs/pci.mak
followed by replacing a few "depends on" clauses with "select"
whenever the symbol is not really related to PCI.
Paolo Bonzini [Fri, 25 Jan 2019 05:48:56 +0000 (06:48 +0100)]
kconfig: introduce CONFIG_TEST_DEVICES
Devices that are mostly used for testing purposes (for example in
endianness-test) will be moved under a new symbol CONFIG_TEST_DEVICES
that can be disabled in the default-configs file. This makes
it easier to drop this code from QEMU if desirable.
Paolo Bonzini [Wed, 23 Jan 2019 06:56:17 +0000 (14:56 +0800)]
minikconfig: implement allnoconfig and defconfig modes
Apart from defconfig (which is a no-op),
allyesconfig/allnoconfig/randcondfig can be implemented simply by ignoring
the RHS of assignments and "default" statements. The RHS is replaced
respectively by "true", "false" or a random value.
However, allyesconfig and randconfig do not quite work, because all the
files for hw/ARCH/Kconfig are sourced and therefore you could end up
enabling some ARM boards in x86 or things like that. This is left for
future work, but I am leaving it in to help debugging minikconf itself.
allnoconfig mode is tied to a new configure option, --without-default-devices.
Paolo Bonzini [Wed, 23 Jan 2019 06:56:01 +0000 (14:56 +0800)]
build: switch to Kconfig
The make_device_config.sh script is replaced by minikconf, which
is modified to support the same command line as its predecessor.
The roots of the parsing are default-configs/*.mak, Kconfig.host and
hw/Kconfig. One difference with make_device_config.sh is that all symbols
have to be defined in a Kconfig file, including those coming from the
configure script. This is the reason for the Kconfig.host file introduced
in the previous patch. Whenever a file in default-configs/*.mak used
$(...) to refer to a config-host.mak symbol, this is replaced by a
Kconfig dependency; this part must be done already in this patch
for bisectability.
Paolo Bonzini [Wed, 23 Jan 2019 06:56:00 +0000 (14:56 +0800)]
kconfig: introduce kconfig files
The Kconfig files were generated mostly with this script:
for i in `grep -ho CONFIG_[A-Z0-9_]* default-configs/* | sort -u`; do
set fnord `git grep -lw $i -- 'hw/*/Makefile.objs' `
shift
if test $# = 1; then
cat >> $(dirname $1)/Kconfig << EOF
config ${i#CONFIG_}
bool
EOF
git add $(dirname $1)/Kconfig
else
echo $i $*
fi
done
sed -i '$d' hw/*/Kconfig
for i in hw/*; do
if test -d $i && ! test -f $i/Kconfig; then
touch $i/Kconfig
git add $i/Kconfig
fi
done
Whenever a symbol is referenced from multiple subdirectories, the
script prints the list of directories that reference the symbol.
These symbols have to be added manually to the Kconfig files.
Kconfig.host and hw/Kconfig were created manually.
Paolo Bonzini [Wed, 23 Jan 2019 06:55:58 +0000 (14:55 +0800)]
minikconfig: add semantic analysis
There are three parts in the semantic analysis:
1) evaluating expressions. This is done as a simple visit
of the Expr nodes.
2) ordering clauses. This is done by constructing a graph of variables.
There is an edge from X to Y if Y depends on X, if X selects Y, or if
X appears in a conditional selection of Y; in other words, if the value
of X can affect the value of Y. Each clause has a "destination" variable
whose value can be affected by the clause, and clauses will be processed
according to a topological sorting of their destination variables.
Defaults are processed after all other clauses with the same destination.
3) deriving the value of the variables. This is done by processing
the clauses in the topological order provided by the previous step.
A "depends on" clause will force a variable to False, a "select" clause
will force a variable to True, an assignment will force a variable
to its RHS. A default will set a variable to its RHS if it has not
been set before. Because all variables have a default, after visiting
all clauses all variables will have been set.
Paolo Bonzini [Wed, 23 Jan 2019 06:55:56 +0000 (14:55 +0800)]
minikconfig: add parser skeleton
This implements a scanner and recursive descent parser for Kconfig-like
configuration files. The only "action" of the parser is for now to
detect undefined variables and process include files.
The main differences between Kconfig and this are:
* only the "bool" type is supported
* variables can only be defined once
* choices are not supported (but they could be added as syntactic
sugar for multiple Boolean values)
* menus and other graphical concepts (prompts, help text) are not
supported
* assignments ("CONFIG_FOO=y", "CONFIG_FOO=n") are parsed as part
of the Kconfig language, not as a separate file.
The idea was originally by Ákos Kovács, but I could not find his
implementation so I had to redo it.
Peter Maydell [Thu, 7 Mar 2019 18:40:43 +0000 (18:40 +0000)]
Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-4.0-pull-request' into staging
- Update netlink types to linux v5.0
- fix accept4/getpeername/getsockname/recvfrom/recvmsg/read
- add/fix ELF_PLATFORM ofr aarch64 and arm
- fix "may be used uninitialized" warnings
- Fix breakpoint support in Nios
- Nicer strace output of chroot() syscall
* remotes/vivier2/tags/linux-user-for-4.0-pull-request:
linux-user: add new netlink types
linux-user: Nicer strace output of chroot() syscall
linux-user: fix "may be used uninitialized" warnings
linux-user: don't short-circuit read with zero length
Fix breakpoint support in Nios II user-mode emulation.
linux-user: fix emulation of accept4/getpeername/getsockname/recvfrom syscalls
linux-user: Fix ELF_PLATFORM for aarch64_be-linux-user
linux-user: Add ELF_PLATFORM for arm
linux-user: fix recvmsg emulation
Convert tests/eepro100-test to a driver node; currently it runs
the PCI nop test only, therefore we're not placing it in tests/libqos.
For now, all nodes share the same constructor and destructor.
Convert tests/nvme-test to a driver node; the code to discover the PCI
device is replaced by generic qgraph code, therefore we're not placing it in tests/libqos.
Paolo Bonzini [Fri, 5 Oct 2018 13:02:44 +0000 (15:02 +0200)]
tests: move virtio entirely to qos-test
The only remaining test that needs libqos-virtio-obj-y is drive_del-test,
which really only needs a function. Move that function to the test
and remove libqos-virtio-obj-y.
Convert tests/virtio-scsi-test in qgraph test node,
virtio-scsi-test. This test consumes a virtio-scsi interface
and checks that its function return the expected values.
Some functions are implemented only for virtio-scsi-pci, so they
don't consume virtio-scsi, but virtio-scsi-pci
Note that this test does not allocate any virtio-scsi structure,
it's all done by the qtest walking graph mechanism
tests/libqos: virtio-scsi driver and interface nodes
Add qgraph nodes for virtio-scsi-pci and virtio-scsi-device.
Both nodes produce virtio-scsi, but virtio-scsi-pci receives
a pci-bus and uses virtio-pci QOSGraphObject and its functions,
while virtio-scsi-device receives a virtio and implements
its own functions
Paolo Bonzini [Thu, 15 Nov 2018 09:08:28 +0000 (10:08 +0100)]
qos-test: vhost-user test node
Convert tests/virtio-net-test in qgraph test node,
virtio-net-test. This test consumes a virtio-net interface
and checks that its function return the expected values.
Note that this test does not allocate any virtio-net structure,
it's all done by the qtest walking graph mechanism. Nevertheless,
vhost-user-test is a bit more complex than the other tests, because
it requires more complicated setup of back-ends and thus almost each
test has a slightly different opts.before function.
Paolo Bonzini [Thu, 15 Nov 2018 09:50:15 +0000 (10:50 +0100)]
vhost-user-test: always use 256 MiB of guest memory
Some tests are using a small amount of RAM for the guest (2 MiB) in order to
save host memory, others are using 512 MiB.
However, pSeries machines only support multiples of 256 MiB. Using 256
MiB of memory does not use more host memory than now, even for the
migration test that starts two guests, and it allows running the test on
pSeries too.
This of course is not optimal, it would be nice to let the qgraph machine object
judge how much memory to provide. This is left for future work, together
with a more generic framework that wraps the QEMU command line.
Convert tests/virtio-net-test in qgraph test node,
virtio-net-test. This test consumes a virtio-net interface
and checks that its function return the expected values.
Some functions are implemented only for virtio-net-pci, so they
don't consume virtio-net, but virtio-net-pci
Note that this test does not allocate any virtio-net structure,
it's all done by the qtest walking graph mechanism
tests/libqos: virtio-net driver and interface nodes
Add qgraph nodes for virtio-net-pci and virtio-net-device.
Both nodes produce virtio-net, but virtio-net-pci receives
a pci-bus and overrides virtio-pci QOSGraphObject and its functions,
while virtio-net-device receives a virtio and implements
its own functions
Paolo Bonzini [Thu, 29 Nov 2018 11:51:08 +0000 (12:51 +0100)]
qos-test: virtio-blk test node
Convert tests/virtio-blk-test in qgraph test node,
virtio-blk-test. This test consumes a virtio-blk interface
and checks that its function return the expected values.
Some functions are implemented only for virtio-blk-pci, so they
don't consume virtio-blk, but virtio-blk-pci
Note that this test does not allocate any virtio-blk structure,
it's all done by the qtest walking graph mechanism. The allocator
is also provided by qgraph; remove malloc-generic.c and malloc-generic.h
which are not used anymore.
tests/libqos: virtio-blk driver and interface nodes
Add qgraph nodes for virtio-blk-pci and virtio-blk-device.
Both nodes produce virtio-blk, but virtio-blk-pci receives
a pci-bus and uses virtio-pci QOSGraphObject and functions,
while virtio-blk-device receives a virtio and implements
its own functions
Convert tests/virtio-rng-test in qgraph test node,
virtio-rng-test. This test consumes a virtio-rng interface
and checks that its function return the expected values.
Some functions are implemented only for virtio-rng-pci, so they
don't consume virtio-rng, but virtio-rng-pci
Note that this test does not allocate any virtio-rng structure,
it's all done by the qtest walking graph mechanism
tests/libqos: virtio-rng driver and interface nodes
Add qgraph nodes for virtio-rng-pci and virtio-rng-device.
Both nodes produce virtio-rng, but virtio-rng-pci receives
a pci-bus and uses virtio-pci QOSGraphObject and functions,
while virtio-rng-device receives a virtio and implements
its own functions
tests/libqos: virtio-balloon driver and interface nodes
Add qgraph nodes for virtio-balloon-pci and virtio-balloon-device.
Both nodes produce virtio-balloon, but virtio-balloon-pci receives
a pci-bus and uses virtio-pci QOSGraphObject and functions,
while virtio-balloon-device receives a virtio and implements
its own functions
Convert tests/virtio-9p-test into a qgraph test node.
This test consumes a virtio-9p interface and checks that its functions
return the expected values.
Note that this test does not allocate any virtio-9p structure,
it's all done by the qtest walking graph mechanism
tests/libqos: virtio-9p driver and interface nodes
Add qgraph nodes for virtio-9p-pci and virtio-9p-device.
Both nodes produce virtio-9p, but virtio-9p-pci receives
a pci-bus and overrides virtio-pci QOSGraphObject and its functions,
while virtio-9p-device receives a virtio and implements
its own functions
qos-test: virtio-console and virtio-serial test node
Convert tests/virtio-console-test and tests/virtio-serial-test
in qgraph test node. This test consumes a virtio-serial interface
and checks that its function return the expected values.
Note that this test does not allocate any virtio-console or
virtio-serial structure, it's all done by the qtest walking graph mechanism
projects / qemu.git / log