tests: Fix up docker cross builds for ppc64 (BE) targets

We currently have docker cross building targets for powerpc (32-bit, BE)
and ppc64el (64-bit, LE), but not for pcp64 (64-bit, BE).  This is an
irritating gap in make check-tcg coverage so correct it.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

Merge remote-tracking branch 'remotes/stsquad/tags/pull-testing-next-280519-2' into staging

Various testing updates

  - semihosting re-factor (used in system tests)
  - aarch64 and alpha system tests
  - editorconfig tweak for .S
  - some docker image updates
  - iotests clean-up (without make check inclusion)

# gpg: Signature made Tue 28 May 2019 17:26:34 BST
# gpg:                using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [full]
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8  DF35 FBD0 DB09 5A9E 2A44

* remotes/stsquad/tags/pull-testing-next-280519-2: (27 commits)
  tests/qemu-iotests: re-format output to for make check-block
  tests/qemu-iotests/group: Re-use the "auto" group for tests that can always run
  Makefile.target: support per-target coverage reports
  Makefile: include per-target build directories in coverage report
  Makefile: fix coverage-report reference to BUILD_DIR
  .travis.yml: enable aarch64-softmmu and alpha-softmmu tcg tests
  tests/tcg/alpha: add system boot.S
  tests/tcg/multiarch: expand system memory test to cover more
  tests/tcg/minilib: support %c format char
  tests/tcg/multiarch: move the system memory test
  tests/tcg/aarch64: add system boot.S
  editorconfig: add settings for .s/.S files
  tests/tcg/multiarch: add hello world system test
  tests/tcg/multiarch: add support for multiarch system tests
  tests/docker: Test more components on the Fedora default image
  tests/docker: add ubuntu 18.04
  MAINTAINERS: update for semihostings new home
  target/mips: convert UHI_plog to use common semihosting code
  target/mips: only build mips-semi for softmmu
  target/arm: correct return values for WRITE/READ in arm-semi
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-may-19-2019-v3' into staging

MIPS queue for May 19th, 2019 - v3

# gpg: Signature made Sun 26 May 2019 17:07:07 BST
# gpg:                using RSA key D4972A8967F75A65
# gpg: Good signature from "Aleksandar Markovic <amarkovic@wavecomp.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 8526 FBF1 5DA3 811F 4A01  DD75 D497 2A89 67F7 5A65

* remotes/amarkovic/tags/mips-queue-may-19-2019-v3:
  BootLinuxSshTest: Test some userspace commands on Malta
  target/mips: realign comments to fix checkpatch warnings
  target/mips: add or remove space to fix checkpatch errors
  linux-user: fix __NR_semtimedop undeclared error
  mips: Decide to map PAGE_EXEC in map_address
  target/mips: Refactor and fix INSERT.<B|H|W|D> instructions
  target/mips: Refactor and fix COPY_U.<B|H|W> instructions
  target/mips: Refactor and fix COPY_S.<B|H|W|D> instructions
  target/mips: Fix MSA instructions ST.<B|H|W|D> on big endian host
  target/mips: Fix MSA instructions LD.<B|H|W|D> on big endian host
  target/mips: Make the results of MOD_<U|S>.<B|H|W|D> the same as on hardware
  target/mips: Make the results of DIV_<U|S>.<B|H|W|D> the same as on hardware

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/palmer/tags/riscv-for-master-4.1-sf0' into staging

RISC-V Patches for the 4.1 Soft Freeze, Part 1

This tag contains a handful of patches that I'd like to target for 4.1:

* An emulation for SiFive's GPIO device.
* A fix to disallow sfence.vma from userspace.
* Additional decodetree cleanups that should have no functional impact.
* C extension emulation fidelity fixes that were noticed as part of that
  cleanup process.
* A new "spike" target, along with the deprecation of a handful of old
  targets and CPUs.
* Some initial infastructure related to the hypervisor extension.
* An emulation fidelity fix that prevents prevents arbitrary bits in the
  SIP CSR from being set.
* A small performance improvement that avoids excessive TLB flushing
  when the ASID does not change.

This time I've used a new testing workflow: I've tested on both 32-bit
and 64-bit builds of OpenEmbedded, via the default OpenSBI-based boot
flow.

# gpg: Signature made Sat 25 May 2019 01:05:57 BST
# gpg:                using RSA key 00CE76D1834960DFCE886DF8EF4CA1502CCBAB41
# gpg:                issuer "palmer@dabbelt.com"
# gpg: Good signature from "Palmer Dabbelt <palmer@dabbelt.com>" [unknown]
# gpg:                 aka "Palmer Dabbelt <palmer@sifive.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 00CE 76D1 8349 60DF CE88  6DF8 EF4C A150 2CCB AB41

* remotes/palmer/tags/riscv-for-master-4.1-sf0: (29 commits)
  target/riscv: Only flush TLB if SATP.ASID changes
  target/riscv: More accurate handling of `sip` CSR
  target/riscv: Add checks for several RVC reserved operands
  target/riscv: Add the HGATP register masks
  target/riscv: Add the HSTATUS register masks
  target/riscv: Add Hypervisor CSR macros
  target/riscv: Allow setting mstatus virtulisation bits
  target/riscv: Add the MPV and MTL mstatus bits
  target/riscv: Improve the scause logic
  target/riscv: Trigger interrupt on MIP update asynchronously
  target/riscv: Mark privilege level 2 as reserved
  riscv: spike: Add a generic spike machine
  target/riscv: Deprecate the generic no MMU CPUs
  target/riscv: Add a base 32 and 64 bit CPU
  target/riscv: Create settable CPU properties
  riscv: virt: Allow specifying a CPU via commandline
  linux-user/riscv: Add the CPU type as a comment
  target/riscv: Remove unused include of riscv_htif.h for virt board riscv
  target/riscv: Remove spaces from register names
  target/riscv: Split gen_arith_imm into functional and temp
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging

Machine Core queue, 2019-05-24

* Display more helpful message when an object type is missing
  (Philippe Mathieu-Daudé)
* Use object_initialize_child for correct reference counting
  (Philippe Mathieu-Daudé)

# gpg: Signature made Fri 24 May 2019 19:31:06 BST
# gpg:                using RSA key 2807936F984DC5A6
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>" [full]
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/machine-next-pull-request:
  hw/intc/nvic: Use object_initialize_child for correct reference counting
  hw/arm/mps2: Use object_initialize_child for correct reference counting
  hw/microblaze/zynqmp: Use object_initialize_child for correct ref. counting
  hw/microblaze/zynqmp: Use object_initialize_child for correct ref. counting
  hw/microblaze/zynqmp: Let the SoC manage the IPI devices
  hw/microblaze/zynqmp: Move the IPI state into the PMUSoC state
  hw/mips: Use object_initialize_child for correct reference counting
  hw/mips: Use object_initialize() on MIPSCPSState
  hw/arm: Use object_initialize_child for correct reference counting
  hw/arm/aspeed: Use object_initialize_child for correct ref. counting
  hw/arm/bcm2835: Use object_initialize_child for correct ref. counting
  hw/arm/bcm2835: Use object_initialize() on PL011State
  hw/arm/bcm2835: Use TYPE_PL011 instead of hardcoded string
  hw/virtio: Use object_initialize_child for correct reference counting
  hw/misc/macio: Use object_initialize_child for correct ref. counting
  hw/ppc/pnv: Use object_initialize_child for correct reference counting
  qom/object: Display more helpful message when an object type is missing

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

tests/qemu-iotests: re-format output to for make check-block

This attempts to clean-up the output to better match the output of the
rest of the QEMU check system when called with -makecheck. This includes:

  - formatting as "  TEST    iotest-FMT: nnn"
  - only dumping config on failure (when -makecheck enabled)

The non-make check output has been cleaned up as well:

  - line re-displayed (\r) at the end
  - fancy colours for pass/fail/skip
  - timestamps always printed (option removed)

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20190503143904.31211-1-alex.bennee@linaro.org>
Tested-by: Thomas Huth <thuth@redhat.com>

tests/qemu-iotests/group: Re-use the "auto" group for tests that can always run

Currently, all tests are in the "auto" group. This is a little bit pointless.
OTOH, we need a group for the tests that we can automatically run during
"make check" each time, too. Tests in this new group are supposed to run
with every possible QEMU configuration, for example they must run with every
QEMU binary (also non-x86), without failing when an optional features is
missing (but reporting "skip" is ok), and be able to run on all kind of host
filesystems and users (i.e. also as "nobody" or "root").
So let's use the "auto" group for this class of tests now. The initial
list has been determined by running the iotests with non-x86 QEMU targets
and with our CI pipelines on Gitlab, Cirrus-CI and Travis (i.e. including
macOS and FreeBSD).

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20190502084506.8009-7-thuth@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

Makefile.target: support per-target coverage reports

Add support for generating a single targets coverage report. Execute:

  make coverage-report

In the target build directory. This coverage report only cares about
target specific blobs so only searches the target build subdirectory.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

Makefile: include per-target build directories in coverage report

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

Makefile: fix coverage-report reference to BUILD_DIR

Commit 337f2311f actually claimed to do this in the commit log but
didn't actually. Oops.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

.travis.yml: enable aarch64-softmmu and alpha-softmmu tcg tests

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/alpha: add system boot.S

This provides the bootstrap and low level helper functions for an
alpha kernel.  We use direct access to the DP264 serial port for
test output, and hard machine halt to exit the emulation.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20190501184306.15208-1-richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

tests/tcg/multiarch: expand system memory test to cover more

Expand the memory test to cover move of the softmmu code. Specifically
we:

  - improve commentary
  - add some helpers (for later BE support)
  - reduce boiler plate into helpers
  - add signed reads at various sizes/offsets
  - required -DCHECK_UNALIGNED

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/minilib: support %c format char

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/multiarch: move the system memory test

There is nothing inherently architecture specific about the memory
test although we may have to manage different restrictions of
unaligned access across architectures.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/aarch64: add system boot.S

This provides the bootstrap and low level helper functions for an
aarch64 kernel. We use semihosting to handle test output and exiting
the emulation. semihosting's parameter passing is a little funky so we
end up using the stack and pointing to that as the parameter block.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

editorconfig: add settings for .s/.S files

We are starting to add assembler foe tests/tcg so lets make sure we
get the mode right.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/multiarch: add hello world system test

This is not really i386 only, we can have the same test for all
architectures supporting system tests.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/tcg/multiarch: add support for multiarch system tests

We can certainly support some common tests for system emulation that
make use of our minimal defined boot.S support. It will still be up to
individual architectures to ensure they build so we provide a
MULTIARCH_TESTS variable that they can tack onto TESTS themselves.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

tests/docker: Test more components on the Fedora default image

Install optional dependencies of QEMU to get better coverage.

The following components are now enabled:

  $ ./configure
  ...
  Multipath support yes
  VNC SASL support  yes
  RDMA support      yes
  PVRDMA support    yes
  libiscsi support  yes
  seccomp support   yes
  libpmem support   yes
  libudev           yes

Note: The udev-devel package is provided by systemd-devel.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190504055440.20406-1-philmd@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Acked-by: Marcel Apfelbaum<marcel.apfelbaum@gmail.com>

tests/docker: add ubuntu 18.04

Based on the ubuntu.docker file.
Used to reproduce the build failure Peter was seeing.
Others might find this useful too ;)

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-Id: <20190503070241.24786-1-kraxel@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>

MAINTAINERS: update for semihostings new home

Seeing as I touched it I should at least keep an eye on it.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

target/mips: convert UHI_plog to use common semihosting code

Rather than printing directly to stdout lets use our common
semihosting code. There is one minor difference in that the output
currently defaults to stderr instead of stdout however this can be
controlled by connecting semihosting to a chardev.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>

target/mips: only build mips-semi for softmmu

The is_uhi gates all semihosting calls and always returns false for
CONFIG_USER_ONLY builds. There is no reason to build and link
mips-semi for these builds so lets fix that.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

target/arm: correct return values for WRITE/READ in arm-semi

The documentation says the write should return the number of bytes not
written on an error (0 means everything was written). Read provides a
buffer length and the return value should be the buffer length - bytes
actually read. Remove the incorrect FIXME's and return the correct
values.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>

target/arm: add LOG_UNIMP messages to arm-semi

Clean-up our unimplemented bits with a proper message.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

target/arm: use the common interface for WRITE0/WRITEC in arm-semi

Now we have a common semihosting console interface use that for our
string output. However ARM is currently unique in also supporting
semihosting for linux-user so we need to replicate the API in
linux-user. If other architectures gain this support we can move the
file later.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

target/arm: fixup some of the commentary for arm-semi

This cleans up a number of the block comments to fit the proper style.
While we are at it we also reference the official specification and
document what the return register value can be.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>

semihosting: enable chardev backed output for console

It will be useful for a number of use-cases to be able to re-direct
output to a file like we do with serial output. This does the wiring
to allow us to treat then semihosting console like just another
character output device.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

semihosting: implement a semihosting console

This provides two functions for handling console output that handle
the common backend behaviour for semihosting.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

semihosting: introduce CONFIG_SEMIHOSTING

There isn't much point building semihosting for platforms that don't
support it. Introduce a new symbol and enable it only for the softmmu
targets that need it.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

semihosting: move semihosting configuration into its own directory

In preparation for having some more common semihosting code let's
excise the current config magic from vl.c into its own file. We shall
later add more conditionals to the build configurations so we can
avoid building this if we don't need it.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

BootLinuxSshTest: Test some userspace commands on Malta

This tests boot a full VM and check the serial console until
the SSH daemon is running, then start a SSH session and run
some commands.

This test can be run using:

  $ avocado --show=ssh run -t arch:mips tests/acceptance/linux_ssh_mips_malta.py
  ssh: Entering interactive session.
  ssh: # uname -a
  ssh: Linux debian-mips 3.2.0-4-4kc-malta #1 Debian 3.2.51-1 mips GNU/Linux
  ssh: # lspci -d 11ab:4620
  ssh: 00:00.0 Host bridge: Marvell Technology Group Ltd. GT-64120/64120A/64121A System Controller (rev 10)
  ssh: # cat /sys/bus/i2c/devices/i2c-0/name
  ssh: SMBus PIIX4 adapter at 1100
  ssh: # cat /proc/mtd
  ssh: dev:    size   erasesize  name
  ssh: mtd0: 00100000 00010000 "YAMON"
  ssh: mtd1: 002e0000 00010000 "User FS"
  ssh: mtd2: 00020000 00010000 "Board Config"
  ssh: # md5sum /dev/mtd2ro
  ssh: 0dfbe8aa4c20b52e1b8bf3cb6cbdf193  /dev/mtd2ro
  ssh: # poweroff

Acked-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20190523161832.22490-5-f4bug@amsat.org>

target/mips: realign comments to fix checkpatch warnings

Realign comments to fix warnings issued by checkpatc.pl tool
 "WARNING: Block comments use a leading /* on a separate line"
within "target/mips/cpu.h" file.

Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <20190413202818.13622-3-jbi.octave@gmail.com>

target/mips: add or remove space to fix checkpatch errors

Add or remove space to fix errors issued by checkpatch.pl tool
"ERROR: spaces required around that..."
"ERROR: space required after that..."
"ERROR: space required before the open parenthesis"
"ERROR: space required after that..."
"ERROR: space prohibited between function name and open parenthesis"
"ERROR: code indent should never use tabs"
"ERROR: line over 90 characters"
within "target/mips/cpu.h" file.

Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <20190413202818.13622-2-jbi.octave@gmail.com>

linux-user: fix __NR_semtimedop undeclared error

In current code, __NR_msgrcv and__NR_semtimedop are supposed to be
defined if __NR_msgsnd is defined.

But linux headers 5.2-rc1 for MIPS define __NR_msgsnd without defining
__NR_semtimedop and it breaks the QEMU build.

__NR_semtimedop is defined in asm-mips/unistd_n64.h and asm-mips/unistd_n32.h
but not in asm-mips/unistd_o32.h.

Commit d9cb4336159a ("linux headers: update against Linux 5.2-rc1") has
updated asm-mips/unistd_o32.h and added __NR_msgsnd but not __NR_semtimedop.
It introduces __NR_semtimedop_time64 instead.

This patch fixes the problem by checking for each __NR_XXX symbol
before defining the corresponding syscall.

Fixes: d9cb4336159a ("linux headers: update against Linux 5.2-rc1")
Reported-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20190523175413.14448-1-laurent@vivier.eu>

mips: Decide to map PAGE_EXEC in map_address

This commit addresses QEMU Bug #1825311:

  mips_cpu_handle_mmu_fault renders all accessed pages executable

It allows finer-grained control over whether the accessed page should
be executable by moving the decision to the underlying map_address
function, which has more information for this.

As a result, pages that have the XI bit set in the TLB and are accessed
for read/write, don't suddenly end up being executable.

Fixes: https://bugs.launchpad.net/qemu/+bug/1825311
Fixes: 2fb58b73746e ('target-mips: add RI and XI fields to TLB entry')
Signed-off-by: Jakub Jermář <jakub.jermar@kernkonzept.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190517123533.868479-1-jakub.jermar@kernkonzept.com>

target/mips: Refactor and fix INSERT.<B|H|W|D> instructions

The old version of the helper for the INSERT.<B|H|W|D> MSA instructions
has been replaced with four helpers that don't use switch, and change
the endianness of the given index, when executed on a big endian host.

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554212605-16457-6-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Refactor and fix COPY_U.<B|H|W> instructions

The old version of the helper for the COPY_U.<B|H|W> MSA instructions
has been replaced with four helpers that don't use switch, and change
the endianness of the given index, when executed on a big endian host.

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554212605-16457-5-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Refactor and fix COPY_S.<B|H|W|D> instructions

The old version of the helper for the COPY_S.<B|H|W|D> MSA instructions
has been replaced with four helpers that don't use switch, and change
the endianness of the given index, when executed on a big endian host.

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554212605-16457-4-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Fix MSA instructions ST.<B|H|W|D> on big endian host

Fix the case when the host is a big endian machine, and change
the approach toward ST.<B|H|W|D> instruction helpers.

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554212605-16457-3-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Fix MSA instructions LD.<B|H|W|D> on big endian host

Fix the case when the host is a big endian machine, and change
the approach toward LD.<B|H|W|D> instruction helpers.

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554212605-16457-2-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Make the results of MOD_<U|S>.<B|H|W|D> the same as on hardware

MSA instructions MOD_<U|S>.<B|H|W|D> when dividing by zero,
didn't return the same value when executed on a referent hardware
(FPGA MIPS 64 r6, little endian) and when executed on QEMU, which
is not a real bug, because the result when dividing by zero is
UNPREDICTABLE [1] (page 255, 256).

[1] MIPS Architecture for Programmers
    Volume IV-j: The MIPS64 SIMD
    Architecture Module, Revision 1.12

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554207110-9113-3-git-send-email-mateja.marjanovic@rt-rk.com>

target/mips: Make the results of DIV_<U|S>.<B|H|W|D> the same as on hardware

MSA instructions DIV_<U|S>.<B|H|W|D> when dividing by zero,
didn't return the same value when executed on a referent hardware
(FPGA MIPS 64 r6, little endian) and when executed on QEMU, which
is not a real bug, because the result when dividing by zero is
UNPREDICTABLE [1] (page 141, 142).

[1] MIPS Architecture for Programmers
    Volume IV-j: The MIPS64 SIMD
    Architecture Module, Revision 1.12

Signed-off-by: Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Message-Id: <1554207110-9113-2-git-send-email-mateja.marjanovic@rt-rk.com>

target/riscv: Only flush TLB if SATP.ASID changes

There is an analogous change for ARM here:
https://patchwork.kernel.org/patch/10649857

Signed-off-by: Jonathan Behrens <jonathan@fintelia.io>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: More accurate handling of `sip` CSR

According to the spec, "All bits besides SSIP, USIP, and UEIP in the sip
register are read-only." Further, if an interrupt is not delegated to mode x,
then "the corresponding bits in xip [...] should appear to be hardwired to
zero. This patch implements both of those requirements.

Signed-off-by: Jonathan Behrens <jonathan@fintelia.io>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add checks for several RVC reserved operands

C.ADDI16SP, C.LWSP, C.JR, C.ADDIW, C.LDSP all have reserved
operands that were not diagnosed.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add the HGATP register masks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add the HSTATUS register masks

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviwed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add Hypervisor CSR macros

Add the 1.10.1 Hypervisor CSRs and remove the 1.9.1 spec versions.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Allow setting mstatus virtulisation bits

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Revieweb-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add the MPV and MTL mstatus bits

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Improve the scause logic

No functional change, just making the code easier to read.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Trigger interrupt on MIP update asynchronously

The requirement of holding the iothread_mutex is burdersome when
swapping the background and foreground registers in the Hypervisor
extension. To avoid the requrirement let's set the interrupt
asynchronously.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Mark privilege level 2 as reserved

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

riscv: spike: Add a generic spike machine

Add a generic spike machine (not tied to a version) and deprecate the
spike mahines that are tied to a specific version. As we can now specify
the CPU via the command line we no londer need specific versions of the
spike machines.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Acked-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Deprecate the generic no MMU CPUs

These can now be specified via the command line so we no longer need
these.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Add a base 32 and 64 bit CPU

At the same time deprecate the ISA string CPUs.

It is dobtful anyone specifies the CPUs, but we are keeping them for the
Spike machine (which is about to be depreated) so we may as well just
mark them as deprecated.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Create settable CPU properties

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

riscv: virt: Allow specifying a CPU via commandline

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

linux-user/riscv: Add the CPU type as a comment

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Remove unused include of riscv_htif.h for virt board riscv

Signed-off-by: Jonathan Behrens <fintelia@gmail.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Remove spaces from register names

These extra spaces make the "-d op" dump look weird.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Split gen_arith_imm into functional and temp

The tcg_gen_fooi_tl functions have some immediate constant
folding built in, which match up with some of the riscv asm
builtin macros, like mv and not.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Split RVC32 and RVC64 insns into separate files

This eliminates all functions in insn_trans/trans_rvc.inc.c,
so the entire file can be removed.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Use pattern groups in insn16.decode

This eliminates about half of the complicated decode
bits within insn_trans/trans_rvc.inc.c.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Merge argument decode for RVC shifti

Special handling for IMM==0 is the only difference between
RVC shifti and RVI shifti.  This can be handled with !function.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Merge argument sets for insn32 and insn16

In some cases this allows us to directly use the insn32
translator function.  In some cases we still need a shim.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Use --static-decode for decodetree

The generated functions are only used within translate.c
and do not need to be global, or declared.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Name the argument sets for all of insn32 formats

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

RISC-V: fix single stepping over ret and other branching instructions

This patch introduces wrappers around the tcg_gen_exit_tb() and
tcg_gen_lookup_and_goto_ptr() functions that handle single stepping,
i.e. call gen_exception_debug() when single stepping is enabled.

Theses functions are then used instead of the originals, bringing single
stepping handling in places where it was previously ignored such as jalr
and system branch instructions (ecall, mret, sret, etc.).

Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

target/riscv: Do not allow sfence.vma from user mode

The 'sfence.vma' instruction is privileged, and should only ever be allowed
when executing in supervisor mode or higher.

Signed-off-by: Jonathan Behrens <fintelia@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

SiFive RISC-V GPIO Device

QEMU model of the GPIO device on the SiFive E300 series SOCs.

The pins are not used by a board definition yet, however this
implementation can already be used to trigger GPIO interrupts from the
software by configuring a pin as both output and input.

Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Reviewed-by: Palmer Dabbelt <palmer@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@sifive.com>

hw/intc/nvic: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script:

 @use_sysbus_init_child_obj_missing_parent@
 expression child_ptr;
 expression child_type;
 expression child_size;
 @@
 -   object_initialize(child_ptr, child_size, child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
     ...
 ?-  object_unref(OBJECT(child_ptr));
 +   sysbus_init_child_obj(OBJECT(PARENT_OBJ), "CHILD_NAME", child_ptr,
 +                         child_size, child_type);

We let NVIC adopt the SysTick timer.

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-17-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm/mps2: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script:

 @use_sysbus_init_child_obj_missing_parent@
 expression child_ptr;
 expression child_type;
 expression child_size;
 @@
 -   object_initialize(child_ptr, child_size, child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
     ...
 ?-  object_unref(OBJECT(child_ptr));
 +   sysbus_init_child_obj(OBJECT(PARENT_OBJ), "CHILD_NAME", child_ptr,
 +                         child_size, child_type);

We let the MPS2 boards adopt the cpu core, the FPGA and the SCC children.

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-16-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/microblaze/zynqmp: Use object_initialize_child for correct ref. counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
?-   object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
?-   object_unref(OBJECT(child_ptr));
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-15-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/microblaze/zynqmp: Use object_initialize_child for correct ref. counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(then manually modified to use numbered IPI name)

 @use_sysbus_init_child_obj_missing_parent@
 expression child_ptr;
 expression child_type;
 expression child_size;
 @@
 -   object_initialize(child_ptr, child_size, child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
     ...
 ?-  object_unref(OBJECT(child_ptr));
 +   sysbus_init_child_obj(OBJECT(PARENT_OBJ), "CHILD_NAME", child_ptr,
 +                         child_size, child_type);

We let the SoC adopt the IPI children.

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-14-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/microblaze/zynqmp: Let the SoC manage the IPI devices

The Inter Processor Interrupt is a block part of the SoC, not the
"machine" (See Zynq UltraScale+ Device TRM UG1085, "Platform
Management Unit", Power Domains and Islands).

Move the IPI management from the machine to the SoC.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-13-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/microblaze/zynqmp: Move the IPI state into the PMUSoC state

The Inter Processor Interrupt is a block part of the SoC, not the
"machine" (talking about machine is borderline with the PMU, since
it is embedded into the ZynqMP SoC, but currentl QEMU doesn't
support multi-arch cores).

Move the IPI state to the SoC state, this will simplify the review
of the next patch.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-12-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/mips: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script:

 @use_sysbus_init_child_obj_missing_parent@
 expression child_ptr;
 expression child_type;
 expression child_size;
 @@
 -   object_initialize(child_ptr, child_size, child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
     ...
 ?-  object_unref(OBJECT(child_ptr));
 +   sysbus_init_child_obj(OBJECT(PARENT_OBJ), "CHILD_NAME", child_ptr,
 +                         child_size, child_type);

We let the Malta/Boston machines adopt the CPS child, and similarly
the CPS adopts the ITU/CPC/GIC/GCR children.

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-11-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/mips: Use object_initialize() on MIPSCPSState

Initialize the MIPSCPSState with object_initialize() instead of
object_new(). This will allow us to add it as children of the
machine container.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-10-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 )

 @use_sysbus_init_child_obj@
 expression parent_obj;
 expression dev;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
 |
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
 -   dev = DEVICE(child_ptr);
 -   qdev_set_parent_bus(dev, sysbus_get_default());
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-9-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm/aspeed: Use object_initialize_child for correct ref. counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 )

 @use_sysbus_init_child_obj@
 expression parent_obj;
 expression dev;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
 |
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
 -   dev = DEVICE(child_ptr);
 -   qdev_set_parent_bus(dev, sysbus_get_default());
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Joel Stanley <joel@jms.id.au>
Message-Id: <20190507163416.24647-8-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm/bcm2835: Use object_initialize_child for correct ref. counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 )

 @use_sysbus_init_child_obj@
 expression parent_obj;
 expression dev;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
     ...
 -   qdev_set_parent_bus(DEVICE(child_ptr), sysbus_get_default());
 |
 -   object_initialize_child(parent_obj, child_name, child_ptr, child_size,
 -                           child_type, errp, NULL);
 +   sysbus_init_child_obj(parent_obj, child_name, child_ptr, child_size,
 +                         child_type);
 -   dev = DEVICE(child_ptr);
 -   qdev_set_parent_bus(dev, sysbus_get_default());
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.
This choice also matches when using sysbus_init_child_obj(),
since its code is:

  void sysbus_init_child_obj(Object *parent,
                             const char *childname, void *child,
                             size_t childsize, const char *childtype)
  {
      object_initialize_child(parent, childname, child, childsize,
                              childtype, &error_abort, NULL);

      qdev_set_parent_bus(DEVICE(child), sysbus_get_default());
  }

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-7-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm/bcm2835: Use object_initialize() on PL011State

To be coherent with the other peripherals contained in the
BCM2835PeripheralState structure, directly allocate the PL011State
(instead of using the pl011 uart as a pointer to a SysBusDevice).

Initialize the PL011State with object_initialize() instead of
object_new().

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-6-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/arm/bcm2835: Use TYPE_PL011 instead of hardcoded string

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-5-philmd@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/virtio: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script:

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-4-philmd@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/misc/macio: Use object_initialize_child for correct ref. counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
 ?-  object_unref(OBJECT(child_ptr));
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-3-philmd@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

hw/ppc/pnv: Use object_initialize_child for correct reference counting

As explained in commit aff39be0ed97:

  Both functions, object_initialize() and object_property_add_child()
  increase the reference counter of the new object, so one of the
  references has to be dropped afterwards to get the reference
  counting right. Otherwise the child object will not be properly
  cleaned up when the parent gets destroyed.
  Thus let's use now object_initialize_child() instead to get the
  reference counting here right.

This patch was generated using the following Coccinelle script
(with a bit of manual fix-up for overly long lines):

 @use_object_initialize_child@
 expression parent_obj;
 expression child_ptr;
 expression child_name;
 expression child_type;
 expression child_size;
 expression errp;
 @@
 (
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                           child_type, &error_abort, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), NULL);
     ...
?-   object_unref(OBJECT(child_ptr));
 |
 -   object_initialize(child_ptr, child_size, child_type);
 +   object_initialize_child(parent_obj, child_name,  child_ptr, child_size,
 +                            child_type, errp, NULL);
     ... when != parent_obj
 -   object_property_add_child(parent_obj, child_name, OBJECT(child_ptr), errp);
     ...
?-   object_unref(OBJECT(child_ptr));
 )

While the object_initialize() function doesn't take an
'Error *errp' argument, the object_initialize_child() does.
Since this code is used when a machine is created (and is not
yet running), we deliberately choose to use the &error_abort
argument instead of ignoring errors if an object creation failed.

Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Inspired-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190507163416.24647-2-philmd@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

qom/object: Display more helpful message when an object type is missing

When writing a new board, adding device which uses other devices
(container) or simply refactoring, one can discover the hard way
his machine misses some devices. In the case of containers, the
error is not obvious:

  $ qemu-system-microblaze -M xlnx-zynqmp-pmu
  **
  ERROR:/source/qemu/qom/object.c:454:object_initialize_with_type: assertion failed: (type != NULL)
  Aborted (core dumped)

And we have to look at the coredump to figure the error:

  (gdb) bt
  #1  0x00007f84773cf895 in abort () at /lib64/libc.so.6
  #2  0x00007f847961fb53 in  () at /lib64/libglib-2.0.so.0
  #3  0x00007f847967a4de in g_assertion_message_expr () at /lib64/libglib-2.0.so.0
  #4  0x000055c4bcac6c11 in object_initialize_with_type (data=data@entry=0x55c4bdf239e0, size=size@entry=2464, type=<optimized out>) at /source/qemu/qom/object.c:454
  #5  0x000055c4bcac6e6d in object_initialize (data=data@entry=0x55c4bdf239e0, size=size@entry=2464, typename=typename@entry=0x55c4bcc7c643 "xlnx.zynqmp_ipi") at /source/qemu/qom/object.c:474
  #6  0x000055c4bc9ea474 in xlnx_zynqmp_pmu_init (machine=0x55c4bdd46000) at /source/qemu/hw/microblaze/xlnx-zynqmp-pmu.c:176
  #7  0x000055c4bca3b6cb in machine_run_board_init (machine=0x55c4bdd46000) at /source/qemu/hw/core/machine.c:1030
  #8  0x000055c4bc95f6d2 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at /source/qemu/vl.c:4479

Since the caller knows the type name requested, we can simply display it
to ease development.

With this patch applied we get:

  $ qemu-system-microblaze -M xlnx-zynqmp-pmu
  qemu-system-microblaze: missing object type 'xlnx.zynqmp_ipi'
  Aborted (core dumped)

Since the assert(type) check in object_initialize_with_type() is
now impossible, remove it.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190427135642.16464-1-philmd@redhat.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>

Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-4.1-pull-request' into staging

Add /proc/hardware and /proc/cpuinfo,
update SIOCXXX ioctls,
fix shmat emulation,
add nanoseconds in stat,
init field fp_abi on mips

# gpg: Signature made Fri 24 May 2019 12:24:36 BST
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier2/tags/linux-user-for-4.1-pull-request:
  linux-user: Pass through nanosecond timestamp components for stat syscalls
  linux-user: Align mmap_find_vma to host page size
  linux-user: Fix shmat emulation by honoring host SHMLBA
  linux-user: Sanitize interp_info and, for mips only, init field fp_abi
  linux-user: Add support for SIOC<G|S>IFPFLAGS ioctls for all targets
  linux-user: Add support for SIOCSPGRP ioctl for all targets
  linux-user: Fix support for SIOCATMARK and SIOCGPGRP ioctls for xtensa
  linux-user: add pseudo /proc/hardware for m68k
  linux-user: add pseudo /proc/cpuinfo for sparc

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

linux-user: Pass through nanosecond timestamp components for stat syscalls

Since Linux 2.6 the stat syscalls have mostly supported nanosecond
components for each of the file-related timestamps.

QEMU user mode emulation currently does not pass through the nanosecond
portion of the timestamp, even when the host system fills in the value.
This results in a mismatch when run on subsecond resolution filesystems
such as ext4 or XFS.

An example of this leading to inconsistency is cross-debootstraping a
full desktop root filesystem of Debian Buster. Recent versions of
fontconfig store the full timestamp (instead of just the second portion)
of the directory in its per-directory cache file, and checks this against
the directory to see if the cache is up-to-date. With QEMU user mode
emulation, the timestamp stored is incorrect, and upon booting the rootfs
natively, fontconfig discovers the mismatch, and proceeds to rebuild the
cache on the comparatively slow machine (low-power ARM vs x86). This
stalls the first attempt to open whatever application that incorporates
fontconfig.

This patch renames the "unused" padding trailing each timestamp element
to its nanosecond counterpart name if such an element exists in the
kernel sources for the given platform. Not all do. Then have the syscall
wrapper fill in the nanosecond portion if the host supports it, as
specified by the _POSIX_C_SOURCE and _XOPEN_SOURCE feature macros.

Recent versions of glibc only use stat64 and newfstatat syscalls on
32-bit and 64-bit platforms respectively. The changes in this patch
were tested by directly calling the stat, stat64 and newfstatat syscalls
directly, in addition to the glibc wrapper, on arm and aarch64 little
endian targets.

Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Message-Id: <20190522162147.26303-1-wens@kernel.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Align mmap_find_vma to host page size

This can avoid stack allocation failures for i386 guest
on ppc64 (64k page) host.

Suggested-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20190519201953.20161-14-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Fix shmat emulation by honoring host SHMLBA

For those hosts with SHMLBA > getpagesize, we don't automatically
select a guest address that is compatible with the host.  We can
achieve this by boosting the alignment of guest_base and by adding
an extra alignment argument to mmap_find_vma.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20190519201953.20161-13-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

Merge remote-tracking branch 'remotes/kraxel/tags/vga-20190524-pull-request' into staging

ramfb: misc improvements.

# gpg: Signature made Fri 24 May 2019 09:56:59 BST
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/vga-20190524-pull-request:
  hw/display/ramfb: initialize fw-config space with xres/ yres
  hw/display/ramfb: lock guest resolution after it's set
  hw/display/ramfb: fix guest memory un-mapping

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20190522' into staging

Misc gvec improvements

# gpg: Signature made Wed 22 May 2019 23:25:48 BST
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth/tags/pull-tcg-20190522:
  tcg/i386: Use MOVDQA for TCG_TYPE_V128 load/store
  tcg/aarch64: Allow immediates for vector ORR and BIC
  tcg/aarch64: Build vector immediates with two insns
  tcg/aarch64: Use MVNI in tcg_out_dupi_vec
  tcg/aarch64: Split up is_fimm
  tcg/aarch64: Support vector bitwise select value
  tcg/i386: Use umin/umax in expanding unsigned compare
  tcg/i386: Remove expansion for missing minmax
  tcg/i386: Support vector comparison select value
  tcg: Add TCG_OPF_NOT_PRESENT if TCG_TARGET_HAS_foo is negative
  tcg: Expand vector minmax using cmp+cmpsel
  tcg: Introduce do_op3_nofail for vector expansion
  tcg: Add support for vector compare select
  tcg: Add support for vector bitwise select
  tcg: Fix missing checks and clears in tcg_gen_gvec_dup_mem
  tcg/i386: Fix dupi/dupm for avx1 and 32-bit hosts

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20190523' into staging

target-arm queue:
 * exynos4210: QOM'ify the Exynos4210 SoC
 * exynos4210: Add DMA support for the Exynos4210
 * arm_gicv3: Fix writes to ICC_CTLR_EL3
 * arm_gicv3: Fix write of ICH_VMCR_EL2.{VBPR0, VBPR1}
 * target/arm: Fix vector operation segfault
 * target/arm: Minor improvements to BFXIL, EXTR

# gpg: Signature made Thu 23 May 2019 15:22:55 BST
# gpg:                using RSA key E1A5C593CD419DE28E8315CF3C2525ED14360CDE
# gpg:                issuer "peter.maydell@linaro.org"
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>" [ultimate]
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>" [ultimate]
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* remotes/pmaydell/tags/pull-target-arm-20190523:
  hw/arm/exynos4210: QOM'ify the Exynos4210 SoC
  hw/arm/exynos4210: Add DMA support for the Exynos4210
  hw/arm/exynos4: Use the IEC binary prefix definitions
  hw/arm/exynos4: Remove unuseful debug code
  hw/intc/arm_gicv3: Fix writes to ICC_CTLR_EL3
  hw/intc/arm_gicv3: Fix write of ICH_VMCR_EL2.{VBPR0, VBPR1}
  arm: Rename hw/arm/arm.h to hw/arm/boot.h
  arm: Remove unnecessary includes of hw/arm/arm.h
  arm: Move system_clock_scale to armv7m_systick.h
  target/arm: Fix vector operation segfault
  target/arm: Simplify BFXIL expansion
  target/arm: Use extract2 for EXTR

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/display/ramfb: initialize fw-config space with xres/ yres

If xres / yres were specified in QEMU command line, write them as an initial
resolution to the fw-config space on guest reset, which a later BIOS / OVMF
patch can take advantage of.

Signed-off-by: HOU Qiming <hqm03ster@gmail.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-id: 20190513115731.17588-4-marcel.apfelbaum@gmail.com
[fixed malformed patch]
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

hw/display/ramfb: lock guest resolution after it's set

Only allow one resolution change per guest boot, which prevents a
crash when the guest writes garbage to the configuration space (e.g.
when rebooting).

Signed-off-by: HOU Qiming <hqm03ster@gmail.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-id: 20190513115731.17588-3-marcel.apfelbaum@gmail.com
[fixed malformed patch]
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

hw/display/ramfb: fix guest memory un-mapping

Pulled back the `qemu_create_displaysurface_guestmem` function to create
the display surface so that the guest memory gets properly unmapped.

Signed-off-by: HOU Qiming <hqm03ster@gmail.com>
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Message-id: 20190513115731.17588-2-marcel.apfelbaum@gmail.com
[rename the new functions and use QEMU coding style]
Signed-off-by: Marcel Apfelbaum <marcel.apfelbaum@gmail.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>