target/riscv: rvb: rotate (left/right)

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-12-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: shift ones

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-11-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: single-bit instructions

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-10-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: add gen_shifti() and gen_shiftiw() helper functions

Add gen_shifti() and gen_shiftiw() helper functions to reuse the same
interfaces for immediate shift instructions.

Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-9-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: sign-extend instructions

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Message-id: 20210505160620.15723-8-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: min/max instructions

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Message-id: 20210505160620.15723-7-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: pack two words into one register

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-6-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: logic-with-negate

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-5-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: count bits set

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Message-id: 20210505160620.15723-4-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rvb: count leading/trailing zeros

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-3-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: reformat @sh format encoding for B-extension

Signed-off-by: Kito Cheng <kito.cheng@sifive.com>
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210505160620.15723-2-frank.chang@sifive.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Pass the same value to oprsz and maxsz.

Since commit e2e7168a214b0ed98dc357bba96816486a289762, if oprsz
is still zero(as we don't use this field), simd_desc will trigger an
assert.

Besides, tcg_gen_gvec_*_ptr calls simd_desc in it's implementation.
Here we pass the value to maxsz and oprsz to bypass the assert.

Signed-off-by: LIU Zhiwei <zhiwei_liu@c-sky.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210521054816.1784297-1-zhiwei_liu@c-sky.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/pmp: Add assert for ePMP operations

Although we construct epmp_operation in such a way that it can only be
between 0 and 15 Coverity complains that we don't handle the other
possible cases. To fix Coverity and make it easier for humans to read
add a default case to the switch statement that calls
g_assert_not_reached().

Fixes: CID 1453108
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: LIU Zhiwei <zhiwei_liu@c-sky.com>
Message-id: ec5f225928eec448278c82fcb1f6805ee61dde82.1621550996.git.alistair.francis@wdc.com

target/riscv: Dump CSR mscratch/sscratch/satp

This dumps the CSR mscratch/sscratch/satp and meanwhile aligns
the output of CSR mtval/stval.

Signed-off-by: Changbin Du <changbin.du@gmail.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210519155738.20486-1-changbin.du@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Remove unnecessary riscv_*_names[] declaration

riscv_excp_names[] and riscv_intr_names[] are only referenced by
target/riscv/cpu.c locally.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210514052435.2203156-1-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Do not include 'pmp.h' in user emulation

Physical Memory Protection is a system feature.
Avoid polluting the user-mode emulation by its definitions.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 20210516205333.696094-1-f4bug@amsat.org
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

docs/system: Move the RISC-V -bios information to removed

QEMU 5.1 changed the behaviour of the default boot for the RISC-V virt
and sifive_u machines. This patch moves that change from the
deprecated.rst file to the removed-features.rst file and the
target-riscv.rst.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Message-id: 4f1c261e7f69045ab8bb8926d85fe1d35e48ea5b.1620081256.git.alistair.francis@wdc.com

target/riscv: fix wfi exception behavior

The wfi exception trigger behavior should take into account user mode,
hstatus.vtw, and the fact the an wfi might raise different types of
exceptions depending on various factors:

If supervisor mode is not present:

- an illegal instruction exception should be generated if user mode
executes and wfi instruction and mstatus.tw = 1.

If supervisor mode is present:

- when a wfi instruction is executed, an illegal exception should be triggered
if either the current mode is user or the mode is supervisor and mstatus.tw is
set.

Plus, if the hypervisor extensions are enabled:

- a virtual instruction exception should be raised when a wfi is executed from
virtual-user or virtual-supervisor and hstatus.vtw is set.

Signed-off-by: Jose Martins <josemartins90@gmail.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210420213656.85148-1-josemartins90@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: microchip_pfsoc: Support direct kernel boot

At present the Microchip Icicle Kit machine only supports using
'-bios' to load the HSS, and does not support '-kernel' for direct
kernel booting just like other RISC-V machines do. One has to use
U-Boot which is chain-loaded by HSS, to load a kernel for testing.
This is not so convenient.

Adding '-kernel' support together with the existing '-bios', we
follow the following table to select which payload we execute:

  -bios |    -kernel | payload
  ------+------------+--------
      N |          N | HSS
      Y | don't care | HSS
      N |          Y | kernel

This ensures backwards compatibility with how we used to expose
'-bios' to users. When '-kernel' is used for direct boot, '-dtb'
must be present to provide a valid device tree for the board,
as we don't generate device tree.

When direct kernel boot is used, the OpenSBI fw_dynamic BIOS image
is used to boot a payload like U-Boot or OS kernel directly.

Documentation is updated to describe the direct kernel boot. Note
as of today there is still no PolarFire SoC support in the upstream
Linux kernel hence the document does not include instructions for
that. It will be updated in the future.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-8-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Use macros for BIOS image names

The OpenSBI BIOS image names are used by many RISC-V machines.
Let's define macros for them.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-7-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

docs/system/riscv: sifive_u: Document '-dtb' usage

Update the 'sifive_u' machine documentation to mention the '-dtb'
option that can be used to pass a custom DTB to QEMU.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-6-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

docs/system/riscv: Correct the indentation level of supported devices

The supported device bullet list has an additional space before each
entry, which makes a wrong indentation level. Correct it.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-5-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Support the official PLIC DT bindings

The official DT bindings of PLIC uses "sifive,plic-1.0.0" as the
compatible string in the upstream Linux kernel. "riscv,plic0" is
now legacy and has to be kept for backward compatibility of legacy
systems.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-4-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Support the official CLINT DT bindings

Linux kernel commit a2770b57d083 ("dt-bindings: timer: Add CLINT bindings")
adds the official DT bindings for CLINT, which uses "sifive,clint0"
as the compatible string. "riscv,clint0" is now legacy and has to
be kept for backward compatibility of legacy systems.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-3-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: virt: Switch to use qemu_fdt_setprop_string_array() helper

Since commit 78da6a1bca22 ("device_tree: add qemu_fdt_setprop_string_array helper"),
we can use the new helper to set the compatible strings for the
SiFive test device node.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-2-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: sifive_u: Switch to use qemu_fdt_setprop_string_array() helper

Since commit 78da6a1bca22 ("device_tree: add qemu_fdt_setprop_string_array helper"),
we can use the new helper to set the clock name for the ethernet
controller node.

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-id: 20210430071302.1489082-1-bmeng.cn@gmail.com
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

Merge remote-tracking branch 'remotes/stsquad/tags/pull-testing-updates-070621-2' into staging

A few testing and configure updates:

  - add the multiarch signals stress test
  - fix display of multi-word compiler stanzas in meson
  - fix quoting of multi-word compiler stazas in configure.sh
  - tag some acceptance tests as TCG only
  - make checkpatch test work harder to find clean diffs
  - split gprof/gconv job to avoid timeouts
  - fix centos8 VM build by adding --source-path
  - make checkpatch aware of .h.inc and .c.inc paths

# gpg: Signature made Mon 07 Jun 2021 14:51:12 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-updates-070621-2:
  scripts/checkpatch.pl: process .c.inc and .h.inc files as C source
  tests/vm: expose --source-path to scripts to find extra files
  gitlab-ci: Split gprof-gcov job
  gitlab: work harder to avoid false positives in checkpatch
  tests/acceptance: tag various arm tests as TCG only
  tests/tcg/configure.sh: tweak quoting of target_compiler
  meson.build: fix cosmetics of compiler display
  tests/tcg: add a multiarch signals test to stress test signal delivery

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

scripts/checkpatch.pl: process .c.inc and .h.inc files as C source

Change the regex used to determine whether a file should be processed as
C source to include .c.inc and .h.inc extensions.

Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Message-Id: <20210520195142.941261-1-matheus.ferst@eldorado.org.br>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

tests/vm: expose --source-path to scripts to find extra files

Currently the centos8 image expects to run an in-src build to find the
kick starter file. Fix this.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210602103527.32021-1-alex.bennee@linaro.org>

gitlab-ci: Split gprof-gcov job

This job is hitting the 70min limit, so split it in 2 tasks.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Willian Rampazzo <willianr@redhat.com>
Message-Id: <20210525082556.4011380-7-f4bug@amsat.org>

gitlab: work harder to avoid false positives in checkpatch

This copies the behaviour of patchew's configuration to make the diff
algorithm generate a minimal diff.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Message-Id: <20210602153247.27651-1-alex.bennee@linaro.org>

tests/acceptance: tag various arm tests as TCG only

We should never be trying to run most of these models under a KVM
environment.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Willian Rampazzo <willianr@redhat.com>
Message-Id: <20210527160319.19834-7-alex.bennee@linaro.org>

tests/tcg/configure.sh: tweak quoting of target_compiler

If you configure the host compiler with a multi-command stanza like:

  --cc="ccache gcc"

then the configure.sh machinery falls over with confusion. Work around
this by ensuring we correctly quote so where we need a complete
evaluation we get it. Of course the has() check needs single variable
so we need to unquote that. This does mean it essentially checks that
just the ccache command exits but if we got past that step we still
check the compiler actually does something.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Message-Id: <20210527160319.19834-4-alex.bennee@linaro.org>

meson.build: fix cosmetics of compiler display

If you specify something like --cc="ccache gcc" on your configure line
the summary output misses the rest of the cmd_array. Do some string
joining to make it complete.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210527160319.19834-3-alex.bennee@linaro.org>

tests/tcg: add a multiarch signals test to stress test signal delivery

This adds a simple signal test that combines the POSIX timer_create
with signal delivery across multiple threads. The aim is to provide a
bit more of a stress test to flush out signal handling issues for
easily than the occasional random crash we sometimes see in linux-test
or threadcount.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210527160319.19834-2-alex.bennee@linaro.org>

Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210604' into staging

Host vector support for arm neon.

# gpg: Signature made Fri 04 Jun 2021 19:56:59 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-gitlab/tags/pull-tcg-20210604:
  tcg/arm: Implement TCG_TARGET_HAS_rotv_vec
  tcg/arm: Implement TCG_TARGET_HAS_roti_vec
  tcg/arm: Implement TCG_TARGET_HAS_shv_vec
  tcg/arm: Implement TCG_TARGET_HAS_bitsel_vec
  tcg/arm: Implement TCG_TARGET_HAS_minmax_vec
  tcg/arm: Implement TCG_TARGET_HAS_sat_vec
  tcg/arm: Implement TCG_TARGET_HAS_mul_vec
  tcg/arm: Implement TCG_TARGET_HAS_shi_vec
  tcg/arm: Implement andc, orc, abs, neg, not vector operations
  tcg/arm: Implement minimal vector operations
  tcg/arm: Implement tcg_out_dup*_vec
  tcg/arm: Implement tcg_out_mov for vector types
  tcg/arm: Implement tcg_out_ld/st for vector types
  tcg/arm: Add host vector framework
  tcg: Change parameters for tcg_target_const_match

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

tcg/arm: Implement TCG_TARGET_HAS_rotv_vec

Implement via expansion, so don't actually set TCG_TARGET_HAS_rotv_vec.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_roti_vec

Implement via expansion, so don't actually set TCG_TARGET_HAS_roti_vec.
For NEON, this is shift-right followed by shift-left-and-insert.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_shv_vec

The three vector shift by vector operations are all implemented via
expansion.  Therefore do not actually set TCG_TARGET_HAS_shv_vec,
as none of shlv_vec, shrv_vec, sarv_vec may actually appear in the
instruction stream, and therefore also do not appear in tcg_target_op_def.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_bitsel_vec

NEON has 3 instructions implementing this 4 argument operation,
with each insn overlapping a different logical input onto the
destination register.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_minmax_vec

This is minimum and maximum, signed and unsigned.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_sat_vec

This is saturating add and subtract, signed and unsigned.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_mul_vec

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement TCG_TARGET_HAS_shi_vec

This consists of the three immediate shifts: shli, shri, sari.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement andc, orc, abs, neg, not vector operations

These logical and arithmetic operations are optional, but are
trivial to accomplish with the existing infrastructure.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement minimal vector operations

Implementing dup2, add, sub, and, or, xor as the minimal set.
This allows us to actually enable neon in the header file.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement tcg_out_dup*_vec

Most of dupi is copied from tcg/aarch64, which has the same
encoding for AdvSimdExpandImm.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement tcg_out_mov for vector types

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Implement tcg_out_ld/st for vector types

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Add host vector framework

Add registers and function stubs.  The functionality
is disabled via use_neon_instructions defined to 0.

We must still include results for the mandatory opcodes in
tcg_target_op_def, as all opcodes are checked during tcg init.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg: Change parameters for tcg_target_const_match

Change the return value to bool, because that's what is should
have been from the start.  Pass the ct mask instead of the whole
TCGArgConstraint, as that's the only part that's relevant.

Change the value argument to int64_t.  We will need the extra
width for 32-bit hosts wanting to match vector constants.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

Merge remote-tracking branch 'remotes/bonzini-gitlab/tags/for-upstream' into staging

* OpenBSD cleanup (Brad)
* fixes for the i386 accel/cpu refactoring (Claudio)
* unmap test for emulated SCSI (Kit)
* fix for iscsi module (myself)
* fix for -readconfig of objects (myself)
* fixes for x86 16-bit task switching (myself)
* fix for x86 MOV from/to CR8 (Richard)

# gpg: Signature made Fri 04 Jun 2021 12:53:32 BST
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini-gitlab/tags/for-upstream:
  vl: plug -object back into -readconfig
  vl: plumb keyval-based options into -readconfig
  qemu-config: parse configuration files to a QDict
  i386: run accel_cpu_instance_init as post_init
  i386: reorder call to cpu_exec_realizefn
  tests/qtest/virtio-scsi-test: add unmap large LBA with 4k blocks test
  target/i386: Fix decode of cr8
  target/i386: tcg: fix switching from 16-bit to 32-bit tasks or vice versa
  target/i386: tcg: fix loading of registers from 16-bit TSS
  target/i386: tcg: fix segment register offsets for 16-bit TSS
  oslib-posix: Remove OpenBSD workaround for fcntl("/dev/null", F_SETFL, O_NONBLOCK) failure
  iscsi: link libm into the module
  meson: allow optional dependencies for block modules

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

Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging

# gpg: Signature made Fri 04 Jun 2021 08:26:16 BST
# gpg:                using RSA key EF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <jasowang@redhat.com>" [marginal]
# 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: 215D 46F4 8246 689E C77F  3562 EF04 965B 398D 6211

* remotes/jasowang/tags/net-pull-request:
  MAINTAINERS: Added eBPF maintainers information.
  docs: Added eBPF documentation.
  virtio-net: Added eBPF RSS to virtio-net.
  ebpf: Added eBPF RSS loader.
  ebpf: Added eBPF RSS program.
  net: Added SetSteeringEBPF method for NetClientState.
  net/tap: Added TUNSETSTEERINGEBPF code.

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

vl: plug -object back into -readconfig

Commit bc2f4fcb1d ("qom: move user_creatable_add_opts logic to vl.c
and QAPIfy it", 2021-03-19) switched the creation of objects from
qemu_opts_foreach to a bespoke QTAILQ in preparation for supporting JSON
syntax in -object.

Unfortunately in doing so it lost support for [object] stanzas in
configuration files and also for "-set object.ID.KEY=VAL".  The latter
is hard to re-establish and probably best solved by deprecating -set.
This patch uses the infrastructure introduced by the previous two
patches in order to parse QOM objects correctly from configuration
files.

Cc: Markus Armbruster <armbru@redhat.com>
Cc: qemu-stable@nongnu.org
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210524105752.3318299-4-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

vl: plumb keyval-based options into -readconfig

Let -readconfig support parsing command line options into QDict or
QemuOpts.  This will be used to add back support for objects in
-readconfig.

Cc: Markus Armbruster <armbru@redhat.com>
Cc: qemu-stable@nongnu.org
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210524105752.3318299-3-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

qemu-config: parse configuration files to a QDict

Change the parser to put the values into a QDict and pass them
to a callback.  qemu_config_parse's QemuOpts creation is
itself turned into a callback function.

This is useful for -readconfig to support keyval-based options;
getting a QDict from the parser removes a roundtrip from
QDict to QemuOpts and then back to QDict.

Unfortunately there is a disadvantage in that semantic errors will
point to the last line of the group, because the entries of the QDict
do not have a location attached.

Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210524105752.3318299-2-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386: run accel_cpu_instance_init as post_init

This fixes host and max cpu initialization, by running the accel cpu
initialization only after all instance init functions are called for all
X86 cpu subclasses.

The bug this is fixing is related to the "max" and "host" i386 cpu
subclasses, which set cpu->max_features, which is then used at cpu
realization time.

In order to properly split the accel-specific max features code that
needs to be executed at cpu instance initialization time,

we cannot call the accel cpu initialization at the end of the x86 base
class initialization, or we will have no way to specialize
"max features" cpu behavior, overriding the "max" cpu class defaults,
and checking for the "max features" flag itself.

This patch moves the accel-specific cpu instance initialization to after
all x86 cpu instance code has been executed, including subclasses,

so that proper initialization of cpu "host" and "max" can be restored.

Fixes: f5cc5a5c ("i386: split cpu accelerators from cpu.c,"...)
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Message-Id: <20210603123001.17843-3-cfontana@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

i386: reorder call to cpu_exec_realizefn

i386 realizefn code is sensitive to ordering, and recent commits
aimed at refactoring it, splitting accelerator-specific code,
broke assumptions which need to be fixed.

We need to:

* process hyper-v enlightements first, as they assume features
  not to be expanded

* only then, expand features

* after expanding features, attempt to check them and modify them in the
  accel-specific realizefn code called by cpu_exec_realizefn().

* after the framework has been called via cpu_exec_realizefn,
  the code can check for what has or hasn't been set by accel-specific
  code, or extend its results, ie:

  - check and evenually set code_urev default
  - modify cpu->mwait after potentially being set from host CPUID.
  - finally check for phys_bits assuming all user and accel-specific
    adjustments have already been taken into account.

Fixes: f5cc5a5c ("i386: split cpu accelerators from cpu.c"...)
Fixes: 30565f10 ("cpu: call AccelCPUClass::cpu_realizefn in"...)
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Message-Id: <20210603123001.17843-2-cfontana@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

tests/qtest/virtio-scsi-test: add unmap large LBA with 4k blocks test

Add test for issue #345

Signed-off-by: Kit Westneat <kit.westneat@gmail.com>
Message-Id: <20210603142022.676395-1-kit.westneat@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Fix decode of cr8

A recent cleanup did not recognize that there are two ways
to encode cr8: one via the LOCK and the other via REX.

Fixes: 7eff2e7c
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/380
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210602035511.96834-1-richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: tcg: fix switching from 16-bit to 32-bit tasks or vice versa

The format of the task state segment is governed by bit 3 in the
descriptor type field.  On a task switch, the format for saving
is given by the current value of TR's type field, while the
format for loading is given by the new descriptor.

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

target/i386: tcg: fix loading of registers from 16-bit TSS

According to the manual, the high 16-bit of the registers are preserved
when switching to a 16-bit task.  Implement this in switch_tss_ra.

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

target/i386: tcg: fix segment register offsets for 16-bit TSS

The TSS offsets in the manuals have only 2-byte slots for the
segment registers.  QEMU incorrectly uses 4-byte slots, so
that SS overlaps the LDT selector.

Resolves: #382
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

oslib-posix: Remove OpenBSD workaround for fcntl("/dev/null", F_SETFL, O_NONBLOCK) failure

OpenBSD prior to 6.3 required a workaround to utilize fcntl(F_SETFL) on memory
devices.

Since modern verions of OpenBSD that are only officialy supported and buildable
on do not have this issue I am garbage collecting this workaround.

Signed-off-by: Brad Smith <brad@comstyle.com>
Message-Id: <YGYECGXQhdamEJgC@humpty.home.comstyle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

iscsi: link libm into the module

Depending on the configuration of QEMU, some binaries might not need libm
at all.  In that case libiscsi, which uses exp(), will fail to load.
Link it in the module explicitly.

Reported-by: Yi Sun <yisun@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

meson: allow optional dependencies for block modules

Right now all dependencies for block modules are passed to
module_ss.add(when: ...), so they are mandatory.  In the next patch we
will need to add a libm dependency to a module, but libm does not exist
on all systems.  So, modify the creation of module_ss and modsrc so that
dependencies can also be passed to module_ss.add(if_true: ...).

While touching the array, remove the useless dependency of the curl
module on glib.  glib is always linked in QEMU and in fact all other
block modules also need it, but they don't have to specify it.

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

Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-fpu-20210603' into staging

Finish conversion of float128 and floatx80 to FloatParts.
Implement float128_muladd and float128_{min,max}*.
Optimize int-to-float conversion with hard-float.

# gpg: Signature made Thu 03 Jun 2021 22:13:10 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-gitlab/tags/pull-fpu-20210603: (29 commits)
  softfloat: Use hard-float for {u}int64_to_float{32,64}
  tests/fp: Enable more tests
  softfloat: Convert modrem operations to FloatParts
  softfloat: Move floatN_log2 to softfloat-parts.c.inc
  softfloat: Convert float32_exp2 to FloatParts
  softfloat: Convert floatx80 compare to FloatParts
  softfloat: Convert floatx80_scalbn to FloatParts
  softfloat: Convert floatx80 to integer to FloatParts
  softfloat: Convert floatx80 float conversions to FloatParts
  softfloat: Convert integer to floatx80 to FloatParts
  softfloat: Convert floatx80_round_to_int to FloatParts
  softfloat: Convert floatx80_round to FloatParts
  softfloat: Convert floatx80_sqrt to FloatParts
  softfloat: Convert floatx80_div to FloatParts
  softfloat: Convert floatx80_mul to FloatParts
  softfloat: Convert floatx80_add/sub to FloatParts
  tests/fp/fp-test: Reverse order of floatx80 precision tests
  softfloat: Adjust parts_uncanon_normal for floatx80
  softfloat: Introduce Floatx80RoundPrec
  softfloat: Reduce FloatFmt
  ...

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

MAINTAINERS: Added eBPF maintainers information.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

docs: Added eBPF documentation.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

virtio-net: Added eBPF RSS to virtio-net.

When RSS is enabled the device tries to load the eBPF program
to select RX virtqueue in the TUN. If eBPF can be loaded
the RSS will function also with vhost (works with kernel 5.8 and later).
Software RSS is used as a fallback with vhost=off when eBPF can't be loaded
or when hash population requested by the guest.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Added eBPF RSS loader.

Added function that loads RSS eBPF program.
Added stub functions for RSS eBPF loader.
Added meson and configuration options.

By default, eBPF feature enabled if libbpf is present in the build system.
libbpf checked in configuration shell script and meson script.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

ebpf: Added eBPF RSS program.

RSS program and Makefile to build it.
The bpftool used to generate '.h' file.
The data in that file may be loaded by libbpf.
EBPF compilation is not required for building qemu.
You can use Makefile if you need to regenerate rss.bpf.skeleton.h.

Signed-off-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

net: Added SetSteeringEBPF method for NetClientState.

For now, that method supported only by Linux TAP.
Linux TAP uses TUNSETSTEERINGEBPF ioctl.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

net/tap: Added TUNSETSTEERINGEBPF code.

Additional code that will be used for eBPF setting steering routine.

Signed-off-by: Andrew Melnychenko <andrew@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>

softfloat: Use hard-float for {u}int64_to_float{32,64}

For the normal case of no additional scaling, this reduces the
profile contribution of int64_to_float64 to the testcase in the
linked issue from 0.81% to 0.04%.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/134
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tests/fp: Enable more tests

Fix the trivial typo in extF80_lt_quiet, and re-enable
all of the floatx80 tests that are now fixed.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-ID: <87bl9iyahr.fsf@linaro.org>
[rth: Squash the fix for lt_quiet, and enable that too.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Convert modrem operations to FloatParts

Rename to parts$N_modrem.  This was the last use of a lot
of the legacy infrastructure, so remove it as required.

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

softfloat: Move floatN_log2 to softfloat-parts.c.inc

Rename to parts$N_log2.  Though this is partly a ruse, since I do not
believe the code will succeed for float128 without work.  Which is ok
for now, because we do not need this for more than float32 and float64.

Since berkeley-testfloat-3 doesn't support log2, compare float64_log2
vs the system log2.  Fix the errors for inputs near 1.0:

test: 3ff00000000000b0  +0x1.00000000000b0p+0
  sf: 3d2fa00000000000  +0x1.fa00000000000p-45
libm: 3d2fbd422b1bd36f  +0x1.fbd422b1bd36fp-45
Error in fraction: 32170028290927 ulp

test: 3feec24f6770b100  +0x1.ec24f6770b100p-1
  sf: bfad3740d13c9ec0  -0x1.d3740d13c9ec0p-5
libm: bfad3740d13c9e98  -0x1.d3740d13c9e98p-5
Error in fraction: 40 ulp

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

softfloat: Convert float32_exp2 to FloatParts

Keep the intermediate results in FloatParts instead of
converting back and forth between float64.  Use muladd
instead of separate mul+add.

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

softfloat: Convert floatx80 compare to FloatParts

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

softfloat: Convert floatx80_scalbn to FloatParts

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

softfloat: Convert floatx80 to integer to FloatParts

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

softfloat: Convert floatx80 float conversions to FloatParts

This is the last use of commonNaNT and all of the routines
that use it, so remove all of them for Werror.

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

softfloat: Convert integer to floatx80 to FloatParts

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

softfloat: Convert floatx80_round_to_int to FloatParts

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

softfloat: Convert floatx80_round to FloatParts

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

softfloat: Convert floatx80_sqrt to FloatParts

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

softfloat: Convert floatx80_div to FloatParts

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

softfloat: Convert floatx80_mul to FloatParts

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

softfloat: Convert floatx80_add/sub to FloatParts

Since this is the first such, this includes all of the
packing and unpacking routines as well.

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

tests/fp/fp-test: Reverse order of floatx80 precision tests

Many qemu softfloat will check floatx80_rounding_precision
even when berkeley testfloat will not.  So begin with
floatx80_precision_x, so that's the one we use
when !FUNC_EFF_ROUNDINGPRECISION.

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

softfloat: Adjust parts_uncanon_normal for floatx80

With floatx80_precision_x, the rounding happens across
the break between words.  Notice this case with

  frac_lsb = round_mask + 1 -> 0

and check the bits in frac_hi as needed.

In addition, since frac_shift == 0, we won't implicitly clear
round_mask via the right-shift, so explicitly clear those bits.
This fixes rounding for floatx80_precision_[sd].

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

softfloat: Introduce Floatx80RoundPrec

Use an enumeration instead of raw 32/64/80 values.

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

softfloat: Reduce FloatFmt

Remove frac_lsb, frac_lsbm1, roundeven_mask.  Compute
these from round_mask in parts$N_uncanon_normal.

With floatx80, round_mask will not be tied to frac_shift.
Everything else is easily computable.

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

softfloat: Split out parts_uncanon_normal

We will need to treat the non-normal cases of floatx80 specially,
so split out the normal case that we can reuse.

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

softfloat: Move sqrt_float to softfloat-parts.c.inc

Rename to parts$N_sqrt.
Reimplement float128_sqrt with FloatParts128.

Reimplement with the inverse sqrt newton-raphson algorithm from musl.
This is significantly faster than even the berkeley sqrt n-r algorithm,
because it does not use division instructions, only multiplication.

Ordinarily, changing algorithms at the same time as migrating code is
a bad idea, but this is the only way I found that didn't break one of
the routines at the same time.

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

softfloat: Move scalbn_decomposed to softfloat-parts.c.inc

Rename to parts$N_scalbn.
Reimplement float128_scalbn with FloatParts128.

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

softfloat: Move compare_floats to softfloat-parts.c.inc

Rename to parts$N_compare.  Rename all of the intermediate
functions to ftype_do_compare.  Rename the hard-float functions
to ftype_hs_compare.  Convert float128 to FloatParts128.

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

softfloat: Implement float128_(min|minnum|minnummag|max|maxnum|maxnummag)

The float128 implementation is straight-forward.
Unfortuantely, we don't have any tests we can simply adjust/unlock.

Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210517142739.38597-24-david@redhat.com>
[rth: Update for changed parts_minmax return value]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move minmax_flags to softfloat-parts.c.inc

Rename to parts$N_minmax.  Combine 3 bool arguments to a bitmask.
Introduce ftype_minmax functions as a common optimization point.
Fold bfloat16 expansions into the same macro as the other types.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>