target/riscv/cpu.c: add riscv_bare_cpu_init()

Next patch will add more bare CPUs. Their cpu_init() functions would be
glorified copy/pastes of rv64i_bare_cpu_init(), differing only by a
riscv_cpu_set_misa() call.

Add a new .instance_init for the TYPE_RISCV_BARE_CPU typ to avoid this
code repetition. While we're at it, add a better explanation on why
we're disabling the timing extensions for bare CPUs.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122123348.973288-2-dbarboza@ventanamicro.com>
[ Changes by AF:
 - Rebase on latest changes
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Enable xtheadsync under user mode

According to xtheadsync[1][2] documentation, it can be used in user mode and
the behavior is same with other priviledges.

[1]:https://github.com/T-head-Semi/thead-extension-spec/blob/master/xtheadsync/sync.adoc
[2]:https://github.com/T-head-Semi/thead-extension-spec/blob/master/xtheadsync/sync_i.adoc

Signed-off-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240204055228.900-1-zhiwei_liu@linux.alibaba.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

qemu-options: enable -smbios option on RISC-V

With SMBIOS support added for RISC-V we also should enable the command line
option.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Message-ID: <20240123184229.10415-5-heinrich.schuchardt@canonical.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: SMBIOS support for RISC-V virt machine

Generate SMBIOS tables for the RISC-V mach-virt.
Add CONFIG_SMBIOS=y to the RISC-V default config.
Set the default processor family in the type 4 table.

The implementation is based on the corresponding ARM and Loongson code.

With the patch the following firmware tables are provided:

    etc/smbios/smbios-anchor
    etc/smbios/smbios-tables

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Message-ID: <20240123184229.10415-4-heinrich.schuchardt@canonical.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

smbios: function to set default processor family

Provide a function to set the default processor family.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Message-ID: <20240123184229.10415-3-heinrich.schuchardt@canonical.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

smbios: add processor-family option

For RISC-V the SMBIOS standard requires specific values of the processor
family value depending on the bitness of the CPU.

Add a processor-family option for SMBIOS table 4.

The value of processor-family may exceed 255 and therefore must be provided
in the Processor Family 2 field. Set the Processor Family field to 0xFE
which signals that the Processor Family 2 is used.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Message-ID: <20240123184229.10415-2-heinrich.schuchardt@canonical.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: support new isa extension detection devicetree properties

A few months ago I submitted a patch to various lists, deprecating
"riscv,isa" with a lengthy commit message [0] that is now commit
aeb71e42caae ("dt-bindings: riscv: deprecate riscv,isa") in the Linux
kernel tree. Primarily, the goal was to replace "riscv,isa" with a new
set of properties that allowed for strictly defining the meaning of
various extensions, where "riscv,isa" was tied to whatever definitions
inflicted upon us by the ISA manual, which have seen some variance over
time.

Two new properties were introduced: "riscv,isa-base" and
"riscv,isa-extensions". The former is a simple string to communicate the
base ISA implemented by a hart and the latter an array of strings used
to communicate the set of ISA extensions supported, per the definitions
of each substring in extensions.yaml [1]. A beneficial side effect was
also the ability to define vendor extensions in a more "official" way,
as the ISA manual and other RVI specifications only covered the format
for vendor extensions in the ISA string, but not the meaning of vendor
extensions, for obvious reasons.

Add support for setting these two new properties in the devicetrees for
the various devicetree platforms supported by QEMU for RISC-V. The Linux
kernel already supports parsing ISA extensions from these new
properties, and documenting them in the dt-binding is a requirement for
new extension detection being added to the kernel.

A side effect of the implementation is that the meaning for elements in
"riscv,isa" and in "riscv,isa-extensions" are now tied together as they
are constructed from the same source. The same applies to the ISA string
provided in ACPI tables, but there does not appear to be any strict
definitions of meanings in ACPI land either.

Link: https://lore.kernel.org/qemu-riscv/20230702-eats-scorebook-c951f170d29f@spud/
Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/riscv/extensions.yaml
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240124-unvarying-foothold-9dde2aaf95d4@spud>
[ Changes by AF:
 - Rebase on recent changes
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: use misa_mxl_max to populate isa string rather than TARGET_LONG_BITS

A cpu may not have the same xlen as the compile time target, and
misa_mxl_max is the source of truth for what the hart supports.

The conversion from misa_mxl_max to xlen already has one user, so
introduce a helper and use that to populate the isa string.

Link: https://lore.kernel.org/qemu-riscv/20240108-efa3f83dcd3997dc0af458d7@orel/
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240124-swear-monthly-56c281f809a6@spud>
[ Changes by AF:
 - Convert to use RISCVCPUClass *mcc
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Expose Zaamo and Zalrsc extensions

Expose the newly added extensions to the guest and allow their control
through the CPU properties.

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240123111030.15074-4-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Check 'A' and split extensions for atomic instructions

Following the pattern for 'M' and Zmmul check if either the 'A'
extension is enabled or the appropriate split extension for the
instruction.

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240123111030.15074-3-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Add Zaamo and Zalrsc extension infrastructure

These extensions represent the atomic operations from A (Zaamo) and the
Load-Reserved/Store-Conditional operations from A (Zalrsc)

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240123111030.15074-2-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: use g_autofree in create_fdt_*

We have a lot of cases where a char or an uint32_t pointer is used once
to alloc a string/array, read/written during the function, and then
g_free() at the end. There's no pointer re-use - a single alloc, a
single g_free().

Use 'g_autofree' to avoid the g_free() calls.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122221529.86562-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: use g_autofree in virt_machine_init()

Move 'soc_name' to the loop, and give it g_autofree, to avoid the manual
g_free().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240122221529.86562-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: use g_autofree in create_fdt_virtio()

Put 'name' declaration inside the loop, with g_autofree, to avoid
manually doing g_free() in each iteration.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240122221529.86562-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: use g_autofree in create_fdt_sockets()

Move 'clust_name' inside the loop, and g_autofree, to avoid having to
g_free() manually in each loop iteration.

'intc_phandles' is also g_autofreed to avoid another manual g_free().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122221529.86562-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: use g_autofree in create_fdt_socket_cpus()

Move all char pointers to the loop. Use g_autofree in all of them to
avoid the g_free() calls.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20240122221529.86562-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/numa.c: use g_autofree in socket_fdt_write_distance_matrix()

Use g_autofree in 'dist_matrix' to avoid the manual g_free().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122221529.86562-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt-acpi-build.c: fix leak in build_rhct()

The 'isa' char pointer isn't being freed after use.

Issue detected by Valgrind:

==38752== 128 bytes in 1 blocks are definitely lost in loss record 3,190 of 3,884
==38752==    at 0x484280F: malloc (vg_replace_malloc.c:442)
==38752==    by 0x5189619: g_malloc (gmem.c:130)
==38752==    by 0x51A5BF2: g_strconcat (gstrfuncs.c:628)
==38752==    by 0x6C1E3E: riscv_isa_string_ext (cpu.c:2321)
==38752==    by 0x6C1E3E: riscv_isa_string (cpu.c:2343)
==38752==    by 0x6BD2EA: build_rhct (virt-acpi-build.c:232)
==38752==    by 0x6BD2EA: virt_acpi_build (virt-acpi-build.c:556)
==38752==    by 0x6BDC86: virt_acpi_setup (virt-acpi-build.c:662)
==38752==    by 0x9C8DC6: notifier_list_notify (notify.c:39)
==38752==    by 0x4A595A: qdev_machine_creation_done (machine.c:1589)
==38752==    by 0x61E052: qemu_machine_creation_done (vl.c:2680)
==38752==    by 0x61E052: qmp_x_exit_preconfig.part.0 (vl.c:2709)
==38752==    by 0x6220C6: qmp_x_exit_preconfig (vl.c:2702)
==38752==    by 0x6220C6: qemu_init (vl.c:3758)
==38752==    by 0x425858: main (main.c:47)

Fixes: ebfd392893 ("hw/riscv/virt: virt-acpi-build.c: Add RHCT Table")
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122221529.86562-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Use RISCVException as return type for all csr ops

The real return value type has been converted to RISCVException,
but some function declarations still not. This patch makes all
csr operation declarations use RISCVExcetion.

Signed-off-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240130110844.437-1-zhiwei_liu@linux.alibaba.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: FCSR doesn't contain vxrm and vxsat

vxrm and vxsat have been moved into a special register vcsr since
RVV v1.0. So remove them from FCSR for vector 1.0.

Signed-off-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240130110945.486-1-zhiwei_liu@linux.alibaba.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Validate misa_mxl_max only once

misa_mxl_max is now a class member and initialized only once for each
class. This also moves the initialization of gdb_core_xml_file which
will be referenced before realization in the future.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240203-riscv-v11-3-a23f4848a628@daynix.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Move misa_mxl_max to class

misa_mxl_max is common for all instances of a RISC-V CPU class so they
are better put into class.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240203-riscv-v11-2-a23f4848a628@daynix.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Remove misa_mxl validation

It is initialized with a simple assignment and there is little room for
error. In fact, the validation is even more complex.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Acked-by: LIU Zhiwei <zhiwei_liu@linux.alibaba.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240203-riscv-v11-1-a23f4848a628@daynix.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/kvm: get/set vector vregs[]

vregs[] have variable size that depends on the current vlenb set by the
host, meaning we can't use our regular kvm_riscv_reg_id() to retrieve
it.

Create a generic kvm_encode_reg_size_id() helper to encode any given
size in bytes into a given kvm reg id. kvm_riscv_vector_reg_id() will
use it to encode vlenb into a given vreg ID.

kvm_riscv_(get|set)_vector() can then get/set all 32 vregs.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240123161714.160149-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/kvm: initialize 'vlenb' via get-reg-list

KVM will check for the correct 'reg_size' when accessing the vector
registers, erroring with EINVAL if we encode the wrong size in reg ID.
Vector registers varies in size with the vector length in bytes, or
'vlenb'. This means that we need the current 'vlenb' being used by the
host, otherwise we won't be able to fetch all vector regs.

We'll deal with 'vlenb' first. Its support was added in Linux 6.8 as a
get-reg-list register. We'll read 'vlenb' via get-reg-list and mark the
register as 'supported'. All 'vlenb' ops via kvm_arch_get_registers()
and kvm_arch_put_registers() will only be done if the reg is supported,
i.e. we fetched it in get-reg-list during init.

If the user sets a new vlenb value using the 'vlen' property, throw an
error if the user value differs from the host.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240123161714.160149-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/kvm: change kvm_reg_id to uint64_t

The field isn't big enough to hold an uint64_t kvm register and Vector
registers will end up overflowing it.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240123161714.160149-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu.c: remove cpu->cfg.vlen

There is no need to keep both 'vlen' and 'vlenb'. All existing code
that requires 'vlen' is retrieving it via 'vlenb << 3'.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-14-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

trans_rvv.c.inc: use vext_get_vlmax() in trans_vrgather_v*()

Use the helper instead of calculating vlmax by hand.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-13-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: change vext_get_vlmax() arguments

We'll re-use the logic froim vext_get_vlmax() in 2 other occurrences in
the next patch, but first we need to make it independent of both 'cpu'
and 'vtype'. To do that, add 'vlenb', 'vsew' and 'lmul' as parameters
instead.

Adapt the two existing callers. In cpu_get_tb_cpu_state(), rename 'sew'
to 'vsew' to be less ambiguous about what we're encoding into *pflags.

In HELPER(vsetvl) the following changes were made:

- add a 'vsew' var to store vsew. Use it in the shift to get 'sew';
- the existing 'lmul' var was renamed to 'vlmul';
- add a new 'lmul' var to store 'lmul' encoded like DisasContext:lmul.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-12-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu.h: use 'vlenb' in vext_get_vlmax()

Rename the existing 'sew' variable to 'vsew' for extra clarity.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-11-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/insn_trans/trans_rvv.c.inc: use 'vlenb' in MAXSZ()

Calculate the maximum vector size possible, 'max_sz', which is the size
in bytes 'vlenb' multiplied by the max value of LMUL (LMUL = 8, when
s->lmul = 3).

'max_sz' is then shifted right by 'scale', expressed as '3 - s->lmul',
which is clearer than doing 'scale = lmul - 3' and then using '-scale'
in the shift right.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-10-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/vector_helper.c: use vlenb in HELPER(vsetvl)

Use the new 'vlenb' CPU config to validate fractional LMUL. The original
comparison is done with 'vlen' and 'sew', both in bits. Adjust the shift
to use vlenb.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-9-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/vector_helper.c: use 'vlenb'

Use 'cpu->cfg.vlenb' instead of 'cpu->cfg.vlen >> 3'.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/insn_trans/trans_rvvk.c.inc: use 'vlenb'

Use s->cfg_ptr->vlenb instead of s->cfg_ptr->vlen / 8.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/insn_trans/trans_rvv.c.inc: use 'vlenb'

Use s->cfg_ptr->vlenb instead of "s->cfg_ptr->vlen / 8"  and
"s->cfg_ptr->vlen >> 3".

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/insn_trans/trans_rvbf16.c.inc: use cpu->cfg.vlenb

Use ctx->cfg_ptr->vlenb instead of ctx->cfg_ptr->vlen / 8.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/gdbstub.c: use 'vlenb' instead of shifting 'vlen'

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/csr.c: use 'vlenb' instead of 'vlen'

As a bonus, we're being more idiomatic using cpu->cfg.vlenb when
reading CSR_VLENB.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: add 'vlenb' field in cpu->cfg

Our usage of 'vlenb' is overwhelming superior than the use of 'vlen'.
We're using 'vlenb' most of the time, having to do 'vlen >> 3' or
'vlen / 8' in every instance.

In hindsight we would be better if the 'vlenb' property  was introduced
instead of 'vlen'. That's not what happened, and now we can't easily get
rid of it due to user scripts all around. What we can do, however, is to
change our internal representation to use 'vlenb'.

Add a 'vlenb' field in cpu->cfg. It'll be set via the existing 'vlen'
property, i.e. setting 'vlen' will also set 'vlenb'.

We'll replace all 'vlen >> 3' code to use 'vlenb' directly. Start with
the single instance we have in target/riscv/cpu.c.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Implement optional CSR mcontext of debug Sdtrig extension

The debug Sdtrig extension defines an CSR "mcontext". This commit
implements its predicate and read/write operations into CSR table.
Its value is reset as 0 when the trigger module is reset.

Signed-off-by: Alvin Chang <alvinga@andestech.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20231219123244.290935-1-alvinga@andestech.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu.c: move 'marchid' to riscv_cpu_properties[]

Keep all class properties in riscv_cpu_properties[].

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-9-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu.c: move 'mimpid' to riscv_cpu_properties[]

Keep all class properties in riscv_cpu_properties[].

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu.c: move 'mvendorid' to riscv_cpu_properties[]

Keep all class properties in riscv_cpu_properties[].

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: remove riscv_cpu_options[]

The array is empty and can be removed.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'cboz_blocksize' to riscv_cpu_properties[]

And remove the now unused kvm_cpu_set_cbomz_blksize() setter.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'cbop_blocksize' to riscv_cpu_properties[]

Do the same we did with 'cbom_blocksize' in the previous patch.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'cbom_blocksize' to riscv_cpu_properties[]

After adding a KVM finalize() implementation, turn cbom_blocksize into a
class property. Follow the same design we used with 'vlen' and 'elen'.

The duplicated 'cbom_blocksize' KVM property can be removed from
kvm_riscv_add_cpu_user_properties().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: create finalize_features() for KVM

To turn cbom_blocksize and cboz_blocksize into class properties we need
KVM specific changes.

KVM is creating its own version of these options with a customized
setter() that prevents users from picking an invalid value during init()
time. This comes at the cost of duplicating each option that KVM
supports. This will keep happening for each new shared option KVM
implements in the future.

We can avoid that by using the same property TCG uses and adding
specific KVM handling during finalize() time, like TCG already does with
riscv_tcg_cpu_finalize_features(). To do that, the common CPU property
offers a way of knowing if an option was user set or not, sparing us
from doing unneeded syscalls.

riscv_kvm_cpu_finalize_features() is then created using the same
KVMScratch CPU we already use during init() time, since finalize() time
is still too early to use the official KVM CPU for it. cbom_blocksize
and cboz_blocksize are then handled during finalize() in the same way
they're handled by their KVM specific setter.

With this change we can proceed with the blocksize changes in the common
code without breaking the KVM driver.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'elen' to riscv_cpu_properties[]

Do the same thing we did with 'vlen' in the previous patch with 'elen'.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-10-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'vlen' to riscv_cpu_properties[]

Turning 'vlen' into a class property will allow its default value to be
overwritten by cpu_init() later on, solving the issue we have now where
CPU specific settings are getting overwritten by the default.

Common validation bits are moved from riscv_cpu_validate_v() to
prop_vlen_set() to be shared with KVM.

And, as done with every option we migrated to riscv_cpu_properties[],
vendor CPUs can't have their 'vlen' value changed.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-9-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rework 'vext_spec'

The same rework did in 'priv_spec' is done for 'vext_spec'. This time is
simpler, since we only accept one value ("v1.0") and we'll always have
env->vext_ver set to VEXT_VERSION_1_00_0, thus we don't need helpers to
convert string to 'vext_ver' back and forth like we needed for
'priv_spec'.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: rework 'priv_spec'

'priv_spec' and 'vext_spec' are two string options used as a fancy way
of setting integers in the CPU state (cpu->env.priv_ver and
cpu->env.vext_ver). It requires us to deal with string parsing and to
store them in cpu_cfg.

We must support these string options, but we don't need to store them.
We have a precedence for this kind of arrangement in target/ppc/compat.c,
ppc_compat_prop_get|set, getters and setters used for the
'max-cpu-compat' class property of the pseries ppc64 machine. We'll do
the same with both 'priv_spec' and 'vext_spec'.

For 'priv_spec', the validation from riscv_cpu_validate_priv_spec() will
be done by the prop_priv_spec_set() setter, while also preventing it to
be changed for vendor CPUs. Add two helpers that converts env->priv_ver
back and forth to its string representation. These helpers allow us to
get a string and set 'env->priv_ver' and return a string giving the
current env->priv_ver value. In other words, make the cpu->cfg.priv_spec
string obsolete.

Last but not the least, move the reworked 'priv_spec' option to
riscv_cpu_properties[].

After all said and done, we don't need to store the 'priv_spec' string in
the CPU state, and we're now protecting vendor CPUs from priv_ver
changes:

$ ./build/qemu-system-riscv64 -M virt -cpu sifive-e51,priv_spec="v1.12.0"
qemu-system-riscv64: can't apply global sifive-e51-riscv-cpu.priv_spec=v1.12.0:
    CPU 'sifive-e51' does not allow changing the value of 'priv_spec'
Current 'priv_spec' val: v1.10.0
$

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'pmp' to riscv_cpu_properties[]

Move 'pmp' to riscv_cpu_properties[], creating a new setter() for it
that forbids 'pmp' to be changed in vendor CPUs, like we did with the
'mmu' option.

We'll also have to manually set 'pmp = true' to generic CPUs that were
still relying on the previous default to set it.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'mmu' to riscv_cpu_properties[]

Commit 7f0bdfb5bfc ("target/riscv/cpu.c: remove cfg setup from
riscv_cpu_init()") already did some of the work by making some
cpu_init() functions to explictly enable their own 'mmu' default.

The generic CPUs didn't get update by that commit, so they are still
relying on the defaults set by the 'mmu' option. But having 'mmu' and
'pmp' being default=true will force CPUs that doesn't implement these
options to set them to 'false' in their cpu_init(), which isn't ideal.

We'll move 'mmu' to riscv_cpu_properties[] without any defaults, i.e.
the default will be 'false'. Compensate it by manually setting 'mmu =
true' to the generic CPUs that requires it.

Implement a setter for it to forbid the 'mmu' setting to be changed for
vendor CPUs. This will allow the option to exist for all CPUs and, at
the same time, protect vendor CPUs from undesired changes:

$ ./build/qemu-system-riscv64 -M virt -cpu sifive-e51,mmu=true
qemu-system-riscv64: can't apply global sifive-e51-riscv-cpu.mmu=true:
   CPU 'sifive-e51' does not allow changing the value of 'mmu'

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: move 'pmu-mask' and 'pmu-num' to riscv_cpu_properties[]

Every property in riscv_cpu_options[] will be migrated to
riscv_cpu_properties[]. This will make their default values init
earlier, allowing cpu_init() functions to overwrite them. We'll also
implement common getters and setters that both accelerators will use,
allowing them to share validations that TCG is doing.

At the same time, some options (namely 'vlen', 'elen' and the cache
blocksizes) need a way of tracking if the user set a value for them.
This is benign for TCG since the cost of always validating these values
are small, but for KVM we need syscalls to read the host values to make
the validations, thus knowing whether the user didn't touch the values
makes a difference.

We'll track user setting for these properties using a hash, like we do
in the TCG driver. All riscv cpu options will update this hash in case
the user sets it. The KVM driver will use this hash to minimize the
amount of syscalls done.

For now, both 'pmu-mask' and 'pmu-num' shouldn't be changed for vendor
CPUs. The existing setter for 'pmu-num' is changed to add this
restriction. New getters and setters are required for 'pmu-mask'

While we're at it, add a 'static' modifier to 'prop_pmu_num' since we're
not exporting it.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: make riscv_cpu_is_vendor() public

We'll use this function in target/riscv/cpu.c to implement setters that
won't allow vendor CPU options to be changed.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/cpu_cfg.h: remove unused fields

user_spec, bext_spec and bext_ver aren't being used.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Add step to validate 'B' extension

If the B extension is enabled warn if the user has disabled any of the
required extensions that are part of the 'B' extension. Conversely
enable the extensions that make up the 'B' extension if it is enabled.

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240111161644.33630-3-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Add infrastructure for 'B' MISA extension

Add the infrastructure for the 'B' extension which is the union of the
Zba, Zbb and Zbs instructions.

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240111161644.33630-2-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Check for 'A' extension on all atomic instructions

Add requirement that 'A' is enabled for all atomic instructions that
lack the check. This makes the 64-bit versions consistent with the
32-bit versions in the same file.

Signed-off-by: Rob Bradford <rbradford@rivosinc.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20240110163959.31291-1-rbradford@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

Merge tag 'pull-tcg-20240205-2' of https://gitlab.com/rth7680/qemu into staging

tcg: Introduce TCG_COND_TST{EQ,NE}
target/alpha: Use TCG_COND_TST{EQ,NE}
target/m68k: Use TCG_COND_TST{EQ,NE} in gen_fcc_cond
target/sparc: Use TCG_COND_TSTEQ in gen_op_mulscc
target/s390x: Use TCG_COND_TSTNE for CC_OP_{TM,ICM}
target/s390x: Improve general case of disas_jcc

# -----BEGIN PGP SIGNATURE-----
#
# iQFRBAABCgA7FiEEekgeeIaLTbaoWgXAZN846K9+IV8FAmXBpTAdHHJpY2hhcmQu
# aGVuZGVyc29uQGxpbmFyby5vcmcACgkQZN846K9+IV/p6gf9HAasTSRECk2cvjW9
# /mcJy0AIaespnI50fG8fm48OoFl0847CdrsJycpZ1spw3W3Wb0cVbMbq/teNMjXZ
# 0SGQJFk9Baq7wMhW7VzhSzJ96pcorpQprp7XBMdheLXqpT4zsM/EuwEAepBk8RUG
# 3kCeo38dswXE681ZafZkd/8pPzII19sQK8eiMpceeYkBsbbep+DDcnE18Ee4kISS
# u0SbuslKVahxd86LKuzrcz0pNFcmFuR5jRP9hmbQ0MfeAn0Pxlndi+ayZNghfgPf
# 3hDjskiionFwxb/OoRj45BssTWfDiluWl7IUsHfegPXCQ2Y+woT5Vq6TVGZn0GqS
# c6RLQQ==
# =TMiE
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 06 Feb 2024 03:19:12 GMT
# 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

* tag 'pull-tcg-20240205-2' of https://gitlab.com/rth7680/qemu: (39 commits)
  tcg/tci: Support TCG_COND_TST{EQ,NE}
  tcg/s390x: Support TCG_COND_TST{EQ,NE}
  tcg/s390x: Add TCG_CT_CONST_CMP
  tcg/s390x: Split constraint A into J+U
  tcg/ppc: Support TCG_COND_TST{EQ,NE}
  tcg/ppc: Add TCG_CT_CONST_CMP
  tcg/ppc: Tidy up tcg_target_const_match
  tcg/ppc: Use cr0 in tcg_to_bc and tcg_to_isel
  tcg/ppc: Sink tcg_to_bc usage into tcg_out_bc
  tcg/sparc64: Support TCG_COND_TST{EQ,NE}
  tcg/sparc64: Pass TCGCond to tcg_out_cmp
  tcg/sparc64: Hoist read of tcg_cond_to_rcond
  tcg/i386: Use TEST r,r to test 8/16/32 bits
  tcg/i386: Improve TSTNE/TESTEQ vs powers of two
  tcg/i386: Support TCG_COND_TST{EQ,NE}
  tcg/i386: Move tcg_cond_to_jcc[] into tcg_out_cmp
  tcg/i386: Pass x86 condition codes to tcg_out_cmov
  tcg/arm: Support TCG_COND_TST{EQ,NE}
  tcg/arm: Split out tcg_out_cmp()
  tcg/aarch64: Generate CBNZ for TSTNE of UINT32_MAX
  ...

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

Merge tag 'pull-request-2024-02-06' of https://gitlab.com/thuth/qemu into staging

* Emulate CVB, CVBY, CVBG and CVDG s390x instructions
* Fix bug in lsi53c895a reentrancy counter
* Deprecate the "power5+" and "power7+" CPU names
* Fix problems in the freebsd VM test

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEJ7iIR+7gJQEY8+q5LtnXdP5wLbUFAmXCCXURHHRodXRoQHJl
# ZGhhdC5jb20ACgkQLtnXdP5wLbXtEA/9HKWMHbWqDAdlrpmfW8lCFaBHgV0+Fqsy
# GlxJykni2BxIWNoR7J6SdAqbgx3E2/7i8IMIUwYXlNBjEs/UQ0ZcnI5k6OfUS24p
# qfbdH717SgsaB9R1vCBhmOGGWYBfe/RqPGIcni/eg+jSxB5cn2XvEv3+ZBckvDsh
# KFuuAa6vvuBVhyXLbkP8Z+LEe27ttIYi5v1dvJ1an4UbFESqxVb0knyuFYpZpY8Y
# h7dZ0hyCid7YT03zVmSADK7anO+epBdzUU3SsKXj2dB9nebSjmkav6lQQBKYHHUg
# THojcWKwFPNK0AojhBuBCqFYgkGGt/9kjwlUt7jfm1TcSemN65XLNYHThRekPuAJ
# Jcze8dcEerbj1xsNWYh4hPvB92laEiyVR5BYFfUkJ9m2IAamPQLHvOT7jzhC3Y9k
# 4wvVcf9QKVtKW0QO54SQjD4A/qQu/4777oH5w83nGuxjUthmHDqZmjDlIRe6lKJt
# gsA+mKn+w9HrtiXOSkoMhK8PAyvCoAef/N7kvHZoHmp6TtfQAjPs4/v2uZMpnd60
# z7Cw50giHpo9lmiZ1Ey2fQvw9orYhNoXAc4XfYGHuYdQFWpCGz1PB2Km8uTPTEUe
# as364ULBqWoFBCRuRndy2+z2e3zhK5THTPCAyHf48M6teMEPa4KTsTCk7MzmfVfx
# C8RsLcmrFPI=
# =eQNc
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 06 Feb 2024 10:27:01 GMT
# gpg:                using RSA key 27B88847EEE0250118F3EAB92ED9D774FE702DB5
# gpg:                issuer "thuth@redhat.com"
# gpg: Good signature from "Thomas Huth <th.huth@gmx.de>" [full]
# gpg:                 aka "Thomas Huth <thuth@redhat.com>" [full]
# gpg:                 aka "Thomas Huth <huth@tuxfamily.org>" [full]
# gpg:                 aka "Thomas Huth <th.huth@posteo.de>" [unknown]
# Primary key fingerprint: 27B8 8847 EEE0 2501 18F3  EAB9 2ED9 D774 FE70 2DB5

* tag 'pull-request-2024-02-06' of https://gitlab.com/thuth/qemu:
  meson: Link with libinotify on FreeBSD
  test-util-filemonitor: Adapt to the FreeBSD inotify rename semantics
  tests/vm/freebsd: Reload the sshd configuration
  tests/vm: Set UseDNS=no in the sshd configuration
  target/s390x: Prefer fast cpu_env() over slower CPU QOM cast macro
  tests/tcg/s390x: Test CONVERT TO BINARY
  tests/tcg/s390x: Test CONVERT TO DECIMAL
  target/s390x: Emulate CVB, CVBY and CVBG
  target/s390x: Emulate CVDG
  docs/about: Deprecate the old "power5+" and "power7+" CPU names
  target/ppc/cpu-models: Rename power5+ and power7+ for new QOM naming rules
  hw/scsi/lsi53c895a: add missing decrement of reentrancy counter

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

Merge tag 'mem-2024-02-06-v3' of https://github.com/davidhildenbrand/qemu into staging

Hi,

"Host Memory Backends" and "Memory devices" queue ("mem"):
- Reintroduce memory region size checks for memory devices; the removal
  lead to some undesired side effects
- Preallocate memory of memory backends in selected configurations
  asynchronously (so we preallocate concurrently), to speed up QEMU
  startup time.

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEG9nKrXNcTDpGDfzKTd4Q9wD/g1oFAmXB3LcRHGRhdmlkQHJl
# ZGhhdC5jb20ACgkQTd4Q9wD/g1plRA/+N8y4aJB+qEwacl5scIpiWShqeBA0aybS
# Rp3796djgjkqozkv7AFGHrOIGiLDtCh4W1JYuML7kLN7IvuJoHSY+AHzfhDiae1l
# eluX/Cs/5rgEninwT9M0yEkgvUybA8+kx+z96hBJgkfJOrdbETc7YVbU5iP/sOOF
# UtfEVWGwT1RJOun0qrgEhHiJCTMcHyJjSEy8D867ymC+knu3OZIz22+axcmpHz6i
# QJFgY40OCP1yxBvPVLR3K/Z0se/FkxG55LwM58j7N/m+VDv4IqZCTbkZb5BTJVla
# 5vKgIrZfZ+XFqrenyMsBnBLgQuyCmDJIDFfxM0A9gOvJbwtf8T4DhL9FoRvVZMDD
# SHBl/EZcViXFDDKVHjotBSA5JoNbjHac5J5jCFu7pRq+2DbzxWHmW6xV7sY9gkSO
# +SdW9hcmF/vF5MKHfoQR2kVLLJ2/EKHiN/xVVsha0+RQDctucrhg1Y9MS2obJV3u
# u2udaVk5UNcfNPuVPwkG8YQ0sIyuDYXOTThwNtsj0tyZ+tGVQmMIlou/GAsrc9PF
# xmqzkCXXyrILrPMQJrYBcdwasBLuEcJMW59BqgxHCVP9NiAQgsNVzXFg4mr3+mVF
# xTrt8wioTvAPoDvXe+BPoaH6AsIY2TqE8j7IqA1Q/IFNf+KLYkPcHknZfzfxSkdW
# woRHVtjrkMo=
# =lW5h
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 06 Feb 2024 07:16:07 GMT
# gpg:                using RSA key 1BD9CAAD735C4C3A460DFCCA4DDE10F700FF835A
# gpg:                issuer "david@redhat.com"
# gpg: Good signature from "David Hildenbrand <david@redhat.com>" [marginal]
# gpg:                 aka "David Hildenbrand <davidhildenbrand@gmail.com>" [full]
# gpg:                 aka "David Hildenbrand <hildenbr@in.tum.de>" [unknown]
# Primary key fingerprint: 1BD9 CAAD 735C 4C3A 460D  FCCA 4DDE 10F7 00FF 835A

* tag 'mem-2024-02-06-v3' of https://github.com/davidhildenbrand/qemu:
  oslib-posix: initialize backend memory objects in parallel
  memory-device: reintroduce memory region size check
  hv-balloon: use get_min_alignment() to express 32 GiB alignment

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

meson: Link with libinotify on FreeBSD

make vm-build-freebsd fails with:

    ld: error: undefined symbol: inotify_init1
    >>> referenced by filemonitor-inotify.c:183 (../src/util/filemonitor-inotify.c:183)
    >>>               util_filemonitor-inotify.c.o:(qemu_file_monitor_new) in archive libqemuutil.a

On FreeBSD the inotify functions are defined in libinotify.so. Add it
to the dependencies.

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20240206002344.12372-5-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

test-util-filemonitor: Adapt to the FreeBSD inotify rename semantics

Unlike on Linux, on FreeBSD renaming a file when the destination
already exists results in an IN_DELETE event for that existing file:

    $ FILEMONITOR_DEBUG=1 build/tests/unit/test-util-filemonitor
    Rename /tmp/test-util-filemonitor-K13LI2/fish/one.txt -> /tmp/test-util-filemonitor-K13LI2/two.txt
    Event id=200000000 event=2 file=one.txt
    Queue event id 200000000 event 2 file one.txt
    Queue event id 100000000 event 2 file two.txt
    Queue event id 100000002 event 2 file two.txt
    Queue event id 100000000 event 0 file two.txt
    Queue event id 100000002 event 0 file two.txt
    Event id=100000000 event=0 file=two.txt
    Expected event 0 but got 2

This difference in behavior is not expected to break the real users, so
teach the test to accept it.

Suggested-by: "Daniel P. Berrange" <berrange@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-4-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/vm/freebsd: Reload the sshd configuration

After console_sshd_config(), the SSH server needs to be nudged to pick
up the new configs. The scripts for the other BSD flavors already do
this with a reboot, but a simple reload is sufficient.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-3-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/vm: Set UseDNS=no in the sshd configuration

make vm-build-freebsd sometimes fails with "Connection timed out during
banner exchange". The client strace shows:

    13:59:30 write(3, "SSH-2.0-OpenSSH_9.3\r\n", 21) = 21
    13:59:30 getpid()                       = 252655
    13:59:30 poll([{fd=3, events=POLLIN}], 1, 5000) = 1 ([{fd=3, revents=POLLIN}])
    13:59:32 read(3, "S", 1)                = 1
    13:59:32 poll([{fd=3, events=POLLIN}], 1, 3625) = 1 ([{fd=3, revents=POLLIN}])
    13:59:32 read(3, "S", 1)                = 1
    13:59:32 poll([{fd=3, events=POLLIN}], 1, 3625) = 1 ([{fd=3, revents=POLLIN}])
    13:59:32 read(3, "H", 1)                = 1

There is a 2s delay during connection, and ConnectTimeout is set to 1.
Raising it makes the issue go away, but we can do better. The server
truss shows:

    888: 27.811414714 socket(PF_INET,SOCK_DGRAM|SOCK_CLOEXEC,0) = 5 (0x5)
    888: 27.811765030 connect(5,{ AF_INET 10.0.2.3:53 },16) = 0 (0x0)
    888: 27.812166941 sendto(5,"\^Z/\^A\0\0\^A\0\0\0\0\0\0\^A2"...,39,0,NULL,0) = 39 (0x27)
    888: 29.363970743 poll({ 5/POLLRDNORM },1,5000) = 1 (0x1)

So the delay is due to a DNS query. Disable DNS queries in the server
config.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-2-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

target/s390x: Prefer fast cpu_env() over slower CPU QOM cast macro

Mechanical patch produced running the command documented
in scripts/coccinelle/cpu_env.cocci_template header.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20240129164514.73104-25-philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/tcg/s390x: Test CONVERT TO BINARY

Check the CVB's, CVBY's, and CVBG's corner cases.

Co-developed-by: Pavel Zbitskiy <pavel.zbitskiy@gmail.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240205205830.6425-5-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/tcg/s390x: Test CONVERT TO DECIMAL

Check the CVD's, CVDY's, and CVDG's corner cases.

Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240205205830.6425-4-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

target/s390x: Emulate CVB, CVBY and CVBG

Convert to Binary - counterparts of the already implemented Convert
to Decimal (CVD*) instructions.
Example from the Principles of Operation: 25594C becomes 63FA.

Co-developed-by: Pavel Zbitskiy <pavel.zbitskiy@gmail.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20240205205830.6425-3-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

target/s390x: Emulate CVDG

CVDG is the same as CVD, except that it converts 64 bits into 128,
rather than 32 into 64. Create a new helper, which uses Int128
wrappers.

Reported-by: Ido Plat <Ido.Plat@ibm.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240205205830.6425-2-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

oslib-posix: initialize backend memory objects in parallel

QEMU initializes preallocated backend memory as the objects are parsed from
the command line. This is not optimal in some cases (e.g. memory spanning
multiple NUMA nodes) because the memory objects are initialized in series.

Allow the initialization to occur in parallel (asynchronously). In order to
ensure optimal thread placement, asynchronous initialization requires prealloc
context threads to be in use.

Signed-off-by: Mark Kanda <mark.kanda@oracle.com>
Message-ID: <20240131165327.3154970-2-mark.kanda@oracle.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>

memory-device: reintroduce memory region size check

We used to check that the memory region size is multiples of the overall
requested address alignment for the device memory address.

We removed that check, because there are cases (i.e., hv-balloon) where
devices unconditionally request an address alignment that has a very large
alignment (i.e., 32 GiB), but the actual memory device size might not be
multiples of that alignment.

However, this change:

(a) allows for some practically impossible DIMM sizes, like "1GB+1 byte".
(b) allows for DIMMs that partially cover hugetlb pages, previously
    reported in [1].

Both scenarios don't make any sense: we might even waste memory.

So let's reintroduce that check, but only check that the
memory region size is multiples of the memory region alignment (i.e.,
page size, huge page size), but not any additional memory device
requirements communicated using md->get_min_alignment().

The following examples now fail again as expected:

(a) 1M with 2M THP
 qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
                     -object memory-backend-ram,id=mem1,size=1M \
                     -device pc-dimm,id=dimm1,memdev=mem1
 -> backend memory size must be multiple of 0x200000

(b) 1G+1byte

 qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
                   -object memory-backend-ram,id=mem1,size=1073741825B \
                   -device pc-dimm,id=dimm1,memdev=mem1
 -> backend memory size must be multiple of 0x200000

(c) Unliagned hugetlb size (2M)

 qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
                   -object memory-backend-file,id=mem1,mem-path=/dev/hugepages/tmp,size=511M \
                   -device pc-dimm,id=dimm1,memdev=mem1
 backend memory size must be multiple of 0x200000

(d) Unliagned hugetlb size (1G)

 qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
                    -object memory-backend-file,id=mem1,mem-path=/dev/hugepages1G/tmp,size=2047M \
                    -device pc-dimm,id=dimm1,memdev=mem1
 -> backend memory size must be multiple of 0x40000000

Note that this fix depends on a hv-balloon change to communicate its
additional alignment requirements using get_min_alignment() instead of
through the memory region.

[1] https://lkml.kernel.org/r/f77d641d500324525ac036fe1827b3070de75fc1.1701088320.git.mprivozn@redhat.com

Message-ID: <20240117135554.787344-3-david@redhat.com>
Reported-by: Zhenyu Zhang <zhenyzha@redhat.com>
Reported-by: Michal Privoznik <mprivozn@redhat.com>
Fixes: eb1b7c4bd413 ("memory-device: Drop size alignment check")
Tested-by: Zhenyu Zhang <zhenyzha@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Signed-off-by: David Hildenbrand <david@redhat.com>

tcg/tci: Support TCG_COND_TST{EQ,NE}

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/s390x: Support TCG_COND_TST{EQ,NE}

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

docs/about: Deprecate the old "power5+" and "power7+" CPU names

For consistency we should drop the names with a "+" in it in the
long run.

Message-ID: <20240117141054.73841-3-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

target/ppc/cpu-models: Rename power5+ and power7+ for new QOM naming rules

The character "+" is now forbidden in QOM device names (see commit
b447378e1217 - "Limit type names to alphanumerical and some few special
characters"). For the "power5+" and "power7+" CPU names, there is
currently a hack in type_name_is_valid() to still allow them for
compatibility reasons. However, there is a much nicer solution for this:
Simply use aliases! This way we can still support the old names without
the need for the ugly hack in type_name_is_valid().

Message-ID: <20240117141054.73841-2-thuth@redhat.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw/scsi/lsi53c895a: add missing decrement of reentrancy counter

When the maximum count of SCRIPTS instructions is reached, the code
stops execution and returns, but fails to decrement the reentrancy
counter. This effectively renders the SCSI controller unusable
because on next entry the reentrancy counter is still above the limit.

This bug was seen on HP-UX 10.20 which seems to trigger SCRIPTS
loops.

Fixes: b987718bbb ("hw/scsi/lsi53c895a: Fix reentrancy issues in the LSI controller (CVE-2023-0330)")
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Message-ID: <20240128202214.2644768-1-svens@stackframe.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Helge Deller <deller@gmx.de>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hv-balloon: use get_min_alignment() to express 32 GiB alignment

Let's implement the get_min_alignment() callback for memory devices, and
copy for the device memory region the alignment of the host memory
region. This mimics what virtio-mem does, and allows for re-introducing
proper alignment checks for the memory region size (where we don't care
about additional device requirements) in memory device core.

Message-ID: <20240117135554.787344-2-david@redhat.com>
Reviewed-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Signed-off-by: David Hildenbrand <david@redhat.com>

tcg/s390x: Add TCG_CT_CONST_CMP

Better constraint for tcg_out_cmp, based on the comparison.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/s390x: Split constraint A into J+U

Signed 33-bit == signed 32-bit + unsigned 32-bit.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/ppc: Support TCG_COND_TST{EQ,NE}

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/ppc: Add TCG_CT_CONST_CMP

Better constraint for tcg_out_cmp, based on the comparison.
We can't yet remove the fallback to load constants into a
scratch because of tcg_out_cmp2, but that path should not
be as frequent.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/ppc: Tidy up tcg_target_const_match

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/ppc: Use cr0 in tcg_to_bc and tcg_to_isel

Using cr0 means we could choose to use rc=1 to compute the condition.
Adjust the tables and tcg_out_cmp that feeds them.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/ppc: Sink tcg_to_bc usage into tcg_out_bc

Rename the current tcg_out_bc function to tcg_out_bc_lab, and
create a new function that takes an integer displacement + link.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/sparc64: Support TCG_COND_TST{EQ,NE}

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/sparc64: Pass TCGCond to tcg_out_cmp

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/sparc64: Hoist read of tcg_cond_to_rcond

Use a non-zero value here (an illegal encoding) as a better
condition than is_unsigned_cond for when MOVR/BPR is usable.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: Use TEST r,r to test 8/16/32 bits

Just like when testing against the sign bits, TEST r,r can be used when the
immediate is 0xff, 0xff00, 0xffff, 0xffffffff.

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

tcg/i386: Improve TSTNE/TESTEQ vs powers of two

Use "test x,x" when the bit is one of the 4 sign bits.
Use "bt imm,x" otherwise.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: Support TCG_COND_TST{EQ,NE}

Merge tcg_out_testi into tcg_out_cmp and adjust the two uses.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: Move tcg_cond_to_jcc[] into tcg_out_cmp

Return the x86 condition codes to use after the compare.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/i386: Pass x86 condition codes to tcg_out_cmov

Hoist the tcg_cond_to_jcc index outside the function.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/arm: Support TCG_COND_TST{EQ,NE}

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231028194522.245170-12-richard.henderson@linaro.org>
[PMD: Split from bigger patch, part 2/2]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20231108145244.72421-2-philmd@linaro.org>

tcg/arm: Split out tcg_out_cmp()

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20231028194522.245170-12-richard.henderson@linaro.org>
[PMD: Split from bigger patch, part 1/2]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20231108145244.72421-1-philmd@linaro.org>

tcg/aarch64: Generate CBNZ for TSTNE of UINT32_MAX

... and the inverse, CBZ for TSTEQ.

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

tcg/aarch64: Generate TBZ, TBNZ

Test the sign bit for LT/GE vs 0, and TSTNE/EQ vs a power of 2.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20240119224737.48943-2-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

tcg/aarch64: Massage tcg_out_brcond()

In order to ease next commit review, modify tcg_out_brcond()
to switch over TCGCond. No logical change intended.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20240119224737.48943-1-philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>