target/xtensa: Explicitly set 2-NaN propagation rule

Set the NaN propagation rule explicitly in xtensa_use_first_nan().

(When we convert the softfloat pickNaNMulAdd routine to also
select a NaN pro

target/xtensa: Explicitly set 2-NaN propagation rule

Set the NaN propagation rule explicitly in xtensa_use_first_nan().

(When we convert the softfloat pickNaNMulAdd routine to also
select a NaN propagation rule at runtime, we will be able to
remove the use_first_nan flag because the propagation rules
will handle everything.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241025141254.2141506-15-peter.maydell@linaro.org

show more ...

target/xtensa: Factor out calls to set_use_first_nan()

In xtensa we currently call set_use_first_nan() in a lot of
places where we want to switch the NaN-propagation handling.
We're about to change

target/xtensa: Factor out calls to set_use_first_nan()

In xtensa we currently call set_use_first_nan() in a lot of
places where we want to switch the NaN-propagation handling.
We're about to change the softfloat API we use to do that,
so start by factoring all the calls out into a single
xtensa_use_first_nan() function.

The bulk of this change was done with
sed -i -e 's/set_use_first_nan(\([^,]*\),[^)]*)/xtensa_use_first_nan(env, \1)/' target/xtensa/fpu_helper.c

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241025141254.2141506-14-peter.maydell@linaro.org

show more ...

include/exec/memop: Rename get_alignment_bits

Rename to use "memop_" prefix, like other functions
that operate on MemOp.

Reviewed-by: Helge Deller <deller@gmx.de>
Reviewed-by: Philippe Mathieu-Daud

include/exec/memop: Rename get_alignment_bits

Rename to use "memop_" prefix, like other functions
that operate on MemOp.

Reviewed-by: Helge Deller <deller@gmx.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

show more ...

target/xtensa: Correct assert condition in handle_interrupt()

In commit ad18376b90c8101 we added an assert that the level value was
in-bounds for the array we're about to index into. However, the
a

target/xtensa: Correct assert condition in handle_interrupt()

In commit ad18376b90c8101 we added an assert that the level value was
in-bounds for the array we're about to index into. However, the
assert condition is wrong -- env->config->interrupt_vector is an
array of uint32_t, so we should bounds check the index against
ARRAY_SIZE(...), not against sizeof().

Resolves: Coverity CID 1507131
Fixes: ad18376b90c8101 ("target/xtensa: Assert that interrupt level is within bounds")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-id: 20240731172246.3682311-1-peter.maydell@linaro.org

show more ...

target/xtensa: Make use of 'segment' in pptlb helper less confusing

Coverity gets confused about the use of the 'segment' variable in the
pptlb helper function: it thinks that we can take a code pat

target/xtensa: Make use of 'segment' in pptlb helper less confusing

Coverity gets confused about the use of the 'segment' variable in the
pptlb helper function: it thinks that we can take a code path where
we first initialize it:
unsigned segment = XTENSA_MPU_PROBE_B; // 0x40000000
and then use that value as a shift count:
} else if (nhits == 1 && (env->sregs[MPUENB] & (1u << segment))) {

In fact this isn't possible, beacuse xtensa_mpu_lookup() is passed
'&segment', and it uses that as an output value, which it will always
set if it returns nonzero. But the way the code is currently written
is confusing to a human reader as well as to Coverity.

Instead of initializing 'segment' at the top of the function with a
value that's only used in the "nhits == 0" code path, use the
constant value directly in that code path, and don't initialize
segment. This matches the way we use xtensa_mpu_lookup() in its
other callsites in get_physical_addr_mpu().

Resolves: Coverity CID 1547589

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Message-id: 20240723151454.1396826-1-peter.maydell@linaro.org

show more ...

target/xtensa: Restrict semihosting to TCG

The semihosting feature depends on TCG (due to the probe_access
API access). Although TCG is the single accelerator currently
available for the xtensa targ

target/xtensa: Restrict semihosting to TCG

The semihosting feature depends on TCG (due to the probe_access
API access). Although TCG is the single accelerator currently
available for the xtensa target, use the Kconfig "imply" directive
which is more correct (if we were to support a different accel).

Reported-by: Anton Johansson <anjo@rev.ng>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20240717105723.58965-8-philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20240718094523.1198645-15-alex.bennee@linaro.org>

show more ...

target: Set TCGCPUOps::cpu_exec_halt to target's has_work implementation

Currently the TCGCPUOps::cpu_exec_halt method is optional, and if it
is not set then the default is to call the CPUClass::has

target: Set TCGCPUOps::cpu_exec_halt to target's has_work implementation

Currently the TCGCPUOps::cpu_exec_halt method is optional, and if it
is not set then the default is to call the CPUClass::has_work
method (which has an identical function signature).

We would like to make the cpu_exec_halt method mandatory so we can
remove the runtime check and fallback handling. In preparation for
that, make all the targets which don't need special handling in their
cpu_exec_halt set it to their cpu_has_work implementation instead of
leaving it unset. (This is every target except for arm and i386.)

In the riscv case this requires us to make the function not
be local to the source file it's defined in.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

show more ...