qemu.git
4 months agotarget/alpha: Set default NaN pattern explicitly
Peter Maydell [Wed, 11 Dec 2024 15:31:05 +0000 (15:31 +0000)]
target/alpha: Set default NaN pattern explicitly

Set the default NaN pattern explicitly for the alpha target.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-40-peter.maydell@linaro.org

4 months agotarget/hppa: Set default NaN pattern explicitly
Peter Maydell [Wed, 11 Dec 2024 15:31:05 +0000 (15:31 +0000)]
target/hppa: Set default NaN pattern explicitly

Set the default NaN pattern explicitly, and remove the ifdef from
parts64_default_nan().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-39-peter.maydell@linaro.org

4 months agotarget/i386: Set default NaN pattern explicitly
Peter Maydell [Wed, 11 Dec 2024 15:31:04 +0000 (15:31 +0000)]
target/i386: Set default NaN pattern explicitly

Set the default NaN pattern explicitly, and remove the ifdef from
parts64_default_nan().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-38-peter.maydell@linaro.org

4 months agotarget/microblaze: Set default NaN pattern explicitly
Peter Maydell [Wed, 11 Dec 2024 15:31:04 +0000 (15:31 +0000)]
target/microblaze: Set default NaN pattern explicitly

Set the default NaN pattern explicitly, and remove the ifdef from
parts64_default_nan().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-37-peter.maydell@linaro.org

4 months agotests/fp: Set default NaN pattern explicitly
Peter Maydell [Wed, 11 Dec 2024 15:31:04 +0000 (15:31 +0000)]
tests/fp: Set default NaN pattern explicitly

Set the default NaN pattern explicitly for the tests/fp code.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-36-peter.maydell@linaro.org

4 months agofpu: Allow runtime choice of default NaN value
Peter Maydell [Wed, 11 Dec 2024 15:31:04 +0000 (15:31 +0000)]
fpu: Allow runtime choice of default NaN value

Currently we hardcode the default NaN value in parts64_default_nan()
using a compile-time ifdef ladder. This is awkward for two cases:
 * for single-QEMU-binary we can't hard-code target-specifics like this
 * for Arm FEAT_AFP the default NaN value depends on FPCR.AH
   (specifically the sign bit is different)

Add a field to float_status to specify the default NaN value; fall
back to the old ifdef behaviour if these are not set.

The default NaN value is specified by setting a uint8_t to a
pattern corresponding to the sign and upper fraction parts of
the NaN; the lower bits of the fraction are set from bit 0 of
the pattern.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-35-peter.maydell@linaro.org

4 months agotarget/arm: Copy entire float_status in is_ebf
Richard Henderson [Wed, 11 Dec 2024 15:31:03 +0000 (15:31 +0000)]
target/arm: Copy entire float_status in is_ebf

Now that float_status has a bunch of fp parameters,
it is easier to copy an existing structure than create
one from scratch.  Begin by copying the structure that
corresponds to the FPSR and make only the adjustments
required for BFloat16 semantics.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20241203203949.483774-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agotarget/ppc: Use env->fp_status in helper_compute_fprf functions
Peter Maydell [Wed, 11 Dec 2024 15:31:03 +0000 (15:31 +0000)]
target/ppc: Use env->fp_status in helper_compute_fprf functions

In the helper_compute_fprf functions, we pass a dummy float_status
in to the is_signaling_nan() function. This is unnecessary, because
we have convenient access to the CPU env pointer here and that
is already set up with the correct values for the snan_bit_is_one
and no_signaling_nans config settings. is_signaling_nan() doesn't
ever update the fp_status with any exception flags, so there is
no reason not to use env->fp_status here.

Use env->fp_status instead of the dummy fp_status.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-34-peter.maydell@linaro.org

4 months agotarget/sparc: Initialize local scratch float_status from env->fp_status
Peter Maydell [Wed, 11 Dec 2024 15:31:03 +0000 (15:31 +0000)]
target/sparc: Initialize local scratch float_status from env->fp_status

In the helper functions flcmps and flcmpd we use a scratch float_status
so that we don't change the CPU state if the comparison raises any
floating point exception flags. Instead of zero-initializing this
scratch float_status, initialize it as a copy of env->fp_status. This
avoids the need to explicitly initialize settings like the NaN
propagation rule or others we might add to softfloat in future.

To do this we need to pass the CPU env pointer in to the helper.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-33-peter.maydell@linaro.org

4 months agotarget/m68k: Init local float_status from env fp_status in gdb get/set reg
Peter Maydell [Wed, 11 Dec 2024 15:31:03 +0000 (15:31 +0000)]
target/m68k: Init local float_status from env fp_status in gdb get/set reg

In cf_fpu_gdb_get_reg() and cf_fpu_gdb_set_reg() we do the conversion
from float64 to floatx80 using a scratch float_status, because we
don't want the conversion to affect the CPU's floating point exception
status. Currently we use a zero-initialized float_status. This will
get steadily more awkward as we add config knobs to float_status
that the target must initialize. Avoid having to add any of that
configuration here by instead initializing our local float_status
from the env->fp_status.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-32-peter.maydell@linaro.org

4 months agotarget/m68k: In frem helper, initialize local float_status from env->fp_status
Peter Maydell [Wed, 11 Dec 2024 15:31:02 +0000 (15:31 +0000)]
target/m68k: In frem helper, initialize local float_status from env->fp_status

In the frem helper, we have a local float_status because we want to
execute the floatx80_div() with a custom rounding mode.  Instead of
zero-initializing the local float_status and then having to set it up
with the m68k standard behaviour (including the NaN propagation rule
and copying the rounding precision from env->fp_status), initialize
it as a complete copy of env->fp_status. This will avoid our having
to add new code in this function for every new config knob we add
to fp_status.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-31-peter.maydell@linaro.org

4 months agotarget/loongarch: Use normal float_status in fclass_s and fclass_d helpers
Peter Maydell [Wed, 11 Dec 2024 15:31:01 +0000 (15:31 +0000)]
target/loongarch: Use normal float_status in fclass_s and fclass_d helpers

In target/loongarch's helper_fclass_s() and helper_fclass_d() we pass
a zero-initialized float_status struct to float32_is_quiet_nan() and
float64_is_quiet_nan(), with the cryptic comment "for
snan_bit_is_one".

This pattern appears to have been copied from target/riscv, where it
is used because the functions there do not have ready access to the
CPU state struct. The comment presumably refers to the fact that the
main reason the is_quiet_nan() functions want the float_state is
because they want to know about the snan_bit_is_one config.

In the loongarch helpers, though, we have the CPU state struct
to hand. Use the usual env->fp_status here. This avoids our needing
to track that we need to update the initializer of the local
float_status structs when the core softfloat code adds new
options for targets to configure their behaviour.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-30-peter.maydell@linaro.org

4 months agosoftfloat: Create floatx80 default NaN from parts64_default_nan
Peter Maydell [Wed, 11 Dec 2024 15:31:00 +0000 (15:31 +0000)]
softfloat: Create floatx80 default NaN from parts64_default_nan

We create our 128-bit default NaN by calling parts64_default_nan()
and then adjusting the result.  We can do the same trick for creating
the floatx80 default NaN, which lets us drop a target ifdef.

floatx80 is used only by:
 i386
 m68k
 arm nwfpe old floating-point emulation emulation support
    (which is essentially dead, especially the parts involving floatx80)
 PPC (only in the xsrqpxp instruction, which just rounds an input
    value by converting to floatx80 and back, so will never generate
    the default NaN)

The floatx80 default NaN as currently implemented is:
 m68k: sign = 0, exp = 1...1, int = 1, frac = 1....1
 i386: sign = 1, exp = 1...1, int = 1, frac = 10...0

These are the same as the parts64_default_nan for these architectures.

This is technically a possible behaviour change for arm linux-user
nwfpe emulation emulation, because the default NaN will now have the
sign bit clear.  But we were already generating a different floatx80
default NaN from the real kernel emulation we are supposedly
following, which appears to use an all-bits-1 value:
 https://elixir.bootlin.com/linux/v6.12/source/arch/arm/nwfpe/softfloat-specialize#L267

This won't affect the only "real" use of the nwfpe emulation, which
is ancient binaries that used it as part of the old floating point
calling convention; that only uses loads and stores of 32 and 64 bit
floats, not any of the floatx80 behaviour the original hardware had.
We also get the nwfpe float64 default NaN value wrong:
 https://elixir.bootlin.com/linux/v6.12/source/arch/arm/nwfpe/softfloat-specialize#L166
so if we ever cared about this obscure corner the right fix would be
to correct that so nwfpe used its own default-NaN setting rather
than the Arm VFP one.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-29-peter.maydell@linaro.org

4 months agotarget/m68k: Don't pass NULL float_status to floatx80_default_nan()
Peter Maydell [Wed, 11 Dec 2024 15:30:59 +0000 (15:30 +0000)]
target/m68k: Don't pass NULL float_status to floatx80_default_nan()

Currently m68k_cpu_reset_hold() calls floatx80_default_nan(NULL)
to get the NaN bit pattern to reset the FPU registers. This
works because it happens that our implementation of
floatx80_default_nan() doesn't actually look at the float_status
pointer except for TARGET_MIPS. However, this isn't guaranteed,
and to be able to remove the ifdef in floatx80_default_nan()
we're going to need a real float_status here.

Rearrange m68k_cpu_reset_hold() so that we initialize env->fp_status
earlier, and thus can pass it to floatx80_default_nan().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-28-peter.maydell@linaro.org

4 months agofpu: Remove use_first_nan field from float_status
Peter Maydell [Wed, 11 Dec 2024 15:30:59 +0000 (15:30 +0000)]
fpu: Remove use_first_nan field from float_status

The use_first_nan field in float_status was an xtensa-specific way to
select at runtime from two different NaN propagation rules.  Now that
xtensa is using the target-agnostic NaN propagation rule selection
that we've just added, we can remove use_first_nan, because there is
no longer any code that reads it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-27-peter.maydell@linaro.org

4 months agotarget/hppa: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:59 +0000 (15:30 +0000)]
target/hppa: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for HPPA, and remove the
ifdef from pickNaNMulAdd().

HPPA is the only target that was using the default branch of the
ifdef ladder (other targets either do not use muladd or set
default_nan_mode), so we can remove the ifdef fallback entirely now
(allowing the "rule not set" case to fall into the default of the
switch statement and assert).

We add a TODO note that the HPPA rule is probably wrong; this is
not a behavioural change for this refactoring.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-26-peter.maydell@linaro.org

4 months agotarget/i386: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:58 +0000 (15:30 +0000)]
target/i386: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for i386.  We had no
i386-specific behaviour in the old ifdef ladder, so we were using the
default "prefer a then b then c" fallback; this is actually the
correct per-the-spec handling for i386.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-25-peter.maydell@linaro.org

4 months agotarget/xtensa: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:58 +0000 (15:30 +0000)]
target/xtensa: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for xtensa, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-24-peter.maydell@linaro.org

4 months agotarget/mips: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:58 +0000 (15:30 +0000)]
target/mips: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for Arm, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-23-peter.maydell@linaro.org

4 months agotarget/sparc: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:58 +0000 (15:30 +0000)]
target/sparc: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for SPARC, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-22-peter.maydell@linaro.org

4 months agotarget/s390x: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
target/s390x: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for s390x, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-21-peter.maydell@linaro.org

4 months agotarget/ppc: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
target/ppc: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for PPC, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-20-peter.maydell@linaro.org

4 months agotarget/loongarch: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
target/loongarch: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for loongarch, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-19-peter.maydell@linaro.org

4 months agotarget/arm: Set Float3NaNPropRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
target/arm: Set Float3NaNPropRule explicitly

Set the Float3NaNPropRule explicitly for Arm, and remove the
ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-18-peter.maydell@linaro.org

4 months agotests/fp: Explicitly set 3-NaN propagation rule
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
tests/fp: Explicitly set 3-NaN propagation rule

Explicitly set a rule in the softfloat tests for propagating NaNs in
the muladd case.  In meson.build we put -DTARGET_ARM in fpcflags, and
so we should select here the Arm rule of float_3nan_prop_s_cab.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-17-peter.maydell@linaro.org

4 months agosoftfloat: Allow runtime choice of NaN propagation for muladd
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
softfloat: Allow runtime choice of NaN propagation for muladd

IEEE 758 does not define a fixed rule for which NaN to pick as the
result if both operands of a 3-operand fused multiply-add operation
are NaNs.  As a result different architectures have ended up with
different rules for propagating NaNs.

QEMU currently hardcodes the NaN propagation logic into the binary
because pickNaNMulAdd() has an ifdef ladder for different targets.
We want to make the propagation rule instead be selectable at
runtime, because:
 * this will let us have multiple targets in one QEMU binary
 * the Arm FEAT_AFP architectural feature includes letting
   the guest select a NaN propagation rule at runtime

In this commit we add an enum for the propagation rule, the field in
float_status, and the corresponding getters and setters.  We change
pickNaNMulAdd to honour this, but because all targets still leave
this field at its default 0 value, the fallback logic will pick the
rule type with the old ifdef ladder.

It's valid not to set a propagation rule if default_nan_mode is
enabled, because in that case there's no need to pick a NaN; all the
callers of pickNaNMulAdd() catch this case and skip calling it.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-16-peter.maydell@linaro.org

4 months agosoftfloat: Pass have_snan to pickNaNMulAdd
Peter Maydell [Wed, 11 Dec 2024 15:30:57 +0000 (15:30 +0000)]
softfloat: Pass have_snan to pickNaNMulAdd

The new implementation of pickNaNMulAdd() will find it convenient
to know whether at least one of the three arguments to the muladd
was a signaling NaN. We already calculate that in the caller,
so pass it in as a new bool have_snan.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-15-peter.maydell@linaro.org

4 months agotarget/hppa: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:56 +0000 (15:30 +0000)]
target/hppa: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the HPPA target,
so we can remove the ifdef from pickNaNMulAdd().

As this is the last target to be converted to explicitly setting
the rule, we can remove the fallback code in pickNaNMulAdd()
entirely.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-14-peter.maydell@linaro.org

4 months agotarget/loongarch: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:56 +0000 (15:30 +0000)]
target/loongarch: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the loongarch target.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-13-peter.maydell@linaro.org

4 months agotarget/x86: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:55 +0000 (15:30 +0000)]
target/x86: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the x86 target.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-12-peter.maydell@linaro.org

4 months agotarget/xtensa: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:55 +0000 (15:30 +0000)]
target/xtensa: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the xtensa target,
so we can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-11-peter.maydell@linaro.org

4 months agotarget/sparc: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:55 +0000 (15:30 +0000)]
target/sparc: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the SPARC target,
so we can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-10-peter.maydell@linaro.org

4 months agotarget/mips: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:54 +0000 (15:30 +0000)]
target/mips: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the MIPS target,
so we can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-9-peter.maydell@linaro.org

4 months agotarget/ppc: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:54 +0000 (15:30 +0000)]
target/ppc: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the PPC target,
so we can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-8-peter.maydell@linaro.org

4 months agotarget/s390: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:54 +0000 (15:30 +0000)]
target/s390: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for s390, so we
can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-7-peter.maydell@linaro.org

4 months agotarget/arm: Set FloatInfZeroNaNRule explicitly
Peter Maydell [Wed, 11 Dec 2024 15:30:53 +0000 (15:30 +0000)]
target/arm: Set FloatInfZeroNaNRule explicitly

Set the FloatInfZeroNaNRule explicitly for the Arm target,
so we can remove the ifdef from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-6-peter.maydell@linaro.org

4 months agotests/fp: Explicitly set inf-zero-nan rule
Peter Maydell [Wed, 11 Dec 2024 15:30:53 +0000 (15:30 +0000)]
tests/fp: Explicitly set inf-zero-nan rule

Explicitly set a rule in the softfloat tests for the inf-zero-nan
muladd special case.  In meson.build we put -DTARGET_ARM in fpcflags,
and so we should select here the Arm rule of
float_infzeronan_dnan_if_qnan.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20241202131347.498124-5-peter.maydell@linaro.org

4 months agosoftfloat: Allow runtime choice of inf * 0 + NaN result
Peter Maydell [Wed, 11 Dec 2024 15:30:53 +0000 (15:30 +0000)]
softfloat: Allow runtime choice of inf * 0 + NaN result

IEEE 758 does not define a fixed rule for what NaN to return in
the case of a fused multiply-add of inf * 0 + NaN. Different
architectures thus do different things:
 * some return the default NaN
 * some return the input NaN
 * Arm returns the default NaN if the input NaN is quiet,
   and the input NaN if it is signalling

We want to make this logic be runtime selected rather than
hardcoded into the binary, because:
 * this will let us have multiple targets in one QEMU binary
 * the Arm FEAT_AFP architectural feature includes letting
   the guest select a NaN propagation rule at runtime

In this commit we add an enum for the propagation rule, the field in
float_status, and the corresponding getters and setters.  We change
pickNaNMulAdd to honour this, but because all targets still leave
this field at its default 0 value, the fallback logic will pick the
rule type with the old ifdef ladder.

Note that four architectures both use the muladd softfloat functions
and did not have a branch of the ifdef ladder to specify their
behaviour (and so were ending up with the "default" case, probably
wrongly): i386, HPPA, SH4 and Tricore.  SH4 and Tricore both set
default_nan_mode, and so will never get into pickNaNMulAdd().  For
HPPA and i386 we retain the same behaviour as the old default-case,
which is to not ever return the default NaN.  This might not be
correct but it is not a behaviour change.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-4-peter.maydell@linaro.org

4 months agofpu: Check for default_nan_mode before calling pickNaNMulAdd
Peter Maydell [Wed, 11 Dec 2024 15:30:53 +0000 (15:30 +0000)]
fpu: Check for default_nan_mode before calling pickNaNMulAdd

If the target sets default_nan_mode then we're always going to return
the default NaN, and pickNaNMulAdd() no longer has any side effects.
For consistency with pickNaN(), check for default_nan_mode before
calling pickNaNMulAdd().

When we convert pickNaNMulAdd() to allow runtime selection of the NaN
propagation rule, this means we won't have to make the targets which
use default_nan_mode also set a propagation rule.

Since RiscV always uses default_nan_mode, this allows us to remove
its ifdef case from pickNaNMulAdd().

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-3-peter.maydell@linaro.org

4 months agofpu: handle raising Invalid for infzero in pick_nan_muladd
Peter Maydell [Wed, 11 Dec 2024 15:30:52 +0000 (15:30 +0000)]
fpu: handle raising Invalid for infzero in pick_nan_muladd

For IEEE fused multiply-add, the (0 * inf) + NaN case should raise
Invalid for the multiplication of 0 by infinity.  Currently we handle
this in the per-architecture ifdef ladder in pickNaNMulAdd().
However, since this isn't really architecture specific we can hoist
it up to the generic code.

For the cases where the infzero test in pickNaNMulAdd was
returning 2, we can delete the check entirely and allow the
code to fall into the normal pick-a-NaN handling, because this
will return 2 anyway (input 'c' being the only NaN in this case).
For the cases where infzero was returning 3 to indicate "return
the default NaN", we must retain that "return 3".

For Arm, this looks like it might be a behaviour change because we
used to set float_flag_invalid | float_flag_invalid_imz only if C is
a quiet NaN.  However, it is not, because Arm target code never looks
at float_flag_invalid_imz, and for the (0 * inf) + SNaN case we
already raised float_flag_invalid via the "abc_mask &
float_cmask_snan" check in pick_nan_muladd.

For any target architecture using the "default implementation" at the
bottom of the ifdef, this is a behaviour change but will be fixing a
bug (where we failed to raise the Invalid exception for (0 * inf +
QNaN).  The architectures using the default case are:
 * hppa
 * i386
 * sh4
 * tricore

The x86, Tricore and SH4 CPU architecture manuals are clear that this
should have raised Invalid; HPPA is a bit vaguer but still seems
clear enough.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20241202131347.498124-2-peter.maydell@linaro.org

4 months agohw/net/lan9118_phy: Add missing 100 mbps full duplex advertisement
Bernhard Beschow [Wed, 11 Dec 2024 15:30:52 +0000 (15:30 +0000)]
hw/net/lan9118_phy: Add missing 100 mbps full duplex advertisement

The real device advertises this mode and the device model already advertises
100 mbps half duplex and 10 mbps full+half duplex. So advertise this mode to
make the model more realistic.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-id: 20241102125724.532843-6-shentey@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/net/lan9118_phy: Reuse MII constants
Bernhard Beschow [Wed, 11 Dec 2024 15:30:52 +0000 (15:30 +0000)]
hw/net/lan9118_phy: Reuse MII constants

Prefer named constants over magic values for better readability.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Message-id: 20241102125724.532843-5-shentey@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/net/lan9118_phy: Fix off-by-one error in MII_ANLPAR register
Bernhard Beschow [Wed, 11 Dec 2024 15:30:52 +0000 (15:30 +0000)]
hw/net/lan9118_phy: Fix off-by-one error in MII_ANLPAR register

Turns 0x70 into 0xe0 (== 0x70 << 1) which adds the missing MII_ANLPAR_TX and
fixes the MSB of selector field to be zero, as specified in the datasheet.

Fixes: 2a424990170b "LAN9118 emulation"
Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20241102125724.532843-4-shentey@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/net/lan9118_phy: Reuse in imx_fec and consolidate implementations
Bernhard Beschow [Wed, 11 Dec 2024 15:30:51 +0000 (15:30 +0000)]
hw/net/lan9118_phy: Reuse in imx_fec and consolidate implementations

imx_fec models the same PHY as lan9118_phy. The code is almost the same with
imx_fec having more logging and tracing. Merge these improvements into
lan9118_phy and reuse in imx_fec to fix the code duplication.

Some migration state how resides in the new device model which breaks migration
compatibility for the following machines:
* imx25-pdk
* sabrelite
* mcimx7d-sabre
* mcimx6ul-evk

Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20241102125724.532843-3-shentey@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/net/lan9118: Extract lan9118_phy
Bernhard Beschow [Wed, 11 Dec 2024 15:30:51 +0000 (15:30 +0000)]
hw/net/lan9118: Extract lan9118_phy

A very similar implementation of the same device exists in imx_fec. Prepare for
a common implementation by extracting a device model into its own files.

Some migration state has been moved into the new device model which breaks
migration compatibility for the following machines:
* smdkc210
* realview-*
* vexpress-*
* kzm
* mps2-*

While breaking migration ABI, fix the size of the MII registers to be 16 bit,
as defined by IEEE 802.3u.

Signed-off-by: Bernhard Beschow <shentey@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20241102125724.532843-2-shentey@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoOpen 10.0 development tree
Peter Maydell [Tue, 10 Dec 2024 17:41:17 +0000 (17:41 +0000)]
Open 10.0 development tree

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoUpdate version for v9.2.0 release
Peter Maydell [Tue, 10 Dec 2024 16:20:54 +0000 (16:20 +0000)]
Update version for v9.2.0 release

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoUpdate version for v9.2.0-rc3 release
Peter Maydell [Tue, 3 Dec 2024 17:56:12 +0000 (17:56 +0000)]
Update version for v9.2.0-rc3 release

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu into staging
Peter Maydell [Tue, 3 Dec 2024 13:44:20 +0000 (13:44 +0000)]
Merge tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu into staging

chardev patch queue

# -----BEGIN PGP SIGNATURE-----
#
# iQJQBAABCAA6FiEEh6m9kz+HxgbSdvYt2ujhCXWWnOUFAmdPCg4cHG1hcmNhbmRy
# ZS5sdXJlYXVAcmVkaGF0LmNvbQAKCRDa6OEJdZac5dmTD/4pNcJ7fJzzxtfBDBvy
# +NWlq7nFDOX471g1B+4V91dOT0TbQLRK0hTmPUGc7FF8sCCDgfZOyHN6GA7TdwXG
# 8+oHqFjxaZcrtrSN67wv8n5tyFNm/1hr1IgqhieKjeVt6eT8PJ8rgFESm4zvDOej
# i2jjhDNN2piu9Lr9smhX3I/kJX70cPXRRX2qBlLEPcMKW280RoTRSEIL1jFdilan
# QyPugQm9R0cSayNg27At2HakFZ5HKgu8EyAvvMFvv1gdoGVdq9D+Gk5xruvBnF/P
# 4g7EnCe/YKWfg887LSjIA2Qcp1xEpUPeumUbCm152ikZ7KnwE9Tn/hPt693GjvMn
# WhyOOuxZmvL4AtJbYheskH7N0b/NB2vUT4N4ONkMeubjzfDu8UHw/rL+WpjAZfXW
# YwzcpVHKXXKazTzzKRLUgsoqfllWLTAQDwU5s46UyVmaBBf5qUSbSnr4ZX2nP0X6
# dP1jM2KcDyaKKowmpYGyQzTVqPwTXfyEdmSNnjirbS2Qyh77snd7rqrxXPz2jKm9
# jfb0yN7BWSEuFguyXL0izeCQsBb/rsrZB/tOYWkUeOzAgpm4GlAQuJ+3uAssgmnu
# aq/bYxGEjC1a+fzg8KUj6S+X+Jfp6+4/Y60M4HAMtMGuy6X1UaCBtH+cntS3JjUj
# 7ZBpTXirstW7KxKZ2mJMa2lkHQ==
# =34fm
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 03 Dec 2024 13:39:26 GMT
# gpg:                using RSA key 87A9BD933F87C606D276F62DDAE8E10975969CE5
# gpg:                issuer "marcandre.lureau@redhat.com"
# gpg: Good signature from "Marc-André Lureau <marcandre.lureau@redhat.com>" [full]
# gpg:                 aka "Marc-André Lureau <marcandre.lureau@gmail.com>" [full]
# Primary key fingerprint: 87A9 BD93 3F87 C606 D276  F62D DAE8 E109 7596 9CE5

* tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu:
  chardev: Remove __-prefixed names
  chardev: Fix record/replay error path NULL deref in device creation

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'hw-misc-20241203' of https://github.com/philmd/qemu into staging
Peter Maydell [Tue, 3 Dec 2024 13:44:05 +0000 (13:44 +0000)]
Merge tag 'hw-misc-20241203' of https://github.com/philmd/qemu into staging

Misc fixes for QEMU v9.2.0

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmdO66kACgkQ4+MsLN6t
# wN4dGw//ZSVZO/cdOc1izC44mCPjnHkhSjX5i12E09QnDfNQ9cef2pG50lsndm3N
# xJwunST+Hqfoh07YT4zuGSO+60BiiPsPse+ytnuHK0E4wLbIfrWzPcf1CMRDk3wf
# O4IniCv8O7BRYccK1iPkFI8ZVqK84m3Bn1tQ51LOy09b6om7jIqNwlxWxGYqkWCR
# l+9RxW/IKWD3OrveBATEhD1lwoUBIYPPzdm6NayEdOJXGOgnvbNbtmuRCR8dCxCS
# RUH3GZu0vvRTpOYjFzeR9Xy+CE1vcOeKsZf35QwV4eX8+UKv30HgF6J8fInkS60m
# EqPA+AU7fi+DN3Ua1Jx38wXHPr8mPTQoVGV6q0UB8b3B22kPQn+Mu8jHLyMOp4rK
# 6JEX5cAOZuwBEk0I8VjwqeDGVYgFwnMHhVuss0N4kCvP0qQcC2JMGMW0Rk4Lxw16
# q0kwbQ0c56+qET57EOf9VW40Yb1q3Zu9t39XfNO/m3KGLkVSPfFZEu5voWD14mZW
# RutPlG/ww3n878Xz06YsqYF0ED/0SiW8U1tmEzg+X9vA/7Z7/0MH0rleNb36a0Fs
# 0aDq/mZBcnFLKy+9rlpy18OolY/N6LZnebCpdQe5wSRn/ioWDc4/GyIaLO9lSE5o
# TFHmGtIIQ2FoeRgtwCSfVNwaA6ILgPRsgXFDOqxCUplgKv6GX2Q=
# =GZ/H
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 03 Dec 2024 11:29:45 GMT
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD  6BB2 E3E3 2C2C DEAD C0DE

* tag 'hw-misc-20241203' of https://github.com/philmd/qemu:
  system: Select HVF by default when no other accelerator is available
  tests/qtest: add test for querying balloon guest stats
  tests/qtest: drop 'fuzz-' prefix from virtio-balloon test
  hw/virtio: fix crash in processing balloon stats
  hw/display/vga: Do not reset 'big_endian_fb' in vga_common_reset()
  target/riscv: Avoid bad shift in riscv_cpu_do_interrupt()
  hw/core/machine: diagnose wrapping of maxmem
  MAINTAINERS: update email addr for Brian Cain
  meson: Add missing SDL dependency to system/main.c
  MAINTAINERS: add myself as the maintainer for LoongArch VirtMachine
  ui/cocoa: Temporarily ignore annoying deprecated declaration warnings
  hw/openrisc/openrisc_sim: keep serial@90000000 as default
  hw/openrisc: Fixed undercounting of TTCR in continuous mode

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'pull-or1k-20241203' of https://github.com/stffrdhrn/qemu into staging
Peter Maydell [Tue, 3 Dec 2024 13:43:57 +0000 (13:43 +0000)]
Merge tag 'pull-or1k-20241203' of https://github.com/stffrdhrn/qemu into staging

OpenRISC updates for 9.2.0

This series has 2 fixes:
 - Fix to keep serial@90000000 as default
 - Fixed undercounting of TTCR in continuous mode

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE2cRzVK74bBA6Je/xw7McLV5mJ+QFAmdO56EACgkQw7McLV5m
# J+T8BRAAxZMH4ykdRJBmYiFVOsYKagcdT6GGBHL44FGeQSr1lNyoU0Rn5r6v5GHe
# Nwq7DTeZlKoVji5GXki53mGrwENXr00m+xfc9ACMoWr5IM6McQUPXlQAQ/50fIGs
# lzXMZH/4EdPIVkkpCi+y8FLYw02oQg61U9G0HW02lQJy4Y4mudtvQFGzJ7f3SIZ3
# EkKn5YLG0bqszq/amFNLQXlbnq3yI5zfcMHhHx0KuDsm2yNhrNA+AJP8tLI3JlxL
# +0YIA+fWuxQzz8Zu9+ckc8VAV83HIgQpXVzI6rQxdSwbmRgUu9ITO09ZmxaDHZF6
# sDI6K3VouyaHJVkvu4coDajpYTjHLE26c9LAlaVBpgdnmnYy4vlndEqbfaBqOouX
# n0N2wJ3IGouIw7AnB9dTaZhM/Uo09hZKDr6kCm3hLfPn2+vi3yxsbwVwLaOpH3G3
# kQ5ZFKjoA7XWOaXGOUMcmhByXkSxja+pSBppB58vJAFyVp73HYIpea3/q1Zd8S4S
# noJoqxDtD2zf26bDBIe83pUEnSnL8fAcsh3rlQP8HrWYhU7ZulVSE1ZvPkPgDpkY
# LVCPautTElsMp2Mg4a2oODGvSDN4/5h2dp6TaK4Qep92HHFOwPZQBQW607VwWR5N
# II8dB/l8PluKkgZ3ymhP1p9JAAZFe9a2cMmegRIiM74PkPty0kk=
# =guIi
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 03 Dec 2024 11:12:33 GMT
# gpg:                using RSA key D9C47354AEF86C103A25EFF1C3B31C2D5E6627E4
# gpg: Good signature from "Stafford Horne <shorne@gmail.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: D9C4 7354 AEF8 6C10 3A25  EFF1 C3B3 1C2D 5E66 27E4

* tag 'pull-or1k-20241203' of https://github.com/stffrdhrn/qemu:
  hw/openrisc: Fixed undercounting of TTCR in continuous mode
  hw/openrisc/openrisc_sim: keep serial@90000000 as default

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agochardev: Remove __-prefixed names
Nicholas Piggin [Wed, 28 Aug 2024 04:33:35 +0000 (14:33 +1000)]
chardev: Remove __-prefixed names

Peter points out double underscore prefix names tend to be reserved
for the system. Clean these up.

Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-ID: <20240828043337.14587-3-npiggin@gmail.com>

4 months agochardev: Fix record/replay error path NULL deref in device creation
Nicholas Piggin [Wed, 28 Aug 2024 04:33:34 +0000 (14:33 +1000)]
chardev: Fix record/replay error path NULL deref in device creation

qemu_chardev_set_replay() was being called in chardev creation to
set up replay parameters even if the chardev is NULL.

A segfault can be reproduced by specifying '-serial chardev:bad' with
an rr=record mode.

Fix this with a NULL pointer check.

Reported-by: Peter Maydell <peter.maydell@linaro.org>
Resolves: Coverity CID 1559470
Fixes: 4c193bb129dae ("chardev: set record/replay on the base device of a muxed device")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-ID: <20240828043337.14587-2-npiggin@gmail.com>

4 months agosystem: Select HVF by default when no other accelerator is available
Philippe Mathieu-Daudé [Tue, 3 Dec 2024 09:36:16 +0000 (10:36 +0100)]
system: Select HVF by default when no other accelerator is available

When testing with a HVF-only binary, we get:

   3/12 qemu:func-quick+func-aarch64 / func-aarch64-version                                      ERROR            0.29s   exit status 1
  stderr:
  Traceback (most recent call last):
    File "tests/functional/test_version.py", line 22, in test_qmp_human_info_version
      self.vm.launch()
    File "machine/machine.py", line 461, in launch
      raise VMLaunchFailure(
  qemu.machine.machine.VMLaunchFailure: ConnectError: Failed to establish session: EOFError
      Exit code: 1
      Command: build/qemu-system-aarch64 -display none -vga none -chardev socket,id=mon,fd=5 -mon chardev=mon,mode=control -machine none -nodefaults
      Output: qemu-system-aarch64: No accelerator selected and no default accelerator available

Fix by checking for HVF in configure_accelerators() and using
it by default when no other accelerator is available.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20241203094232.62232-1-philmd@linaro.org>

4 months agotests/qtest: add test for querying balloon guest stats
Daniel P. Berrangé [Fri, 29 Nov 2024 13:55:07 +0000 (13:55 +0000)]
tests/qtest: add test for querying balloon guest stats

This test would have identified the crash caused by the addition of new
balloon stats fields.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-ID: <20241129135507.699030-4-berrange@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agotests/qtest: drop 'fuzz-' prefix from virtio-balloon test
Daniel P. Berrangé [Fri, 29 Nov 2024 13:55:06 +0000 (13:55 +0000)]
tests/qtest: drop 'fuzz-' prefix from virtio-balloon test

This test file is expected to be extended for arbitrary virtio-balloon
related tests, not merely those discovered by fuzzing.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Message-ID: <20241129135507.699030-3-berrange@redhat.com>
[PMD: Update MAINTAINERS]
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
4 months agohw/virtio: fix crash in processing balloon stats
Daniel P. Berrangé [Fri, 29 Nov 2024 13:55:05 +0000 (13:55 +0000)]
hw/virtio: fix crash in processing balloon stats

balloon_stats_get_all will iterate over guest stats upto the max
VIRTIO_BALLOON_S_NR value, calling visit_type_uint64 to populate
the QObject dict. The dict keys are obtained from the static
array balloon_stat_names which is VIRTIO_BALLOON_S_NR in size.

Unfortunately the way that array is declared results in any
unassigned stats getting a NULL name, which will then cause
visit_type_uint64 to trigger an assert in qobject_output_add_obj.

The balloon_stat_names array was fortunately fully populated with
names until recently:

  commit 0d2eeef77a33315187df8519491a900bde4a3d83
  Author: Bibo Mao <maobibo@loongson.cn>
  Date:   Mon Oct 28 10:38:09 2024 +0800

    linux-headers: Update to Linux v6.12-rc5

pulled a change to include/standard-headers/linux/virtio_balloon.h
which increased VIRTIO_BALLOON_S_NR by 6, and failed to add the new
names to balloon_stat_names.

This commit fills in the missing names, and uses a static assert to
guarantee that any future changes to VIRTIO_BALLOON_S_NR will cause
a build failure until balloon_stat_names is updated.

This problem was detected by the Cockpit Project's automated
integration tests on QEMU 9.2.0-rc1.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2329448
Fixes: 0d2eeef77a3 ("linux-headers: Update to Linux v6.12-rc5")
Reported-by: Martin Pitt <mpitt@redhat.com>
Reviewed-by: Richard W.M. Jones <rjones@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Michael Tokarev <mjt@tls.msk.ru>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Message-ID: <20241129135507.699030-2-berrange@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agohw/display/vga: Do not reset 'big_endian_fb' in vga_common_reset()
Philippe Mathieu-Daudé [Thu, 28 Nov 2024 17:54:09 +0000 (18:54 +0100)]
hw/display/vga: Do not reset 'big_endian_fb' in vga_common_reset()

The 'pci-vga' device allow setting a 'big-endian-framebuffer'
property since commit 3c2784fc864 ("vga: Expose framebuffer
byteorder as a QOM property"). Similarly, the 'virtio-vga'
device since commit 8be61ce2ce3 ("virtio-vga: implement
big-endian-framebuffer property").

Both call vga_common_reset() in their reset handler, respectively
pci_secondary_vga_reset() and virtio_vga_base_reset_hold(), which
reset 'big_endian_fb', overwritting the property. This is not
correct: the hardware is expected to keep its configured
endianness during resets.

Move 'big_endian_fb' assignment from vga_common_reset() to
vga_common_init() which is called once when the common VGA state
is initialized.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Message-Id: <20241129101721.17836-2-philmd@linaro.org>

4 months agotarget/riscv: Avoid bad shift in riscv_cpu_do_interrupt()
Peter Maydell [Thu, 28 Nov 2024 10:38:31 +0000 (10:38 +0000)]
target/riscv: Avoid bad shift in riscv_cpu_do_interrupt()

In riscv_cpu_do_interrupt() we use the 'cause' value we got out of
cs->exception as a shift value.  However this value can be larger
than 31, which means that "1 << cause" is undefined behaviour,
because we do the shift on an 'int' type.

This causes the undefined behaviour sanitizer to complain
on one of the check-tcg tests:

$ UBSAN_OPTIONS=print_stacktrace=1:abort_on_error=1:halt_on_error=1 ./build/clang/qemu-system-riscv64 -M virt -semihosting -display none -device loader,file=build/clang/tests/tcg/riscv64-softmmu/issue1060
../../target/riscv/cpu_helper.c:1805:38: runtime error: shift exponent 63 is too large for 32-bit type 'int'
    #0 0x55f2dc026703 in riscv_cpu_do_interrupt /mnt/nvmedisk/linaro/qemu-from-laptop/qemu/build/clang/../../target/riscv/cpu_helper.c:1805:38
    #1 0x55f2dc3d170e in cpu_handle_exception /mnt/nvmedisk/linaro/qemu-from-laptop/qemu/build/clang/../../accel/tcg/cpu-exec.c:752:9

In this case cause is RISCV_EXCP_SEMIHOST, which is 0x3f.

Use 1ULL instead to ensure that the shift is in range.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Fixes: 1697837ed9 ("target/riscv: Add M-mode virtual interrupt and IRQ filtering support.")
Fixes: 40336d5b1d ("target/riscv: Add HS-mode virtual interrupt and IRQ filtering support.")
Reviewed-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: <20241128103831.3452572-1-peter.maydell@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agohw/core/machine: diagnose wrapping of maxmem
Daniel P. Berrangé [Wed, 27 Nov 2024 11:40:57 +0000 (11:40 +0000)]
hw/core/machine: diagnose wrapping of maxmem

The 'maxmem' parameter parsed on the command line is held in uint64_t
and then assigned to the MachineState field that is 'ram_addr_t'. This
assignment will wrap on 32-bit hosts, silently changing the user's
config request if it were over-sized.

Improve the existing diagnositics for validating 'size', and add the
same diagnostics for 'maxmem'

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Tested-by: Ani Sinha <anisinha@redhat.com>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Message-ID: <20241127114057.255995-1-berrange@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agoMAINTAINERS: update email addr for Brian Cain
Brian Cain [Sat, 23 Nov 2024 16:46:40 +0000 (08:46 -0800)]
MAINTAINERS: update email addr for Brian Cain

Also: add mapping for "quic_bcain@quicinc.com" which was ~briefly
used for some replies to mailing list traffic.

Signed-off-by: Brian Cain <bcain@quicinc.com>
Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
Tested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20241123164641.364748-2-bcain@quicinc.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agomeson: Add missing SDL dependency to system/main.c
Philippe Mathieu-Daudé [Wed, 20 Nov 2024 11:36:43 +0000 (12:36 +0100)]
meson: Add missing SDL dependency to system/main.c

When building QEMU configure with --disable-gtk --disable-cocoa
on macOS we get:

  User interface
    Cocoa support                   : NO
    SDL support                     : YES 2.30.5
    SDL image support               : NO
    GTK support                     : NO
    pixman                          : YES 0.42.2
    VTE support                     : NO
    PNG support                     : YES 1.6.43
    VNC support                     : YES
    VNC SASL support                : YES
    VNC JPEG support                : YES 3.0.3
    spice protocol support          : YES 0.14.4
      spice server support          : NO
    curses support                  : YES
    brlapi support                  : NO

  User defined options
    cocoa                           : disabled
    docs                            : disabled
    gtk                             : disabled

  ../system/main.c:30:10: fatal error: 'SDL.h' file not found
     30 | #include <SDL.h>
        |          ^~~~~~~
  1 error generated.

Fix by adding the SDL dependency to main.c it's CFLAGS contains
the SDL include directory.

Fixes: 64ed6f92ff ("meson: link emulators without Makefile.target")
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20241120114943.85080-1-philmd@linaro.org>

4 months agoMAINTAINERS: add myself as the maintainer for LoongArch VirtMachine
Bibo Mao [Tue, 12 Nov 2024 07:37:14 +0000 (15:37 +0800)]
MAINTAINERS: add myself as the maintainer for LoongArch VirtMachine

Song Gao is will be sick leave for a long time, I apply for maintainer
for LoongArch Virt Machine during this period, LoongArch TCG keeps unchanged
since I am not familiar with it. The maintainer duty will transfer to him
after he comes back to work.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Acked-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20241112073714.1953481-1-maobibo@loongson.cn>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agoui/cocoa: Temporarily ignore annoying deprecated declaration warnings
Philippe Mathieu-Daudé [Tue, 17 Sep 2024 12:07:56 +0000 (14:07 +0200)]
ui/cocoa: Temporarily ignore annoying deprecated declaration warnings

These warnings are breaking some build configurations since 2 months
now (https://gitlab.com/qemu-project/qemu/-/issues/2575):

  ui/cocoa.m:662:14: error: 'CVDisplayLinkCreateWithCGDisplay' is deprecated: first deprecated in macOS 15.0 - use NSView.displayLink(target:selector:), NSWindow.displayLink(target:selector:), or NSScreen.displayLink(target:selector:)  [-Werror,-Wdeprecated-declarations]
    662 |         if (!CVDisplayLinkCreateWithCGDisplay(display, &displayLink)) {
        |              ^
  /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/CoreVideo.framework/Headers/CVDisplayLink.h:89:20: note: 'CVDisplayLinkCreateWithCGDisplay' has been explicitly marked deprecated here
     89 | CV_EXPORT CVReturn CVDisplayLinkCreateWithCGDisplay(
        |                    ^
  ui/cocoa.m:663:29: error: 'CVDisplayLinkGetNominalOutputVideoRefreshPeriod' is deprecated: first deprecated in macOS 15.0 - use NSView.displayLink(target:selector:), NSWindow.displayLink(target:selector:), or NSScreen.displayLink(target:selector:)  [-Werror,-Wdeprecated-declarations]
    663 |             CVTime period = CVDisplayLinkGetNominalOutputVideoRefreshPeriod(displayLink);
        |                             ^
  /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/CoreVideo.framework/Headers/CVDisplayLink.h:182:18: note: 'CVDisplayLinkGetNominalOutputVideoRefreshPeriod' has been explicitly marked deprecated here
    182 | CV_EXPORT CVTime CVDisplayLinkGetNominalOutputVideoRefreshPeriod( CVDisplayLinkRef CV_NONNULL displayLink );
        |                  ^
  ui/cocoa.m:664:13: error: 'CVDisplayLinkRelease' is deprecated: first deprecated in macOS 15.0 - use NSView.displayLink(target:selector:), NSWindow.displayLink(target:selector:), or NSScreen.displayLink(target:selector:)  [-Werror,-Wdeprecated-declarations]
    664 |             CVDisplayLinkRelease(displayLink);
        |             ^
  /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/System/Library/Frameworks/CoreVideo.framework/Headers/CVDisplayLink.h:249:16: note: 'CVDisplayLinkRelease' has been explicitly marked deprecated here
    249 | CV_EXPORT void CVDisplayLinkRelease( CV_RELEASES_ARGUMENT CVDisplayLinkRef CV_NULLABLE displayLink );
        |                ^
  3 errors generated.

For the next release, ignore the warnings using #pragma directives.
At least until we figure the correct new API usage.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Phil Dennis-Jordan <phil@philjordan.eu>
Tested-by: Phil Dennis-Jordan <phil@philjordan.eu>
Message-Id: <20241121131954.98949-1-philmd@linaro.org>

4 months agohw/openrisc/openrisc_sim: keep serial@90000000 as default
Ahmad Fatoum [Thu, 22 Aug 2024 16:38:38 +0000 (18:38 +0200)]
hw/openrisc/openrisc_sim: keep serial@90000000 as default

We used to only have a single UART on the platform and it was located at
address 0x90000000. When the number of UARTs was increased to 4, the
first UART remained at it's location, but instead of being the first one
to be registered, it became the last.

This caused QEMU to pick 0x90000300 as the default UART, which broke
software that hardcoded the address of 0x90000000 and expected it's
output to be visible when the user configured only a single console.

This caused regressions[1] in the barebox test suite when updating to a
newer QEMU. As there seems to be no good reason to register the UARTs in
inverse order, let's register them by ascending address, so existing
software can remain oblivious to the additional UART ports.

Changing the order of uart registration alone breaks Linux which
was choosing the UART at 0x90000300 as the default for ttyS0.  To fix
Linux we fix three things in the device tree:

 1. Define stdout-path only one time for the first registered UART
    instead of incorrectly defining for each UART.
 2. Change the UART alias name from 'uart0' to 'serial0' as almost all
    Linux tty drivers look for an alias starting with "serial".
 3. Add the UART nodes so they appear in the final DTB in the
    order starting with the lowest address and working upwards.

In summary these changes mean that the QEMU default UART (serial_hd(0))
is now setup where:

 * serial_hd(0) is the lowest-address UART
 * serial_hd(0) is listed first in the DTB
 * serial_hd(0) is the /chosen/stdout-path one
 * the /aliases/serial0 alias points at serial_hd(0)

[1]: https://lore.barebox.org/barebox/707e7c50-aad1-4459-8796-0cc54bab32e2@pengutronix.de/T/#m5da26e8a799033301489a938b5d5667b81cef6ad

[stafford: Change to serial0 alias and update change message, reverse
 uart registration order]

Fixes: 777784bda468 ("hw/openrisc: support 4 serial ports in or1ksim")
Cc: qemu-stable@nongnu.org
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-ID: <20241203110536.402131-2-shorne@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agohw/openrisc: Fixed undercounting of TTCR in continuous mode
Joel Holdsworth [Fri, 7 Jun 2024 22:29:33 +0000 (15:29 -0700)]
hw/openrisc: Fixed undercounting of TTCR in continuous mode

In the existing design, TTCR is prone to undercounting when running in
continuous mode. This manifests as a timer interrupt appearing to
trigger a few cycles prior to the deadline set in SPR_TTMR_TP.

When the timer triggers, the virtual time delta in nanoseconds between
the time when the timer was set, and when it triggers is calculated.
This nanoseconds value is then divided by TIMER_PERIOD (50) to compute
an increment of cycles to apply to TTCR.

However, this calculation rounds down the number of cycles causing the
undercounting.

A simplistic solution would be to instead round up the number of cycles,
however this will result in the accumulation of timing error over time.

This patch corrects the issue by calculating the time delta in
nanoseconds between when the timer was last reset and the timer event.
This approach allows the TTCR value to be rounded up, but without
accumulating error over time.

Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
[stafford: Incremented version in vmstate_or1k_timer, checkpatch fixes]
Signed-off-by: Stafford Horne <shorne@gmail.com>
Message-ID: <20241203110536.402131-3-shorne@gmail.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
4 months agohw/openrisc: Fixed undercounting of TTCR in continuous mode
Joel Holdsworth [Fri, 7 Jun 2024 22:29:33 +0000 (15:29 -0700)]
hw/openrisc: Fixed undercounting of TTCR in continuous mode

In the existing design, TTCR is prone to undercounting when running in
continuous mode. This manifests as a timer interrupt appearing to
trigger a few cycles prior to the deadline set in SPR_TTMR_TP.

When the timer triggers, the virtual time delta in nanoseconds between
the time when the timer was set, and when it triggers is calculated.
This nanoseconds value is then divided by TIMER_PERIOD (50) to compute
an increment of cycles to apply to TTCR.

However, this calculation rounds down the number of cycles causing the
undercounting.

A simplistic solution would be to instead round up the number of cycles,
however this will result in the accumulation of timing error over time.

This patch corrects the issue by calculating the time delta in
nanoseconds between when the timer was last reset and the timer event.
This approach allows the TTCR value to be rounded up, but without
accumulating error over time.

Signed-off-by: Joel Holdsworth <jholdsworth@nvidia.com>
[stafford: Incremented version in vmstate_or1k_timer, checkpatch fixes]
Signed-off-by: Stafford Horne <shorne@gmail.com>
4 months agohw/openrisc/openrisc_sim: keep serial@90000000 as default
Ahmad Fatoum [Thu, 22 Aug 2024 16:38:38 +0000 (18:38 +0200)]
hw/openrisc/openrisc_sim: keep serial@90000000 as default

We used to only have a single UART on the platform and it was located at
address 0x90000000. When the number of UARTs was increased to 4, the
first UART remained at it's location, but instead of being the first one
to be registered, it became the last.

This caused QEMU to pick 0x90000300 as the default UART, which broke
software that hardcoded the address of 0x90000000 and expected it's
output to be visible when the user configured only a single console.

This caused regressions[1] in the barebox test suite when updating to a
newer QEMU. As there seems to be no good reason to register the UARTs in
inverse order, let's register them by ascending address, so existing
software can remain oblivious to the additional UART ports.

Changing the order of uart registration alone breaks Linux which
was choosing the UART at 0x90000300 as the default for ttyS0.  To fix
Linux we fix three things in the device tree:

 1. Define stdout-path only one time for the first registered UART
    instead of incorrectly defining for each UART.
 2. Change the UART alias name from 'uart0' to 'serial0' as almost all
    Linux tty drivers look for an alias starting with "serial".
 3. Add the UART nodes so they appear in the final DTB in the
    order starting with the lowest address and working upwards.

In summary these changes mean that the QEMU default UART (serial_hd(0))
is now setup where:

 * serial_hd(0) is the lowest-address UART
 * serial_hd(0) is listed first in the DTB
 * serial_hd(0) is the /chosen/stdout-path one
 * the /aliases/serial0 alias points at serial_hd(0)

[1]: https://lore.barebox.org/barebox/707e7c50-aad1-4459-8796-0cc54bab32e2@pengutronix.de/T/#m5da26e8a799033301489a938b5d5667b81cef6ad

Fixes: 777784bda468 ("hw/openrisc: support 4 serial ports in or1ksim")
Cc: qemu-stable@nongnu.org
Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
[stafford: Change to serial0 alias and update change message, reverse
 uart registration order]
Signed-off-by: Stafford Horne <shorne@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'pull-nvme-20241203' of https://gitlab.com/birkelund/qemu into staging
Peter Maydell [Tue, 3 Dec 2024 10:22:21 +0000 (10:22 +0000)]
Merge tag 'pull-nvme-20241203' of https://gitlab.com/birkelund/qemu into staging

nvme queue

# -----BEGIN PGP SIGNATURE-----
#
# iQEzBAABCgAdFiEEUigzqnXi3OaiR2bATeGvMW1PDekFAmdOpR0ACgkQTeGvMW1P
# DenTOgf/S4+yjo1pkOIyWoDEQZ434eOwXeUqcM8taGwwRB/IyulFBbGexkY8Feu3
# iCYA78FTtGcTCIKWc0ze5TsA2abbrX1mB1IxR3keeYPdQyZPacfxXG5qjjeygU0N
# 2Ws6UZBFZDNBkLxasT6yHBl9oHriz91IjksA62FYr1pXGHfLbluWTyKYghPfoY4t
# ifc0tTSinhsFjKixRvuyUo7AWd+jm29QJbcfhckcPTVZ52VE5FUGg0bWvdLk/zWY
# K4qlPVa2MngV3fSY04Qt1lHuLUCegiHMCjyJHvN3S2nxhJLSWLyE/RU9zUJCXFsd
# HvJ7nKxq8GODjLNghsjOzdL1rxQPdA==
# =yOQB
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 03 Dec 2024 06:28:45 GMT
# gpg:                using RSA key 522833AA75E2DCE6A24766C04DE1AF316D4F0DE9
# gpg: Good signature from "Klaus Jensen <its@irrelevant.dk>" [full]
# gpg:                 aka "Klaus Jensen <k.jensen@samsung.com>" [full]
# Primary key fingerprint: DDCA 4D9C 9EF9 31CC 3468  4272 63D5 6FC5 E55D A838
#      Subkey fingerprint: 5228 33AA 75E2 DCE6 A247  66C0 4DE1 AF31 6D4F 0DE9

* tag 'pull-nvme-20241203' of https://gitlab.com/birkelund/qemu:
  hw/nvme: take a reference on the subsystem on vf realization
  hw/nvme: SR-IOV VFs must hardwire pci interrupt pin register to zero
  hw/nvme: fix use/unuse of msix vectors
  hw/nvme: fix msix_uninit with exclusive bar

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'migration-20241202-pull-request' of https://gitlab.com/peterx/qemu into...
Peter Maydell [Tue, 3 Dec 2024 10:22:13 +0000 (10:22 +0000)]
Merge tag 'migration-20241202-pull-request' of https://gitlab.com/peterx/qemu into staging

Migration pull for 9.2-rc3

- Prasad's one fix on thread name typo

# -----BEGIN PGP SIGNATURE-----
#
# iIgEABYKADAWIQS5GE3CDMRX2s990ak7X8zN86vXBgUCZ04mxBIccGV0ZXJ4QHJl
# ZGhhdC5jb20ACgkQO1/MzfOr1wbyegEAxWBfoGpPoysYOe+ijwMZKgsXylWzHrfJ
# HGlVck7f1RQA/jAwS8aR3t6QZwhL42lXfAlGPX2ebz4IL+f52qs6/nQP
# =L+tL
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 02 Dec 2024 21:29:40 GMT
# gpg:                using EDDSA key B9184DC20CC457DACF7DD1A93B5FCCCDF3ABD706
# gpg:                issuer "peterx@redhat.com"
# gpg: Good signature from "Peter Xu <xzpeter@gmail.com>" [marginal]
# gpg:                 aka "Peter Xu <peterx@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: B918 4DC2 0CC4 57DA CF7D  D1A9 3B5F CCCD F3AB D706

* tag 'migration-20241202-pull-request' of https://gitlab.com/peterx/qemu:
  migration: correct multifd receive thread name

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/nvme: take a reference on the subsystem on vf realization
Klaus Jensen [Mon, 11 Nov 2024 11:14:49 +0000 (12:14 +0100)]
hw/nvme: take a reference on the subsystem on vf realization

Make sure we grab a reference on the subsystem when a VF is realized.
Otherwise, the subsytem will be unrealized automatically when the VFs
are unregistered and unreffed.

This fixes a latent bug but was not exposed until commit 08f632848008
("pcie: Release references of virtual functions"). This was then fixed
(or rather, hidden) by commit c613ad25125b ("pcie_sriov: Do not manually
unrealize"), but that was then reverted (due to other issues) in commit
b0fdaee5d1ed, exposing the bug yet again.

Cc: qemu-stable@nongnu.org
Fixes: 08f632848008 ("pcie: Release references of virtual functions")
Reviewed-by: Jesper Wendel Devantier <foss@defmacro.it>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
4 months agohw/nvme: SR-IOV VFs must hardwire pci interrupt pin register to zero
Klaus Jensen [Mon, 11 Nov 2024 11:15:10 +0000 (12:15 +0100)]
hw/nvme: SR-IOV VFs must hardwire pci interrupt pin register to zero

The PCI Interrupt Pin Register does not apply to VFs and MUST be
hardwired to zero.

Fixes: 44c2c09488db ("hw/nvme: Add support for SR-IOV")
Reviewed-by: Jesper Wendel Devantier <foss@defmacro.it>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
4 months agohw/nvme: fix use/unuse of msix vectors
Klaus Jensen [Sun, 10 Nov 2024 19:26:23 +0000 (20:26 +0100)]
hw/nvme: fix use/unuse of msix vectors

Only call msix_{un,}use_vector() when interrupts are actually enabled
for a completion queue.

Reviewed-by: Jesper Wendel Devantier <foss@defmacro.it>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
4 months agohw/nvme: fix msix_uninit with exclusive bar
Klaus Jensen [Sun, 10 Nov 2024 13:04:27 +0000 (14:04 +0100)]
hw/nvme: fix msix_uninit with exclusive bar

Commit fa905f65c554 introduced a machine compatibility parameter to
enable an exclusive bar for msix. It failed to account for this when
cleaning up. Make sure that if an exclusive bar is enabled, we use the
proper cleanup routine.

Cc: qemu-stable@nongnu.org
Fixes: fa905f65c554 ("hw/nvme: add machine compatibility parameter to enable msix exclusive bar")
Reviewed-by: Jesper Wendel Devantier <foss@defmacro.it>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
4 months agomigration: correct multifd receive thread name
Prasad Pandit [Wed, 27 Nov 2024 11:15:28 +0000 (16:45 +0530)]
migration: correct multifd receive thread name

Multifd receive threads run on the destination side.
Correct the thread name marco to indicate the same.

Fixes: e620b1e4770b ("migration: Put thread names together with macros")
Signed-off-by: Prasad Pandit <pjp@fedoraproject.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Link: https://lore.kernel.org/r/20241127111528.167330-1-ppandit@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
4 months agoMerge tag 'pull-request-2024-12-02' of https://gitlab.com/thuth/qemu into staging
Peter Maydell [Mon, 2 Dec 2024 16:16:14 +0000 (16:16 +0000)]
Merge tag 'pull-request-2024-12-02' of https://gitlab.com/thuth/qemu into staging

* Fix timeouts in arm-sx1 test
* Check for the "pc" machine in the virtio_version and acpi-bits tests

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEJ7iIR+7gJQEY8+q5LtnXdP5wLbUFAmdNvDsRHHRodXRoQHJl
# ZGhhdC5jb20ACgkQLtnXdP5wLbXEDBAAkGHlPrQ9JVDjR+SHn4Bjl7DOQbKuLq1T
# MXPSOUvYtX7/F0WOMBXQOXj86iv3HphvlQD/kFYfdogC8trE1gLyk2vGQnwkr0FK
# cZRQZb9wdg8XYJEP++59WEQL8MIS+TcQkjqpUXbT5T3ZSmODc1FMEZinsvChM/C/
# XC6VcRAibpQ6k7189ORWZfmjlbnxbS95+NJ2FZ/c2oyFbA2fJneef5RWcp/0/1hx
# isjRDI9wTb+l12JcRn06Osh9KyjUbz3qaN7HIrsMoMHYtsYOCn4r/MKD56iGp2Aw
# G86nwGyNIZBcfadnL792gu6xda8A915aWIIn6WzRTnGvWKseoub2hHDYrADW0wvV
# 8lsMa2uHoLp1SLRDE4TYWgzsbgO2ueC73T/QPpmkaAdYaRaFpKl3+oT19UsbtqUg
# xzUpR1nM8sdOF9g5OF971z10LaVFRE710UWuiWVRlS3gSXMvQ93PR7rKhM6bM/yM
# x+fP9/X9wka8ZFXQ67XGqPxE4xrXN0JD4wCKH5KysIO9hJ+c2GkZitFV7VB9r365
# VdbWItw/lTznIuQ+54Zm6xB9kskwnp6XAPzSgQwJEwD8wTDRM7vo0G+CTHYtKJ68
# VZVgwkbSMoI0XDMsAoxS4V+wtuOj7UWTYLskHkV7Tqp87oq7cIgEJy3CnvIjWZ0d
# Uh+L0I4qz60=
# =oeWj
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 02 Dec 2024 13:55:07 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-2024-12-02' of https://gitlab.com/thuth/qemu:
  tests/functional: increase timeouts for arm sx1 test
  tests/functional/test_virtio_version: Check for the availability of the machine
  tests/functional/test_acpi_bits: Turn the test into a QemuSystemTest

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agotests/functional: increase timeouts for arm sx1 test
Daniel P. Berrangé [Fri, 29 Nov 2024 17:30:59 +0000 (17:30 +0000)]
tests/functional: increase timeouts for arm sx1 test

When under high load the test VM does not complete running in the
default 30 second timeout. Double it to give more headroom.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Message-ID: <20241129173120.761728-2-berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months agotests/functional/test_virtio_version: Check for the availability of the machine
Thomas Huth [Thu, 28 Nov 2024 12:01:42 +0000 (13:01 +0100)]
tests/functional/test_virtio_version: Check for the availability of the machine

Use self_set_machine() to set and check for the availability of the
default pc machine (so that the test is not failing if the machine
has not been included in the QEMU binary).

Message-ID: <20241128120142.593408-1-thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months agotests/functional/test_acpi_bits: Turn the test into a QemuSystemTest
Thomas Huth [Thu, 28 Nov 2024 11:50:19 +0000 (12:50 +0100)]
tests/functional/test_acpi_bits: Turn the test into a QemuSystemTest

By using QemuSystemTest as a base class, we can use the set_machine()
command to check whether the required machine is available in the
binary (otherwise this test is failing when QEMU has been compiled
without the default 'pc' machine type).

Message-ID: <20241128115019.591362-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Ani Sinha <anisinha@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months agoMerge tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu into staging
Peter Maydell [Mon, 2 Dec 2024 12:31:39 +0000 (12:31 +0000)]
Merge tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu into staging

chardev patch queue

# -----BEGIN PGP SIGNATURE-----
#
# iQJQBAABCAA6FiEEh6m9kz+HxgbSdvYt2ujhCXWWnOUFAmdNp4EcHG1hcmNhbmRy
# ZS5sdXJlYXVAcmVkaGF0LmNvbQAKCRDa6OEJdZac5WsaD/92DwK0nKM7HqbqQOHN
# lR715V8ve16+mWYZnWA4jQvclFUigg0Xbxl6d6adsK5QnW2nDGgjf3cXsCngUx7K
# jNCmynQrpmUx7a7BRgXylEl5zN/HLeVvk2Qb4Jg8EAkRK0utK1Aerlx6CV2fzbR/
# zDslP+C7ELk9evNwWx6PZRgeOii0sL7lto67J4a5Iri2IU4yi9zI/irEIxeE4cDT
# P3KFXYsgbm86VC6HaOBWw5+d5JyPr2hVHxvKMpI4YNyU/f2RHdhfqgsLS1gms2Yz
# JBuh8DpIq0E+rdQp4ofH5SBPuOFxg7SgDFAyqgHerDd4uAwUafyOfAEVy2hpZmNT
# 3JlavSMb2SFhWboMUu9B+gRHC+sc7YCoPZEhaIaciW3XaC/S1JJi/fvB8gcugBDs
# lgTUIDEsXLMcgXaFNKbRXpuLG5eXRnzaQ4Y9Gr/3yEvY8eSZ6+x3EDwSMGfg/Ta1
# OW+h09AvWfJR5eZ13x+5CCMFRs8RgQN/D+mxYt3cZ/SdNSBci/4By54IGEgGCCM8
# xoTAqcd2kbpHg2S/MTx7O7OPNC38TbUaOWDR6u0qO0ZXqZA6OdlC+g/Z1uu7ZMqA
# FkOTQZuDDgOOhpsmLy4v3nKmFBQNJAPVyfph88DE1QyrU5nVCS55lyTy/ZH2hayL
# RCVbI20jrgEJoi44EFdFu6QP/Q==
# =0vex
# -----END PGP SIGNATURE-----
# gpg: Signature made Mon 02 Dec 2024 12:26:41 GMT
# gpg:                using RSA key 87A9BD933F87C606D276F62DDAE8E10975969CE5
# gpg:                issuer "marcandre.lureau@redhat.com"
# gpg: Good signature from "Marc-André Lureau <marcandre.lureau@redhat.com>" [full]
# gpg:                 aka "Marc-André Lureau <marcandre.lureau@gmail.com>" [full]
# Primary key fingerprint: 87A9 BD93 3F87 C606 D276  F62D DAE8 E109 7596 9CE5

* tag 'chr-pull-request' of https://gitlab.com/marcandre.lureau/qemu:
  chardev/char-mux: make boolean bit check instead of find_next_bit()
  chardev/char-mux: shift unsigned long to avoid 32-bit overflow

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agochardev/char-mux: make boolean bit check instead of find_next_bit()
Roman Penyaev [Fri, 29 Nov 2024 10:32:39 +0000 (11:32 +0100)]
chardev/char-mux: make boolean bit check instead of find_next_bit()

This patch simplifies (and makes less confusing) bit checks by
replacing `find_next_bit()` calls with boolean AND operation.

Resolves: Coverity CID 1563776
Signed-off-by: Roman Penyaev <r.peniaev@gmail.com>
Reviewed-by: "Marc-André Lureau" <marcandre.lureau@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Clément Mathieu--Drif <clement.mathieu--drif@eviden.com>
Cc: qemu-devel@nongnu.org
Message-ID: <20241129103239.464061-3-r.peniaev@gmail.com>

4 months agochardev/char-mux: shift unsigned long to avoid 32-bit overflow
Roman Penyaev [Fri, 29 Nov 2024 10:32:38 +0000 (11:32 +0100)]
chardev/char-mux: shift unsigned long to avoid 32-bit overflow

Allthough the size of MAX_MUX is equal to 4 and likely will never
change, this patch changes type of constant to unsigned long to
be on the safe side.

Also add a static compile check that MAX_MUX never bigger than
`sizeof(d->mux_bitset) * BITS_PER_BYTE`.

Signed-off-by: Roman Penyaev <r.peniaev@gmail.com>
Reviewed-by: "Marc-André Lureau" <marcandre.lureau@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Clément Mathieu--Drif <clement.mathieu--drif@eviden.com>
Cc: qemu-devel@nongnu.org
Message-ID: <20241129103239.464061-2-r.peniaev@gmail.com>

4 months agoMerge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging
Peter Maydell [Fri, 29 Nov 2024 10:09:05 +0000 (10:09 +0000)]
Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging

* amd_iommu: fix clang failure on non-KVM targets
* target/i386/hvf: fix advertised 1G page support
* megasas: fix CDB length

# -----BEGIN PGP SIGNATURE-----
#
# iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmdIokMUHHBib256aW5p
# QHJlZGhhdC5jb20ACgkQv/vSX3jHroOpbAf/dxoD/vCFfuMPanIbP4quQGWFTKAo
# x4qlis07s730EWoNzdT42eER+GWcZiT4LhSwAt3EPqjANpnNwcoOGsc5hlJp8emI
# /UbJII90AKLhoJV2gj1xGbRwN1vBJmuBrUQSrLC+5tEr5gFLdHixnlH4pjXGfuF6
# 9eAutgxjRzMiYczsPCev5gbtnvetodP2u9okeQwpQd7N9siCbKx+pB6a67KdvOYn
# tmH0se4ai5w9J4Xnagus/goA5VK5dli6ukvYhLps/LQHj2r7XT6AcPMO7KwFlJkW
# INW0qEzYqa95iH5pNW2lgmJcCn3rp7g1WFVCCocpecp1WpsweVyzMRhGsA==
# =DHEO
# -----END PGP SIGNATURE-----
# gpg: Signature made Thu 28 Nov 2024 17:02:59 GMT
# 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

* tag 'for-upstream' of https://gitlab.com/bonzini/qemu:
  scsi: megasas: Internal cdbs have 16-byte length
  hvf: complete 1G page support
  amd_iommu: Fix kvm_enable_x2apic link error with clang in non-KVM builds

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'pull-9p-20241128' of https://github.com/cschoenebeck/qemu into staging
Peter Maydell [Fri, 29 Nov 2024 10:08:53 +0000 (10:08 +0000)]
Merge tag 'pull-9p-20241128' of https://github.com/cschoenebeck/qemu into staging

* Fix open-unlink-fstat idiom on Linux guests.

* Add test to verify this behaviour.

* Cleanup patches.

# -----BEGIN PGP SIGNATURE-----
#
# iQJLBAABCgA1FiEEltjREM96+AhPiFkBNMK1h2Wkc5UFAmdIvDkXHHFlbXVfb3Nz
# QGNydWRlYnl0ZS5jb20ACgkQNMK1h2Wkc5X8ixAApDPStDxYf1CGdLirInHGp77i
# 0MlBsuaP00f8bZyCLJCFgax2+ogXD72Ptw2thDDMtkMsg9lqZwOtG5I4cJGC3TK2
# J4ZXpg/mg0bY+4o2gvnyeKv8BFl5wE91pdIeFX8ufQ+L2WE+fasWOn38TFB/T/8Z
# 1naN4A8Mu5F9myJ+F6pIYlJfkgbZniNib9BgSMG8pYI6uayWD+YVjR139ozWCf1c
# vhFFpLrwW4j3DOC0WblghQmiMwhXo1QxNAEq0x31/eoD1+calJAwhWsLWksuVIqR
# 6wbGPfNVozgk9l7owYB5Gams5zVJRfLD5LCAitUx2qqMMzxuD3QldLjOmFA/8XdG
# +2/ROBeXJ51blCAMFdp9IwTKzimvuWVL3kXbcQ3n+D459iBZzqW+9w4EYVYShpp6
# uwAAkW9fwVR/U7ERG3n8D6Cw1B9Scvtjksw/VCe9XUNFp6H66K/OXy8NFmnZZk9K
# K9SYkKOVixwZDqMoGoLsoxx0DbakYL+lBYrl6qVZUPRLOjJ+JvLAoblJ0ZmUgsl2
# lXG7vO96+LyRvVjqPoi2D7+MHrmFoeRgWjzZqFqWOakXBHCKcCEVzpAoB4eYyQrj
# rXC5BNhdu9yIa7Dy7V6tFoXPdN1is90bJs92DYTsOG1KdU2DviAUSZk4MjTJzQWN
# 3fvOcZPFq74228CWrN4=
# =XP1U
# -----END PGP SIGNATURE-----
# gpg: Signature made Thu 28 Nov 2024 18:53:45 GMT
# gpg:                using RSA key 96D8D110CF7AF8084F88590134C2B58765A47395
# gpg:                issuer "qemu_oss@crudebyte.com"
# gpg: Good signature from "Christian Schoenebeck <qemu_oss@crudebyte.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: ECAB 1A45 4014 1413 BA38  4926 30DB 47C3 A012 D5F4
#      Subkey fingerprint: 96D8 D110 CF7A F808 4F88  5901 34C2 B587 65A4 7395

* tag 'pull-9p-20241128' of https://github.com/cschoenebeck/qemu:
  tests/9p: also check 'Tgetattr' in 'use-after-unlink' test
  9pfs: fix 'Tgetattr' after unlink
  9pfs: remove obsolete comment in v9fs_getattr()
  tests/9p: add missing Rgetattr response name
  tests/9p: fix Rreaddir response name
  tests/9p: add 'use-after-unlink' test
  9pfs: cleanup V9fsFidState

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agotests/9p: also check 'Tgetattr' in 'use-after-unlink' test
Christian Schoenebeck [Sun, 24 Nov 2024 16:05:32 +0000 (17:05 +0100)]
tests/9p: also check 'Tgetattr' in 'use-after-unlink' test

This verifies expected behaviour of previous bug fix patch.

Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <7017658155c517b9665b75333a97c79aa2d4f3df.1732465720.git.qemu_oss@crudebyte.com>

4 months ago9pfs: fix 'Tgetattr' after unlink
Christian Schoenebeck [Sun, 24 Nov 2024 15:50:03 +0000 (16:50 +0100)]
9pfs: fix 'Tgetattr' after unlink

With a valid file ID (FID) of an open file, it should be possible to send
a 'Tgettattr' 9p request and successfully receive a 'Rgetattr' response,
even if the file has been removed in the meantime. Currently this would
fail with ENOENT.

I.e. this fixes the following misbehaviour with a 9p Linux client:

  open("/home/tst/filename", O_RDWR|O_CREAT|O_EXCL, 0600) = 3
  unlink("/home/tst/filename") = 0
  fstat(3, 0x23aa1a8) = -1 ENOENT (No such file or directory)

Expected results:

  open("/home/tst/filename", O_RDWR|O_CREAT|O_EXCL, 0600) = 3
  unlink("/home/tst/filename") = 0
  fstat(3, {st_mode=S_IFREG|0600, st_size=0, ...}) = 0

This is because 9p server is always using a path name based lstat() call
which fails as soon as the file got removed. So to fix this, use fstat()
whenever we have an open file descriptor already.

Fixes: 00ede4c2529b ("virtio-9p: getattr server implementation...")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/103
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <4c41ad47f449a5cc8bfa9285743e029080d5f324.1732465720.git.qemu_oss@crudebyte.com>

4 months ago9pfs: remove obsolete comment in v9fs_getattr()
Christian Schoenebeck [Sun, 24 Nov 2024 15:06:40 +0000 (16:06 +0100)]
9pfs: remove obsolete comment in v9fs_getattr()

The comment claims that we'd only support basic Tgetattr fields. This is
no longer true, so remove this comment.

Fixes: e06a765efbe3 ("hw/9pfs: Add st_gen support in getattr reply")
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <fb364d12045217a4c6ccd0dd6368103ddb80698b.1732465720.git.qemu_oss@crudebyte.com>

4 months agotests/9p: add missing Rgetattr response name
Christian Schoenebeck [Sun, 24 Nov 2024 14:49:55 +0000 (15:49 +0100)]
tests/9p: add missing Rgetattr response name

'Tgetattr' 9p request and its 'Rgetattr' response types are already used
by test client, however this response type is yet missing in function
rmessage_name(), so add it.

Fixes: a6821b828404 ("tests/9pfs: compare QIDs in fs_walk_none() test")
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <e183da80d390cfd7d55bdbce92f0ff6e3e5cdced.1732465720.git.qemu_oss@crudebyte.com>

4 months agotests/9p: fix Rreaddir response name
Christian Schoenebeck [Sun, 24 Nov 2024 13:34:31 +0000 (14:34 +0100)]
tests/9p: fix Rreaddir response name

All 9p response types are prefixed with an "R", therefore fix
"READDIR" -> "RREADDIR" in function rmessage_name().

Fixes: 4829469fd9ff ("tests/virtio-9p: added readdir test")
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <daad7af58b403aaa2487c566032beca36664b30e.1732465720.git.qemu_oss@crudebyte.com>

4 months agotests/9p: add 'use-after-unlink' test
Christian Schoenebeck [Wed, 21 Feb 2024 14:13:13 +0000 (15:13 +0100)]
tests/9p: add 'use-after-unlink' test

After removing a file from the file system, we should still be able to
work with the file if we already had it open before removal.

As a first step we verify that it is possible to write to an unlinked
file, as this is what already works. This test is extended later on
after having fixed other use cases after unlink that are not working
yet.

Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <3d6449d4df25bcdd3e807eff169f46f1385e5257.1732465720.git.qemu_oss@crudebyte.com>

4 months ago9pfs: cleanup V9fsFidState
Christian Schoenebeck [Thu, 21 Nov 2024 10:52:48 +0000 (11:52 +0100)]
9pfs: cleanup V9fsFidState

Drop V9fsFidState's 'next' member, which is no longer used since:

  f5265c8f917e ('9pfs: use GHashTable for fid table')

Fixes: f5265c8f917e ('9pfs: use GHashTable for fid table')
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <E1tE4v2-0051EH-Ni@kylie.crudebyte.com>

4 months agoscsi: megasas: Internal cdbs have 16-byte length
Guenter Roeck [Tue, 28 Feb 2023 17:11:29 +0000 (09:11 -0800)]
scsi: megasas: Internal cdbs have 16-byte length

Host drivers do not necessarily set cdb_len in megasas io commands.
With commits 6d1511cea0 ("scsi: Reject commands if the CDB length
exceeds buf_len") and fe9d8927e2 ("scsi: Add buf_len parameter to
scsi_req_new()"), this results in failures to boot Linux from affected
SCSI drives because cdb_len is set to 0 by the host driver.
Set the cdb length to its actual size to solve the problem.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Fiona Ebner <f.ebner@proxmox.com>
Link: https://lore.kernel.org/r/20230228171129.4094709-1-linux@roeck-us.net
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 months agohvf: complete 1G page support
Alexander Graf [Thu, 20 Apr 2023 22:52:58 +0000 (00:52 +0200)]
hvf: complete 1G page support

Hvf on x86 only supported 2MiB large pages, but never bothered to strip
out the 1GiB page size capability from -cpu host. With QEMU 8.0.0 this
became a problem because OVMF started to use 1GiB pages by default.

Let's just unconditionally add 1GiB page walk support to the walker.

With this fix applied, I can successfully run OVMF again.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1603
Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reported-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Reported-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Phil Dennis-Jordan <phil@philjordan.eu>
Link: https://lore.kernel.org/r/20230420225258.58009-1-agraf@csgraf.de
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 months agoamd_iommu: Fix kvm_enable_x2apic link error with clang in non-KVM builds
Sairaj Kodilkar [Thu, 14 Nov 2024 11:45:09 +0000 (17:15 +0530)]
amd_iommu: Fix kvm_enable_x2apic link error with clang in non-KVM builds

Commit b12cb3819 (amd_iommu: Check APIC ID > 255 for XTSup) throws
linking error for the `kvm_enable_x2apic` when kvm is disabled
and Clang is used for compilation.

This issue comes up because Clang does not remove the function callsite
(kvm_enable_x2apic in this case) during optimization when if condition
have variable. Intel IOMMU driver solves this issue by creating separate
if condition for checking variables, which causes call site being
optimized away by virtue of `kvm_irqchip_is_split()` being defined as 0.
Implement same solution for the AMD driver.

Fixes: b12cb3819baf (amd_iommu: Check APIC ID > 255 for XTSup)
Signed-off-by: Sairaj Kodilkar <sarunkod@amd.com>
Signed-off-by: Santosh Shukla <santosh.shukla@amd.com>
Tested-by: Phil Dennis-Jordan <phil@philjordan.eu>
Link: https://lore.kernel.org/r/20241114114509.15350-1-sarunkod@amd.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
4 months agoMerge tag 'for_upstream' of https://git.kernel.org/pub/scm/virt/kvm/mst/qemu into...
Peter Maydell [Thu, 28 Nov 2024 10:50:20 +0000 (10:50 +0000)]
Merge tag 'for_upstream' of https://git./virt/kvm/mst/qemu into staging

virtio,pc,pci: bug fixes, new test

Some small bug fixes, notably a fix for a regression
in cpu hotplug after migration. I also included a
new test, just to help make sure we don't regress cxl.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# -----BEGIN PGP SIGNATURE-----
#
# iQFDBAABCAAtFiEEXQn9CHHI+FuUyooNKB8NuNKNVGkFAmdHJRIPHG1zdEByZWRo
# YXQuY29tAAoJECgfDbjSjVRpTkQIAJYFpFTPRnO8mA6gurfheB7Jt0ywAMrjKWfg
# uEkfZXXSQeCS8NBNPoZt7S8AE6xHE2a4b5lNWiS4a4coFmgTjtKPM8YsU01riyRk
# EasRxynGua2XGUWGK93r9L27v9zGz/vRC0Lujmw3VAUKGeL7a17KzmxwXLXe+DzS
# PgcI/H5hqoCDalT8aF6wOEDaWIHeo4dauDubYavW/+yaPtUvmy9MBkXbIV4iYqT5
# V6geeYIKW/yE/GHxxXOw/RE1FgpiZhebtQP26jPTSk0z/JaV5S0DNYs07joXmbaU
# fW5LSLgH3+oDI/GIhvsZ6hP87rVXBdaAogeJqT8SsuChBR55TpY=
# =B7KB
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 27 Nov 2024 13:56:34 GMT
# gpg:                using RSA key 5D09FD0871C8F85B94CA8A0D281F0DB8D28D5469
# gpg:                issuer "mst@redhat.com"
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>" [full]
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>" [full]
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17  0970 C350 3912 AFBE 8E67
#      Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA  8A0D 281F 0DB8 D28D 5469

* tag 'for_upstream' of https://git.kernel.org/pub/scm/virt/kvm/mst/qemu:
  vhost: fail device start if iotlb update fails
  bios-tables-test: Add data for complex numa test (GI, GP etc)
  bios-tables-test: Add complex SRAT / HMAT test for GI GP
  bios-tables-test: Allow for new acpihmat-generic-x test data.
  qapi/qom: Change Since entry for AcpiGenericPortProperties to 9.2
  hw/acpi: Fix size of HID in build_append_srat_acpi_device_handle()
  qapi: fix device-sync-config since-version
  hw/cxl: Check for zero length features in cmd_features_set_feature()
  tests/acpi: update expected blobs
  Revert "hw/acpi: Make CPUs ACPI `presence` conditional during vCPU hot-unplug"
  Revert "hw/acpi: Update ACPI `_STA` method with QOM vCPU ACPI Hotplug states"
  qtest: allow ACPI DSDT Table changes
  vhost_net: fix assertion triggered by batch of host notifiers processing

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agoMerge tag 'pull-request-2024-11-27' of https://gitlab.com/thuth/qemu into staging
Peter Maydell [Wed, 27 Nov 2024 13:35:54 +0000 (13:35 +0000)]
Merge tag 'pull-request-2024-11-27' of https://gitlab.com/thuth/qemu into staging

* Two small doc updates
* Fix the flaky loongarch64 and sh4 functional tests
* Refuse to compile with old XCode versions that don't work anymore
* Remove an unused function from PCI code

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEJ7iIR+7gJQEY8+q5LtnXdP5wLbUFAmdG9WcRHHRodXRoQHJl
# ZGhhdC5jb20ACgkQLtnXdP5wLbVwTRAAgOWsUiqscFk7x+G1EcxE21Cii1s3kj2t
# BYjzsl+sHzu2/aiazYfRbYOwUoXDU5kshfA0Df5KIQs2o1VBFt+v9z5vEAMjZ1tT
# j3LfazCUb4pmwS4wiD1DeEgrbCzrQ/B5grvyJsf7eB9pIxnGt+R1FGdpyDoKv8HB
# fTm6tkWKf6Mn8vJm/S+dJkoZu4ToBIOl2Q/s3/NxSf6a07FvY1QERCHndKGI90kE
# 0jodq+lb1UotaqicmlenSY33FGWWH2aeiSWGnsZRzDPLedmTBvY6hnkHtB2u+0kF
# D13ZF8oxc1pdviMXu978YRdKHmQacQ0RZVoJ6UxEf7tsDI4dEI87enVTBueGgv3z
# iz4wDDg6i/5kn12eOkjv3YgzcSF37kS/4aHltxyyuZBVT1tBsZ++zz5eCfTcO9wa
# mYElg1KDmWjmzs/HwtRb/07Bws7TIxxBUICKCA39GFpAKq6mssNAWujmhtB18AAC
# 8+tyCOz1KbseJR1oO2+I5biWBLOfAtHrBdSYIIM7yn+FWXC81kYhmT+/6hhkvl0u
# OHamRK8Kmpi71wRkm+3C/J+A1XDirXywteNIpsJXaAIEdSFzWRZdVMKblaHtqxET
# DwMdYRZ9u45phg5b+lTbDLBgsSf93NnrqyyIG7/6mS2CLwQ1x6o+NnleIJluNxSC
# kROh/xwUydo=
# =6+P3
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 27 Nov 2024 10:33:11 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-2024-11-27' of https://gitlab.com/thuth/qemu:
  hw/pci: Remove unused pci_irq_pulse() method
  tests/functional: Remove sleep workarounds from sh4 test
  .gitlab-ci.d/cirrus: Remove the wrong CPU and RAM settings from the macOS job
  meson.build: Refuse XCode versions < v15.0
  tests/functional: Fix the running test case causes loongarch64 to hang
  docs: Document that hvf on Arm is supported
  docs/devel/testing/functional: Clarify that we have to use the build folder

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 months agohw/pci: Remove unused pci_irq_pulse() method
Philippe Mathieu-Daudé [Fri, 22 Nov 2024 10:34:18 +0000 (11:34 +0100)]
hw/pci: Remove unused pci_irq_pulse() method

Last use of pci_irq_pulse() was removed 7 years ago in commit
5e9aa92eb1 ("hw/block: Fix pin-based interrupt behaviour of NVMe").

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20241122103418.539-1-philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months agotests/functional: Remove sleep workarounds from sh4 test
Cédric Le Goater [Fri, 22 Nov 2024 14:18:27 +0000 (15:18 +0100)]
tests/functional: Remove sleep workarounds from sh4 test

These were introduced in the avocado tests to workaround read issues
when interacting with console. They are no longer necessary and we can
use the expected login string instead.

Test always passes now. Remove skipUnless test on QEMU_TEST_FLAKY_TESTS.

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20241122141827.2039984-1-clg@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months ago.gitlab-ci.d/cirrus: Remove the wrong CPU and RAM settings from the macOS job
Thomas Huth [Mon, 25 Nov 2024 12:43:42 +0000 (13:43 +0100)]
.gitlab-ci.d/cirrus: Remove the wrong CPU and RAM settings from the macOS job

The macOS runner ignores them and always uses 4 CPUs and 12 GiB of
RAM, so remove our setting to avoid wrong expectations.

Message-ID: <20241125124342.187594-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 months agomeson.build: Refuse XCode versions < v15.0
Thomas Huth [Tue, 26 Nov 2024 08:10:54 +0000 (09:10 +0100)]
meson.build: Refuse XCode versions < v15.0

According to our support policy, we only support the two latest
major versions of macOS, and we already removed compatibility code
for older versions. However, it's still possible that people install
an older version of XCode on a recent version of macOS - which won't
be able to compile QEMU anymore, see for example the ticket here:

 https://gitlab.com/qemu-project/qemu/-/issues/2694

Thus let's set the expectations right and refuse older versions of
XCode that do not match the two latest versions of macOS anymore.

Message-ID: <20241126081054.244365-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>