[Qemu-devel] [PATCH for-1.1] target-arm: Fix crash when passed "-cpu foo"

[Qemu-devel] [PATCH for-1.1] target-arm: Fix crash when passed "-cpu foo"

[Peter Maydell]

The macro definition of cpu_init meant that if cpu_arm_init()
returned NULL this wouldn't result in cpu_init() itself returning
NULL. This had the effect that "-cpu foo" for some unknown CPU
name 'foo' would cause ARM targets to segfault rather than
generating a useful error message. Fix this by making cpu_init
a simple inline function.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
I did a quick grep and I think ARM is the only target at the moment
where we've made this change to cpu_init -- is that right, Andreas?

 target-arm/cpu.h |   10 +++++++++-
 1 files changed, 9 insertions(+), 1 deletions(-)

diff --git a/target-arm/cpu.h b/target-arm/cpu.h
index 5eac070..d01285f 100644
--- a/target-arm/cpu.h
+++ b/target-arm/cpu.h
@@ -458,7 +458,15 @@ void cpu_arm_set_cp_io(CPUARMState *env, int cpnum,
 #define TARGET_PHYS_ADDR_SPACE_BITS 32
 #define TARGET_VIRT_ADDR_SPACE_BITS 32
 
-#define cpu_init(model) (&cpu_arm_init(model)->env)
+static inline CPUARMState *cpu_init(const char *cpu_model)
+{
+    ARMCPU *cpu = cpu_arm_init(cpu_model);
+    if (cpu) {
+        return &cpu->env;
+    }
+    return NULL;
+}
+
 #define cpu_exec cpu_arm_exec
 #define cpu_gen_code cpu_arm_gen_code
 #define cpu_signal_handler cpu_arm_signal_handler
-- 
1.7.1

Re: [Qemu-devel] [PATCH for-1.1] target-arm: Fix crash when passed "-cpu foo"

[Andreas Färber]

Am 09.05.2012 16:54, schrieb Peter Maydell:
> The macro definition of cpu_init meant that if cpu_arm_init()
> returned NULL this wouldn't result in cpu_init() itself returning
> NULL. This had the effect that "-cpu foo" for some unknown CPU
> name 'foo' would cause ARM targets to segfault rather than
> generating a useful error message. Fix this by making cpu_init
> a simple inline function.

Ouch.

> Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
> ---
> I did a quick grep and I think ARM is the only target at the moment
> where we've made this change to cpu_init -- is that right, Andreas?

Affirmative for master. But I need to respin my qom-cpu-reset series
(QOM CPUState part 3) and the followups I have queued.

>  target-arm/cpu.h |   10 +++++++++-
>  1 files changed, 9 insertions(+), 1 deletions(-)
> 
> diff --git a/target-arm/cpu.h b/target-arm/cpu.h
> index 5eac070..d01285f 100644
> --- a/target-arm/cpu.h
> +++ b/target-arm/cpu.h
> @@ -458,7 +458,15 @@ void cpu_arm_set_cp_io(CPUARMState *env, int cpnum,
>  #define TARGET_PHYS_ADDR_SPACE_BITS 32
>  #define TARGET_VIRT_ADDR_SPACE_BITS 32
>  
> -#define cpu_init(model) (&cpu_arm_init(model)->env)
> +static inline CPUARMState *cpu_init(const char *cpu_model)
> +{
> +    ARMCPU *cpu = cpu_arm_init(cpu_model);
> +    if (cpu) {

Personally I prefer an explicit NULL check for pointers, but it's the
same either way.

> +        return &cpu->env;
> +    }
> +    return NULL;
> +}
> +
>  #define cpu_exec cpu_arm_exec
>  #define cpu_gen_code cpu_arm_gen_code
>  #define cpu_signal_handler cpu_arm_signal_handler

Acked-by: Andreas Färber <afaerber@suse.de>

Thanks,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg