We already have cpu_abort() to dump CPU states and abort.

Restrict hw_error() to peripheral errors, hoping we can completely
remove it by proper functions from "error-report.h" in the future.

Suggested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
---
 softmmu/cpus.c | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/softmmu/cpus.c b/softmmu/cpus.c
index a802e899abb..c96a04d7f18 100644
--- a/softmmu/cpus.c
+++ b/softmmu/cpus.c
@@ -913,16 +913,11 @@ static void stop_tcg_kick_timer(void)
 void hw_error(const char *fmt, ...)
 {
     va_list ap;
-    CPUState *cpu;
 
     va_start(ap, fmt);
     fprintf(stderr, "qemu: hardware error: ");
     vfprintf(stderr, fmt, ap);
     fprintf(stderr, "\n");
-    CPU_FOREACH(cpu) {
-        fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
-        cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
-    }
     va_end(ap);
     abort();
 }
-- 
2.26.2

On 01/09/2020 13.23, Philippe Mathieu-Daudé wrote:
> We already have cpu_abort() to dump CPU states and abort.
> 
> Restrict hw_error() to peripheral errors, hoping we can completely
> remove it by proper functions from "error-report.h" in the future.
> 
> Suggested-by: Thomas Huth <thuth@redhat.com>

IIRC I rather suggested to rename the function to "cpu_hw_error" and
only use it for real CPU problems...
But I think your approach here is fine as well. Please replace the
"Suggested-by" with "Reviewed-by" now :-)

 Thomas


> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> ---
>  softmmu/cpus.c | 5 -----
>  1 file changed, 5 deletions(-)
> 
> diff --git a/softmmu/cpus.c b/softmmu/cpus.c
> index a802e899abb..c96a04d7f18 100644
> --- a/softmmu/cpus.c
> +++ b/softmmu/cpus.c
> @@ -913,16 +913,11 @@ static void stop_tcg_kick_timer(void)
>  void hw_error(const char *fmt, ...)
>  {
>      va_list ap;
> -    CPUState *cpu;
>  
>      va_start(ap, fmt);
>      fprintf(stderr, "qemu: hardware error: ");
>      vfprintf(stderr, fmt, ap);
>      fprintf(stderr, "\n");
> -    CPU_FOREACH(cpu) {
> -        fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
> -        cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
> -    }

You could argue that cpu_abort() only prints the state of one CPU and
not of all. But I doubt that dumping the state of *all* CPUs is really
helpful in any of the contexts where hw_error() is used. So I think it's
fine to remove this CPU_FOREACH loop here.

>      va_end(ap);
>      abort();
>  }
> 

On 9/2/20 8:33 AM, Thomas Huth wrote:
> On 01/09/2020 13.23, Philippe Mathieu-Daudé wrote:
>> We already have cpu_abort() to dump CPU states and abort.
>>
>> Restrict hw_error() to peripheral errors, hoping we can completely
>> remove it by proper functions from "error-report.h" in the future.
>>
>> Suggested-by: Thomas Huth <thuth@redhat.com>
> 
> IIRC I rather suggested to rename the function to "cpu_hw_error" and
> only use it for real CPU problems...
> But I think your approach here is fine as well. Please replace the
> "Suggested-by" with "Reviewed-by" now :-)
> 
>  Thomas
> 
> 
>> Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
>> ---
>>  softmmu/cpus.c | 5 -----
>>  1 file changed, 5 deletions(-)
>>
>> diff --git a/softmmu/cpus.c b/softmmu/cpus.c
>> index a802e899abb..c96a04d7f18 100644
>> --- a/softmmu/cpus.c
>> +++ b/softmmu/cpus.c
>> @@ -913,16 +913,11 @@ static void stop_tcg_kick_timer(void)
>>  void hw_error(const char *fmt, ...)
>>  {
>>      va_list ap;
>> -    CPUState *cpu;
>>  
>>      va_start(ap, fmt);
>>      fprintf(stderr, "qemu: hardware error: ");
>>      vfprintf(stderr, fmt, ap);
>>      fprintf(stderr, "\n");
>> -    CPU_FOREACH(cpu) {
>> -        fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
>> -        cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
>> -    }
> 
> You could argue that cpu_abort() only prints the state of one CPU and
> not of all. But I doubt that dumping the state of *all* CPUs is really
> helpful in any of the contexts where hw_error() is used. So I think it's
> fine to remove this CPU_FOREACH loop here.

I find this not very practical when using more than 8 vCPUs.
Personally I stopped looking at this output because it is too noisy.

Are there developers interested in having cpu_abort() dumping all
vCPU registers? I can move that there. Or the interested ones can
also do it later if they find it useful :)

> 
>>      va_end(ap);
>>      abort();
>>  }
>>
> 
>