[kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[Alexandru Elisei]

The prototype for the halt() function is incorrect, because halt() doesn't
take any arguments. Fix it by using the prototype from smp.h.

Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
---
 lib/arm/io.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/lib/arm/io.c b/lib/arm/io.c
index 8226b765bdc5..6d3d7afed002 100644
--- a/lib/arm/io.c
+++ b/lib/arm/io.c
@@ -15,11 +15,10 @@
 #include <asm/psci.h>
 #include <asm/spinlock.h>
 #include <asm/io.h>
+#include <asm/smp.h>
 
 #include "io.h"
 
-extern void halt(int code);
-
 static struct spinlock uart_lock;
 /*
  * Use this guess for the uart base in order to make an attempt at
@@ -93,6 +92,6 @@ void exit(int code)
 {
 	chr_testdev_exit(code);
 	psci_system_off();
-	halt(code);
+	halt();
 	__builtin_unreachable();
 }
-- 
2.17.0

Re: [kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[Sean Christopherson]

On Mon, Apr 08, 2019 at 04:11:25PM +0100, Alexandru Elisei wrote:
> The prototype for the halt() function is incorrect, because halt() doesn't
> take any arguments. Fix it by using the prototype from smp.h.
> 
> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
> ---

Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>

Re: [kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[Andrew Jones]

On Mon, Apr 08, 2019 at 04:11:25PM +0100, Alexandru Elisei wrote:
> The prototype for the halt() function is incorrect, because halt() doesn't
> take any arguments. Fix it by using the prototype from smp.h.
> 
> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
> ---
>  lib/arm/io.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/lib/arm/io.c b/lib/arm/io.c
> index 8226b765bdc5..6d3d7afed002 100644
> --- a/lib/arm/io.c
> +++ b/lib/arm/io.c
> @@ -15,11 +15,10 @@
>  #include <asm/psci.h>
>  #include <asm/spinlock.h>
>  #include <asm/io.h>
> +#include <asm/smp.h>
>  
>  #include "io.h"
>  
> -extern void halt(int code);
> -
>  static struct spinlock uart_lock;
>  /*
>   * Use this guess for the uart base in order to make an attempt at
> @@ -93,6 +92,6 @@ void exit(int code)
>  {
>  	chr_testdev_exit(code);
>  	psci_system_off();
> -	halt(code);
> +	halt();
>  	__builtin_unreachable();
>  }
> -- 
> 2.17.0
>

I don't mind this change, because per the code it is the "correct"
thing to do. However, I was being a bit tricky here when I wrote it.
By changing the prototype to take 'code' as argument we guarantee
that 'code' will be in x0/r0 when we halt, giving us a last chance
to see it when inspecting the halted unit test state.

Anyway, like I said, I'm fine with the cleanup, but the prototype
abuse does serve a purpose - maybe just not a good enough purpose
to justify the weirdness.

Thanks,
drew

Re: [kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[Alexandru Elisei]

On 4/9/19 8:40 AM, Andrew Jones wrote:
> On Mon, Apr 08, 2019 at 04:11:25PM +0100, Alexandru Elisei wrote:
>> The prototype for the halt() function is incorrect, because halt() doesn't
>> take any arguments. Fix it by using the prototype from smp.h.
>>
>> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
>> ---
>>  lib/arm/io.c | 5 ++---
>>  1 file changed, 2 insertions(+), 3 deletions(-)
>>
>> diff --git a/lib/arm/io.c b/lib/arm/io.c
>> index 8226b765bdc5..6d3d7afed002 100644
>> --- a/lib/arm/io.c
>> +++ b/lib/arm/io.c
>> @@ -15,11 +15,10 @@
>>  #include <asm/psci.h>
>>  #include <asm/spinlock.h>
>>  #include <asm/io.h>
>> +#include <asm/smp.h>
>>  
>>  #include "io.h"
>>  
>> -extern void halt(int code);
>> -
>>  static struct spinlock uart_lock;
>>  /*
>>   * Use this guess for the uart base in order to make an attempt at
>> @@ -93,6 +92,6 @@ void exit(int code)
>>  {
>>  	chr_testdev_exit(code);
>>  	psci_system_off();
>> -	halt(code);
>> +	halt();
>>  	__builtin_unreachable();
>>  }
>> -- 
>> 2.17.0
>>
> I don't mind this change, because per the code it is the "correct"
> thing to do. However, I was being a bit tricky here when I wrote it.
> By changing the prototype to take 'code' as argument we guarantee
> that 'code' will be in x0/r0 when we halt, giving us a last chance
> to see it when inspecting the halted unit test state.
>
> Anyway, like I said, I'm fine with the cleanup, but the prototype
> abuse does serve a purpose - maybe just not a good enough purpose
> to justify the weirdness.

Now it makes sense. I didn't think it was intentional, but now that you have
mentioned it, the same pattern is used by powerpc.

Perhaps a comment explaining that having different prototypes was on purpose
would be the best solution?

>
> Thanks,
> drew

Re: [kvm-unit-tests PATCH] lib: arm: Use correct halt() prototype from smp.h

[Andrew Jones]

On Tue, Apr 09, 2019 at 10:15:59AM +0100, Alexandru Elisei wrote:
> On 4/9/19 8:40 AM, Andrew Jones wrote:
> > On Mon, Apr 08, 2019 at 04:11:25PM +0100, Alexandru Elisei wrote:
> >> The prototype for the halt() function is incorrect, because halt() doesn't
> >> take any arguments. Fix it by using the prototype from smp.h.
> >>
> >> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
> >> ---
> >>  lib/arm/io.c | 5 ++---
> >>  1 file changed, 2 insertions(+), 3 deletions(-)
> >>
> >> diff --git a/lib/arm/io.c b/lib/arm/io.c
> >> index 8226b765bdc5..6d3d7afed002 100644
> >> --- a/lib/arm/io.c
> >> +++ b/lib/arm/io.c
> >> @@ -15,11 +15,10 @@
> >>  #include <asm/psci.h>
> >>  #include <asm/spinlock.h>
> >>  #include <asm/io.h>
> >> +#include <asm/smp.h>
> >>  
> >>  #include "io.h"
> >>  
> >> -extern void halt(int code);
> >> -
> >>  static struct spinlock uart_lock;
> >>  /*
> >>   * Use this guess for the uart base in order to make an attempt at
> >> @@ -93,6 +92,6 @@ void exit(int code)
> >>  {
> >>  	chr_testdev_exit(code);
> >>  	psci_system_off();
> >> -	halt(code);
> >> +	halt();
> >>  	__builtin_unreachable();
> >>  }
> >> -- 
> >> 2.17.0
> >>
> > I don't mind this change, because per the code it is the "correct"
> > thing to do. However, I was being a bit tricky here when I wrote it.
> > By changing the prototype to take 'code' as argument we guarantee
> > that 'code' will be in x0/r0 when we halt, giving us a last chance
> > to see it when inspecting the halted unit test state.
> >
> > Anyway, like I said, I'm fine with the cleanup, but the prototype
> > abuse does serve a purpose - maybe just not a good enough purpose
> > to justify the weirdness.
> 
> Now it makes sense. I didn't think it was intentional, but now that you have
> mentioned it, the same pattern is used by powerpc.

No surprise there. I wrote that too, based on the arm code :)

> 
> Perhaps a comment explaining that having different prototypes was on purpose
> would be the best solution?

Yes, either a comment explaining the weirdness or your patch to
remove it would be a good idea in order to avoid future head
scratching. I'll send a patch that adds comments to both arm and
powerpc.

Thanks,
drew