Re: [PATCH v8 04/22] s390/zcrypt: Integrate ap_asm.h into include/asm/ap.h.

[Harald Freudenberger]

On 10.08.2018 10:49, Cornelia Huck wrote:
> On Thu, 9 Aug 2018 12:06:56 -0400
> Tony Krowiak <akrowiak@xxxxxxxxxxxxx> wrote:
>
>> On 08/09/2018 05:17 AM, Harald Freudenberger wrote:
>>> On 09.08.2018 11:06, Cornelia Huck wrote:  
>>>> On Wed,  8 Aug 2018 10:44:14 -0400
>>>> Tony Krowiak <akrowiak@xxxxxxxxxxxxxxxxxx> wrote:
>>>>  
>>>>> From: Harald Freudenberger <freude@xxxxxxxxxx>
>>>>>
>>>>> Move all the inline functions from the ap bus header
>>>>> file ap_asm.h into the in-kernel api header file
>>>>> arch/s390/include/asm/ap.h so that KVM can make use
>>>>> of all the low level AP functions.
>>>>>
>>>>> Signed-off-by: Harald Freudenberger <freude@xxxxxxxxxx>
>>>>> Signed-off-by: Christian Borntraeger <borntraeger@xxxxxxxxxx>  
>>>> You should add your own s-o-b if you are sending on patches written by
>>>> others (even if it does not matter in the end, when they are merged
>>>> through a different path anyway.)
>>>>  
>>>>> ---
>>>>>   arch/s390/include/asm/ap.h     |  284 ++++++++++++++++++++++++++++++++++++----
>>>>>   drivers/s390/crypto/ap_asm.h   |  261 ------------------------------------
>>>>>   drivers/s390/crypto/ap_bus.c   |   21 +---
>>>>>   drivers/s390/crypto/ap_bus.h   |    1 +
>>>>>   drivers/s390/crypto/ap_card.c  |    1 -
>>>>>   drivers/s390/crypto/ap_queue.c |    1 -
>>>>>   6 files changed, 259 insertions(+), 310 deletions(-)
>>>>>   delete mode 100644 drivers/s390/crypto/ap_asm.h
>>>>>
>>>>> diff --git a/arch/s390/include/asm/ap.h b/arch/s390/include/asm/ap.h
>>>>> index c1bedb4..046e044 100644
>>>>> --- a/arch/s390/include/asm/ap.h
>>>>> +++ b/arch/s390/include/asm/ap.h
>>>>> @@ -47,6 +47,50 @@ struct ap_queue_status {
>>>>>   };
>>>>>   
>>>>>   /**
>>>>> + * ap_intructions_available() - Test if AP instructions are available.
>>>>> + *
>>>>> + * Returns 0 if the AP instructions are installed.  
>>>> Stumbled over this when I was looking at the usage in patch 7: if I see
>>>> a function called '_available' return 0, I'd assume that whatever the
>>>> function tests for is *not* available.
>>>>
>>>> Rather call this function ap_instructions_check_availability() (and
>>>> keep the return code convention), or switch this to return 0 if not
>>>> available and !0 if available?  
>>> Good catch, Cony you are right. I'll fix this to return 1 if AP instructions
>>> are available and 0 if not. However, this patch will come via Martin's pipe
>>> to the Linus Torwald kernel sources.  
>> Is your intent to simply indicate whether the AP instructions are 
>> available or
>> not; or is the intention to indicate whether the AP instructions are 
>> available
>> and if not, they why? In the former, then I agree that a boolean should be
>> returned; however, if the case is the latter, then what you have is fine but
>> maybe the function name should be changed as Connie suggests.
> So, can this actually fail for any reason other than "instructions not
> installed"? Even if it did, the end result is that the instructions are
> not usable -- I don't think the "why" would be interesting at that
> point.
I can not think of any other reason why the PQAP(TAPQ) would fail
other than the AP instructions are not available at all. However,
the old implementation returned -ENODEV on failure and 0 on
success. The new implementation now returns 1 for success
and 0 for failure. This is just one of a couple of functions related
to ap xxx available. I'll rework them all to return a boolean value
soon.
>
>>>>> + */
>>>>> +static inline int ap_instructions_available(void)
>>>>> +{
>>>>> +	register unsigned long reg0 asm ("0") = AP_MKQID(0, 0);
>>>>> +	register unsigned long reg1 asm ("1") = -ENODEV;
>>>>> +	register unsigned long reg2 asm ("2");
>>>>> +
>>>>> +	asm volatile(
>>>>> +		"   .long 0xb2af0000\n"		/* PQAP(TAPQ) */
>>>>> +		"0: la    %0,0\n"
>>>>> +		"1:\n"
>>>>> +		EX_TABLE(0b, 1b)
>>>>> +		: "+d" (reg1), "=d" (reg2)
>>>>> +		: "d" (reg0)
>>>>> +		: "cc");
>>>>> +	return reg1;
>>>>> +}