Re: [PATCH v2] s390/cpacf: Unpoison instruction results

On 8/13/26 16:06, Harald Freudenberger wrote:
> On 2026-08-07 14:20, Ilya Leoshkevich wrote:
>> Stop KMSAN from complaining about CPACF outputs being uninitialized.
>>
>> Do not unpoison variable-length parameter blocks: mapping function
>> codes (like CPACF_KIMD_SHA_256) to lengths is ugly. Let the callers
>> handle this once need arises.
>>
>> Signed-off-by: Ilya Leoshkevich <[email protected]>
>> ---
>> v1:
>> https://lore.kernel.org/linux-s390/20260806190223.3292580-1- 
>> [email protected]/
>> v1 -> v2: Unpoison outputs of more instructions (Sashiko).
>>
> 
> See my comment below.
> 
>>  arch/s390/include/asm/cpacf.h | 7 +++++++
>>  1 file changed, 7 insertions(+)
>>
>> diff --git a/arch/s390/include/asm/cpacf.h b/arch/s390/include/asm/ 
>> cpacf.h
>> index a83683169d98d..13d6a6344802a 100644
>> --- a/arch/s390/include/asm/cpacf.h
>> +++ b/arch/s390/include/asm/cpacf.h
>> @@ -301,6 +301,7 @@ static __always_inline void __cpacf_query(unsigned
>> int opcode,
>>                        cpacf_mask_t *mask)
>>  {
>>      __cpacf_query_insn(opcode, mask, CPACF_FC_QUERY);
>> +    kmsan_unpoison_memory(mask, sizeof(*mask));
>>  }
>>
>>  static __always_inline int __cpacf_check_opcode(unsigned int opcode)
>> @@ -370,6 +371,7 @@ static __always_inline int
>> cpacf_query_func(unsigned int opcode,
>>  static __always_inline void __cpacf_qai(unsigned int opcode, 
>> cpacf_qai_t *qai)
>>  {
>>      __cpacf_query_insn(opcode, qai, CPACF_FC_QUERY_AUTH_INFO);
>> +    kmsan_unpoison_memory(qai, sizeof(*qai));
>>  }
>>
>>  /**
>> @@ -422,6 +424,7 @@ static inline int cpacf_km(unsigned long func, 
>> void *param,
>>            [opc] "i" (CPACF_KM)
>>          : "cc", "memory", "0", "1");
>>
>> +    kmsan_unpoison_memory(dest, src_len - s.odd);
>>      return src_len - s.odd;
>>  }
>>
>> @@ -454,6 +457,7 @@ static inline int cpacf_kmc(unsigned long func, 
>> void *param,
>>            [opc] "i" (CPACF_KMC)
>>          : "cc", "memory", "0", "1");
>>
>> +    kmsan_unpoison_memory(dest, src_len - s.odd);
>>      return src_len - s.odd;
>>  }
>>
>> @@ -587,6 +591,7 @@ static inline int cpacf_kmctr(unsigned long func,
>> void *param, u8 *dest,
>>            [opc] "i" (CPACF_KMCTR)
>>          : "cc", "memory", "0", "1");
>>
>> +    kmsan_unpoison_memory(dest, src_len - s.odd);
>>      return src_len - s.odd;
>>  }
>>
>> @@ -619,6 +624,7 @@ static inline void cpacf_prno(unsigned long func,
>> void *param,
>>          : [fc] "d" (func), [pba] "d" ((unsigned long)param),
>>            [seed] "d" (s.pair), [opc] "i" (CPACF_PRNO)
>>          : "cc", "memory", "0", "1");
>> +    kmsan_unpoison_memory(dest, dest_len);
>>  }
>>
>>  /**
>> @@ -731,6 +737,7 @@ static inline void cpacf_kma(unsigned long func,
>> void *param, u8 *dest,
>>          : [fc] "d" (func), [pba] "d" ((unsigned long)param),
>>            [opc] "i" (CPACF_KMA)
>>          : "cc", "memory", "0", "1");
>> +    kmsan_unpoison_memory(dest, src_len);
> 
> I am unhappy with this kma thing. This would only work properly with
> CC 0. But CC 1 and 2 are not handled properly. However, I had a closer
> look onto the AR for kma and ... well this inline function should
> handle CC 1 and CC 2 somehow and it should similar to the other inline
> functions return the number of bytes processed like for example cpacf_km().
> I'll open up a bugzilla to fix this kma inline function.
> If you leave out this kma hunk you can add my R-b.

Will do, thanks!

May I ask you to add a proper KMSAN annotation while fixing CC1/CC2
cases? Alternatively I can do it myself afterwards, of course.

> 
>>  }
>>
>>  #endif    /* _ASM_S390_CPACF_H */