Re: [PATCH v2] xen/arm64: Remove READ/WRITE_SYSREG32 helper macros

[Michal Orzel <michal.orzel@xxxxxxx>]

On 12.07.2021 10:33, Julien Grall wrote:
> 
> 
> On 12/07/2021 07:26, Michal Orzel wrote:
>> Hi Julien,
> 
> Hi Michal,
> 
>> On 09.07.2021 17:21, Julien Grall wrote:
>>> Hi Michal,
>>>
>>> On 09/07/2021 13:40, Michal Orzel wrote:
>>>> AArch64 system registers are 64bit whereas AArch32 ones
>>>> are 32bit or 64bit. MSR/MRS are expecting 64bit values thus
>>>> we should get rid of helpers READ/WRITE_SYSREG32
>>>> in favour of using READ/WRITE_SYSREG.
>>>>
>>>> The last place in code making use of READ/WRITE_SYSREG32
>>>> on arm64 is in TVM_REG macro defining functions vreg_emulate_<register>.
>>>> Implement a macro WRITE_SYSREG_SZ which expands as follows:
>>>> -on arm64: WRITE_SYSREG
>>>> -on arm32: WRITE_SYSREG{32/64}
>>>>
>>>> As there are no other places in the code using these helpers
>>>> on arm64 - remove them.
>>>>
>>>> Signed-off-by: Michal Orzel <michal.orzel@xxxxxxx>
>>>> ---
>>>> Changes since v1:
>>>> -implement WRITE_SYSREG_SZ instead of duplicating the TVM_REG
>>>> ---
>>>>    xen/arch/arm/vcpreg.c               | 12 +++++++++++-
>>>>    xen/include/asm-arm/arm64/sysregs.h |  4 ----
>>>>    2 files changed, 11 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/xen/arch/arm/vcpreg.c b/xen/arch/arm/vcpreg.c
>>>> index f0cdcc8a54..10c4846954 100644
>>>> --- a/xen/arch/arm/vcpreg.c
>>>> +++ b/xen/arch/arm/vcpreg.c
>>>> @@ -47,6 +47,16 @@
>>>>     *
>>>>     */
>>>>    +#ifdef CONFIG_ARM_64
>>>> +#define WRITE_SYSREG_SZ(sz, val, sysreg)     WRITE_SYSREG(val, sysreg)
>>>
>>> I think you want to cast to (uint##sz##_t) to avoid any surprise. I think...
>>>
>> But the val will always be of type uint##sz##_t because it is passed from ...
>>>> +#else
>>>> +/*
>>>> + * WRITE_SYSREG{32/64} on arm32 is defined as variadic macro which imposes
>>>> + * on the below macro to be defined like that as well.
>>>> + */
>>>> +#define WRITE_SYSREG_SZ(sz, val, sysreg...)  WRITE_SYSREG##sz(val, sysreg)
>>>> +#endif
>>>> +
>>>>    /* The name is passed from the upper macro to workaround macro 
>>>> expansion. */
>>>>    #define TVM_REG(sz, func, reg...)                                       
>>>>     \
>>>>    static bool func(struct cpu_user_regs *regs, uint##sz##_t *r, bool 
>>>> read)    \
>>>> @@ -55,7 +65,7 @@ static bool func(struct cpu_user_regs *regs, 
>>>> uint##sz##_t *r, bool read)    \
>> ... here(*r argument).
>> So I do not think that such casting makes sense e.g
>> uint##sz##_t foo;
>> WRITE_SYSREG((uint##sz##_t)foo, bar);
> 
> I agree that this doesn't make sense for the *current* use. However, when 
> writing code, we need to think how one could use it in the future...
> 
> When I read the name, I would expect that if I call it with sz == 32, then 
> bottom 32-bit value to be written and the top 32-bit zeroed. But this is not 
> the case...
> 
> You are relying on the developper and reviewer to notice that only 32-bit 
> value should be passed when sz == 32.
> 
> I am not particularly keen on relying on this when a simple cast could 
> prevent any future misuse. Alternatively, I would be happy to consider 
> checking the type of the value at build time.
> 
> Cheers,
> 

Ok, you are right that it does not cost much to add it, so let's stick with:
#ifdef CONFIG_ARM_64
#define WRITE_SYSREG_SZ(sz, val, sysreg) WRITE_SYSREG((uint##sz##_t)val, sysreg)
#else

I will push v3 with this change.

Cheers,
Michal