Re: [PATCH v7 10/12] crypto: x86/aesni - Use the proper data type in struct aesni_xts_ctx

[Chang S. Bae]

On 5/25/2023 11:54 PM, Eric Biggers wrote:
> On Wed, May 24, 2023 at 09:57:15AM -0700, Chang S. Bae wrote:
> > +static inline unsigned long aes_align_addr(unsigned long addr)
> > +{
> > +	return (crypto_tfm_ctx_alignment() >= AESNI_ALIGN) ?
> > +	       ALIGN(addr, 1) : ALIGN(addr, AESNI_ALIGN);
> > +}
>
> Wouldn't it be simpler to make this take and return 'void *'?  Then the callers
> wouldn't need to cast to/from 'unsigned long'.
>
> Also, aligning to a 1-byte boundary is a no-op.
>
> So, please consider the following:
>
> static inline void *aes_align_addr(void *addr)
> {
> 	if (crypto_tfm_ctx_alignment() >= AES_ALIGN)
> 		return addr;
> 	return PTR_ALIGN(addr, AES_ALIGN);
> }
>
> Also, aesni_rfc4106_gcm_ctx_get() and generic_gcmaes_ctx_get() should call this
> too, rather than duplicating similar code.

Indeed, your suggestion can improve the overall code there.

Thanks!
Chang