Re: [RFC PATCH 07/10] x86/fpu: Rellocate fpstate on save_fpregs_to_fpstate

[Dave Hansen]

On 12/2/21 4:36 PM, Jiaxun Yang wrote:
> --- a/arch/x86/kernel/fpu/core.c
> +++ b/arch/x86/kernel/fpu/core.c
> @@ -112,6 +112,22 @@ static void update_avx_timestamp(struct fpu *fpu)
>  		fpu->avx512_timestamp = jiffies;
>  }
>  
> +/* Update xstate size if it more dynamic features are opted-in. */
> +static inline void xstate_update_size(struct fpu *fpu)
> +{
> +	int err;
> +	struct fpstate *fpstate = fpu->fpstate;
> +	u64 fpsmask = fpstate->xfeatures;
> +	u64 curmask = fpsmask | xfeatures_in_use();
> +
> +	if (fpu_state_size_dynamic()) {
> +		if (fpsmask != curmask) {
> +			err = fpstate_realloc(fpu, curmask);
> +			WARN_ON_FPU(err);
> +		}
> +	}
> +}
> +
>  /*
>   * Save the FPU register state in fpu->fpstate->regs. The register state is
>   * preserved.
> @@ -129,6 +145,7 @@ static void update_avx_timestamp(struct fpu *fpu)
>  void save_fpregs_to_fpstate(struct fpu *fpu)
>  {
>  	if (likely(use_xsave())) {
> +		xstate_update_size(fpu);
>  		os_xsave(fpu->fpstate);
>  		update_avx_timestamp(fpu);
>  		return;

Have you considered what exactly happens when you hit that WARN_ON_FPU()
which otherwise ignores the allocation error?  Have you considered what
happens on the os_xsave() that follows it immediately?  How about what
happens the next time this task runs after that failure?