Re: [PATCH] x86, xsave: fix non-lazy allocation of the xsave area

[H. Peter Anvin]

On 04/13/2011 03:58 AM, Hans Rosenfeld wrote:
> A single static xsave area just for init is not enough, since there are
> more user processes that are directly executed by kernel threads. Use
> fpu_alloc(), and SIGKILL the process if that fails.
> 
> Signed-off-by: Hans Rosenfeld <hans.rosenfeld@xxxxxxx>
> ---
>  arch/x86/include/asm/i387.h |    9 +++------
>  1 files changed, 3 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/x86/include/asm/i387.h b/arch/x86/include/asm/i387.h
> index 989c0ac..833b6f1 100644
> --- a/arch/x86/include/asm/i387.h
> +++ b/arch/x86/include/asm/i387.h
> @@ -329,15 +329,12 @@ static inline void fpu_copy(struct fpu *dst, struct fpu *src)
>  }
>  
>  extern void fpu_finit(struct fpu *fpu);
> -static union thread_xstate __init_xstate, *init_xstate = &__init_xstate;
>  
>  static inline void fpu_clear(struct fpu *fpu)
>  {
> -	if (!fpu_allocated(fpu)) {
> -		BUG_ON(init_xstate == NULL);
> -		fpu->state = init_xstate;
> -		init_xstate = NULL;
> -	}
> +	if (!fpu_allocated(fpu) && fpu_alloc(fpu))
> +		do_group_exit(SIGKILL);
> +
>  	memset(fpu->state, 0, xstate_size);
>  	fpu_finit(fpu);
>  	set_used_math();

Ideally this should be done earlier, while it is still possible to
ENOMEM the exec.  Specifically, it probably should be done from a new
arch hook at the top in flush_old_exec().  I'm not sure how much it
matters in practice, because if we are that memory-constrained we'll
probably die shortly anyway, and to a kernel thread it is probably not
that much of a difference if the exec'd process dies with SIGKILL or if
it gets ENOMEM from the exec() -- it will typically be visible only from
the parent thread anyway.

	-hpa

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/