Re: [PATCH v2 8/8] x86/fpu/xstate: Restore supervisor xstates for __fpu__restore_sig()

[Yu-cheng Yu]

On Fri, 2020-02-21 at 18:58 +0100, Borislav Petkov wrote:
> On Tue, Jan 21, 2020 at 12:18:43PM -0800, Yu-cheng Yu wrote:
> > [...]
> >  arch/x86/kernel/fpu/signal.c | 26 +++++++++++++++++++++-----
> >  1 file changed, 21 insertions(+), 5 deletions(-)
> > 
> > diff --git a/arch/x86/kernel/fpu/signal.c b/arch/x86/kernel/fpu/signal.c
> > index e3781a4a52a8..0d3e06a772b0 100644
> > --- a/arch/x86/kernel/fpu/signal.c
> > +++ b/arch/x86/kernel/fpu/signal.c
> > @@ -327,14 +327,22 @@ static int __fpu__restore_sig(void __user *buf, void __user *buf_fx, int size)
> >  	}
> >  
> >  	/*
> > -	 * The current state of the FPU registers does not matter. By setting
> > -	 * TIF_NEED_FPU_LOAD unconditionally it is ensured that the our xstate
> > -	 * is not modified on context switch and that the xstate is considered
> > +	 * Supervisor xstates are not modified by user space input, and
> > +	 * need to be saved and restored.  Save the whole because doing
> > +	 * partial XSAVES changes xcomp_bv.
> > +	 * By setting TIF_NEED_FPU_LOAD it is ensured that our xstate is
> > +	 * not modified on context switch and that the xstate is considered
> 
> Reflow those comments to 80 cols. There's room to the right.
> 
> >  	 * to be loaded again on return to userland (overriding last_cpu avoids
> >  	 * the optimisation).
> >  	 */
> > -	set_thread_flag(TIF_NEED_FPU_LOAD);
> > +	fpregs_lock();
> > +	if (!test_thread_flag(TIF_NEED_FPU_LOAD)) {
> > +		if (xfeatures_mask_supervisor())
> > +			copy_xregs_to_kernel(&fpu->state.xsave);
> > +		set_thread_flag(TIF_NEED_FPU_LOAD);
> 
> So the code sets TIF_NEED_FPU_LOAD unconditionally, why are you changing
> this?
> 
> Why don't you simply do:
> 
> 		set_thread_flag(TIF_NEED_FPU_LOAD);
> 		fpregs_lock();
> 		if (xfeatures_mask_supervisor())
> 			copy_xregs_to_kernel(&fpu->state.xsave);
> 		fpregs_unlock();

If TIF_NEED_FPU_LOAD is set, then xstates are already in the xsave buffer. 
We can skip saving them again.

Yu-cheng