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

[Yu-cheng Yu]

On Fri, 2020-02-28 at 13:17 +0100, Borislav Petkov wrote:
> On Thu, Feb 27, 2020 at 02:52:12PM -0800, Yu-cheng Yu wrote:
> > > 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.
> 
> Ok, then pls use test_and_set_thread_flag().
> 
> Also, in talking to Sebastian about this on IRC, he raised a valid
> concern: if we are going to save supervisor states here, then
> copy_xregs_to_kernel() should better save *only* supervisor states
> because we're not interested in the user states - they're going to be
> overwritten with the states from the stack.

Yes, saving only supervisor states is optimal, but doing XSAVES with a
partial RFBM changes xcomp_bv.

Yu-cheng