Re: [PATCH v5 04/27] x86/fpu/xstate: Add XSAVES system states for shadow stack

[Matthew Wilcox]

On Thu, Nov 08, 2018 at 01:48:54PM -0800, Dave Hansen wrote:
> On 11/8/18 1:22 PM, Andy Lutomirski wrote:
> >> +struct cet_kernel_state {
> >> +       u64 kernel_ssp; /* kernel shadow stack */
> >> +       u64 pl1_ssp;    /* ring-1 shadow stack */
> >> +       u64 pl2_ssp;    /* ring-2 shadow stack */
> >> +} __packed;
> >> +
> > Why are these __packed?  It seems like it'll generate bad code for no
> > obvious purpose.
> 
> It's a hardware-defined in-memory structure.  Granted, we'd need a
> really wonky compiler to make that anything *other* than a nicely-packed
> 24-byte structure, but the __packed makes it explicit.
> 
> It is probably a really useful long-term thing to stop using __packed
> and start using "__hw_defined" or something that #defines down to __packed.

packed doesn't mean "don't leave gaps".  It means:

'packed'
     The 'packed' attribute specifies that a variable or structure field
     should have the smallest possible alignment--one byte for a
     variable, and one bit for a field, unless you specify a larger
     value with the 'aligned' attribute.

So Andy's right.  It tells the compiler, "this struct will not be naturally aligned, it will be aligned to a 1-byte boundary".  Which is silly.  If we have

struct b {
	unsigned long x;
} __packed;

struct a {
	char c;
	struct b b;
};

we want struct b to start at offset 8, but with __packed, it will start
at offset 1.

Delete __packed.  It doesn't do what you think it does.