Re: [RFC][PATCH 11/34] x86/cpu/intel: Prepare MKTME for "address configuration" infrastructure

[Kirill A. Shutemov]

On Fri, Feb 23, 2024 at 08:22:16AM -0800, Dave Hansen wrote:
> On 2/23/24 03:33, Kirill A. Shutemov wrote:
> > On Thu, Feb 22, 2024 at 10:39:41AM -0800, Dave Hansen wrote:
> >> From: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> >>
> >> Intel also does memory encryption and also fiddles with the physical
> >> address bits.  This is currently called for *each* CPU, but practically
> >> only done on the boot CPU because of 'mktme_status'.
> >>
> >> Move it from the "each CPU" ->c_init() function to ->c_bsp_init() where
> >> the whole thing only gets called once ever.  This also necessitates moving
> >> detect_tme() and its entourage around in the file.
> > The state machine around mktme_state doesn't make sense if we only call it
> > on boot CPU, so detect_tme() can be drastically simplified. I can do it on
> > top of the patchset.
> 
> That would be great.  Looking at it again, the (tme_activate !=
> tme_activate_cpu0) block is total cruft now.  It probably just needs to
> get moved to secondary CPU startup.

I have never saw the check to be useful. I think it can be just dropped.

The patch below makes detect_tme() only enumerate TME and MKTME. And
report number of keyid bits. Kernel doesn't care about anything else.

Any comments?