Re: [PATCH v15 17/23] x86/kexec: Flush cache of TDX private memory

[Huang, Kai]

On Mon, 2023-11-27 at 19:33 +0000, Huang, Kai wrote:
> On Mon, 2023-11-27 at 10:13 -0800, Hansen, Dave wrote:
> > On 11/9/23 03:55, Kai Huang wrote:
> > ...
> > > --- a/arch/x86/kernel/reboot.c
> > > +++ b/arch/x86/kernel/reboot.c
> > > @@ -31,6 +31,7 @@
> > >  #include <asm/realmode.h>
> > >  #include <asm/x86_init.h>
> > >  #include <asm/efi.h>
> > > +#include <asm/tdx.h>
> > >  
> > >  /*
> > >   * Power off function, if any
> > > @@ -741,6 +742,20 @@ void native_machine_shutdown(void)
> > >  	local_irq_disable();
> > >  	stop_other_cpus();
> > >  #endif
> > > +	/*
> > > +	 * stop_other_cpus() has flushed all dirty cachelines of TDX
> > > +	 * private memory on remote cpus.  Unlike SME, which does the
> > > +	 * cache flush on _this_ cpu in the relocate_kernel(), flush
> > > +	 * the cache for _this_ cpu here.  This is because on the
> > > +	 * platforms with "partial write machine check" erratum the
> > > +	 * kernel needs to convert all TDX private pages back to normal
> > > +	 * before booting to the new kernel in kexec(), and the cache
> > > +	 * flush must be done before that.  If the kernel took SME's way,
> > > +	 * it would have to muck with the relocate_kernel() assembly to
> > > +	 * do memory conversion.
> > > +	 */
> > > +	if (platform_tdx_enabled())
> > > +		native_wbinvd();
> > 
> > Why can't the TDX host code just set host_mem_enc_active=1?
> > 
> > Sure, you'll end up *using* the SME WBINVD support, but then you don't
> > have two different WBINVD call sites.  You also don't have to mess with
> > a single line of assembly.
> 
> I wanted to avoid changing the assembly.
> 
> Perhaps the comment isn't very clear.  Flushing cache (on the CPU running kexec)
> when the host_mem_enc_active=1 is handled in the relocate_kernel() assembly,
> which happens at very late stage right before jumping to the new kernel. 
> However for TDX when the platform has erratum we need to convert TDX private
> pages back to normal, which must be done after flushing cache.  If we reuse
> host_mem_enc_active=1, then we will need to change the assembly code to do that.
> 

Forgot to say doing TDX page conversion in the relocate_assembly() isn't easy
because the cache flushing when host_mem_enc_active=1 happens after kernel has
switched to the identity mapping table, so we will need to do hacks like fixing
up symbol address etc.