Re: [PATCH v3 2/6] x86/microcode/core: Take a snapshot before and after applying microcode

[Ashok Raj <ashok.raj@intel.com>]

On Wed, Jan 04, 2023 at 07:56:52PM +0100, Borislav Petkov wrote:
> On Tue, Jan 03, 2023 at 10:02:08AM -0800, Ashok Raj wrote:
> > Fixes: 1008c52c09dc ("x86/CPU: Add a microcode loader callback")
> 
> Why a Fixes tag? Do you have a failure scenario for current kernels?

When we reload the same microcode there is no change in CPUID bits, but if
the kernel has turned off a feature, in my case it was SGX. So bsp copy has
SGX off, but new get_cpu_cap() seems to think SGX is there since this is
unfiltered by the kernel. 

This results in a miscompare. I have noticed even when i load a brand new
patch but has no change in CPUID, report there might have been a change in
CPUID.

> 
> If so, then it would need stable backporting.
> 
> If so, it would need the previous patch too.

I think so, but your call.


> 
> > diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
> > index 387578049de0..ac2e67156b9b 100644
> > --- a/arch/x86/include/asm/processor.h
> > +++ b/arch/x86/include/asm/processor.h
> > @@ -697,6 +697,7 @@ bool xen_set_default_idle(void);
> >  #endif
> >  
> >  void __noreturn stop_this_cpu(void *dummy);
> > +void microcode_store_cpu_caps(struct cpuinfo_x86 *info);
> 
> s/microcode_store_cpu_caps/store_cpu_caps/g

Yes, i'll change.

> 
> >  void microcode_check(struct cpuinfo_x86 *info);
> >  
> >  enum l1tf_mitigations {
> > diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
> > index b9c7529c920e..7c86c6fd07ae 100644
> > --- a/arch/x86/kernel/cpu/common.c
> > +++ b/arch/x86/kernel/cpu/common.c
> > @@ -2297,28 +2297,43 @@ void cpu_init_secondary(void)
> >  #endif
> >  
> >  #ifdef CONFIG_MICROCODE_LATE_LOADING
> > +
> > +void microcode_store_cpu_caps(struct cpuinfo_x86 *info)
> > +{
> > +	/* Reload CPUID max function as it might've changed. */
> 
> Might've changed how?

This comment existed in the previous microcode_check(). I just moved it
around during the refactor.

I suppose new microcode can bring new features and this max function
enumeration can also change?

> 
> > +	info->cpuid_level = cpuid_eax(0);
> > +
> > +	/*
> > +	 * Copy all capability leafs to pick up the synthetic ones so that
> > +	 * memcmp() below doesn't fail on that...
> 
> split that comment and put the second part...
> 
> > +	 */
> > +	memcpy(info->x86_capability, &boot_cpu_data.x86_capability,
> > +	       sizeof(info->x86_capability));
> > +
> 
> ... here:

Will do.

> 
> 	/*
> 	 * ... the ones coming from CPUID will get overwritten here:
> 	 */
> 
> > +	get_cpu_cap(info);
> > +}
> > +
> >  /*
> >   * The microcode loader calls this upon late microcode load to recheck features,
> >   * only when microcode has been updated. Caller holds microcode_mutex and CPU
> >   * hotplug lock.
> >   */
> > -void microcode_check(struct cpuinfo_x86 *info)
> > +void microcode_check(struct cpuinfo_x86 *orig)
> 					   ^^^^^
> 
> Yeah, what dhansen said.

Yes, I'm changing that as well.