Re: [Xen-devel] [PATCH 1/3] x86/boot: Remove cached CPUID data from the trampoline

[Jan Beulich <jbeulich@xxxxxxxx>]

On 19.11.2019 16:15, Andrew Cooper wrote:
> On 13/11/2019 13:29, Jan Beulich wrote:
>> On 13.11.2019 14:22, Andrew Cooper wrote:
>>> I am not convinced the behaviour is worth changing, and I don't have
>>> time for this scope creep.
>> There's no scope creep here at all.
> 
> Yes - it really is scope creep.
> 
> This patch does not change the behaviour of Xen in the case of poor
> virtualisation of the bit.  Xen will still crash either way.

So I have to apologize. What I didn't notice is

        if (disable & MSR_IA32_MISC_ENABLE_XD_DISABLE) {
                write_efer(read_efer() | EFER_NX);
                printk(KERN_INFO
                       "re-enabled NX (Execute Disable) protection\n");
        }

in early_init_intel(). I simply didn't expect we'd already have
such a blind EFER write. I therefore agree now that this is a
pre-existing bug that you don't make any worse.

Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>

> I have explained, repeatedly now, why I am not inclined to fix this. It
> is a bug which doesn't exist in practice.

I should have been looking more closely; the lack of sufficient
context did misguide me.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/mailman/listinfo/xen-devel