Re: [Xen-devel] [PATCH 4/4] x86/boot: Size the boot/directmap mappings dynamically

[Jan Beulich <jbeulich@suse.com>]

On 14.01.2020 18:27, Andrew Cooper wrote:
> On 14/01/2020 17:02, Jan Beulich wrote:
>> On 13.01.2020 18:50, Andrew Cooper wrote:
>>> --- a/xen/arch/x86/boot/head.S
>>> +++ b/xen/arch/x86/boot/head.S
>>> @@ -687,14 +687,19 @@ trampoline_setup:
>>>           * handling/walking), and identity map Xen into bootmap (needed for
>>>           * the transition into long mode), using 2M superpages.
>>>           */
>>> -        lea     sym_esi(start),%ebx
>>> -        lea     (1<<L2_PAGETABLE_SHIFT)*7+(PAGE_HYPERVISOR_RWX|_PAGE_PSE)(%ebx),%eax
>>> -        shr     $(L2_PAGETABLE_SHIFT-3),%ebx
>>> -        mov     $8,%ecx
>>> -1:      mov     %eax,sym_fs(l2_bootmap)-8(%ebx,%ecx,8)
>>> -        mov     %eax,sym_fs(l2_directmap)-8(%ebx,%ecx,8)
>>> -        sub     $(1<<L2_PAGETABLE_SHIFT),%eax
>>> -        loop    1b
>>> +        lea     sym_esi(_start), %ecx
>>> +        lea     -1 + sym_esi(_end), %edx
>> This looks pretty odd - does
>>
>>         lea     sym_esi(_end) - 1, %edx
>>
>> not work?
> 
> No:
> 
> head.S: Assembler messages:
> head.S:521: Error: junk `(%esi)-1' after expression
> 
> but it is not at all surprising when you expand the macro:
> 
> lea (_end - start)(%esi) - 1, %edx
> 
> The expression for the displacement ends up split across both sides of
> the SIB.

Hmm, seems I've mis-remembered that stuff ahead of ( and after
) gets concatenated.

>>> +        lea     _PAGE_PSE + PAGE_HYPERVISOR_RWX(%ecx), %eax /* PTE to write. */
>>> +        shr     $L2_PAGETABLE_SHIFT, %ecx                   /* First slot to write. */
>>> +        shr     $L2_PAGETABLE_SHIFT, %edx                   /* Final slot to write. */
>>> +
>>> +1:      mov     %eax, sym_offs(l2_bootmap)  (%esi, %ecx, 8)
>>> +        mov     %eax, sym_offs(l2_directmap)(%esi, %ecx, 8)
>> I guess I could have noticed this on the previous patch already:
>> This would look better as
>>
>> 1:      mov     %eax, sym_esi(l2_bootmap,   %ecx, 8)
>>         mov     %eax, sym_esi(l2_directmap, %ecx, 8)
>>
>> Can sym_esi() perhaps be made
>>
>> #define sym_esi(sym, extra...)      sym_offs(sym)(%esi, ## extra)
>>
>> ?
> 
> I considered and dismissed this approach.  Yes, the code is slightly
> shorter, but at the expense of readability.
> 
> The advantage of the longhand version is that it is obvious which half
> is the displacement expression, and which half is the SIB.
> 
> The reduced version leaves a distinct possibility of %ecx being mistaken
> as the base register, rather than the index.

With it being sym_esi() that gets used, I don't see any such risk.
But anyway, if you're convinced of the longer form being better,
so be it then.

>>> --- a/xen/arch/x86/xen.lds.S
>>> +++ b/xen/arch/x86/xen.lds.S
>>> @@ -384,6 +384,3 @@ ASSERT((trampoline_end - trampoline_start) < TRAMPOLINE_SPACE - MBI_SPACE_MIN,
>>>      "not enough room for trampoline and mbi data")
>>>  ASSERT((wakeup_stack - wakeup_stack_start) >= WAKEUP_STACK_MIN,
>>>      "wakeup stack too small")
>>> -
>>> -/* Plenty of boot code assumes that Xen isn't larger than 16M. */
>>> -ASSERT(_end - _start <= MB(16), "Xen too large for early-boot assumptions")
>> Following your reply to the cover letter, this can't be dropped just yet.
> 
> Correct.
> 
>> Even when that remaining issue got addressed, I think it would be better
>> to keep it, altering the bound to GB(1).
> 
> A 1G check wouldn't be correct.
> 
> We've already got a more suitable one, which is the check that Xen
> doesn't encroach into the stubs range.

Oh, right. If only that check was correct. I guess it ought to be
using &, not |, and perhaps also __image_base__ == XEN_VIRT_START.
I'll give this a try and send a patch unless in the course of
doing so I realize there's a reason for it being the way it is.

Jan

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