Re: [tip:x86/urgent] x86/boot/compressed/64: Fix boot on machines with broken E820 table

[Kirill A. Shutemov]

On Mon, Aug 26, 2019 at 09:15:39AM +0200, Borislav Petkov wrote:
> On Sun, Aug 25, 2019 at 10:33:15PM -0500, Gustavo A. R. Silva wrote:
> > Hi all,
> > 
> > On 8/19/19 9:16 AM, tip-bot for Kirill A. Shutemov wrote:
> > [..]
> > > 
> > > diff --git a/arch/x86/boot/compressed/pgtable_64.c b/arch/x86/boot/compressed/pgtable_64.c
> > > index 5f2d03067ae5..2faddeb0398a 100644
> > > --- a/arch/x86/boot/compressed/pgtable_64.c
> > > +++ b/arch/x86/boot/compressed/pgtable_64.c
> > > @@ -72,6 +72,8 @@ static unsigned long find_trampoline_placement(void)
> > >  
> > >  	/* Find the first usable memory region under bios_start. */
> > >  	for (i = boot_params->e820_entries - 1; i >= 0; i--) {
> > > +		unsigned long new;
> > > +
> > >  		entry = &boot_params->e820_table[i];
> > >  
> > >  		/* Skip all entries above bios_start. */
> > > @@ -84,15 +86,20 @@ static unsigned long find_trampoline_placement(void)
> > >  
> > >  		/* Adjust bios_start to the end of the entry if needed. */
> > >  		if (bios_start > entry->addr + entry->size)
> > 
> > Notice that if this condition happens to be false, we end up with an
> > uninitialized variable *new*.
> 
> Yap, good catch.

:facepalm:

> > What would be the right value to assign to *new* at declaration under
> > this condition?
> 
> Looking at the changed flow of the loop, how we use new instead of
> bios_start and how we assign new back to bios_start, I think we should
> do:
> 
> 		unsigned long new = bios_start;
> 
> at the beginning...

Right.

What about this: