Re: [RFC][PATCH -mm 4/4] Hibernation: Use temporary page tables for kernel text mapping on x86_64

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Wednesday, 22 August 2007 10:29, Pavel Machek wrote:
> Hi!
> 
> > Use temporary page tables for the kernel text mapping during hibernation restore
> > on x86_64.
> > 
> > Without the patch, the original boot kernel's page tables that represent the
> > kernel text mapping are used while the core of the image kernel is being
> > restored.  However, in principle, if the boot kernel is not identical to the
> > image kernel, the location of these page tables in the image kernel need not be
> > the same, so we should create a safe copy of the kernel text mapping prior to
> > restoring the core of the image kernel.
> 
> 
> > @@ -190,25 +190,42 @@ static int res_phys_pud_init(pud_t *pud,
> >  	return 0;
> >  }
> >  
> > +static int res_kernel_text_pud_init(pud_t *pud, unsigned long start)
> > +{
> > +	pmd_t *pmd;
> > +	unsigned long paddr;
> > +
> > +	pmd = (pmd_t *)get_safe_page(GFP_ATOMIC);
> > +	if (!pmd)
> > +		return -ENOMEM;
> > +	set_pud(pud + pud_index(start), __pud(__pa(pmd) | _KERNPG_TABLE));
> > +	for (paddr = 0; paddr < KERNEL_TEXT_SIZE; pmd++, paddr += PMD_SIZE) {
> > +		unsigned long pe;
> > +
> > +		pe = __PAGE_KERNEL_LARGE_EXEC | _PAGE_GLOBAL | paddr;
> > +		pe &= __supported_pte_mask;
> > +		set_pmd(pmd, __pmd(pe));
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> >  static int set_up_temporary_mappings(void)
> >  {
> >  	unsigned long start, end, next;
> > +	pud_t *pud;
> >  	int error;
> >  
> >  	temp_level4_pgt = (pgd_t *)get_safe_page(GFP_ATOMIC);
> >  	if (!temp_level4_pgt)
> >  		return -ENOMEM;
> >  
> > -	/* It is safe to reuse the original kernel mapping */
> > -	set_pgd(temp_level4_pgt + pgd_index(__START_KERNEL_map),
> > -		init_level4_pgt[pgd_index(__START_KERNEL_map)]);
> > -
> >  	/* Set up the direct mapping from scratch */
> >  	start = (unsigned long)pfn_to_kaddr(0);
> >  	end = (unsigned long)pfn_to_kaddr(end_pfn);
> >  
> >  	for (; start < end; start = next) {
> > -		pud_t *pud = (pud_t *)get_safe_page(GFP_ATOMIC);
> > +		pud = (pud_t *)get_safe_page(GFP_ATOMIC);
> >  		if (!pud)
> >  			return -ENOMEM;
> 
> I believe we leak temp_level4_pgt here.
> 
> > +	/* Set up the kernel text mapping from scratch */
> > +	pud = (pud_t *)get_safe_page(GFP_ATOMIC);
> > +	if (!pud)
> > +		return -ENOMEM;
> 
> ...and some more memory here :-(. 

No, these pages will be freed by swsusp_free(), thanks to the get_safe_page().

I'm not _that_ careless. ;-)

Greetings,
Rafael