[CRIU] ppc64le soft dirty tracking on 4.14

[Laurent Dufour]

On 04/10/2017 01:01, Andrei Vagin wrote:
> On Mon, Sep 25, 2017 at 06:08:26PM +0200, Adrian Reber wrote:
>> Hello Laurent,
>>
>> I had a closer look and soft dirty tracking works correctly. I was just
>> confused by the page size. The file was 65536 and not the expected
>> 4096.
>>
>> CRIU still cannot detect if CONFIG_MEM_SOFT_DIRTY is enabled or not.
>> The check always returns true and the criu pre-dump command is always
>> successful. In the case where CONFIG_MEM_SOFT_DIRTY=n CRIU still does
>> pre-dumps but it seems to always dump the complete memory.

I don't think this is a good idea to change the value of _PAGE_SOFT_DIRTY
based on CONFIG_MEM_SOFT_DIRTY.

If processings are done on the soft dirty bit while CONFIG_MEM_SOFT_DIRTY
is not set, this is a bug.

Short looking to the code shows that the pte_*soft_dirty() services are
define only when CONFIG_HAVE_ARCH_SOFT_DIRTY is set.

#ifdef CONFIG_HAVE_ARCH_SOFT_DIRTY
static inline bool pte_soft_dirty(pte_t pte)
{
        return !!(pte_raw(pte) & cpu_to_be64(_PAGE_SOFT_DIRTY));
}

static inline pte_t pte_mksoft_dirty(pte_t pte)
{
        return __pte(pte_val(pte) | _PAGE_SOFT_DIRTY);
}

static inline pte_t pte_clear_soft_dirty(pte_t pte)
{
        return __pte(pte_val(pte) & ~_PAGE_SOFT_DIRTY);
}
#endif /* CONFIG_HAVE_ARCH_SOFT_DIRTY */

So I guess the issue happens when CONFIG_MEM_SOFT_DIRTY=n and
CONFIG_HAVE_ARCH_SOFT_DIRTY=y, so there should be a path where a check  to
CONFIG_MEM_SOFT_DIRTY is missing.

> I think it is a bug in a kernel and it can be fixed by a patch like
> this:
> 
> 
> --- a/arch/powerpc/include/asm/book3s/64/pgtable.h
> +++ b/arch/powerpc/include/asm/book3s/64/pgtable.h
> @@ -78,7 +78,12 @@
>   */
>  #define _PAGE_PA_MAX           53
> 
> +#ifdef CONFIG_MEM_SOFT_DIRTY
>  #define _PAGE_SOFT_DIRTY       _RPAGE_SW3 /* software: software dirty tracking */
> +#elif
> +#define _PAGE_SOFT_DIRTY       0
> +#endif
> +
>  #define _PAGE_SPECIAL          _RPAGE_SW2 /* software: special page */
>  #define _PAGE_DEVMAP           _RPAGE_SW1 /* software: ZONE_DEVICE page */
>  #define __HAVE_ARCH_PTE_DEVMAP
> 
> I haven't try to test it, just got an idea from arch/x86/include/asm/pgtable_types.h.
> 
> Without this patch, the kernel has some logic about _PAGE_SOFT_DIRTY,
> even when CONFIG_MEM_SOFT_DIRTY isn't set.
> 
> 
>>
>> 'criu/criu check --feature mem_dirty_track' is always successful.
>>
>> Disabling CONFIG_PPC_RADIX_MMU makes also no difference to me.
>>
>> So it seems the check for CONFIG_MEM_SOFT_DIRTY in CRIU on ppc64le is
>> broken. The kernel does that right thing.
>>
>> 		Adrian
>> _______________________________________________
>> CRIU mailing list
>> CRIU at openvz.org
>> https://lists.openvz.org/mailman/listinfo/criu
>