Re: [PATCH 19/37] x86, mm: simplify get_user_pages() PTE bit handling

[Dave Hansen <dave@sr71.net>]

On 11/27/2015 02:12 AM, Thomas Gleixner wrote:
> On Mon, 16 Nov 2015, Dave Hansen wrote:
>> +static inline int pte_allows_gup(pte_t pte, int write)
>> +{
>> +	/*
>> +	 * 'entry_flags' can come from a pte, pmd or pud.  Make
>> +	 * sure to only check flags here that are valid *and* the
>> +	 * same value on all 3 types.  (PAT is currently the only
>> +	 * one where that is true and is not checked here).
> 
> I have a hard time to understand that comment.
> 
>        /*
>         * 'entry_flags' can come from a pte, pmd or pud. We only check
> 	* _PAGE_PRESENT, _PAGE_USER and _PAGE_RW here, which are the
> 	* same for all types.
>         */
> 
> Is that what you wanted to say?

Yeah, that's a much better way to say it.  I'll fix it up.

>> +	 */
>> +	if (!(pte_flags(pte) & (_PAGE_PRESENT|_PAGE_USER)))
>> +		return 0;
>> +	if (write && !(pte_write(pte)))
>> +		return 0;
>> +	return 1;
>> +}
>> +
>> +static inline int pmd_allows_gup(pmd_t pmd, int write)
>> +{
>> +	return pte_allows_gup(*(pte_t *)&pmd, write);
>> +}
>> +
>> +static inline int pud_allows_gup(pud_t pud, int write)
>> +{
>> +	return pte_allows_gup(*(pte_t *)&pud, write);
>> +}
>> +
> 
>>  static noinline int gup_huge_pmd(pmd_t pmd, unsigned long addr,
>>  		unsigned long end, int write, struct page **pages, int *nr)
>>  {
>> -	unsigned long mask;
>>  	struct page *head, *page;
>>  	int refs;
>>  
>> -	mask = _PAGE_PRESENT|_PAGE_USER;
>> -	if (write)
>> -		mask |= _PAGE_RW;
>> -	if ((pmd_flags(pmd) & mask) != mask)
>> +	if (!pmd_allows_gup(pmd, write))
> 
> Why do you need that extra indirection here of rereading the pmd?
> You have the pmd already.

The intention there was not to re-read the PMD, but only to cast the
structure over to a pte_t.  I expected the compiler to be able to figure
out what was going on and not actually re-read anything.

Is that a bad assumption?