Re: [PATCH V3] mm: Standardize printing for pgtable entries

[David Hildenbrand (Arm)]

On 7/13/26 04:37, Anshuman Khandual wrote:
> 
> 
> On 11/07/26 10:28 AM, Matthew Wilcox wrote:
>> On Thu, Jul 09, 2026 at 10:13:34AM +0530, Anshuman Khandual wrote:
>>> @@ -532,34 +575,34 @@ static void __print_bad_page_map_pgtable(struct mm_struct *mm, unsigned long add
>>>  	 * see locking requirements for print_bad_page_map().
>>>  	 */
>>>  	pgdp = pgd_offset(mm, addr);
>>> -	pgdv = pgd_val(*pgdp);
>>> +	ptval_to_str(pgd_str, pgd_val(*pgdp));
>>>  
>>>  	if (!pgd_present(*pgdp) || pgd_leaf(*pgdp)) {
>>> -		pr_alert("pgd:%08llx\n", pgdv);
>>> +		pr_alert("pgd:%s\n", pgd_str);
>>
>> Why can't we do this as:
>>
>> 		pr_alert("pgd:%*phN\n", sizeof(pgdv), &pgdv);
> 
> %*phN dumps raw memory bytes and will have different formats on little
> vs big endian platforms which is not desired here. Debug print here is
> expected to report pgtable entry canonical value where various platform
> specific fields can be inspected.

If we'd want to go that route I guess we'd have to add a new specifier that will
print the pointed number in the size specified by the second argument. We could
start by supporting 8/16/32/64/128 bit only, no weird stuff in between.

-- 
Cheers,

David