Re: [PATCH 2/2] mm/madvise: Use set_pte() to write page tables

[Ryan Roberts <ryan.roberts@arm.com>]

On 15/12/2025 10:37, Lorenzo Stoakes wrote:
> On Thu, Dec 11, 2025 at 09:43:58AM +0000, Ryan Roberts wrote:
>> On 11/12/2025 08:11, Samuel Holland wrote:
>>> Generic code must always use the architecture-provided helper function
>>> to write page tables.
>>>
>>> Fixes: 662df3e5c376 ("mm: madvise: implement lightweight guard page mechanism")
>>> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
>>> ---
>>>
>>>  mm/madvise.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/mm/madvise.c b/mm/madvise.c
>>> index b617b1be0f535..4da9c32f8738a 100644
>>> --- a/mm/madvise.c
>>> +++ b/mm/madvise.c
>>> @@ -1114,7 +1114,7 @@ static int guard_install_set_pte(unsigned long addr, unsigned long next,
>>>  	unsigned long *nr_pages = (unsigned long *)walk->private;
>>>
>>>  	/* Simply install a PTE marker, this causes segfault on access. */
>>> -	*ptep = make_pte_marker(PTE_MARKER_GUARD);
>>> +	set_pte(ptep, make_pte_marker(PTE_MARKER_GUARD));
>>
>> No! As I explained in my response on the other thread (which you linked in the
>> cover letter), it is correct as is and should not be changed to set_pte().
> 
> Yup agreed, esp. given this is my code :)
> 
> Also some arches don't define set_pte()... it seems set_xxx() functions not
> really intended to be used outside of arch code - see
> e.g. https://elixir.bootlin.com/linux/v6.18.1/A/ident/set_pte
> 
>>
>> Copy/pasting my explanation:
>>
>> | I tried "fixing" this before. But it's correct as is. ptep is pointing to a
>> | value on the stack. See [2].
>> |
>> | https://lore.kernel.org/linux-mm/2308a4d0-273e-4cf8-9c9f-3008c42b6d18@arm.com/
>>
>> If you go look at where this function is called from, you'll see that it's a
>> pointer to a stack variable:
>>
>>
>> ---8<---
>> static int walk_pte_range_inner(pte_t *pte, unsigned long addr,
>> 				unsigned long end, struct mm_walk *walk)
>> {
>> 	const struct mm_walk_ops *ops = walk->ops;
>> 	int err = 0;
>>
>> 	for (;;) {
>> 		if (ops->install_pte && pte_none(ptep_get(pte))) {
>> 			pte_t new_pte;
>>
>> 			err = ops->install_pte(addr, addr + PAGE_SIZE, &new_pte,
>> 					       walk);
>> ---8<---
>>
>> I agree that it's extremely confusing. Perhaps, at a minimum, we should come up
>> with some kind of naming convention for this and update this and the other
>> couple of places that pass pointers to stack-based pXX_t around?
>>
>> e.g. instead of calling it "ptep", call it "ptevalp" or something like that?
> 
> Not sure that'd clarify, we already have a bit of an inconsistent mess with all
> this :(
> 
> Given it's a stack variable I'm not sure using a helper is in any way helpful
> other than I suppose to account for people grepping around for incorrect page
> table manipulation code?

I've proposed an approach to clean all of this up. I'd appreciate your opinion
if you get a few mins:

https://lore.kernel.org/all/a063f6c5-2785-4a9f-8079-25edb3e54cef@arm.com/

> 
>>
>> Thanks,
>> Ryan
>>
>>
>>>  	(*nr_pages)++;
>>>
>>>  	return 0;
>>
> 
> Cheers, Lorenzo