Re: [RFC] Tagging of vmalloc pages for supporting the pmalloc allocator

[Igor Stoppa]

On 09/08/17 02:15, Jerome Glisse wrote:
> On Tue, Aug 08, 2017 at 03:59:36PM +0300, Igor Stoppa wrote:

[...]

>> I am tempted to add
>>
>> #define VM_PMALLOC		0x00000100

[...]

> VM_PMALLOC sounds fine to me also adding a comment there pointing to
> pmalloc documentation would be a good thing to do. The above are flags
> that are use only inside vmalloc context and so there is no issue
> here of conflicting with other potential user.

ok, will do

>>
>> Unless it's acceptable to check the private field in the page struct.
>> It would bear the pmalloc magic number.
> 
> I thought you wanted to do:
>   check struct page mapping field
>   check vmap->flags for VM_PMALLOC
> 
> bool is_pmalloc(unsigned long addr)
> {
>     struct page *page;
>     struct vm_struct *vm_struct;
> 
>     if (!is_vmalloc_addr(addr))
>         return false;
>     page = vmalloc_to_page(addr);
>     if (!page)
>         return false;
>     if (page->mapping != pmalloc_magic_key)

page->private  ?
I thought mapping would not work in the cases you mentioned?

>         return false;
> 
>     vm_struct = find_vm_area(addr);
>     if (!vm_struct)
>         return false;
> 
>     return vm_struct->flags & VM_PMALLOC;
> }
> 
> Did you change your plan ?

No, the code I have is almost 1:1 what you wrote.
Apart from mapping <=> private

In my previous mail I referred to page->private.

Maybe I was not very clear in what I wrote, but I'm almost 100% aligned
with your snippet.

>> I'm thinking to use a pointer to one of pmalloc data items, as signature.
> 
> What ever is easier for you. Note that dereferencing such pointer before
> asserting this is really a pmalloc page would be hazardous.

Yes, it's not even needed in this scenario.
It was just a way to ensure that it would be a value that cannot be come
out accidentally as pointer value, since it is already taken.

--
igor