On Wed, Sep 03, 2025 at 09:56:20AM -0300, Jason Gunthorpe wrote:
> On Wed, Sep 03, 2025 at 09:30:17AM +0300, Mike Rapoport wrote:
> > +int kho_preserve_vmalloc(void *ptr, phys_addr_t *preservation)
> > +{
> > + struct kho_vmalloc_chunk *chunk, *first_chunk;
> > + struct vm_struct *vm = find_vm_area(ptr);
> > + int err;
> > +
> > + if (!vm)
> > + return -EINVAL;
> > +
> > + /* we don't support HUGE_VMAP yet */
> > + if (get_vm_area_page_order(vm))
> > + return -EOPNOTSUPP;
>
> This is a compatability problem.. Should have some way to indicate
> that future kernels have an incompatible serialization so restore can
> fail..
We can add version or flags to kho_vmalloc_chunk, e.g. make it
struct kho_vmalloc_hdr {
DECLARE_KHOSER_PTR(next, struct kho_vmalloc_chunk *);
unsigned int total_pages; /* only valid in the first chunk */
unsigned short version; /* only valid in the first chunk */
unsigned short num_elms;
};
I'm thinking about actually adding support for HUGE_VMAP for the next
resping, but version/flags seems useful anyway.
> > + chunk = new_vmalloc_chunk(NULL);
> > + if (!chunk)
> > + return -ENOMEM;
> > + first_chunk = chunk;
> > + first_chunk->hdr.total_pages = vm->nr_pages;
> > +
> > + for (int i = 0; i < vm->nr_pages; i++) {
> > + phys_addr_t phys = page_to_phys(vm->pages[i]);
> > +
> > + err = kho_preserve_phys(phys, PAGE_SIZE);
>
> Don't call kho_preserve_phy if you already have a page!
Ok, I'll add kho_preserve_page() ;-P.
Now seriously, by no means this is a folio, so it's either
kho_preserve_phys() or __kho_preserve_order(). I don't mind switching to
latter, but I really see no point doing it.
> We should be getting rid of kho_preserve_phys() :(
How do you suggest to preserve memblock?
> Jason
--
Sincerely yours,
Mike.
