On 03/05/2018 01:37 PM, Khalid Aziz wrote:
>> How big can this storage get, btw? Superficially it seems like it might
>> be able to be gigantic for a large, sparse VMA.
>>
> Tags are stored only for the pages being swapped out, not for the pages
> in entire vma. Each tag storage page can hold tags for 128 pages (each
> page has 128 4-bit tags, hence 64 bytes are needed to store tags for an
> entire page allowing each page to store tags for 128 pages). Sparse VMA
> does not cause any problems since holes do not have corresponding pages
> that will be swapped out. Tag storage pages are freed once all the pages
> they store tags for have been swapped back in, except for a small number
> of pages (maximum of 8) marked for emergency tag storage.
With a linear scan holding a process-wide spinlock? If you have a fast
swap device, does this become the bottleneck when swapping ADI-tagged
memory?
FWIW, this tag storage is complex and subtle enough code that it
deserves to be in its own well-documented patch, not buried in a
thousand-line patch.
> +tag_storage_desc_t *find_tag_store(struct mm_struct *mm,
> + struct vm_area_struct *vma,
> + unsigned long addr)
> +{
> + tag_storage_desc_t *tag_desc = NULL;
> + unsigned long i, max_desc, flags;
> +
> + /* Check if this vma already has tag storage descriptor
> + * allocated for it.
> + */
> + max_desc = PAGE_SIZE/sizeof(tag_storage_desc_t);
> + if (mm->context.tag_store) {
> + tag_desc = mm->context.tag_store;
> + spin_lock_irqsave(&mm->context.tag_lock, flags);
> + for (i = 0; i < max_desc; i++) {
> + if ((addr >= tag_desc->start) &&
> + ((addr + PAGE_SIZE - 1) <= tag_desc->end))
> + break;
> + tag_desc++;
> + }
> + spin_unlock_irqrestore(&mm->context.tag_lock, flags);
--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
