Re: [PATCH v13 2/5] fs/proc/task_mmu: Implement IOCTL to get and optionally clear info about PTEs

[Muhammad Usama Anjum <usama.anjum@xxxxxxxxxxxxx>]

On Apr 18, 2023 12:42 AM, Peter Xu <peterx@xxxxxxxxxx> wrote:

Muhammad,

On Mon, Apr 17, 2023 at 05:56:27PM +0500, Muhammad Usama Anjum wrote:
> +static int pagemap_scan_pmd_entry(pmd_t *pmd, unsigned long start,
> +   unsigned long end, struct mm_walk *walk)
> +{
> + struct pagemap_scan_private *p = walk->private;
> + struct vm_area_struct *vma = walk->vma;
> + unsigned long addr = end;
> + pte_t *pte, *orig_pte;
> + spinlock_t *ptl;
> + bool is_written;
> + int ret = 0;
> +
> +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> + ptl = pmd_trans_huge_lock(pmd, vma);
> + if (ptl) {
> + unsigned long n_pages = (end - start)/PAGE_SIZE;
> +
> + if (p->max_pages && n_pages > p->max_pages - p->found_pages)
> + n_pages = p->max_pages - p->found_pages;
> +
> + is_written = !is_pmd_uffd_wp(*pmd);
> +
> + /*
> + * Break huge page into small pages if the WP operation need to
> + * be performed is on a portion of the huge page.
> + */
> + if (is_written && PM_SCAN_DO_UFFD_WP(p) &&
> +     n_pages < HPAGE_SIZE/PAGE_SIZE) {
> + spin_unlock(ptl);
> + split_huge_pmd(vma, pmd, start);
> + goto process_smaller_pages;
> + }
> +
> + ret = pagemap_scan_output(is_written, vma->vm_file,
> +   pmd_present(*pmd), is_swap_pmd(*pmd),
> +   p, start, n_pages);
> +
> + if (ret >= 0 && is_written && PM_SCAN_DO_UFFD_WP(p))
> + make_uffd_wp_pmd(vma, addr, pmd);
> +
> + spin_unlock(ptl);
> + return ret;
> + }
> +process_smaller_pages:
> + if (pmd_trans_unstable(pmd))
> + return 0;
> +#endif
> +
> + pte = pte_offset_map_lock(vma->vm_mm, pmd, start, &ptl);
> + for (addr = start; addr < end && !ret; pte++, addr += PAGE_SIZE) {
> + is_written = !is_pte_uffd_wp(*pte);
> +
> + ret = pagemap_scan_output(is_written, vma->vm_file,
> +   pte_present(*pte), is_swap_pte(*pte),
> +   p, addr, 1);
> +
> + if (ret >= 0 && is_written && PM_SCAN_DO_UFFD_WP(p))
> + make_uffd_wp_pte(vma, addr, pte);
> + }
> + pte_unmap_unlock(orig_pte, ptl);

IIUC tlb flushes, mmu notifications are still missing here, am I right?

Tlb flushes and mmu notifications have been moves to do_pagemap_scan(). We flush and generate notifications only one time, instead of doing these on granular ranges.

Thanks,

> +
> + cond_resched();
> + return ret;
> +}

--
Peter Xu