Re: [PATCH v21 2/5] fs/proc/task_mmu: Implement IOCTL to get and optionally clear info about PTEs

[Andrei Vagin <avagin@gmail.com>]

On Tue, Jun 27, 2023 at 02:00:31PM +0500, Muhammad Usama Anjum wrote:
> Hi Andrei and Michal,
> 
> Lets resolve last two points. Please reply below.
> 
> On 6/27/23 6:46 AM, Andrei Vagin wrote:
> ...
> >> +#ifdef CONFIG_HUGETLB_PAGE
> >> +static int pagemap_scan_hugetlb_entry(pte_t *ptep, unsigned long hmask,
> >> +				      unsigned long start, unsigned long end,
> >> +				      struct mm_walk *walk)
> >> +{
> >> +	unsigned long n_pages = (end - start)/PAGE_SIZE;
> >> +	struct pagemap_scan_private *p = walk->private;
> >> +	struct vm_area_struct *vma = walk->vma;
> >> +	bool is_written, is_interesting = true;
> >> +	struct hstate *h = hstate_vma(vma);
> >> +	unsigned long bitmap;
> >> +	spinlock_t *ptl;
> >> +	int ret = 0;
> >> +	pte_t ptent;
> >> +
> >> +	if (IS_PM_SCAN_WP(p->flags) && n_pages < HPAGE_SIZE/PAGE_SIZE)
> >> +		return -EINVAL;
> >> +
> >> +	if (n_pages > p->max_pages - p->found_pages)
> >> +		n_pages = p->max_pages - p->found_pages;
> >> +
> >> +	if (IS_PM_SCAN_WP(p->flags)) {
> >> +		i_mmap_lock_write(vma->vm_file->f_mapping);
> >> +		ptl = huge_pte_lock(h, vma->vm_mm, ptep);
> >> +	}
> >> +
> >> +	ptent = huge_ptep_get(ptep);
> >> +	is_written = !is_huge_pte_uffd_wp(ptent);
> >> +
> >> +	/*
> >> +	 * Partial hugetlb page clear isn't supported
> >> +	 */
> >> +	if (is_written && IS_PM_SCAN_WP(p->flags) &&
> >> +	    n_pages < HPAGE_SIZE/PAGE_SIZE) {
> > 
> > should it be done only if is_interesting is set?
> This can be good optimization. We shouldn't return error before finding if
> page is interesting. I'll update.
> 
> > 
> >> +		ret = PM_SCAN_END_WALK;
> >> +		goto unlock_and_return;
> >> +	}
> >> +
> >> +	bitmap = PM_SCAN_FLAGS(is_written, pagemap_scan_is_huge_file(ptent),
> >> +			       pte_present(ptent), is_swap_pte(ptent),
> >> +			       pte_present(ptent) && is_zero_pfn(pte_pfn(ptent)));
> >> +
> >> +	if (IS_PM_SCAN_GET(p->flags)) {
> >> +		is_interesting = pagemap_scan_is_interesting_page(bitmap, p);
> >> +		if (is_interesting)
> >> +			ret = pagemap_scan_output(bitmap, p, start, n_pages);
> >> +	}
> >> +
> >> +	if (IS_PM_SCAN_WP(p->flags) && is_written && is_interesting &&
> >> +	    ret >= 0) {
> >> +		make_uffd_wp_huge_pte(vma, start, ptep, ptent);
> >> +		flush_hugetlb_tlb_range(vma, start, end);
> >> +	}
> >> +
> >> +unlock_and_return:
> >> +	if (IS_PM_SCAN_WP(p->flags)) {
> >> +		spin_unlock(ptl);
> >> +		i_mmap_unlock_write(vma->vm_file->f_mapping);
> >> +	}
> >> +
> >> +	return ret;
> >> +}
> ...
> >> +
> >> +static long do_pagemap_scan(struct mm_struct *mm, unsigned long __arg)
> >> +{
> >> +	struct pm_scan_arg __user *uarg = (struct pm_scan_arg __user *)__arg;
> >> +	unsigned long long start, end, walk_start, walk_end;
> >> +	unsigned long empty_slots, vec_index = 0;
> >> +	struct mmu_notifier_range range;
> >> +	struct page_region __user *vec;
> >> +	struct pagemap_scan_private p;
> >> +	struct pm_scan_arg arg;
> >> +	int ret = 0;
> >> +
> >> +	if (copy_from_user(&arg, uarg, sizeof(arg)))
> >> +		return -EFAULT;
> >> +
> >> +	start = untagged_addr((unsigned long)arg.start);
> >> +	vec = (struct page_region *)untagged_addr((unsigned long)arg.vec);
> >> +
> >> +	ret = pagemap_scan_args_valid(&arg, start, vec);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	end = start + arg.len;
> >> +	p.max_pages = (arg.max_pages) ? arg.max_pages : ULONG_MAX;
> >> +	p.found_pages = 0;
> >> +	p.required_mask = arg.required_mask;
> >> +	p.anyof_mask = arg.anyof_mask;
> >> +	p.excluded_mask = arg.excluded_mask;
> >> +	p.return_mask = arg.return_mask;
> >> +	p.flags = arg.flags;
> >> +	p.flags |= ((p.required_mask | p.anyof_mask | p.excluded_mask) &
> >> +		    PAGE_IS_WRITTEN) ? PM_SCAN_REQUIRE_UFFD : 0;
> >> +	p.cur_buf.start = p.cur_buf.len = p.cur_buf.flags = 0;
> >> +	p.vec_buf = NULL;
> >> +	p.vec_buf_len = PAGEMAP_WALK_SIZE >> PAGE_SHIFT;
> >> +
> >> +	/*
> >> +	 * Allocate smaller buffer to get output from inside the page walk
> >> +	 * functions and walk page range in PAGEMAP_WALK_SIZE size chunks. As
> >> +	 * we want to return output to user in compact form where no two
> >> +	 * consecutive regions should be continuous and have the same flags.
> >> +	 * So store the latest element in p.cur_buf between different walks and
> >> +	 * store the p.cur_buf at the end of the walk to the user buffer.
> >> +	 */
> >> +	if (IS_PM_SCAN_GET(p.flags)) {
> >> +		p.vec_buf = kmalloc_array(p.vec_buf_len, sizeof(*p.vec_buf),
> >> +					  GFP_KERNEL);
> >> +		if (!p.vec_buf)
> >> +			return -ENOMEM;
> >> +	}
> >> +
> >> +	if (IS_PM_SCAN_WP(p.flags)) {
> >> +		mmu_notifier_range_init(&range, MMU_NOTIFY_PROTECTION_VMA, 0,
> >> +					mm, start, end);
> >> +		mmu_notifier_invalidate_range_start(&range);
> >> +	}
> >> +
> >> +	walk_start = walk_end = start;
> >> +	while (walk_end < end && !ret) {
> >> +		if (IS_PM_SCAN_GET(p.flags)) {
> >> +			p.vec_buf_index = 0;
> >> +
> >> +			/*
> >> +			 * All data is copied to cur_buf first. When more data
> >> +			 * is found, we push cur_buf to vec_buf and copy new
> >> +			 * data to cur_buf. Subtract 1 from length as the
> >> +			 * index of cur_buf isn't counted in length.
> >> +			 */
> >> +			empty_slots = arg.vec_len - vec_index;
> >> +			p.vec_buf_len = min(p.vec_buf_len, empty_slots - 1);
> >> +		}
> >> +
> >> +		walk_end = (walk_start + PAGEMAP_WALK_SIZE) & PAGEMAP_WALK_MASK;
> >> +		if (walk_end > end)
> >> +			walk_end = end;
> >> +
> > 
> > If this loop can run for a long time, we need to interrupt it in case of
> > pending signals.
> > 
> > If you think we don't need to do that, pls explain in the commit
> > message, so that maintainers don't miss this part and double check that
> > everything is alright here.
> This can be done. I'll add to the commit message that we are walking over
> entire range passed.
> 
> > 
> >> +		ret = mmap_read_lock_killable(mm);
> >> +		if (ret)
> > 
> > If any pages have been handled, we need to report them to user-space. It
> > isn't acceptable to return a error in such cases.
> This will return error only when task has gotten some serios signal and it
> is giong to be killed. In this scenerio, we shouldn't care about returning
> gracefully. Why do you think we should return gracefully in this case?

You are right, it can be interrupted only by a fatal signal. You can
ignore this comment.

> 
> > 
> > And we need to report an address where it stopped scanning.
> > We can do that by adding zero length vector.
> I don't want to do multiplexing the ending address in vec. Can we add
> end_addr variable in struct pm_scan_arg to always return the ending address?
> 
> struct pm_scan_arg {
> 	...
> 	_u64 end_addr;
> };
> 
> 
> > 
> > 
> >> +			goto free_data;
> >> +		ret = walk_page_range(mm, walk_start, walk_end,
> >> +				      &pagemap_scan_ops, &p);
> >> +		mmap_read_unlock(mm);
> >> +
> >> +		if (ret && ret != PM_SCAN_FOUND_MAX_PAGES &&
> >> +		    ret != PM_SCAN_END_WALK)
> >> +			goto free_data;
> >> +
> >> +		walk_start = walk_end;
> >> +		if (IS_PM_SCAN_GET(p.flags) && p.vec_buf_index) {
> >> +			if (copy_to_user(&vec[vec_index], p.vec_buf,
> >> +					 p.vec_buf_index * sizeof(*p.vec_buf))) {
> >> +				/*
> >> +				 * Return error even though the OP succeeded
> >> +				 */
> >> +				ret = -EFAULT;
> >> +				goto free_data;
> >> +			}
> >> +			vec_index += p.vec_buf_index;
> > 
> > Should we set ret to zero here if it is equal PM_SCAN_END_WALK.
> No, PM_SCAN_END_WALK is just internal code to stop the page walk and return
> immedtitely. When we get this return value, we stop this loop and return to
> user with whatever data we have in user buffer.

but PM_SCAN_END_WALK is returned when p.vec_buf is full, so we can
restart the loop after coping vec_buf to the user buffer, can't we?

> 
> > 
> >> +		}
> >> +	}
> >> +
> >> +	if (p.cur_buf.len) {
> >> +		if (copy_to_user(&vec[vec_index], &p.cur_buf, sizeof(p.cur_buf))) {
> >> +			ret = -EFAULT;
> >> +			goto free_data;
> >> +		}
> >> +		vec_index++;
> >> +	}
> >> +
> >> +	ret = vec_index;
> >> +
> >> +free_data:
> >> +	if (IS_PM_SCAN_WP(p.flags))
> >> +		mmu_notifier_invalidate_range_end(&range);
> >> +
> >> +	kfree(p.vec_buf);
> >> +	return ret;
> >> +}
> >> +
> ...
> 
> -- 
> BR,
> Muhammad Usama Anjum