synchronously scan and reclaim empty user PTE pages

static unsigned long zap_pte_range(struct mmu_gather *tlb,
struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

mm/memory.c +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb,
  		pte_t ptent = ptep_get(pte);
  		int max_nr;
 
-		nr = 1;
-		if (pte_none(ptent))
-			continue;
-
  		if (need_resched())
  			break;

mm/memory.c +0 -0

  		if (pte_present(ptent)) {
  			nr = zap_present_ptes(tlb, vma, pte, ptent, max_nr,

Quoted 13 lines

  static unsigned long zap_pte_range(struct mmu_gather *tlb, struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

mm/memory.c +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct 
 mmu_gather *tlb,
           pte_t ptent = ptep_get(pte);
           int max_nr;
-        nr = 1;
-        if (pte_none(ptent))
-            continue;
-
           if (need_resched())

mm/memory.c static unsigned long zap_pte_range(struct · line 1717

Quoted 10 lines

mm/memory.c +8 -0

               break;
+        nr = skip_none_ptes(pte, addr, end);
+        if (nr) {
+            addr += PAGE_SIZE * nr;
+            if (addr == end)
+                break;
+            pte += nr;
+        }
+
           max_nr = (end - addr) / PAGE_SIZE;

I dislike calculating max_nr twice, once here and once in skip_non_ptes.

Further, you're missing to update ptent here.

avoid another ptep_get().

Quoted 3 lines

mm/memory.c +0 -0

           if (pte_present(ptent)) {
               nr = zap_present_ptes(tlb, vma, pte, ptent, max_nr,

Quoted 26 lines · David Hildenbrand

Quoted 13 lines

  static unsigned long zap_pte_range(struct mmu_gather *tlb, struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

(quoted diff) +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
           pte_t ptent = ptep_get(pte);
           int max_nr;
-        nr = 1;
-        if (pte_none(ptent))
-            continue;
-
           if (need_resched())

Review static unsigned long zap_pte_range(struct · line 1717

Quoted 10 lines

(quoted diff) +8 -0

               break;
+        nr = skip_none_ptes(pte, addr, end);
+        if (nr) {
+            addr += PAGE_SIZE * nr;
+            if (addr == end)
+                break;
+            pte += nr;
+        }
+
           max_nr = (end - addr) / PAGE_SIZE;

I dislike calculating max_nr twice, once here and once in skip_non_ptes.

Quoted 1 line

Further, you're missing to update ptent here.

Oh, my bad. However, with [PATCH v3 5/9], there will be no problem, but there are still two ptep_get() and max_nr calculation.

If you inline it you can

Quoted 1 line

avoid another ptep_get().

Do you mean to inline the skip_none_ptes() into do_zap_pte_range()?

Quoted 31 lines · Qi Zheng

Quoted 26 lines · David Hildenbrand

Quoted 13 lines

   static unsigned long zap_pte_range(struct mmu_gather *tlb, struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

(quoted diff) +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
            pte_t ptent = ptep_get(pte);
            int max_nr;
-        nr = 1;
-        if (pte_none(ptent))
-            continue;
-
            if (need_resched())

Review static unsigned long zap_pte_range(struct · line 1717

Quoted 10 lines

(quoted diff) +8 -0

                break;
+        nr = skip_none_ptes(pte, addr, end);
+        if (nr) {
+            addr += PAGE_SIZE * nr;
+            if (addr == end)
+                break;
+            pte += nr;
+        }
+
            max_nr = (end - addr) / PAGE_SIZE;

I dislike calculating max_nr twice, once here and once in skip_non_ptes.

Quoted 1 line

Further, you're missing to update ptent here.

Quoted 7 lines

Oh, my bad. However, with [PATCH v3 5/9], there will be no problem, but there are still two ptep_get() and max_nr calculation.

If you inline it you can

Quoted 1 line

avoid another ptep_get().

Do you mean to inline the skip_none_ptes() into do_zap_pte_range()?

Effectively moving this patch after #5, and have it be something like:

mm/memory.c +15 -2

index 1949f5e0fece5..4f5d1e4c6688e 100644
@@ -1667,8 +1667,21 @@ static inline int do_zap_pte_range(struct 
 mmu_gather *tlb,
         pte_t ptent = ptep_get(pte);
         int max_nr = (end - addr) / PAGE_SIZE;
 
-       if (pte_none(ptent))
-               return 1;
+       /* Skip all consecutive pte_none(). */
+       if (pte_none(ptent)) {
+               int nr;
+
+               for (nr = 1; nr < max_nr; nr++) {
+                       ptent = ptep_get(pte + nr);
+                       if (!pte_none(ptent))
+                               break;
+               }
+               max_nr -= nr;
+               if (!max_nr)
+                       return nr;
+               pte += nr;
+               addr += nr * PAGE_SIZE;
+       }
 
         if (pte_present(ptent))

                return zap_present_ptes(tlb, vma, pte, ptent, max_nr,

In the context of this patch this makes most sense.

Regarding "count_pte_none" comment, I assume you talk about patch #7.

Can't you simply return the number of pte_none that you skipped here using another input variable, if really required?

Quoted 73 lines · David Hildenbrand

Quoted 31 lines · Qi Zheng

Quoted 26 lines · David Hildenbrand

Quoted 13 lines

   static unsigned long zap_pte_range(struct mmu_gather *tlb, struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

(quoted diff) +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
            pte_t ptent = ptep_get(pte);
            int max_nr;
-        nr = 1;
-        if (pte_none(ptent))
-            continue;
-
            if (need_resched())

Review static unsigned long zap_pte_range(struct · line 1717

Quoted 10 lines

(quoted diff) +8 -0

                break;
+        nr = skip_none_ptes(pte, addr, end);
+        if (nr) {
+            addr += PAGE_SIZE * nr;
+            if (addr == end)
+                break;
+            pte += nr;
+        }
+
            max_nr = (end - addr) / PAGE_SIZE;

I dislike calculating max_nr twice, once here and once in skip_non_ptes.

Quoted 1 line

Further, you're missing to update ptent here.

Quoted 7 lines

Oh, my bad. However, with [PATCH v3 5/9], there will be no problem, but there are still two ptep_get() and max_nr calculation.

If you inline it you can

Quoted 1 line

avoid another ptep_get().

Do you mean to inline the skip_none_ptes() into do_zap_pte_range()?

Effectively moving this patch after #5, and have it be something like:

mm/memory.c +15 -2

index 1949f5e0fece5..4f5d1e4c6688e 100644
@@ -1667,8 +1667,21 @@ static inline int do_zap_pte_range(struct
 mmu_gather *tlb,
          pte_t ptent = ptep_get(pte);
          int max_nr = (end - addr) / PAGE_SIZE;
 
-       if (pte_none(ptent))
-               return 1;
+       /* Skip all consecutive pte_none(). */
+       if (pte_none(ptent)) {
+               int nr;
+
+               for (nr = 1; nr < max_nr; nr++) {
+                       ptent = ptep_get(pte + nr);
+                       if (!pte_none(ptent))
+                               break;
+               }
+               max_nr -= nr;
+               if (!max_nr)
+                       return nr;
+               pte += nr;
+               addr += nr * PAGE_SIZE;
+       }
 
          if (pte_present(ptent))

                return zap_present_ptes(tlb, vma, pte, ptent, max_nr,

Quoted 3 lines

In the context of this patch this makes most sense.

Regarding "count_pte_none" comment, I assume you talk about patch #7.

Yes.

Quoted 1 line

Quoted 3 lines

Can't you simply return the number of pte_none that you skipped here using another
input variable, if really required?

Suppose we add an input variable nr_skip to do_zap_pte_range(), you mean to return the above nr to zap_pte_range() through:

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,

   				      rss, &force_flush, &force_break);
      if (can_reclaim_pt) {
          none_nr += count_pte_none(pte, nr);
          none_nr += nr_skip;
      }

Right?

Quoted 83 lines · Qi Zheng

Quoted 73 lines · David Hildenbrand

Quoted 31 lines · Qi Zheng

Quoted 26 lines · David Hildenbrand

Quoted 13 lines

    static unsigned long zap_pte_range(struct mmu_gather *tlb, struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

(quoted diff) +0 -4

@@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
             pte_t ptent = ptep_get(pte);
             int max_nr;
-        nr = 1;
-        if (pte_none(ptent))
-            continue;
-
             if (need_resched())

Review static unsigned long zap_pte_range(struct · line 1717

Quoted 10 lines

(quoted diff) +8 -0

                 break;
+        nr = skip_none_ptes(pte, addr, end);
+        if (nr) {
+            addr += PAGE_SIZE * nr;
+            if (addr == end)
+                break;
+            pte += nr;
+        }
+
             max_nr = (end - addr) / PAGE_SIZE;

I dislike calculating max_nr twice, once here and once in skip_non_ptes.

Quoted 1 line

Further, you're missing to update ptent here.

Quoted 7 lines

Oh, my bad. However, with [PATCH v3 5/9], there will be no problem, but there are still two ptep_get() and max_nr calculation.

If you inline it you can

Quoted 1 line

avoid another ptep_get().

Do you mean to inline the skip_none_ptes() into do_zap_pte_range()?

Effectively moving this patch after #5, and have it be something like:

mm/memory.c +15 -2

index 1949f5e0fece5..4f5d1e4c6688e 100644
@@ -1667,8 +1667,21 @@ static inline int do_zap_pte_range(struct
 mmu_gather *tlb,
           pte_t ptent = ptep_get(pte);
           int max_nr = (end - addr) / PAGE_SIZE;
 
-       if (pte_none(ptent))
-               return 1;
+       /* Skip all consecutive pte_none(). */
+       if (pte_none(ptent)) {
+               int nr;
+
+               for (nr = 1; nr < max_nr; nr++) {
+                       ptent = ptep_get(pte + nr);
+                       if (!pte_none(ptent))
+                               break;
+               }
+               max_nr -= nr;
+               if (!max_nr)
+                       return nr;
+               pte += nr;
+               addr += nr * PAGE_SIZE;
+       }
 
           if (pte_present(ptent))

                  return zap_present_ptes(tlb, vma, pte, ptent, max_nr,

Quoted 3 lines

In the context of this patch this makes most sense.

Regarding "count_pte_none" comment, I assume you talk about patch #7.

Yes.

Quoted 1 line

Quoted 6 lines

Quoted 3 lines

Can't you simply return the number of pte_none that you skipped here using another
input variable, if really required?

Suppose we add an input variable nr_skip to do_zap_pte_range(), you mean to return the above nr to zap_pte_range() through:

Maybe "cur_none_nr" or something similar.

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr, rss, &force_flush, &force_break); if (can_reclaim_pt) {
          none_nr += count_pte_none(pte, nr);
          none_nr += nr_skip;
      }

Quoted 1 line

Right?

Yes. I did not look closely at the patch that adds the counting of

pte_none though (to digest why it is required :) ).

Quoted 14 lines

Quoted 13 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                         rss, &force_flush, &force_break); if (can_reclaim_pt) {
          none_nr += count_pte_none(pte, nr);
          none_nr += nr_skip;
      }

Right?

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Looking forward to your more review feedback on this series.

Quoted 20 lines

Quoted 12 lines

Quoted 12 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr, rss, &force_flush, &force_break); if (can_reclaim_pt) {
           none_nr += count_pte_none(pte, nr);
           none_nr += nr_skip;
       }

Quoted 3 lines

Quoted 1 line

Right?

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Quoted 2 lines

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Quoted 3 lines

Looking forward to your more review feedback on this series.

Thanks for all your hard work on this, I'm only able to make slow progress because I keep getting distracted by all different kinds of things :(

Quoted 26 lines · David Hildenbrand

Quoted 14 lines

Quoted 14 lines

Quoted 13 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                          rss, &force_flush, &force_break); if (can_reclaim_pt) {
           none_nr += count_pte_none(pte, nr);
           none_nr += nr_skip;
       }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

Quoted 5 lines

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 36 lines · Qi Zheng

Quoted 32 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                           rss, &force_flush, &force_break); if (can_reclaim_pt) {
            none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
        }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Quoted 2 lines

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

I'm not sure if well need cont_pte_none(), but I'll have to take a look at your new patch to see how this fits together with doing the pte_none detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Quoted 40 lines · David Hildenbrand

Quoted 30 lines · Qi Zheng

Quoted 27 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                           rss, &force_flush, &force_break); if (can_reclaim_pt) {
            none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
        }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

Quoted 8 lines

Quoted 5 lines

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 6 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look at your new patch to see how this fits together with doing the pte_none detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages to not be reclaimed.

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim

     the PTE page;

b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Quoted 65 lines · Qi Zheng

Quoted 40 lines · David Hildenbrand

Quoted 30 lines · Qi Zheng

Quoted 27 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                            rss, &force_flush, &force_break); if (can_reclaim_pt) {
             none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
         }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

Quoted 8 lines

Quoted 5 lines

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 6 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look at your new patch to see how this fits together with doing the pte_none detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages to not be reclaimed.

Quoted 5 lines

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

Assume we have

nr = do_zap_pte_range(&any_skipped)

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

If any_skipped == true, we have at least one !pte_none() entry among the "nr" entries. -> We cannot and must not reclaim.

Quoted 66 lines · David Hildenbrand

Quoted 65 lines · Qi Zheng

Quoted 40 lines · David Hildenbrand

Quoted 30 lines · Qi Zheng

Quoted 27 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                            rss, &force_flush, &force_break); if (can_reclaim_pt) {
             none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
         }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

Quoted 8 lines

Quoted 5 lines

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 6 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look at your new patch to see how this fits together with doing the pte_none detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages to not be reclaimed.

Quoted 7 lines

Quoted 5 lines

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim
     the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 6 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

Quoted 6 lines

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

I prefer that "nr" should not include pte_none().

Quoted 92 lines · Qi Zheng

Quoted 66 lines · David Hildenbrand

Quoted 65 lines · Qi Zheng

Quoted 40 lines · David Hildenbrand

Quoted 30 lines · Qi Zheng

Quoted 27 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                             rss, &force_flush, &force_break); if (can_reclaim_pt) {
              none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
          }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but there are cases where we "process an entry but not zap it".

Quoted 8 lines

Quoted 5 lines

What you really only want to know is "was any entry not zapped", which could be a simple input boolean variable passed into do_zap_pte_range?

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 6 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look at your new patch to see how this fits together with doing the pte_none detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages to not be reclaimed.

Quoted 7 lines

Quoted 5 lines

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 12 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

Quoted 2 lines

I prefer that "nr" should not include pte_none().

Why? do_zap_pte_range() should tell you how far to advance, nothing less, nothing more.

Let's just keep it simple and avoid count_pte_none().

I'm probably missing something important?

Quoted 106 lines · David Hildenbrand

Quoted 92 lines · Qi Zheng

Quoted 72 lines · David Hildenbrand

Quoted 71 lines · Qi Zheng

Quoted 43 lines · David Hildenbrand

Quoted 36 lines · Qi Zheng

Quoted 33 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                             rss, &force_flush, &force_break); if (can_reclaim_pt) {
              none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
          }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but
there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which
could be a simple input boolean variable passed into do_zap_pte_range?

Quoted 5 lines

Quoted 2 lines

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 8 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look
at your new patch to see how this fits together with doing the pte_none
detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages
to not be reclaimed.

Quoted 8 lines

Quoted 6 lines

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim
      the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 6 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

Quoted 9 lines

Quoted 6 lines

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

I prefer that "nr" should not include pte_none().

Why? do_zap_pte_range() should tell you how far to advance, nothing less, nothing more.

Quoted 3 lines

Let's just keep it simple and avoid count_pte_none().

I'm probably missing something important?

As we discussed before, we should skip all consecutive none ptes, pte and addr are already incremented before returning.

Quoted 109 lines · Qi Zheng

Quoted 106 lines · David Hildenbrand

Quoted 92 lines · Qi Zheng

Quoted 72 lines · David Hildenbrand

Quoted 71 lines · Qi Zheng

Quoted 43 lines · David Hildenbrand

Quoted 36 lines · Qi Zheng

Quoted 33 lines · David Hildenbrand

Quoted 15 lines

Quoted 15 lines

Quoted 14 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
                              rss, &force_flush, &force_break); if (can_reclaim_pt) {
               none_nr += count_pte_none(pte, nr); none_nr += nr_skip;
           }

Right?

Quoted 8 lines

Quoted 1 line

Yes. I did not look closely at the patch that adds the counting of

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but
there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which
could be a simple input boolean variable passed into do_zap_pte_range?

Quoted 5 lines

Quoted 2 lines

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

Quoted 8 lines

I'm not sure if well need cont_pte_none(), but I'll have to take a look
at your new patch to see how this fits together with doing the pte_none
detection+skipping in do_zap_pte_range().

I was wondering if you cannot simply avoid the additional scanning and simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages
to not be reclaimed.

Quoted 8 lines

Quoted 6 lines

So the most reliable and efficient method may be:

a. If there is a zap that is skipped, stop scanning and do not reclaim the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 6 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

Quoted 9 lines

Quoted 6 lines

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

I prefer that "nr" should not include pte_none().

Why? do_zap_pte_range() should tell you how far to advance, nothing less, nothing more.

Quoted 6 lines

Quoted 3 lines

Let's just keep it simple and avoid count_pte_none().

I'm probably missing something important?

As we discussed before, we should skip all consecutive none ptes, pte and addr are already incremented before returning.

It's probably best to send the resulting patch so I can either understand why count_pte_none() is required or comment on how to get rid of it.

Quoted 117 lines · David Hildenbrand

Quoted 116 lines · Qi Zheng

Quoted 113 lines · David Hildenbrand

Quoted 97 lines · Qi Zheng

Quoted 77 lines · David Hildenbrand

Quoted 76 lines · Qi Zheng

Quoted 50 lines · David Hildenbrand

Quoted 37 lines · Qi Zheng

Quoted 34 lines · David Hildenbrand

Quoted 18 lines

Quoted 13 lines

Quoted 13 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details,
&skip_nr,
                              rss, &force_flush, &force_break);
           if (can_reclaim_pt) {
               none_nr += count_pte_none(pte, nr);
               none_nr += nr_skip;
           }

Quoted 4 lines

Quoted 1 line

Right?

Yes. I did not look closely at the patch that adds the counting of

Quoted 6 lines

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but
there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which
could be a simple input boolean variable passed into do_zap_pte_range?

Quoted 5 lines

Quoted 2 lines

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

I'm not sure if well need cont_pte_none(), but I'll have to take a look
at your new patch to see how this fits together with doing the pte_none
detection+skipping in do_zap_pte_range().

Quoted 3 lines

I was wondering if you cannot simply avoid the additional scanning and
simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I
did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages
to not be reclaimed.

So the most reliable and efficient method may be:

Quoted 7 lines

Quoted 5 lines

a. If there is a zap that is skipped, stop scanning and do not reclaim
       the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 6 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

Quoted 9 lines

Quoted 6 lines

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

I prefer that "nr" should not include pte_none().

Why? do_zap_pte_range() should tell you how far to advance, nothing less, nothing more.

Let's just keep it simple and avoid count_pte_none().

Quoted 8 lines

Quoted 4 lines

Quoted 1 line

I'm probably missing something important?

As we discussed before, we should skip all consecutive none ptes, pte and addr are already incremented before returning.

It's probably best to send the resulting patch so I can either understand why count_pte_none() is required or comment on how to get rid of it.

Something like this:

mm/memory.c +0 -0

index bd9ebe0f4471f..e9bec3cd49d44 100644
@@ -1657,6 +1657,66 @@ static inline int zap_nonpresent_ptes(struct
 mmu_gather *tlb,
          return nr;
   }

mm/memory.c +6 -0

+               for (nr = 1; nr < max_nr; nr++) {
+                       ptent = ptep_get(pte + nr);
+                       if (!pte_none(ptent))
+                               break;
+               }
+               max_nr -= nr;

mm/memory.c +37 -18

+
+       return zap_nonpresent_ptes(tlb, vma, pte, ptent, max_nr, addr,
+                                  details, rss, any_skipped);
+}
+
+static inline int count_pte_none(pte_t *pte, int nr)
+{
+       int none_nr = 0;
+
+       /*
+        * If PTE_MARKER_UFFD_WP is enabled, the uffd-wp PTEs may be
+        * re-installed, so we need to check pte_none() one by one.
+        * Otherwise, checking a single PTE in a batch is sufficient.
+        */
+#ifdef CONFIG_PTE_MARKER_UFFD_WP
+       for (;;) {
+               if (pte_none(ptep_get(pte)))
+                       none_nr++;
+               if (--nr == 0)
+                       break;
+               pte++;
+       }
+#else
+       if (pte_none(ptep_get(pte)))
+               none_nr = nr;
+#endif
+       return none_nr;
+}
+
+
   static unsigned long zap_pte_range(struct mmu_gather *tlb,
                                  struct vm_area_struct *vma, pmd_t *pmd,
                                  unsigned long addr, unsigned long end,

@@ -1667,6 +1727,7 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
          int rss[NR_MM_COUNTERS];
          spinlock_t *ptl;
          pte_t *start_pte;
+       bool can_reclaim_pt;
          pte_t *pte;
          int nr;

@@ -1679,28 +1740,22 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
          flush_tlb_batched_pending(mm);
          arch_enter_lazy_mmu_mode();
          do {
-               pte_t ptent = ptep_get(pte);
-               int max_nr;
-
-               nr = 1;
-               if (pte_none(ptent))
-                       continue;
+               bool any_skipped;
 
                  if (need_resched())
                          break;
 
-               max_nr = (end - addr) / PAGE_SIZE;
-               if (pte_present(ptent)) {
-                       nr = zap_present_ptes(tlb, vma, pte, ptent, max_nr,
-                                             addr, details, rss,
 &force_flush,
-                                             &force_break);
-                       if (unlikely(force_break)) {
-                               addr += nr * PAGE_SIZE;
-                               break;
-                       }
-               } else {
-                       nr = zap_nonpresent_ptes(tlb, vma, pte, ptent,
 max_nr,
-                                                addr, details, rss);
+               nr = do_zap_pte_range(tlb, vma, pte, addr, end, details,
+                                     rss, &force_flush, &force_break,
+                                     &any_skipped);
+               if (can_reclaim_pt) {
+                       VM_BUG_ON(!any_skipped && count_pte_none(pte,

nr) == nr);

Quoted 140 lines · Qi Zheng

Quoted 117 lines · David Hildenbrand

Quoted 116 lines · Qi Zheng

Quoted 113 lines · David Hildenbrand

Quoted 97 lines · Qi Zheng

Quoted 77 lines · David Hildenbrand

Quoted 76 lines · Qi Zheng

Quoted 50 lines · David Hildenbrand

Quoted 37 lines · Qi Zheng

Quoted 34 lines · David Hildenbrand

Quoted 18 lines

Quoted 17 lines

*nr_skip = nr;

and then:

zap_pte_range
--> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details,
&skip_nr,
                               rss, &force_flush, &force_break);
            if (can_reclaim_pt) {
                none_nr += count_pte_none(pte, nr);
                none_nr += nr_skip;
            }

Right?

Yes. I did not look closely at the patch that adds the
counting of

Quoted 6 lines

Got it.

Quoted 1 line

pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect empty PTE page.

Okay, so the problem is that "nr" would be "all processed entries" but
there are cases where we "process an entry but not zap it".

What you really only want to know is "was any entry not zapped", which
could be a simple input boolean variable passed into do_zap_pte_range?

Quoted 5 lines

Quoted 2 lines

Because as soon as any entry was processed but  no zapped, you can immediately give up on reclaiming that table.

Yes, we can set can_reclaim_pt to false when a !pte_none() entry is found in count_pte_none().

I'm not sure if well need cont_pte_none(), but I'll have to take a look
at your new patch to see how this fits together with doing the pte_none
detection+skipping in do_zap_pte_range().

Quoted 3 lines

I was wondering if you cannot simply avoid the additional scanning and
simply set "can_reclaim_pt" if you skip a zap.

Maybe we can return the information whether the zap was skipped from zap_present_ptes() and zap_nonpresent_ptes() through parameters like I
did in [PATCH v1 3/7] and [PATCH v1 4/7].

In theory, we can detect empty PTE pages in the following two ways:

1) If no zap is skipped, it means that all pte entries have been zap, and the PTE page must be empty.
2) If all pte entries are detected to be none, then the PTE page is empty.

In the error case, 1) may cause non-empty PTE pages to be reclaimed (which is unacceptable), while the 2) will at most cause empty PTE pages
to not be reclaimed.

So the most reliable and efficient method may be:

Quoted 7 lines

Quoted 5 lines

a. If there is a zap that is skipped, stop scanning and do not reclaim
        the PTE page;
b. Otherwise, as now, detect the empty PTE page through count_pte_none()

Is there a need for count_pte_none() that I am missing?

When any_skipped == false, at least add VM_BUG_ON() to recheck none ptes.

Quoted 6 lines

Assume we have

nr = do_zap_pte_range(&any_skipped)

Quoted 9 lines

Quoted 6 lines

If "nr" is the number of processed entries (including pte_none()), and "any_skipped" is set whenever we skipped to zap a !pte_none entry, we can detect what we need, no?

If any_skipped == false after the call, we now have "nr" pte_none() entries. -> We can continue trying to reclaim

I prefer that "nr" should not include pte_none().

Why? do_zap_pte_range() should tell you how far to advance, nothing less, nothing more.

Let's just keep it simple and avoid count_pte_none().

Quoted 8 lines

Quoted 3 lines

Quoted 1 line

I'm probably missing something important?

As we discussed before, we should skip all consecutive none ptes,

  > pte and addr are already incremented before returning.

It's probably best to send the resulting patch so I can either understand why count_pte_none() is required or comment on how to get rid of it.

Something like this:

mm/memory.c +0 -0

index bd9ebe0f4471f..e9bec3cd49d44 100644
@@ -1657,6 +1657,66 @@ static inline int zap_nonpresent_ptes(struct
 mmu_gather *tlb,
           return nr;
    }
 

mm/memory.c static inline int zap_nonpresent_ptes(struct · line 1670

Quoted 10 lines

mm/memory.c +9 -0

+static inline int do_zap_pte_range(struct mmu_gather *tlb,
+                                  struct vm_area_struct *vma, pte_t 
 *pte,
+                                  unsigned long addr, unsigned long end,
+                                  struct zap_details *details, int *rss,
+                                  bool *force_flush, bool *force_break,
+                                  bool *any_skipped)
+{
+       pte_t ptent = ptep_get(pte);
+       int max_nr = (end - addr) / PAGE_SIZE;

I'd do here:

      int nr = 0;

mm/memory.c static inline int zap_nonpresent_ptes(struct mmu_gather *tlb, · line 1673

Quoted 4 lines

mm/memory.c +4 -0

+
+       /* Skip all consecutive pte_none(). */
+       if (pte_none(ptent)) {
+

instead of the "int nr" here

Quoted 6 lines

mm/memory.c +6 -0

+               for (nr = 1; nr < max_nr; nr++) {
+                       ptent = ptep_get(pte + nr);
+                       if (!pte_none(ptent))
+                               break;
+               }
+               max_nr -= nr;

mm/memory.c static inline int zap_nonpresent_ptes(struct mmu_gather *tlb, · line 1689

Quoted 10 lines

mm/memory.c +10 -0

+               if (!max_nr)
+                       return 0;
+               pte += nr;
+               addr += nr * PAGE_SIZE;
+       }
+
+       if (pte_present(ptent))
+               return zap_present_ptes(tlb, vma, pte, ptent, max_nr,
+                                       addr, details, rss, force_flush,
+                                       force_break, any_skipped);

and here:

if (pte_present(ptent))
    nr += zap_present_ptes(...)
else
    nr += zap_nonpresent_ptes();
return nr;

So you really return the number of processed items -- how much the caller should advance.

Quoted 79 lines

mm/memory.c +30 -0

+
+       return zap_nonpresent_ptes(tlb, vma, pte, ptent, max_nr, addr,
+                                  details, rss, any_skipped);
+}
+
+static inline int count_pte_none(pte_t *pte, int nr)
+{
+       int none_nr = 0;
+
+       /*
+        * If PTE_MARKER_UFFD_WP is enabled, the uffd-wp PTEs may be
+        * re-installed, so we need to check pte_none() one by one.
+        * Otherwise, checking a single PTE in a batch is sufficient.
+        */
+#ifdef CONFIG_PTE_MARKER_UFFD_WP
+       for (;;) {
+               if (pte_none(ptep_get(pte)))
+                       none_nr++;
+               if (--nr == 0)
+                       break;
+               pte++;
+       }
+#else
+       if (pte_none(ptep_get(pte)))
+               none_nr = nr;
+#endif
+       return none_nr;
+}
+
+
    static unsigned long zap_pte_range(struct mmu_gather *tlb,

                                  struct vm_area_struct *vma, pmd_t *pmd, unsigned long addr, unsigned long end,

mm/memory.c +7 -18

@@ -1667,6 +1727,7 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
           int rss[NR_MM_COUNTERS];
           spinlock_t *ptl;
           pte_t *start_pte;
+       bool can_reclaim_pt;
           pte_t *pte;
           int nr;

@@ -1679,28 +1740,22 @@ static unsigned long zap_pte_range(struct
 mmu_gather *tlb,
           flush_tlb_batched_pending(mm);
           arch_enter_lazy_mmu_mode();
           do {
-               pte_t ptent = ptep_get(pte);
-               int max_nr;
-
-               nr = 1;
-               if (pte_none(ptent))
-                       continue;
+               bool any_skipped;
 
                   if (need_resched())
                           break;
 
-               max_nr = (end - addr) / PAGE_SIZE;
-               if (pte_present(ptent)) {
-                       nr = zap_present_ptes(tlb, vma, pte, ptent, 
 max_nr,
-                                             addr, details, rss,
 &force_flush,
-                                             &force_break);
-                       if (unlikely(force_break)) {
-                               addr += nr * PAGE_SIZE;
-                               break;
-                       }
-               } else {
-                       nr = zap_nonpresent_ptes(tlb, vma, pte, ptent,
 max_nr,
-                                                addr, details, rss);
+               nr = do_zap_pte_range(tlb, vma, pte, addr, end, details,
+                                     rss, &force_flush, &force_break,
+                                     &any_skipped);
+               if (can_reclaim_pt) {
+                       VM_BUG_ON(!any_skipped && count_pte_none(pte,

Quoted 1 line

nr) == nr);

Okay, so this is really only for debugging then? So we can safely drop that for now.

I assume it would make sense to check when reclaiming a page table with CONFIG_DEBUG_VM, that the table is actually all-pte_none instead?

(as a side note: no VM_BUG_ON, please :) )