mm: swapoff prototype: frontswap handling added

The prototype of the new swapoff (without the quadratic complexity) presently ignores the frontswap case. Pass the count of pages_to_unuse down the page table walks in try_to_unuse(), and return from the walk when the desired number of pages has been swapped back in.

Quoted 5 lines · On Thu, 19 Mar 2015 03:55:45 -0700 Kelley Nielsen <[email protected]> wrote:

The prototype of the new swapoff (without the quadratic complexity) presently ignores the frontswap case. Pass the count of pages_to_unuse down the page table walks in try_to_unuse(), and return from the walk when the desired number of pages has been swapped back in.

Does this fix the "TODO" in [1/2]?

Do you think this patchset is ready for testing (while Hugh reviews it :)), or is there some deeper reason behind the "RFC"?

The prototype of the new swapoff (without the quadratic complexity) presently ignores the frontswap case. Pass the count of pages_to_unuse down the page table walks in try_to_unuse(), and return from the walk when the desired number of pages has been swapped back in.

Signed-off-by Kelley Nielsen <[email protected]>

---
Changes since v6:

- In try_to_unuse(), copy pages_to_unuse to local fs_pages_to_unuse and pass by reference (instead of value) down the page table walk
- create #define FRONTSWAP_PAGES_UNUSED to pass special code back up the chain from unuse_pte_range()
- Test for this return code in try_to_unuse() and exit before orphans cleanup and retries
- In unuse_pte_range, move decrement of frontswap pages count next to call to delete_from_swap_cache(), so that it happens on every unused swap slot instead of every pte unmap
- Remove unnecessary goto after call to unuse_mm() in try_to_unuse()
- Remove blank line in shmem.c/shmem_unuse_inode()

---
 mm/swapfile.c | 60 +++++++++++++++++++++++++++++++++++++++++------------------
 1 file changed, 42 insertions(+), 18 deletions(-)

mm/swapfile.c +2 -1

 static int unuse_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 				unsigned long addr, unsigned long end,
-				unsigned int type)
+				unsigned int type,
+				unsigned long *fs_pages_to_unuse)
 {
 	struct page *page;
 	swp_entry_t entry;

mm/swapfile.c +9 -1

 		swap_map = &si->swap_map[offset];
 		if (!swap_count(*swap_map))

@@ -1205,11 +1209,18 @@ static int unuse_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 		if (PageSwapCache(page) && (swap_count(*swap_map) == 0)) {
 			wait_on_page_writeback(page);
 			delete_from_swap_cache(page);
+			if (*fs_pages_to_unuse > 0) {
+				(*fs_pages_to_unuse)--;
+				if (*fs_pages_to_unuse == 0)
+					ret = FRONTSWAP_PAGES_UNUSED;
+			}
 		}
 
 		SetPageDirty(page);
 		unlock_page(page);
 		page_cache_release(page);
+		if (ret == FRONTSWAP_PAGES_UNUSED)
+			goto out;
 try_next:
 		pte = pte_offset_map(pmd, addr);
 	} while (pte++, addr += PAGE_SIZE, addr != end);

@@ -1220,7 +1231,8 @@ out:
 
 static inline int unuse_pmd_range(struct vm_area_struct *vma, pud_t *pud,
 				unsigned long addr, unsigned long end,
-				unsigned int type)
+				unsigned int type,
+				unsigned long *fs_pages_to_unuse)
 {
 	pmd_t *pmd;
 	unsigned long next;

mm/swapfile.c +2 -1

 			return ret;
 	} while (pmd++, addr = next, addr != end);
 	return 0;

@@ -1240,7 +1253,8 @@ static inline int unuse_pmd_range(struct vm_area_struct *vma, pud_t *pud,
 
 static inline int unuse_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
 				unsigned long addr, unsigned long end,
-				unsigned int type)
+				unsigned int type,
+				unsigned long *fs_pages_to_unuse)
 {
 	pud_t *pud;
 	unsigned long next;

mm/swapfile.c +2 -1

 			return ret;
 	} while (pud++, addr = next, addr != end);
 	return 0;
 }
 
-static int unuse_vma(struct vm_area_struct *vma, unsigned int type)
+static int unuse_vma(struct vm_area_struct *vma, unsigned int type,
+		unsigned long *fs_pages_to_unuse)
 {
 	pgd_t *pgd;
 	unsigned long addr, end, next;

mm/swapfile.c +3 -3

 			return ret;
 	} while (pgd++, addr = next, addr != end);
 	return 0;
 }
 
-static int unuse_mm(struct mm_struct *mm, unsigned int type)
+static int unuse_mm(struct mm_struct *mm, unsigned int type,
+		unsigned long *fs_pages_to_unuse)
 {
 	struct vm_area_struct *vma;
 	int ret = 0;

@@ -1287,7 +1305,7 @@ static int unuse_mm(struct mm_struct *mm, unsigned int type)
 	down_read(&mm->mmap_sem);
 	for (vma = mm->mmap; vma; vma = vma->vm_next) {
 		if (vma->anon_vma) {
-			ret = unuse_vma(vma, type);
+			ret = unuse_vma(vma, type, fs_pages_to_unuse);
 			if (ret)
 				break;
 		}

@@ -1342,7 +1360,6 @@ static unsigned int find_next_to_unuse(struct swap_info_struct *si,
 	return i;
 }
 
-/* TODO: frontswap */
 #define MAX_RETRIES 3
 int try_to_unuse(unsigned int type, bool frontswap,
 		 unsigned long pages_to_unuse)

@@ -1354,10 +1371,14 @@ int try_to_unuse(unsigned int type, bool frontswap,
 	struct swap_info_struct *si = swap_info[type];

mm/swapfile.c +7 -4

 retry:
 	retval = shmem_unuse(type);
 	if (retval)

@@ -1381,9 +1402,7 @@ retry:
 		mmput(prev_mm);
 		prev_mm = mm;
 
-		retval = unuse_mm(mm, type);
-		if (retval)
-			goto out_put;
+		retval = unuse_mm(mm, type, &fs_pages_to_unuse);
 
 		/*
 		 * Make sure that we aren't completely killing

@@ -1396,8 +1415,13 @@ retry:
 
 out_put:
 	mmput(prev_mm);
-	if (retval)
+	if (retval < 0)
+		goto out;
+	if (retval == FRONTSWAP_PAGES_UNUSED) {
+		retval = 0;
 		goto out;
+	}
+
 	while ((i = find_next_to_unuse(si, i, frontswap)) != 0) {
 		/*
 		 * under global memory pressure, swap entries

mm/swapfile.c +0 -0

 				retval = -EBUSY;
 				goto out;
 			}

Signature · 1 line

1.8.3.2