2.4.13-ac2

 mm

diff -u --new-file --recursive --exclude-from /usr/src/exclude linux.vanilla/mm/swapfile.c linux.ac/mm/swapfile.c
--- linux.vanilla/mm/swapfile.c   Thu Oct 25 16:26:39 2001
+++ linux.ac/mm/swapfile.c   Sun Oct 21 18:54:37 2001
@@ -14,7 +14,6 @@
 #include <linux/vmalloc.h>
 #include <linux/pagemap.h>
 #include <linux/shm.h>
-#include <linux/compiler.h>
 
 #include <asm/pgtable.h>
 
@@ -32,6 +31,25 @@
 
 struct swap_info_struct swap_info[MAX_SWAPFILES];
 
+/*
+ * When swap space gets filled up, we will set this flag.
+ * This will make do_swap_page(), in the page fault path,
+ * free swap entries on swapin so we'll reclaim swap space
+ * in order to be able to swap something out.
+ *
+ * At the moment we start reclaiming when swap usage goes
+ * over 80% of swap space.
+ *
+ * XXX: Random numbers, fixme.
+ */
+#define SWAP_FULL_PCT 80
+int vm_swap_full (void)
+{
+   int swap_used = total_swap_pages - nr_swap_pages;
+
+   return swap_used * 100 > total_swap_pages * SWAP_FULL_PCT;
+}
+
 #define SWAPFILE_CLUSTER 256
 
 static inline int scan_swap_map(struct swap_info_struct *si)
@@ -209,15 +227,15 @@
  * share this swap entry, so be cautious and let do_wp_page work out
  * what to do if a write is requested later.
  */
-/* mmlist_lock and vma->vm_mm->page_table_lock are held */
+/* BKL, mmlist_lock and vma->vm_mm->page_table_lock are held */
 static inline void unuse_pte(struct vm_area_struct * vma, unsigned long address,
    pte_t *dir, swp_entry_t entry, struct page* page)
 {
    pte_t pte = *dir;
 
-   if (likely(pte_to_swp_entry(pte).val != entry.val))
+   if (pte_to_swp_entry(pte).val != entry.val)
       return;
-   if (unlikely(pte_none(pte) || pte_present(pte)))
+   if (pte_none(pte) || pte_present(pte))
       return;
    get_page(page);
    set_pte(dir, pte_mkold(mk_pte(page, vma->vm_page_prot)));
@@ -225,7 +243,7 @@
    ++vma->vm_mm->rss;
 }
 
-/* mmlist_lock and vma->vm_mm->page_table_lock are held */
+/* BKL, mmlist_lock and vma->vm_mm->page_table_lock are held */
 static inline void unuse_pmd(struct vm_area_struct * vma, pmd_t *dir,
    unsigned long address, unsigned long size, unsigned long offset,
    swp_entry_t entry, struct page* page)
@@ -253,7 +271,7 @@
    } while (address && (address < end));
 }
 
-/* mmlist_lock and vma->vm_mm->page_table_lock are held */
+/* BKL, mmlist_lock and vma->vm_mm->page_table_lock are held */
 static inline void unuse_pgd(struct vm_area_struct * vma, pgd_t *dir,
    unsigned long address, unsigned long size,
    swp_entry_t entry, struct page* page)
@@ -284,7 +302,7 @@
    } while (address && (address < end));
 }
 
-/* mmlist_lock and vma->vm_mm->page_table_lock are held */
+/* BKL, mmlist_lock and vma->vm_mm->page_table_lock are held */
 static void unuse_vma(struct vm_area_struct * vma, pgd_t *pgdir,
          swp_entry_t entry, struct page* page)
 {
@@ -424,6 +442,10 @@
 
       /*
        * Don't hold on to start_mm if it looks like exiting.
+       * Can mmput ever block? if so, then we cannot risk
+       * it between deleting the page from the swap cache,
+       * and completing the search through mms (and cannot
+       * use it to avoid the long hold on mmlist_lock there).
        */
       if (atomic_read(&start_mm->mm_users) == 1) {
          mmput(start_mm);
@@ -432,15 +454,18 @@
       }
 
       /*
-       * Wait for and lock page.  When do_swap_page races with
-       * try_to_unuse, do_swap_page can handle the fault much
-       * faster than try_to_unuse can locate the entry.  This
-       * apparently redundant "wait_on_page" lets try_to_unuse
-       * defer to do_swap_page in such a case - in some tests,
-       * do_swap_page and try_to_unuse repeatedly compete.
+       * Wait for and lock page.  Remove it from swap cache
+       * so try_to_swap_out won't bump swap count.  Mark dirty
+       * so try_to_swap_out will preserve it without us having
+       * to mark any present ptes as dirty: so we can skip
+       * searching processes once swap count has all gone.
        */
-      wait_on_page(page);
       lock_page(page);
+      if (PageSwapCache(page))
+         delete_from_swap_cache(page);
+      SetPageDirty(page);
+      UnlockPage(page);
+      flush_page_to_ram(page);
 
       /*
        * Remove all references to entry, without blocking.
@@ -448,22 +473,20 @@
        * to search, but use it as a reminder to search shmem.
        */
       swcount = *swap_map;
-      if (swcount > 1) {
-         flush_page_to_ram(page);
+      if (swcount) {
          if (start_mm == &init_mm)
             shmem_unuse(entry, page);
          else
             unuse_process(start_mm, entry, page);
       }
-      if (*swap_map > 1) {
+      if (*swap_map) {
          int set_start_mm = (*swap_map >= swcount);
          struct list_head *p = &start_mm->mmlist;
          struct mm_struct *new_start_mm = start_mm;
          struct mm_struct *mm;
 
          spin_lock(&mmlist_lock);
-         while (*swap_map > 1 &&
-               (p = p->next) != &start_mm->mmlist) {
+         while (*swap_map && (p = p->next) != &start_mm->mmlist) {
             mm = list_entry(p, struct mm_struct, mmlist);
             swcount = *swap_map;
             if (mm == &init_mm) {
@@ -481,6 +504,7 @@
          mmput(start_mm);
          start_mm = new_start_mm;
       }
+      page_cache_release(page);
 
       /*
        * How could swap count reach 0x7fff when the maximum
@@ -499,52 +523,23 @@
          swap_list_lock();
          swap_device_lock(si);
          nr_swap_pages++;
-         *swap_map = 1;
+         *swap_map = 0;
          swap_device_unlock(si);
          swap_list_unlock();
          reset_overflow = 1;
       }
 
       /*
-       * If a reference remains (rare), we would like to leave
-       * the page in the swap cache; but try_to_swap_out could
-       * then re-duplicate the entry once we drop page lock,
-       * so we might loop indefinitely; also, that page could
-       * not be swapped out to other storage meanwhile.  So:
-       * delete from cache even if there's another reference,
-       * after ensuring that the data has been saved to disk -
-       * since if the reference remains (rarer), it will be
-       * read from disk into another page.  Splitting into two
-       * pages would be incorrect if swap supported "shared
-       * private" pages, but they are handled by tmpfs files.
-       * Note shmem_unuse already deleted its from swap cache.
-       */
-      swcount = *swap_map;
-      if ((swcount > 0) != PageSwapCache(page))
-         BUG();
-      if ((swcount > 1) && PageDirty(page)) {
-         rw_swap_page(WRITE, page);
-         lock_page(page);
-      }
-      if (PageSwapCache(page))
-         delete_from_swap_cache(page);
-
-      /*
-       * So we could skip searching mms once swap count went
-       * to 1, we did not mark any present ptes as dirty: must
-       * mark page dirty so try_to_swap_out will preserve it.
-       */
-      SetPageDirty(page);
-      UnlockPage(page);
-      page_cache_release(page);
-
-      /*
        * Make sure that we aren't completely killing
        * interactive performance.  Interruptible check on
        * signal_pending() would be nice, but changes the spec?
        */
       if (current->need_resched)
          schedule();
+      else {
+         unlock_kernel();
+         lock_kernel();
+      }
    }
 
    mmput(start_mm);
@@ -576,8 +571,14 @@
    for (type = swap_list.head; type >= 0; type = swap_info[type].next) {
       p = swap_info + type;
       if ((p->flags & SWP_WRITEOK) == SWP_WRITEOK) {
-         if (p->swap_file == nd.dentry)
-           break;
+         if (p->swap_file) {
+            if (p->swap_file == nd.dentry)
+              break;
+         } else {
+            if (S_ISBLK(nd.dentry->d_inode->i_mode)
+                && (p->swap_device == nd.dentry->d_inode->i_rdev))
+              break;
+         }
       }
       prev = type;
    }
@@ -600,9 +601,7 @@
    total_swap_pages -= p->pages;
    p->flags = SWP_USED;
    swap_list_unlock();
-   unlock_kernel();
    err = try_to_unuse(type);
-   lock_kernel();
    if (err) {
       /* re-insert swap space back into swap_list */
       swap_list_lock();
@@ -781,23 +780,28 @@
       if (!dev || (blk_size[MAJOR(dev)] &&
            !blk_size[MAJOR(dev)][MINOR(dev)]))
          goto bad_swap;
+      error = -EBUSY;
+      for (i = 0 ; i < nr_swapfiles ; i++) {
+         if (i == type)
+            continue;
+         if (dev == swap_info[i].swap_device)
+            goto bad_swap;
+      }
       swapfilesize = 0;
       if (blk_size[MAJOR(dev)])
          swapfilesize = blk_size[MAJOR(dev)][MINOR(dev)]
             >> (PAGE_SHIFT - 10);
-   } else if (S_ISREG(swap_inode->i_mode))
+   } else if (S_ISREG(swap_inode->i_mode)) {
+      error = -EBUSY;
+      for (i = 0 ; i < nr_swapfiles ; i++) {
+         if (i == type || !swap_info[i].swap_file)
+            continue;
+         if (swap_inode == swap_info[i].swap_file->d_inode)
+            goto bad_swap;
+      }
       swapfilesize = swap_inode->i_size >> PAGE_SHIFT;
-   else
+   } else
       goto bad_swap;
-
-   error = -EBUSY;
-   for (i = 0 ; i < nr_swapfiles ; i++) {
-      struct swap_info_struct *q = &swap_info[i];
-      if (i == type || !q->swap_file)
-         continue;
-      if (swap_inode->i_mapping == q->swap_file->d_inode->i_mapping)
-         goto bad_swap;
-   }
 
    swap_header = (void *) __get_free_page(GFP_USER);
    if (!swap_header) {