9p: Convert to using the netfs helper lib to do reads and caching
authorDavid Howells <dhowells@redhat.com>
Tue, 2 Nov 2021 08:29:55 +0000 (08:29 +0000)
committerDominique Martinet <asmadeus@codewreck.org>
Wed, 3 Nov 2021 08:45:02 +0000 (17:45 +0900)
Convert the 9p filesystem to use the netfs helper lib to handle readpage,
readahead and write_begin, converting those into a common issue_op for the
filesystem itself to handle.  The netfs helper lib also handles reading
from fscache if a cache is available, and interleaving reads from both
sources.

This change also switches from the old fscache I/O API to the new one,
meaning that fscache no longer keeps track of netfs pages and instead does
async DIO between the backing files and the 9p file pagecache.  As a part
of this change, the handling of PG_fscache changes.  It now just means that
the cache has a write I/O operation in progress on a page (PG_locked
is used for a read I/O op).

Note that this is a cut-down version of the fscache rewrite and does not
change any of the cookie and cache coherency handling.

Changes
=======
ver #4:
  - Rebase on top of folios.
  - Don't use wait_on_page_bit_killable().

ver #3:
  - v9fs_req_issue_op() needs to terminate the subrequest.
  - v9fs_write_end() needs to call SetPageUptodate() a bit more often.
  - It's not CONFIG_{AFS,V9FS}_FSCACHE[1]
  - v9fs_init_rreq() should take a ref on the p9_fid and the cleanup should
    drop it [from Dominique Martinet].

Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-and-tested-by: Dominique Martinet <asmadeus@codewreck.org>
cc: v9fs-developer@lists.sourceforge.net
cc: linux-cachefs@redhat.com
Link: https://lore.kernel.org/r/YUm+xucHxED+1MJp@codewreck.org/
Link: https://lore.kernel.org/r/163162772646.438332.16323773205855053535.stgit@warthog.procyon.org.uk/
Link: https://lore.kernel.org/r/163189109885.2509237.7153668924503399173.stgit@warthog.procyon.org.uk/
Link: https://lore.kernel.org/r/163363943896.1980952.1226527304649419689.stgit@warthog.procyon.org.uk/
Link: https://lore.kernel.org/r/163551662876.1877519.14706391695553204156.stgit@warthog.procyon.org.uk/
Link: https://lore.kernel.org/r/163584179557.4023316.11089762304657644342.stgit@warthog.procyon.org.uk
Signed-off-by: Dominique Martinet <asmadeus@codewreck.org>
fs/9p/Kconfig
fs/9p/cache.c
fs/9p/cache.h
fs/9p/v9fs.h
fs/9p/vfs_addr.c
fs/9p/vfs_file.c

index 09fd4a1..d7bc934 100644 (file)
@@ -2,6 +2,7 @@
 config 9P_FS
        tristate "Plan 9 Resource Sharing Support (9P2000)"
        depends on INET && NET_9P
+       select NETFS_SUPPORT
        help
          If you say Y here, you will get experimental support for
          Plan 9 resource sharing via the 9P2000 protocol.
index 1769a44..077f0a4 100644 (file)
@@ -199,140 +199,3 @@ void v9fs_cache_inode_reset_cookie(struct inode *inode)
 
        mutex_unlock(&v9inode->fscache_lock);
 }
-
-int __v9fs_fscache_release_page(struct page *page, gfp_t gfp)
-{
-       struct inode *inode = page->mapping->host;
-       struct v9fs_inode *v9inode = V9FS_I(inode);
-
-       BUG_ON(!v9inode->fscache);
-
-       return fscache_maybe_release_page(v9inode->fscache, page, gfp);
-}
-
-void __v9fs_fscache_invalidate_page(struct page *page)
-{
-       struct inode *inode = page->mapping->host;
-       struct v9fs_inode *v9inode = V9FS_I(inode);
-
-       BUG_ON(!v9inode->fscache);
-
-       if (PageFsCache(page)) {
-               fscache_wait_on_page_write(v9inode->fscache, page);
-               BUG_ON(!PageLocked(page));
-               fscache_uncache_page(v9inode->fscache, page);
-       }
-}
-
-static void v9fs_vfs_readpage_complete(struct page *page, void *data,
-                                      int error)
-{
-       if (!error)
-               SetPageUptodate(page);
-
-       unlock_page(page);
-}
-
-/*
- * __v9fs_readpage_from_fscache - read a page from cache
- *
- * Returns 0 if the pages are in cache and a BIO is submitted,
- * 1 if the pages are not in cache and -error otherwise.
- */
-
-int __v9fs_readpage_from_fscache(struct inode *inode, struct page *page)
-{
-       int ret;
-       const struct v9fs_inode *v9inode = V9FS_I(inode);
-
-       p9_debug(P9_DEBUG_FSC, "inode %p page %p\n", inode, page);
-       if (!v9inode->fscache)
-               return -ENOBUFS;
-
-       ret = fscache_read_or_alloc_page(v9inode->fscache,
-                                        page,
-                                        v9fs_vfs_readpage_complete,
-                                        NULL,
-                                        GFP_KERNEL);
-       switch (ret) {
-       case -ENOBUFS:
-       case -ENODATA:
-               p9_debug(P9_DEBUG_FSC, "page/inode not in cache %d\n", ret);
-               return 1;
-       case 0:
-               p9_debug(P9_DEBUG_FSC, "BIO submitted\n");
-               return ret;
-       default:
-               p9_debug(P9_DEBUG_FSC, "ret %d\n", ret);
-               return ret;
-       }
-}
-
-/*
- * __v9fs_readpages_from_fscache - read multiple pages from cache
- *
- * Returns 0 if the pages are in cache and a BIO is submitted,
- * 1 if the pages are not in cache and -error otherwise.
- */
-
-int __v9fs_readpages_from_fscache(struct inode *inode,
-                                 struct address_space *mapping,
-                                 struct list_head *pages,
-                                 unsigned *nr_pages)
-{
-       int ret;
-       const struct v9fs_inode *v9inode = V9FS_I(inode);
-
-       p9_debug(P9_DEBUG_FSC, "inode %p pages %u\n", inode, *nr_pages);
-       if (!v9inode->fscache)
-               return -ENOBUFS;
-
-       ret = fscache_read_or_alloc_pages(v9inode->fscache,
-                                         mapping, pages, nr_pages,
-                                         v9fs_vfs_readpage_complete,
-                                         NULL,
-                                         mapping_gfp_mask(mapping));
-       switch (ret) {
-       case -ENOBUFS:
-       case -ENODATA:
-               p9_debug(P9_DEBUG_FSC, "pages/inodes not in cache %d\n", ret);
-               return 1;
-       case 0:
-               BUG_ON(!list_empty(pages));
-               BUG_ON(*nr_pages != 0);
-               p9_debug(P9_DEBUG_FSC, "BIO submitted\n");
-               return ret;
-       default:
-               p9_debug(P9_DEBUG_FSC, "ret %d\n", ret);
-               return ret;
-       }
-}
-
-/*
- * __v9fs_readpage_to_fscache - write a page to the cache
- *
- */
-
-void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page)
-{
-       int ret;
-       const struct v9fs_inode *v9inode = V9FS_I(inode);
-
-       p9_debug(P9_DEBUG_FSC, "inode %p page %p\n", inode, page);
-       ret = fscache_write_page(v9inode->fscache, page,
-                                i_size_read(&v9inode->vfs_inode), GFP_KERNEL);
-       p9_debug(P9_DEBUG_FSC, "ret =  %d\n", ret);
-       if (ret != 0)
-               v9fs_uncache_page(inode, page);
-}
-
-/*
- * wait for a page to complete writing to the cache
- */
-void __v9fs_fscache_wait_on_page_write(struct inode *inode, struct page *page)
-{
-       const struct v9fs_inode *v9inode = V9FS_I(inode);
-       p9_debug(P9_DEBUG_FSC, "inode %p page %p\n", inode, page);
-       if (PageFsCache(page))
-               fscache_wait_on_page_write(v9inode->fscache, page);
-}
index 00f107a..7480b4b 100644 (file)
@@ -7,9 +7,10 @@
 
 #ifndef _9P_CACHE_H
 #define _9P_CACHE_H
-#ifdef CONFIG_9P_FSCACHE
+#define FSCACHE_USE_NEW_IO_API
 #include <linux/fscache.h>
-#include <linux/spinlock.h>
+
+#ifdef CONFIG_9P_FSCACHE
 
 extern struct fscache_netfs v9fs_cache_netfs;
 extern const struct fscache_cookie_def v9fs_cache_session_index_def;
@@ -27,64 +28,6 @@ extern void v9fs_cache_inode_reset_cookie(struct inode *inode);
 extern int __v9fs_cache_register(void);
 extern void __v9fs_cache_unregister(void);
 
-extern int __v9fs_fscache_release_page(struct page *page, gfp_t gfp);
-extern void __v9fs_fscache_invalidate_page(struct page *page);
-extern int __v9fs_readpage_from_fscache(struct inode *inode,
-                                       struct page *page);
-extern int __v9fs_readpages_from_fscache(struct inode *inode,
-                                        struct address_space *mapping,
-                                        struct list_head *pages,
-                                        unsigned *nr_pages);
-extern void __v9fs_readpage_to_fscache(struct inode *inode, struct page *page);
-extern void __v9fs_fscache_wait_on_page_write(struct inode *inode,
-                                             struct page *page);
-
-static inline int v9fs_fscache_release_page(struct page *page,
-                                           gfp_t gfp)
-{
-       return __v9fs_fscache_release_page(page, gfp);
-}
-
-static inline void v9fs_fscache_invalidate_page(struct page *page)
-{
-       __v9fs_fscache_invalidate_page(page);
-}
-
-static inline int v9fs_readpage_from_fscache(struct inode *inode,
-                                            struct page *page)
-{
-       return __v9fs_readpage_from_fscache(inode, page);
-}
-
-static inline int v9fs_readpages_from_fscache(struct inode *inode,
-                                             struct address_space *mapping,
-                                             struct list_head *pages,
-                                             unsigned *nr_pages)
-{
-       return __v9fs_readpages_from_fscache(inode, mapping, pages,
-                                            nr_pages);
-}
-
-static inline void v9fs_readpage_to_fscache(struct inode *inode,
-                                           struct page *page)
-{
-       if (PageFsCache(page))
-               __v9fs_readpage_to_fscache(inode, page);
-}
-
-static inline void v9fs_uncache_page(struct inode *inode, struct page *page)
-{
-       struct v9fs_inode *v9inode = V9FS_I(inode);
-       fscache_uncache_page(v9inode->fscache, page);
-       BUG_ON(PageFsCache(page));
-}
-
-static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
-                                                  struct page *page)
-{
-       return __v9fs_fscache_wait_on_page_write(inode, page);
-}
-
 #else /* CONFIG_9P_FSCACHE */
 
 static inline void v9fs_cache_inode_get_cookie(struct inode *inode)
@@ -99,39 +42,5 @@ static inline void v9fs_cache_inode_set_cookie(struct inode *inode, struct file
 {
 }
 
-static inline int v9fs_fscache_release_page(struct page *page,
-                                           gfp_t gfp) {
-       return 1;
-}
-
-static inline void v9fs_fscache_invalidate_page(struct page *page) {}
-
-static inline int v9fs_readpage_from_fscache(struct inode *inode,
-                                            struct page *page)
-{
-       return -ENOBUFS;
-}
-
-static inline int v9fs_readpages_from_fscache(struct inode *inode,
-                                             struct address_space *mapping,
-                                             struct list_head *pages,
-                                             unsigned *nr_pages)
-{
-       return -ENOBUFS;
-}
-
-static inline void v9fs_readpage_to_fscache(struct inode *inode,
-                                           struct page *page)
-{}
-
-static inline void v9fs_uncache_page(struct inode *inode, struct page *page)
-{}
-
-static inline void v9fs_fscache_wait_on_page_write(struct inode *inode,
-                                                  struct page *page)
-{
-       return;
-}
-
 #endif /* CONFIG_9P_FSCACHE */
 #endif /* _9P_CACHE_H */
index 4ca56c5..92124b2 100644 (file)
@@ -124,6 +124,15 @@ static inline struct v9fs_inode *V9FS_I(const struct inode *inode)
        return container_of(inode, struct v9fs_inode, vfs_inode);
 }
 
+static inline struct fscache_cookie *v9fs_inode_cookie(struct v9fs_inode *v9inode)
+{
+#ifdef CONFIG_9P_FSCACHE
+       return v9inode->fscache;
+#else
+       return NULL;
+#endif
+}
+
 extern int v9fs_show_options(struct seq_file *m, struct dentry *root);
 
 struct p9_fid *v9fs_session_init(struct v9fs_session_info *, const char *,
index 1c4f1b3..2a2368d 100644 (file)
@@ -19,7 +19,7 @@
 #include <linux/idr.h>
 #include <linux/sched.h>
 #include <linux/uio.h>
-#include <linux/bvec.h>
+#include <linux/netfs.h>
 #include <net/9p/9p.h>
 #include <net/9p/client.h>
 
 #include "fid.h"
 
 /**
- * v9fs_fid_readpage - read an entire page in from 9P
- * @data: Opaque pointer to the fid being read
- * @page: structure to page
- *
+ * v9fs_req_issue_op - Issue a read from 9P
+ * @subreq: The read to make
  */
-static int v9fs_fid_readpage(void *data, struct page *page)
+static void v9fs_req_issue_op(struct netfs_read_subrequest *subreq)
 {
-       struct p9_fid *fid = data;
-       struct inode *inode = page->mapping->host;
-       struct bio_vec bvec = {.bv_page = page, .bv_len = PAGE_SIZE};
+       struct netfs_read_request *rreq = subreq->rreq;
+       struct p9_fid *fid = rreq->netfs_priv;
        struct iov_iter to;
-       int retval, err;
+       loff_t pos = subreq->start + subreq->transferred;
+       size_t len = subreq->len   - subreq->transferred;
+       int total, err;
 
-       p9_debug(P9_DEBUG_VFS, "\n");
+       iov_iter_xarray(&to, READ, &rreq->mapping->i_pages, pos, len);
 
-       BUG_ON(!PageLocked(page));
+       total = p9_client_read(fid, pos, &to, &err);
+       netfs_subreq_terminated(subreq, err ?: total, false);
+}
 
-       retval = v9fs_readpage_from_fscache(inode, page);
-       if (retval == 0)
-               return retval;
+/**
+ * v9fs_init_rreq - Initialise a read request
+ * @rreq: The read request
+ * @file: The file being read from
+ */
+static void v9fs_init_rreq(struct netfs_read_request *rreq, struct file *file)
+{
+       struct p9_fid *fid = file->private_data;
 
-       iov_iter_bvec(&to, READ, &bvec, 1, PAGE_SIZE);
+       refcount_inc(&fid->count);
+       rreq->netfs_priv = fid;
+}
 
-       retval = p9_client_read(fid, page_offset(page), &to, &err);
-       if (err) {
-               v9fs_uncache_page(inode, page);
-               retval = err;
-               goto done;
-       }
+/**
+ * v9fs_req_cleanup - Cleanup request initialized by v9fs_init_rreq
+ * @mapping: unused mapping of request to cleanup
+ * @priv: private data to cleanup, a fid, guaranted non-null.
+ */
+static void v9fs_req_cleanup(struct address_space *mapping, void *priv)
+{
+       struct p9_fid *fid = priv;
 
-       zero_user(page, retval, PAGE_SIZE - retval);
-       flush_dcache_page(page);
-       SetPageUptodate(page);
+       p9_client_clunk(fid);
+}
 
-       v9fs_readpage_to_fscache(inode, page);
-       retval = 0;
+/**
+ * v9fs_is_cache_enabled - Determine if caching is enabled for an inode
+ * @inode: The inode to check
+ */
+static bool v9fs_is_cache_enabled(struct inode *inode)
+{
+       struct fscache_cookie *cookie = v9fs_inode_cookie(V9FS_I(inode));
 
-done:
-       unlock_page(page);
-       return retval;
+       return fscache_cookie_enabled(cookie) && !hlist_empty(&cookie->backing_objects);
+}
+
+/**
+ * v9fs_begin_cache_operation - Begin a cache operation for a read
+ * @rreq: The read request
+ */
+static int v9fs_begin_cache_operation(struct netfs_read_request *rreq)
+{
+       struct fscache_cookie *cookie = v9fs_inode_cookie(V9FS_I(rreq->inode));
+
+       return fscache_begin_read_operation(rreq, cookie);
 }
 
+static const struct netfs_read_request_ops v9fs_req_ops = {
+       .init_rreq              = v9fs_init_rreq,
+       .is_cache_enabled       = v9fs_is_cache_enabled,
+       .begin_cache_operation  = v9fs_begin_cache_operation,
+       .issue_op               = v9fs_req_issue_op,
+       .cleanup                = v9fs_req_cleanup,
+};
+
 /**
  * v9fs_vfs_readpage - read an entire page in from 9P
- *
- * @filp: file being read
+ * @file: file being read
  * @page: structure to page
  *
  */
-
-static int v9fs_vfs_readpage(struct file *filp, struct page *page)
+static int v9fs_vfs_readpage(struct file *file, struct page *page)
 {
-       return v9fs_fid_readpage(filp->private_data, page);
+       return netfs_readpage(file, page, &v9fs_req_ops, NULL);
 }
 
 /**
- * v9fs_vfs_readpages - read a set of pages from 9P
- *
- * @filp: file being read
- * @mapping: the address space
- * @pages: list of pages to read
- * @nr_pages: count of pages to read
- *
+ * v9fs_vfs_readahead - read a set of pages from 9P
+ * @ractl: The readahead parameters
  */
-
-static int v9fs_vfs_readpages(struct file *filp, struct address_space *mapping,
-                            struct list_head *pages, unsigned nr_pages)
+static void v9fs_vfs_readahead(struct readahead_control *ractl)
 {
-       int ret = 0;
-       struct inode *inode;
-
-       inode = mapping->host;
-       p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, filp);
-
-       ret = v9fs_readpages_from_fscache(inode, mapping, pages, &nr_pages);
-       if (ret == 0)
-               return ret;
-
-       ret = read_cache_pages(mapping, pages, v9fs_fid_readpage,
-                       filp->private_data);
-       p9_debug(P9_DEBUG_VFS, "  = %d\n", ret);
-       return ret;
+       netfs_readahead(ractl, &v9fs_req_ops, NULL);
 }
 
 /**
@@ -125,7 +134,14 @@ static int v9fs_release_page(struct page *page, gfp_t gfp)
 {
        if (PagePrivate(page))
                return 0;
-       return v9fs_fscache_release_page(page, gfp);
+#ifdef CONFIG_9P_FSCACHE
+       if (PageFsCache(page)) {
+               if (!(gfp & __GFP_DIRECT_RECLAIM) || !(gfp & __GFP_FS))
+                       return 0;
+               wait_on_page_fscache(page);
+       }
+#endif
+       return 1;
 }
 
 /**
@@ -138,21 +154,16 @@ static int v9fs_release_page(struct page *page, gfp_t gfp)
 static void v9fs_invalidate_page(struct page *page, unsigned int offset,
                                 unsigned int length)
 {
-       /*
-        * If called with zero offset, we should release
-        * the private state assocated with the page
-        */
-       if (offset == 0 && length == PAGE_SIZE)
-               v9fs_fscache_invalidate_page(page);
+       wait_on_page_fscache(page);
 }
 
 static int v9fs_vfs_writepage_locked(struct page *page)
 {
        struct inode *inode = page->mapping->host;
        struct v9fs_inode *v9inode = V9FS_I(inode);
+       loff_t start = page_offset(page);
        loff_t size = i_size_read(inode);
        struct iov_iter from;
-       struct bio_vec bvec;
        int err, len;
 
        if (page->index == size >> PAGE_SHIFT)
@@ -160,17 +171,14 @@ static int v9fs_vfs_writepage_locked(struct page *page)
        else
                len = PAGE_SIZE;
 
-       bvec.bv_page = page;
-       bvec.bv_offset = 0;
-       bvec.bv_len = len;
-       iov_iter_bvec(&from, WRITE, &bvec, 1, len);
+       iov_iter_xarray(&from, WRITE, &page->mapping->i_pages, start, len);
 
        /* We should have writeback_fid always set */
        BUG_ON(!v9inode->writeback_fid);
 
        set_page_writeback(page);
 
-       p9_client_write(v9inode->writeback_fid, page_offset(page), &from, &err);
+       p9_client_write(v9inode->writeback_fid, start, &from, &err);
 
        end_page_writeback(page);
        return err;
@@ -208,14 +216,13 @@ static int v9fs_vfs_writepage(struct page *page, struct writeback_control *wbc)
 static int v9fs_launder_page(struct page *page)
 {
        int retval;
-       struct inode *inode = page->mapping->host;
 
-       v9fs_fscache_wait_on_page_write(inode, page);
        if (clear_page_dirty_for_io(page)) {
                retval = v9fs_vfs_writepage_locked(page);
                if (retval)
                        return retval;
        }
+       wait_on_page_fscache(page);
        return 0;
 }
 
@@ -260,35 +267,24 @@ static int v9fs_write_begin(struct file *filp, struct address_space *mapping,
                            loff_t pos, unsigned len, unsigned flags,
                            struct page **pagep, void **fsdata)
 {
-       int retval = 0;
+       int retval;
        struct page *page;
-       struct v9fs_inode *v9inode;
-       pgoff_t index = pos >> PAGE_SHIFT;
-       struct inode *inode = mapping->host;
-
+       struct v9fs_inode *v9inode = V9FS_I(mapping->host);
 
        p9_debug(P9_DEBUG_VFS, "filp %p, mapping %p\n", filp, mapping);
 
-       v9inode = V9FS_I(inode);
-start:
-       page = grab_cache_page_write_begin(mapping, index, flags);
-       if (!page) {
-               retval = -ENOMEM;
-               goto out;
-       }
        BUG_ON(!v9inode->writeback_fid);
-       if (PageUptodate(page))
-               goto out;
 
-       if (len == PAGE_SIZE)
-               goto out;
+       /* Prefetch area to be written into the cache if we're caching this
+        * file.  We need to do this before we get a lock on the page in case
+        * there's more than one writer competing for the same cache block.
+        */
+       retval = netfs_write_begin(filp, mapping, pos, len, flags, &page, fsdata,
+                                  &v9fs_req_ops, NULL);
+       if (retval < 0)
+               return retval;
 
-       retval = v9fs_fid_readpage(v9inode->writeback_fid, page);
-       put_page(page);
-       if (!retval)
-               goto start;
-out:
-       *pagep = page;
+       *pagep = find_subpage(page, pos / PAGE_SIZE);
        return retval;
 }
 
@@ -305,10 +301,11 @@ static int v9fs_write_end(struct file *filp, struct address_space *mapping,
                if (unlikely(copied < len)) {
                        copied = 0;
                        goto out;
-               } else if (len == PAGE_SIZE) {
-                       SetPageUptodate(page);
                }
+
+               SetPageUptodate(page);
        }
+
        /*
         * No need to use i_size_read() here, the i_size
         * cannot change under us because we hold the i_mutex.
@@ -328,7 +325,7 @@ out:
 
 const struct address_space_operations v9fs_addr_operations = {
        .readpage = v9fs_vfs_readpage,
-       .readpages = v9fs_vfs_readpages,
+       .readahead = v9fs_vfs_readahead,
        .set_page_dirty = __set_page_dirty_nobuffers,
        .writepage = v9fs_vfs_writepage,
        .write_begin = v9fs_write_begin,
index 246235e..8d5e0ef 100644 (file)
@@ -537,14 +537,23 @@ v9fs_vm_page_mkwrite(struct vm_fault *vmf)
        p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
                 page, (unsigned long)filp->private_data);
 
+       v9inode = V9FS_I(inode);
+
+       /* Wait for the page to be written to the cache before we allow it to
+        * be modified.  We then assume the entire page will need writing back.
+        */
+#ifdef CONFIG_9P_FSCACHE
+       if (PageFsCache(page) &&
+           wait_on_page_fscache_killable(page) < 0)
+               return VM_FAULT_RETRY;
+#endif
+
        /* Update file times before taking page lock */
        file_update_time(filp);
 
-       v9inode = V9FS_I(inode);
-       /* make sure the cache has finished storing the page */
-       v9fs_fscache_wait_on_page_write(inode, page);
        BUG_ON(!v9inode->writeback_fid);
-       lock_page(page);
+       if (lock_page_killable(page) < 0)
+               return VM_FAULT_RETRY;
        if (page->mapping != inode->i_mapping)
                goto out_unlock;
        wait_for_stable_page(page);