7 files changed, 106 insertions, 151 deletions

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index eeb71e5de27a..26a9d10d75e9 100644
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -276,8 +276,10 @@ static void finish_read(struct ceph_osd_request *req)
 	for (i = 0; i < num_pages; i++) {
 		struct page *page = osd_data->pages[i];
 
-		if (rc < 0 && rc != -ENOENT)
+		if (rc < 0 && rc != -ENOENT) {
+			ceph_fscache_readpage_cancel(inode, page);
 			goto unlock;
+		}
 		if (bytes < (int)PAGE_SIZE) {
 			/* zero (remainder of) page */
 			int s = bytes < 0 ? 0 : bytes;
@@ -535,8 +537,6 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 	    CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb))
 		set_bdi_congested(&fsc->backing_dev_info, BLK_RW_ASYNC);
 
-	ceph_readpage_to_fscache(inode, page);
-
 	set_page_writeback(page);
 	err = ceph_osdc_writepages(osdc, ceph_vino(inode),
 				   &ci->i_layout, snapc,
diff --git a/fs/ceph/cache.c b/fs/ceph/cache.c
index c052b5bf219b..238c55b01723 100644
--- a/fs/ceph/cache.c
+++ b/fs/ceph/cache.c
@@ -25,6 +25,7 @@
 #include "cache.h"
 
 struct ceph_aux_inode {
+	u64 		version;
 	struct timespec	mtime;
 	loff_t          size;
 };
@@ -69,15 +70,8 @@ int ceph_fscache_register_fs(struct ceph_fs_client* fsc)
 	fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index,
 					      &ceph_fscache_fsid_object_def,
 					      fsc, true);
-
-	if (fsc->fscache == NULL) {
+	if (!fsc->fscache)
 		pr_err("Unable to resgister fsid: %p fscache cookie", fsc);
-		return 0;
-	}
-
-	fsc->revalidate_wq = alloc_workqueue("ceph-revalidate", 0, 1);
-	if (fsc->revalidate_wq == NULL)
-		return -ENOMEM;
 
 	return 0;
 }
@@ -105,6 +99,7 @@ static uint16_t ceph_fscache_inode_get_aux(const void *cookie_netfs_data,
 	const struct inode* inode = &ci->vfs_inode;
 
 	memset(&aux, 0, sizeof(aux));
+	aux.version = ci->i_version;
 	aux.mtime = inode->i_mtime;
 	aux.size = i_size_read(inode);
 
@@ -131,6 +126,7 @@ static enum fscache_checkaux ceph_fscache_inode_check_aux(
 		return FSCACHE_CHECKAUX_OBSOLETE;
 
 	memset(&aux, 0, sizeof(aux));
+	aux.version = ci->i_version;
 	aux.mtime = inode->i_mtime;
 	aux.size = i_size_read(inode);
 
@@ -181,32 +177,26 @@ static const struct fscache_cookie_def ceph_fscache_inode_object_def = {
 	.now_uncached	= ceph_fscache_inode_now_uncached,
 };
 
-void ceph_fscache_register_inode_cookie(struct ceph_fs_client* fsc,
-					struct ceph_inode_info* ci)
+void ceph_fscache_register_inode_cookie(struct inode *inode)
 {
-	struct inode* inode = &ci->vfs_inode;
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 
 	/* No caching for filesystem */
 	if (fsc->fscache == NULL)
 		return;
 
 	/* Only cache for regular files that are read only */
-	if ((ci->vfs_inode.i_mode & S_IFREG) == 0)
+	if (!S_ISREG(inode->i_mode))
 		return;
 
-	/* Avoid multiple racing open requests */
-	inode_lock(inode);
-
-	if (ci->fscache)
-		goto done;
-
-	ci->fscache = fscache_acquire_cookie(fsc->fscache,
-					     &ceph_fscache_inode_object_def,
-					     ci, true);
-	fscache_check_consistency(ci->fscache);
-done:
+	inode_lock_nested(inode, I_MUTEX_CHILD);
+	if (!ci->fscache) {
+		ci->fscache = fscache_acquire_cookie(fsc->fscache,
+					&ceph_fscache_inode_object_def,
+					ci, false);
+	}
 	inode_unlock(inode);
-
 }
 
 void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
@@ -222,6 +212,34 @@ void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
 	fscache_relinquish_cookie(cookie, 0);
 }
 
+static bool ceph_fscache_can_enable(void *data)
+{
+	struct inode *inode = data;
+	return !inode_is_open_for_write(inode);
+}
+
+void ceph_fscache_file_set_cookie(struct inode *inode, struct file *filp)
+{
+	struct ceph_inode_info *ci = ceph_inode(inode);
+
+	if (!fscache_cookie_valid(ci->fscache))
+		return;
+
+	if (inode_is_open_for_write(inode)) {
+		dout("fscache_file_set_cookie %p %p disabling cache\n",
+		     inode, filp);
+		fscache_disable_cookie(ci->fscache, false);
+		fscache_uncache_all_inode_pages(ci->fscache, inode);
+	} else {
+		fscache_enable_cookie(ci->fscache, ceph_fscache_can_enable,
+				inode);
+		if (fscache_cookie_enabled(ci->fscache)) {
+			dout("fscache_file_set_cookie %p %p enabing cache\n",
+			     inode, filp);
+		}
+	}
+}
+
 static void ceph_vfs_readpage_complete(struct page *page, void *data, int error)
 {
 	if (!error)
@@ -238,8 +256,7 @@ static void ceph_vfs_readpage_complete_unlock(struct page *page, void *data, int
 
 static inline bool cache_valid(struct ceph_inode_info *ci)
 {
-	return ((ceph_caps_issued(ci) & CEPH_CAP_FILE_CACHE) &&
-		(ci->i_fscache_gen == ci->i_rdcache_gen));
+	return ci->i_fscache_gen == ci->i_rdcache_gen;
 }
 
 
@@ -332,69 +349,27 @@ void ceph_invalidate_fscache_page(struct inode* inode, struct page *page)
 
 void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc)
 {
-	if (fsc->revalidate_wq)
-		destroy_workqueue(fsc->revalidate_wq);
-
 	fscache_relinquish_cookie(fsc->fscache, 0);
 	fsc->fscache = NULL;
 }
 
-static void ceph_revalidate_work(struct work_struct *work)
-{
-	int issued;
-	u32 orig_gen;
-	struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info,
-						  i_revalidate_work);
-	struct inode *inode = &ci->vfs_inode;
-
-	spin_lock(&ci->i_ceph_lock);
-	issued = __ceph_caps_issued(ci, NULL);
-	orig_gen = ci->i_rdcache_gen;
-	spin_unlock(&ci->i_ceph_lock);
-
-	if (!(issued & CEPH_CAP_FILE_CACHE)) {
-		dout("revalidate_work lost cache before validation %p\n",
-		     inode);
-		goto out;
-	}
-
-	if (!fscache_check_consistency(ci->fscache))
-		fscache_invalidate(ci->fscache);
-
-	spin_lock(&ci->i_ceph_lock);
-	/* Update the new valid generation (backwards sanity check too) */
-	if (orig_gen > ci->i_fscache_gen) {
-		ci->i_fscache_gen = orig_gen;
-	}
-	spin_unlock(&ci->i_ceph_lock);
-
-out:
-	iput(&ci->vfs_inode);
-}
-
-void ceph_queue_revalidate(struct inode *inode)
+/*
+ * caller should hold CEPH_CAP_FILE_{RD,CACHE}
+ */
+void ceph_fscache_revalidate_cookie(struct ceph_inode_info *ci)
 {
-	struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
-	struct ceph_inode_info *ci = ceph_inode(inode);
-
-	if (fsc->revalidate_wq == NULL || ci->fscache == NULL)
+	if (cache_valid(ci))
 		return;
 
-	ihold(inode);
-
-	if (queue_work(ceph_sb_to_client(inode->i_sb)->revalidate_wq,
-		       &ci->i_revalidate_work)) {
-		dout("ceph_queue_revalidate %p\n", inode);
-	} else {
-		dout("ceph_queue_revalidate %p failed\n)", inode);
-		iput(inode);
+	/* resue i_truncate_mutex. There should be no pending
+	 * truncate while the caller holds CEPH_CAP_FILE_RD */
+	mutex_lock(&ci->i_truncate_mutex);
+	if (!cache_valid(ci)) {
+		if (fscache_check_consistency(ci->fscache))
+			fscache_invalidate(ci->fscache);
+		spin_lock(&ci->i_ceph_lock);
+		ci->i_fscache_gen = ci->i_rdcache_gen;
+		spin_unlock(&ci->i_ceph_lock);
 	}
-}
-
-void ceph_fscache_inode_init(struct ceph_inode_info *ci)
-{
-	ci->fscache = NULL;
-	/* The first load is verifed cookie open time */
-	ci->i_fscache_gen = 1;
-	INIT_WORK(&ci->i_revalidate_work, ceph_revalidate_work);
+	mutex_unlock(&ci->i_truncate_mutex);
 }
diff --git a/fs/ceph/cache.h b/fs/ceph/cache.h
index 5ac591bd012b..7e72c7594f0c 100644
--- a/fs/ceph/cache.h
+++ b/fs/ceph/cache.h
@@ -34,10 +34,10 @@ void ceph_fscache_unregister(void);
 int ceph_fscache_register_fs(struct ceph_fs_client* fsc);
 void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc);
 
-void ceph_fscache_inode_init(struct ceph_inode_info *ci);
-void ceph_fscache_register_inode_cookie(struct ceph_fs_client* fsc,
-					struct ceph_inode_info* ci);
+void ceph_fscache_register_inode_cookie(struct inode *inode);
 void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci);
+void ceph_fscache_file_set_cookie(struct inode *inode, struct file *filp);
+void ceph_fscache_revalidate_cookie(struct ceph_inode_info *ci);
 
 int ceph_readpage_from_fscache(struct inode *inode, struct page *page);
 int ceph_readpages_from_fscache(struct inode *inode,
@@ -46,12 +46,11 @@ int ceph_readpages_from_fscache(struct inode *inode,
 				unsigned *nr_pages);
 void ceph_readpage_to_fscache(struct inode *inode, struct page *page);
 void ceph_invalidate_fscache_page(struct inode* inode, struct page *page);
-void ceph_queue_revalidate(struct inode *inode);
 
-static inline void ceph_fscache_update_objectsize(struct inode *inode)
+static inline void ceph_fscache_inode_init(struct ceph_inode_info *ci)
 {
-	struct ceph_inode_info *ci = ceph_inode(inode);
-	fscache_attr_changed(ci->fscache);
+	ci->fscache = NULL;
+	ci->i_fscache_gen = 0;
 }
 
 static inline void ceph_fscache_invalidate(struct inode *inode)
@@ -88,6 +87,11 @@ static inline void ceph_fscache_readpages_cancel(struct inode *inode,
 	return fscache_readpages_cancel(ci->fscache, pages);
 }
 
+static inline void ceph_disable_fscache_readpage(struct ceph_inode_info *ci)
+{
+	ci->i_fscache_gen = ci->i_rdcache_gen - 1;
+}
+
 #else
 
 static inline int ceph_fscache_register(void)
@@ -112,8 +116,20 @@ static inline void ceph_fscache_inode_init(struct ceph_inode_info *ci)
 {
 }
 
-static inline void ceph_fscache_register_inode_cookie(struct ceph_fs_client* parent_fsc,
-						      struct ceph_inode_info* ci)
+static inline void ceph_fscache_register_inode_cookie(struct inode *inode)
+{
+}
+
+static inline void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
+{
+}
+
+static inline void ceph_fscache_file_set_cookie(struct inode *inode,
+						struct file *filp)
+{
+}
+
+static inline void ceph_fscache_revalidate_cookie(struct ceph_inode_info *ci)
 {
 }
 
@@ -141,10 +157,6 @@ static inline void ceph_readpage_to_fscache(struct inode *inode,
 {
 }
 
-static inline void ceph_fscache_update_objectsize(struct inode *inode)
-{
-}
-
 static inline void ceph_fscache_invalidate(struct inode *inode)
 {
 }
@@ -154,10 +166,6 @@ static inline void ceph_invalidate_fscache_page(struct inode *inode,
 {
 }
 
-static inline void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
-{
-}
-
 static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp)
 {
 	return 1;
@@ -173,7 +181,7 @@ static inline void ceph_fscache_readpages_cancel(struct inode *inode,
 {
 }
 
-static inline void ceph_queue_revalidate(struct inode *inode)
+static inline void ceph_disable_fscache_readpage(struct ceph_inode_info *ci)
 {
 }
 
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index c17b5d76d75e..6f60d0a3d0f9 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -2393,6 +2393,9 @@ again:
 				snap_rwsem_locked = true;
 			}
 			*got = need | (have & want);
+			if ((need & CEPH_CAP_FILE_RD) &&
+			    !(*got & CEPH_CAP_FILE_CACHE))
+				ceph_disable_fscache_readpage(ci);
 			__take_cap_refs(ci, *got, true);
 			ret = 1;
 		}
@@ -2554,6 +2557,9 @@ int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 		break;
 	}
 
+	if ((_got & CEPH_CAP_FILE_RD) && (_got & CEPH_CAP_FILE_CACHE))
+		ceph_fscache_revalidate_cookie(ci);
+
 	*got = _got;
 	return 0;
 }
@@ -2795,7 +2801,6 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 	bool writeback = false;
 	bool queue_trunc = false;
 	bool queue_invalidate = false;
-	bool queue_revalidate = false;
 	bool deleted_inode = false;
 	bool fill_inline = false;
 
@@ -2837,8 +2842,6 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 				ci->i_rdcache_revoking = ci->i_rdcache_gen;
 			}
 		}
-
-		ceph_fscache_invalidate(inode);
 	}
 
 	/* side effects now are allowed */
@@ -2880,11 +2883,6 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 		}
 	}
 
-	/* Do we need to revalidate our fscache cookie. Don't bother on the
-	 * first cache cap as we already validate at cookie creation time. */
-	if ((issued & CEPH_CAP_FILE_CACHE) && ci->i_rdcache_gen > 1)
-		queue_revalidate = true;
-
 	if (newcaps & CEPH_CAP_ANY_RD) {
 		/* ctime/mtime/atime? */
 		ceph_decode_timespec(&mtime, &grant->mtime);
@@ -2993,11 +2991,8 @@ static void handle_cap_grant(struct ceph_mds_client *mdsc,
 	if (fill_inline)
 		ceph_fill_inline_data(inode, NULL, inline_data, inline_len);
 
-	if (queue_trunc) {
+	if (queue_trunc)
 		ceph_queue_vmtruncate(inode);
-		ceph_queue_revalidate(inode);
-	} else if (queue_revalidate)
-		ceph_queue_revalidate(inode);
 
 	if (writeback)
 		/*
@@ -3199,10 +3194,8 @@ static void handle_cap_trunc(struct inode *inode,
 					  truncate_seq, truncate_size, size);
 	spin_unlock(&ci->i_ceph_lock);
 
-	if (queue_trunc) {
+	if (queue_trunc)
 		ceph_queue_vmtruncate(inode);
-		ceph_fscache_invalidate(inode);
-	}
 }
 
 /*
diff --git a/fs/ceph/export.c b/fs/ceph/export.c
index 6e72c98162d5..1780218a48f0 100644
--- a/fs/ceph/export.c
+++ b/fs/ceph/export.c
@@ -95,10 +95,8 @@ static struct dentry *__fh_to_dentry(struct super_block *sb, u64 ino)
 	}
 
 	dentry = d_obtain_alias(inode);
-	if (IS_ERR(dentry)) {
-		iput(inode);
+	if (IS_ERR(dentry))
 		return dentry;
-	}
 	err = ceph_init_dentry(dentry);
 	if (err < 0) {
 		dput(dentry);
@@ -167,10 +165,8 @@ static struct dentry *__get_parent(struct super_block *sb,
 		return ERR_PTR(-ENOENT);
 
 	dentry = d_obtain_alias(inode);
-	if (IS_ERR(dentry)) {
-		iput(inode);
+	if (IS_ERR(dentry))
 		return dentry;
-	}
 	err = ceph_init_dentry(dentry);
 	if (err < 0) {
 		dput(dentry);
@@ -210,7 +206,7 @@ static struct dentry *ceph_fh_to_parent(struct super_block *sb,
 
 	dout("fh_to_parent %llx\n", cfh->parent_ino);
 	dentry = __get_parent(sb, NULL, cfh->ino);
-	if (IS_ERR(dentry) && PTR_ERR(dentry) == -ENOENT)
+	if (unlikely(dentry == ERR_PTR(-ENOENT)))
 		dentry = __fh_to_dentry(sb, cfh->parent_ino);
 	return dentry;
 }
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index a888df6f2d71..0daaf7ceedc5 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -137,23 +137,11 @@ static int ceph_init_file(struct inode *inode, struct file *file, int fmode)
 {
 	struct ceph_file_info *cf;
 	int ret = 0;
-	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb);
-	struct ceph_mds_client *mdsc = fsc->mdsc;
 
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFREG:
-		/* First file open request creates the cookie, we want to keep
-		 * this cookie around for the filetime of the inode as not to
-		 * have to worry about fscache register / revoke / operation
-		 * races.
-		 *
-		 * Also, if we know the operation is going to invalidate data
-		 * (non readonly) just nuke the cache right away.
-		 */
-		ceph_fscache_register_inode_cookie(mdsc->fsc, ci);
-		if ((fmode & CEPH_FILE_MODE_WR))
-			ceph_fscache_invalidate(inode);
+		ceph_fscache_register_inode_cookie(inode);
+		ceph_fscache_file_set_cookie(inode, file);
 	case S_IFDIR:
 		dout("init_file %p %p 0%o (regular)\n", inode, file,
 		     inode->i_mode);
@@ -406,7 +394,7 @@ int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
 	if ((flags & O_CREAT) && !req->r_reply_info.head->is_dentry)
 		err = ceph_handle_notrace_create(dir, dentry);
 
-	if (d_unhashed(dentry)) {
+	if (d_in_lookup(dentry)) {
 		dn = ceph_finish_lookup(req, dentry, err);
 		if (IS_ERR(dn))
 			err = PTR_ERR(dn);
@@ -1349,7 +1337,7 @@ static ssize_t ceph_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	}
 
 retry_snap:
-	if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL)) {
+	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL)) {
 		err = -ENOSPC;
 		goto out;
 	}
@@ -1407,7 +1395,6 @@ retry_snap:
 			iov_iter_advance(from, written);
 		ceph_put_snap_context(snapc);
 	} else {
-		loff_t old_size = i_size_read(inode);
 		/*
 		 * No need to acquire the i_truncate_mutex. Because
 		 * the MDS revokes Fwb caps before sending truncate
@@ -1418,8 +1405,6 @@ retry_snap:
 		written = generic_perform_write(file, from, pos);
 		if (likely(written >= 0))
 			iocb->ki_pos = pos + written;
-		if (i_size_read(inode) > old_size)
-			ceph_fscache_update_objectsize(inode);
 		inode_unlock(inode);
 	}
 
@@ -1440,7 +1425,7 @@ retry_snap:
 	ceph_put_cap_refs(ci, got);
 
 	if (written >= 0) {
-		if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_NEARFULL))
+		if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_NEARFULL))
 			iocb->ki_flags |= IOCB_DSYNC;
 
 		written = generic_write_sync(iocb, written);
@@ -1672,8 +1657,8 @@ static long ceph_fallocate(struct file *file, int mode,
 		goto unlock;
 	}
 
-	if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL) &&
-		!(mode & FALLOC_FL_PUNCH_HOLE)) {
+	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) &&
+	    !(mode & FALLOC_FL_PUNCH_HOLE)) {
 		ret = -ENOSPC;
 		goto unlock;
 	}
diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index 0130a8592191..0168b49fb6ad 100644
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -103,7 +103,6 @@ struct ceph_fs_client {
 
 #ifdef CONFIG_CEPH_FSCACHE
 	struct fscache_cookie *fscache;
-	struct workqueue_struct *revalidate_wq;
 #endif
 };
 
@@ -360,8 +359,7 @@ struct ceph_inode_info {
 
 #ifdef CONFIG_CEPH_FSCACHE
 	struct fscache_cookie *fscache;
-	u32 i_fscache_gen; /* sequence, for delayed fscache validate */
-	struct work_struct i_revalidate_work;
+	u32 i_fscache_gen;
 #endif
 	struct inode vfs_inode; /* at end */
 };