ceph: directory operations

Directory operations, including lookup, are defined here.  We take
advantage of lookup intents when possible.  For the most part, we just
need to build the proper requests for the metadata server(s) and
pass things off to the mds_client.

The results of most operations are normally incorporated into the
client's cache when the reply is parsed by ceph_fill_trace().
However, if the MDS replies without a trace (e.g., when retrying an
update after an MDS failure recovery), some operation-specific cleanup
may be needed.

We can validate cached dentries in two ways.  A per-dentry lease may
be issued by the MDS, or a per-directory cap may be issued that acts
as a lease on the entire directory.  In the latter case, a 'gen' value
is used to determine which dentries belong to the currently leased
directory contents.

We normally prepopulate the dcache and icache with readdir results.
This makes subsequent lookups and getattrs avoid any server
interaction.  It also lets us satisfy readdir operation by peeking at
the dcache IFF we hold the per-directory cap/lease, previously
performed a readdir, and haven't dropped any of the resulting
dentries.

Signed-off-by: Sage Weil <sage@newdream.net>

1 files changed, 1212 insertions, 0 deletions

diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
new file mode 100644
index 00000000000..7bb8db524e5
--- /dev/null
+++ b/fs/ceph/dir.c
@@ -0,0 +1,1212 @@
+#include "ceph_debug.h"
+
+#include <linux/spinlock.h>
+#include <linux/fs_struct.h>
+#include <linux/namei.h>
+#include <linux/sched.h>
+
+#include "super.h"
+
+/*
+ * Directory operations: readdir, lookup, create, link, unlink,
+ * rename, etc.
+ */
+
+/*
+ * Ceph MDS operations are specified in terms of a base ino and
+ * relative path.  Thus, the client can specify an operation on a
+ * specific inode (e.g., a getattr due to fstat(2)), or as a path
+ * relative to, say, the root directory.
+ *
+ * Normally, we limit ourselves to strict inode ops (no path component)
+ * or dentry operations (a single path component relative to an ino).  The
+ * exception to this is open_root_dentry(), which will open the mount
+ * point by name.
+ */
+
+const struct inode_operations ceph_dir_iops;
+const struct file_operations ceph_dir_fops;
+struct dentry_operations ceph_dentry_ops;
+
+/*
+ * Initialize ceph dentry state.
+ */
+int ceph_init_dentry(struct dentry *dentry)
+{
+	struct ceph_dentry_info *di;
+
+	if (dentry->d_fsdata)
+		return 0;
+
+	if (ceph_snap(dentry->d_parent->d_inode) == CEPH_NOSNAP)
+		dentry->d_op = &ceph_dentry_ops;
+	else if (ceph_snap(dentry->d_parent->d_inode) == CEPH_SNAPDIR)
+		dentry->d_op = &ceph_snapdir_dentry_ops;
+	else
+		dentry->d_op = &ceph_snap_dentry_ops;
+
+	di = kmem_cache_alloc(ceph_dentry_cachep, GFP_NOFS);
+	if (!di)
+		return -ENOMEM;          /* oh well */
+
+	spin_lock(&dentry->d_lock);
+	if (dentry->d_fsdata) /* lost a race */
+		goto out_unlock;
+	di->dentry = dentry;
+	di->lease_session = NULL;
+	dentry->d_fsdata = di;
+	dentry->d_time = jiffies;
+	ceph_dentry_lru_add(dentry);
+out_unlock:
+	spin_unlock(&dentry->d_lock);
+	return 0;
+}
+
+
+
+/*
+ * for readdir, we encode the directory frag and offset within that
+ * frag into f_pos.
+ */
+static unsigned fpos_frag(loff_t p)
+{
+	return p >> 32;
+}
+static unsigned fpos_off(loff_t p)
+{
+	return p & 0xffffffff;
+}
+
+/*
+ * When possible, we try to satisfy a readdir by peeking at the
+ * dcache.  We make this work by carefully ordering dentries on
+ * d_u.d_child when we initially get results back from the MDS, and
+ * falling back to a "normal" sync readdir if any dentries in the dir
+ * are dropped.
+ *
+ * I_COMPLETE tells indicates we have all dentries in the dir.  It is
+ * defined IFF we hold CEPH_CAP_FILE_SHARED (which will be revoked by
+ * the MDS if/when the directory is modified).
+ */
+static int __dcache_readdir(struct file *filp,
+			    void *dirent, filldir_t filldir)
+{
+	struct inode *inode = filp->f_dentry->d_inode;
+	struct ceph_file_info *fi = filp->private_data;
+	struct dentry *parent = filp->f_dentry;
+	struct inode *dir = parent->d_inode;
+	struct list_head *p;
+	struct dentry *dentry, *last;
+	struct ceph_dentry_info *di;
+	int err = 0;
+
+	/* claim ref on last dentry we returned */
+	last = fi->dentry;
+	fi->dentry = NULL;
+
+	dout("__dcache_readdir %p at %llu (last %p)\n", dir, filp->f_pos,
+	     last);
+
+	spin_lock(&dcache_lock);
+
+	/* start at beginning? */
+	if (filp->f_pos == 2 || (last &&
+				 filp->f_pos < ceph_dentry(last)->offset)) {
+		if (list_empty(&parent->d_subdirs))
+			goto out_unlock;
+		p = parent->d_subdirs.prev;
+		dout(" initial p %p/%p\n", p->prev, p->next);
+	} else {
+		p = last->d_u.d_child.prev;
+	}
+
+more:
+	dentry = list_entry(p, struct dentry, d_u.d_child);
+	di = ceph_dentry(dentry);
+	while (1) {
+		dout(" p %p/%p d_subdirs %p/%p\n", p->prev, p->next,
+		     parent->d_subdirs.prev, parent->d_subdirs.next);
+		if (p == &parent->d_subdirs) {
+			fi->at_end = 1;
+			goto out_unlock;
+		}
+		if (!d_unhashed(dentry) && dentry->d_inode &&
+		    filp->f_pos <= di->offset)
+			break;
+		dout(" skipping %p %.*s at %llu (%llu)%s%s\n", dentry,
+		     dentry->d_name.len, dentry->d_name.name, di->offset,
+		     filp->f_pos, d_unhashed(dentry) ? " unhashed" : "",
+		     !dentry->d_inode ? " null" : "");
+		p = p->prev;
+		dentry = list_entry(p, struct dentry, d_u.d_child);
+		di = ceph_dentry(dentry);
+	}
+
+	atomic_inc(&dentry->d_count);
+	spin_unlock(&dcache_lock);
+	spin_unlock(&inode->i_lock);
+
+	dout(" %llu (%llu) dentry %p %.*s %p\n", di->offset, filp->f_pos,
+	     dentry, dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
+	filp->f_pos = di->offset;
+	err = filldir(dirent, dentry->d_name.name,
+		      dentry->d_name.len, di->offset,
+		      dentry->d_inode->i_ino,
+		      dentry->d_inode->i_mode >> 12);
+
+	if (last) {
+		if (err < 0) {
+			/* remember our position */
+			fi->dentry = last;
+			fi->next_offset = di->offset;
+		} else {
+			dput(last);
+		}
+		last = NULL;
+	}
+
+	spin_lock(&inode->i_lock);
+	spin_lock(&dcache_lock);
+
+	if (err < 0)
+		goto out_unlock;
+
+	last = dentry;
+
+	p = p->prev;
+	filp->f_pos++;
+
+	/* make sure a dentry wasn't dropped while we didn't have dcache_lock */
+	if ((ceph_inode(dir)->i_ceph_flags & CEPH_I_COMPLETE))
+		goto more;
+	dout(" lost I_COMPLETE on %p; falling back to mds\n", dir);
+	err = -EAGAIN;
+
+out_unlock:
+	spin_unlock(&dcache_lock);
+
+	if (last) {
+		spin_unlock(&inode->i_lock);
+		dput(last);
+		spin_lock(&inode->i_lock);
+	}
+
+	return err;
+}
+
+/*
+ * make note of the last dentry we read, so we can
+ * continue at the same lexicographical point,
+ * regardless of what dir changes take place on the
+ * server.
+ */
+static int note_last_dentry(struct ceph_file_info *fi, const char *name,
+			    int len)
+{
+	kfree(fi->last_name);
+	fi->last_name = kmalloc(len+1, GFP_NOFS);
+	if (!fi->last_name)
+		return -ENOMEM;
+	memcpy(fi->last_name, name, len);
+	fi->last_name[len] = 0;
+	dout("note_last_dentry '%s'\n", fi->last_name);
+	return 0;
+}
+
+static int ceph_readdir(struct file *filp, void *dirent, filldir_t filldir)
+{
+	struct ceph_file_info *fi = filp->private_data;
+	struct inode *inode = filp->f_dentry->d_inode;
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	struct ceph_client *client = ceph_inode_to_client(inode);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	unsigned frag = fpos_frag(filp->f_pos);
+	int off = fpos_off(filp->f_pos);
+	int err;
+	u32 ftype;
+	struct ceph_mds_reply_info_parsed *rinfo;
+	const int max_entries = client->mount_args.max_readdir;
+
+	dout("readdir %p filp %p frag %u off %u\n", inode, filp, frag, off);
+	if (fi->at_end)
+		return 0;
+
+	/* always start with . and .. */
+	if (filp->f_pos == 0) {
+		/* note dir version at start of readdir so we can tell
+		 * if any dentries get dropped */
+		fi->dir_release_count = ci->i_release_count;
+
+		dout("readdir off 0 -> '.'\n");
+		if (filldir(dirent, ".", 1, ceph_make_fpos(0, 0),
+			    inode->i_ino, inode->i_mode >> 12) < 0)
+			return 0;
+		filp->f_pos = 1;
+		off = 1;
+	}
+	if (filp->f_pos == 1) {
+		dout("readdir off 1 -> '..'\n");
+		if (filldir(dirent, "..", 2, ceph_make_fpos(0, 1),
+			    filp->f_dentry->d_parent->d_inode->i_ino,
+			    inode->i_mode >> 12) < 0)
+			return 0;
+		filp->f_pos = 2;
+		off = 2;
+	}
+
+	/* can we use the dcache? */
+	spin_lock(&inode->i_lock);
+	if ((filp->f_pos == 2 || fi->dentry) &&
+	    !ceph_test_opt(client, NOASYNCREADDIR) &&
+	    (ci->i_ceph_flags & CEPH_I_COMPLETE) &&
+	    __ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1)) {
+		err = __dcache_readdir(filp, dirent, filldir);
+		if (err != -EAGAIN) {
+			spin_unlock(&inode->i_lock);
+			return err;
+		}
+	}
+	spin_unlock(&inode->i_lock);
+	if (fi->dentry) {
+		err = note_last_dentry(fi, fi->dentry->d_name.name,
+				       fi->dentry->d_name.len);
+		if (err)
+			return err;
+		dput(fi->dentry);
+		fi->dentry = NULL;
+	}
+
+	/* proceed with a normal readdir */
+
+more:
+	/* do we have the correct frag content buffered? */
+	if (fi->frag != frag || fi->last_readdir == NULL) {
+		struct ceph_mds_request *req;
+		int op = ceph_snap(inode) == CEPH_SNAPDIR ?
+			CEPH_MDS_OP_LSSNAP : CEPH_MDS_OP_READDIR;
+
+		/* discard old result, if any */
+		if (fi->last_readdir)
+			ceph_mdsc_put_request(fi->last_readdir);
+
+		/* requery frag tree, as the frag topology may have changed */
+		frag = ceph_choose_frag(ceph_inode(inode), frag, NULL, NULL);
+
+		dout("readdir fetching %llx.%llx frag %x offset '%s'\n",
+		     ceph_vinop(inode), frag, fi->last_name);
+		req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
+		if (IS_ERR(req))
+			return PTR_ERR(req);
+		req->r_inode = igrab(inode);
+		req->r_dentry = dget(filp->f_dentry);
+		/* hints to request -> mds selection code */
+		req->r_direct_mode = USE_AUTH_MDS;
+		req->r_direct_hash = ceph_frag_value(frag);
+		req->r_direct_is_hash = true;
+		req->r_path2 = kstrdup(fi->last_name, GFP_NOFS);
+		req->r_readdir_offset = fi->next_offset;
+		req->r_args.readdir.frag = cpu_to_le32(frag);
+		req->r_args.readdir.max_entries = cpu_to_le32(max_entries);
+		req->r_num_caps = max_entries;
+		err = ceph_mdsc_do_request(mdsc, NULL, req);
+		if (err < 0) {
+			ceph_mdsc_put_request(req);
+			return err;
+		}
+		dout("readdir got and parsed readdir result=%d"
+		     " on frag %x, end=%d, complete=%d\n", err, frag,
+		     (int)req->r_reply_info.dir_end,
+		     (int)req->r_reply_info.dir_complete);
+
+		if (!req->r_did_prepopulate) {
+			dout("readdir !did_prepopulate");
+			fi->dir_release_count--;    /* preclude I_COMPLETE */
+		}
+
+		/* note next offset and last dentry name */
+		fi->offset = fi->next_offset;
+		fi->last_readdir = req;
+
+		if (req->r_reply_info.dir_end) {
+			kfree(fi->last_name);
+			fi->last_name = NULL;
+			fi->next_offset = 0;
+		} else {
+			rinfo = &req->r_reply_info;
+			err = note_last_dentry(fi,
+				       rinfo->dir_dname[rinfo->dir_nr-1],
+				       rinfo->dir_dname_len[rinfo->dir_nr-1]);
+			if (err)
+				return err;
+			fi->next_offset += rinfo->dir_nr;
+		}
+	}
+
+	rinfo = &fi->last_readdir->r_reply_info;
+	dout("readdir frag %x num %d off %d chunkoff %d\n", frag,
+	     rinfo->dir_nr, off, fi->offset);
+	while (off - fi->offset >= 0 && off - fi->offset < rinfo->dir_nr) {
+		u64 pos = ceph_make_fpos(frag, off);
+		struct ceph_mds_reply_inode *in =
+			rinfo->dir_in[off - fi->offset].in;
+		dout("readdir off %d (%d/%d) -> %lld '%.*s' %p\n",
+		     off, off - fi->offset, rinfo->dir_nr, pos,
+		     rinfo->dir_dname_len[off - fi->offset],
+		     rinfo->dir_dname[off - fi->offset], in);
+		BUG_ON(!in);
+		ftype = le32_to_cpu(in->mode) >> 12;
+		if (filldir(dirent,
+			    rinfo->dir_dname[off - fi->offset],
+			    rinfo->dir_dname_len[off - fi->offset],
+			    pos,
+			    le64_to_cpu(in->ino),
+			    ftype) < 0) {
+			dout("filldir stopping us...\n");
+			return 0;
+		}
+		off++;
+		filp->f_pos = pos + 1;
+	}
+
+	if (fi->last_name) {
+		ceph_mdsc_put_request(fi->last_readdir);
+		fi->last_readdir = NULL;
+		goto more;
+	}
+
+	/* more frags? */
+	if (!ceph_frag_is_rightmost(frag)) {
+		frag = ceph_frag_next(frag);
+		off = 0;
+		filp->f_pos = ceph_make_fpos(frag, off);
+		dout("readdir next frag is %x\n", frag);
+		goto more;
+	}
+	fi->at_end = 1;
+
+	/*
+	 * if dir_release_count still matches the dir, no dentries
+	 * were released during the whole readdir, and we should have
+	 * the complete dir contents in our cache.
+	 */
+	spin_lock(&inode->i_lock);
+	if (ci->i_release_count == fi->dir_release_count) {
+		dout(" marking %p complete\n", inode);
+		ci->i_ceph_flags |= CEPH_I_COMPLETE;
+		ci->i_max_offset = filp->f_pos;
+	}
+	spin_unlock(&inode->i_lock);
+
+	dout("readdir %p filp %p done.\n", inode, filp);
+	return 0;
+}
+
+static void reset_readdir(struct ceph_file_info *fi)
+{
+	if (fi->last_readdir) {
+		ceph_mdsc_put_request(fi->last_readdir);
+		fi->last_readdir = NULL;
+	}
+	kfree(fi->last_name);
+	fi->next_offset = 2;  /* compensate for . and .. */
+	if (fi->dentry) {
+		dput(fi->dentry);
+		fi->dentry = NULL;
+	}
+	fi->at_end = 0;
+}
+
+static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int origin)
+{
+	struct ceph_file_info *fi = file->private_data;
+	struct inode *inode = file->f_mapping->host;
+	loff_t old_offset = offset;
+	loff_t retval;
+
+	mutex_lock(&inode->i_mutex);
+	switch (origin) {
+	case SEEK_END:
+		offset += inode->i_size + 2;   /* FIXME */
+		break;
+	case SEEK_CUR:
+		offset += file->f_pos;
+	}
+	retval = -EINVAL;
+	if (offset >= 0 && offset <= inode->i_sb->s_maxbytes) {
+		if (offset != file->f_pos) {
+			file->f_pos = offset;
+			file->f_version = 0;
+			fi->at_end = 0;
+		}
+		retval = offset;
+
+		/*
+		 * discard buffered readdir content on seekdir(0), or
+		 * seek to new frag, or seek prior to current chunk.
+		 */
+		if (offset == 0 ||
+		    fpos_frag(offset) != fpos_frag(old_offset) ||
+		    fpos_off(offset) < fi->offset) {
+			dout("dir_llseek dropping %p content\n", file);
+			reset_readdir(fi);
+		}
+
+		/* bump dir_release_count if we did a forward seek */
+		if (offset > old_offset)
+			fi->dir_release_count--;
+	}
+	mutex_unlock(&inode->i_mutex);
+	return retval;
+}
+
+/*
+ * Process result of a lookup/open request.
+ *
+ * Mainly, make sure we return the final req->r_dentry (if it already
+ * existed) in place of the original VFS-provided dentry when they
+ * differ.
+ *
+ * Gracefully handle the case where the MDS replies with -ENOENT and
+ * no trace (which it may do, at its discretion, e.g., if it doesn't
+ * care to issue a lease on the negative dentry).
+ */
+struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
+				  struct dentry *dentry, int err)
+{
+	struct ceph_client *client = ceph_client(dentry->d_sb);
+	struct inode *parent = dentry->d_parent->d_inode;
+
+	/* .snap dir? */
+	if (err == -ENOENT &&
+	    ceph_vino(parent).ino != CEPH_INO_ROOT && /* no .snap in root dir */
+	    strcmp(dentry->d_name.name, client->mount_args.snapdir_name) == 0) {
+		struct inode *inode = ceph_get_snapdir(parent);
+		dout("ENOENT on snapdir %p '%.*s', linking to snapdir %p\n",
+		     dentry, dentry->d_name.len, dentry->d_name.name, inode);
+		d_add(dentry, inode);
+		err = 0;
+	}
+
+	if (err == -ENOENT) {
+		/* no trace? */
+		err = 0;
+		if (!req->r_reply_info.head->is_dentry) {
+			dout("ENOENT and no trace, dentry %p inode %p\n",
+			     dentry, dentry->d_inode);
+			if (dentry->d_inode) {
+				d_drop(dentry);
+				err = -ENOENT;
+			} else {
+				d_add(dentry, NULL);
+			}
+		}
+	}
+	if (err)
+		dentry = ERR_PTR(err);
+	else if (dentry != req->r_dentry)
+		dentry = dget(req->r_dentry);   /* we got spliced */
+	else
+		dentry = NULL;
+	return dentry;
+}
+
+/*
+ * Look up a single dir entry.  If there is a lookup intent, inform
+ * the MDS so that it gets our 'caps wanted' value in a single op.
+ */
+static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
+				  struct nameidata *nd)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int op;
+	int err;
+
+	dout("lookup %p dentry %p '%.*s'\n",
+	     dir, dentry, dentry->d_name.len, dentry->d_name.name);
+
+	if (dentry->d_name.len > NAME_MAX)
+		return ERR_PTR(-ENAMETOOLONG);
+
+	err = ceph_init_dentry(dentry);
+	if (err < 0)
+		return ERR_PTR(err);
+
+	/* open (but not create!) intent? */
+	if (nd &&
+	    (nd->flags & LOOKUP_OPEN) &&
+	    (nd->flags & LOOKUP_CONTINUE) == 0 && /* only open last component */
+	    !(nd->intent.open.flags & O_CREAT)) {
+		int mode = nd->intent.open.create_mode & ~current->fs->umask;
+		return ceph_lookup_open(dir, dentry, nd, mode, 1);
+	}
+
+	/* can we conclude ENOENT locally? */
+	if (dentry->d_inode == NULL) {
+		struct ceph_inode_info *ci = ceph_inode(dir);
+		struct ceph_dentry_info *di = ceph_dentry(dentry);
+
+		spin_lock(&dir->i_lock);
+		dout(" dir %p flags are %d\n", dir, ci->i_ceph_flags);
+		if (strncmp(dentry->d_name.name,
+			    client->mount_args.snapdir_name,
+			    dentry->d_name.len) &&
+		    (ci->i_ceph_flags & CEPH_I_COMPLETE) &&
+		    (__ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1))) {
+			di->offset = ci->i_max_offset++;
+			spin_unlock(&dir->i_lock);
+			dout(" dir %p complete, -ENOENT\n", dir);
+			d_add(dentry, NULL);
+			di->lease_shared_gen = ci->i_shared_gen;
+			return NULL;
+		}
+		spin_unlock(&dir->i_lock);
+	}
+
+	op = ceph_snap(dir) == CEPH_SNAPDIR ?
+		CEPH_MDS_OP_LOOKUPSNAP : CEPH_MDS_OP_LOOKUP;
+	req = ceph_mdsc_create_request(mdsc, op, USE_ANY_MDS);
+	if (IS_ERR(req))
+		return ERR_PTR(PTR_ERR(req));
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	/* we only need inode linkage */
+	req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
+	req->r_locked_dir = dir;
+	err = ceph_mdsc_do_request(mdsc, NULL, req);
+	dentry = ceph_finish_lookup(req, dentry, err);
+	ceph_mdsc_put_request(req);  /* will dput(dentry) */
+	dout("lookup result=%p\n", dentry);
+	return dentry;
+}
+
+/*
+ * If we do a create but get no trace back from the MDS, follow up with
+ * a lookup (the VFS expects us to link up the provided dentry).
+ */
+int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry)
+{
+	struct dentry *result = ceph_lookup(dir, dentry, NULL);
+
+	if (result && !IS_ERR(result)) {
+		/*
+		 * We created the item, then did a lookup, and found
+		 * it was already linked to another inode we already
+		 * had in our cache (and thus got spliced).  Link our
+		 * dentry to that inode, but don't hash it, just in
+		 * case the VFS wants to dereference it.
+		 */
+		BUG_ON(!result->d_inode);
+		d_instantiate(dentry, result->d_inode);
+		return 0;
+	}
+	return PTR_ERR(result);
+}
+
+static int ceph_mknod(struct inode *dir, struct dentry *dentry,
+		      int mode, dev_t rdev)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int err;
+
+	if (ceph_snap(dir) != CEPH_NOSNAP)
+		return -EROFS;
+
+	dout("mknod in dir %p dentry %p mode 0%o rdev %d\n",
+	     dir, dentry, mode, rdev);
+	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_MKNOD, USE_AUTH_MDS);
+	if (IS_ERR(req)) {
+		d_drop(dentry);
+		return PTR_ERR(req);
+	}
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	req->r_locked_dir = dir;
+	req->r_args.mknod.mode = cpu_to_le32(mode);
+	req->r_args.mknod.rdev = cpu_to_le32(rdev);
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	err = ceph_mdsc_do_request(mdsc, dir, req);
+	if (!err && !req->r_reply_info.head->is_dentry)
+		err = ceph_handle_notrace_create(dir, dentry);
+	ceph_mdsc_put_request(req);
+	if (err)
+		d_drop(dentry);
+	return err;
+}
+
+static int ceph_create(struct inode *dir, struct dentry *dentry, int mode,
+		       struct nameidata *nd)
+{
+	dout("create in dir %p dentry %p name '%.*s'\n",
+	     dir, dentry, dentry->d_name.len, dentry->d_name.name);
+
+	if (ceph_snap(dir) != CEPH_NOSNAP)
+		return -EROFS;
+
+	if (nd) {
+		BUG_ON((nd->flags & LOOKUP_OPEN) == 0);
+		dentry = ceph_lookup_open(dir, dentry, nd, mode, 0);
+		/* hrm, what should i do here if we get aliased? */
+		if (IS_ERR(dentry))
+			return PTR_ERR(dentry);
+		return 0;
+	}
+
+	/* fall back to mknod */
+	return ceph_mknod(dir, dentry, (mode & ~S_IFMT) | S_IFREG, 0);
+}
+
+static int ceph_symlink(struct inode *dir, struct dentry *dentry,
+			    const char *dest)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int err;
+
+	if (ceph_snap(dir) != CEPH_NOSNAP)
+		return -EROFS;
+
+	dout("symlink in dir %p dentry %p to '%s'\n", dir, dentry, dest);
+	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_SYMLINK, USE_AUTH_MDS);
+	if (IS_ERR(req)) {
+		d_drop(dentry);
+		return PTR_ERR(req);
+	}
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	req->r_path2 = kstrdup(dest, GFP_NOFS);
+	req->r_locked_dir = dir;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	err = ceph_mdsc_do_request(mdsc, dir, req);
+	if (!err && !req->r_reply_info.head->is_dentry)
+		err = ceph_handle_notrace_create(dir, dentry);
+	ceph_mdsc_put_request(req);
+	if (err)
+		d_drop(dentry);
+	return err;
+}
+
+static int ceph_mkdir(struct inode *dir, struct dentry *dentry, int mode)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int err = -EROFS;
+	int op;
+
+	if (ceph_snap(dir) == CEPH_SNAPDIR) {
+		/* mkdir .snap/foo is a MKSNAP */
+		op = CEPH_MDS_OP_MKSNAP;
+		dout("mksnap dir %p snap '%.*s' dn %p\n", dir,
+		     dentry->d_name.len, dentry->d_name.name, dentry);
+	} else if (ceph_snap(dir) == CEPH_NOSNAP) {
+		dout("mkdir dir %p dn %p mode 0%o\n", dir, dentry, mode);
+		op = CEPH_MDS_OP_MKDIR;
+	} else {
+		goto out;
+	}
+	req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
+	if (IS_ERR(req)) {
+		err = PTR_ERR(req);
+		goto out;
+	}
+
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	req->r_locked_dir = dir;
+	req->r_args.mkdir.mode = cpu_to_le32(mode);
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	err = ceph_mdsc_do_request(mdsc, dir, req);
+	if (!err && !req->r_reply_info.head->is_dentry)
+		err = ceph_handle_notrace_create(dir, dentry);
+	ceph_mdsc_put_request(req);
+out:
+	if (err < 0)
+		d_drop(dentry);
+	return err;
+}
+
+static int ceph_link(struct dentry *old_dentry, struct inode *dir,
+		     struct dentry *dentry)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int err;
+
+	if (ceph_snap(dir) != CEPH_NOSNAP)
+		return -EROFS;
+
+	dout("link in dir %p old_dentry %p dentry %p\n", dir,
+	     old_dentry, dentry);
+	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LINK, USE_AUTH_MDS);
+	if (IS_ERR(req)) {
+		d_drop(dentry);
+		return PTR_ERR(req);
+	}
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	req->r_old_dentry = dget(old_dentry); /* or inode? hrm. */
+	req->r_locked_dir = dir;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	err = ceph_mdsc_do_request(mdsc, dir, req);
+	if (err)
+		d_drop(dentry);
+	else if (!req->r_reply_info.head->is_dentry)
+		d_instantiate(dentry, igrab(old_dentry->d_inode));
+	ceph_mdsc_put_request(req);
+	return err;
+}
+
+/*
+ * For a soon-to-be unlinked file, drop the AUTH_RDCACHE caps.  If it
+ * looks like the link count will hit 0, drop any other caps (other
+ * than PIN) we don't specifically want (due to the file still being
+ * open).
+ */
+static int drop_caps_for_unlink(struct inode *inode)
+{
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	int drop = CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL;
+
+	spin_lock(&inode->i_lock);
+	if (inode->i_nlink == 1) {
+		drop |= ~(__ceph_caps_wanted(ci) | CEPH_CAP_PIN);
+		ci->i_ceph_flags |= CEPH_I_NODELAY;
+	}
+	spin_unlock(&inode->i_lock);
+	return drop;
+}
+
+/*
+ * rmdir and unlink are differ only by the metadata op code
+ */
+static int ceph_unlink(struct inode *dir, struct dentry *dentry)
+{
+	struct ceph_client *client = ceph_sb_to_client(dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct inode *inode = dentry->d_inode;
+	struct ceph_mds_request *req;
+	int err = -EROFS;
+	int op;
+
+	if (ceph_snap(dir) == CEPH_SNAPDIR) {
+		/* rmdir .snap/foo is RMSNAP */
+		dout("rmsnap dir %p '%.*s' dn %p\n", dir, dentry->d_name.len,
+		     dentry->d_name.name, dentry);
+		op = CEPH_MDS_OP_RMSNAP;
+	} else if (ceph_snap(dir) == CEPH_NOSNAP) {
+		dout("unlink/rmdir dir %p dn %p inode %p\n",
+		     dir, dentry, inode);
+		op = ((dentry->d_inode->i_mode & S_IFMT) == S_IFDIR) ?
+			CEPH_MDS_OP_RMDIR : CEPH_MDS_OP_UNLINK;
+	} else
+		goto out;
+	req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
+	if (IS_ERR(req)) {
+		err = PTR_ERR(req);
+		goto out;
+	}
+	req->r_dentry = dget(dentry);
+	req->r_num_caps = 2;
+	req->r_locked_dir = dir;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	req->r_inode_drop = drop_caps_for_unlink(inode);
+	err = ceph_mdsc_do_request(mdsc, dir, req);
+	if (!err && !req->r_reply_info.head->is_dentry)
+		d_delete(dentry);
+	ceph_mdsc_put_request(req);
+out:
+	return err;
+}
+
+static int ceph_rename(struct inode *old_dir, struct dentry *old_dentry,
+		       struct inode *new_dir, struct dentry *new_dentry)
+{
+	struct ceph_client *client = ceph_sb_to_client(old_dir->i_sb);
+	struct ceph_mds_client *mdsc = &client->mdsc;
+	struct ceph_mds_request *req;
+	int err;
+
+	if (ceph_snap(old_dir) != ceph_snap(new_dir))
+		return -EXDEV;
+	if (ceph_snap(old_dir) != CEPH_NOSNAP ||
+	    ceph_snap(new_dir) != CEPH_NOSNAP)
+		return -EROFS;
+	dout("rename dir %p dentry %p to dir %p dentry %p\n",
+	     old_dir, old_dentry, new_dir, new_dentry);
+	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_RENAME, USE_AUTH_MDS);
+	if (IS_ERR(req))
+		return PTR_ERR(req);
+	req->r_dentry = dget(new_dentry);
+	req->r_num_caps = 2;
+	req->r_old_dentry = dget(old_dentry);
+	req->r_locked_dir = new_dir;
+	req->r_old_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_old_dentry_unless = CEPH_CAP_FILE_EXCL;
+	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
+	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
+	/* release LINK_RDCACHE on source inode (mds will lock it) */
+	req->r_old_inode_drop = CEPH_CAP_LINK_SHARED;
+	if (new_dentry->d_inode)
+		req->r_inode_drop = drop_caps_for_unlink(new_dentry->d_inode);
+	err = ceph_mdsc_do_request(mdsc, old_dir, req);
+	if (!err && !req->r_reply_info.head->is_dentry) {
+		/*
+		 * Normally d_move() is done by fill_trace (called by
+		 * do_request, above).  If there is no trace, we need
+		 * to do it here.
+		 */
+		d_move(old_dentry, new_dentry);
+	}
+	ceph_mdsc_put_request(req);
+	return err;
+}
+
+
+/*
+ * Check if dentry lease is valid.  If not, delete the lease.  Try to
+ * renew if the least is more than half up.
+ */
+static int dentry_lease_is_valid(struct dentry *dentry)
+{
+	struct ceph_dentry_info *di;
+	struct ceph_mds_session *s;
+	int valid = 0;
+	u32 gen;
+	unsigned long ttl;
+	struct ceph_mds_session *session = NULL;
+	struct inode *dir = NULL;
+	u32 seq = 0;
+
+	spin_lock(&dentry->d_lock);
+	di = ceph_dentry(dentry);
+	if (di && di->lease_session) {
+		s = di->lease_session;
+		spin_lock(&s->s_cap_lock);
+		gen = s->s_cap_gen;
+		ttl = s->s_cap_ttl;
+		spin_unlock(&s->s_cap_lock);
+
+		if (di->lease_gen == gen &&
+		    time_before(jiffies, dentry->d_time) &&
+		    time_before(jiffies, ttl)) {
+			valid = 1;
+			if (di->lease_renew_after &&
+			    time_after(jiffies, di->lease_renew_after)) {
+				/* we should renew */
+				dir = dentry->d_parent->d_inode;
+				session = ceph_get_mds_session(s);
+				seq = di->lease_seq;
+				di->lease_renew_after = 0;
+				di->lease_renew_from = jiffies;
+			}
+		} else {
+			__ceph_mdsc_drop_dentry_lease(dentry);
+		}
+	}
+	spin_unlock(&dentry->d_lock);
+
+	if (session) {
+		ceph_mdsc_lease_send_msg(session, dir, dentry,
+					 CEPH_MDS_LEASE_RENEW, seq);
+		ceph_put_mds_session(session);
+	}
+	dout("dentry_lease_is_valid - dentry %p = %d\n", dentry, valid);
+	return valid;
+}
+
+/*
+ * Check if directory-wide content lease/cap is valid.
+ */
+static int dir_lease_is_valid(struct inode *dir, struct dentry *dentry)
+{
+	struct ceph_inode_info *ci = ceph_inode(dir);
+	struct ceph_dentry_info *di = ceph_dentry(dentry);
+	int valid = 0;
+
+	spin_lock(&dir->i_lock);
+	if (ci->i_shared_gen == di->lease_shared_gen)
+		valid = __ceph_caps_issued_mask(ci, CEPH_CAP_FILE_SHARED, 1);
+	spin_unlock(&dir->i_lock);
+	dout("dir_lease_is_valid dir %p v%u dentry %p v%u = %d\n",
+	     dir, (unsigned)ci->i_shared_gen, dentry,
+	     (unsigned)di->lease_shared_gen, valid);
+	return valid;
+}
+
+/*
+ * Check if cached dentry can be trusted.
+ */
+static int ceph_d_revalidate(struct dentry *dentry, struct nameidata *nd)
+{
+	struct inode *dir = dentry->d_parent->d_inode;
+
+	dout("d_revalidate %p '%.*s' inode %p\n", dentry,
+	     dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
+
+	/* always trust cached snapped dentries, snapdir dentry */
+	if (ceph_snap(dir) != CEPH_NOSNAP) {
+		dout("d_revalidate %p '%.*s' inode %p is SNAPPED\n", dentry,
+		     dentry->d_name.len, dentry->d_name.name, dentry->d_inode);
+		goto out_touch;
+	}
+	if (dentry->d_inode && ceph_snap(dentry->d_inode) == CEPH_SNAPDIR)
+		goto out_touch;
+
+	if (dentry_lease_is_valid(dentry) ||
+	    dir_lease_is_valid(dir, dentry))
+		goto out_touch;
+
+	dout("d_revalidate %p invalid\n", dentry);
+	d_drop(dentry);
+	return 0;
+out_touch:
+	ceph_dentry_lru_touch(dentry);
+	return 1;
+}
+
+/*
+ * When a dentry is released, clear the dir I_COMPLETE if it was part
+ * of the current dir gen.
+ */
+static void ceph_dentry_release(struct dentry *dentry)
+{
+	struct ceph_dentry_info *di = ceph_dentry(dentry);
+	struct inode *parent_inode = dentry->d_parent->d_inode;
+
+	if (parent_inode) {
+		struct ceph_inode_info *ci = ceph_inode(parent_inode);
+
+		spin_lock(&parent_inode->i_lock);
+		if (ci->i_shared_gen == di->lease_shared_gen) {
+			dout(" clearing %p complete (d_release)\n",
+			     parent_inode);
+			ci->i_ceph_flags &= ~CEPH_I_COMPLETE;
+			ci->i_release_count++;
+		}
+		spin_unlock(&parent_inode->i_lock);
+	}
+	if (di) {
+		ceph_dentry_lru_del(dentry);
+		if (di->lease_session)
+			ceph_put_mds_session(di->lease_session);
+		kmem_cache_free(ceph_dentry_cachep, di);
+		dentry->d_fsdata = NULL;
+	}
+}
+
+static int ceph_snapdir_d_revalidate(struct dentry *dentry,
+					  struct nameidata *nd)
+{
+	/*
+	 * Eventually, we'll want to revalidate snapped metadata
+	 * too... probably...
+	 */
+	return 1;
+}
+
+
+
+/*
+ * read() on a dir.  This weird interface hack only works if mounted
+ * with '-o dirstat'.
+ */
+static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size,
+			     loff_t *ppos)
+{
+	struct ceph_file_info *cf = file->private_data;
+	struct inode *inode = file->f_dentry->d_inode;
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	int left;
+
+	if (!ceph_test_opt(ceph_client(inode->i_sb), DIRSTAT))
+		return -EISDIR;
+
+	if (!cf->dir_info) {
+		cf->dir_info = kmalloc(1024, GFP_NOFS);
+		if (!cf->dir_info)
+			return -ENOMEM;
+		cf->dir_info_len =
+			sprintf(cf->dir_info,
+				"entries:   %20lld\n"
+				" files:    %20lld\n"
+				" subdirs:  %20lld\n"
+				"rentries:  %20lld\n"
+				" rfiles:   %20lld\n"
+				" rsubdirs: %20lld\n"
+				"rbytes:    %20lld\n"
+				"rctime:    %10ld.%09ld\n",
+				ci->i_files + ci->i_subdirs,
+				ci->i_files,
+				ci->i_subdirs,
+				ci->i_rfiles + ci->i_rsubdirs,
+				ci->i_rfiles,
+				ci->i_rsubdirs,
+				ci->i_rbytes,
+				(long)ci->i_rctime.tv_sec,
+				(long)ci->i_rctime.tv_nsec);
+	}
+
+	if (*ppos >= cf->dir_info_len)
+		return 0;
+	size = min_t(unsigned, size, cf->dir_info_len-*ppos);
+	left = copy_to_user(buf, cf->dir_info + *ppos, size);
+	if (left == size)
+		return -EFAULT;
+	*ppos += (size - left);
+	return size - left;
+}
+
+/*
+ * an fsync() on a dir will wait for any uncommitted directory
+ * operations to commit.
+ */
+static int ceph_dir_fsync(struct file *file, struct dentry *dentry,
+			  int datasync)
+{
+	struct inode *inode = dentry->d_inode;
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	struct list_head *head = &ci->i_unsafe_dirops;
+	struct ceph_mds_request *req;
+	u64 last_tid;
+	int ret = 0;
+
+	dout("dir_fsync %p\n", inode);
+	spin_lock(&ci->i_unsafe_lock);
+	if (list_empty(head))
+		goto out;
+
+	req = list_entry(head->prev,
+			 struct ceph_mds_request, r_unsafe_dir_item);
+	last_tid = req->r_tid;
+
+	do {
+		ceph_mdsc_get_request(req);
+		spin_unlock(&ci->i_unsafe_lock);
+		dout("dir_fsync %p wait on tid %llu (until %llu)\n",
+		     inode, req->r_tid, last_tid);
+		if (req->r_timeout) {
+			ret = wait_for_completion_timeout(
+				&req->r_safe_completion, req->r_timeout);
+			if (ret > 0)
+				ret = 0;
+			else if (ret == 0)
+				ret = -EIO;  /* timed out */
+		} else {
+			wait_for_completion(&req->r_safe_completion);
+		}
+		spin_lock(&ci->i_unsafe_lock);
+		ceph_mdsc_put_request(req);
+
+		if (ret || list_empty(head))
+			break;
+		req = list_entry(head->next,
+				 struct ceph_mds_request, r_unsafe_dir_item);
+	} while (req->r_tid < last_tid);
+out:
+	spin_unlock(&ci->i_unsafe_lock);
+	return ret;
+}
+
+/*
+ * We maintain a private dentry LRU.
+ *
+ * FIXME: this needs to be changed to a per-mds lru to be useful.
+ */
+void ceph_dentry_lru_add(struct dentry *dn)
+{
+	struct ceph_dentry_info *di = ceph_dentry(dn);
+	struct ceph_mds_client *mdsc;
+	dout("dentry_lru_add %p %p\t%.*s\n",
+			di, dn, dn->d_name.len, dn->d_name.name);
+
+	if (di) {
+		mdsc = &ceph_client(dn->d_sb)->mdsc;
+		spin_lock(&mdsc->dentry_lru_lock);
+		list_add_tail(&di->lru, &mdsc->dentry_lru);
+		mdsc->num_dentry++;
+		spin_unlock(&mdsc->dentry_lru_lock);
+	}
+}
+
+void ceph_dentry_lru_touch(struct dentry *dn)
+{
+	struct ceph_dentry_info *di = ceph_dentry(dn);
+	struct ceph_mds_client *mdsc;
+	dout("dentry_lru_touch %p %p\t%.*s\n",
+			di, dn, dn->d_name.len, dn->d_name.name);
+
+	if (di) {
+		mdsc = &ceph_client(dn->d_sb)->mdsc;
+		spin_lock(&mdsc->dentry_lru_lock);
+		list_move_tail(&di->lru, &mdsc->dentry_lru);
+		spin_unlock(&mdsc->dentry_lru_lock);
+	}
+}
+
+void ceph_dentry_lru_del(struct dentry *dn)
+{
+	struct ceph_dentry_info *di = ceph_dentry(dn);
+	struct ceph_mds_client *mdsc;
+
+	dout("dentry_lru_del %p %p\t%.*s\n",
+			di, dn, dn->d_name.len, dn->d_name.name);
+	if (di) {
+		mdsc = &ceph_client(dn->d_sb)->mdsc;
+		spin_lock(&mdsc->dentry_lru_lock);
+		list_del_init(&di->lru);
+		mdsc->num_dentry--;
+		spin_unlock(&mdsc->dentry_lru_lock);
+	}
+}
+
+const struct file_operations ceph_dir_fops = {
+	.read = ceph_read_dir,
+	.readdir = ceph_readdir,
+	.llseek = ceph_dir_llseek,
+	.open = ceph_open,
+	.release = ceph_release,
+	.unlocked_ioctl = ceph_ioctl,
+	.fsync = ceph_dir_fsync,
+};
+
+const struct inode_operations ceph_dir_iops = {
+	.lookup = ceph_lookup,
+	.permission = ceph_permission,
+	.getattr = ceph_getattr,
+	.setattr = ceph_setattr,
+	.setxattr = ceph_setxattr,
+	.getxattr = ceph_getxattr,
+	.listxattr = ceph_listxattr,
+	.removexattr = ceph_removexattr,
+	.mknod = ceph_mknod,
+	.symlink = ceph_symlink,
+	.mkdir = ceph_mkdir,
+	.link = ceph_link,
+	.unlink = ceph_unlink,
+	.rmdir = ceph_unlink,
+	.rename = ceph_rename,
+	.create = ceph_create,
+};
+
+struct dentry_operations ceph_dentry_ops = {
+	.d_revalidate = ceph_d_revalidate,
+	.d_release = ceph_dentry_release,
+};
+
+struct dentry_operations ceph_snapdir_dentry_ops = {
+	.d_revalidate = ceph_snapdir_d_revalidate,
+};
+
+struct dentry_operations ceph_snap_dentry_ops = {
+};