9p: readdir implementation for 9p2000.L

This patch implements the kernel part of readdir() implementation for 9p2000.L

    Change from V3: Instead of inode, server now sends qids for each dirent

    SYNOPSIS

    size[4] Treaddir tag[2] fid[4] offset[8] count[4]
    size[4] Rreaddir tag[2] count[4] data[count]

    DESCRIPTION

    The readdir request asks the server to read the directory specified by 'fid'
    at an offset specified by 'offset' and return as many dirent structures as
    possible that fit into count bytes. Each dirent structure is laid out as
    follows.

            qid.type[1]
              the type of the file (directory, etc.), represented as a bit
              vector corresponding to the high 8 bits of the file's mode
              word.

            qid.vers[4]
              version number for given path

            qid.path[8]
              the file server's unique identification for the file

            offset[8]
              offset into the next dirent.

            type[1]
              type of this directory entry.

            name[256]
              name of this directory entry.

    This patch adds v9fs_dir_readdir_dotl() as the readdir() call for 9p2000.L.
    This function sends P9_TREADDIR command to the server. In response the server
    sends a buffer filled with dirent structures. This is different from the
    existing v9fs_dir_readdir() call which receives stat structures from the server.
    This results in significant speedup of readdir() on large directories.
    For example, doing 'ls >/dev/null' on a directory with 10000 files on my
    laptop takes 1.088 seconds with the existing code, but only takes 0.339 seconds
    with the new readdir.

Signed-off-by: Sripathi Kodi <sripathik@in.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p/vfs_dir.c

@@ -87,29 +87,19 @@ static void p9stat_init(struct p9_wstat *stbuf)
 }
 
 /**
- * v9fs_dir_readdir - read a directory
+ * v9fs_alloc_rdir_buf - Allocate buffer used for read and readdir
  * @filp: opened file structure
- * @dirent: directory structure ???
- * @filldir: function to populate directory structure ???
+ * @buflen: Length in bytes of buffer to allocate
  *
  */
 
-static int v9fs_dir_readdir(struct file *filp, void *dirent, filldir_t filldir)
+static int v9fs_alloc_rdir_buf(struct file *filp, int buflen)
 {
-	int over;
-	struct p9_wstat st;
-	int err = 0;
-	struct p9_fid *fid;
-	int buflen;
-	int reclen = 0;
 	struct p9_rdir *rdir;
+	struct p9_fid *fid;
+	int err = 0;
 
-	P9_DPRINTK(P9_DEBUG_VFS, "name %s\n", filp->f_path.dentry->d_name.name);
 	fid = filp->private_data;
-
-	buflen = fid->clnt->msize - P9_IOHDRSZ;
-
-	/* allocate rdir on demand */
 	if (!fid->rdir) {
 		rdir = kmalloc(sizeof(struct p9_rdir) + buflen, GFP_KERNEL);
 
@@ -128,6 +118,36 @@ static int v9fs_dir_readdir(struct file *filp, void *dirent, filldir_t filldir)
 		spin_unlock(&filp->f_dentry->d_lock);
 		kfree(rdir);
 	}
+exit:
+	return err;
+}
+
+/**
+ * v9fs_dir_readdir - read a directory
+ * @filp: opened file structure
+ * @dirent: directory structure ???
+ * @filldir: function to populate directory structure ???
+ *
+ */
+
+static int v9fs_dir_readdir(struct file *filp, void *dirent, filldir_t filldir)
+{
+	int over;
+	struct p9_wstat st;
+	int err = 0;
+	struct p9_fid *fid;
+	int buflen;
+	int reclen = 0;
+	struct p9_rdir *rdir;
+
+	P9_DPRINTK(P9_DEBUG_VFS, "name %s\n", filp->f_path.dentry->d_name.name);
+	fid = filp->private_data;
+
+	buflen = fid->clnt->msize - P9_IOHDRSZ;
+
+	err = v9fs_alloc_rdir_buf(filp, buflen);
+	if (err)
+		goto exit;
 	rdir = (struct p9_rdir *) fid->rdir;
 
 	err = mutex_lock_interruptible(&rdir->mutex);
@@ -176,6 +196,88 @@ exit:
 	return err;
 }
 
+/**
+ * v9fs_dir_readdir_dotl - read a directory
+ * @filp: opened file structure
+ * @dirent: buffer to fill dirent structures
+ * @filldir: function to populate dirent structures
+ *
+ */
+static int v9fs_dir_readdir_dotl(struct file *filp, void *dirent,
+						filldir_t filldir)
+{
+	int over;
+	int err = 0;
+	struct p9_fid *fid;
+	int buflen;
+	struct p9_rdir *rdir;
+	struct p9_dirent curdirent;
+	u64 oldoffset = 0;
+
+	P9_DPRINTK(P9_DEBUG_VFS, "name %s\n", filp->f_path.dentry->d_name.name);
+	fid = filp->private_data;
+
+	buflen = fid->clnt->msize - P9_READDIRHDRSZ;
+
+	err = v9fs_alloc_rdir_buf(filp, buflen);
+	if (err)
+		goto exit;
+	rdir = (struct p9_rdir *) fid->rdir;
+
+	err = mutex_lock_interruptible(&rdir->mutex);
+	if (err)
+		return err;
+
+	while (err == 0) {
+		if (rdir->tail == rdir->head) {
+			err = p9_client_readdir(fid, rdir->buf, buflen,
+								filp->f_pos);
+			if (err <= 0)
+				goto unlock_and_exit;
+
+			rdir->head = 0;
+			rdir->tail = err;
+		}
+
+		while (rdir->head < rdir->tail) {
+
+			err = p9dirent_read(rdir->buf + rdir->head,
+						buflen - rdir->head, &curdirent,
+						fid->clnt->proto_version);
+			if (err < 0) {
+				P9_DPRINTK(P9_DEBUG_VFS, "returned %d\n", err);
+				err = -EIO;
+				goto unlock_and_exit;
+			}
+
+			/* d_off in dirent structure tracks the offset into
+			 * the next dirent in the dir. However, filldir()
+			 * expects offset into the current dirent. Hence
+			 * while calling filldir send the offset from the
+			 * previous dirent structure.
+			 */
+			over = filldir(dirent, curdirent.d_name,
+					strlen(curdirent.d_name),
+					oldoffset, v9fs_qid2ino(&curdirent.qid),
+					curdirent.d_type);
+			oldoffset = curdirent.d_off;
+
+			if (over) {
+				err = 0;
+				goto unlock_and_exit;
+			}
+
+			filp->f_pos = curdirent.d_off;
+			rdir->head += err;
+		}
+	}
+
+unlock_and_exit:
+	mutex_unlock(&rdir->mutex);
+exit:
+	return err;
+}
+
 
 /**
  * v9fs_dir_release - close a directory
@@ -207,7 +309,7 @@ const struct file_operations v9fs_dir_operations = {
 const struct file_operations v9fs_dir_operations_dotl = {
 	.read = generic_read_dir,
 	.llseek = generic_file_llseek,
-	.readdir = v9fs_dir_readdir,
+	.readdir = v9fs_dir_readdir_dotl,
 	.open = v9fs_file_open,
 	.release = v9fs_dir_release,
 };

include/net/9p/9p.h

@@ -133,6 +133,8 @@ enum p9_msg_t {
 	P9_RSTATFS,
 	P9_TRENAME = 20,
 	P9_RRENAME,
+	P9_TREADDIR = 40,
+	P9_RREADDIR,
 	P9_TVERSION = 100,
 	P9_RVERSION,
 	P9_TAUTH = 102,
@@ -275,6 +277,9 @@ enum p9_qid_t {
 /* ample room for Twrite/Rread header */
 #define P9_IOHDRSZ	24
 
+/* Room for readdir header */
+#define P9_READDIRHDRSZ	24
+
 /**
  * struct p9_str - length prefixed string type
  * @len: length of the string
@@ -485,6 +490,18 @@ struct p9_rwrite {
 	u32 count;
 };
 
+struct p9_treaddir {
+	u32 fid;
+	u64 offset;
+	u32 count;
+};
+
+struct p9_rreaddir {
+	u32 count;
+	u8 *data;
+};
+
+
 struct p9_tclunk {
 	u32 fid;
 };

include/net/9p/client.h

@@ -195,6 +195,21 @@ struct p9_fid {
 	struct list_head dlist;	/* list of all fids attached to a dentry */
 };
 
+/**
+ * struct p9_dirent - directory entry structure
+ * @qid: The p9 server qid for this dirent
+ * @d_off: offset to the next dirent
+ * @d_type: type of file
+ * @d_name: file name
+ */
+
+struct p9_dirent {
+	struct p9_qid qid;
+	u64 d_off;
+	unsigned char d_type;
+	char d_name[256];
+};
+
 int p9_client_statfs(struct p9_fid *fid, struct p9_rstatfs *sb);
 int p9_client_rename(struct p9_fid *fid, struct p9_fid *newdirfid, char *name);
 int p9_client_version(struct p9_client *);
@@ -217,6 +232,9 @@ int p9_client_read(struct p9_fid *fid, char *data, char __user *udata,
 							u64 offset, u32 count);
 int p9_client_write(struct p9_fid *fid, char *data, const char __user *udata,
 							u64 offset, u32 count);
+int p9_client_readdir(struct p9_fid *fid, char *data, u32 count, u64 offset);
+int p9dirent_read(char *buf, int len, struct p9_dirent *dirent,
+							int proto_version);
 struct p9_wstat *p9_client_stat(struct p9_fid *fid);
 int p9_client_wstat(struct p9_fid *fid, struct p9_wstat *wst);
 

net/9p/client.c

@@ -1432,3 +1432,50 @@ error:
 }
 EXPORT_SYMBOL(p9_client_rename);
 
+int p9_client_readdir(struct p9_fid *fid, char *data, u32 count, u64 offset)
+{
+	int err, rsize, total;
+	struct p9_client *clnt;
+	struct p9_req_t *req;
+	char *dataptr;
+
+	P9_DPRINTK(P9_DEBUG_9P, ">>> TREADDIR fid %d offset %llu count %d\n",
+				fid->fid, (long long unsigned) offset, count);
+
+	err = 0;
+	clnt = fid->clnt;
+	total = 0;
+
+	rsize = fid->iounit;
+	if (!rsize || rsize > clnt->msize-P9_READDIRHDRSZ)
+		rsize = clnt->msize - P9_READDIRHDRSZ;
+
+	if (count < rsize)
+		rsize = count;
+
+	req = p9_client_rpc(clnt, P9_TREADDIR, "dqd", fid->fid, offset, rsize);
+	if (IS_ERR(req)) {
+		err = PTR_ERR(req);
+		goto error;
+	}
+
+	err = p9pdu_readf(req->rc, clnt->proto_version, "D", &count, &dataptr);
+	if (err) {
+		p9pdu_dump(1, req->rc);
+		goto free_and_error;
+	}
+
+	P9_DPRINTK(P9_DEBUG_9P, "<<< RREADDIR count %d\n", count);
+
+	if (data)
+		memmove(data, dataptr, count);
+
+	p9_free_req(clnt, req);
+	return count;
+
+free_and_error:
+	p9_free_req(clnt, req);
+error:
+	return err;
+}
+EXPORT_SYMBOL(p9_client_readdir);

net/9p/protocol.c

@@ -580,3 +580,30 @@ void p9pdu_reset(struct p9_fcall *pdu)
 	pdu->offset = 0;
 	pdu->size = 0;
 }
+
+int p9dirent_read(char *buf, int len, struct p9_dirent *dirent,
+						int proto_version)
+{
+	struct p9_fcall fake_pdu;
+	int ret;
+	char *nameptr;
+
+	fake_pdu.size = len;
+	fake_pdu.capacity = len;
+	fake_pdu.sdata = buf;
+	fake_pdu.offset = 0;
+
+	ret = p9pdu_readf(&fake_pdu, proto_version, "Qqbs", &dirent->qid,
+			&dirent->d_off, &dirent->d_type, &nameptr);
+	if (ret) {
+		P9_DPRINTK(P9_DEBUG_9P, "<<< p9dirent_read failed: %d\n", ret);
+		p9pdu_dump(1, &fake_pdu);
+		goto out;
+	}
+
+	strcpy(dirent->d_name, nameptr);
+
+out:
+	return fake_pdu.offset;
+}
+EXPORT_SYMBOL(p9dirent_read);