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>
diff --git a/fs/9p/vfs_dir.c b/fs/9p/vfs_dir.c
index 36d961f..16c8a2a 100644
--- a/fs/9p/vfs_dir.c
+++ b/fs/9p/vfs_dir.c
@@ -87,6 +87,42 @@
 }
 
 /**
+ * v9fs_alloc_rdir_buf - Allocate buffer used for read and readdir
+ * @filp: opened file structure
+ * @buflen: Length in bytes of buffer to allocate
+ *
+ */
+
+static int v9fs_alloc_rdir_buf(struct file *filp, int buflen)
+{
+	struct p9_rdir *rdir;
+	struct p9_fid *fid;
+	int err = 0;
+
+	fid = filp->private_data;
+	if (!fid->rdir) {
+		rdir = kmalloc(sizeof(struct p9_rdir) + buflen, GFP_KERNEL);
+
+		if (rdir == NULL) {
+			err = -ENOMEM;
+			goto exit;
+		}
+		spin_lock(&filp->f_dentry->d_lock);
+		if (!fid->rdir) {
+			rdir->buf = (uint8_t *)rdir + sizeof(struct p9_rdir);
+			mutex_init(&rdir->mutex);
+			rdir->head = rdir->tail = 0;
+			fid->rdir = (void *) rdir;
+			rdir = NULL;
+		}
+		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 ???
@@ -109,25 +145,9 @@
 
 	buflen = fid->clnt->msize - P9_IOHDRSZ;
 
-	/* allocate rdir on demand */
-	if (!fid->rdir) {
-		rdir = kmalloc(sizeof(struct p9_rdir) + buflen, GFP_KERNEL);
-
-		if (rdir == NULL) {
-			err = -ENOMEM;
-			goto exit;
-		}
-		spin_lock(&filp->f_dentry->d_lock);
-		if (!fid->rdir) {
-			rdir->buf = (uint8_t *)rdir + sizeof(struct p9_rdir);
-			mutex_init(&rdir->mutex);
-			rdir->head = rdir->tail = 0;
-			fid->rdir = (void *) rdir;
-			rdir = NULL;
-		}
-		spin_unlock(&filp->f_dentry->d_lock);
-		kfree(rdir);
-	}
+	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 @@
 	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_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,
 };