1 files changed, 464 insertions, 967 deletions

diff --git a/net/9p/trans_fd.c b/net/9p/trans_fd.c
index 6dabbdb66651..be65d8242fd2 100644
--- a/net/9p/trans_fd.c
+++ b/net/9p/trans_fd.c
@@ -39,12 +39,11 @@
 #include <linux/file.h>
 #include <linux/parser.h>
 #include <net/9p/9p.h>
+#include <net/9p/client.h>
 #include <net/9p/transport.h>
 
 #define P9_PORT 564
 #define MAX_SOCK_BUF (64*1024)
-#define ERREQFLUSH	1
-#define SCHED_TIMEOUT	10
 #define MAXPOLLWADDR	2
 
 /**
@@ -61,7 +60,6 @@ struct p9_fd_opts {
 	u16 port;
 };
 
-
 /**
  * struct p9_trans_fd - transport state
  * @rd: reference to file to read from
@@ -100,60 +98,22 @@ enum {
 	Wpending = 8,		/* can write */
 };
 
-enum {
-	None,
-	Flushing,
-	Flushed,
-};
-
-struct p9_req;
-typedef void (*p9_conn_req_callback)(struct p9_req *req, void *a);
-
-/**
- * struct p9_req - fd mux encoding of an rpc transaction
- * @lock: protects req_list
- * @tag: numeric tag for rpc transaction
- * @tcall: request &p9_fcall structure
- * @rcall: response &p9_fcall structure
- * @err: error state
- * @cb: callback for when response is received
- * @cba: argument to pass to callback
- * @flush: flag to indicate RPC has been flushed
- * @req_list: list link for higher level objects to chain requests
- *
- */
-
-struct p9_req {
-	spinlock_t lock;
-	int tag;
-	struct p9_fcall *tcall;
-	struct p9_fcall *rcall;
-	int err;
-	p9_conn_req_callback cb;
-	void *cba;
-	int flush;
-	struct list_head req_list;
-};
-
-struct p9_mux_poll_task {
-	struct task_struct *task;
-	struct list_head mux_list;
-	int muxnum;
+struct p9_poll_wait {
+	struct p9_conn *conn;
+	wait_queue_t wait;
+	wait_queue_head_t *wait_addr;
 };
 
 /**
  * struct p9_conn - fd mux connection state information
- * @lock: protects mux_list (?)
  * @mux_list: list link for mux to manage multiple connections (?)
- * @poll_task: task polling on this connection
- * @msize: maximum size for connection (dup)
- * @extended: 9p2000.u flag (dup)
- * @trans: reference to transport instance for this connection
- * @tagpool: id accounting for transactions
+ * @client: reference to client instance for this connection
  * @err: error state
  * @req_list: accounting for requests which have been sent
  * @unsent_req_list: accounting for requests that haven't been sent
- * @rcall: current response &p9_fcall structure
+ * @req: current request being processed (if any)
+ * @tmp_buf: temporary buffer to read in header
+ * @rsize: amount to read for current frame
  * @rpos: read position in current frame
  * @rbuf: current read buffer
  * @wpos: write position for current frame
@@ -169,409 +129,300 @@ struct p9_mux_poll_task {
  */
 
 struct p9_conn {
-	spinlock_t lock; /* protect lock structure */
 	struct list_head mux_list;
-	struct p9_mux_poll_task *poll_task;
-	int msize;
-	unsigned char extended;
-	struct p9_trans *trans;
-	struct p9_idpool *tagpool;
+	struct p9_client *client;
 	int err;
 	struct list_head req_list;
 	struct list_head unsent_req_list;
-	struct p9_fcall *rcall;
+	struct p9_req_t *req;
+	char tmp_buf[7];
+	int rsize;
 	int rpos;
 	char *rbuf;
 	int wpos;
 	int wsize;
 	char *wbuf;
-	wait_queue_t poll_wait[MAXPOLLWADDR];
-	wait_queue_head_t *poll_waddr[MAXPOLLWADDR];
+	struct list_head poll_pending_link;
+	struct p9_poll_wait poll_wait[MAXPOLLWADDR];
 	poll_table pt;
 	struct work_struct rq;
 	struct work_struct wq;
 	unsigned long wsched;
 };
 
-/**
- * struct p9_mux_rpc - fd mux rpc accounting structure
- * @m: connection this request was issued on
- * @err: error state
- * @tcall: request &p9_fcall
- * @rcall: response &p9_fcall
- * @wqueue: wait queue that client is blocked on for this rpc
- *
- * Bug: isn't this information duplicated elsewhere like &p9_req
- */
-
-struct p9_mux_rpc {
-	struct p9_conn *m;
-	int err;
-	struct p9_fcall *tcall;
-	struct p9_fcall *rcall;
-	wait_queue_head_t wqueue;
-};
-
-static int p9_poll_proc(void *);
-static void p9_read_work(struct work_struct *work);
-static void p9_write_work(struct work_struct *work);
-static void p9_pollwait(struct file *filp, wait_queue_head_t *wait_address,
-								poll_table *p);
-static int p9_fd_write(struct p9_trans *trans, void *v, int len);
-static int p9_fd_read(struct p9_trans *trans, void *v, int len);
-
-static DEFINE_MUTEX(p9_mux_task_lock);
+static DEFINE_SPINLOCK(p9_poll_lock);
+static LIST_HEAD(p9_poll_pending_list);
 static struct workqueue_struct *p9_mux_wq;
+static struct task_struct *p9_poll_task;
 
-static int p9_mux_num;
-static int p9_mux_poll_task_num;
-static struct p9_mux_poll_task p9_mux_poll_tasks[100];
-
-static void p9_conn_destroy(struct p9_conn *);
-static unsigned int p9_fd_poll(struct p9_trans *trans,
-						struct poll_table_struct *pt);
-
-#ifdef P9_NONBLOCK
-static int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
-	p9_conn_req_callback cb, void *a);
-#endif /* P9_NONBLOCK */
-
-static void p9_conn_cancel(struct p9_conn *m, int err);
-
-static u16 p9_mux_get_tag(struct p9_conn *m)
+static void p9_mux_poll_stop(struct p9_conn *m)
 {
-	int tag;
+	unsigned long flags;
+	int i;
 
-	tag = p9_idpool_get(m->tagpool);
-	if (tag < 0)
-		return P9_NOTAG;
-	else
-		return (u16) tag;
-}
+	for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
+		struct p9_poll_wait *pwait = &m->poll_wait[i];
 
-static void p9_mux_put_tag(struct p9_conn *m, u16 tag)
-{
-	if (tag != P9_NOTAG && p9_idpool_check(tag, m->tagpool))
-		p9_idpool_put(tag, m->tagpool);
+		if (pwait->wait_addr) {
+			remove_wait_queue(pwait->wait_addr, &pwait->wait);
+			pwait->wait_addr = NULL;
+		}
+	}
+
+	spin_lock_irqsave(&p9_poll_lock, flags);
+	list_del_init(&m->poll_pending_link);
+	spin_unlock_irqrestore(&p9_poll_lock, flags);
 }
 
 /**
- * p9_mux_calc_poll_procs - calculates the number of polling procs
- * @muxnum: number of mounts
+ * p9_conn_cancel - cancel all pending requests with error
+ * @m: mux data
+ * @err: error code
  *
- * Calculation is based on the number of mounted v9fs filesystems.
- * The current implementation returns sqrt of the number of mounts.
  */
 
-static int p9_mux_calc_poll_procs(int muxnum)
+static void p9_conn_cancel(struct p9_conn *m, int err)
 {
-	int n;
-
-	if (p9_mux_poll_task_num)
-		n = muxnum / p9_mux_poll_task_num +
-		    (muxnum % p9_mux_poll_task_num ? 1 : 0);
-	else
-		n = 1;
-
-	if (n > ARRAY_SIZE(p9_mux_poll_tasks))
-		n = ARRAY_SIZE(p9_mux_poll_tasks);
-
-	return n;
-}
+	struct p9_req_t *req, *rtmp;
+	unsigned long flags;
+	LIST_HEAD(cancel_list);
 
-static int p9_mux_poll_start(struct p9_conn *m)
-{
-	int i, n;
-	struct p9_mux_poll_task *vpt, *vptlast;
-	struct task_struct *pproc;
-
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p muxnum %d procnum %d\n", m, p9_mux_num,
-		p9_mux_poll_task_num);
-	mutex_lock(&p9_mux_task_lock);
-
-	n = p9_mux_calc_poll_procs(p9_mux_num + 1);
-	if (n > p9_mux_poll_task_num) {
-		for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
-			if (p9_mux_poll_tasks[i].task == NULL) {
-				vpt = &p9_mux_poll_tasks[i];
-				P9_DPRINTK(P9_DEBUG_MUX, "create proc %p\n",
-									vpt);
-				pproc = kthread_create(p9_poll_proc, vpt,
-								"v9fs-poll");
-
-				if (!IS_ERR(pproc)) {
-					vpt->task = pproc;
-					INIT_LIST_HEAD(&vpt->mux_list);
-					vpt->muxnum = 0;
-					p9_mux_poll_task_num++;
-					wake_up_process(vpt->task);
-				}
-				break;
-			}
-		}
+	P9_DPRINTK(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
 
-		if (i >= ARRAY_SIZE(p9_mux_poll_tasks))
-			P9_DPRINTK(P9_DEBUG_ERROR,
-					"warning: no free poll slots\n");
-	}
+	spin_lock_irqsave(&m->client->lock, flags);
 
-	n = (p9_mux_num + 1) / p9_mux_poll_task_num +
-	    ((p9_mux_num + 1) % p9_mux_poll_task_num ? 1 : 0);
-
-	vptlast = NULL;
-	for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++) {
-		vpt = &p9_mux_poll_tasks[i];
-		if (vpt->task != NULL) {
-			vptlast = vpt;
-			if (vpt->muxnum < n) {
-				P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
-				list_add(&m->mux_list, &vpt->mux_list);
-				vpt->muxnum++;
-				m->poll_task = vpt;
-				memset(&m->poll_waddr, 0,
-							sizeof(m->poll_waddr));
-				init_poll_funcptr(&m->pt, p9_pollwait);
-				break;
-			}
-		}
+	if (m->err) {
+		spin_unlock_irqrestore(&m->client->lock, flags);
+		return;
 	}
 
-	if (i >= ARRAY_SIZE(p9_mux_poll_tasks)) {
-		if (vptlast == NULL) {
-			mutex_unlock(&p9_mux_task_lock);
-			return -ENOMEM;
-		}
+	m->err = err;
 
-		P9_DPRINTK(P9_DEBUG_MUX, "put in proc %d\n", i);
-		list_add(&m->mux_list, &vptlast->mux_list);
-		vptlast->muxnum++;
-		m->poll_task = vptlast;
-		memset(&m->poll_waddr, 0, sizeof(m->poll_waddr));
-		init_poll_funcptr(&m->pt, p9_pollwait);
+	list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
+		req->status = REQ_STATUS_ERROR;
+		if (!req->t_err)
+			req->t_err = err;
+		list_move(&req->req_list, &cancel_list);
 	}
-
-	p9_mux_num++;
-	mutex_unlock(&p9_mux_task_lock);
-
-	return 0;
-}
-
-static void p9_mux_poll_stop(struct p9_conn *m)
-{
-	int i;
-	struct p9_mux_poll_task *vpt;
-
-	mutex_lock(&p9_mux_task_lock);
-	vpt = m->poll_task;
-	list_del(&m->mux_list);
-	for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
-		if (m->poll_waddr[i] != NULL) {
-			remove_wait_queue(m->poll_waddr[i], &m->poll_wait[i]);
-			m->poll_waddr[i] = NULL;
-		}
+	list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
+		req->status = REQ_STATUS_ERROR;
+		if (!req->t_err)
+			req->t_err = err;
+		list_move(&req->req_list, &cancel_list);
 	}
-	vpt->muxnum--;
-	if (!vpt->muxnum) {
-		P9_DPRINTK(P9_DEBUG_MUX, "destroy proc %p\n", vpt);
-		kthread_stop(vpt->task);
-		vpt->task = NULL;
-		p9_mux_poll_task_num--;
+	spin_unlock_irqrestore(&m->client->lock, flags);
+
+	list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
+		list_del(&req->req_list);
+		P9_DPRINTK(P9_DEBUG_ERROR, "call back req %p\n", req);
+		p9_client_cb(m->client, req);
 	}
-	p9_mux_num--;
-	mutex_unlock(&p9_mux_task_lock);
 }
 
-/**
- * p9_conn_create - allocate and initialize the per-session mux data
- * @trans: transport structure
- *
- * Note: Creates the polling task if this is the first session.
- */
-
-static struct p9_conn *p9_conn_create(struct p9_trans *trans)
+static unsigned int
+p9_fd_poll(struct p9_client *client, struct poll_table_struct *pt)
 {
-	int i, n;
-	struct p9_conn *m;
+	int ret, n;
+	struct p9_trans_fd *ts = NULL;
 
-	P9_DPRINTK(P9_DEBUG_MUX, "transport %p msize %d\n", trans,
-								trans->msize);
-	m = kzalloc(sizeof(struct p9_conn), GFP_KERNEL);
-	if (!m)
-		return ERR_PTR(-ENOMEM);
+	if (client && client->status == Connected)
+		ts = client->trans;
 
-	spin_lock_init(&m->lock);
-	INIT_LIST_HEAD(&m->mux_list);
-	m->msize = trans->msize;
-	m->extended = trans->extended;
-	m->trans = trans;
-	m->tagpool = p9_idpool_create();
-	if (IS_ERR(m->tagpool)) {
-		kfree(m);
-		return ERR_PTR(-ENOMEM);
-	}
+	if (!ts)
+		return -EREMOTEIO;
 
-	INIT_LIST_HEAD(&m->req_list);
-	INIT_LIST_HEAD(&m->unsent_req_list);
-	INIT_WORK(&m->rq, p9_read_work);
-	INIT_WORK(&m->wq, p9_write_work);
-	n = p9_mux_poll_start(m);
-	if (n) {
-		kfree(m);
-		return ERR_PTR(n);
-	}
+	if (!ts->rd->f_op || !ts->rd->f_op->poll)
+		return -EIO;
 
-	n = p9_fd_poll(trans, &m->pt);
-	if (n & POLLIN) {
-		P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
-		set_bit(Rpending, &m->wsched);
-	}
+	if (!ts->wr->f_op || !ts->wr->f_op->poll)
+		return -EIO;
 
-	if (n & POLLOUT) {
-		P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
-		set_bit(Wpending, &m->wsched);
-	}
+	ret = ts->rd->f_op->poll(ts->rd, pt);
+	if (ret < 0)
+		return ret;
 
-	for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++) {
-		if (IS_ERR(m->poll_waddr[i])) {
-			p9_mux_poll_stop(m);
-			kfree(m);
-			return (void *)m->poll_waddr;	/* the error code */
-		}
+	if (ts->rd != ts->wr) {
+		n = ts->wr->f_op->poll(ts->wr, pt);
+		if (n < 0)
+			return n;
+		ret = (ret & ~POLLOUT) | (n & ~POLLIN);
 	}
 
-	return m;
+	return ret;
 }
 
 /**
- * p9_mux_destroy - cancels all pending requests and frees mux resources
- * @m: mux to destroy
+ * p9_fd_read- read from a fd
+ * @client: client instance
+ * @v: buffer to receive data into
+ * @len: size of receive buffer
  *
  */
 
-static void p9_conn_destroy(struct p9_conn *m)
+static int p9_fd_read(struct p9_client *client, void *v, int len)
 {
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p prev %p next %p\n", m,
-		m->mux_list.prev, m->mux_list.next);
+	int ret;
+	struct p9_trans_fd *ts = NULL;
 
-	p9_mux_poll_stop(m);
-	cancel_work_sync(&m->rq);
-	cancel_work_sync(&m->wq);
+	if (client && client->status != Disconnected)
+		ts = client->trans;
 
-	p9_conn_cancel(m, -ECONNRESET);
+	if (!ts)
+		return -EREMOTEIO;
 
-	m->trans = NULL;
-	p9_idpool_destroy(m->tagpool);
-	kfree(m);
+	if (!(ts->rd->f_flags & O_NONBLOCK))
+		P9_DPRINTK(P9_DEBUG_ERROR, "blocking read ...\n");
+
+	ret = kernel_read(ts->rd, ts->rd->f_pos, v, len);
+	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
+		client->status = Disconnected;
+	return ret;
 }
 
 /**
- * p9_pollwait - add poll task to the wait queue
- * @filp: file pointer being polled
- * @wait_address: wait_q to block on
- * @p: poll state
+ * p9_read_work - called when there is some data to be read from a transport
+ * @work: container of work to be done
  *
- * called by files poll operation to add v9fs-poll task to files wait queue
  */
 
-static void
-p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
+static void p9_read_work(struct work_struct *work)
 {
-	int i;
+	int n, err;
 	struct p9_conn *m;
 
-	m = container_of(p, struct p9_conn, pt);
-	for (i = 0; i < ARRAY_SIZE(m->poll_waddr); i++)
-		if (m->poll_waddr[i] == NULL)
-			break;
+	m = container_of(work, struct p9_conn, rq);
 
-	if (i >= ARRAY_SIZE(m->poll_waddr)) {
-		P9_DPRINTK(P9_DEBUG_ERROR, "not enough wait_address slots\n");
+	if (m->err < 0)
 		return;
-	}
 
-	m->poll_waddr[i] = wait_address;
+	P9_DPRINTK(P9_DEBUG_TRANS, "start mux %p pos %d\n", m, m->rpos);
 
-	if (!wait_address) {
-		P9_DPRINTK(P9_DEBUG_ERROR, "no wait_address\n");
-		m->poll_waddr[i] = ERR_PTR(-EIO);
+	if (!m->rbuf) {
+		m->rbuf = m->tmp_buf;
+		m->rpos = 0;
+		m->rsize = 7; /* start by reading header */
+	}
+
+	clear_bit(Rpending, &m->wsched);
+	P9_DPRINTK(P9_DEBUG_TRANS, "read mux %p pos %d size: %d = %d\n", m,
+					m->rpos, m->rsize, m->rsize-m->rpos);
+	err = p9_fd_read(m->client, m->rbuf + m->rpos,
+						m->rsize - m->rpos);
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p got %d bytes\n", m, err);
+	if (err == -EAGAIN) {
+		clear_bit(Rworksched, &m->wsched);
 		return;
 	}
 
-	init_waitqueue_entry(&m->poll_wait[i], m->poll_task->task);
-	add_wait_queue(wait_address, &m->poll_wait[i]);
-}
+	if (err <= 0)
+		goto error;
 
-/**
- * p9_poll_mux - polls a mux and schedules read or write works if necessary
- * @m: connection to poll
- *
- */
+	m->rpos += err;
 
-static void p9_poll_mux(struct p9_conn *m)
-{
-	int n;
+	if ((!m->req) && (m->rpos == m->rsize)) { /* header read in */
+		u16 tag;
+		P9_DPRINTK(P9_DEBUG_TRANS, "got new header\n");
 
-	if (m->err < 0)
-		return;
+		n = le32_to_cpu(*(__le32 *) m->rbuf); /* read packet size */
+		if (n >= m->client->msize) {
+			P9_DPRINTK(P9_DEBUG_ERROR,
+				"requested packet size too big: %d\n", n);
+			err = -EIO;
+			goto error;
+		}
 
-	n = p9_fd_poll(m->trans, NULL);
-	if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
-		P9_DPRINTK(P9_DEBUG_MUX, "error mux %p err %d\n", m, n);
-		if (n >= 0)
-			n = -ECONNRESET;
-		p9_conn_cancel(m, n);
-	}
+		tag = le16_to_cpu(*(__le16 *) (m->rbuf+5)); /* read tag */
+		P9_DPRINTK(P9_DEBUG_TRANS,
+			"mux %p pkt: size: %d bytes tag: %d\n", m, n, tag);
 
-	if (n & POLLIN) {
-		set_bit(Rpending, &m->wsched);
-		P9_DPRINTK(P9_DEBUG_MUX, "mux %p can read\n", m);
-		if (!test_and_set_bit(Rworksched, &m->wsched)) {
-			P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
-			queue_work(p9_mux_wq, &m->rq);
+		m->req = p9_tag_lookup(m->client, tag);
+		if (!m->req) {
+			P9_DPRINTK(P9_DEBUG_ERROR, "Unexpected packet tag %d\n",
+								 tag);
+			err = -EIO;
+			goto error;
 		}
-	}
 
-	if (n & POLLOUT) {
-		set_bit(Wpending, &m->wsched);
-		P9_DPRINTK(P9_DEBUG_MUX, "mux %p can write\n", m);
-		if ((m->wsize || !list_empty(&m->unsent_req_list))
-		    && !test_and_set_bit(Wworksched, &m->wsched)) {
-			P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
-			queue_work(p9_mux_wq, &m->wq);
+		if (m->req->rc == NULL) {
+			m->req->rc = kmalloc(sizeof(struct p9_fcall) +
+						m->client->msize, GFP_KERNEL);
+			if (!m->req->rc) {
+				m->req = NULL;
+				err = -ENOMEM;
+				goto error;
+			}
 		}
+		m->rbuf = (char *)m->req->rc + sizeof(struct p9_fcall);
+		memcpy(m->rbuf, m->tmp_buf, m->rsize);
+		m->rsize = n;
 	}
+
+	/* not an else because some packets (like clunk) have no payload */
+	if ((m->req) && (m->rpos == m->rsize)) { /* packet is read in */
+		P9_DPRINTK(P9_DEBUG_TRANS, "got new packet\n");
+		spin_lock(&m->client->lock);
+		list_del(&m->req->req_list);
+		spin_unlock(&m->client->lock);
+		p9_client_cb(m->client, m->req);
+
+		m->rbuf = NULL;
+		m->rpos = 0;
+		m->rsize = 0;
+		m->req = NULL;
+	}
+
+	if (!list_empty(&m->req_list)) {
+		if (test_and_clear_bit(Rpending, &m->wsched))
+			n = POLLIN;
+		else
+			n = p9_fd_poll(m->client, NULL);
+
+		if (n & POLLIN) {
+			P9_DPRINTK(P9_DEBUG_TRANS, "sched read work %p\n", m);
+			queue_work(p9_mux_wq, &m->rq);
+		} else
+			clear_bit(Rworksched, &m->wsched);
+	} else
+		clear_bit(Rworksched, &m->wsched);
+
+	return;
+error:
+	p9_conn_cancel(m, err);
+	clear_bit(Rworksched, &m->wsched);
 }
 
 /**
- * p9_poll_proc - poll worker thread
- * @a: thread state and arguments
- *
- * polls all v9fs transports for new events and queues the appropriate
- * work to the work queue
+ * p9_fd_write - write to a socket
+ * @client: client instance
+ * @v: buffer to send data from
+ * @len: size of send buffer
  *
  */
 
-static int p9_poll_proc(void *a)
+static int p9_fd_write(struct p9_client *client, void *v, int len)
 {
-	struct p9_conn *m, *mtmp;
-	struct p9_mux_poll_task *vpt;
+	int ret;
+	mm_segment_t oldfs;
+	struct p9_trans_fd *ts = NULL;
 
-	vpt = a;
-	P9_DPRINTK(P9_DEBUG_MUX, "start %p %p\n", current, vpt);
-	while (!kthread_should_stop()) {
-		set_current_state(TASK_INTERRUPTIBLE);
+	if (client && client->status != Disconnected)
+		ts = client->trans;
 
-		list_for_each_entry_safe(m, mtmp, &vpt->mux_list, mux_list) {
-			p9_poll_mux(m);
-		}
+	if (!ts)
+		return -EREMOTEIO;
 
-		P9_DPRINTK(P9_DEBUG_MUX, "sleeping...\n");
-		schedule_timeout(SCHED_TIMEOUT * HZ);
-	}
+	if (!(ts->wr->f_flags & O_NONBLOCK))
+		P9_DPRINTK(P9_DEBUG_ERROR, "blocking write ...\n");
 
-	__set_current_state(TASK_RUNNING);
-	P9_DPRINTK(P9_DEBUG_MUX, "finish\n");
-	return 0;
+	oldfs = get_fs();
+	set_fs(get_ds());
+	/* The cast to a user pointer is valid due to the set_fs() */
+	ret = vfs_write(ts->wr, (void __user *)v, len, &ts->wr->f_pos);
+	set_fs(oldfs);
+
+	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
+		client->status = Disconnected;
+	return ret;
 }
 
 /**
@@ -584,7 +435,7 @@ static void p9_write_work(struct work_struct *work)
 {
 	int n, err;
 	struct p9_conn *m;
-	struct p9_req *req;
+	struct p9_req_t *req;
 
 	m = container_of(work, struct p9_conn, wq);
 
@@ -599,25 +450,23 @@ static void p9_write_work(struct work_struct *work)
 			return;
 		}
 
-		spin_lock(&m->lock);
-again:
-		req = list_entry(m->unsent_req_list.next, struct p9_req,
+		spin_lock(&m->client->lock);
+		req = list_entry(m->unsent_req_list.next, struct p9_req_t,
 			       req_list);
+		req->status = REQ_STATUS_SENT;
 		list_move_tail(&req->req_list, &m->req_list);
-		if (req->err == ERREQFLUSH)
-			goto again;
 
-		m->wbuf = req->tcall->sdata;
-		m->wsize = req->tcall->size;
+		m->wbuf = req->tc->sdata;
+		m->wsize = req->tc->size;
 		m->wpos = 0;
-		spin_unlock(&m->lock);
+		spin_unlock(&m->client->lock);
 	}
 
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p pos %d size %d\n", m, m->wpos,
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p pos %d size %d\n", m, m->wpos,
 								m->wsize);
 	clear_bit(Wpending, &m->wsched);
-	err = p9_fd_write(m->trans, m->wbuf + m->wpos, m->wsize - m->wpos);
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p sent %d bytes\n", m, err);
+	err = p9_fd_write(m->client, m->wbuf + m->wpos, m->wsize - m->wpos);
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p sent %d bytes\n", m, err);
 	if (err == -EAGAIN) {
 		clear_bit(Wworksched, &m->wsched);
 		return;
@@ -638,10 +487,10 @@ again:
 		if (test_and_clear_bit(Wpending, &m->wsched))
 			n = POLLOUT;
 		else
-			n = p9_fd_poll(m->trans, NULL);
+			n = p9_fd_poll(m->client, NULL);
 
 		if (n & POLLOUT) {
-			P9_DPRINTK(P9_DEBUG_MUX, "schedule write work %p\n", m);
+			P9_DPRINTK(P9_DEBUG_TRANS, "sched write work %p\n", m);
 			queue_work(p9_mux_wq, &m->wq);
 		} else
 			clear_bit(Wworksched, &m->wsched);
@@ -655,504 +504,197 @@ error:
 	clear_bit(Wworksched, &m->wsched);
 }
 
-static void process_request(struct p9_conn *m, struct p9_req *req)
+static int p9_pollwake(wait_queue_t *wait, unsigned mode, int sync, void *key)
 {
-	int ecode;
-	struct p9_str *ename;
-
-	if (!req->err && req->rcall->id == P9_RERROR) {
-		ecode = req->rcall->params.rerror.errno;
-		ename = &req->rcall->params.rerror.error;
-
-		P9_DPRINTK(P9_DEBUG_MUX, "Rerror %.*s\n", ename->len,
-								ename->str);
-
-		if (m->extended)
-			req->err = -ecode;
+	struct p9_poll_wait *pwait =
+		container_of(wait, struct p9_poll_wait, wait);
+	struct p9_conn *m = pwait->conn;
+	unsigned long flags;
+	DECLARE_WAITQUEUE(dummy_wait, p9_poll_task);
 
-		if (!req->err) {
-			req->err = p9_errstr2errno(ename->str, ename->len);
+	spin_lock_irqsave(&p9_poll_lock, flags);
+	if (list_empty(&m->poll_pending_link))
+		list_add_tail(&m->poll_pending_link, &p9_poll_pending_list);
+	spin_unlock_irqrestore(&p9_poll_lock, flags);
 
-			/* string match failed */
-			if (!req->err) {
-				PRINT_FCALL_ERROR("unknown error", req->rcall);
-				req->err = -ESERVERFAULT;
-			}
-		}
-	} else if (req->tcall && req->rcall->id != req->tcall->id + 1) {
-		P9_DPRINTK(P9_DEBUG_ERROR,
-				"fcall mismatch: expected %d, got %d\n",
-				req->tcall->id + 1, req->rcall->id);
-		if (!req->err)
-			req->err = -EIO;
-	}
+	/* perform the default wake up operation */
+	return default_wake_function(&dummy_wait, mode, sync, key);
 }
 
 /**
- * p9_read_work - called when there is some data to be read from a transport
- * @work: container of work to be done
+ * p9_pollwait - add poll task to the wait queue
+ * @filp: file pointer being polled
+ * @wait_address: wait_q to block on
+ * @p: poll state
  *
+ * called by files poll operation to add v9fs-poll task to files wait queue
  */
 
-static void p9_read_work(struct work_struct *work)
+static void
+p9_pollwait(struct file *filp, wait_queue_head_t *wait_address, poll_table *p)
 {
-	int n, err;
-	struct p9_conn *m;
-	struct p9_req *req, *rptr, *rreq;
-	struct p9_fcall *rcall;
-	char *rbuf;
-
-	m = container_of(work, struct p9_conn, rq);
-
-	if (m->err < 0)
-		return;
-
-	rcall = NULL;
-	P9_DPRINTK(P9_DEBUG_MUX, "start mux %p pos %d\n", m, m->rpos);
+	struct p9_conn *m = container_of(p, struct p9_conn, pt);
+	struct p9_poll_wait *pwait = NULL;
+	int i;
 
-	if (!m->rcall) {
-		m->rcall =
-		    kmalloc(sizeof(struct p9_fcall) + m->msize, GFP_KERNEL);
-		if (!m->rcall) {
-			err = -ENOMEM;
-			goto error;
+	for (i = 0; i < ARRAY_SIZE(m->poll_wait); i++) {
+		if (m->poll_wait[i].wait_addr == NULL) {
+			pwait = &m->poll_wait[i];
+			break;
 		}
-
-		m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
-		m->rpos = 0;
 	}
 
-	clear_bit(Rpending, &m->wsched);
-	err = p9_fd_read(m->trans, m->rbuf + m->rpos, m->msize - m->rpos);
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p got %d bytes\n", m, err);
-	if (err == -EAGAIN) {
-		clear_bit(Rworksched, &m->wsched);
+	if (!pwait) {
+		P9_DPRINTK(P9_DEBUG_ERROR, "not enough wait_address slots\n");
 		return;
 	}
 
-	if (err <= 0)
-		goto error;
-
-	m->rpos += err;
-	while (m->rpos > 4) {
-		n = le32_to_cpu(*(__le32 *) m->rbuf);
-		if (n >= m->msize) {
-			P9_DPRINTK(P9_DEBUG_ERROR,
-				"requested packet size too big: %d\n", n);
-			err = -EIO;
-			goto error;
-		}
-
-		if (m->rpos < n)
-			break;
-
-		err =
-		    p9_deserialize_fcall(m->rbuf, n, m->rcall, m->extended);
-		if (err < 0)
-			goto error;
-
-#ifdef CONFIG_NET_9P_DEBUG
-		if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
-			char buf[150];
-
-			p9_printfcall(buf, sizeof(buf), m->rcall,
-				m->extended);
-			printk(KERN_NOTICE ">>> %p %s\n", m, buf);
-		}
-#endif
-
-		rcall = m->rcall;
-		rbuf = m->rbuf;
-		if (m->rpos > n) {
-			m->rcall = kmalloc(sizeof(struct p9_fcall) + m->msize,
-					   GFP_KERNEL);
-			if (!m->rcall) {
-				err = -ENOMEM;
-				goto error;
-			}
-
-			m->rbuf = (char *)m->rcall + sizeof(struct p9_fcall);
-			memmove(m->rbuf, rbuf + n, m->rpos - n);
-			m->rpos -= n;
-		} else {
-			m->rcall = NULL;
-			m->rbuf = NULL;
-			m->rpos = 0;
-		}
-
-		P9_DPRINTK(P9_DEBUG_MUX, "mux %p fcall id %d tag %d\n", m,
-							rcall->id, rcall->tag);
-
-		req = NULL;
-		spin_lock(&m->lock);
-		list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
-			if (rreq->tag == rcall->tag) {
-				req = rreq;
-				if (req->flush != Flushing)
-					list_del(&req->req_list);
-				break;
-			}
-		}
-		spin_unlock(&m->lock);
-
-		if (req) {
-			req->rcall = rcall;
-			process_request(m, req);
-
-			if (req->flush != Flushing) {
-				if (req->cb)
-					(*req->cb) (req, req->cba);
-				else
-					kfree(req->rcall);
-			}
-		} else {
-			if (err >= 0 && rcall->id != P9_RFLUSH)
-				P9_DPRINTK(P9_DEBUG_ERROR,
-				  "unexpected response mux %p id %d tag %d\n",
-				  m, rcall->id, rcall->tag);
-			kfree(rcall);
-		}
-	}
-
-	if (!list_empty(&m->req_list)) {
-		if (test_and_clear_bit(Rpending, &m->wsched))
-			n = POLLIN;
-		else
-			n = p9_fd_poll(m->trans, NULL);
-
-		if (n & POLLIN) {
-			P9_DPRINTK(P9_DEBUG_MUX, "schedule read work %p\n", m);
-			queue_work(p9_mux_wq, &m->rq);
-		} else
-			clear_bit(Rworksched, &m->wsched);
-	} else
-		clear_bit(Rworksched, &m->wsched);
-
-	return;
-
-error:
-	p9_conn_cancel(m, err);
-	clear_bit(Rworksched, &m->wsched);
+	pwait->conn = m;
+	pwait->wait_addr = wait_address;
+	init_waitqueue_func_entry(&pwait->wait, p9_pollwake);
+	add_wait_queue(wait_address, &pwait->wait);
 }
 
 /**
- * p9_send_request - send 9P request
- * The function can sleep until the request is scheduled for sending.
- * The function can be interrupted. Return from the function is not
- * a guarantee that the request is sent successfully. Can return errors
- * that can be retrieved by PTR_ERR macros.
- *
- * @m: mux data
- * @tc: request to be sent
- * @cb: callback function to call when response is received
- * @cba: parameter to pass to the callback function
+ * p9_conn_create - allocate and initialize the per-session mux data
+ * @client: client instance
  *
+ * Note: Creates the polling task if this is the first session.
  */
 
-static struct p9_req *p9_send_request(struct p9_conn *m,
-					  struct p9_fcall *tc,
-					  p9_conn_req_callback cb, void *cba)
+static struct p9_conn *p9_conn_create(struct p9_client *client)
 {
 	int n;
-	struct p9_req *req;
-
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p task %p tcall %p id %d\n", m, current,
-		tc, tc->id);
-	if (m->err < 0)
-		return ERR_PTR(m->err);
-
-	req = kmalloc(sizeof(struct p9_req), GFP_KERNEL);
-	if (!req)
-		return ERR_PTR(-ENOMEM);
-
-	if (tc->id == P9_TVERSION)
-		n = P9_NOTAG;
-	else
-		n = p9_mux_get_tag(m);
+	struct p9_conn *m;
 
-	if (n < 0) {
-		kfree(req);
+	P9_DPRINTK(P9_DEBUG_TRANS, "client %p msize %d\n", client,
+								client->msize);
+	m = kzalloc(sizeof(struct p9_conn), GFP_KERNEL);
+	if (!m)
 		return ERR_PTR(-ENOMEM);
-	}
 
-	p9_set_tag(tc, n);
+	INIT_LIST_HEAD(&m->mux_list);
+	m->client = client;
 
-#ifdef CONFIG_NET_9P_DEBUG
-	if ((p9_debug_level&P9_DEBUG_FCALL) == P9_DEBUG_FCALL) {
-		char buf[150];
+	INIT_LIST_HEAD(&m->req_list);
+	INIT_LIST_HEAD(&m->unsent_req_list);
+	INIT_WORK(&m->rq, p9_read_work);
+	INIT_WORK(&m->wq, p9_write_work);
+	INIT_LIST_HEAD(&m->poll_pending_link);
+	init_poll_funcptr(&m->pt, p9_pollwait);
 
-		p9_printfcall(buf, sizeof(buf), tc, m->extended);
-		printk(KERN_NOTICE "<<< %p %s\n", m, buf);
+	n = p9_fd_poll(client, &m->pt);
+	if (n & POLLIN) {
+		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can read\n", m);
+		set_bit(Rpending, &m->wsched);
 	}
-#endif
-
-	spin_lock_init(&req->lock);
-	req->tag = n;
-	req->tcall = tc;
-	req->rcall = NULL;
-	req->err = 0;
-	req->cb = cb;
-	req->cba = cba;
-	req->flush = None;
-
-	spin_lock(&m->lock);
-	list_add_tail(&req->req_list, &m->unsent_req_list);
-	spin_unlock(&m->lock);
-
-	if (test_and_clear_bit(Wpending, &m->wsched))
-		n = POLLOUT;
-	else
-		n = p9_fd_poll(m->trans, NULL);
 
-	if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
-		queue_work(p9_mux_wq, &m->wq);
+	if (n & POLLOUT) {
+		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can write\n", m);
+		set_bit(Wpending, &m->wsched);
+	}
 
-	return req;
+	return m;
 }
 
-static void p9_mux_free_request(struct p9_conn *m, struct p9_req *req)
-{
-	p9_mux_put_tag(m, req->tag);
-	kfree(req);
-}
+/**
+ * p9_poll_mux - polls a mux and schedules read or write works if necessary
+ * @m: connection to poll
+ *
+ */
 
-static void p9_mux_flush_cb(struct p9_req *freq, void *a)
+static void p9_poll_mux(struct p9_conn *m)
 {
-	int tag;
-	struct p9_conn *m;
-	struct p9_req *req, *rreq, *rptr;
-
-	m = a;
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p rc %p err %d oldtag %d\n", m,
-		freq->tcall, freq->rcall, freq->err,
-		freq->tcall->params.tflush.oldtag);
-
-	spin_lock(&m->lock);
-	tag = freq->tcall->params.tflush.oldtag;
-	req = NULL;
-	list_for_each_entry_safe(rreq, rptr, &m->req_list, req_list) {
-		if (rreq->tag == tag) {
-			req = rreq;
-			list_del(&req->req_list);
-			break;
-		}
-	}
-	spin_unlock(&m->lock);
+	int n;
 
-	if (req) {
-		spin_lock(&req->lock);
-		req->flush = Flushed;
-		spin_unlock(&req->lock);
+	if (m->err < 0)
+		return;
 
-		if (req->cb)
-			(*req->cb) (req, req->cba);
-		else
-			kfree(req->rcall);
+	n = p9_fd_poll(m->client, NULL);
+	if (n < 0 || n & (POLLERR | POLLHUP | POLLNVAL)) {
+		P9_DPRINTK(P9_DEBUG_TRANS, "error mux %p err %d\n", m, n);
+		if (n >= 0)
+			n = -ECONNRESET;
+		p9_conn_cancel(m, n);
 	}
 
-	kfree(freq->tcall);
-	kfree(freq->rcall);
-	p9_mux_free_request(m, freq);
-}
-
-static int
-p9_mux_flush_request(struct p9_conn *m, struct p9_req *req)
-{
-	struct p9_fcall *fc;
-	struct p9_req *rreq, *rptr;
-
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p req %p tag %d\n", m, req, req->tag);
-
-	/* if a response was received for a request, do nothing */
-	spin_lock(&req->lock);
-	if (req->rcall || req->err) {
-		spin_unlock(&req->lock);
-		P9_DPRINTK(P9_DEBUG_MUX,
-			"mux %p req %p response already received\n", m, req);
-		return 0;
+	if (n & POLLIN) {
+		set_bit(Rpending, &m->wsched);
+		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can read\n", m);
+		if (!test_and_set_bit(Rworksched, &m->wsched)) {
+			P9_DPRINTK(P9_DEBUG_TRANS, "sched read work %p\n", m);
+			queue_work(p9_mux_wq, &m->rq);
+		}
 	}
 
-	req->flush = Flushing;
-	spin_unlock(&req->lock);
-
-	spin_lock(&m->lock);
-	/* if the request is not sent yet, just remove it from the list */
-	list_for_each_entry_safe(rreq, rptr, &m->unsent_req_list, req_list) {
-		if (rreq->tag == req->tag) {
-			P9_DPRINTK(P9_DEBUG_MUX,
-			   "mux %p req %p request is not sent yet\n", m, req);
-			list_del(&rreq->req_list);
-			req->flush = Flushed;
-			spin_unlock(&m->lock);
-			if (req->cb)
-				(*req->cb) (req, req->cba);
-			return 0;
+	if (n & POLLOUT) {
+		set_bit(Wpending, &m->wsched);
+		P9_DPRINTK(P9_DEBUG_TRANS, "mux %p can write\n", m);
+		if ((m->wsize || !list_empty(&m->unsent_req_list))
+		    && !test_and_set_bit(Wworksched, &m->wsched)) {
+			P9_DPRINTK(P9_DEBUG_TRANS, "sched write work %p\n", m);
+			queue_work(p9_mux_wq, &m->wq);
 		}
 	}
-	spin_unlock(&m->lock);
-
-	clear_thread_flag(TIF_SIGPENDING);
-	fc = p9_create_tflush(req->tag);
-	p9_send_request(m, fc, p9_mux_flush_cb, m);
-	return 1;
-}
-
-static void
-p9_conn_rpc_cb(struct p9_req *req, void *a)
-{
-	struct p9_mux_rpc *r;
-
-	P9_DPRINTK(P9_DEBUG_MUX, "req %p r %p\n", req, a);
-	r = a;
-	r->rcall = req->rcall;
-	r->err = req->err;
-
-	if (req->flush != None && !req->err)
-		r->err = -ERESTARTSYS;
-
-	wake_up(&r->wqueue);
 }
 
 /**
- * p9_fd_rpc- sends 9P request and waits until a response is available.
- *	The function can be interrupted.
- * @t: transport data
- * @tc: request to be sent
- * @rc: pointer where a pointer to the response is stored
+ * p9_fd_request - send 9P request
+ * The function can sleep until the request is scheduled for sending.
+ * The function can be interrupted. Return from the function is not
+ * a guarantee that the request is sent successfully.
+ *
+ * @client: client instance
+ * @req: request to be sent
  *
  */
 
-int
-p9_fd_rpc(struct p9_trans *t, struct p9_fcall *tc, struct p9_fcall **rc)
+static int p9_fd_request(struct p9_client *client, struct p9_req_t *req)
 {
-	struct p9_trans_fd *p = t->priv;
-	struct p9_conn *m = p->conn;
-	int err, sigpending;
-	unsigned long flags;
-	struct p9_req *req;
-	struct p9_mux_rpc r;
-
-	r.err = 0;
-	r.tcall = tc;
-	r.rcall = NULL;
-	r.m = m;
-	init_waitqueue_head(&r.wqueue);
-
-	if (rc)
-		*rc = NULL;
-
-	sigpending = 0;
-	if (signal_pending(current)) {
-		sigpending = 1;
-		clear_thread_flag(TIF_SIGPENDING);
-	}
-
-	req = p9_send_request(m, tc, p9_conn_rpc_cb, &r);
-	if (IS_ERR(req)) {
-		err = PTR_ERR(req);
-		P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
-		return err;
-	}
+	int n;
+	struct p9_trans_fd *ts = client->trans;
+	struct p9_conn *m = ts->conn;
 
-	err = wait_event_interruptible(r.wqueue, r.rcall != NULL || r.err < 0);
-	if (r.err < 0)
-		err = r.err;
-
-	if (err == -ERESTARTSYS && m->trans->status == Connected
-							&& m->err == 0) {
-		if (p9_mux_flush_request(m, req)) {
-			/* wait until we get response of the flush message */
-			do {
-				clear_thread_flag(TIF_SIGPENDING);
-				err = wait_event_interruptible(r.wqueue,
-					r.rcall || r.err);
-			} while (!r.rcall && !r.err && err == -ERESTARTSYS &&
-				m->trans->status == Connected && !m->err);
-
-			err = -ERESTARTSYS;
-		}
-		sigpending = 1;
-	}
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p task %p tcall %p id %d\n", m,
+						current, req->tc, req->tc->id);
+	if (m->err < 0)
+		return m->err;
 
-	if (sigpending) {
-		spin_lock_irqsave(&current->sighand->siglock, flags);
-		recalc_sigpending();
-		spin_unlock_irqrestore(&current->sighand->siglock, flags);
-	}
+	spin_lock(&client->lock);
+	req->status = REQ_STATUS_UNSENT;
+	list_add_tail(&req->req_list, &m->unsent_req_list);
+	spin_unlock(&client->lock);
 
-	if (rc)
-		*rc = r.rcall;
+	if (test_and_clear_bit(Wpending, &m->wsched))
+		n = POLLOUT;
 	else
-		kfree(r.rcall);
+		n = p9_fd_poll(m->client, NULL);
 
-	p9_mux_free_request(m, req);
-	if (err > 0)
-		err = -EIO;
+	if (n & POLLOUT && !test_and_set_bit(Wworksched, &m->wsched))
+		queue_work(p9_mux_wq, &m->wq);
 
-	return err;
+	return 0;
 }
 
-#ifdef P9_NONBLOCK
-/**
- * p9_conn_rpcnb - sends 9P request without waiting for response.
- * @m: mux data
- * @tc: request to be sent
- * @cb: callback function to be called when response arrives
- * @a: value to pass to the callback function
- *
- */
-
-int p9_conn_rpcnb(struct p9_conn *m, struct p9_fcall *tc,
-		   p9_conn_req_callback cb, void *a)
+static int p9_fd_cancel(struct p9_client *client, struct p9_req_t *req)
 {
-	int err;
-	struct p9_req *req;
+	struct p9_trans_fd *ts = client->trans;
+	struct p9_conn *m = ts->conn;
+	int ret = 1;
 
-	req = p9_send_request(m, tc, cb, a);
-	if (IS_ERR(req)) {
-		err = PTR_ERR(req);
-		P9_DPRINTK(P9_DEBUG_MUX, "error %d\n", err);
-		return PTR_ERR(req);
-	}
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p req %p\n", m, req);
 
-	P9_DPRINTK(P9_DEBUG_MUX, "mux %p tc %p tag %d\n", m, tc, req->tag);
-	return 0;
-}
-#endif /* P9_NONBLOCK */
+	spin_lock(&client->lock);
+	list_del(&req->req_list);
 
-/**
- * p9_conn_cancel - cancel all pending requests with error
- * @m: mux data
- * @err: error code
- *
- */
-
-void p9_conn_cancel(struct p9_conn *m, int err)
-{
-	struct p9_req *req, *rtmp;
-	LIST_HEAD(cancel_list);
-
-	P9_DPRINTK(P9_DEBUG_ERROR, "mux %p err %d\n", m, err);
-	m->err = err;
-	spin_lock(&m->lock);
-	list_for_each_entry_safe(req, rtmp, &m->req_list, req_list) {
-		list_move(&req->req_list, &cancel_list);
+	if (req->status == REQ_STATUS_UNSENT) {
+		req->status = REQ_STATUS_FLSHD;
+		ret = 0;
 	}
-	list_for_each_entry_safe(req, rtmp, &m->unsent_req_list, req_list) {
-		list_move(&req->req_list, &cancel_list);
-	}
-	spin_unlock(&m->lock);
 
-	list_for_each_entry_safe(req, rtmp, &cancel_list, req_list) {
-		list_del(&req->req_list);
-		if (!req->err)
-			req->err = err;
+	spin_unlock(&client->lock);
 
-		if (req->cb)
-			(*req->cb) (req, req->cba);
-		else
-			kfree(req->rcall);
-	}
+	return ret;
 }
 
 /**
@@ -1216,7 +758,7 @@ static int parse_opts(char *params, struct p9_fd_opts *opts)
 	return 0;
 }
 
-static int p9_fd_open(struct p9_trans *trans, int rfd, int wfd)
+static int p9_fd_open(struct p9_client *client, int rfd, int wfd)
 {
 	struct p9_trans_fd *ts = kmalloc(sizeof(struct p9_trans_fd),
 					   GFP_KERNEL);
@@ -1234,13 +776,13 @@ static int p9_fd_open(struct p9_trans *trans, int rfd, int wfd)
 		return -EIO;
 	}
 
-	trans->priv = ts;
-	trans->status = Connected;
+	client->trans = ts;
+	client->status = Connected;
 
 	return 0;
 }
 
-static int p9_socket_open(struct p9_trans *trans, struct socket *csocket)
+static int p9_socket_open(struct p9_client *client, struct socket *csocket)
 {
 	int fd, ret;
 
@@ -1251,137 +793,65 @@ static int p9_socket_open(struct p9_trans *trans, struct socket *csocket)
 		return fd;
 	}
 
-	ret = p9_fd_open(trans, fd, fd);
+	ret = p9_fd_open(client, fd, fd);
 	if (ret < 0) {
 		P9_EPRINTK(KERN_ERR, "p9_socket_open: failed to open fd\n");
 		sockfd_put(csocket);
 		return ret;
 	}
 
-	((struct p9_trans_fd *)trans->priv)->rd->f_flags |= O_NONBLOCK;
+	((struct p9_trans_fd *)client->trans)->rd->f_flags |= O_NONBLOCK;
 
 	return 0;
 }
 
 /**
- * p9_fd_read- read from a fd
- * @trans: transport instance state
- * @v: buffer to receive data into
- * @len: size of receive buffer
- *
- */
-
-static int p9_fd_read(struct p9_trans *trans, void *v, int len)
-{
-	int ret;
-	struct p9_trans_fd *ts = NULL;
-
-	if (trans && trans->status != Disconnected)
-		ts = trans->priv;
-
-	if (!ts)
-		return -EREMOTEIO;
-
-	if (!(ts->rd->f_flags & O_NONBLOCK))
-		P9_DPRINTK(P9_DEBUG_ERROR, "blocking read ...\n");
-
-	ret = kernel_read(ts->rd, ts->rd->f_pos, v, len);
-	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
-		trans->status = Disconnected;
-	return ret;
-}
-
-/**
- * p9_fd_write - write to a socket
- * @trans: transport instance state
- * @v: buffer to send data from
- * @len: size of send buffer
+ * p9_mux_destroy - cancels all pending requests and frees mux resources
+ * @m: mux to destroy
  *
  */
 
-static int p9_fd_write(struct p9_trans *trans, void *v, int len)
-{
-	int ret;
-	mm_segment_t oldfs;
-	struct p9_trans_fd *ts = NULL;
-
-	if (trans && trans->status != Disconnected)
-		ts = trans->priv;
-
-	if (!ts)
-		return -EREMOTEIO;
-
-	if (!(ts->wr->f_flags & O_NONBLOCK))
-		P9_DPRINTK(P9_DEBUG_ERROR, "blocking write ...\n");
-
-	oldfs = get_fs();
-	set_fs(get_ds());
-	/* The cast to a user pointer is valid due to the set_fs() */
-	ret = vfs_write(ts->wr, (void __user *)v, len, &ts->wr->f_pos);
-	set_fs(oldfs);
-
-	if (ret <= 0 && ret != -ERESTARTSYS && ret != -EAGAIN)
-		trans->status = Disconnected;
-	return ret;
-}
-
-static unsigned int
-p9_fd_poll(struct p9_trans *trans, struct poll_table_struct *pt)
+static void p9_conn_destroy(struct p9_conn *m)
 {
-	int ret, n;
-	struct p9_trans_fd *ts = NULL;
-
-	if (trans && trans->status == Connected)
-		ts = trans->priv;
-
-	if (!ts)
-		return -EREMOTEIO;
-
-	if (!ts->rd->f_op || !ts->rd->f_op->poll)
-		return -EIO;
-
-	if (!ts->wr->f_op || !ts->wr->f_op->poll)
-		return -EIO;
+	P9_DPRINTK(P9_DEBUG_TRANS, "mux %p prev %p next %p\n", m,
+		m->mux_list.prev, m->mux_list.next);
 
-	ret = ts->rd->f_op->poll(ts->rd, pt);
-	if (ret < 0)
-		return ret;
+	p9_mux_poll_stop(m);
+	cancel_work_sync(&m->rq);
+	cancel_work_sync(&m->wq);
 
-	if (ts->rd != ts->wr) {
-		n = ts->wr->f_op->poll(ts->wr, pt);
-		if (n < 0)
-			return n;
-		ret = (ret & ~POLLOUT) | (n & ~POLLIN);
-	}
+	p9_conn_cancel(m, -ECONNRESET);
 
-	return ret;
+	m->client = NULL;
+	kfree(m);
 }
 
 /**
- * p9_fd_close - shutdown socket
- * @trans: private socket structure
+ * p9_fd_close - shutdown file descriptor transport
+ * @client: client instance
  *
  */
 
-static void p9_fd_close(struct p9_trans *trans)
+static void p9_fd_close(struct p9_client *client)
 {
 	struct p9_trans_fd *ts;
 
-	if (!trans)
+	if (!client)
 		return;
 
-	ts = xchg(&trans->priv, NULL);
-
+	ts = client->trans;
 	if (!ts)
 		return;
 
+	client->status = Disconnected;
+
 	p9_conn_destroy(ts->conn);
 
-	trans->status = Disconnected;
 	if (ts->rd)
 		fput(ts->rd);
 	if (ts->wr)
 		fput(ts->wr);
+
 	kfree(ts);
 }
 
@@ -1402,31 +872,23 @@ static inline int valid_ipaddr4(const char *buf)
 	return 0;
 }
 
-static struct p9_trans *
-p9_trans_create_tcp(const char *addr, char *args, int msize, unsigned char dotu)
+static int
+p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
 {
 	int err;
-	struct p9_trans *trans;
 	struct socket *csocket;
 	struct sockaddr_in sin_server;
 	struct p9_fd_opts opts;
-	struct p9_trans_fd *p;
+	struct p9_trans_fd *p = NULL; /* this gets allocated in p9_fd_open */
 
 	err = parse_opts(args, &opts);
 	if (err < 0)
-		return ERR_PTR(err);
+		return err;
 
 	if (valid_ipaddr4(addr) < 0)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	csocket = NULL;
-	trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
-	if (!trans)
-		return ERR_PTR(-ENOMEM);
-	trans->msize = msize;
-	trans->extended = dotu;
-	trans->rpc = p9_fd_rpc;
-	trans->close = p9_fd_close;
 
 	sin_server.sin_family = AF_INET;
 	sin_server.sin_addr.s_addr = in_aton(addr);
@@ -1449,45 +911,38 @@ p9_trans_create_tcp(const char *addr, char *args, int msize, unsigned char dotu)
 		goto error;
 	}
 
-	err = p9_socket_open(trans, csocket);
+	err = p9_socket_open(client, csocket);
 	if (err < 0)
 		goto error;
 
-	p = (struct p9_trans_fd *) trans->priv;
-	p->conn = p9_conn_create(trans);
+	p = (struct p9_trans_fd *) client->trans;
+	p->conn = p9_conn_create(client);
 	if (IS_ERR(p->conn)) {
 		err = PTR_ERR(p->conn);
 		p->conn = NULL;
 		goto error;
 	}
 
-	return trans;
+	return 0;
 
 error:
 	if (csocket)
 		sock_release(csocket);
 
-	kfree(trans);
-	return ERR_PTR(err);
+	kfree(p);
+
+	return err;
 }
 
-static struct p9_trans *
-p9_trans_create_unix(const char *addr, char *args, int msize,
-							unsigned char dotu)
+static int
+p9_fd_create_unix(struct p9_client *client, const char *addr, char *args)
 {
 	int err;
 	struct socket *csocket;
 	struct sockaddr_un sun_server;
-	struct p9_trans *trans;
-	struct p9_trans_fd *p;
+	struct p9_trans_fd *p = NULL; /* this gets allocated in p9_fd_open */
 
 	csocket = NULL;
-	trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
-	if (!trans)
-		return ERR_PTR(-ENOMEM);
-
-	trans->rpc = p9_fd_rpc;
-	trans->close = p9_fd_close;
 
 	if (strlen(addr) > UNIX_PATH_MAX) {
 		P9_EPRINTK(KERN_ERR, "p9_trans_unix: address too long: %s\n",
@@ -1508,79 +963,69 @@ p9_trans_create_unix(const char *addr, char *args, int msize,
 		goto error;
 	}
 
-	err = p9_socket_open(trans, csocket);
+	err = p9_socket_open(client, csocket);
 	if (err < 0)
 		goto error;
 
-	trans->msize = msize;
-	trans->extended = dotu;
-	p = (struct p9_trans_fd *) trans->priv;
-	p->conn = p9_conn_create(trans);
+	p = (struct p9_trans_fd *) client->trans;
+	p->conn = p9_conn_create(client);
 	if (IS_ERR(p->conn)) {
 		err = PTR_ERR(p->conn);
 		p->conn = NULL;
 		goto error;
 	}
 
-	return trans;
+	return 0;
 
 error:
 	if (csocket)
 		sock_release(csocket);
 
-	kfree(trans);
-	return ERR_PTR(err);
+	kfree(p);
+	return err;
 }
 
-static struct p9_trans *
-p9_trans_create_fd(const char *name, char *args, int msize,
-							unsigned char extended)
+static int
+p9_fd_create(struct p9_client *client, const char *addr, char *args)
 {
 	int err;
-	struct p9_trans *trans;
 	struct p9_fd_opts opts;
-	struct p9_trans_fd *p;
+	struct p9_trans_fd *p = NULL; /* this get allocated in p9_fd_open */
 
 	parse_opts(args, &opts);
 
 	if (opts.rfd == ~0 || opts.wfd == ~0) {
 		printk(KERN_ERR "v9fs: Insufficient options for proto=fd\n");
-		return ERR_PTR(-ENOPROTOOPT);
+		return -ENOPROTOOPT;
 	}
 
-	trans = kmalloc(sizeof(struct p9_trans), GFP_KERNEL);
-	if (!trans)
-		return ERR_PTR(-ENOMEM);
-
-	trans->rpc = p9_fd_rpc;
-	trans->close = p9_fd_close;
-
-	err = p9_fd_open(trans, opts.rfd, opts.wfd);
+	err = p9_fd_open(client, opts.rfd, opts.wfd);
 	if (err < 0)
 		goto error;
 
-	trans->msize = msize;
-	trans->extended = extended;
-	p = (struct p9_trans_fd *) trans->priv;
-	p->conn = p9_conn_create(trans);
+	p = (struct p9_trans_fd *) client->trans;
+	p->conn = p9_conn_create(client);
 	if (IS_ERR(p->conn)) {
 		err = PTR_ERR(p->conn);
 		p->conn = NULL;
 		goto error;
 	}
 
-	return trans;
+	return 0;
 
 error:
-	kfree(trans);
-	return ERR_PTR(err);
+	kfree(p);
+	return err;
 }
 
 static struct p9_trans_module p9_tcp_trans = {
 	.name = "tcp",
 	.maxsize = MAX_SOCK_BUF,
 	.def = 1,
-	.create = p9_trans_create_tcp,
+	.create = p9_fd_create_tcp,
+	.close = p9_fd_close,
+	.request = p9_fd_request,
+	.cancel = p9_fd_cancel,
 	.owner = THIS_MODULE,
 };
 
@@ -1588,7 +1033,10 @@ static struct p9_trans_module p9_unix_trans = {
 	.name = "unix",
 	.maxsize = MAX_SOCK_BUF,
 	.def = 0,
-	.create = p9_trans_create_unix,
+	.create = p9_fd_create_unix,
+	.close = p9_fd_close,
+	.request = p9_fd_request,
+	.cancel = p9_fd_cancel,
 	.owner = THIS_MODULE,
 };
 
@@ -1596,23 +1044,71 @@ static struct p9_trans_module p9_fd_trans = {
 	.name = "fd",
 	.maxsize = MAX_SOCK_BUF,
 	.def = 0,
-	.create = p9_trans_create_fd,
+	.create = p9_fd_create,
+	.close = p9_fd_close,
+	.request = p9_fd_request,
+	.cancel = p9_fd_cancel,
 	.owner = THIS_MODULE,
 };
 
-int p9_trans_fd_init(void)
+/**
+ * p9_poll_proc - poll worker thread
+ * @a: thread state and arguments
+ *
+ * polls all v9fs transports for new events and queues the appropriate
+ * work to the work queue
+ *
+ */
+
+static int p9_poll_proc(void *a)
 {
-	int i;
+	unsigned long flags;
+
+	P9_DPRINTK(P9_DEBUG_TRANS, "start %p\n", current);
+ repeat:
+	spin_lock_irqsave(&p9_poll_lock, flags);
+	while (!list_empty(&p9_poll_pending_list)) {
+		struct p9_conn *conn = list_first_entry(&p9_poll_pending_list,
+							struct p9_conn,
+							poll_pending_link);
+		list_del_init(&conn->poll_pending_link);
+		spin_unlock_irqrestore(&p9_poll_lock, flags);
+
+		p9_poll_mux(conn);
+
+		spin_lock_irqsave(&p9_poll_lock, flags);
+	}
+	spin_unlock_irqrestore(&p9_poll_lock, flags);
+
+	set_current_state(TASK_INTERRUPTIBLE);
+	if (list_empty(&p9_poll_pending_list)) {
+		P9_DPRINTK(P9_DEBUG_TRANS, "sleeping...\n");
+		schedule();
+	}
+	__set_current_state(TASK_RUNNING);
+
+	if (!kthread_should_stop())
+		goto repeat;
 
-	for (i = 0; i < ARRAY_SIZE(p9_mux_poll_tasks); i++)
-		p9_mux_poll_tasks[i].task = NULL;
+	P9_DPRINTK(P9_DEBUG_TRANS, "finish\n");
+	return 0;
+}
 
+int p9_trans_fd_init(void)
+{
 	p9_mux_wq = create_workqueue("v9fs");
 	if (!p9_mux_wq) {
 		printk(KERN_WARNING "v9fs: mux: creating workqueue failed\n");
 		return -ENOMEM;
 	}
 
+	p9_poll_task = kthread_run(p9_poll_proc, NULL, "v9fs-poll");
+	if (IS_ERR(p9_poll_task)) {
+		destroy_workqueue(p9_mux_wq);
+		printk(KERN_WARNING "v9fs: mux: creating poll task failed\n");
+		return PTR_ERR(p9_poll_task);
+	}
+
 	v9fs_register_trans(&p9_tcp_trans);
 	v9fs_register_trans(&p9_unix_trans);
 	v9fs_register_trans(&p9_fd_trans);
@@ -1622,6 +1118,7 @@ int p9_trans_fd_init(void)
 
 void p9_trans_fd_exit(void)
 {
+	kthread_stop(p9_poll_task);
 	v9fs_unregister_trans(&p9_tcp_trans);
 	v9fs_unregister_trans(&p9_unix_trans);
 	v9fs_unregister_trans(&p9_fd_trans);