Merge tag 'for-linus-3.11-merge-window-part-2' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs

Pull second round of 9p patches from Eric Van Hensbergen:
 "Several of these patches were rebased in order to correct style
  issues.  Only stylistic changes were made versus the patches which
  were in linux-next for two weeks.  The rebases have been in linux-next
  for 3 days and have passed my regressions.

  The bulk of these are RDMA fixes and improvements.  There's also some
  additions on the extended attributes front to support some additional
  namespaces and a new option for TCP to force allocation of mount
  requests from a priviledged port"

* tag 'for-linus-3.11-merge-window-part-2' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  fs/9p: Remove the unused variable "err" in v9fs_vfs_getattr()
  9P: Add cancelled() to the transport functions.
  9P/RDMA: count posted buffers without a pending request
  9P/RDMA: Improve error handling in rdma_request
  9P/RDMA: Do not free req->rc in error handling in rdma_request()
  9P/RDMA: Use a semaphore to protect the RQ
  9P/RDMA: Protect against duplicate replies
  9P/RDMA: increase P9_RDMA_MAXSIZE to 1MB
  9pnet: refactor struct p9_fcall alloc code
  9P/RDMA: rdma_request() needs not allocate req->rc
  9P: Fix fcall allocation for rdma
  fs/9p: xattr: add trusted and security namespaces
  net/9p: add privport option to 9p tcp transport

3 files changed, 167 insertions, 76 deletions

diff --git a/net/9p/client.c b/net/9p/client.c
index 01f1779eba8..8b93cae2d11 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -204,6 +204,17 @@ free_and_return:
 	return ret;
 }
 
+struct p9_fcall *p9_fcall_alloc(int alloc_msize)
+{
+	struct p9_fcall *fc;
+	fc = kmalloc(sizeof(struct p9_fcall) + alloc_msize, GFP_NOFS);
+	if (!fc)
+		return NULL;
+	fc->capacity = alloc_msize;
+	fc->sdata = (char *) fc + sizeof(struct p9_fcall);
+	return fc;
+}
+
 /**
  * p9_tag_alloc - lookup/allocate a request by tag
  * @c: client session to lookup tag within
@@ -256,39 +267,36 @@ p9_tag_alloc(struct p9_client *c, u16 tag, unsigned int max_size)
 	col = tag % P9_ROW_MAXTAG;
 
 	req = &c->reqs[row][col];
-	if (!req->tc) {
+	if (!req->wq) {
 		req->wq = kmalloc(sizeof(wait_queue_head_t), GFP_NOFS);
-		if (!req->wq) {
-			pr_err("Couldn't grow tag array\n");
-			return ERR_PTR(-ENOMEM);
-		}
+		if (!req->wq)
+			goto grow_failed;
 		init_waitqueue_head(req->wq);
-		req->tc = kmalloc(sizeof(struct p9_fcall) + alloc_msize,
-				  GFP_NOFS);
-		req->rc = kmalloc(sizeof(struct p9_fcall) + alloc_msize,
-				  GFP_NOFS);
-		if ((!req->tc) || (!req->rc)) {
-			pr_err("Couldn't grow tag array\n");
-			kfree(req->tc);
-			kfree(req->rc);
-			kfree(req->wq);
-			req->tc = req->rc = NULL;
-			req->wq = NULL;
-			return ERR_PTR(-ENOMEM);
-		}
-		req->tc->capacity = alloc_msize;
-		req->rc->capacity = alloc_msize;
-		req->tc->sdata = (char *) req->tc + sizeof(struct p9_fcall);
-		req->rc->sdata = (char *) req->rc + sizeof(struct p9_fcall);
 	}
 
+	if (!req->tc)
+		req->tc = p9_fcall_alloc(alloc_msize);
+	if (!req->rc)
+		req->rc = p9_fcall_alloc(alloc_msize);
+	if (!req->tc || !req->rc)
+		goto grow_failed;
+
 	p9pdu_reset(req->tc);
 	p9pdu_reset(req->rc);
 
 	req->tc->tag = tag-1;
 	req->status = REQ_STATUS_ALLOC;
 
-	return &c->reqs[row][col];
+	return req;
+
+grow_failed:
+	pr_err("Couldn't grow tag array\n");
+	kfree(req->tc);
+	kfree(req->rc);
+	kfree(req->wq);
+	req->tc = req->rc = NULL;
+	req->wq = NULL;
+	return ERR_PTR(-ENOMEM);
 }
 
 /**
@@ -648,12 +656,20 @@ static int p9_client_flush(struct p9_client *c, struct p9_req_t *oldreq)
 		return PTR_ERR(req);
 
 
-	/* if we haven't received a response for oldreq,
-	   remove it from the list. */
+	/*
+	 * if we haven't received a response for oldreq,
+	 * remove it from the list, and notify the transport
+	 * layer that the reply will never arrive.
+	 */
 	spin_lock(&c->lock);
-	if (oldreq->status == REQ_STATUS_FLSH)
+	if (oldreq->status == REQ_STATUS_FLSH) {
 		list_del(&oldreq->req_list);
-	spin_unlock(&c->lock);
+		spin_unlock(&c->lock);
+		if (c->trans_mod->cancelled)
+			c->trans_mod->cancelled(c, req);
+	} else {
+		spin_unlock(&c->lock);
+	}
 
 	p9_free_req(c, req);
 	return 0;
diff --git a/net/9p/trans_fd.c b/net/9p/trans_fd.c
index 02efb25c295..3ffda1b3799 100644
--- a/net/9p/trans_fd.c
+++ b/net/9p/trans_fd.c
@@ -63,6 +63,7 @@ struct p9_fd_opts {
 	int rfd;
 	int wfd;
 	u16 port;
+	int privport;
 };
 
 /**
@@ -87,12 +88,15 @@ struct p9_trans_fd {
 enum {
 	/* Options that take integer arguments */
 	Opt_port, Opt_rfdno, Opt_wfdno, Opt_err,
+	/* Options that take no arguments */
+	Opt_privport,
 };
 
 static const match_table_t tokens = {
 	{Opt_port, "port=%u"},
 	{Opt_rfdno, "rfdno=%u"},
 	{Opt_wfdno, "wfdno=%u"},
+	{Opt_privport, "privport"},
 	{Opt_err, NULL},
 };
 
@@ -161,6 +165,9 @@ static DEFINE_SPINLOCK(p9_poll_lock);
 static LIST_HEAD(p9_poll_pending_list);
 static DECLARE_WORK(p9_poll_work, p9_poll_workfn);
 
+static unsigned int p9_ipport_resv_min = P9_DEF_MIN_RESVPORT;
+static unsigned int p9_ipport_resv_max = P9_DEF_MAX_RESVPORT;
+
 static void p9_mux_poll_stop(struct p9_conn *m)
 {
 	unsigned long flags;
@@ -741,7 +748,7 @@ static int parse_opts(char *params, struct p9_fd_opts *opts)
 		if (!*p)
 			continue;
 		token = match_token(p, tokens, args);
-		if (token != Opt_err) {
+		if ((token != Opt_err) && (token != Opt_privport)) {
 			r = match_int(&args[0], &option);
 			if (r < 0) {
 				p9_debug(P9_DEBUG_ERROR,
@@ -759,6 +766,9 @@ static int parse_opts(char *params, struct p9_fd_opts *opts)
 		case Opt_wfdno:
 			opts->wfd = option;
 			break;
+		case Opt_privport:
+			opts->privport = 1;
+			break;
 		default:
 			continue;
 		}
@@ -898,6 +908,24 @@ static inline int valid_ipaddr4(const char *buf)
 	return 0;
 }
 
+static int p9_bind_privport(struct socket *sock)
+{
+	struct sockaddr_in cl;
+	int port, err = -EINVAL;
+
+	memset(&cl, 0, sizeof(cl));
+	cl.sin_family = AF_INET;
+	cl.sin_addr.s_addr = INADDR_ANY;
+	for (port = p9_ipport_resv_max; port >= p9_ipport_resv_min; port--) {
+		cl.sin_port = htons((ushort)port);
+		err = kernel_bind(sock, (struct sockaddr *)&cl, sizeof(cl));
+		if (err != -EADDRINUSE)
+			break;
+	}
+	return err;
+}
+
+
 static int
 p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
 {
@@ -926,6 +954,16 @@ p9_fd_create_tcp(struct p9_client *client, const char *addr, char *args)
 		return err;
 	}
 
+	if (opts.privport) {
+		err = p9_bind_privport(csocket);
+		if (err < 0) {
+			pr_err("%s (%d): problem binding to privport\n",
+			       __func__, task_pid_nr(current));
+			sock_release(csocket);
+			return err;
+		}
+	}
+
 	err = csocket->ops->connect(csocket,
 				    (struct sockaddr *)&sin_server,
 				    sizeof(struct sockaddr_in), 0);
diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c
index 2c69ddd691a..928f2bb9bf8 100644
--- a/net/9p/trans_rdma.c
+++ b/net/9p/trans_rdma.c
@@ -57,9 +57,7 @@
 #define P9_RDMA_IRD		0
 #define P9_RDMA_ORD		0
 #define P9_RDMA_TIMEOUT		30000		/* 30 seconds */
-#define P9_RDMA_MAXSIZE		(4*4096)	/* Min SGE is 4, so we can
-						 * safely advertise a maxsize
-						 * of 64k */
+#define P9_RDMA_MAXSIZE		(1024*1024)	/* 1MB */
 
 /**
  * struct p9_trans_rdma - RDMA transport instance
@@ -75,7 +73,9 @@
  * @sq_depth: The depth of the Send Queue
  * @sq_sem: Semaphore for the SQ
  * @rq_depth: The depth of the Receive Queue.
- * @rq_count: Count of requests in the Receive Queue.
+ * @rq_sem: Semaphore for the RQ
+ * @excess_rc : Amount of posted Receive Contexts without a pending request.
+ *		See rdma_request()
  * @addr: The remote peer's address
  * @req_lock: Protects the active request list
  * @cm_done: Completion event for connection management tracking
@@ -100,7 +100,8 @@ struct p9_trans_rdma {
 	int sq_depth;
 	struct semaphore sq_sem;
 	int rq_depth;
-	atomic_t rq_count;
+	struct semaphore rq_sem;
+	atomic_t excess_rc;
 	struct sockaddr_in addr;
 	spinlock_t req_lock;
 
@@ -296,6 +297,13 @@ handle_recv(struct p9_client *client, struct p9_trans_rdma *rdma,
 	if (!req)
 		goto err_out;
 
+	/* Check that we have not yet received a reply for this request.
+	 */
+	if (unlikely(req->rc)) {
+		pr_err("Duplicate reply for request %d", tag);
+		goto err_out;
+	}
+
 	req->rc = c->rc;
 	req->status = REQ_STATUS_RCVD;
 	p9_client_cb(client, req);
@@ -336,8 +344,8 @@ static void cq_comp_handler(struct ib_cq *cq, void *cq_context)
 
 		switch (c->wc_op) {
 		case IB_WC_RECV:
-			atomic_dec(&rdma->rq_count);
 			handle_recv(client, rdma, c, wc.status, wc.byte_len);
+			up(&rdma->rq_sem);
 			break;
 
 		case IB_WC_SEND:
@@ -421,32 +429,33 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 	struct p9_rdma_context *c = NULL;
 	struct p9_rdma_context *rpl_context = NULL;
 
+	/* When an error occurs between posting the recv and the send,
+	 * there will be a receive context posted without a pending request.
+	 * Since there is no way to "un-post" it, we remember it and skip
+	 * post_recv() for the next request.
+	 * So here,
+	 * see if we are this `next request' and need to absorb an excess rc.
+	 * If yes, then drop and free our own, and do not recv_post().
+	 **/
+	if (unlikely(atomic_read(&rdma->excess_rc) > 0)) {
+		if ((atomic_sub_return(1, &rdma->excess_rc) >= 0)) {
+			/* Got one ! */
+			kfree(req->rc);
+			req->rc = NULL;
+			goto dont_need_post_recv;
+		} else {
+			/* We raced and lost. */
+			atomic_inc(&rdma->excess_rc);
+		}
+	}
+
 	/* Allocate an fcall for the reply */
 	rpl_context = kmalloc(sizeof *rpl_context, GFP_NOFS);
 	if (!rpl_context) {
 		err = -ENOMEM;
-		goto err_close;
-	}
-
-	/*
-	 * If the request has a buffer, steal it, otherwise
-	 * allocate a new one.  Typically, requests should already
-	 * have receive buffers allocated and just swap them around
-	 */
-	if (!req->rc) {
-		req->rc = kmalloc(sizeof(struct p9_fcall)+client->msize,
-				  GFP_NOFS);
-		if (req->rc) {
-			req->rc->sdata = (char *) req->rc +
-						sizeof(struct p9_fcall);
-			req->rc->capacity = client->msize;
-		}
+		goto recv_error;
 	}
 	rpl_context->rc = req->rc;
-	if (!rpl_context->rc) {
-		err = -ENOMEM;
-		goto err_free2;
-	}
 
 	/*
 	 * Post a receive buffer for this request. We need to ensure
@@ -455,29 +464,35 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 	 * outstanding request, so we must keep a count to avoid
 	 * overflowing the RQ.
 	 */
-	if (atomic_inc_return(&rdma->rq_count) <= rdma->rq_depth) {
-		err = post_recv(client, rpl_context);
-		if (err)
-			goto err_free1;
-	} else
-		atomic_dec(&rdma->rq_count);
+	if (down_interruptible(&rdma->rq_sem)) {
+		err = -EINTR;
+		goto recv_error;
+	}
 
+	err = post_recv(client, rpl_context);
+	if (err) {
+		p9_debug(P9_DEBUG_FCALL, "POST RECV failed\n");
+		goto recv_error;
+	}
 	/* remove posted receive buffer from request structure */
 	req->rc = NULL;
 
+dont_need_post_recv:
 	/* Post the request */
 	c = kmalloc(sizeof *c, GFP_NOFS);
 	if (!c) {
 		err = -ENOMEM;
-		goto err_free1;
+		goto send_error;
 	}
 	c->req = req;
 
 	c->busa = ib_dma_map_single(rdma->cm_id->device,
 				    c->req->tc->sdata, c->req->tc->size,
 				    DMA_TO_DEVICE);
-	if (ib_dma_mapping_error(rdma->cm_id->device, c->busa))
-		goto error;
+	if (ib_dma_mapping_error(rdma->cm_id->device, c->busa)) {
+		err = -EIO;
+		goto send_error;
+	}
 
 	sge.addr = c->busa;
 	sge.length = c->req->tc->size;
@@ -491,22 +506,32 @@ static int rdma_request(struct p9_client *client, struct p9_req_t *req)
 	wr.sg_list = &sge;
 	wr.num_sge = 1;
 
-	if (down_interruptible(&rdma->sq_sem))
-		goto error;
+	if (down_interruptible(&rdma->sq_sem)) {
+		err = -EINTR;
+		goto send_error;
+	}
 
-	return ib_post_send(rdma->qp, &wr, &bad_wr);
+	err = ib_post_send(rdma->qp, &wr, &bad_wr);
+	if (err)
+		goto send_error;
 
- error:
+	/* Success */
+	return 0;
+
+ /* Handle errors that happened during or while preparing the send: */
+ send_error:
 	kfree(c);
-	kfree(rpl_context->rc);
-	kfree(rpl_context);
-	p9_debug(P9_DEBUG_ERROR, "EIO\n");
-	return -EIO;
- err_free1:
-	kfree(rpl_context->rc);
- err_free2:
+	p9_debug(P9_DEBUG_ERROR, "Error %d in rdma_request()\n", err);
+
+	/* Ach.
+	 *  We did recv_post(), but not send. We have one recv_post in excess.
+	 */
+	atomic_inc(&rdma->excess_rc);
+	return err;
+
+ /* Handle errors that happened during or while preparing post_recv(): */
+ recv_error:
 	kfree(rpl_context);
- err_close:
 	spin_lock_irqsave(&rdma->req_lock, flags);
 	if (rdma->state < P9_RDMA_CLOSING) {
 		rdma->state = P9_RDMA_CLOSING;
@@ -551,7 +576,8 @@ static struct p9_trans_rdma *alloc_rdma(struct p9_rdma_opts *opts)
 	spin_lock_init(&rdma->req_lock);
 	init_completion(&rdma->cm_done);
 	sema_init(&rdma->sq_sem, rdma->sq_depth);
-	atomic_set(&rdma->rq_count, 0);
+	sema_init(&rdma->rq_sem, rdma->rq_depth);
+	atomic_set(&rdma->excess_rc, 0);
 
 	return rdma;
 }
@@ -562,6 +588,17 @@ static int rdma_cancel(struct p9_client *client, struct p9_req_t *req)
 	return 1;
 }
 
+/* A request has been fully flushed without a reply.
+ * That means we have posted one buffer in excess.
+ */
+static int rdma_cancelled(struct p9_client *client, struct p9_req_t *req)
+{
+	struct p9_trans_rdma *rdma = client->trans;
+
+	atomic_inc(&rdma->excess_rc);
+	return 0;
+}
+
 /**
  * trans_create_rdma - Transport method for creating atransport instance
  * @client: client instance