3 files changed, 238 insertions, 127 deletions

diff --git a/include/net/9p/client.h b/include/net/9p/client.h
index c35fb548e7cf..140cf1d58452 100644
--- a/include/net/9p/client.h
+++ b/include/net/9p/client.h
@@ -26,6 +26,9 @@
 #ifndef NET_9P_CLIENT_H
 #define NET_9P_CLIENT_H
 
+/* Number of requests per row */
+#define P9_ROW_MAXTAG 255
+
 /**
  * enum p9_trans_status - different states of underlying transports
  * @Connected: transport is connected and healthy
@@ -43,6 +46,62 @@ enum p9_trans_status {
 };
 
 /**
+ * enum p9_req_status_t - virtio request status
+ * @REQ_STATUS_IDLE: request slot unused
+ * @REQ_STATUS_ALLOC: request has been allocated but not sent
+ * @REQ_STATUS_SENT: request sent to server
+ * @REQ_STATUS_FLSH: a flush has been sent for this request
+ * @REQ_STATUS_RCVD: response received from server
+ * @REQ_STATUS_FLSHD: request has been flushed
+ * @REQ_STATUS_ERR: request encountered an error on the client side
+ *
+ * The @REQ_STATUS_IDLE state is used to mark a request slot as unused
+ * but use is actually tracked by the idpool structure which handles tag
+ * id allocation.
+ *
+ */
+
+enum p9_req_status_t {
+        REQ_STATUS_IDLE,
+        REQ_STATUS_ALLOC,
+        REQ_STATUS_SENT,
+        REQ_STATUS_FLSH,
+        REQ_STATUS_RCVD,
+        REQ_STATUS_FLSHD,
+        REQ_STATUS_ERROR,
+};
+
+/**
+ * struct p9_req_t - request slots
+ * @status: status of this request slot
+ * @t_err: transport error
+ * @wq: wait_queue for the client to block on for this request
+ * @tc: the request fcall structure
+ * @rc: the response fcall structure
+ * @aux: transport specific data (provided for trans_fd migration)
+ *
+ * Transport use an array to track outstanding requests
+ * instead of a list.  While this may incurr overhead during initial
+ * allocation or expansion, it makes request lookup much easier as the
+ * tag id is a index into an array.  (We use tag+1 so that we can accomodate
+ * the -1 tag for the T_VERSION request).
+ * This also has the nice effect of only having to allocate wait_queues
+ * once, instead of constantly allocating and freeing them.  Its possible
+ * other resources could benefit from this scheme as well.
+ *
+ */
+
+struct p9_req_t {
+        int status;
+        int t_err;
+        wait_queue_head_t *wq;
+        struct p9_fcall *tc;
+        struct p9_fcall *rc;
+        u16 flush_tag;
+        void *aux;
+};
+
+/**
  * struct p9_client - per client instance state
  * @lock: protect @fidlist
  * @msize: maximum data size negotiated by protocol
@@ -52,9 +111,20 @@ enum p9_trans_status {
  * @conn: connection state information used by trans_fd
  * @fidpool: fid handle accounting for session
  * @fidlist: List of active fid handles
+ * @tagpool - transaction id accounting for session
+ * @reqs - 2D array of requests
+ * @max_tag - current maximum tag id allocated
  *
  * The client structure is used to keep track of various per-client
  * state that has been instantiated.
+ * In order to minimize per-transaction overhead we use a
+ * simple array to lookup requests instead of a hash table
+ * or linked list.  In order to support larger number of
+ * transactions, we make this a 2D array, allocating new rows
+ * when we need to grow the total number of the transactions.
+ *
+ * Each row is 256 requests and we'll support up to 256 rows for
+ * a total of 64k concurrent requests per session.
  *
  * Bugs: duplicated data and potentially unnecessary elements.
  */
@@ -70,6 +140,10 @@ struct p9_client {
 
         struct p9_idpool *fidpool;
         struct list_head fidlist;
+
+        struct p9_idpool *tagpool;
+        struct p9_req_t *reqs[P9_ROW_MAXTAG];
+        int max_tag;
 };
 
 /**
@@ -131,4 +205,7 @@ struct p9_stat *p9_client_stat(struct p9_fid *fid);
 int p9_client_wstat(struct p9_fid *fid, struct p9_wstat *wst);
 struct p9_stat *p9_client_dirread(struct p9_fid *fid, u64 offset);
 
+struct p9_req_t *p9_tag_alloc(struct p9_client *, u16);
+struct p9_req_t *p9_tag_lookup(struct p9_client *, u16);
+
 #endif /* NET_9P_CLIENT_H */
diff --git a/net/9p/client.c b/net/9p/client.c
index 712d4f336adc..867031934f75 100644
--- a/net/9p/client.c
+++ b/net/9p/client.c
@@ -120,6 +120,154 @@ static int parse_opts(char *opts, struct p9_client *clnt)
         return ret;
 }
 
+/**
+ * p9_tag_alloc - lookup/allocate a request by tag
+ * @c: client session to lookup tag within
+ * @tag: numeric id for transaction
+ *
+ * this is a simple array lookup, but will grow the
+ * request_slots as necessary to accomodate transaction
+ * ids which did not previously have a slot.
+ *
+ * this code relies on the client spinlock to manage locks, its
+ * possible we should switch to something else, but I'd rather
+ * stick with something low-overhead for the common case.
+ *
+ */
+
+struct p9_req_t *p9_tag_alloc(struct p9_client *c, u16 tag)
+{
+        unsigned long flags;
+        int row, col;
+
+        /* This looks up the original request by tag so we know which
+         * buffer to read the data into */
+        tag++;
+
+        if (tag >= c->max_tag) {
+                spin_lock_irqsave(&c->lock, flags);
+                /* check again since original check was outside of lock */
+                while (tag >= c->max_tag) {
+                        row = (tag / P9_ROW_MAXTAG);
+                        c->reqs[row] = kcalloc(P9_ROW_MAXTAG,
+                                        sizeof(struct p9_req_t), GFP_ATOMIC);
+
+                        if (!c->reqs[row]) {
+                                printk(KERN_ERR "Couldn't grow tag array\n");
+                                BUG();
+                        }
+                        for (col = 0; col < P9_ROW_MAXTAG; col++) {
+                                c->reqs[row][col].status = REQ_STATUS_IDLE;
+                                c->reqs[row][col].flush_tag = P9_NOTAG;
+                                c->reqs[row][col].wq = kmalloc(
+                                        sizeof(wait_queue_head_t), GFP_ATOMIC);
+                                if (!c->reqs[row][col].wq) {
+                                        printk(KERN_ERR
+                                                "Couldn't grow tag array\n");
+                                        BUG();
+                                }
+                                init_waitqueue_head(c->reqs[row][col].wq);
+                        }
+                        c->max_tag += P9_ROW_MAXTAG;
+                }
+                spin_unlock_irqrestore(&c->lock, flags);
+        }
+        row = tag / P9_ROW_MAXTAG;
+        col = tag % P9_ROW_MAXTAG;
+
+        c->reqs[row][col].status = REQ_STATUS_ALLOC;
+        c->reqs[row][col].flush_tag = P9_NOTAG;
+
+        return &c->reqs[row][col];
+}
+EXPORT_SYMBOL(p9_tag_alloc);
+
+/**
+ * p9_tag_lookup - lookup a request by tag
+ * @c: client session to lookup tag within
+ * @tag: numeric id for transaction
+ *
+ */
+
+struct p9_req_t *p9_tag_lookup(struct p9_client *c, u16 tag)
+{
+        int row, col;
+
+        /* This looks up the original request by tag so we know which
+         * buffer to read the data into */
+        tag++;
+
+        BUG_ON(tag >= c->max_tag);
+
+        row = tag / P9_ROW_MAXTAG;
+        col = tag % P9_ROW_MAXTAG;
+
+        return &c->reqs[row][col];
+}
+EXPORT_SYMBOL(p9_tag_lookup);
+
+/**
+ * p9_tag_init - setup tags structure and contents
+ * @tags: tags structure from the client struct
+ *
+ * This initializes the tags structure for each client instance.
+ *
+ */
+
+static int p9_tag_init(struct p9_client *c)
+{
+        int err = 0;
+
+        c->tagpool = p9_idpool_create();
+        if (IS_ERR(c->tagpool)) {
+                err = PTR_ERR(c->tagpool);
+                c->tagpool = NULL;
+                goto error;
+        }
+
+        p9_idpool_get(c->tagpool); /* reserve tag 0 */
+
+        c->max_tag = 0;
+error:
+        return err;
+}
+
+/**
+ * p9_tag_cleanup - cleans up tags structure and reclaims resources
+ * @tags: tags structure from the client struct
+ *
+ * This frees resources associated with the tags structure
+ *
+ */
+static void p9_tag_cleanup(struct p9_client *c)
+{
+        int row, col;
+
+        /* check to insure all requests are idle */
+        for (row = 0; row < (c->max_tag/P9_ROW_MAXTAG); row++) {
+                for (col = 0; col < P9_ROW_MAXTAG; col++) {
+                        if (c->reqs[row][col].status != REQ_STATUS_IDLE) {
+                                P9_DPRINTK(P9_DEBUG_MUX,
+                                  "Attempting to cleanup non-free tag %d,%d\n",
+                                  row, col);
+                                /* TODO: delay execution of cleanup */
+                                return;
+                        }
+                }
+        }
+
+        if (c->tagpool)
+                p9_idpool_destroy(c->tagpool);
+
+        /* free requests associated with tags */
+        for (row = 0; row < (c->max_tag/P9_ROW_MAXTAG); row++) {
+                for (col = 0; col < P9_ROW_MAXTAG; col++)
+                        kfree(c->reqs[row][col].wq);
+                kfree(c->reqs[row]);
+        }
+        c->max_tag = 0;
+}
+
 static struct p9_fid *p9_fid_create(struct p9_client *clnt)
 {
         int err;
@@ -209,6 +357,8 @@ struct p9_client *p9_client_create(const char *dev_name, char *options)
                 goto error;
         }
 
+        p9_tag_init(clnt);
+
         err = parse_opts(options, clnt);
         if (err < 0)
                 goto error;
@@ -285,6 +435,8 @@ void p9_client_destroy(struct p9_client *clnt)
         if (clnt->fidpool)
                 p9_idpool_destroy(clnt->fidpool);
 
+        p9_tag_cleanup(clnt);
+
         kfree(clnt);
 }
 EXPORT_SYMBOL(p9_client_destroy);
diff --git a/net/9p/trans_virtio.c b/net/9p/trans_virtio.c
index 72493f04a76d..36bce45e4e44 100644
--- a/net/9p/trans_virtio.c
+++ b/net/9p/trans_virtio.c
@@ -1,12 +1,10 @@
 /*
- * The Guest 9p transport driver
+ * The Virtio 9p transport driver
  *
  * This is a block based transport driver based on the lguest block driver
  * code.
  *
- */
-/*
- *  Copyright (C) 2007 Eric Van Hensbergen, IBM Corporation
+ *  Copyright (C) 2007, 2008 Eric Van Hensbergen, IBM Corporation
  *
  *  Based on virtio console driver
  *  Copyright (C) 2006, 2007 Rusty Russell, IBM Corporation
@@ -54,49 +52,6 @@ static DEFINE_MUTEX(virtio_9p_lock);
 /* global which tracks highest initialized channel */
 static int chan_index;
 
-#define P9_INIT_MAXTAG  16
-
-/**
- * enum p9_req_status_t - virtio request status
- * @REQ_STATUS_IDLE: request slot unused
- * @REQ_STATUS_SENT: request sent to server
- * @REQ_STATUS_RCVD: response received from server
- * @REQ_STATUS_FLSH: request has been flushed
- *
- * The @REQ_STATUS_IDLE state is used to mark a request slot as unused
- * but use is actually tracked by the idpool structure which handles tag
- * id allocation.
- *
- */
-
-enum p9_req_status_t {
-        REQ_STATUS_IDLE,
-        REQ_STATUS_SENT,
-        REQ_STATUS_RCVD,
-        REQ_STATUS_FLSH,
-};
-
-/**
- * struct p9_req_t - virtio request slots
- * @status: status of this request slot
- * @wq: wait_queue for the client to block on for this request
- *
- * The virtio transport uses an array to track outstanding requests
- * instead of a list.  While this may incurr overhead during initial
- * allocation or expansion, it makes request lookup much easier as the
- * tag id is a index into an array.  (We use tag+1 so that we can accomodate
- * the -1 tag for the T_VERSION request).
- * This also has the nice effect of only having to allocate wait_queues
- * once, instead of constantly allocating and freeing them.  Its possible
- * other resources could benefit from this scheme as well.
- *
- */
-
-struct p9_req_t {
-        int status;
-        wait_queue_head_t *wq;
-};
-
 /**
  * struct virtio_chan - per-instance transport information
  * @initialized: whether the channel is initialized
@@ -121,67 +76,14 @@ static struct virtio_chan {
 
         spinlock_t lock;
 
+        struct p9_client *client;
         struct virtio_device *vdev;
         struct virtqueue *vq;
 
-        struct p9_idpool *tagpool;
-        struct p9_req_t *reqs;
-        int max_tag;
-
         /* Scatterlist: can be too big for stack. */
         struct scatterlist sg[VIRTQUEUE_NUM];
 } channels[MAX_9P_CHAN];
 
-/**
- * p9_lookup_tag - Lookup requests by tag
- * @c: virtio channel to lookup tag within
- * @tag: numeric id for transaction
- *
- * this is a simple array lookup, but will grow the
- * request_slots as necessary to accomodate transaction
- * ids which did not previously have a slot.
- *
- * Bugs: there is currently no upper limit on request slots set
- * here, but that should be constrained by the id accounting.
- */
-
-static struct p9_req_t *p9_lookup_tag(struct virtio_chan *c, u16 tag)
-{
-        /* This looks up the original request by tag so we know which
-         * buffer to read the data into */
-        tag++;
-
-        while (tag >= c->max_tag) {
-                int old_max = c->max_tag;
-                int count;
-
-                if (c->max_tag)
-                        c->max_tag *= 2;
-                else
-                        c->max_tag = P9_INIT_MAXTAG;
-
-                c->reqs = krealloc(c->reqs, sizeof(struct p9_req_t)*c->max_tag,
-                                                                GFP_ATOMIC);
-                if (!c->reqs) {
-                        printk(KERN_ERR "Couldn't grow tag array\n");
-                        BUG();
-                }
-                for (count = old_max; count < c->max_tag; count++) {
-                        c->reqs[count].status = REQ_STATUS_IDLE;
-                        c->reqs[count].wq = kmalloc(sizeof(wait_queue_head_t),
-                                                                GFP_ATOMIC);
-                        if (!c->reqs[count].wq) {
-                                printk(KERN_ERR "Couldn't grow tag array\n");
-                                BUG();
-                        }
-                        init_waitqueue_head(c->reqs[count].wq);
-                }
-        }
-
-        return &c->reqs[tag];
-}
-
-
 /* How many bytes left in this page. */
 static unsigned int rest_of_page(void *data)
 {
@@ -200,22 +102,10 @@ static unsigned int rest_of_page(void *data)
 static void p9_virtio_close(struct p9_client *client)
 {
         struct virtio_chan *chan = client->trans;
-        int count;
-        unsigned long flags;
-
-        spin_lock_irqsave(&chan->lock, flags);
-        p9_idpool_destroy(chan->tagpool);
-        for (count = 0; count < chan->max_tag; count++)
-                kfree(chan->reqs[count].wq);
-        kfree(chan->reqs);
-        chan->max_tag = 0;
-        spin_unlock_irqrestore(&chan->lock, flags);
 
         mutex_lock(&virtio_9p_lock);
         chan->inuse = false;
         mutex_unlock(&virtio_9p_lock);
-
-        client->trans = NULL;
 }
 
 /**
@@ -241,7 +131,7 @@ static void req_done(struct virtqueue *vq)
 
         spin_lock_irqsave(&chan->lock, flags);
         while ((rc = chan->vq->vq_ops->get_buf(chan->vq, &len)) != NULL) {
-                req = p9_lookup_tag(chan, rc->tag);
+                req = p9_tag_lookup(chan->client, rc->tag);
                 req->status = REQ_STATUS_RCVD;
                 wake_up(req->wq);
         }
@@ -311,13 +201,13 @@ p9_virtio_rpc(struct p9_client *c, struct p9_fcall *tc, struct p9_fcall **rc)
 
         n = P9_NOTAG;
         if (tc->id != P9_TVERSION) {
-                n = p9_idpool_get(chan->tagpool);
+                n = p9_idpool_get(c->tagpool);
                 if (n < 0)
                         return -ENOMEM;
         }
 
         spin_lock_irqsave(&chan->lock, flags);
-        req = p9_lookup_tag(chan, n);
+        req = p9_tag_alloc(c, n);
         spin_unlock_irqrestore(&chan->lock, flags);
 
         p9_set_tag(tc, n);
@@ -357,8 +247,8 @@ p9_virtio_rpc(struct p9_client *c, struct p9_fcall *tc, struct p9_fcall **rc)
         }
 #endif
 
-        if (n != P9_NOTAG && p9_idpool_check(n, chan->tagpool))
-                p9_idpool_put(n, chan->tagpool);
+        if (n != P9_NOTAG && p9_idpool_check(n, c->tagpool))
+                p9_idpool_put(n, c->tagpool);
 
         req->status = REQ_STATUS_IDLE;
 
@@ -463,16 +353,8 @@ p9_virtio_create(struct p9_client *client, const char *devname, char *args)
                 return -ENODEV;
         }
 
-        chan->tagpool = p9_idpool_create();
-        if (IS_ERR(chan->tagpool)) {
-                printk(KERN_ERR "9p: couldn't allocate tagpool\n");
-                return -ENOMEM;
-        }
-        p9_idpool_get(chan->tagpool); /* reserve tag 0 */
-        chan->max_tag = 0;
-        chan->reqs = NULL;
-
         client->trans = (void *)chan;
+        chan->client = client;
 
         return 0;
 }