From f5545d24b8aa9fccd8071203e83bc9f4b26e17a6 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Sun, 2 Apr 2006 14:39:19 -0700 Subject: IPoIB: Always build debugging code unless CONFIG_EMBEDDED=y Don't allow CONFIG_INFINIBAND_IPOIB_DEBUG to be disabled unless CONFIG_EMBEDDED is selected. We want users (and especially distros) to have this turned on unless they really need to save space, because by the time we want debugging output, it's usually too late to rebuild a kernel. The debugging output can be controlled at runtime via the debug_level module parameter in sysfs. Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/Kconfig | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/Kconfig b/drivers/infiniband/ulp/ipoib/Kconfig index 8d2e04cac68..13d6d01c72c 100644 --- a/drivers/infiniband/ulp/ipoib/Kconfig +++ b/drivers/infiniband/ulp/ipoib/Kconfig @@ -10,8 +10,9 @@ config INFINIBAND_IPOIB group: . config INFINIBAND_IPOIB_DEBUG - bool "IP-over-InfiniBand debugging" + bool "IP-over-InfiniBand debugging" if EMBEDDED depends on INFINIBAND_IPOIB + default y ---help--- This option causes debugging code to be compiled into the IPoIB driver. The output can be turned on via the -- cgit v1.2.3 From ce1823f0323be9f38bbe0df229a5bba025404923 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 3 Apr 2006 09:31:04 -0700 Subject: IB/srp: Fix memory leak in options parsing Fix memory leak if parsing destination GID fails. Coverity bug 1042 Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/srp/ib_srp.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index fd8a95a9c5d..5f2b3f6e4c4 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -1434,6 +1434,7 @@ static int srp_parse_options(const char *buf, struct srp_target_port *target) p = match_strdup(args); if (strlen(p) != 32) { printk(KERN_WARNING PFX "bad dest GID parameter '%s'\n", p); + kfree(p); goto out; } -- cgit v1.2.3 From d2e0655ede1d91c3a586455d03a4a2d57e659830 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Tue, 4 Apr 2006 19:59:40 +0300 Subject: IPoIB: Consolidate private neighbour data handling Consolidate IPoIB's private neighbour data handling into ipoib_neigh_alloc() and ipoib_neigh_free(). This will make it easier to keep track of the neighbour structures that IPoIB is handling, and is a nice cleanup of the code: add/remove: 2/1 grow/shrink: 1/8 up/down: 100/-178 (-78) function old new delta ipoib_neigh_alloc - 61 +61 ipoib_neigh_free - 36 +36 ipoib_mcast_join_finish 1288 1291 +3 path_rec_completion 575 573 -2 ipoib_mcast_join_task 664 660 -4 ipoib_neigh_destructor 101 92 -9 ipoib_neigh_setup_dev 14 3 -11 ipoib_neigh_setup 17 - -17 path_free 238 215 -23 ipoib_mcast_free 329 306 -23 ipoib_mcast_send 718 684 -34 neigh_add_path 705 650 -55 Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 3 ++ drivers/infiniband/ulp/ipoib/ipoib_main.c | 41 ++++++++++++++++++-------- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 7 ++--- 3 files changed, 34 insertions(+), 17 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index b640107fb73..374109df730 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -230,6 +230,9 @@ static inline struct ipoib_neigh **to_ipoib_neigh(struct neighbour *neigh) INFINIBAND_ALEN, sizeof(void *)); } +struct ipoib_neigh *ipoib_neigh_alloc(struct neighbour *neigh); +void ipoib_neigh_free(struct ipoib_neigh *neigh); + extern struct workqueue_struct *ipoib_workqueue; /* functions */ diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 9b0bd7c746c..8f6607bf426 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -252,8 +252,8 @@ static void path_free(struct net_device *dev, struct ipoib_path *path) */ if (neigh->ah) ipoib_put_ah(neigh->ah); - *to_ipoib_neigh(neigh->neighbour) = NULL; - kfree(neigh); + + ipoib_neigh_free(neigh); } spin_unlock_irqrestore(&priv->lock, flags); @@ -481,7 +481,7 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev) struct ipoib_path *path; struct ipoib_neigh *neigh; - neigh = kmalloc(sizeof *neigh, GFP_ATOMIC); + neigh = ipoib_neigh_alloc(skb->dst->neighbour); if (!neigh) { ++priv->stats.tx_dropped; dev_kfree_skb_any(skb); @@ -489,8 +489,6 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev) } skb_queue_head_init(&neigh->queue); - neigh->neighbour = skb->dst->neighbour; - *to_ipoib_neigh(skb->dst->neighbour) = neigh; /* * We can only be called from ipoib_start_xmit, so we're @@ -503,7 +501,7 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev) path = path_rec_create(dev, (union ib_gid *) (skb->dst->neighbour->ha + 4)); if (!path) - goto err; + goto err_path; __path_add(dev, path); } @@ -521,17 +519,17 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev) __skb_queue_tail(&neigh->queue, skb); if (!path->query && path_rec_start(dev, path)) - goto err; + goto err_list; } spin_unlock(&priv->lock); return; -err: - *to_ipoib_neigh(skb->dst->neighbour) = NULL; +err_list: list_del(&neigh->list); - kfree(neigh); +err_path: + ipoib_neigh_free(neigh); ++priv->stats.tx_dropped; dev_kfree_skb_any(skb); @@ -763,8 +761,7 @@ static void ipoib_neigh_destructor(struct neighbour *n) if (neigh->ah) ah = neigh->ah; list_del(&neigh->list); - *to_ipoib_neigh(n) = NULL; - kfree(neigh); + ipoib_neigh_free(neigh); } spin_unlock_irqrestore(&priv->lock, flags); @@ -773,6 +770,26 @@ static void ipoib_neigh_destructor(struct neighbour *n) ipoib_put_ah(ah); } +struct ipoib_neigh *ipoib_neigh_alloc(struct neighbour *neighbour) +{ + struct ipoib_neigh *neigh; + + neigh = kmalloc(sizeof *neigh, GFP_ATOMIC); + if (!neigh) + return NULL; + + neigh->neighbour = neighbour; + *to_ipoib_neigh(neighbour) = neigh; + + return neigh; +} + +void ipoib_neigh_free(struct ipoib_neigh *neigh) +{ + *to_ipoib_neigh(neigh->neighbour) = NULL; + kfree(neigh); +} + static int ipoib_neigh_setup_dev(struct net_device *dev, struct neigh_parms *parms) { parms->neigh_destructor = ipoib_neigh_destructor; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 93c462eaf4f..a8395ef06c1 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -114,8 +114,7 @@ static void ipoib_mcast_free(struct ipoib_mcast *mcast) */ if (neigh->ah) ipoib_put_ah(neigh->ah); - *to_ipoib_neigh(neigh->neighbour) = NULL; - kfree(neigh); + ipoib_neigh_free(neigh); } spin_unlock_irqrestore(&priv->lock, flags); @@ -772,13 +771,11 @@ out: if (skb->dst && skb->dst->neighbour && !*to_ipoib_neigh(skb->dst->neighbour)) { - struct ipoib_neigh *neigh = kmalloc(sizeof *neigh, GFP_ATOMIC); + struct ipoib_neigh *neigh = ipoib_neigh_alloc(skb->dst->neighbour); if (neigh) { kref_get(&mcast->ah->ref); neigh->ah = mcast->ah; - neigh->neighbour = skb->dst->neighbour; - *to_ipoib_neigh(skb->dst->neighbour) = neigh; list_add_tail(&neigh->list, &mcast->neigh_list); } } -- cgit v1.2.3 From bf6a9e31cfa768ce0a8e18474b3ca808641d9243 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Mon, 10 Apr 2006 09:43:47 -0700 Subject: IB: simplify static rate encoding Push translation of static rate to HCA format into low-level drivers, where it belongs. For static rate encoding, use encoding of rate field from IB standard PathRecord, with addition of value 0, for backwards compatibility with current usage. The changes are: - Add enum ib_rate to midlayer includes. - Get rid of static rate translation in IPoIB; just use static rate directly from Path and MulticastGroup records. - Update mthca driver to translate absolute static rate into the format used by hardware. This also fixes mthca's static rate handling for HCAs that are capable of 4X DDR. Signed-off-by: Jack Morgenstein Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_fs.c | 2 +- drivers/infiniband/ulp/ipoib/ipoib_main.c | 11 ++--------- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 10 +--------- 3 files changed, 4 insertions(+), 19 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_fs.c b/drivers/infiniband/ulp/ipoib/ipoib_fs.c index 685258e3403..5dde380e8db 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_fs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_fs.c @@ -213,7 +213,7 @@ static int ipoib_path_seq_show(struct seq_file *file, void *iter_ptr) gid_buf, path.pathrec.dlid ? "yes" : "no"); if (path.pathrec.dlid) { - rate = ib_sa_rate_enum_to_int(path.pathrec.rate) * 25; + rate = ib_rate_to_mult(path.pathrec.rate) * 25; seq_printf(file, " DLID: 0x%04x\n" diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 8f6607bf426..9cb9e430aaa 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -373,16 +373,9 @@ static void path_rec_completion(int status, struct ib_ah_attr av = { .dlid = be16_to_cpu(pathrec->dlid), .sl = pathrec->sl, - .port_num = priv->port + .port_num = priv->port, + .static_rate = pathrec->rate }; - int path_rate = ib_sa_rate_enum_to_int(pathrec->rate); - - if (path_rate > 0 && priv->local_rate > path_rate) - av.static_rate = (priv->local_rate - 1) / path_rate; - - ipoib_dbg(priv, "static_rate %d for local port %dX, path %dX\n", - av.static_rate, priv->local_rate, - ib_sa_rate_enum_to_int(pathrec->rate)); ah = ipoib_create_ah(dev, priv->pd, &av); } diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index a8395ef06c1..07b9826b519 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -250,6 +250,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, .port_num = priv->port, .sl = mcast->mcmember.sl, .ah_flags = IB_AH_GRH, + .static_rate = mcast->mcmember.rate, .grh = { .flow_label = be32_to_cpu(mcast->mcmember.flow_label), .hop_limit = mcast->mcmember.hop_limit, @@ -257,17 +258,8 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast, .traffic_class = mcast->mcmember.traffic_class } }; - int path_rate = ib_sa_rate_enum_to_int(mcast->mcmember.rate); - av.grh.dgid = mcast->mcmember.mgid; - if (path_rate > 0 && priv->local_rate > path_rate) - av.static_rate = (priv->local_rate - 1) / path_rate; - - ipoib_dbg_mcast(priv, "static_rate %d for local port %dX, mcmember %dX\n", - av.static_rate, priv->local_rate, - ib_sa_rate_enum_to_int(mcast->mcmember.rate)); - ah = ipoib_create_ah(dev, priv->pd, &av); if (!ah) { ipoib_warn(priv, "ib_address_create failed\n"); -- cgit v1.2.3 From f2de3b06126ddb07d0e4617225d74dce0855add3 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Wed, 5 Apr 2006 14:59:40 +0300 Subject: IPoIB: Wait for join to finish before freeing mcast struct ipoib_mcast_restart_task() might free an mcast object while a join request is still outstanding, leading to an oops when the query completes. Fix this by waiting for query to complete, similar to what ipoib_stop_thread() is doing. The wait for mcast completion code is consolidated in wait_for_mcast_join(). Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_multicast.c | 41 +++++++++++++------------- 1 file changed, 20 insertions(+), 21 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c index 07b9826b519..1dae4b23825 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_multicast.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_multicast.c @@ -609,6 +609,22 @@ int ipoib_mcast_start_thread(struct net_device *dev) return 0; } +static void wait_for_mcast_join(struct ipoib_dev_priv *priv, + struct ipoib_mcast *mcast) +{ + spin_lock_irq(&priv->lock); + if (mcast && mcast->query) { + ib_sa_cancel_query(mcast->query_id, mcast->query); + mcast->query = NULL; + spin_unlock_irq(&priv->lock); + ipoib_dbg_mcast(priv, "waiting for MGID " IPOIB_GID_FMT "\n", + IPOIB_GID_ARG(mcast->mcmember.mgid)); + wait_for_completion(&mcast->done); + } + else + spin_unlock_irq(&priv->lock); +} + int ipoib_mcast_stop_thread(struct net_device *dev, int flush) { struct ipoib_dev_priv *priv = netdev_priv(dev); @@ -628,28 +644,10 @@ int ipoib_mcast_stop_thread(struct net_device *dev, int flush) if (flush) flush_workqueue(ipoib_workqueue); - spin_lock_irq(&priv->lock); - if (priv->broadcast && priv->broadcast->query) { - ib_sa_cancel_query(priv->broadcast->query_id, priv->broadcast->query); - priv->broadcast->query = NULL; - spin_unlock_irq(&priv->lock); - ipoib_dbg_mcast(priv, "waiting for bcast\n"); - wait_for_completion(&priv->broadcast->done); - } else - spin_unlock_irq(&priv->lock); + wait_for_mcast_join(priv, priv->broadcast); - list_for_each_entry(mcast, &priv->multicast_list, list) { - spin_lock_irq(&priv->lock); - if (mcast->query) { - ib_sa_cancel_query(mcast->query_id, mcast->query); - mcast->query = NULL; - spin_unlock_irq(&priv->lock); - ipoib_dbg_mcast(priv, "waiting for MGID " IPOIB_GID_FMT "\n", - IPOIB_GID_ARG(mcast->mcmember.mgid)); - wait_for_completion(&mcast->done); - } else - spin_unlock_irq(&priv->lock); - } + list_for_each_entry(mcast, &priv->multicast_list, list) + wait_for_mcast_join(priv, mcast); return 0; } @@ -902,6 +900,7 @@ void ipoib_mcast_restart_task(void *dev_ptr) /* We have to cancel outside of the spinlock */ list_for_each_entry_safe(mcast, tmcast, &remove_list, list) { + wait_for_mcast_join(priv, mcast); ipoib_mcast_leave(mcast->dev, mcast); ipoib_mcast_free(mcast); } -- cgit v1.2.3 From 0f4852513fb07405ce88da40d8c497060561246e Mon Sep 17 00:00:00 2001 From: Shirley Ma Date: Mon, 10 Apr 2006 09:43:58 -0700 Subject: IPoIB: Make send and receive queue sizes tunable Make IPoIB's send and receive queue sizes tunable via module parameters ("send_queue_size" and "recv_queue_size"). This allows the queue sizes to be enlarged to fix disastrously bad performance on some platforms and workloads, without bloating memory usage when large queues aren't needed. Signed-off-by: Shirley Ma Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib.h | 4 ++++ drivers/infiniband/ulp/ipoib/ipoib_ib.c | 22 +++++++++++----------- drivers/infiniband/ulp/ipoib/ipoib_main.c | 28 ++++++++++++++++++++++------ drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 6 +++--- 4 files changed, 40 insertions(+), 20 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h index 374109df730..12a1e0572ef 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib.h +++ b/drivers/infiniband/ulp/ipoib/ipoib.h @@ -65,6 +65,8 @@ enum { IPOIB_RX_RING_SIZE = 128, IPOIB_TX_RING_SIZE = 64, + IPOIB_MAX_QUEUE_SIZE = 8192, + IPOIB_MIN_QUEUE_SIZE = 2, IPOIB_NUM_WC = 4, @@ -332,6 +334,8 @@ static inline void ipoib_unregister_debugfs(void) { } #define ipoib_warn(priv, format, arg...) \ ipoib_printk(KERN_WARNING, priv, format , ## arg) +extern int ipoib_sendq_size; +extern int ipoib_recvq_size; #ifdef CONFIG_INFINIBAND_IPOIB_DEBUG extern int ipoib_debug_level; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index ed65202878d..a54da42849a 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -161,7 +161,7 @@ static int ipoib_ib_post_receives(struct net_device *dev) struct ipoib_dev_priv *priv = netdev_priv(dev); int i; - for (i = 0; i < IPOIB_RX_RING_SIZE; ++i) { + for (i = 0; i < ipoib_recvq_size; ++i) { if (ipoib_alloc_rx_skb(dev, i)) { ipoib_warn(priv, "failed to allocate receive buffer %d\n", i); return -ENOMEM; @@ -187,7 +187,7 @@ static void ipoib_ib_handle_wc(struct net_device *dev, if (wr_id & IPOIB_OP_RECV) { wr_id &= ~IPOIB_OP_RECV; - if (wr_id < IPOIB_RX_RING_SIZE) { + if (wr_id < ipoib_recvq_size) { struct sk_buff *skb = priv->rx_ring[wr_id].skb; dma_addr_t addr = priv->rx_ring[wr_id].mapping; @@ -252,9 +252,9 @@ static void ipoib_ib_handle_wc(struct net_device *dev, struct ipoib_tx_buf *tx_req; unsigned long flags; - if (wr_id >= IPOIB_TX_RING_SIZE) { + if (wr_id >= ipoib_sendq_size) { ipoib_warn(priv, "completion event with wrid %d (> %d)\n", - wr_id, IPOIB_TX_RING_SIZE); + wr_id, ipoib_sendq_size); return; } @@ -275,7 +275,7 @@ static void ipoib_ib_handle_wc(struct net_device *dev, spin_lock_irqsave(&priv->tx_lock, flags); ++priv->tx_tail; if (netif_queue_stopped(dev) && - priv->tx_head - priv->tx_tail <= IPOIB_TX_RING_SIZE / 2) + priv->tx_head - priv->tx_tail <= ipoib_sendq_size >> 1) netif_wake_queue(dev); spin_unlock_irqrestore(&priv->tx_lock, flags); @@ -344,13 +344,13 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, * means we have to make sure everything is properly recorded and * our state is consistent before we call post_send(). */ - tx_req = &priv->tx_ring[priv->tx_head & (IPOIB_TX_RING_SIZE - 1)]; + tx_req = &priv->tx_ring[priv->tx_head & (ipoib_sendq_size - 1)]; tx_req->skb = skb; addr = dma_map_single(priv->ca->dma_device, skb->data, skb->len, DMA_TO_DEVICE); pci_unmap_addr_set(tx_req, mapping, addr); - if (unlikely(post_send(priv, priv->tx_head & (IPOIB_TX_RING_SIZE - 1), + if (unlikely(post_send(priv, priv->tx_head & (ipoib_sendq_size - 1), address->ah, qpn, addr, skb->len))) { ipoib_warn(priv, "post_send failed\n"); ++priv->stats.tx_errors; @@ -363,7 +363,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb, address->last_send = priv->tx_head; ++priv->tx_head; - if (priv->tx_head - priv->tx_tail == IPOIB_TX_RING_SIZE) { + if (priv->tx_head - priv->tx_tail == ipoib_sendq_size) { ipoib_dbg(priv, "TX ring full, stopping kernel net queue\n"); netif_stop_queue(dev); } @@ -488,7 +488,7 @@ static int recvs_pending(struct net_device *dev) int pending = 0; int i; - for (i = 0; i < IPOIB_RX_RING_SIZE; ++i) + for (i = 0; i < ipoib_recvq_size; ++i) if (priv->rx_ring[i].skb) ++pending; @@ -527,7 +527,7 @@ int ipoib_ib_dev_stop(struct net_device *dev) */ while ((int) priv->tx_tail - (int) priv->tx_head < 0) { tx_req = &priv->tx_ring[priv->tx_tail & - (IPOIB_TX_RING_SIZE - 1)]; + (ipoib_sendq_size - 1)]; dma_unmap_single(priv->ca->dma_device, pci_unmap_addr(tx_req, mapping), tx_req->skb->len, @@ -536,7 +536,7 @@ int ipoib_ib_dev_stop(struct net_device *dev) ++priv->tx_tail; } - for (i = 0; i < IPOIB_RX_RING_SIZE; ++i) + for (i = 0; i < ipoib_recvq_size; ++i) if (priv->rx_ring[i].skb) { dma_unmap_single(priv->ca->dma_device, pci_unmap_addr(&priv->rx_ring[i], diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 9cb9e430aaa..5bf7e263454 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -41,6 +41,7 @@ #include #include #include +#include #include /* For ARPHRD_xxx */ @@ -53,6 +54,14 @@ MODULE_AUTHOR("Roland Dreier"); MODULE_DESCRIPTION("IP-over-InfiniBand net driver"); MODULE_LICENSE("Dual BSD/GPL"); +int ipoib_sendq_size __read_mostly = IPOIB_TX_RING_SIZE; +int ipoib_recvq_size __read_mostly = IPOIB_RX_RING_SIZE; + +module_param_named(send_queue_size, ipoib_sendq_size, int, 0444); +MODULE_PARM_DESC(send_queue_size, "Number of descriptors in send queue"); +module_param_named(recv_queue_size, ipoib_recvq_size, int, 0444); +MODULE_PARM_DESC(recv_queue_size, "Number of descriptors in receive queue"); + #ifdef CONFIG_INFINIBAND_IPOIB_DEBUG int ipoib_debug_level; @@ -795,20 +804,19 @@ int ipoib_dev_init(struct net_device *dev, struct ib_device *ca, int port) struct ipoib_dev_priv *priv = netdev_priv(dev); /* Allocate RX/TX "rings" to hold queued skbs */ - - priv->rx_ring = kzalloc(IPOIB_RX_RING_SIZE * sizeof (struct ipoib_rx_buf), + priv->rx_ring = kzalloc(ipoib_recvq_size * sizeof *priv->rx_ring, GFP_KERNEL); if (!priv->rx_ring) { printk(KERN_WARNING "%s: failed to allocate RX ring (%d entries)\n", - ca->name, IPOIB_RX_RING_SIZE); + ca->name, ipoib_recvq_size); goto out; } - priv->tx_ring = kzalloc(IPOIB_TX_RING_SIZE * sizeof (struct ipoib_tx_buf), + priv->tx_ring = kzalloc(ipoib_sendq_size * sizeof *priv->tx_ring, GFP_KERNEL); if (!priv->tx_ring) { printk(KERN_WARNING "%s: failed to allocate TX ring (%d entries)\n", - ca->name, IPOIB_TX_RING_SIZE); + ca->name, ipoib_sendq_size); goto out_rx_ring_cleanup; } @@ -876,7 +884,7 @@ static void ipoib_setup(struct net_device *dev) dev->hard_header_len = IPOIB_ENCAP_LEN + INFINIBAND_ALEN; dev->addr_len = INFINIBAND_ALEN; dev->type = ARPHRD_INFINIBAND; - dev->tx_queue_len = IPOIB_TX_RING_SIZE * 2; + dev->tx_queue_len = ipoib_sendq_size * 2; dev->features = NETIF_F_VLAN_CHALLENGED | NETIF_F_LLTX; /* MTU will be reset when mcast join happens */ @@ -1128,6 +1136,14 @@ static int __init ipoib_init_module(void) { int ret; + ipoib_recvq_size = roundup_pow_of_two(ipoib_recvq_size); + ipoib_recvq_size = min(ipoib_recvq_size, IPOIB_MAX_QUEUE_SIZE); + ipoib_recvq_size = max(ipoib_recvq_size, IPOIB_MIN_QUEUE_SIZE); + + ipoib_sendq_size = roundup_pow_of_two(ipoib_sendq_size); + ipoib_sendq_size = min(ipoib_sendq_size, IPOIB_MAX_QUEUE_SIZE); + ipoib_sendq_size = max(ipoib_sendq_size, IPOIB_MIN_QUEUE_SIZE); + ret = ipoib_register_debugfs(); if (ret) return ret; diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index 5f0388027b2..1d49d1643c5 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -159,8 +159,8 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) struct ipoib_dev_priv *priv = netdev_priv(dev); struct ib_qp_init_attr init_attr = { .cap = { - .max_send_wr = IPOIB_TX_RING_SIZE, - .max_recv_wr = IPOIB_RX_RING_SIZE, + .max_send_wr = ipoib_sendq_size, + .max_recv_wr = ipoib_recvq_size, .max_send_sge = 1, .max_recv_sge = 1 }, @@ -175,7 +175,7 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca) } priv->cq = ib_create_cq(priv->ca, ipoib_ib_completion, NULL, dev, - IPOIB_TX_RING_SIZE + IPOIB_RX_RING_SIZE + 1); + ipoib_sendq_size + ipoib_recvq_size + 1); if (IS_ERR(priv->cq)) { printk(KERN_WARNING "%s: failed to create CQ\n", ca->name); goto out_free_pd; -- cgit v1.2.3 From a30bb96c6f5aca6513e4dbd94962da03d14b20a9 Mon Sep 17 00:00:00 2001 From: Eli Cohen Date: Wed, 5 Apr 2006 15:59:34 +0300 Subject: IPoIB: Close race in ipoib_flush_paths() ib_sa_cancel_query() must be called with priv->lock held since a completion might arrive and set path->query to NULL. Signed-off-by: Eli Cohen Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_main.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 5bf7e263454..996c6e16a46 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -346,14 +346,15 @@ void ipoib_flush_paths(struct net_device *dev) list_for_each_entry(path, &remove_list, list) rb_erase(&path->rb_node, &priv->path_tree); - spin_unlock_irqrestore(&priv->lock, flags); - list_for_each_entry_safe(path, tp, &remove_list, list) { if (path->query) ib_sa_cancel_query(path->query_id, path->query); + spin_unlock_irqrestore(&priv->lock, flags); wait_for_completion(&path->done); path_free(dev, path); + spin_lock_irqsave(&priv->lock, flags); } + spin_unlock_irqrestore(&priv->lock, flags); } static void path_rec_completion(int status, -- cgit v1.2.3 From f697f74a6b189702474b2fd457e1f9365fa213e3 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 10 Apr 2006 09:43:59 -0700 Subject: IPoIB: Use spin_lock_irq() instead of spin_lock_irqsave() We know ipoib_flush_paths() is called from plain process context with interrupts enabled, since it does wait_for_completion(). So there's no need to use spin_lock_irqsave() -- spin_lock_irq() is fine. Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_main.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 996c6e16a46..cb078a7d0bf 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -336,9 +336,8 @@ void ipoib_flush_paths(struct net_device *dev) struct ipoib_dev_priv *priv = netdev_priv(dev); struct ipoib_path *path, *tp; LIST_HEAD(remove_list); - unsigned long flags; - spin_lock_irqsave(&priv->lock, flags); + spin_lock_irq(&priv->lock); list_splice(&priv->path_list, &remove_list); INIT_LIST_HEAD(&priv->path_list); @@ -349,12 +348,12 @@ void ipoib_flush_paths(struct net_device *dev) list_for_each_entry_safe(path, tp, &remove_list, list) { if (path->query) ib_sa_cancel_query(path->query_id, path->query); - spin_unlock_irqrestore(&priv->lock, flags); + spin_unlock_irq(&priv->lock); wait_for_completion(&path->done); path_free(dev, path); - spin_lock_irqsave(&priv->lock, flags); + spin_lock_irq(&priv->lock); } - spin_unlock_irqrestore(&priv->lock, flags); + spin_unlock_irq(&priv->lock); } static void path_rec_completion(int status, -- cgit v1.2.3