12 files changed, 77 insertions, 32 deletions

diff --git a/drivers/infiniband/hw/amso1100/c2_cq.c b/drivers/infiniband/hw/amso1100/c2_cq.c
index bb17cce3cb59..f5c45b194f53 100644
--- a/drivers/infiniband/hw/amso1100/c2_cq.c
+++ b/drivers/infiniband/hw/amso1100/c2_cq.c
@@ -133,7 +133,7 @@ static inline int c2_poll_one(struct c2_dev *c2dev,
         struct c2_qp *qp;
         int is_recv = 0;
 
-        ce = (struct c2wr_ce *) c2_mq_consume(&cq->mq);
+        ce = c2_mq_consume(&cq->mq);
         if (!ce) {
                 return -EAGAIN;
         }
@@ -146,7 +146,7 @@ static inline int c2_poll_one(struct c2_dev *c2dev,
         while ((qp =
                 (struct c2_qp *) (unsigned long) ce->qp_user_context) == NULL) {
                 c2_mq_free(&cq->mq);
-                ce = (struct c2wr_ce *) c2_mq_consume(&cq->mq);
+                ce = c2_mq_consume(&cq->mq);
                 if (!ce)
                         return -EAGAIN;
         }
diff --git a/drivers/infiniband/hw/cxgb3/cxio_wr.h b/drivers/infiniband/hw/cxgb3/cxio_wr.h
index ff9be1a13106..32e3b1461d81 100644
--- a/drivers/infiniband/hw/cxgb3/cxio_wr.h
+++ b/drivers/infiniband/hw/cxgb3/cxio_wr.h
@@ -176,7 +176,7 @@ struct t3_send_wr {
         struct t3_sge sgl[T3_MAX_SGE];  /* 4+ */
 };
 
-#define T3_MAX_FASTREG_DEPTH 24
+#define T3_MAX_FASTREG_DEPTH 10
 #define T3_MAX_FASTREG_FRAG 10
 
 struct t3_fastreg_wr {
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index 160ef482712d..e2a63214008a 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -40,6 +40,7 @@
 #include <linux/spinlock.h>
 #include <linux/ethtool.h>
 #include <linux/rtnetlink.h>
+#include <linux/inetdevice.h>
 
 #include <asm/io.h>
 #include <asm/irq.h>
@@ -1152,12 +1153,39 @@ static int iwch_query_device(struct ib_device *ibdev,
 static int iwch_query_port(struct ib_device *ibdev,
                            u8 port, struct ib_port_attr *props)
 {
+        struct iwch_dev *dev;
+        struct net_device *netdev;
+        struct in_device *inetdev;
+
         PDBG("%s ibdev %p\n", __func__, ibdev);
 
+        dev = to_iwch_dev(ibdev);
+        netdev = dev->rdev.port_info.lldevs[port-1];
+
         memset(props, 0, sizeof(struct ib_port_attr));
         props->max_mtu = IB_MTU_4096;
-        props->active_mtu = IB_MTU_2048;
-        props->state = IB_PORT_ACTIVE;
+        if (netdev->mtu >= 4096)
+                props->active_mtu = IB_MTU_4096;
+        else if (netdev->mtu >= 2048)
+                props->active_mtu = IB_MTU_2048;
+        else if (netdev->mtu >= 1024)
+                props->active_mtu = IB_MTU_1024;
+        else if (netdev->mtu >= 512)
+                props->active_mtu = IB_MTU_512;
+        else
+                props->active_mtu = IB_MTU_256;
+
+        if (!netif_carrier_ok(netdev))
+                props->state = IB_PORT_DOWN;
+        else {
+                inetdev = in_dev_get(netdev);
+                if (inetdev->ifa_list)
+                        props->state = IB_PORT_ACTIVE;
+                else
+                        props->state = IB_PORT_INIT;
+                in_dev_put(inetdev);
+        }
+
         props->port_cap_flags =
             IB_PORT_CM_SUP |
             IB_PORT_SNMP_TUNNEL_SUP |
diff --git a/drivers/infiniband/hw/ehca/ehca_irq.c b/drivers/infiniband/hw/ehca/ehca_irq.c
index 99bcbd7ffb0a..4b89b791be6a 100644
--- a/drivers/infiniband/hw/ehca/ehca_irq.c
+++ b/drivers/infiniband/hw/ehca/ehca_irq.c
@@ -479,13 +479,13 @@ void ehca_tasklet_neq(unsigned long data)
         struct ehca_eqe *eqe;
         u64 ret;
 
-        eqe = (struct ehca_eqe *)ehca_poll_eq(shca, &shca->neq);
+        eqe = ehca_poll_eq(shca, &shca->neq);
 
         while (eqe) {
                 if (!EHCA_BMASK_GET(NEQE_COMPLETION_EVENT, eqe->entry))
                         parse_ec(shca, eqe->entry);
 
-                eqe = (struct ehca_eqe *)ehca_poll_eq(shca, &shca->neq);
+                eqe = ehca_poll_eq(shca, &shca->neq);
         }
 
         ret = hipz_h_reset_event(shca->ipz_hca_handle,
@@ -572,8 +572,7 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
         eqe_cnt = 0;
         do {
                 u32 token;
-                eqe_cache[eqe_cnt].eqe =
-                        (struct ehca_eqe *)ehca_poll_eq(shca, eq);
+                eqe_cache[eqe_cnt].eqe = ehca_poll_eq(shca, eq);
                 if (!eqe_cache[eqe_cnt].eqe)
                         break;
                 eqe_value = eqe_cache[eqe_cnt].eqe->entry;
@@ -637,7 +636,7 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
                 goto unlock_irq_spinlock;
         do {
                 struct ehca_eqe *eqe;
-                eqe = (struct ehca_eqe *)ehca_poll_eq(shca, &shca->eq);
+                eqe = ehca_poll_eq(shca, &shca->eq);
                 if (!eqe)
                         break;
                 process_eqe(shca, eqe);
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 20724aee76f4..c4a02648c8af 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -1585,12 +1585,16 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
                                 break;
 
                         case IB_WR_LOCAL_INV:
+                                ctrl->srcrb_flags |=
+                                        cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
                                 set_local_inv_seg(wqe, wr->ex.invalidate_rkey);
                                 wqe  += sizeof (struct mlx4_wqe_local_inval_seg);
                                 size += sizeof (struct mlx4_wqe_local_inval_seg) / 16;
                                 break;
 
                         case IB_WR_FAST_REG_MR:
+                                ctrl->srcrb_flags |=
+                                        cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
                                 set_fmr_seg(wqe, wr);
                                 wqe  += sizeof (struct mlx4_wqe_fmr_seg);
                                 size += sizeof (struct mlx4_wqe_fmr_seg) / 16;
diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c
index 6d55f9d748f6..8c2ed994d540 100644
--- a/drivers/infiniband/hw/mthca/mthca_cmd.c
+++ b/drivers/infiniband/hw/mthca/mthca_cmd.c
@@ -1059,7 +1059,7 @@ int mthca_QUERY_DEV_LIM(struct mthca_dev *dev,
         MTHCA_GET(field, outbox, QUERY_DEV_LIM_RSVD_MTT_OFFSET);
         if (mthca_is_memfree(dev))
                 dev_lim->reserved_mtts = ALIGN((1 << (field >> 4)) * sizeof(u64),
-                                               MTHCA_MTT_SEG_SIZE) / MTHCA_MTT_SEG_SIZE;
+                                               dev->limits.mtt_seg_size) / dev->limits.mtt_seg_size;
         else
                 dev_lim->reserved_mtts = 1 << (field >> 4);
         MTHCA_GET(field, outbox, QUERY_DEV_LIM_MAX_MRW_SZ_OFFSET);
diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h
index 252590116df5..9ef611f6dd36 100644
--- a/drivers/infiniband/hw/mthca/mthca_dev.h
+++ b/drivers/infiniband/hw/mthca/mthca_dev.h
@@ -159,6 +159,7 @@ struct mthca_limits {
         int      reserved_eqs;
         int      num_mpts;
         int      num_mtt_segs;
+        int      mtt_seg_size;
         int      fmr_reserved_mtts;
         int      reserved_mtts;
         int      reserved_mrws;
diff --git a/drivers/infiniband/hw/mthca/mthca_eq.c b/drivers/infiniband/hw/mthca/mthca_eq.c
index 28f0e0c40d7d..90e4e450a120 100644
--- a/drivers/infiniband/hw/mthca/mthca_eq.c
+++ b/drivers/infiniband/hw/mthca/mthca_eq.c
@@ -641,9 +641,11 @@ static void mthca_free_irqs(struct mthca_dev *dev)
         if (dev->eq_table.have_irq)
                 free_irq(dev->pdev->irq, dev);
         for (i = 0; i < MTHCA_NUM_EQ; ++i)
-                if (dev->eq_table.eq[i].have_irq)
+                if (dev->eq_table.eq[i].have_irq) {
                         free_irq(dev->eq_table.eq[i].msi_x_vector,
                                  dev->eq_table.eq + i);
+                        dev->eq_table.eq[i].have_irq = 0;
+                }
 }
 
 static int mthca_map_reg(struct mthca_dev *dev,
diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c
index 1d83cf7caf38..13da9f1d24c0 100644
--- a/drivers/infiniband/hw/mthca/mthca_main.c
+++ b/drivers/infiniband/hw/mthca/mthca_main.c
@@ -125,6 +125,10 @@ module_param_named(fmr_reserved_mtts, hca_profile.fmr_reserved_mtts, int, 0444);
 MODULE_PARM_DESC(fmr_reserved_mtts,
                  "number of memory translation table segments reserved for FMR");
 
+static int log_mtts_per_seg = ilog2(MTHCA_MTT_SEG_SIZE / 8);
+module_param_named(log_mtts_per_seg, log_mtts_per_seg, int, 0444);
+MODULE_PARM_DESC(log_mtts_per_seg, "Log2 number of MTT entries per segment (1-5)");
+
 static char mthca_version[] __devinitdata =
         DRV_NAME ": Mellanox InfiniBand HCA driver v"
         DRV_VERSION " (" DRV_RELDATE ")\n";
@@ -162,6 +166,7 @@ static int mthca_dev_lim(struct mthca_dev *mdev, struct mthca_dev_lim *dev_lim)
         int err;
         u8 status;
 
+        mdev->limits.mtt_seg_size = (1 << log_mtts_per_seg) * 8;
         err = mthca_QUERY_DEV_LIM(mdev, dev_lim, &status);
         if (err) {
                 mthca_err(mdev, "QUERY_DEV_LIM command failed, aborting.\n");
@@ -460,11 +465,11 @@ static int mthca_init_icm(struct mthca_dev *mdev,
         }
 
         /* CPU writes to non-reserved MTTs, while HCA might DMA to reserved mtts */
-        mdev->limits.reserved_mtts = ALIGN(mdev->limits.reserved_mtts * MTHCA_MTT_SEG_SIZE,
-                                           dma_get_cache_alignment()) / MTHCA_MTT_SEG_SIZE;
+        mdev->limits.reserved_mtts = ALIGN(mdev->limits.reserved_mtts * mdev->limits.mtt_seg_size,
+                                           dma_get_cache_alignment()) / mdev->limits.mtt_seg_size;
 
         mdev->mr_table.mtt_table = mthca_alloc_icm_table(mdev, init_hca->mtt_base,
-                                                         MTHCA_MTT_SEG_SIZE,
+                                                         mdev->limits.mtt_seg_size,
                                                          mdev->limits.num_mtt_segs,
                                                          mdev->limits.reserved_mtts,
                                                          1, 0);
@@ -1315,6 +1320,12 @@ static void __init mthca_validate_profile(void)
                 printk(KERN_WARNING PFX "Corrected fmr_reserved_mtts to %d.\n",
                        hca_profile.fmr_reserved_mtts);
         }
+
+        if ((log_mtts_per_seg < 1) || (log_mtts_per_seg > 5)) {
+                printk(KERN_WARNING PFX "bad log_mtts_per_seg (%d). Using default - %d\n",
+                       log_mtts_per_seg, ilog2(MTHCA_MTT_SEG_SIZE / 8));
+                log_mtts_per_seg = ilog2(MTHCA_MTT_SEG_SIZE / 8);
+        }
 }
 
 static int __init mthca_init(void)
diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c
index 882e6b735915..d606edf10858 100644
--- a/drivers/infiniband/hw/mthca/mthca_mr.c
+++ b/drivers/infiniband/hw/mthca/mthca_mr.c
@@ -220,7 +220,7 @@ static struct mthca_mtt *__mthca_alloc_mtt(struct mthca_dev *dev, int size,
 
         mtt->buddy = buddy;
         mtt->order = 0;
-        for (i = MTHCA_MTT_SEG_SIZE / 8; i < size; i <<= 1)
+        for (i = dev->limits.mtt_seg_size / 8; i < size; i <<= 1)
                 ++mtt->order;
 
         mtt->first_seg = mthca_alloc_mtt_range(dev, mtt->order, buddy);
@@ -267,7 +267,7 @@ static int __mthca_write_mtt(struct mthca_dev *dev, struct mthca_mtt *mtt,
 
         while (list_len > 0) {
                 mtt_entry[0] = cpu_to_be64(dev->mr_table.mtt_base +
-                                           mtt->first_seg * MTHCA_MTT_SEG_SIZE +
+                                           mtt->first_seg * dev->limits.mtt_seg_size +
                                            start_index * 8);
                 mtt_entry[1] = 0;
                 for (i = 0; i < list_len && i < MTHCA_MAILBOX_SIZE / 8 - 2; ++i)
@@ -326,7 +326,7 @@ static void mthca_tavor_write_mtt_seg(struct mthca_dev *dev,
         u64 __iomem *mtts;
         int i;
 
-        mtts = dev->mr_table.tavor_fmr.mtt_base + mtt->first_seg * MTHCA_MTT_SEG_SIZE +
+        mtts = dev->mr_table.tavor_fmr.mtt_base + mtt->first_seg * dev->limits.mtt_seg_size +
                 start_index * sizeof (u64);
         for (i = 0; i < list_len; ++i)
                 mthca_write64_raw(cpu_to_be64(buffer_list[i] | MTHCA_MTT_FLAG_PRESENT),
@@ -345,10 +345,10 @@ static void mthca_arbel_write_mtt_seg(struct mthca_dev *dev,
         /* For Arbel, all MTTs must fit in the same page. */
         BUG_ON(s / PAGE_SIZE != (s + list_len * sizeof(u64) - 1) / PAGE_SIZE);
         /* Require full segments */
-        BUG_ON(s % MTHCA_MTT_SEG_SIZE);
+        BUG_ON(s % dev->limits.mtt_seg_size);
 
         mtts = mthca_table_find(dev->mr_table.mtt_table, mtt->first_seg +
-                                s / MTHCA_MTT_SEG_SIZE, &dma_handle);
+                                s / dev->limits.mtt_seg_size, &dma_handle);
 
         BUG_ON(!mtts);
 
@@ -479,7 +479,7 @@ int mthca_mr_alloc(struct mthca_dev *dev, u32 pd, int buffer_size_shift,
         if (mr->mtt)
                 mpt_entry->mtt_seg =
                         cpu_to_be64(dev->mr_table.mtt_base +
-                                    mr->mtt->first_seg * MTHCA_MTT_SEG_SIZE);
+                                    mr->mtt->first_seg * dev->limits.mtt_seg_size);
 
         if (0) {
                 mthca_dbg(dev, "Dumping MPT entry %08x:\n", mr->ibmr.lkey);
@@ -626,7 +626,7 @@ int mthca_fmr_alloc(struct mthca_dev *dev, u32 pd,
                 goto err_out_table;
         }
 
-        mtt_seg = mr->mtt->first_seg * MTHCA_MTT_SEG_SIZE;
+        mtt_seg = mr->mtt->first_seg * dev->limits.mtt_seg_size;
 
         if (mthca_is_memfree(dev)) {
                 mr->mem.arbel.mtts = mthca_table_find(dev->mr_table.mtt_table,
@@ -908,7 +908,7 @@ int mthca_init_mr_table(struct mthca_dev *dev)
                          dev->mr_table.mtt_base);
 
                 dev->mr_table.tavor_fmr.mtt_base =
-                        ioremap(addr, mtts * MTHCA_MTT_SEG_SIZE);
+                        ioremap(addr, mtts * dev->limits.mtt_seg_size);
                 if (!dev->mr_table.tavor_fmr.mtt_base) {
                         mthca_warn(dev, "MTT ioremap for FMR failed.\n");
                         err = -ENOMEM;
diff --git a/drivers/infiniband/hw/mthca/mthca_profile.c b/drivers/infiniband/hw/mthca/mthca_profile.c
index d168c2540611..8edb28a9a0e7 100644
--- a/drivers/infiniband/hw/mthca/mthca_profile.c
+++ b/drivers/infiniband/hw/mthca/mthca_profile.c
@@ -94,7 +94,7 @@ s64 mthca_make_profile(struct mthca_dev *dev,
         profile[MTHCA_RES_RDB].size  = MTHCA_RDB_ENTRY_SIZE;
         profile[MTHCA_RES_MCG].size  = MTHCA_MGM_ENTRY_SIZE;
         profile[MTHCA_RES_MPT].size  = dev_lim->mpt_entry_sz;
-        profile[MTHCA_RES_MTT].size  = MTHCA_MTT_SEG_SIZE;
+        profile[MTHCA_RES_MTT].size  = dev->limits.mtt_seg_size;
         profile[MTHCA_RES_UAR].size  = dev_lim->uar_scratch_entry_sz;
         profile[MTHCA_RES_UDAV].size = MTHCA_AV_SIZE;
         profile[MTHCA_RES_UARC].size = request->uarc_size;
@@ -232,7 +232,7 @@ s64 mthca_make_profile(struct mthca_dev *dev,
                         dev->limits.num_mtt_segs = profile[i].num;
                         dev->mr_table.mtt_base   = profile[i].start;
                         init_hca->mtt_base       = profile[i].start;
-                        init_hca->mtt_seg_sz     = ffs(MTHCA_MTT_SEG_SIZE) - 7;
+                        init_hca->mtt_seg_sz     = ffs(dev->limits.mtt_seg_size) - 7;
                         break;
                 case MTHCA_RES_UAR:
                         dev->limits.num_uars       = profile[i].num;
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index b832a7b814a2..4a84d02ece06 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -667,7 +667,7 @@ static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_
                 i = 0;
                 while (((nes_read32(nesdev->regs+NES_SOFTWARE_RESET) & 0x00000040) == 0) && i++ < 10000)
                         mdelay(1);
-                if (i >= 10000) {
+                if (i > 10000) {
                         nes_debug(NES_DBG_INIT, "Did not see full soft reset done.\n");
                         return 0;
                 }
@@ -675,7 +675,7 @@ static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_
                 i = 0;
                 while ((nes_read_indexed(nesdev, NES_IDX_INT_CPU_STATUS) != 0x80) && i++ < 10000)
                         mdelay(1);
-                if (i >= 10000) {
+                if (i > 10000) {
                         printk(KERN_ERR PFX "Internal CPU not ready, status = %02X\n",
                                nes_read_indexed(nesdev, NES_IDX_INT_CPU_STATUS));
                         return 0;
@@ -701,7 +701,7 @@ static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_
         i = 0;
         while (((nes_read32(nesdev->regs+NES_SOFTWARE_RESET) & 0x00000040) == 0) && i++ < 10000)
                 mdelay(1);
-        if (i >= 10000) {
+        if (i > 10000) {
                 nes_debug(NES_DBG_INIT, "Did not see port soft reset done.\n");
                 return 0;
         }
@@ -711,7 +711,7 @@ static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_
         while (((u32temp = (nes_read_indexed(nesdev, NES_IDX_ETH_SERDES_COMMON_STATUS0)
                         & 0x0000000f)) != 0x0000000f) && i++ < 5000)
                 mdelay(1);
-        if (i >= 5000) {
+        if (i > 5000) {
                 nes_debug(NES_DBG_INIT, "Serdes 0 not ready, status=%x\n", u32temp);
                 return 0;
         }
@@ -722,7 +722,7 @@ static unsigned int nes_reset_adapter_ne020(struct nes_device *nesdev, u8 *OneG_
                 while (((u32temp = (nes_read_indexed(nesdev, NES_IDX_ETH_SERDES_COMMON_STATUS1)
                                 & 0x0000000f)) != 0x0000000f) && i++ < 5000)
                         mdelay(1);
-                if (i >= 5000) {
+                if (i > 5000) {
                         nes_debug(NES_DBG_INIT, "Serdes 1 not ready, status=%x\n", u32temp);
                         return 0;
                 }
@@ -792,7 +792,7 @@ static int nes_init_serdes(struct nes_device *nesdev, u8 hw_rev, u8 port_count,
                 while (((u32temp = (nes_read_indexed(nesdev, NES_IDX_ETH_SERDES_COMMON_STATUS0)
                                 & 0x0000000f)) != 0x0000000f) && i++ < 5000)
                         mdelay(1);
-                if (i >= 5000) {
+                if (i > 5000) {
                         nes_debug(NES_DBG_PHY, "Init: serdes 0 not ready, status=%x\n", u32temp);
                         return 1;
                 }
@@ -815,7 +815,7 @@ static int nes_init_serdes(struct nes_device *nesdev, u8 hw_rev, u8 port_count,
                         while (((u32temp = (nes_read_indexed(nesdev, NES_IDX_ETH_SERDES_COMMON_STATUS1)
                                 & 0x0000000f)) != 0x0000000f) && (i++ < 5000))
                                 mdelay(1);
-                        if (i >= 5000) {
+                        if (i > 5000) {
                                 printk("%s: Init: serdes 1 not ready, status=%x\n", __func__, u32temp);
                                 /* return 1; */
                         }