From 7e6edb9b2e0bcfb2a588db390c44d120213c57ae Mon Sep 17 00:00:00 2001 From: Matan Barak Date: Thu, 31 Jul 2014 11:01:28 +0300 Subject: IB/core: Add user MR re-registration support Memory re-registration is a feature that enables changing the attributes of a memory region registered by user-space, including PD, translation (address and length) and access flags. Add the required support in uverbs and the kernel verbs API. Signed-off-by: Matan Barak Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- include/rdma/ib_verbs.h | 10 +++++++++- include/uapi/rdma/ib_user_verbs.h | 16 ++++++++++++++++ 2 files changed, 25 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 7ccef342f724..ed44cc07a7b3 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1097,7 +1097,8 @@ struct ib_mr_attr { enum ib_mr_rereg_flags { IB_MR_REREG_TRANS = 1, IB_MR_REREG_PD = (1<<1), - IB_MR_REREG_ACCESS = (1<<2) + IB_MR_REREG_ACCESS = (1<<2), + IB_MR_REREG_SUPPORTED = ((IB_MR_REREG_ACCESS << 1) - 1) }; /** @@ -1547,6 +1548,13 @@ struct ib_device { u64 virt_addr, int mr_access_flags, struct ib_udata *udata); + int (*rereg_user_mr)(struct ib_mr *mr, + int flags, + u64 start, u64 length, + u64 virt_addr, + int mr_access_flags, + struct ib_pd *pd, + struct ib_udata *udata); int (*query_mr)(struct ib_mr *mr, struct ib_mr_attr *mr_attr); int (*dereg_mr)(struct ib_mr *mr); diff --git a/include/uapi/rdma/ib_user_verbs.h b/include/uapi/rdma/ib_user_verbs.h index cbfdd4ca9510..26daf55ff76e 100644 --- a/include/uapi/rdma/ib_user_verbs.h +++ b/include/uapi/rdma/ib_user_verbs.h @@ -276,6 +276,22 @@ struct ib_uverbs_reg_mr_resp { __u32 rkey; }; +struct ib_uverbs_rereg_mr { + __u64 response; + __u32 mr_handle; + __u32 flags; + __u64 start; + __u64 length; + __u64 hca_va; + __u32 pd_handle; + __u32 access_flags; +}; + +struct ib_uverbs_rereg_mr_resp { + __u32 lkey; + __u32 rkey; +}; + struct ib_uverbs_dereg_mr { __u32 mr_handle; }; -- cgit v1.2.3 From e630664c8383f300c4146d7613d61e5a8eb1f8e3 Mon Sep 17 00:00:00 2001 From: Matan Barak Date: Thu, 31 Jul 2014 11:01:29 +0300 Subject: mlx4_core: Add helper functions to support MR re-registration Add few helper functions to support a mechanism of getting an MPT, modifying it and updating the HCA with the modified object. The code takes 2 paths, one for directly changing the MPT (and sometimes its related MTTs) and another one which queries the MPT and updates the HCA via fw command SW2HW_MPT. The first path is used in native mode; the second path is slower and is used only in SRIOV. Signed-off-by: Jack Morgenstein Signed-off-by: Matan Barak Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- drivers/net/ethernet/mellanox/mlx4/mlx4.h | 2 + drivers/net/ethernet/mellanox/mlx4/mr.c | 160 +++++++++++++++++++++ .../net/ethernet/mellanox/mlx4/resource_tracker.c | 26 +++- include/linux/mlx4/device.h | 16 +++ 4 files changed, 202 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4.h b/drivers/net/ethernet/mellanox/mlx4/mlx4.h index 1d8af7336807..b40d587974fa 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mlx4.h +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4.h @@ -279,6 +279,8 @@ struct mlx4_icm_table { #define MLX4_MPT_FLAG_PHYSICAL (1 << 9) #define MLX4_MPT_FLAG_REGION (1 << 8) +#define MLX4_MPT_PD_MASK (0x1FFFFUL) +#define MLX4_MPT_PD_VF_MASK (0xFE0000UL) #define MLX4_MPT_PD_FLAG_FAST_REG (1 << 27) #define MLX4_MPT_PD_FLAG_RAE (1 << 28) #define MLX4_MPT_PD_FLAG_EN_INV (3 << 24) diff --git a/drivers/net/ethernet/mellanox/mlx4/mr.c b/drivers/net/ethernet/mellanox/mlx4/mr.c index 2839abb878a6..7d717eccb7b0 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mr.c +++ b/drivers/net/ethernet/mellanox/mlx4/mr.c @@ -298,6 +298,131 @@ static int mlx4_HW2SW_MPT(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *mailbox MLX4_CMD_TIME_CLASS_B, MLX4_CMD_WRAPPED); } +int mlx4_mr_hw_get_mpt(struct mlx4_dev *dev, struct mlx4_mr *mmr, + struct mlx4_mpt_entry ***mpt_entry) +{ + int err; + int key = key_to_hw_index(mmr->key) & (dev->caps.num_mpts - 1); + struct mlx4_cmd_mailbox *mailbox = NULL; + + /* Make sure that at this point we have single-threaded access only */ + + if (mmr->enabled != MLX4_MPT_EN_HW) + return -EINVAL; + + err = mlx4_HW2SW_MPT(dev, NULL, key); + + if (err) { + mlx4_warn(dev, "HW2SW_MPT failed (%d).", err); + mlx4_warn(dev, "Most likely the MR has MWs bound to it.\n"); + return err; + } + + mmr->enabled = MLX4_MPT_EN_SW; + + if (!mlx4_is_mfunc(dev)) { + **mpt_entry = mlx4_table_find( + &mlx4_priv(dev)->mr_table.dmpt_table, + key, NULL); + } else { + mailbox = mlx4_alloc_cmd_mailbox(dev); + if (IS_ERR_OR_NULL(mailbox)) + return PTR_ERR(mailbox); + + err = mlx4_cmd_box(dev, 0, mailbox->dma, key, + 0, MLX4_CMD_QUERY_MPT, + MLX4_CMD_TIME_CLASS_B, + MLX4_CMD_WRAPPED); + + if (err) + goto free_mailbox; + + *mpt_entry = (struct mlx4_mpt_entry **)&mailbox->buf; + } + + if (!(*mpt_entry) || !(**mpt_entry)) { + err = -ENOMEM; + goto free_mailbox; + } + + return 0; + +free_mailbox: + mlx4_free_cmd_mailbox(dev, mailbox); + return err; +} +EXPORT_SYMBOL_GPL(mlx4_mr_hw_get_mpt); + +int mlx4_mr_hw_write_mpt(struct mlx4_dev *dev, struct mlx4_mr *mmr, + struct mlx4_mpt_entry **mpt_entry) +{ + int err; + + if (!mlx4_is_mfunc(dev)) { + /* Make sure any changes to this entry are flushed */ + wmb(); + + *(u8 *)(*mpt_entry) = MLX4_MPT_STATUS_HW; + + /* Make sure the new status is written */ + wmb(); + + err = mlx4_SYNC_TPT(dev); + } else { + int key = key_to_hw_index(mmr->key) & (dev->caps.num_mpts - 1); + + struct mlx4_cmd_mailbox *mailbox = + container_of((void *)mpt_entry, struct mlx4_cmd_mailbox, + buf); + + err = mlx4_SW2HW_MPT(dev, mailbox, key); + } + + mmr->pd = be32_to_cpu((*mpt_entry)->pd_flags) & MLX4_MPT_PD_MASK; + if (!err) + mmr->enabled = MLX4_MPT_EN_HW; + return err; +} +EXPORT_SYMBOL_GPL(mlx4_mr_hw_write_mpt); + +void mlx4_mr_hw_put_mpt(struct mlx4_dev *dev, + struct mlx4_mpt_entry **mpt_entry) +{ + if (mlx4_is_mfunc(dev)) { + struct mlx4_cmd_mailbox *mailbox = + container_of((void *)mpt_entry, struct mlx4_cmd_mailbox, + buf); + mlx4_free_cmd_mailbox(dev, mailbox); + } +} +EXPORT_SYMBOL_GPL(mlx4_mr_hw_put_mpt); + +int mlx4_mr_hw_change_pd(struct mlx4_dev *dev, struct mlx4_mpt_entry *mpt_entry, + u32 pdn) +{ + u32 pd_flags = be32_to_cpu(mpt_entry->pd_flags); + /* The wrapper function will put the slave's id here */ + if (mlx4_is_mfunc(dev)) + pd_flags &= ~MLX4_MPT_PD_VF_MASK; + mpt_entry->pd_flags = cpu_to_be32((pd_flags & ~MLX4_MPT_PD_MASK) | + (pdn & MLX4_MPT_PD_MASK) + | MLX4_MPT_PD_FLAG_EN_INV); + return 0; +} +EXPORT_SYMBOL_GPL(mlx4_mr_hw_change_pd); + +int mlx4_mr_hw_change_access(struct mlx4_dev *dev, + struct mlx4_mpt_entry *mpt_entry, + u32 access) +{ + u32 flags = (be32_to_cpu(mpt_entry->flags) & ~MLX4_PERM_MASK) | + (access & MLX4_PERM_MASK); + + mpt_entry->flags = cpu_to_be32(flags); + return 0; +} +EXPORT_SYMBOL_GPL(mlx4_mr_hw_change_access); + static int mlx4_mr_alloc_reserved(struct mlx4_dev *dev, u32 mridx, u32 pd, u64 iova, u64 size, u32 access, int npages, int page_shift, struct mlx4_mr *mr) @@ -463,6 +588,41 @@ int mlx4_mr_free(struct mlx4_dev *dev, struct mlx4_mr *mr) } EXPORT_SYMBOL_GPL(mlx4_mr_free); +void mlx4_mr_rereg_mem_cleanup(struct mlx4_dev *dev, struct mlx4_mr *mr) +{ + mlx4_mtt_cleanup(dev, &mr->mtt); +} +EXPORT_SYMBOL_GPL(mlx4_mr_rereg_mem_cleanup); + +int mlx4_mr_rereg_mem_write(struct mlx4_dev *dev, struct mlx4_mr *mr, + u64 iova, u64 size, int npages, + int page_shift, struct mlx4_mpt_entry *mpt_entry) +{ + int err; + + mpt_entry->start = cpu_to_be64(mr->iova); + mpt_entry->length = cpu_to_be64(mr->size); + mpt_entry->entity_size = cpu_to_be32(mr->mtt.page_shift); + + err = mlx4_mtt_init(dev, npages, page_shift, &mr->mtt); + if (err) + return err; + + if (mr->mtt.order < 0) { + mpt_entry->flags |= cpu_to_be32(MLX4_MPT_FLAG_PHYSICAL); + mpt_entry->mtt_addr = 0; + } else { + mpt_entry->mtt_addr = cpu_to_be64(mlx4_mtt_addr(dev, + &mr->mtt)); + if (mr->mtt.page_shift == 0) + mpt_entry->mtt_sz = cpu_to_be32(1 << mr->mtt.order); + } + mr->enabled = MLX4_MPT_EN_SW; + + return 0; +} +EXPORT_SYMBOL_GPL(mlx4_mr_rereg_mem_write); + int mlx4_mr_enable(struct mlx4_dev *dev, struct mlx4_mr *mr) { struct mlx4_cmd_mailbox *mailbox; diff --git a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c index 0efc1368e5a8..1089367fed22 100644 --- a/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c +++ b/drivers/net/ethernet/mellanox/mlx4/resource_tracker.c @@ -2613,12 +2613,34 @@ int mlx4_QUERY_MPT_wrapper(struct mlx4_dev *dev, int slave, if (err) return err; - if (mpt->com.from_state != RES_MPT_HW) { + if (mpt->com.from_state == RES_MPT_MAPPED) { + /* In order to allow rereg in SRIOV, we need to alter the MPT entry. To do + * that, the VF must read the MPT. But since the MPT entry memory is not + * in the VF's virtual memory space, it must use QUERY_MPT to obtain the + * entry contents. To guarantee that the MPT cannot be changed, the driver + * must perform HW2SW_MPT before this query and return the MPT entry to HW + * ownership fofollowing the change. The change here allows the VF to + * perform QUERY_MPT also when the entry is in SW ownership. + */ + struct mlx4_mpt_entry *mpt_entry = mlx4_table_find( + &mlx4_priv(dev)->mr_table.dmpt_table, + mpt->key, NULL); + + if (NULL == mpt_entry || NULL == outbox->buf) { + err = -EINVAL; + goto out; + } + + memcpy(outbox->buf, mpt_entry, sizeof(*mpt_entry)); + + err = 0; + } else if (mpt->com.from_state == RES_MPT_HW) { + err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd); + } else { err = -EBUSY; goto out; } - err = mlx4_DMA_wrapper(dev, slave, vhcr, inbox, outbox, cmd); out: put_res(dev, slave, id, RES_MPT); diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index 35b51e7af886..bac002167ace 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -262,6 +262,7 @@ enum { MLX4_PERM_REMOTE_WRITE = 1 << 13, MLX4_PERM_ATOMIC = 1 << 14, MLX4_PERM_BIND_MW = 1 << 15, + MLX4_PERM_MASK = 0xFC00 }; enum { @@ -1243,4 +1244,19 @@ int mlx4_vf_smi_enabled(struct mlx4_dev *dev, int slave, int port); int mlx4_vf_get_enable_smi_admin(struct mlx4_dev *dev, int slave, int port); int mlx4_vf_set_enable_smi_admin(struct mlx4_dev *dev, int slave, int port, int enable); +int mlx4_mr_hw_get_mpt(struct mlx4_dev *dev, struct mlx4_mr *mmr, + struct mlx4_mpt_entry ***mpt_entry); +int mlx4_mr_hw_write_mpt(struct mlx4_dev *dev, struct mlx4_mr *mmr, + struct mlx4_mpt_entry **mpt_entry); +int mlx4_mr_hw_change_pd(struct mlx4_dev *dev, struct mlx4_mpt_entry *mpt_entry, + u32 pdn); +int mlx4_mr_hw_change_access(struct mlx4_dev *dev, + struct mlx4_mpt_entry *mpt_entry, + u32 access); +void mlx4_mr_hw_put_mpt(struct mlx4_dev *dev, + struct mlx4_mpt_entry **mpt_entry); +void mlx4_mr_rereg_mem_cleanup(struct mlx4_dev *dev, struct mlx4_mr *mr); +int mlx4_mr_rereg_mem_write(struct mlx4_dev *dev, struct mlx4_mr *mr, + u64 iova, u64 size, int npages, + int page_shift, struct mlx4_mpt_entry *mpt_entry); #endif /* MLX4_DEVICE_H */ -- cgit v1.2.3 From 114840c3d29b9cbd867faa69595a2aee6f6b54a2 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Sun, 1 Jun 2014 11:53:50 +0300 Subject: mlx4_core: Add support for secure-host and SMP firewall Secure-host is the general term for the capability of a device to protect itself and the subnet from malicious host software. This is achieved by: 1. Not allowing un-trusted entities to access device configuration registers, directly (through pci_cr or pci_conf) and indirectly (through MADs). 2. Hiding M_Key from untrusted entities. 3. Preventing the modification of GUID0 by un-trusted entities 4. Not allowing drivers on untrusted hosts to receive nor to transmit packets over QP0 (SMP Firewall). The secure-host capability depends on firmware handling all QP0 packets, and not passing these packets up to the driver. Any information required by the driver for proper operation (e.g., SM lid) is passed via events generated by the firmware while processing QP0 MADs. Driver support mainly requires using the MAD_DEMUX FW command at startup, where the feature is enabled/disabled through a procedure described in the Mellanox HCA tools package. Signed-off-by: Jack Morgenstein Signed-off-by: Or Gerlitz [ Fix error path in mlx4_setup_hca to go to err_mcg_table_free. - Roland ] Signed-off-by: Roland Dreier --- drivers/net/ethernet/mellanox/mlx4/cmd.c | 9 +++ drivers/net/ethernet/mellanox/mlx4/fw.c | 91 ++++++++++++++++++++++++++++++- drivers/net/ethernet/mellanox/mlx4/main.c | 5 ++ drivers/net/ethernet/mellanox/mlx4/mlx4.h | 1 + include/linux/mlx4/cmd.h | 7 +++ include/linux/mlx4/device.h | 1 + 6 files changed, 113 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/drivers/net/ethernet/mellanox/mlx4/cmd.c b/drivers/net/ethernet/mellanox/mlx4/cmd.c index 5d940a26055c..65a4a0f88ea0 100644 --- a/drivers/net/ethernet/mellanox/mlx4/cmd.c +++ b/drivers/net/ethernet/mellanox/mlx4/cmd.c @@ -1310,6 +1310,15 @@ static struct mlx4_cmd_info cmd_info[] = { .verify = NULL, .wrapper = mlx4_MAD_IFC_wrapper }, + { + .opcode = MLX4_CMD_MAD_DEMUX, + .has_inbox = false, + .has_outbox = false, + .out_is_imm = false, + .encode_slave_id = false, + .verify = NULL, + .wrapper = mlx4_CMD_EPERM_wrapper + }, { .opcode = MLX4_CMD_QUERY_IF_STAT, .has_inbox = false, diff --git a/drivers/net/ethernet/mellanox/mlx4/fw.c b/drivers/net/ethernet/mellanox/mlx4/fw.c index 688e1eabab29..494753e44ae3 100644 --- a/drivers/net/ethernet/mellanox/mlx4/fw.c +++ b/drivers/net/ethernet/mellanox/mlx4/fw.c @@ -136,7 +136,8 @@ static void dump_dev_cap_flags2(struct mlx4_dev *dev, u64 flags) [7] = "FSM (MAC anti-spoofing) support", [8] = "Dynamic QP updates support", [9] = "Device managed flow steering IPoIB support", - [10] = "TCP/IP offloads/flow-steering for VXLAN support" + [10] = "TCP/IP offloads/flow-steering for VXLAN support", + [11] = "MAD DEMUX (Secure-Host) support" }; int i; @@ -571,6 +572,7 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) #define QUERY_DEV_CAP_MAX_ICM_SZ_OFFSET 0xa0 #define QUERY_DEV_CAP_FW_REASSIGN_MAC 0x9d #define QUERY_DEV_CAP_VXLAN 0x9e +#define QUERY_DEV_CAP_MAD_DEMUX_OFFSET 0xb0 dev_cap->flags2 = 0; mailbox = mlx4_alloc_cmd_mailbox(dev); @@ -748,6 +750,11 @@ int mlx4_QUERY_DEV_CAP(struct mlx4_dev *dev, struct mlx4_dev_cap *dev_cap) MLX4_GET(dev_cap->max_counters, outbox, QUERY_DEV_CAP_MAX_COUNTERS_OFFSET); + MLX4_GET(field32, outbox, + QUERY_DEV_CAP_MAD_DEMUX_OFFSET); + if (field32 & (1 << 0)) + dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_MAD_DEMUX; + MLX4_GET(field32, outbox, QUERY_DEV_CAP_EXT_2_FLAGS_OFFSET); if (field32 & (1 << 16)) dev_cap->flags2 |= MLX4_DEV_CAP_FLAG2_UPDATE_QP; @@ -2016,3 +2023,85 @@ void mlx4_opreq_action(struct work_struct *work) out: mlx4_free_cmd_mailbox(dev, mailbox); } + +static int mlx4_check_smp_firewall_active(struct mlx4_dev *dev, + struct mlx4_cmd_mailbox *mailbox) +{ +#define MLX4_CMD_MAD_DEMUX_SET_ATTR_OFFSET 0x10 +#define MLX4_CMD_MAD_DEMUX_GETRESP_ATTR_OFFSET 0x20 +#define MLX4_CMD_MAD_DEMUX_TRAP_ATTR_OFFSET 0x40 +#define MLX4_CMD_MAD_DEMUX_TRAP_REPRESS_ATTR_OFFSET 0x70 + + u32 set_attr_mask, getresp_attr_mask; + u32 trap_attr_mask, traprepress_attr_mask; + + MLX4_GET(set_attr_mask, mailbox->buf, + MLX4_CMD_MAD_DEMUX_SET_ATTR_OFFSET); + mlx4_dbg(dev, "SMP firewall set_attribute_mask = 0x%x\n", + set_attr_mask); + + MLX4_GET(getresp_attr_mask, mailbox->buf, + MLX4_CMD_MAD_DEMUX_GETRESP_ATTR_OFFSET); + mlx4_dbg(dev, "SMP firewall getresp_attribute_mask = 0x%x\n", + getresp_attr_mask); + + MLX4_GET(trap_attr_mask, mailbox->buf, + MLX4_CMD_MAD_DEMUX_TRAP_ATTR_OFFSET); + mlx4_dbg(dev, "SMP firewall trap_attribute_mask = 0x%x\n", + trap_attr_mask); + + MLX4_GET(traprepress_attr_mask, mailbox->buf, + MLX4_CMD_MAD_DEMUX_TRAP_REPRESS_ATTR_OFFSET); + mlx4_dbg(dev, "SMP firewall traprepress_attribute_mask = 0x%x\n", + traprepress_attr_mask); + + if (set_attr_mask && getresp_attr_mask && trap_attr_mask && + traprepress_attr_mask) + return 1; + + return 0; +} + +int mlx4_config_mad_demux(struct mlx4_dev *dev) +{ + struct mlx4_cmd_mailbox *mailbox; + int secure_host_active; + int err; + + /* Check if mad_demux is supported */ + if (!(dev->caps.flags2 & MLX4_DEV_CAP_FLAG2_MAD_DEMUX)) + return 0; + + mailbox = mlx4_alloc_cmd_mailbox(dev); + if (IS_ERR(mailbox)) { + mlx4_warn(dev, "Failed to allocate mailbox for cmd MAD_DEMUX"); + return -ENOMEM; + } + + /* Query mad_demux to find out which MADs are handled by internal sma */ + err = mlx4_cmd_box(dev, 0, mailbox->dma, 0x01 /* subn mgmt class */, + MLX4_CMD_MAD_DEMUX_QUERY_RESTR, MLX4_CMD_MAD_DEMUX, + MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE); + if (err) { + mlx4_warn(dev, "MLX4_CMD_MAD_DEMUX: query restrictions failed (%d)\n", + err); + goto out; + } + + secure_host_active = mlx4_check_smp_firewall_active(dev, mailbox); + + /* Config mad_demux to handle all MADs returned by the query above */ + err = mlx4_cmd(dev, mailbox->dma, 0x01 /* subn mgmt class */, + MLX4_CMD_MAD_DEMUX_CONFIG, MLX4_CMD_MAD_DEMUX, + MLX4_CMD_TIME_CLASS_B, MLX4_CMD_NATIVE); + if (err) { + mlx4_warn(dev, "MLX4_CMD_MAD_DEMUX: configure failed (%d)\n", err); + goto out; + } + + if (secure_host_active) + mlx4_warn(dev, "HCA operating in secure-host mode. SMP firewall activated.\n"); +out: + mlx4_free_cmd_mailbox(dev, mailbox); + return err; +} diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c index 82ab427290c3..f2c8e8ba23fe 100644 --- a/drivers/net/ethernet/mellanox/mlx4/main.c +++ b/drivers/net/ethernet/mellanox/mlx4/main.c @@ -1831,6 +1831,11 @@ static int mlx4_setup_hca(struct mlx4_dev *dev) mlx4_err(dev, "Failed to initialize multicast group table, aborting\n"); goto err_mr_table_free; } + err = mlx4_config_mad_demux(dev); + if (err) { + mlx4_err(dev, "Failed in config_mad_demux, aborting\n"); + goto err_mcg_table_free; + } } err = mlx4_init_eq_table(dev); diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4.h b/drivers/net/ethernet/mellanox/mlx4/mlx4.h index 1d8af7336807..310b30b4682c 100644 --- a/drivers/net/ethernet/mellanox/mlx4/mlx4.h +++ b/drivers/net/ethernet/mellanox/mlx4/mlx4.h @@ -1311,5 +1311,6 @@ void mlx4_init_quotas(struct mlx4_dev *dev); int mlx4_get_slave_num_gids(struct mlx4_dev *dev, int slave, int port); /* Returns the VF index of slave */ int mlx4_get_vf_indx(struct mlx4_dev *dev, int slave); +int mlx4_config_mad_demux(struct mlx4_dev *dev); #endif /* MLX4_H */ diff --git a/include/linux/mlx4/cmd.h b/include/linux/mlx4/cmd.h index c8450366c130..379c02648ab3 100644 --- a/include/linux/mlx4/cmd.h +++ b/include/linux/mlx4/cmd.h @@ -116,6 +116,7 @@ enum { /* special QP and management commands */ MLX4_CMD_CONF_SPECIAL_QP = 0x23, MLX4_CMD_MAD_IFC = 0x24, + MLX4_CMD_MAD_DEMUX = 0x203, /* multicast commands */ MLX4_CMD_READ_MCG = 0x25, @@ -185,6 +186,12 @@ enum { MLX4_SET_PORT_VXLAN = 0xB }; +enum { + MLX4_CMD_MAD_DEMUX_CONFIG = 0, + MLX4_CMD_MAD_DEMUX_QUERY_STATE = 1, + MLX4_CMD_MAD_DEMUX_QUERY_RESTR = 2, /* Query mad demux restrictions */ +}; + enum { MLX4_CMD_WRAPPED, MLX4_CMD_NATIVE diff --git a/include/linux/mlx4/device.h b/include/linux/mlx4/device.h index 35b51e7af886..cee9561e8ef6 100644 --- a/include/linux/mlx4/device.h +++ b/include/linux/mlx4/device.h @@ -172,6 +172,7 @@ enum { MLX4_DEV_CAP_FLAG2_UPDATE_QP = 1LL << 8, MLX4_DEV_CAP_FLAG2_DMFS_IPOIB = 1LL << 9, MLX4_DEV_CAP_FLAG2_VXLAN_OFFLOADS = 1LL << 10, + MLX4_DEV_CAP_FLAG2_MAD_DEMUX = 1LL << 11, }; enum { -- cgit v1.2.3 From 0f29b46d49b0ca50536632c6a33986c3171f5ea1 Mon Sep 17 00:00:00 2001 From: Ira Weiny Date: Fri, 8 Aug 2014 19:00:55 -0400 Subject: IB/mad: add new ioctl to ABI to support new registration options Registrations options are specified through flags. Definitions of flags will be in subsequent patches. Signed-off-by: Ira Weiny Signed-off-by: Roland Dreier --- Documentation/infiniband/user_mad.txt | 13 ++-- drivers/infiniband/core/agent.c | 4 +- drivers/infiniband/core/cm.c | 5 +- drivers/infiniband/core/mad.c | 4 +- drivers/infiniband/core/sa_query.c | 2 +- drivers/infiniband/core/user_mad.c | 120 +++++++++++++++++++++++++++++++- drivers/infiniband/hw/mlx4/mad.c | 2 +- drivers/infiniband/hw/mthca/mthca_mad.c | 2 +- drivers/infiniband/hw/qib/qib_mad.c | 2 +- drivers/infiniband/ulp/srpt/ib_srpt.c | 2 +- include/rdma/ib_mad.h | 7 +- include/uapi/rdma/ib_user_mad.h | 39 +++++++++++ 12 files changed, 186 insertions(+), 16 deletions(-) (limited to 'include') diff --git a/Documentation/infiniband/user_mad.txt b/Documentation/infiniband/user_mad.txt index 8a366959f5cc..7aca13a54a3a 100644 --- a/Documentation/infiniband/user_mad.txt +++ b/Documentation/infiniband/user_mad.txt @@ -26,6 +26,11 @@ Creating MAD agents ioctl. Also, all agents registered through a file descriptor will be unregistered when the descriptor is closed. + 2014 -- a new registration ioctl is now provided which allows additional + fields to be provided during registration. + Users of this registration call are implicitly setting the use of + pkey_index (see below). + Receiving MADs MADs are received using read(). The receive side now supports @@ -104,10 +109,10 @@ P_Key Index Handling The old ib_umad interface did not allow setting the P_Key index for MADs that are sent and did not provide a way for obtaining the P_Key index of received MADs. A new layout for struct ib_user_mad_hdr - with a pkey_index member has been defined; however, to preserve - binary compatibility with older applications, this new layout will - not be used unless the IB_USER_MAD_ENABLE_PKEY ioctl is called - before a file descriptor is used for anything else. + with a pkey_index member has been defined; however, to preserve binary + compatibility with older applications, this new layout will not be used + unless one of IB_USER_MAD_ENABLE_PKEY or IB_USER_MAD_REGISTER_AGENT2 ioctl's + are called before a file descriptor is used for anything else. In September 2008, the IB_USER_MAD_ABI_VERSION will be incremented to 6, the new layout of struct ib_user_mad_hdr will be used by diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c index 8e32c5abd09d..f6d29614cb01 100644 --- a/drivers/infiniband/core/agent.c +++ b/drivers/infiniband/core/agent.c @@ -161,7 +161,7 @@ int ib_agent_port_open(struct ib_device *device, int port_num) port_priv->agent[0] = ib_register_mad_agent(device, port_num, IB_QPT_SMI, NULL, 0, &agent_send_handler, - NULL, NULL); + NULL, NULL, 0); if (IS_ERR(port_priv->agent[0])) { ret = PTR_ERR(port_priv->agent[0]); goto error2; @@ -172,7 +172,7 @@ int ib_agent_port_open(struct ib_device *device, int port_num) port_priv->agent[1] = ib_register_mad_agent(device, port_num, IB_QPT_GSI, NULL, 0, &agent_send_handler, - NULL, NULL); + NULL, NULL, 0); if (IS_ERR(port_priv->agent[1])) { ret = PTR_ERR(port_priv->agent[1]); goto error3; diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index c3239170d8b7..e28a494e2a3a 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -3753,7 +3753,7 @@ static void cm_add_one(struct ib_device *ib_device) struct cm_port *port; struct ib_mad_reg_req reg_req = { .mgmt_class = IB_MGMT_CLASS_CM, - .mgmt_class_version = IB_CM_CLASS_VERSION + .mgmt_class_version = IB_CM_CLASS_VERSION, }; struct ib_port_modify port_modify = { .set_port_cap_mask = IB_PORT_CM_SUP @@ -3801,7 +3801,8 @@ static void cm_add_one(struct ib_device *ib_device) 0, cm_send_handler, cm_recv_handler, - port); + port, + 0); if (IS_ERR(port->mad_agent)) goto error2; diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index f46995d4f54b..988bbda67952 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -198,7 +198,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, u8 rmpp_version, ib_mad_send_handler send_handler, ib_mad_recv_handler recv_handler, - void *context) + void *context, + u32 registration_flags) { struct ib_mad_port_private *port_priv; struct ib_mad_agent *ret = ERR_PTR(-EINVAL); @@ -359,6 +360,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, mad_agent_priv->agent.context = context; mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp; mad_agent_priv->agent.port_num = port_num; + mad_agent_priv->agent.flags = registration_flags; spin_lock_init(&mad_agent_priv->lock); INIT_LIST_HEAD(&mad_agent_priv->send_list); INIT_LIST_HEAD(&mad_agent_priv->wait_list); diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index 233eaf541f55..c38f030f0dc9 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -1184,7 +1184,7 @@ static void ib_sa_add_one(struct ib_device *device) sa_dev->port[i].agent = ib_register_mad_agent(device, i + s, IB_QPT_GSI, NULL, 0, send_handler, - recv_handler, sa_dev); + recv_handler, sa_dev, 0); if (IS_ERR(sa_dev->port[i].agent)) goto err; diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index ea90a3ebb9c2..11af1c61c135 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -647,6 +647,7 @@ static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg, found: if (ureq.mgmt_class) { + memset(&req, 0, sizeof(req)); req.mgmt_class = ureq.mgmt_class; req.mgmt_class_version = ureq.mgmt_class_version; memcpy(req.oui, ureq.oui, sizeof req.oui); @@ -667,7 +668,7 @@ found: ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI, ureq.mgmt_class ? &req : NULL, ureq.rmpp_version, - send_handler, recv_handler, file); + send_handler, recv_handler, file, 0); if (IS_ERR(agent)) { ret = PTR_ERR(agent); agent = NULL; @@ -705,6 +706,119 @@ out: return ret; } +static int ib_umad_reg_agent2(struct ib_umad_file *file, void __user *arg) +{ + struct ib_user_mad_reg_req2 ureq; + struct ib_mad_reg_req req; + struct ib_mad_agent *agent = NULL; + int agent_id; + int ret; + + mutex_lock(&file->port->file_mutex); + mutex_lock(&file->mutex); + + if (!file->port->ib_dev) { + dev_notice(file->port->dev, + "ib_umad_reg_agent2: invalid device\n"); + ret = -EPIPE; + goto out; + } + + if (copy_from_user(&ureq, arg, sizeof(ureq))) { + ret = -EFAULT; + goto out; + } + + if (ureq.qpn != 0 && ureq.qpn != 1) { + dev_notice(file->port->dev, + "ib_umad_reg_agent2: invalid QPN %d specified\n", + ureq.qpn); + ret = -EINVAL; + goto out; + } + + if (ureq.flags & ~IB_USER_MAD_REG_FLAGS_CAP) { + dev_notice(file->port->dev, + "ib_umad_reg_agent2 failed: invalid registration flags specified 0x%x; supported 0x%x\n", + ureq.flags, IB_USER_MAD_REG_FLAGS_CAP); + ret = -EINVAL; + + if (put_user((u32)IB_USER_MAD_REG_FLAGS_CAP, + (u32 __user *) (arg + offsetof(struct + ib_user_mad_reg_req2, flags)))) + ret = -EFAULT; + + goto out; + } + + for (agent_id = 0; agent_id < IB_UMAD_MAX_AGENTS; ++agent_id) + if (!__get_agent(file, agent_id)) + goto found; + + dev_notice(file->port->dev, + "ib_umad_reg_agent2: Max Agents (%u) reached\n", + IB_UMAD_MAX_AGENTS); + ret = -ENOMEM; + goto out; + +found: + if (ureq.mgmt_class) { + memset(&req, 0, sizeof(req)); + req.mgmt_class = ureq.mgmt_class; + req.mgmt_class_version = ureq.mgmt_class_version; + if (ureq.oui & 0xff000000) { + dev_notice(file->port->dev, + "ib_umad_reg_agent2 failed: oui invalid 0x%08x\n", + ureq.oui); + ret = -EINVAL; + goto out; + } + req.oui[2] = ureq.oui & 0x0000ff; + req.oui[1] = (ureq.oui & 0x00ff00) >> 8; + req.oui[0] = (ureq.oui & 0xff0000) >> 16; + memcpy(req.method_mask, ureq.method_mask, + sizeof(req.method_mask)); + } + + agent = ib_register_mad_agent(file->port->ib_dev, file->port->port_num, + ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI, + ureq.mgmt_class ? &req : NULL, + ureq.rmpp_version, + send_handler, recv_handler, file, + ureq.flags); + if (IS_ERR(agent)) { + ret = PTR_ERR(agent); + agent = NULL; + goto out; + } + + if (put_user(agent_id, + (u32 __user *)(arg + + offsetof(struct ib_user_mad_reg_req2, id)))) { + ret = -EFAULT; + goto out; + } + + if (!file->already_used) { + file->already_used = 1; + file->use_pkey_index = 1; + } + + file->agent[agent_id] = agent; + ret = 0; + +out: + mutex_unlock(&file->mutex); + + if (ret && agent) + ib_unregister_mad_agent(agent); + + mutex_unlock(&file->port->file_mutex); + + return ret; +} + + static int ib_umad_unreg_agent(struct ib_umad_file *file, u32 __user *arg) { struct ib_mad_agent *agent = NULL; @@ -760,6 +874,8 @@ static long ib_umad_ioctl(struct file *filp, unsigned int cmd, return ib_umad_unreg_agent(filp->private_data, (__u32 __user *) arg); case IB_USER_MAD_ENABLE_PKEY: return ib_umad_enable_pkey(filp->private_data); + case IB_USER_MAD_REGISTER_AGENT2: + return ib_umad_reg_agent2(filp->private_data, (void __user *) arg); default: return -ENOIOCTLCMD; } @@ -776,6 +892,8 @@ static long ib_umad_compat_ioctl(struct file *filp, unsigned int cmd, return ib_umad_unreg_agent(filp->private_data, compat_ptr(arg)); case IB_USER_MAD_ENABLE_PKEY: return ib_umad_enable_pkey(filp->private_data); + case IB_USER_MAD_REGISTER_AGENT2: + return ib_umad_reg_agent2(filp->private_data, compat_ptr(arg)); default: return -ENOIOCTLCMD; } diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 287ad0564acd..82a7dd87089b 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -891,7 +891,7 @@ int mlx4_ib_mad_init(struct mlx4_ib_dev *dev) agent = ib_register_mad_agent(&dev->ib_dev, p + 1, q ? IB_QPT_GSI : IB_QPT_SMI, NULL, 0, send_handler, - NULL, NULL); + NULL, NULL, 0); if (IS_ERR(agent)) { ret = PTR_ERR(agent); goto err; diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index b6f7f457fc55..8881fa376e06 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -294,7 +294,7 @@ int mthca_create_agents(struct mthca_dev *dev) agent = ib_register_mad_agent(&dev->ib_dev, p + 1, q ? IB_QPT_GSI : IB_QPT_SMI, NULL, 0, send_handler, - NULL, NULL); + NULL, NULL, 0); if (IS_ERR(agent)) { ret = PTR_ERR(agent); goto err; diff --git a/drivers/infiniband/hw/qib/qib_mad.c b/drivers/infiniband/hw/qib/qib_mad.c index 22c720e5740d..636be117b578 100644 --- a/drivers/infiniband/hw/qib/qib_mad.c +++ b/drivers/infiniband/hw/qib/qib_mad.c @@ -2476,7 +2476,7 @@ int qib_create_agents(struct qib_ibdev *dev) ibp = &dd->pport[p].ibport_data; agent = ib_register_mad_agent(&dev->ibdev, p + 1, IB_QPT_SMI, NULL, 0, send_handler, - NULL, NULL); + NULL, NULL, 0); if (IS_ERR(agent)) { ret = PTR_ERR(agent); goto err; diff --git a/drivers/infiniband/ulp/srpt/ib_srpt.c b/drivers/infiniband/ulp/srpt/ib_srpt.c index fe09f2788b15..8a8311e35cbd 100644 --- a/drivers/infiniband/ulp/srpt/ib_srpt.c +++ b/drivers/infiniband/ulp/srpt/ib_srpt.c @@ -563,7 +563,7 @@ static int srpt_refresh_port(struct srpt_port *sport) ®_req, 0, srpt_mad_send_handler, srpt_mad_recv_handler, - sport); + sport, 0); if (IS_ERR(sport->mad_agent)) { ret = PTR_ERR(sport->mad_agent); sport->mad_agent = NULL; diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h index 3d81b90cc315..876f497f8b0c 100644 --- a/include/rdma/ib_mad.h +++ b/include/rdma/ib_mad.h @@ -355,6 +355,7 @@ typedef void (*ib_mad_recv_handler)(struct ib_mad_agent *mad_agent, * @hi_tid: Access layer assigned transaction ID for this client. * Unsolicited MADs sent by this client will have the upper 32-bits * of their TID set to this value. + * @flags: registration flags * @port_num: Port number on which QP is registered * @rmpp_version: If set, indicates the RMPP version used by this agent. */ @@ -367,6 +368,7 @@ struct ib_mad_agent { ib_mad_snoop_handler snoop_handler; void *context; u32 hi_tid; + u32 flags; u8 port_num; u8 rmpp_version; }; @@ -426,6 +428,7 @@ struct ib_mad_recv_wc { * in the range from 0x30 to 0x4f. Otherwise not used. * @method_mask: The caller will receive unsolicited MADs for any method * where @method_mask = 1. + * */ struct ib_mad_reg_req { u8 mgmt_class; @@ -451,6 +454,7 @@ struct ib_mad_reg_req { * @recv_handler: The completion callback routine invoked for a received * MAD. * @context: User specified context associated with the registration. + * @registration_flags: Registration flags to set for this agent */ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, u8 port_num, @@ -459,7 +463,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, u8 rmpp_version, ib_mad_send_handler send_handler, ib_mad_recv_handler recv_handler, - void *context); + void *context, + u32 registration_flags); enum ib_mad_snoop_flags { /*IB_MAD_SNOOP_POSTED_SENDS = 1,*/ diff --git a/include/uapi/rdma/ib_user_mad.h b/include/uapi/rdma/ib_user_mad.h index d6fce1cbdb90..c00b081dc10e 100644 --- a/include/uapi/rdma/ib_user_mad.h +++ b/include/uapi/rdma/ib_user_mad.h @@ -191,6 +191,42 @@ struct ib_user_mad_reg_req { __u8 rmpp_version; }; +/** + * ib_user_mad_reg_req2 - MAD registration request + * + * @id - Set by the _kernel_; used by userspace to identify the + * registered agent in future requests. + * @qpn - Queue pair number; must be 0 or 1. + * @mgmt_class - Indicates which management class of MADs should be + * receive by the caller. This field is only required if + * the user wishes to receive unsolicited MADs, otherwise + * it should be 0. + * @mgmt_class_version - Indicates which version of MADs for the given + * management class to receive. + * @res - Ignored. + * @flags - additional registration flags; Must be in the set of + * flags defined in IB_USER_MAD_REG_FLAGS_CAP + * @method_mask - The caller wishes to receive unsolicited MADs for the + * methods whose bit(s) is(are) set. + * @oui - Indicates IEEE OUI to use when mgmt_class is a vendor + * class in the range from 0x30 to 0x4f. Otherwise not + * used. + * @rmpp_version - If set, indicates the RMPP version to use. + */ +#define IB_USER_MAD_REG_FLAGS_CAP (0) +struct ib_user_mad_reg_req2 { + __u32 id; + __u32 qpn; + __u8 mgmt_class; + __u8 mgmt_class_version; + __u16 res; + __u32 flags; + __u64 method_mask[2]; + __u32 oui; + __u8 rmpp_version; + __u8 reserved[3]; +}; + #define IB_IOCTL_MAGIC 0x1b #define IB_USER_MAD_REGISTER_AGENT _IOWR(IB_IOCTL_MAGIC, 1, \ @@ -200,4 +236,7 @@ struct ib_user_mad_reg_req { #define IB_USER_MAD_ENABLE_PKEY _IO(IB_IOCTL_MAGIC, 3) +#define IB_USER_MAD_REGISTER_AGENT2 _IOWR(IB_IOCTL_MAGIC, 4, \ + struct ib_user_mad_reg_req2) + #endif /* IB_USER_MAD_H */ -- cgit v1.2.3 From 1471cb6ca67990a306500e69e52ffb28c93ccbbc Mon Sep 17 00:00:00 2001 From: Ira Weiny Date: Fri, 8 Aug 2014 19:00:56 -0400 Subject: IB/mad: Add user space RMPP support Using the new registration mechanism, define a flag that indicates the user wishes to process RMPP messages in user space rather than have the kernel process them. Signed-off-by: Ira Weiny Signed-off-by: Roland Dreier --- drivers/infiniband/core/mad.c | 74 ++++++++++++++++++++++++++------------ drivers/infiniband/core/user_mad.c | 34 +++++++++++------- include/rdma/ib_mad.h | 11 ++++++ include/uapi/rdma/ib_user_mad.h | 5 ++- 4 files changed, 88 insertions(+), 36 deletions(-) (limited to 'include') diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 988bbda67952..74c30f4c557e 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -283,6 +283,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, goto error1; } } + /* Make sure class supplied is consistent with QP type */ if (qp_type == IB_QPT_SMI) { if ((mad_reg_req->mgmt_class != @@ -309,6 +310,8 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, /* No registration request supplied */ if (!send_handler) goto error1; + if (registration_flags & IB_MAD_USER_RMPP) + goto error1; } /* Validate device and port */ @@ -907,6 +910,12 @@ static int alloc_send_rmpp_list(struct ib_mad_send_wr_private *send_wr, return 0; } +int ib_mad_kernel_rmpp_agent(struct ib_mad_agent *agent) +{ + return agent->rmpp_version && !(agent->flags & IB_MAD_USER_RMPP); +} +EXPORT_SYMBOL(ib_mad_kernel_rmpp_agent); + struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent, u32 remote_qpn, u16 pkey_index, int rmpp_active, @@ -923,10 +932,12 @@ struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent, pad = get_pad_size(hdr_len, data_len); message_size = hdr_len + data_len + pad; - if ((!mad_agent->rmpp_version && - (rmpp_active || message_size > sizeof(struct ib_mad))) || - (!rmpp_active && message_size > sizeof(struct ib_mad))) - return ERR_PTR(-EINVAL); + if (ib_mad_kernel_rmpp_agent(mad_agent)) { + if (!rmpp_active && message_size > sizeof(struct ib_mad)) + return ERR_PTR(-EINVAL); + } else + if (rmpp_active || message_size > sizeof(struct ib_mad)) + return ERR_PTR(-EINVAL); size = rmpp_active ? hdr_len : sizeof(struct ib_mad); buf = kzalloc(sizeof *mad_send_wr + size, gfp_mask); @@ -1180,7 +1191,7 @@ int ib_post_send_mad(struct ib_mad_send_buf *send_buf, &mad_agent_priv->send_list); spin_unlock_irqrestore(&mad_agent_priv->lock, flags); - if (mad_agent_priv->agent.rmpp_version) { + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { ret = ib_send_rmpp_mad(mad_send_wr); if (ret >= 0 && ret != IB_RMPP_RESULT_CONSUMED) ret = ib_send_mad(mad_send_wr); @@ -1730,6 +1741,7 @@ static int is_data_mad(struct ib_mad_agent_private *mad_agent_priv, rmpp_mad = (struct ib_rmpp_mad *)mad_hdr; return !mad_agent_priv->agent.rmpp_version || + !ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent) || !(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & IB_MGMT_RMPP_FLAG_ACTIVE) || (rmpp_mad->rmpp_hdr.rmpp_type == IB_MGMT_RMPP_TYPE_DATA); @@ -1857,7 +1869,7 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, INIT_LIST_HEAD(&mad_recv_wc->rmpp_list); list_add(&mad_recv_wc->recv_buf.list, &mad_recv_wc->rmpp_list); - if (mad_agent_priv->agent.rmpp_version) { + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { mad_recv_wc = ib_process_rmpp_recv_wc(mad_agent_priv, mad_recv_wc); if (!mad_recv_wc) { @@ -1872,23 +1884,39 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, mad_send_wr = ib_find_send_mad(mad_agent_priv, mad_recv_wc); if (!mad_send_wr) { spin_unlock_irqrestore(&mad_agent_priv->lock, flags); - ib_free_recv_mad(mad_recv_wc); - deref_mad_agent(mad_agent_priv); - return; - } - ib_mark_mad_done(mad_send_wr); - spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + if (!ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent) + && ib_is_mad_class_rmpp(mad_recv_wc->recv_buf.mad->mad_hdr.mgmt_class) + && (ib_get_rmpp_flags(&((struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad)->rmpp_hdr) + & IB_MGMT_RMPP_FLAG_ACTIVE)) { + /* user rmpp is in effect + * and this is an active RMPP MAD + */ + mad_recv_wc->wc->wr_id = 0; + mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, + mad_recv_wc); + atomic_dec(&mad_agent_priv->refcount); + } else { + /* not user rmpp, revert to normal behavior and + * drop the mad */ + ib_free_recv_mad(mad_recv_wc); + deref_mad_agent(mad_agent_priv); + return; + } + } else { + ib_mark_mad_done(mad_send_wr); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); - /* Defined behavior is to complete response before request */ - mad_recv_wc->wc->wr_id = (unsigned long) &mad_send_wr->send_buf; - mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, - mad_recv_wc); - atomic_dec(&mad_agent_priv->refcount); + /* Defined behavior is to complete response before request */ + mad_recv_wc->wc->wr_id = (unsigned long) &mad_send_wr->send_buf; + mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, + mad_recv_wc); + atomic_dec(&mad_agent_priv->refcount); - mad_send_wc.status = IB_WC_SUCCESS; - mad_send_wc.vendor_err = 0; - mad_send_wc.send_buf = &mad_send_wr->send_buf; - ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc); + mad_send_wc.status = IB_WC_SUCCESS; + mad_send_wc.vendor_err = 0; + mad_send_wc.send_buf = &mad_send_wr->send_buf; + ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc); + } } else { mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, mad_recv_wc); @@ -2128,7 +2156,7 @@ void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr, mad_agent_priv = mad_send_wr->mad_agent_priv; spin_lock_irqsave(&mad_agent_priv->lock, flags); - if (mad_agent_priv->agent.rmpp_version) { + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { ret = ib_process_rmpp_send_wc(mad_send_wr, mad_send_wc); if (ret == IB_RMPP_RESULT_CONSUMED) goto done; @@ -2524,7 +2552,7 @@ static int retry_send(struct ib_mad_send_wr_private *mad_send_wr) mad_send_wr->timeout = msecs_to_jiffies(mad_send_wr->send_buf.timeout_ms); - if (mad_send_wr->mad_agent_priv->agent.rmpp_version) { + if (ib_mad_kernel_rmpp_agent(&mad_send_wr->mad_agent_priv->agent)) { ret = ib_retry_rmpp(mad_send_wr); switch (ret) { case IB_RMPP_RESULT_UNHANDLED: diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index 11af1c61c135..928cdd20e2d1 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -506,13 +506,15 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf, rmpp_mad = (struct ib_rmpp_mad *) packet->mad.data; hdr_len = ib_get_mad_data_offset(rmpp_mad->mad_hdr.mgmt_class); - if (!ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class)) { - copy_offset = IB_MGMT_MAD_HDR; - rmpp_active = 0; - } else { + + if (ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class) + && ib_mad_kernel_rmpp_agent(agent)) { copy_offset = IB_MGMT_RMPP_HDR; rmpp_active = ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & - IB_MGMT_RMPP_FLAG_ACTIVE; + IB_MGMT_RMPP_FLAG_ACTIVE; + } else { + copy_offset = IB_MGMT_MAD_HDR; + rmpp_active = 0; } data_len = count - hdr_size(file) - hdr_len; @@ -558,14 +560,22 @@ static ssize_t ib_umad_write(struct file *filp, const char __user *buf, rmpp_mad->mad_hdr.tid = *tid; } - spin_lock_irq(&file->send_lock); - ret = is_duplicate(file, packet); - if (!ret) + if (!ib_mad_kernel_rmpp_agent(agent) + && ib_is_mad_class_rmpp(rmpp_mad->mad_hdr.mgmt_class) + && (ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & IB_MGMT_RMPP_FLAG_ACTIVE)) { + spin_lock_irq(&file->send_lock); list_add_tail(&packet->list, &file->send_list); - spin_unlock_irq(&file->send_lock); - if (ret) { - ret = -EINVAL; - goto err_msg; + spin_unlock_irq(&file->send_lock); + } else { + spin_lock_irq(&file->send_lock); + ret = is_duplicate(file, packet); + if (!ret) + list_add_tail(&packet->list, &file->send_list); + spin_unlock_irq(&file->send_lock); + if (ret) { + ret = -EINVAL; + goto err_msg; + } } ret = ib_post_send_mad(packet->msg, NULL); diff --git a/include/rdma/ib_mad.h b/include/rdma/ib_mad.h index 876f497f8b0c..9bb99e983f58 100644 --- a/include/rdma/ib_mad.h +++ b/include/rdma/ib_mad.h @@ -40,6 +40,7 @@ #include #include +#include /* Management base version */ #define IB_MGMT_BASE_VERSION 1 @@ -359,6 +360,9 @@ typedef void (*ib_mad_recv_handler)(struct ib_mad_agent *mad_agent, * @port_num: Port number on which QP is registered * @rmpp_version: If set, indicates the RMPP version used by this agent. */ +enum { + IB_MAD_USER_RMPP = IB_USER_MAD_USER_RMPP, +}; struct ib_mad_agent { struct ib_device *device; struct ib_qp *qp; @@ -666,4 +670,11 @@ void *ib_get_rmpp_segment(struct ib_mad_send_buf *send_buf, int seg_num); */ void ib_free_send_mad(struct ib_mad_send_buf *send_buf); +/** + * ib_mad_kernel_rmpp_agent - Returns if the agent is performing RMPP. + * @agent: the agent in question + * @return: true if agent is performing rmpp, false otherwise. + */ +int ib_mad_kernel_rmpp_agent(struct ib_mad_agent *agent); + #endif /* IB_MAD_H */ diff --git a/include/uapi/rdma/ib_user_mad.h b/include/uapi/rdma/ib_user_mad.h index c00b081dc10e..09f809f323ea 100644 --- a/include/uapi/rdma/ib_user_mad.h +++ b/include/uapi/rdma/ib_user_mad.h @@ -213,7 +213,10 @@ struct ib_user_mad_reg_req { * used. * @rmpp_version - If set, indicates the RMPP version to use. */ -#define IB_USER_MAD_REG_FLAGS_CAP (0) +enum { + IB_USER_MAD_USER_RMPP = (1 << 0), +}; +#define IB_USER_MAD_REG_FLAGS_CAP (IB_USER_MAD_USER_RMPP) struct ib_user_mad_reg_req2 { __u32 id; __u32 qpn; -- cgit v1.2.3 From db1044d458a287c18c4d413adc4ad12e92e253b5 Mon Sep 17 00:00:00 2001 From: Doug Ledford Date: Tue, 12 Aug 2014 19:20:11 -0400 Subject: RDMA/uapi: Include socket.h in rdma_user_cm.h added struct sockaddr_storage to rdma_user_cm.h without also adding an include for linux/socket.h to make sure it is defined. Systemtap needs the header files to build standalone and cannot rely on other files to pre-include other headers, so add linux/socket.h to the list of includes in this file. Fixes: ee7aed4528f ("RDMA/ucma: Support querying for AF_IB addresses") Signed-off-by: Doug Ledford Signed-off-by: Roland Dreier --- include/uapi/rdma/rdma_user_cm.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/uapi/rdma/rdma_user_cm.h b/include/uapi/rdma/rdma_user_cm.h index 99b80abf360a..3066718eb120 100644 --- a/include/uapi/rdma/rdma_user_cm.h +++ b/include/uapi/rdma/rdma_user_cm.h @@ -34,6 +34,7 @@ #define RDMA_USER_CM_H #include +#include #include #include #include -- cgit v1.2.3