From fd75c789abf7948e16fe50917a6acb809927719a Mon Sep 17 00:00:00 2001 From: Nir Muchtar Date: Fri, 20 May 2011 11:46:10 -0700 Subject: RDMA: Add error handling to ib_core_init() Fail RDMA midlayer initialization if sysfs setup fails. Signed-off-by: Nir Muchtar Signed-off-by: Roland Dreier --- drivers/infiniband/core/device.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index f793bf2f5da7..46a7a3febc12 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -725,16 +725,24 @@ static int __init ib_core_init(void) return -ENOMEM; ret = ib_sysfs_setup(); - if (ret) + if (ret) { printk(KERN_WARNING "Couldn't create InfiniBand device class\n"); + goto err; + } ret = ib_cache_setup(); if (ret) { printk(KERN_WARNING "Couldn't set up InfiniBand P_Key/GID cache\n"); - ib_sysfs_cleanup(); - destroy_workqueue(ib_wq); + goto err_sysfs; } + return 0; + +err_sysfs: + ib_sysfs_cleanup(); + +err: + destroy_workqueue(ib_wq); return ret; } -- cgit v1.2.3 From b2cbae2c248776d81cc265ff7d48405b6a4cc463 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Fri, 20 May 2011 11:46:11 -0700 Subject: RDMA: Add netlink infrastructure Add basic RDMA netlink infrastructure that allows for registration of RDMA clients for which data is to be exported and supplies message construction callbacks. Signed-off-by: Nir Muchtar [ Reorganize a few things, add CONFIG_NET dependency. - Roland ] Signed-off-by: Roland Dreier --- drivers/infiniband/Kconfig | 1 + drivers/infiniband/core/Makefile | 2 +- drivers/infiniband/core/device.c | 13 ++- drivers/infiniband/core/netlink.c | 190 ++++++++++++++++++++++++++++++++++++++ drivers/infiniband/hw/qib/Kconfig | 2 +- include/linux/netlink.h | 1 + include/rdma/rdma_netlink.h | 64 +++++++++++++ 7 files changed, 270 insertions(+), 3 deletions(-) create mode 100644 drivers/infiniband/core/netlink.c create mode 100644 include/rdma/rdma_netlink.h (limited to 'drivers') diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig index 6e35eccc9caa..0f9a84c1046a 100644 --- a/drivers/infiniband/Kconfig +++ b/drivers/infiniband/Kconfig @@ -2,6 +2,7 @@ menuconfig INFINIBAND tristate "InfiniBand support" depends on PCI || BROKEN depends on HAS_IOMEM + depends on NET ---help--- Core support for InfiniBand (IB). Make sure to also select any protocols you wish to use as well as drivers for your diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile index cb1ab3ea4998..c8bbaef1becb 100644 --- a/drivers/infiniband/core/Makefile +++ b/drivers/infiniband/core/Makefile @@ -8,7 +8,7 @@ obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o ib_ucm.o \ $(user_access-y) ib_core-y := packer.o ud_header.o verbs.o sysfs.o \ - device.o fmr_pool.o cache.o + device.o fmr_pool.o cache.o netlink.o ib_core-$(CONFIG_INFINIBAND_USER_MEM) += umem.o ib_mad-y := mad.o smi.o agent.o mad_rmpp.o diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 46a7a3febc12..4007f721d25d 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -38,6 +38,7 @@ #include #include #include +#include #include "core_priv.h" @@ -730,14 +731,23 @@ static int __init ib_core_init(void) goto err; } + ret = ibnl_init(); + if (ret) { + printk(KERN_WARNING "Couldn't init IB netlink interface\n"); + goto err_sysfs; + } + ret = ib_cache_setup(); if (ret) { printk(KERN_WARNING "Couldn't set up InfiniBand P_Key/GID cache\n"); - goto err_sysfs; + goto err_nl; } return 0; +err_nl: + ibnl_cleanup(); + err_sysfs: ib_sysfs_cleanup(); @@ -749,6 +759,7 @@ err: static void __exit ib_core_cleanup(void) { ib_cache_cleanup(); + ibnl_cleanup(); ib_sysfs_cleanup(); /* Make sure that any pending umem accounting work is done. */ destroy_workqueue(ib_wq); diff --git a/drivers/infiniband/core/netlink.c b/drivers/infiniband/core/netlink.c new file mode 100644 index 000000000000..4a5abaf0a25c --- /dev/null +++ b/drivers/infiniband/core/netlink.c @@ -0,0 +1,190 @@ +/* + * Copyright (c) 2010 Voltaire Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#define pr_fmt(fmt) "%s:%s: " fmt, KBUILD_MODNAME, __func__ + +#include +#include +#include +#include + +struct ibnl_client { + struct list_head list; + int index; + int nops; + const struct ibnl_client_cbs *cb_table; +}; + +static DEFINE_MUTEX(ibnl_mutex); +static struct sock *nls; +static LIST_HEAD(client_list); + +int ibnl_add_client(int index, int nops, + const struct ibnl_client_cbs cb_table[]) +{ + struct ibnl_client *cur; + struct ibnl_client *nl_client; + + nl_client = kmalloc(sizeof *nl_client, GFP_KERNEL); + if (!nl_client) + return -ENOMEM; + + nl_client->index = index; + nl_client->nops = nops; + nl_client->cb_table = cb_table; + + mutex_lock(&ibnl_mutex); + + list_for_each_entry(cur, &client_list, list) { + if (cur->index == index) { + pr_warn("Client for %d already exists\n", index); + mutex_unlock(&ibnl_mutex); + kfree(nl_client); + return -EINVAL; + } + } + + list_add_tail(&nl_client->list, &client_list); + + mutex_unlock(&ibnl_mutex); + + return 0; +} +EXPORT_SYMBOL(ibnl_add_client); + +int ibnl_remove_client(int index) +{ + struct ibnl_client *cur, *next; + + mutex_lock(&ibnl_mutex); + list_for_each_entry_safe(cur, next, &client_list, list) { + if (cur->index == index) { + list_del(&(cur->list)); + mutex_unlock(&ibnl_mutex); + kfree(cur); + return 0; + } + } + pr_warn("Can't remove callback for client idx %d. Not found\n", index); + mutex_unlock(&ibnl_mutex); + + return -EINVAL; +} +EXPORT_SYMBOL(ibnl_remove_client); + +void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq, + int len, int client, int op) +{ + unsigned char *prev_tail; + + prev_tail = skb_tail_pointer(skb); + *nlh = NLMSG_NEW(skb, 0, seq, RDMA_NL_GET_TYPE(client, op), + len, NLM_F_MULTI); + (*nlh)->nlmsg_len = skb_tail_pointer(skb) - prev_tail; + return NLMSG_DATA(*nlh); + +nlmsg_failure: + nlmsg_trim(skb, prev_tail); + return NULL; +} +EXPORT_SYMBOL(ibnl_put_msg); + +int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh, + int len, void *data, int type) +{ + unsigned char *prev_tail; + + prev_tail = skb_tail_pointer(skb); + NLA_PUT(skb, type, len, data); + nlh->nlmsg_len += skb_tail_pointer(skb) - prev_tail; + return 0; + +nla_put_failure: + nlmsg_trim(skb, prev_tail - nlh->nlmsg_len); + return -EMSGSIZE; +} +EXPORT_SYMBOL(ibnl_put_attr); + +static int ibnl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) +{ + struct ibnl_client *client; + int type = nlh->nlmsg_type; + int index = RDMA_NL_GET_CLIENT(type); + int op = RDMA_NL_GET_OP(type); + + list_for_each_entry(client, &client_list, list) { + if (client->index == index) { + if (op < 0 || op >= client->nops || + !client->cb_table[RDMA_NL_GET_OP(op)].dump) + return -EINVAL; + return netlink_dump_start(nls, skb, nlh, + client->cb_table[op].dump, + NULL); + } + } + + pr_info("Index %d wasn't found in client list\n", index); + return -EINVAL; +} + +static void ibnl_rcv(struct sk_buff *skb) +{ + mutex_lock(&ibnl_mutex); + netlink_rcv_skb(skb, &ibnl_rcv_msg); + mutex_unlock(&ibnl_mutex); +} + +int __init ibnl_init(void) +{ + nls = netlink_kernel_create(&init_net, NETLINK_RDMA, 0, ibnl_rcv, + NULL, THIS_MODULE); + if (!nls) { + pr_warn("Failed to create netlink socket\n"); + return -ENOMEM; + } + + return 0; +} + +void ibnl_cleanup(void) +{ + struct ibnl_client *cur, *next; + + mutex_lock(&ibnl_mutex); + list_for_each_entry_safe(cur, next, &client_list, list) { + list_del(&(cur->list)); + kfree(cur); + } + mutex_unlock(&ibnl_mutex); + + netlink_kernel_release(nls); +} diff --git a/drivers/infiniband/hw/qib/Kconfig b/drivers/infiniband/hw/qib/Kconfig index 7c03a70c55a2..8349f9c5064c 100644 --- a/drivers/infiniband/hw/qib/Kconfig +++ b/drivers/infiniband/hw/qib/Kconfig @@ -1,6 +1,6 @@ config INFINIBAND_QIB tristate "QLogic PCIe HCA support" - depends on 64BIT && NET + depends on 64BIT ---help--- This is a low-level driver for QLogic PCIe QLE InfiniBand host channel adapters. This driver does not support the QLogic diff --git a/include/linux/netlink.h b/include/linux/netlink.h index 4c4ac3f3ce5a..a9dd89552f9c 100644 --- a/include/linux/netlink.h +++ b/include/linux/netlink.h @@ -24,6 +24,7 @@ /* leave room for NETLINK_DM (DM Events) */ #define NETLINK_SCSITRANSPORT 18 /* SCSI Transports */ #define NETLINK_ECRYPTFS 19 +#define NETLINK_RDMA 20 #define MAX_LINKS 32 diff --git a/include/rdma/rdma_netlink.h b/include/rdma/rdma_netlink.h new file mode 100644 index 000000000000..c983a193cecf --- /dev/null +++ b/include/rdma/rdma_netlink.h @@ -0,0 +1,64 @@ +#ifndef _RDMA_NETLINK_H +#define _RDMA_NETLINK_H + +#define RDMA_NL_GET_CLIENT(type) ((type & (((1 << 6) - 1) << 10)) >> 10) +#define RDMA_NL_GET_OP(type) (type & ((1 << 10) - 1)) +#define RDMA_NL_GET_TYPE(client, op) ((client << 10) + op) + +#ifdef __KERNEL__ + +#include + +struct ibnl_client_cbs { + int (*dump)(struct sk_buff *skb, struct netlink_callback *nlcb); +}; + +int ibnl_init(void); +void ibnl_cleanup(void); + +/** + * Add a a client to the list of IB netlink exporters. + * @index: Index of the added client + * @nops: Number of supported ops by the added client. + * @cb_table: A table for op->callback + * + * Returns 0 on success or a negative error code. + */ +int ibnl_add_client(int index, int nops, + const struct ibnl_client_cbs cb_table[]); + +/** + * Remove a client from IB netlink. + * @index: Index of the removed IB client. + * + * Returns 0 on success or a negative error code. + */ +int ibnl_remove_client(int index); + +/** + * Put a new message in a supplied skb. + * @skb: The netlink skb. + * @nlh: Pointer to put the header of the new netlink message. + * @seq: The message sequence number. + * @len: The requested message length to allocate. + * @client: Calling IB netlink client. + * @op: message content op. + * Returns the allocated buffer on success and NULL on failure. + */ +void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq, + int len, int client, int op); +/** + * Put a new attribute in a supplied skb. + * @skb: The netlink skb. + * @nlh: Header of the netlink message to append the attribute to. + * @len: The length of the attribute data. + * @data: The attribute data to put. + * @type: The attribute type. + * Returns the 0 and a negative error code on failure. + */ +int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh, + int len, void *data, int type); + +#endif /* __KERNEL__ */ + +#endif /* _RDMA_NETLINK_H */ -- cgit v1.2.3 From 04ea2f81973f55db715bfdac7dd258f8a8485a6d Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 23 May 2011 10:48:43 -0700 Subject: RDMA/ucma: Add .nodename/.mode to tell userspace where to create device node We want udev to create a device node under /dev/infiniband with permission 0666 for rdma_cm, so add that info to our struct miscdevice. Signed-off-by: Roland Dreier Acked-by: Sean Hefty --- drivers/infiniband/core/ucma.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index b3fa798525b2..3170899dab01 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -1338,9 +1338,11 @@ static const struct file_operations ucma_fops = { }; static struct miscdevice ucma_misc = { - .minor = MISC_DYNAMIC_MINOR, - .name = "rdma_cm", - .fops = &ucma_fops, + .minor = MISC_DYNAMIC_MINOR, + .name = "rdma_cm", + .nodename = "infiniband/rdma_cm", + .mode = 0666, + .fops = &ucma_fops, }; static ssize_t show_abi_version(struct device *dev, -- cgit v1.2.3 From 71c29bd5c23559c9cabcca3bfb3152ecc5efcd75 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 23 May 2011 11:10:05 -0700 Subject: IB/uverbs: Add devnode method to set path/mode We want udev to create a device node under /dev/infiniband with permission 0666 for uverbsX devices, so add a devnode method to set the appropriate info. Signed-off-by: Roland Dreier --- drivers/infiniband/core/uverbs_main.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index ec83e9fe387b..e49a85f8a44d 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -824,6 +824,12 @@ static void ib_uverbs_remove_one(struct ib_device *device) kfree(uverbs_dev); } +static char *uverbs_devnode(struct device *dev, mode_t *mode) +{ + *mode = 0666; + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + static int __init ib_uverbs_init(void) { int ret; @@ -842,6 +848,8 @@ static int __init ib_uverbs_init(void) goto out_chrdev; } + uverbs_class->devnode = uverbs_devnode; + ret = class_create_file(uverbs_class, &class_attr_abi_version.attr); if (ret) { printk(KERN_ERR "user_verbs: couldn't create abi_version attribute\n"); -- cgit v1.2.3 From c8367c4cd9de512d296fc557f121be62a43987f3 Mon Sep 17 00:00:00 2001 From: Ira Weiny Date: Thu, 19 May 2011 18:19:28 -0700 Subject: IB/mad: Return EPROTONOSUPPORT when an RDMA device lacks the QP required We had a script which was looping through the devices returned from ibstat and attempted to register a SMI agent on an ethernet device. This caused a kernel panic for IBoE devices that don't have QP0. Fix this by checking if the QP exists before using it. Signed-off-by: Ira Weiny Acked-by: Sean Hefty Signed-off-by: Roland Dreier --- drivers/infiniband/core/mad.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 822cfdcd9f78..b4d8672a3e4e 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -276,6 +276,13 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, goto error1; } + /* Verify the QP requested is supported. For example, Ethernet devices + * will not have QP0 */ + if (!port_priv->qp_info[qpn].qp) { + ret = ERR_PTR(-EPROTONOSUPPORT); + goto error1; + } + /* Allocate structures */ mad_agent_priv = kzalloc(sizeof *mad_agent_priv, GFP_KERNEL); if (!mad_agent_priv) { -- cgit v1.2.3 From c3af0980ce01dfb8e946990c3ae4c3727b1e9f26 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 23 May 2011 11:21:47 -0700 Subject: IB: Add devnode methods to cm_class and umad_class We want the ucmX, umadX and issmX device nodes to show up under /dev/infiniband, and additionally ucmX should have mode 0666. Add appropriate devnode methods to their class structs for this. Signed-off-by: Roland Dreier --- drivers/infiniband/core/cm.c | 8 ++++++++ drivers/infiniband/core/user_mad.c | 7 +++++++ 2 files changed, 15 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index f804e28e1ebb..f62f52fb9ece 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -3639,8 +3639,16 @@ static struct kobj_type cm_port_obj_type = { .release = cm_release_port_obj }; +static char *cm_devnode(struct device *dev, mode_t *mode) +{ + *mode = 0666; + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + struct class cm_class = { + .owner = THIS_MODULE, .name = "infiniband_cm", + .devnode = cm_devnode, }; EXPORT_SYMBOL(cm_class); diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c index cd1996d0ad08..8d261b6ea5fe 100644 --- a/drivers/infiniband/core/user_mad.c +++ b/drivers/infiniband/core/user_mad.c @@ -1176,6 +1176,11 @@ static void ib_umad_remove_one(struct ib_device *device) kref_put(&umad_dev->ref, ib_umad_release_dev); } +static char *umad_devnode(struct device *dev, mode_t *mode) +{ + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + static int __init ib_umad_init(void) { int ret; @@ -1194,6 +1199,8 @@ static int __init ib_umad_init(void) goto out_chrdev; } + umad_class->devnode = umad_devnode; + ret = class_create_file(umad_class, &class_attr_abi_version.attr); if (ret) { printk(KERN_ERR "user_mad: couldn't create abi_version attribute\n"); -- cgit v1.2.3 From 737b94eb41cb99250ccce9148ca411b55d4dc96a Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Mon, 23 May 2011 11:30:04 -0700 Subject: IB/srp: Fix integer -> pointer cast warnings Fix drivers/infiniband/ulp/srp/ib_srp.c: In function 'srp_handle_recv': drivers/infiniband/ulp/srp/ib_srp.c:1150: warning: cast to pointer from integer of different size drivers/infiniband/ulp/srp/ib_srp.c: In function 'srp_send_completion': drivers/infiniband/ulp/srp/ib_srp.c:1234: warning: cast to pointer from integer of different size by adding an intermediate cast to uintptr_t. Signed-off-by: Roland Dreier Acked-by: David Dillow --- drivers/infiniband/ulp/srp/ib_srp.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 376d640487d2..ee165fdcb596 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -1147,7 +1147,7 @@ static void srp_process_aer_req(struct srp_target_port *target, static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc) { struct ib_device *dev = target->srp_host->srp_dev->dev; - struct srp_iu *iu = (struct srp_iu *) wc->wr_id; + struct srp_iu *iu = (struct srp_iu *) (uintptr_t) wc->wr_id; int res; u8 opcode; @@ -1231,7 +1231,7 @@ static void srp_send_completion(struct ib_cq *cq, void *target_ptr) break; } - iu = (struct srp_iu *) wc.wr_id; + iu = (struct srp_iu *) (uintptr_t) wc.wr_id; list_add(&iu->list, &target->free_tx); } } -- cgit v1.2.3 From c337374bf23b88620bcc66a7a09f141cc640f548 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Fri, 20 May 2011 16:25:05 +0000 Subject: RDMA/cxgb4: Use completion objects for event blocking There exists a race condition when using wait_queue_head_t objects that are declared on the stack. This was being done in a few places where we are sending work requests to the FW and awaiting replies, but we don't have an endpoint structure with an embedded c4iw_wr_wait struct. So the code was allocating it locally on the stack. Bad design. The race is: 1) thread on cpuX declares the wait_queue_head_t on the stack, then posts a firmware WR with that wait object ptr as the cookie to be returned in the WR reply. This thread will proceed to block in wait_event_timeout() but before it does: 2) An interrupt runs on cpuY with the WR reply. fw6_msg() handles this and calls c4iw_wake_up(). c4iw_wake_up() sets the condition variable in the c4iw_wr_wait object to TRUE and will call wake_up(), but before it calls wake_up(): 3) The thread on cpuX calls c4iw_wait_for_reply(), which calls wait_event_timeout(). The wait_event_timeout() macro checks the condition variable and returns immediately since it is TRUE. So this thread never blocks/sleeps. The function then returns effectively deallocating the c4iw_wr_wait object that was on the stack. 4) So at this point cpuY has a pointer to the c4iw_wr_wait object that is no longer valid. Further its pointing to a stack frame that might now be in use by some other context/thread. So cpuY continues execution and calls wake_up() on a ptr to a wait object that as been effectively deallocated. This race, when it hits, can cause a crash in wake_up(), which I've seen under heavy stress. It can also corrupt the referenced stack which can cause any number of failures. The fix: Use struct completion, which supports on-stack declarations. Completions use a spinlock around setting the condition to true and the wake up so that steps 2 and 4 above are atomic and step 3 can never happen in-between. Signed-off-by: Steve Wise --- drivers/infiniband/hw/cxgb4/iw_cxgb4.h | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h index 35d2a5dd9bb4..4f045375c8e2 100644 --- a/drivers/infiniband/hw/cxgb4/iw_cxgb4.h +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h @@ -35,7 +35,7 @@ #include #include #include -#include +#include #include #include #include @@ -131,28 +131,21 @@ static inline int c4iw_num_stags(struct c4iw_rdev *rdev) #define C4IW_WR_TO (10*HZ) -enum { - REPLY_READY = 0, -}; - struct c4iw_wr_wait { - wait_queue_head_t wait; - unsigned long status; + struct completion completion; int ret; }; static inline void c4iw_init_wr_wait(struct c4iw_wr_wait *wr_waitp) { wr_waitp->ret = 0; - wr_waitp->status = 0; - init_waitqueue_head(&wr_waitp->wait); + init_completion(&wr_waitp->completion); } static inline void c4iw_wake_up(struct c4iw_wr_wait *wr_waitp, int ret) { wr_waitp->ret = ret; - set_bit(REPLY_READY, &wr_waitp->status); - wake_up(&wr_waitp->wait); + complete(&wr_waitp->completion); } static inline int c4iw_wait_for_reply(struct c4iw_rdev *rdev, @@ -164,8 +157,7 @@ static inline int c4iw_wait_for_reply(struct c4iw_rdev *rdev, int ret; do { - ret = wait_event_timeout(wr_waitp->wait, - test_and_clear_bit(REPLY_READY, &wr_waitp->status), to); + ret = wait_for_completion_timeout(&wr_waitp->completion, to); if (!ret) { printk(KERN_ERR MOD "%s - Device %s not responding - " "tid %u qpid %u\n", func, -- cgit v1.2.3 From 807838686eb9e40d73b8a3f2384881358f51fff0 Mon Sep 17 00:00:00 2001 From: Steve Wise Date: Fri, 13 May 2011 18:37:18 +0000 Subject: RDMA/cxgb3: Don't post zero-byte read if endpoint is going away tx_ack() wasn't checking the endpoint state and consequently would attempt to post the p2p 0B read on an endpoint/QP that is closing or aborting. This causes a NULL pointer dereference crash. Signed-off-by: Steve Wise Signed-off-by: Roland Dreier --- drivers/infiniband/hw/cxgb3/iwch_cm.c | 26 +++++++++++++++++--------- drivers/infiniband/hw/cxgb3/iwch_provider.h | 2 +- drivers/infiniband/hw/cxgb3/iwch_qp.c | 6 +++--- 3 files changed, 21 insertions(+), 13 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/cxgb3/iwch_cm.c b/drivers/infiniband/hw/cxgb3/iwch_cm.c index 3216bcad7e82..ad998c0b51e2 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_cm.c +++ b/drivers/infiniband/hw/cxgb3/iwch_cm.c @@ -913,7 +913,7 @@ static void process_mpa_reply(struct iwch_ep *ep, struct sk_buff *skb) goto err; if (peer2peer && iwch_rqes_posted(ep->com.qp) == 0) { - iwch_post_zb_read(ep->com.qp); + iwch_post_zb_read(ep); } goto out; @@ -1077,6 +1077,8 @@ static int tx_ack(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) struct iwch_ep *ep = ctx; struct cpl_wr_ack *hdr = cplhdr(skb); unsigned int credits = ntohs(hdr->credits); + unsigned long flags; + int post_zb = 0; PDBG("%s ep %p credits %u\n", __func__, ep, credits); @@ -1086,28 +1088,34 @@ static int tx_ack(struct t3cdev *tdev, struct sk_buff *skb, void *ctx) return CPL_RET_BUF_DONE; } + spin_lock_irqsave(&ep->com.lock, flags); BUG_ON(credits != 1); dst_confirm(ep->dst); if (!ep->mpa_skb) { PDBG("%s rdma_init wr_ack ep %p state %u\n", - __func__, ep, state_read(&ep->com)); + __func__, ep, ep->com.state); if (ep->mpa_attr.initiator) { PDBG("%s initiator ep %p state %u\n", - __func__, ep, state_read(&ep->com)); - if (peer2peer) - iwch_post_zb_read(ep->com.qp); + __func__, ep, ep->com.state); + if (peer2peer && ep->com.state == FPDU_MODE) + post_zb = 1; } else { PDBG("%s responder ep %p state %u\n", - __func__, ep, state_read(&ep->com)); - ep->com.rpl_done = 1; - wake_up(&ep->com.waitq); + __func__, ep, ep->com.state); + if (ep->com.state == MPA_REQ_RCVD) { + ep->com.rpl_done = 1; + wake_up(&ep->com.waitq); + } } } else { PDBG("%s lsm ack ep %p state %u freeing skb\n", - __func__, ep, state_read(&ep->com)); + __func__, ep, ep->com.state); kfree_skb(ep->mpa_skb); ep->mpa_skb = NULL; } + spin_unlock_irqrestore(&ep->com.lock, flags); + if (post_zb) + iwch_post_zb_read(ep); return CPL_RET_BUF_DONE; } diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.h b/drivers/infiniband/hw/cxgb3/iwch_provider.h index c5406da3f4cd..9a342c9b220d 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_provider.h +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.h @@ -332,7 +332,7 @@ int iwch_bind_mw(struct ib_qp *qp, struct ib_mw_bind *mw_bind); int iwch_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc); int iwch_post_terminate(struct iwch_qp *qhp, struct respQ_msg_t *rsp_msg); -int iwch_post_zb_read(struct iwch_qp *qhp); +int iwch_post_zb_read(struct iwch_ep *ep); int iwch_register_device(struct iwch_dev *dev); void iwch_unregister_device(struct iwch_dev *dev); void stop_read_rep_timer(struct iwch_qp *qhp); diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c index 1b4cd09f74dc..ecd313f359a4 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_qp.c +++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c @@ -738,7 +738,7 @@ static inline void build_term_codes(struct respQ_msg_t *rsp_msg, } } -int iwch_post_zb_read(struct iwch_qp *qhp) +int iwch_post_zb_read(struct iwch_ep *ep) { union t3_wr *wqe; struct sk_buff *skb; @@ -761,10 +761,10 @@ int iwch_post_zb_read(struct iwch_qp *qhp) wqe->read.local_len = cpu_to_be32(0); wqe->read.local_to = cpu_to_be64(1); wqe->send.wrh.op_seop_flags = cpu_to_be32(V_FW_RIWR_OP(T3_WR_READ)); - wqe->send.wrh.gen_tid_len = cpu_to_be32(V_FW_RIWR_TID(qhp->ep->hwtid)| + wqe->send.wrh.gen_tid_len = cpu_to_be32(V_FW_RIWR_TID(ep->hwtid)| V_FW_RIWR_LEN(flit_cnt)); skb->priority = CPL_PRIORITY_DATA; - return iwch_cxgb3_ofld_send(qhp->rhp->rdev.t3cdev_p, skb); + return iwch_cxgb3_ofld_send(ep->com.qp->rhp->rdev.t3cdev_p, skb); } /* -- cgit v1.2.3 From 52f81dbaf1378faf64c3ecea5129cebf826ef126 Mon Sep 17 00:00:00 2001 From: Liu Yuan Date: Fri, 1 Apr 2011 06:23:49 +0000 Subject: RDMA/nes: Add a check for strict_strtoul() It should check if strict_strtoul() succeeds before using 'wqm_quanta_value'. Signed-off-by: Liu Yuan [ Convert to kstrtoul() directly while we're here. - Roland ] Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c index 13de1192927c..2d668c69f6d9 100644 --- a/drivers/infiniband/hw/nes/nes.c +++ b/drivers/infiniband/hw/nes/nes.c @@ -1138,7 +1138,9 @@ static ssize_t nes_store_wqm_quanta(struct device_driver *ddp, u32 i = 0; struct nes_device *nesdev; - strict_strtoul(buf, 0, &wqm_quanta_value); + if (kstrtoul(buf, 0, &wqm_quanta_value) < 0) + return -EINVAL; + list_for_each_entry(nesdev, &nes_dev_list, list) { if (i == ee_flsh_adapter) { nesdev->nesadapter->wqm_quanta = wqm_quanta_value; -- cgit v1.2.3 From 550e5ca77e96989c5e19f60e017205b2bcc615a5 Mon Sep 17 00:00:00 2001 From: Nir Muchtar Date: Fri, 20 May 2011 11:46:11 -0700 Subject: RDMA/cma: Export enum cma_state in Move cma.c's internal definition of enum cma_state to enum rdma_cm_state in an exported header so that it can be exported via RDMA netlink. Signed-off-by: Nir Muchtar Signed-off-by: Roland Dreier --- drivers/infiniband/core/cma.c | 170 ++++++++++++++++++++---------------------- include/rdma/rdma_cm.h | 14 ++++ 2 files changed, 94 insertions(+), 90 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 99dde874fbbd..451d39e19cb4 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -89,20 +89,6 @@ struct cma_device { struct list_head id_list; }; -enum cma_state { - CMA_IDLE, - CMA_ADDR_QUERY, - CMA_ADDR_RESOLVED, - CMA_ROUTE_QUERY, - CMA_ROUTE_RESOLVED, - CMA_CONNECT, - CMA_DISCONNECT, - CMA_ADDR_BOUND, - CMA_LISTEN, - CMA_DEVICE_REMOVAL, - CMA_DESTROYING -}; - struct rdma_bind_list { struct idr *ps; struct hlist_head owners; @@ -126,7 +112,7 @@ struct rdma_id_private { struct list_head mc_list; int internal_id; - enum cma_state state; + enum rdma_cm_state state; spinlock_t lock; struct mutex qp_mutex; @@ -165,8 +151,8 @@ struct cma_multicast { struct cma_work { struct work_struct work; struct rdma_id_private *id; - enum cma_state old_state; - enum cma_state new_state; + enum rdma_cm_state old_state; + enum rdma_cm_state new_state; struct rdma_cm_event event; }; @@ -217,7 +203,7 @@ struct sdp_hah { #define CMA_VERSION 0x00 #define SDP_MAJ_VERSION 0x2 -static int cma_comp(struct rdma_id_private *id_priv, enum cma_state comp) +static int cma_comp(struct rdma_id_private *id_priv, enum rdma_cm_state comp) { unsigned long flags; int ret; @@ -229,7 +215,7 @@ static int cma_comp(struct rdma_id_private *id_priv, enum cma_state comp) } static int cma_comp_exch(struct rdma_id_private *id_priv, - enum cma_state comp, enum cma_state exch) + enum rdma_cm_state comp, enum rdma_cm_state exch) { unsigned long flags; int ret; @@ -241,11 +227,11 @@ static int cma_comp_exch(struct rdma_id_private *id_priv, return ret; } -static enum cma_state cma_exch(struct rdma_id_private *id_priv, - enum cma_state exch) +static enum rdma_cm_state cma_exch(struct rdma_id_private *id_priv, + enum rdma_cm_state exch) { unsigned long flags; - enum cma_state old; + enum rdma_cm_state old; spin_lock_irqsave(&id_priv->lock, flags); old = id_priv->state; @@ -413,7 +399,7 @@ static void cma_deref_id(struct rdma_id_private *id_priv) } static int cma_disable_callback(struct rdma_id_private *id_priv, - enum cma_state state) + enum rdma_cm_state state) { mutex_lock(&id_priv->handler_mutex); if (id_priv->state != state) { @@ -437,7 +423,7 @@ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, if (!id_priv) return ERR_PTR(-ENOMEM); - id_priv->state = CMA_IDLE; + id_priv->state = RDMA_CM_IDLE; id_priv->id.context = context; id_priv->id.event_handler = event_handler; id_priv->id.ps = ps; @@ -858,16 +844,16 @@ static void cma_cancel_listens(struct rdma_id_private *id_priv) } static void cma_cancel_operation(struct rdma_id_private *id_priv, - enum cma_state state) + enum rdma_cm_state state) { switch (state) { - case CMA_ADDR_QUERY: + case RDMA_CM_ADDR_QUERY: rdma_addr_cancel(&id_priv->id.route.addr.dev_addr); break; - case CMA_ROUTE_QUERY: + case RDMA_CM_ROUTE_QUERY: cma_cancel_route(id_priv); break; - case CMA_LISTEN: + case RDMA_CM_LISTEN: if (cma_any_addr((struct sockaddr *) &id_priv->id.route.addr.src_addr) && !id_priv->cma_dev) cma_cancel_listens(id_priv); @@ -918,10 +904,10 @@ static void cma_leave_mc_groups(struct rdma_id_private *id_priv) void rdma_destroy_id(struct rdma_cm_id *id) { struct rdma_id_private *id_priv; - enum cma_state state; + enum rdma_cm_state state; id_priv = container_of(id, struct rdma_id_private, id); - state = cma_exch(id_priv, CMA_DESTROYING); + state = cma_exch(id_priv, RDMA_CM_DESTROYING); cma_cancel_operation(id_priv, state); /* @@ -1015,9 +1001,9 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) int ret = 0; if ((ib_event->event != IB_CM_TIMEWAIT_EXIT && - cma_disable_callback(id_priv, CMA_CONNECT)) || + cma_disable_callback(id_priv, RDMA_CM_CONNECT)) || (ib_event->event == IB_CM_TIMEWAIT_EXIT && - cma_disable_callback(id_priv, CMA_DISCONNECT))) + cma_disable_callback(id_priv, RDMA_CM_DISCONNECT))) return 0; memset(&event, 0, sizeof event); @@ -1048,7 +1034,8 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) event.status = -ETIMEDOUT; /* fall through */ case IB_CM_DREQ_RECEIVED: case IB_CM_DREP_RECEIVED: - if (!cma_comp_exch(id_priv, CMA_CONNECT, CMA_DISCONNECT)) + if (!cma_comp_exch(id_priv, RDMA_CM_CONNECT, + RDMA_CM_DISCONNECT)) goto out; event.event = RDMA_CM_EVENT_DISCONNECTED; break; @@ -1075,7 +1062,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) if (ret) { /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.ib = NULL; - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; @@ -1132,7 +1119,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id, rdma_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid); id_priv = container_of(id, struct rdma_id_private, id); - id_priv->state = CMA_CONNECT; + id_priv->state = RDMA_CM_CONNECT; return id_priv; destroy_id: @@ -1172,7 +1159,7 @@ static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id, } id_priv = container_of(id, struct rdma_id_private, id); - id_priv->state = CMA_CONNECT; + id_priv->state = RDMA_CM_CONNECT; return id_priv; err: rdma_destroy_id(id); @@ -1201,7 +1188,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) int offset, ret; listen_id = cm_id->context; - if (cma_disable_callback(listen_id, CMA_LISTEN)) + if (cma_disable_callback(listen_id, RDMA_CM_LISTEN)) return -ECONNABORTED; memset(&event, 0, sizeof event); @@ -1243,7 +1230,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) * while we're accessing the cm_id. */ mutex_lock(&lock); - if (cma_comp(conn_id, CMA_CONNECT) && + if (cma_comp(conn_id, RDMA_CM_CONNECT) && !cma_is_ud_ps(conn_id->id.ps)) ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0); mutex_unlock(&lock); @@ -1257,7 +1244,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) conn_id->cm_id.ib = NULL; release_conn_id: - cma_exch(conn_id, CMA_DESTROYING); + cma_exch(conn_id, RDMA_CM_DESTROYING); mutex_unlock(&conn_id->handler_mutex); rdma_destroy_id(&conn_id->id); @@ -1328,7 +1315,7 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event) struct sockaddr_in *sin; int ret = 0; - if (cma_disable_callback(id_priv, CMA_CONNECT)) + if (cma_disable_callback(id_priv, RDMA_CM_CONNECT)) return 0; memset(&event, 0, sizeof event); @@ -1371,7 +1358,7 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event) if (ret) { /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.iw = NULL; - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; @@ -1393,7 +1380,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, struct ib_device_attr attr; listen_id = cm_id->context; - if (cma_disable_callback(listen_id, CMA_LISTEN)) + if (cma_disable_callback(listen_id, RDMA_CM_LISTEN)) return -ECONNABORTED; /* Create a new RDMA id for the new IW CM ID */ @@ -1406,7 +1393,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, } conn_id = container_of(new_cm_id, struct rdma_id_private, id); mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING); - conn_id->state = CMA_CONNECT; + conn_id->state = RDMA_CM_CONNECT; dev = ip_dev_find(&init_net, iw_event->local_addr.sin_addr.s_addr); if (!dev) { @@ -1461,7 +1448,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, if (ret) { /* User wants to destroy the CM ID */ conn_id->cm_id.iw = NULL; - cma_exch(conn_id, CMA_DESTROYING); + cma_exch(conn_id, RDMA_CM_DESTROYING); mutex_unlock(&conn_id->handler_mutex); cma_deref_id(conn_id); rdma_destroy_id(&conn_id->id); @@ -1554,7 +1541,7 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv, dev_id_priv = container_of(id, struct rdma_id_private, id); - dev_id_priv->state = CMA_ADDR_BOUND; + dev_id_priv->state = RDMA_CM_ADDR_BOUND; memcpy(&id->route.addr.src_addr, &id_priv->id.route.addr.src_addr, ip_addr_size((struct sockaddr *) &id_priv->id.route.addr.src_addr)); @@ -1601,8 +1588,8 @@ static void cma_query_handler(int status, struct ib_sa_path_rec *path_rec, route->num_paths = 1; *route->path_rec = *path_rec; } else { - work->old_state = CMA_ROUTE_QUERY; - work->new_state = CMA_ADDR_RESOLVED; + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ADDR_RESOLVED; work->event.event = RDMA_CM_EVENT_ROUTE_ERROR; work->event.status = status; } @@ -1660,7 +1647,7 @@ static void cma_work_handler(struct work_struct *_work) goto out; if (id_priv->id.event_handler(&id_priv->id, &work->event)) { - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); destroy = 1; } out: @@ -1678,12 +1665,12 @@ static void cma_ndev_work_handler(struct work_struct *_work) int destroy = 0; mutex_lock(&id_priv->handler_mutex); - if (id_priv->state == CMA_DESTROYING || - id_priv->state == CMA_DEVICE_REMOVAL) + if (id_priv->state == RDMA_CM_DESTROYING || + id_priv->state == RDMA_CM_DEVICE_REMOVAL) goto out; if (id_priv->id.event_handler(&id_priv->id, &work->event)) { - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); destroy = 1; } @@ -1707,8 +1694,8 @@ static int cma_resolve_ib_route(struct rdma_id_private *id_priv, int timeout_ms) work->id = id_priv; INIT_WORK(&work->work, cma_work_handler); - work->old_state = CMA_ROUTE_QUERY; - work->new_state = CMA_ROUTE_RESOLVED; + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ROUTE_RESOLVED; work->event.event = RDMA_CM_EVENT_ROUTE_RESOLVED; route->path_rec = kmalloc(sizeof *route->path_rec, GFP_KERNEL); @@ -1737,7 +1724,8 @@ int rdma_set_ib_paths(struct rdma_cm_id *id, int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp_exch(id_priv, CMA_ADDR_RESOLVED, CMA_ROUTE_RESOLVED)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ROUTE_RESOLVED)) return -EINVAL; id->route.path_rec = kmemdup(path_rec, sizeof *path_rec * num_paths, @@ -1750,7 +1738,7 @@ int rdma_set_ib_paths(struct rdma_cm_id *id, id->route.num_paths = num_paths; return 0; err: - cma_comp_exch(id_priv, CMA_ROUTE_RESOLVED, CMA_ADDR_RESOLVED); + cma_comp_exch(id_priv, RDMA_CM_ROUTE_RESOLVED, RDMA_CM_ADDR_RESOLVED); return ret; } EXPORT_SYMBOL(rdma_set_ib_paths); @@ -1765,8 +1753,8 @@ static int cma_resolve_iw_route(struct rdma_id_private *id_priv, int timeout_ms) work->id = id_priv; INIT_WORK(&work->work, cma_work_handler); - work->old_state = CMA_ROUTE_QUERY; - work->new_state = CMA_ROUTE_RESOLVED; + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ROUTE_RESOLVED; work->event.event = RDMA_CM_EVENT_ROUTE_RESOLVED; queue_work(cma_wq, &work->work); return 0; @@ -1830,8 +1818,8 @@ static int cma_resolve_iboe_route(struct rdma_id_private *id_priv) goto err2; } - work->old_state = CMA_ROUTE_QUERY; - work->new_state = CMA_ROUTE_RESOLVED; + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ROUTE_RESOLVED; work->event.event = RDMA_CM_EVENT_ROUTE_RESOLVED; work->event.status = 0; @@ -1853,7 +1841,7 @@ int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms) int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp_exch(id_priv, CMA_ADDR_RESOLVED, CMA_ROUTE_QUERY)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, RDMA_CM_ROUTE_QUERY)) return -EINVAL; atomic_inc(&id_priv->refcount); @@ -1882,7 +1870,7 @@ int rdma_resolve_route(struct rdma_cm_id *id, int timeout_ms) return 0; err: - cma_comp_exch(id_priv, CMA_ROUTE_QUERY, CMA_ADDR_RESOLVED); + cma_comp_exch(id_priv, RDMA_CM_ROUTE_QUERY, RDMA_CM_ADDR_RESOLVED); cma_deref_id(id_priv); return ret; } @@ -1941,14 +1929,16 @@ static void addr_handler(int status, struct sockaddr *src_addr, memset(&event, 0, sizeof event); mutex_lock(&id_priv->handler_mutex); - if (!cma_comp_exch(id_priv, CMA_ADDR_QUERY, CMA_ADDR_RESOLVED)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, + RDMA_CM_ADDR_RESOLVED)) goto out; if (!status && !id_priv->cma_dev) status = cma_acquire_dev(id_priv); if (status) { - if (!cma_comp_exch(id_priv, CMA_ADDR_RESOLVED, CMA_ADDR_BOUND)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ADDR_BOUND)) goto out; event.event = RDMA_CM_EVENT_ADDR_ERROR; event.status = status; @@ -1959,7 +1949,7 @@ static void addr_handler(int status, struct sockaddr *src_addr, } if (id_priv->id.event_handler(&id_priv->id, &event)) { - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); mutex_unlock(&id_priv->handler_mutex); cma_deref_id(id_priv); rdma_destroy_id(&id_priv->id); @@ -2004,8 +1994,8 @@ static int cma_resolve_loopback(struct rdma_id_private *id_priv) work->id = id_priv; INIT_WORK(&work->work, cma_work_handler); - work->old_state = CMA_ADDR_QUERY; - work->new_state = CMA_ADDR_RESOLVED; + work->old_state = RDMA_CM_ADDR_QUERY; + work->new_state = RDMA_CM_ADDR_RESOLVED; work->event.event = RDMA_CM_EVENT_ADDR_RESOLVED; queue_work(cma_wq, &work->work); return 0; @@ -2034,13 +2024,13 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (id_priv->state == CMA_IDLE) { + if (id_priv->state == RDMA_CM_IDLE) { ret = cma_bind_addr(id, src_addr, dst_addr); if (ret) return ret; } - if (!cma_comp_exch(id_priv, CMA_ADDR_BOUND, CMA_ADDR_QUERY)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_ADDR_QUERY)) return -EINVAL; atomic_inc(&id_priv->refcount); @@ -2056,7 +2046,7 @@ int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, return 0; err: - cma_comp_exch(id_priv, CMA_ADDR_QUERY, CMA_ADDR_BOUND); + cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); cma_deref_id(id_priv); return ret; } @@ -2070,7 +2060,7 @@ int rdma_set_reuseaddr(struct rdma_cm_id *id, int reuse) id_priv = container_of(id, struct rdma_id_private, id); spin_lock_irqsave(&id_priv->lock, flags); - if (id_priv->state == CMA_IDLE) { + if (id_priv->state == RDMA_CM_IDLE) { id_priv->reuseaddr = reuse; ret = 0; } else { @@ -2177,7 +2167,7 @@ static int cma_check_port(struct rdma_bind_list *bind_list, if (id_priv == cur_id) continue; - if ((cur_id->state == CMA_LISTEN) || + if ((cur_id->state == RDMA_CM_LISTEN) || !reuseaddr || !cur_id->reuseaddr) { cur_addr = (struct sockaddr *) &cur_id->id.route.addr.src_addr; if (cma_any_addr(cur_addr)) @@ -2280,14 +2270,14 @@ int rdma_listen(struct rdma_cm_id *id, int backlog) int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (id_priv->state == CMA_IDLE) { + if (id_priv->state == RDMA_CM_IDLE) { ((struct sockaddr *) &id->route.addr.src_addr)->sa_family = AF_INET; ret = rdma_bind_addr(id, (struct sockaddr *) &id->route.addr.src_addr); if (ret) return ret; } - if (!cma_comp_exch(id_priv, CMA_ADDR_BOUND, CMA_LISTEN)) + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_LISTEN)) return -EINVAL; if (id_priv->reuseaddr) { @@ -2319,7 +2309,7 @@ int rdma_listen(struct rdma_cm_id *id, int backlog) return 0; err: id_priv->backlog = 0; - cma_comp_exch(id_priv, CMA_LISTEN, CMA_ADDR_BOUND); + cma_comp_exch(id_priv, RDMA_CM_LISTEN, RDMA_CM_ADDR_BOUND); return ret; } EXPORT_SYMBOL(rdma_listen); @@ -2333,7 +2323,7 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) return -EAFNOSUPPORT; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp_exch(id_priv, CMA_IDLE, CMA_ADDR_BOUND)) + if (!cma_comp_exch(id_priv, RDMA_CM_IDLE, RDMA_CM_ADDR_BOUND)) return -EINVAL; ret = cma_check_linklocal(&id->route.addr.dev_addr, addr); @@ -2360,7 +2350,7 @@ err2: if (id_priv->cma_dev) cma_release_dev(id_priv); err1: - cma_comp_exch(id_priv, CMA_ADDR_BOUND, CMA_IDLE); + cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE); return ret; } EXPORT_SYMBOL(rdma_bind_addr); @@ -2433,7 +2423,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, struct ib_cm_sidr_rep_event_param *rep = &ib_event->param.sidr_rep_rcvd; int ret = 0; - if (cma_disable_callback(id_priv, CMA_CONNECT)) + if (cma_disable_callback(id_priv, RDMA_CM_CONNECT)) return 0; memset(&event, 0, sizeof event); @@ -2479,7 +2469,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, if (ret) { /* Destroy the CM ID by returning a non-zero value. */ id_priv->cm_id.ib = NULL; - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return ret; @@ -2645,7 +2635,7 @@ int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp_exch(id_priv, CMA_ROUTE_RESOLVED, CMA_CONNECT)) + if (!cma_comp_exch(id_priv, RDMA_CM_ROUTE_RESOLVED, RDMA_CM_CONNECT)) return -EINVAL; if (!id->qp) { @@ -2672,7 +2662,7 @@ int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) return 0; err: - cma_comp_exch(id_priv, CMA_CONNECT, CMA_ROUTE_RESOLVED); + cma_comp_exch(id_priv, RDMA_CM_CONNECT, RDMA_CM_ROUTE_RESOLVED); return ret; } EXPORT_SYMBOL(rdma_connect); @@ -2758,7 +2748,7 @@ int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp(id_priv, CMA_CONNECT)) + if (!cma_comp(id_priv, RDMA_CM_CONNECT)) return -EINVAL; if (!id->qp && conn_param) { @@ -2887,8 +2877,8 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast) int ret; id_priv = mc->id_priv; - if (cma_disable_callback(id_priv, CMA_ADDR_BOUND) && - cma_disable_callback(id_priv, CMA_ADDR_RESOLVED)) + if (cma_disable_callback(id_priv, RDMA_CM_ADDR_BOUND) && + cma_disable_callback(id_priv, RDMA_CM_ADDR_RESOLVED)) return 0; mutex_lock(&id_priv->qp_mutex); @@ -2912,7 +2902,7 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast) ret = id_priv->id.event_handler(&id_priv->id, &event); if (ret) { - cma_exch(id_priv, CMA_DESTROYING); + cma_exch(id_priv, RDMA_CM_DESTROYING); mutex_unlock(&id_priv->handler_mutex); rdma_destroy_id(&id_priv->id); return 0; @@ -3095,8 +3085,8 @@ int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr, int ret; id_priv = container_of(id, struct rdma_id_private, id); - if (!cma_comp(id_priv, CMA_ADDR_BOUND) && - !cma_comp(id_priv, CMA_ADDR_RESOLVED)) + if (!cma_comp(id_priv, RDMA_CM_ADDR_BOUND) && + !cma_comp(id_priv, RDMA_CM_ADDR_RESOLVED)) return -EINVAL; mc = kmalloc(sizeof *mc, GFP_KERNEL); @@ -3261,19 +3251,19 @@ static void cma_add_one(struct ib_device *device) static int cma_remove_id_dev(struct rdma_id_private *id_priv) { struct rdma_cm_event event; - enum cma_state state; + enum rdma_cm_state state; int ret = 0; /* Record that we want to remove the device */ - state = cma_exch(id_priv, CMA_DEVICE_REMOVAL); - if (state == CMA_DESTROYING) + state = cma_exch(id_priv, RDMA_CM_DEVICE_REMOVAL); + if (state == RDMA_CM_DESTROYING) return 0; cma_cancel_operation(id_priv, state); mutex_lock(&id_priv->handler_mutex); /* Check for destruction from another callback. */ - if (!cma_comp(id_priv, CMA_DEVICE_REMOVAL)) + if (!cma_comp(id_priv, RDMA_CM_DEVICE_REMOVAL)) goto out; memset(&event, 0, sizeof event); diff --git a/include/rdma/rdma_cm.h b/include/rdma/rdma_cm.h index 169f7a53fb0c..d5b2265b7ce8 100644 --- a/include/rdma/rdma_cm.h +++ b/include/rdma/rdma_cm.h @@ -111,6 +111,20 @@ struct rdma_cm_event { } param; }; +enum rdma_cm_state { + RDMA_CM_IDLE, + RDMA_CM_ADDR_QUERY, + RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ROUTE_QUERY, + RDMA_CM_ROUTE_RESOLVED, + RDMA_CM_CONNECT, + RDMA_CM_DISCONNECT, + RDMA_CM_ADDR_BOUND, + RDMA_CM_LISTEN, + RDMA_CM_DEVICE_REMOVAL, + RDMA_CM_DESTROYING +}; + struct rdma_cm_id; /** -- cgit v1.2.3 From b26f9b9949013fec31b23c426fc463164ae08891 Mon Sep 17 00:00:00 2001 From: Sean Hefty Date: Thu, 1 Apr 2010 17:08:41 +0000 Subject: RDMA/cma: Pass QP type into rdma_create_id() The RDMA CM currently infers the QP type from the port space selected by the user. In the future (eg with RDMA_PS_IB or XRC), there may not be a 1-1 correspondence between port space and QP type. For netlink export of RDMA CM state, we want to export the QP type to userspace, so it is cleaner to explicitly associate a QP type to an ID. Modify rdma_create_id() to allow the user to specify the QP type, and use it to make our selections of datagram versus connected mode. Signed-off-by: Sean Hefty Signed-off-by: Roland Dreier --- drivers/infiniband/core/cma.c | 35 +++++++++++++++----------------- drivers/infiniband/core/ucma.c | 27 ++++++++++++++++++++---- drivers/infiniband/ulp/iser/iser_verbs.c | 2 +- include/rdma/rdma_cm.h | 5 ++++- net/9p/trans_rdma.c | 3 ++- net/rds/ib.c | 2 +- net/rds/ib_cm.c | 2 +- net/rds/iw.c | 2 +- net/rds/iw_cm.c | 2 +- net/rds/rdma_transport.c | 3 ++- net/sunrpc/xprtrdma/svc_rdma_transport.c | 3 ++- net/sunrpc/xprtrdma/verbs.c | 2 +- 12 files changed, 55 insertions(+), 33 deletions(-) (limited to 'drivers') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 451d39e19cb4..44be1c9ed05b 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -265,11 +265,6 @@ static inline void sdp_set_ip_ver(struct sdp_hh *hh, u8 ip_ver) hh->ip_version = (ip_ver << 4) | (hh->ip_version & 0xF); } -static inline int cma_is_ud_ps(enum rdma_port_space ps) -{ - return (ps == RDMA_PS_UDP || ps == RDMA_PS_IPOIB); -} - static void cma_attach_to_dev(struct rdma_id_private *id_priv, struct cma_device *cma_dev) { @@ -415,7 +410,8 @@ static int cma_has_cm_dev(struct rdma_id_private *id_priv) } struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, - void *context, enum rdma_port_space ps) + void *context, enum rdma_port_space ps, + enum ib_qp_type qp_type) { struct rdma_id_private *id_priv; @@ -427,6 +423,7 @@ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, id_priv->id.context = context; id_priv->id.event_handler = event_handler; id_priv->id.ps = ps; + id_priv->id.qp_type = qp_type; spin_lock_init(&id_priv->lock); mutex_init(&id_priv->qp_mutex); init_completion(&id_priv->comp); @@ -494,7 +491,7 @@ int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd, if (IS_ERR(qp)) return PTR_ERR(qp); - if (cma_is_ud_ps(id_priv->id.ps)) + if (id->qp_type == IB_QPT_UD) ret = cma_init_ud_qp(id_priv, qp); else ret = cma_init_conn_qp(id_priv, qp); @@ -622,7 +619,7 @@ static int cma_ib_init_qp_attr(struct rdma_id_private *id_priv, qp_attr->port_num = id_priv->id.port_num; *qp_attr_mask = IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_PORT; - if (cma_is_ud_ps(id_priv->id.ps)) { + if (id_priv->id.qp_type == IB_QPT_UD) { ret = cma_set_qkey(id_priv); if (ret) return ret; @@ -645,7 +642,7 @@ int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr, id_priv = container_of(id, struct rdma_id_private, id); switch (rdma_node_get_transport(id_priv->id.device->node_type)) { case RDMA_TRANSPORT_IB: - if (!id_priv->cm_id.ib || cma_is_ud_ps(id_priv->id.ps)) + if (!id_priv->cm_id.ib || (id_priv->id.qp_type == IB_QPT_UD)) ret = cma_ib_init_qp_attr(id_priv, qp_attr, qp_attr_mask); else ret = ib_cm_init_qp_attr(id_priv->cm_id.ib, qp_attr, @@ -1088,7 +1085,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id, goto err; id = rdma_create_id(listen_id->event_handler, listen_id->context, - listen_id->ps); + listen_id->ps, ib_event->param.req_rcvd.qp_type); if (IS_ERR(id)) goto err; @@ -1139,7 +1136,7 @@ static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id, int ret; id = rdma_create_id(listen_id->event_handler, listen_id->context, - listen_id->ps); + listen_id->ps, IB_QPT_UD); if (IS_ERR(id)) return NULL; @@ -1194,7 +1191,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) memset(&event, 0, sizeof event); offset = cma_user_data_offset(listen_id->id.ps); event.event = RDMA_CM_EVENT_CONNECT_REQUEST; - if (cma_is_ud_ps(listen_id->id.ps)) { + if (listen_id->id.qp_type == IB_QPT_UD) { conn_id = cma_new_udp_id(&listen_id->id, ib_event); event.param.ud.private_data = ib_event->private_data + offset; event.param.ud.private_data_len = @@ -1230,8 +1227,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event) * while we're accessing the cm_id. */ mutex_lock(&lock); - if (cma_comp(conn_id, RDMA_CM_CONNECT) && - !cma_is_ud_ps(conn_id->id.ps)) + if (cma_comp(conn_id, RDMA_CM_CONNECT) && (conn_id->id.qp_type != IB_QPT_UD)) ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0); mutex_unlock(&lock); mutex_unlock(&conn_id->handler_mutex); @@ -1386,7 +1382,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id, /* Create a new RDMA id for the new IW CM ID */ new_cm_id = rdma_create_id(listen_id->id.event_handler, listen_id->id.context, - RDMA_PS_TCP); + RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(new_cm_id)) { ret = -ENOMEM; goto out; @@ -1535,7 +1531,8 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv, struct rdma_cm_id *id; int ret; - id = rdma_create_id(cma_listen_handler, id_priv, id_priv->id.ps); + id = rdma_create_id(cma_listen_handler, id_priv, id_priv->id.ps, + id_priv->id.qp_type); if (IS_ERR(id)) return; @@ -2645,7 +2642,7 @@ int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) switch (rdma_node_get_transport(id->device->node_type)) { case RDMA_TRANSPORT_IB: - if (cma_is_ud_ps(id->ps)) + if (id->qp_type == IB_QPT_UD) ret = cma_resolve_ib_udp(id_priv, conn_param); else ret = cma_connect_ib(id_priv, conn_param); @@ -2758,7 +2755,7 @@ int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) switch (rdma_node_get_transport(id->device->node_type)) { case RDMA_TRANSPORT_IB: - if (cma_is_ud_ps(id->ps)) + if (id->qp_type == IB_QPT_UD) ret = cma_send_sidr_rep(id_priv, IB_SIDR_SUCCESS, conn_param->private_data, conn_param->private_data_len); @@ -2819,7 +2816,7 @@ int rdma_reject(struct rdma_cm_id *id, const void *private_data, switch (rdma_node_get_transport(id->device->node_type)) { case RDMA_TRANSPORT_IB: - if (cma_is_ud_ps(id->ps)) + if (id->qp_type == IB_QPT_UD) ret = cma_send_sidr_rep(id_priv, IB_SIDR_REJECT, private_data, private_data_len); else diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c index b3fa798525b2..7109d5d23ba5 100644 --- a/drivers/infiniband/core/ucma.c +++ b/drivers/infiniband/core/ucma.c @@ -367,13 +367,28 @@ done: return ret; } -static ssize_t ucma_create_id(struct ucma_file *file, - const char __user *inbuf, - int in_len, int out_len) +static int ucma_get_qp_type(struct rdma_ucm_create_id *cmd, enum ib_qp_type *qp_type) +{ + switch (cmd->ps) { + case RDMA_PS_TCP: + *qp_type = IB_QPT_RC; + return 0; + case RDMA_PS_UDP: + case RDMA_PS_IPOIB: + *qp_type = IB_QPT_UD; + return 0; + default: + return -EINVAL; + } +} + +static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) { struct rdma_ucm_create_id cmd; struct rdma_ucm_create_id_resp resp; struct ucma_context *ctx; + enum ib_qp_type qp_type; int ret; if (out_len < sizeof(resp)) @@ -382,6 +397,10 @@ static ssize_t ucma_create_id(struct ucma_file *file, if (copy_from_user(&cmd, inbuf, sizeof(cmd))) return -EFAULT; + ret = ucma_get_qp_type(&cmd, &qp_type); + if (ret) + return ret; + mutex_lock(&file->mut); ctx = ucma_alloc_ctx(file); mutex_unlock(&file->mut); @@ -389,7 +408,7 @@ static ssize_t ucma_create_id(struct ucma_file *file, return -ENOMEM; ctx->uid = cmd.uid; - ctx->cm_id = rdma_create_id(ucma_event_handler, ctx, cmd.ps); + ctx->cm_id = rdma_create_id(ucma_event_handler, ctx, cmd.ps, qp_type); if (IS_ERR(ctx->cm_id)) { ret = PTR_ERR(ctx->cm_id); goto err1; diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 9876865732f7..ede1475bee09 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -548,7 +548,7 @@ int iser_connect(struct iser_conn *ib_conn, iser_conn_get(ib_conn); /* ref ib conn's cma id */ ib_conn->cma_id = rdma_create_id(iser_cma_handler, (void *)ib_conn, - RDMA_PS_TCP); + RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(ib_conn->cma_id)) { err = PTR_ERR(ib_conn->cma_id); iser_err("rdma_create_id failed: %d\n", err); diff --git a/include/rdma/rdma_cm.h b/include/rdma/rdma_cm.h index d5b2265b7ce8..26977c149c41 100644 --- a/include/rdma/rdma_cm.h +++ b/include/rdma/rdma_cm.h @@ -144,6 +144,7 @@ struct rdma_cm_id { rdma_cm_event_handler event_handler; struct rdma_route route; enum rdma_port_space ps; + enum ib_qp_type qp_type; u8 port_num; }; @@ -154,9 +155,11 @@ struct rdma_cm_id { * returned rdma_id. * @context: User specified context associated with the id. * @ps: RDMA port space. + * @qp_type: type of queue pair associated with the id. */ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, - void *context, enum rdma_port_space ps); + void *context, enum rdma_port_space ps, + enum ib_qp_type qp_type); /** * rdma_destroy_id - Destroys an RDMA identifier. diff --git a/net/9p/trans_rdma.c b/net/9p/trans_rdma.c index 150e0c4bbf40..3640e83eef8f 100644 --- a/net/9p/trans_rdma.c +++ b/net/9p/trans_rdma.c @@ -591,7 +591,8 @@ rdma_create_trans(struct p9_client *client, const char *addr, char *args) return -ENOMEM; /* Create the RDMA CM ID */ - rdma->cm_id = rdma_create_id(p9_cm_event_handler, client, RDMA_PS_TCP); + rdma->cm_id = rdma_create_id(p9_cm_event_handler, client, RDMA_PS_TCP, + IB_QPT_RC); if (IS_ERR(rdma->cm_id)) goto error; diff --git a/net/rds/ib.c b/net/rds/ib.c index cce19f95c624..3b83086bcc30 100644 --- a/net/rds/ib.c +++ b/net/rds/ib.c @@ -325,7 +325,7 @@ static int rds_ib_laddr_check(__be32 addr) /* Create a CMA ID and try to bind it. This catches both * IB and iWARP capable NICs. */ - cm_id = rdma_create_id(NULL, NULL, RDMA_PS_TCP); + cm_id = rdma_create_id(NULL, NULL, RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(cm_id)) return PTR_ERR(cm_id); diff --git a/net/rds/ib_cm.c b/net/rds/ib_cm.c index ee369d201a65..fd453dd5124b 100644 --- a/net/rds/ib_cm.c +++ b/net/rds/ib_cm.c @@ -587,7 +587,7 @@ int rds_ib_conn_connect(struct rds_connection *conn) /* XXX I wonder what affect the port space has */ /* delegate cm event handler to rdma_transport */ ic->i_cm_id = rdma_create_id(rds_rdma_cm_event_handler, conn, - RDMA_PS_TCP); + RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(ic->i_cm_id)) { ret = PTR_ERR(ic->i_cm_id); ic->i_cm_id = NULL; diff --git a/net/rds/iw.c b/net/rds/iw.c index 5a9676fe594f..f7474844f096 100644 --- a/net/rds/iw.c +++ b/net/rds/iw.c @@ -226,7 +226,7 @@ static int rds_iw_laddr_check(__be32 addr) /* Create a CMA ID and try to bind it. This catches both * IB and iWARP capable NICs. */ - cm_id = rdma_create_id(NULL, NULL, RDMA_PS_TCP); + cm_id = rdma_create_id(NULL, NULL, RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(cm_id)) return PTR_ERR(cm_id); diff --git a/net/rds/iw_cm.c b/net/rds/iw_cm.c index 3a60a15d1b4a..c12db66f24c7 100644 --- a/net/rds/iw_cm.c +++ b/net/rds/iw_cm.c @@ -522,7 +522,7 @@ int rds_iw_conn_connect(struct rds_connection *conn) /* XXX I wonder what affect the port space has */ /* delegate cm event handler to rdma_transport */ ic->i_cm_id = rdma_create_id(rds_rdma_cm_event_handler, conn, - RDMA_PS_TCP); + RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(ic->i_cm_id)) { ret = PTR_ERR(ic->i_cm_id); ic->i_cm_id = NULL; diff --git a/net/rds/rdma_transport.c b/net/rds/rdma_transport.c index 4195a0539829..f8760e1b6688 100644 --- a/net/rds/rdma_transport.c +++ b/net/rds/rdma_transport.c @@ -158,7 +158,8 @@ static int rds_rdma_listen_init(void) struct rdma_cm_id *cm_id; int ret; - cm_id = rdma_create_id(rds_rdma_cm_event_handler, NULL, RDMA_PS_TCP); + cm_id = rdma_create_id(rds_rdma_cm_event_handler, NULL, RDMA_PS_TCP, + IB_QPT_RC); if (IS_ERR(cm_id)) { ret = PTR_ERR(cm_id); printk(KERN_ERR "RDS/RDMA: failed to setup listener, " diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c index 1a10dcd999ea..afff475642ca 100644 --- a/net/sunrpc/xprtrdma/svc_rdma_transport.c +++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c @@ -695,7 +695,8 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, return ERR_PTR(-ENOMEM); xprt = &cma_xprt->sc_xprt; - listen_id = rdma_create_id(rdma_listen_handler, cma_xprt, RDMA_PS_TCP); + listen_id = rdma_create_id(rdma_listen_handler, cma_xprt, RDMA_PS_TCP, + IB_QPT_RC); if (IS_ERR(listen_id)) { ret = PTR_ERR(listen_id); dprintk("svcrdma: rdma_create_id failed = %d\n", ret); diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index d4297dc43dc4..80f8da344df5 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -387,7 +387,7 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, init_completion(&ia->ri_done); - id = rdma_create_id(rpcrdma_conn_upcall, xprt, RDMA_PS_TCP); + id = rdma_create_id(rpcrdma_conn_upcall, xprt, RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(id)) { rc = PTR_ERR(id); dprintk("RPC: %s: rdma_create_id() failed %i\n", -- cgit v1.2.3 From 753f618ae03a699cc30044d64eea84165016dc6e Mon Sep 17 00:00:00 2001 From: Nir Muchtar Date: Mon, 3 Jan 2011 15:33:53 +0000 Subject: RDMA/cma: Add support for netlink statistics export Add callbacks and data types for statistics export of all current devices/ids. The schema for RDMA CM is a series of netlink messages. Each one contains an rdma_cm_stat struct. Additionally, two netlink attributes are created for the addresses for each message (if applicable). Their types used are: RDMA_NL_RDMA_CM_ATTR_SRC_ADDR (The source address for this ID) RDMA_NL_RDMA_CM_ATTR_DST_ADDR (The destination address for this ID) sockaddr_* structs are encapsulated within these attributes. In other words, every transaction contains a series of messages like: -------message 1------- struct rdma_cm_id_stats { __u32 qp_num; __u32 bound_dev_if; __u32 port_space; __s32 pid; __u8 cm_state; __u8 node_type; __u8 port_num; __u8 reserved; } RDMA_NL_RDMA_CM_ATTR_SRC_ADDR attribute - contains the source address RDMA_NL_RDMA_CM_ATTR_DST_ADDR attribute - contains the destination address -------end 1------- -------message 2------- struct rdma_cm_id_stats RDMA_NL_RDMA_CM_ATTR_SRC_ADDR attribute RDMA_NL_RDMA_CM_ATTR_DST_ADDR attribute -------end 2------- Signed-off-by: Nir Muchtar Signed-off-by: Roland Dreier --- drivers/infiniband/core/cma.c | 99 +++++++++++++++++++++++++++++++++++++++++++ include/rdma/rdma_netlink.h | 28 ++++++++++++ 2 files changed, 127 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 44be1c9ed05b..5dd9acb92c3a 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -47,6 +47,7 @@ #include #include +#include #include #include #include @@ -3315,6 +3316,99 @@ static void cma_remove_one(struct ib_device *device) kfree(cma_dev); } +static int cma_get_id_stats(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct nlmsghdr *nlh; + struct rdma_cm_id_stats *id_stats; + struct rdma_id_private *id_priv; + struct rdma_cm_id *id = NULL; + struct cma_device *cma_dev; + int i_dev = 0, i_id = 0; + + /* + * We export all of the IDs as a sequence of messages. Each + * ID gets its own netlink message. + */ + mutex_lock(&lock); + + list_for_each_entry(cma_dev, &dev_list, list) { + if (i_dev < cb->args[0]) { + i_dev++; + continue; + } + + i_id = 0; + list_for_each_entry(id_priv, &cma_dev->id_list, list) { + if (i_id < cb->args[1]) { + i_id++; + continue; + } + + id_stats = ibnl_put_msg(skb, &nlh, cb->nlh->nlmsg_seq, + sizeof *id_stats, RDMA_NL_RDMA_CM, + RDMA_NL_RDMA_CM_ID_STATS); + if (!id_stats) + goto out; + + memset(id_stats, 0, sizeof *id_stats); + id = &id_priv->id; + id_stats->node_type = id->route.addr.dev_addr.dev_type; + id_stats->port_num = id->port_num; + id_stats->bound_dev_if = + id->route.addr.dev_addr.bound_dev_if; + + if (id->route.addr.src_addr.ss_family == AF_INET) { + if (ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_in), + &id->route.addr.src_addr, + RDMA_NL_RDMA_CM_ATTR_SRC_ADDR)) { + goto out; + } + if (ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_in), + &id->route.addr.dst_addr, + RDMA_NL_RDMA_CM_ATTR_DST_ADDR)) { + goto out; + } + } else if (id->route.addr.src_addr.ss_family == AF_INET6) { + if (ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_in6), + &id->route.addr.src_addr, + RDMA_NL_RDMA_CM_ATTR_SRC_ADDR)) { + goto out; + } + if (ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_in6), + &id->route.addr.dst_addr, + RDMA_NL_RDMA_CM_ATTR_DST_ADDR)) { + goto out; + } + } + + id_stats->port_space = id->ps; + id_stats->cm_state = id_priv->state; + id_stats->qp_num = id_priv->qp_num; + id_stats->qp_type = id->qp_type; + + i_id++; + } + + cb->args[1] = 0; + i_dev++; + } + +out: + mutex_unlock(&lock); + cb->args[0] = i_dev; + cb->args[1] = i_id; + + return skb->len; +} + +static const struct ibnl_client_cbs cma_cb_table[] = { + [RDMA_NL_RDMA_CM_ID_STATS] = { .dump = cma_get_id_stats }, +}; + static int __init cma_init(void) { int ret; @@ -3330,6 +3424,10 @@ static int __init cma_init(void) ret = ib_register_client(&cma_client); if (ret) goto err; + + if (ibnl_add_client(RDMA_NL_RDMA_CM, RDMA_NL_RDMA_CM_NUM_OPS, cma_cb_table)) + printk(KERN_WARNING "RDMA CMA: failed to add netlink callback\n"); + return 0; err: @@ -3342,6 +3440,7 @@ err: static void __exit cma_cleanup(void) { + ibnl_remove_client(RDMA_NL_RDMA_CM); ib_unregister_client(&cma_client); unregister_netdevice_notifier(&cma_nb); rdma_addr_unregister_client(&addr_client); diff --git a/include/rdma/rdma_netlink.h b/include/rdma/rdma_netlink.h index c983a193cecf..3c5363ab867b 100644 --- a/include/rdma/rdma_netlink.h +++ b/include/rdma/rdma_netlink.h @@ -1,10 +1,38 @@ #ifndef _RDMA_NETLINK_H #define _RDMA_NETLINK_H +#include + +enum { + RDMA_NL_RDMA_CM = 1 +}; + #define RDMA_NL_GET_CLIENT(type) ((type & (((1 << 6) - 1) << 10)) >> 10) #define RDMA_NL_GET_OP(type) (type & ((1 << 10) - 1)) #define RDMA_NL_GET_TYPE(client, op) ((client << 10) + op) +enum { + RDMA_NL_RDMA_CM_ID_STATS = 0, + RDMA_NL_RDMA_CM_NUM_OPS +}; + +enum { + RDMA_NL_RDMA_CM_ATTR_SRC_ADDR = 1, + RDMA_NL_RDMA_CM_ATTR_DST_ADDR, + RDMA_NL_RDMA_CM_NUM_ATTR, +}; + +struct rdma_cm_id_stats { + __u32 qp_num; + __u32 bound_dev_if; + __u32 port_space; + __s32 pid; + __u8 cm_state; + __u8 node_type; + __u8 port_num; + __u8 qp_type; +}; + #ifdef __KERNEL__ #include -- cgit v1.2.3 From 83e9502d8db142822f3302e6a46a45082d3a27b7 Mon Sep 17 00:00:00 2001 From: Nir Muchtar Date: Thu, 13 Jan 2011 13:56:04 +0000 Subject: RDMA/cma: Save PID of ID's owner Save the PID associated with an RDMA CM ID for reporting via netlink. Signed-off-by: Nir Muchtar Signed-off-by: Roland Dreier --- drivers/infiniband/core/cma.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers') diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c index 5dd9acb92c3a..b6a33b3c516d 100644 --- a/drivers/infiniband/core/cma.c +++ b/drivers/infiniband/core/cma.c @@ -133,6 +133,7 @@ struct rdma_id_private { u32 seq_num; u32 qkey; u32 qp_num; + pid_t owner; u8 srq; u8 tos; u8 reuseaddr; @@ -420,6 +421,7 @@ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler, if (!id_priv) return ERR_PTR(-ENOMEM); + id_priv->owner = task_pid_nr(current); id_priv->state = RDMA_CM_IDLE; id_priv->id.context = context; id_priv->id.event_handler = event_handler; @@ -2746,6 +2748,9 @@ int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) int ret; id_priv = container_of(id, struct rdma_id_private, id); + + id_priv->owner = task_pid_nr(current); + if (!cma_comp(id_priv, RDMA_CM_CONNECT)) return -EINVAL; @@ -3385,6 +3390,7 @@ static int cma_get_id_stats(struct sk_buff *skb, struct netlink_callback *cb) } } + id_stats->pid = id_priv->owner; id_stats->port_space = id->ps; id_stats->cm_state = id_priv->state; id_stats->qp_num = id_priv->qp_num; -- cgit v1.2.3