cregit-Linux how code gets into the kernel

Release 4.11 net/rds/ib.c

Directory: net/rds
/*
 * Copyright (c) 2006 Oracle.  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.
 *
 */
#include <linux/kernel.h>
#include <linux/in.h>
#include <linux/if.h>
#include <linux/netdevice.h>
#include <linux/inetdevice.h>
#include <linux/if_arp.h>
#include <linux/delay.h>
#include <linux/slab.h>
#include <linux/module.h>

#include "rds_single_path.h"
#include "rds.h"
#include "ib.h"
#include "ib_mr.h"


static unsigned int rds_ib_mr_1m_pool_size = RDS_MR_1M_POOL_SIZE;

static unsigned int rds_ib_mr_8k_pool_size = RDS_MR_8K_POOL_SIZE;

unsigned int rds_ib_retry_count = RDS_IB_DEFAULT_RETRY_COUNT;

module_param(rds_ib_mr_1m_pool_size, int, 0444);
MODULE_PARM_DESC(rds_ib_mr_1m_pool_size, " Max number of 1M mr per HCA");
module_param(rds_ib_mr_8k_pool_size, int, 0444);
MODULE_PARM_DESC(rds_ib_mr_8k_pool_size, " Max number of 8K mr per HCA");
module_param(rds_ib_retry_count, int, 0444);
MODULE_PARM_DESC(rds_ib_retry_count, " Number of hw retries before reporting an error");

/*
 * we have a clumsy combination of RCU and a rwsem protecting this list
 * because it is used both in the get_mr fast path and while blocking in
 * the FMR flushing path.
 */

DECLARE_RWSEM(rds_ib_devices_lock);

struct list_head rds_ib_devices;

/* NOTE: if also grabbing ibdev lock, grab this first */

DEFINE_SPINLOCK(ib_nodev_conns_lock);

LIST_HEAD(ib_nodev_conns);


static void rds_ib_nodev_connect(void) { struct rds_ib_connection *ic; spin_lock(&ib_nodev_conns_lock); list_for_each_entry(ic, &ib_nodev_conns, ib_node) rds_conn_connect_if_down(ic->conn); spin_unlock(&ib_nodev_conns_lock); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown3997.50%150.00%
Stephen Hemminger12.50%150.00%
Total40100.00%2100.00%


static void rds_ib_dev_shutdown(struct rds_ib_device *rds_ibdev) { struct rds_ib_connection *ic; unsigned long flags; spin_lock_irqsave(&rds_ibdev->spinlock, flags); list_for_each_entry(ic, &rds_ibdev->conn_list, ib_node) rds_conn_drop(ic->conn); spin_unlock_irqrestore(&rds_ibdev->spinlock, flags); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown5498.18%150.00%
Stephen Hemminger11.82%150.00%
Total55100.00%2100.00%

/* * rds_ib_destroy_mr_pool() blocks on a few things and mrs drop references * from interrupt context so we push freing off into a work struct in krdsd. */
static void rds_ib_dev_free(struct work_struct *work) { struct rds_ib_ipaddr *i_ipaddr, *i_next; struct rds_ib_device *rds_ibdev = container_of(work, struct rds_ib_device, free_work); if (rds_ibdev->mr_8k_pool) rds_ib_destroy_mr_pool(rds_ibdev->mr_8k_pool); if (rds_ibdev->mr_1m_pool) rds_ib_destroy_mr_pool(rds_ibdev->mr_1m_pool); if (rds_ibdev->pd) ib_dealloc_pd(rds_ibdev->pd); list_for_each_entry_safe(i_ipaddr, i_next, &rds_ibdev->ipaddr_list, list) { list_del(&i_ipaddr->list); kfree(i_ipaddr); } kfree(rds_ibdev->vector_load); kfree(rds_ibdev); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown8880.00%133.33%
Santosh Shilimkar2220.00%266.67%
Total110100.00%3100.00%


void rds_ib_dev_put(struct rds_ib_device *rds_ibdev) { BUG_ON(atomic_read(&rds_ibdev->refcount) <= 0); if (atomic_dec_and_test(&rds_ibdev->refcount)) queue_work(rds_wq, &rds_ibdev->free_work); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown43100.00%1100.00%
Total43100.00%1100.00%


static void rds_ib_add_one(struct ib_device *device) { struct rds_ib_device *rds_ibdev; /* Only handle IB (no iWARP) devices */ if (device->node_type != RDMA_NODE_IB_CA) return; rds_ibdev = kzalloc_node(sizeof(struct rds_ib_device), GFP_KERNEL, ibdev_to_node(device)); if (!rds_ibdev) return; spin_lock_init(&rds_ibdev->spinlock); atomic_set(&rds_ibdev->refcount, 1); INIT_WORK(&rds_ibdev->free_work, rds_ib_dev_free); rds_ibdev->max_wrs = device->attrs.max_qp_wr; rds_ibdev->max_sge = min(device->attrs.max_sge, RDS_IB_MAX_SGE); rds_ibdev->has_fr = (device->attrs.device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS); rds_ibdev->has_fmr = (device->alloc_fmr && device->dealloc_fmr && device->map_phys_fmr && device->unmap_fmr); rds_ibdev->use_fastreg = (rds_ibdev->has_fr && !rds_ibdev->has_fmr); rds_ibdev->fmr_max_remaps = device->attrs.max_map_per_fmr?: 32; rds_ibdev->max_1m_mrs = device->attrs.max_mr ? min_t(unsigned int, (device->attrs.max_mr / 2), rds_ib_mr_1m_pool_size) : rds_ib_mr_1m_pool_size; rds_ibdev->max_8k_mrs = device->attrs.max_mr ? min_t(unsigned int, ((device->attrs.max_mr / 2) * RDS_MR_8K_SCALE), rds_ib_mr_8k_pool_size) : rds_ib_mr_8k_pool_size; rds_ibdev->max_initiator_depth = device->attrs.max_qp_init_rd_atom; rds_ibdev->max_responder_resources = device->attrs.max_qp_rd_atom; rds_ibdev->vector_load = kzalloc(sizeof(int) * device->num_comp_vectors, GFP_KERNEL); if (!rds_ibdev->vector_load) { pr_err("RDS/IB: %s failed to allocate vector memory\n", __func__); goto put_dev; } rds_ibdev->dev = device; rds_ibdev->pd = ib_alloc_pd(device, 0); if (IS_ERR(rds_ibdev->pd)) { rds_ibdev->pd = NULL; goto put_dev; } rds_ibdev->mr_1m_pool = rds_ib_create_mr_pool(rds_ibdev, RDS_IB_MR_1M_POOL); if (IS_ERR(rds_ibdev->mr_1m_pool)) { rds_ibdev->mr_1m_pool = NULL; goto put_dev; } rds_ibdev->mr_8k_pool = rds_ib_create_mr_pool(rds_ibdev, RDS_IB_MR_8K_POOL); if (IS_ERR(rds_ibdev->mr_8k_pool)) { rds_ibdev->mr_8k_pool = NULL; goto put_dev; } rdsdebug("RDS/IB: max_mr = %d, max_wrs = %d, max_sge = %d, fmr_max_remaps = %d, max_1m_mrs = %d, max_8k_mrs = %d\n", device->attrs.max_fmr, rds_ibdev->max_wrs, rds_ibdev->max_sge, rds_ibdev->fmr_max_remaps, rds_ibdev->max_1m_mrs, rds_ibdev->max_8k_mrs); pr_info("RDS/IB: %s: %s supported and preferred\n", device->name, rds_ibdev->use_fastreg ? "FRMR" : "FMR"); INIT_LIST_HEAD(&rds_ibdev->ipaddr_list); INIT_LIST_HEAD(&rds_ibdev->conn_list); down_write(&rds_ib_devices_lock); list_add_tail_rcu(&rds_ibdev->list, &rds_ib_devices); up_write(&rds_ib_devices_lock); atomic_inc(&rds_ibdev->refcount); ib_set_client_data(device, &rds_ib_client, rds_ibdev); atomic_inc(&rds_ibdev->refcount); rds_ib_nodev_connect(); put_dev: rds_ib_dev_put(rds_ibdev); }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover20441.46%321.43%
Santosh Shilimkar19439.43%535.71%
Zach Brown6413.01%321.43%
Or Gerlitz275.49%17.14%
Christoph Hellwig20.41%17.14%
Stephen Hemminger10.20%17.14%
Total492100.00%14100.00%

/* * New connections use this to find the device to associate with the * connection. It's not in the fast path so we're not concerned about the * performance of the IB call. (As of this writing, it uses an interrupt * blocking spinlock to serialize walking a per-device list of all registered * clients.) * * RCU is used to handle incoming connections racing with device teardown. * Rather than use a lock to serialize removal from the client_data and * getting a new reference, we use an RCU grace period. The destruction * path removes the device from client_data and then waits for all RCU * readers to finish. * * A new connection can get NULL from this if its arriving on a * device that is in the process of being removed. */
struct rds_ib_device *rds_ib_get_client_data(struct ib_device *device) { struct rds_ib_device *rds_ibdev; rcu_read_lock(); rds_ibdev = ib_get_client_data(device, &rds_ib_client); if (rds_ibdev) atomic_inc(&rds_ibdev->refcount); rcu_read_unlock(); return rds_ibdev; }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover3368.75%150.00%
Zach Brown1531.25%150.00%
Total48100.00%2100.00%

/* * The IB stack is letting us know that a device is going away. This can * happen if the underlying HCA driver is removed or if PCI hotplug is removing * the pci function, for example. * * This can be called at any time and can be racing with any other RDS path. */
static void rds_ib_remove_one(struct ib_device *device, void *client_data) { struct rds_ib_device *rds_ibdev = client_data; if (!rds_ibdev) return; rds_ib_dev_shutdown(rds_ibdev); /* stop connection attempts from getting a reference to this device. */ ib_set_client_data(device, &rds_ib_client, NULL); down_write(&rds_ib_devices_lock); list_del_rcu(&rds_ibdev->list); up_write(&rds_ib_devices_lock); /* * This synchronize rcu is waiting for readers of both the ib * client data and the devices list to finish before we drop * both of those references. */ synchronize_rcu(); rds_ib_dev_put(rds_ibdev); rds_ib_dev_put(rds_ibdev); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown4456.41%350.00%
Andy Grover2734.62%116.67%
Haggai Eran67.69%116.67%
Stephen Hemminger11.28%116.67%
Total78100.00%6100.00%

struct ib_client rds_ib_client = { .name = "rds_ib", .add = rds_ib_add_one, .remove = rds_ib_remove_one };
static int rds_ib_conn_info_visitor(struct rds_connection *conn, void *buffer) { struct rds_info_rdma_connection *iinfo = buffer; struct rds_ib_connection *ic; /* We will only ever look at IB transports */ if (conn->c_trans != &rds_ib_transport) return 0; iinfo->src_addr = conn->c_laddr; iinfo->dst_addr = conn->c_faddr; memset(&iinfo->src_gid, 0, sizeof(iinfo->src_gid)); memset(&iinfo->dst_gid, 0, sizeof(iinfo->dst_gid)); if (rds_conn_state(conn) == RDS_CONN_UP) { struct rds_ib_device *rds_ibdev; struct rdma_dev_addr *dev_addr; ic = conn->c_transport_data; dev_addr = &ic->i_cm_id->route.addr.dev_addr; rdma_addr_get_sgid(dev_addr, (union ib_gid *) &iinfo->src_gid); rdma_addr_get_dgid(dev_addr, (union ib_gid *) &iinfo->dst_gid); rds_ibdev = ic->rds_ibdev; iinfo->max_send_wr = ic->i_send_ring.w_nr; iinfo->max_recv_wr = ic->i_recv_ring.w_nr; iinfo->max_send_sge = rds_ibdev->max_sge; rds_ib_get_mr_info(rds_ibdev, iinfo); } return 1; }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover20198.53%133.33%
Sean Hefty20.98%133.33%
Zach Brown10.49%133.33%
Total204100.00%3100.00%


static void rds_ib_ic_info(struct socket *sock, unsigned int len, struct rds_info_iterator *iter, struct rds_info_lengths *lens) { rds_for_each_conn_info(sock, len, iter, lens, rds_ib_conn_info_visitor, sizeof(struct rds_info_rdma_connection)); }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover44100.00%1100.00%
Total44100.00%1100.00%

/* * Early RDS/IB was built to only bind to an address if there is an IPoIB * device with that address set. * * If it were me, I'd advocate for something more flexible. Sending and * receiving should be device-agnostic. Transports would try and maintain * connections between peers who have messages queued. Userspace would be * allowed to influence which paths have priority. We could call userspace * asserting this policy "routing". */
static int rds_ib_laddr_check(struct net *net, __be32 addr) { int ret; struct rdma_cm_id *cm_id; struct sockaddr_in sin; /* Create a CMA ID and try to bind it. This catches both * IB and iWARP capable NICs. */ cm_id = rdma_create_id(&init_net, NULL, NULL, RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(cm_id)) return PTR_ERR(cm_id); memset(&sin, 0, sizeof(sin)); sin.sin_family = AF_INET; sin.sin_addr.s_addr = addr; /* rdma_bind_addr will only succeed for IB & iWARP devices */ ret = rdma_bind_addr(cm_id, (struct sockaddr *)&sin); /* due to this, we will claim to support iWARP devices unless we check node_type. */ if (ret || !cm_id->device || cm_id->device->node_type != RDMA_NODE_IB_CA) ret = -EADDRNOTAVAIL; rdsdebug("addr %pI4 ret %d node type %d\n", &addr, ret, cm_id->device ? cm_id->device->node_type : -1); rdma_destroy_id(cm_id); return ret; }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover13185.62%116.67%
Dan Carpenter74.58%116.67%
Sasha Levin53.27%116.67%
Sowmini Varadhan53.27%116.67%
Guy Shapiro31.96%116.67%
Sean Hefty21.31%116.67%
Total153100.00%6100.00%


static void rds_ib_unregister_client(void) { ib_unregister_client(&rds_ib_client); /* wait for rds_ib_dev_free() to complete */ flush_workqueue(rds_wq); }

Contributors

PersonTokensPropCommitsCommitProp
Zach Brown20100.00%1100.00%
Total20100.00%1100.00%


void rds_ib_exit(void) { rds_info_deregister_func(RDS_INFO_IB_CONNECTIONS, rds_ib_ic_info); rds_ib_unregister_client(); rds_ib_destroy_nodev_conns(); rds_ib_sysctl_exit(); rds_ib_recv_exit(); rds_trans_unregister(&rds_ib_transport); rds_ib_mr_exit(); }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover2982.86%120.00%
Santosh Shilimkar38.57%240.00%
Zach Brown38.57%240.00%
Total35100.00%5100.00%

struct rds_transport rds_ib_transport = { .laddr_check = rds_ib_laddr_check, .xmit_path_complete = rds_ib_xmit_path_complete, .xmit = rds_ib_xmit, .xmit_rdma = rds_ib_xmit_rdma, .xmit_atomic = rds_ib_xmit_atomic, .recv_path = rds_ib_recv_path, .conn_alloc = rds_ib_conn_alloc, .conn_free = rds_ib_conn_free, .conn_path_connect = rds_ib_conn_path_connect, .conn_path_shutdown = rds_ib_conn_path_shutdown, .inc_copy_to_user = rds_ib_inc_copy_to_user, .inc_free = rds_ib_inc_free, .cm_initiate_connect = rds_ib_cm_initiate_connect, .cm_handle_connect = rds_ib_cm_handle_connect, .cm_connect_complete = rds_ib_cm_connect_complete, .stats_info_copy = rds_ib_stats_info_copy, .exit = rds_ib_exit, .get_mr = rds_ib_get_mr, .sync_mr = rds_ib_sync_mr, .free_mr = rds_ib_free_mr, .flush_mrs = rds_ib_flush_mrs, .t_owner = THIS_MODULE, .t_name = "infiniband", .t_type = RDS_TRANS_IB };
int rds_ib_init(void) { int ret; INIT_LIST_HEAD(&rds_ib_devices); ret = rds_ib_mr_init(); if (ret) goto out; ret = ib_register_client(&rds_ib_client); if (ret) goto out_mr_exit; ret = rds_ib_sysctl_init(); if (ret) goto out_ibreg; ret = rds_ib_recv_init(); if (ret) goto out_sysctl; rds_trans_register(&rds_ib_transport); rds_info_register_func(RDS_INFO_IB_CONNECTIONS, rds_ib_ic_info); goto out; out_sysctl: rds_ib_sysctl_exit(); out_ibreg: rds_ib_unregister_client(); out_mr_exit: rds_ib_mr_exit(); out: return ret; }

Contributors

PersonTokensPropCommitsCommitProp
Andy Grover8380.58%125.00%
Santosh Shilimkar1817.48%250.00%
Zach Brown21.94%125.00%
Total103100.00%4100.00%

MODULE_LICENSE("GPL");

Overall Contributors

PersonTokensPropCommitsCommitProp
Andy Grover99658.25%720.00%
Zach Brown38222.34%514.29%
Santosh Shilimkar25014.62%617.14%
Or Gerlitz271.58%12.86%
Sowmini Varadhan160.94%514.29%
Dan Carpenter70.41%12.86%
Haggai Eran60.35%12.86%
Sasha Levin50.29%12.86%
Sean Hefty40.23%25.71%
Stephen Hemminger40.23%12.86%
Paul Gortmaker30.18%12.86%
Tejun Heo30.18%12.86%
Guy Shapiro30.18%12.86%
Zhu Yanjun20.12%12.86%
Christoph Hellwig20.12%12.86%
Total1710100.00%35100.00%
Directory: net/rds
Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.
Created with cregit.