【RDMA】RDMA 编程实例(rdma_cm API)

目录

RDMA编程基础

说明

1. RDMA的学习环境搭建

2. RDMA与socket的类比

3. RDMA服务器的代码流程

main()

{

}

实例

用法

Makefile

服务端server.c

客户端client.c

更多讲解教程

WRITE|READ编程(RDMA read and write with IB verbs)

LINUX 编程例子


作者:bandaoyu  链接:https://blog.csdn.net/bandaoyu/article/details/116062334

RDMA编程基础

存储大师班 | RDMA简介与编程基础 -https://zhuanlan.zhihu.com/p/387549948

bandaoyuRDMA编程入门

说明

1. RDMA的学习环境搭建

RDMA需要专门的RDMA网卡或者InfiniBand卡才能使用,学习RDMA而又没有这些硬件设备,可以使用一个软件RDMA模拟环境,softiwarp ,
- 这是加载地址:https://github.com/zrlio/softiwarp
- 这是安装教程:http://www.reflectionsofthevoid.com/2011/03/how-to-install-soft-iwarp-on-ubuntu.html

更多的rdmacm实例:,
https://github.com/tarickb/the-geek-in-the-corner
需要注意的是,这个例子里面缺省用的是IPv6连接,如果希望在IPv4环境下测试,需要先改代码用IPv4地址。


2. RDMA与socket的类比

和Socket连接类似,RDMA连接也分为可靠连接和不可靠连接。然而也不完全相同,Socket的可靠连接就是TCP连接,是流式的;不可靠连接也就是UDP,是消息式的。对于RDMA来说,无论是可靠连接和不可靠连接,都是消息式的。
编程角度看,RDMA代码也分为Server端,Client端,也有bind, listen, connect, accept,等动作,然而细节上仍有不少区别。

大家可以关注一下mellonx的vma,貌似可以直接用socket api通信,方便很多:【RDMA】降低CPU除了RDMA (vbers)还是VMA ?|使用socket进行RDMA编程?_bandaoyu的note-CSDN博客前言看介绍,像是mellonx针对其kernel bypass网卡(RDMA网卡)提供的一个lib库,该lib库对外提供socket api,使得用户的程序不需要修改就可以直接使用kernel bypass网卡(如RDMA网卡)。我们都知道RDMA 网卡目前使用的是rdma_cm和vbers api编程,和socket不一样,如果能用socket对RDMA编程,那确实是很大的利好。官网介绍什么是VMA?Mellanox Interconnect Community官方介绍:Mhttps://blog.csdn.net/bandaoyu/article/details/120726746

rdma_cm API说明:

https://linux.die.net/man/3/rdma_create_id (推荐)

https://www.ibm.com/docs/en/aix/7.2?topic=operations-rdma-listen (内容少)

rdma_cm API 管理连接(建立连接和销毁)+vbers api 管理收发

        RDMA主机使用queue pairs(QP)进行通信;主机创建由发送队列SQ和接收队列RQ组成的QP,并使用verbs API将操作post 到这些队列。(所以rdma_cm是管理连接的,收发还是verbs API?)

3. RDMA服务器的代码流程

main()

{

  1. channel=rdma_create_event_channel
    这一步是创建一个event channel,event channel是RDMA设备在操作完成后,或者有连接请求等事件发生时,用来通知应用程序的通道。其内部就是一个file descriptor, 因此可以进行poll等操作。

  2. rdma_create_id(channel, **id,……)
    这一步创建一个rdma_cm_id, 概念上等价与socket编程时的listen socket。

  3. rdma_bind_addr(id,addr)
    和socket编程一样,也要先绑定一个本地的地址和端口,以进行listen操作。

  4. rdma_listen(id,block)
    开始侦听客户端的连接请求

  5. rdma_get_cm_event(channel,&event)
    这个调用就是作用在第一步创建的event channel上面,要从event channel中获取一个事件。这是个阻塞调用,只有有事件时才会返回。在一切正常的情况下,函数返回时会得到一个 RDMA_CM_EVENT_CONNECT_REQUEST事件,也就是说,有客户端发起连接了。
    在事件的参数里面,会有一个新的rdma_cm_id传入。这点和socket是不同的,socket只有在accept后才有新的socket fd创建。

on_event()

{

     on_connect_request()//RDMA_CM_EVENT_CONNECT_REQUEST

    {

        build_context()

      {

6.ibv_alloc_pd

创建一个protection domain。protection domain可以看作是一个内存保护单位,在内存区域和队列直接建立一个关联关系,防止未授权的访问。

7.ibv_create_comp_channel

和之前创建的event channel类似,这也是一个event channel,但只用来报告【完成队列】里面的事件。当【完成队列】里有新的任务完成时,就通过这个channel向应用程序报告。

8.ibv_create_cq

创建【完成队列】,创建时就指定使用第6步的channel。

}//--end build_context()

9.rdma_create_qp
创建一个queue pair, 一个queue pair包括一个发送queue和一个接收queue. 指定使用前面创建的cq作为完成队列。该qp创建时就指定关联到第6步创建的pd上。

10.ibv_reg_mr
注册内存区域。RDMA使用的内存,必须事先进行注册。这个是可以理解的,DMA的内存在边界对齐,能否被swap等方面,都有要求。

11.rdma_accept
至此,做好了全部的准备工作,可以调用accept接受客户端的这个请求了。 –:)长出一口气 ~~ 且慢,

}

//--end on_connect_request()

12.rdma_ack_cm_event
对于每个从event channel得到的事件,都要调用ack函数,否则会产生内存泄漏。这一步的ack是对应第5步的get。每一次get调用,都要有对应的ack调用。

13.rdma_get_cm_event
继续调用rdma_get_cm_event, 一切正常的话我们此时应该得到 RDMA_CM_EVENT_ESTABLISHED 事件,表示连接已经建立起来。不需要做额外的处理,直接rdma_ack_cm_event就行了

}//--end on_event()

终于可以开始进行数据传输了 ==== (如何传输下篇再说)

参考:http://10.165.104.246:8080/#/c/43882/

4. 关闭连接

  1. 断开连接
    rdma_get_cm_event返回RDMA_CM_EVENT_DISCONNECTED事件时,表示客户端断开了连接,server端要进行对应的清理。此时可以调用rdma_ack_cm_event释放事件资源。然后依次调用下面的函数,释放连接资源,内存资源,队列资源。

  2. rdma_disconnect

  3. rdma_destroy_qp

  4. ibv_dereg_mr

  5. rdma_destroy_id
    释放同客户端连接的rdma_cm_id

  6. rdma_destroy_id
    释放用于侦听的rdma_cm_id

  7. rdma_destroy_event_channel
    释放 event channel

}

// end main

实例

源码地址- https://github.com/tarickb/the-geek-in-the-corner

用法

[root@localhost 01_basic-client-server]# ./server 
listening on port 42956.

client <server-address> <server-port>
 

Makefile

.PHONY: clean

CFLAGS  := -Wall -g
LDLIBS  := ${LDLIBS} -lrdmacm -libverbs -lpthread

APPS    := server client

all: ${APPS}


clean:
    rm -f ${APPS}

注意:makefile 没有-L 指定lib的路径,所以 -lrdmacm -libverbs -lpthread 对应的库 librdmacm.so libibverbs.so libpthread.so 应放在默认的路径下/usr/lib 或/usr/lib64

server、clicent源码下载:RDMA编程实例源码(rdma_cmAPI)_rdma源码,rdma实例-网络基础文档类资源-CSDN下载

服务端server.c

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <rdma/rdma_cma.h>

#define TEST_NZ(x) do { if ( (x)) die("error: " #x " failed (returned non-zero)." ); } while (0)
#define TEST_Z(x)  do { if (!(x)) die("error: " #x " failed (returned zero/null)."); } while (0)

const int BUFFER_SIZE = 1024;

struct context {
  struct ibv_context *ctx;
  struct ibv_pd *pd;
  struct ibv_cq *cq;
  struct ibv_comp_channel *comp_channel;

  pthread_t cq_poller_thread;
};

struct connection {
  struct ibv_qp *qp;

  struct ibv_mr *recv_mr;
  struct ibv_mr *send_mr;

  char *recv_region;
  char *send_region;
};

static void die(const char *reason);

static void build_context(struct ibv_context *verbs);
static void build_qp_attr(struct ibv_qp_init_attr *qp_attr);
static void * poll_cq(void *);
static void post_receives(struct connection *conn);
static void register_memory(struct connection *conn);

static void on_completion(struct ibv_wc *wc);
static int on_connect_request(struct rdma_cm_id *id);
static int on_connection(void *context);
static int on_disconnect(struct rdma_cm_id *id);
static int on_event(struct rdma_cm_event *event);

static struct context *s_ctx = NULL;

int main(int argc, char **argv)
{
#if _USE_IPV6
  struct sockaddr_in6 addr;
#else
  struct sockaddr_in addr;
#endif
  struct rdma_cm_event *event = NULL;
  struct rdma_cm_id *listener = NULL;
  struct rdma_event_channel *ec = NULL;
  uint16_t port = 0;

  memset(&addr, 0, sizeof(addr));
#if _USE_IPV6
  addr.sin6_family = AF_INET6;
#else
  addr.sin_family = AF_INET;
#endif

  TEST_Z(ec = rdma_create_event_channel());
  TEST_NZ(rdma_create_id(ec, &listener, NULL, RDMA_PS_TCP));
  TEST_NZ(rdma_bind_addr(listener, (struct sockaddr *)&addr));
  TEST_NZ(rdma_listen(listener, 10)); /* backlog=10 is arbitrary */

  port = ntohs(rdma_get_src_port(listener)); //rdma_get_src_port 返回listener对应的tcp 端口

  printf("listening on port %d.\n", port);

  while (rdma_get_cm_event(ec, &event) == 0) {
    struct rdma_cm_event event_copy;

    memcpy(&event_copy, event, sizeof(*event));
    rdma_ack_cm_event(event);

    if (on_event(&event_copy))
      break;
  }

  rdma_destroy_id(listener);
  rdma_destroy_event_channel(ec);

  return 0;
}

void die(const char *reason)
{
  fprintf(stderr, "%s\n", reason);
  exit(EXIT_FAILURE);
}

void build_context(struct ibv_context *verbs)
{
  if (s_ctx) {
    if (s_ctx->ctx != verbs)
      die("cannot handle events in more than one context.");

    return;
  }

  s_ctx = (struct context *)malloc(sizeof(struct context));

  s_ctx->ctx = verbs;

  TEST_Z(s_ctx->pd = ibv_alloc_pd(s_ctx->ctx));
  TEST_Z(s_ctx->comp_channel = ibv_create_comp_channel(s_ctx->ctx));
  TEST_Z(s_ctx->cq = ibv_create_cq(s_ctx->ctx, 10, NULL, s_ctx->comp_channel, 0)); /* cqe=10 is arbitrary */
  TEST_NZ(ibv_req_notify_cq(s_ctx->cq, 0)); #完成完成队列与完成通道的关联

  TEST_NZ(pthread_create(&s_ctx->cq_poller_thread, NULL, poll_cq, NULL));
}

void build_qp_attr(struct ibv_qp_init_attr *qp_attr)
{
  memset(qp_attr, 0, sizeof(*qp_attr));

  qp_attr->send_cq = s_ctx->cq;
  qp_attr->recv_cq = s_ctx->cq;
  qp_attr->qp_type = IBV_QPT_RC;

  qp_attr->cap.max_send_wr = 10;
  qp_attr->cap.max_recv_wr = 10;
  qp_attr->cap.max_send_sge = 1;
  qp_attr->cap.max_recv_sge = 1;
}

void * poll_cq(void *ctx)
{
  struct ibv_cq *cq;
  struct ibv_wc wc;

  while (1) {
    TEST_NZ(ibv_get_cq_event(s_ctx->comp_channel, &cq, &ctx));
    ibv_ack_cq_events(cq, 1);
    TEST_NZ(ibv_req_notify_cq(cq, 0));

    while (ibv_poll_cq(cq, 1, &wc))
      on_completion(&wc);
  }

  return NULL;
}

void post_receives(struct connection *conn)
{
  struct ibv_recv_wr wr, *bad_wr = NULL;
  struct ibv_sge sge;

  wr.wr_id = (uintptr_t)conn;
  wr.next = NULL;
  wr.sg_list = &sge;
  wr.num_sge = 1;

  sge.addr = (uintptr_t)conn->recv_region;
  sge.length = BUFFER_SIZE;
  sge.lkey = conn->recv_mr->lkey;

  TEST_NZ(ibv_post_recv(conn->qp, &wr, &bad_wr));
}

void register_memory(struct connection *conn)
{
  conn->send_region = malloc(BUFFER_SIZE);
  conn->recv_region = malloc(BUFFER_SIZE);

  TEST_Z(conn->send_mr = ibv_reg_mr(
    s_ctx->pd,
    conn->send_region,
    BUFFER_SIZE,
    0));

  TEST_Z(conn->recv_mr = ibv_reg_mr(
    s_ctx->pd,
    conn->recv_region,
    BUFFER_SIZE,
    IBV_ACCESS_LOCAL_WRITE));
}

void on_completion(struct ibv_wc *wc)
{
  if (wc->status != IBV_WC_SUCCESS)
    die("on_completion: status is not IBV_WC_SUCCESS.");

  if (wc->opcode & IBV_WC_RECV) {
    struct connection *conn = (struct connection *)(uintptr_t)wc->wr_id;

    printf("received message: %s\n", conn->recv_region);

  } else if (wc->opcode == IBV_WC_SEND) {
    printf("send completed successfully.\n");
  }
}

int on_connect_request(struct rdma_cm_id *id)
{
  struct ibv_qp_init_attr qp_attr;
  struct rdma_conn_param cm_params;
  struct connection *conn;

  printf("received connection request.\n");

  build_context(id->verbs);
  build_qp_attr(&qp_attr);

  TEST_NZ(rdma_create_qp(id, s_ctx->pd, &qp_attr));

  id->context = conn = (struct connection *)malloc(sizeof(struct connection));
  conn->qp = id->qp;

  register_memory(conn);
  post_receives(conn);

  memset(&cm_params, 0, sizeof(cm_params));
  TEST_NZ(rdma_accept(id, &cm_params));

  return 0;
}

int on_connection(void *context)
{
  struct connection *conn = (struct connection *)context;
  struct ibv_send_wr wr, *bad_wr = NULL;
  struct ibv_sge sge;

  snprintf(conn->send_region, BUFFER_SIZE, "message from passive/server side with pid %d", getpid());

  printf("connected. posting send...\n");

  memset(&wr, 0, sizeof(wr));

  wr.opcode = IBV_WR_SEND;
  wr.sg_list = &sge;
  wr.num_sge = 1;
  wr.send_flags = IBV_SEND_SIGNALED;

  sge.addr = (uintptr_t)conn->send_region;
  sge.length = BUFFER_SIZE;
  sge.lkey = conn->send_mr->lkey;

  TEST_NZ(ibv_post_send(conn->qp, &wr, &bad_wr));

  return 0;
}

int on_disconnect(struct rdma_cm_id *id)
{
  struct connection *conn = (struct connection *)id->context;

  printf("peer disconnected.\n");

  rdma_destroy_qp(id);

  ibv_dereg_mr(conn->send_mr);
  ibv_dereg_mr(conn->recv_mr);

  free(conn->send_region);
  free(conn->recv_region);

  free(conn);

  rdma_destroy_id(id);

  return 0;
}

int on_event(struct rdma_cm_event *event)
{
  int r = 0;

  if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST)
    r = on_connect_request(event->id);
  else if (event->event == RDMA_CM_EVENT_ESTABLISHED)
    r = on_connection(event->id->context);
  else if (event->event == RDMA_CM_EVENT_DISCONNECTED)
    r = on_disconnect(event->id);
  else
    die("on_event: unknown event.");

  return r;
}

客户端client.c

#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <rdma/rdma_cma.h>

#define TEST_NZ(x) do { if ( (x)) die("error: " #x " failed (returned non-zero)." ); } while (0)
#define TEST_Z(x)  do { if (!(x)) die("error: " #x " failed (returned zero/null)."); } while (0)

const int BUFFER_SIZE = 1024;
const int TIMEOUT_IN_MS = 500; /* ms */

struct context {
  struct ibv_context *ctx;
  struct ibv_pd *pd;
  struct ibv_cq *cq;
  struct ibv_comp_channel *comp_channel;

  pthread_t cq_poller_thread;
};

struct connection {
  struct rdma_cm_id *id;
  struct ibv_qp *qp;

  struct ibv_mr *recv_mr;
  struct ibv_mr *send_mr;

  char *recv_region;
  char *send_region;

  int num_completions;
};

static void die(const char *reason);

static void build_context(struct ibv_context *verbs);
static void build_qp_attr(struct ibv_qp_init_attr *qp_attr);
static void * poll_cq(void *);
static void post_receives(struct connection *conn);
static void register_memory(struct connection *conn);

static int on_addr_resolved(struct rdma_cm_id *id);
static void on_completion(struct ibv_wc *wc);
static int on_connection(void *context);
static int on_disconnect(struct rdma_cm_id *id);
static int on_event(struct rdma_cm_event *event);
static int on_route_resolved(struct rdma_cm_id *id);

static struct context *s_ctx = NULL;

int main(int argc, char **argv)
{
  struct addrinfo *addr;
  struct rdma_cm_event *event = NULL;
  struct rdma_cm_id *conn= NULL;
  struct rdma_event_channel *ec = NULL;

  if (argc != 3)
    die("usage: client <server-address> <server-port>");

  TEST_NZ(getaddrinfo(argv[1], argv[2], NULL, &addr));

  TEST_Z(ec = rdma_create_event_channel());
  TEST_NZ(rdma_create_id(ec, &conn, NULL, RDMA_PS_TCP));
  TEST_NZ(rdma_resolve_addr(conn, NULL, addr->ai_addr, TIMEOUT_IN_MS));

  freeaddrinfo(addr);

  while (rdma_get_cm_event(ec, &event) == 0) {
    struct rdma_cm_event event_copy;

    memcpy(&event_copy, event, sizeof(*event));
    rdma_ack_cm_event(event);

    if (on_event(&event_copy))
      break;
  }

  rdma_destroy_event_channel(ec);

  return 0;
}

void die(const char *reason)
{
  fprintf(stderr, "%s\n", reason);
  exit(EXIT_FAILURE);
}

void build_context(struct ibv_context *verbs)
{
  if (s_ctx) {
    if (s_ctx->ctx != verbs)
      die("cannot handle events in more than one context.");

    return;
  }

  s_ctx = (struct context *)malloc(sizeof(struct context));

  s_ctx->ctx = verbs;

  TEST_Z(s_ctx->pd = ibv_alloc_pd(s_ctx->ctx));
  TEST_Z(s_ctx->comp_channel = ibv_create_comp_channel(s_ctx->ctx));
  TEST_Z(s_ctx->cq = ibv_create_cq(s_ctx->ctx, 10, NULL, s_ctx->comp_channel, 0)); /* cqe=10 is arbitrary */
  TEST_NZ(ibv_req_notify_cq(s_ctx->cq, 0));

  TEST_NZ(pthread_create(&s_ctx->cq_poller_thread, NULL, poll_cq, NULL));
}

void build_qp_attr(struct ibv_qp_init_attr *qp_attr)
{
  memset(qp_attr, 0, sizeof(*qp_attr));

  qp_attr->send_cq = s_ctx->cq;
  qp_attr->recv_cq = s_ctx->cq;
  qp_attr->qp_type = IBV_QPT_RC;

  qp_attr->cap.max_send_wr = 10;
  qp_attr->cap.max_recv_wr = 10;
  qp_attr->cap.max_send_sge = 1;
  qp_attr->cap.max_recv_sge = 1;
}

void * poll_cq(void *ctx)
{
  struct ibv_cq *cq;
  struct ibv_wc wc;

  while (1) {
    TEST_NZ(ibv_get_cq_event(s_ctx->comp_channel, &cq, &ctx));
    ibv_ack_cq_events(cq, 1);
    TEST_NZ(ibv_req_notify_cq(cq, 0));

    while (ibv_poll_cq(cq, 1, &wc))
      on_completion(&wc);
  }

  return NULL;
}

void post_receives(struct connection *conn)
{
  struct ibv_recv_wr wr, *bad_wr = NULL;
  struct ibv_sge sge;

  wr.wr_id = (uintptr_t)conn;
  wr.next = NULL;
  wr.sg_list = &sge;
  wr.num_sge = 1;

  sge.addr = (uintptr_t)conn->recv_region;
  sge.length = BUFFER_SIZE;
  sge.lkey = conn->recv_mr->lkey;

  TEST_NZ(ibv_post_recv(conn->qp, &wr, &bad_wr));
}

void register_memory(struct connection *conn)
{
  conn->send_region = malloc(BUFFER_SIZE);
  conn->recv_region = malloc(BUFFER_SIZE);

  TEST_Z(conn->send_mr = ibv_reg_mr(
    s_ctx->pd, 
    conn->send_region, 
    BUFFER_SIZE, 
    0));

  TEST_Z(conn->recv_mr = ibv_reg_mr(
    s_ctx->pd, 
    conn->recv_region, 
    BUFFER_SIZE, 
    IBV_ACCESS_LOCAL_WRITE));
}

int on_addr_resolved(struct rdma_cm_id *id)
{
  struct ibv_qp_init_attr qp_attr;
  struct connection *conn;

  printf("address resolved.\n");

  build_context(id->verbs);
  build_qp_attr(&qp_attr);

  TEST_NZ(rdma_create_qp(id, s_ctx->pd, &qp_attr));

  id->context = conn = (struct connection *)malloc(sizeof(struct connection));

  conn->id = id;
  conn->qp = id->qp;
  conn->num_completions = 0;

  register_memory(conn);
  post_receives(conn);

  TEST_NZ(rdma_resolve_route(id, TIMEOUT_IN_MS));

  return 0;
}

void on_completion(struct ibv_wc *wc)
{
  struct connection *conn = (struct connection *)(uintptr_t)wc->wr_id;

  if (wc->status != IBV_WC_SUCCESS)
    die("on_completion: status is not IBV_WC_SUCCESS.");

  if (wc->opcode & IBV_WC_RECV)
    printf("received message: %s\n", conn->recv_region);
  else if (wc->opcode == IBV_WC_SEND)
    printf("send completed successfully.\n");
  else
    die("on_completion: completion isn't a send or a receive.");

  if (++conn->num_completions == 2)
    rdma_disconnect(conn->id);
}

int on_connection(void *context)
{
  struct connection *conn = (struct connection *)context;
  struct ibv_send_wr wr, *bad_wr = NULL;
  struct ibv_sge sge;

  snprintf(conn->send_region, BUFFER_SIZE, "message from active/client side with pid %d", getpid());

  printf("connected. posting send...\n");

  memset(&wr, 0, sizeof(wr));

  wr.wr_id = (uintptr_t)conn;
  wr.opcode = IBV_WR_SEND;
  wr.sg_list = &sge;
  wr.num_sge = 1;
  wr.send_flags = IBV_SEND_SIGNALED;

  sge.addr = (uintptr_t)conn->send_region;
  sge.length = BUFFER_SIZE;
  sge.lkey = conn->send_mr->lkey;

  TEST_NZ(ibv_post_send(conn->qp, &wr, &bad_wr));

  return 0;
}

int on_disconnect(struct rdma_cm_id *id)
{
  struct connection *conn = (struct connection *)id->context;

  printf("disconnected.\n");

  rdma_destroy_qp(id);

  ibv_dereg_mr(conn->send_mr);
  ibv_dereg_mr(conn->recv_mr);

  free(conn->send_region);
  free(conn->recv_region);

  free(conn);

  rdma_destroy_id(id);

  return 1; /* exit event loop */
}

int on_event(struct rdma_cm_event *event)
{
  int r = 0;

  if (event->event == RDMA_CM_EVENT_ADDR_RESOLVED)
    r = on_addr_resolved(event->id);
  else if (event->event == RDMA_CM_EVENT_ROUTE_RESOLVED)
    r = on_route_resolved(event->id);
  else if (event->event == RDMA_CM_EVENT_ESTABLISHED)
    r = on_connection(event->id->context);
  else if (event->event == RDMA_CM_EVENT_DISCONNECTED)
    r = on_disconnect(event->id);
  else
    die("on_event: unknown event.");

  return r;
}

int on_route_resolved(struct rdma_cm_id *id)
{
  struct rdma_conn_param cm_params;

  printf("route resolved.\n");

  memset(&cm_params, 0, sizeof(cm_params));
  TEST_NZ(rdma_connect(id, &cm_params));

  return 0;
}

更多讲解教程

InfiniBand, Verbs, RDMA | https://thegeekinthecorner.wordpress.com/category/infiniband-verbs-rdma/

RDMA read and write with IB verbs | https://thegeekinthecorner.wordpress.com/2010/09/28/rdma-read-and-write-with-ib-verbs/

http://www.hpcadvisorycouncil.com/pdf/building-an-rdma-capable-application-with-ib-verbs.pdf

WRITE|READ编程(RDMA read and write with IB verbs)

(本文讲解的示例代码在:RDMA read and write with IB verbs | The Geek in the Corner


在《4. RDMA操作类型|WRITE|READ》后半段:https://blog.csdn.net/bandaoyu/article/details/112861454

LINUX 编程例子

https://community.mellanox.com/s/topic/0TO50000000g1zhGAA/linux-programming?tabset-dea0d=2

4、rdma_listen()和rdma_request()是否要在一起执行?

不一定。rdma_listen()不是阻塞式方法,只是监听连接请求,不会产生新的连接rdma_cm_id,如果想产生新的rdma_cm_id,需要执行rdma_get_cm_event()获取RDMA_CM_EVENT_CONNECT_REQUEST事件,或者执行rdma_get_request()获取连接请求事件。所以rdma_get_cm_event()和rdma_get_request()的原理是相同的,都是为了获取RDMA_CM_EVENT_CONNECT_REQUEST事件,表示有客户端连接,然后会产生新的rdma_cm_id,也就是连接rdma_cm_id。对于传统的TCP/IP通信,listen阶段只是监听,accept之后才产生新的socket fd,但是RDMA中,在listen阶段即可产生新的rdma_cm_id,更为准确来说,是在listen调用之后,只要有RDMA_CM_EVENT_CONNECT_REQUEST事件产生,就会生成新的rdma_cm_id。尝试过,如果只有listen,没有request,就会报错。可见,listen之后需要有相应的处理机制才可以。

5、为什么要设置IBV_SEND_INLINE?

原因:https://blog.csdn.net/bandaoyu/article/details/119207147

int send_flags,描述了WR的属性,其值为0或者一个或多个flags的按位异或。

IBV_SEND_FENCE - 为此WR设置围栏指示器。这意味着这个WR的处理将被阻止,直到所有之前发布的RDMA Read和Atomic WR都将被完成。仅对运输服务类型为IBV_QPT_RC的QP有效

IBV_SEND_SIGNALED - 设置此WR的完成通知指示符。这意味着如果QP是使用sq_sig_all = 0创建的,那么当WR的处理结束时,将会产生一个工作完成。如果QP是使用sq_sig_all = 1创建的,则不会对此标志产生任何影响

IBV_SEND_SOLICITED - 为此WR设置请求事件指示器。这意味着,当这个WR中的消息将在远程QP中结束时,将会创建一个请求事件,如果在远程侧,用户正在等待请求事件,它将被唤醒。与仅用于发送和RDMA写入的立即操作码相关

IBV_SEND_INLINE - sg_list中指定的内存缓冲区将内联放置在发送请求中。这意味着低级驱动程序(即CPU)将读取数据而不是RDMA设备。这意味着L_Key将不会被检查,实际上这些内存缓冲区甚至不需要被注册,并且在ibv_post_send()结束之后可以立即重用。仅对发送和RDMA写操作码有效。由于在该代码中没有涉及到key的交换,所以也无法使用RDMA传输,所以还是使用了CPU读取数据,既然是CPU读取,那么也就不需要注册内存缓冲区了,这个标记只能用于发送和写操作。

6、操作码和对应的QP传输服务类型的关系?

@UESTC

CQE's wr_id could be:
1)BEACON_WRID
2)&RDMAConnectedSocketImpl::qp
3)Chunks address start from Cluster::chunk_base
When assuming qp as Chunk through CQE's wr_id, it's possible to misjudge
&(qp->ib_physical_port) into Cluster::[base, end) because there're 4 bytes
random data filled in the higher 4 bytes address around ib_pysical_port due
to the address alignement requirement of structure member.

Fix this case by checking whether wr_id value is in the allocated Chunk space.
Fixes: https://tracker.ceph.com/issues/44346

msg/async/rdma: use wr_id address to check valid chunk by rosinL ·https://github.com/ceph/ceph/pull/36908

posted on 2022-10-04 01:23  bdy  阅读(885)  评论(0编辑  收藏  举报

导航