X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Ftap%2Ftap_netlink.c;h=75af3404b03593fceeed140b60ababf3a490f99d;hb=a8354c99a912e1c2513e113ae07c9e91237f7951;hp=82c8dc0e63e51aa44ef1088f8fd7fe51a913144f;hpb=acfe2bd440b4e059d871f8ee71abb708c10b8ee8;p=dpdk.git diff --git a/drivers/net/tap/tap_netlink.c b/drivers/net/tap/tap_netlink.c index 82c8dc0e63..75af3404b0 100644 --- a/drivers/net/tap/tap_netlink.c +++ b/drivers/net/tap/tap_netlink.c @@ -1,6 +1,6 @@ /* SPDX-License-Identifier: BSD-3-Clause * Copyright 2017 6WIND S.A. - * Copyright 2017 Mellanox. + * Copyright 2017 Mellanox Technologies, Ltd */ #include @@ -9,11 +9,19 @@ #include #include #include +#include #include #include #include +#include "tap_log.h" + +/* Compatibility with glibc < 2.24 */ +#ifndef SOL_NETLINK +#define SOL_NETLINK 270 +#endif + /* Must be quite large to support dumping a huge list of QDISC or filters. */ #define BUF_SIZE (32 * 1024) /* Size of the buffer to receive kernel messages */ #define SNDBUF_SIZE 32768 /* Send buffer size for the netlink socket */ @@ -42,22 +50,34 @@ tap_nl_init(uint32_t nl_groups) .nl_family = AF_NETLINK, .nl_groups = nl_groups, }; +#ifdef NETLINK_EXT_ACK + int one = 1; +#endif fd = socket(AF_NETLINK, SOCK_RAW | SOCK_CLOEXEC, NETLINK_ROUTE); if (fd < 0) { - RTE_LOG(ERR, PMD, "Unable to create a netlink socket\n"); + TAP_LOG(ERR, "Unable to create a netlink socket"); return -1; } if (setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &sndbuf_size, sizeof(int))) { - RTE_LOG(ERR, PMD, "Unable to set socket buffer send size\n"); + TAP_LOG(ERR, "Unable to set socket buffer send size"); + close(fd); return -1; } if (setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf_size, sizeof(int))) { - RTE_LOG(ERR, PMD, "Unable to set socket buffer receive size\n"); + TAP_LOG(ERR, "Unable to set socket buffer receive size"); + close(fd); return -1; } + +#ifdef NETLINK_EXT_ACK + /* Ask for extended ACK response. on older kernel will ignore request. */ + setsockopt(fd, SOL_NETLINK, NETLINK_EXT_ACK, &one, sizeof(one)); +#endif + if (bind(fd, (struct sockaddr *)&local, sizeof(local)) < 0) { - RTE_LOG(ERR, PMD, "Unable to bind to the netlink socket\n"); + TAP_LOG(ERR, "Unable to bind to the netlink socket"); + close(fd); return -1; } return fd; @@ -76,7 +96,7 @@ int tap_nl_final(int nlsk_fd) { if (close(nlsk_fd)) { - RTE_LOG(ERR, PMD, "Failed to close netlink socket: %s (%d)\n", + TAP_LOG(ERR, "Failed to close netlink socket: %s (%d)", strerror(errno), errno); return -1; } @@ -97,33 +117,92 @@ tap_nl_final(int nlsk_fd) int tap_nl_send(int nlsk_fd, struct nlmsghdr *nh) { - /* man 7 netlink EXAMPLE */ - struct sockaddr_nl sa = { - .nl_family = AF_NETLINK, - }; - struct iovec iov = { - .iov_base = nh, - .iov_len = nh->nlmsg_len, - }; - struct msghdr msg = { - .msg_name = &sa, - .msg_namelen = sizeof(sa), - .msg_iov = &iov, - .msg_iovlen = 1, - }; int send_bytes; nh->nlmsg_pid = 0; /* communication with the kernel uses pid 0 */ nh->nlmsg_seq = (uint32_t)rte_rand(); - send_bytes = sendmsg(nlsk_fd, &msg, 0); + +retry: + send_bytes = send(nlsk_fd, nh, nh->nlmsg_len, 0); if (send_bytes < 0) { - RTE_LOG(ERR, PMD, "Failed to send netlink message: %s (%d)\n", + if (errno == EINTR) + goto retry; + + TAP_LOG(ERR, "Failed to send netlink message: %s (%d)", strerror(errno), errno); return -1; } return send_bytes; } +#ifdef NETLINK_EXT_ACK +static const struct nlattr * +tap_nl_attr_first(const struct nlmsghdr *nh, size_t offset) +{ + return (const struct nlattr *)((const char *)nh + NLMSG_SPACE(offset)); +} + +static const struct nlattr * +tap_nl_attr_next(const struct nlattr *attr) +{ + return (const struct nlattr *)((const char *)attr + + NLMSG_ALIGN(attr->nla_len)); +} + +static bool +tap_nl_attr_ok(const struct nlattr *attr, int len) +{ + if (len < (int)sizeof(struct nlattr)) + return false; /* missing header */ + if (attr->nla_len < sizeof(struct nlattr)) + return false; /* attribute length should include itself */ + if ((int)attr->nla_len > len) + return false; /* attribute is truncated */ + return true; +} + + +/* Decode extended errors from kernel */ +static void +tap_nl_dump_ext_ack(const struct nlmsghdr *nh, const struct nlmsgerr *err) +{ + const struct nlattr *attr; + const char *tail = (const char *)nh + NLMSG_ALIGN(nh->nlmsg_len); + size_t hlen = sizeof(*err); + + /* no TLVs, no extended response */ + if (!(nh->nlmsg_flags & NLM_F_ACK_TLVS)) + return; + + if (!(nh->nlmsg_flags & NLM_F_CAPPED)) + hlen += err->msg.nlmsg_len - NLMSG_HDRLEN; + + for (attr = tap_nl_attr_first(nh, hlen); + tap_nl_attr_ok(attr, tail - (const char *)attr); + attr = tap_nl_attr_next(attr)) { + uint16_t type = attr->nla_type & NLA_TYPE_MASK; + + if (type == NLMSGERR_ATTR_MSG) { + const char *msg = (const char *)attr + + NLMSG_ALIGN(sizeof(*attr)); + + if (err->error) + TAP_LOG(ERR, "%s", msg); + else + + TAP_LOG(WARNING, "%s", msg); + break; + } + } +} +#else +/* + * External ACK support was added in Linux kernel 4.17 + * on older kernels, just ignore that part of message + */ +#define tap_nl_dump_ext_ack(nh, err) do { } while (0) +#endif + /** * Check that the kernel sends an appropriate ACK in response * to an tap_nl_send(). @@ -157,36 +236,29 @@ tap_nl_recv_ack(int nlsk_fd) int tap_nl_recv(int nlsk_fd, int (*cb)(struct nlmsghdr *, void *arg), void *arg) { - /* man 7 netlink EXAMPLE */ - struct sockaddr_nl sa; char buf[BUF_SIZE]; - struct iovec iov = { - .iov_base = buf, - .iov_len = sizeof(buf), - }; - struct msghdr msg = { - .msg_name = &sa, - .msg_namelen = sizeof(sa), - .msg_iov = &iov, - /* One message at a time */ - .msg_iovlen = 1, - }; int multipart = 0; int ret = 0; do { struct nlmsghdr *nh; - int recv_bytes = 0; + int recv_bytes; - recv_bytes = recvmsg(nlsk_fd, &msg, 0); - if (recv_bytes < 0) +retry: + recv_bytes = recv(nlsk_fd, buf, sizeof(buf), 0); + if (recv_bytes < 0) { + if (errno == EINTR) + goto retry; return -1; + } + for (nh = (struct nlmsghdr *)buf; NLMSG_OK(nh, (unsigned int)recv_bytes); nh = NLMSG_NEXT(nh, recv_bytes)) { if (nh->nlmsg_type == NLMSG_ERROR) { struct nlmsgerr *err_data = NLMSG_DATA(nh); + tap_nl_dump_ext_ack(nh, err_data); if (err_data->error < 0) { errno = -err_data->error; return -1; @@ -300,9 +372,8 @@ tap_nlattr_nested_start(struct nlmsg *msg, uint16_t type) tail = rte_zmalloc(NULL, sizeof(struct nested_tail), 0); if (!tail) { - RTE_LOG(ERR, PMD, - "Couldn't allocate memory for nested netlink" - " attribute\n"); + TAP_LOG(ERR, + "Couldn't allocate memory for nested netlink attribute"); return -1; }