From patchwork Thu Feb 1 20:51:16 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Steve Wise X-Patchwork-Id: 10217449 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 5FC0B60329 for ; Tue, 13 Feb 2018 20:10:42 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 4E41028DC9 for ; Tue, 13 Feb 2018 20:10:42 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 4317628EE3; Tue, 13 Feb 2018 20:10:42 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-3.5 required=2.0 tests=BAYES_00, DATE_IN_PAST_96_XX, RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 6341A28DC9 for ; Tue, 13 Feb 2018 20:10:41 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S965625AbeBMUKj (ORCPT ); Tue, 13 Feb 2018 15:10:39 -0500 Received: from opengridcomputing.com ([70.118.0.34]:40003 "EHLO smtp.opengridcomputing.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965745AbeBMUKi (ORCPT ); Tue, 13 Feb 2018 15:10:38 -0500 Received: by smtp.opengridcomputing.com (Postfix, from userid 503) id CEC922B9D2; Tue, 13 Feb 2018 14:10:37 -0600 (CST) Message-Id: In-Reply-To: References: From: Steve Wise Date: Thu, 1 Feb 2018 12:51:16 -0800 Subject: [PATCH v1 rdma-next 5/6] RDMA/nldev: provide detailed MR information To: jgg@mellanox.com, dledford@redhat.com Cc: linux-rdma@vger.kernel.org, leon@kernel.org Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Implement the RDMA nldev netlink interface for dumping detailed MR information. Signed-off-by: Steve Wise --- drivers/infiniband/core/nldev.c | 174 +++++++++++++++++++++++++++++++++++ drivers/infiniband/core/restrack.c | 48 +++++++--- drivers/infiniband/core/uverbs_cmd.c | 6 ++ drivers/infiniband/core/verbs.c | 3 + include/rdma/ib_verbs.h | 5 + include/rdma/restrack.h | 4 + include/uapi/rdma/rdma_netlink.h | 10 ++ 7 files changed, 238 insertions(+), 12 deletions(-) diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c index 34fb0d3..8d96f3e 100644 --- a/drivers/infiniband/core/nldev.c +++ b/drivers/infiniband/core/nldev.c @@ -94,6 +94,13 @@ [RDMA_NLDEV_ATTR_RES_CQE] = { .type = NLA_U32 }, [RDMA_NLDEV_ATTR_RES_USECNT] = { .type = NLA_U64 }, [RDMA_NLDEV_ATTR_RES_POLL_CTX] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_MR] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_MR_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_RKEY] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_LKEY] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_IOVA] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_RES_MRLEN] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_RES_PGSIZE] = { .type = NLA_U32 }, }; static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device) @@ -206,6 +213,7 @@ static int fill_res_info(struct sk_buff *msg, struct ib_device *device) [RDMA_RESTRACK_CQ] = "cq", [RDMA_RESTRACK_QP] = "qp", [RDMA_RESTRACK_CM_ID] = "cm_id", + [RDMA_RESTRACK_MR] = "mr", }; struct rdma_restrack_root *res = &device->res; @@ -446,6 +454,51 @@ static int fill_res_cq_entry(struct sk_buff *msg, return -EMSGSIZE; } +static int fill_res_mr_entry(struct sk_buff *msg, + struct ib_mr *mr) +{ + struct rdma_restrack_entry *res = &mr->res; + struct nlattr *entry_attr; + + entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_MR_ENTRY); + if (!entry_attr) + goto out; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RKEY, mr->rkey)) + goto err; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LKEY, mr->lkey)) + goto err; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_IOVA, mr->iova, 0)) + goto err; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_MRLEN, mr->length, 0)) + goto err; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PGSIZE, mr->page_size)) + goto err; + + /* + * Existence of task means that it is user MR and netlink + * user is invited to go and read /proc/PID/comm to get name + * of the task file and res->task_com should be NULL. + */ + if (rdma_is_kernel_res(res)) { + if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME, + res->kern_name)) + goto err; + } else { + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID, + task_pid_vnr(res->task))) + goto err; + } + + nla_nest_end(msg, entry_attr); + return 0; + +err: + nla_nest_cancel(msg, entry_attr); +out: + return -EMSGSIZE; +} + static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, struct netlink_ext_ack *extack) { @@ -1098,6 +1151,124 @@ static int nldev_res_get_cq_dumpit(struct sk_buff *skb, return ret; } +static int nldev_res_get_mr_dumpit(struct sk_buff *skb, + struct netlink_callback *cb) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct rdma_restrack_entry *res; + int err, ret = 0, idx = 0; + struct nlattr *table_attr; + struct ib_device *device; + int start = cb->args[0]; + struct ib_mr *mr = NULL; + struct nlmsghdr *nlh; + u32 index; + + err = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, NULL); + /* + * Right now, we are expecting the device index to get MR information, + * but it is possible to extend this code to return all devices in + * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX. + * if it doesn't exist, we will iterate over all devices. + * + * But it is not needed for now. + */ + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(index); + if (!device) + return -EINVAL; + + nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_MR_GET), + 0, NLM_F_MULTI); + + if (fill_nldev_handle(skb, device)) { + ret = -EMSGSIZE; + goto err; + } + + table_attr = nla_nest_start(skb, RDMA_NLDEV_ATTR_RES_MR); + if (!table_attr) { + ret = -EMSGSIZE; + goto err; + } + + down_read(&device->res.rwsem); + hash_for_each_possible(device->res.hash, res, node, RDMA_RESTRACK_MR) { + if (idx < start) + goto next; + + if ((rdma_is_kernel_res(res) && + task_active_pid_ns(current) != &init_pid_ns) || + (!rdma_is_kernel_res(res) && + task_active_pid_ns(current) != + task_active_pid_ns(res->task))) + /* + * 1. Kernel MRs should be visible in init namspace only + * 2. Present only MRs visible in the current namespace + */ + goto next; + + if (!rdma_restrack_get(res)) + /* + * Resource is under release now, but we are not + * relesing lock now, so it will be released in + * our next pass, once we will get ->next pointer. + */ + goto next; + + mr = container_of(res, struct ib_mr, res); + + up_read(&device->res.rwsem); + ret = fill_res_mr_entry(skb, mr); + down_read(&device->res.rwsem); + /* + * Return resource back, but it won't be released till + * the &device->res.rwsem will be released for write. + */ + rdma_restrack_put(res); + + if (ret == -EMSGSIZE) + /* + * There is a chance to optimize here. + * It can be done by using list_prepare_entry + * and list_for_each_entry_continue afterwards. + */ + break; + if (ret) + goto res_err; +next: idx++; + } + up_read(&device->res.rwsem); + + nla_nest_end(skb, table_attr); + nlmsg_end(skb, nlh); + cb->args[0] = idx; + + /* + * No more MRs to fill, cancel the message and + * return 0 to mark end of dumpit. + */ + if (!mr) + goto err; + + put_device(&device->dev); + return skb->len; + +res_err: + nla_nest_cancel(skb, table_attr); + up_read(&device->res.rwsem); + +err: + nlmsg_cancel(skb, nlh); + put_device(&device->dev); + return ret; +} + static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = { [RDMA_NLDEV_CMD_GET] = { .doit = nldev_get_doit, @@ -1130,6 +1301,9 @@ static int nldev_res_get_cq_dumpit(struct sk_buff *skb, [RDMA_NLDEV_CMD_RES_CQ_GET] = { .dump = nldev_res_get_cq_dumpit, }, + [RDMA_NLDEV_CMD_RES_MR_GET] = { + .dump = nldev_res_get_mr_dumpit, + }, }; void __init nldev_init(void) diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c index 6385914..d3ad0ab 100644 --- a/drivers/infiniband/core/restrack.c +++ b/drivers/infiniband/core/restrack.c @@ -43,22 +43,36 @@ int rdma_restrack_count(struct rdma_restrack_root *res, static void set_kern_name(struct rdma_restrack_entry *res) { - enum rdma_restrack_type type = res->type; - struct ib_qp *qp; + struct ib_pd *pd = NULL; - if (type != RDMA_RESTRACK_QP) - /* Other types already have this name embedded in */ - return; + switch (res->type) { + case RDMA_RESTRACK_QP: { + struct ib_qp *qp; - qp = container_of(res, struct ib_qp, res); - if (!qp->pd) { - WARN_ONCE(true, "XRC QPs are not supported\n"); - /* Survive, despite the programmer's error */ - res->kern_name = " "; - return; + qp = container_of(res, struct ib_qp, res); + if (qp->pd) { + pd = qp->pd; + } else { + WARN_ONCE(true, "XRC QPs are not supported\n"); + /* Survive, despite the programmer's error */ + res->kern_name = " "; + } + break; + } + case RDMA_RESTRACK_MR: { + struct ib_mr *mr; + + mr = container_of(res, struct ib_mr, res); + pd = mr->pd; + break; + } + default: + /* Other types set kern_name directly */ + break; } - res->kern_name = qp->pd->res.kern_name; + if (pd) + res->kern_name = pd->res.kern_name; } static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) @@ -70,6 +84,7 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) struct ib_pd *pd; struct ib_cq *cq; struct ib_qp *qp; + struct ib_mr *mr; switch (type) { case RDMA_RESTRACK_PD: @@ -92,6 +107,10 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) cm_id = container_of(res, struct rdma_cm_id, res); dev = cm_id->device; break; + case RDMA_RESTRACK_MR: + mr = container_of(res, struct ib_mr, res); + dev = mr->device; + break; default: WARN_ONCE(true, "Wrong resource tracking type %u\n", type); return NULL; @@ -108,6 +127,7 @@ static bool res_is_user(struct rdma_restrack_entry *res) struct ib_pd *pd; struct ib_cq *cq; struct ib_qp *qp; + struct ib_mr *mr; bool is_user = false; switch (type) { @@ -132,6 +152,10 @@ static bool res_is_user(struct rdma_restrack_entry *res) cm_id = container_of(res, struct rdma_cm_id, res); is_user = !cm_id->caller; break; + case RDMA_RESTRACK_MR: + mr = container_of(res, struct ib_mr, res); + is_user = mr->pd->uobject; + break; default: WARN_ONCE(true, "Wrong resource tracking type %u\n", type); } diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c index 256934d..3f026c4 100644 --- a/drivers/infiniband/core/uverbs_cmd.c +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -694,6 +694,8 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file, mr->pd = pd; mr->uobject = uobj; atomic_inc(&pd->usecnt); + mr->res.type = RDMA_RESTRACK_MR; + rdma_restrack_add(&mr->res); uobj->object = mr; @@ -819,6 +821,7 @@ ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file, struct ib_uverbs_dereg_mr cmd; struct ib_uobject *uobj; int ret = -EINVAL; + struct ib_mr *mr; if (copy_from_user(&cmd, buf, sizeof cmd)) return -EFAULT; @@ -828,6 +831,9 @@ ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file, if (IS_ERR(uobj)) return PTR_ERR(uobj); + mr = uobj->object; + rdma_restrack_del(&mr->res); + ret = uobj_remove_commit(uobj); return ret ?: in_len; diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c index 16ebc63..c3265f7 100644 --- a/drivers/infiniband/core/verbs.c +++ b/drivers/infiniband/core/verbs.c @@ -1623,6 +1623,7 @@ int ib_dereg_mr(struct ib_mr *mr) struct ib_pd *pd = mr->pd; int ret; + rdma_restrack_del(&mr->res); ret = mr->device->dereg_mr(mr); if (!ret) atomic_dec(&pd->usecnt); @@ -1659,6 +1660,8 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd, mr->uobject = NULL; atomic_inc(&pd->usecnt); mr->need_inval = false; + mr->res.type = RDMA_RESTRACK_MR; + rdma_restrack_add(&mr->res); } return mr; diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h index 5263c86..33d8c5d 100644 --- a/include/rdma/ib_verbs.h +++ b/include/rdma/ib_verbs.h @@ -1771,6 +1771,11 @@ struct ib_mr { struct ib_uobject *uobject; /* user */ struct list_head qp_entry; /* FR */ }; + + /* + * Implementation details of the RDMA core, don't use in drivers: + */ + struct rdma_restrack_entry res; }; struct ib_mw { diff --git a/include/rdma/restrack.h b/include/rdma/restrack.h index a794e0e..bfd1140 100644 --- a/include/rdma/restrack.h +++ b/include/rdma/restrack.h @@ -37,6 +37,10 @@ enum rdma_restrack_type { */ RDMA_RESTRACK_CM_ID, /** + * @RDMA_RESTRACK_MR: Memory Region (MR) + */ + RDMA_RESTRACK_MR, + /** * @RDMA_RESTRACK_MAX: Last entry, used for array dclarations */ RDMA_RESTRACK_MAX diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h index fa677ef..6adaeaa 100644 --- a/include/uapi/rdma/rdma_netlink.h +++ b/include/uapi/rdma/rdma_netlink.h @@ -244,6 +244,8 @@ enum rdma_nldev_command { RDMA_NLDEV_CMD_RES_CQ_GET, /* can dump */ + RDMA_NLDEV_CMD_RES_MR_GET, /* can dump */ + RDMA_NLDEV_NUM_OPS }; @@ -390,6 +392,14 @@ enum rdma_nldev_attr { RDMA_NLDEV_ATTR_RES_USECNT, /* u64 */ RDMA_NLDEV_ATTR_RES_POLL_CTX, /* u8 */ + RDMA_NLDEV_ATTR_RES_MR, /* nested table */ + RDMA_NLDEV_ATTR_RES_MR_ENTRY, /* nested table */ + RDMA_NLDEV_ATTR_RES_RKEY, /* u32 */ + RDMA_NLDEV_ATTR_RES_LKEY, /* u32 */ + RDMA_NLDEV_ATTR_RES_IOVA, /* u64 */ + RDMA_NLDEV_ATTR_RES_MRLEN, /* u64 */ + RDMA_NLDEV_ATTR_RES_PGSIZE, /* u32 */ + RDMA_NLDEV_ATTR_MAX }; #endif /* _UAPI_RDMA_NETLINK_H */