From: Xinghai Cen <cenxinghai@h-partners.com> driver inclusion category: bugfix bugzilla: https://gitee.com/src-openeuler/rdma-core/issues/IB66RT ------------------------------------------------------------------ Changes to be committed: modified: 0040-libhns-Fix-memory-leakage-when-DCA-is-enabled.patch new file: 0045-libhns-fix-incorrectly-using-fixed-pagesize.patch new file: 0046-libhns-fix-missing-new-IO-support-for-DCA.patch Signed-off-by: Xinghai Cen <cenxinghai@h-partners.com> --- ...x-memory-leakage-when-DCA-is-enabled.patch | 12 +- ...fix-incorrectly-using-fixed-pagesize.patch | 110 ++++++++++++++++++ ...s-fix-missing-new-IO-support-for-DCA.patch | 55 +++++++++ rdma-core.spec | 10 +- 4 files changed, 181 insertions(+), 6 deletions(-) create mode 100644 0045-libhns-fix-incorrectly-using-fixed-pagesize.patch create mode 100644 0046-libhns-fix-missing-new-IO-support-for-DCA.patch diff --git a/0040-libhns-Fix-memory-leakage-when-DCA-is-enabled.patch b/0040-libhns-Fix-memory-leakage-when-DCA-is-enabled.patch index f046264..58cd8ec 100644 --- a/0040-libhns-Fix-memory-leakage-when-DCA-is-enabled.patch +++ b/0040-libhns-Fix-memory-leakage-when-DCA-is-enabled.patch @@ -1,4 +1,4 @@ -From af20dd32df73ff72d35a430a5fb87ac42d70cdf4 Mon Sep 17 00:00:00 2001 +From f8e29f955dd5399bd227c4de532f6d09872a254a Mon Sep 17 00:00:00 2001 From: wenglianfa <wenglianfa@huawei.com> Date: Thu, 25 Jul 2024 11:06:01 +0800 Subject: [PATCH] libhns: Fix memory leakage when DCA is enabled @@ -17,18 +17,20 @@ Fixes: 2783884a97e7 ("libhns: Add support for attaching QP's WQE buffer") Signed-off-by: wenglianfa <wenglianfa@huawei.com> Signed-off-by: Xinghai Cen <cenxinghai@h-partners.com> --- - providers/hns/hns_roce_u_verbs.c | 3 ++- - 1 file changed, 2 insertions(+), 1 deletion(-) + providers/hns/hns_roce_u_verbs.c | 5 ++++- + 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/providers/hns/hns_roce_u_verbs.c b/providers/hns/hns_roce_u_verbs.c -index e30880c..c733b21 100644 +index e30880c..154e800 100644 --- a/providers/hns/hns_roce_u_verbs.c +++ b/providers/hns/hns_roce_u_verbs.c -@@ -1357,7 +1357,8 @@ static void qp_free_wqe(struct hns_roce_qp *qp) +@@ -1357,7 +1357,10 @@ static void qp_free_wqe(struct hns_roce_qp *qp) if (qp->rq.wqe_cnt) free(qp->rq.wrid); - hns_roce_free_buf(&qp->buf); ++ if (qp->dca_wqe.bufs) ++ free(qp->dca_wqe.bufs); + else + hns_roce_free_buf(&qp->buf); } diff --git a/0045-libhns-fix-incorrectly-using-fixed-pagesize.patch b/0045-libhns-fix-incorrectly-using-fixed-pagesize.patch new file mode 100644 index 0000000..717b340 --- /dev/null +++ b/0045-libhns-fix-incorrectly-using-fixed-pagesize.patch @@ -0,0 +1,110 @@ +From 7bd22fed52a1828b0d44a990b52266e9e1d92b5d Mon Sep 17 00:00:00 2001 +From: Chengchang Tang <tangchengchang@huawei.com> +Date: Tue, 30 Jan 2024 21:00:46 +0800 +Subject: [PATCH 45/46] libhns: fix incorrectly using fixed pagesize + +driver inclusion +category: bugfix +bugzilla: https://gitee.com/src-openeuler/rdma-core/issues/IB66RT + +------------------------------------------------------------------ + +Currently, actually used page size is fixed, causing the flexible wqe +buffer size feature to not take effect. + +Fixes: 9ab7600d832b ("libhns: Add support for attaching QP's WQE buffer") +Signed-off-by: Chengchang Tang <tangchengchang@huawei.com> +Signed-off-by: Xinghai Cen <cenxinghai@h-partners.com> +--- + providers/hns/hns_roce_u_verbs.c | 24 +++++++++++++----------- + 1 file changed, 13 insertions(+), 11 deletions(-) + +diff --git a/providers/hns/hns_roce_u_verbs.c b/providers/hns/hns_roce_u_verbs.c +index bce215e..848f836 100644 +--- a/providers/hns/hns_roce_u_verbs.c ++++ b/providers/hns/hns_roce_u_verbs.c +@@ -1296,14 +1296,14 @@ static void free_recv_rinl_buf(struct hns_roce_rinl_buf *rinl_buf) + + static void get_best_multi_region_pg_shift(struct hns_roce_device *hr_dev, + struct hns_roce_context *ctx, +- struct hns_roce_qp *qp) ++ struct hns_roce_qp *qp, bool dca_en) + { + uint32_t ext_sge_size; + uint32_t sq_size; + uint32_t rq_size; + uint8_t pg_shift; + +- if (!(ctx->config & HNS_ROCE_UCTX_RSP_DYN_QP_PGSZ)) { ++ if (!(ctx->config & HNS_ROCE_UCTX_RSP_DYN_QP_PGSZ || dca_en)) { + qp->pageshift = HNS_HW_PAGE_SHIFT; + return; + } +@@ -1334,7 +1334,7 @@ static void get_best_multi_region_pg_shift(struct hns_roce_device *hr_dev, + + static int calc_qp_buff_size(struct hns_roce_device *hr_dev, + struct hns_roce_context *ctx, +- struct hns_roce_qp *qp) ++ struct hns_roce_qp *qp, bool dca_en) + { + struct hns_roce_wq *sq = &qp->sq; + struct hns_roce_wq *rq = &qp->rq; +@@ -1342,7 +1342,7 @@ static int calc_qp_buff_size(struct hns_roce_device *hr_dev, + unsigned int size; + + qp->buf_size = 0; +- get_best_multi_region_pg_shift(hr_dev, ctx, qp); ++ get_best_multi_region_pg_shift(hr_dev, ctx, qp, dca_en); + page_size = 1 << qp->pageshift; + + /* SQ WQE */ +@@ -1384,7 +1384,7 @@ static inline bool check_qp_support_dca(struct hns_roce_dca_ctx *dca_ctx, + if (hns_attr && + (hns_attr->comp_mask & HNSDV_QP_INIT_ATTR_MASK_QP_CREATE_FLAGS) && + (hns_attr->create_flags & HNSDV_QP_CREATE_ENABLE_DCA_MODE)) +- return true; ++ return dca_ctx->max_size > 0; + + return false; + } +@@ -1408,9 +1408,12 @@ static int qp_alloc_wqe(struct ibv_qp_init_attr_ex *attr, + struct hns_roce_qp *qp, struct hns_roce_context *ctx) + { + struct hns_roce_device *hr_dev = to_hr_dev(ctx->ibv_ctx.context.device); ++ bool dca_en = check_qp_support_dca(&ctx->dca_ctx, attr, hns_attr); ++ int ret; + +- if (calc_qp_buff_size(hr_dev, ctx, qp)) +- return -EINVAL; ++ ret = calc_qp_buff_size(hr_dev, ctx, qp, dca_en); ++ if (ret) ++ return ret; + + qp->sq.wrid = malloc(qp->sq.wqe_cnt * sizeof(uint64_t)); + if (!qp->sq.wrid) +@@ -1428,19 +1431,18 @@ static int qp_alloc_wqe(struct ibv_qp_init_attr_ex *attr, + goto err_alloc; + } + +- if (check_qp_support_dca(&ctx->dca_ctx, attr, hns_attr) && +- ctx->dca_ctx.max_size > 0) { ++ if (check_qp_support_dca(&ctx->dca_ctx, attr, hns_attr)) { + /* when DCA is enabled, use a buffer list to store page addr */ + qp->buf.buf = NULL; + qp->dca_wqe.max_cnt = hr_hw_page_count(qp->buf_size); +- qp->dca_wqe.shift = HNS_HW_PAGE_SHIFT; ++ qp->dca_wqe.shift = qp->pageshift; + qp->dca_wqe.bufs = calloc(qp->dca_wqe.max_cnt, sizeof(void *)); + if (!qp->dca_wqe.bufs) + goto err_alloc; + verbs_debug(&ctx->ibv_ctx, "alloc DCA buf.\n"); + } else { + if (hns_roce_alloc_buf(&qp->buf, qp->buf_size, +- HNS_HW_PAGE_SIZE)) ++ 1 << qp->pageshift)) + goto err_alloc; + } + +-- +2.33.0 + diff --git a/0046-libhns-fix-missing-new-IO-support-for-DCA.patch b/0046-libhns-fix-missing-new-IO-support-for-DCA.patch new file mode 100644 index 0000000..1b98200 --- /dev/null +++ b/0046-libhns-fix-missing-new-IO-support-for-DCA.patch @@ -0,0 +1,55 @@ +From 199b2f78ff9eeeb25acc78f9da495ae58877807a Mon Sep 17 00:00:00 2001 +From: Chengchang Tang <tangchengchang@huawei.com> +Date: Tue, 30 Jan 2024 21:28:44 +0800 +Subject: [PATCH 46/46] libhns: fix missing new IO support for DCA + +driver inclusion +category: bugfix +bugzilla: https://gitee.com/src-openeuler/rdma-core/issues/IBSL67 + +------------------------------------------------------------------ + +New IO related support has been missed for DCA. + +Fixes: 9ab7600d832b ("libhns: Add support for attaching QP's WQE buffer") +Signed-off-by: Chengchang Tang <tangchengchang@huawei.com> +Signed-off-by: Xinghai Cen <cenxinghai@h-partners.com> +--- + providers/hns/hns_roce_u_hw_v2.c | 7 +++++++ + 1 file changed, 7 insertions(+) + +diff --git a/providers/hns/hns_roce_u_hw_v2.c b/providers/hns/hns_roce_u_hw_v2.c +index aadea7a..3137111 100644 +--- a/providers/hns/hns_roce_u_hw_v2.c ++++ b/providers/hns/hns_roce_u_hw_v2.c +@@ -2191,6 +2191,8 @@ static int wc_start_poll_cq(struct ibv_cq_ex *current, + } + + err = hns_roce_poll_one(ctx, &qp, cq, NULL); ++ if (qp && check_dca_detach_enable(qp)) ++ dca_detach_qp_buf(ctx, qp); + + start_poll_done: + if (err != V2_CQ_OK) +@@ -2210,6 +2212,8 @@ static int wc_next_poll_cq(struct ibv_cq_ex *current) + return hns_roce_poll_one_swc(cq, NULL); + + err = hns_roce_poll_one(ctx, &qp, cq, NULL); ++ if (qp && check_dca_detach_enable(qp)) ++ dca_detach_qp_buf(ctx, qp); + if (err != V2_CQ_OK) + return err; + +@@ -2408,6 +2412,9 @@ init_rc_wqe(struct hns_roce_qp *qp, uint64_t wr_id, unsigned int opcode) + hr_reg_write_bool(wqe, RCWQE_SE, send_flags & IBV_SEND_SOLICITED); + hr_reg_clear(wqe, RCWQE_INLINE); + ++ if (check_qp_dca_enable(qp)) ++ fill_rc_dca_fields(qp->verbs_qp.qp.qp_num, wqe); ++ + qp->sq.wrid[wqe_idx] = wr_id; + qp->cur_wqe = wqe; + +-- +2.33.0 + diff --git a/rdma-core.spec b/rdma-core.spec index b202c8d..0edb02b 100644 --- a/rdma-core.spec +++ b/rdma-core.spec @@ -1,6 +1,6 @@ Name: rdma-core Version: 50.0 -Release: 22 +Release: 23 Summary: RDMA core userspace libraries and daemons License: GPL-2.0-only OR BSD-2-Clause AND BSD-3-Clause Url: https://github.com/linux-rdma/rdma-core @@ -50,6 +50,8 @@ patch41: 0041-libhns-Fix-coredump-during-QP-destruction-when-send_.patch patch42: 0042-libhns-Add-error-logs-to-help-diagnosis.patch patch43: 0043-libhns-Fix-missing-fields-for-SRQ-WC.patch patch44: 0044-libxscale-Add-Yunsilicon-User-Space-RDMA-Driver.patch +patch45: 0045-libhns-fix-incorrectly-using-fixed-pagesize.patch +patch46: 0046-libhns-fix-missing-new-IO-support-for-DCA.patch BuildRequires: binutils cmake >= 2.8.11 gcc libudev-devel pkgconfig pkgconfig(libnl-3.0) BuildRequires: pkgconfig(libnl-route-3.0) systemd systemd-devel @@ -629,6 +631,12 @@ fi %doc %{_docdir}/%{name}-%{version}/70-persistent-ipoib.rules %changelog +* Tue Mar 11 2025 Xinghai Cen <cenxinghai@h-partners.com> - 50.0-23 +- Type: bugfix +- ID: NA +- SUG: NA +- DESC: Fix some bugs for libhns + * Wed Feb 26 2025 Xin Tian <tianx@yunsilicon.com> - 50.0-22 - Type: requirement - ID: NA -- 2.33.0