From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga12.intel.com (mga12.intel.com [192.55.52.136]) by dpdk.org (Postfix) with ESMTP id 0AF002C52 for ; Thu, 7 Mar 2019 13:02:20 +0100 (CET) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga001.jf.intel.com ([10.7.209.18]) by fmsmga106.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 07 Mar 2019 04:02:20 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.58,451,1544515200"; d="scan'208";a="169544713" Received: from tjozwiax-mobl1.ger.corp.intel.com (HELO localhost.localdomain) ([10.103.104.44]) by orsmga001.jf.intel.com with ESMTP; 07 Mar 2019 04:02:19 -0800 From: Tomasz Jozwiak To: dev@dpdk.org, fiona.trahe@intel.com, tomaszx.jozwiak@intel.com Date: Thu, 7 Mar 2019 13:02:16 +0100 Message-Id: <1551960136-30358-2-git-send-email-tomaszx.jozwiak@intel.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1551960136-30358-1-git-send-email-tomaszx.jozwiak@intel.com> References: <1551439054-17658-1-git-send-email-tomaszx.jozwiak@intel.com> <1551960136-30358-1-git-send-email-tomaszx.jozwiak@intel.com> Subject: [dpdk-dev] [PATCH v4 1/1] compress/qat: add dynamic sgl allocation X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 07 Mar 2019 12:02:21 -0000 This patch adds dynamic SGL allocation instead of static one. The number of element in SGL can be adjusted in each operation depend of the request. Signed-off-by: Tomasz Jozwiak --- config/common_base | 1 - doc/guides/compressdevs/qat_comp.rst | 1 - doc/guides/cryptodevs/qat.rst | 5 ---- drivers/compress/qat/qat_comp.c | 58 ++++++++++++++++++++++++++++++++---- drivers/compress/qat/qat_comp.h | 13 ++++---- drivers/compress/qat/qat_comp_pmd.c | 49 +++++++++++++++++++++++++----- 6 files changed, 101 insertions(+), 26 deletions(-) diff --git a/config/common_base b/config/common_base index 0b09a93..91c7b73 100644 --- a/config/common_base +++ b/config/common_base @@ -549,7 +549,6 @@ CONFIG_RTE_LIBRTE_PMD_QAT_SYM=n # Max. number of QuickAssist devices, which can be detected and attached # CONFIG_RTE_PMD_QAT_MAX_PCI_DEVICES=48 -CONFIG_RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS=16 CONFIG_RTE_PMD_QAT_COMP_IM_BUFFER_SIZE=65536 # diff --git a/doc/guides/compressdevs/qat_comp.rst b/doc/guides/compressdevs/qat_comp.rst index 5631cb1..6f583a4 100644 --- a/doc/guides/compressdevs/qat_comp.rst +++ b/doc/guides/compressdevs/qat_comp.rst @@ -35,7 +35,6 @@ Limitations * Compressdev level 0, no compression, is not supported. * Queue pairs are not thread-safe (that is, within a single queue pair, RX and TX from different lcores is not supported). * No BSD support as BSD QAT kernel driver not available. -* Number of segments in mbuf chains in the op must be <= RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS from the config file. * When using Deflate dynamic huffman encoding for compression, the input size (op.src.length) must be < CONFIG_RTE_PMD_QAT_COMP_IM_BUFFER_SIZE from the config file, see :ref:`building_qat_config` for more details. diff --git a/doc/guides/cryptodevs/qat.rst b/doc/guides/cryptodevs/qat.rst index b7eace1..03bd0c1 100644 --- a/doc/guides/cryptodevs/qat.rst +++ b/doc/guides/cryptodevs/qat.rst @@ -156,7 +156,6 @@ These are the build configuration options affecting QAT, and their default value CONFIG_RTE_LIBRTE_PMD_QAT=y CONFIG_RTE_LIBRTE_PMD_QAT_SYM=n CONFIG_RTE_PMD_QAT_MAX_PCI_DEVICES=48 - CONFIG_RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS=16 CONFIG_RTE_PMD_QAT_COMP_IM_BUFFER_SIZE=65536 CONFIG_RTE_LIBRTE_PMD_QAT must be enabled for any QAT PMD to be built. @@ -174,10 +173,6 @@ Note, there are separate config items for max cryptodevs CONFIG_RTE_CRYPTO_MAX_D and max compressdevs CONFIG_RTE_COMPRESS_MAX_DEVS, if necessary these should be adjusted to handle the total of QAT and other devices which the process will use. -QAT allocates internal structures to handle SGLs. For the compression service -CONFIG_RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS can be changed if more segments are needed. -An extra (max_inflight_ops x 16) bytes per queue_pair will be used for every increment. - QAT compression PMD needs intermediate buffers to support Deflate compression with Dynamic Huffman encoding. CONFIG_RTE_PMD_QAT_COMP_IM_BUFFER_SIZE specifies the size of a single buffer, the PMD will allocate a multiple of these, diff --git a/drivers/compress/qat/qat_comp.c b/drivers/compress/qat/qat_comp.c index 32ca753..c021f4a 100644 --- a/drivers/compress/qat/qat_comp.c +++ b/drivers/compress/qat/qat_comp.c @@ -1,5 +1,5 @@ /* SPDX-License-Identifier: BSD-3-Clause - * Copyright(c) 2018 Intel Corporation + * Copyright(c) 2018-2019 Intel Corporation */ #include @@ -55,22 +55,70 @@ qat_comp_build_request(void *in_op, uint8_t *out_msg, ICP_QAT_FW_COMN_PTR_TYPE_SET(comp_req->comn_hdr.comn_req_flags, QAT_COMN_PTR_TYPE_SGL); + if (unlikely(op->m_src->nb_segs > cookie->src_nb_elems)) { + /* we need to allocate more elements in SGL*/ + void *tmp; + + tmp = rte_realloc_socket(cookie->qat_sgl_src_d, + sizeof(struct qat_sgl) + + sizeof(struct qat_flat_buf) * + op->m_src->nb_segs, 64, + rte_socket_id()); + + if (unlikely(tmp == NULL)) { + QAT_DP_LOG(ERR, "QAT PMD can't allocate memory" + " for %d elements of SGL", + op->m_src->nb_segs); + op->status = RTE_COMP_OP_STATUS_INVALID_ARGS; + return -ENOMEM; + } + /* new SGL is valid now */ + cookie->qat_sgl_src_d = (struct qat_sgl *)tmp; + cookie->src_nb_elems = op->m_src->nb_segs; + cookie->qat_sgl_src_phys_addr = + rte_malloc_virt2iova(cookie->qat_sgl_src_d); + } + ret = qat_sgl_fill_array(op->m_src, op->src.offset, - &cookie->qat_sgl_src, + cookie->qat_sgl_src_d, op->src.length, - RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS); + (const uint16_t)cookie->src_nb_elems); if (ret) { QAT_DP_LOG(ERR, "QAT PMD Cannot fill source sgl array"); op->status = RTE_COMP_OP_STATUS_INVALID_ARGS; return ret; } + if (unlikely(op->m_dst->nb_segs > cookie->dst_nb_elems)) { + /* we need to allocate more elements in SGL*/ + struct qat_sgl *tmp; + + tmp = rte_realloc_socket(cookie->qat_sgl_dst_d, + sizeof(struct qat_sgl) + + sizeof(struct qat_flat_buf) * + op->m_dst->nb_segs, 64, + rte_socket_id()); + + if (unlikely(tmp == NULL)) { + QAT_DP_LOG(ERR, "QAT PMD can't allocate memory" + " for %d elements of SGL", + op->m_dst->nb_segs); + op->status = RTE_COMP_OP_STATUS_INVALID_ARGS; + return -EINVAL; + } + /* new SGL is valid now */ + cookie->qat_sgl_dst_d = (struct qat_sgl *)tmp; + cookie->dst_nb_elems = op->m_dst->nb_segs; + cookie->qat_sgl_dst_phys_addr = + rte_malloc_virt2iova(cookie->qat_sgl_dst_d); + } + ret = qat_sgl_fill_array(op->m_dst, op->dst.offset, - &cookie->qat_sgl_dst, + cookie->qat_sgl_dst_d, comp_req->comp_pars.out_buffer_sz, - RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS); + (const uint16_t)cookie->dst_nb_elems); if (ret) { QAT_DP_LOG(ERR, "QAT PMD Cannot fill dest. sgl array"); op->status = RTE_COMP_OP_STATUS_INVALID_ARGS; diff --git a/drivers/compress/qat/qat_comp.h b/drivers/compress/qat/qat_comp.h index 19f48df..2465f12 100644 --- a/drivers/compress/qat/qat_comp.h +++ b/drivers/compress/qat/qat_comp.h @@ -1,5 +1,5 @@ /* SPDX-License-Identifier: BSD-3-Clause - * Copyright(c) 2015-2018 Intel Corporation + * Copyright(c) 2015-2019 Intel Corporation */ #ifndef _QAT_COMP_H_ @@ -37,16 +37,15 @@ struct qat_inter_sgl { struct qat_flat_buf buffers[QAT_NUM_BUFS_IN_IM_SGL]; } __rte_packed __rte_cache_aligned; -struct qat_comp_sgl { - qat_sgl_hdr; - struct qat_flat_buf buffers[RTE_PMD_QAT_COMP_SGL_MAX_SEGMENTS]; -} __rte_packed __rte_cache_aligned; struct qat_comp_op_cookie { - struct qat_comp_sgl qat_sgl_src; - struct qat_comp_sgl qat_sgl_dst; phys_addr_t qat_sgl_src_phys_addr; phys_addr_t qat_sgl_dst_phys_addr; + /* dynamically created SGLs */ + uint16_t src_nb_elems; + uint16_t dst_nb_elems; + struct qat_sgl *qat_sgl_src_d; + struct qat_sgl *qat_sgl_dst_d; }; struct qat_comp_xform { diff --git a/drivers/compress/qat/qat_comp_pmd.c b/drivers/compress/qat/qat_comp_pmd.c index 27c8856..f034a19 100644 --- a/drivers/compress/qat/qat_comp_pmd.c +++ b/drivers/compress/qat/qat_comp_pmd.c @@ -1,10 +1,14 @@ /* SPDX-License-Identifier: BSD-3-Clause - * Copyright(c) 2015-2018 Intel Corporation + * Copyright(c) 2015-2019 Intel Corporation */ +#include + #include "qat_comp.h" #include "qat_comp_pmd.h" +#define QAT_PMD_COMP_SGL_DEF_SEGMENTS 16 + static const struct rte_compressdev_capabilities qat_comp_gen_capabilities[] = { {/* COMPRESSION - deflate */ .algo = RTE_COMP_ALGO_DEFLATE, @@ -60,6 +64,10 @@ static int qat_comp_qp_release(struct rte_compressdev *dev, uint16_t queue_pair_id) { struct qat_comp_dev_private *qat_private = dev->data->dev_private; + struct qat_qp **qp_addr = + (struct qat_qp **)&(dev->data->queue_pairs[queue_pair_id]); + struct qat_qp *qp = (struct qat_qp *)*qp_addr; + uint32_t i; QAT_LOG(DEBUG, "Release comp qp %u on device %d", queue_pair_id, dev->data->dev_id); @@ -67,6 +75,14 @@ qat_comp_qp_release(struct rte_compressdev *dev, uint16_t queue_pair_id) qat_private->qat_dev->qps_in_use[QAT_SERVICE_COMPRESSION][queue_pair_id] = NULL; + for (i = 0; i < qp->nb_descriptors; i++) { + + struct qat_comp_op_cookie *cookie = qp->op_cookies[i]; + + rte_free(cookie->qat_sgl_src_d); + rte_free(cookie->qat_sgl_dst_d); + } + return qat_qp_release((struct qat_qp **) &(dev->data->queue_pairs[queue_pair_id])); } @@ -122,15 +138,34 @@ qat_comp_qp_setup(struct rte_compressdev *dev, uint16_t qp_id, struct qat_comp_op_cookie *cookie = qp->op_cookies[i]; + cookie->qat_sgl_src_d = rte_zmalloc_socket(NULL, + sizeof(struct qat_sgl) + + sizeof(struct qat_flat_buf) * + QAT_PMD_COMP_SGL_DEF_SEGMENTS, + 64, socket_id); + + cookie->qat_sgl_dst_d = rte_zmalloc_socket(NULL, + sizeof(struct qat_sgl) + + sizeof(struct qat_flat_buf) * + QAT_PMD_COMP_SGL_DEF_SEGMENTS, + 64, socket_id); + + if (cookie->qat_sgl_src_d == NULL || + cookie->qat_sgl_dst_d == NULL) { + QAT_LOG(ERR, "Can't allocate SGL" + " for device %s", + qat_private->qat_dev->name); + return -ENOMEM; + } + cookie->qat_sgl_src_phys_addr = - rte_mempool_virt2iova(cookie) + - offsetof(struct qat_comp_op_cookie, - qat_sgl_src); + rte_malloc_virt2iova(cookie->qat_sgl_src_d); cookie->qat_sgl_dst_phys_addr = - rte_mempool_virt2iova(cookie) + - offsetof(struct qat_comp_op_cookie, - qat_sgl_dst); + rte_malloc_virt2iova(cookie->qat_sgl_dst_d); + + cookie->dst_nb_elems = cookie->src_nb_elems = + QAT_PMD_COMP_SGL_DEF_SEGMENTS; } return ret; -- 2.7.4