From patchwork Fri May 6 14:38:53 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Cabiddu, Giovanni" X-Patchwork-Id: 570947 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id C4376C433EF for ; Fri, 6 May 2022 14:39:14 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1442284AbiEFOmy (ORCPT ); Fri, 6 May 2022 10:42:54 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:35308 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1442288AbiEFOmx (ORCPT ); Fri, 6 May 2022 10:42:53 -0400 Received: from mga12.intel.com (mga12.intel.com [192.55.52.136]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 5B59C6A427; Fri, 6 May 2022 07:39:09 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1651847949; x=1683383949; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=PHirnwRewAXRkYbi7irfJOicze95UPl0UU4GfkNeB54=; b=YDWs6jIi7eaJ0QYAgR8CVH0DHRcqvM7avwYSnC7sXPW9dIT5m/31Y8Z4 M6H/MwhiEGfI+ynlnSrQDj9XQlpbzkeImLRhJrV3PoeEIVk3hN4P328yV wqsf9skPp7X4pBIdMVQZeAqBvY5evZJljugyuIsR+wl87Jlya7wxNGol+ voyUDyfRtN6cxLJnJdqaiUkJI6/PqttN1AtLwzo8TvN5dLbsmxbU4dSgw DvkeSV2g9uASGfRDlf88E2Q2hiWMmHW7mYkUDGhYPZl1C2SokGLdVOiz+ uFO0JLIBNNH0cs28kMKxJUmWjoKvPO+cwysHcLppaC2tT3lhMe7iE19vc w==; X-IronPort-AV: E=McAfee;i="6400,9594,10339"; a="248387454" X-IronPort-AV: E=Sophos;i="5.91,203,1647327600"; d="scan'208";a="248387454" Received: from orsmga006.jf.intel.com ([10.7.209.51]) by fmsmga106.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 06 May 2022 07:39:09 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.91,203,1647327600"; d="scan'208";a="537914799" Received: from silpixa00400314.ir.intel.com (HELO silpixa00400314.ger.corp.intel.com) ([10.237.222.76]) by orsmga006.jf.intel.com with ESMTP; 06 May 2022 07:39:06 -0700 From: Giovanni Cabiddu To: herbert@gondor.apana.org.au Cc: linux-crypto@vger.kernel.org, qat-linux@intel.com, vdronov@redhat.com, Giovanni Cabiddu , stable@vger.kernel.org, Mikulas Patocka , Marco Chiappero , Wojciech Ziemba Subject: [PATCH v2 01/11] crypto: qat - use pre-allocated buffers in datapath Date: Fri, 6 May 2022 15:38:53 +0100 Message-Id: <20220506143903.31776-2-giovanni.cabiddu@intel.com> X-Mailer: git-send-email 2.35.1 In-Reply-To: <20220506143903.31776-1-giovanni.cabiddu@intel.com> References: <20220506143903.31776-1-giovanni.cabiddu@intel.com> MIME-Version: 1.0 Organization: Intel Research and Development Ireland Ltd - Co. Reg. #308263 - Collinstown Industrial Park, Leixlip, County Kildare - Ireland Precedence: bulk List-ID: X-Mailing-List: stable@vger.kernel.org In order to do DMAs, the QAT device requires that the scatterlist structures are mapped and translated into a format that the firmware can understand. This is defined as the composition of a scatter gather list (SGL) descriptor header, the struct qat_alg_buf_list, plus a variable number of flat buffer descriptors, the struct qat_alg_buf. The allocation and mapping of these data structures is done each time a request is received from the skcipher and aead APIs. In an OOM situation, this behaviour might lead to a dead-lock if an allocation fails. Based on the conversation in [1], increase the size of the aead and skcipher request contexts to include an SGL descriptor that can handle a maximum of 4 flat buffers. If requests exceed 4 entries buffers, memory is allocated dynamically. [1] https://lore.kernel.org/linux-crypto/20200722072932.GA27544@gondor.apana.org.au/ Cc: stable@vger.kernel.org Fixes: d370cec32194 ("crypto: qat - Intel(R) QAT crypto interface") Reported-by: Mikulas Patocka Signed-off-by: Giovanni Cabiddu Reviewed-by: Marco Chiappero Reviewed-by: Wojciech Ziemba --- drivers/crypto/qat/qat_common/qat_algs.c | 64 +++++++++++++--------- drivers/crypto/qat/qat_common/qat_crypto.h | 24 ++++++++ 2 files changed, 61 insertions(+), 27 deletions(-) diff --git a/drivers/crypto/qat/qat_common/qat_algs.c b/drivers/crypto/qat/qat_common/qat_algs.c index f998ed58457c..ec635fe44c1f 100644 --- a/drivers/crypto/qat/qat_common/qat_algs.c +++ b/drivers/crypto/qat/qat_common/qat_algs.c @@ -46,19 +46,6 @@ static DEFINE_MUTEX(algs_lock); static unsigned int active_devs; -struct qat_alg_buf { - u32 len; - u32 resrvd; - u64 addr; -} __packed; - -struct qat_alg_buf_list { - u64 resrvd; - u32 num_bufs; - u32 num_mapped_bufs; - struct qat_alg_buf bufers[]; -} __packed __aligned(64); - /* Common content descriptor */ struct qat_alg_cd { union { @@ -693,7 +680,10 @@ static void qat_alg_free_bufl(struct qat_crypto_instance *inst, bl->bufers[i].len, DMA_BIDIRECTIONAL); dma_unmap_single(dev, blp, sz, DMA_TO_DEVICE); - kfree(bl); + + if (!qat_req->buf.sgl_src_valid) + kfree(bl); + if (blp != blpout) { /* If out of place operation dma unmap only data */ int bufless = blout->num_bufs - blout->num_mapped_bufs; @@ -704,7 +694,9 @@ static void qat_alg_free_bufl(struct qat_crypto_instance *inst, DMA_BIDIRECTIONAL); } dma_unmap_single(dev, blpout, sz_out, DMA_TO_DEVICE); - kfree(blout); + + if (!qat_req->buf.sgl_dst_valid) + kfree(blout); } } @@ -721,15 +713,24 @@ static int qat_alg_sgl_to_bufl(struct qat_crypto_instance *inst, dma_addr_t blp = DMA_MAPPING_ERROR; dma_addr_t bloutp = DMA_MAPPING_ERROR; struct scatterlist *sg; - size_t sz_out, sz = struct_size(bufl, bufers, n + 1); + size_t sz_out, sz = struct_size(bufl, bufers, n); + int node = dev_to_node(&GET_DEV(inst->accel_dev)); if (unlikely(!n)) return -EINVAL; - bufl = kzalloc_node(sz, GFP_ATOMIC, - dev_to_node(&GET_DEV(inst->accel_dev))); - if (unlikely(!bufl)) - return -ENOMEM; + qat_req->buf.sgl_src_valid = false; + qat_req->buf.sgl_dst_valid = false; + + if (n > QAT_MAX_BUFF_DESC) { + bufl = kzalloc_node(sz, GFP_ATOMIC, node); + if (unlikely(!bufl)) + return -ENOMEM; + } else { + bufl = &qat_req->buf.sgl_src.sgl_hdr; + memset(bufl, 0, sizeof(struct qat_alg_buf_list)); + qat_req->buf.sgl_src_valid = true; + } for_each_sg(sgl, sg, n, i) bufl->bufers[i].addr = DMA_MAPPING_ERROR; @@ -760,12 +761,18 @@ static int qat_alg_sgl_to_bufl(struct qat_crypto_instance *inst, struct qat_alg_buf *bufers; n = sg_nents(sglout); - sz_out = struct_size(buflout, bufers, n + 1); + sz_out = struct_size(buflout, bufers, n); sg_nctr = 0; - buflout = kzalloc_node(sz_out, GFP_ATOMIC, - dev_to_node(&GET_DEV(inst->accel_dev))); - if (unlikely(!buflout)) - goto err_in; + + if (n > QAT_MAX_BUFF_DESC) { + buflout = kzalloc_node(sz_out, GFP_ATOMIC, node); + if (unlikely(!buflout)) + goto err_in; + } else { + buflout = &qat_req->buf.sgl_dst.sgl_hdr; + memset(buflout, 0, sizeof(struct qat_alg_buf_list)); + qat_req->buf.sgl_dst_valid = true; + } bufers = buflout->bufers; for_each_sg(sglout, sg, n, i) @@ -810,7 +817,9 @@ static int qat_alg_sgl_to_bufl(struct qat_crypto_instance *inst, dma_unmap_single(dev, buflout->bufers[i].addr, buflout->bufers[i].len, DMA_BIDIRECTIONAL); - kfree(buflout); + + if (!qat_req->buf.sgl_dst_valid) + kfree(buflout); err_in: if (!dma_mapping_error(dev, blp)) @@ -823,7 +832,8 @@ static int qat_alg_sgl_to_bufl(struct qat_crypto_instance *inst, bufl->bufers[i].len, DMA_BIDIRECTIONAL); - kfree(bufl); + if (!qat_req->buf.sgl_src_valid) + kfree(bufl); dev_err(dev, "Failed to map buf for dma\n"); return -ENOMEM; diff --git a/drivers/crypto/qat/qat_common/qat_crypto.h b/drivers/crypto/qat/qat_common/qat_crypto.h index b6a4c95ae003..0928f159ea99 100644 --- a/drivers/crypto/qat/qat_common/qat_crypto.h +++ b/drivers/crypto/qat/qat_common/qat_crypto.h @@ -21,6 +21,26 @@ struct qat_crypto_instance { atomic_t refctr; }; +#define QAT_MAX_BUFF_DESC 4 + +struct qat_alg_buf { + u32 len; + u32 resrvd; + u64 addr; +} __packed; + +struct qat_alg_buf_list { + u64 resrvd; + u32 num_bufs; + u32 num_mapped_bufs; + struct qat_alg_buf bufers[]; +} __packed; + +struct qat_alg_fixed_buf_list { + struct qat_alg_buf_list sgl_hdr; + struct qat_alg_buf descriptors[QAT_MAX_BUFF_DESC]; +} __packed __aligned(64); + struct qat_crypto_request_buffs { struct qat_alg_buf_list *bl; dma_addr_t blp; @@ -28,6 +48,10 @@ struct qat_crypto_request_buffs { dma_addr_t bloutp; size_t sz; size_t sz_out; + bool sgl_src_valid; + bool sgl_dst_valid; + struct qat_alg_fixed_buf_list sgl_src; + struct qat_alg_fixed_buf_list sgl_dst; }; struct qat_crypto_request;