Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.8 required=3.0 tests=DKIM_INVALID,DKIM_SIGNED, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY, SPF_PASS autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 55AB3C10F11 for ; Wed, 10 Apr 2019 20:07:12 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 1B3B22075B for ; Wed, 10 Apr 2019 20:07:12 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=fail reason="signature verification failed" (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="Mlj2hWDl" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726085AbfDJUHL (ORCPT ); Wed, 10 Apr 2019 16:07:11 -0400 Received: from mail-it1-f193.google.com ([209.85.166.193]:52554 "EHLO mail-it1-f193.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726623AbfDJUHL (ORCPT ); Wed, 10 Apr 2019 16:07:11 -0400 Received: by mail-it1-f193.google.com with SMTP id x132so5661615itf.2; Wed, 10 Apr 2019 13:07:10 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=sender:subject:from:to:date:message-id:in-reply-to:references :user-agent:mime-version:content-transfer-encoding; bh=T73l4MNjLo8EXHOhIrQ5HeZlHAH/77uD80btH/xxBEA=; b=Mlj2hWDl42wiiL9YB1p/El4Lk5MlUz5TExYsxycfr10VgOc7jT4UB7vn2f4Rfv4NdX uVMWgpR5CJwcZ5D+fOPii7MscwkEDLXHpacAYuHrY68YiYHILVOpIDEGwcmboIPfhcJv EZjnnnIh/EKWEWzYRKlw9iDdFbNv1VcNdUc3VyOaZSfzmJeN5zeGE2wZmJa49mlX8ckh ZVEyrADrtPzFnCTyulHTKPnuEy0kOEYIVKd4aQ55I3gA7df4ysKD07lOMpRyc2Sh/B7c yPc1UDGNVfkqbygcCCo+ut3b6v97MfSwe2/T5rV6fhyxzdjap562oddVBb2CVdTZIBHm lHSw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:sender:subject:from:to:date:message-id :in-reply-to:references:user-agent:mime-version :content-transfer-encoding; bh=T73l4MNjLo8EXHOhIrQ5HeZlHAH/77uD80btH/xxBEA=; b=Bs2aENP0eZMaYVdSBrGcjFa0z0LazokBCf1uINV+YkJDNjXre4zJSbnOgsbQG0j/fq z8evGwL/fT89neCvE5BnexBv2gqPMIJmX0SVyi06IBDaXH+x7jWRHQ3ik9eJUxptNKpC 1p933i5I3iILvVFdeqBnqWdHr0ySeDO4WMDMwLPaGCMk232Yw+oTt9sPgYiVLLhlm4YJ h8oHLqbP3TQ1IQJpWcB1StGlAgzEjHf+2CE0t4vp8EKNR0REgXVpznXYhomPKliMv9Px kI2IU19xFHtmnzTaByzYsxOinN6d0eexVJNpWf2eaQLgYMFbCusJfgqnTNuQooLWYI00 SU3g== X-Gm-Message-State: APjAAAWClLChygo4fkz9GIq6rBKapsqWE1TWbsBxSKlFL+pn/504wb0B dNUskUt2YTXf4xZYhRqtIBGiZWF5 X-Google-Smtp-Source: APXvYqzCxnyORi1sVchV+l/C27JhnxuCn6FUdU+8QsnXe822omhwcY5rvhGVz4JcX8hyjYq25wdgng== X-Received: by 2002:a24:9197:: with SMTP id i145mr4805001ite.117.1554926830149; Wed, 10 Apr 2019 13:07:10 -0700 (PDT) Received: from gateway.1015granger.net (c-68-61-232-219.hsd1.mi.comcast.net. [68.61.232.219]) by smtp.gmail.com with ESMTPSA id y203sm1515778itb.22.2019.04.10.13.07.09 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 10 Apr 2019 13:07:09 -0700 (PDT) Received: from manet.1015granger.net (manet.1015granger.net [192.168.1.51]) by gateway.1015granger.net (8.14.7/8.14.7) with ESMTP id x3AK78YL004525; Wed, 10 Apr 2019 20:07:08 GMT Subject: [PATCH v1 07/19] xprtrdma: Allocate req's regbufs at xprt create time From: Chuck Lever To: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org Date: Wed, 10 Apr 2019 16:07:08 -0400 Message-ID: <20190410200708.11522.60558.stgit@manet.1015granger.net> In-Reply-To: <20190410200446.11522.21145.stgit@manet.1015granger.net> References: <20190410200446.11522.21145.stgit@manet.1015granger.net> User-Agent: StGit/0.17.1-dirty MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Sender: linux-nfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-nfs@vger.kernel.org Allocating an rpcrdma_req's regbufs at xprt create time enables a pair of micro-optimizations: First, if these regbufs are always there, we can eliminate two conditional branches from the hot xprt_rdma_allocate path. Second, by allocating a 1KB buffer, it places a lower bound on the size of these buffers, without adding yet another conditional branch. The lower bound reduces the number of hardway re- allocations. In fact, for some workloads it completely eliminates hardway allocations. Signed-off-by: Chuck Lever --- net/sunrpc/xprtrdma/backchannel.c | 18 ++++-------------- net/sunrpc/xprtrdma/transport.c | 4 ++-- net/sunrpc/xprtrdma/verbs.c | 34 ++++++++++++++++++++++++++-------- net/sunrpc/xprtrdma/xprt_rdma.h | 2 +- 4 files changed, 33 insertions(+), 25 deletions(-) diff --git a/net/sunrpc/xprtrdma/backchannel.c b/net/sunrpc/xprtrdma/backchannel.c index 6170ec7..e1a125a 100644 --- a/net/sunrpc/xprtrdma/backchannel.c +++ b/net/sunrpc/xprtrdma/backchannel.c @@ -28,10 +28,10 @@ static int rpcrdma_bc_setup_reqs(struct rpcrdma_xprt *r_xprt, unsigned int i; for (i = 0; i < (count << 1); i++) { - struct rpcrdma_regbuf *rb; size_t size; - req = rpcrdma_req_create(r_xprt, GFP_KERNEL); + size = min_t(size_t, r_xprt->rx_data.inline_rsize, PAGE_SIZE); + req = rpcrdma_req_create(r_xprt, size, GFP_KERNEL); if (!req) return -ENOMEM; rqst = &req->rl_slot; @@ -42,20 +42,10 @@ static int rpcrdma_bc_setup_reqs(struct rpcrdma_xprt *r_xprt, spin_lock(&xprt->bc_pa_lock); list_add(&rqst->rq_bc_pa_list, &xprt->bc_pa_list); spin_unlock(&xprt->bc_pa_lock); - - size = r_xprt->rx_data.inline_rsize; - rb = rpcrdma_alloc_regbuf(size, DMA_TO_DEVICE, GFP_KERNEL); - if (!rb) - goto out_fail; - req->rl_sendbuf = rb; - xdr_buf_init(&rqst->rq_snd_buf, rdmab_data(rb), - min_t(size_t, size, PAGE_SIZE)); + xdr_buf_init(&rqst->rq_snd_buf, rdmab_data(req->rl_sendbuf), + size); } return 0; - -out_fail: - rpcrdma_req_destroy(req); - return -ENOMEM; } /** diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c index e3b5b91..09a4693 100644 --- a/net/sunrpc/xprtrdma/transport.c +++ b/net/sunrpc/xprtrdma/transport.c @@ -591,7 +591,7 @@ void xprt_rdma_close(struct rpc_xprt *xprt) { struct rpcrdma_regbuf *rb; - if (req->rl_sendbuf && rdmab_length(req->rl_sendbuf) >= size) + if (likely(rdmab_length(req->rl_sendbuf) >= size)) return true; rb = rpcrdma_alloc_regbuf(size, DMA_TO_DEVICE, flags); @@ -621,7 +621,7 @@ void xprt_rdma_close(struct rpc_xprt *xprt) { struct rpcrdma_regbuf *rb; - if (req->rl_recvbuf && rdmab_length(req->rl_recvbuf) >= size) + if (likely(rdmab_length(req->rl_recvbuf) >= size)) return true; rb = rpcrdma_alloc_regbuf(size, DMA_NONE, flags); diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index ca2d6d8..e4644fd 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -998,11 +998,13 @@ struct rpcrdma_sendctx *rpcrdma_sendctx_get_locked(struct rpcrdma_buffer *buf) /** * rpcrdma_req_create - Allocate an rpcrdma_req object * @r_xprt: controlling r_xprt + * @size: initial size, in bytes, of send and receive buffers * @flags: GFP flags passed to memory allocators * * Returns an allocated and fully initialized rpcrdma_req or NULL. */ -struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, gfp_t flags) +struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size, + gfp_t flags) { struct rpcrdma_buffer *buffer = &r_xprt->rx_buf; struct rpcrdma_regbuf *rb; @@ -1010,22 +1012,37 @@ struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, gfp_t flags) req = kzalloc(sizeof(*req), flags); if (req == NULL) - return NULL; + goto out1; rb = rpcrdma_alloc_regbuf(RPCRDMA_HDRBUF_SIZE, DMA_TO_DEVICE, flags); - if (!rb) { - kfree(req); - return NULL; - } + if (!rb) + goto out2; req->rl_rdmabuf = rb; xdr_buf_init(&req->rl_hdrbuf, rdmab_data(rb), rdmab_length(rb)); + + req->rl_sendbuf = rpcrdma_alloc_regbuf(size, DMA_TO_DEVICE, flags); + if (!req->rl_sendbuf) + goto out3; + + req->rl_recvbuf = rpcrdma_alloc_regbuf(size, DMA_NONE, flags); + if (!req->rl_recvbuf) + goto out4; + req->rl_buffer = buffer; INIT_LIST_HEAD(&req->rl_registered); - spin_lock(&buffer->rb_lock); list_add(&req->rl_all, &buffer->rb_allreqs); spin_unlock(&buffer->rb_lock); return req; + +out4: + kfree(req->rl_sendbuf); +out3: + kfree(req->rl_rdmabuf); +out2: + kfree(req); +out1: + return NULL; } static bool rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt, bool temp) @@ -1090,7 +1107,8 @@ static bool rpcrdma_rep_create(struct rpcrdma_xprt *r_xprt, bool temp) for (i = 0; i < buf->rb_max_requests; i++) { struct rpcrdma_req *req; - req = rpcrdma_req_create(r_xprt, GFP_KERNEL); + req = rpcrdma_req_create(r_xprt, RPCRDMA_V1_DEF_INLINE_SIZE, + GFP_KERNEL); if (!req) goto out; list_add(&req->rl_list, &buf->rb_send_bufs); diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h index 73f9e54..202294a 100644 --- a/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/net/sunrpc/xprtrdma/xprt_rdma.h @@ -529,7 +529,7 @@ int rpcrdma_ep_post(struct rpcrdma_ia *, struct rpcrdma_ep *, /* * Buffer calls - xprtrdma/verbs.c */ -struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, +struct rpcrdma_req *rpcrdma_req_create(struct rpcrdma_xprt *r_xprt, size_t size, gfp_t flags); void rpcrdma_req_destroy(struct rpcrdma_req *req); int rpcrdma_buffer_create(struct rpcrdma_xprt *);