From patchwork Fri Jun 9 10:39:46 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ekansh Gupta X-Patchwork-Id: 105555 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a05:6358:3046:b0:115:7a1d:dabb with SMTP id p6csp905465rwl; Fri, 9 Jun 2023 04:02:36 -0700 (PDT) X-Google-Smtp-Source: ACHHUZ7BYXeKi2OisjvTJEdJXs1Il21AaDOsMXZmLcf3H4QwYdG1UXAfKYsdzB8dkLgG8NpXoDDV X-Received: by 2002:a05:6a20:4326:b0:10f:1d33:d667 with SMTP id h38-20020a056a20432600b0010f1d33d667mr1218456pzk.5.1686308555595; Fri, 09 Jun 2023 04:02:35 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1686308555; cv=none; d=google.com; s=arc-20160816; b=AZNCj2NmGPvmShhP1iQRYwoAcZVodpaAgKW7mxgWADCNFUaadA4hKIfvIxOPCmrxw0 A7koF5xtYDCMOI+hOaaqHCBtJZhR+OGJVKDMKXtyg5yFPVE77BDhcZ1SwjdIUdDvKv05 ZWtm4WQH2J1aG015kkvFpfnx2Bm4fCxU40/XQEvMgeu1lELcv2X3KB+cv0PCKHvFfYdQ kijySIJeIio6T+dT4GWtNcB2eOCLbRL0pcZfEQZkMQSqn52qxY7HPd2iJv6WG8OLFjGc T8juvl2mSUfnmLjzG3gjK8HkvdJVdhtIKmkN+t+WX5YZcF0PyJ8KHuzs5wqUNh5aq8yp s9eg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:mime-version:references:in-reply-to:message-id :date:subject:cc:to:from:dkim-signature; bh=NN6eiIuHt5ju4Koc6y3iuQjGcM9R8729RQjnsafk9CA=; b=pT38r+Zo0VRcRoc2bFcUJzy3UDzExbyy02jdzcJXtzR95fotRU1oFXOwHAAqZI20dB nHD6O4DXYmnaTcPgU6F4AtZts/TaMRdE85se71GhQg12dnZx2njziNZ9/7LhUYYiQzs0 KimZSrPXAQnLaia5aICGNiN0D417Ivo8CcO5hYdC+yaqhnaeFVmIEojMKmRNmdJmJQFJ spckPXk4i1gcwiAhBipt8uCNKfxld6srcwhm5uxGH9FUYP66Hyid/Fav3PZYXDPf3X1q 10v0wJ370nbVkWV2sR5hY1ZK6k4ydooP2oUxAXb5UahpzcAjvouagA2mpEI2nqFV5uAi qfoA== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@quicinc.com header.s=qcppdkim1 header.b=VlaMyI0N; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=quicinc.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id w187-20020a627bc4000000b0065304909e29si2296631pfc.187.2023.06.09.04.02.22; Fri, 09 Jun 2023 04:02:35 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; dkim=pass header.i=@quicinc.com header.s=qcppdkim1 header.b=VlaMyI0N; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=quicinc.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238739AbjFIKnN (ORCPT + 99 others); Fri, 9 Jun 2023 06:43:13 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:42870 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240154AbjFIKl7 (ORCPT ); Fri, 9 Jun 2023 06:41:59 -0400 Received: from mx0b-0031df01.pphosted.com (mx0b-0031df01.pphosted.com [205.220.180.131]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 02EB21BFF; Fri, 9 Jun 2023 03:40:00 -0700 (PDT) Received: from pps.filterd (m0279868.ppops.net [127.0.0.1]) by mx0a-0031df01.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 359ATMXj001184; Fri, 9 Jun 2023 10:39:59 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=quicinc.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=qcppdkim1; bh=NN6eiIuHt5ju4Koc6y3iuQjGcM9R8729RQjnsafk9CA=; b=VlaMyI0N1snyZs1JzHwNCn3DBdXRXAzd5XgVpvSKOwHPh/1UfomFG1oXUW/X/+DFA5xI iYOvwveA7GYliPCMsKQqHvthBA27txkZpunTDZP1JBTKEFVrKgqzvAixko7HeoD2lpAo xresGEjt2QlIh+dhEypZVwuUtutG78CdiiRikU9xfjvOU+40wLz6xEPGg5afHHkRPoIY N51TYiZrsQFTLMErd/IzSidjQNRKzGu70WrOhfslSkDfYHSdI+D46l7JsNTW+OKDepAZ vir8eBjTDCEEnyDTdGDULYXVxFzhtrGI1TQNKbQhGC8A+V+ueIDmibBLk+UGss20BFZQ vA== Received: from nalasppmta05.qualcomm.com (Global_NAT1.qualcomm.com [129.46.96.20]) by mx0a-0031df01.pphosted.com (PPS) with ESMTPS id 3r3w7drkv0-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 09 Jun 2023 10:39:58 +0000 Received: from nalasex01b.na.qualcomm.com (nalasex01b.na.qualcomm.com [10.47.209.197]) by NALASPPMTA05.qualcomm.com (8.17.1.5/8.17.1.5) with ESMTPS id 359AdvQN018347 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 9 Jun 2023 10:39:57 GMT Received: from ekangupt-linux.qualcomm.com (10.80.80.8) by nalasex01b.na.qualcomm.com (10.47.209.197) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.2.986.42; Fri, 9 Jun 2023 03:39:54 -0700 From: Ekansh Gupta To: , CC: Ekansh Gupta , , , , , Subject: [PATCH v1 1/2] misc: fastrpc: Add cached buffer support Date: Fri, 9 Jun 2023 16:09:46 +0530 Message-ID: <1686307187-15199-2-git-send-email-quic_ekangupt@quicinc.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1686307187-15199-1-git-send-email-quic_ekangupt@quicinc.com> References: <1686307187-15199-1-git-send-email-quic_ekangupt@quicinc.com> MIME-Version: 1.0 X-Originating-IP: [10.80.80.8] X-ClientProxiedBy: nasanex01a.na.qualcomm.com (10.52.223.231) To nalasex01b.na.qualcomm.com (10.47.209.197) X-QCInternal: smtphost X-Proofpoint-Virus-Version: vendor=nai engine=6200 definitions=5800 signatures=585085 X-Proofpoint-GUID: tEjLTE9JCckode59PBlJfLTN2Lclo_S1 X-Proofpoint-ORIG-GUID: tEjLTE9JCckode59PBlJfLTN2Lclo_S1 X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.957,Hydra:6.0.573,FMLib:17.11.176.26 definitions=2023-06-09_06,2023-06-08_01,2023-05-22_02 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 malwarescore=0 bulkscore=0 spamscore=0 impostorscore=0 phishscore=0 adultscore=0 mlxlogscore=999 lowpriorityscore=0 suspectscore=0 clxscore=1015 priorityscore=1501 mlxscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2305260000 definitions=main-2306090089 X-Spam-Status: No, score=-2.1 required=5.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,SPF_HELO_NONE,SPF_PASS, T_SCC_BODY_TEXT_LINE,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1768222679698201370?= X-GMAIL-MSGID: =?utf-8?q?1768222679698201370?= FastRPC driver uses DMA API for region allocation for any buf_alloc request and the DMA region is unmap and freed once the usage is complete. Cached buffer support enables caching of certain types of buffers instead of freeing which help in reducing the overhead of DMA API call for every buffer allocation request. A list of cached buffer is maintained which will get reused when needed and this buffer list will get freed during device release. Co-developed-by: Anandu Krishnan E Signed-off-by: Anandu Krishnan E Signed-off-by: Ekansh Gupta --- drivers/misc/fastrpc.c | 133 +++++++++++++++++++++++++++++++++++++------- include/uapi/misc/fastrpc.h | 8 +++ 2 files changed, 122 insertions(+), 19 deletions(-) diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c index 30d4d04..a961a66 100644 --- a/drivers/misc/fastrpc.c +++ b/drivers/misc/fastrpc.c @@ -41,6 +41,10 @@ #define INIT_FILE_NAMELEN_MAX (128) #define FASTRPC_DEVICE_NAME "fastrpc" +/* Maximum buffers cached in cached buffer list */ +#define FASTRPC_MAX_CACHED_BUFS (32) +#define FASTRPC_MAX_CACHE_BUF_SIZE (8*1024*1024) + /* Add memory to static PD pool, protection thru XPU */ #define ADSP_MMAP_HEAP_ADDR 4 /* MAP static DMA buffer on DSP User PD */ @@ -195,6 +199,7 @@ struct fastrpc_buf { struct dma_buf *dmabuf; struct device *dev; void *virt; + u32 type; u64 phys; u64 size; /* Lock for dma buf attachments */ @@ -293,6 +298,7 @@ struct fastrpc_user { struct list_head maps; struct list_head pending; struct list_head mmaps; + struct list_head cached_bufs; struct fastrpc_channel_ctx *cctx; struct fastrpc_session_ctx *sctx; @@ -300,6 +306,8 @@ struct fastrpc_user { int tgid; int pd; + /* total cached buffers */ + u32 num_cached_buf; bool is_secure_dev; /* Lock for lists */ spinlock_t lock; @@ -391,17 +399,95 @@ static int fastrpc_map_lookup(struct fastrpc_user *fl, int fd, return ret; } -static void fastrpc_buf_free(struct fastrpc_buf *buf) +static void __fastrpc_buf_free(struct fastrpc_buf *buf) { dma_free_coherent(buf->dev, buf->size, buf->virt, FASTRPC_PHYS(buf->phys)); kfree(buf); } +static void fastrpc_cached_buf_list_add(struct fastrpc_buf *buf) +{ + struct fastrpc_user *fl = buf->fl; + + if (buf->size < FASTRPC_MAX_CACHE_BUF_SIZE) { + spin_lock(&fl->lock); + if (fl->num_cached_buf > FASTRPC_MAX_CACHED_BUFS) { + spin_unlock(&fl->lock); + goto skip_buf_cache; + } + + list_add_tail(&buf->node, &fl->cached_bufs); + fl->num_cached_buf++; + buf->type = -1; + spin_unlock(&fl->lock); + } + return; + +skip_buf_cache: + __fastrpc_buf_free(buf); +} + +static void fastrpc_buf_free(struct fastrpc_buf *buf, bool cache) +{ + if (cache) + fastrpc_cached_buf_list_add(buf); + else + __fastrpc_buf_free(buf); +} + +static inline bool fastrpc_get_cached_buf(struct fastrpc_user *fl, + size_t size, int buf_type, struct fastrpc_buf **obuf) +{ + bool found = false; + struct fastrpc_buf *buf, *n, *cbuf = NULL; + + if (buf_type == USER_BUF) + return found; + + /* find the smallest buffer that fits in the cache */ + spin_lock(&fl->lock); + list_for_each_entry_safe(buf, n, &fl->cached_bufs, node) { + if (buf->size >= size && (!cbuf || cbuf->size > buf->size)) + cbuf = buf; + } + if (cbuf) { + list_del_init(&cbuf->node); + fl->num_cached_buf--; + } + spin_unlock(&fl->lock); + if (cbuf) { + cbuf->type = buf_type; + *obuf = cbuf; + found = true; + } + + return found; +} + +static void fastrpc_cached_buf_list_free(struct fastrpc_user *fl) +{ + struct fastrpc_buf *buf, *n, *free; + + do { + free = NULL; + spin_lock(&fl->lock); + list_for_each_entry_safe(buf, n, &fl->cached_bufs, node) { + list_del(&buf->node); + fl->num_cached_buf--; + free = buf; + break; + } + spin_unlock(&fl->lock); + if (free) + fastrpc_buf_free(free, false); + } while (free); +} + static int __fastrpc_buf_alloc(struct fastrpc_user *fl, struct device *dev, u64 size, struct fastrpc_buf **obuf) { - struct fastrpc_buf *buf; + struct fastrpc_buf *buf = NULL; buf = kzalloc(sizeof(*buf), GFP_KERNEL); if (!buf) @@ -432,16 +518,23 @@ static int __fastrpc_buf_alloc(struct fastrpc_user *fl, struct device *dev, } static int fastrpc_buf_alloc(struct fastrpc_user *fl, struct device *dev, - u64 size, struct fastrpc_buf **obuf) + u64 size, u32 buf_type, struct fastrpc_buf **obuf) { int ret; struct fastrpc_buf *buf; + if (fastrpc_get_cached_buf(fl, size, buf_type, obuf)) + return 0; ret = __fastrpc_buf_alloc(fl, dev, size, obuf); - if (ret) - return ret; + if (ret == -ENOMEM) { + fastrpc_cached_buf_list_free(fl); + ret = __fastrpc_buf_alloc(fl, dev, size, obuf); + if (ret) + return ret; + } buf = *obuf; + buf->type = buf_type; if (fl->sctx && fl->sctx->sid) buf->phys += ((u64)fl->sctx->sid << 32); @@ -490,7 +583,7 @@ static void fastrpc_context_free(struct kref *ref) fastrpc_map_put(ctx->maps[i]); if (ctx->buf) - fastrpc_buf_free(ctx->buf); + fastrpc_buf_free(ctx->buf, true); spin_lock_irqsave(&cctx->lock, flags); idr_remove(&cctx->ctx_idr, ctx->ctxid >> 4); @@ -674,7 +767,7 @@ static void fastrpc_release(struct dma_buf *dmabuf) { struct fastrpc_buf *buffer = dmabuf->priv; - fastrpc_buf_free(buffer); + fastrpc_buf_free(buffer, false); } static int fastrpc_dma_buf_attach(struct dma_buf *dmabuf, @@ -951,7 +1044,7 @@ static int fastrpc_get_args(u32 kernel, struct fastrpc_invoke_ctx *ctx) ctx->msg_sz = pkt_size; - err = fastrpc_buf_alloc(ctx->fl, dev, pkt_size, &ctx->buf); + err = fastrpc_buf_alloc(ctx->fl, dev, pkt_size, METADATA_BUF, &ctx->buf); if (err) return err; @@ -1334,7 +1427,7 @@ static int fastrpc_init_create_static_process(struct fastrpc_user *fl, fl->cctx->remote_heap->phys, fl->cctx->remote_heap->size, err); } err_map: - fastrpc_buf_free(fl->cctx->remote_heap); + fastrpc_buf_free(fl->cctx->remote_heap, false); err_name: kfree(name); err: @@ -1403,7 +1496,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, memlen = ALIGN(max(INIT_FILELEN_MAX, (int)init.filelen * 4), 1024 * 1024); err = fastrpc_buf_alloc(fl, fl->sctx->dev, memlen, - &imem); + INITMEM_BUF, &imem); if (err) goto err_alloc; @@ -1450,7 +1543,7 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, err_invoke: fl->init_mem = NULL; - fastrpc_buf_free(imem); + fastrpc_buf_free(imem, false); err_alloc: fastrpc_map_put(map); err: @@ -1521,7 +1614,7 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) spin_unlock_irqrestore(&cctx->lock, flags); if (fl->init_mem) - fastrpc_buf_free(fl->init_mem); + fastrpc_buf_free(fl->init_mem, false); list_for_each_entry_safe(ctx, n, &fl->pending, node) { list_del(&ctx->node); @@ -1533,9 +1626,10 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) list_for_each_entry_safe(buf, b, &fl->mmaps, node) { list_del(&buf->node); - fastrpc_buf_free(buf); + fastrpc_buf_free(buf, false); } + fastrpc_cached_buf_list_free(fl); fastrpc_session_free(cctx, fl->sctx); fastrpc_channel_ctx_put(cctx); @@ -1570,6 +1664,7 @@ static int fastrpc_device_open(struct inode *inode, struct file *filp) INIT_LIST_HEAD(&fl->maps); INIT_LIST_HEAD(&fl->mmaps); INIT_LIST_HEAD(&fl->user); + INIT_LIST_HEAD(&fl->cached_bufs); fl->tgid = current->tgid; fl->cctx = cctx; fl->is_secure_dev = fdevice->secure; @@ -1600,7 +1695,7 @@ static int fastrpc_dmabuf_alloc(struct fastrpc_user *fl, char __user *argp) if (copy_from_user(&bp, argp, sizeof(bp))) return -EFAULT; - err = fastrpc_buf_alloc(fl, fl->sctx->dev, bp.size, &buf); + err = fastrpc_buf_alloc(fl, fl->sctx->dev, bp.size, USER_BUF, &buf); if (err) return err; exp_info.ops = &fastrpc_dma_buf_ops; @@ -1610,7 +1705,7 @@ static int fastrpc_dmabuf_alloc(struct fastrpc_user *fl, char __user *argp) buf->dmabuf = dma_buf_export(&exp_info); if (IS_ERR(buf->dmabuf)) { err = PTR_ERR(buf->dmabuf); - fastrpc_buf_free(buf); + fastrpc_buf_free(buf, false); return err; } @@ -1805,7 +1900,7 @@ static int fastrpc_req_munmap_impl(struct fastrpc_user *fl, struct fastrpc_buf * spin_lock(&fl->lock); list_del(&buf->node); spin_unlock(&fl->lock); - fastrpc_buf_free(buf); + fastrpc_buf_free(buf, false); } else { dev_err(dev, "unmmap\tpt 0x%09lx ERROR\n", buf->raddr); } @@ -1866,7 +1961,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp) return -EINVAL; } - err = fastrpc_buf_alloc(fl, fl->sctx->dev, req.size, &buf); + err = fastrpc_buf_alloc(fl, fl->sctx->dev, req.size, USER_BUF, &buf); if (err) { dev_err(dev, "failed to allocate buffer\n"); return err; @@ -1935,7 +2030,7 @@ static int fastrpc_req_mmap(struct fastrpc_user *fl, char __user *argp) err_assign: fastrpc_req_munmap_impl(fl, buf); err_invoke: - fastrpc_buf_free(buf); + fastrpc_buf_free(buf, false); return err; } @@ -2380,7 +2475,7 @@ static void fastrpc_rpmsg_remove(struct rpmsg_device *rpdev) list_del(&buf->node); if (cctx->remote_heap) - fastrpc_buf_free(cctx->remote_heap); + fastrpc_buf_free(cctx->remote_heap, false); of_platform_depopulate(&rpdev->dev); diff --git a/include/uapi/misc/fastrpc.h b/include/uapi/misc/fastrpc.h index f33d914..91c7c4f 100644 --- a/include/uapi/misc/fastrpc.h +++ b/include/uapi/misc/fastrpc.h @@ -64,6 +64,14 @@ enum fastrpc_proc_attr { FASTRPC_MODE_PRIVILEGED = (1 << 6), }; + /* Types of fastrpc DMA bufs sent to DSP */ + enum fastrpc_buf_type { + METADATA_BUF, + COPYDATA_BUF, + INITMEM_BUF, + USER_BUF, +}; + /* Fastrpc attribute for memory protection of buffers */ #define FASTRPC_ATTR_SECUREMAP (1) From patchwork Fri Jun 9 10:39:47 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ekansh Gupta X-Patchwork-Id: 105547 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:994d:0:b0:3d9:f83d:47d9 with SMTP id k13csp852481vqr; Fri, 9 Jun 2023 03:44:54 -0700 (PDT) X-Google-Smtp-Source: ACHHUZ6hQvL4VYLMY4DzToIWZhC+HOJAL0Cjlz7IbRU/21xl8dOsGFmZdQ6eJEtuktZLUO66ONOY X-Received: by 2002:a05:6a20:7da6:b0:10c:d5dd:c223 with SMTP id v38-20020a056a207da600b0010cd5ddc223mr1529050pzj.15.1686307494355; Fri, 09 Jun 2023 03:44:54 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1686307494; cv=none; d=google.com; s=arc-20160816; b=Zz+JaX7BhNX5wQIKuOYHjAgvlG6ek0GByy3HPeyH+VlGW4PyeQrpm7XDDp/XP8EQ7P NVFh6miMHuB76l7ojv4hZUh7z8amcbprl5PWW64qMwFrfvv1qHJW0EIbpCtWD/zarXAP ChbBMnld7SZw4CRwegUX1JhhS8e94MgQV/RCoEV4sK9is41rm/mlpQQ+QGVO+QD0/bpU su+EQQ2yDdMiwLJ2a/JYWAdIwuurEua/2c3Zy2qOg3DZxWzt/x3GGgXiVpNksQY0PRHH 6+M8m4q4oVtoRAtA1EmmIQrnM4mpdn1ch/kAInsv+9lKnWucAanmELx3TKm+FoqOy14R 7KcQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:mime-version:references:in-reply-to:message-id :date:subject:cc:to:from:dkim-signature; bh=1Mx5zaCc4BAFmXKZXl06MSVCN40RKvwkVuuo/nEoGLI=; b=vgiGSJxmkCkVWp5lmRFEaEv6oBoZZe1e9FvsUHY5+NjP6RkKrzdEa3ZI39maAWIpZz RdrPyt6h/+9X1G+FmB8QdbTlUosITPp7LM797nxm1JRe7tIEWPeTX8CEJxwBp5gXeuPZ 8O6BdvoPIQfsD6JbK1GhUaXlqzRhiF8yVlWGHROZ+p9naGNHlXm0zWyss41pOsBO+XyE dUtJaaEBzEAcZfyalKsrBBo8ctqp+wv7MabXqASJX7euGNjXF+Ai2GMnhfdop+rEDo6w Cdggvd8iudcgT4rzYIGdIXcPjQPZyi4Hus5kX+QWd06dKoCC1j687EQXufv5PzgGpFnO xWpw== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@quicinc.com header.s=qcppdkim1 header.b=iyljq8RH; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=quicinc.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id bv2-20020a632e02000000b00534897b70f9si2332566pgb.29.2023.06.09.03.44.39; Fri, 09 Jun 2023 03:44:54 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; dkim=pass header.i=@quicinc.com header.s=qcppdkim1 header.b=iyljq8RH; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=quicinc.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S238791AbjFIKnS (ORCPT + 99 others); Fri, 9 Jun 2023 06:43:18 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:40998 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240131AbjFIKl7 (ORCPT ); Fri, 9 Jun 2023 06:41:59 -0400 Received: from mx0b-0031df01.pphosted.com (mx0b-0031df01.pphosted.com [205.220.180.131]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 353A01FDC; Fri, 9 Jun 2023 03:40:04 -0700 (PDT) Received: from pps.filterd (m0279870.ppops.net [127.0.0.1]) by mx0a-0031df01.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 359AJMaq016557; Fri, 9 Jun 2023 10:40:02 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=quicinc.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-type; s=qcppdkim1; bh=1Mx5zaCc4BAFmXKZXl06MSVCN40RKvwkVuuo/nEoGLI=; b=iyljq8RHyll58dtbJthgsQAuycKgXm103r5Ivs+ck/r/yJen1b+rAw20zzCsbA5FvvOG N1bEZZBFqsDnvDlZWYGI+Rj3ddryOeE9kh7u/YzGH2+/zdFxVrTiy/p0Je6OSmhYokXj v0pWnNEDIlN37KqA2x+fjC3x7owdNWngNIrGhx2rce1l25LwwiaevZ7dZLEnMEsdYIFG VoAgUwhRdCdr3EOX5SA8TIVc4SWyO+0TKmP0NqvIvDGQ6hUsFbLszFh8cdctJCdYUe+X tfWR6uAffjCvJlbKLid+vJqHw8Sw/vn32/9GFEiDcNNL03UDcrgFG5089f5G3KGa5yXU Uw== Received: from nalasppmta04.qualcomm.com (Global_NAT1.qualcomm.com [129.46.96.20]) by mx0a-0031df01.pphosted.com (PPS) with ESMTPS id 3r3vu4gn90-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 09 Jun 2023 10:40:02 +0000 Received: from nalasex01b.na.qualcomm.com (nalasex01b.na.qualcomm.com [10.47.209.197]) by NALASPPMTA04.qualcomm.com (8.17.1.5/8.17.1.5) with ESMTPS id 359Ae0hL027348 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Fri, 9 Jun 2023 10:40:01 GMT Received: from ekangupt-linux.qualcomm.com (10.80.80.8) by nalasex01b.na.qualcomm.com (10.47.209.197) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.2.986.42; Fri, 9 Jun 2023 03:39:57 -0700 From: Ekansh Gupta To: , CC: Ekansh Gupta , , , , , Subject: [PATCH v1 2/2] misc: fastrpc: Add persistent header buffer support Date: Fri, 9 Jun 2023 16:09:47 +0530 Message-ID: <1686307187-15199-3-git-send-email-quic_ekangupt@quicinc.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1686307187-15199-1-git-send-email-quic_ekangupt@quicinc.com> References: <1686307187-15199-1-git-send-email-quic_ekangupt@quicinc.com> MIME-Version: 1.0 X-Originating-IP: [10.80.80.8] X-ClientProxiedBy: nasanex01a.na.qualcomm.com (10.52.223.231) To nalasex01b.na.qualcomm.com (10.47.209.197) X-QCInternal: smtphost X-Proofpoint-Virus-Version: vendor=nai engine=6200 definitions=5800 signatures=585085 X-Proofpoint-GUID: vB4PLFwChLabtF-Jc0c9zWpveYkVfoPD X-Proofpoint-ORIG-GUID: vB4PLFwChLabtF-Jc0c9zWpveYkVfoPD X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.957,Hydra:6.0.573,FMLib:17.11.176.26 definitions=2023-06-09_06,2023-06-09_01,2023-05-22_02 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 bulkscore=0 mlxlogscore=999 lowpriorityscore=0 phishscore=0 priorityscore=1501 clxscore=1015 suspectscore=0 spamscore=0 mlxscore=0 adultscore=0 malwarescore=0 impostorscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2305260000 definitions=main-2306090089 X-Spam-Status: No, score=-2.1 required=5.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,SPF_HELO_NONE,SPF_PASS, T_SCC_BODY_TEXT_LINE,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1768221567193678635?= X-GMAIL-MSGID: =?utf-8?q?1768221567193678635?= This change enables the usage of persistent header buffers which are allocated during process initialization and is mapped to CDSP. This mechanism helps in reducing the RPC overhead for remote calls. For any user process remote RPC call to CDSP, it is tried to use persistent header buffers. In case all persistent buffers are already in use, there is a fallback to cached buffer usage and allocation using DMA API. Co-developed-by: Anandu Krishnan E Signed-off-by: Anandu Krishnan E Signed-off-by: Ekansh Gupta --- drivers/misc/fastrpc.c | 215 +++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 173 insertions(+), 42 deletions(-) diff --git a/drivers/misc/fastrpc.c b/drivers/misc/fastrpc.c index a961a66..0132b63 100644 --- a/drivers/misc/fastrpc.c +++ b/drivers/misc/fastrpc.c @@ -44,6 +44,8 @@ /* Maximum buffers cached in cached buffer list */ #define FASTRPC_MAX_CACHED_BUFS (32) #define FASTRPC_MAX_CACHE_BUF_SIZE (8*1024*1024) +/* Max no. of persistent headers pre-allocated per user process */ +#define FASTRPC_MAX_PERSISTENT_HEADERS (25) /* Add memory to static PD pool, protection thru XPU */ #define ADSP_MMAP_HEAP_ADDR 4 @@ -54,7 +56,9 @@ /* Add memory to userPD pool, for user heap */ #define ADSP_MMAP_ADD_PAGES 0x1000 /* Add memory to userPD pool, for LLC heap */ -#define ADSP_MMAP_ADD_PAGES_LLC 0x3000, +#define ADSP_MMAP_ADD_PAGES_LLC 0x3000 +/* Map persistent header buffer on DSP */ +#define ADSP_MMAP_PERSIST_HDR 0x4000 #define DSP_UNSUPPORTED_API (0x80000414) /* MAX NUMBER of DSP ATTRIBUTES SUPPORTED */ @@ -208,6 +212,7 @@ struct fastrpc_buf { /* mmap support */ struct list_head node; /* list of user requested mmaps */ uintptr_t raddr; + bool in_use; }; struct fastrpc_dma_buf_attachment { @@ -303,11 +308,17 @@ struct fastrpc_user { struct fastrpc_channel_ctx *cctx; struct fastrpc_session_ctx *sctx; struct fastrpc_buf *init_mem; + /* Pre-allocated header buffer */ + struct fastrpc_buf *pers_hdr_buf; + /* Pre-allocated buffer divided into N chunks */ + struct fastrpc_buf *hdr_bufs; int tgid; int pd; /* total cached buffers */ u32 num_cached_buf; + /* total persistent headers */ + u32 num_pers_hdrs; bool is_secure_dev; /* Lock for lists */ spinlock_t lock; @@ -399,6 +410,37 @@ static int fastrpc_map_lookup(struct fastrpc_user *fl, int fd, return ret; } +static bool fastrpc_get_persistent_buf(struct fastrpc_user *fl, + size_t size, int buf_type, struct fastrpc_buf **obuf) +{ + u32 i = 0; + bool found = false; + struct fastrpc_buf *buf = NULL; + + spin_lock(&fl->lock); + /* + * Persistent header buffer can be used only if + * metadata length is less than 1 page size. + */ + if (!fl->num_pers_hdrs || buf_type != METADATA_BUF || size > PAGE_SIZE) { + spin_unlock(&fl->lock); + return found; + } + + for (i = 0; i < fl->num_pers_hdrs; i++) { + buf = &fl->hdr_bufs[i]; + /* If buffer not in use, then assign it for requested alloc */ + if (!buf->in_use) { + buf->in_use = true; + *obuf = buf; + found = true; + break; + } + } + spin_unlock(&fl->lock); + return found; +} + static void __fastrpc_buf_free(struct fastrpc_buf *buf) { dma_free_coherent(buf->dev, buf->size, buf->virt, @@ -430,6 +472,15 @@ static void fastrpc_cached_buf_list_add(struct fastrpc_buf *buf) static void fastrpc_buf_free(struct fastrpc_buf *buf, bool cache) { + struct fastrpc_user *fl = buf->fl; + + if (buf->in_use) { + /* Don't free persistent header buf. Just mark as available */ + spin_lock(&fl->lock); + buf->in_use = false; + spin_unlock(&fl->lock); + return; + } if (cache) fastrpc_cached_buf_list_add(buf); else @@ -523,6 +574,8 @@ static int fastrpc_buf_alloc(struct fastrpc_user *fl, struct device *dev, int ret; struct fastrpc_buf *buf; + if (fastrpc_get_persistent_buf(fl, size, buf_type, obuf)) + return 0; if (fastrpc_get_cached_buf(fl, size, buf_type, obuf)) return 0; ret = __fastrpc_buf_alloc(fl, dev, size, obuf); @@ -1305,6 +1358,107 @@ static int fastrpc_internal_invoke(struct fastrpc_user *fl, u32 kernel, return err; } +static int fastrpc_mem_map_to_dsp(struct fastrpc_user *fl, int fd, int offset, + u32 flags, u32 va, u64 phys, + size_t size, uintptr_t *raddr) +{ + struct fastrpc_invoke_args args[4] = { [0 ... 3] = { 0 } }; + struct fastrpc_mem_map_req_msg req_msg = { 0 }; + struct fastrpc_mmap_rsp_msg rsp_msg = { 0 }; + struct fastrpc_phy_page pages = { 0 }; + struct device *dev = fl->sctx->dev; + int err = 0; + u32 sc; + + req_msg.pgid = fl->tgid; + req_msg.fd = fd; + req_msg.offset = offset; + req_msg.vaddrin = va; + req_msg.flags = flags; + req_msg.num = sizeof(pages); + req_msg.data_len = 0; + + args[0].ptr = (u64) (uintptr_t) &req_msg; + args[0].length = sizeof(req_msg); + + pages.addr = phys; + pages.size = size; + + args[1].ptr = (u64) (uintptr_t) &pages; + args[1].length = sizeof(pages); + + args[2].ptr = (u64) (uintptr_t) &pages; + args[2].length = 0; + + args[3].ptr = (u64) (uintptr_t) &rsp_msg; + args[3].length = sizeof(rsp_msg); + + sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MEM_MAP, 3, 1); + err = fastrpc_internal_invoke(fl, true, FASTRPC_INIT_HANDLE, sc, &args[0]); + if (err) { + dev_err(dev, "mem mmap error, fd %d, vaddr %llx, size %lld\n", + fd, va, size); + return err; + } + *raddr = rsp_msg.vaddr; + + return 0; +} + +static int fastrpc_create_persistent_headers(struct fastrpc_user *fl) +{ + int err = 0; + int i = 0; + u64 virtb = 0; + struct device *dev = fl->sctx->dev; + struct fastrpc_buf *hdr_bufs, *buf, *pers_hdr_buf = NULL; + u32 num_pers_hdrs = 0; + size_t hdr_buf_alloc_len = 0; + + /* + * Pre-allocate memory for persistent header buffers based + * on concurrency info passed by user. Upper limit enforced. + */ + num_pers_hdrs = FASTRPC_MAX_PERSISTENT_HEADERS; + hdr_buf_alloc_len = num_pers_hdrs * PAGE_SIZE; + err = fastrpc_buf_alloc(fl, dev, hdr_buf_alloc_len, + METADATA_BUF, &pers_hdr_buf); + if (err) + return err; + + virtb = (u64) (uintptr_t)(pers_hdr_buf->virt); + err = fastrpc_mem_map_to_dsp(fl, -1, 0, + ADSP_MMAP_PERSIST_HDR, 0, (u64) (uintptr_t)(pers_hdr_buf->phys), + pers_hdr_buf->size, &pers_hdr_buf->raddr); + if (err) + goto err_dsp_map; + + hdr_bufs = kcalloc(num_pers_hdrs, sizeof(struct fastrpc_buf), + GFP_KERNEL); + if (!hdr_bufs) + return -ENOMEM; + + spin_lock(&fl->lock); + fl->pers_hdr_buf = pers_hdr_buf; + fl->num_pers_hdrs = num_pers_hdrs; + fl->hdr_bufs = hdr_bufs; + for (i = 0; i < num_pers_hdrs; i++) { + buf = &fl->hdr_bufs[i]; + buf->fl = fl; + buf->virt = (void *)(virtb + (i * PAGE_SIZE)); + buf->phys = pers_hdr_buf->phys + (i * PAGE_SIZE); + buf->size = PAGE_SIZE; + buf->type = pers_hdr_buf->type; + buf->in_use = false; + } + spin_unlock(&fl->lock); + + return 0; +err_dsp_map: + fastrpc_buf_free(pers_hdr_buf, 0); + return err; +} + static bool is_session_rejected(struct fastrpc_user *fl, bool unsigned_pd_request) { /* Check if the device node is non-secure and channel is secure*/ @@ -1537,6 +1691,12 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl, if (err) goto err_invoke; + if (fl->cctx->domain_id == CDSP_DOMAIN_ID) { + err = fastrpc_create_persistent_headers(fl); + if (err) + goto err_invoke; + } + kfree(args); return 0; @@ -1629,6 +1789,10 @@ static int fastrpc_device_release(struct inode *inode, struct file *file) fastrpc_buf_free(buf, false); } + if (fl->pers_hdr_buf) + fastrpc_buf_free(fl->pers_hdr_buf, false); + kfree(fl->hdr_bufs); + fastrpc_cached_buf_list_free(fl); fastrpc_session_free(cctx, fl->sctx); fastrpc_channel_ctx_put(cctx); @@ -2089,16 +2253,11 @@ static int fastrpc_req_mem_unmap(struct fastrpc_user *fl, char __user *argp) static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp) { - struct fastrpc_invoke_args args[4] = { [0 ... 3] = { 0 } }; - struct fastrpc_mem_map_req_msg req_msg = { 0 }; - struct fastrpc_mmap_rsp_msg rsp_msg = { 0 }; struct fastrpc_mem_unmap req_unmap = { 0 }; - struct fastrpc_phy_page pages = { 0 }; - struct fastrpc_mem_map req; + struct fastrpc_mem_map req = {0}; struct device *dev = fl->sctx->dev; struct fastrpc_map *map = NULL; int err; - u32 sc; if (copy_from_user(&req, argp, sizeof(req))) return -EFAULT; @@ -2110,57 +2269,29 @@ static int fastrpc_req_mem_map(struct fastrpc_user *fl, char __user *argp) return err; } - req_msg.pgid = fl->tgid; - req_msg.fd = req.fd; - req_msg.offset = req.offset; - req_msg.vaddrin = req.vaddrin; map->va = (void *) (uintptr_t) req.vaddrin; - req_msg.flags = req.flags; - req_msg.num = sizeof(pages); - req_msg.data_len = 0; - - args[0].ptr = (u64) (uintptr_t) &req_msg; - args[0].length = sizeof(req_msg); - - pages.addr = map->phys; - pages.size = map->size; - - args[1].ptr = (u64) (uintptr_t) &pages; - args[1].length = sizeof(pages); - - args[2].ptr = (u64) (uintptr_t) &pages; - args[2].length = 0; - - args[3].ptr = (u64) (uintptr_t) &rsp_msg; - args[3].length = sizeof(rsp_msg); - - sc = FASTRPC_SCALARS(FASTRPC_RMID_INIT_MEM_MAP, 3, 1); - err = fastrpc_internal_invoke(fl, true, FASTRPC_INIT_HANDLE, sc, &args[0]); + /* map to dsp, get virtual adrress for the user*/ + err = fastrpc_mem_map_to_dsp(fl, map->fd, req.offset, + req.flags, req.vaddrin, map->phys, + map->size, (uintptr_t *)&req.vaddrout); if (err) { - dev_err(dev, "mem mmap error, fd %d, vaddr %llx, size %lld\n", - req.fd, req.vaddrin, map->size); + dev_err(dev, "failed to map buffer on dsp, fd = %d\n", map->fd); goto err_invoke; } /* update the buffer to be able to deallocate the memory on the DSP */ - map->raddr = rsp_msg.vaddr; - - /* let the client know the address to use */ - req.vaddrout = rsp_msg.vaddr; + map->raddr = req.vaddrout; if (copy_to_user((void __user *)argp, &req, sizeof(req))) { /* unmap the memory and release the buffer */ - req_unmap.vaddr = (uintptr_t) rsp_msg.vaddr; + req_unmap.vaddr = (uintptr_t)req.vaddrout; req_unmap.length = map->size; fastrpc_req_mem_unmap_impl(fl, &req_unmap); return -EFAULT; } - return 0; - err_invoke: fastrpc_map_put(map); - return err; }