From mboxrd@z Thu Jan 1 00:00:00 1970 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=-9.9 required=3.0 tests=DKIM_SIGNED,DKIM_VALID, DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_PATCH,MAILING_LIST_MULTI, SIGNED_OFF_BY,SPF_HELO_NONE,SPF_PASS,USER_AGENT_GIT 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 B8093C47257 for ; Mon, 4 May 2020 09:26:51 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 9658321582 for ; Mon, 4 May 2020 09:26:51 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=ideasonboard.com header.i=@ideasonboard.com header.b="BZrUWm0I" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1728434AbgEDJ0v (ORCPT ); Mon, 4 May 2020 05:26:51 -0400 Received: from perceval.ideasonboard.com ([213.167.242.64]:56776 "EHLO perceval.ideasonboard.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1728422AbgEDJ0u (ORCPT ); Mon, 4 May 2020 05:26:50 -0400 Received: from pendragon.bb.dnainternet.fi (81-175-216-236.bb.dnainternet.fi [81.175.216.236]) by perceval.ideasonboard.com (Postfix) with ESMTPSA id 1F40D4F7; Mon, 4 May 2020 11:26:46 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=ideasonboard.com; s=mail; t=1588584406; bh=h1f4LLJwwL0fBSoFl0xi5O2F/8Lszy9bphrat8ky49o=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=BZrUWm0IeMGoJ1sKI5Z9OTvAHnF8bWGP6a2HY2vfYisU3rr1HpCXX0U4dxrEx2Dux e9ljkZOGexON6HKVREzNFucVYCUfX01voAM/EOCzbpAeDOp11bpADdHokXFJB8UBqM lR2TvxSZ+BYdY8x6BuMKJfqM9Svwxs79wnvZow5s= From: Laurent Pinchart To: linux-media@vger.kernel.org Cc: Kieran Bingham , Jacopo Mondi , =?UTF-8?q?Niklas=20S=C3=B6derlund?= , Naushir Patuck , Dave Stevenson , Oliver Gjoneski Subject: [PATCH v2 28/34] staging: vchiq_2835_arm: Implement a DMA pool for small bulk transfers Date: Mon, 4 May 2020 12:26:05 +0300 Message-Id: <20200504092611.9798-29-laurent.pinchart@ideasonboard.com> X-Mailer: git-send-email 2.26.2 In-Reply-To: <20200504092611.9798-1-laurent.pinchart@ideasonboard.com> References: <20200504092611.9798-1-laurent.pinchart@ideasonboard.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org From: Oliver Gjoneski During a bulk transfer we request a DMA allocation to hold the scatter-gather list. Most of the time, this allocation is small (<< PAGE_SIZE), however it can be requested at a high enough frequency to cause fragmentation and/or stress the CMA allocator (think time spent in compaction here, or during allocations elsewhere). Implement a pool to serve up small DMA allocations, falling back to a coherent allocation if the request is greater than VCHIQ_DMA_POOL_SIZE. Signed-off-by: Oliver Gjoneski Signed-off-by: Jacopo Mondi --- .../interface/vchiq_arm/vchiq_2835_arm.c | 38 ++++++++++++++++--- 1 file changed, 33 insertions(+), 5 deletions(-) diff --git a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c index c18c6ca0b6c0..3e422a7eb3f1 100644 --- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c +++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_2835_arm.c @@ -7,6 +7,7 @@ #include #include #include +#include #include #include #include @@ -28,6 +29,8 @@ #define BELL0 0x00 #define BELL2 0x08 +#define VCHIQ_DMA_POOL_SIZE PAGE_SIZE + struct vchiq_2835_state { int inited; struct vchiq_arm_state arm_state; @@ -37,6 +40,7 @@ struct vchiq_pagelist_info { struct pagelist *pagelist; size_t pagelist_buffer_size; dma_addr_t dma_addr; + bool is_from_pool; enum dma_data_direction dma_dir; unsigned int num_pages; unsigned int pages_need_release; @@ -57,6 +61,7 @@ static void __iomem *g_regs; * of 32. */ static unsigned int g_cache_line_size = 32; +static struct dma_pool *g_dma_pool; static unsigned int g_fragments_size; static char *g_fragments_base; static char *g_free_fragments; @@ -159,6 +164,14 @@ int vchiq_platform_init(struct platform_device *pdev, struct vchiq_state *state) } g_dev = dev; + g_dma_pool = dmam_pool_create("vchiq_scatter_pool", dev, + VCHIQ_DMA_POOL_SIZE, g_cache_line_size, + 0); + if (!g_dma_pool) { + dev_err(dev, "failed to create dma pool"); + return -ENOMEM; + } + vchiq_log_info(vchiq_arm_log_level, "vchiq_init - done (slots %pK, phys %pad)", vchiq_slot_zero, &slot_phys); @@ -293,9 +306,14 @@ cleanup_pagelistinfo(struct vchiq_pagelist_info *pagelistinfo) for (i = 0; i < pagelistinfo->num_pages; i++) put_page(pagelistinfo->pages[i]); } - - dma_free_coherent(g_dev, pagelistinfo->pagelist_buffer_size, - pagelistinfo->pagelist, pagelistinfo->dma_addr); + if (pagelistinfo->is_from_pool) { + dma_pool_free(g_dma_pool, pagelistinfo->pagelist, + pagelistinfo->dma_addr); + } else { + dma_free_coherent(g_dev, pagelistinfo->pagelist_buffer_size, + pagelistinfo->pagelist, + pagelistinfo->dma_addr); + } } /* There is a potential problem with partial cache lines (pages?) @@ -315,6 +333,7 @@ create_pagelist(char __user *buf, size_t count, unsigned short type) u32 *addrs; unsigned int num_pages, offset, i, k; int actual_pages; + bool is_from_pool; size_t pagelist_size; struct scatterlist *scatterlist, *sg; int dma_buffers; @@ -341,8 +360,16 @@ create_pagelist(char __user *buf, size_t count, unsigned short type) /* Allocate enough storage to hold the page pointers and the page * list */ - pagelist = dma_alloc_coherent(g_dev, pagelist_size, &dma_addr, - GFP_KERNEL); + if (pagelist_size > VCHIQ_DMA_POOL_SIZE) { + pagelist = dma_alloc_coherent(g_dev, + pagelist_size, + &dma_addr, + GFP_KERNEL); + is_from_pool = false; + } else { + pagelist = dma_pool_alloc(g_dma_pool, GFP_KERNEL, &dma_addr); + is_from_pool = true; + } vchiq_log_trace(vchiq_arm_log_level, "%s - %pK", __func__, pagelist); @@ -363,6 +390,7 @@ create_pagelist(char __user *buf, size_t count, unsigned short type) pagelistinfo->pagelist = pagelist; pagelistinfo->pagelist_buffer_size = pagelist_size; pagelistinfo->dma_addr = dma_addr; + pagelistinfo->is_from_pool = is_from_pool; pagelistinfo->dma_dir = (type == PAGELIST_WRITE) ? DMA_TO_DEVICE : DMA_FROM_DEVICE; pagelistinfo->num_pages = num_pages; -- Regards, Laurent Pinchart