From patchwork Tue Jul 23 10:50:03 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Guennadi Liakhovetski X-Patchwork-Id: 2831933 Return-Path: X-Original-To: patchwork-linux-sh@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork1.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.19.201]) by patchwork1.web.kernel.org (Postfix) with ESMTP id B795B9F243 for ; Tue, 23 Jul 2013 10:54:42 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 6D26F201D5 for ; Tue, 23 Jul 2013 10:54:41 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 01D0D201CE for ; Tue, 23 Jul 2013 10:54:40 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756713Ab3GWKuw (ORCPT ); Tue, 23 Jul 2013 06:50:52 -0400 Received: from moutng.kundenserver.de ([212.227.126.186]:51339 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756667Ab3GWKuX (ORCPT ); Tue, 23 Jul 2013 06:50:23 -0400 Received: from axis700.grange (dslb-094-220-153-077.pools.arcor-ip.net [94.220.153.77]) by mrelayeu.kundenserver.de (node=mreu2) with ESMTP (Nemesis) id 0LbNoU-1UGzny06Is-00lCYy; Tue, 23 Jul 2013 12:50:12 +0200 Received: from 6a.grange (6a.grange [192.168.1.11]) by axis700.grange (Postfix) with ESMTPS id CF76240BC5; Tue, 23 Jul 2013 12:50:10 +0200 (CEST) Received: from lyakh by 6a.grange with local (Exim 4.72) (envelope-from ) id 1V1aAU-0007EP-Ge; Tue, 23 Jul 2013 12:50:10 +0200 From: Guennadi Liakhovetski To: linux-kernel@vger.kernel.org Cc: Magnus Damm , Simon Horman , Laurent Pinchart , Vinod Koul , linux-sh@vger.kernel.org, Sergei Shtylyov , Guennadi Liakhovetski Subject: [PATCH/RFC v4 09/15] DMA: shdma: support referencing specific DMACs within a multiplexer in DT Date: Tue, 23 Jul 2013 12:50:03 +0200 Message-Id: <1374576609-27748-10-git-send-email-g.liakhovetski@gmx.de> X-Mailer: git-send-email 1.7.2.5 In-Reply-To: <1374576609-27748-1-git-send-email-g.liakhovetski@gmx.de> References: <1374576609-27748-1-git-send-email-g.liakhovetski@gmx.de> X-Provags-ID: V02:K0:TnxXA+7BTVlpimVWIzgNDJVFXujiEAaQMGLbxcMlPFQ EvNhTv2uR9KE3IA2QYc2/MaDs4UXOJj1k4/KB/Mnu0A+WJvRBr PnIxepPUjyQqifQnhRdfaSrYF3u9dDslws21DIwFtO6CLoOjrc vN7RUY3P54sLqN3XESKj26gI1aiP+Fv2lxMIHK1c4wayyNdYtU EAN9Qmp9UJNiFO/aFNJ/Nua2hgdLa953GEX1qcZnx2qhOFM6AR q58AzXpJhXVqrweemCh/6o40IH8eSkqfPfny19Z86Z9rYzivYi dNqjeXGYqX+diBxHZp4sfLoqIUeXNMv7BU8EWvz0CRxBKQv9Kf lyecpWgr8i9vSBFX9pQveN7GkYPnZFBKEHf+OvKx04vlXmZksW ONryz3Hlj8ZU2Rdzck65kh/jNHArJTZxZ8= Sender: linux-sh-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-sh@vger.kernel.org X-Spam-Status: No, score=-6.9 required=5.0 tests=BAYES_00,FREEMAIL_FROM, RCVD_IN_DNSWL_HI, RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=unavailable version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Currently shdma DT nodes have to be placed under a multiplexer node. DMA slave DT nodes then use that multiplexer's phandle in their "dmas" properties. However, sometimes it can be necessary to let DMA slaves only use a specific DMAC instance. In this case it would be logical to just use the respective phandle in that slave's "dmas" property. For this to work the referenced DMAC has to register a struct of_dma instance, which isn't presently done. Instead the driver currently only registers one struct of_dma for the multiplexer. This patch adds support for such configurations. To enable this option a "#dma-cells" property also must be added to the respective DMAC DT node. Signed-off-by: Guennadi Liakhovetski --- As mentioned in the cover letter patch, this is an RFC. We are certain, that we want to be able to reference specific DMACs within multiplexers. However, with the current implementation this cannot work, because single DMAC DT nodes don't have of_dma objects associated with them, currently there is only one such object per multiplexer. The proposal is to implement such direct referencing by adding a "#dma-cells" property to respective DT nodes, in which case an of_dma object will be allocated for them too. Documentation/devicetree/bindings/dma/shdma.txt | 16 ++++++ drivers/dma/sh/shdma.h | 7 +++ drivers/dma/sh/shdmac.c | 66 +++++++++++++++++++++++ 3 files changed, 89 insertions(+), 0 deletions(-) diff --git a/Documentation/devicetree/bindings/dma/shdma.txt b/Documentation/devicetree/bindings/dma/shdma.txt index 6b015d6..aeed9b8 100644 --- a/Documentation/devicetree/bindings/dma/shdma.txt +++ b/Documentation/devicetree/bindings/dma/shdma.txt @@ -27,6 +27,22 @@ Required properties: "renesas,shdma-r8a7740" for the DMACs (not RTDMAC) on r8a7740 "renesas,shdma-sh73a0" for the DMACs on sh73a0 +Optional properties: +- #dma-cells: this property is only needed in one specific case: if DMA slaves + have to be able to request channels specifically from this DMAC, + not from anyone from the multiplexer. In such a case the board + .dts file can contain code, similar to this: + +&dma1 { + #dma-cells = <1>; +}; + +&mmc0 { + dmas = <&dma1 0xd1 + &dma1 0xd2>; + dma-names = "tx", "rx"; +}; + Example: dmac: dma-mux0 { compatible = "renesas,shdma-mux"; diff --git a/drivers/dma/sh/shdma.h b/drivers/dma/sh/shdma.h index 8394424..991316f 100644 --- a/drivers/dma/sh/shdma.h +++ b/drivers/dma/sh/shdma.h @@ -23,6 +23,7 @@ #define SH_DMAE_TCR_MAX 0x00FFFFFF /* 16MB */ struct device; +struct device_node; struct sh_dmae_chan { struct shdma_chan shdma_chan; @@ -33,6 +34,11 @@ struct sh_dmae_chan { int pm_error; }; +struct sh_dmae_filter_info { + u32 hw_req; + struct device_node *of_node; +}; + struct sh_dmae_device { struct shdma_dev shdma_dev; struct sh_dmae_chan *chan[SH_DMAE_MAX_CHANNELS]; @@ -42,6 +48,7 @@ struct sh_dmae_device { void __iomem *dmars; unsigned int chcr_offset; u32 chcr_ie_bit; + struct sh_dmae_filter_info filter_info; }; struct sh_dmae_regs { diff --git a/drivers/dma/sh/shdmac.c b/drivers/dma/sh/shdmac.c index 9ee3c28..fcaed8d 100644 --- a/drivers/dma/sh/shdmac.c +++ b/drivers/dma/sh/shdmac.c @@ -22,6 +22,7 @@ #include #include #include +#include #include #include #include @@ -665,6 +666,63 @@ static const struct shdma_ops sh_dmae_shdma_ops = { .get_partial = sh_dmae_get_partial, }; +static bool sh_dmae_chan_filter(struct dma_chan *chan, void *arg) +{ + struct sh_dmae_filter_info *info = arg; + struct shdma_chan *schan = to_shdma_chan(chan); + int match = info->hw_req; + + if (match < 0) + /* No slave requested - arbitrary channel */ + return true; + + dev_dbg(schan->dev, "%s(): trying %s for 0x%x\n", __func__, + info->of_node->full_name, match); + + if (schan->dev->of_node != info->of_node) + return false; + + return !sh_dmae_set_slave(schan, match, true); +} + +static struct dma_chan *sh_dmae_of_xlate(struct of_phandle_args *dma_spec, + struct of_dma *ofdma) +{ + struct sh_dmae_filter_info *info = ofdma->of_dma_data; + u32 id = dma_spec->args[0]; + dma_cap_mask_t mask; + struct dma_chan *chan; + + if (dma_spec->args_count != 1) + return NULL; + + dma_cap_zero(mask); + /* Only slave DMA channels can be allocated via DT */ + dma_cap_set(DMA_SLAVE, mask); + + info->hw_req = id; + info->of_node = dma_spec->np; + + chan = dma_request_channel(mask, sh_dmae_chan_filter, info); + if (chan) + to_shdma_chan(chan)->hw_req = id; + + return chan; +} + +static int sh_dmae_of_add(struct device *dev, struct sh_dmae_device *shdev) +{ + u32 cells; + int ret = of_property_read_u32(dev->of_node, "#dma-cells", &cells); + + dev_dbg(dev, "%s(): %u (%d)\n", __func__, cells, ret); + if (ret < 0 || !cells) + return 0; + + return of_dma_controller_register(dev->of_node, + sh_dmae_of_xlate, &shdev->filter_info); +} + static const struct of_device_id sh_dmae_of_match[] = { {.compatible = "renesas,shdma-r8a73a4", .data = r8a73a4_shdma_devid,}, {.compatible = "renesas,shdma-r8a7740", .data = r8a7740_shdma_devid,}, @@ -845,6 +903,10 @@ static int sh_dmae_probe(struct platform_device *pdev) } while (irq_cnt < pdata->channel_num && chanirq_res); } + err = sh_dmae_of_add(&pdev->dev, shdev); + if (err < 0) + goto of_add_err; + /* Create DMA Channel */ for (i = 0; i < irq_cnt; i++) { err = sh_dmae_chan_probe(shdev, i, chan_irq[i], chan_flag[i]); @@ -869,6 +931,8 @@ edmadevreg: pm_runtime_get(&pdev->dev); chan_probe_err: + of_dma_controller_free(pdev->dev.of_node); +of_add_err: sh_dmae_chan_remove(shdev); #if defined(CONFIG_CPU_SH4) || defined(CONFIG_ARCH_SHMOBILE) @@ -895,6 +959,8 @@ static int sh_dmae_remove(struct platform_device *pdev) struct sh_dmae_device *shdev = platform_get_drvdata(pdev); struct dma_device *dma_dev = &shdev->shdma_dev.dma_dev; + of_dma_controller_free(pdev->dev.of_node); + dma_async_device_unregister(dma_dev); spin_lock_irq(&sh_dmae_lock);