Received: by 2002:a25:4158:0:0:0:0:0 with SMTP id o85csp3606620yba; Tue, 9 Apr 2019 00:31:49 -0700 (PDT) X-Google-Smtp-Source: APXvYqwf5iiFeqNZAI/crA9TlFez402yIN/xlOju1XXYXOEfINRA0XCS4waO71jgdZVp23S3GnzP X-Received: by 2002:a17:902:6b05:: with SMTP id o5mr35312773plk.90.1554795108919; Tue, 09 Apr 2019 00:31:48 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1554795108; cv=none; d=google.com; s=arc-20160816; b=sNT3cotBscTK6LtVr8D/LG9ks6dG4Rfiw4noKaA/N40KLgIBIQlHYwj4IJgdu4R4TW tqmO72naYXHVqoCZmb2d4688Xe0s6HF/5OcnBVIeneL/RmACEHd6Y1Xj/MG0GYm0wqAe zA/Kt8xBnn+5F3u4PmiCdvf4fFbUTltfY5qypq6RLaX3WwNYU3CG7uvOGFkmbZVWYmAh sKS+l27wUvkziudlrLWum7BMPctDvymA0fCTn7J366RtRxl7UAcin9GrHSpbOWYpRhV9 o1AGqR2XGfN3zkLYiGusQwcT/nC1P2OCIDvEOIo/5ckiHvXyaIq5w2Xk9pwr6iCZT3R8 HzKA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:references:in-reply-to:message-id:date :subject:cc:to:from; bh=yP2kGsXmWHbY+KUuTIgjb1eOTX4mAh8Xk3eXpv5bz50=; b=iJTr0ZGLnbBN0sXWazpQwikdZ46xtMxxbNcicwb1tFOM1I9W1/qJX8TbuDYqty84Nn 5700Sev9Y3hnEa8nGj346RD19u3lctEhlUZXVh7RlFc8RVoxHi5BusxFSAAuTxtze0Sk 5DPOtBCFq7EyEDoT0zTxO3ivcKbgg1CiT8ffu+22g9VNkFy6KF4UIBHvoAUjUkPQYLtT Ba7M6hzU2RTf50ggANr/Dn2idgeYrJE1TPtscPx7U85jdgBKO+N5/5KvDBuAMnAPqqjI Ym3/rDy/UF5Pk3BMDjz6fC0aPpaW8+m2kBx9u4DHeZLKnW+N8iQM6OjXERqgD6LRnWes XmQw== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=nxp.com Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id 3si27562040pgs.42.2019.04.09.00.31.33; Tue, 09 Apr 2019 00:31:48 -0700 (PDT) Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=nxp.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726686AbfDIH3O (ORCPT + 99 others); Tue, 9 Apr 2019 03:29:14 -0400 Received: from inva020.nxp.com ([92.121.34.13]:46514 "EHLO inva020.nxp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726464AbfDIH3O (ORCPT ); Tue, 9 Apr 2019 03:29:14 -0400 Received: from inva020.nxp.com (localhost [127.0.0.1]) by inva020.eu-rdc02.nxp.com (Postfix) with ESMTP id A943E1A018E; Tue, 9 Apr 2019 09:29:06 +0200 (CEST) Received: from invc005.ap-rdc01.nxp.com (invc005.ap-rdc01.nxp.com [165.114.16.14]) by inva020.eu-rdc02.nxp.com (Postfix) with ESMTP id 4E45A1A00DC; Tue, 9 Apr 2019 09:29:03 +0200 (CEST) Received: from titan.ap.freescale.net (TITAN.ap.freescale.net [10.192.208.233]) by invc005.ap-rdc01.nxp.com (Postfix) with ESMTP id 310D340302; Tue, 9 Apr 2019 15:28:59 +0800 (SGT) From: Peng Ma To: vkoul@kernel.org, dan.j.williams@intel.com, leoyang.li@nxp.com Cc: linux-kernel@vger.kernel.org, dmaengine@vger.kernel.org, Peng Ma Subject: [V3 2/2] dmaengine: fsl-dpaa2-qdma: Add NXP dpaa2 qDMA controller driver for Layerscape SoCs Date: Tue, 9 Apr 2019 15:22:12 +0800 Message-Id: <20190409072212.15860-2-peng.ma@nxp.com> X-Mailer: git-send-email 2.14.1 In-Reply-To: <20190409072212.15860-1-peng.ma@nxp.com> References: <20190409072212.15860-1-peng.ma@nxp.com> X-Virus-Scanned: ClamAV using ClamSMTP Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org DPPA2(Data Path Acceleration Architecture 2) qDMA The qDMA supports channel virtualization by allowing DMA jobs to be enqueued into different frame queues. Core can initiate a DMA transaction by preparing a frame descriptor(FD) for each DMA job and enqueuing this job to a frame queue. through a hardware portal. The qDMA prefetches DMA jobs from the frame queues. It then schedules and dispatches to internal DMA hardware engines, which generate read and write requests. Both qDMA source data and destination data can be either contiguous or non-contiguous using one or more scatter/gather tables. The qDMA supports global bandwidth flow control where all DMA transactions are stalled if the bandwidth threshold has been reached. Also supported are transaction based read throttling. Add NXP dppa2 qDMA to support some of Layerscape SoCs. such as: LS1088A, LS208xA, LX2, etc. Signed-off-by: Peng Ma --- changed for v3: - Add depends on arm64 for dpaa2 qdma driver - The dpaa2_io_service_[de]register functions have a new parameter So update all calls to some functions drivers/dma/Kconfig | 2 + drivers/dma/Makefile | 1 + drivers/dma/fsl-dpaa2-qdma/Kconfig | 9 + drivers/dma/fsl-dpaa2-qdma/Makefile | 3 + drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.c | 782 +++++++++++++++++++++++++++++++ drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.h | 152 ++++++ 6 files changed, 949 insertions(+), 0 deletions(-) create mode 100644 drivers/dma/fsl-dpaa2-qdma/Kconfig create mode 100644 drivers/dma/fsl-dpaa2-qdma/Makefile create mode 100644 drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.c create mode 100644 drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.h diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig index eaf78f4..08aae01 100644 --- a/drivers/dma/Kconfig +++ b/drivers/dma/Kconfig @@ -671,6 +671,8 @@ source "drivers/dma/sh/Kconfig" source "drivers/dma/ti/Kconfig" +source "drivers/dma/fsl-dpaa2-qdma/Kconfig" + # clients comment "DMA Clients" depends on DMA_ENGINE diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile index 6126e1c..2499ed8 100644 --- a/drivers/dma/Makefile +++ b/drivers/dma/Makefile @@ -75,6 +75,7 @@ obj-$(CONFIG_UNIPHIER_MDMAC) += uniphier-mdmac.o obj-$(CONFIG_XGENE_DMA) += xgene-dma.o obj-$(CONFIG_ZX_DMA) += zx_dma.o obj-$(CONFIG_ST_FDMA) += st_fdma.o +obj-$(CONFIG_FSL_DPAA2_QDMA) += fsl-dpaa2-qdma/ obj-y += mediatek/ obj-y += qcom/ diff --git a/drivers/dma/fsl-dpaa2-qdma/Kconfig b/drivers/dma/fsl-dpaa2-qdma/Kconfig new file mode 100644 index 0000000..258ed6b --- /dev/null +++ b/drivers/dma/fsl-dpaa2-qdma/Kconfig @@ -0,0 +1,9 @@ +menuconfig FSL_DPAA2_QDMA + tristate "NXP DPAA2 QDMA" + depends on ARM64 + depends on FSL_MC_BUS && FSL_MC_DPIO + select DMA_ENGINE + select DMA_VIRTUAL_CHANNELS + help + NXP Data Path Acceleration Architecture 2 QDMA driver, + using the NXP MC bus driver. diff --git a/drivers/dma/fsl-dpaa2-qdma/Makefile b/drivers/dma/fsl-dpaa2-qdma/Makefile new file mode 100644 index 0000000..c1d0226 --- /dev/null +++ b/drivers/dma/fsl-dpaa2-qdma/Makefile @@ -0,0 +1,3 @@ +# SPDX-License-Identifier: GPL-2.0 +# Makefile for the NXP DPAA2 qDMA controllers +obj-$(CONFIG_FSL_DPAA2_QDMA) += dpaa2-qdma.o dpdmai.o diff --git a/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.c b/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.c new file mode 100644 index 0000000..0cdde0f --- /dev/null +++ b/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.c @@ -0,0 +1,782 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright 2014-2018 NXP + +/* + * Author: Changming Huang + * + * Driver for the NXP QDMA engine with QMan mode. + * Channel virtualization is supported through enqueuing of DMA jobs to, + * or dequeuing DMA jobs from different work queues with QMan portal. + * This module can be found on NXP LS2 SoCs. + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../virt-dma.h" +#include "dpdmai_cmd.h" +#include "dpdmai.h" +#include "dpaa2-qdma.h" + +static bool smmu_disable = true; + +static struct dpaa2_qdma_chan *to_dpaa2_qdma_chan(struct dma_chan *chan) +{ + return container_of(chan, struct dpaa2_qdma_chan, vchan.chan); +} + +static struct dpaa2_qdma_comp *to_fsl_qdma_comp(struct virt_dma_desc *vd) +{ + return container_of(vd, struct dpaa2_qdma_comp, vdesc); +} + +static int dpaa2_qdma_alloc_chan_resources(struct dma_chan *chan) +{ + struct dpaa2_qdma_chan *dpaa2_chan = to_dpaa2_qdma_chan(chan); + struct dpaa2_qdma_engine *dpaa2_qdma = dpaa2_chan->qdma; + struct device *dev = &dpaa2_qdma->priv->dpdmai_dev->dev; + + dpaa2_chan->fd_pool = dma_pool_create("fd_pool", dev, + FD_POOL_SIZE, 32, 0); + if (!dpaa2_chan->fd_pool) + return -ENOMEM; + + return dpaa2_qdma->desc_allocated++; +} + +static void dpaa2_qdma_free_chan_resources(struct dma_chan *chan) +{ + struct dpaa2_qdma_chan *dpaa2_chan = to_dpaa2_qdma_chan(chan); + struct dpaa2_qdma_engine *dpaa2_qdma = dpaa2_chan->qdma; + unsigned long flags; + + LIST_HEAD(head); + + spin_lock_irqsave(&dpaa2_chan->vchan.lock, flags); + vchan_get_all_descriptors(&dpaa2_chan->vchan, &head); + spin_unlock_irqrestore(&dpaa2_chan->vchan.lock, flags); + + vchan_dma_desc_free_list(&dpaa2_chan->vchan, &head); + + dpaa2_dpdmai_free_comp(dpaa2_chan, &dpaa2_chan->comp_used); + dpaa2_dpdmai_free_comp(dpaa2_chan, &dpaa2_chan->comp_free); + + dma_pool_destroy(dpaa2_chan->fd_pool); + dpaa2_qdma->desc_allocated--; +} + +/* + * Request a command descriptor for enqueue. + */ +static struct dpaa2_qdma_comp * +dpaa2_qdma_request_desc(struct dpaa2_qdma_chan *dpaa2_chan) +{ + struct dpaa2_qdma_comp *comp_temp = NULL; + unsigned long flags; + + spin_lock_irqsave(&dpaa2_chan->queue_lock, flags); + if (list_empty(&dpaa2_chan->comp_free)) { + spin_unlock_irqrestore(&dpaa2_chan->queue_lock, flags); + comp_temp = kzalloc(sizeof(*comp_temp), GFP_KERNEL); + if (!comp_temp) + goto err; + comp_temp->fd_virt_addr = + dma_pool_alloc(dpaa2_chan->fd_pool, GFP_NOWAIT, + &comp_temp->fd_bus_addr); + if (!comp_temp->fd_virt_addr) + goto err; + + comp_temp->fl_virt_addr = + (void *)((struct dpaa2_fd *) + comp_temp->fd_virt_addr + 1); + comp_temp->fl_bus_addr = comp_temp->fd_bus_addr + + sizeof(struct dpaa2_fd); + comp_temp->desc_virt_addr = + (void *)((struct dpaa2_fl_entry *) + comp_temp->fl_virt_addr + 3); + comp_temp->desc_bus_addr = comp_temp->fl_bus_addr + + sizeof(struct dpaa2_fl_entry) * 3; + + comp_temp->qchan = dpaa2_chan; + return comp_temp; + } + comp_temp = list_first_entry(&dpaa2_chan->comp_free, + struct dpaa2_qdma_comp, list); + list_del(&comp_temp->list); + spin_unlock_irqrestore(&dpaa2_chan->queue_lock, flags); + + comp_temp->qchan = dpaa2_chan; +err: + return comp_temp; +} + +static void +dpaa2_qdma_populate_fd(u32 format, struct dpaa2_qdma_comp *dpaa2_comp) +{ + struct dpaa2_fd *fd; + + fd = (struct dpaa2_fd *)dpaa2_comp->fd_virt_addr; + memset(fd, 0, sizeof(struct dpaa2_fd)); + + /* fd populated */ + dpaa2_fd_set_addr(fd, dpaa2_comp->fl_bus_addr); + /* Bypass memory translation, Frame list format, short length disable */ + /* we need to disable BMT if fsl-mc use iova addr */ + if (smmu_disable) + dpaa2_fd_set_bpid(fd, QMAN_FD_BMT_ENABLE); + dpaa2_fd_set_format(fd, QMAN_FD_FMT_ENABLE | QMAN_FD_SL_DISABLE); + + dpaa2_fd_set_frc(fd, format | QDMA_SER_CTX); +} + +/* first frame list for descriptor buffer */ +static void +dpaa2_qdma_populate_first_framel(struct dpaa2_fl_entry *f_list, + struct dpaa2_qdma_comp *dpaa2_comp, + bool wrt_changed) +{ + struct dpaa2_qdma_sd_d *sdd; + + sdd = (struct dpaa2_qdma_sd_d *)dpaa2_comp->desc_virt_addr; + memset(sdd, 0, 2 * (sizeof(*sdd))); + + /* source descriptor CMD */ + sdd->cmd = cpu_to_le32(QDMA_SD_CMD_RDTTYPE_COHERENT); + sdd++; + + /* dest descriptor CMD */ + if (wrt_changed) + sdd->cmd = cpu_to_le32(LX2160_QDMA_DD_CMD_WRTTYPE_COHERENT); + else + sdd->cmd = cpu_to_le32(QDMA_DD_CMD_WRTTYPE_COHERENT); + + memset(f_list, 0, sizeof(struct dpaa2_fl_entry)); + + /* first frame list to source descriptor */ + dpaa2_fl_set_addr(f_list, dpaa2_comp->desc_bus_addr); + dpaa2_fl_set_len(f_list, 0x20); + dpaa2_fl_set_format(f_list, QDMA_FL_FMT_SBF | QDMA_FL_SL_LONG); + + /* bypass memory translation */ + if (smmu_disable) + f_list->bpid = cpu_to_le16(QDMA_FL_BMT_ENABLE); +} + +/* source and destination frame list */ +static void +dpaa2_qdma_populate_frames(struct dpaa2_fl_entry *f_list, + dma_addr_t dst, dma_addr_t src, + size_t len, uint8_t fmt) +{ + /* source frame list to source buffer */ + memset(f_list, 0, sizeof(struct dpaa2_fl_entry)); + + dpaa2_fl_set_addr(f_list, src); + dpaa2_fl_set_len(f_list, len); + + /* single buffer frame or scatter gather frame */ + dpaa2_fl_set_format(f_list, (fmt | QDMA_FL_SL_LONG)); + + /* bypass memory translation */ + if (smmu_disable) + f_list->bpid = cpu_to_le16(QDMA_FL_BMT_ENABLE); + + f_list++; + + /* destination frame list to destination buffer */ + memset(f_list, 0, sizeof(struct dpaa2_fl_entry)); + + dpaa2_fl_set_addr(f_list, dst); + dpaa2_fl_set_len(f_list, len); + dpaa2_fl_set_format(f_list, (fmt | QDMA_FL_SL_LONG)); + /* single buffer frame or scatter gather frame */ + dpaa2_fl_set_final(f_list, QDMA_FL_F); + /* bypass memory translation */ + if (smmu_disable) + f_list->bpid = cpu_to_le16(QDMA_FL_BMT_ENABLE); +} + +static struct dma_async_tx_descriptor +*dpaa2_qdma_prep_memcpy(struct dma_chan *chan, dma_addr_t dst, + dma_addr_t src, size_t len, ulong flags) +{ + struct dpaa2_qdma_chan *dpaa2_chan = to_dpaa2_qdma_chan(chan); + struct dpaa2_qdma_engine *dpaa2_qdma; + struct dpaa2_qdma_comp *dpaa2_comp; + struct dpaa2_fl_entry *f_list; + bool wrt_changed; + u32 format; + + dpaa2_qdma = dpaa2_chan->qdma; + dpaa2_comp = dpaa2_qdma_request_desc(dpaa2_chan); + wrt_changed = (bool)dpaa2_qdma->qdma_wrtype_fixup; + +#ifdef LONG_FORMAT + format = QDMA_FD_LONG_FORMAT; +#else + format = QDMA_FD_SHORT_FORMAT; +#endif + /* populate Frame descriptor */ + dpaa2_qdma_populate_fd(format, dpaa2_comp); + + f_list = (struct dpaa2_fl_entry *)dpaa2_comp->fl_virt_addr; + +#ifdef LONG_FORMAT + /* first frame list for descriptor buffer (logn format) */ + dpaa2_qdma_populate_first_framel(f_list, dpaa2_comp, wrt_changed); + + f_list++; +#endif + + dpaa2_qdma_populate_frames(f_list, dst, src, len, QDMA_FL_FMT_SBF); + + return vchan_tx_prep(&dpaa2_chan->vchan, &dpaa2_comp->vdesc, flags); +} + +static enum +dma_status dpaa2_qdma_tx_status(struct dma_chan *chan, + dma_cookie_t cookie, + struct dma_tx_state *txstate) +{ + return dma_cookie_status(chan, cookie, txstate); +} + +static void dpaa2_qdma_issue_pending(struct dma_chan *chan) +{ + struct dpaa2_qdma_chan *dpaa2_chan = to_dpaa2_qdma_chan(chan); + struct dpaa2_qdma_engine *dpaa2_qdma = dpaa2_chan->qdma; + struct dpaa2_qdma_priv *priv = dpaa2_qdma->priv; + struct dpaa2_qdma_comp *dpaa2_comp; + struct virt_dma_desc *vdesc; + struct dpaa2_fd *fd; + unsigned long flags; + int err; + + spin_lock_irqsave(&dpaa2_chan->queue_lock, flags); + spin_lock(&dpaa2_chan->vchan.lock); + if (vchan_issue_pending(&dpaa2_chan->vchan)) { + vdesc = vchan_next_desc(&dpaa2_chan->vchan); + if (!vdesc) + goto err_enqueue; + dpaa2_comp = to_fsl_qdma_comp(vdesc); + + fd = (struct dpaa2_fd *)dpaa2_comp->fd_virt_addr; + + list_del(&vdesc->node); + list_add_tail(&dpaa2_comp->list, &dpaa2_chan->comp_used); + + /* TOBO: priority hard-coded to zero */ + err = dpaa2_io_service_enqueue_fq(NULL, + priv->tx_queue_attr[0].fqid, fd); + if (err) { + list_del(&dpaa2_comp->list); + list_add_tail(&dpaa2_comp->list, + &dpaa2_chan->comp_free); + } + } +err_enqueue: + spin_unlock(&dpaa2_chan->vchan.lock); + spin_unlock_irqrestore(&dpaa2_chan->queue_lock, flags); +} + +static int __cold dpaa2_qdma_setup(struct fsl_mc_device *ls_dev) +{ + struct dpaa2_qdma_priv_per_prio *ppriv; + struct device *dev = &ls_dev->dev; + struct dpaa2_qdma_priv *priv; + u8 prio_def = DPDMAI_PRIO_NUM; + int err; + int i; + + priv = dev_get_drvdata(dev); + + priv->dev = dev; + priv->dpqdma_id = ls_dev->obj_desc.id; + + /*Get the handle for the DPDMAI this interface is associate with */ + err = dpdmai_open(priv->mc_io, 0, priv->dpqdma_id, &ls_dev->mc_handle); + if (err) { + dev_err(dev, "dpdmai_open() failed\n"); + return err; + } + dev_info(dev, "Opened dpdmai object successfully\n"); + + err = dpdmai_get_attributes(priv->mc_io, 0, ls_dev->mc_handle, + &priv->dpdmai_attr); + if (err) { + dev_err(dev, "dpdmai_get_attributes() failed\n"); + return err; + } + + if (priv->dpdmai_attr.version.major > DPDMAI_VER_MAJOR) { + dev_err(dev, "DPDMAI major version mismatch\n" + "Found %u.%u, supported version is %u.%u\n", + priv->dpdmai_attr.version.major, + priv->dpdmai_attr.version.minor, + DPDMAI_VER_MAJOR, DPDMAI_VER_MINOR); + } + + if (priv->dpdmai_attr.version.minor > DPDMAI_VER_MINOR) { + dev_err(dev, "DPDMAI minor version mismatch\n" + "Found %u.%u, supported version is %u.%u\n", + priv->dpdmai_attr.version.major, + priv->dpdmai_attr.version.minor, + DPDMAI_VER_MAJOR, DPDMAI_VER_MINOR); + } + + priv->num_pairs = min(priv->dpdmai_attr.num_of_priorities, prio_def); + ppriv = kcalloc(priv->num_pairs, sizeof(*ppriv), GFP_KERNEL); + if (!ppriv) { + dev_err(dev, "kzalloc for ppriv failed\n"); + return -1; + } + priv->ppriv = ppriv; + + for (i = 0; i < priv->num_pairs; i++) { + err = dpdmai_get_rx_queue(priv->mc_io, 0, ls_dev->mc_handle, + i, &priv->rx_queue_attr[i]); + if (err) { + dev_err(dev, "dpdmai_get_rx_queue() failed\n"); + return err; + } + ppriv->rsp_fqid = priv->rx_queue_attr[i].fqid; + + err = dpdmai_get_tx_queue(priv->mc_io, 0, ls_dev->mc_handle, + i, &priv->tx_queue_attr[i]); + if (err) { + dev_err(dev, "dpdmai_get_tx_queue() failed\n"); + return err; + } + ppriv->req_fqid = priv->tx_queue_attr[i].fqid; + ppriv->prio = i; + ppriv->priv = priv; + ppriv++; + } + + return 0; +} + +static void dpaa2_qdma_fqdan_cb(struct dpaa2_io_notification_ctx *ctx) +{ + struct dpaa2_qdma_priv_per_prio *ppriv = container_of(ctx, + struct dpaa2_qdma_priv_per_prio, nctx); + struct dpaa2_qdma_comp *dpaa2_comp, *_comp_tmp; + struct dpaa2_qdma_priv *priv = ppriv->priv; + u32 n_chans = priv->dpaa2_qdma->n_chans; + struct dpaa2_qdma_chan *qchan; + const struct dpaa2_fd *fd_eq; + const struct dpaa2_fd *fd; + struct dpaa2_dq *dq; + int is_last = 0; + int found; + u8 status; + int err; + int i; + + do { + err = dpaa2_io_service_pull_fq(NULL, ppriv->rsp_fqid, + ppriv->store); + } while (err); + + while (!is_last) { + do { + dq = dpaa2_io_store_next(ppriv->store, &is_last); + } while (!is_last && !dq); + if (!dq) { + dev_err(priv->dev, "FQID returned no valid frames!\n"); + continue; + } + + /* obtain FD and process the error */ + fd = dpaa2_dq_fd(dq); + + status = dpaa2_fd_get_ctrl(fd) & 0xff; + if (status) + dev_err(priv->dev, "FD error occurred\n"); + found = 0; + for (i = 0; i < n_chans; i++) { + qchan = &priv->dpaa2_qdma->chans[i]; + spin_lock(&qchan->queue_lock); + if (list_empty(&qchan->comp_used)) { + spin_unlock(&qchan->queue_lock); + continue; + } + list_for_each_entry_safe(dpaa2_comp, _comp_tmp, + &qchan->comp_used, list) { + fd_eq = (struct dpaa2_fd *) + dpaa2_comp->fd_virt_addr; + + if (le64_to_cpu(fd_eq->simple.addr) == + le64_to_cpu(fd->simple.addr)) { + spin_lock(&qchan->vchan.lock); + vchan_cookie_complete(& + dpaa2_comp->vdesc); + spin_unlock(&qchan->vchan.lock); + found = 1; + break; + } + } + spin_unlock(&qchan->queue_lock); + if (found) + break; + } + } + + dpaa2_io_service_rearm(NULL, ctx); +} + +static int __cold dpaa2_qdma_dpio_setup(struct dpaa2_qdma_priv *priv) +{ + struct dpaa2_qdma_priv_per_prio *ppriv; + struct device *dev = priv->dev; + int err, i, num; + + num = priv->num_pairs; + ppriv = priv->ppriv; + for (i = 0; i < num; i++) { + ppriv->nctx.is_cdan = 0; + ppriv->nctx.desired_cpu = 1; + ppriv->nctx.id = ppriv->rsp_fqid; + ppriv->nctx.cb = dpaa2_qdma_fqdan_cb; + err = dpaa2_io_service_register(NULL, &ppriv->nctx, dev); + if (err) { + dev_err(dev, "Notification register failed\n"); + goto err_service; + } + + ppriv->store = + dpaa2_io_store_create(DPAA2_QDMA_STORE_SIZE, dev); + if (!ppriv->store) { + dev_err(dev, "dpaa2_io_store_create() failed\n"); + goto err_store; + } + + ppriv++; + } + return 0; + +err_store: + dpaa2_io_service_deregister(NULL, &ppriv->nctx, dev); +err_service: + ppriv--; + while (ppriv >= priv->ppriv) { + dpaa2_io_service_deregister(NULL, &ppriv->nctx, dev); + dpaa2_io_store_destroy(ppriv->store); + ppriv--; + } + return -1; +} + +static void dpaa2_dpmai_store_free(struct dpaa2_qdma_priv *priv) +{ + struct dpaa2_qdma_priv_per_prio *ppriv = priv->ppriv; + int i; + + for (i = 0; i < priv->num_pairs; i++) { + dpaa2_io_store_destroy(ppriv->store); + ppriv++; + } +} + +static void dpaa2_dpdmai_dpio_free(struct dpaa2_qdma_priv *priv) +{ + struct dpaa2_qdma_priv_per_prio *ppriv = priv->ppriv; + struct device *dev = priv->dev; + int i; + + for (i = 0; i < priv->num_pairs; i++) { + dpaa2_io_service_deregister(NULL, &ppriv->nctx, dev); + ppriv++; + } +} + +static int __cold dpaa2_dpdmai_bind(struct dpaa2_qdma_priv *priv) +{ + int err; + int i, num; + struct device *dev = priv->dev; + struct dpaa2_qdma_priv_per_prio *ppriv; + struct dpdmai_rx_queue_cfg rx_queue_cfg; + struct fsl_mc_device *ls_dev = to_fsl_mc_device(dev); + + num = priv->num_pairs; + ppriv = priv->ppriv; + for (i = 0; i < num; i++) { + rx_queue_cfg.options = DPDMAI_QUEUE_OPT_USER_CTX | + DPDMAI_QUEUE_OPT_DEST; + rx_queue_cfg.user_ctx = ppriv->nctx.qman64; + rx_queue_cfg.dest_cfg.dest_type = DPDMAI_DEST_DPIO; + rx_queue_cfg.dest_cfg.dest_id = ppriv->nctx.dpio_id; + rx_queue_cfg.dest_cfg.priority = ppriv->prio; + err = dpdmai_set_rx_queue(priv->mc_io, 0, ls_dev->mc_handle, + rx_queue_cfg.dest_cfg.priority, + &rx_queue_cfg); + if (err) { + dev_err(dev, "dpdmai_set_rx_queue() failed\n"); + return err; + } + + ppriv++; + } + + return 0; +} + +static int __cold dpaa2_dpdmai_dpio_unbind(struct dpaa2_qdma_priv *priv) +{ + int i; + int err = 0; + struct device *dev = priv->dev; + struct fsl_mc_device *ls_dev = to_fsl_mc_device(dev); + struct dpaa2_qdma_priv_per_prio *ppriv = priv->ppriv; + + for (i = 0; i < priv->num_pairs; i++) { + ppriv->nctx.qman64 = 0; + ppriv->nctx.dpio_id = 0; + ppriv++; + } + + err = dpdmai_reset(priv->mc_io, 0, ls_dev->mc_handle); + if (err) + dev_err(dev, "dpdmai_reset() failed\n"); + + return err; +} + +static void dpaa2_dpdmai_free_comp(struct dpaa2_qdma_chan *qchan, + struct list_head *head) +{ + struct dpaa2_qdma_comp *comp_tmp, *_comp_tmp; + + list_for_each_entry_safe(comp_tmp, _comp_tmp, + head, list) { + dma_pool_free(qchan->fd_pool, + comp_tmp->fd_virt_addr, + comp_tmp->fd_bus_addr); + list_del(&comp_tmp->list); + kfree(comp_tmp); + } +} + +static void dpaa2_dpdmai_free_channels(struct dpaa2_qdma_engine *dpaa2_qdma) +{ + struct dpaa2_qdma_chan *qchan; + int num, i; + + num = dpaa2_qdma->n_chans; + for (i = 0; i < num; i++) { + qchan = &dpaa2_qdma->chans[i]; + dpaa2_dpdmai_free_comp(qchan, &qchan->comp_used); + dpaa2_dpdmai_free_comp(qchan, &qchan->comp_free); + dma_pool_destroy(qchan->fd_pool); + } +} + +static void dpaa2_qdma_free_desc(struct virt_dma_desc *vdesc) +{ + struct dpaa2_qdma_comp *dpaa2_comp; + struct dpaa2_qdma_chan *qchan; + + dpaa2_comp = to_fsl_qdma_comp(vdesc); + qchan = dpaa2_comp->qchan; + list_del(&dpaa2_comp->list); + list_add_tail(&dpaa2_comp->list, &qchan->comp_free); +} + +static int dpaa2_dpdmai_init_channels(struct dpaa2_qdma_engine *dpaa2_qdma) +{ + struct dpaa2_qdma_chan *dpaa2_chan; + int i; + + INIT_LIST_HEAD(&dpaa2_qdma->dma_dev.channels); + for (i = 0; i < dpaa2_qdma->n_chans; i++) { + dpaa2_chan = &dpaa2_qdma->chans[i]; + dpaa2_chan->qdma = dpaa2_qdma; + dpaa2_chan->vchan.desc_free = dpaa2_qdma_free_desc; + vchan_init(&dpaa2_chan->vchan, &dpaa2_qdma->dma_dev); + spin_lock_init(&dpaa2_chan->queue_lock); + INIT_LIST_HEAD(&dpaa2_chan->comp_used); + INIT_LIST_HEAD(&dpaa2_chan->comp_free); + } + return 0; +} + +static int dpaa2_qdma_probe(struct fsl_mc_device *dpdmai_dev) +{ + struct dpaa2_qdma_priv *priv; + struct device *dev = &dpdmai_dev->dev; + struct dpaa2_qdma_engine *dpaa2_qdma; + int err; + + priv = kzalloc(sizeof(*priv), GFP_KERNEL); + if (!priv) + return -ENOMEM; + dev_set_drvdata(dev, priv); + priv->dpdmai_dev = dpdmai_dev; + + priv->iommu_domain = iommu_get_domain_for_dev(dev); + if (priv->iommu_domain) + smmu_disable = false; + + /* obtain a MC portal */ + err = fsl_mc_portal_allocate(dpdmai_dev, 0, &priv->mc_io); + if (err) { + if (err == -ENXIO) + err = -EPROBE_DEFER; + else + dev_err(dev, "MC portal allocation failed\n"); + goto err_mcportal; + } + + /* DPDMAI initialization */ + err = dpaa2_qdma_setup(dpdmai_dev); + if (err) { + dev_err(dev, "dpaa2_dpdmai_setup() failed\n"); + goto err_dpdmai_setup; + } + + /* DPIO */ + err = dpaa2_qdma_dpio_setup(priv); + if (err) { + dev_err(dev, "dpaa2_dpdmai_dpio_setup() failed\n"); + goto err_dpio_setup; + } + + /* DPDMAI binding to DPIO */ + err = dpaa2_dpdmai_bind(priv); + if (err) { + dev_err(dev, "dpaa2_dpdmai_bind() failed\n"); + goto err_bind; + } + + /* DPDMAI enable */ + err = dpdmai_enable(priv->mc_io, 0, dpdmai_dev->mc_handle); + if (err) { + dev_err(dev, "dpdmai_enable() faile\n"); + goto err_enable; + } + + dpaa2_qdma = kzalloc(sizeof(*dpaa2_qdma), GFP_KERNEL); + if (!dpaa2_qdma) { + err = -ENOMEM; + goto err_eng; + } + + priv->dpaa2_qdma = dpaa2_qdma; + dpaa2_qdma->priv = priv; + + dpaa2_qdma->desc_allocated = 0; + dpaa2_qdma->n_chans = NUM_CH; + + dpaa2_dpdmai_init_channels(dpaa2_qdma); + + if (soc_device_match(soc_fixup_tuning)) + dpaa2_qdma->qdma_wrtype_fixup = true; + else + dpaa2_qdma->qdma_wrtype_fixup = false; + + dma_cap_set(DMA_PRIVATE, dpaa2_qdma->dma_dev.cap_mask); + dma_cap_set(DMA_SLAVE, dpaa2_qdma->dma_dev.cap_mask); + dma_cap_set(DMA_MEMCPY, dpaa2_qdma->dma_dev.cap_mask); + + dpaa2_qdma->dma_dev.dev = dev; + dpaa2_qdma->dma_dev.device_alloc_chan_resources = + dpaa2_qdma_alloc_chan_resources; + dpaa2_qdma->dma_dev.device_free_chan_resources = + dpaa2_qdma_free_chan_resources; + dpaa2_qdma->dma_dev.device_tx_status = dpaa2_qdma_tx_status; + dpaa2_qdma->dma_dev.device_prep_dma_memcpy = dpaa2_qdma_prep_memcpy; + dpaa2_qdma->dma_dev.device_issue_pending = dpaa2_qdma_issue_pending; + + err = dma_async_device_register(&dpaa2_qdma->dma_dev); + if (err) { + dev_err(dev, "Can't register NXP QDMA engine.\n"); + goto err_eng; + } + + return 0; + +err_eng: + dpdmai_disable(priv->mc_io, 0, dpdmai_dev->mc_handle); +err_enable: + dpaa2_dpdmai_dpio_unbind(priv); +err_bind: + dpaa2_dpmai_store_free(priv); + dpaa2_dpdmai_dpio_free(priv); +err_dpio_setup: + dpdmai_close(priv->mc_io, 0, dpdmai_dev->mc_handle); +err_dpdmai_setup: + fsl_mc_portal_free(priv->mc_io); +err_mcportal: + kfree(priv->ppriv); + kfree(priv); + dev_set_drvdata(dev, NULL); + return err; +} + +static int dpaa2_qdma_remove(struct fsl_mc_device *ls_dev) +{ + struct device *dev; + struct dpaa2_qdma_priv *priv; + struct dpaa2_qdma_engine *dpaa2_qdma; + + dev = &ls_dev->dev; + priv = dev_get_drvdata(dev); + dpaa2_qdma = priv->dpaa2_qdma; + + dpdmai_disable(priv->mc_io, 0, ls_dev->mc_handle); + dpaa2_dpdmai_dpio_unbind(priv); + dpaa2_dpmai_store_free(priv); + dpaa2_dpdmai_dpio_free(priv); + dpdmai_close(priv->mc_io, 0, ls_dev->mc_handle); + fsl_mc_portal_free(priv->mc_io); + dev_set_drvdata(dev, NULL); + dpaa2_dpdmai_free_channels(dpaa2_qdma); + + dma_async_device_unregister(&dpaa2_qdma->dma_dev); + kfree(priv); + kfree(dpaa2_qdma); + + return 0; +} + +static const struct fsl_mc_device_id dpaa2_qdma_id_table[] = { + { + .vendor = FSL_MC_VENDOR_FREESCALE, + .obj_type = "dpdmai", + }, + { .vendor = 0x0 } +}; + +static struct fsl_mc_driver dpaa2_qdma_driver = { + .driver = { + .name = "dpaa2-qdma", + .owner = THIS_MODULE, + }, + .probe = dpaa2_qdma_probe, + .remove = dpaa2_qdma_remove, + .match_id_table = dpaa2_qdma_id_table +}; + +static int __init dpaa2_qdma_driver_init(void) +{ + return fsl_mc_driver_register(&(dpaa2_qdma_driver)); +} +late_initcall(dpaa2_qdma_driver_init); + +static void __exit fsl_qdma_exit(void) +{ + fsl_mc_driver_unregister(&(dpaa2_qdma_driver)); +} +module_exit(fsl_qdma_exit); + +MODULE_ALIAS("platform:fsl-dpaa2-qdma"); +MODULE_LICENSE("GPL v2"); +MODULE_DESCRIPTION("NXP Layerscape DPAA2 qDMA engine driver"); diff --git a/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.h b/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.h new file mode 100644 index 0000000..3fcade8 --- /dev/null +++ b/drivers/dma/fsl-dpaa2-qdma/dpaa2-qdma.h @@ -0,0 +1,152 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* Copyright 2014-2018 NXP */ + +#ifndef __DPAA2_QDMA_H +#define __DPAA2_QDMA_H + +#define LONG_FORMAT 1 + +#define DPAA2_QDMA_STORE_SIZE 16 +#define NUM_CH 8 + +struct dpaa2_qdma_sd_d { + u32 rsv:32; + union { + struct { + u32 ssd:12; /* souce stride distance */ + u32 sss:12; /* souce stride size */ + u32 rsv1:8; + } sdf; + struct { + u32 dsd:12; /* Destination stride distance */ + u32 dss:12; /* Destination stride size */ + u32 rsv2:8; + } ddf; + } df; + u32 rbpcmd; /* Route-by-port command */ + u32 cmd; +} __attribute__((__packed__)); + +/* Source descriptor command read transaction type for RBP=0: */ +/* coherent copy of cacheable memory */ +#define QDMA_SD_CMD_RDTTYPE_COHERENT (0xb << 28) +/* Destination descriptor command write transaction type for RBP=0: */ +/* coherent copy of cacheable memory */ +#define QDMA_DD_CMD_WRTTYPE_COHERENT (0x6 << 28) +#define LX2160_QDMA_DD_CMD_WRTTYPE_COHERENT (0xb << 28) + +#define QMAN_FD_FMT_ENABLE BIT(0) /* frame list table enable */ +#define QMAN_FD_BMT_ENABLE BIT(15) /* bypass memory translation */ +#define QMAN_FD_BMT_DISABLE (0) /* bypass memory translation */ +#define QMAN_FD_SL_DISABLE (0) /* short lengthe disabled */ +#define QMAN_FD_SL_ENABLE BIT(14) /* short lengthe enabled */ + +#define QDMA_FINAL_BIT_DISABLE (0) /* final bit disable */ +#define QDMA_FINAL_BIT_ENABLE BIT(31) /* final bit enable */ + +#define QDMA_FD_SHORT_FORMAT BIT(11) /* short format */ +#define QDMA_FD_LONG_FORMAT (0) /* long format */ +#define QDMA_SER_DISABLE (8) /* no notification */ +#define QDMA_SER_CTX BIT(8) /* notification by FQD_CTX[fqid] */ +#define QDMA_SER_DEST (2 << 8) /* notification by destination desc */ +#define QDMA_SER_BOTH (3 << 8) /* soruce and dest notification */ +#define QDMA_FD_SPF_ENALBE BIT(30) /* source prefetch enable */ + +#define QMAN_FD_VA_ENABLE BIT(14) /* Address used is virtual address */ +#define QMAN_FD_VA_DISABLE (0)/* Address used is a real address */ +/* Flow Context: 49bit physical address */ +#define QMAN_FD_CBMT_ENABLE BIT(15) +#define QMAN_FD_CBMT_DISABLE (0) /* Flow Context: 64bit virtual address */ +#define QMAN_FD_SC_DISABLE (0) /* stashing control */ + +#define QDMA_FL_FMT_SBF (0x0) /* Single buffer frame */ +#define QDMA_FL_FMT_SGE (0x2) /* Scatter gather frame */ +#define QDMA_FL_BMT_ENABLE BIT(15) /* enable bypass memory translation */ +#define QDMA_FL_BMT_DISABLE (0x0) /* enable bypass memory translation */ +#define QDMA_FL_SL_LONG (0x0)/* long length */ +#define QDMA_FL_SL_SHORT (0x1) /* short length */ +#define QDMA_FL_F (0x1)/* last frame list bit */ + +/*Description of Frame list table structure*/ +struct dpaa2_qdma_chan { + struct virt_dma_chan vchan; + struct virt_dma_desc vdesc; + enum dma_status status; + struct dpaa2_qdma_engine *qdma; + + struct mutex dpaa2_queue_mutex; + spinlock_t queue_lock; + struct dma_pool *fd_pool; + + struct list_head comp_used; + struct list_head comp_free; + +}; + +struct dpaa2_qdma_comp { + dma_addr_t fd_bus_addr; + dma_addr_t fl_bus_addr; + dma_addr_t desc_bus_addr; + void *fd_virt_addr; + void *fl_virt_addr; + void *desc_virt_addr; + struct dpaa2_qdma_chan *qchan; + struct virt_dma_desc vdesc; + struct list_head list; +}; + +struct dpaa2_qdma_engine { + struct dma_device dma_dev; + u32 n_chans; + struct dpaa2_qdma_chan chans[NUM_CH]; + int qdma_wrtype_fixup; + int desc_allocated; + + struct dpaa2_qdma_priv *priv; +}; + +/* + * dpaa2_qdma_priv - driver private data + */ +struct dpaa2_qdma_priv { + int dpqdma_id; + + struct iommu_domain *iommu_domain; + struct dpdmai_attr dpdmai_attr; + struct device *dev; + struct fsl_mc_io *mc_io; + struct fsl_mc_device *dpdmai_dev; + u8 num_pairs; + + struct dpaa2_qdma_engine *dpaa2_qdma; + struct dpaa2_qdma_priv_per_prio *ppriv; + + struct dpdmai_rx_queue_attr rx_queue_attr[DPDMAI_PRIO_NUM]; + struct dpdmai_tx_queue_attr tx_queue_attr[DPDMAI_PRIO_NUM]; +}; + +struct dpaa2_qdma_priv_per_prio { + int req_fqid; + int rsp_fqid; + int prio; + + struct dpaa2_io_store *store; + struct dpaa2_io_notification_ctx nctx; + + struct dpaa2_qdma_priv *priv; +}; + +static struct soc_device_attribute soc_fixup_tuning[] = { + { .family = "QorIQ LX2160A"}, + { }, +}; + +/* FD pool size: one FD + 3 Frame list + 2 source/destination descriptor */ +#define FD_POOL_SIZE (sizeof(struct dpaa2_fd) + \ + sizeof(struct dpaa2_fl_entry) * 3 + \ + sizeof(struct dpaa2_qdma_sd_d) * 2) + +static void dpaa2_dpdmai_free_channels(struct dpaa2_qdma_engine *dpaa2_qdma); +static void dpaa2_dpdmai_free_comp(struct dpaa2_qdma_chan *qchan, + struct list_head *head); +#endif /* __DPAA2_QDMA_H */ -- 1.7.1