From patchwork Mon Jun 15 13:29:18 2015 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Marciniszyn, Mike" X-Patchwork-Id: 6609311 Return-Path: X-Original-To: patchwork-linux-rdma@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork2.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.29.136]) by patchwork2.web.kernel.org (Postfix) with ESMTP id 2D7F2C0433 for ; Mon, 15 Jun 2015 13:29:29 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id ABA0A205DA for ; Mon, 15 Jun 2015 13:29:26 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id BF962205D1 for ; Mon, 15 Jun 2015 13:29:23 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755553AbbFON3W (ORCPT ); Mon, 15 Jun 2015 09:29:22 -0400 Received: from mga14.intel.com ([192.55.52.115]:21950 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755164AbbFON3V (ORCPT ); Mon, 15 Jun 2015 09:29:21 -0400 Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by fmsmga103.fm.intel.com with ESMTP; 15 Jun 2015 06:29:20 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.13,618,1427785200"; d="scan'208";a="727636971" Received: from sedona.ch.intel.com ([143.182.228.65]) by fmsmga001.fm.intel.com with ESMTP; 15 Jun 2015 06:29:21 -0700 Received: from phlsvsles11.ph.intel.com (phlsvsles11.ph.intel.com [10.228.195.43]) by sedona.ch.intel.com (8.13.6/8.14.3/Standard MailSET/Hub) with ESMTP id t5FDTJ7Z032431; Mon, 15 Jun 2015 06:29:19 -0700 Received: from phlsvslse11.ph.intel.com (localhost [127.0.0.1]) by phlsvsles11.ph.intel.com with ESMTP id t5FDTIOr030654; Mon, 15 Jun 2015 09:29:18 -0400 Subject: [PATCH v2 37/49] IB/hfi1: add sdma header file To: dledford@redhat.com From: Mike Marciniszyn Cc: linux-rdma@vger.kernel.org Date: Mon, 15 Jun 2015 09:29:18 -0400 Message-ID: <20150615132918.29741.4055.stgit@phlsvslse11.ph.intel.com> In-Reply-To: <20150615132434.29741.35258.stgit@phlsvslse11.ph.intel.com> References: <20150615132434.29741.35258.stgit@phlsvslse11.ph.intel.com> User-Agent: StGit/0.16 MIME-Version: 1.0 Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org X-Spam-Status: No, score=-7.4 required=5.0 tests=BAYES_00, 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 Signed-off-by: Andrew Friedley Signed-off-by: Arthur Kepner Signed-off-by: Brendan Cunningham Signed-off-by: Brian Welty Signed-off-by: Caz Yokoyama Signed-off-by: Dean Luick Signed-off-by: Dennis Dalessandro Signed-off-by: Easwar Hariharan Signed-off-by: Harish Chegondi Signed-off-by: Ira Weiny Signed-off-by: Jim Snow Signed-off-by: John Gregor Signed-off-by: Jubin John Signed-off-by: Kaike Wan Signed-off-by: Kevin Pine Signed-off-by: Kyle Liddell Signed-off-by: Mike Marciniszyn Signed-off-by: Mitko Haralanov Signed-off-by: Ravi Krishnaswamy Signed-off-by: Sadanand Warrier Signed-off-by: Sanath Kumar Signed-off-by: Sudeep Dutt Signed-off-by: Vlad Danushevsky --- drivers/infiniband/hw/hfi1/sdma.h | 1122 +++++++++++++++++++++++++++++++++++++ 1 file changed, 1122 insertions(+) create mode 100644 drivers/infiniband/hw/hfi1/sdma.h -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html diff --git a/drivers/infiniband/hw/hfi1/sdma.h b/drivers/infiniband/hw/hfi1/sdma.h new file mode 100644 index 0000000..2d2ff07 --- /dev/null +++ b/drivers/infiniband/hw/hfi1/sdma.h @@ -0,0 +1,1122 @@ +#ifndef _HFI1_SDMA_H +#define _HFI1_SDMA_H +/* + * + * This file is provided under a dual BSD/GPLv2 license. When using or + * redistributing this file, you may do so under either license. + * + * GPL LICENSE SUMMARY + * + * Copyright(c) 2015 Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of version 2 of the GNU General Public License as + * published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * BSD LICENSE + * + * Copyright(c) 2015 Intel Corporation. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * + * - Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * - Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in + * the documentation and/or other materials provided with the + * distribution. + * - Neither the name of Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived + * from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS + * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR + * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT + * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, + * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE + * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + */ + +#include +#include +#include +#include +#include + +#include "hfi.h" +#include "verbs.h" + +/* increased for AHG */ +#define NUM_DESC 6 +/* Hardware limit */ +#define MAX_DESC 64 +/* Hardware limit for SDMA packet size */ +#define MAX_SDMA_PKT_SIZE ((16 * 1024) - 1) + + +#define SDMA_TXREQ_S_OK 0 +#define SDMA_TXREQ_S_SENDERROR 1 +#define SDMA_TXREQ_S_ABORTED 2 +#define SDMA_TXREQ_S_SHUTDOWN 3 + +/* flags bits */ +#define SDMA_TXREQ_F_URGENT 0x0001 +#define SDMA_TXREQ_F_AHG_COPY 0x0002 +#define SDMA_TXREQ_F_USE_AHG 0x0004 + +#define SDMA_MAP_NONE 0 +#define SDMA_MAP_SINGLE 1 +#define SDMA_MAP_PAGE 2 + +#define SDMA_AHG_VALUE_MASK 0xffff +#define SDMA_AHG_VALUE_SHIFT 0 +#define SDMA_AHG_INDEX_MASK 0xf +#define SDMA_AHG_INDEX_SHIFT 16 +#define SDMA_AHG_FIELD_LEN_MASK 0xf +#define SDMA_AHG_FIELD_LEN_SHIFT 20 +#define SDMA_AHG_FIELD_START_MASK 0x1f +#define SDMA_AHG_FIELD_START_SHIFT 24 +#define SDMA_AHG_UPDATE_ENABLE_MASK 0x1 +#define SDMA_AHG_UPDATE_ENABLE_SHIFT 31 + +/* AHG modes */ + +/* + * Be aware the ordering and values + * for SDMA_AHG_APPLY_UPDATE[123] + * are assumed in generating a skip + * count in submit_tx() in sdma.c + */ +#define SDMA_AHG_NO_AHG 0 +#define SDMA_AHG_COPY 1 +#define SDMA_AHG_APPLY_UPDATE1 2 +#define SDMA_AHG_APPLY_UPDATE2 3 +#define SDMA_AHG_APPLY_UPDATE3 4 + +/* + * Bits defined in the send DMA descriptor. + */ +#define SDMA_DESC0_FIRST_DESC_FLAG (1ULL<<63) +#define SDMA_DESC0_LAST_DESC_FLAG (1ULL<<62) +#define SDMA_DESC0_BYTE_COUNT_SHIFT 48 +#define SDMA_DESC0_BYTE_COUNT_WIDTH 14 +#define SDMA_DESC0_BYTE_COUNT_MASK \ + ((1ULL<descq_tail == sde->descq_head; +} + +static inline u16 sdma_descq_freecnt(struct sdma_engine *sde) +{ + return sde->descq_cnt - + (sde->descq_tail - + ACCESS_ONCE(sde->descq_head)) - 1; +} + +static inline u16 sdma_descq_inprocess(struct sdma_engine *sde) +{ + return sde->descq_cnt - sdma_descq_freecnt(sde); +} + +/* + * Either head_lock or tail lock required to see + * a steady state. + */ +static inline int __sdma_running(struct sdma_engine *engine) +{ + return engine->state.current_state == sdma_state_s99_running; +} + + +/** + * sdma_running() - state suitability test + * @engine: sdma engine + * + * sdma_running probes the internal state to determine if it is suitable + * for submitting packets. + * + * Return: + * 1 - ok to submit, 0 - not ok to submit + * + */ +static inline int sdma_running(struct sdma_engine *engine) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&engine->tail_lock, flags); + ret = __sdma_running(engine); + spin_unlock_irqrestore(&engine->tail_lock, flags); + return ret; +} + +void _sdma_txreq_ahgadd( + struct sdma_txreq *tx, + u8 num_ahg, + u8 ahg_entry, + u32 *ahg, + u8 ahg_hlen); + + +/** + * sdma_txinit_ahg() - initialize an sdma_txreq struct with AHG + * @tx: tx request to initialize + * @flags: flags to key last descriptor additions + * @tlen: total packet length (pbc + headers + data) + * @ahg_entry: ahg entry to use (0 - 31) + * @num_ahg: ahg descriptor for first descriptor (0 - 9) + * @ahg: array of AHG descriptors (up to 9 entries) + * @ahg_hlen: number of bytes from ASIC entry to use + * @cb: callback + * + * The allocation of the sdma_txreq and it enclosing structure is user + * dependent. This routine must be called to initialize the user independent + * fields. + * + * The currently supported flags are SDMA_TXREQ_F_URGENT, + * SDMA_TXREQ_F_AHG_COPY, and SDMA_TXREQ_F_USE_AHG. + * + * SDMA_TXREQ_F_URGENT is used for latency sensitive situations where the + * completion is desired as soon as possible. + * + * SDMA_TXREQ_F_AHG_COPY causes the header in the first descriptor to be + * copied to chip entry. SDMA_TXREQ_F_USE_AHG causes the code to add in + * the AHG descriptors into the first 1 to 3 descriptors. + * + * Completions of submitted requests can be gotten on selected + * txreqs by giving a completion routine callback to sdma_txinit() or + * sdma_txinit_ahg(). The environment in which the callback runs + * can be from an ISR, a tasklet, or a thread, so no sleeping + * kernel routines can be used. Aspects of the sdma ring may + * be locked so care should be taken with locking. + * + * The callback pointer can be NULL to avoid any callback for the packet + * being submitted. The callback will be provided this tx, a status, and a flag. + * + * The status will be one of SDMA_TXREQ_S_OK, SDMA_TXREQ_S_SENDERROR, + * SDMA_TXREQ_S_ABORTED, or SDMA_TXREQ_S_SHUTDOWN. + * + * The flag, if the is the iowait had been used, indicates the iowait + * sdma_busy count has reached zero. + * + * user data portion of tlen should be precise. The sdma_txadd_* entrances + * will pad with a descriptor references 1 - 3 bytes when the number of bytes + * specified in tlen have been supplied to the sdma_txreq. + * + * ahg_hlen is used to determine the number of on-chip entry bytes to + * use as the header. This is for cases where the stored header is + * larger than the header to be used in a packet. This is typical + * for verbs where an RDMA_WRITE_FIRST is larger than the packet in + * and RDMA_WRITE_MIDDLE. + * + */ +static inline int sdma_txinit_ahg( + struct sdma_txreq *tx, + u16 flags, + u16 tlen, + u8 ahg_entry, + u8 num_ahg, + u32 *ahg, + u8 ahg_hlen, + void (*cb)(struct sdma_txreq *, int, int)) +{ + if (tlen == 0) + return -ENODATA; + if (tlen > MAX_SDMA_PKT_SIZE) + return -EMSGSIZE; + tx->desc_limit = ARRAY_SIZE(tx->descs); + tx->descp = &tx->descs[0]; + INIT_LIST_HEAD(&tx->list); + tx->num_desc = 0; + tx->flags = flags; + tx->complete = cb; + tx->coalesce_buf = NULL; + tx->wait = NULL; + tx->tlen = tx->packet_len = tlen; + tx->descs[0].qw[0] = SDMA_DESC0_FIRST_DESC_FLAG; + tx->descs[0].qw[1] = 0; + if (flags & SDMA_TXREQ_F_AHG_COPY) + tx->descs[0].qw[1] |= + (((u64)ahg_entry & SDMA_DESC1_HEADER_INDEX_MASK) + << SDMA_DESC1_HEADER_INDEX_SHIFT) | + (((u64)SDMA_AHG_COPY & SDMA_DESC1_HEADER_MODE_MASK) + << SDMA_DESC1_HEADER_MODE_SHIFT); + else if (flags & SDMA_TXREQ_F_USE_AHG && num_ahg) + _sdma_txreq_ahgadd(tx, num_ahg, ahg_entry, ahg, ahg_hlen); + return 0; +} + +/** + * sdma_txinit() - initialize an sdma_txreq struct (no AHG) + * @tx: tx request to initialize + * @flags: flags to key last descriptor additions + * @tlen: total packet length (pbc + headers + data) + * @cb: callback pointer + * + * The allocation of the sdma_txreq and it enclosing structure is user + * dependent. This routine must be called to initialize the user + * independent fields. + * + * The currently supported flags is SDMA_TXREQ_F_URGENT. + * + * SDMA_TXREQ_F_URGENT is used for latency sensitive situations where the + * completion is desired as soon as possible. + * + * Completions of submitted requests can be gotten on selected + * txreqs by giving a completion routine callback to sdma_txinit() or + * sdma_txinit_ahg(). The environment in which the callback runs + * can be from an ISR, a tasklet, or a thread, so no sleeping + * kernel routines can be used. The head size of the sdma ring may + * be locked so care should be taken with locking. + * + * The callback pointer can be NULL to avoid any callback for the packet + * being submitted. + * + * The callback, if non-NULL, will be provided this tx and a status. The + * status will be one of SDMA_TXREQ_S_OK, SDMA_TXREQ_S_SENDERROR, + * SDMA_TXREQ_S_ABORTED, or SDMA_TXREQ_S_SHUTDOWN. + * + */ +static inline int sdma_txinit( + struct sdma_txreq *tx, + u16 flags, + u16 tlen, + void (*cb)(struct sdma_txreq *, int, int)) +{ + return sdma_txinit_ahg(tx, flags, tlen, 0, 0, NULL, 0, cb); +} + +/* helpers - don't use */ +static inline int sdma_mapping_type(struct sdma_desc *d) +{ + return (d->qw[1] & SDMA_DESC1_GENERATION_SMASK) + >> SDMA_DESC1_GENERATION_SHIFT; +} + +static inline size_t sdma_mapping_len(struct sdma_desc *d) +{ + return (d->qw[0] & SDMA_DESC0_BYTE_COUNT_SMASK) + >> SDMA_DESC0_BYTE_COUNT_SHIFT; +} + +static inline dma_addr_t sdma_mapping_addr(struct sdma_desc *d) +{ + return (d->qw[0] & SDMA_DESC0_PHY_ADDR_SMASK) + >> SDMA_DESC0_PHY_ADDR_SHIFT; +} + +static inline void make_tx_sdma_desc( + struct sdma_txreq *tx, + int type, + dma_addr_t addr, + size_t len) +{ + struct sdma_desc *desc = &tx->descp[tx->num_desc]; + + if (!tx->num_desc) { + /* qw[0] zero; qw[1] first, ahg mode already in from init */ + desc->qw[1] |= ((u64)type & SDMA_DESC1_GENERATION_MASK) + << SDMA_DESC1_GENERATION_SHIFT; + } else { + desc->qw[0] = 0; + desc->qw[1] = ((u64)type & SDMA_DESC1_GENERATION_MASK) + << SDMA_DESC1_GENERATION_SHIFT; + } + desc->qw[0] |= (((u64)addr & SDMA_DESC0_PHY_ADDR_MASK) + << SDMA_DESC0_PHY_ADDR_SHIFT) | + (((u64)len & SDMA_DESC0_BYTE_COUNT_MASK) + << SDMA_DESC0_BYTE_COUNT_SHIFT); +} + +/* helper to extend txreq */ +int _extend_sdma_tx_descs(struct hfi1_devdata *, struct sdma_txreq *); +int _pad_sdma_tx_descs(struct hfi1_devdata *, struct sdma_txreq *); +void sdma_txclean(struct hfi1_devdata *, struct sdma_txreq *); + +/* helpers used by public routines */ +static inline void _sdma_close_tx(struct hfi1_devdata *dd, + struct sdma_txreq *tx) +{ + tx->descp[tx->num_desc].qw[0] |= + SDMA_DESC0_LAST_DESC_FLAG; + tx->descp[tx->num_desc].qw[1] |= + dd->default_desc1; + if (tx->flags & SDMA_TXREQ_F_URGENT) + tx->descp[tx->num_desc].qw[1] |= + (SDMA_DESC1_HEAD_TO_HOST_FLAG| + SDMA_DESC1_INT_REQ_FLAG); +} + +static inline int _sdma_txadd_daddr( + struct hfi1_devdata *dd, + int type, + struct sdma_txreq *tx, + dma_addr_t addr, + u16 len) +{ + int rval = 0; + + if ((unlikely(tx->num_desc == tx->desc_limit))) { + rval = _extend_sdma_tx_descs(dd, tx); + if (rval) + return rval; + } + make_tx_sdma_desc( + tx, + type, + addr, len); + BUG_ON(len > tx->tlen); + tx->tlen -= len; + /* special cases for last */ + if (!tx->tlen) { + if (tx->packet_len & (sizeof(u32) - 1)) + rval = _pad_sdma_tx_descs(dd, tx); + else + _sdma_close_tx(dd, tx); + } + tx->num_desc++; + return rval; +} + +/** + * sdma_txadd_page() - add a page to the sdma_txreq + * @dd: the device to use for mapping + * @tx: tx request to which the page is added + * @page: page to map + * @offset: offset within the page + * @len: length in bytes + * + * This is used to add a page/offset/length descriptor. + * + * The mapping/unmapping of the page/offset/len is automatically handled. + * + * Return: + * 0 - success, -ENOSPC - mapping fail, -ENOMEM - couldn't + * extend descriptor array or couldn't allocate coalesce + * buffer. + * + */ +static inline int sdma_txadd_page( + struct hfi1_devdata *dd, + struct sdma_txreq *tx, + struct page *page, + unsigned long offset, + u16 len) +{ + dma_addr_t addr = + dma_map_page( + &dd->pcidev->dev, + page, + offset, + len, + DMA_TO_DEVICE); + if (unlikely(dma_mapping_error(&dd->pcidev->dev, addr))) { + sdma_txclean(dd, tx); + return -ENOSPC; + } + return _sdma_txadd_daddr( + dd, SDMA_MAP_PAGE, tx, addr, len); +} + +/** + * sdma_txadd_daddr() - add a dma address to the sdma_txreq + * @dd: the device to use for mapping + * @tx: sdma_txreq to which the page is added + * @addr: dma address mapped by caller + * @len: length in bytes + * + * This is used to add a descriptor for memory that is already dma mapped. + * + * In this case, there is no unmapping as part of the progress processing for + * this memory location. + * + * Return: + * 0 - success, -ENOMEM - couldn't extend descriptor array + */ + +static inline int sdma_txadd_daddr( + struct hfi1_devdata *dd, + struct sdma_txreq *tx, + dma_addr_t addr, + u16 len) +{ + return _sdma_txadd_daddr(dd, SDMA_MAP_NONE, tx, addr, len); +} + +/** + * sdma_txadd_kvaddr() - add a kernel virtual address to sdma_txreq + * @dd: the device to use for mapping + * @tx: sdma_txreq to which the page is added + * @kvaddr: the kernel virtual address + * @len: length in bytes + * + * This is used to add a descriptor referenced by the indicated kvaddr and + * len. + * + * The mapping/unmapping of the kvaddr and len is automatically handled. + * + * Return: + * 0 - success, -ENOSPC - mapping fail, -ENOMEM - couldn't extend + * descriptor array + */ +static inline int sdma_txadd_kvaddr( + struct hfi1_devdata *dd, + struct sdma_txreq *tx, + void *kvaddr, + u16 len) +{ + dma_addr_t addr = + dma_map_single( + &dd->pcidev->dev, + kvaddr, + len, + DMA_TO_DEVICE); + if (unlikely(dma_mapping_error(&dd->pcidev->dev, addr))) { + sdma_txclean(dd, tx); + return -ENOSPC; + } + return _sdma_txadd_daddr( + dd, SDMA_MAP_SINGLE, tx, addr, len); +} + +struct iowait; + +int sdma_send_txreq(struct sdma_engine *sde, + struct iowait *wait, + struct sdma_txreq *tx); +int sdma_send_txlist(struct sdma_engine *sde, + struct iowait *wait, + struct list_head *tx_list); + +int sdma_ahg_alloc(struct sdma_engine *sde); +void sdma_ahg_free(struct sdma_engine *sde, int ahg_index); + +/** + * sdma_build_ahg - build ahg descriptor + * @data + * @dwindex + * @startbit + * @bits + * + * Build and return a 32 bit descriptor. + */ +static inline u32 sdma_build_ahg_descriptor( + u16 data, + u8 dwindex, + u8 startbit, + u8 bits) +{ + return (u32)(1UL << SDMA_AHG_UPDATE_ENABLE_SHIFT | + ((startbit & SDMA_AHG_FIELD_START_MASK) << + SDMA_AHG_FIELD_START_SHIFT) | + ((bits & SDMA_AHG_FIELD_LEN_MASK) << + SDMA_AHG_FIELD_LEN_SHIFT) | + ((dwindex & SDMA_AHG_INDEX_MASK) << + SDMA_AHG_INDEX_SHIFT) | + ((data & SDMA_AHG_VALUE_MASK) << + SDMA_AHG_VALUE_SHIFT)); +} + +/** + * sdma_progress - use seq number of detect head progress + * @sde: sdma_engine to check + * @seq: base seq count + * @tx: txreq for which we need to check descriptor availability + * + * This is used in the appropriate spot in the sleep routine + * to check for potential ring progress. This routine gets the + * seqcount before queuing the iowait structure for progress. + * + * If the seqcount indicates that progress needs to be checked, + * re-submission is detected by checking whether the descriptor + * queue has enough descriptor for the txreq. + */ +static inline unsigned sdma_progress(struct sdma_engine *sde, unsigned seq, + struct sdma_txreq *tx) +{ + if (read_seqretry(&sde->head_lock, seq)) { + sde->desc_avail = sdma_descq_freecnt(sde); + if (tx->num_desc > sde->desc_avail) + return 0; + return 1; + } + return 0; +} + +/** + * sdma_iowait_schedule() - initialize wait structure + * @sde: sdma_engine to schedule + * @wait: wait struct to schedule + * + * This function initializes the iowait + * structure embedded in the QP or PQ. + * + */ +static inline void sdma_iowait_schedule( + struct sdma_engine *sde, + struct iowait *wait) +{ + iowait_schedule(wait, sde->wq); +} + +/* for use by interrupt handling */ +void sdma_engine_error(struct sdma_engine *sde, u64 status); +void sdma_engine_interrupt(struct sdma_engine *sde, u64 status); + +/* + * + * The diagram below details the relationship of the mapping structures + * + * Since the mapping now allows for non-uniform engines per vl, the + * number of engines for a vl is either the vl_engines[vl] or + * a computation based on num_sdma/num_vls: + * + * For example: + * nactual = vl_engines ? vl_engines[vl] : num_sdma/num_vls + * + * n = roundup to next highest power of 2 using nactual + * + * In the case where there are num_sdma/num_vls doesn't divide + * evenly, the extras are added from the last vl downward. + * + * For the case where n > nactual, the engines are assigned + * in a round robin fashion wrapping back to the first engine + * for a particular vl. + * + * dd->sdma_map + * | sdma_map_elem[0] + * | +--------------------+ + * v | mask | + * sdma_vl_map |--------------------| + * +--------------------------+ | sde[0] -> eng 1 | + * | list (RCU) | |--------------------| + * |--------------------------| ->| sde[1] -> eng 2 | + * | mask | --/ |--------------------| + * |--------------------------| -/ | * | + * | actual_vls (max 8) | -/ |--------------------| + * |--------------------------| --/ | sde[n] -> eng n | + * | vls (max 8) | -/ +--------------------+ + * |--------------------------| --/ + * | map[0] |-/ + * |--------------------------| +--------------------+ + * | map[1] |--- | mask | + * |--------------------------| \---- |--------------------| + * | * | \-- | sde[0] -> eng 1+n | + * | * | \---- |--------------------| + * | * | \->| sde[1] -> eng 2+n | + * |--------------------------| |--------------------| + * | map[vls - 1] |- | * | + * +--------------------------+ \- |--------------------| + * \- | sde[m] -> eng m+n | + * \ +--------------------+ + * \- + * \ + * \- +--------------------+ + * \- | mask | + * \ |--------------------| + * \- | sde[0] -> eng 1+m+n| + * \- |--------------------| + * >| sde[1] -> eng 2+m+n| + * |--------------------| + * | * | + * |--------------------| + * | sde[o] -> eng o+m+n| + * +--------------------+ + * + */ + +/** + * struct sdma_map_elem - mapping for a vl + * @mask - selector mask + * @sde - array of engines for this vl + * + * The mask is used to "mod" the selector + * to produce index into the trailing + * array of sdes. + */ +struct sdma_map_elem { + u32 mask; + struct sdma_engine *sde[0]; +}; + +/** + * struct sdma_map_el - mapping for a vl + * @list - rcu head for free callback + * @mask - vl mask to "mod" the vl to produce an index to map array + * @actual_vls - number of vls + * @vls - number of vls rounded to next power of 2 + * @map - array of sdma_map_elem entries + * + * This is the parent mapping structure. The trailing + * members of the struct point to sdma_map_elem entries, which + * in turn point to an array of sde's for that vl. + */ +struct sdma_vl_map { + struct rcu_head list; + u32 mask; + u8 actual_vls; + u8 vls; + struct sdma_map_elem *map[0]; +}; + +int sdma_map_init( + struct hfi1_devdata *dd, + u8 port, + u8 num_vls, + u8 *vl_engines); + +/* slow path */ +void _sdma_engine_progress_schedule(struct sdma_engine *sde); + +/** + * sdma_engine_progress_schedule() - schedule progress on engine + * @sde: sdma_engine to schedule progress + * + * This is the fast path. + * + */ +static inline void sdma_engine_progress_schedule( + struct sdma_engine *sde) +{ + if (!sde || sdma_descq_inprocess(sde) < (sde->descq_cnt / 8)) + return; + _sdma_engine_progress_schedule(sde); +} + +struct sdma_engine *sdma_select_engine_sc( + struct hfi1_devdata *dd, + u32 selector, + u8 sc5); + +struct sdma_engine *sdma_select_engine_vl( + struct hfi1_devdata *dd, + u32 selector, + u8 vl); + +void sdma_seqfile_dump_sde(struct seq_file *s, struct sdma_engine *); + +#ifdef CONFIG_SDMA_VERBOSITY +void sdma_dumpstate(struct sdma_engine *); +#endif +static inline char *slashstrip(char *s) +{ + char *r = s; + + while (*s) + if (*s++ == '/') + r = s; + return r; +} + +u16 sdma_get_descq_cnt(void); + +extern uint mod_num_sdma; + +void sdma_update_lmc(struct hfi1_devdata *dd, u64 mask, u32 lid); + +#endif