From patchwork Tue Dec 1 22:19:51 2009 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chien Tung X-Patchwork-Id: 64059 Received: from vger.kernel.org (vger.kernel.org [209.132.176.167]) by demeter.kernel.org (8.14.2/8.14.2) with ESMTP id nB1MJlFO031857 for ; Tue, 1 Dec 2009 22:19:59 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753688AbZLAWTv (ORCPT ); Tue, 1 Dec 2009 17:19:51 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754046AbZLAWTv (ORCPT ); Tue, 1 Dec 2009 17:19:51 -0500 Received: from mga14.intel.com ([143.182.124.37]:30492 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753688AbZLAWTu (ORCPT ); Tue, 1 Dec 2009 17:19:50 -0500 Received: from azsmga001.ch.intel.com ([10.2.17.19]) by azsmga102.ch.intel.com with ESMTP; 01 Dec 2009 14:19:56 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.47,316,1257148800"; d="scan'208";a="217648675" Received: from ctung-mobl.amr.corp.intel.com (HELO ctung-MOBL.intel.com) ([10.232.237.216]) by azsmga001.ch.intel.com with SMTP; 01 Dec 2009 14:19:51 -0800 Received: by ctung-MOBL.intel.com (sSMTP sendmail emulation); Tue, 01 Dec 2009 16:19:51 -0600 Date: Tue, 1 Dec 2009 16:19:51 -0600 From: Chien Tung To: Roland Dreier Cc: linux-rdma@vger.kernel.org Subject: [PATCH 4/9] RDMA/nes: Correct fast memory registration implementation Message-ID: <20091201221951.GA7840@ctung-MOBL> Mime-Version: 1.0 Content-Disposition: inline User-Agent: Mutt/1.4.2.2i Sender: linux-rdma-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-rdma@vger.kernel.org diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 3512d6d..3d9bbff 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -424,8 +424,9 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) { nesadapter->base_pd = 1; - nesadapter->device_cap_flags = - IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW; + nesadapter->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | + IB_DEVICE_MEM_WINDOW | + IB_DEVICE_MEM_MGT_EXTENSIONS; nesadapter->allocated_qps = (unsigned long *)&(((unsigned char *)nesadapter) [(sizeof(struct nes_adapter)+(sizeof(unsigned long)-1))&(~(sizeof(unsigned long)-1))]); diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index f28a41b..8a4c738 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -546,11 +546,23 @@ enum nes_iwarp_sq_fmr_wqe_word_idx { NES_IWARP_SQ_FMR_WQE_PBL_LENGTH_IDX = 14, }; +enum nes_iwarp_sq_fmr_opcodes { + NES_IWARP_SQ_FMR_WQE_ZERO_BASED = (1<<6), + NES_IWARP_SQ_FMR_WQE_PAGE_SIZE_4K = (0<<7), + NES_IWARP_SQ_FMR_WQE_PAGE_SIZE_2M = (1<<7), + NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_LOCAL_READ = (1<<16), + NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_LOCAL_WRITE = (1<<17), + NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_REMOTE_READ = (1<<18), + NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_REMOTE_WRITE = (1<<19), + NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_WINDOW_BIND = (1<<20), +}; + +#define NES_IWARP_SQ_FMR_WQE_MR_LENGTH_HIGH_MASK 0xFF; + enum nes_iwarp_sq_locinv_wqe_word_idx { NES_IWARP_SQ_LOCINV_WQE_INV_STAG_IDX = 6, }; - enum nes_iwarp_rq_wqe_word_idx { NES_IWARP_RQ_WQE_TOTAL_PAYLOAD_IDX = 1, NES_IWARP_RQ_WQE_COMP_CTX_LOW_IDX = 2, @@ -1153,6 +1165,19 @@ struct nes_pbl { /* TODO: need to add list for two level tables */ }; +#define NES_4K_PBL_CHUNK_SIZE 4096 + +struct nes_fast_mr_wqe_pbl { + u64 *kva; + dma_addr_t paddr; +}; + +struct nes_ib_fast_reg_page_list { + struct ib_fast_reg_page_list ibfrpl; + struct nes_fast_mr_wqe_pbl nes_wqe_pbl; + u64 pbl; +}; + struct nes_listener { struct work_struct work; struct workqueue_struct *wq; diff --git a/drivers/infiniband/hw/nes/nes_user.h b/drivers/infiniband/hw/nes/nes_user.h index cc90c14..ce62f3c 100644 --- a/drivers/infiniband/hw/nes/nes_user.h +++ b/drivers/infiniband/hw/nes/nes_user.h @@ -86,6 +86,7 @@ enum iwnes_memreg_type { IWNES_MEMREG_TYPE_CQ = 0x0002, IWNES_MEMREG_TYPE_MW = 0x0003, IWNES_MEMREG_TYPE_FMR = 0x0004, + IWNES_MEMREG_TYPE_FMEM = 0x0005, }; struct nes_mem_reg_req { diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index 499dd78..42ec5d0 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -275,342 +275,236 @@ static int nes_bind_mw(struct ib_qp *ibqp, struct ib_mw *ibmw, } -/** - * nes_alloc_fmr +/* + * nes_alloc_fast_mr */ -static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd, - int ibmr_access_flags, - struct ib_fmr_attr *ibfmr_attr) +static int alloc_fast_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, + u32 stag, u32 page_count) { - unsigned long flags; - struct nes_pd *nespd = to_nespd(ibpd); - struct nes_vnic *nesvnic = to_nesvnic(ibpd->device); - struct nes_device *nesdev = nesvnic->nesdev; - struct nes_adapter *nesadapter = nesdev->nesadapter; - struct nes_fmr *nesfmr; - struct nes_cqp_request *cqp_request; struct nes_hw_cqp_wqe *cqp_wqe; + struct nes_cqp_request *cqp_request; + unsigned long flags; int ret; - u32 stag; - u32 stag_index = 0; - u32 next_stag_index = 0; - u32 driver_key = 0; + struct nes_adapter *nesadapter = nesdev->nesadapter; u32 opcode = 0; - u8 stag_key = 0; - int i=0; - struct nes_vpbl vpbl; - - get_random_bytes(&next_stag_index, sizeof(next_stag_index)); - stag_key = (u8)next_stag_index; - - driver_key = 0; - - next_stag_index >>= 8; - next_stag_index %= nesadapter->max_mr; - - ret = nes_alloc_resource(nesadapter, nesadapter->allocated_mrs, - nesadapter->max_mr, &stag_index, &next_stag_index); - if (ret) { - goto failed_resource_alloc; - } - - nesfmr = kzalloc(sizeof(*nesfmr), GFP_KERNEL); - if (!nesfmr) { - ret = -ENOMEM; - goto failed_fmr_alloc; - } - - nesfmr->nesmr.mode = IWNES_MEMREG_TYPE_FMR; - if (ibfmr_attr->max_pages == 1) { - /* use zero length PBL */ - nesfmr->nesmr.pbl_4k = 0; - nesfmr->nesmr.pbls_used = 0; - } else if (ibfmr_attr->max_pages <= 32) { - /* use PBL 256 */ - nesfmr->nesmr.pbl_4k = 0; - nesfmr->nesmr.pbls_used = 1; - } else if (ibfmr_attr->max_pages <= 512) { - /* use 4K PBLs */ - nesfmr->nesmr.pbl_4k = 1; - nesfmr->nesmr.pbls_used = 1; - } else { - /* use two level 4K PBLs */ - /* add support for two level 256B PBLs */ - nesfmr->nesmr.pbl_4k = 1; - nesfmr->nesmr.pbls_used = 1 + (ibfmr_attr->max_pages >> 9) + - ((ibfmr_attr->max_pages & 511) ? 1 : 0); - } - /* Register the region with the adapter */ - spin_lock_irqsave(&nesadapter->pbl_lock, flags); - - /* track PBL resources */ - if (nesfmr->nesmr.pbls_used != 0) { - if (nesfmr->nesmr.pbl_4k) { - if (nesfmr->nesmr.pbls_used > nesadapter->free_4kpbl) { - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - ret = -ENOMEM; - goto failed_vpbl_avail; - } else { - nesadapter->free_4kpbl -= nesfmr->nesmr.pbls_used; - } - } else { - if (nesfmr->nesmr.pbls_used > nesadapter->free_256pbl) { - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - ret = -ENOMEM; - goto failed_vpbl_avail; - } else { - nesadapter->free_256pbl -= nesfmr->nesmr.pbls_used; - } - } - } - - /* one level pbl */ - if (nesfmr->nesmr.pbls_used == 0) { - nesfmr->root_vpbl.pbl_vbase = NULL; - nes_debug(NES_DBG_MR, "zero level pbl \n"); - } else if (nesfmr->nesmr.pbls_used == 1) { - /* can change it to kmalloc & dma_map_single */ - nesfmr->root_vpbl.pbl_vbase = pci_alloc_consistent(nesdev->pcidev, 4096, - &nesfmr->root_vpbl.pbl_pbase); - if (!nesfmr->root_vpbl.pbl_vbase) { - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - ret = -ENOMEM; - goto failed_vpbl_alloc; - } - nesfmr->leaf_pbl_cnt = 0; - nes_debug(NES_DBG_MR, "one level pbl, root_vpbl.pbl_vbase=%p \n", - nesfmr->root_vpbl.pbl_vbase); - } - /* two level pbl */ - else { - nesfmr->root_vpbl.pbl_vbase = pci_alloc_consistent(nesdev->pcidev, 8192, - &nesfmr->root_vpbl.pbl_pbase); - if (!nesfmr->root_vpbl.pbl_vbase) { - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - ret = -ENOMEM; - goto failed_vpbl_alloc; - } - - nesfmr->leaf_pbl_cnt = nesfmr->nesmr.pbls_used-1; - nesfmr->root_vpbl.leaf_vpbl = kzalloc(sizeof(*nesfmr->root_vpbl.leaf_vpbl)*1024, GFP_ATOMIC); - if (!nesfmr->root_vpbl.leaf_vpbl) { - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - ret = -ENOMEM; - goto failed_leaf_vpbl_alloc; - } - - nes_debug(NES_DBG_MR, "two level pbl, root_vpbl.pbl_vbase=%p" - " leaf_pbl_cnt=%d root_vpbl.leaf_vpbl=%p\n", - nesfmr->root_vpbl.pbl_vbase, nesfmr->leaf_pbl_cnt, nesfmr->root_vpbl.leaf_vpbl); - - for (i=0; ileaf_pbl_cnt; i++) - nesfmr->root_vpbl.leaf_vpbl[i].pbl_vbase = NULL; - - for (i=0; ileaf_pbl_cnt; i++) { - vpbl.pbl_vbase = pci_alloc_consistent(nesdev->pcidev, 4096, - &vpbl.pbl_pbase); - - if (!vpbl.pbl_vbase) { - ret = -ENOMEM; - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - goto failed_leaf_vpbl_pages_alloc; - } - - nesfmr->root_vpbl.pbl_vbase[i].pa_low = cpu_to_le32((u32)vpbl.pbl_pbase); - nesfmr->root_vpbl.pbl_vbase[i].pa_high = cpu_to_le32((u32)((((u64)vpbl.pbl_pbase)>>32))); - nesfmr->root_vpbl.leaf_vpbl[i] = vpbl; - - nes_debug(NES_DBG_MR, "pbase_low=0x%x, pbase_high=0x%x, vpbl=%p\n", - nesfmr->root_vpbl.pbl_vbase[i].pa_low, - nesfmr->root_vpbl.pbl_vbase[i].pa_high, - &nesfmr->root_vpbl.leaf_vpbl[i]); - } - } - nesfmr->ib_qp = NULL; - nesfmr->access_rights =0; + u16 major_code; + u64 region_length = page_count * PAGE_SIZE; - stag = stag_index << 8; - stag |= driver_key; - stag += (u32)stag_key; - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); cqp_request = nes_get_cqp_request(nesdev); if (cqp_request == NULL) { nes_debug(NES_DBG_MR, "Failed to get a cqp_request.\n"); - ret = -ENOMEM; - goto failed_leaf_vpbl_pages_alloc; + return -ENOMEM; } + nes_debug(NES_DBG_MR, "alloc_fast_reg_mr: page_count = %d, " + "region_length = %llu\n", + page_count, region_length); cqp_request->waiting = 1; cqp_wqe = &cqp_request->cqp_wqe; - nes_debug(NES_DBG_MR, "Registering STag 0x%08X, index = 0x%08X\n", - stag, stag_index); - - opcode = NES_CQP_ALLOCATE_STAG | NES_CQP_STAG_VA_TO | NES_CQP_STAG_MR; - - if (nesfmr->nesmr.pbl_4k == 1) - opcode |= NES_CQP_STAG_PBL_BLK_SIZE; - - if (ibmr_access_flags & IB_ACCESS_REMOTE_WRITE) { - opcode |= NES_CQP_STAG_RIGHTS_REMOTE_WRITE | - NES_CQP_STAG_RIGHTS_LOCAL_WRITE | NES_CQP_STAG_REM_ACC_EN; - nesfmr->access_rights |= - NES_CQP_STAG_RIGHTS_REMOTE_WRITE | NES_CQP_STAG_RIGHTS_LOCAL_WRITE | - NES_CQP_STAG_REM_ACC_EN; + spin_lock_irqsave(&nesadapter->pbl_lock, flags); + if (nesadapter->free_4kpbl > 0) { + nesadapter->free_4kpbl--; + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + } else { + /* No 4kpbl's available: */ + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + nes_debug(NES_DBG_MR, "Out of Pbls\n"); + nes_free_cqp_request(nesdev, cqp_request); + return -ENOMEM; } - if (ibmr_access_flags & IB_ACCESS_REMOTE_READ) { - opcode |= NES_CQP_STAG_RIGHTS_REMOTE_READ | - NES_CQP_STAG_RIGHTS_LOCAL_READ | NES_CQP_STAG_REM_ACC_EN; - nesfmr->access_rights |= - NES_CQP_STAG_RIGHTS_REMOTE_READ | NES_CQP_STAG_RIGHTS_LOCAL_READ | - NES_CQP_STAG_REM_ACC_EN; - } + opcode = NES_CQP_ALLOCATE_STAG | NES_CQP_STAG_MR | + NES_CQP_STAG_PBL_BLK_SIZE | NES_CQP_STAG_VA_TO | + NES_CQP_STAG_REM_ACC_EN; + /* + * The current OFED API does not support the zero based TO option. + * If added then need to changed the NES_CQP_STAG_VA* option. Also, + * the API does not support that ability to have the MR set for local + * access only when created and not allow the SQ op to override. Given + * this the remote enable must be set here. + */ nes_fill_init_cqp_wqe(cqp_wqe, nesdev); set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_OPCODE_IDX, opcode); - set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_LEN_HIGH_PD_IDX, (nespd->pd_id & 0x00007fff)); - set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, stag); + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PBL_BLK_COUNT_IDX, 1); - cqp_wqe->wqe_words[NES_CQP_STAG_WQE_PBL_BLK_COUNT_IDX] = - cpu_to_le32((nesfmr->nesmr.pbls_used>1) ? - (nesfmr->nesmr.pbls_used-1) : nesfmr->nesmr.pbls_used); + cqp_wqe->wqe_words[NES_CQP_STAG_WQE_LEN_HIGH_PD_IDX] = + cpu_to_le32((u32)(region_length >> 8) & 0xff000000); + cqp_wqe->wqe_words[NES_CQP_STAG_WQE_LEN_HIGH_PD_IDX] |= + cpu_to_le32(nespd->pd_id & 0x00007fff); + + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, stag); + set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_VA_LOW_IDX, 0); + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_LEN_LOW_IDX, 0); + set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PA_LOW_IDX, 0); + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PBL_LEN_IDX, (page_count * 8)); + cqp_wqe->wqe_words[NES_CQP_WQE_OPCODE_IDX] |= cpu_to_le32(NES_CQP_STAG_PBL_BLK_SIZE); + barrier(); atomic_set(&cqp_request->refcount, 2); nes_post_cqp_request(nesdev, cqp_request); /* Wait for CQP */ - ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0), - NES_EVENT_TIMEOUT); - nes_debug(NES_DBG_MR, "Register STag 0x%08X completed, wait_event_timeout ret = %u," - " CQP Major:Minor codes = 0x%04X:0x%04X.\n", - stag, ret, cqp_request->major_code, cqp_request->minor_code); - - if ((!ret) || (cqp_request->major_code)) { - nes_put_cqp_request(nesdev, cqp_request); - ret = (!ret) ? -ETIME : -EIO; - goto failed_leaf_vpbl_pages_alloc; - } + ret = wait_event_timeout(cqp_request->waitq, + (0 != cqp_request->request_done), + NES_EVENT_TIMEOUT); + + nes_debug(NES_DBG_MR, "Allocate STag 0x%08X completed, " + "wait_event_timeout ret = %u, CQP Major:Minor codes = " + "0x%04X:0x%04X.\n", stag, ret, cqp_request->major_code, + cqp_request->minor_code); + major_code = cqp_request->major_code; nes_put_cqp_request(nesdev, cqp_request); - nesfmr->nesmr.ibfmr.lkey = stag; - nesfmr->nesmr.ibfmr.rkey = stag; - nesfmr->attr = *ibfmr_attr; - - return &nesfmr->nesmr.ibfmr; - - failed_leaf_vpbl_pages_alloc: - /* unroll all allocated pages */ - for (i=0; ileaf_pbl_cnt; i++) { - if (nesfmr->root_vpbl.leaf_vpbl[i].pbl_vbase) { - pci_free_consistent(nesdev->pcidev, 4096, nesfmr->root_vpbl.leaf_vpbl[i].pbl_vbase, - nesfmr->root_vpbl.leaf_vpbl[i].pbl_pbase); - } - } - if (nesfmr->root_vpbl.leaf_vpbl) - kfree(nesfmr->root_vpbl.leaf_vpbl); - - failed_leaf_vpbl_alloc: - if (nesfmr->leaf_pbl_cnt == 0) { - if (nesfmr->root_vpbl.pbl_vbase) - pci_free_consistent(nesdev->pcidev, 4096, nesfmr->root_vpbl.pbl_vbase, - nesfmr->root_vpbl.pbl_pbase); - } else - pci_free_consistent(nesdev->pcidev, 8192, nesfmr->root_vpbl.pbl_vbase, - nesfmr->root_vpbl.pbl_pbase); - failed_vpbl_alloc: - if (nesfmr->nesmr.pbls_used != 0) { + if (!ret || major_code) { spin_lock_irqsave(&nesadapter->pbl_lock, flags); - if (nesfmr->nesmr.pbl_4k) - nesadapter->free_4kpbl += nesfmr->nesmr.pbls_used; - else - nesadapter->free_256pbl += nesfmr->nesmr.pbls_used; + nesadapter->free_4kpbl++; spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); } -failed_vpbl_avail: - kfree(nesfmr); - - failed_fmr_alloc: - nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index); - - failed_resource_alloc: - return ERR_PTR(ret); + if (!ret) + return -ETIME; + else if (major_code) + return -EIO; + return 0; } - -/** - * nes_dealloc_fmr +/* + * nes_alloc_fast_reg_mr */ -static int nes_dealloc_fmr(struct ib_fmr *ibfmr) +struct ib_mr *nes_alloc_fast_reg_mr(struct ib_pd *ibpd, int max_page_list_len) { - unsigned long flags; - struct nes_mr *nesmr = to_nesmr_from_ibfmr(ibfmr); - struct nes_fmr *nesfmr = to_nesfmr(nesmr); - struct nes_vnic *nesvnic = to_nesvnic(ibfmr->device); + struct nes_pd *nespd = to_nespd(ibpd); + struct nes_vnic *nesvnic = to_nesvnic(ibpd->device); struct nes_device *nesdev = nesvnic->nesdev; struct nes_adapter *nesadapter = nesdev->nesadapter; - int i = 0; - int rc; - /* free the resources */ - if (nesfmr->leaf_pbl_cnt == 0) { - /* single PBL case */ - if (nesfmr->root_vpbl.pbl_vbase) - pci_free_consistent(nesdev->pcidev, 4096, nesfmr->root_vpbl.pbl_vbase, - nesfmr->root_vpbl.pbl_pbase); - } else { - for (i = 0; i < nesfmr->leaf_pbl_cnt; i++) { - pci_free_consistent(nesdev->pcidev, 4096, nesfmr->root_vpbl.leaf_vpbl[i].pbl_vbase, - nesfmr->root_vpbl.leaf_vpbl[i].pbl_pbase); - } - kfree(nesfmr->root_vpbl.leaf_vpbl); - pci_free_consistent(nesdev->pcidev, 8192, nesfmr->root_vpbl.pbl_vbase, - nesfmr->root_vpbl.pbl_pbase); - } - nesmr->ibmw.device = ibfmr->device; - nesmr->ibmw.pd = ibfmr->pd; - nesmr->ibmw.rkey = ibfmr->rkey; - nesmr->ibmw.uobject = NULL; + u32 next_stag_index; + u8 stag_key = 0; + u32 driver_key = 0; + int err = 0; + u32 stag_index = 0; + struct nes_mr *nesmr; + u32 stag; + int ret; + struct ib_mr *ibmr; +/* + * Note: Set to always use a fixed length single page entry PBL. This is to allow + * for the fast_reg_mr operation to always know the size of the PBL. + */ + if (max_page_list_len > (NES_4K_PBL_CHUNK_SIZE / sizeof(u64))) + return ERR_PTR(-E2BIG); - rc = nes_dealloc_mw(&nesmr->ibmw); + get_random_bytes(&next_stag_index, sizeof(next_stag_index)); + stag_key = (u8)next_stag_index; + next_stag_index >>= 8; + next_stag_index %= nesadapter->max_mr; - if ((rc == 0) && (nesfmr->nesmr.pbls_used != 0)) { - spin_lock_irqsave(&nesadapter->pbl_lock, flags); - if (nesfmr->nesmr.pbl_4k) { - nesadapter->free_4kpbl += nesfmr->nesmr.pbls_used; - WARN_ON(nesadapter->free_4kpbl > nesadapter->max_4kpbl); - } else { - nesadapter->free_256pbl += nesfmr->nesmr.pbls_used; - WARN_ON(nesadapter->free_256pbl > nesadapter->max_256pbl); - } - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + err = nes_alloc_resource(nesadapter, nesadapter->allocated_mrs, + nesadapter->max_mr, &stag_index, + &next_stag_index); + if (err) + return ERR_PTR(err); + + nesmr = kzalloc(sizeof(*nesmr), GFP_KERNEL); + if (!nesmr) { + nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index); + return ERR_PTR(-ENOMEM); } - return rc; -} + stag = stag_index << 8; + stag |= driver_key; + stag += (u32)stag_key; + nes_debug(NES_DBG_MR, "Allocating STag 0x%08X index = 0x%08X\n", + stag, stag_index); -/** - * nes_map_phys_fmr + ret = alloc_fast_reg_mr(nesdev, nespd, stag, max_page_list_len); + + if (ret == 0) { + nesmr->ibmr.rkey = stag; + nesmr->ibmr.lkey = stag; + nesmr->mode = IWNES_MEMREG_TYPE_FMEM; + ibmr = &nesmr->ibmr; + } else { + kfree(nesmr); + nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index); + ibmr = ERR_PTR(-ENOMEM); + } + return ibmr; +} + +/* + * nes_alloc_fast_reg_page_list */ -static int nes_map_phys_fmr(struct ib_fmr *ibfmr, u64 *page_list, - int list_len, u64 iova) +static struct ib_fast_reg_page_list *nes_alloc_fast_reg_page_list( + struct ib_device *ibdev, + int page_list_len) { - return 0; -} + struct nes_vnic *nesvnic = to_nesvnic(ibdev); + struct nes_device *nesdev = nesvnic->nesdev; + struct ib_fast_reg_page_list *pifrpl; + struct nes_ib_fast_reg_page_list *pnesfrpl; + if (page_list_len > (NES_4K_PBL_CHUNK_SIZE / sizeof(u64))) + return ERR_PTR(-E2BIG); + /* + * Allocate the ib_fast_reg_page_list structure, the + * nes_fast_bpl structure, and the PLB table. + */ + pnesfrpl = kmalloc(sizeof(struct nes_ib_fast_reg_page_list) + + page_list_len * sizeof(u64), GFP_KERNEL); + + if (!pnesfrpl) + return ERR_PTR(-ENOMEM); -/** - * nes_unmap_frm + pifrpl = &pnesfrpl->ibfrpl; + pifrpl->page_list = &pnesfrpl->pbl; + pifrpl->max_page_list_len = page_list_len; + /* + * Allocate the WQE PBL + */ + pnesfrpl->nes_wqe_pbl.kva = pci_alloc_consistent(nesdev->pcidev, + page_list_len * sizeof(u64), + &pnesfrpl->nes_wqe_pbl.paddr); + + if (!pnesfrpl->nes_wqe_pbl.kva) { + kfree(pnesfrpl); + return ERR_PTR(-ENOMEM); + } + nes_debug(NES_DBG_MR, "nes_alloc_fast_reg_pbl: nes_frpl = %p, " + "ibfrpl = %p, ibfrpl.page_list = %p, pbl.kva = %p, " + "pbl.paddr= %p\n", pnesfrpl, &pnesfrpl->ibfrpl, + pnesfrpl->ibfrpl.page_list, pnesfrpl->nes_wqe_pbl.kva, + (void *)pnesfrpl->nes_wqe_pbl.paddr); + + return pifrpl; +} + +/* + * nes_free_fast_reg_page_list */ -static int nes_unmap_fmr(struct list_head *ibfmr_list) +static void nes_free_fast_reg_page_list(struct ib_fast_reg_page_list *pifrpl) { - return 0; + struct nes_vnic *nesvnic = to_nesvnic(pifrpl->device); + struct nes_device *nesdev = nesvnic->nesdev; + struct nes_ib_fast_reg_page_list *pnesfrpl; + + pnesfrpl = container_of(pifrpl, struct nes_ib_fast_reg_page_list, ibfrpl); + /* + * Free the WQE PBL. + */ + pci_free_consistent(nesdev->pcidev, + pifrpl->max_page_list_len * sizeof(u64), + pnesfrpl->nes_wqe_pbl.kva, + pnesfrpl->nes_wqe_pbl.paddr); + /* + * Free the PBL structure + */ + kfree(pnesfrpl); } - - /** * nes_query_device */ @@ -3514,6 +3408,91 @@ static int nes_post_send(struct ib_qp *ibqp, struct ib_send_wr *ib_wr, NES_IWARP_SQ_LOCINV_WQE_INV_STAG_IDX, ib_wr->ex.invalidate_rkey); break; + case IB_WR_FAST_REG_MR: + { + int i; + int flags = ib_wr->wr.fast_reg.access_flags; + struct nes_ib_fast_reg_page_list *pnesfrpl = + container_of(ib_wr->wr.fast_reg.page_list, + struct nes_ib_fast_reg_page_list, + ibfrpl); + u64 *src_page_list = pnesfrpl->ibfrpl.page_list; + u64 *dst_page_list = pnesfrpl->nes_wqe_pbl.kva; + + if (ib_wr->wr.fast_reg.page_list_len > + (NES_4K_PBL_CHUNK_SIZE / sizeof(u64))) { + nes_debug(NES_DBG_IW_TX, "SQ_FMR: bad page_list_len\n"); + err = -EINVAL; + break; + } + wqe_misc = NES_IWARP_SQ_OP_FAST_REG; + set_wqe_64bit_value(wqe->wqe_words, + NES_IWARP_SQ_FMR_WQE_VA_FBO_LOW_IDX, + ib_wr->wr.fast_reg.iova_start); + set_wqe_32bit_value(wqe->wqe_words, + NES_IWARP_SQ_FMR_WQE_LENGTH_LOW_IDX, + ib_wr->wr.fast_reg.length); + set_wqe_32bit_value(wqe->wqe_words, + NES_IWARP_SQ_FMR_WQE_MR_STAG_IDX, + ib_wr->wr.fast_reg.rkey); + /* Set page size: */ + if (ib_wr->wr.fast_reg.page_shift == 12) { + wqe_misc |= NES_IWARP_SQ_FMR_WQE_PAGE_SIZE_4K; + } else if (ib_wr->wr.fast_reg.page_shift == 21) { + wqe_misc |= NES_IWARP_SQ_FMR_WQE_PAGE_SIZE_2M; + } else { + nes_debug(NES_DBG_IW_TX, "Invalid page shift," + " ib_wr=%u, max=1\n", ib_wr->num_sge); + err = -EINVAL; + break; + } + /* Set access_flags */ + wqe_misc |= NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_LOCAL_READ; + if (flags & IB_ACCESS_LOCAL_WRITE) + wqe_misc |= NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_LOCAL_WRITE; + + if (flags & IB_ACCESS_REMOTE_WRITE) + wqe_misc |= NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_REMOTE_WRITE; + + if (flags & IB_ACCESS_REMOTE_READ) + wqe_misc |= NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_REMOTE_READ; + + if (flags & IB_ACCESS_MW_BIND) + wqe_misc |= NES_IWARP_SQ_FMR_WQE_RIGHTS_ENABLE_WINDOW_BIND; + + /* Fill in PBL info: */ + if (ib_wr->wr.fast_reg.page_list_len > + pnesfrpl->ibfrpl.max_page_list_len) { + nes_debug(NES_DBG_IW_TX, "Invalid page list length," + " ib_wr=%p, value=%u, max=%u\n", + ib_wr, ib_wr->wr.fast_reg.page_list_len, + pnesfrpl->ibfrpl.max_page_list_len); + err = -EINVAL; + break; + } + + set_wqe_64bit_value(wqe->wqe_words, + NES_IWARP_SQ_FMR_WQE_PBL_ADDR_LOW_IDX, + pnesfrpl->nes_wqe_pbl.paddr); + + set_wqe_32bit_value(wqe->wqe_words, + NES_IWARP_SQ_FMR_WQE_PBL_LENGTH_IDX, + ib_wr->wr.fast_reg.page_list_len * 8); + + for (i = 0; i < ib_wr->wr.fast_reg.page_list_len; i++) + dst_page_list[i] = cpu_to_le64(src_page_list[i]); + + nes_debug(NES_DBG_IW_TX, "SQ_FMR: iova_start: %p, " + "length: %d, rkey: %0x, pgl_paddr: %p, " + "page_list_len: %u\n", + (void *)ib_wr->wr.fast_reg.iova_start, + ib_wr->wr.fast_reg.length, + ib_wr->wr.fast_reg.rkey, + (void *)pnesfrpl->nes_wqe_pbl.paddr, + ib_wr->wr.fast_reg.page_list_len, + wqe_misc); + break; + } default: /* error */ err = -EINVAL; @@ -3752,6 +3731,9 @@ static int nes_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) case NES_IWARP_SQ_OP_LOCINV: entry->opcode = IB_WR_LOCAL_INV; break; + case NES_IWARP_SQ_OP_FAST_REG: + entry->opcode = IB_WC_FAST_REG_MR; + break; } nesqp->hwqp.sq_tail = (wqe_index+1)&(nesqp->hwqp.sq_size - 1); @@ -3922,10 +3904,9 @@ struct nes_ib_device *nes_init_ofa_device(struct net_device *netdev) nesibdev->ibdev.dealloc_mw = nes_dealloc_mw; nesibdev->ibdev.bind_mw = nes_bind_mw; - nesibdev->ibdev.alloc_fmr = nes_alloc_fmr; - nesibdev->ibdev.unmap_fmr = nes_unmap_fmr; - nesibdev->ibdev.dealloc_fmr = nes_dealloc_fmr; - nesibdev->ibdev.map_phys_fmr = nes_map_phys_fmr; + nesibdev->ibdev.alloc_fast_reg_mr = nes_alloc_fast_reg_mr; + nesibdev->ibdev.alloc_fast_reg_page_list = nes_alloc_fast_reg_page_list; + nesibdev->ibdev.free_fast_reg_page_list = nes_free_fast_reg_page_list; nesibdev->ibdev.attach_mcast = nes_multicast_attach; nesibdev->ibdev.detach_mcast = nes_multicast_detach;