From patchwork Wed Apr 26 09:28:02 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Cai Huoqing X-Patchwork-Id: 87777 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b0ea:0:b0:3b6:4342:cba0 with SMTP id b10csp121016vqo; Wed, 26 Apr 2023 02:53:09 -0700 (PDT) X-Google-Smtp-Source: AKy350aRW2TjJs+fGNLONvvZu/kF7ZgTrtjp0Rn8Shaq8jrWJPuCnW+p40OFEIbpwGhQyk7uKlkk X-Received: by 2002:a17:902:ecc8:b0:1a6:f5d5:b809 with SMTP id a8-20020a170902ecc800b001a6f5d5b809mr26559876plh.1.1682502789357; Wed, 26 Apr 2023 02:53:09 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1682502789; cv=none; d=google.com; s=arc-20160816; b=iuIPohPeiMmLsnpZHW5l4CA7/RcxB3YJ24A2giSGXeaD4pqIhmq0WzXp0WwtzxF/TS nwcXYRLB6rE8OjCIw5jahWUnVZ1NJVZIKqnEHgHzRZ8mvkgIToSfBL7Zki8uf0rZdUSs zNUejyF1gGXXesC+Hbd87ce3NfJScqQP9I38LEw+bp+XjAWiwdBOzZ8Hnq5xIL0CY773 tM+U9JCl0yQ4ivRZ3Cy5GAKEFq380kLiX23RSkZC4gJl+sE1PlmLbrqmSrJVskpep/Fm cZTRKMyjVsRTGtY3UqIemQZB+sfg+xZ7qNrYNpTENdrHxE5lZkkjj37zLpI4ggTk8jIQ 3sng== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :message-id:date:subject:cc:to:from:dkim-signature; bh=ypU1MXYO6v9EAMmNXSPncGg9hMlm/McPdVQ+roObBUg=; b=s+6TIL8CocntnS8HPi62VVuQ63P0oGBKcMDDS04ZypVKDK0x6NVLdDPNS2Rc2KTBCj /jZ9dNL6ZOSL5cWYGlNLZI566gc5rr0zOCaZTkiqrfEQ/tPkVQ87dg4R+dJNcmt8WbS8 Qt/QzBwzomPYl2NvTjMpJUaiAonFeGZJYhaprUXDv9Em548SqRtniQthvW2cKD5pTxDq RaZ1rGwJSo+oPAjNBEJZ38DspmFZNCX7U+zomnc5EW6ju1WWopvcMeJiPEIftz5JtMUW XqQ+gdCuc503Q3HuQdELbF87ZgjzqGgGMlIp/9iLxytjqmO8XyMvH5ngcgL0nf2V9dpn qInw== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@linux.dev header.s=key1 header.b=QtC18PWo; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=linux.dev Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id x4-20020a170902a38400b001a92c3eb083si15424554pla.279.2023.04.26.02.52.56; Wed, 26 Apr 2023 02:53:09 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; dkim=pass header.i=@linux.dev header.s=key1 header.b=QtC18PWo; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=linux.dev Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S240356AbjDZJ2o (ORCPT + 99 others); Wed, 26 Apr 2023 05:28:44 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:53218 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240231AbjDZJ2Z (ORCPT ); Wed, 26 Apr 2023 05:28:25 -0400 Received: from out-53.mta1.migadu.com (out-53.mta1.migadu.com [95.215.58.53]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id C3ED949D4 for ; Wed, 26 Apr 2023 02:28:19 -0700 (PDT) X-Report-Abuse: Please report any abuse attempt to abuse@migadu.com and include these headers. DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linux.dev; s=key1; t=1682501297; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version: content-transfer-encoding:content-transfer-encoding; bh=ypU1MXYO6v9EAMmNXSPncGg9hMlm/McPdVQ+roObBUg=; b=QtC18PWo301eE6Xmcf5IlUsHQuW5tnzOXGWpw/VBmXJg7eOgNTWotMGC5/sI2oyw6nbc3F 1x+E9186eYJtxS/KGPFG4lUxPmtislNMN315rmSVSYL01lgATWO+wembhoZr8+IWdYc0PV EJpAf31qcZT8R4Ae5+/sci2vzARocIw= From: Cai Huoqing To: cai.huoqing@linux.dev Cc: Oded Gabbay , Ohad Sharabi , Greg Kroah-Hartman , dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org Subject: [PATCH] accel/habanalabs: Make use of rhashtable Date: Wed, 26 Apr 2023 17:28:02 +0800 Message-Id: <20230426092813.44635-1-cai.huoqing@linux.dev> MIME-Version: 1.0 X-Migadu-Flow: FLOW_OUT X-Spam-Status: No, score=-2.1 required=5.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,SPF_HELO_NONE,SPF_PASS, TO_EQ_FM_DIRECT_MX,T_SCC_BODY_TEXT_LINE,URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1764232044865780953?= X-GMAIL-MSGID: =?utf-8?q?1764232044865780953?= Using rhashtable to accelerate the search for userptr by address, instead of using a list. Preferably, the lookup complexity of a hash table is O(1). This patch will speedup the method hl_userptr_is_pinned by rhashtable_lookup_fast. Signed-off-by: Cai Huoqing --- .../habanalabs/common/command_submission.c | 16 ++++++--- drivers/accel/habanalabs/common/habanalabs.h | 19 +++++----- drivers/accel/habanalabs/common/memory.c | 35 +++++++++++++------ drivers/accel/habanalabs/gaudi/gaudi.c | 16 +++++---- drivers/accel/habanalabs/goya/goya.c | 14 +++++--- 5 files changed, 66 insertions(+), 34 deletions(-) diff --git a/drivers/accel/habanalabs/common/command_submission.c b/drivers/accel/habanalabs/common/command_submission.c index af9d2e22c6e7..35c2ab934396 100644 --- a/drivers/accel/habanalabs/common/command_submission.c +++ b/drivers/accel/habanalabs/common/command_submission.c @@ -312,7 +312,7 @@ static int cs_parser(struct hl_fpriv *hpriv, struct hl_cs_job *job) parser.job_id = job->id; parser.hw_queue_id = job->hw_queue_id; - parser.job_userptr_list = &job->userptr_list; + parser.job_userptr_ht = &job->userptr_ht; parser.patched_cb = NULL; parser.user_cb = job->user_cb; parser.user_cb_size = job->user_cb_size; @@ -351,7 +351,7 @@ static void hl_complete_job(struct hl_device *hdev, struct hl_cs_job *job) struct hl_cs *cs = job->cs; if (is_cb_patched(hdev, job)) { - hl_userptr_delete_list(hdev, &job->userptr_list); + hl_userptr_delete_list(hdev, &job->userptr_ht); /* * We might arrive here from rollback and patched CB wasn't @@ -1284,6 +1284,7 @@ struct hl_cs_job *hl_cs_allocate_job(struct hl_device *hdev, enum hl_queue_type queue_type, bool is_kernel_allocated_cb) { struct hl_cs_job *job; + int rc; job = kzalloc(sizeof(*job), GFP_ATOMIC); if (!job) @@ -1296,13 +1297,20 @@ struct hl_cs_job *hl_cs_allocate_job(struct hl_device *hdev, job->queue_type = queue_type; job->is_kernel_allocated_cb = is_kernel_allocated_cb; - if (is_cb_patched(hdev, job)) - INIT_LIST_HEAD(&job->userptr_list); + if (is_cb_patched(hdev, job)) { + rc = rhashtable_init(&job->userptr_ht, &hl_userptr_rht_params); + if (rc) + goto free_job; + } if (job->queue_type == QUEUE_TYPE_EXT) INIT_WORK(&job->finish_work, job_wq_completion); return job; + +free_job: + kfree(job); + return NULL; } static enum hl_cs_type hl_cs_get_cs_type(u32 cs_type_flags) diff --git a/drivers/accel/habanalabs/common/habanalabs.h b/drivers/accel/habanalabs/common/habanalabs.h index eaae69a9f817..9c876d1480d2 100644 --- a/drivers/accel/habanalabs/common/habanalabs.h +++ b/drivers/accel/habanalabs/common/habanalabs.h @@ -19,6 +19,7 @@ #include #include #include +#include #include #include #include @@ -540,6 +541,8 @@ struct hl_hints_range { u64 end_addr; }; +extern const struct rhashtable_params hl_userptr_rht_params; + /** * struct asic_fixed_properties - ASIC specific immutable properties. * @hw_queues_props: H/W queues properties. @@ -1915,7 +1918,7 @@ struct hl_ctx_mgr { /** * struct hl_userptr - memory mapping chunk information * @vm_type: type of the VM. - * @job_node: linked-list node for hanging the object on the Job's list. + * @job_node: hashtable node for hanging the object on the Job's list. * @pages: pointer to struct page array * @npages: size of @pages array * @sgt: pointer to the scatter-gather table that holds the pages. @@ -1928,7 +1931,7 @@ struct hl_ctx_mgr { */ struct hl_userptr { enum vm_type vm_type; /* must be first */ - struct list_head job_node; + struct rhash_head job_node; struct page **pages; unsigned int npages; struct sg_table *sgt; @@ -2028,7 +2031,7 @@ struct hl_cs { * @patched_cb: in case of patching, this is internal CB which is submitted on * the queue instead of the CB we got from the IOCTL. * @finish_work: workqueue object to run when job is completed. - * @userptr_list: linked-list of userptr mappings that belong to this job and + * @userptr_ht: hashtable of userptr mappings that belong to this job and * wait for completion. * @debugfs_list: node in debugfs list of command submission jobs. * @refcount: reference counter for usage of the CS job. @@ -2056,7 +2059,7 @@ struct hl_cs_job { struct hl_cb *user_cb; struct hl_cb *patched_cb; struct work_struct finish_work; - struct list_head userptr_list; + struct rhashtable userptr_ht; struct list_head debugfs_list; struct kref refcount; enum hl_queue_type queue_type; @@ -2075,7 +2078,7 @@ struct hl_cs_job { * @user_cb: the CB we got from the user. * @patched_cb: in case of patching, this is internal CB which is submitted on * the queue instead of the CB we got from the IOCTL. - * @job_userptr_list: linked-list of userptr mappings that belong to the related + * @job_userptr_ht: hashtable of userptr mappings that belong to the related * job and wait for completion. * @cs_sequence: the sequence number of the related CS. * @queue_type: the type of the H/W queue this job is submitted to. @@ -2098,7 +2101,7 @@ struct hl_cs_job { struct hl_cs_parser { struct hl_cb *user_cb; struct hl_cb *patched_cb; - struct list_head *job_userptr_list; + struct rhashtable *job_userptr_ht; u64 cs_sequence; enum hl_queue_type queue_type; u32 ctx_id; @@ -3760,9 +3763,9 @@ int hl_pin_host_memory(struct hl_device *hdev, u64 addr, u64 size, struct hl_userptr *userptr); void hl_unpin_host_memory(struct hl_device *hdev, struct hl_userptr *userptr); void hl_userptr_delete_list(struct hl_device *hdev, - struct list_head *userptr_list); + struct rhashtable *userptr_ht); bool hl_userptr_is_pinned(struct hl_device *hdev, u64 addr, u32 size, - struct list_head *userptr_list, + struct rhashtable *userptr_ht, struct hl_userptr **userptr); int hl_mmu_init(struct hl_device *hdev); diff --git a/drivers/accel/habanalabs/common/memory.c b/drivers/accel/habanalabs/common/memory.c index a7b6a273ce21..e5e7912b3b34 100644 --- a/drivers/accel/habanalabs/common/memory.c +++ b/drivers/accel/habanalabs/common/memory.c @@ -23,6 +23,13 @@ MODULE_IMPORT_NS(DMA_BUF); #define MEM_HANDLE_INVALID ULONG_MAX +const struct rhashtable_params hl_userptr_rht_params = { + .head_offset = offsetof(struct hl_userptr, job_node), + .key_offset = offsetof(struct hl_userptr, addr), + .key_len = sizeof(u64), + .automatic_shrinking = true, +}; + static int allocate_timestamps_buffers(struct hl_fpriv *hpriv, struct hl_mem_in *args, u64 *handle); @@ -2483,7 +2490,6 @@ int hl_pin_host_memory(struct hl_device *hdev, u64 addr, u64 size, userptr->size = size; userptr->addr = addr; userptr->dma_mapped = false; - INIT_LIST_HEAD(&userptr->job_node); rc = get_user_memory(hdev, addr, size, npages, start, offset, userptr); @@ -2522,8 +2528,6 @@ void hl_unpin_host_memory(struct hl_device *hdev, struct hl_userptr *userptr) unpin_user_pages_dirty_lock(userptr->pages, userptr->npages, true); kvfree(userptr->pages); - list_del(&userptr->job_node); - sg_free_table(userptr->sgt); kfree(userptr->sgt); } @@ -2531,23 +2535,31 @@ void hl_unpin_host_memory(struct hl_device *hdev, struct hl_userptr *userptr) /** * hl_userptr_delete_list() - clear userptr list. * @hdev: pointer to the habanalabs device structure. - * @userptr_list: pointer to the list to clear. + * @userptr_ht: pointer to the hashtable to clear. * * This function does the following: * - Iterates over the list and unpins the host memory and frees the userptr * structure. */ void hl_userptr_delete_list(struct hl_device *hdev, - struct list_head *userptr_list) + struct rhashtable *userptr_ht) { - struct hl_userptr *userptr, *tmp; + struct hl_userptr *userptr; + struct rhashtable_iter hti; + struct rhash_head *pos; - list_for_each_entry_safe(userptr, tmp, userptr_list, job_node) { + rhashtable_walk_enter(userptr_ht, &hti); + rhashtable_walk_start(&hti); + while ((pos = rhashtable_walk_next(&hti))) { + if (PTR_ERR(pos) == -EAGAIN) + continue; + rhashtable_remove_fast(userptr_ht, hti.p, hl_userptr_rht_params); + userptr = rhashtable_walk_peek(&hti); hl_unpin_host_memory(hdev, userptr); kfree(userptr); } - INIT_LIST_HEAD(userptr_list); + rhashtable_destroy(userptr_ht); } /** @@ -2555,7 +2567,7 @@ void hl_userptr_delete_list(struct hl_device *hdev, * @hdev: pointer to the habanalabs device structure. * @addr: user address to check. * @size: user block size to check. - * @userptr_list: pointer to the list to clear. + * @userptr_ht: pointer to the hashtable to clear. * @userptr: pointer to userptr to check. * * This function does the following: @@ -2563,10 +2575,11 @@ void hl_userptr_delete_list(struct hl_device *hdev, * pinned. If so, returns true, otherwise returns false. */ bool hl_userptr_is_pinned(struct hl_device *hdev, u64 addr, - u32 size, struct list_head *userptr_list, + u32 size, struct rhashtable *userptr_ht, struct hl_userptr **userptr) { - list_for_each_entry((*userptr), userptr_list, job_node) { + (*userptr) = rhashtable_lookup_fast(userptr_ht, &addr, hl_userptr_rht_params); + if (*userptr) { if ((addr == (*userptr)->addr) && (size == (*userptr)->size)) return true; } diff --git a/drivers/accel/habanalabs/gaudi/gaudi.c b/drivers/accel/habanalabs/gaudi/gaudi.c index a29aa8f7b6f3..1e1433042413 100644 --- a/drivers/accel/habanalabs/gaudi/gaudi.c +++ b/drivers/accel/habanalabs/gaudi/gaudi.c @@ -1031,7 +1031,7 @@ static int _gaudi_init_tpc_mem(struct hl_device *hdev, } free_job: - hl_userptr_delete_list(hdev, &job->userptr_list); + hl_userptr_delete_list(hdev, &job->userptr_ht); hl_debugfs_remove_job(hdev, job); kfree(job); atomic_dec(&cb->cs_cnt); @@ -4901,7 +4901,7 @@ static int gaudi_pin_memory_before_cs(struct hl_device *hdev, int rc; if (hl_userptr_is_pinned(hdev, addr, le32_to_cpu(user_dma_pkt->tsize), - parser->job_userptr_list, &userptr)) + parser->job_userptr_ht, &userptr)) goto already_pinned; userptr = kzalloc(sizeof(*userptr), GFP_KERNEL); @@ -4913,7 +4913,10 @@ static int gaudi_pin_memory_before_cs(struct hl_device *hdev, if (rc) goto free_userptr; - list_add_tail(&userptr->job_node, parser->job_userptr_list); + rc = rhashtable_insert_fast(parser->job_userptr_ht, + &userptr->job_node, hl_userptr_rht_params); + if (rc) + goto unpin_memory; rc = hdev->asic_funcs->asic_dma_map_sgtable(hdev, userptr->sgt, dir); if (rc) { @@ -4931,7 +4934,8 @@ static int gaudi_pin_memory_before_cs(struct hl_device *hdev, return 0; unpin_memory: - list_del(&userptr->job_node); + rhashtable_remove_fast(parser->job_userptr_ht, + &userptr->job_node, hl_userptr_rht_params); hl_unpin_host_memory(hdev, userptr); free_userptr: kfree(userptr); @@ -5175,7 +5179,7 @@ static int gaudi_patch_dma_packet(struct hl_device *hdev, if ((!skip_host_mem_pin) && (!hl_userptr_is_pinned(hdev, addr, le32_to_cpu(user_dma_pkt->tsize), - parser->job_userptr_list, &userptr))) { + parser->job_userptr_ht, &userptr))) { dev_err(hdev->dev, "Userptr 0x%llx + 0x%x NOT mapped\n", addr, user_dma_pkt->tsize); return -EFAULT; @@ -5472,7 +5476,7 @@ static int gaudi_parse_cb_no_mmu(struct hl_device *hdev, free_userptr: if (rc) - hl_userptr_delete_list(hdev, parser->job_userptr_list); + hl_userptr_delete_list(hdev, parser->job_userptr_ht); return rc; } diff --git a/drivers/accel/habanalabs/goya/goya.c b/drivers/accel/habanalabs/goya/goya.c index fb0ac9df841a..bfcbb9e8b126 100644 --- a/drivers/accel/habanalabs/goya/goya.c +++ b/drivers/accel/habanalabs/goya/goya.c @@ -3347,7 +3347,7 @@ static int goya_pin_memory_before_cs(struct hl_device *hdev, int rc; if (hl_userptr_is_pinned(hdev, addr, le32_to_cpu(user_dma_pkt->tsize), - parser->job_userptr_list, &userptr)) + parser->job_userptr_ht, &userptr)) goto already_pinned; userptr = kzalloc(sizeof(*userptr), GFP_KERNEL); @@ -3359,7 +3359,10 @@ static int goya_pin_memory_before_cs(struct hl_device *hdev, if (rc) goto free_userptr; - list_add_tail(&userptr->job_node, parser->job_userptr_list); + rc = rhashtable_insert_fast(parser->job_userptr_ht, + &userptr->job_node, hl_userptr_rht_params); + if (rc) + goto unpin_memory; rc = hdev->asic_funcs->asic_dma_map_sgtable(hdev, userptr->sgt, dir); if (rc) { @@ -3377,7 +3380,8 @@ static int goya_pin_memory_before_cs(struct hl_device *hdev, return 0; unpin_memory: - list_del(&userptr->job_node); + rhashtable_remove_fast(parser->job_userptr_ht, + &userptr->job_node, hl_userptr_rht_params); hl_unpin_host_memory(hdev, userptr); free_userptr: kfree(userptr); @@ -3806,7 +3810,7 @@ static int goya_patch_dma_packet(struct hl_device *hdev, if ((!skip_host_mem_pin) && (hl_userptr_is_pinned(hdev, addr, le32_to_cpu(user_dma_pkt->tsize), - parser->job_userptr_list, &userptr) == false)) { + parser->job_userptr_ht, &userptr) == false)) { dev_err(hdev->dev, "Userptr 0x%llx + 0x%x NOT mapped\n", addr, user_dma_pkt->tsize); return -EFAULT; @@ -4104,7 +4108,7 @@ static int goya_parse_cb_no_mmu(struct hl_device *hdev, free_userptr: if (rc) - hl_userptr_delete_list(hdev, parser->job_userptr_list); + hl_userptr_delete_list(hdev, parser->job_userptr_ht); return rc; }