Message ID | 20240813090518.3252469-2-link@vivo.com (mailing list archive) |
---|---|
State | Not Applicable |
Headers |
Received: from sv.mirrors.kernel.org ([139.178.88.99]) by linuxtv.org with esmtps (TLS1.3) tls TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 (Exim 4.96) (envelope-from <linux-media+bounces-16179-patchwork=linuxtv.org@vger.kernel.org>) id 1sdnTy-0005IT-1v for patchwork@linuxtv.org; Tue, 13 Aug 2024 09:06:23 +0000 Received: from smtp.subspace.kernel.org (wormhole.subspace.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by sv.mirrors.kernel.org (Postfix) with ESMTPS id B1B3A28202E for <patchwork@linuxtv.org>; Tue, 13 Aug 2024 09:06:21 +0000 (UTC) Received: from localhost.localdomain (localhost.localdomain [127.0.0.1]) by smtp.subspace.kernel.org (Postfix) with ESMTP id AA1DF187348; Tue, 13 Aug 2024 09:05:49 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=vivo.com header.i=@vivo.com header.b="QABmRsCH" X-Original-To: linux-media@vger.kernel.org Received: from APC01-PSA-obe.outbound.protection.outlook.com (mail-psaapc01on2055.outbound.protection.outlook.com [40.107.255.55]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 821C817CA09; Tue, 13 Aug 2024 09:05:47 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=fail smtp.client-ip=40.107.255.55 ARC-Seal: i=2; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1723539949; cv=fail; b=WkhqcLSjZdFoKt5p+az+QceszU4Lm8gnhHeYTDXkX+l5azBno6JRtwcpoEI/iE+565Y2sCZatxVZcARBPue1JXCH0X42E+XqWRma/uJa+/dFpBs61lnytcLa/dIAiS/OsyTQlWV+rmDiqt00PMWlXd1yjiBCJTTpi0QVOM8CFk0= ARC-Message-Signature: i=2; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1723539949; c=relaxed/simple; bh=scfOv+PWvJejjDlKBnCQ2d4wLLpQaS8qxaRfhoC8pLU=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: Content-Type:MIME-Version; b=aH8gzjR2+ZR+AXprYkkYR4WnIu+BW3Yaia5JC1udRYFA0A7zwzltnkzouMBctsd/0MoadmSrc1AaXqWie3VBA9SKE444z4nWG2ZPE6AeBUJb5at9Tp+QcwZGhk54YNwLnAOF3M6lv9pt+CaqUmKhBazfx/5jzGwEhSOEuiZXw3k= ARC-Authentication-Results: i=2; smtp.subspace.kernel.org; dmarc=pass (p=quarantine dis=none) header.from=vivo.com; spf=pass smtp.mailfrom=vivo.com; dkim=pass (2048-bit key) header.d=vivo.com header.i=@vivo.com header.b=QABmRsCH; arc=fail smtp.client-ip=40.107.255.55 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=quarantine dis=none) header.from=vivo.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=vivo.com ARC-Seal: i=1; a=rsa-sha256; s=arcselector10001; d=microsoft.com; cv=none; b=LG9eTygCQeq/sOk5rTPfTdGhno93PFD+fnfYa3tVCIp5uKziW58F1b6q7WxTwZ8MefAp0uP2GjsZZS7dMtUAz3/h8OFWEBfOqZ6+QHkx+Aaqnxg37p55yPdF1CngK7ltefwZQfS1HgoHUZ+JFQVq+tDtBkqYtisPpPI8NvZfxbK6I5za89aHiSWDfAnv+mbysfobuGib3Mowugi2+612w0PTBY6i/AZ2Hib2aNa5DWThOg4dGhTgk9SZo92bLb/F1hSlpNuv8/jloJ1OixVT6KGa6JQ1j+O5Y4yDE5GDVB9KoAoTggu1VDJeGcfsu0fN/whmrI/FwB4HR48OHcEJCQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector10001; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=rg6KYKgpaSeM55sTXqXgdPKjcJnZRfusdSSul92Ad0s=; b=JWkDCFNPxoUlZGYuKPkxcARgxm7+57zOwlvF07pN1CevMggaE2H6YLxk7yJZrSpqb+gAPo6qLl9EHC3qGWgBPgrpceuO9EgOveNuWkcUZZHxEEt9mq2OF6cDghX22ZFqVX0NX27omGkZ0F6rCUDMuDqAl6XrmRAoFxhj7r+maZKD8tiJ6JiepaxxGCNgm/P/IlVV/wwaJk1F6mMo7CSDNJEyoQPWjAwsGiWRCgzcSMCfazGBwyqQLyJUdldq6Mid2d8UL4CCt2dU1eg7mb/NUq2X+MR3HjbuxhPySb7GGsIirapv1xPSqkKDVrivUvOMPTac35SJelQBkkYA6UoffQ== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=vivo.com; dmarc=pass action=none header.from=vivo.com; dkim=pass header.d=vivo.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=vivo.com; s=selector2; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=rg6KYKgpaSeM55sTXqXgdPKjcJnZRfusdSSul92Ad0s=; b=QABmRsCHTS82EhXw0tPBbYONtqL9G3fO4tu5QwiOqYfq7GMzb9J/MT5OcMDDKMplFJQ/Xr89H2jZRklhI02Zkk27PesSGdcJZ8gfiyrGNTlx9WsVhVfJB6OfVcLzexTwKE0+nNH4s1H39t6D0OKicKLtJcFfr9A5BD5SIfW/mVgNvMvjqW4m9RJvZlh/0yaDw8PvQyTM6Ew1STmrf9Bmqmo92FcxX7YB4dwCsfbvVuypCePxaWCx1ReUj++mQeMNxHvoKh4LYB478jPESaRILevDElko0ZU4I4NRAlXffoaPa2Qqgjw2cl0fMlJv8fRyYuPH+lb97uCr8ToErCoPQQ== Authentication-Results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=vivo.com; Received: from PUZPR06MB5676.apcprd06.prod.outlook.com (2603:1096:301:f8::10) by JH0PR06MB7198.apcprd06.prod.outlook.com (2603:1096:990:98::7) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.7849.20; Tue, 13 Aug 2024 09:05:37 +0000 Received: from PUZPR06MB5676.apcprd06.prod.outlook.com ([fe80::a00b:f422:ac44:636f]) by PUZPR06MB5676.apcprd06.prod.outlook.com ([fe80::a00b:f422:ac44:636f%4]) with mapi id 15.20.7849.019; Tue, 13 Aug 2024 09:05:37 +0000 From: Huan Yang <link@vivo.com> To: Gerd Hoffmann <kraxel@redhat.com>, Sumit Semwal <sumit.semwal@linaro.org>, =?utf-8?q?Christian_K=C3=B6nig?= <christian.koenig@amd.com>, dri-devel@lists.freedesktop.org, linux-media@vger.kernel.org, linaro-mm-sig@lists.linaro.org, linux-kernel@vger.kernel.org Cc: opensource.kernel@vivo.com, Huan Yang <link@vivo.com>, Vivek Kasireddy <vivek.kasireddy@intel.com> Subject: [PATCH v3 1/5] udmabuf: cancel mmap page fault, direct map it Date: Tue, 13 Aug 2024 17:05:07 +0800 Message-ID: <20240813090518.3252469-2-link@vivo.com> X-Mailer: git-send-email 2.45.2 In-Reply-To: <20240813090518.3252469-1-link@vivo.com> References: <20240813090518.3252469-1-link@vivo.com> Content-Transfer-Encoding: 8bit Content-Type: text/plain X-ClientProxiedBy: SI2PR01CA0041.apcprd01.prod.exchangelabs.com (2603:1096:4:193::15) To PUZPR06MB5676.apcprd06.prod.outlook.com (2603:1096:301:f8::10) Precedence: bulk X-Mailing-List: linux-media@vger.kernel.org List-Id: <linux-media.vger.kernel.org> List-Subscribe: <mailto:linux-media+subscribe@vger.kernel.org> List-Unsubscribe: <mailto:linux-media+unsubscribe@vger.kernel.org> MIME-Version: 1.0 X-MS-PublicTrafficType: Email X-MS-TrafficTypeDiagnostic: PUZPR06MB5676:EE_|JH0PR06MB7198:EE_ X-MS-Office365-Filtering-Correlation-Id: 16218051-0a07-4a6e-b125-08dcbb7718c7 X-MS-Exchange-SenderADCheck: 1 X-MS-Exchange-AntiSpam-Relay: 0 X-Microsoft-Antispam: BCL:0;ARA:13230040|366016|376014|52116014|1800799024|38350700014; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:PUZPR06MB5676.apcprd06.prod.outlook.com;PTR:;CAT:NONE;SFS:(13230040)(366016)(376014)(52116014)(1800799024)(38350700014);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: 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 X-OriginatorOrg: vivo.com X-MS-Exchange-CrossTenant-Network-Message-Id: 16218051-0a07-4a6e-b125-08dcbb7718c7 X-MS-Exchange-CrossTenant-AuthSource: PUZPR06MB5676.apcprd06.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 13 Aug 2024 09:05:37.2723 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 923e42dc-48d5-4cbe-b582-1a797a6412ed X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: ITe3cNikEuCiTXnm65tG/7DB17pDCw6BcMSgt7PPj/bRpe0xnYaJDBzaPO0+ndB0apmrqSRYZBS4nC39tZ+rYg== X-MS-Exchange-Transport-CrossTenantHeadersStamped: JH0PR06MB7198 X-LSpam-Score: -8.6 (--------) X-LSpam-Report: No, score=-8.6 required=5.0 tests=ARC_SIGNED=0.001,ARC_VALID=-0.1,BAYES_00=-1.9,DKIM_SIGNED=0.1,DKIM_VALID=-0.1,DKIM_VALID_AU=-0.1,DMARC_PASS=-0.001,HEADER_FROM_DIFFERENT_DOMAINS=0.5,MAILING_LIST_MULTI=-1,RCVD_IN_DNSWL_MED=-2.3,RCVD_IN_VALIDITY_CERTIFIED=-3,RCVD_IN_VALIDITY_RPBL=1.31,RCVD_IN_VALIDITY_SAFE=-2,SPF_HELO_NONE=0.001,SPF_PASS=-0.001 autolearn=ham autolearn_force=no |
Series |
udmbuf bug fix and some improvements
|
|
Commit Message
Huan Yang
Aug. 13, 2024, 9:05 a.m. UTC
The current udmabuf mmap uses a page fault to populate the vma.
However, the current udmabuf has already obtained and pinned the folio
upon completion of the creation.This means that the physical memory has
already been acquired, rather than being accessed dynamically. The
current page fault method only saves some page table memory.
As a result, the page fault has lost its purpose as a demanding
page. Due to the fact that page fault requires trapping into kernel mode
and filling in when accessing the corresponding virtual address in mmap,
when creating a large size udmabuf, this represents a considerable
overhead.
The current patch removes the page fault method of mmap and
instead fills pfn directly when mmap is triggered.
Signed-off-by: Huan Yang <link@vivo.com>
Suggested-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
---
drivers/dma-buf/udmabuf.c | 37 +++++++++++++++----------------------
1 file changed, 15 insertions(+), 22 deletions(-)
Comments
Hi Huan, > > The current udmabuf mmap uses a page fault to populate the vma. > > However, the current udmabuf has already obtained and pinned the folio > upon completion of the creation.This means that the physical memory has > already been acquired, rather than being accessed dynamically. The > current page fault method only saves some page table memory. > > As a result, the page fault has lost its purpose as a demanding > page. Due to the fact that page fault requires trapping into kernel mode > and filling in when accessing the corresponding virtual address in mmap, > when creating a large size udmabuf, this represents a considerable > overhead. > > The current patch removes the page fault method of mmap and > instead fills pfn directly when mmap is triggered. > > Signed-off-by: Huan Yang <link@vivo.com> > Suggested-by: Vivek Kasireddy <vivek.kasireddy@intel.com> > --- > drivers/dma-buf/udmabuf.c | 37 +++++++++++++++---------------------- > 1 file changed, 15 insertions(+), 22 deletions(-) > > diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c > index 047c3cd2ceff..d39f9e1cd532 100644 > --- a/drivers/dma-buf/udmabuf.c > +++ b/drivers/dma-buf/udmabuf.c > @@ -38,36 +38,29 @@ struct udmabuf_folio { > struct list_head list; > }; > > -static vm_fault_t udmabuf_vm_fault(struct vm_fault *vmf) > -{ > - struct vm_area_struct *vma = vmf->vma; > - struct udmabuf *ubuf = vma->vm_private_data; > - pgoff_t pgoff = vmf->pgoff; > - unsigned long pfn; > - > - if (pgoff >= ubuf->pagecount) > - return VM_FAULT_SIGBUS; > - > - pfn = folio_pfn(ubuf->folios[pgoff]); > - pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; > - > - return vmf_insert_pfn(vma, vmf->address, pfn); > -} > - > -static const struct vm_operations_struct udmabuf_vm_ops = { > - .fault = udmabuf_vm_fault, > -}; So, what I was suggesting earlier is that it would be OK to populate the whole vma after first fault because userspace can simply call mmap() but choose not to use the returned pointer for various reasons. This is what Qemu's virtio-gpu module does and in this case we'd be unnecessarily populating the vma. Therefore, my request to you is to try to benchmark your userspace to see if there is a significant difference in performance when you populate the vma during mmap() vs doing it after first fault (which means moving the for loop to udmabuf_vm_fault()). > - > static int mmap_udmabuf(struct dma_buf *buf, struct vm_area_struct > *vma) > { > struct udmabuf *ubuf = buf->priv; > + unsigned long addr; > + unsigned long end; > + unsigned long pgoff; > + int ret; Looks like ret's type needs to be vm_fault_t. > > if ((vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) == 0) > return -EINVAL; > > - vma->vm_ops = &udmabuf_vm_ops; > - vma->vm_private_data = ubuf; > vm_flags_set(vma, VM_PFNMAP | VM_DONTEXPAND | > VM_DONTDUMP); > + > + for (pgoff = vma->vm_pgoff, end = vma->vm_end, addr = vma- I think initializing these variables above at the declaration time looks better than initializing them in the for loop, IMO. Thanks, Vivek > >vm_start; > + addr < end; pgoff++, addr += PAGE_SIZE) { > + unsigned long pfn = folio_pfn(ubuf->folios[pgoff]); > + > + pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; > + ret = vmf_insert_pfn(vma, addr, pfn); > + if (ret & VM_FAULT_ERROR) > + return vm_fault_to_errno(ret, 0); > + } > + > return 0; > } > > -- > 2.45.2
在 2024/8/17 8:53, Kasireddy, Vivek 写道: > Hi Huan, > >> The current udmabuf mmap uses a page fault to populate the vma. >> >> However, the current udmabuf has already obtained and pinned the folio >> upon completion of the creation.This means that the physical memory has >> already been acquired, rather than being accessed dynamically. The >> current page fault method only saves some page table memory. >> >> As a result, the page fault has lost its purpose as a demanding >> page. Due to the fact that page fault requires trapping into kernel mode >> and filling in when accessing the corresponding virtual address in mmap, >> when creating a large size udmabuf, this represents a considerable >> overhead. >> >> The current patch removes the page fault method of mmap and >> instead fills pfn directly when mmap is triggered. >> >> Signed-off-by: Huan Yang <link@vivo.com> >> Suggested-by: Vivek Kasireddy <vivek.kasireddy@intel.com> >> --- >> drivers/dma-buf/udmabuf.c | 37 +++++++++++++++---------------------- >> 1 file changed, 15 insertions(+), 22 deletions(-) >> >> diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c >> index 047c3cd2ceff..d39f9e1cd532 100644 >> --- a/drivers/dma-buf/udmabuf.c >> +++ b/drivers/dma-buf/udmabuf.c >> @@ -38,36 +38,29 @@ struct udmabuf_folio { >> struct list_head list; >> }; >> >> -static vm_fault_t udmabuf_vm_fault(struct vm_fault *vmf) >> -{ >> - struct vm_area_struct *vma = vmf->vma; >> - struct udmabuf *ubuf = vma->vm_private_data; >> - pgoff_t pgoff = vmf->pgoff; >> - unsigned long pfn; >> - >> - if (pgoff >= ubuf->pagecount) >> - return VM_FAULT_SIGBUS; >> - >> - pfn = folio_pfn(ubuf->folios[pgoff]); >> - pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; >> - >> - return vmf_insert_pfn(vma, vmf->address, pfn); >> -} >> - >> -static const struct vm_operations_struct udmabuf_vm_ops = { >> - .fault = udmabuf_vm_fault, >> -}; > So, what I was suggesting earlier is that it would be OK to populate the whole > vma after first fault because userspace can simply call mmap() but choose not > to use the returned pointer for various reasons. This is what Qemu's virtio-gpu > module does and in this case we'd be unnecessarily populating the vma. I may get your point. Fill pgtable when access is better than fill when invoke mmap? This is reasonable. And I'll try to test it too. IMO, there won't be much of a difference in performance. > > Therefore, my request to you is to try to benchmark your userspace to see if > there is a significant difference in performance when you populate the vma > during mmap() vs doing it after first fault (which means moving the for loop > to udmabuf_vm_fault()). > >> - >> static int mmap_udmabuf(struct dma_buf *buf, struct vm_area_struct >> *vma) >> { >> struct udmabuf *ubuf = buf->priv; >> + unsigned long addr; >> + unsigned long end; >> + unsigned long pgoff; >> + int ret; > Looks like ret's type needs to be vm_fault_t. > >> if ((vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) == 0) >> return -EINVAL; >> >> - vma->vm_ops = &udmabuf_vm_ops; >> - vma->vm_private_data = ubuf; >> vm_flags_set(vma, VM_PFNMAP | VM_DONTEXPAND | >> VM_DONTDUMP); >> + >> + for (pgoff = vma->vm_pgoff, end = vma->vm_end, addr = vma- > I think initializing these variables above at the declaration time looks better > than initializing them in the for loop, IMO. Yes, even though initializing in the loop declaration can better indicate which variables the loop needs, here it makes the loop declaration too long. I'll change it in next version. Thanks. > > Thanks, > Vivek > >>> vm_start; >> + addr < end; pgoff++, addr += PAGE_SIZE) { >> + unsigned long pfn = folio_pfn(ubuf->folios[pgoff]); >> + >> + pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; >> + ret = vmf_insert_pfn(vma, addr, pfn); >> + if (ret & VM_FAULT_ERROR) >> + return vm_fault_to_errno(ret, 0); >> + } >> + >> return 0; >> } >> >> -- >> 2.45.2
diff --git a/drivers/dma-buf/udmabuf.c b/drivers/dma-buf/udmabuf.c index 047c3cd2ceff..d39f9e1cd532 100644 --- a/drivers/dma-buf/udmabuf.c +++ b/drivers/dma-buf/udmabuf.c @@ -38,36 +38,29 @@ struct udmabuf_folio { struct list_head list; }; -static vm_fault_t udmabuf_vm_fault(struct vm_fault *vmf) -{ - struct vm_area_struct *vma = vmf->vma; - struct udmabuf *ubuf = vma->vm_private_data; - pgoff_t pgoff = vmf->pgoff; - unsigned long pfn; - - if (pgoff >= ubuf->pagecount) - return VM_FAULT_SIGBUS; - - pfn = folio_pfn(ubuf->folios[pgoff]); - pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; - - return vmf_insert_pfn(vma, vmf->address, pfn); -} - -static const struct vm_operations_struct udmabuf_vm_ops = { - .fault = udmabuf_vm_fault, -}; - static int mmap_udmabuf(struct dma_buf *buf, struct vm_area_struct *vma) { struct udmabuf *ubuf = buf->priv; + unsigned long addr; + unsigned long end; + unsigned long pgoff; + int ret; if ((vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) == 0) return -EINVAL; - vma->vm_ops = &udmabuf_vm_ops; - vma->vm_private_data = ubuf; vm_flags_set(vma, VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP); + + for (pgoff = vma->vm_pgoff, end = vma->vm_end, addr = vma->vm_start; + addr < end; pgoff++, addr += PAGE_SIZE) { + unsigned long pfn = folio_pfn(ubuf->folios[pgoff]); + + pfn += ubuf->offsets[pgoff] >> PAGE_SHIFT; + ret = vmf_insert_pfn(vma, addr, pfn); + if (ret & VM_FAULT_ERROR) + return vm_fault_to_errno(ret, 0); + } + return 0; }