From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-5.5 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_AGENT_SANE_1 autolearn=unavailable autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 9A1A7C433E0 for ; Mon, 22 Jun 2020 15:31:03 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 4EB9620732 for ; Mon, 22 Jun 2020 15:31:03 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 4EB9620732 Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=arm.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id D38C16B0006; Mon, 22 Jun 2020 11:31:02 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id CE8076B0007; Mon, 22 Jun 2020 11:31:02 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id BAF816B0008; Mon, 22 Jun 2020 11:31:02 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0234.hostedemail.com [216.40.44.234]) by kanga.kvack.org (Postfix) with ESMTP id 9E6E26B0006 for ; Mon, 22 Jun 2020 11:31:02 -0400 (EDT) Received: from smtpin15.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay03.hostedemail.com (Postfix) with ESMTP id 1C3A5824559C for ; Mon, 22 Jun 2020 15:31:02 +0000 (UTC) X-FDA: 76957235964.15.ball27_0c025aa26e33 Received: from filter.hostedemail.com (10.5.16.251.rfc1918.com [10.5.16.251]) by smtpin15.hostedemail.com (Postfix) with ESMTP id 5B3C8180260D0 for ; Mon, 22 Jun 2020 15:30:50 +0000 (UTC) X-HE-Tag: ball27_0c025aa26e33 X-Filterd-Recvd-Size: 6529 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by imf44.hostedemail.com (Postfix) with ESMTP for ; Mon, 22 Jun 2020 15:30:49 +0000 (UTC) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id BA49C31B; Mon, 22 Jun 2020 08:30:48 -0700 (PDT) Received: from [10.57.9.128] (unknown [10.57.9.128]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id C4B6A3F6CF; Mon, 22 Jun 2020 08:30:46 -0700 (PDT) Subject: Re: kernel BUG at mm/huge_memory.c:2613! To: Joerg Roedel , Andrea Arcangeli Cc: Roman Gushchin , Yang Shi , iommu@lists.linux-foundation.org, Linux Kernel Mailing List , Linux MM , Michal Hocko , Johannes Weiner , "Kirill A. Shutemov" , Andrew Morton , Wei Yang , Will Deacon References: <20200619001938.GA135965@carbon.dhcp.thefacebook.com> <20200619011449.GC135965@carbon.dhcp.thefacebook.com> <20200619024026.GB21081@redhat.com> <20200622124646.GI3701@8bytes.org> From: Robin Murphy Message-ID: Date: Mon, 22 Jun 2020 16:30:41 +0100 User-Agent: Mozilla/5.0 (Windows NT 10.0; rv:68.0) Gecko/20100101 Thunderbird/68.9.0 MIME-Version: 1.0 In-Reply-To: <20200622124646.GI3701@8bytes.org> Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-GB Content-Transfer-Encoding: 7bit X-Rspamd-Queue-Id: 5B3C8180260D0 X-Spamd-Result: default: False [0.00 / 100.00] X-Rspamd-Server: rspam03 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On 2020-06-22 13:46, Joerg Roedel wrote: > + Robin > > Robin, any idea on this? After a bit of archaeology, this dates back to the original review: https://lore.kernel.org/linux-arm-kernel/54C285D4.3070802@arm.com/ https://lore.kernel.org/linux-arm-kernel/54DA2666.9030003@arm.com/ In summary: originally this inherited from other arch code that did simply strip __GFP_COMP; that was deemed questionable because of the nonsensical comment about CONFIG_HUGETLBFS that was stuck to it; the current code is like it is because in 5 and a half years nobody said that it's wrong :) If there actually *are* good reasons for stripping __GFP_COMP, then I've certainly no objection to doing so. Robin. > On Thu, Jun 18, 2020 at 10:40:26PM -0400, Andrea Arcangeli wrote: >> Hello, >> >> On Thu, Jun 18, 2020 at 06:14:49PM -0700, Roman Gushchin wrote: >>> I agree. The whole >>> >>> page = alloc_pages_node(nid, alloc_flags, order); >>> if (!page) >>> continue; >>> if (!order) >>> break; >>> if (!PageCompound(page)) { >>> split_page(page, order); >>> break; >>> } else if (!split_huge_page(page)) { >>> break; >>> } >>> >>> looks very suspicious to me. >>> My wild guess is that gfp flags changed somewhere above, so we hit >>> the branch which was never hit before. >> >> Right to be suspicious about the above: split_huge_page on a regular >> page allocated by a driver was never meant to work. >> >> The PageLocked BUG_ON is just a symptom of a bigger issue, basically >> split_huge_page it may survive, but it'll stay compound and in turn it >> must be freed as compound. >> >> The respective free method doesn't even contemplate freeing compound >> pages, the only way the free method can survive, is by removing >> __GFP_COMP forcefully in the allocation that was perhaps set here >> (there are that many __GFP_COMP in that directory): >> >> static void snd_malloc_dev_pages(struct snd_dma_buffer *dmab, size_t size) >> { >> gfp_t gfp_flags; >> >> gfp_flags = GFP_KERNEL >> | __GFP_COMP /* compound page lets parts be mapped */ >> >> And I'm not sure what the comment means here, compound or non compound >> doesn't make a difference when you map it, it's not a THP, the >> mappings must be handled manually so nothing should check PG_compound >> anyway in the mapping code. >> >> Something like this may improve things, it's an untested quick hack, >> but this assumes it's always a bug to setup a compound page for these >> DMA allocations and given the API it's probably a correct >> assumption.. Compound is slower, unless you need it, you can avoid it >> and then split_page will give contiguous memory page granular. Ideally >> the code shouldn't call split_page at all and it should free it all at >> once by keeping track of the order and by returning the order to the >> caller, something the API can't do right now as it returns a plain >> array that can only represent individual small pages. >> >> Once this is resolved, you may want to check your config, iommu passthrough >> sounds more optimal for a soundcard. >> >> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c >> index f68a62c3c32b..3dfbc010fa83 100644 >> --- a/drivers/iommu/dma-iommu.c >> +++ b/drivers/iommu/dma-iommu.c >> @@ -499,6 +499,10 @@ static struct page **__iommu_dma_alloc_pages(struct device *dev, >> >> /* IOMMU can map any pages, so himem can also be used here */ >> gfp |= __GFP_NOWARN | __GFP_HIGHMEM; >> + if (unlikely(gfp & __GFP_COMP)) { >> + WARN(); >> + gfp &= ~__GFP_COMP; >> + } >> >> while (count) { >> struct page *page = NULL; >> @@ -522,13 +526,8 @@ static struct page **__iommu_dma_alloc_pages(struct device *dev, >> continue; >> if (!order) >> break; >> - if (!PageCompound(page)) { >> - split_page(page, order); >> - break; >> - } else if (!split_huge_page(page)) { >> - break; >> - } >> - __free_pages(page, order); >> + split_page(page, order); >> + break; >> } >> if (!page) { >> __iommu_dma_free_pages(pages, i); >> diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c >> index 6850d13aa98c..378f5a36ec5f 100644 >> --- a/sound/core/memalloc.c >> +++ b/sound/core/memalloc.c >> @@ -28,7 +28,6 @@ static void snd_malloc_dev_pages(struct snd_dma_buffer *dmab, size_t size) >> gfp_t gfp_flags; >> >> gfp_flags = GFP_KERNEL >> - | __GFP_COMP /* compound page lets parts be mapped */ >> | __GFP_NORETRY /* don't trigger OOM-killer */ >> | __GFP_NOWARN; /* no stack trace print - this call is non-critical */ >> dmab->area = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr,