From nobody Fri May 17 11:58:55 2024 Delivered-To: importer@patchew.org Received-SPF: pass (zohomail.com: domain of lists.xenproject.org designates 192.237.175.120 as permitted sender) client-ip=192.237.175.120; envelope-from=xen-devel-bounces@lists.xenproject.org; helo=lists.xenproject.org; Authentication-Results: mx.zohomail.com; spf=pass (zohomail.com: domain of lists.xenproject.org designates 192.237.175.120 as permitted sender) smtp.mailfrom=xen-devel-bounces@lists.xenproject.org; dmarc=fail(p=none dis=none) header.from=arm.com Return-Path: Received: from lists.xenproject.org (lists.xenproject.org [192.237.175.120]) by mx.zohomail.com with SMTPS id 1666025545665442.06585456541177; Mon, 17 Oct 2022 09:52:25 -0700 (PDT) Received: from list by lists.xenproject.org with outflank-mailman.424519.671998 (Exim 4.92) (envelope-from ) id 1okTLB-0002tS-28; Mon, 17 Oct 2022 16:51:49 +0000 Received: by outflank-mailman (output) from mailman id 424519.671998; Mon, 17 Oct 2022 16:51:49 +0000 Received: from localhost ([127.0.0.1] helo=lists.xenproject.org) by lists.xenproject.org with esmtp (Exim 4.92) (envelope-from ) id 1okTLA-0002tL-Vr; Mon, 17 Oct 2022 16:51:48 +0000 Received: by outflank-mailman (input) for mailman id 424519; Mon, 17 Oct 2022 16:51:47 +0000 Received: from se1-gles-sth1-in.inumbo.com ([159.253.27.254] helo=se1-gles-sth1.inumbo.com) by lists.xenproject.org with esmtp (Exim 4.92) (envelope-from ) id 1okTL9-0002tF-HN for xen-devel@lists.xenproject.org; Mon, 17 Oct 2022 16:51:47 +0000 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by se1-gles-sth1.inumbo.com (Halon) with ESMTP id fb50595b-4e3b-11ed-91b4-6bf2151ebd3b; Mon, 17 Oct 2022 18:51:44 +0200 (CEST) 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 3481C113E; Mon, 17 Oct 2022 09:51:50 -0700 (PDT) Received: from entos-skylake.shanghai.arm.com (entos-skylake.shanghai.arm.com [10.169.212.207]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPA id A64573F792; Mon, 17 Oct 2022 09:51:41 -0700 (PDT) X-Outflank-Mailman: Message body and most headers restored to incoming version X-BeenThere: xen-devel@lists.xenproject.org List-Id: Xen developer discussion List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Errors-To: xen-devel-bounces@lists.xenproject.org Precedence: list Sender: "Xen-devel" X-Inumbo-ID: fb50595b-4e3b-11ed-91b4-6bf2151ebd3b From: Henry Wang To: xen-devel@lists.xenproject.org Cc: Henry Wang , Stefano Stabellini , Julien Grall , Bertrand Marquis , Wei Chen , Volodymyr Babchuk Subject: [PATCH v4] xen/arm: p2m: Populate pages for GICv2 mapping in arch_domain_create() Date: Mon, 17 Oct 2022 16:51:33 +0000 Message-Id: <20221017165133.17066-1-Henry.Wang@arm.com> X-Mailer: git-send-email 2.17.1 X-ZM-MESSAGEID: 1666025546828100001 Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Hardware using GICv2 needs to create a P2M mapping of 8KB GICv2 area when the domain is created. Considering the worst case of page tables which requires 6 P2M pages as the two pages will be consecutive but not necessarily in the same L3 page table and keep a buffer, populate 16 pages as the default value to the P2M pages pool in arch_domain_create() at the domain creation stage to satisfy the GICv2 requirement. For GICv3, the above-mentioned P2M mapping is not necessary, but since the allocated 16 pages here would not be lost, hence populate these pages unconditionally. With the default 16 P2M pages populated, there would be a case that failures would happen in the domain creation with P2M pages already in use. To properly free the P2M for this case, firstly support the optionally preemption of p2m_teardown(), then call p2m_teardown() and p2m_set_allocation(d, 0, NULL) non-preemptively in p2m_final_teardown(). As non-preemptive p2m_teardown() should only return 0, use a BUG_ON to confirm that. Since p2m_final_teardown() is called either after domain_relinquish_resources() where relinquish_p2m_mapping() has been called, or from failure path of domain_create()/arch_domain_create() where mappings that require p2m_put_l3_page() should never be created, relinquish_p2m_mapping() is not added in p2m_final_teardown(), add in-code comments to refer this. Fixes: cbea5a1149ca ("xen/arm: Allocate and free P2M pages from the P2M poo= l") Suggested-by: Julien Grall Signed-off-by: Henry Wang --- This should also be backported to 4.13, 4.14, 4.15 and 4.16. v4 changes: - Move the initial population of 16 default pages to the end of p2m_init(), add if(rc) return rc; after p2m_alloc_table() v3 changes: - Move the population of default pages to p2m_init(). - Use a loop over p2m_teardown_allocation() to implement the non-preemptive p2m_teardown_allocation() and avoid open-coding. - Reorder assertions in p2m_final_teardown(). - Add p2m_teardown() will always return 0 if called non-preemptively in doc, move the page_list_empty(&p2m->pages) check to p2m_teardown() and use a BUG_ON to confirm p2m_teardown() will return 0 in p2m_final_teardown(). - Add a comment in p2m_final_teardown() to mention relinquish_p2m_mapping() does not need to be called, also update commit message. v2 changes: - Move the p2m_set_allocation(d, 0, NULL); to p2m_final_teardown(). - Support optionally preemption of p2m_teardown(), and make the calling of p2m_teardown() preemptively when relinquish the resources, non-preemptive= ly in p2m_final_teardown(). - Refactor the error handling to make the code use less spin_unlock. - Explain the worst case of page tables and the unconditional population of pages in commit message. - Mention the unconditional population of pages in in-code comment. --- xen/arch/arm/domain.c | 2 +- xen/arch/arm/include/asm/p2m.h | 14 ++++++++++---- xen/arch/arm/p2m.c | 34 ++++++++++++++++++++++++++++++++-- 3 files changed, 43 insertions(+), 7 deletions(-) diff --git a/xen/arch/arm/domain.c b/xen/arch/arm/domain.c index 2c84e6dbbb..38e22f12af 100644 --- a/xen/arch/arm/domain.c +++ b/xen/arch/arm/domain.c @@ -1064,7 +1064,7 @@ int domain_relinquish_resources(struct domain *d) return ret; =20 PROGRESS(p2m): - ret =3D p2m_teardown(d); + ret =3D p2m_teardown(d, true); if ( ret ) return ret; =20 diff --git a/xen/arch/arm/include/asm/p2m.h b/xen/arch/arm/include/asm/p2m.h index 42bfd548c4..c8f14d13c2 100644 --- a/xen/arch/arm/include/asm/p2m.h +++ b/xen/arch/arm/include/asm/p2m.h @@ -194,14 +194,18 @@ int p2m_init(struct domain *d); =20 /* * The P2M resources are freed in two parts: - * - p2m_teardown() will be called when relinquish the resources. It - * will free large resources (e.g. intermediate page-tables) that - * requires preemption. + * - p2m_teardown() will be called preemptively when relinquish the + * resources, in which case it will free large resources (e.g. intermed= iate + * page-tables) that requires preemption. * - p2m_final_teardown() will be called when domain struct is been * freed. This *cannot* be preempted and therefore one small * resources should be freed here. + * Note that p2m_final_teardown() will also call p2m_teardown(), to prope= rly + * free the P2M when failures happen in the domain creation with P2M pages + * already in use. In this case p2m_teardown() is called non-preemptively= and + * p2m_teardown() will always return 0. */ -int p2m_teardown(struct domain *d); +int p2m_teardown(struct domain *d, bool allow_preemption); void p2m_final_teardown(struct domain *d); =20 /* @@ -266,6 +270,8 @@ mfn_t p2m_get_entry(struct p2m_domain *p2m, gfn_t gfn, /* * Direct set a p2m entry: only for use by the P2M code. * The P2M write lock should be taken. + * TODO: Add a check in __p2m_set_entry() to avoid creating a mapping in + * arch_domain_create() that requires p2m_put_l3_page() to be called. */ int p2m_set_entry(struct p2m_domain *p2m, gfn_t sgfn, diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c index f17500ddf3..0b4e24218e 100644 --- a/xen/arch/arm/p2m.c +++ b/xen/arch/arm/p2m.c @@ -1685,7 +1685,7 @@ static void p2m_free_vmid(struct domain *d) spin_unlock(&vmid_alloc_lock); } =20 -int p2m_teardown(struct domain *d) +int p2m_teardown(struct domain *d, bool allow_preemption) { struct p2m_domain *p2m =3D p2m_get_hostp2m(d); unsigned long count =3D 0; @@ -1693,6 +1693,9 @@ int p2m_teardown(struct domain *d) unsigned int i; int rc =3D 0; =20 + if ( page_list_empty(&p2m->pages) ) + return 0; + p2m_write_lock(p2m); =20 /* @@ -1716,7 +1719,7 @@ int p2m_teardown(struct domain *d) p2m_free_page(p2m->domain, pg); count++; /* Arbitrarily preempt every 512 iterations */ - if ( !(count % 512) && hypercall_preempt_check() ) + if ( allow_preemption && !(count % 512) && hypercall_preempt_check= () ) { rc =3D -ERESTART; break; @@ -1736,7 +1739,20 @@ void p2m_final_teardown(struct domain *d) if ( !p2m->domain ) return; =20 + /* + * No need to call relinquish_p2m_mapping() here because + * p2m_final_teardown() is called either after domain_relinquish_resou= rces() + * where relinquish_p2m_mapping() has been called, or from failure pat= h of + * domain_create()/arch_domain_create() where mappings that require + * p2m_put_l3_page() should never be created. For the latter case, als= o see + * comment on top of the p2m_set_entry() for more info. + */ + + BUG_ON(p2m_teardown(d, false)); ASSERT(page_list_empty(&p2m->pages)); + + while ( p2m_teardown_allocation(d) =3D=3D -ERESTART ) + continue; /* No preemption support here */ ASSERT(page_list_empty(&d->arch.paging.p2m_freelist)); =20 if ( p2m->root ) @@ -1784,6 +1800,8 @@ int p2m_init(struct domain *d) !iommu_has_feature(d, IOMMU_FEAT_COHERENT_WALK); =20 rc =3D p2m_alloc_table(d); + if ( rc !=3D 0 ) + return rc; =20 /* * Make sure that the type chosen to is able to store the an vCPU ID @@ -1803,6 +1821,18 @@ int p2m_init(struct domain *d) */ p2m->domain =3D d; =20 + /* + * Hardware using GICv2 needs to create a P2M mapping of 8KB GICv2 area + * when the domain is created. Considering the worst case for page + * tables and keep a buffer, populate 16 pages to the P2M pages pool h= ere. + * For GICv3, the above-mentioned P2M mapping is not necessary, but si= nce + * the allocated 16 pages here would not be lost, hence populate these + * pages unconditionally. + */ + spin_lock(&d->arch.paging.lock); + rc =3D p2m_set_allocation(d, 16, NULL); + spin_unlock(&d->arch.paging.lock); + return rc; } =20 --=20 2.17.1