Received-SPF: pass (google.com: domain of linux-kernel+bounces-53072-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Mon, 5 Feb 2024 13:12:30 -0400
From: Jason Gunthorpe <jgg@ziepe.ca>
To: James Gowans <jgowans@amazon.com>
Cc: linux-kernel@vger.kernel.org, Eric Biederman <ebiederm@xmission.com>,
	kexec@lists.infradead.org, Joerg Roedel <joro@8bytes.org>,
	Will Deacon <will@kernel.org>, iommu@lists.linux.dev,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	Christian Brauner <brauner@kernel.org>,
	linux-fsdevel@vger.kernel.org, Paolo Bonzini <pbonzini@redhat.com>,
	Sean Christopherson <seanjc@google.com>, kvm@vger.kernel.org,
	Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
	Alexander Graf <graf@amazon.com>,
	David Woodhouse <dwmw@amazon.co.uk>,
	"Jan H . Schoenherr" <jschoenh@amazon.de>,
	Usama Arif <usama.arif@bytedance.com>,
	Anthony Yznaga <anthony.yznaga@oracle.com>,
	Stanislav Kinsburskii <skinsburskii@linux.microsoft.com>,
	madvenka@linux.microsoft.com, steven.sistare@oracle.com,
	yuleixzhang@tencent.com
Subject: Re: [RFC 14/18] intel-iommu: Allocate domain pgtable pages from
 pkernfs
Message-ID: <20240205171230.GB31743@ziepe.ca>
References: <20240205120203.60312-1-jgowans@amazon.com>
 <20240205120203.60312-15-jgowans@amazon.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240205120203.60312-15-jgowans@amazon.com>

On Mon, Feb 05, 2024 at 12:01:59PM +0000, James Gowans wrote:
> @@ -946,7 +946,13 @@ static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
>  		if (!dma_pte_present(pte)) {
>  			uint64_t pteval;
>  
> -			tmp_page = alloc_pgtable_page(domain->nid, gfp);
> +			if (domain->pgtables_allocator.vaddr)
> +				iommu_alloc_page_from_region(
> +						&domain->pgtables_allocator,
> +						&tmp_page,
> +						NULL);

I'm really worried about this change - I plan to redo all of this page
table handling code so it makes use of struct page members for things
like RCU free and more.

Does this end up making the entire struct page owned by the
filesystem?

Jason