Received-SPF: pass (google.com: domain of linux-kernel+bounces-90836-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Mon, 4 Mar 2024 15:32:03 +0000
From: Will Deacon <will@kernel.org>
To: Michael Kelley <mhklinux@outlook.com>
Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"kernel-team@android.com" <kernel-team@android.com>,
	"iommu@lists.linux.dev" <iommu@lists.linux.dev>,
	Christoph Hellwig <hch@lst.de>,
	Marek Szyprowski <m.szyprowski@samsung.com>,
	Robin Murphy <robin.murphy@arm.com>,
	Petr Tesarik <petr.tesarik1@huawei-partners.com>,
	Dexuan Cui <decui@microsoft.com>,
	Nicolin Chen <nicolinc@nvidia.com>
Subject: Re: [PATCH v5 6/6] swiotlb: Remove pointless stride adjustment for
 allocations >= PAGE_SIZE
Message-ID: <20240304153203.GA21077@willie-the-truck>
References: <20240228133930.15400-1-will@kernel.org>
 <20240228133930.15400-7-will@kernel.org>
 <SN6PR02MB4157A62353559DA8DB8BC4ADD45F2@SN6PR02MB4157.namprd02.prod.outlook.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <SN6PR02MB4157A62353559DA8DB8BC4ADD45F2@SN6PR02MB4157.namprd02.prod.outlook.com>
User-Agent: Mutt/1.10.1 (2018-07-13)

On Thu, Feb 29, 2024 at 06:07:32AM +0000, Michael Kelley wrote:
> From: Will Deacon <will@kernel.org> Sent: Wednesday, February 28, 2024 5:40 AM
> > 
> > For swiotlb allocations >= PAGE_SIZE, the slab search historically
> > adjusted the stride to avoid checking unaligned slots. However, this is
> > no longer needed now that the code around it has evolved and the
> > stride is calculated from the required alignment.
> > 
> > Either 'alloc_align_mask' is used to specify the allocation alignment or
> > the DMA 'min_align_mask' is used to align the allocation with 'orig_addr'.
> > At least one of these masks is always non-zero.
> 
> I think the patch is correct, but this justification is not.  alloc_align_mask
> and the DMA min_align_mask are often both zero.  While the NVMe
> PCI driver sets min_align_mask, SCSI disk drivers do not (except for the
> Hyper-V synthetic SCSI driver).   When both are zero, presumably
> there are no alignment requirements, so a stride of 1 is appropriate.

Sorry, yes, I messed up the commit message here as I was trying to reason
through the allocation case separately from the mapping case. However, I
need to digest the rest of this thread before doing the obvious fix...

Will