Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 25 Jul 2023 10:41:59 -0300
From:   Jason Gunthorpe <jgg@ziepe.ca>
To:     Dawei Li <set_pte_at@outlook.com>
Cc:     will@kernel.org, robin.murphy@arm.com, joro@8bytes.org,
        linux-arm-kernel@lists.infradead.org, iommu@lists.linux.dev,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] iommu/arm-smmu-v3: Change vmid alloc strategy from
 bitmap to ida
Message-ID: <ZL/RJwoJkB3muoTC@ziepe.ca>
References: <TYCP286MB2323E0C525FF9F94E3B07C7ACA35A@TYCP286MB2323.JPNP286.PROD.OUTLOOK.COM>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <TYCP286MB2323E0C525FF9F94E3B07C7ACA35A@TYCP286MB2323.JPNP286.PROD.OUTLOOK.COM>
Precedence: bulk

On Sun, Jul 16, 2023 at 12:16:21AM +0800, Dawei Li wrote:
> For current implementation of vmid allocation of arm smmu-v3, a per-smmu
> devide bitmap of 64K bits(8K bytes) is allocated on behalf of possible VMID
> range, which is two pages for some architectures. Besides that, its memory
> consumption is 'static', despite of how many VMIDs are allocated actually.

Is there an actual problem here? Allocating a single 8k page at early
boot doesn't seem like a burden? Are there alot of smmu instances?

> So an IDA based implementation is introduced to address this issue, which
> is capable of self-expanding on the actual need of VMID allocation.

However, I agree IDA is just generally better and generally drivers
shouldn't be open coding it

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason