Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
Message-ID: <42860c87-cf4d-0413-c3ae-b74ee9e7e5e6@linux.intel.com>
Date:   Mon, 25 Sep 2023 10:48:21 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.15.1
Cc:     baolu.lu@linux.intel.com,
        Jean-Philippe Brucker <jean-philippe@linaro.org>,
        Niklas Schnelle <schnelle@linux.ibm.com>,
        Joerg Roedel <joro@8bytes.org>, Will Deacon <will@kernel.org>,
        virtualization@lists.linux-foundation.org, iommu@lists.linux.dev,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 1/2] iommu/virtio: Make use of ops->iotlb_sync_map
Content-Language: en-US
To:     Jason Gunthorpe <jgg@ziepe.ca>, Robin Murphy <robin.murphy@arm.com>
References: <20230918-viommu-sync-map-v2-0-f33767f6cf7a@linux.ibm.com>
 <20230918-viommu-sync-map-v2-1-f33767f6cf7a@linux.ibm.com>
 <ae7e513b-eb86-97e2-bed0-3cca91b8c959@arm.com>
 <20230919081519.GA3860249@myrica> <20230919144649.GT13795@ziepe.ca>
 <20230922075719.GB1361815@myrica> <20230922124130.GD13795@ziepe.ca>
 <900b644e-6e21-1038-2252-3dc86cbf0a32@arm.com>
 <20230922162714.GH13795@ziepe.ca>
 <123c53c3-d259-9c20-9aa6-0c216d7eb3c0@arm.com>
 <20230922233309.GI13795@ziepe.ca>
From:   Baolu Lu <baolu.lu@linux.intel.com>
In-Reply-To: <20230922233309.GI13795@ziepe.ca>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 9/23/23 7:33 AM, Jason Gunthorpe wrote:
> On Fri, Sep 22, 2023 at 07:07:40PM +0100, Robin Murphy wrote:
> 
>> virtio isn't setting ops->pgsize_bitmap for the sake of direct mappings
>> either; it sets it once it's discovered any instance, since apparently it's
>> assuming that all instances must support identical page sizes, and thus once
>> it's seen one it can work "normally" per the core code's assumptions. It's
>> also I think the only driver which has a "finalise" bodge but*can*  still
>> properly support map-before-attach, by virtue of having to replay mappings
>> to every new endpoint anyway.
> Well it can't quite do that since it doesn't know the geometry - it
> all is sort of guessing and hoping it doesn't explode on replay. If it
> knows the geometry it wouldn't need finalize...

The ultimate solution to this problem seems to be to add device pointer
to the parameter of ops->domain_alloc. So once the domain is allocated,
it is fully initialized. Attaching this domain to a device that is not
compatible will return -EINVAL, then the caller has to allocate a new
domain for this device.

I feel that this is not an AMD specific problem, other iommu drivers
will also encounter the similar problem sooner or later.

Best regards,
baolu