Received-SPF: pass (google.com: domain of linux-kernel+bounces-165907-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Message-ID: <ed0e7138-c5db-40fb-ac12-36df007d0b4f@linux.intel.com>
Date: Thu, 2 May 2024 09:57:44 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Cc: baolu.lu@linux.intel.com, Palmer Dabbelt <palmer@dabbelt.com>,
 Albert Ou <aou@eecs.berkeley.edu>, Anup Patel <apatel@ventanamicro.com>,
 Sunil V L <sunilvl@ventanamicro.com>, Nick Kossifidis <mick@ics.forth.gr>,
 Sebastien Boeuf <seb@rivosinc.com>, Rob Herring <robh+dt@kernel.org>,
 Krzysztof Kozlowski <krzk+dt@kernel.org>, Conor Dooley
 <conor+dt@kernel.org>, devicetree@vger.kernel.org, iommu@lists.linux.dev,
 linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org,
 linux@rivosinc.com
Subject: Re: [PATCH v3 5/7] iommu/riscv: Device directory management.
To: Tomasz Jeznach <tjeznach@rivosinc.com>, Joerg Roedel <joro@8bytes.org>,
 Will Deacon <will@kernel.org>, Robin Murphy <robin.murphy@arm.com>,
 Paul Walmsley <paul.walmsley@sifive.com>
References: <cover.1714494653.git.tjeznach@rivosinc.com>
 <ce3b82a20db0b776685269674ce9b7a926d5680d.1714494653.git.tjeznach@rivosinc.com>
 <104358de-ea86-4e25-8942-ba285d6f317b@linux.intel.com>
Content-Language: en-US
From: Baolu Lu <baolu.lu@linux.intel.com>
In-Reply-To: <104358de-ea86-4e25-8942-ba285d6f317b@linux.intel.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

On 5/2/24 9:38 AM, Baolu Lu wrote:
> On 5/1/24 4:01 AM, Tomasz Jeznach wrote:
>> @@ -128,6 +489,7 @@ void riscv_iommu_remove(struct riscv_iommu_device 
>> *iommu)
>>   {
>>       iommu_device_unregister(&iommu->iommu);
>>       iommu_device_sysfs_remove(&iommu->iommu);
>> +    riscv_iommu_iodir_set_mode(iommu, RISCV_IOMMU_DDTP_MODE_OFF);
>>   }
>>   int riscv_iommu_init(struct riscv_iommu_device *iommu)
>> @@ -138,12 +500,13 @@ int riscv_iommu_init(struct riscv_iommu_device 
>> *iommu)
>>       if (rc)
>>           return dev_err_probe(iommu->dev, rc, "unexpected device 
>> state\n");
>> -    /*
>> -     * Placeholder for a complete IOMMU device initialization.
>> -     * For now, only bare minimum: enable global identity mapping 
>> mode and register sysfs.
>> -     */
>> -    riscv_iommu_writeq(iommu, RISCV_IOMMU_REG_DDTP,
>> -               FIELD_PREP(RISCV_IOMMU_DDTP_MODE, 
>> RISCV_IOMMU_DDTP_MODE_BARE));
>> +    rc = riscv_iommu_iodir_alloc(iommu);
>> +    if (rc)
>> +        goto err_init;
>> +
>> +    rc = riscv_iommu_iodir_set_mode(iommu, RISCV_IOMMU_DDTP_MODE_MAX);
>> +    if (rc)
>> +        goto err_init;
>>       rc = iommu_device_sysfs_add(&iommu->iommu, NULL, NULL, 
>> "riscv-iommu@%s",
>>                       dev_name(iommu->dev));
> 
> The device directory root page might be allocated in
> riscv_iommu_iodir_alloc(),
> 
> +    if (!iommu->ddt_root) {
> +        iommu->ddt_root = riscv_iommu_get_pages(iommu, 0);
> +        iommu->ddt_phys = __pa(iommu->ddt_root);
> +    }
> 
> But I didn't find any place to free it in the error paths. Did I
> overlook anything?

I found the answer by myself. devres_alloc() is used so the page memory
resources are managed by the driver core automatically. Please ignore
the above comment.

Not sure about the pages for paging domain. If all pages release is
deferred, it may cause big memory consumption. We have a trade-off here.
On one hand, this might speed up the domain mapping since all pages
might be preallocated; on the other hand, it will cost more memory.

Best regards,
baolu