Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Message-ID: <b6af3fe8-db9a-b5dc-199f-21c05d7664a2@huawei.com>
Date:   Tue, 12 Apr 2022 08:57:39 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.6.1
Subject: Re: [PATCH 1/4] scsi: core: constify pointer to scsi_host_template
To:     "Ewan D. Milne" <emilne@redhat.com>,
        Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>,
        "James E.J. Bottomley" <jejb@linux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@oracle.com>,
        "Alim Akhtar" <alim.akhtar@samsung.com>,
        Avri Altman <avri.altman@wdc.com>,
        "Doug Gilbert" <dgilbert@interlog.com>,
        <linux-scsi@vger.kernel.org>, <linux-kernel@vger.kernel.org>
CC:     <james.smart@broadcom.com>
References: <20220408103027.311624-1-krzysztof.kozlowski@linaro.org>
 <2a88a992-641a-b3ff-fe39-7a61fff87cb6@huawei.com>
 <4c3be5b6-50ef-9e9a-6cee-9642df943342@linaro.org>
 <7b3885e3-dbae-ff0b-21dc-c28d635d950b@huawei.com>
 <c121430b1b5c8f5816b2b42b9178d00889260c90.camel@redhat.com>
From:   John Garry <john.garry@huawei.com>
In-Reply-To: <c121430b1b5c8f5816b2b42b9178d00889260c90.camel@redhat.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 08/04/2022 20:31, Ewan D. Milne wrote:
> On Fri, 2022-04-08 at 13:57 +0100, John Garry wrote:
>> On 08/04/2022 13:32, Krzysztof Kozlowski wrote:
>>> On 08/04/2022 14:14, John Garry wrote:
>>>> On 08/04/2022 11:30, Krzysztof Kozlowski wrote:
>>>>> Several pointers to 'struct scsi_host_template' do not modify it, so
>>>>> made them const for safety.
>>>>>
>>>> Is this standard practice? What is so special here?
>>> This is standard practice and there is nothing special here. Pointers to
>>> const are preferred because:
>>> 1. They add safety if data is actually const. This is not yet the case,
>>> but scsi_host_template allocation could be made const with some effort.
> 
> This seems unlikely, because some drivers, e.g. vmw_pvscsi and scsi_debug,
> modify the scsi_host_template based on things like module parameters.
>

The standard flow is:

shost = scsi_host_alloc(sht, )

// modify shost, like
shost->cmd_per_lun = 5;

scsi_add_host(shost)

Is there some reason for which those two drivers can't follow that?

>>
>> To me this seems better, but I think that some drivers might modify
>> their scsi_host_template (so not possible)
> 
> Several drivers modify scsi_host_template, e.g. .can_queue, .cmd_per_lun
> 
> There is also code in lpfc_create_port() that initializes a scsi_host_template
> that is embedded in the lpfc_hba struct.  I don't think it gets modified after
> scsi_add_host() but it seems like driver maintainers might expect to be able
> to do so, in general.
> 

Even so, I don't see why other drivers cannot declare their 
scsi_host_template as const. C would have no problem with sht not be 
being const for this:

struct Scsi_Host *scsi_host_alloc(const struct scsi_host_template *sht, )

thanks,
John