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;
Message-ID: <d6db1784-0b21-9bdb-8528-3f814da188b0@mm-sol.com>
Date:   Mon, 18 Jul 2022 10:43:56 +0300
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.9.1
Subject: Re: [PATCH] PCI: qcom: Add support for modular builds
Content-Language: en-US
To:     Bjorn Helgaas <helgaas@kernel.org>,
        Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Cc:     Johan Hovold <johan+linaro@kernel.org>,
        Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
        Rob Herring <robh@kernel.org>,
        =?UTF-8?Q?Krzysztof_Wilczy=c5=84ski?= <kw@linux.com>,
        Bjorn Helgaas <bhelgaas@google.com>,
        Andy Gross <agross@kernel.org>,
        Bjorn Andersson <bjorn.andersson@linaro.org>,
        linux-pci@vger.kernel.org, linux-arm-msm@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20220715165601.GA1139849@bhelgaas>
From:   Stanimir Varbanov <svarbanov@mm-sol.com>
In-Reply-To: <20220715165601.GA1139849@bhelgaas>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 7/15/22 19:56, Bjorn Helgaas wrote:
> On Thu, Jul 14, 2022 at 04:05:41PM +0300, Dmitry Baryshkov wrote:
>> On 14/07/2022 15:19, Stanimir Varbanov wrote:
>>> Please take a look why we made it built-in first [1].
>>>
>>> If arguments there are still valid I don't see why to make it a module
>>> again.
>>>
>>> [1] https://lkml.org/lkml/2016/8/24/694
>>
>> It looks like there is a move to make all non-essential drivers buildable as
>> modules. For example, the Kirin, dra7xx, Meson PCI controllers are now
>> buildable as modules. So I think we can follow that and allow building the
>> pcie-qcom as a module.
> 
> IIUC the arguments in [1] are that:
> 
>   - Kconfig is bool, so it can't be built as a module
>   - there's no sensible use case for unbind
> 
> Those described the situation at the time, and there's no point in
> having .remove() and using module_platform_driver() if Kconfig is
> bool.
> 
> But they don't seem like arguments for why the driver couldn't be
> *made* modular.

I guess the core of the problem was lack of dw_pcie_host_deinit() at
that time.
> 
> I think drivers *should* be modular unless there's a technical reason
> they can't be.

I agree.

> 
> Bjorn

-- 
regards,
Stan