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: <60c2291e-6e74-1df1-692f-00d02f8a83cd@gmail.com>
Date:   Thu, 9 Jun 2022 15:24:34 -0500
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 v1 00/16] ACPI: Get rid of the list of children in struct
 acpi_device
Content-Language: en-US
To:     Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        "Rafael J. Wysocki" <rjw@rjwysocki.net>,
        Frank Rowand <frank.rowand@sony.com>,
        Rob Herring <robh@kernel.org>
Cc:     Linux ACPI <linux-acpi@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Linux PM <linux-pm@vger.kernel.org>,
        Mika Westerberg <mika.westerberg@linux.intel.com>,
        Hans de Goede <hdegoede@redhat.com>,
        Sakari Ailus <sakari.ailus@linux.intel.com>
References: <1843211.tdWV9SEqCh@kreacher>
 <YqINyDTfpNBGDYlb@smile.fi.intel.com>
From:   Frank Rowand <frowand.list@gmail.com>
In-Reply-To: <YqINyDTfpNBGDYlb@smile.fi.intel.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 6/9/22 11:12, Andy Shevchenko wrote:
> On Thu, Jun 09, 2022 at 03:44:27PM +0200, Rafael J. Wysocki wrote:
>> Hi All,
>>
>> Confusingly enough, the ACPI subsystem stores the information on the given ACPI
>> device's children in two places: as the list of children in struct acpi_device
>> and (as a result of device registration) in the list of children in the embedded
>> struct device.
>>
>> These two lists agree with each other most of the time, but not always (like in
>> error paths in some cases), and the list of children in struct acpi_device is
>> not generally safe to use without locking.  In principle, it should always be
>> walked under acpi_device_lock, but in practice holding acpi_scan_lock is
>> sufficient for that too.  However, its users may not know whether or not
>> they operate under acpi_scan_lock and at least in some cases it is not accessed
>> in a safe way (note that ACPI devices may go away as a result of hot-remove,
> 
>> unlike OF nodes).
> 
> Hmm... Does it true for DT overlays? Not an expert in DT overlays, though,
> adding Rob and Frank.

DT nodes can be removed.  The devicetree code uses devtree_lock and of_mutex
as needed for protection.

-Frank

> 
>> For this reason, it is better to consolidate the code that needs to walk the
>> children of an ACPI device which is the purpose of this patch series.
>>
>> Overall, it switches over all of the users of the list of children in struct
>> acpi_device to using helpers based on the driver core's mechanics and finally
>> drops that list, but some extra cleanups are done on the way.
>>
>> Please refer to the patch changelogs for details.
> 
> I'm going to look the individual patches.
>