Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] coresight: use put_device() instead of kfree()
To:     Mathieu Poirier <mathieu.poirier@linaro.org>
References: <d5a326b6c284f7389301b265716dc58fda492c98.1521358552.git.arvind.yadav.cs@gmail.com>
 <CANLsYkymOF2oG7TmKbRTakYKw0ECN_DXfkX5b8WjmS0Lz0sfig@mail.gmail.com>
 <5AB9ACDA.9090207@gmail.com>
 <CANLsYkw0v-vVpf1M1Ae+y=Hq5cqqp+_Sjq41pDS+=Xv9zesHAw@mail.gmail.com>
Cc:     linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org
From:   arvindY <arvind.yadav.cs@gmail.com>
Message-ID: <5ABA7111.3020908@gmail.com>
Date:   Tue, 27 Mar 2018 21:58:01 +0530
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.6.0
MIME-Version: 1.0
In-Reply-To: <CANLsYkw0v-vVpf1M1Ae+y=Hq5cqqp+_Sjq41pDS+=Xv9zesHAw@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


On Tuesday 27 March 2018 09:37 PM, Mathieu Poirier wrote:
> On 26 March 2018 at 20:30, arvindY <arvind.yadav.cs@gmail.com> wrote:
>>
>> On Tuesday 27 March 2018 03:16 AM, Mathieu Poirier wrote:
>>> drivers/hwtracing/coresight/coresight.c
>>> On 18 March 2018 at 01:38, Arvind Yadav <arvind.yadav.cs@gmail.com> wrote:
>>>> Never directly free @dev after calling device_register(), even
>>>> if it returned an error. Always use put_device() to give up the
>>>> reference initialized.
>>>>
>>>> Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
>>>> ---
>>>>    drivers/hwtracing/coresight/coresight.c | 8 ++++----
>>>>    1 file changed, 4 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/drivers/hwtracing/coresight/coresight.c
>>>> b/drivers/hwtracing/coresight/coresight.c
>>>> index 389c4ba..132dfbc 100644
>>>> --- a/drivers/hwtracing/coresight/coresight.c
>>>> +++ b/drivers/hwtracing/coresight/coresight.c
>>>> @@ -1026,8 +1026,10 @@ struct coresight_device *coresight_register(struct
>>>> coresight_desc *desc)
>>>>           dev_set_name(&csdev->dev, "%s", desc->pdata->name);
>>>>
>>>>           ret = device_register(&csdev->dev);
>>>> -       if (ret)
>>>> -               goto err_device_register;
>>>> +       if (ret) {
>>>> +               put_device(&csdev->dev);
>>>> +               goto err_kzalloc_csdev;
>>>> +       }
>>>>
>>>>           mutex_lock(&coresight_mutex);
>>>>
>>>> @@ -1038,8 +1040,6 @@ struct coresight_device *coresight_register(struct
>>>> coresight_desc *desc)
>>>>
>>>>           return csdev;
>>>>
>>>> -err_device_register:
>>>> -       kfree(conns);
>>> Apologies for the late reply, I was travelling.
>>>
>>> I suggest to simply replace kfree() with put_device() in order to
>>> concentrate the error handling code in the same area and make sure
>>> that memory allocated for @conns and @refcnts is freed.
>>>
>>> Thanks,
>>> Mathieu
>>
>> If you will see the comment for device_register(drivers/base/core.c)
>> there is mentioned that 'NOTE: _Never_ directly free @dev
>> after calling this function, even if it returned an error!
>> Always use put_device() to give up the reference initialized
>> in this function instead.
> I have read the notice and in full agreement with the put_device() part.
>
>> Here  put_device() will decrement the last reference and then
>> free the memory by calling dev->release.  Internally
>> put_device() -> kobject_put() -> kobject_cleanup() which is
>> responsible to call 'dev -> release' and also free other kobject
>> resources.
> Memory would be automatically freed if it would have been allocated
> with the devm_XYZ() helpers but it is not the case here.  As such it
> has to be freed explicitly.  Reading your patch again (and the jetlag
> somewhat fading) I think you've done the right thing except for the
> "goto" that should still point to "err_device_register".
>

Take rest :)
'goto' should not point to "err_device_register" because
put_device() will call 'dev->release' which is coresight_device_release().
It's release @conns, @refcnt and @csdev . If we will keep same 'goto
then kfree() will be redundant for all.


>> we should always avoid kfree() if device_register()
>> returned an error. Otherwise it'll not do clean up of other
>> kobject resources.
>>
>> ~arvind
>>>>    err_kzalloc_conns:
>>>>           kfree(refcnts);
>>>>    err_kzalloc_refcnts:
>>>> --
>>>> 2.7.4
>>>>