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: <461c6576-71df-c16a-3c4f-4691ae84064f@redhat.com>
Date:   Tue, 23 May 2023 19:56:02 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.10.0
Subject: Re: [PATCH] drivers: mpt3sas: mpt3sas_debugfs: return value check of
 `mpt3sas_debugfs_root`
Content-Language: en-US
To:     Dan Carpenter <dan.carpenter@linaro.org>
Cc:     Dongliang Mu <dzm91@hust.edu.cn>, Jing Xu <U202112064@hust.edu.cn>,
        Sathya Prakash <sathya.prakash@broadcom.com>,
        Sreekanth Reddy <sreekanth.reddy@broadcom.com>,
        Suganath Prabu Subramani 
        <suganath-prabu.subramani@broadcom.com>,
        "James E.J. Bottomley" <jejb@linux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@oracle.com>,
        hust-os-kernel-patches@googlegroups.com,
        MPT-FusionLinux.pdl@broadcom.com, linux-scsi@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20230423122535.31019-1-U202112064@hust.edu.cn>
 <6e69b57c-80ae-8b6e-cb5f-9e05da46ecd6@redhat.com>
 <1484408f-f68e-4354-ab59-56af9cd1ef14@kili.mountain>
 <b7154e2c-0438-87d1-9edc-7eb1aad40cd1@hust.edu.cn>
 <81d236bb-3913-4eef-bf71-6d17535d6d79@kili.mountain>
 <892bc614-9e2e-904b-29e0-62daeb855f79@redhat.com>
 <3c4b372f-db4b-43b4-b5ab-7f4860cf6f20@kili.mountain>
From:   Tomas Henzl <thenzl@redhat.com>
In-Reply-To: <3c4b372f-db4b-43b4-b5ab-7f4860cf6f20@kili.mountain>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 5/23/23 16:57, Dan Carpenter wrote:
> On Tue, May 23, 2023 at 04:48:12PM +0200, Tomas Henzl wrote:
>> On 5/8/23 16:38, Dan Carpenter wrote:
>>> On Mon, May 08, 2023 at 09:40:41PM +0800, Dongliang Mu wrote:
>>>>>>> diff --git a/drivers/scsi/mpt3sas/mpt3sas_debugfs.c b/drivers/scsi/mpt3sas/mpt3sas_debugfs.c
>>>>>>> index a6ab1db81167..c92e08c130b9 100644
>>>>>>> --- a/drivers/scsi/mpt3sas/mpt3sas_debugfs.c
>>>>>>> +++ b/drivers/scsi/mpt3sas/mpt3sas_debugfs.c
>>>>>>> @@ -99,8 +99,6 @@ static const struct file_operations mpt3sas_debugfs_iocdump_fops = {
>>>>>>>   void mpt3sas_init_debugfs(void)
>>>>>>>   {
>>>>>>>   	mpt3sas_debugfs_root = debugfs_create_dir("mpt3sas", NULL);
>>>>>>> -	if (!mpt3sas_debugfs_root)
>>>>>>> -		pr_info("mpt3sas: Cannot create debugfs root\n");
>>>>>> Hi Jing,
>>>>>> most drivers just ignore the return value but here the author wanted to
>>>>>> have the information logged.
>>>>>> Can you instead of removing the message modify the 'if' condition so it
>>>>>> suits the author's intention?
>>>>>
>>>>> This code was always just wrong.
>>>>>
>>>>> The history of this is slightly complicated and boring.  These days it's
>>>>> harmless dead code so I guess it's less bad than before.
>>>>
>>>> Hi Dan and Tomas,
>>>>
>>>> Any conclusion about this patch? The student Jing Xu is not sure about how
>>>> to revise this patch.
>>>
>>> The correct fix is to delete the code.
>>>
>>> Debugfs code has error checking built in and was never supposed to be
>>> checked for errors in normal driver code.
>>>
>>> Originally, debugfs returned a mix of error pointers and NULL.  In the
>>> kernel, when you have a mix of error pointers and NULL, then the NULL
>>> means that the feature has been disabled deliberately.  It's not an
>>> error, we should not print a message.
>>>
>>> So a different, correct-ish way to write write debugfs error handling
>>> was to say:
>>>
>>> 	mpt3sas_debugfs_root = debugfs_create_dir("mpt3sas", NULL);
>>> 	if (IS_ERR(mpt3sas_debugfs_root))
>>> 		return PTR_ERR(mpt3sas_debugfs_root);
>> I'm fine with this as well, I could wish we get a fix for the exact same
>> case of debugfs_create_dir in mpt3sas_setup_debugfs and ideally all the
>> debugfs_create* in  mpt3sas_debugfs.c in a single patch. But this patch
>> is ok even if that wasn't possible.
>> tomash
> 
> No, you didn't read until the end.  That will break the driver badly.
> 
> This *used* to be a correct-ish way that *used* to work but it was never
> the what Greg wanted.  So to discourage people from doing it, Greg made
> it *impossible* to check for if debugfs has failed.  Literally, the only
> correct thing to do now is to delete the debugfs checking.

I put my comment in on a wrong place I meant the original patch without
the check, sorry for the confusion. The only thing I'd like to see is to
have corrected all debugfs_create* and that is also optional.

> 
> regards,
> dan carpenter
> 
>