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] zcrypt: handle AP Info notification from CHSC SEI command
To:     Harald Freudenberger <freude@linux.ibm.com>,
        Heiko Carstens <heiko.carstens@de.ibm.com>
Cc:     Sebastian Ott <sebott@linux.ibm.com>, linux-s390@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        Martin Schwidefsky <schwidefsky@de.ibm.com>,
        oberpar@linux.ibm.com, pmorel@linux.ibm.com, pasic@linux.ibm.com,
        cohuck@redhat.com
References: <1548870526-30595-1-git-send-email-akrowiak@linux.ibm.com>
 <alpine.LFD.2.21.1901301912380.2372@schleppi>
 <88b72363-9dca-4653-c326-9c37b6349a32@linux.ibm.com>
 <20190201090159.GC3770@osiris>
 <a02f611b-370c-85b3-9e3d-c7dc2462004d@linux.ibm.com>
 <e42e9ddb-e771-2789-727c-917ca3e176f2@linux.ibm.com>
From:   Tony Krowiak <akrowiak@linux.ibm.com>
Date:   Tue, 5 Feb 2019 15:26:09 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.2.1
MIME-Version: 1.0
In-Reply-To: <e42e9ddb-e771-2789-727c-917ca3e176f2@linux.ibm.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Message-Id: <70880c70-382b-2753-be46-4167a6e6b15f@linux.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 2/4/19 5:06 AM, Harald Freudenberger wrote:
> On 01.02.19 16:38, Tony Krowiak wrote:
>> On 2/1/19 4:01 AM, Heiko Carstens wrote:
>>> On Thu, Jan 31, 2019 at 06:28:39PM -0500, Tony Krowiak wrote:
>>>> On 1/30/19 1:32 PM, Sebastian Ott wrote:
>>>>> On Wed, 30 Jan 2019, Tony Krowiak wrote:
>>>>>> +#if IS_ENABLED(CONFIG_ZCRYPT)
>>>>>> +void ap_bus_cfg_chg(void);
>>>>>> +#else
>>>>>> +#error "no CONFIG_ZCRYPT"
>>>>>      ^
>>>>> I don't think that's the right thing to do here.
>>>>
>>>> I'd like to leave it. If somebody edits .config
>>>> and sets CONFIG_ZCRYPT=n, then the build will
>>>> fail. The preprocessor error above tells them
>>>> why.
>>>
>>> No, the kernel build should never fail if a config option is not set.
>>> Also the above should be "#ifdef CONFIG_ZCRYPT".
>>
>> Will do.
>>
>>>
>>> In addition (this isn't quoted unfortunately) the alternative function
>>> in the header file is missing the "inline" attribute. Can you please
>>> add that too?
>>
>> I can.
>>
>>>
>>> static inline void ap_bus_cfg_chg(void) { }
>>>
>>>>>> +* A config change has happened, Force an ap bus rescan.
>>>>>> +*/
>>>>>> +void ap_bus_cfg_chg(void)
>>>>>> +{
>>>>>> +    AP_DBF(DBF_INFO, "%s config change, forcing bus rescan\n", __func__);
>>>>>> +
>>>>>> +    ap_bus_force_rescan();
>>>>>> +}
>>>>>> +EXPORT_SYMBOL(ap_bus_cfg_chg);
>>>>>
>>>>> There is no need for the export symbol - you don't call that function
>>>> >from module code.
>>>>> As an unrelated question, just to be sure: ap_bus.c is compiled as
>>>>> built-in even with ZCRYPT=m, right?
>>>>
>>>> No. If you edit .config and set CONFIG_ZCRYPT=m, ap_bus.c will be built
>>>> into the zcrypt.ko module. Through some other magic, the zcrypt module
>>>> is loaded when linux boots.
>>>
>>> If that happens, then we have a build problem that needs to be
>>> fixed. What exactly are you doing to get the ap code linked into the
>>> zcrypt module?
>>
>> To tell you the truth, I don't know. The build configuration precedes my
>> creating this patch by many years. I only discovered these anomalies by
>> playing with the CONFIG_ZCRYPT option in response to Sebastian's
>> comments. I will have to take this up with our internal IBM maintainer.
>>
>>>
>>
> As Martin already pointed out the ap bus code is statically build into the kernel
> with this line in the Makefile:
>    obj-$(subst m,y,$(CONFIG_ZCRYPT)) += ap.o
> So CONFIG_ZCRYPT may have the value 'm' or 'y'
> and this is also the reason why a simple #ifdef CONFIG_ZCRYPT does not
> work, instead #if IS_ENABLED(CONFIG_ZCRYPT) has to be used.
> 
> Only the ap stuff (which is: ap_bus.o, ap_card.o and ap_queue.o) is statically
> build into the kernel. The zcrypt kernel module is loaded as a dependency
> when a crypto card is detected. So detecting a CEX4 forces kernel module
> zcrypt_cex4.ko which forces the zcrypt.ko to insert into the kernel.
> So all global functions in ap_bus.c are available when CONFIG_ZCRYPT is
> enabled. However, I thought the symbol still needs an EXPORT_SYMBOL()
> statement. Otherwise I should throw out all these statements.
> 
> Harald

Thank you for explaining that Harald.

Tony K

>