Re: [Xenomai] [Xenomai-git] Gilles Chanteperdrix : hal/x86: fix smi message

[Philippe Gerum <rpm@xenomai.org>]

On 09/29/2013 06:10 PM, Jan Kiszka wrote:
> On 2013-09-29 18:00, Philippe Gerum wrote:
>> On 09/29/2013 05:46 PM, Jan Kiszka wrote:
>>> On 2013-09-29 17:45, Philippe Gerum wrote:
>>>> On 09/29/2013 05:37 PM, Jan Kiszka wrote:
>>>>> On 2013-09-29 16:56, Philippe Gerum wrote:
>>>>>> On 09/29/2013 12:12 PM, Jan Kiszka wrote:
>>>>>>> On 2013-09-28 18:30, Philippe Gerum wrote:
>>>>>>>> On 09/28/2013 01:12 PM, Gilles Chanteperdrix wrote:
>>>>>>>>> On 09/28/2013 11:52 AM, Philippe Gerum wrote:
>>>>>>>>>> On 09/28/2013 10:36 AM, Gilles Chanteperdrix wrote:
>>>>>>>>>>> On 09/28/2013 10:28 AM, Philippe Gerum wrote:
>>>>>>>>>>>> On 09/27/2013 10:19 PM, git repository hosting wrote:
>>>>>>>>>>>>> Module: xenomai-gch
>>>>>>>>>>>>> Branch: for-forge
>>>>>>>>>>>>> Commit: 283fabdf63e23f03a45565fa1f9937f9a5825a59
>>>>>>>>>>>>> URL:
>>>>>>>>>>>>> http://git.xenomai.org/?p=xenomai-gch.git;a=commit;h=283fabdf63e23f03a45565fa1f9937f9a5825a59
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Author: Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
>>>>>>>>>>>>> Date:   Sun Sep 15 15:59:52 2013 +0200
>>>>>>>>>>>>>
>>>>>>>>>>>>> hal/x86: fix smi message
>>>>>>>>>>>>>
>>>>>>>>>>>>> ---
>>>>>>>>>>>>>
>>>>>>>>>>>>>         kernel/cobalt/arch/x86/smi.c |    2 +-
>>>>>>>>>>>>>         1 files changed, 1 insertions(+), 1 deletions(-)
>>>>>>>>>>>>>
>>>>>>>>>>>>> diff --git a/kernel/cobalt/arch/x86/smi.c
>>>>>>>>>>>>> b/kernel/cobalt/arch/x86/smi.c
>>>>>>>>>>>>> index e06734a..124d311 100644
>>>>>>>>>>>>> --- a/kernel/cobalt/arch/x86/smi.c
>>>>>>>>>>>>> +++ b/kernel/cobalt/arch/x86/smi.c
>>>>>>>>>>>>> @@ -123,7 +123,7 @@ void mach_x86_smi_init(void)
>>>>>>>>>>>>>             if (smi_state == 0) {
>>>>>>>>>>>>>                 printk("Xenomai: SMI-enabled chipset found, but
>>>>>>>>>>>>> SMI "
>>>>>>>>>>>>>                        "workaround disabled\n"
>>>>>>>>>>>>> -               "         (see xeno_hal.smi parameter). You may
>>>>>>>>>>>>> encounter\n"
>>>>>>>>>>>>> +               "         (see x86_hal.smi parameter). You may
>>>>>>>>>>>>> encounter\n"
>>>>>>>>>>>>>                        "         high interrupt latencies!\n");
>>>>>>>>>>>>>                 pci_dev_put(dev);
>>>>>>>>>>>>>                 return;
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> Strictly speaking, the HAL disappeared from x3, since most of
>>>>>>>>>>>> the
>>>>>>>>>>>> arch-independant interface is already provided by the interrupt
>>>>>>>>>>>> pipeline, and the event driven simulator was dropped. We should
>>>>>>>>>>>> not
>>>>>>>>>>>> reintroduce it in the naming.
>>>>>>>>>>>>
>>>>>>>>>>> The parameter name is due to this Makefile chunk:
>>>>>>>>>>>
>>>>>>>>>>> obj-$(CONFIG_XENOMAI) += x86_hal.o
>>>>>>>>>>> x86_hal-y := machine.o mayday.o thread.o smi.o
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> This is a left over, like the message you fixed. Should be
>>>>>>>>>> x86_machine
>>>>>>>>>> or something along these lines.
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> x86 alone maybe?
>>>>>>>>>
>>>>>>>>
>>>>>>>> Whatever fits. It's just that the code living under
>>>>>>>> kernel/cobalt/arch/*
>>>>>>>> is referred to as the machine layer elsewhere.
>>>>>>>
>>>>>>> Shouldn't this just be added to the Xenomai module (xenomai-y)?
>>>>>>>
>>>>>>
>>>>>> xenomai-y is supposed to include all arch-independent bits, for this
>>>>>> reason the machine layer is rather pulled from core-y in arch/*.
>>>>>
>>>>> If you want a different module name, then please make at least sure
>>>>> that
>>>>> it includes "xenomai", "xenomai_arch" or so. Current proposals are way
>>>>> too generic and confusing for the users.
>>>>>
>>>>
>>>> The point is such information is absolutely not visible to the user.
>>>
>>> And what else controls where the module parameters end up under
>>> /sys/modules?
>>>
>>
>> I'm discussing the internal sub-module name from the Kbuild hierarchy,
>> which is by no mean exported to the user
>
> Sorry, but it is:
>
> 	.module_name    = KBUILD_MODNAME,
>
> So the KBUILD module name defines the parameter prefixes.

Ack, you are right, I overlooked the x86 params, which is unfortunate 
since they sparked the discussion... I moved all others to xenomai-y 
already.

  And that is
> what we should discuss here. I don't mind the kbuild structure otherwise.
>

Merging your proposal and mine, this would give the change below. I 
don't have any preference between "xeno", "xenomai", whatever fits. 
However, we should keep something in the name that clearly gives a hint 
about the machine/arch-dependent nature of the parameter (i.e.
not merging those bits to xenomai-y).

e.g.

diff --git a/kernel/cobalt/arch/arm/Makefile 
b/kernel/cobalt/arch/arm/Makefile
index 25d33a5..d62976c 100644
--- a/kernel/cobalt/arch/arm/Makefile
+++ b/kernel/cobalt/arch/arm/Makefile
@@ -1,5 +1,5 @@
-obj-$(CONFIG_XENOMAI) += arm_hal.o
+obj-$(CONFIG_XENOMAI) += xeno_machine.o

-arm_hal-y := machine.o mayday.o thread.o switch.o syscall.o
+xeno_machine-y := machine.o mayday.o thread.o switch.o syscall.o

  ccflags-y := -Iarch/arm/xenomai/include -Iinclude/xenomai

-- 
Philippe.