Re: [PATCH 1/6 v10] ARM: Add basic architecture support for VIA/WonderMedia 85xx SoC's

[Ryan Mallon <ryan@bluewatersys.com>]

On 12/23/2010 11:02 AM, Alexey Charkov wrote:
> 2010/12/23 Ryan Mallon <ryan@bluewatersys.com>:
>> On 12/23/2010 10:18 AM, Alexey Charkov wrote:
>>> This adds support for the family of Systems-on-Chip produced initially
>>> by VIA and now its subsidiary WonderMedia that have recently become
>>> widespread in lower-end Chinese ARM-based tablets and netbooks.
>>>
>>> Support is included for both VT8500 and WM8505, selectable by a
>>> configuration switch at kernel build time.
>>>
>>> Included are basic machine initialization files, register and
>>> interrupt definitions, support for the on-chip interrupt controller,
>>> high-precision OS timer, GPIO lines, necessary macros for early debug,
>>> pulse-width-modulated outputs control, as well as platform device
>>> configurations for the specific drivers implemented elsewhere.
>>>
>>> Signed-off-by: Alexey Charkov <alchark@gmail.com>
>>> ---
>>>
>>> Welcome the jubilee tenth revision of this patch ;-)
>>>
>>> I've tried to incorporate the suggestions by Ryan and Arnd, hope that
>>> there is nothing left out. There was a massive reorganization of code
>>> to remove less-than-obvious magic with MMIO registers and irqs being
>>> held in huge structs, now they are again macro definitions. Those
>>> macros are, however, only included in single isolated files, and
>>> actual values to use are chosen at runtime by calling the respective
>>> routines at machine initialization. There are also stylistic changes
>>> all around, where Ryan suggested.
>>>
>>> As a result, i8042 should again be adjusted a bit to reflect the new
>>> place to find respective register/irq definitions, that one will be
>>> sent in the respective thread shortly.
>>>
>>> Best regards,
>>> Alexey
>>
>> <snip>
>>
>>> --- /dev/null
>>> +++ b/arch/arm/mach-vt8500/devices-vt8500.c
>>> @@ -0,0 +1,117 @@
>>> +/* linux/arch/arm/mach-vt8500/devices-vt8500.c
>>> + *
>>> + * Copyright (C) 2010 Alexey Charkov <alchark@gmail.com>
>>> + *
>>> + * This software is licensed under the terms of the GNU General Public
>>> + * License version 2, as published by the Free Software Foundation, and
>>> + * may be copied, distributed, and modified under those terms.
>>> + *
>>> + * This program is distributed in the hope that it will be useful,
>>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>>> + * GNU General Public License for more details.
>>> + *
>>> + */
>>> +
>>> +#include <linux/platform_device.h>
>>> +
>>> +#include <mach/vt8500_regs.h>
>>> +#include <mach/vt8500_irqs.h>
>>> +#include <mach/i8042.h>
>>> +#include "devices.h"
>>> +
>>> +static inline struct resource WMT_MMIO_RES(u32 start, u32 size)
>>> +{
>>> +     struct resource tmp = {
>>> +             .flags = IORESOURCE_MEM,
>>> +             .start = start,
>>> +             .end = start + size - 1,
>>> +     };
>>> +
>>> +     return tmp;
>>> +}
>>
>> These functions can be marked __init (though I guess they already are if
>> marked inline?). They should also have lower case names since they are
>> proper functions.
> 
> As inline functions are unrolled into the caller at compile time, and
> the caller is __init, I would expect their code to be freed after init
> as well. I could be wrong, though :)

If the compiler decided not to inline them for whatever reason they
would stick around. They should also probably be marked __init to make
it clear that they are not permanent functions.

>> Do these functions generate warnings about returning temporary values
>> off the stack? If so, they could be rewritten as:
>>
> 
> Should those be compile-time or run-time? I did not see any.

I thought you might get compile time warnings. Returning things off the
stack in general is error prone. I think passing a pointer to the
resource into the function (as below) maybe a bit better.

>> static __init void wmt_mmio_res(struct resource *res,
>>                                u32 start, u32 size)
>> {
>>        res->flags = IORESOURCE_MEM;
>>        res->start = start;
>>        res->end   = start + size - 1;
>> }
>>
>> You could also make these functions static inline in devices.h to avoid
>> having to define them for each board.
>>
> 
> Agreed.
> 
>>> +
>>> +static inline struct resource WMT_IRQ_RES(int irq)
>>> +{
>>> +     struct resource tmp = {
>>> +             .flags = IORESOURCE_IRQ,
>>> +             .start = irq,
>>> +             .end = irq,
>>> +     };
>>> +
>>> +     return tmp;
>>> +}
>>> +
>>> +#define WMT_RES_ADD(__device, __resource, __num) \
>>> +if (platform_device_add_resources(__device, __resource, __num)) \
>>> +     pr_err("Failed to assign resources to __device##\n");
>>
>> This could be written as a proper function. The resource add is unlikely
>> to fail. Maybe keep the warning but don't worry about printing the
>> device name?
>>
> 
> There is memory allocation inside platform_device_add_resources, so
> probably there is scope for failure. I could add unlikely(), though.

I meant more in terms of losing the device name in the error which seems
to be the sole point of the macro. i.e it could be rewritten as:

static void wmt_add_resource(struct platform_device *pdev,
				const struct resource *res,
				unsigned num)
{
	if (platform_device_add_resources(pdev, res, num))
		pr_err("Failed to add device resource\n");
}

Because the add resource failing would indicate some serious problem you
may even want to BUG().

~Ryan

-- 
Bluewater Systems Ltd - ARM Technology Solution Centre

Ryan Mallon         		5 Amuri Park, 404 Barbadoes St
ryan@bluewatersys.com         	PO Box 13 889, Christchurch 8013
http://www.bluewatersys.com	New Zealand
Phone: +64 3 3779127		Freecall: Australia 1800 148 751
Fax:   +64 3 3779135			  USA 1800 261 2934