Re: [PATCH 1/2] asm-generic/io.h: provide default ioremap/iounmap for !HAS_IOMEM

[Rob Herring]

On Wed, Mar 30, 2016 at 3:08 PM, Geert Uytterhoeven
<geert@xxxxxxxxxxxxxx> wrote:
> Hi Rob,
>
> On Tue, Mar 29, 2016 at 10:13 PM, Rob Herring <robh@xxxxxxxxxx> wrote:
>> On Tue, Mar 29, 2016 at 2:37 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
>>> On Tuesday 29 March 2016 13:23:00 Rob Herring wrote:
>>>> Drivers shouldn't have to care about HAS_IOMEM to compile and having to
>>>> causes a Kconfig mess:
>>>>
>>>> warning: (MEDIA_SUBDRV_AUTOSELECT && VIDEO_CX231XX && INV_MPU6050_I2C) selects I2C_MUX which has unmet direct dependencies (I2C && HAS_IOMEM)
>>>> warning: (ST_IRQCHIP && STMMAC_PLATFORM && DWMAC_IPQ806X && DWMAC_LPC18XX && DWMAC_ROCKCHIP && DWMAC_SOCFPGA && DWMAC_STI && TI_CPSW && PINCTRL_ROCKCHIP && PINCTRL_DOVE && POWER_RESET_KEYSTONE && S3C2410_WATCHDOG && VIDEO_OMAP3 && VIDEO_S5P_FIMC && USB_XHCI_MTK && RTC_DRV_AT91SAM9 && LPC18XX_DMAMUX && VIDEO_OMAP4 && HWSPINLOCK_QCOM && ATMEL_ST && QCOM_GSBI && PHY_HI6220_USB) selects MFD_SYSCON which has unmet direct dependencies (HAS_IOMEM)
>>>>
>>>> Reuse the !MMU variants for !HAS_IOMEM as they are sufficient for our
>>>> needs. This fixes build errors for UM allyesconfig:
>>>>
>>>> drivers/mfd/syscon.c:89:2: error: implicit declaration of function 'iounmap' [-Werror=implicit-function-declaration]
>>>>      iounmap(base);
>>>>
>>>> Reported-by: Fengguang Wu <fengguang.wu@xxxxxxxxx>
>>>> Cc: Arnd Bergmann <arnd@xxxxxxxx>
>>>> Cc: linux-arch@xxxxxxxxxxxxxxx
>>>> Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
>>>
>>> (adding Richard and the UML list to cc)
>>>
>>> I actually prototyped a patch that did the opposite: remove the readl/writel/...
>>> definitions when HAS_IOMEM is unset. I didn't get far enough to submit it,
>>> but see below for what I did.
>>
>> Ewww. Why do the opposite of what we do for every other Kconfig symbol
>> which is provide empty functions? It really only functions as a
>> disable on UML flag which runs counter to enabling drivers to build
>> for all arches.
>
> Usually the empty function fall into one of two classes:
>   1. They return an error, so drivers will abort their initialization,
>   2. They are a plain no-op, for functions with harmless side-effects.
>
> The !MMU versions are not dummies, but assume a transparent translation.
> This may lead to drivers continuing their initialization and crashing the
> system.

I thought about this, but how could you even get to the point of
having some physical address passed to the driver? Perhaps some old
ISA driver, but I'm not sure memory ranges were ever fixed (only i/o).
We're really only talking about UML here.

It can be done though, it's just more ifdef'ery.

>> I actually started a patch to remove the HAS_IOMEM dependency
>> everywhere (or just the per driver cases). It didn't break as bad as I
>> expected, but became more than I wanted to fix. Mainly, all the devm_
>> variants also need empty versions or to be always enabled.
>
> Should these dependencies on HAS_IOMEM be changed to "HAS_IOMEM ||
> COMPILE_TEST"?

Why? Most are either HAS_IOMEM or HAS_IOMEM && COMPILE_TEST.

Rob