Re: [PATCH] Re: [PATCH 4/4 v2] [RFC] UIO: generic platform driver

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Mon, Apr 14, 2008 at 11:54:45AM +0200, Uwe Kleine-König wrote:
> Hi Russell,
> 
> Russell King - ARM Linux wrote:
> > On Mon, Apr 14, 2008 at 09:48:58AM +0200, Uwe Kleine-König wrote:
> > > > > > > > But what about this:
> > > > > > > > 
> > > > > > > > ERROR: "clk_get" [drivers/uio/uio_pdrv.ko] undefined!
> > > > > > > > ERROR: "clk_enable" [drivers/uio/uio_pdrv.ko] undefined!
> > > > > > > > ERROR: "clk_disable" [drivers/uio/uio_pdrv.ko] undefined!
> > > > > > > > ERROR: "clk_put" [drivers/uio/uio_pdrv.ko] undefined!
> > > > > > > > 
> > > > > > > > Do you have any extra patches applied?
> > > > > > > Yes I have, but nothing special.  This is part of a generic API defined
> > > > > > > in include/linux/clk.h.  One of it's use it to abstract away some
> > > > > > > platform dependencies.  There are several architectures that define
> > > > > > > it[1].
> > > > > > 
> > > > > > I know. Unfortunately, I tested on x86_64, and it doesn't compile.
> > > > > > If it's depending on something, then this dependency should be added in
> > > > > > Kconfig. If it can be selected in the configuration, I expect it to
> > > > > > compile (and work).
> > > > > Maybe adding a dummy implementation that is compiled for machines that
> > > > > don't provide a native one.  Currently there is no cpp symbol that tells
> > > > > if an machine supports the API.
> > > > > 
> > > > > @Russell: Do you have an opinion regarding this!?
> > > > 
> > > > Only that the kernels Kconfig is turning into a real complicated mess
> > > > of dependencies IMHO.
> > > > 
> > > > We could add a HAVE_CLK and add that to the dependency of all the drivers
> > > > which use linux/clk.h.  The problem will be finding all those drivers and
> > > > their corresponding Kconfig entries.
> > > > 
> > > > My feeling is that we're just going to end up creating another Kconfig
> > > > symbol which folk half-heartedly use.
> > >
> > > I don't like that either.  What do you think about the patch below?
> > > It doesn't introduce a new symbol that needs much care and attention.
> > > This way the clk API is available on all configurations provided that 
> > > CONFIG_DUMMY_CLK is set correctly.  If CONFIG_DUMMY_CLK is set wrong it
> > > should result in a compile error.  Either because there are two
> > > implementations of clk_get or none.
> > 
> > Hang on.  I'm lost.  What are we talking about here?  I thought the
> > thread was about the one liner patch for UIO to arch/arm/Kconfi
> > (which still hasn't hit the patch system so is still on target for
> > being missed...)
> No, the topic here is a generic uio platform driver.  It uses the clk
> API and Hans criticised that is doesn't compile on x86 (because there is
> no implementation of the clk API).  So I suggested to implement a dummy
> for that.
> 
> This is completly independant of the inclusion of drivers/uio/Kconfig in
> arch/arm/Kconfig.  I will send a patch for that.
> 
> > What's this drivers/uio/uio_pdrv.ko module, and why doesn't it appear
> > in the LKML archive of this thread?
> Don't know why lkml.org didn't link these.  The start of the thread can
> be found at
> 
> 	http://lkml.org/lkml/2008/4/10/110

Thanks.  Well, tbh, I don't know which way to go on this.  Each of the
suggested ways have their downsides.

However:

+       pdata->clk = clk_get(&pdev->dev, DRIVER_NAME);

seems wrong - "uio" as a clock name?

+               /* XXX: better use dev_dbg, but which device should I use?
+                * info->uio_dev->dev isn't accessible here as struct uio_device+                * is opaque.
+                */

why not store a copy of 'dev' in struct uio_platdata ?

+       uiomem = &uioinfo->mem[0];
+       for (i = 0; i < pdev->num_resources; ++i) {
...
+               ++uiomem;
+       }

Who's to say there's pdev->num_resources entries in the 'mem' array?
Shouldn't this loop also be limited to MAX_UIO_MAPS iterations (or
maybe complain if there's more than MAX_UIO_MAPS)?

+/* XXX: I thought there already exists something like STRINGIFY, but I could not
+ * find it ...
+ */
+#ifndef STRINGIFY
+#define STRINGIFY(x) __STRINGIFY_HELPER(x)
+#define __STRINGIFY_HELPER(x) #x
+#endif

#include <linux/stringify.h> ?