Re: [PATCH] IRQ: ease out-of-tree migration to new irq_handler prototype

[William D Waddington <william.waddington@beezmo.com>]

Alexey Dobriyan wrote:
> On Sat, Nov 04, 2006 at 12:06:53PM -0800, William D Waddington wrote:
> 
>>Alexey Dobriyan wrote:
>>
>>>On Sat, Nov 04, 2006 at 10:29:37AM -0800, William D Waddington wrote:
>>>
>>>
>>>>Ease out-of-tree driver migration to new irq_handler prototype.
>>>>Define empty 3rd argument macro for use in multi kernel version
>>>>out-of-tree drivers going forward.  Backportable drives can do:
>>>>
>>>>(in a header)
>>>>#ifndef __PT_REGS
>>>># define __PT_REGS , struct pt_regs *regs
>>>>#endif
>>>
>>>
>>>Backportable drivers should check kernel version themselves and define
>>>__PT_REGS themselves.
>>
>>I think I provided too much information :(  It would be sufficiently
>>helpful to just #define __PT_REGS <nothing> in  interrupt.h to make
>>things easier for low-life out-of-tree maintainers.  There isn't any
>>need to actualy detect version.  Just detect __PT_REGS already defined.
> 
> 
> Out-of-tree maintainer will have to change his code ANYWAY. And while he
> is doing that, he can spend 10 seconds to add 5-line version check.

I'm a little out of my depth here (obviously) but why is it "better" to
require a version check rather than the kernel simply flagging the
function change?  Rather like HAVE_COMPAT_IOCTL.

With the proposed 1-line kernel patch my drivers build against kernels
back to 2.6.9 (the earliest I have around here) without any version
checking.

> More, if you've followed pt_regs removal patches, you'd noticed that
> some of them were not trivial. In this case version check is least of
> his worries.

Ah. I don't use the *regs arg.  The one line patch just keeps the
compiler happy. And provides a "tidy" way to detect the interface
change if necessary.

>>The "in a header" above referred to the driver's header - #ifdefs in
>>executable code really looks nasty IMHO.
>>
>>The "#define __PT_REGS , ..." comment below was intended to be a
>>"helpful" note to driver writers.  Like I said, TMI.
>>
>>
>>>>(in code body)
>>>>static irqreturn_t irq_handler(int irq, void *dev_id __PT_REGS)
>>>
>>>
>>>>+/*
>>>>+ * Irq handler migration helper - empty 3rd argument
>>>>+ * #define __PT_REGS , struct pt_regs *regs
>>>>+ * for older kernel versions
>>>>+ */
>>>>+
>>>>+#define __PT_REGS
>>
>>How should I tidy this up - if it is acceptable at all?
> 
> 
> No, this is not acceptable.

Oh well.

Thanks for your time,
Bill
-- 
--------------------------------------------
William D Waddington
Bainbridge Island, WA, USA
william.waddington@beezmo.com
--------------------------------------------
"Even bugs...are unexpected signposts on
the long road of creativity..." - Ken Burtch