Something changed in the OF parser in the v3.16 merge window
making it be strict about passing the number of IRQ cells
correctly and disturbing the irqdomain xlate function guard
to crash when subdevices try to obtain IRQs like this:

------------[ cut here ]------------
WARNING: CPU: 0 PID: 1 at
/home/linus/linux-stericsson/kernel/irq/irqdomain.c:676
irq_domain_xlate_twocell+0x40/0x48()
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted
 3.15.0-07915-gf6d059821ce9-dirty #46
[<c0014660>] (unwind_backtrace)
 from [<c0011424>] (show_stack+0x10/0x14)
[<c0011424>] (show_stack)
 from [<c0432630>] (dump_stack+0x9c/0xd4)
[<c0432630>] (dump_stack)
 from [<c001d5c0>] (warn_slowpath_common+0x6c/0x88)
[<c001d5c0>] (warn_slowpath_common)
 from [<c001d678>] (warn_slowpath_null+0x1c/0x24)
[<c001d678>] (warn_slowpath_null)
 from [<c005acd0>] (irq_domain_xlate_twocell+0x40/0x48)
[<c005acd0>] (irq_domain_xlate_twocell)
 from [<c005b658>] (irq_create_of_mapping+0x64/0x110)
[<c005b658>] (irq_create_of_mapping)
 from [<c02e147c>] (of_irq_get+0x38/0x48)
[<c02e147c>] (of_irq_get)
 from [<c01f8910>] (tc3589x_gpio_probe+0x38/0x1e4)
[<c01f8910>] (tc3589x_gpio_probe)
 from [<c022eedc>] (platform_drv_probe+0x18/0x48)
[<c022eedc>] (platform_drv_probe)
 from [<c022d80c>] (driver_probe_device+0x118/0x24c)
[<c022d80c>] (driver_probe_device)
 from [<c022bf20>] (bus_for_each_drv+0x58/0x8c)
[<c022bf20>] (bus_for_each_drv)
 from [<c022d6c4>] (device_attach+0x74/0x88)
[<c022d6c4>] (device_attach)
 from [<c022cdac>] (bus_probe_device+0x84/0xa8)
[<c022cdac>] (bus_probe_device)
 from [<c022b35c>] (device_add+0x440/0x520)
[<c022b35c>] (device_add)
 from [<c022ec50>] (platform_device_add+0xb4/0x218)
[<c022ec50>] (platform_device_add)
 from [<c0243508>] (mfd_add_device+0x220/0x31c)
[<c0243508>] (mfd_add_device)
 from [<c02436a8>] (mfd_add_devices+0xa4/0x100)
[<c02436a8>] (mfd_add_devices)
 from [<c024312c>] (tc3589x_probe+0x334/0x3c0)
[<c024312c>] (tc3589x_probe)
 from [<c022d80c>] (driver_probe_device+0x118/0x24c)

The TC3589x device trees specify the MFD core device
as having one interrupt cell (cannot specify flags) so the
twocell translation function is clearly wrong, changing it to
onecell, as it should be, fixes the regression.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Lee Jones <lee.jones@linaro.org>