Protect EHCI clock reference counter with a lock.

• Only activate EHCI and required PHY clocks, not OHCI or USB DRD.
• Add EHCI clock ref count and only deactivate EHCI clocks when there are no users.
• Rename a31_clk_usb_activate/deactivate to a31_clk_ehci_activate/deactivate.

• Use GPIO_MAP_GPIOS to find pin number.

In D5467#116566, @manu_bidouilliste.com wrote:

Should we rename now a10_ehci.c to aw_ehci.c ?
And start doing incremental renaming file for when we change the drivers ?

• Removed hard-coded GPIO pin numbers from a10_ehci.c
• Added new awusbphy driver to handle GPIO pin settings from FDT
• Use compat data to handle differences between SoC families
• Add A31 USB clock deactivate function

• Fix a comment

• Hide edid_print with bootverbose
• Fix size of struct resource * array in a10fb_softc
• No need for struct resource * array in a10hdmi_softc
• Style fixes
• Use pcm_getbuffersize to determine a10hdmiaudio DMA buffer size

#define<tab>

Added test for sensor type in axp209_sensor, cleanup whitespace.

• Fix cubieboard2.dts merge issue (codec is now enabled in sun7i-a20-cubieboard.dts)
• include <arm/allwinner/a10_clk.h> instead of "a10_clk.h"
• Add DRQ types for A10 (same as A20)

Add comment explaining 24.576MHz audio source clock.

Make DDMA-specific parameters configurable.

Rename "pamute-gpios" property to "allwinner,pa-gpios" to match Linux mainline.
PA mute GPIO not required for Cubieboard2.

Remove unused headers and sc_dev softc member.

Use bus_alloc_resources

Moved comment as suggested.

Reducing scope of this diff to the DMA controller driver only. Changes based on previous feedback.

In D5050#107261, @andrew wrote:

Why are you adding both drivers in one review? I would expect the audio driver could depend on DMA, but not the reverse.