This indentation looks different than above. Is it correct?

Missing newline.

Why imx_hdmi_read_1 and not RD1?

Missing newline after void.

Extra indentation.

Why 0xFF?

static void\n

What's special about 45250000?

Many magic values

What are all these magic numbers for?

if (err != 0) {

Ditto

Shouldn't err be set?

And here.

Is this needed? I don't see how err could be non-zero here.

Extra space after the (

the freebsd idiom for this is #define EDID_DEBUG_not (although I haven't seen it as much recently as in years past)

the DI_COUNTER and DI_SYNC things seem strangely indented, as if they're subordinate to some missing IPU_DI_something register definitions.

I think if offset were defined as u_int the compiler would have a better chance of converting these divide ops into shifting and masking.

this check should probably be before the 1 << len. Also, this function is called from all those macros above, right? so len > 32 is a programming error, not a runtime error? if so, KASSERT is better.

and likewise in ​ipu_ch_param_get_value() below

a lot of this arithmetic is missing the spaces between operators throughout all these functions

should this be "microcode"?

s/User/Use/

could just pass BUS_DMA_ZERO to bus_dmamem_alloc()

why do we need this have_ipu thing enforcing that only one instance can exist? I don't see any global or static vars or anything that imply there can be only one.

Because if there are more than one IPU instance they all handle HDMI connect event and create framebuffer. This will be handled later when I add support for multiple video paths.