On Wed, Mar 06, 2024 at 04:43:11PM +0100, Herve Codina wrote:

On Wed, 6 Mar 2024 15:43:04 +0200
Andy Shevchenko [off-list ref] wrote:

quoted

On Wed, Mar 06, 2024 at 05:06:12AM -0800, Yury Norov wrote:

quoted

On Wed, Mar 06, 2024 at 09:07:20AM +0100, Herve Codina wrote:  

...

quoted

quoted

quoted

+	DECLARE_BITMAP(ts_mask_avail, 64);
+	DECLARE_BITMAP(ts_mask, 64);
+	DECLARE_BITMAP(map, 64);  

quoted

quoted

+	bitmap_from_u64(ts_mask_avail, ts_info->rx_ts_mask_avail);
+	bitmap_from_u64(map, slot_map);  

quoted

We've got a BITMAP_FROM_U64() for this:

	DECLARE_BITMAP(ts_mask_avail, 64) = { BITMAP_FROM_U64(ts_info->rx_ts_mask_avail) };
	DECLARE_BITMAP(map, 64) = { BITMAP_FROM_U64(slot_map) };  

This looks ugly. Can we rather provide a macro that does this under the hood?

Roughly:

#define DEFINE_BITMAP_64(name, src)				\
	DECLARE_BITMAP(name, 64) = { BITMAP_FROM_U64(src) }

Well, the construction I used:
	DECLARE_BITMAP(foo, 64);
	...
	bitmap_from_u64(foo, init_value);
	...
can be found in several places in the kernel.

Having the DEFINE_BITMAP_64() macro can be a way to remove this
construction but I am not sure that this should be done in this
series.

I also think that this can be done later, above is just a pure suggestion how.

IMHO, a specific series introducing the macro and updating pieces of
code in the kernel everywhere it is needed to replace this construction
would make much more sense.

Right.

-- 
With Best Regards,
Andy Shevchenko