On 2018-06-18 13:54, Wolfram Sang wrote:

quoted

quoted

I wonder if i2c_lock_segment() and i2c_lock_root_adapter() are really
more readable and convenient than i2c_lock_bus() with the flag. I think
the flags have speaking names, too.

Is that an idea to remove these functions altogether and start using
i2c_lock_bus()?

That would be fine with me. I don't have a strong opinion and agree that
both are readable enough...

It would make for a reduction of the number of lines so that's nice, but
the macro in drivers/i2c/busses/i2c-gpio.c (patch 11) would not fit in
the current \-width (or whatever you'd call that line of backslashes to
the right in a multi-line macro).

Does anyone have a strong opinion?

I have a strong opinion on making i2c.h less bloated. And yes, less
number of lines is nice, too. I think that surely pays off the
whitespace exception.

Ok, I have rebased onto v4.18-rc1, killed the i2c-tegra hunk and converted
i2c_lock_root(foo) over to i2c_lock_bus(foo, I2C_LOCK_ROOT_ADAPTER) and
i2c_lock_segment(foo) over to i2c_lock_bus(foo, I2C_LOCK_SEGMENT). And I
of course killed a bunch of locking helpers in i2c.h.

I doing build tests now, will post a v2 in the morning.

Cheers,
Peter