Re: [PATCH] coresight: do not use the BIT() macro in the UAPI header

[Eugene Syromiatnikov]

On Tue, Mar 24, 2020 at 11:19:38AM +0100, Greg Kroah-Hartman wrote:
> On Tue, Mar 24, 2020 at 10:53:04AM +0100, Eugene Syromiatnikov wrote:
> > On Tue, Mar 24, 2020 at 07:28:53AM +0100, Greg Kroah-Hartman wrote:
> > > On Tue, Mar 24, 2020 at 05:22:13AM +0100, Eugene Syromiatnikov wrote:
> > > > The BIT() macro definition is not available for the UAPI headers
> > > > (moreover, it can be defined differently in the user space); replace
> > > > its usage with the _BITUL() macro that is defined in <linux/const.h>.
> > > 
> > > Why is somehow _BITUL() ok to use here instead?
> > 
> > It is provided in an UAPI header (include/uapi/linux/const.h)
> > and is appropriately prefixed with an underscore to avoid conflicts.
> 
> Because no one uses _ in their own macros?  :)

Well, it is a reserved prefix (ANSI C99, 4.1.2 "Standard headers": "All
other identifiers that begin with an underscore and either an upper-case
letter or another underscore are reserved"), so valid C files shouldn't
use them.

> Anyway, this is fine, but if it's really the way forward, I think a lot
> of files will end up being changed...

There are 5 cases for using BIT() in UAPI headers so far (rtc.h[1],
serio.h[2], psci.h[3], pkt_sched.h[4], coresight-stm.h), two of them were
conversions from the open-coded variant; the usage of _BITUL in pkt_sched.h
made me think that it is the better approach since people tend to use
BIT-like macro anyway, so, by increasing a number of cases it may raise
awareness of the UAPI specifics.

[1] https://lore.kernel.org/lkml/20200324041209.GA30727@xxxxxxxxxxxxxxxxx/
[2] https://lore.kernel.org/lkml/20200324041341.GA32335@xxxxxxxxxxxxxxxxx/
[3] https://lore.kernel.org/lkml/20200324041526.GA1978@xxxxxxxxxxxxxxxxx/
[4] https://lore.kernel.org/lkml/20200324041920.GA7068@xxxxxxxxxxxxxxxxx/

> 
> thanks,
> 
> greg k-h
>