Re: [-next PATCH 2/4] treewide: Use DEVICE_ATTR_RW

From: Julia Lawall
Date: Wed Dec 20 2017 - 08:26:37 EST




On Wed, 20 Dec 2017, Joe Perches wrote:

> On Wed, 2017-12-20 at 10:59 +0100, Greg Kroah-Hartman wrote:
> > > > Why you didn't send that patch to the sysfs maintainer is a bit odd... :)
> > >
> > > So here's an opportunity for you:
> > >
> > > The sysfs maintainer hasn't added include/linux/sysfs.h to
> > > the list of maintained files...
> > >
> > > DRIVER CORE, KOBJECTS, DEBUGFS AND SYSFS
> > > M: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > T: git git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core.git
> > > S: Supported
> > > F: Documentation/kobject.txt
> > > F: drivers/base/
> > > F: fs/debugfs/
> > > F: fs/sysfs/
> > > F: include/linux/debugfs.h
> > > F: include/linux/kobj*
> > > F: lib/kobj*
> >
> > Heh, good point, but using get_maintainer.pl does put me at the top of
> > the list that you should be cc:ing:
> >
> > $ ./scripts/get_maintainer.pl --file include/linux/sysfs.h
> > Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> (commit_signer:3/3=100%,authored:2/3=67%,added_lines:7/8=88%)
> > Kate Stewart <kstewart@xxxxxxxxxxxxxxxxxxx> (commit_signer:1/3=33%)
> > Thomas Gleixner <tglx@xxxxxxxxxxxxx> (commit_signer:1/3=33%)
> > Philippe Ombredanne <pombredanne@xxxxxxxx> (commit_signer:1/3=33%)
> > Nick Desaulniers <nick.desaulniers@xxxxxxxxx> (commit_signer:1/3=33%,authored:1/3=33%,added_lines:1/8=12%,removed_lines:1/1=100%)
> > linux-kernel@xxxxxxxxxxxxxxx (open list)
>
> The script I use to send patches adds --nogit --nogit-fallback
> to copy only listed maintainers because people that send cleanup
> patches don't generally like to get random patches.
>
> > > btw: there are many uses of a reversed declaration style of DEVICE_ATTR
> > >
> > > Here's another thing that could be done for more DEVICE_ATTR_<FOO> uses.
> > >
> > > ===
> > >
> > > Some DEVICE_ATTR definitions use a reversed static function form from
> > > the typical. Convert them to use the more common macro form so it is
> > > easier to grep for the style.
> []
> > > $ git grep --name-only -w DEVICE_ATTR | \
> > > xargs perl -i dev_attr_rw_backwards.perl
> > Ah, nice, I love perl :
>
> That was a bad copy/paste of the script.
>
> The actual script for RW is:
>
> $ cat dev_attr_rw_backwards.perl
> local $/;
> while (<>) {
> my $file = $_;
> while ($file =~ m/\bDEVICE_ATTR\s*\(\s*(\w+)\s*,/g) {
> my $var = $1;
> if ($file =~ s/\bDEVICE_ATTR\s*\(\s*${var}\s*,\s*\(?(\s*S_IRUGO\s*\|\s*S_IWUSR|\s*S_IWUSR\s*\|\s*S_IRUGO\s*|\s*0644\s*)\)?\s*,\s*show_${var}\s*,\s*store_${var}\s*\)/DEVICE_ATTR_RW(${var})/g) {
> $file =~ s/\bshow_${var}\b/${var}_show/g;
> $file =~ s/\bstore_${var}\b/${var}_store/g;
> }
> }
> print $file;
> }
>
> There are 3 different perl scripts for rw, ro, and wo.
> and these scripts, because of function renaming and
> possible reuse of the original function names by other
> string concatenated macros, create some bad conversions
> so they need some manual cleanups too.
>
> Perhaps coccinelle could do a better job of it, but
> likely string concatenation macro uses are going to
> be hard to deal with in any case.

I made a rule for this at one point, but there are cases where the
functions have the wrong names, and then these functions may be used
elsewhere.

julia