On Sun, 2020-08-30 at 18:25 +0300, Denis Efremov wrote:
>
> On 8/30/20 3:43 AM, Joe Perches wrote:
> > $ cat sysfs_emit.cocci
> > @@
> > identifier d_show =~ "^.*show.*$";
>
> I think this additional pattern will allow to take more functions into the scope.
>
> @da@
> identifier show, store;
> expression name, mode;
> @@
>
> (
> DEVICE_ATTR(name, mode, show, store)
> DEVICE_ATTR_PREALLOC(name, mode, show, store)
> DEVICE_ATTR_IGNORE_LOCKDEP(name, mode, show, store)
> )
Thanks Denis.
I'll try that out too.
A trivial grep shows there are at least 130+
DEVICE_ATTR functions that have a show function
that doesn't include "show" in the function name.
$ grep-2.5.4 -rP --include=*.[ch] '\bDEVICE_ATTR\s*\(\s*\w+\s*,\s*[^,]+,\s*[^,]*,[^;]+;' * | \
perl -p -e 's/[[:space:]]*//g; s/;/;\n/g' | \
cut -f3 -d, | \
grep -v show | \
sort | uniq | wc -l
139
> @@
> // I think device_show_ulong, device_show_int, device_show_bool
> // functions deserve explicit handling because they are somewhat
> // reference implementations.
Those reference implementations could be send as
a separate patch but this preliminary script does
already handle them.
I do like the idea below of renaming the show
functions without _show in the name adding _show.
> identifier d_show = { da.show, device_show_ulong, device_show_int, device_show_bool };
> identifier dev, attr, buf;
> @@
>
> * ssize_t d_show(struct device *dev, struct device_attribute *attr, char *buf)
> {
> ...
> }
>
>
> I tried also to handle DEVICE_ATTR_RW, but I failed to use fresh identifier.
> This doesn't work:
>
> @darw@
> identifier name;
> @@
>
> (
> DEVICE_ATTR_RW(name)
> DEVICE_ATTR_RO(name)
> DEVICE_ATTR_WO(name)
> )
>
> @@
> identifier darw.name;
> fresh identifier d_show = name ## "_show"; // <== parse error
> identifier dev, attr, buf;
> @@
>
> * ssize_t d_show(struct device *dev, struct device_attribute *attr, char *buf)
> {
> ...
> }
>
>
> Regards,
> Denis
