RE: [PATCH] staging: unisys: visorbus: constify attribute_group structures.

From: Kershner, David A
Date: Mon Jul 17 2017 - 17:37:07 EST


> -----Original Message-----
> From: Greg KH [mailto:greg@xxxxxxxxx]
> Sent: Monday, July 17, 2017 8:38 AM
> To: Arvind Yadav <arvind.yadav.cs@xxxxxxxxx>
> Cc: Kershner, David A <David.Kershner@xxxxxxxxxx>; Sell, Timothy C
> <Timothy.Sell@xxxxxxxxxx>; Thompson, Bryan E.
> <bryan.thompson@xxxxxxxxxx>; jon.frisch@xxxxxxxxxx; Binder, David
> Anthony <David.Binder@xxxxxxxxxx>; *S-Par-Maintainer
> <SParMaintainer@xxxxxxxxxx>; devel@xxxxxxxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH] staging: unisys: visorbus: constify attribute_group
> structures.
>
> On Mon, Jul 17, 2017 at 05:43:14PM +0530, Arvind Yadav wrote:
> > Hi Greg,
> >
> >
> > On Monday 17 July 2017 04:15 PM, Greg KH wrote:
> > > On Mon, Jul 17, 2017 at 02:55:37PM +0530, Arvind Yadav wrote:
> > > > attribute_groups are not supposed to change at runtime. All functions
> > > > working with attribute_groups provided by <linux/sysfs.h> work
> > > > with const attribute_group. So mark the non-const structs as const.
> > > >
> > > > Signed-off-by: Arvind Yadav <arvind.yadav.cs@xxxxxxxxx>
> > > > ---
> > > > drivers/staging/unisys/visorbus/visorbus_main.c | 4 ++--
> > > > drivers/staging/unisys/visorbus/visorchipset.c | 2 +-
> > > > 2 files changed, 3 insertions(+), 3 deletions(-)
> > > Why not just use the ATTRIBUTE_GROUPS() macro for these? Or is there
> > > something that is preventing that?
> > Yes, we can use. if we are only initializing '.attrs'.
> > ATTRIBUTE_GROUPS() will not work if we will initialize other member of
> > attribute_group like 'bin_attrs', 'is_visible', and 'name'.
>
> That means you should redo this patch :)
>
> Also, your changelog text had a typo, it is "attribute_group", not
> "attribute_groups".
>

Greg, are you recommending that we shouldn't be setting the attribute_group
.name field? What does it pick up if we don't specify it?

Also, for our attribute_groups in visorchipset, we are defining it with two
different attribute_group variables. Are you allowed to use two different
attribute_group variables in an attribute_groups, or is this frowned upon and
we should flatten it out to just one? An example that we used in the kernel was:

static const struct attribute_group *l2_cache_pmu_attr_grps[] = {
&l2_cache_pmu_format_group,
&l2_cache_pmu_cpumask_group,
NULL,
};

Thanks,
David Kershner

> thanks,
>
> greg k-h