Re: [PATCH] HID: Reorder fields in 'struct hid_field'

From: Dan Carpenter
Date: Tue Jun 20 2023 - 01:09:15 EST


On Mon, Jun 19, 2023 at 08:20:08PM +0200, Christophe JAILLET wrote:
> Le 19/06/2023 à 07:18, Dan Carpenter a écrit :
> > On Sun, Jun 18, 2023 at 10:08:07AM +0200, Christophe JAILLET wrote:
> > > diff --git a/include/linux/hid.h b/include/linux/hid.h
> > > index 39e21e3815ad..5be5e671c263 100644
> > > --- a/include/linux/hid.h
> > > +++ b/include/linux/hid.h
> > > @@ -480,9 +480,9 @@ struct hid_field {
> > > __s32 physical_maximum;
> > > __s32 unit_exponent;
> > > unsigned unit;
> > > - bool ignored; /* this field is ignored in this event */
> > > struct hid_report *report; /* associated report */
> > > unsigned index; /* index into report->field[] */
> > > + bool ignored; /* this field is ignored in this event */
> > > /* hidinput data */
> > > struct hid_input *hidinput; /* associated input structure */
> > > __u16 dpad; /* dpad input code */
> >
> > You could move the dpad next to the ignored to save another 4 bytes.
> > I think it is still grouped logically that way but I don't really know
> > what dpad is so I might be wrong.
> >
> > struct hid_report *report; /* associated report */
> > unsigned index; /* index into report->field[] */
> > bool ignored; /* this field is ignored in this event */
> > /* hidinput data */
> > __u16 dpad; /* dpad input code */
> > struct hid_input *hidinput; /* associated input structure */
> >
> > regards,
> > dan carpenter
> >
> >
>
> In fact, not really,
> It would fill a hole better, but would generate some padding at the end of
> the struct:
>
> bool ignored; /* 108 1 */
>
> /* XXX 1 byte hole, try to pack */
>
> __u16 dpad; /* 110 2 */
> struct hid_input * hidinput; /* 112 8 */
> unsigned int slot_idx; /* 120 4 */
>
> /* size: 128, cachelines: 2, members: 25 */
> /* sum members: 119, holes: 2, sum holes: 5 */
> /* padding: 4 */
> };

Ah. Right. I should have looked at that more closely.

regards,
dan carpenter