Re: [PATCH next 0/3] debugfs: introduce debugfs_create_single/seq[,_data]

[Greg KH]

On Fri, Nov 29, 2019 at 11:19:38PM +0100, Greg KH wrote:
> On Fri, Nov 29, 2019 at 11:16:38PM +0800, Kefeng Wang wrote:
> > 
> > 
> > On 2019/11/29 22:21, Greg KH wrote:
> > > On Fri, Nov 29, 2019 at 05:27:49PM +0800, Kefeng Wang wrote:
> > >> Like proc_create_single/seq[,_data] in procfs, we could provide similar debugfs
> > >> helper to reduce losts of boilerplate code.
> > >>
> > >> debugfs_create_single[,_data]
> > >>   creates a file in debugfs with the extra data and a seq_file show callback.
> > >> debugfs_create_seq[,_data]
> > >>   creates a file in debugfs with the extra data and a seq_operations.
> > >>
> > >> There is a object dynamically allocated in the helper, which is used to store
> > >> extra data, we need free it when remove the debugfs file.
> > >>
> > >> If the change is acceptable, we could change the caller one by one.
> > > 
> > > I would like to see a user of this and how you would convert it, in
> > > order to see if this is worth it or not.
> > 
> > I have some diff patches, the conversion is in progress. current statistics
> > are as follows,
> > 
> > 1) debugfs: switch to debugfs_create_seq[,_data]
> > 19 files changed, 85 insertions(+), 620 deletions(-)
> > 2) debugfs: switch to debugfs_create_single[,_data]
> > 70 files changed, 249 insertions(+), 1482 deletions(-)
> > 
> > Here are some examples,
> > 1) debugfs_create_seq
> > diff --git a/mm/vmstat.c b/mm/vmstat.c
> > index 78d53378db99..62c26772f24c 100644
> > --- a/mm/vmstat.c
> > +++ b/mm/vmstat.c
> > @@ -2057,18 +2057,6 @@ static const struct seq_operations unusable_op = {
> >  	.show	= unusable_show,
> >  };
> > 
> > -static int unusable_open(struct inode *inode, struct file *file)
> > -{
> > -	return seq_open(file, &unusable_op);
> > -}
> > -
> > -static const struct file_operations unusable_file_ops = {
> > -	.open		= unusable_open,
> > -	.read		= seq_read,
> > -	.llseek		= seq_lseek,
> > -	.release	= seq_release,
> > -};
> > -
> 
> Can't this file just use the normal file macro/interface for debugfs
> files instead?  Hm, maybe not, it seems the celf code wants to do much
> the same as above, but is seq_read() really needed for these?

I refer to DEFINE_SIMPLE_ATTRIBUTE(), sorry for not saying that here.

And if they do not work, how about creating:
	DEFINE_SEQ_ATTRIBUTE()
in much the same way for the whole kernel to use.

thanks,

greg k-h