Re: [PATCH 1/3] btrfs: don't allow trans ioctl on a directory

[Josef Bacik <josef@toxicpanda.com>]

On Mon, Jul 24, 2017 at 02:42:29PM +0200, David Sterba wrote:
> On Fri, Jul 21, 2017 at 01:29:07PM -0400, josef@toxicpanda.com wrote:
> > From: Josef Bacik <jbacik@fb.com>
> > 
> > We need to use file->private_data for readdir on directories, so just
> > don't allow user space transactions on directories.
> > 
> > Signed-off-by: Josef Bacik <jbacik@fb.com>
> > ---
> >  fs/btrfs/ioctl.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
> > index bedeec6..ddb3811 100644
> > --- a/fs/btrfs/ioctl.c
> > +++ b/fs/btrfs/ioctl.c
> > @@ -3968,6 +3968,9 @@ static long btrfs_ioctl_trans_start(struct file *file)
> >  	struct btrfs_trans_handle *trans;
> >  	int ret;
> >  
> > +	if (S_ISDIR(inode->i_mode))
> > +		return -EINVAL;
> 
> You can't do this, starting a transaction on a directory needs to work.
> The most natural way to run the ioctl is on the mount point.
> 
> The file private data would need to be able to hold multipe values, so
> you can add
> 
> struct btrfs_inode {
> 	...
> 	struct priv_data {
> 		void *for_readdir;
> 		void *for_tranc_ioctl;
> 	};
> 	...
> };
> 
> then set file->file_private = &btrfs_inode->priv_data; and update all
> uses to check for the embedded pointers.
> 

Blah I really want to just jetison the user space transaction stuff altogether
so I was hoping this would be a first step.  But yeah we can do it your way too.
Thanks,

Josef