Re: [PATCH RFC] libfs: getdents() should return 0 after reaching EOD

[Chuck Lever <chuck.lever@oracle.com>]

On Tue, Nov 14, 2023 at 06:29:15PM +0100, Christian Brauner wrote:
> On Tue, Nov 14, 2023 at 10:49:37AM -0500, Chuck Lever wrote:
> > From: Chuck Lever <chuck.lever@oracle.com>
> > 
> > The new directory offset helpers don't conform with the convention
> > of getdents() returning no more entries once a directory file
> > descriptor has reached the current end-of-directory.
> > 
> > To address this, copy the logic from dcache_readdir() to mark the
> > open directory file descriptor once EOD has been reached. Rewinding
> > resets the mark.
> > 
> > Reported-by: Tavian Barnes <tavianator@tavianator.com>
> > Closes: https://lore.kernel.org/linux-fsdevel/20231113180616.2831430-1-tavianator@tavianator.com/
> > Fixes: 6faddda69f62 ("libfs: Add directory operations for stable offsets")
> > Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
> > ---
> >  fs/libfs.c |   13 ++++++++++---
> >  1 file changed, 10 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/libfs.c b/fs/libfs.c
> > index e9440d55073c..1c866b087f0c 100644
> > --- a/fs/libfs.c
> > +++ b/fs/libfs.c
> > @@ -428,7 +428,7 @@ static bool offset_dir_emit(struct dir_context *ctx, struct dentry *dentry)
> >  			  inode->i_ino, fs_umode_to_dtype(inode->i_mode));
> >  }
> >  
> > -static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
> > +static void *offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
> >  {
> >  	struct offset_ctx *so_ctx = inode->i_op->get_offset_ctx(inode);
> >  	XA_STATE(xas, &so_ctx->xa, ctx->pos);
> > @@ -437,7 +437,8 @@ static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
> >  	while (true) {
> >  		dentry = offset_find_next(&xas);
> >  		if (!dentry)
> > -			break;
> > +			/* readdir has reached the current EOD */
> > +			return (void *)0x10;
> >  
> >  		if (!offset_dir_emit(ctx, dentry)) {
> >  			dput(dentry);
> > @@ -447,6 +448,7 @@ static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx)
> >  		dput(dentry);
> >  		ctx->pos = xas.xa_index + 1;
> >  	}
> > +	return NULL;
> >  }
> >  
> >  /**
> > @@ -479,7 +481,12 @@ static int offset_readdir(struct file *file, struct dir_context *ctx)
> >  	if (!dir_emit_dots(file, ctx))
> >  		return 0;
> >  
> > -	offset_iterate_dir(d_inode(dir), ctx);
> > +	if (ctx->pos == 2)
> > +		file->private_data = NULL;
> > +	else if (file->private_data == (void *)0x10)
> > +		return 0;
> > +
> > +	file->private_data = offset_iterate_dir(d_inode(dir), ctx);
> 
> I think it's usually best practice to only modify the file->private_data
> pointer during f_op->open and f_op->close but not override
> file->private_data once the file is visible to other threads. I think
> here it might not matter because access to file->private_data is
> serialized on f_pos_lock and it's not used by anything else.

I freely admit that using file->private_data this way is ugly. I was
hoping to find one bit somewhere that I could use to mark the file
descriptor, and file->private_data seemed the most handy.

We could go back to allocating a phony dentry (d_alloc_cursor) and
place the end-of-directory flag in there.

-- 
Chuck Lever