Re: [PATCH][RFC] splice support

[Linus Torvalds]

On Wed, 29 Mar 2006, Jeff Garzik wrote:
> 
> 1) What are the consequences of doing
> 
> 	if (f_op->splice_write)
> 		f_op->splice_write(...);
> 	else
> 		generic_file_splice_write(...);
> 
> to cause sys_splice() to default to supported?

I'd actually much prefer a number of filesystems just adding he 
"generic_file_splice_write()" thing. If it works for them (and it usually 
will), it's a one-liner. And it won't do wrong things on filesystems that 
have special rules (inode re-validate for networked filesystems etc).

> 2) Do you really have to test f_op itself for NULL?  Is that a stealth
> closed-file check or something?  I would be surprised if f_op was ever really
> NULL.

Hmm.. I agree that f_op probably should never be NULL (a struct file with 
a NULL f_op is pretty useless), but it is a test that we historically have 
had. So it's probably best to keep for consistency, and if somebody wants 
to, they can clean up all the other tests too (in the read/write/lseek 
paths).

I'm inclined to apply this patch (well, I'd like the fixed one). The whole 
splice() thing has been rolling around in my head for years, and the pipe 
support infrastructure for it has been around for over a year now in 
preparation for this.

And the patch actually looks pretty clean to me.

			Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/