linux-kernel - Re: [PATCH 04/56] fs: Macros to define splice file

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Thu, 13 Nov 2014 14:24:22 -0800
From:	josh@...htriplett.org
To:	Richard Weinberger <richard.weinberger@...il.com>,
	Al Viro <viro@...IV.linux.org.uk>
Cc:	Pieter Smith <pieter@...sman.nl>,
	Jeff Layton <jlayton@...chiereds.net>,
	"J. Bruce Fields" <bfields@...ldses.org>,
	linux-fsdevel <linux-fsdevel@...r.kernel.org>,
	open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 04/56] fs: Macros to define splice file_operations

On Thu, Nov 13, 2014 at 10:49:07PM +0100, Richard Weinberger wrote:
> On Thu, Nov 13, 2014 at 10:22 PM, Pieter Smith <pieter@...sman.nl> wrote:
> > Provides a CONFIG_SYSCALL_SPLICE compatible way of defining the .splice_read
> > and .splice_write file_operations so that they can later be compiled out when
> > the kernel is configured without the splice-family syscalls
[...]
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -1512,6 +1512,32 @@ struct file_operations {
> >         int (*show_fdinfo)(struct seq_file *m, struct file *f);
> >  };
> >
> > +#ifdef CONFIG_SYSCALL_SPLICE
> > +
> > +/*
> > + * Define and init the splice_read member of a file_operations struct
> > + */
> > +#define SPLICE_READ_INIT(read) .splice_read = read,
> > +
> > +/*
> > + * Define and init the splice_read member of a file_operations struct
> > + */
> > +#define SPLICE_WRITE_INIT(write) .splice_write = write,
> 
> This is ugly like hell.
> Why can't you do something like __exit_p()?

On Thu, Nov 13, 2014 at 09:51:39PM +0000, Al Viro wrote:
> This (and subsequent stuff making use of that) is bloody pointless.  You
> save 2 words per file_operations instance, at the cost of making things
> uglier and harder to grep.  NAK.

Given the large number of uses of these, I agree that it doesn't seem
worth the tradeoff, particularly since very few file_operations
structures will exist on any individual tiny configuration.  I think we
should go with a wrapper similar to __exit_p (splice_p?), which just
becomes NULL when !CONFIG_SYSCALL_SPLICE.  Removing the actual pointers
from file_operations can wait until we have compiler support for tagging
specific fields in a structure (like splice_read and splice_write) as
dead.

Similarly, you shouldn't wrap the functions that get assigned to those
pointers with #ifdef; instead, mark them as __maybe_unused, which
doesn't even add any lines of code.  The compiler will then
automatically throw them out when not used, without emiting a warning.

That should drastically reduce the number of changes, and in particular
eliminate almost all of the ifdefs.

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