Re: [RFC PATCH v2 2/3] fs: define a firmware security filesystem named fwsecurityfs

[Christian Brauner <brauner@kernel.org>]

On Mon, Jun 27, 2022 at 09:37:28AM +0200, Greg Kroah-Hartman wrote:
> On Sun, Jun 26, 2022 at 11:48:06AM -0400, Mimi Zohar wrote:
> > On Thu, 2022-06-23 at 09:23 -0400, James Bottomley wrote:
> > > On Thu, 2022-06-23 at 10:54 +0200, Greg Kroah-Hartman wrote:
> > > [...]
> > > > > diff --git a/fs/fwsecurityfs/inode.c b/fs/fwsecurityfs/inode.c
> > > > > new file mode 100644
> > > > > index 000000000000..5d06dc0de059
> > > > > --- /dev/null
> > > > > +++ b/fs/fwsecurityfs/inode.c
> > > > > @@ -0,0 +1,159 @@
> > > > > +// SPDX-License-Identifier: GPL-2.0-only
> > > > > +/*
> > > > > + * Copyright (C) 2022 IBM Corporation
> > > > > + * Author: Nayna Jain <nayna@linux.ibm.com>
> > > > > + */
> > > > > +
> > > > > +#include <linux/sysfs.h>
> > > > > +#include <linux/kobject.h>
> > > > > +#include <linux/fs.h>
> > > > > +#include <linux/fs_context.h>
> > > > > +#include <linux/mount.h>
> > > > > +#include <linux/pagemap.h>
> > > > > +#include <linux/init.h>
> > > > > +#include <linux/namei.h>
> > > > > +#include <linux/security.h>
> > > > > +#include <linux/lsm_hooks.h>
> > > > > +#include <linux/magic.h>
> > > > > +#include <linux/ctype.h>
> > > > > +#include <linux/fwsecurityfs.h>
> > > > > +
> > > > > +#include "internal.h"
> > > > > +
> > > > > +int fwsecurityfs_remove_file(struct dentry *dentry)
> > > > > +{
> > > > > +	drop_nlink(d_inode(dentry));
> > > > > +	dput(dentry);
> > > > > +	return 0;
> > > > > +};
> > > > > +EXPORT_SYMBOL_GPL(fwsecurityfs_remove_file);
> > > > > +
> > > > > +int fwsecurityfs_create_file(const char *name, umode_t mode,
> > > > > +					u16 filesize, struct dentry
> > > > > *parent,
> > > > > +					struct dentry *dentry,
> > > > > +					const struct file_operations
> > > > > *fops)
> > > > > +{
> > > > > +	struct inode *inode;
> > > > > +	int error;
> > > > > +	struct inode *dir;
> > > > > +
> > > > > +	if (!parent)
> > > > > +		return -EINVAL;
> > > > > +
> > > > > +	dir = d_inode(parent);
> > > > > +	pr_debug("securityfs: creating file '%s'\n", name);
> > > > 
> > > > Did you forget to call simple_pin_fs() here or anywhere else?
> > > > 
> > > > And this can be just one function with the directory creation file,
> > > > just check the mode and you will be fine.  Look at securityfs as an
> > > > example of how to make this simpler.
> > > 
> > > Actually, before you go down this route can you consider the namespace
> > > ramifications.  In fact we're just having to rework securityfs to pull
> > > out all the simple_pin_... calls because simple_pin_... is completely
> > > inimical to namespaces.

I described this at length in the securityfs namespacing thread at
various points. simple_pin_*() should be avoided if possible. Ideally
the filesystem will just be cleaned up on umount. There might be a
reason to make it survive umounts if you have state that stays around
and somehow is intimately tied to that filesystem.

> > > 
> > > The first thing to consider is if you simply use securityfs you'll
> > > inherit all the simple_pin_... removal work and be namespace ready.  It
> > > could be that creating a new filesystem that can't be namespaced is the
> > > right thing to do here, but at least ask the question: would we ever
> > > want any of these files to be presented selectively inside containers? 
> > > If the answer is "yes" then simple_pin_... is the wrong interface.
> > 
> > Greg, the securityfs changes James is referring to are part of the IMA
> > namespacing patch set:
> > https://lore.kernel.org/linux-integrity/20220420140633.753772-1-stefanb@linux.ibm.com/
> > 
> > I'd really appreciate your reviewing the first two patches:
> > [PATCH v12 01/26] securityfs: rework dentry creation
> > [PATCH v12 02/26] securityfs: Extend securityfs with namespacing
> > support
> 
> Looks like others have already reviewed them, they seem sane to me if
> they past testing.

Thanks for taking a look.