linux-kernel - Re: [patch 5/5] VFS: allow filesystem to override mknod capability checks

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

Message-Id: <E1IYkAN-0001mh-00@dorka.pomaz.szeredi.hu>
Date:	Fri, 21 Sep 2007 17:11:39 +0200
From:	Miklos Szeredi <miklos@...redi.hu>
To:	hch@...radead.org
CC:	miklos@...redi.hu, hch@...radead.org, akpm@...ux-foundation.org,
	linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org
Subject: Re: [patch 5/5] VFS: allow filesystem to override mknod capability checks

> On Fri, Sep 21, 2007 at 04:48:58PM +0200, Miklos Szeredi wrote:
> > Ah, OK.  Well, that's what fuse would do with the above change.  So
> > you are basically saying, the change is OK, but we want proper
> > unprivileged mounts first.
> 
> Yes, that and that it should be a mount flag, not a file_system_type
> flag.
> 
> > I'm sure we don't want it by default.
> > 
> > For example if user bind mounts / onto /home/user/myroot (with 'nodev'
> > of couse), we still don't want mknod to work on that mount, for
> > obvious reasons.
> 
> True, we'll have to deny it if there is any non-privilegued mount of
> the backing device possible.   At this point it's getting rather nasty
> and I wonder whether it's really worth it..

I think the assumption, that we want this as a generic service is
false.

We want this as a special service for a few filesystems, such as the
unprivileged userspace loopback mounting I was talking about.

So my thinking is: if an unprivileged filesystem explicitly asks for
this, then it should be allowed.  It could be a per-superblock flag
instead of a per fs-type flag, if that sounds better.

My fuse implementation would have been exactly the same: the ->mknod()
implementation would check a per filesystem flag, and if it's not set,
check the permissions as normal mknod() would.  Here's the relevant
patch snippet:

Index: linux/fs/fuse/dir.c
===================================================================
--- linux.orig/fs/fuse/dir.c	2007-09-21 13:45:23.000000000 +0200
+++ linux/fs/fuse/dir.c	2007-09-21 13:45:25.000000000 +0200
@@ -486,7 +486,13 @@ static int fuse_mknod(struct inode *dir,
 {
 	struct fuse_mknod_in inarg;
 	struct fuse_conn *fc = get_fuse_conn(dir);
-	struct fuse_req *req = fuse_get_req(fc);
+	struct fuse_req *req;
+
+	if (!fc->mknod_nocheck &&
+	    ((S_ISCHR(mode) || S_ISBLK(mode)) && !capable(CAP_MKNOD)))
+		return -EPERM;
+
+	req = fuse_get_req(fc);
 	if (IS_ERR(req))
 		return PTR_ERR(req);
 

-
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/