Re: [PATCH] fuse: Fix IOC_[GS]ETFLAGS argument size brokenness

[Darrick J. Wong]

On Fri, Dec 20, 2013 at 01:19:27AM -0500, Richard Hansen wrote:
> On 2013-12-19 18:27, Darrick J. Wong wrote:
> > The IOC_[GS]ETFLAGS ioctls, despite being defined to take a "long"
> > parameter, actually take "int" parameters.  FUSE unfortunately assumed
> > that the ioctl definitions never lie, and transfers a long's worth of
> > data in and out of userspace, which causes stack smashing in chattr,
> > and other bugs elsewhere.
> > 
> > So, special-case this in FUSE, and document this int/long quirk in
> > include/uapi/linux/fs.h.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > ---
> >  fs/fuse/file.c          |   11 +++++++++++
> >  include/uapi/linux/fs.h |    1 +
> >  2 files changed, 12 insertions(+)
> > 
> > diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> > index 7e70506..5fa8181 100644
> > --- a/fs/fuse/file.c
> > +++ b/fs/fuse/file.c
> > @@ -2385,6 +2385,17 @@ long fuse_do_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
> >  		iov->iov_base = (void __user *)arg;
> >  		iov->iov_len = _IOC_SIZE(cmd);
> >  
> > +		/*
> > +		 * The IOC_[GS]ETFLAGS ioctls take int parameters even though
> > +		 * the ioctl definition specifies long.  Userland has been
> > +		 * expecting int for ages (and chattr segfaults on FUSE
> > +		 * filesystems), so special case that here.  The IOC32
> > +		 * variants were declared with int, so they don't need this.
> > +		 */
> > +		if (cmd == FS_IOC_GETFLAGS || cmd == FS_IOC_SETFLAGS) {
> > +			iov->iov_len = sizeof(int);
> > +		}
> > +
> >  		if (_IOC_DIR(cmd) & _IOC_WRITE) {
> >  			in_iov = iov;
> >  			in_iovs = 1;
> > diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> > index 6c28b61..bc8aa8e 100644
> > --- a/include/uapi/linux/fs.h
> > +++ b/include/uapi/linux/fs.h
> > @@ -154,6 +154,7 @@ struct inodes_stat_t {
> >  #define FITHAW		_IOWR('X', 120, int)	/* Thaw */
> >  #define FITRIM		_IOWR('X', 121, struct fstrim_range)	/* Trim */
> >  
> > +/* IOC_[GS]ETFLAGS take an int argument despite being defined to take long. */
> >  #define	FS_IOC_GETFLAGS			_IOR('f', 1, long)
> >  #define	FS_IOC_SETFLAGS			_IOW('f', 2, long)
> >  #define	FS_IOC_GETVERSION		_IOR('v', 1, long)
> 
> This change to include/uapi/linux/fs.h is similar to but not as thorough as:
> 
> http://mid.gmane.org/1385548839-16617-1-git-send-email-aurelien@xxxxxxxxxxx
> https://lkml.org/lkml/2013/11/27/93
> 
> (The above linked patch is the patch Aurelien referenced in the
> "Argument type for FS_IOC_GETFLAGS/FS_IOC_SETFLAGS ioctls" thread on
> linux-fsdevel; see
> <http://mid.gmane.org/20131127231430.GG29912@xxxxxxxxxxxxxxxx>.)

Aha!  I thought I'd seen something like that on the list, though it doesn't
seem to be in -rc.  Thank you for pointing that out.

And yes, now that I look at all the [GS]ETVERSION implementations, I think FUSE
could special-case those two ioctls too.

--D
> 
> -Richard
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
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/