Re: eCryptfs: Request for review

[Anton Altaparmakov <aia21@cam.ac.uk>]

On Wed, 26 Oct 2005, Michael Thompson wrote:
> Hello again,
> 
> In preperation of our submission to the LKML, I have begun running
> sparse on the code, and have discovered that due to the nature of our
> stacked filesystem, we have encountered the following sparse message:
> 
> fs/ecryptfs/inode.c:843:60: warning: incorrect type in argument 2
> (different address spaces)
> fs/ecryptfs/inode.c:843:60:    expected char [noderef] *<noident><asn:1>
> fs/ecryptfs/inode.c:843:60:    got char *[assigned] lower_buf
> 
> This is in the function ecryptfs_readlink, which follows the inode
> operations definition of:
> int (*readlink) (struct dentry *, char __user *,int);
> 
> The reason we are getting these messages is that we are calling the
> readlink operation of the lower filesystem, but passing it a buffer
> which we allocate. Spare is picking up on the fact that the lower fs's
> readlink expects a buffer of type (char __user*), but we provide
> (char*).
> 
> Is it acceptable to do the following:
> char *buffer;
> ...
> lower_fs_inode->i_op->readlink(lower_dentry, (char __user*)lower_buf, bufsiz);

Yes, that is absolutely fine to do.

But I do hope you are doing the necessary get_fs/set_fs magic if you are 
passing kernel pointers instead of user space ones...  I haven't seen your 
code so apologies if I am stating the obvious.  If you OTOH don't know 
what I am talking about you better learn it fast, e.g. look at 
drivers/block/loop.c::__do_lo_send_write() which does it.

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://linux-ntfs.sf.net/ & http://www-stu.christs.cam.ac.uk/~aia21/