mbox series

[0/2] virtiofsd: Add capability to block xattrs

Message ID 20210826211937.317558-1-vgoyal@redhat.com (mailing list archive)
Headers show
Series virtiofsd: Add capability to block xattrs | expand

Message

Vivek Goyal Aug. 26, 2021, 9:19 p.m. UTC
As of now we have a knob "-o xattr/no_xattr" which either enables
all xattrs or disables all xattrs.

We need something more fine grained where we can selectively disable
only certain xattrs (and not all).

For example, in some cases we want to disable "security.selinux"
xattr. This is equivalent to virtiofs not supporting security.selinux
and guest kernel will fallback to a single label for whole fs
(virtiofs_t).

So add an option "-o block_xattr=<list-of-xattrs>" which will allow
specifying a list of xattrs to block.

Vivek Goyal (2):
  virtiofsd: Add an array to keep track of blocked xattrs
  virtiofsd: Add option "block_xattr=" to block certain xattrs

 docs/tools/virtiofsd.rst         |  17 ++++
 tools/virtiofsd/helper.c         |   3 +
 tools/virtiofsd/passthrough_ll.c | 166 ++++++++++++++++++++++++++++---
 3 files changed, 171 insertions(+), 15 deletions(-)

Comments

Dr. David Alan Gilbert Sept. 22, 2021, 11 a.m. UTC | #1
* Vivek Goyal (vgoyal@redhat.com) wrote:
> As of now we have a knob "-o xattr/no_xattr" which either enables
> all xattrs or disables all xattrs.

Hi Vivek,
  Thanks for this.

> We need something more fine grained where we can selectively disable
> only certain xattrs (and not all).
> 
> For example, in some cases we want to disable "security.selinux"
> xattr. This is equivalent to virtiofs not supporting security.selinux
> and guest kernel will fallback to a single label for whole fs
> (virtiofs_t).
> 
> So add an option "-o block_xattr=<list-of-xattrs>" which will allow
> specifying a list of xattrs to block.

This is quite interesting; I'd not noticed you had the exisitng blocking
mechanism, however, as discussed, I think my preference is if this could
be done as a modification of the xattrmap it would avoid another set of
options.

The mapping code already has 'type's of:

  prefix, ok, bad

I think you just need to add a 'reject' type
that produces the error code you need.

Dave

> Vivek Goyal (2):
>   virtiofsd: Add an array to keep track of blocked xattrs
>   virtiofsd: Add option "block_xattr=" to block certain xattrs
> 
>  docs/tools/virtiofsd.rst         |  17 ++++
>  tools/virtiofsd/helper.c         |   3 +
>  tools/virtiofsd/passthrough_ll.c | 166 ++++++++++++++++++++++++++++---
>  3 files changed, 171 insertions(+), 15 deletions(-)
> 
> -- 
> 2.31.1
>
Vivek Goyal Sept. 22, 2021, 12:30 p.m. UTC | #2
On Wed, Sep 22, 2021 at 12:00:17PM +0100, Dr. David Alan Gilbert wrote:
> * Vivek Goyal (vgoyal@redhat.com) wrote:
> > As of now we have a knob "-o xattr/no_xattr" which either enables
> > all xattrs or disables all xattrs.
> 
> Hi Vivek,
>   Thanks for this.
> 
> > We need something more fine grained where we can selectively disable
> > only certain xattrs (and not all).
> > 
> > For example, in some cases we want to disable "security.selinux"
> > xattr. This is equivalent to virtiofs not supporting security.selinux
> > and guest kernel will fallback to a single label for whole fs
> > (virtiofs_t).
> > 
> > So add an option "-o block_xattr=<list-of-xattrs>" which will allow
> > specifying a list of xattrs to block.
> 
> This is quite interesting; I'd not noticed you had the exisitng blocking
> mechanism,

Yes, that's for blocking posix acl xattrs if needed. If xattr map
support blocking, then we could probably insert an internal rule
to block posix acl xattrs. But that's more of a cleanup exercise
I will take up some other time.

> however, as discussed, I think my preference is if this could
> be done as a modification of the xattrmap it would avoid another set of
> options.
> 
> The mapping code already has 'type's of:
> 
>   prefix, ok, bad
> 
> I think you just need to add a 'reject' type
> that produces the error code you need.

How about "unsupported" and then return -EOPNOTSUPP? 

I am looking at selinux kernel code and it expect -EOPNOTSUPP to decide
that selinux xattr is not supported and looks into fallback options.

static int sb_check_xattr_support(struct super_block *sb)
{
        rc = __vfs_getxattr(root, root_inode, XATTR_NAME_SELINUX, NULL, 0);
        if (rc < 0 && rc != -ENODATA) {
                if (rc == -EOPNOTSUPP) {
                        pr_warn("SELinux: (dev %s, type %s) has no security xattr handler\n",
                                sb->s_id, sb->s_type->name);
                        goto fallback;
...
...
}

Vivek