On Sat, 2023-08-05 at 14:43 +0200, Christian Brauner wrote:
> On Fri, Aug 04, 2023 at 12:09:34PM -0400, Jeff Layton wrote:
> > From: David Howells <dhowells@xxxxxxxxxx>
> >
> > When NFS superblocks are created by automounting, their LSM parameters
> > aren't set in the fs_context struct prior to sget_fc() being called,
> > leading to failure to match existing superblocks.
> >
> > This bug leads to messages like the following appearing in dmesg when
> > fscache is enabled:
> >
> > NFS: Cache volume key already in use (nfs,4.2,2,108,106a8c0,1,,,,100000,100000,2ee,3a98,1d4c,3a98,1)
> >
> > Fix this by adding a new LSM hook to load fc->security for submount
> > creation when alloc_fs_context() is creating the fs_context for it.
> >
> > Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
> > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > Fixes: 9bc61ab18b1d ("vfs: Introduce fs_context, switch vfs_kern_mount() to it.")
> > Fixes: 779df6a5480f ("NFS: Ensure security label is set for root inode)
> > Tested-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > Acked-by: Casey Schaufler <casey@xxxxxxxxxxxxxxxx>
> > Acked-by: "Christian Brauner (Microsoft)" <brauner@xxxxxxxxxx>
> > Link: https://lore.kernel.org/r/165962680944.3334508.6610023900349142034.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v1
> > Link: https://lore.kernel.org/r/165962729225.3357250.14350728846471527137.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v2
> > Link: https://lore.kernel.org/r/165970659095.2812394.6868894171102318796.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v3
> > Link: https://lore.kernel.org/r/166133579016.3678898.6283195019480567275.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v4
> > Link: https://lore.kernel.org/r/217595.1662033775@xxxxxxxxxxxxxxxxxxxxxx/ # v5
> > ---
> > ver #7)
> > - Drop lsm_set boolean
> > - Link to v6: https://lore.kernel.org/r/20230802-master-v6-1-45d48299168b@xxxxxxxxxx
> >
> > ver #6)
> > - Rebase onto v6.5.0-rc4
> >
> > ver #5)
> > - Removed unused variable.
> > - Only allocate smack_mnt_opts if we're dealing with a submount.
> >
> > ver #4)
> > - When doing a FOR_SUBMOUNT mount, don't set the root label in SELinux or
> > Smack.
> >
> > ver #3)
> > - Made LSM parameter extraction dependent on fc->purpose ==
> > FS_CONTEXT_FOR_SUBMOUNT. Shouldn't happen on FOR_RECONFIGURE.
> >
> > ver #2)
> > - Added Smack support
> > - Made LSM parameter extraction dependent on reference != NULL.
> > ---
> > fs/fs_context.c | 4 ++++
> > include/linux/lsm_hook_defs.h | 1 +
> > include/linux/security.h | 6 +++++
> > security/security.c | 14 +++++++++++
> > security/selinux/hooks.c | 25 ++++++++++++++++++++
> > security/smack/smack_lsm.c | 54 +++++++++++++++++++++++++++++++++++++++++++
> > 6 files changed, 104 insertions(+)
> >
> > diff --git a/fs/fs_context.c b/fs/fs_context.c
> > index 851214d1d013..a523aea956c4 100644
> > --- a/fs/fs_context.c
> > +++ b/fs/fs_context.c
> > @@ -282,6 +282,10 @@ static struct fs_context *alloc_fs_context(struct file_system_type *fs_type,
> > break;
> > }
> >
> > + ret = security_fs_context_init(fc, reference);
> > + if (ret < 0)
> > + goto err_fc;
> > +
> > /* TODO: Make all filesystems support this unconditionally */
> > init_fs_context = fc->fs_type->init_fs_context;
> > if (!init_fs_context)
> > diff --git a/include/linux/lsm_hook_defs.h b/include/linux/lsm_hook_defs.h
> > index 7308a1a7599b..7ce3550154b1 100644
> > --- a/include/linux/lsm_hook_defs.h
> > +++ b/include/linux/lsm_hook_defs.h
> > @@ -54,6 +54,7 @@ LSM_HOOK(int, 0, bprm_creds_from_file, struct linux_binprm *bprm, struct file *f
> > LSM_HOOK(int, 0, bprm_check_security, struct linux_binprm *bprm)
> > LSM_HOOK(void, LSM_RET_VOID, bprm_committing_creds, struct linux_binprm *bprm)
> > LSM_HOOK(void, LSM_RET_VOID, bprm_committed_creds, struct linux_binprm *bprm)
> > +LSM_HOOK(int, 0, fs_context_init, struct fs_context *fc, struct dentry *reference)
> > LSM_HOOK(int, 0, fs_context_dup, struct fs_context *fc,
> > struct fs_context *src_sc)
> > LSM_HOOK(int, -ENOPARAM, fs_context_parse_param, struct fs_context *fc,
> > diff --git a/include/linux/security.h b/include/linux/security.h
> > index 32828502f09e..61fda06fac9d 100644
> > --- a/include/linux/security.h
> > +++ b/include/linux/security.h
> > @@ -293,6 +293,7 @@ int security_bprm_creds_from_file(struct linux_binprm *bprm, struct file *file);
> > int security_bprm_check(struct linux_binprm *bprm);
> > void security_bprm_committing_creds(struct linux_binprm *bprm);
> > void security_bprm_committed_creds(struct linux_binprm *bprm);
> > +int security_fs_context_init(struct fs_context *fc, struct dentry *reference);
> > int security_fs_context_dup(struct fs_context *fc, struct fs_context *src_fc);
> > int security_fs_context_parse_param(struct fs_context *fc, struct fs_parameter *param);
> > int security_sb_alloc(struct super_block *sb);
> > @@ -629,6 +630,11 @@ static inline void security_bprm_committed_creds(struct linux_binprm *bprm)
> > {
> > }
> >
> > +static inline int security_fs_context_init(struct fs_context *fc,
> > + struct dentry *reference)
>
> I think that's the wrong way of doing this hook. The security hook
> really doesn't belong into alloc_fs_context().
>
> I think what we want is a dedicated helper similar to vfs_dup_context():
>
> // Only pass the superblock. There's no need for the dentry. I would
> // avoid even passing fs_context but if that's preferred then sure.
> security_fs_context_submount(struct fs_context *fc, const struct super_block *sb)
>
> vfs_submount_fs_context(struct file_system_type *fs_type, struct dentry *reference)
> {
> fc = fs_context_for_submount(fs_type, reference);
>
> security_fs_context_for_submount(fc, reference->d_sb);
> }
>
> This automatically ensures it's only called for submounts, the LSM
> doesn't need to care about fc->purpose and this isn't called
> in a pure allocation function for all allocation calls.
>
> The we should switch all callers over to that new helper and unexport
> that fs_context_for_submount() thing completely. Yes, that's more work
> but that's the correct thing to do. And we need to audit fuse, cifs,
> afs, and nfs anyway that they work fine with the new security hook.*
>
It's the same prototype. We could just move the hook call to the end of
fs_context_for_submount, and that would be less churn for its callers.
Or were you wanting to do that to make this a more gradual changeover
for some reason?
I will rework the security hook to take a sb pointer instead though.
>
> [1]: If really needed, then any additional fs specific work that needs
> to be done during submount allocation should probably probably be
> done in a new callback.
>
> struct fs_context_operations {
> void (*free)(struct fs_context *fc);
> int (*dup)(struct fs_context *fc, struct fs_context *src_fc);
> + int (*submount)(struct fs_context *fc, const struct super_block *sb);
--
Jeff Layton <jlayton@xxxxxxxxxx>
