These ioctls are equivalent to fcntl(fd, F_SETFD, flags), which SELinux
always allows too. Furthermore, a failed FIOCLEX could result in a file
descriptor being leaked to a process that should not have access to it.
As this patch removes access controls, a policy capability needs to be
enabled in policy to always allow these ioctls.
Based-on-patch-by: Demi Marie Obenour <[email protected]>
Signed-off-by: Richard Haines <[email protected]>
---
V2 Change: Control via a policy capability. See this thread for discussion:
https://lore.kernel.org/selinux/CAHC9VhQEPxYP_KU56gAGNHKQaxucY8gSsHiUB42PVgADBAccRQ@mail.gmail.com/T/#t
With this patch and the polcap enabled, the selinux-testsuite will fail:
ioctl/test at line 47 - Will need a fix.
security/selinux/hooks.c | 7 +++++++
security/selinux/include/policycap.h | 1 +
security/selinux/include/policycap_names.h | 3 ++-
security/selinux/include/security.h | 7 +++++++
4 files changed, 17 insertions(+), 1 deletion(-)
diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
index 5b6895e4f..030c41652 100644
--- a/security/selinux/hooks.c
+++ b/security/selinux/hooks.c
@@ -3745,6 +3745,13 @@ static int selinux_file_ioctl(struct file *file, unsigned int cmd,
CAP_OPT_NONE, true);
break;
+ case FIOCLEX:
+ case FIONCLEX:
+ /* Must always succeed if polcap set, else default: */
+ if (selinux_policycap_ioctl_skip_cloexec())
+ break;
+ fallthrough;
+
/* default case assumes that the command will go
* to the file's ioctl() function.
*/
diff --git a/security/selinux/include/policycap.h b/security/selinux/include/policycap.h
index 2ec038efb..44d73dc32 100644
--- a/security/selinux/include/policycap.h
+++ b/security/selinux/include/policycap.h
@@ -11,6 +11,7 @@ enum {
POLICYDB_CAPABILITY_CGROUPSECLABEL,
POLICYDB_CAPABILITY_NNP_NOSUID_TRANSITION,
POLICYDB_CAPABILITY_GENFS_SECLABEL_SYMLINKS,
+ POLICYDB_CAPABILITY_IOCTL_CLOEXEC,
__POLICYDB_CAPABILITY_MAX
};
#define POLICYDB_CAPABILITY_MAX (__POLICYDB_CAPABILITY_MAX - 1)
diff --git a/security/selinux/include/policycap_names.h b/security/selinux/include/policycap_names.h
index b89289f09..ebd64afe1 100644
--- a/security/selinux/include/policycap_names.h
+++ b/security/selinux/include/policycap_names.h
@@ -12,7 +12,8 @@ const char *selinux_policycap_names[__POLICYDB_CAPABILITY_MAX] = {
"always_check_network",
"cgroup_seclabel",
"nnp_nosuid_transition",
- "genfs_seclabel_symlinks"
+ "genfs_seclabel_symlinks",
+ "ioctl_skip_cloexec"
};
#endif /* _SELINUX_POLICYCAP_NAMES_H_ */
diff --git a/security/selinux/include/security.h b/security/selinux/include/security.h
index ac0ece013..8a789c22b 100644
--- a/security/selinux/include/security.h
+++ b/security/selinux/include/security.h
@@ -219,6 +219,13 @@ static inline bool selinux_policycap_genfs_seclabel_symlinks(void)
return READ_ONCE(state->policycap[POLICYDB_CAPABILITY_GENFS_SECLABEL_SYMLINKS]);
}
+static inline bool selinux_policycap_ioctl_skip_cloexec(void)
+{
+ struct selinux_state *state = &selinux_state;
+
+ return READ_ONCE(state->policycap[POLICYDB_CAPABILITY_IOCTL_CLOEXEC]);
+}
+
struct selinux_policy_convert_data;
struct selinux_load_state {
--
2.35.1
On Tue, 2022-02-22 at 18:28 -0500, Paul Moore wrote:
> On Mon, Feb 21, 2022 at 8:15 AM Richard Haines
> <[email protected]> wrote:
> >
> > These ioctls are equivalent to fcntl(fd, F_SETFD, flags), which
> > SELinux
> > always allows too. Furthermore, a failed FIOCLEX could result in a
> > file
> > descriptor being leaked to a process that should not have access to
> > it.
> >
> > As this patch removes access controls, a policy capability needs to
> > be
> > enabled in policy to always allow these ioctls.
> >
> > Based-on-patch-by: Demi Marie Obenour <[email protected]>
> > Signed-off-by: Richard Haines <[email protected]>
> > ---
> > V2 Change: Control via a policy capability. See this thread for
> > discussion:
> > https://lore.kernel.org/selinux/CAHC9VhQEPxYP_KU56gAGNHKQaxucY8gSsHiUB42PVgADBAccRQ@mail.gmail.com/T/#t
> >
> > With this patch and the polcap enabled, the selinux-testsuite will
> > fail:
> > ioctl/test at line 47 - Will need a fix.
> >
> > security/selinux/hooks.c | 7 +++++++
> > security/selinux/include/policycap.h | 1 +
> > security/selinux/include/policycap_names.h | 3 ++-
> > security/selinux/include/security.h | 7 +++++++
> > 4 files changed, 17 insertions(+), 1 deletion(-)
>
> Thanks Richard for putting together the v2 of this patch.
>
> As far as the test is concerned, it seems like the quick-n-dirty fix
> is to simply remove the ioctl(FIOCLEX) test in test_noioctl.c; is
> everyone okay with that? At least that is what I'm going to do with
> my local copy that I use to validate the kernel-secnext builds unless
> someone has a better patch :)
To fix this I was planning to submit a patch that would change the
ioctl(FIOCLEX) tests to ioctl(FS_IOC_GETFSLABEL) as that would continue
to test the xperms.
>
> > diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> > index 5b6895e4f..030c41652 100644
> > --- a/security/selinux/hooks.c
> > +++ b/security/selinux/hooks.c
> > @@ -3745,6 +3745,13 @@ static int selinux_file_ioctl(struct file
> > *file, unsigned int cmd,
> > CAP_OPT_NONE, true);
> > break;
> >
> > + case FIOCLEX:
> > + case FIONCLEX:
> > + /* Must always succeed if polcap set, else default:
> > */
> > + if (selinux_policycap_ioctl_skip_cloexec())
> > + break;
> > + fallthrough;
> > +
>
> The break/fallthrough looks like it might be a little more fragile
> than necessary, how about something like this:
>
> case FIOCLEX:
> case FIONCLEX:
> if (!selinux_policycap_ioctl_skip_cloexec())
> error = ioctl_has_perm(cred, file, FILE__IOCTL, (u16) cmd);
> break;
>
> Yes, it does duplicate the default ioctl_has_perm() call, but since
> we
> are effectively deprecating this and locking the FIOCLEX/FIONCLEX
> behavior with this policy capability it seems okay to me (and
> preferable to relying on the fallthrough).
>
> Thoughts?
Yes I did ponder this and in my first attempt I had this before the
switch():
/* Must always succeed if polcap set */
if (selinux_policycap_ioctl_skip_cloexec() &&
(cmd == FIOCLEX || cmd == FIONCLEX))
return 0;
switch (cmd) {
case FIONREAD:
case FIBMAP:
but changed to within the switch(), anyway I'm happy to resubmit a
patch either way.
>
On Wed, Feb 23, 2022 at 4:36 PM Paul Moore <[email protected]> wrote:
> On Wed, Feb 23, 2022 at 7:43 AM Richard Haines
> <[email protected]> wrote:
> > On Wed, 2022-02-23 at 13:12 +0100, Ondrej Mosnacek wrote:
> > > On Wed, Feb 23, 2022 at 12:58 PM Richard Haines
> > > <[email protected]> wrote:
> > > > On Tue, 2022-02-22 at 18:28 -0500, Paul Moore wrote:
> > > > > On Mon, Feb 21, 2022 at 8:15 AM Richard Haines
> > > > > <[email protected]> wrote:
> > > > > >
> > > > > > These ioctls are equivalent to fcntl(fd, F_SETFD, flags), which
> > > > > > SELinux
> > > > > > always allows too. Furthermore, a failed FIOCLEX could result
> > > > > > in a
> > > > > > file
> > > > > > descriptor being leaked to a process that should not have
> > > > > > access to
> > > > > > it.
> > > > > >
> > > > > > As this patch removes access controls, a policy capability
> > > > > > needs to
> > > > > > be
> > > > > > enabled in policy to always allow these ioctls.
> > > > > >
> > > > > > Based-on-patch-by: Demi Marie Obenour <[email protected]>
> > > > > > Signed-off-by: Richard Haines <[email protected]>
> > > > > > ---
> > > > > > V2 Change: Control via a policy capability. See this thread for
> > > > > > discussion:
> > > > > > https://lore.kernel.org/selinux/CAHC9VhQEPxYP_KU56gAGNHKQaxucY8gSsHiUB42PVgADBAccRQ@mail.gmail.com/T/#t
> > > > > >
> > > > > > With this patch and the polcap enabled, the selinux-testsuite
> > > > > > will
> > > > > > fail:
> > > > > > ioctl/test at line 47 - Will need a fix.
> > > > > >
> > > > > > security/selinux/hooks.c | 7 +++++++
> > > > > > security/selinux/include/policycap.h | 1 +
> > > > > > security/selinux/include/policycap_names.h | 3 ++-
> > > > > > security/selinux/include/security.h | 7 +++++++
> > > > > > 4 files changed, 17 insertions(+), 1 deletion(-)
> > > > >
> > > > > Thanks Richard for putting together the v2 of this patch.
> > > > >
> > > > > As far as the test is concerned, it seems like the quick-n-dirty
> > > > > fix
> > > > > is to simply remove the ioctl(FIOCLEX) test in test_noioctl.c; is
> > > > > everyone okay with that? At least that is what I'm going to do
> > > > > with
> > > > > my local copy that I use to validate the kernel-secnext builds
> > > > > unless
> > > > > someone has a better patch :)
> > > >
> > > > To fix this I was planning to submit a patch that would change the
> > > > ioctl(FIOCLEX) tests to ioctl(FS_IOC_GETFSLABEL) as that would
> > > > continue
> > > > to test the xperms.
> > >
> > > That one seems to be implemented only by some filesystems. Is there
> > > any more generic one we could use?
> >
> > What about FS_IOC_GETFLAGS
>
> Unless I'm mistaken, FIGETBSZ should be largely fs independent.
Bah, nevermind, FIGETBSZ ends up in a FILE__GETATTR check.
FS_IOC_GETFLAGS has the same problem.
How about FIOQSIZE?
--
paul-moore.com
On Wed, Feb 23, 2022 at 7:43 AM Richard Haines
<[email protected]> wrote:
> On Wed, 2022-02-23 at 13:12 +0100, Ondrej Mosnacek wrote:
> > On Wed, Feb 23, 2022 at 12:58 PM Richard Haines
> > <[email protected]> wrote:
> > > On Tue, 2022-02-22 at 18:28 -0500, Paul Moore wrote:
> > > > On Mon, Feb 21, 2022 at 8:15 AM Richard Haines
> > > > <[email protected]> wrote:
> > > > >
> > > > > These ioctls are equivalent to fcntl(fd, F_SETFD, flags), which
> > > > > SELinux
> > > > > always allows too. Furthermore, a failed FIOCLEX could result
> > > > > in a
> > > > > file
> > > > > descriptor being leaked to a process that should not have
> > > > > access to
> > > > > it.
> > > > >
> > > > > As this patch removes access controls, a policy capability
> > > > > needs to
> > > > > be
> > > > > enabled in policy to always allow these ioctls.
> > > > >
> > > > > Based-on-patch-by: Demi Marie Obenour <[email protected]>
> > > > > Signed-off-by: Richard Haines <[email protected]>
> > > > > ---
> > > > > V2 Change: Control via a policy capability. See this thread for
> > > > > discussion:
> > > > > https://lore.kernel.org/selinux/CAHC9VhQEPxYP_KU56gAGNHKQaxucY8gSsHiUB42PVgADBAccRQ@mail.gmail.com/T/#t
> > > > >
> > > > > With this patch and the polcap enabled, the selinux-testsuite
> > > > > will
> > > > > fail:
> > > > > ioctl/test at line 47 - Will need a fix.
> > > > >
> > > > > security/selinux/hooks.c | 7 +++++++
> > > > > security/selinux/include/policycap.h | 1 +
> > > > > security/selinux/include/policycap_names.h | 3 ++-
> > > > > security/selinux/include/security.h | 7 +++++++
> > > > > 4 files changed, 17 insertions(+), 1 deletion(-)
> > > >
> > > > Thanks Richard for putting together the v2 of this patch.
> > > >
> > > > As far as the test is concerned, it seems like the quick-n-dirty
> > > > fix
> > > > is to simply remove the ioctl(FIOCLEX) test in test_noioctl.c; is
> > > > everyone okay with that? At least that is what I'm going to do
> > > > with
> > > > my local copy that I use to validate the kernel-secnext builds
> > > > unless
> > > > someone has a better patch :)
> > >
> > > To fix this I was planning to submit a patch that would change the
> > > ioctl(FIOCLEX) tests to ioctl(FS_IOC_GETFSLABEL) as that would
> > > continue
> > > to test the xperms.
> >
> > That one seems to be implemented only by some filesystems. Is there
> > any more generic one we could use?
>
> What about FS_IOC_GETFLAGS
Unless I'm mistaken, FIGETBSZ should be largely fs independent.
> > > > > diff --git a/security/selinux/hooks.c
> > > > > b/security/selinux/hooks.c
> > > > > index 5b6895e4f..030c41652 100644
> > > > > --- a/security/selinux/hooks.c
> > > > > +++ b/security/selinux/hooks.c
> > > > > @@ -3745,6 +3745,13 @@ static int selinux_file_ioctl(struct
> > > > > file
> > > > > *file, unsigned int cmd,
> > > > > CAP_OPT_NONE,
> > > > > true);
> > > > > break;
> > > > >
> > > > > + case FIOCLEX:
> > > > > + case FIONCLEX:
> > > > > + /* Must always succeed if polcap set, else
> > > > > default:
> > > > > */
> > > > > + if (selinux_policycap_ioctl_skip_cloexec())
> > > > > + break;
> > > > > + fallthrough;
> > > > > +
> > > >
> > > > The break/fallthrough looks like it might be a little more
> > > > fragile
> > > > than necessary, how about something like this:
> > > >
> > > > case FIOCLEX:
> > > > case FIONCLEX:
> > > > if (!selinux_policycap_ioctl_skip_cloexec())
> > > > error = ioctl_has_perm(cred, file, FILE__IOCTL, (u16) cmd);
> > > > break;
> > > >
> > > > Yes, it does duplicate the default ioctl_has_perm() call, but
> > > > since
> > > > we
> > > > are effectively deprecating this and locking the FIOCLEX/FIONCLEX
> > > > behavior with this policy capability it seems okay to me (and
> > > > preferable to relying on the fallthrough).
> > > >
> > > > Thoughts?
> > >
> > > Yes I did ponder this and in my first attempt I had this before the
> > > switch():
> > >
> > > /* Must always succeed if polcap set */
> > > if (selinux_policycap_ioctl_skip_cloexec() &&
> > > (cmd == FIOCLEX || cmd == FIONCLEX))
> > > return 0;
> > >
> > > switch (cmd) {
> > > case FIONREAD:
> > > case FIBMAP:
> > >
> > > but changed to within the switch(), anyway I'm happy to resubmit a
> > > patch either way.
> >
> > I agree with Paul's suggestion. Better to duplicate the simple call
> > than to complicate the code flow.
>
> Okay will use Paul's.
Thanks guys.
--
paul-moore.com