Skip to contentSkip to conversation

SELinux: always check SIGCHLD in selinux_task_wait

PATCH In review Eric Paris 2 participants 3 messages Oct 22, 2007 SELinux security module

lore
Share & tools
lore

Eric Paris

author
#0 lore raw

PATCH 1/1 SELinux: always check SIGCHLD in selinux_task_wait

When checking if we can wait on a child we were looking at p->exit_signal and trying to make the decision based on if the signal would eventually be allowed. One big flaw is that p->exit_signal is -1 for NPTL threads and so aignal_to_av was not actually checking SIGCHLD which is what would have been sent. Even is exit_signal was set to something strange it wouldn't change the fact that the child was there and needed to be waited on. This patch just assumes wait is based on SIGCHLD. Specific permission checks are made when the child actually attempts to send a signal.

This resolves the problem of things like using GDB on confined domains such as in RH BZ 232371. The confined domain did not have permission to send a generic signal (exit_signal == -1) back to the unconfined GDB. With this patch the GDB wait works and since the actual signal sent is allowed everything functions as it should.

---

security/selinux/hooks.c | 2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)
security/selinux/hooks.c +1 -1
index 24e1b18..078fe82 100644
@@ -2979,7 +2979,7 @@ static int selinux_task_wait(struct task_struct *p)
 {
 	u32 perm;
 
-	perm = signal_to_av(p->exit_signal);
+	perm = signal_to_av(SIGCHLD);
 
 	return task_has_perm(p, current, perm);
 }

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.

#1 lore raw
Quoted 28 lines · Eric Paris

When checking if we can wait on a child we were looking at p->exit_signal and trying to make the decision based on if the signal would eventually be allowed. One big flaw is that p->exit_signal is -1 for NPTL threads and so aignal_to_av was not actually checking SIGCHLD which is what would have been sent. Even is exit_signal was set to something strange it wouldn't change the fact that the child was there and needed to be waited on. This patch just assumes wait is based on SIGCHLD. Specific permission checks are made when the child actually attempts to send a signal.

This resolves the problem of things like using GDB on confined domains such as in RH BZ 232371. The confined domain did not have permission to send a generic signal (exit_signal == -1) back to the unconfined GDB. With this patch the GDB wait works and since the actual signal sent is allowed everything functions as it should.

---

security/selinux/hooks.c | 2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)
security/selinux/hooks.c +0 -0
index 24e1b18..078fe82 100644

security/selinux/hooks.c static int selinux_task_wait(struct task_struct *p) · line 2985

Quoted 8 lines
security/selinux/hooks.c +1 -1
@@ -2979,7 +2979,7 @@ static int selinux_task_wait(struct task_struct *p)
 {
 	u32 perm;
 
-	perm = signal_to_av(p->exit_signal);
+	perm = signal_to_av(SIGCHLD);
 
 	return task_has_perm(p, current, perm);
 }

Just make it:

	return task_has_perm(p, current, PROCESS__SIGCHLD);
Signature · 8 lines
Stephen Smalley
National Security Agency


--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to [email protected] with
the words "unsubscribe selinux" without quotes as the message.
author
#2 lore raw
Quoted 32 lines · Stephen Smalley
Quoted 31 lines · Eric Paris

When checking if we can wait on a child we were looking at p->exit_signal and trying to make the decision based on if the signal would eventually be allowed. One big flaw is that p->exit_signal is -1 for NPTL threads and so aignal_to_av was not actually checking SIGCHLD which is what would have been sent. Even is exit_signal was set to something strange it wouldn't change the fact that the child was there and needed to be waited on. This patch just assumes wait is based on SIGCHLD. Specific permission checks are made when the child actually attempts to send a signal.

This resolves the problem of things like using GDB on confined domains such as in RH BZ 232371. The confined domain did not have permission to send a generic signal (exit_signal == -1) back to the unconfined GDB. With this patch the GDB wait works and since the actual signal sent is allowed everything functions as it should.

---

security/selinux/hooks.c | 2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)
security/selinux/hooks.c +0 -0
index 24e1b18..078fe82 100644
@@ -2979,7 +2979,7 @@ static int selinux_task_wait(struct task_struct *p)
 {
 	u32 perm;
 
Quoted 9 lines

security/selinux/hooks.c static int selinux_task_wait(struct task_struct *p) · line 2985

Quoted 5 lines
security/selinux/hooks.c +1 -1
-	perm = signal_to_av(p->exit_signal);
+	perm = signal_to_av(SIGCHLD);
 
 	return task_has_perm(p, current, perm);
 }

Just make it:

	return task_has_perm(p, current, PROCESS__SIGCHLD);

ok, will resend as soon as it compiles and boots. I was sticking with it because we don't use PROCESS__SIG* anywhere outside of signal_to_av and we use constant input to that function in selinux_file_send_sigiotask (although for very different reasons). Is the compiler smart enough to compile out the inlined switch statement when given a constant? Doesn't matter, new patch coming.

-Eric

--
This message was distributed to subscribers of the selinux mailing list. If you no longer wish to subscribe, send mail to [email protected] with the words "unsubscribe selinux" without quotes as the message.