Skip to contentSkip to conversation

selinux: put the mmap() DAC controls before the MAC controls

PATCH In review Paul Moore 4 participants 14 messages Feb 27, 2014 – Feb 28, 2014 SELinux security module

lore
Share & tools
lore

Paul Moore

author
#0 lore raw

PATCH 1/1 selinux: put the mmap() DAC controls before the MAC controls

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

Example:

 # cat mmap_test.c
  #include <stdlib.h>
  #include <stdio.h>
  #include <errno.h>
  #include <sys/mman.h>
  int main(int argc, char *argv[])
  {
        int rc;
        void *mem;
        mem = mmap(0x0, 4096,
                   PROT_READ | PROT_WRITE,
                   MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
        if (mem == MAP_FAILED)
                return errno;
        printf("mem = %p\n", mem);
        munmap(mem, 4096);

        return 0;
  }
 # gcc -g -O0 -o mmap_test mmap_test.c
 # ./mmap_test
 mem = (nil)
 # ausearch -m AVC | grep mmap_zero
 type=AVC msg=audit(...): avc:  denied  { mmap_zero }
   for pid=1025 comm="mmap_test"
   scontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tclass=memprotect

This patch corrects things so that when the above example is run by a user without CAP_SYS_RAWIO the SELinux AVC is no longer generated as the DAC capability check fails before the SELinux permission check.

---
 security/selinux/hooks.c |   20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)
security/selinux/hooks.c +8 -12
index 57b0b49..e3664ae 100644
@@ -3205,24 +3205,20 @@ error:
 
 static int selinux_mmap_addr(unsigned long addr)
 {
-	int rc = 0;
-	u32 sid = current_sid();
+	int rc;
+
+	/* do DAC check on address space usage */
+	rc = cap_mmap_addr(addr);
+	if (rc)
+		return rc;
 
-	/*
-	 * notice that we are intentionally putting the SELinux check before
-	 * the secondary cap_file_mmap check.  This is such a likely attempt
-	 * at bad behaviour/exploit that we always want to get the AVC, even
-	 * if DAC would have also denied the operation.
-	 */
 	if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
+		u32 sid = current_sid();
 		rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
 				  MEMPROTECT__MMAP_ZERO, NULL);
-		if (rc)
-			return rc;
 	}
 
-	/* do DAC check on address space usage */
-	return cap_mmap_addr(addr);
+	return rc;
 }
 
 static int selinux_mmap_file(struct file *file, unsigned long reqprot,
#1 lore raw
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Quoted 87 lines

Example:

 # cat mmap_test.c
  #include <stdlib.h>
  #include <stdio.h>
  #include <errno.h>
  #include <sys/mman.h>
  int main(int argc, char *argv[])
  {
        int rc;
        void *mem;
        mem = mmap(0x0, 4096,
                   PROT_READ | PROT_WRITE,
                   MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
        if (mem == MAP_FAILED)
                return errno;
        printf("mem = %p\n", mem);
        munmap(mem, 4096);

        return 0;
  }
 # gcc -g -O0 -o mmap_test mmap_test.c
 # ./mmap_test
 mem = (nil)
 # ausearch -m AVC | grep mmap_zero
 type=AVC msg=audit(...): avc:  denied  { mmap_zero }
   for pid=1025 comm="mmap_test"
   scontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tclass=memprotect

This patch corrects things so that when the above example is run by a user without CAP_SYS_RAWIO the SELinux AVC is no longer generated as the DAC capability check fails before the SELinux permission check.

---
 security/selinux/hooks.c |   20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)
security/selinux/hooks.c +8 -12
index 57b0b49..e3664ae 100644
@@ -3205,24 +3205,20 @@ error:
 
 static int selinux_mmap_addr(unsigned long addr)
 {
-	int rc = 0;
-	u32 sid = current_sid();
+	int rc;
+
+	/* do DAC check on address space usage */
+	rc = cap_mmap_addr(addr);
+	if (rc)
+		return rc;
 
-	/*
-	 * notice that we are intentionally putting the SELinux check before
-	 * the secondary cap_file_mmap check.  This is such a likely attempt
-	 * at bad behaviour/exploit that we always want to get the AVC, even
-	 * if DAC would have also denied the operation.
-	 */
 	if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
+		u32 sid = current_sid();
 		rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
 				  MEMPROTECT__MMAP_ZERO, NULL);
-		if (rc)
-			return rc;
 	}
 
-	/* do DAC check on address space usage */
-	return cap_mmap_addr(addr);
+	return rc;
 }
 
 static int selinux_mmap_file(struct file *file, unsigned long reqprot,

_______________________________________________
Selinux mailing list
[email protected]
To unsubscribe, send email to [email protected].
To get help, send an email containing "help" to [email protected].

#2 lore raw
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

And if so, is there in fact any real value in retaining the SELinux check at all? Do we in fact ever allow sys_rawio and not mmap_zero (or if we do, is there a real reason for it)?

Quoted 95 lines
Quoted 82 lines

Example:

 # cat mmap_test.c
  #include <stdlib.h>
  #include <stdio.h>
  #include <errno.h>
  #include <sys/mman.h>
  int main(int argc, char *argv[])
  {
        int rc;
        void *mem;
        mem = mmap(0x0, 4096,
                   PROT_READ | PROT_WRITE,
                   MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
        if (mem == MAP_FAILED)
                return errno;
        printf("mem = %p\n", mem);
        munmap(mem, 4096);

        return 0;
  }
 # gcc -g -O0 -o mmap_test mmap_test.c
 # ./mmap_test
 mem = (nil)
 # ausearch -m AVC | grep mmap_zero
 type=AVC msg=audit(...): avc:  denied  { mmap_zero }
   for pid=1025 comm="mmap_test"
   scontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tclass=memprotect

This patch corrects things so that when the above example is run by a user without CAP_SYS_RAWIO the SELinux AVC is no longer generated as the DAC capability check fails before the SELinux permission check.

---
 security/selinux/hooks.c |   20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)
security/selinux/hooks.c +8 -12
index 57b0b49..e3664ae 100644
@@ -3205,24 +3205,20 @@ error:
 
 static int selinux_mmap_addr(unsigned long addr)
 {
-	int rc = 0;
-	u32 sid = current_sid();
+	int rc;
+
+	/* do DAC check on address space usage */
+	rc = cap_mmap_addr(addr);
+	if (rc)
+		return rc;
 
-	/*
-	 * notice that we are intentionally putting the SELinux check before
-	 * the secondary cap_file_mmap check.  This is such a likely attempt
-	 * at bad behaviour/exploit that we always want to get the AVC, even
-	 * if DAC would have also denied the operation.
-	 */
 	if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
+		u32 sid = current_sid();
 		rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
 				  MEMPROTECT__MMAP_ZERO, NULL);
-		if (rc)
-			return rc;
 	}
 
-	/* do DAC check on address space usage */
-	return cap_mmap_addr(addr);
+	return rc;
 }
 
 static int selinux_mmap_file(struct file *file, unsigned long reqprot,
Quoted 6 lines

_______________________________________________
Selinux mailing list
[email protected]
To unsubscribe, send email to [email protected].
To get help, send an email containing "help" to [email protected].

_______________________________________________
Selinux mailing list
[email protected]
To unsubscribe, send email to [email protected].
To get help, send an email containing "help" to [email protected].

author
#3 lore raw
In reply to Stephen Smalley
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Do you have a strong objection to this patch?

Signature · 2 lines
paul moore
security and virtualization @ redhat
#4 lore raw
Quoted 19 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Quoted 5 lines

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

#5 lore raw
Quoted 24 lines · Stephen Smalley
Quoted 23 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Quoted 7 lines
Quoted 1 line

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Wouldn't removing mmap_zero be contrary to SELinux having fine-grained access controls, since CAP_SYS_RAWIO is broader than this specific case?

Signature · 3 lines
Chris PeBenito
Tresys Technology, LLC
www.tresys.com | oss.tresys.com
#6 lore raw
Quoted 27 lines · Christopher J. PeBenito
Quoted 26 lines · Stephen Smalley
Quoted 24 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Do you have a strong objection to this patch?

Quoted 7 lines
Quoted 5 lines

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Wouldn't removing mmap_zero be contrary to SELinux having fine-grained access controls, since CAP_SYS_RAWIO is broader than this specific case?

Show me a real world scenario where I would want to allow sys_rawio but not mmap_zero and where doing so would provide real security benefit.

author
#7 lore raw
In reply to Stephen Smalley
Quoted 24 lines · Stephen Smalley
Quoted 23 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Quoted 7 lines
Quoted 1 line

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Seems like a reasonable argument to me. I pinged Eric to get his thoughts on the issue since he added the check originally; if he is okay with removing it, I'll go ahead do it.

Signature · 2 lines
paul moore
security and virtualization @ redhat
#8 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 33 lines · Paul Moore
Quoted 24 lines · Stephen Smalley
Quoted 23 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Quoted 7 lines
Quoted 1 line

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Quoted 4 lines

Seems like a reasonable argument to me. I pinged Eric to get his thoughts on the issue since he added the check originally; if he is okay with removing it, I'll go ahead do it.

The only thing is this is a nice debugging tool for the kernel. Finding apps that accidentally mmap_zero.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlMPk1kACgkQrlYvE4MpobPVwACfZQGC8tldE6F5PXKLeYgELrYT t28An21+GMg/0Jipe+8TLiKcHuBSYLM1
=TwuX
-----END PGP SIGNATURE-----

#9 lore raw
Quoted 34 lines · Daniel J Walsh
Quoted 32 lines · Paul Moore
Quoted 31 lines · Stephen Smalley
Quoted 19 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Quoted 5 lines

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Quoted 6 lines
Quoted 3 lines

Seems like a reasonable argument to me. I pinged Eric to get his thoughts on the issue since he added the check originally; if he is okay with removing it, I'll go ahead do it.

The only thing is this is a nice debugging tool for the kernel. Finding apps that accidentally mmap_zero.

You'll still see sys_rawio avc denials and the audit syscall record will show that it was mmap of a low address.

#10 lore raw
Quoted 39 lines · Stephen Smalley
Quoted 38 lines · Daniel J Walsh
Quoted 36 lines · Paul Moore
Quoted 24 lines · Stephen Smalley
Quoted 23 lines · Paul Moore
Quoted 9 lines · Stephen Smalley
Quoted 5 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Quoted 7 lines
Quoted 1 line

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Seems like a reasonable argument to me. I pinged Eric to get his thoughts on the issue since he added the check originally; if he is okay with removing it, I'll go ahead do it.

Quoted 5 lines
Quoted 2 lines

The only thing is this is a nice debugging tool for the kernel. Finding apps that accidentally mmap_zero.

You'll still see sys_rawio avc denials and the audit syscall record will show that it was mmap of a low address.

Looking at Fedora policy, there are differences in what domains are allowed mmap_zero vs sys_rawio, although I don't know how intentional/meaningful that is.

sesearch -A -p mmap_zero
vs
sesearch -A -p sys_rawio

Why for example does cupsd_t have sys_rawio? That's rather disturbing.

I guess you should keep the separate check until those differences are resolved.

#11 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 63 lines · Stephen Smalley
Quoted 43 lines · Stephen Smalley
Quoted 42 lines · Daniel J Walsh
Quoted 40 lines · Paul Moore
Quoted 27 lines · Stephen Smalley
Quoted 26 lines · Paul Moore
Quoted 10 lines · Stephen Smalley
Quoted 6 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

So you think that the explanation given in the comment for the current ordering is no longer valid?

Yes and no. Arguably there is still some value in it but there are enough problems with it as-is that I think the value is starting to be outweighed by the pain it is causing (Dan can be very annoying when he wants something <g>). For those users who still want notification of processes trying to mmap() low addresses, I think an audit watch is a much better approach. I don't think SELinux shouldn't be acting as an intrustion detection tool when we have other things that do that job.

Let's also not forget that the MAC-before-DAC approach goes against the general approach to applying SELinux controls, so there is some argument to be had for consistency as well.

Quoted 8 lines
Quoted 1 line

Do you have a strong objection to this patch?

No, although I do wonder if we ought to just dispense with mmap_zero altogether at this point. It made sense when there was no capability check or if the capability was one of the extremely broad ones (e.g. CAP_SYS_ADMIN), but I don't really see why we can't be just as restrictive with CAP_SYS_RAWIO / sys_rawio as with mmap_zero.

Seems like a reasonable argument to me. I pinged Eric to get his thoughts on the issue since he added the check originally; if he is okay with removing it, I'll go ahead do it.

Quoted 5 lines
Quoted 2 lines

The only thing is this is a nice debugging tool for the kernel. Finding apps that accidentally mmap_zero.

You'll still see sys_rawio avc denials and the audit syscall record will show that it was mmap of a low address.

Looking at Fedora policy, there are differences in what domains are allowed mmap_zero vs sys_rawio, although I don't know how intentional/meaningful that is.

sesearch -A -p mmap_zero vs sesearch -A -p sys_rawio

Why for example does cupsd_t have sys_rawio? That's rather disturbing.

I guess you should keep the separate check until those differences are resolved.

Quoted 6 lines

_______________________________________________ Selinux mailing list [email protected] To unsubscribe, send email to [email protected]. To get help, send an email containing "help" to [email protected].

Don't know why cups has it but I will remove it.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlMPpiwACgkQrlYvE4MpobOnrACfUE/quImyDenbAQ0b3xytW5FU 2tEAniONTu0sIUbuqxObgofqZb/J+JQx
=oLp9
-----END PGP SIGNATURE-----

author
#12 lore raw
In reply to Stephen Smalley
Quoted 4 lines · Stephen Smalley

Looking at Fedora policy, there are differences in what domains are allowed mmap_zero vs sys_rawio, although I don't know how intentional/meaningful that is.

Quoted 8 lines

sesearch -A -p mmap_zero
vs
sesearch -A -p sys_rawio

Why for example does cupsd_t have sys_rawio? That's rather disturbing.

I guess you should keep the separate check until those differences are resolved.

No more comments overnight so I'm going to go ahead and push this to next with your ack added. We can always remove the SELinux check at a later date.

Thanks.

Signature · 2 lines
paul moore
security and virtualization @ redhat
Acked-by on v1 1/1
#13 lore raw
In reply to Paul Moore
Quoted 39 lines · Paul Moore

It turns out that doing the SELinux MAC checks for mmap() before the DAC checks was causing users and the SELinux policy folks headaches as users were seeing a lot of SELinux AVC denials for the memprotect:mmap_zero permission that would have also been denied by the normal DAC capability checks (CAP_SYS_RAWIO).

Example:

 # cat mmap_test.c
  #include <stdlib.h>
  #include <stdio.h>
  #include <errno.h>
  #include <sys/mman.h>
  int main(int argc, char *argv[])
  {
        int rc;
        void *mem;
        mem = mmap(0x0, 4096,
                   PROT_READ | PROT_WRITE,
                   MAP_PRIVATE | MAP_ANONYMOUS | MAP_FIXED, -1, 0);
        if (mem == MAP_FAILED)
                return errno;
        printf("mem = %p\n", mem);
        munmap(mem, 4096);

        return 0;
  }
 # gcc -g -O0 -o mmap_test mmap_test.c
 # ./mmap_test
 mem = (nil)
 # ausearch -m AVC | grep mmap_zero
 type=AVC msg=audit(...): avc:  denied  { mmap_zero }
   for pid=1025 comm="mmap_test"
   scontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tcontext=unconfined_u:unconfined_r:unconfined_t:s0-s0:c0.c1023
   tclass=memprotect
Quoted 5 lines

This patch corrects things so that when the above example is run by a user without CAP_SYS_RAWIO the SELinux AVC is no longer generated as the DAC capability check fails before the SELinux permission check.

Quoted 49 lines
---
 security/selinux/hooks.c |   20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)
security/selinux/hooks.c +8 -12
index 57b0b49..e3664ae 100644
@@ -3205,24 +3205,20 @@ error:
 
 static int selinux_mmap_addr(unsigned long addr)
 {
-	int rc = 0;
-	u32 sid = current_sid();
+	int rc;
+
+	/* do DAC check on address space usage */
+	rc = cap_mmap_addr(addr);
+	if (rc)
+		return rc;
 
-	/*
-	 * notice that we are intentionally putting the SELinux check before
-	 * the secondary cap_file_mmap check.  This is such a likely attempt
-	 * at bad behaviour/exploit that we always want to get the AVC, even
-	 * if DAC would have also denied the operation.
-	 */
 	if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
+		u32 sid = current_sid();
 		rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
 				  MEMPROTECT__MMAP_ZERO, NULL);
-		if (rc)
-			return rc;
 	}
 
-	/* do DAC check on address space usage */
-	return cap_mmap_addr(addr);
+	return rc;
 }
 
 static int selinux_mmap_file(struct file *file, unsigned long reqprot,

_______________________________________________
Selinux mailing list
[email protected]
To unsubscribe, send email to [email protected].
To get help, send an email containing "help" to [email protected].