Skip to contentSkip to conversation

xfs: don't generate selinux audit messages for capability testing

PATCH RESEND In review Darrick J. Wong 4 participants 7 messages Mar 1, 2022 – Mar 4, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 xfs: don't generate selinux audit messages for capability testing

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsmap.c  |    4 ++--
 fs/xfs/xfs_ioctl.c  |    2 +-
 fs/xfs/xfs_iops.c   |    2 +-
 kernel/capability.c |    1 +
 4 files changed, 5 insertions(+), 4 deletions(-)
fs/xfs/xfs_fsmap.c +2 -2
index 48287caad28b..10e1cb71439e 100644
@@ -864,8 +864,8 @@ xfs_getfsmap(
 	    !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
 		return -EINVAL;
 
-	use_rmap = capable(CAP_SYS_ADMIN) &&
-		   xfs_has_rmapbt(mp);
+	use_rmap = xfs_has_rmapbt(mp) &&
+		   has_capability_noaudit(current, CAP_SYS_ADMIN);
 	head->fmh_entries = 0;
 
 	/* Set up our device handlers. */
fs/xfs/xfs_ioctl.c +1 -1
index 2515fe8299e1..83481005317a 100644
@@ -1189,7 +1189,7 @@ xfs_ioctl_setattr_get_trans(
 		goto out_error;
 
 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_error;
 
fs/xfs/xfs_iops.c +1 -1
index b79b3846e71b..a65217f787cf 100644
@@ -723,7 +723,7 @@ xfs_setattr_nonsize(
 	}
 
 	error = xfs_trans_alloc_ichange(ip, udqp, gdqp, NULL,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_dqrele;
 
kernel/capability.c +1 -0
index 46a361dde042..765194f5d678 100644
@@ -360,6 +360,7 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
 {
 	return has_ns_capability_noaudit(t, &init_user_ns, cap);
 }
+EXPORT_SYMBOL(has_capability_noaudit);
 
 static bool ns_capable_common(struct user_namespace *ns,
 			      int cap,
maintainer
#1 lore raw
Quoted 26 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsmap.c  |    4 ++--
 fs/xfs/xfs_ioctl.c  |    2 +-
 fs/xfs/xfs_iops.c   |    2 +-
 kernel/capability.c |    1 +
 4 files changed, 5 insertions(+), 4 deletions(-)
fs/xfs/xfs_fsmap.c +0 -0
index 48287caad28b..10e1cb71439e 100644

fs/xfs/xfs_fsmap.c xfs_getfsmap( · line 867

Quoted 6 lines
fs/xfs/xfs_fsmap.c +1 -2
@@ -864,8 +864,8 @@ xfs_getfsmap(
 	    !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
 		return -EINVAL;
 
-	use_rmap = capable(CAP_SYS_ADMIN) &&
-		   xfs_has_rmapbt(mp);
+	use_rmap = xfs_has_rmapbt(mp) &&

Hm, I'm failing to find where xfs_has_rmapbt() is defined. I just wanted to make sure it doesn't have any side effects that you'd want to avoid in the no-capability case. (Seems very unlikely that it would, given the name)

Quoted 42 lines
fs/xfs/xfs_fsmap.c +1 -0
+		   has_capability_noaudit(current, CAP_SYS_ADMIN);
 	head->fmh_entries = 0;
 
 	/* Set up our device handlers. */
fs/xfs/xfs_ioctl.c +1 -1
index 2515fe8299e1..83481005317a 100644
@@ -1189,7 +1189,7 @@ xfs_ioctl_setattr_get_trans(
 		goto out_error;
 
 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_error;
 
fs/xfs/xfs_iops.c +1 -1
index b79b3846e71b..a65217f787cf 100644
@@ -723,7 +723,7 @@ xfs_setattr_nonsize(
 	}
 
 	error = xfs_trans_alloc_ichange(ip, udqp, gdqp, NULL,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_dqrele;
 
kernel/capability.c +1 -0
index 46a361dde042..765194f5d678 100644
@@ -360,6 +360,7 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
 {
 	return has_ns_capability_noaudit(t, &init_user_ns, cap);
 }
+EXPORT_SYMBOL(has_capability_noaudit);
 
 static bool ns_capable_common(struct user_namespace *ns,
 			      int cap,
#2 lore raw
Quoted 30 lines · Serge E. Hallyn
Quoted 29 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsmap.c  |    4 ++--
 fs/xfs/xfs_ioctl.c  |    2 +-
 fs/xfs/xfs_iops.c   |    2 +-
 kernel/capability.c |    1 +
 4 files changed, 5 insertions(+), 4 deletions(-)
fs/xfs/xfs_fsmap.c +0 -0
index 48287caad28b..10e1cb71439e 100644
@@ -864,8 +864,8 @@ xfs_getfsmap(
 	    !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
 		return -EINVAL;
 
Quoted 8 lines

fs/xfs/xfs_fsmap.c xfs_getfsmap( · line 867

Quoted 3 lines
fs/xfs/xfs_fsmap.c +1 -2
-	use_rmap = capable(CAP_SYS_ADMIN) &&
-		   xfs_has_rmapbt(mp);
+	use_rmap = xfs_has_rmapbt(mp) &&

Hm, I'm failing to find where xfs_has_rmapbt() is defined. I just wanted to make sure it doesn't have any side effects that you'd want to avoid in the no-capability case. (Seems very unlikely that it would, given the name)

fs/xfs/xfs_mount.h:495:__XFS_HAS_FEAT(rmapbt, RMAPBT)

To expand on that a little -- it's a convenience predicate that tells us whether or not the mounted xfs filesystem supports the reverse mapping btree feature. The predicate itself has no side effects, of course, so the rearranging of the two sides of the && operator so that we do the cheaper check first (like this code probably should have done from the start).

--D

Quoted 43 lines
Quoted 43 lines
fs/xfs/xfs_fsmap.c +1 -0
+		   has_capability_noaudit(current, CAP_SYS_ADMIN);
 	head->fmh_entries = 0;
 
 	/* Set up our device handlers. */
fs/xfs/xfs_ioctl.c +1 -1
index 2515fe8299e1..83481005317a 100644
@@ -1189,7 +1189,7 @@ xfs_ioctl_setattr_get_trans(
 		goto out_error;
 
 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_error;
 
fs/xfs/xfs_iops.c +1 -1
index b79b3846e71b..a65217f787cf 100644
@@ -723,7 +723,7 @@ xfs_setattr_nonsize(
 	}
 
 	error = xfs_trans_alloc_ichange(ip, udqp, gdqp, NULL,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_dqrele;
 
kernel/capability.c +1 -0
index 46a361dde042..765194f5d678 100644
@@ -360,6 +360,7 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
 {
 	return has_ns_capability_noaudit(t, &init_user_ns, cap);
 }
+EXPORT_SYMBOL(has_capability_noaudit);
 
 static bool ns_capable_common(struct user_namespace *ns,
 			      int cap,
maintainer Acked-by
#3 lore raw
Quoted 42 lines · Darrick J. Wong
Quoted 30 lines · Serge E. Hallyn
Quoted 29 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsmap.c  |    4 ++--
 fs/xfs/xfs_ioctl.c  |    2 +-
 fs/xfs/xfs_iops.c   |    2 +-
 kernel/capability.c |    1 +
 4 files changed, 5 insertions(+), 4 deletions(-)
fs/xfs/xfs_fsmap.c +0 -0
index 48287caad28b..10e1cb71439e 100644
@@ -864,8 +864,8 @@ xfs_getfsmap(
 	    !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
 		return -EINVAL;
 
Quoted 8 lines

fs/xfs/xfs_fsmap.c xfs_getfsmap( · line 867

Quoted 3 lines
fs/xfs/xfs_fsmap.c +1 -2
-	use_rmap = capable(CAP_SYS_ADMIN) &&
-		   xfs_has_rmapbt(mp);
+	use_rmap = xfs_has_rmapbt(mp) &&

Hm, I'm failing to find where xfs_has_rmapbt() is defined. I just wanted to make sure it doesn't have any side effects that you'd want to avoid in the no-capability case. (Seems very unlikely that it would, given the name)

fs/xfs/xfs_mount.h:495:__XFS_HAS_FEAT(rmapbt, RMAPBT)

Quoted 8 lines

To expand on that a little -- it's a convenience predicate that tells us whether or not the mounted xfs filesystem supports the reverse mapping btree feature. The predicate itself has no side effects, of course, so the rearranging of the two sides of the && operator so that we do the cheaper check first (like this code probably should have done from the start).

--D

THank you - looks good then.

Quoted 43 lines
Quoted 43 lines
Quoted 43 lines
fs/xfs/xfs_fsmap.c +1 -0
+		   has_capability_noaudit(current, CAP_SYS_ADMIN);
 	head->fmh_entries = 0;
 
 	/* Set up our device handlers. */
fs/xfs/xfs_ioctl.c +1 -1
index 2515fe8299e1..83481005317a 100644
@@ -1189,7 +1189,7 @@ xfs_ioctl_setattr_get_trans(
 		goto out_error;
 
 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_error;
 
fs/xfs/xfs_iops.c +1 -1
index b79b3846e71b..a65217f787cf 100644
@@ -723,7 +723,7 @@ xfs_setattr_nonsize(
 	}
 
 	error = xfs_trans_alloc_ichange(ip, udqp, gdqp, NULL,
-			capable(CAP_FOWNER), &tp);
+			has_capability_noaudit(current, CAP_FOWNER), &tp);
 	if (error)
 		goto out_dqrele;
 
kernel/capability.c +1 -0
index 46a361dde042..765194f5d678 100644
@@ -360,6 +360,7 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
 {
 	return has_ns_capability_noaudit(t, &init_user_ns, cap);
 }
+EXPORT_SYMBOL(has_capability_noaudit);
 
 static bool ns_capable_common(struct user_namespace *ns,
 			      int cap,
Reviewed-by on v1 1/1
#4 lore raw
In reply to Darrick J. Wong
Quoted 68 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsmap.c  |    4 ++--
 fs/xfs/xfs_ioctl.c  |    2 +-
 fs/xfs/xfs_iops.c   |    2 +-
 kernel/capability.c |    1 +
 4 files changed, 5 insertions(+), 4 deletions(-)
fs/xfs/xfs_fsmap.c +2 -2
index 48287caad28b..10e1cb71439e 100644
@@ -864,8 +864,8 @@ xfs_getfsmap(
            !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
                return -EINVAL;
 
-       use_rmap = capable(CAP_SYS_ADMIN) &&
-                  xfs_has_rmapbt(mp);
+       use_rmap = xfs_has_rmapbt(mp) &&
+                  has_capability_noaudit(current, CAP_SYS_ADMIN);
        head->fmh_entries = 0;
 
        /* Set up our device handlers. */
fs/xfs/xfs_ioctl.c +1 -1
index 2515fe8299e1..83481005317a 100644
@@ -1189,7 +1189,7 @@ xfs_ioctl_setattr_get_trans(
                goto out_error;
 
        error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
-                       capable(CAP_FOWNER), &tp);
+                       has_capability_noaudit(current, CAP_FOWNER), &tp);
        if (error)
                goto out_error;
 
fs/xfs/xfs_iops.c +1 -1
index b79b3846e71b..a65217f787cf 100644
@@ -723,7 +723,7 @@ xfs_setattr_nonsize(
        }
 
        error = xfs_trans_alloc_ichange(ip, udqp, gdqp, NULL,
-                       capable(CAP_FOWNER), &tp);
+                       has_capability_noaudit(current, CAP_FOWNER), &tp);
        if (error)
                goto out_dqrele;
 
kernel/capability.c +0 -0
index 46a361dde042..765194f5d678 100644

kernel/capability.c bool has_capability_noaudit(struct task_struct *t, int cap) · line 366

Quoted 7 lines
kernel/capability.c +1 -0
@@ -360,6 +360,7 @@ bool has_capability_noaudit(struct task_struct *t, int cap)
 {
        return has_ns_capability_noaudit(t, &init_user_ns, cap);
 }
+EXPORT_SYMBOL(has_capability_noaudit);
 
 static bool ns_capable_common(struct user_namespace *ns,
                              int cap,

Thank you for respinning the patch[1]! And sorry that I couldn't find the time to do that :/

(Yes, we should still clean up and document the capability functions, but if no one has the energy, let's at least do the minimal fix.)

[1] Link for reference:
https://lore.kernel.org/linux-xfs/[email protected]/T/

--
Ondrej Mosnacek
Software Engineer, Linux Security - SELinux kernel Red Hat, Inc.

Reviewed-by on v1 1/1
#5 lore raw
In reply to Darrick J. Wong
Quoted 7 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

Quoted 7 lines

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>

Thanks Darrick. This looks technically correct to me as well.

You might want to add a:

Fixes 7317a03df703f ("xfs: refactor inode ownership change transaction/inode/quota allocation idiom")

because I /think/ that's the commit that moved the capable() checks out from under quota tests, and made the problem more visible.

And maybe:

Fixes ea9a46e1c4925 ("xfs: only return detailed fsmap info if the caller has CAP_SYS_ADMIN")

as well?

It's not strictly fixing the former; AFAICT the problem existed when quota was enabled already, so I'll leave all that to your discretion.

Thanks,
-Eric

#6 lore raw
Quoted 27 lines · Eric Sandeen
Quoted 7 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

There are a few places where we test the current process' capability set to decide if we're going to be more or less generous with resource acquisition for a system call. If the process doesn't have the capability, we can continue the call, albeit in a degraded mode.

Quoted 7 lines

These are /not/ the actual security decisions, so it's not proper to use capable(), which (in certain selinux setups) causes audit messages to get logged. Switch them to has_capability_noaudit.

Cc Ondrej Mosnacek <[email protected]>
Cc Dave Chinner <[email protected]>

Thanks Darrick. This looks technically correct to me as well.

You might want to add a:

Fixes 7317a03df703f ("xfs: refactor inode ownership change transaction/inode/quota allocation idiom")

because I /think/ that's the commit that moved the capable() checks out from under quota tests, and made the problem more visible.

And maybe:

Quoted 8 lines
Fixes ea9a46e1c4925 ("xfs: only return detailed fsmap info if the caller has CAP_SYS_ADMIN")

as well?

It's not strictly fixing the former; AFAICT the problem existed when quota was enabled already, so I'll leave all that to your discretion.

Thank you.

--D

Quoted 2 lines

Thanks,
-Eric