lsm,audit,selinux: Introduce a new audit data type LSM_AUDIT_DATA_FILE

Hi,

Right now LSM_AUDIT_DATA_PATH type contains "struct path" in union "u" of common_audit_data. This information is used to print path of file at the same time it is also used to get to dentry and inode. And this inode information is used to get to superblock and device and print device information.

This does not work well for layered filesystems like overlay where dentry contained in path is overlay dentry and not the real dentry of underlying file system. That means inode retrieved from dentry is also overlay inode and not the real inode.

seliux helpers like file_path_has_perm() are doing checks on inode retrieved from file_inode(). This returns the real inode and not the overlay inode. That means we are doing check on real inode but for audit purposes we are printing details of overlay inode and that can be confusing while debugging.

Hence, introduce a new type LSM_AUDIT_DATA_FILE which carries file information and inode retrieved is real inode using file_inode(). That way right avc denied information is given to user.

For example, following is one example avc before the patch.

type=AVC msg=audit(1473360868.399:214): avc: denied { read open } for pid=1765 comm="cat" path="/root/git/selinux-testsuite/tests/overlay/container1/merged/readfile" dev="overlay" ino=21443 scontext=unconfined_u:unconfined_r:test_overlay_client_t:s0:c10,c20 tcontext=unconfined_u:object_r:test_overlay_files_ro_t:s0 tclass=file permissive=0

It looks as follows after the patch.

type=AVC msg=audit(1473360017.388:282): avc: denied { read open } for pid=2530 comm="cat" path="/root/git/selinux-testsuite/tests/overlay/container1/merged/readfile" dev="dm-0" ino=2377915 scontext=unconfined_u:unconfined_r:test_overlay_client_t:s0:c10,c20 tcontext=unconfined_u:object_r:test_overlay_files_ro_t:s0 tclass=file permissive=0

Notice that now dev information points to "dm-0" device instead of "overlay" device. This makes it clear that check failed on underlying inode and not on the overlay inode.

Signed-off-by Vivek Goyal <[email protected]>

---
 include/linux/lsm_audit.h |    2 ++
 security/lsm_audit.c      |   13 +++++++++++++
 security/selinux/hooks.c  |   16 ++++++++--------
 3 files changed, 23 insertions(+), 8 deletions(-)

Index: pcmoore-linux/include/linux/lsm_audit.h
===================================================================

include/linux/lsm_audit.h 2016-09-08 14:56:14.066159741 -0400 renamed from include/linux/lsm_audit.h 2016-09-08 14:56:10.173159741 -0400 +2 -0

@@ -59,6 +59,7 @@ struct common_audit_data {
 #define LSM_AUDIT_DATA_INODE   9
 #define LSM_AUDIT_DATA_DENTRY  10
 #define LSM_AUDIT_DATA_IOCTL_OP        11
+#define LSM_AUDIT_DATA_FILE    12
        union   {
                struct path path;
                struct dentry *dentry;

@@ -75,6 +76,7 @@ struct common_audit_data {
 #endif
                char *kmod_name;
                struct lsm_ioctlop_audit *op;
+               struct file *file;
        } u;
        /* this union contains LSM specific data */
        union {

Index: pcmoore-linux/security/lsm_audit.c
===================================================================

security/lsm_audit.c 2016-09-08 14:56:14.067159741 -0400 renamed from security/lsm_audit.c 2016-09-08 14:56:10.173159741 -0400 +13 -0

@@ -245,6 +245,19 @@ static void dump_common_audit_data(struc
                }
                break;
        }
+       case LSM_AUDIT_DATA_FILE: {
+               struct inode *inode;
+
+               audit_log_d_path(ab, " path=", &a->u.file->f_path);
+
+               inode = file_inode(a->u.file);
+               if (inode) {
+                       audit_log_format(ab, " dev=");
+                       audit_log_untrustedstring(ab, inode->i_sb->s_id);
+                       audit_log_format(ab, " ino=%lu", inode->i_ino);
+               }
+               break;
+       }
        case LSM_AUDIT_DATA_IOCTL_OP: {
                struct inode *inode;
 

Index: pcmoore-linux/security/selinux/hooks.c
===================================================================

security/selinux/hooks.c 2016-09-09 09:15:37.001159741 -0400 renamed from security/selinux/hooks.c 2016-09-08 14:56:10.173159741 -0400 +8 -8

@@ -1761,8 +1761,8 @@ static inline int file_path_has_perm(con
 {
        struct common_audit_data ad;
 
-       ad.type = LSM_AUDIT_DATA_PATH;
-       ad.u.path = file->f_path;
+       ad.type = LSM_AUDIT_DATA_FILE;
+       ad.u.file = file;
        return inode_has_perm(cred, file_inode(file), av, &ad);
 }
 

@@ -1784,8 +1784,8 @@ static int file_has_perm(const struct cr
        u32 sid = cred_sid(cred);
        int rc;
 
-       ad.type = LSM_AUDIT_DATA_PATH;
-       ad.u.path = file->f_path;
+       ad.type = LSM_AUDIT_DATA_FILE;
+       ad.u.file = file;
 
        if (sid != fsec->sid) {
                rc = avc_has_perm(sid, fsec->sid,

@@ -2365,8 +2365,8 @@ static int selinux_bprm_set_creds(struct
                        new_tsec->sid = old_tsec->sid;
        }
 
-       ad.type = LSM_AUDIT_DATA_PATH;
-       ad.u.path = bprm->file->f_path;
+       ad.type = LSM_AUDIT_DATA_FILE;
+       ad.u.file = bprm->file;
 
        if (new_tsec->sid == old_tsec->sid) {
                rc = avc_has_perm(old_tsec->sid, isec->sid,

@@ -3833,8 +3833,8 @@ static int selinux_kernel_module_from_fi
 
        /* finit_module */
 
-       ad.type = LSM_AUDIT_DATA_PATH;
-       ad.u.path = file->f_path;
+       ad.type = LSM_AUDIT_DATA_FILE;
+       ad.u.file = file;
 
        fsec = file->f_security;
        if (sid != fsec->sid) {

Quoted 35 lines · Vivek Goyal

Hi,

Right now LSM_AUDIT_DATA_PATH type contains "struct path" in union "u" of common_audit_data. This information is used to print path of file at the same time it is also used to get to dentry and inode. And this inode information is used to get to superblock and device and print device information.

This does not work well for layered filesystems like overlay where dentry contained in path is overlay dentry and not the real dentry of underlying file system. That means inode retrieved from dentry is also overlay inode and not the real inode.

seliux helpers like file_path_has_perm() are doing checks on inode retrieved from file_inode(). This returns the real inode and not the overlay inode. That means we are doing check on real inode but for audit purposes we are printing details of overlay inode and that can be confusing while debugging.

Hence, introduce a new type LSM_AUDIT_DATA_FILE which carries file information and inode retrieved is real inode using file_inode(). That way right avc denied information is given to user.

For example, following is one example avc before the patch.

type=AVC msg=audit(1473360868.399:214): avc: denied { read open } for pid=1765 comm="cat" path="/root/git/selinux-testsuite/tests/overlay/container1/merged/readfile" dev="overlay" ino=21443 scontext=unconfined_u:unconfined_r:test_overlay_client_t:s0:c10,c20 tcontext=unconfined_u:object_r:test_overlay_files_ro_t:s0 tclass=file permissive=0

It looks as follows after the patch.

type=AVC msg=audit(1473360017.388:282): avc: denied { read open } for pid=2530 comm="cat" path="/root/git/selinux-testsuite/tests/overlay/container1/merged/readfile" dev="dm-0" ino=2377915 scontext=unconfined_u:unconfined_r:test_overlay_client_t:s0:c10,c20 tcontext=unconfined_u:object_r:test_overlay_files_ro_t:s0 tclass=file permissive=0

Notice that now dev information points to "dm-0" device instead of "overlay" device. This makes it clear that check failed on underlying inode and not on the overlay inode.

Quoted 6 lines

Signed-off-by Vivek Goyal <[email protected]>

---
 include/linux/lsm_audit.h |    2 ++
 security/lsm_audit.c      |   13 +++++++++++++
 security/selinux/hooks.c  |   16 ++++++++--------
 3 files changed, 23 insertions(+), 8 deletions(-)

Hi Vivek,

Sorry for the delay, this fell through the my normal filters as it didn't go to the SELinux and/or audit mailing list. However, this patch looks reasonable to me and I think it is something we want in 4.9 with the rest of the LSM/overlayfs bits. I'm building a test kernel right now and assuming all goes well, I'll push this up to James since it looks like we've got another week before the merge window opens.