fs: Avoid premature clearing of file capabilities
Subsystems (36)
Jan Kara
reviewerv2 cover fs: Avoid premature clearing of file capabilities
Hello,
I'm resending the patches since Al did not respond to the series yet (I've previously posted it in [1] on Aug 3. Al? If Al does not respond, can you Andrew please feed to patches to Linus? Thanks.
This patch series is my attempt to fix an issue when user can clear capabilites of arbitrary file he can look up for example by running chown on it (this got assigned CVE-2015-1350). The problem is that we call security_inode_killpriv() before checking permissions in inode_change_ok(). This patch set moves that call into inode_change_ok() after permissions are checked - the only trouble is that we need to give dentry instead of inode there and that is not completely trivial in some cases - I'm still missing a review from Ceph people to verify I didn't miss anything.
Anyway, I guess the changes are fine for merging. Al, you please merge the patches? Thanks!
Changes since v1:
* Added acks for XFS and FUSE patches
Honza
Jan Kara
reviewerPATCH v2 1/5 xfs: Propagate dentry down to inode_change_ok()
To avoid clearing of capabilities or security related extended attributes too early, inode_change_ok() will need to take dentry instead of inode. Propagate dentry down to functions calling inode_change_ok(). This is rather straightforward except for xfs_set_mode() function which does not have dentry easily available. Luckily that function does not call inode_change_ok() anyway so we just have to do a little dance with function prototypes.
--- fs/xfs/xfs_file.c | 2 +- fs/xfs/xfs_inode.c | 2 +- fs/xfs/xfs_ioctl.c | 2 +- fs/xfs/xfs_iops.c | 94 ++++++++++++++++++++++++++++++++++++------------------ fs/xfs/xfs_iops.h | 3 +- 5 files changed, 68 insertions(+), 35 deletions(-)
fs/xfs/xfs_file.c +1 -1
@@ -984,7 +984,7 @@ xfs_file_fallocate( iattr.ia_valid =ATTR_SIZE ; iattr.ia_size = new_size; - error =xfs_setattr_size (ip, &iattr); + error =xfs_vn_setattr_size (file_dentry (file), &iattr);if (error) goto out_unlock; }
fs/xfs/xfs_inode.c +1 -1
@@ -1710,7 +1710,7 @@ xfs_inactive_truncate(/* * Log the inode size first to prevent stale data exposure in the event * of a system crash before the truncate completes. See the related -* comment in xfs_setattr_size() for details. +* comment in xfs_vn_setattr_size() for details. */ ip->i_d.di_size =0 ;xfs_trans_log_inode (tp, ip,XFS_ILOG_CORE );
fs/xfs/xfs_ioctl.c +1 -1
@@ -720,7 +720,7 @@ xfs_ioc_space( iattr.ia_valid =ATTR_SIZE ; iattr.ia_size = bf->l_start; - error =xfs_setattr_size (ip, &iattr); + error =xfs_vn_setattr_size (file_dentry (filp), &iattr);break ;default :ASSERT (0 );
fs/xfs/xfs_iops.c +63 -31
143 lines not on this page · raw message
fs/xfs/xfs_iops.h +2 -1
@@ -33,6 +33,7 @@ extern ssize_t xfs_vn_listxattr(struct dentry *, char *data, size_t size);extern void xfs_setattr_time (struct xfs_inode *ip,struct iattr *iattr);extern int xfs_setattr_nonsize (struct xfs_inode *ip,struct iattr *vap,int flags); -extern int xfs_setattr_size (struct xfs_inode *ip,struct iattr *vap); +extern int xfs_vn_setattr_nonsize (struct dentry *dentry,struct iattr *vap); +extern int xfs_vn_setattr_size (struct dentry *dentry,struct iattr *vap);#endif /* __XFS_IOPS_H__ */
Signature · 1 line
2.6.6
Jan Kara
reviewerPATCH v2 2/5 ceph: Propagate dentry down to inode_change_ok()
In reply to Jan KaraTo avoid clearing of capabilities or security related extended attributes too early, inode_change_ok() will need to take dentry instead of inode. ceph_setattr() has the dentry easily available but __ceph_setattr() is also called from ceph_set_acl() where dentry is not easily available. Luckily that call path does not need inode_change_ok() to be called anyway. So reorganize functions a bit so that inode_change_ok() is called only from paths where dentry is available.
--- fs/ceph/acl.c | 5 +++++ fs/ceph/inode.c | 19 +++++++++++-------- 2 files changed, 16 insertions(+), 8 deletions(-)
fs/ceph/acl.c +5 -0
@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) goto out_free; } +if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + } +if (new_mode != old_mode) { newattrs.ia_mode = new_mode; newattrs.ia_valid =ATTR_MODE ;
fs/ceph/inode.c +11 -8
@@ -1905,13 +1905,6 @@ int __ceph_setattr(struct inode *inode, struct iattr *attr)int inode_dirty_flags =0 ;bool lock_snap_rwsem = false; -if (ceph_snap (inode) !=CEPH_NOSNAP ) -return -EROFS ; - - err =inode_change_ok (inode, attr); -if (err !=0 ) -return err; - prealloc_cf =ceph_alloc_cap_flush ();if (!prealloc_cf)return -ENOMEM ;
@@ -2124,7 +2117,17 @@ out_put:*/ int ceph_setattr (struct dentry *dentry,struct iattr *attr) { -return __ceph_setattr (d_inode (dentry), attr); +struct inode *inode =d_inode (dentry); +int err; + +if (ceph_snap (inode) !=CEPH_NOSNAP ) +return -EROFS ; + + err =inode_change_ok (inode, attr); +if (err !=0 ) +return err; + +return __ceph_setattr (inode, attr); }/*
Signature · 1 line
2.6.6
Re: [PATCH 2/5] ceph: Propagate dentry down to inode_change_ok()
Quoted 22 lines · Jan Kara
To avoid clearing of capabilities or security related extended attributes too early,
inode_change_ok()will need to take dentry instead of inode.ceph_setattr()has the dentry easily available but__ceph_setattr()is also called fromceph_set_acl()where dentry is not easily available. Luckily that call path does not needinode_change_ok()to be called anyway. So reorganize functions a bit so thatinode_change_ok()is called only from paths where dentry is available.Quoted 1 line
Signed-off-by Jan Kara <[email protected]>---
fs/ceph/acl.c | 5 +++++
fs/ceph/inode.c | 19 +++++++++++--------2 files changed, 16 insertions(+), 8 deletions(-)fs/ceph/acl.c +0
-0@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) > goto out_free; > }
Quoted 5 lines
fs/ceph/acl.c int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) · line 133
Quoted 4 lines
fs/ceph/acl.c +4
-0+if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + }+
So to make sure I understand: What's the expected behavior when someone changes the ACL in such a way that the mode gets changed? Should security_inode_killpriv be getting called in that case? If so, where does that happen, as I don't see notify_change being called in this codepath?
Quoted 32 lines
if (
new_mode!=old_mode) {Quoted 2 lines
newattrs.ia_mode =
new_mode;
newattrs.ia_valid =ATTR_MODE;fs/ceph/inode.c +0
-0@@ -1905,13 +1905,6 @@ int __ceph_setattr(struct inode *inode, struct iattr *attr) > int inode_dirty_flags = 0; > bool lock_snap_rwsem = false;fs/ceph/inode.c int __ceph_setattr(struct inode *inode, struct iattr *attr) · line 1909
Quoted 2 lines
fs/ceph/inode.c +0
-2-if (ceph_snap (inode) !=CEPH_NOSNAP ) -return -EROFS ;-
fs/ceph/inode.c int __ceph_setattr(struct inode *inode, struct iattr *attr) · line 1913
Quoted 3 lines
fs/ceph/inode.c +0
-3- err =inode_change_ok (inode, attr); -if (err !=0 ) -return err;-
Quoted 3 lines
prealloc_cf=ceph_alloc_cap_flush();
if (!prealloc_cf)
return-ENOMEM;fs/ceph/inode.c +1
-0@@ -2124,7 +2117,17 @@ out_put:*/ int ceph_setattr(struct dentry *dentry, struct iattr *attr) { > - return __ceph_setattr(d_inode(dentry), attr); > + struct inode *inode = d_inode(dentry); > + int err; + > + if (ceph_snap(inode) != CEPH_NOSNAP)fs/ceph/inode.c out_put: · line 2124
Quoted 1 line
fs/ceph/inode.c +1
-0+return -EROFS ;+
Quoted 8 lines
fs/ceph/inode.c out_put: · line 2128
Quoted 3 lines
fs/ceph/inode.c +3
-0+ err =inode_change_ok (inode, attr); +if (err !=0 ) +return err;+
fs/ceph/inode.c out_put: · line 2130
Quoted 1 line
fs/ceph/inode.c +1
-0+return __ceph_setattr (inode, attr);}
/*
Looks reasonable though:
Jan Kara
reviewerRe: [PATCH 2/5] ceph: Propagate dentry down to inode_change_ok()
Quoted 29 lines · Jeff Layton
Quoted 22 lines · Jan Kara
To avoid clearing of capabilities or security related extended attributes too early,
inode_change_ok()will need to take dentry instead of inode.ceph_setattr()has the dentry easily available but__ceph_setattr()is also called fromceph_set_acl()where dentry is not easily available. Luckily that call path does not needinode_change_ok()to be called anyway. So reorganize functions a bit so thatinode_change_ok()is called only from paths where dentry is available.Quoted 1 line
Signed-off-by Jan Kara <[email protected]>---
fs/ceph/acl.c | 5 +++++
fs/ceph/inode.c | 19 +++++++++++--------2 files changed, 16 insertions(+), 8 deletions(-)fs/ceph/acl.c +0
-0@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) > goto out_free; > }Quoted 5 lines
fs/ceph/acl.c int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) · line 133
Quoted 4 lines
fs/ceph/acl.c +4
-0+if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + }+
Quoted 5 lines
So to make sure I understand: What's the expected behavior when someone changes the ACL in such a way that the mode gets changed? Should
security_inode_killprivbe getting called in that case? If so, where does that happen, as I don't seenotify_changebeing called in this codepath?
No. security_inode_killpriv() is supposed to be called when either contents of the file changes (truncate, write) or when owner of the file changes. It is not called for permission changes.
...
Quoted 3 lines
Looks reasonable though:
Acked-by Jeff Layton <[email protected]>
Thanks for review!
Honza
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR
Re: [PATCH 2/5] ceph: Propagate dentry down to inode_change_ok()
Quoted 41 lines · Jan Kara
Quoted 34 lines · Jeff Layton
Quoted 24 lines · Jan Kara
To avoid clearing of capabilities or security related extended attributes too early,
inode_change_ok()will need to take dentry instead of inode.ceph_setattr()has the dentry easily available but__ceph_setattr()is also called fromceph_set_acl()where dentry is not easily available. Luckily that call path does not needinode_change_ok()to be called anyway. So reorganize functions a bit so thatinode_change_ok()is called only from paths where dentry is available.Quoted 2 lines
Signed-off-by Jan Kara <[email protected]>---
fs/ceph/acl.c | 5 +++++
fs/ceph/inode.c | 19 +++++++++++--------2 files changed, 16 insertions(+), 8 deletions(-)fs/ceph/acl.c +0
-0@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type)Quoted 3 lines
goto
out_free;
}Quoted 7 lines
fs/ceph/acl.c int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) · line 133
Quoted 5 lines
fs/ceph/acl.c +4
-0+if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + }+
Quoted 5 lines
So to make sure I understand: What's the expected behavior when someone changes the ACL in such a way that the mode gets changed? Should
security_inode_killprivbe getting called in that case? If so, where does that happen, as I don't seenotify_changebeing called in this codepath?
Quoted 5 lines
No.
security_inode_killpriv()is supposed to be called when either contents of the file changes (truncate, write) or when owner of the file changes. It is not called for permission changes....
Huh, ok...
I would have thought that changing the mode of the file should remove capabilities though. Consider:
Suppose we have a binary with extra file capabilities that is only executable by group owner. Attacker is able to change the permissions such that it's executable by world, and now anyone can get that capability since it wasn't revoked. IOW, I wonder if we may be trying to follow suit a little too closely with how the setuid bit works.
When you call chmod(), you have to reinstate the setuid bit anyway if you want to keep it so not revoking it is ok there. Capabilities are stored separately though, so you don't need to make the same conscious decision to keep them there. Maybe the kernel should be revoking them when the mode changes?
Quoted 8 lines
Quoted 4 lines
Looks reasonable though:
Acked-by Jeff Layton <[email protected]>Thanks for review!
Honza
Signature · 1 line
Jeff Layton <[email protected]>
Jan Kara
reviewerRe: [PATCH 2/5] ceph: Propagate dentry down to inode_change_ok()
Quoted 59 lines · Jeff Layton
Quoted 41 lines · Jan Kara
Quoted 34 lines · Jeff Layton
Quoted 24 lines · Jan Kara
To avoid clearing of capabilities or security related extended attributes too early,
inode_change_ok()will need to take dentry instead of inode.ceph_setattr()has the dentry easily available but__ceph_setattr()is also called fromceph_set_acl()where dentry is not easily available. Luckily that call path does not needinode_change_ok()to be called anyway. So reorganize functions a bit so thatinode_change_ok()is called only from paths where dentry is available.Quoted 2 lines
Signed-off-by Jan Kara <[email protected]>---
fs/ceph/acl.c | 5 +++++
fs/ceph/inode.c | 19 +++++++++++--------2 files changed, 16 insertions(+), 8 deletions(-)fs/ceph/acl.c +0
-0@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type)Quoted 3 lines
goto
out_free;
}Quoted 7 lines
fs/ceph/acl.c int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) · line 133
Quoted 5 lines
fs/ceph/acl.c +4
-0+if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + }+
Quoted 5 lines
So to make sure I understand: What's the expected behavior when someone changes the ACL in such a way that the mode gets changed? Should
security_inode_killprivbe getting called in that case? If so, where does that happen, as I don't seenotify_changebeing called in this codepath?Quoted 5 lines
No.
security_inode_killpriv()is supposed to be called when either contents of the file changes (truncate, write) or when owner of the file changes. It is not called for permission changes....
Huh, ok...
I would have thought that changing the mode of the file should remove capabilities though. Consider:
Suppose we have a binary with extra file capabilities that is only executable by group owner. Attacker is able to change the permissions such that it's executable by world, and now anyone can get that capability since it wasn't revoked. IOW, I wonder if we may be trying to follow suit a little too closely with how the setuid bit works.
Quoted 5 lines
When you call
chmod(), you have to reinstate the setuid bit anyway if you want to keep it so not revoking it is ok there. Capabilities are stored separately though, so you don't need to make the same conscious decision to keep them there. Maybe the kernel should be revoking them when the mode changes?
So this is certainly an orthogonal issue and I don't want to complicate this patch set with it. But yeah, your arguments make some sense - inadvertedly exposing an excutable with additional capabilities may be an issue. However we behave like this for a long time so I'm not sure we have the liberty of changing the behavior. Anyway, I've added linux-security-module to CC if they have anything to say to this.
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR
Re: [PATCH 2/5] ceph: Propagate dentry down to inode_change_ok()
In reply to Jan KaraQuoted 58 lines · Jan Kara
To avoid clearing of capabilities or security related extended attributes too early,
inode_change_ok()will need to take dentry instead of inode.ceph_setattr()has the dentry easily available but__ceph_setattr()is also called fromceph_set_acl()where dentry is not easily available. Luckily that call path does not needinode_change_ok()to be called anyway. So reorganize functions a bit so thatinode_change_ok()is called only from paths where dentry is available.Signed-off-by Jan Kara <[email protected]>---
fs/ceph/acl.c| 5 +++++fs/ceph/inode.c| 19 +++++++++++--------2 files changed, 16 insertions(+), 8 deletions(-)fs/ceph/acl.c +5
-0@@ -127,6 +127,11 @@ int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type) goto out_free; } +if (ceph_snap (inode) !=CEPH_NOSNAP ) { + ret = -EROFS ; + goto out_free; + } +if (new_mode != old_mode) { newattrs.ia_mode = new_mode; newattrs.ia_valid =ATTR_MODE ;fs/ceph/inode.c +2
-8@@ -1905,13 +1905,6 @@ int __ceph_setattr(struct inode *inode, struct iattr *attr)int inode_dirty_flags =0 ;bool lock_snap_rwsem = false; -if (ceph_snap (inode) !=CEPH_NOSNAP ) -return -EROFS ; - - err =inode_change_ok (inode, attr); -if (err !=0 ) -return err; - prealloc_cf =ceph_alloc_cap_flush ();if (!prealloc_cf)return -ENOMEM ;@@ -2124,7 +2117,17 @@ out_put:*/ int ceph_setattr (struct dentry *dentry,struct iattr *attr) { -return __ceph_setattr (d_inode (dentry), attr); +struct inode *inode =d_inode (dentry); +int err;
fs/ceph/inode.c out_put: · line 2131
Quoted 10 lines
fs/ceph/inode.c +9
-0+ +if (ceph_snap (inode) !=CEPH_NOSNAP ) +return -EROFS ; + + err =inode_change_ok (inode, attr); +if (err !=0 ) +return err; + +return __ceph_setattr (inode, attr); }
Quoted 5 lines
fs/ceph/inode.c +0
-0/* Signature · 1 line
2.6.6
Jan Kara
reviewerPATCH v2 3/5 fuse: Propagate dentry down to inode_change_ok()
In reply to Jan KaraTo avoid clearing of capabilities or security related extended attributes too early, inode_change_ok() will need to take dentry instead of inode. Propagate it down to fuse_do_setattr().
--- fs/fuse/dir.c | 7 ++++--- fs/fuse/file.c | 2 +- fs/fuse/fuse_i.h | 2 +- 3 files changed, 6 insertions(+), 5 deletions(-)
fs/fuse/dir.c +4 -3
@@ -1591,9 +1591,10 @@ int fuse_flush_times(struct inode *inode, struct fuse_file *ff)* vmtruncate() doesn't allow for this case, so do the rlimit checking * and the actual truncation by hand. */ -int fuse_do_setattr (struct inode *inode,struct iattr *attr, +int fuse_do_setattr (struct dentry *dentry,struct iattr *attr,struct file *file) { +struct inode *inode =d_inode (dentry);struct fuse_conn *fc =get_fuse_conn (inode);struct fuse_inode *fi =get_fuse_inode (inode);FUSE_ARGS (args);
@@ -1707,9 +1708,9 @@ static int fuse_setattr(struct dentry *entry, struct iattr *attr)return -EACCES ;if (attr->ia_valid &ATTR_FILE ) -return fuse_do_setattr (inode, attr, attr->ia_file); +return fuse_do_setattr (entry, attr, attr->ia_file);else -return fuse_do_setattr (inode, attr,NULL ); +return fuse_do_setattr (entry, attr,NULL ); } static int fuse_getattr(struct vfsmount *mnt, struct dentry *entry,
fs/fuse/file.c +1 -1
@@ -2841,7 +2841,7 @@ static void fuse_do_truncate(struct file *file) attr.ia_file = file; attr.ia_valid |=ATTR_FILE ; -fuse_do_setattr (inode, &attr, file); +fuse_do_setattr (file_dentry (file), &attr, file); }static inline loff_t fuse_round_up (loff_t off)
fs/fuse/fuse_i.h +1 -1
@@ -958,7 +958,7 @@ bool fuse_write_update_size(struct inode *inode, loff_t pos);int fuse_flush_times (struct inode *inode,struct fuse_file *ff);int fuse_write_inode (struct inode *inode,struct writeback_control *wbc); -int fuse_do_setattr (struct inode *inode,struct iattr *attr, +int fuse_do_setattr (struct dentry *dentry,struct iattr *attr,struct file *file);void fuse_set_initialized (struct fuse_conn *fc);
Signature · 1 line
2.6.6
Jan Kara
reviewerPATCH v2 4/5 fs: Give dentry to inode_change_ok() instead of inode
In reply to Jan Karainode_change_ok() will be resposible for clearing capabilities and IMA extended attributes and as such will need dentry. Give it as an argument to inode_change_ok() instead of an inode. Also rename inode_change_ok() to setattr_prepare() to better relect that it does also some modifications in addition to checks.
--- Documentation/filesystems/porting | 4 ++-- drivers/staging/lustre/lustre/llite/llite_lib.c | 2 +- fs/9p/vfs_inode.c | 2 +- fs/9p/vfs_inode_dotl.c | 2 +- fs/adfs/inode.c | 2 +- fs/affs/inode.c | 2 +- fs/attr.c | 15 +++++++++------ fs/btrfs/inode.c | 2 +- fs/ceph/inode.c | 2 +- fs/cifs/inode.c | 4 ++-- fs/ecryptfs/inode.c | 2 +- fs/exofs/inode.c | 2 +- fs/ext2/inode.c | 2 +- fs/ext4/inode.c | 2 +- fs/f2fs/file.c | 2 +- fs/fat/file.c | 2 +- fs/fuse/dir.c | 2 +- fs/gfs2/inode.c | 2 +- fs/hfs/inode.c | 2 +- fs/hfsplus/inode.c | 2 +- fs/hostfs/hostfs_kern.c | 2 +- fs/hpfs/inode.c | 2 +- fs/hugetlbfs/inode.c | 2 +- fs/jffs2/fs.c | 2 +- fs/jfs/file.c | 2 +- fs/kernfs/inode.c | 2 +- fs/libfs.c | 2 +- fs/logfs/file.c | 2 +- fs/minix/file.c | 2 +- fs/ncpfs/inode.c | 2 +- fs/nfsd/nfsproc.c | 8 +++----- fs/nilfs2/inode.c | 2 +- fs/ntfs/inode.c | 2 +- fs/ocfs2/dlmfs/dlmfs.c | 2 +- fs/ocfs2/file.c | 2 +- fs/omfs/file.c | 2 +- fs/orangefs/inode.c | 2 +- fs/overlayfs/inode.c | 2 +- fs/proc/base.c | 2 +- fs/proc/generic.c | 2 +- fs/proc/proc_sysctl.c | 2 +- fs/ramfs/file-nommu.c | 2 +- fs/reiserfs/inode.c | 2 +- fs/sysv/file.c | 2 +- fs/ubifs/file.c | 2 +- fs/udf/file.c | 2 +- fs/ufs/inode.c | 2 +- fs/utimes.c | 4 ++-- fs/xfs/xfs_iops.c | 10 ++++------ include/linux/fs.h | 2 +- mm/shmem.c | 2 +- 51 files changed, 67 insertions(+), 68 deletions(-)
Documentation/filesystems/porting +2 -2
@@ -287,8 +287,8 @@ implementing on-disk size changes. Start with a copy of the old inode_setattr and vmtruncate, and the reorder the vmtruncate + foofs_vmtruncate sequence to be in order of zeroing blocks using block_truncate_page or similar helpers, size update and on finally on-disk truncation which should not fail. -inode_change_ok now includes the size checks for ATTR_SIZE and must be called -in the beginning of ->setattr unconditionally. +setattr_prepare (which used to be inode_change_ok) now includes the size checks +for ATTR_SIZE and must be called in the beginning of ->setattr unconditionally. [mandatory]
drivers/staging/lustre/lustre/llite/llite_lib.c +1 -1
@@ -1192,7 +1192,7 @@ int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import) attr->ia_valid |=ATTR_MTIME |ATTR_CTIME ; } -/* POSIX: check before ATTR_*TIME_SET set (from inode_change_ok) */ +/* POSIX: check before ATTR_*TIME_SET set (from setattr_prepare) */ if (attr->ia_valid &TIMES_SET_FLAGS ) {if ((!uid_eq (current_fsuid (), inode->i_uid)) && !capable (CFS_CAP_FOWNER ))
fs/9p/vfs_inode.c +1 -1
@@ -1094,7 +1094,7 @@ static int v9fs_vfs_setattr(struct dentry *dentry, struct iattr *iattr)struct p9_wstat wstat;p9_debug (P9_DEBUG_VFS ,"\n" ); - retval =inode_change_ok (d_inode (dentry), iattr); + retval =setattr_prepare (dentry, iattr);if (retval)return retval;
fs/9p/vfs_inode_dotl.c +1 -1
@@ -558,7 +558,7 @@ int v9fs_vfs_setattr_dotl(struct dentry *dentry, struct iattr *iattr)p9_debug (P9_DEBUG_VFS ,"\n" ); - retval =inode_change_ok (inode, iattr); + retval =setattr_prepare (dentry, iattr);if (retval)return retval;
fs/adfs/inode.c +1 -1
@@ -303,7 +303,7 @@ adfs_notify_change(struct dentry *dentry, struct iattr *attr)unsigned int ia_valid = attr->ia_valid;int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);/* * we can't change the UID or GID of any file -
fs/affs/inode.c +1 -1
@@ -219,7 +219,7 @@ affs_notify_change(struct dentry *dentry, struct iattr *attr)pr_debug ("notify_change(%lu,0x%x)\n" , inode->i_ino, attr->ia_valid); - error =inode_change_ok (inode,attr); + error =setattr_prepare (dentry, attr);if (error) goto out;
fs/attr.c +9 -6
@@ -17,19 +17,22 @@#include <linux/ima.h> /** -* inode_change_ok - check if attribute changes to an inode are allowed -* @inode: inode to check +* setattr_prepare - check if attribute changes to a dentry are allowed +* @dentry: dentry to check * @attr: attributes to change * * Check if we are allowed to change the attributes contained in @attr -* in the given inode. This includes the normal unix access permission -* checks, as well as checks for rlimits and others. +* in the given dentry. This includes the normal unix access permission +* checks, as well as checks for rlimits and others. The function also clears +* SGID bit from mode if user is not allowed to set it. Also file capabilities +* and IMA extended attributes are cleared if ATTR_KILL_PRIV is set. * * Should be called as the first thing in ->setattr implementations, * possibly after taking additional locks. */ -int inode_change_ok (const struct inode *inode,struct iattr *attr) +int setattr_prepare (struct dentry *dentry,struct iattr *attr) { +struct inode *inode =d_inode (dentry);unsigned int ia_valid = attr->ia_valid;/*
@@ -79,7 +82,7 @@ int inode_change_ok(const struct inode *inode, struct iattr *attr)return 0 ; } -EXPORT_SYMBOL (inode_change_ok); +EXPORT_SYMBOL (setattr_prepare);/** * inode_newsize_ok - may this inode be truncated to a given size
fs/btrfs/inode.c +1 -1
@@ -5072,7 +5072,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)if (btrfs_root_readonly (root))return -EROFS ; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/ceph/inode.c +1 -1
@@ -2123,7 +2123,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)if (ceph_snap (inode) !=CEPH_NOSNAP )return -EROFS ; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err !=0 )return err;
fs/cifs/inode.c +2 -2
@@ -2154,7 +2154,7 @@ cifs_setattr_unix(struct dentry *direntry, struct iattr *attrs)if (cifs_sb->mnt_cifs_flags &CIFS_MOUNT_NO_PERM ) attrs->ia_valid |=ATTR_FORCE ; - rc =inode_change_ok (inode, attrs); + rc =setattr_prepare (direntry, attrs);if (rc <0 ) goto out;
@@ -2294,7 +2294,7 @@ cifs_setattr_nounix(struct dentry *direntry, struct iattr *attrs)if (cifs_sb->mnt_cifs_flags &CIFS_MOUNT_NO_PERM ) attrs->ia_valid |=ATTR_FORCE ; - rc =inode_change_ok (inode, attrs); + rc =setattr_prepare (direntry, attrs);if (rc <0 ) {free_xid (xid);return rc;
fs/ecryptfs/inode.c +1 -1
@@ -927,7 +927,7 @@ static int ecryptfs_setattr(struct dentry *dentry, struct iattr *ia) }mutex_unlock (&crypt_stat->cs_mutex); - rc =inode_change_ok (inode, ia); + rc =setattr_prepare (dentry, ia);if (rc) goto out;if (ia->ia_valid &ATTR_SIZE ) {
fs/exofs/inode.c +1 -1
@@ -1034,7 +1034,7 @@ int exofs_setattr(struct dentry *dentry, struct iattr *iattr)if (unlikely (error))return error; - error =inode_change_ok (inode, iattr); + error =setattr_prepare (dentry, iattr);if (unlikely (error))return error;
fs/ext2/inode.c +1 -1
@@ -1580,7 +1580,7 @@ int ext2_setattr(struct dentry *dentry, struct iattr *iattr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, iattr); + error =setattr_prepare (dentry, iattr);if (error)return error;
fs/ext4/inode.c +1 -1
@@ -5073,7 +5073,7 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)int orphan =0 ;const unsigned int ia_valid = attr->ia_valid; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/f2fs/file.c +1 -1
@@ -680,7 +680,7 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int err; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/fat/file.c +1 -1
@@ -450,7 +450,7 @@ int fat_setattr(struct dentry *dentry, struct iattr *attr) attr->ia_valid &= ~TIMES_SET_FLAGS ; } - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr); attr->ia_valid = ia_valid;if (error) {if (sbi->options.quiet)
fs/fuse/dir.c +1 -1
@@ -1609,7 +1609,7 @@ int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,if (!(fc->flags &FUSE_DEFAULT_PERMISSIONS )) attr->ia_valid |=ATTR_FORCE ; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/gfs2/inode.c +1 -1
@@ -1932,7 +1932,7 @@ static int gfs2_setattr(struct dentry *dentry, struct iattr *attr)if (IS_IMMUTABLE (inode) ||IS_APPEND (inode)) goto out; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error) goto out;
fs/hfs/inode.c +1 -1
@@ -605,7 +605,7 @@ int hfs_inode_setattr(struct dentry *dentry, struct iattr * attr)struct hfs_sb_info *hsb =HFS_SB (inode->i_sb);int error; - error =inode_change_ok (inode, attr);/* basic permission checks */ + error =setattr_prepare (dentry, attr);/* basic permission checks */ if (error)return error;
fs/hfsplus/inode.c +1 -1
@@ -245,7 +245,7 @@ static int hfsplus_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/hostfs/hostfs_kern.c +1 -1
@@ -812,7 +812,7 @@ static int hostfs_setattr(struct dentry *dentry, struct iattr *attr)int fd =HOSTFS_I (inode)->fd; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/hpfs/inode.c +1 -1
@@ -273,7 +273,7 @@ int hpfs_setattr(struct dentry *dentry, struct iattr *attr)if ((attr->ia_valid &ATTR_SIZE ) && attr->ia_size > inode->i_size) goto out_unlock; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error) goto out_unlock;
fs/hugetlbfs/inode.c +1 -1
@@ -672,7 +672,7 @@ static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr)BUG_ON (!inode); - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/jffs2/fs.c +1 -1
@@ -193,7 +193,7 @@ int jffs2_setattr(struct dentry *dentry, struct iattr *iattr)struct inode *inode =d_inode (dentry);int rc; - rc =inode_change_ok (inode, iattr); + rc =setattr_prepare (dentry, iattr);if (rc)return rc;
fs/jfs/file.c +1 -1
@@ -103,7 +103,7 @@ int jfs_setattr(struct dentry *dentry, struct iattr *iattr)struct inode *inode =d_inode (dentry);int rc; - rc =inode_change_ok (inode, iattr); + rc =setattr_prepare (dentry, iattr);if (rc)return rc;
fs/kernfs/inode.c +1 -1
@@ -122,7 +122,7 @@ int kernfs_iop_setattr(struct dentry *dentry, struct iattr *iattr)return -EINVAL ;mutex_lock (&kernfs_mutex); - error =inode_change_ok (inode, iattr); + error =setattr_prepare (dentry, iattr);if (error) goto out;
fs/libfs.c +1 -1
@@ -394,7 +394,7 @@ int simple_setattr(struct dentry *dentry, struct iattr *iattr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, iattr); + error =setattr_prepare (dentry, iattr);if (error)return error;
fs/logfs/file.c +1 -1
@@ -244,7 +244,7 @@ static int logfs_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int err =0 ; - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/minix/file.c +1 -1
@@ -26,7 +26,7 @@ static int minix_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/ncpfs/inode.c +1 -1
@@ -884,7 +884,7 @@ int ncp_notify_change(struct dentry *dentry, struct iattr *attr)/* ageing the dentry to force validation */ ncp_age_dentry (server, dentry); - result =inode_change_ok (inode, attr); + result =setattr_prepare (dentry, attr);if (result <0 ) goto out;
fs/nfsd/nfsproc.c +3 -5
@@ -74,10 +74,10 @@ nfsd_proc_setattr(struct svc_rqst *rqstp, struct nfsd_sattrargs *argp,* which only requires access, and "set-[ac]time-to-X" which * requires ownership. * So if it looks like it might be "set both to the same time which -* is close to now", and if inode_change_ok fails, then we +* is close to now", and if setattr_prepare fails, then we * convert to "set to now" instead of "set to explicit time" * -* We only call inode_change_ok as the last test as technically +* We only call setattr_prepare as the last test as technically * it is not an interface that we should be using. */ #define BOTH_TIME_SET (ATTR_ATIME_SET | ATTR_MTIME_SET)
@@ -92,17 +92,15 @@ nfsd_proc_setattr(struct svc_rqst *rqstp, struct nfsd_sattrargs *argp,* request is. We require it be within 30 minutes of now. */ time_t delta = iap->ia_atime.tv_sec -get_seconds (); -struct inode *inode; nfserr =fh_verify (rqstp, fhp,0 ,NFSD_MAY_NOP );if (nfserr) goto done; - inode =d_inode (fhp->fh_dentry);if (delta <0 ) delta = -delta;if (delta <MAX_TOUCH_TIME_ERROR && -inode_change_ok (inode, iap) !=0 ) { +setattr_prepare (fhp->fh_dentry, iap) !=0 ) {/* * Turn off ATTR_[AM]TIME_SET but leave ATTR_[AM]TIME. * This will cause notify_change to set these times
fs/nilfs2/inode.c +1 -1
@@ -829,7 +829,7 @@ int nilfs_setattr(struct dentry *dentry, struct iattr *iattr)struct super_block *sb = inode->i_sb;int err; - err =inode_change_ok (inode, iattr); + err =setattr_prepare (dentry, iattr);if (err)return err;
fs/ntfs/inode.c +1 -1
@@ -2893,7 +2893,7 @@ int ntfs_setattr(struct dentry *dentry, struct iattr *attr)int err;unsigned int ia_valid = attr->ia_valid; - err =inode_change_ok (vi, attr); + err =setattr_prepare (dentry, attr);if (err) goto out;/* We do not support NTFS ACLs yet. */
fs/ocfs2/dlmfs/dlmfs.c +1 -1
@@ -211,7 +211,7 @@ static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry); attr->ia_valid &= ~ATTR_SIZE ; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/ocfs2/file.c +1 -1
@@ -1155,7 +1155,7 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)if (!(attr->ia_valid &OCFS2_VALID_ATTRS ))return 0 ; - status =inode_change_ok (inode, attr); + status =setattr_prepare (dentry, attr);if (status)return status;
fs/omfs/file.c +1 -1
@@ -349,7 +349,7 @@ static int omfs_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/orangefs/inode.c +1 -1
@@ -219,7 +219,7 @@ int orangefs_setattr(struct dentry *dentry, struct iattr *iattr)"orangefs_setattr: called on %s\n" , dentry->d_name.name); - ret =inode_change_ok (inode, iattr); + ret =setattr_prepare (dentry, iattr);if (ret) goto out;
fs/overlayfs/inode.c +1 -1
@@ -53,7 +53,7 @@ int ovl_setattr(struct dentry *dentry, struct iattr *attr)* inode_newsize_ok() will always check against MAX_LFS_FILESIZE and not * check for a swapfile (which this won't be anyway). */ - err =inode_change_ok (dentry->d_inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/proc/base.c +1 -1
@@ -709,7 +709,7 @@ int proc_setattr(struct dentry *dentry, struct iattr *attr)if (attr->ia_valid &ATTR_MODE )return -EPERM ; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/proc/generic.c +1 -1
@@ -105,7 +105,7 @@ static int proc_notify_change(struct dentry *dentry, struct iattr *iattr)struct proc_dir_entry *de =PDE (inode);int error; - error =inode_change_ok (inode, iattr); + error =setattr_prepare (dentry, iattr);if (error)return error;
fs/proc/proc_sysctl.c +1 -1
@@ -754,7 +754,7 @@ static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)if (attr->ia_valid & (ATTR_MODE |ATTR_UID |ATTR_GID ))return -EPERM ; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/ramfs/file-nommu.c +1 -1
@@ -169,7 +169,7 @@ static int ramfs_nommu_setattr(struct dentry *dentry, struct iattr *ia)int ret =0 ;/* POSIX UID/GID verification for setting inode attributes */ - ret =inode_change_ok (inode, ia); + ret =setattr_prepare (dentry, ia);if (ret)return ret;
fs/reiserfs/inode.c +1 -1
@@ -3312,7 +3312,7 @@ int reiserfs_setattr(struct dentry *dentry, struct iattr *attr)unsigned int ia_valid;int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/sysv/file.c +1 -1
@@ -33,7 +33,7 @@ static int sysv_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/ubifs/file.c +1 -1
@@ -1262,7 +1262,7 @@ int ubifs_setattr(struct dentry *dentry, struct iattr *attr)dbg_gen ("ino %lu, mode %#x, ia_valid %#x" , inode->i_ino, inode->i_mode, attr->ia_valid); - err =inode_change_ok (inode, attr); + err =setattr_prepare (dentry, attr);if (err)return err;
fs/udf/file.c +1 -1
@@ -247,7 +247,7 @@ static int udf_setattr(struct dentry *dentry, struct iattr *attr)struct inode *inode =d_inode (dentry);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/ufs/inode.c +1 -1
@@ -1208,7 +1208,7 @@ int ufs_setattr(struct dentry *dentry, struct iattr *attr)unsigned int ia_valid = attr->ia_valid;int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
fs/utimes.c +2 -2
@@ -81,7 +81,7 @@ static int utimes_common(struct path *path, struct timespec *times) newattrs.ia_valid |=ATTR_MTIME_SET ; }/* -* Tell inode_change_ok(), that this is an explicit time +* Tell setattr_prepare(), that this is an explicit time * update, even if neither ATTR_ATIME_SET nor ATTR_MTIME_SET * were used. */
@@ -90,7 +90,7 @@ static int utimes_common(struct path *path, struct timespec *times)/* * If times is NULL (or both times are UTIME_NOW), * then we need to check permissions, because -* inode_change_ok() won't do it. +* setattr_prepare() won't do it. */ error = -EPERM ;if (IS_IMMUTABLE (inode))
fs/xfs/xfs_iops.c +4 -6
@@ -547,9 +547,7 @@ xfs_vn_change_ok(struct dentry *dentry, struct iattr *iattr) { - struct inode *inode = d_inode(dentry); -struct xfs_inode *ip =XFS_I (inode); -struct xfs_mount *mp = ip->i_mount; + struct xfs_mount *mp = XFS_I(d_inode(dentry))->i_mount;if (mp->m_flags &XFS_MOUNT_RDONLY )return -EROFS ;
@@ -557,14 +555,14 @@ xfs_vn_change_ok(if (XFS_FORCED_SHUTDOWN (mp))return -EIO ; -return inode_change_ok (inode, iattr); +return setattr_prepare (dentry, iattr); }/* * Set non-size attributes of an inode. * * Caution: The caller of this function is responsible for calling -* inode_change_ok() or otherwise verifying the change is fine. +* setattr_prepare() or otherwise verifying the change is fine. */ int xfs_setattr_nonsize(
@@ -772,7 +770,7 @@ xfs_vn_setattr_nonsize(* Truncate file. Must have write permission and not be a directory. * * Caution: The caller of this function is responsible for calling -* inode_change_ok() or otherwise verifying the change is fine. +* setattr_prepare() or otherwise verifying the change is fine. */ int xfs_setattr_size(
include/linux/fs.h +1 -1
@@ -2995,7 +2995,7 @@ extern int buffer_migrate_page(struct address_space *,#define buffer_migrate_page NULL#endif -extern int inode_change_ok (const struct inode *,struct iattr *); +extern int setattr_prepare (struct dentry *,struct iattr *);extern int inode_newsize_ok (const struct inode *,loff_t offset);extern void setattr_copy (struct inode *inode,const struct iattr *attr);
mm/shmem.c +1 -1
@@ -959,7 +959,7 @@ static int shmem_setattr(struct dentry *dentry, struct iattr *attr)struct shmem_sb_info *sbinfo =SHMEM_SB (inode->i_sb);int error; - error =inode_change_ok (inode, attr); + error =setattr_prepare (dentry, attr);if (error)return error;
Signature · 1 line
2.6.6
Jan Kara
reviewerPATCH v2 5/5 fs: Avoid premature clearing of capabilities
In reply to Jan KaraCurrently, notify_change() clears capabilities or IMA attributes by calling security_inode_killpriv() before calling into ->setattr. Thus it happens before any other permission checks in inode_change_ok() and user is thus allowed to trigger clearing of capabilities or IMA attributes for any file he can look up e.g. by calling chown for that file. This is unexpected and can lead to user DoSing a system.
Fix the problem by calling security_inode_killpriv() at the end of inode_change_ok() instead of from notify_change(). At that moment we are sure user has permissions to do the requested change.
References: CVE-2015-1350
--- fs/attr.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-)
fs/attr.c +14 -6
@@ -47,7 +47,7 @@ int setattr_prepare(struct dentry *dentry, struct iattr *attr)/* If force is set do it anyway. */ if (ia_valid &ATTR_FORCE ) -return 0 ; + goto kill_priv;/* Make sure a caller can chown. */ if ((ia_valid & ATTR_UID) &&
@@ -80,6 +80,16 @@ int setattr_prepare(struct dentry *dentry, struct iattr *attr)return -EPERM ; } +kill_priv: +/* User has permission for the change */ +if (ia_valid &ATTR_KILL_PRIV ) { +int error; + + error =security_inode_killpriv (dentry); +if (error) +return error; + } +return 0 ; }EXPORT_SYMBOL (setattr_prepare);
@@ -220,13 +230,11 @@ int notify_change(struct dentry * dentry, struct iattr * attr, struct inode **deif (!(ia_valid &ATTR_MTIME_SET )) attr->ia_mtime = now;if (ia_valid &ATTR_KILL_PRIV ) { - attr->ia_valid &= ~ATTR_KILL_PRIV ; - ia_valid &= ~ATTR_KILL_PRIV ; error =security_inode_need_killpriv (dentry); -if (error >0 ) - error =security_inode_killpriv (dentry); -if (error) +if (error <0 )return error; +if (error ==0 ) + ia_valid = attr->ia_valid &= ~ATTR_KILL_PRIV ; }/*
Signature · 1 line
2.6.6
Quoted 5 lines · Jan Kara
Hello,
I'm resending the patches since Al did not respond to the series yet (I've previously posted it in [1] on Aug 3. Al? If Al does not respond, can you Andrew please feed to patches to Linus? Thanks.
Yes, please get the series in. I thought I reviewed it before, but here it comes again: