[PATCH v2 0/3] evm: disable EVM on overlayfs

[PATCH v2 0/3] evm: disable EVM on overlayfs

[Mimi Zohar]

EVM verifies the existing 'security.evm' value, before allowing it
to be updated.  The EVM HMAC and the original file signatures contain
filesystem specific metadata (e.g. i_ino, i_generation and s_uuid).

This poses a challenge when transitioning from the lower backing file
to the upper backing file.

Until a complete solution is developed, disable EVM on overlayfs.

Changelog v2:
Addressed Amir's comments:
- Simplified security_inode_copy_up_xattr() return.
- Identified filesystems that don't support EVM based on a new SB_I flag.

Mimi Zohar (3):
  evm: don't copy up 'security.evm' xattr
  evm: add support to disable EVM on unsupported filesystems
  overlay: disable EVM

 fs/overlayfs/super.c              |  1 +
 include/linux/evm.h               |  6 +++++
 include/linux/fs.h                |  1 +
 security/integrity/evm/evm_main.c | 42 ++++++++++++++++++++++++++++++-
 security/security.c               |  2 +-
 5 files changed, 50 insertions(+), 2 deletions(-)

-- 
2.39.3

[PATCH v2 1/3] evm: don't copy up 'security.evm' xattr

[Mimi Zohar]

The security.evm HMAC and the original file signatures contain
filesystem specific data.  As a result, the HMAC and signature
are not the same on the stacked and backing filesystems.

Don't copy up 'security.evm'.

Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
---
 include/linux/evm.h               | 6 ++++++
 security/integrity/evm/evm_main.c | 7 +++++++
 security/security.c               | 2 +-
 3 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/include/linux/evm.h b/include/linux/evm.h
index 01fc495a83e2..36ec884320d9 100644
--- a/include/linux/evm.h
+++ b/include/linux/evm.h
@@ -31,6 +31,7 @@ extern void evm_inode_post_setxattr(struct dentry *dentry,
 				    const char *xattr_name,
 				    const void *xattr_value,
 				    size_t xattr_value_len);
+extern int evm_inode_copy_up_xattr(const char *name);
 extern int evm_inode_removexattr(struct mnt_idmap *idmap,
 				 struct dentry *dentry, const char *xattr_name);
 extern void evm_inode_post_removexattr(struct dentry *dentry,
@@ -117,6 +118,11 @@ static inline void evm_inode_post_setxattr(struct dentry *dentry,
 	return;
 }
 
+static inline int  evm_inode_copy_up_xattr(const char *name)
+{
+	return 0;
+}
+
 static inline int evm_inode_removexattr(struct mnt_idmap *idmap,
 					struct dentry *dentry,
 					const char *xattr_name)
diff --git a/security/integrity/evm/evm_main.c b/security/integrity/evm/evm_main.c
index 894570fe39bc..02adba635b02 100644
--- a/security/integrity/evm/evm_main.c
+++ b/security/integrity/evm/evm_main.c
@@ -863,6 +863,13 @@ void evm_inode_post_setattr(struct dentry *dentry, int ia_valid)
 		evm_update_evmxattr(dentry, NULL, NULL, 0);
 }
 
+int evm_inode_copy_up_xattr(const char *name)
+{
+	if (strcmp(name, XATTR_NAME_EVM) == 0)
+		return 1; /* Discard */
+	return -EOPNOTSUPP;
+}
+
 /*
  * evm_inode_init_security - initializes security.evm HMAC value
  */
diff --git a/security/security.c b/security/security.c
index dcb3e7014f9b..f00ec4d988b8 100644
--- a/security/security.c
+++ b/security/security.c
@@ -2539,7 +2539,7 @@ int security_inode_copy_up_xattr(const char *name)
 			return rc;
 	}
 
-	return LSM_RET_DEFAULT(inode_copy_up_xattr);
+	return evm_inode_copy_up_xattr(name);
 }
 EXPORT_SYMBOL(security_inode_copy_up_xattr);
 
-- 
2.39.3

[PATCH v2 2/3] evm: add support to disable EVM on unsupported filesystems

[Mimi Zohar]

Identify EVM unsupported filesystems by defining a new flag
SB_I_EVM_UNSUPPORTED.

Don't verify, write, remove or update 'security.evm' on unsupported
filesystems.

Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
---
 include/linux/fs.h                |  1 +
 security/integrity/evm/evm_main.c | 35 ++++++++++++++++++++++++++++++-
 2 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/include/linux/fs.h b/include/linux/fs.h
index 98b7a7a8c42e..db9350a734ef 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1164,6 +1164,7 @@ extern int send_sigurg(struct fown_struct *fown);
 #define SB_I_USERNS_VISIBLE		0x00000010 /* fstype already mounted */
 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE	0x00000020
 #define SB_I_UNTRUSTED_MOUNTER		0x00000040
+#define SB_I_EVM_UNSUPPORTED		0x00000050
 
 #define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
 #define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
diff --git a/security/integrity/evm/evm_main.c b/security/integrity/evm/evm_main.c
index 02adba635b02..cc7956d7878b 100644
--- a/security/integrity/evm/evm_main.c
+++ b/security/integrity/evm/evm_main.c
@@ -151,6 +151,17 @@ static int evm_find_protected_xattrs(struct dentry *dentry)
 	return count;
 }
 
+static int is_unsupported_fs(struct dentry *dentry)
+{
+	struct inode *inode = d_backing_inode(dentry);
+
+	if (inode->i_sb->s_iflags & SB_I_EVM_UNSUPPORTED) {
+		pr_info_once("%s not supported\n", inode->i_sb->s_type->name);
+		return 1;
+	}
+	return 0;
+}
+
 /*
  * evm_verify_hmac - calculate and compare the HMAC with the EVM xattr
  *
@@ -181,6 +192,9 @@ static enum integrity_status evm_verify_hmac(struct dentry *dentry,
 		     iint->evm_status == INTEGRITY_PASS_IMMUTABLE))
 		return iint->evm_status;
 
+	if (is_unsupported_fs(dentry))
+		return INTEGRITY_UNKNOWN;
+
 	/* if status is not PASS, try to check again - against -ENOMEM */
 
 	/* first need to know the sig type */
@@ -408,6 +422,9 @@ enum integrity_status evm_verifyxattr(struct dentry *dentry,
 	if (!evm_key_loaded() || !evm_protected_xattr(xattr_name))
 		return INTEGRITY_UNKNOWN;
 
+	if (is_unsupported_fs(dentry))
+		return INTEGRITY_UNKNOWN;
+
 	if (!iint) {
 		iint = integrity_iint_find(d_backing_inode(dentry));
 		if (!iint)
@@ -491,15 +508,21 @@ static int evm_protect_xattr(struct mnt_idmap *idmap,
 	if (strcmp(xattr_name, XATTR_NAME_EVM) == 0) {
 		if (!capable(CAP_SYS_ADMIN))
 			return -EPERM;
+		if (is_unsupported_fs(dentry))
+			return -EPERM;
 	} else if (!evm_protected_xattr(xattr_name)) {
 		if (!posix_xattr_acl(xattr_name))
 			return 0;
+		if (is_unsupported_fs(dentry))
+			return 0;
+
 		evm_status = evm_verify_current_integrity(dentry);
 		if ((evm_status == INTEGRITY_PASS) ||
 		    (evm_status == INTEGRITY_NOXATTRS))
 			return 0;
 		goto out;
-	}
+	} else if (is_unsupported_fs(dentry))
+		return 0;
 
 	evm_status = evm_verify_current_integrity(dentry);
 	if (evm_status == INTEGRITY_NOXATTRS) {
@@ -750,6 +773,9 @@ void evm_inode_post_setxattr(struct dentry *dentry, const char *xattr_name,
 	if (!(evm_initialized & EVM_INIT_HMAC))
 		return;
 
+	if (is_unsupported_fs(dentry))
+		return;
+
 	evm_update_evmxattr(dentry, xattr_name, xattr_value, xattr_value_len);
 }
 
@@ -814,8 +840,12 @@ int evm_inode_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
 	if (evm_initialized & EVM_ALLOW_METADATA_WRITES)
 		return 0;
 
+	if (is_unsupported_fs(dentry))
+		return 0;
+
 	if (!(ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID)))
 		return 0;
+
 	evm_status = evm_verify_current_integrity(dentry);
 	/*
 	 * Writing attrs is safe for portable signatures, as portable signatures
@@ -859,6 +889,9 @@ void evm_inode_post_setattr(struct dentry *dentry, int ia_valid)
 	if (!(evm_initialized & EVM_INIT_HMAC))
 		return;
 
+	if (is_unsupported_fs(dentry))
+		return;
+
 	if (ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
 		evm_update_evmxattr(dentry, NULL, NULL, 0);
 }
-- 
2.39.3

Re: [PATCH v2 2/3] evm: add support to disable EVM on unsupported filesystems

[Mimi Zohar]

On Tue, 2023-12-19 at 12:52 -0500, Mimi Zohar wrote:

> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 98b7a7a8c42e..db9350a734ef 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -1164,6 +1164,7 @@ extern int send_sigurg(struct fown_struct *fown);
>  #define SB_I_USERNS_VISIBLE		0x00000010 /* fstype already mounted */
>  #define SB_I_IMA_UNVERIFIABLE_SIGNATURE	0x00000020
>  #define SB_I_UNTRUSTED_MOUNTER		0x00000040
> +#define SB_I_EVM_UNSUPPORTED		0x00000050

This needs to be fixed.

Mimi

>  
>  #define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
>  #define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
> diff --git a/security/integrity/evm/evm_main.c b/security/integrity/evm/evm_main.c

Re: [PATCH v2 2/3] evm: add support to disable EVM on unsupported filesystems

[Amir Goldstein]

On Tue, Dec 19, 2023 at 9:10 PM Mimi Zohar <zohar@linux.ibm.com> wrote:
>
> On Tue, 2023-12-19 at 12:52 -0500, Mimi Zohar wrote:
>
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index 98b7a7a8c42e..db9350a734ef 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -1164,6 +1164,7 @@ extern int send_sigurg(struct fown_struct *fown);
> >  #define SB_I_USERNS_VISIBLE          0x00000010 /* fstype already mounted */
> >  #define SB_I_IMA_UNVERIFIABLE_SIGNATURE      0x00000020
> >  #define SB_I_UNTRUSTED_MOUNTER               0x00000040
> > +#define SB_I_EVM_UNSUPPORTED         0x00000050
>
> This needs to be fixed.
>

With this fixed, you may add:

Acked-by: Amir Goldstein <amir73il@gmail.com>

Thanks,
Amir.

[PATCH v2 3/3] overlay: disable EVM

[Mimi Zohar]

Until a complete solution is developed, update 'sb->s_iflags' to
disable EVM.

Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
---
 fs/overlayfs/super.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
index a0967bb25003..e3d9c6c80a47 100644
--- a/fs/overlayfs/super.c
+++ b/fs/overlayfs/super.c
@@ -1454,6 +1454,7 @@ int ovl_fill_super(struct super_block *sb, struct fs_context *fc)
 	 * lead to unexpected results.
 	 */
 	sb->s_iflags |= SB_I_NOUMASK;
+	sb->s_iflags |= SB_I_EVM_UNSUPPORTED;
 
 	err = -ENOMEM;
 	root_dentry = ovl_get_root(sb, ctx->upper.dentry, oe);
-- 
2.39.3

Re: [PATCH v2 3/3] overlay: disable EVM

[Amir Goldstein]

On Tue, Dec 19, 2023 at 7:52 PM Mimi Zohar <zohar@linux.ibm.com> wrote:
>
> Until a complete solution is developed, update 'sb->s_iflags' to
> disable EVM.
>
> Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>

Acked-by: Amir Goldstein <amir73il@gmail.com>

> ---
>  fs/overlayfs/super.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
> index a0967bb25003..e3d9c6c80a47 100644
> --- a/fs/overlayfs/super.c
> +++ b/fs/overlayfs/super.c
> @@ -1454,6 +1454,7 @@ int ovl_fill_super(struct super_block *sb, struct fs_context *fc)
>          * lead to unexpected results.
>          */
>         sb->s_iflags |= SB_I_NOUMASK;
> +       sb->s_iflags |= SB_I_EVM_UNSUPPORTED;
>
>         err = -ENOMEM;
>         root_dentry = ovl_get_root(sb, ctx->upper.dentry, oe);
> --
> 2.39.3
>

Re: [PATCH v2 0/3] evm: disable EVM on overlayfs

[Christian Brauner]

On Tue, Dec 19, 2023 at 12:52:03PM -0500, Mimi Zohar wrote:
> EVM verifies the existing 'security.evm' value, before allowing it
> to be updated.  The EVM HMAC and the original file signatures contain
> filesystem specific metadata (e.g. i_ino, i_generation and s_uuid).
> 
> This poses a challenge when transitioning from the lower backing file
> to the upper backing file.
> 
> Until a complete solution is developed, disable EVM on overlayfs.
> 
> Changelog v2:
> Addressed Amir's comments:
> - Simplified security_inode_copy_up_xattr() return.
> - Identified filesystems that don't support EVM based on a new SB_I flag.

We're wasting a flag for a single filesystem but we do have enough of
them left so I think this is ok,

Reviewed-by: Christian Brauner <brauner@kernel.org>

Re: [PATCH v2 0/3] evm: disable EVM on overlayfs

[Mimi Zohar]

On Wed, 2023-12-20 at 13:35 +0100, Christian Brauner wrote:
> On Tue, Dec 19, 2023 at 12:52:03PM -0500, Mimi Zohar wrote:
> > EVM verifies the existing 'security.evm' value, before allowing it
> > to be updated.  The EVM HMAC and the original file signatures contain
> > filesystem specific metadata (e.g. i_ino, i_generation and s_uuid).
> > 
> > This poses a challenge when transitioning from the lower backing file
> > to the upper backing file.
> > 
> > Until a complete solution is developed, disable EVM on overlayfs.
> > 
> > Changelog v2:
> > Addressed Amir's comments:
> > - Simplified security_inode_copy_up_xattr() return.
> > - Identified filesystems that don't support EVM based on a new SB_I flag.
> 
> We're wasting a flag for a single filesystem but we do have enough of
> them left so I think this is ok,

Thanks, Christian.

> 
> Reviewed-by: Christian Brauner <brauner@kernel.org>