2008-07-30 12:06:44

by Miklos Szeredi

Subject: [patch] vfs: make security_inode_setattr() calling consistent

From: Miklos Szeredi <[email protected]>

Call security_inode_setattr() consistetly before inode_change_ok().
It doesn't make sense to try to "optimize" the i_op->setattr == NULL
case, as most filesystem do define their own setattr function.

Signed-off-by: Miklos Szeredi <[email protected]>
---
fs/attr.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)

Index: linux-2.6/fs/attr.c
===================================================================
--- linux-2.6.orig/fs/attr.c 2008-07-30 13:52:35.000000000 +0200
+++ linux-2.6/fs/attr.c 2008-07-30 13:52:41.000000000 +0200
@@ -159,17 +159,17 @@ int notify_change(struct dentry * dentry
if (!(attr->ia_valid & ~(ATTR_KILL_SUID | ATTR_KILL_SGID)))
return 0;

+ error = security_inode_setattr(dentry, attr);
+ if (error)
+ return error;
+
if (ia_valid & ATTR_SIZE)
down_write(&dentry->d_inode->i_alloc_sem);

if (inode->i_op && inode->i_op->setattr) {
- error = security_inode_setattr(dentry, attr);
- if (!error)
- error = inode->i_op->setattr(dentry, attr);
+ error = inode->i_op->setattr(dentry, attr);
} else {
error = inode_change_ok(inode, attr);
- if (!error)
- error = security_inode_setattr(dentry, attr);
if (!error) {
if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
(ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid))

2008-07-30 14:38:07

by Stephen Smalley

Subject: Re: [patch] vfs: make security_inode_setattr() calling consistent

On Wed, 2008-07-30 at 14:06 +0200, Miklos Szeredi wrote:
> From: Miklos Szeredi <[email protected]>
>
> Call security_inode_setattr() consistetly before inode_change_ok().
> It doesn't make sense to try to "optimize" the i_op->setattr == NULL
> case, as most filesystem do define their own setattr function.
>
> Signed-off-by: Miklos Szeredi <[email protected]>

Unfortunate since we'd prefer to have the DAC checks applied first, and
since inode_change_ok() may alter the ia_mode in response to those
checks, but it does seem inconsistent at present and it doesn't yield
any difference in the ltp selinux testsuite results.

Acked-by: Stephen Smalley <[email protected]>

> ---
> fs/attr.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
> Index: linux-2.6/fs/attr.c
> ===================================================================
> --- linux-2.6.orig/fs/attr.c 2008-07-30 13:52:35.000000000 +0200
> +++ linux-2.6/fs/attr.c 2008-07-30 13:52:41.000000000 +0200
> @@ -159,17 +159,17 @@ int notify_change(struct dentry * dentry
> if (!(attr->ia_valid & ~(ATTR_KILL_SUID | ATTR_KILL_SGID)))
> return 0;
>
> + error = security_inode_setattr(dentry, attr);
> + if (error)
> + return error;
> +
> if (ia_valid & ATTR_SIZE)
> down_write(&dentry->d_inode->i_alloc_sem);
>
> if (inode->i_op && inode->i_op->setattr) {
> - error = security_inode_setattr(dentry, attr);
> - if (!error)
> - error = inode->i_op->setattr(dentry, attr);
> + error = inode->i_op->setattr(dentry, attr);
> } else {
> error = inode_change_ok(inode, attr);
> - if (!error)
> - error = security_inode_setattr(dentry, attr);
> if (!error) {
> if ((ia_valid & ATTR_UID && attr->ia_uid != inode->i_uid) ||
> (ia_valid & ATTR_GID && attr->ia_gid != inode->i_gid))
--
Stephen Smalley
National Security Agency

2008-07-30 21:53:25

by James Morris

Subject: Re: [patch] vfs: make security_inode_setattr() calling consistent

On Wed, 30 Jul 2008, Stephen Smalley wrote:

> Unfortunate since we'd prefer to have the DAC checks applied first, and
> since inode_change_ok() may alter the ia_mode in response to those
> checks, but it does seem inconsistent at present and it doesn't yield
> any difference in the ltp selinux testsuite results.

No LSM seems to be using ia_mode, so that aspect is ok. The DAC checks
are generally not being applied first currently, so common case behaviour
doesn't change with this patch.

Acked-by: James Morris <[email protected]>

--
James Morris
<[email protected]>