[PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[David Howells]

nfsd_open() gets an unrefcounted pointer to the current process's effective
credentials at the top of the function, then calls nfsd_setuser() via
fh_verify() - which may replace and destroy the current process's effective
credentials - and then passes the unrefcounted pointer to dentry_open() - but
the credentials may have been destroyed by this point.

Instead, the value from current_cred() should be passed directly to
dentry_open() as one of its arguments, rather than being cached in a variable.

Possibly fh_verify() should return the creds to use.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/nfsd/vfs.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)


diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 4145083..23341c1 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -678,7 +678,6 @@ __be32
 nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 			int access, struct file **filp)
 {
-	const struct cred *cred = current_cred();
 	struct dentry	*dentry;
 	struct inode	*inode;
 	int		flags = O_RDONLY|O_LARGEFILE;
@@ -733,7 +732,7 @@ nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 		vfs_dq_init(inode);
 	}
 	*filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
-			    flags, cred);
+			    flags, current_cred());
 	if (IS_ERR(*filp))
 		host_err = PTR_ERR(*filp);
 	else

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[Steve Dickson]

On 07/02/2009 09:35 AM, David Howells wrote:
> nfsd_open() gets an unrefcounted pointer to the current process's effective
> credentials at the top of the function, then calls nfsd_setuser() via
> fh_verify() - which may replace and destroy the current process's effective
> credentials - and then passes the unrefcounted pointer to dentry_open() - but
> the credentials may have been destroyed by this point.
> 
> Instead, the value from current_cred() should be passed directly to
> dentry_open() as one of its arguments, rather than being cached in a variable.
> 
> Possibly fh_verify() should return the creds to use.
> 
> Signed-off-by: David Howells <dhowells@redhat.com>

Tested-and-Verified-By: Steve Dickson <steved@redhat.com>

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[J. Bruce Fields]

On Thu, Jul 02, 2009 at 02:35:32PM +0100, David Howells wrote:
> nfsd_open() gets an unrefcounted pointer to the current process's effective
> credentials at the top of the function, then calls nfsd_setuser() via
> fh_verify() - which may replace and destroy the current process's effective
> credentials - and then passes the unrefcounted pointer to dentry_open() - but
> the credentials may have been destroyed by this point.
> 
> Instead, the value from current_cred() should be passed directly to
> dentry_open() as one of its arguments, rather than being cached in a variable.

Got it, thanks.  How easy is it to hit this case?  (Just wondering
whether to submit for stable.)

--b.

> 
> Possibly fh_verify() should return the creds to use.
> 
> Signed-off-by: David Howells <dhowells@redhat.com>
> ---
> 
>  fs/nfsd/vfs.c |    3 +--
>  1 files changed, 1 insertions(+), 2 deletions(-)
> 
> 
> diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
> index 4145083..23341c1 100644
> --- a/fs/nfsd/vfs.c
> +++ b/fs/nfsd/vfs.c
> @@ -678,7 +678,6 @@ __be32
>  nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
>  			int access, struct file **filp)
>  {
> -	const struct cred *cred = current_cred();
>  	struct dentry	*dentry;
>  	struct inode	*inode;
>  	int		flags = O_RDONLY|O_LARGEFILE;
> @@ -733,7 +732,7 @@ nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
>  		vfs_dq_init(inode);
>  	}
>  	*filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
> -			    flags, cred);
> +			    flags, current_cred());
>  	if (IS_ERR(*filp))
>  		host_err = PTR_ERR(*filp);
>  	else
> 

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[David Howells]

J. Bruce Fields <bfields@fieldses.org> wrote:

> Got it, thanks.  How easy is it to hit this case?  (Just wondering
> whether to submit for stable.)

Actually, one of the problems with it is that the file is opened using the old
creds, cached across the call to nfsd_setuser().

David

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[J. Bruce Fields]

On Thu, Jul 02, 2009 at 10:31:33PM +0100, David Howells wrote:
> J. Bruce Fields <bfields@fieldses.org> wrote:
> 
> > Got it, thanks.  How easy is it to hit this case?  (Just wondering
> > whether to submit for stable.)
> 
> Actually, one of the problems with it is that the file is opened using the old
> creds, cached across the call to nfsd_setuser().

Oops, got it.  That answers that question....

--b.

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[J. Bruce Fields]

On Thu, Jul 02, 2009 at 07:09:55PM -0400, bfields wrote:
> On Thu, Jul 02, 2009 at 10:31:33PM +0100, David Howells wrote:
> > J. Bruce Fields <bfields@fieldses.org> wrote:
> > 
> > > Got it, thanks.  How easy is it to hit this case?  (Just wondering
> > > whether to submit for stable.)
> > 
> > Actually, one of the problems with it is that the file is opened using the old
> > creds, cached across the call to nfsd_setuser().
> 
> Oops, got it.  That answers that question....

Also: the problem was introduced by
745ca2475a6ac596e3d8d37c2759c0fbe2586227, right?  Or does it not
actually manifest till a little later?

--b.

Re: [PATCH] NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[David Howells]

J. Bruce Fields <bfields@fieldses.org> wrote:

> Also: the problem was introduced by
> 745ca2475a6ac596e3d8d37c2759c0fbe2586227, right?

Yes.

David

NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()

[J. Bruce Fields]

The following patch is appropriate for 2.6.31 (and stable), and is also
available from

  git://linux-nfs.org/~bfields/linux.git for-2.6.31

--b.

commit 033a666ccb842ab4134fcd0c861d5ba9f5d6bf3a
Author: David Howells <dhowells@redhat.com>
Date:   Thu Jul 2 14:35:32 2009 +0100

    NFSD: Don't hold unrefcounted creds over call to nfsd_setuser()
    
    nfsd_open() gets an unrefcounted pointer to the current process's effective
    credentials at the top of the function, then calls nfsd_setuser() via
    fh_verify() - which may replace and destroy the current process's effective
    credentials - and then passes the unrefcounted pointer to dentry_open() - but
    the credentials may have been destroyed by this point.
    
    Instead, the value from current_cred() should be passed directly to
    dentry_open() as one of its arguments, rather than being cached in a variable.
    
    Possibly fh_verify() should return the creds to use.
    
    This is a regression introduced by
    745ca2475a6ac596e3d8d37c2759c0fbe2586227 "CRED: Pass credentials through
    dentry_open()".
    
    Signed-off-by: David Howells <dhowells@redhat.com>
    Tested-and-Verified-By: Steve Dickson <steved@redhat.com>
    Cc: stable@kernel.org
    Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 4145083..23341c1 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -678,7 +678,6 @@ __be32
 nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 			int access, struct file **filp)
 {
-	const struct cred *cred = current_cred();
 	struct dentry	*dentry;
 	struct inode	*inode;
 	int		flags = O_RDONLY|O_LARGEFILE;
@@ -733,7 +732,7 @@ nfsd_open(struct svc_rqst *rqstp, struct svc_fh *fhp, int type,
 		vfs_dq_init(inode);
 	}
 	*filp = dentry_open(dget(dentry), mntget(fhp->fh_export->ex_path.mnt),
-			    flags, cred);
+			    flags, current_cred());
 	if (IS_ERR(*filp))
 		host_err = PTR_ERR(*filp);
 	else