| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <1241804120.4843.7.camel@dyn9002018117.watson.ibm.com>
Date: Fri, 08 May 2009 13:35:20 -0400
From: Mimi Zohar <zohar@...ux.vnet.ibm.com>
To: Andrew Morton <akpm@...ux-foundation.org>
Cc: linux-kernel@...r.kernel.org, hooanon05@...oo.co.jp,
jmorris@...ei.org, safford@...son.ibm.com,
"J. Bruce Fields" <bfields@...ldses.org>, linux-nfs@...r.kernel.org
Subject: Re: integrity: nfsd imbalance bug fix
On Wed, 2009-04-29 at 14:18 -0700, Andrew Morton wrote:
> On Tue, 28 Apr 2009 14:32:14 -0400
> Mimi Zohar <zohar@...ux.vnet.ibm.com> wrote:
>
> > The number of calls to ima_path_check()/ima_file_free()
> > should be balanced. An extra call to fput(), indicates
> > the file could have been accessed without first being
> > measured.
> >
> > An nfsd exported file is opened/closed by the kernel
> > causing an integrity imbalance message.
> >
> > - rename and export opencount_get to ima_opencount_get
> > - replace ima_shm_check calls with ima_opencount_get
> > - add call to increment opencount for files opened by nfsd.
> > - add call to measure exported files in nfsd_permission().
> > - export ima_path_check
> >
>
> The patch looks fragile to me. It sprinkles IMA-specific operations
> over random unrelated subsystems. There is hence a decent chance of
> breakage in the future.
>
> Suppose some other new piece of kernel code opens/closes a file. The
> developer didn't think to add the IMA hooks and whoops, we have a bug.
Agreed. Exactly for that reason we have the imbalance message. The
imbalance message is not a bug per se, but an indication that the
file hasn't been integrity (permission) checked.
> It would be really really better if we could add the IMA hooks in a
> single place. That might require the addition of a new function, and
> that's fine. d_instantiate_kernel() or init_file_kernel() or whatever
> - that's fine. It still has the risk that new code will forget to use
> the in-kernel variant, but we have a better chance of detecting it.
Definitely. Originally the integrity checking was in inode_permission().
But measuring a file requires a dentry and mount point.
I've updated the patch to fix a 'counts' error and divided the patch
into two, separating the IMA code changes from the IMA calls in nfsd.
> >
> > Index: security-testing-2.6/fs/nfsd/vfs.c
> > ===================================================================
> > --- security-testing-2.6.orig/fs/nfsd/vfs.c
> > +++ security-testing-2.6/fs/nfsd/vfs.c
> > @@ -55,6 +55,7 @@
> > #include <linux/security.h>
> > #endif /* CONFIG_NFSD_V4 */
> > #include <linux/jhash.h>
> > +#include <linux/ima.h>
> >
> > #include <asm/uaccess.h>
> >
> > @@ -735,6 +736,8 @@ nfsd_open(struct svc_rqst *rqstp, struct
> > flags, cred);
> > if (IS_ERR(*filp))
> > host_err = PTR_ERR(*filp);
> > + else
> > + ima_opencount_get(*filp);
>
> This suggests dentry_open_kernel().
Good name, but making dentry_open_kernel() the default behavior,
would mask the real bug - not doing integrity (permission) checking.
>
> > out_nfserr:
> > err = nfserrno(host_err);
> > out:
> > @@ -2096,7 +2099,13 @@ nfsd_permission(struct svc_rqst *rqstp,
> > if (err == -EACCES && S_ISREG(inode->i_mode) &&
> > acc == (NFSD_MAY_READ | NFSD_MAY_OWNER_OVERRIDE))
> > err = inode_permission(inode, MAY_EXEC);
> > + if (err)
> > + goto nfsd_out;
> >
> > + err = ima_path_check(&exp->ex_path,
> > + acc & (MAY_READ | MAY_WRITE | MAY_EXEC));
>
> hm, dunno what to do about that.
The real problem is that this belongs in inode_permission(), but an
inode is not enough. It requires a dentry and mount point to
measure the file.
> > + return err;
> > +nfsd_out:
> > return err? nfserrno(err) : 0;
> > }
> >
> > Index: security-testing-2.6/security/integrity/ima/ima_main.c
> > ===================================================================
> > --- security-testing-2.6.orig/security/integrity/ima/ima_main.c
> > +++ security-testing-2.6/security/integrity/ima/ima_main.c
> > @@ -206,6 +206,7 @@ out:
> > kref_put(&iint->refcount, iint_free);
> > return 0;
> > }
> > +EXPORT_SYMBOL_GPL(ima_path_check);
> >
> > static int process_measurement(struct file *file, const unsigned char *filename,
> > int mask, int function)
> > @@ -234,7 +235,16 @@ out:
> > return rc;
> > }
> >
> > -static void opencount_get(struct file *file)
> > +/*
> > + * ima_opencount_get - incr opencount for files opened by the kernel
> > + *
> > + * - IPC shm and shmat create/fput a file.
> > + * - nfsd opens/closes exported files.
> > + *
> > + * Increment the opencount for these files to prevent unnecessary
> > + * imbalance messages.
> > + */
> > +void ima_opencount_get(struct file *file)
> > {
> > struct inode *inode = file->f_dentry->d_inode;
> > struct ima_iint_cache *iint;
> > @@ -248,6 +258,7 @@ static void opencount_get(struct file *f
> > iint->opencount++;
> > mutex_unlock(&iint->mutex);
> > }
> > +EXPORT_SYMBOL_GPL(ima_opencount_get);
> >
> > /**
> > * ima_file_mmap - based on policy, collect/store measurement.
> > @@ -272,18 +283,6 @@ int ima_file_mmap(struct file *file, uns
> > return 0;
> > }
> >
> > -/*
> > - * ima_shm_check - IPC shm and shmat create/fput a file
> > - *
> > - * Maintain the opencount for these files to prevent unnecessary
> > - * imbalance messages.
> > - */
> > -void ima_shm_check(struct file *file)
> > -{
> > - opencount_get(file);
> > - return;
> > -}
> > -
> > /**
> > * ima_bprm_check - based on policy, collect/store measurement.
> > * @bprm: contains the linux_binprm structure
> > Index: security-testing-2.6/include/linux/ima.h
> > ===================================================================
> > --- security-testing-2.6.orig/include/linux/ima.h
> > +++ security-testing-2.6/include/linux/ima.h
> > @@ -20,7 +20,7 @@ extern void ima_inode_free(struct inode
> > extern int ima_path_check(struct path *path, int mask);
> > extern void ima_file_free(struct file *file);
> > extern int ima_file_mmap(struct file *file, unsigned long prot);
> > -extern void ima_shm_check(struct file *file);
> > +extern void ima_opencount_get(struct file *file);
> >
> > #else
> > static inline int ima_bprm_check(struct linux_binprm *bprm)
> > @@ -53,7 +53,7 @@ static inline int ima_file_mmap(struct f
> > return 0;
> > }
> >
> > -static inline void ima_shm_check(struct file *file)
> > +static inline void ima_opencount_get(struct file *file)
> > {
> > return;
> > }
> > Index: security-testing-2.6/ipc/shm.c
> > ===================================================================
> > --- security-testing-2.6.orig/ipc/shm.c
> > +++ security-testing-2.6/ipc/shm.c
> > @@ -384,7 +384,7 @@ static int newseg(struct ipc_namespace *
> > error = PTR_ERR(file);
> > if (IS_ERR(file))
> > goto no_file;
> > - ima_shm_check(file);
> > + ima_opencount_get(file);
> >
> > id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
> > if (id < 0) {
> > @@ -891,7 +891,7 @@ long do_shmat(int shmid, char __user *sh
> > file = alloc_file(path.mnt, path.dentry, f_mode, &shm_file_operations);
> > if (!file)
> > goto out_free;
> > - ima_shm_check(file);
> > + ima_opencount_get(file);
> >
> > file->private_data = sfd;
> > file->f_mapping = shp->shm_file->f_mapping;
> > Index: security-testing-2.6/mm/shmem.c
> > ===================================================================
> > --- security-testing-2.6.orig/mm/shmem.c
> > +++ security-testing-2.6/mm/shmem.c
> > @@ -2680,7 +2680,7 @@ int shmem_zero_setup(struct vm_area_stru
> > if (IS_ERR(file))
> > return PTR_ERR(file);
> >
> > - ima_shm_check(file);
> > + ima_opencount_get(file);
> > if (vma->vm_file)
> > fput(vma->vm_file);
> > vma->vm_file = file;
>
> Maybe do the IMA operations in (or under) shmem_file_setup() and
> hugetlb_file_setup()?
Again, that would hide the real issue of not having done integrity
(permission) checking.
Mimi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists