[v3] kernfs: fix memory leak of kernfs_iattrs in __kernfs_new_node

[PATCH v3] kernfs: fix memory leak of kernfs_iattrs in __kernfs_new_node

Posted by Will Rosenberg 1 week ago

There exists a memory leak of kernfs_iattrs contained as an element
of kernfs_node allocated in __kernfs_new_node(). __kernfs_setattr()
allocates kernfs_iattrs as a sub-object, and the LSM security check
incorrectly errors out and does not free the kernfs_iattrs sub-object.

Make an additional error out case that properly frees kernfs_iattrs if
security_kernfs_init_security() fails.

Fixes: e19dfdc83b60 ("kernfs: initialize security of newly created nodes")
Co-developed-by: Oliver Rosenberg <olrose55@gmail.com>
Signed-off-by: Oliver Rosenberg <olrose55@gmail.com>
Signed-off-by: Will Rosenberg <whrosenb@asu.edu>
---

Notes:
    V1 -> V2: meant as a RESEND, but the commit message and notes were also made more succinct. Patch remained unchanged. v1 was not sent to LKML by mistake.
    V2 -> V3: Update Fixes tag in commit message. Patch remains unchanged.

 fs/kernfs/dir.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index a670ba3e565e..5a40bfee7055 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -675,11 +675,13 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
 	if (parent) {
 		ret = security_kernfs_init_security(parent, kn);
 		if (ret)
-			goto err_out3;
+			goto err_out4;
 	}
 
 	return kn;
 
+ err_out4:
+	kfree(kn->iattr);
  err_out3:
 	spin_lock(&root->kernfs_idr_lock);
 	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));

base-commit: dcb6fa37fd7bc9c3d2b066329b0d27dedf8becaa
-- 
2.34.1

Re: [PATCH v3] kernfs: fix memory leak of kernfs_iattrs in __kernfs_new_node

Posted by Paul Moore 1 week ago

On Mon, Nov 24, 2025 at 1:37 PM Will Rosenberg <whrosenb@asu.edu> wrote:
>
> There exists a memory leak of kernfs_iattrs contained as an element
> of kernfs_node allocated in __kernfs_new_node(). __kernfs_setattr()
> allocates kernfs_iattrs as a sub-object, and the LSM security check
> incorrectly errors out and does not free the kernfs_iattrs sub-object.
>
> Make an additional error out case that properly frees kernfs_iattrs if
> security_kernfs_init_security() fails.
>
> Fixes: e19dfdc83b60 ("kernfs: initialize security of newly created nodes")
> Co-developed-by: Oliver Rosenberg <olrose55@gmail.com>
> Signed-off-by: Oliver Rosenberg <olrose55@gmail.com>
> Signed-off-by: Will Rosenberg <whrosenb@asu.edu>
> ---
>
> Notes:
>     V1 -> V2: meant as a RESEND, but the commit message and notes were also made more succinct. Patch remained unchanged. v1 was not sent to LKML by mistake.
>     V2 -> V3: Update Fixes tag in commit message. Patch remains unchanged.
>
>  fs/kernfs/dir.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
> index a670ba3e565e..5a40bfee7055 100644
> --- a/fs/kernfs/dir.c
> +++ b/fs/kernfs/dir.c
> @@ -675,11 +675,13 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
>         if (parent) {
>                 ret = security_kernfs_init_security(parent, kn);
>                 if (ret)
> -                       goto err_out3;
> +                       goto err_out4;
>         }
>
>         return kn;
>
> + err_out4:
> +       kfree(kn->iattr);

Should this be kmem_cache_free(kernfs_iattrs_cache, kn->iattr)
instead?  I suspect it would also probably be a good idea to call
simple_xattrs_free() first.

I also noticed there is a reasonable amount of duplicated kernfs
cleanup code spread around fs/kernfs/dir.c, we could probably benefit
from some helper functions here, but that would be a follow-on patch;
keeping a small bugfix patch to ease stable tree backporting is a good
thing.

[HINT FOR THE FUTURE: for patches like this, look for a function that
normally does resource cleanup for the object and do what it does, in
this case kernfs_free_rcu() is a good example.]

>   err_out3:
>         spin_lock(&root->kernfs_idr_lock);
>         idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));
>
> base-commit: dcb6fa37fd7bc9c3d2b066329b0d27dedf8becaa
> --
> 2.34.1

-- 
paul-moore.com

Re: [PATCH v3] kernfs: fix memory leak of kernfs_iattrs in __kernfs_new_node

Posted by Will Rosenberg 6 days, 8 hours ago

On Mon, Nov 24, 2025 at 03:41:51PM -0500, Paul Moore wrote:
>
> Should this be kmem_cache_free(kernfs_iattrs_cache, kn->iattr)
> instead?  I suspect it would also probably be a good idea to call
> simple_xattrs_free() first.
>
> I also noticed there is a reasonable amount of duplicated kernfs
> cleanup code spread around fs/kernfs/dir.c, we could probably benefit
> from some helper functions here, but that would be a follow-on patch;
> keeping a small bugfix patch to ease stable tree backporting is a good
> thing.

I agree. Helper functions could also help prevent this type of bug from
re-emerging.

> [HINT FOR THE FUTURE: for patches like this, look for a function that
> normally does resource cleanup for the object and do what it does, in
> this case kernfs_free_rcu() is a good example.]

Thank you for all the help. I updated kfree to kmem_cache_free() and
added simple_xattrs_free() just to be safe for future changes.

--
will rosenberg

[PATCH v4] kernfs: fix memory leak of kernfs_iattrs in __kernfs_new_node

Posted by Will Rosenberg 6 days, 8 hours ago

There exists a memory leak of kernfs_iattrs contained as an element
of kernfs_node allocated in __kernfs_new_node(). __kernfs_setattr()
allocates kernfs_iattrs as a sub-object, and the LSM security check
incorrectly errors out and does not free the kernfs_iattrs sub-object.

Make an additional error out case that properly frees kernfs_iattrs if
security_kernfs_init_security() fails.

Fixes: e19dfdc83b60 ("kernfs: initialize security of newly created nodes")
Co-developed-by: Oliver Rosenberg <olrose55@gmail.com>
Signed-off-by: Oliver Rosenberg <olrose55@gmail.com>
Signed-off-by: Will Rosenberg <whrosenb@asu.edu>
---

Notes:
    V1 -> V2: meant as a RESEND, but the commit message and notes were also made more succinct. Patch remained unchanged. v1 was not sent to LKML by mistake.
    V2 -> V3: Update Fixes tag in commit message. Patch remains unchanged.
    V3 -> V4: Use kmem_cache_free() in place of kfree(). To be safe and consistent, call simple_xattrs_free() in case simple_xattr is allocated in the future.

 fs/kernfs/dir.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index a670ba3e565e..5c0efd6b239f 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -675,11 +675,14 @@ static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
 	if (parent) {
 		ret = security_kernfs_init_security(parent, kn);
 		if (ret)
-			goto err_out3;
+			goto err_out4;
 	}
 
 	return kn;
 
+ err_out4:
+	simple_xattrs_free(&kn->iattr->xattrs, NULL);
+	kmem_cache_free(kernfs_iattrs_cache, kn->iattr);
  err_out3:
 	spin_lock(&root->kernfs_idr_lock);
 	idr_remove(&root->ino_idr, (u32)kernfs_ino(kn));

base-commit: dcb6fa37fd7bc9c3d2b066329b0d27dedf8becaa
-- 
2.34.1