On Mon, Dec 13, 2021 at 11:17:35AM -0800, David Vernet wrote:
> When enabling a KLP patch with `klp_enable_patch`, we invoke
> `klp_init_patch_early` to initialize the kobjects for the patch itself, as
> well as the `struct klp_object*`'s and `struct klp_func*`'s that comprise
> it. However, there are some paths where we may fail to do an
> early-initialization of an object or its functions if certain conditions
> are not met, such as an object having a `NULL` funcs pointer. In these
> paths, we may currently leak the `struct klp_patch*`'s kobject, as well as
> any of its objects or functions, as we don't free the patch in
> `klp_enable_patch` if `klp_init_patch_early` returns an error code. For
> example, if we added the following object entry to the sample livepatch
> code, it would cause us to leak the vmlinux `klp_object`, and its `struct
> klp_func` which updates `cmdline_proc_show`:
>
> ```
> static struct klp_object objs[] = {
> {
> .name = "kvm",
> }, { }
> };
> ```
>
> Without this change, if we enable `CONFIG_DEBUG_KOBJECT` and try to `kpatch
> load livepatch-sample.ko`, we don't observe the kobjects being released
> (though of course we do observe `insmod` failing to insert the module).
> With the change, we do observe that the `kobject` for the patch and its
> `vmlinux` object are released.
>
> Signed-off-by: David Vernet <void@xxxxxxxxxxxxx>
Thanks for reporting the issue and submitting the patch!
The patch description needs a few tweaks. In the kernel we don't use
Markdown for patch descriptions.
A function can be postfixed with a trailing pair of parentheses, like
klp_enable_patch().
Other symbols can be enclosed with single quotes, like 'struct
klp_object'.
I'd also recommend avoiding the excessive use of "we", in favor of more
imperative-type language.
See Documentation/process/submitting-patches.rst for more details. It's
also a good idea to look at some kernel commit logs to get a general
idea of the kernel patch description style.
> @@ -1052,10 +1052,7 @@ int klp_enable_patch(struct klp_patch *patch)
> }
>
> ret = klp_init_patch_early(patch);
> - if (ret) {
> - mutex_unlock(&klp_mutex);
> - return ret;
> - }
> + goto err;
>
> ret = klp_init_patch(patch);
> if (ret)
I don't think the fix will be quite that simple. For example, if
klp_init_patch_early() fails, that means try_module_get() hasn't been
done, so klp_free_patch_finish() will wrongly do a module_put().
--
Josh
