Re: [PATCH 00/11] fix memory leak while kset_register() fails

From: Yang Yingliang
Date: Fri Oct 21 2022 - 05:13:50 EST



On 2022/10/21 16:36, Greg KH wrote:
On Fri, Oct 21, 2022 at 04:24:23PM +0800, Yang Yingliang wrote:
On 2022/10/21 13:37, Greg KH wrote:
On Fri, Oct 21, 2022 at 01:29:31AM -0400, Luben Tuikov wrote:
On 2022-10-20 22:20, Yang Yingliang wrote:
The previous discussion link:
https://lore.kernel.org/lkml/0db486eb-6927-927e-3629-958f8f211194@xxxxxxxxxx/T/
The very first discussion on this was here:

https://www.spinics.net/lists/dri-devel/msg368077.html

Please use this link, and not the that one up there you which quoted above,
and whose commit description is taken verbatim from the this link.

kset_register() is currently used in some places without calling
kset_put() in error path, because the callers think it should be
kset internal thing to do, but the driver core can not know what
caller doing with that memory at times. The memory could be freed
both in kset_put() and error path of caller, if it is called in
kset_register().
As I explained in the link above, the reason there's
a memory leak is that one cannot call kset_register() without
the kset->kobj.name being set--kobj_add_internal() returns -EINVAL,
in this case, i.e. kset_register() fails with -EINVAL.

Thus, the most common usage is something like this:

kobj_set_name(&kset->kobj, format, ...);
kset->kobj.kset = parent_kset;
kset->kobj.ktype = ktype;
res = kset_register(kset);

So, what is being leaked, is the memory allocated in kobj_set_name(),
by the common idiom shown above. This needs to be mentioned in
the documentation, at least, in case, in the future this is absolved
in kset_register() redesign, etc.
Based on this, can kset_register() just clean up from itself when an
error happens? Ideally that would be the case, as the odds of a kset
being embedded in a larger structure is probably slim, but we would have
to search the tree to make sure.
I have search the whole tree, the kset used in bus_register() - patch #3,
kset_create_and_add() - patch #4
__class_register() - patch #5,  fw_cfg_build_symlink() - patch #6 and
amdgpu_discovery.c - patch #10
is embedded in a larger structure. In these cases, we can not call
kset_put() in error path in kset_register()
Yes you can as the kobject in the kset should NOT be controling the
lifespan of those larger objects.
Read through the code the only leak in this case is the name, so can we free it
directly in kset_register():

--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -844,8 +844,11 @@ int kset_register(struct kset *k)

        kset_init(k);
        err = kobject_add_internal(&k->kobj);
-       if (err)
+       if (err) {
+               kfree_const(k->kobj.name);
+               k->kobj.name = NULL;
                return err;
+       }
        kobject_uevent(&k->kobj, KOBJ_ADD);
        return 0;
 }

or unset ktype of kobject, then call kset_put():

--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -844,8 +844,11 @@ int kset_register(struct kset *k)

        kset_init(k);
        err = kobject_add_internal(&k->kobj);
-       if (err)
+       if (err) {
+               k->kobj.ktype = NULL;
+               kset_put(k);
                return err;
+       }
        kobject_uevent(&k->kobj, KOBJ_ADD);
        return 0;
 }


If it is, please point out the call chain here as I don't think that
should be possible.

Note all of this is a mess because the kobject name stuff was added much
later, after the driver model had been created and running for a while.
We missed this error path when adding the dynamic kobject name logic,
thank for looking into this.

If you could test the patch posted with your error injection systems,
that could make this all much simpler to solve.

thanks,

greg k-h
.