Re: [PATCH v2 1/2] of: overlay: Fix memory leak in of_overlay_apply() error path

[Geert Uytterhoeven]

Hi Frank,

On Tue, Dec 5, 2017 at 3:07 AM, Frank Rowand <frowand.list@xxxxxxxxx> wrote:
> On 12/04/17 10:47, Geert Uytterhoeven wrote:
>> If of_resolve_phandles() fails, free_overlay_changeset() is called in
>> the error path.  However, that function returns early if the list hasn't
>> been initialized yet, before freeing the object.
>>
>> Explicitly calling kfree() instead would solve that issue. However, that
>> complicates matter, by having to consider which of two different methods
>> to use to dispose of the same object.
>>
>> Hence make free_overlay_changeset() consider initialization state of the
>> different parts of the object, making it always safe to call (once!) to
>> dispose of a (partially) initialized overlay_changeset:
>>   - Only destroy the changeset if the list was initialized,
>>   - Ignore uninitialized IDs (zero).
>>
>> Reported-by: Colin King <colin.king@xxxxxxxxxxxxx>
>> Fixes: f948d6d8b792bb90 ("of: overlay: avoid race condition between applying multiple overlays")
>> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
>> ---
>>  drivers/of/overlay.c | 7 +++----
>>  1 file changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/of/overlay.c b/drivers/of/overlay.c
>> index 3b7a3980ff50d6bf..312cd658bec0083b 100644
>> --- a/drivers/of/overlay.c
>> +++ b/drivers/of/overlay.c
>> @@ -630,11 +630,10 @@ static void free_overlay_changeset(struct overlay_changeset *ovcs)
>>  {
>>       int i;
>>
>> -     if (!ovcs->cset.entries.next)
>> -             return;
>> -     of_changeset_destroy(&ovcs->cset);
>> +     if (ovcs->cset.entries.next)
>> +             of_changeset_destroy(&ovcs->cset);
>>
>
> OK
>
>> -     if (ovcs->id)
>> +     if (ovcs->id > 0)
>
> Instead of this change, could you please make a change in init_overlay_changeset()?
>
> Current init_overlay_changeset():
>
>         ovcs->id = idr_alloc(&ovcs_idr, ovcs, 1, 0, GFP_KERNEL);
>         if (ovcs->id <= 0)
>                 return ovcs->id;
>
> My proposed version:
>
>         ret = idr_alloc(&ovcs_idr, ovcs, 1, 0, GFP_KERNEL);
>         if (ret <= 0)
>                 return ret;
>         ovcs->id = ret;

Sure.

>>               idr_remove(&ovcs_idr, ovcs->id);
>>
>>       for (i = 0; i < ovcs->count; i++) {
>>
>
> Also, the previous version of the patch, and the discussion around the resulting
> bug make me think that I should not have moved 'kfree(ovcs)' into
> free_overlay_changeset(), because that kfree is then not very visible in the
> error path of of_overlay_apply().  Could you remove 'kfree(ovcs)' from
> free_overlay_changeset(), and instead call it immediately after each call
> to free_overlay_changeset()?

Actually I like that free_overlay_changeset() takes care of the deallocation,
especially in light of the kojectification op top from bbb-overlays, which
means you cannot just call kfree(ovcs) anymore (I know this won't go upstream
anytime soon, but I need overlay configfs for my development and testing).

Perhaps the allocation of ovcs should be moved into free_overlay_changeset(),
and the latter being renamed to alloc_overlay_changeset()?
That way allocation and freeing become symmetrical.
It would move the allocation under the mutexes, though.

What do you think?

Thanks!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds