Date:   Mon, 13 Dec 2021 19:13:10 -0800
From:   Josh Poimboeuf <jpoimboe@redhat.com>
To:     David Vernet <void@manifault.com>
Cc:     pmladek@suse.com, linux-doc@vger.kernel.org,
        live-patching@vger.kernel.org, linux-kernel@vger.kernel.org,
        jikos@kernel.org, mbenes@suse.cz, joe.lawrence@redhat.com,
        corbet@lwn.net, yhs@fb.com, songliubraving@fb.com
Subject: Re: [PATCH] livepatch: Fix leak on klp_init_patch_early failure path
Message-ID: <20211214031310.p6kmbvd73kn6j7x3@treble>
References: <20211213191734.3238783-1-void@manifault.com>
 <20211213201022.dhalhtc2bpey55gh@treble>
 <YbfQHjoUO5GTvImR@dev0025.ash9.facebook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <YbfQHjoUO5GTvImR@dev0025.ash9.facebook.com>
Precedence: bulk

On Mon, Dec 13, 2021 at 02:58:38PM -0800, David Vernet wrote:
> > 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().
> 
> Ugh, good point and thank you for catching that. Another problem with the
> current patch is that we'll call kobject_put() on the patch even if we
> never call kobject_init on the patch due to patch->objs being NULL.
> 
> Perhaps we should pull try_module_get() and the NULL check for patch->objs
> out of klp_init_patch_early()? It feels a bit more intuitive to me if
> klp_init_patch_early() were only be responsible for initializing kobjects
> for the patch and its objects / funcs anyways.
> 
> Testing it locally seems to work fine. Let me know if this sounds
> reasonable to you, and I'll send out a v2 patch with the fixes to both the
> patch description, and logic.

Sounds reasonable to me.  Thanks.

-- 
Josh