Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20220120103714.32108-1-linmq006@gmail.com>
In-Reply-To: <20220120103714.32108-1-linmq006@gmail.com>
From:   Jinpu Wang <jinpu.wang@ionos.com>
Date:   Thu, 20 Jan 2022 15:06:55 +0100
Message-ID: <CAMGffE=mzZVCg9hoiMRmjnnKqMb9cfC9YTcCkhHMo9BjHOr+Tg@mail.gmail.com>
Subject: Re: [PATCH] RDMA/rtrs: Fix double free in alloc_clt
To:     Miaoqian Lin <linmq006@gmail.com>
Cc:     "Md. Haris Iqbal" <haris.iqbal@ionos.com>,
        Jason Gunthorpe <jgg@ziepe.ca>,
        Danil Kipnis <danil.kipnis@ionos.com>,
        linux-rdma@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Jan 20, 2022 at 11:37 AM Miaoqian Lin <linmq006@gmail.com> wrote:
>
> Callback function rtrs_clt_dev_release() in put_device()
> calls kfree(clt); to free memory. We shouldn't call kfree(clt) again.
>
right, there is a bug, but the fix is not right.
> Fixes: 6a98d71daea1 ("RDMA/rtrs: client: main functionality")
> Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
> ---
>  drivers/infiniband/ulp/rtrs/rtrs-clt.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> index 7c3f98e57889..61723f48fbd4 100644
> --- a/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c
> @@ -2741,7 +2741,7 @@ static struct rtrs_clt_sess *alloc_clt(const char *sessname, size_t paths_num,
>         err = device_register(&clt->dev);
>         if (err) {
>                 put_device(&clt->dev);
> -               goto err;
> +               goto err_free_cpu;
>         }
>
>         clt->kobj_paths = kobject_create_and_add("paths", &clt->dev.kobj);
> @@ -2764,6 +2764,9 @@ static struct rtrs_clt_sess *alloc_clt(const char *sessname, size_t paths_num,
>  err:
>         free_percpu(clt->pcpu_path);
>         kfree(clt);
> +       clt->pcpu_path = NULL;
> +err_free_cpu:
> +       free_percpu(clt->pcpu_path);
this still lead to use after free, because clt could be free already.
the right fix should move the free_percpu() to the release call back
before kfree(clt);
>         return ERR_PTR(err);
>  }
>
> --
> 2.17.1
>