Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.32 as permitted sender) client-ip=23.128.96.32;
Message-ID: <3ea3d92db5c4c077a76b29dc5a89c4d491695752.camel@gmail.com>
Subject: Re: [PATCH] iio: triggered-buffer: prevent possible freeing of
 wrong buffer
From:   Nuno =?ISO-8859-1?Q?S=E1?= <noname.nuno@gmail.com>
To:     David Lechner <dlechner@baylibre.com>,
        Jonathan Cameron <jic23@kernel.org>
Cc:     linux-iio@vger.kernel.org, linux-kernel@vger.kernel.org
Date:   Thu, 02 Nov 2023 10:02:15 +0100
In-Reply-To: <20231031210521.1661552-1-dlechner@baylibre.com>
References: <20231031210521.1661552-1-dlechner@baylibre.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.50.1 
MIME-Version: 1.0
Precedence: bulk

On Tue, 2023-10-31 at 16:05 -0500, David Lechner wrote:
> Commit ee708e6baacd ("iio: buffer: introduce support for attaching more
> IIO buffers") introduced support for multiple buffers per indio_dev but
> left indio_dev->buffer for a few legacy use cases.
>=20
> In the case of the triggered buffer, iio_triggered_buffer_cleanup()
> still assumes that indio_dev->buffer points to the buffer allocated by
> iio_triggered_buffer_setup_ext(). However, since
> iio_triggered_buffer_setup_ext() now calls iio_device_attach_buffer()
> to attach the buffer, indio_dev->buffer will only point to the buffer
> allocated by iio_device_attach_buffer() if it the first buffer attached.
>=20
> This adds a check to make sure that no other buffer has been attached
> yet to ensure that indio_dev->buffer will be assigned when
> iio_device_attach_buffer() is called.
>=20
> Fixes: ee708e6baacd ("iio: buffer: introduce support for attaching more I=
IO
> buffers")
> Signed-off-by: David Lechner <dlechner@baylibre.com>
> ---
> =C2=A0drivers/iio/buffer/industrialio-triggered-buffer.c | 10 ++++++++++
> =C2=A01 file changed, 10 insertions(+)
>=20
> diff --git a/drivers/iio/buffer/industrialio-triggered-buffer.c
> b/drivers/iio/buffer/industrialio-triggered-buffer.c
> index c7671b1f5ead..c06515987e7a 100644
> --- a/drivers/iio/buffer/industrialio-triggered-buffer.c
> +++ b/drivers/iio/buffer/industrialio-triggered-buffer.c
> @@ -46,6 +46,16 @@ int iio_triggered_buffer_setup_ext(struct iio_dev
> *indio_dev,
> =C2=A0	struct iio_buffer *buffer;
> =C2=A0	int ret;
> =C2=A0
> +	/*
> +	 * iio_triggered_buffer_cleanup() assumes that the buffer allocated
> here
> +	 * is assigned to indio_dev->buffer but this is only the case if this
> +	 * function is the first caller to iio_device_attach_buffer(). If
> +	 * indio_dev->buffer is already set then we can't proceed otherwise
> the
> +	 * cleanup function will try to free a buffer that was not allocated
> here.
> +	 */
> +	if (indio_dev->buffer)
> +		return -EADDRINUSE;
> +

Hmmm, good catch! But I think this is just workarounding the real problem
because like this, you can only have a triggered buffer by device. This sho=
uld
be fine as we don't really have any multi buffer user so far but ideally it
should be possible.

Long term we might want to think about moving 'pollfunc' to be a per buffer
thing. Not sure how much trouble that would be given that a trigger is also=
 per
device and I don't know if it would make sense to have a trigger per buffer=
?!
Ideally, given the multi buffer concept, I would say it makes sense but it =
might
be difficult to accomplish. So better to think about it only if there's a r=
eal
usecase for it.=C2=A0

On thing that I guess it could be done is to change the triggered API so it
returns a buffer and so iio_triggered_buffer_cleanup() would also get a poi=
nter
to the buffer it allocated (similar to what DMA buffer's are doing). But th=
at's
indeed also bigger change... Bahh, I'm likely over complicating things for =
now.
Fell free to:

Acked-by: Nuno Sa <nuno.sa@analog.com>