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;
To:     Florian Fainelli <f.fainelli@gmail.com>,
        Mark Brown <broonie@kernel.org>, Ray Jui <rjui@broadcom.com>,
        Scott Branden <sbranden@broadcom.com>,
        bcm-kernel-feedback-list@broadcom.com,
        Nicolas Saenz Julienne <nsaenzjulienne@suse.de>,
        linux-spi@vger.kernel.org, linux-rpi-kernel@lists.infradead.org,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
References: <20210420083402.6950-1-joe.burmeister@devtank.co.uk>
 <c087ba2c-7839-02d1-a522-b104d8ffb8d2@gmail.com>
From:   Joe Burmeister <joe.burmeister@devtank.co.uk>
Subject: Re: [PATCH] spi: bcm2835: Fix buffer overflow with CS able to go
 beyond limit.
Message-ID: <7c9f9376-1a80-b624-7b9e-0f6d04437c02@devtank.co.uk>
Date:   Thu, 22 Apr 2021 21:10:21 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.9.0
MIME-Version: 1.0
In-Reply-To: <c087ba2c-7839-02d1-a522-b104d8ffb8d2@gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Content-Language: en-GB
Precedence: bulk

> On 4/20/2021 1:34 AM, Joe Burmeister wrote:
>> It was previoulsy possible to have a device tree with more chips than
>> the driver supports and go off the end of CS arrays.
> Do you mind walking me through the code how that could have happened? W=
e
> have spi_register_controller() call of_spi_get_gpio_numbers() which has=

> the following:
>
> ctlr->num_chipselect =3D max_t(int, nb, ctlr->num_chipselect);
>
> such that what the controller has is the maximum between the number of
> 'cs-gpios' properties parsed and what was already populated in
> ctrl->num_chipselect during bcm2835_spi_probe(), which for this driver
> is BCM2835_SPI_NUM_CS (3).

If you make a initial device tree (or add overlay in the rpi's=20
config.txt) with more on the bus than BCM2835_SPI_NUM_CS (in my case 8
devices), you get into this trampling memory state. As the devices are
added, once the chip_select is equal to or greater than
BCM2835_SPI_NUM_CS, it's writing off the end of the arrays.

There is no protection from this happening. By the looks of it, this
isn't the only driver this could happen with, but it is the one I have
hardware for to test. There are also drivers that look like they don't
have a problem going well beyond the limit they gave.

There is protection in spi_add_device, which will catch extra added
later, but not ones in the device tree when the spi controller was
registered.

>> This patches inforces CS limit but sets that limit to the max of the
>> default limit and what is in the device tree when driver is loaded.
>>
>> Signed-off-by: Joe Burmeister <joe.burmeister@devtank.co.uk>
> You have changed many more things that just enforcing a limit on
> BCM2835_SPI_NUM_CS you have now made all chip-select related data
> structuresd dynamically allocated and you have changed a number of
> prints to use the shorthand "dev" instead of &pdev->dev.
The change to dynamic allocated arrays is just to support what is given
in the device=C2=A0 tree rather than increase and enforce the CS limit ju=
st
for my case.

The shorthand is of course not required. I'll drop it on resubmitting.

Regards,

Joe