linux-kernel - Re: [PATCH] iio: accel: bmc150: Do not configure IRQ registers if no IRQ connected

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <07d91b36-dbeb-42b3-8dd7-b0771df9b306@mailbox.org>
Date: Sat, 21 Jun 2025 19:24:22 +0200
From: Marek Vasut <marek.vasut@...lbox.org>
To: Jonathan Cameron <jic23@...nel.org>, Hans de Goede <hansg@...nel.org>
Cc: Marek Vasut <marek.vasut+bmc150@...lbox.org>, linux-iio@...r.kernel.org,
 Nuno Sá <nuno.sa@...log.com>,
 Andy Shevchenko <andy@...nel.org>, David Lechner <dlechner@...libre.com>,
 Julien Stephan <jstephan@...libre.com>, Peter Zijlstra
 <peterz@...radead.org>, Salvatore Bonaccorso <carnil@...ian.org>,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH] iio: accel: bmc150: Do not configure IRQ registers if no
 IRQ connected

On 6/21/25 7:17 PM, Jonathan Cameron wrote:
> On Mon, 16 Jun 2025 14:42:54 +0200
> Hans de Goede <hansg@...nel.org> wrote:
> 
>> Hi,
>>
>> On 13-Jun-25 14:45, Marek Vasut wrote:
>>> The BMC150 on Onemix 2S does not have IRQ line described in ACPI tables,
>>> which leads to bmc150_accel_core_probe() being called with irq=0, which
>>> leads to bmc150_accel_interrupts_setup() never being called, which leads
>>> to struct bmc150_accel_data *data ->interrupts[i].info being left unset
>>> to NULL. Later, userspace can indirectly trigger bmc150_accel_set_interrupt()
>>> which depends on struct bmc150_accel_data *data ->interrupts[i].info being
>>> non-NULL, and which triggers NULL pointer dereference. This is triggered
>>> e.g. from iio-sensor-proxy.
>>>
>>> Fix this by skipping the IRQ register configuration in case there is no
>>> IRQ connected in hardware, in a manner similar to what the driver did in
>>> the very first commit which added the driver.
>>
>> ...
>>
>>> Fixes: 8e22f477e143 ("iio: bmc150: refactor interrupt enabling")
>>> Signed-off-by: Marek Vasut <marek.vasut+bmc150@...lbox.org>
>>> ---
>>> Cc: "Nuno Sá" <nuno.sa@...log.com>
>>> Cc: Andy Shevchenko <andy@...nel.org>
>>> Cc: David Lechner <dlechner@...libre.com>
>>> Cc: Jonathan Cameron <jic23@...nel.org>
>>> Cc: Julien Stephan <jstephan@...libre.com>
>>> Cc: Peter Zijlstra <peterz@...radead.org>
>>> Cc: Salvatore Bonaccorso <carnil@...ian.org>
>>> Cc: linux-iio@...r.kernel.org
>>> Cc: linux-kernel@...r.kernel.org
>>> ---
>>>   drivers/iio/accel/bmc150-accel-core.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/iio/accel/bmc150-accel-core.c b/drivers/iio/accel/bmc150-accel-core.c
>>> index 744a034bb8b5..1c3583ade2b4 100644
>>> --- a/drivers/iio/accel/bmc150-accel-core.c
>>> +++ b/drivers/iio/accel/bmc150-accel-core.c
>>> @@ -550,6 +550,9 @@ static int bmc150_accel_set_interrupt(struct bmc150_accel_data *data, int i,
>>>   	if (ret < 0)
>>>   		return ret;
>>>   
>>> +	if (!info)
>>> +		return 0;
>>> +
>>>   	/* map the interrupt to the appropriate pins */
>>>   	ret = regmap_update_bits(data->regmap, info->map_reg, info->map_bitmask,
>>>   				 (state ? info->map_bitmask : 0));
>>
>> AFAIK the proper fix would be to not register any IIO-triggers. This fix will
>> avoid the problem, but userspace might still try to use non-working triggers
>> which will now silently fail.
>>
>> I'm not an IIO expert, but IIRC other drivers simply skip registering their triggers
>> when there is no interrupt support.
> 
> Absolutely. It is annoyingly common for devices to have some or none of the interrupt
> lines actually wired so drivers should not present the interfaces if they aren't.
> It is acceptable for a new driver to just fail to probe if handling the device with no
> interrupts is particularly complex but in general at least some functionality tends
> to be easy to implement so we prefer that.
I haven't gotten to this again ... yet ... I can try and add some sort 
of polling fallback maybe ?