linux-kernel - A potential data race in drivers/media/platform/s5p-mfc/

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>] [day] [month] [year] [list]

Date:   Tue, 23 Mar 2021 09:37:24 +0300
From:   Pavel Andrianov <andrianov@...ras.ru>
To:     Andrzej Hajda <a.hajda@...sung.com>
Cc:     Mauro Carvalho Chehab <mchehab@...nel.org>,
        linux-arm-kernel@...ts.infradead.org, linux-media@...r.kernel.org,
        linux-kernel@...r.kernel.org, ldv-project@...uxtesting.org
Subject: A potential data race in drivers/media/platform/s5p-mfc/

Hi,

s5p_mfc_probe[1] registers an interrupt handler s5p_mfc_irq before compete initialization. For example, the interrupt handler operates
with mfc_ops, which are set up in [2]. So, potentially, the interrupt handler may be executed in parallel with initialization. The question is if the device can produce the interrupts. Its registers are initialized in [3] and there are nothing like "enabling interrupts". So, likely, they are activated. And if interrupts can come, then this is a data race.

Best regards,
Pavel Andrianov
Linux Verification Center, ISPRAS
web:http://linuxtesting.org

[1]https://elixir.bootlin.com/linux/v5.4.106/source/drivers/media/platform/s5p-mfc/s5p_mfc.c#L1299
[2]https://elixir.bootlin.com/linux/v5.4.106/source/drivers/media/platform/s5p-mfc/s5p_mfc_opr.c#L19
[3]https://elixir.bootlin.com/linux/v5.4.106/source/drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c#L2229