Cmt: [SRU][F][PATCH v2 0/3] CVE-2024-40967

[Magali Lemes]

On 14/11/2024 20:44, Yuxuan Luo wrote:
> v2: cherry picked prerequisites instead of backport in a wrong way.
> 
> [Impact]
> A potential deadlock might occur if too much time were spent for serial
> IMX subsystem, leading to denial of service.
> 
> [Backport]
> There is one conflict that needs to be solved manually when backporting
> the fix commit:
> 
> 1. Declaration for usr2 variable: it is introduced by 18ee37e1382a
>     (“serial: drop irq-flags initialisations”) and merely a change of the
>     order of declaration of variables, apply it manually.
> 
> Besides, read_poll_timeout_atomic macro is not in the tree yet, include
> it by cherry picking the following two prerequisites:
> 1. 57a29df34146 (“iopoll: Introduce read_poll_timeout_atomic macro”)
> 2. 5f5323a14cad (“iopoll: introduce read_poll_timeout macro”)
> 
> [Test]
> Compile and boot tested only due to lack of arm64 hardware.
> 
> [Where problems could occur]
> Regression might occur on all IMX CPU users.
> 
> Dejin Zheng (1):
>    iopoll: introduce read_poll_timeout macro
> 

I guess we don't strictly need this commit, right? As the fix commit 
only uses read_poll_timeout_atomic, and read_poll_timeout_atomic also 
doesn't have any dependency on read_poll_timeout. Any particular reason 
for also including this commit?

> Esben Haabendal (1):
>    serial: imx: Introduce timeout when waiting on transmitter empty
> 
> Kai-Heng Feng (1):
>    iopoll: Introduce read_poll_timeout_atomic macro
> 
>   drivers/tty/serial/imx.c |  7 +--
>   include/linux/iopoll.h   | 92 ++++++++++++++++++++++++++++++++++------
>   2 files changed, 84 insertions(+), 15 deletions(-)
>