ACK/cmnt: [Cover Letter] Out of order reads can fetch a NULL pointer causing a kernel crash

Kleber Souza kleber.souza at canonical.com
Tue Apr 27 10:43:58 UTC 2021


On 26.04.21 23:41, Guilherme G. Piccoli wrote:
> BugLink: https://bugs.launchpad.net/bugs/1926184
> 
> 
> [NOTE]
> * We have reports of this issue happening in AWS instances, so we sent this
> with a priority "flag" for linux-aws - the tests (see below) were performed
> with -aws kernels. This will/should reach all Ubuntu kernels organically
> via the regular upstream stable process, but nothing prevent us to merge
> it now, from this submission.
> 
> [Impact]
> * Out of order reads can fetch a NULL pointer causing a kernel crash. Affects
> kernels from v4.15 to v5.11.
> 
> [Fix]
> * commit 84a24bf8c52e ("locking/qrwlock: Fix ordering in queued_write_lock_slowpath()")
> 
> * For Bionic, we also require an additional patch:
> commit fcfdfe30e324 ("locking/barriers: Introduce smp_cond_load_relaxed() and atomic_cond_read_relaxed()")
> 
> [Where problems could occur]
> 
> * This is quite a subtle bug. It is more prevalent on arm64. Regression
> possibility seems quite low. The worst impact could be a minor performance
> degradation.
> 
> * Kernels 4.15 / 5.4 / 5.8 / 5.11 were successfully built and booted with this
> patch. Also, by using ftrace we could see that the changed function was executed
> with success multiple times.
> 
> 

Hi Guilherme,

For the next submissions, could you please add to the cover letter the affected
kernels as well? Ideally it would also have the "[SRU]" tag as well. This makes
it easier to identify for which kernels the patches need to be applied.

Otherwise the patches look good.

Acked-by: Kleber Sacilotto de Souza <kleber.souza at canonical.com>

Thanks



More information about the kernel-team mailing list