On 2016-11-18 13:26, Agustin Benito Bethencourt wrote:

Hi,

as you probably know, the more features we backport, the higher will be
the maintenance cost overtime so as a strategy, we need to be very
conservative.

I sent a mail some days ago about the features backported already by
Linaro for their LSK (Linaro Stable Kernel) for your evaluation. I bring
today some potential backports related with security features and
hardware support that has been identified by Ben H.

* UBSAN support:
<https://git.kernel.org/linus/c6d308534aef6c99904bf5862066360ae067abc4>

* Increased user-space ASLR range for ARM:
<https://lwn.net/Articles/667790/>

* Page poisoning on free:
<https://git.kernel.org/linus/8823b1dbc05fab1a8bec275eeae4709257c2661d>,
<https://git.kernel.org/linus/1414c7f4f7d72d138fff35f00151d15749b5beda>

* SLAB/SLUB freelist randomisation

* Hardened usercopy

* Do Members use SLUB? if not, we should take a look at KASAN support
for SLAB

* drm/tilcdc update? (many bug fixes post-4.4)

* AM33xx RTC support:

<https://git.kernel.org/linus/461932dfb54ebaf7da438fd8b769a01ce97a9360>,
<https://git.kernel.org/linus/b5a553c08bec14a058501df3fa6eb39f63f00a98>

To add two features areas from our requirement list:

- Distributed Switch Architecture, basically the level of 4.8 would be
sufficient. However, there might be too many dependencies on
networking core changes. If we pick 4.9 as next SLTS, then this
becomes obsolete.

- Graphic support for AM57xx from more recent kernels, but things may
even still depend on TI's vendor kernel (in which case it would be
too early).

Both are a bit vague yet, but I'm trying to clarify more details.
Comments are already welcome, of course.

Jan