Self introduction
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 1 line · AKASHI Takahiro
Hi Kees and all,
Hi! Sorry for not replying sooner -- this email seemed to strangely not make its way into the kernel-hardening list, even though it looks from the headers like you sent it there. Regardless, on to my late reply...
Quoted 6 lines
I'm looking for any tasks that I could contribute in kernel-hardening areas. As I'm working for Linaro, arm/arm64-related items (or arch-specific implementation of more generic features) would be better fitted, but I'm open to broader topics, including complementary tasks for other guys.
Awesome! Yeah, it'll be great to have you helping. Are there any areas you especially enjoy working on?
Quoted 2 lines
Do you have any suggestions about such items for me to start small (or can be even big :) with?
I think a great area would be looking at arm and arm64 support for catching atomic_t wrap-around. This kind of protection would have stopped many of the public attacks from last year, since accidents with refcounting on atomic_t tend to be a precursor to use-after-free vulnerabilities.
This idea has been implemented in PaX/Grsecurity for a while now as PAX_REFCOUNT. An earlier effort to extract it was started but he ran out of time to work on it. Recently the folks from Intel have been looking at the x86 side of this (and took on the burden of preparing the upstreaming of the marking of the expected-wrap uses in the kernel), but we'll still need other architectures to support this feature. There's an excellent write-up on the feature here:
https://forums.grsecurity.net/viewtopic.php?f=7&t=4173
There isn't yet an implementation for arm64 (though maybe it'll be identical/similar to the existing arm implementation).
I think Intel will be posting an RFC series soonish, which would likely be the right place to start from.
-Kees
Signature · 2 lines
Kees Cook Nexus Security
Hi,
Quoted 28 lines · Kees Cook
Quoted 1 line · AKASHI Takahiro
Hi Kees and all,
Hi! Sorry for not replying sooner -- this email seemed to strangely not make its way into the kernel-hardening list, even though it looks from the headers like you sent it there. Regardless, on to my late reply...
Quoted 6 lines
I'm looking for any tasks that I could contribute in kernel-hardening areas. As I'm working for Linaro, arm/arm64-related items (or arch-specific implementation of more generic features) would be better fitted, but I'm open to broader topics, including complementary tasks for other guys.
Awesome! Yeah, it'll be great to have you helping. Are there any areas you especially enjoy working on?
Quoted 2 lines
Do you have any suggestions about such items for me to start small (or can be even big :) with?
I think a great area would be looking at arm and arm64 support for catching
atomic_twrap-around. This kind of protection would have stopped many of the public attacks from last year, since accidents with refcounting onatomic_ttend to be a precursor to use-after-free vulnerabilities.
Quoted 3 lines
This idea has been implemented in PaX/Grsecurity for a while now as
PAX_REFCOUNT. An earlier effort to extract it was started but he ran out of time to work on it.
Sorry about not communicating better about this. I started working on v3 (porting to linux-next) a few days ago and have the code moved, but still have to work through a few compile issues.
If you're interested in helping, take a look:
https://github.com/dwindsor/linux-next/tree/strict-refcount-v3
I'll continue working through this, but help would certainly be appreciated.
Thanks,
David
Quoted 42 lines · David Windsor
Hi,
Quoted 29 lines · Kees Cook
Quoted 1 line · AKASHI Takahiro
Hi Kees and all,
Hi! Sorry for not replying sooner -- this email seemed to strangely not make its way into the kernel-hardening list, even though it looks from the headers like you sent it there. Regardless, on to my late reply...
Quoted 6 lines
I'm looking for any tasks that I could contribute in kernel-hardening areas. As I'm working for Linaro, arm/arm64-related items (or arch-specific implementation of more generic features) would be better fitted, but I'm open to broader topics, including complementary tasks for other guys.
Awesome! Yeah, it'll be great to have you helping. Are there any areas you especially enjoy working on?
Quoted 2 lines
Do you have any suggestions about such items for me to start small (or can be even big :) with?
I think a great area would be looking at arm and arm64 support for catching
atomic_twrap-around. This kind of protection would have stopped many of the public attacks from last year, since accidents with refcounting onatomic_ttend to be a precursor to use-after-free vulnerabilities.Quoted 3 lines
This idea has been implemented in PaX/Grsecurity for a while now as
PAX_REFCOUNT. An earlier effort to extract it was started but he ran out of time to work on it.Sorry about not communicating better about this. I started working on v3 (porting to linux-next) a few days ago and have the code moved, but still have to work through a few compile issues.
Quoted 5 lines
If you're interested in helping, take a look:
https://github.com/dwindsor/linux-next/tree/strict-refcount-v3
I'll continue working through this, but help would certainly be appreciated.
No worries, thanks for the heads-up! I think getting everyone on the same page will be good. We can all compare notes when Intel posts patches too.
-Kees
Signature · 2 lines
Kees Cook Nexus Security
Quoted 9 lines
Sorry about not communicating better about this. I started working on v3 (porting to linux-next) a few days ago and have the code moved, but still have to work through a few compile issues.
If you're interested in helping, take a look:
https://github.com/dwindsor/linux-next/tree/strict-refcount-v3
I'll continue working through this, but help would certainly be appreciated.
Hi David!
We have started to work on this about 3 weeks ago and currently have working code with additional tests. Sorry for not letting you know, I guess everyone thought that you are not working on this anymore. Just yesterday Hans figured out the proper explanantion for PERCPU_COUNT_BIAS change, so we will be posting series hopefully today-tomorrow after writing better commit messages.
Best Regards,
Elena.
Quoted 6 lines · Elena
Quoted 6 lines
Sorry about not communicating better about this. I started working on v3 (porting to linux-next) a few days ago and have the code moved, but still have to work through a few compile issues.
If you're interested in helping, take a look:
Quoted 8 lines
Quoted 3 lines
https://github.com/dwindsor/linux-next/tree/strict-refcount-v3
I'll continue working through this, but help would certainly be appreciated.
Hi David!
We have started to work on this about 3 weeks ago and currently have working code with additional tests. Sorry for not letting you know, I guess everyone thought that you are not working on this anymore. Just yesterday Hans figured out the proper explanantion for
PERCPU_COUNT_BIASchange, so we will be posting series hopefully today-tomorrow after writing better commit messages.
Great! Yeah, I think having verbose and clear commit messages will be an important part of this series.
-Kees
Signature · 2 lines
Kees Cook Nexus Security