RFE: Add Motorola 68000 support

[glaubitz]

This pull request adds support for the Motorola 68000 architecture to libseccomp.

Patches for seccomp support in the kernel have been posted on the LKML [1] but not merged yet. Plus, libseccomp currently fails to build from source after the syscall.csv file has been updated to include the syscalls for m68k. The problem seem to be a number of syscalls which exist exclusively on m68k.

For this reason, I'm marking this pull request as WIP.

[1] https://marc.info/?l=linux-m68k&m=167349572928947&w=2

[coveralls]

Coverage: 89.738% (+0.01%) from 89.724% when pulling a23ea38e5d63d38afa7d7864bc1d4234f94845f9 on glaubitz:m68k into 5aceb9d68a14e055c7a702483d612b537244e276 on seccomp:main.

[glaubitz]

I had to add to add the three syscalls atomic_barrier, atomic_cmpxchg_32 and getpagesize to include/seccomp-syscalls.h. Testsuite passes now.

[drakenclimber]

Thanks so much for the pull request, @glaubitz. Let us know when m68k support has been merged into the kernel, and I'll take a deeper look. :+1:

[pcmoore]

Hi @glaubitz, a quick question for you: are there any modern, common Linux distros which support m68k?

[glaubitz]

On Feb 6, 2023, at 11:36 PM, Paul Moore @.***> wrote: Hi @glaubitz, a quick question for you: are there any modern, common Linux distros which support m68k? Yes, Debian and Gentoo.The port is very actively maintained and has a vibrant hobbyist around it with even new hardware being developed such as the Pi-Storm and Apollo accelerator cards for the Amiga.

[pcmoore]

Great, thanks!

[glaubitz]

seccomp support for m68k was just merged today into Linus' tree:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/arch/m68k?id=6baaade15594b28195da369962208b1f658e7342

[drakenclimber]

@glaubitz, looks like we can now review these changes and start to work on getting them into libseccomp. Correct?

If so, would you mind updating your patchset to resolve the conflicts. (m86k collided with LoongArch, but the collisions look fairly trivial.) Once you do that, I'll look over the code. Thanks so much!

[glaubitz]

@glaubitz, looks like we can now review these changes and start to work on getting them into libseccomp. Correct?

Yes.

If so, would you mind updating your patchset to resolve the conflicts. (m86k collided with LoongArch, but the collisions look fairly trivial.) Once you do that, I'll look over the code. Thanks so much!

Rebase should be done now.

[glaubitz]

On Mar 31, 2023, at 9:52 PM, Paul Moore @.***> wrote: @pcmoore requested changes on this pull request.

In include/seccomp-syscalls.h:

@@ -824,6 +839,12 @@

define __SNR_getitimer __NR_getitimer

+#ifdef NR_atomic_getpagesize +#define SNR_getpagesize NR_getpagesize +#else +#define __SNR_getpagesize PNR_getpagesize

I would have expected the above code to look like this:

ifdef __NR_atomic_getpagesize

define SNR_atomic_getpagesize NR_atomic_getpagesize

else

define SNR_atomic_getpagesize PNR_atomic_getpagesize

... is this a typo or is there some m68k cleverness going on here? It’s a typo. Will fix that in a bit. Thanks a lot for catching this.

[glaubitz]

Linux 6.3 which includes seccomp support for m68k was just released today ;-).

[drakenclimber]

Linux 6.3 which includes seccomp support for m68k was just released today ;-).

Congrats! I have carved out time this week to review the changes. Thanks for your patience, @glaubitz

[drakenclimber]

Pushed to main in commits dd5c9c24e8ba and aa168d49243b. Thanks, @glaubitz