Allow setting a per-Jail fallback ABI brand.

[nyan_myuji.xyz]

Summary

Similar to kern.elfXX.fallback_brand sysctl, this allow setting a fallback ELF32/64
brand, but in a per-Jail basis. If the fallback brand in Jail does not work, it will attempt
to use the system-wise kern.elfXX.fallback_brand ABI.

Useful in running statically linked Penguin binaries with wrong ABI value in
their EFI header in a Jail, without brandelf (which modifies the file) nor
setting systemwise kern.elfXX.fallback_brand sysctl.

Event: Kitchener-Waterloo Hackathon 202305

[nyan_myuji.xyz]

• fix sysctl parametetr description

[freebsd_igalic.co]

please extend the man page as well, to mention this setting

[nyan_myuji.xyz]

Man page update

[lwhsu]

usr.sbin/jail/jail.8

28

Don't forget bump the date. :)

765

If you really want a blank line, you need .Pp

775

.Xr brandelf 1

[nyan_myuji.xyz]

• use .Xr in manpage

[nyan_myuji.xyz]

• use .Pp instead of empty lines in manpage

[nyan_myuji.xyz]

• update date

[freebsd_igalic.co]

looks good otherwise!

usr.sbin/jail/jail.8

773

cause —> causes

[jamie]

sys/sys/jail.h

209

This would be better as a replacement for pr_spare.

[jamie]

Rather than add a separate value different from the kern.fallback_elf_brand sysctl, it would make sense for the jail parameter to be tied to the sysctl itself, such as the securelevel parameter is. It's complicated somewhat by the fact that there are three similar sysctls.

[nyan_myuji.xyz]

• Use pr_spare for pr_elf_fallback_brand

[nyan_myuji.xyz]

In D40262#917097, @jamie wrote:

Rather than add a separate value different from the kern.fallback_elf_brand sysctl, it would make sense for the jail parameter to be tied to the sysctl itself, such as the securelevel parameter is. It's complicated somewhat by the fact that there are three similar sysctls.

I think that will be a good idea. However right now the kern.elfXX.fallback_elf_brand are tied to 2 global variables which their name are generated by the __elfN(xxx) macro. We could technically remove them and tie all 3 to the jail one? Maybe that belongs to a separate review?

[jamie]

In D40262#917102, @nyan_myuji.xyz wrote:

However right now the kern.elfXX.fallback_elf_brand are tied to 2 global variables which their name are generated by the __elfN(xxx) macro. We could technically remove them and tie all 3 to the jail one? Maybe that belongs to a separate review?

Interesting. I didn't expect that kern.fallback_elf_brand was a just an alias for kern.elfXX.fallback_brand. But then I also didn't expect it would have a comment /* XXX compatibility, remove for 6.0 */ ;-). This does complicate things, since you can't just use struct prison for dynamic sysctls like the per-arch elf fallbacks, but would need to use OSD(9) instead. So yes, perhaps that's something for another review.

But then, since there's not a non-arch-specific fallback value (though ther kern.fallback_elf_brand sysctl alias makes it look like there is), I'm not sure we should be introducing one for jail only. But maybe it would be good to make such a global fallback, tied to the jail as I mention, and use kern.fallback_elf_brand for that. And then perhaps that old XXX comment could go away.

[dchagin]

sys/kern/imgact_elf.c

349

I suggest drop it below, ie, initialize prison_fallback_brand before using

[nyan_myuji.xyz]

• assign value to prison_fallback_brand varibale right before it is used