libc: Let elf_aux_info() handle AT_KPRELOAD

[jrtc27]

What's the use case? So a static libc can also use it in future?

[markj]

In D33702#761792, @jrtc27 wrote:

What's the use case? So a static libc can also use it in future?

I wanted to change tests/sys/kern/kern_copyin.c to use this instead of the KERN_PROC_PS_STRINGS sysctl to derive the max user address. That would still be kind of gross, maybe the right approach is to add a new sysctl to fetch that value. But is there any reason not to expose the AT_KPRELOAD value this way?

[jrtc27]

It's a bit scary from a CHERI perspective, but probably ok...

[jrtc27]

But yes, adding a sysctl for the actual thing you want rather than inferring it from ps_strings or kpreload+PAGE_SIZE seems like the right thing to do, regardless of whether this is exposed

[kib]

It definitely should not be AT_KPRELOAD to get the U/K boundary. For instance, shared page base might get randomized.

[markj]

https://reviews.freebsd.org/D33708