Implement prlimit64() system call.

[dchagin]

style(9)

[trasz]

sys/amd64/linux32/syscalls.master

558

const? Also, FreeBSD rlim_t is int64, while Linux 'struct rlimit64' seems to have uint64 there.

Same thing below.

[dchagin]

Add a note about rlim signedness.

About const - I would prefer to not use 'const' in masters,
thereto clang complains to discard qualifers in copyout()

[dchagin]

I'm wrong, we do not conversion for new limits, but we should convert RLIM_INFINITY.

[trasz]

Erm, what exactly does the clang say? I mean, in copyout(9) declaration there is a "const" for the kernel pointer, so it shouldn't complain about it?

[trasz]

I'd prefer explicit conversion - even if it's just memcpy in one of the directions - but it's ok for now.

[trasz]

Erm, wait. You're missing access check before kern_proc_setrlimit(). I _think_ it should be PGET_CANDEBUG (as opposed to PGET_CANSEE for retrieving limits); that's what sysctl_kern_proc_rlimit() uses.

[dchagin]

Yes, yoa are right! I was thinking why kern_proc_setrlimit()
does not check access. ie, sys_setrlimit.

fixed by copiyng code fro, sysctl code, thnak you very match!

[dchagin]

typo fixed.

[trasz]

sys/compat/linux/linux_misc.c

1986

Does it still do the same thing for PID 0?