The riscv BPF JIT doesn't emit proper kCFI prologues for BPF programs
and struct_ops trampolines when CONFIG_CFI_CLANG is enabled.

This causes CFI failures when calling BPF programs and can even crash
the kernel due to invalid memory accesses.

Example crash:

root@rv-selftester:~/bpf# ./test_progs -a dummy_st_ops

 Unable to handle kernel paging request at virtual address ffffffff78204ffc
 Oops [#1]
 Modules linked in: bpf_testmod(OE) [....]
 CPU: 3 PID: 356 Comm: test_progs Tainted: P           OE      6.8.0-rc1 #1
 Hardware name: riscv-virtio,qemu (DT)
 epc : bpf_struct_ops_test_run+0x28c/0x5fc
  ra : bpf_struct_ops_test_run+0x26c/0x5fc
 epc : ffffffff82958010 ra : ffffffff82957ff0 sp : ff200000007abc80
  gp : ffffffff868d6218 tp : ff6000008d87b840 t0 : 000000000000000f
  t1 : 0000000000000000 t2 : 000000002005793e s0 : ff200000007abcf0
  s1 : ff6000008a90fee0 a0 : 0000000000000000 a1 : 0000000000000000
  a2 : 0000000000000000 a3 : 0000000000000000 a4 : 0000000000000000
  a5 : ffffffff868dba26 a6 : 0000000000000001 a7 : 0000000052464e43
  s2 : 00007ffffc0a95f0 s3 : ff6000008a90fe80 s4 : ff60000084c24c00
  s5 : ffffffff78205000 s6 : ff60000088750648 s7 : ff20000000035008
  s8 : fffffffffffffff4 s9 : ffffffff86200610 s10: 0000000000000000
  s11: 0000000000000000 t3 : ffffffff8483dc30 t4 : ffffffff8483dc10
  t5 : ffffffff8483dbf0 t6 : ffffffff8483dbd0
 status: 0000000200000120 badaddr: ffffffff78204ffc cause: 000000000000000d
 [<ffffffff82958010>] bpf_struct_ops_test_run+0x28c/0x5fc
 [<ffffffff805083ee>] bpf_prog_test_run+0x170/0x548
 [<ffffffff805029c8>] __sys_bpf+0x2d2/0x378
 [<ffffffff804ff570>] __riscv_sys_bpf+0x5c/0x120
 [<ffffffff8000e8fe>] syscall_handler+0x62/0xe4
 [<ffffffff83362df6>] do_trap_ecall_u+0xc6/0x27c
 [<ffffffff833822c4>] ret_from_exception+0x0/0x64
 Code: b603 0109 b683 0189 b703 0209 8493 0609 157d 8d65 (a303) ffca
 ---[ end trace 0000000000000000 ]---
 Kernel panic - not syncing: Fatal exception
 SMP: stopping secondary CPUs

Implement proper kCFI prologues for the BPF programs and callbacks and
drop __nocfi for riscv64. Fix the trampoline generation code to emit kCFI
prologue when a struct_ops trampoline is being prepared.
Signed-off-by: Puranjay Mohan <puranjay12@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/bpf/20240303170207.82201-2-puranjay12@gmail.com