Can Linux UML kernels be configured to support BPF tracepoints?
15:38 05 May 2026

When loading using libbpf a BTF object that defines a probe on SEC("tracepoint/syscalls/sys_enter_execve") on an UML kernel (tested with 6.18.26), EINVAL is returned from __sys_bpf -> bpf_prog_load -> find_prog_type where the program type is BPF_PROG_TYPE_TRACING.

I enabled these using scripts/config but CONFIG_TRACING is still not listed in the .config file:

BPF
BPF_SYSCALL
DEBUG_INFO_BTF
BPF_EVENTS
TRACEPOINTS
DEBUG_FS
KPROBES
KPROBE_EVENTS
EVENT_TRACING
TRACING
FTRACE
FUNCTION_TRACER
DYNAMIC_FTRACE

Do the UML Kconfig rules force CONFIG_TRACING or at least BPF tracing probes to be disabled ?

linux kernel trace bpf user-mode-linux