On 31/7/26 08:07, Andrii Nakryiko wrote: > On Sat, Jul 25, 2026 at 6:27 AM Leon Hwang <[email protected]> wrote: >> >> When CONFIG_FUNCTION_ERROR_INJECTION is disabled, a sleepable tracing prog >> is allowed to attach to '__x64_'-alike prefix symbols. >> >> It is because the verifier does not verify whether the symbol is a kernel >> function or a bpf prog. That said, a sleepable tracing prog is allowed to >> attach to a bpf prog target whose name has '__x64_'-alike prefix. >> >> For example, a sleepable fentry prog attaches to a '__x64_sys_nop' XDP > > we do have addr, so we should be able to distinguish between attaching > to kernel function vs BPF program, no?
Yes, we can distinguish a bpf prog from a kernel function by addr. I'd prefer passing 'tgt_prog' to btf_id_allow_sleepable() as a simple change. > >> prog, and copies buffer from a user pointer with bpf_copy_from_user() >> helper. After attaching the XDP prog to lo interface, the kernel BUG >> could be triggered by 'ping -c 1 -W 1 127.0.0.1': >> >> [ 3.460756] BUG: sleeping function called from invalid context at >> kernel/bpf/trampoline.c:1324 >> >> Fix it by disallowing sleepable tracing prog always when its target is >> bpf prog. > > what happens when we freplace sleepable BPF program/subprogram with > another sleepable BPF subprogram? And same question for sleepable > fentry/fexit program attaching to sleepable BPF program? Is it > something that just cannot work or we can actually allow that? Currently, sleepable freplace/fentry/fexit progs are already disallowed from attaching to bpf prog, because btf_id_allow_sleepable() returns -EINVAL for them (except for this BUG). Since no one has proposed relaxing the restriction, I'd like to keep it as-is when fixing this BUG. I'm not sure whether allowing it would introduce any issue. Thanks, Leon > [...]
