Re: eBPF verifier does not load libxdp dispatcher eBPF program

[Chris Ward]

The 5.15.0 kernel (built by 'git checkout v5.15' from the kernel.org
torvalds tree) fails in the same way that the 6.2.0-rc5+ kernel fails.
So it seems that something Canonical did for the Ubuntu 20.04 kernel
causes eBPF to work correctly.

On Mon, 23 Jan 2023 at 11:06, Chris Ward <tjcw01@xxxxxxxxx> wrote:
>
> I am trying to use the 'bleeding edge' kernel to determine whether a
> problem I see has already been fixed, but with this kernel the eBPF
> verifier will not load the dispatcher program that is contained within
> libxdp. I am testing kernel commit hash 2475bf0 which fails, and the
> kernel in Ubuntu 22.04 (5.15.0-58-generic) works properly. I am
> running the test case from
> https://github.com/tjcw/bpf-examples/tree/tjcw-explore-sameeth ; to
> build it go to the AF_XDP-filter directory and type 'make', and to run
> it go to the AF_XDP-filter/runscripts/iperf3-namespace directory and
> type 'sudo FILTER=af_xdp_kern PORT=50000 ./run.sh' .
> The lines from the run output indicating the failure are
> libbpf: prog 'xdp_dispatcher': BPF program load failed: Invalid argument
> libbpf: prog 'xdp_dispatcher': -- BEGIN PROG LOAD LOG --
> Func#11 is safe for any args that match its prototype
> btf_vmlinux is malformed
> reg type unsupported for arg#0 function xdp_dispatcher#29
> 0: R1=ctx(off=0,imm=0) R10=fp0
> ; int xdp_dispatcher(struct xdp_md *ctx)
> 0: (bf) r6 = r1                       ; R1=ctx(off=0,imm=0)
> R6_w=ctx(off=0,imm=0)
> 1: (b7) r0 = 2                        ; R0_w=2
> ; __u8 num_progs_enabled = conf.num_progs_enabled;
> 2: (18) r8 = 0xffffb2f6c06d8000       ; R8_w=map_value(off=0,ks=4,vs=84,imm=0)
> 4: (71) r7 = *(u8 *)(r8 +0)           ; R7=1
> R8=map_value(off=0,ks=4,vs=84,imm=0)
> ; if (num_progs_enabled < 1)
> 5: (15) if r7 == 0x0 goto pc+141      ; R7=1
> ; ret = prog0(ctx);
> 6: (bf) r1 = r6                       ; R1_w=ctx(off=0,imm=0)
> R6=ctx(off=0,imm=0)
> 7: (85) call pc+140
> btf_vmlinux is malformed
> R1 type=ctx expected=fp
> Caller passes invalid args into func#1
> processed 84 insns (limit 1000000) max_states_per_insn 0 total_states
> 9 peak_states 9 mark_read 1
> -- END PROG LOAD LOG --
> libbpf: prog 'xdp_dispatcher': failed to load: -22
> libbpf: failed to load object 'xdp-dispatcher.o'
> libxdp: Failed to load dispatcher: Invalid argument
> libxdp: Falling back to loading single prog without dispatcher
>
> Can this regression be fixed before kernel 6.2 ships ?