Re: xdp_iphash_to_cpu_cmdline: Error loading BTF: Invalid argument(22)

["Ethy H. Brito" <ethy.brito@inexo.com.br>]

On Thu, 24 Jun 2021 22:16:37 +0200
Jesper Dangaard Brouer <brouer@redhat.com> wrote:

> On Thu, 24 Jun 2021 17:04:58 -0300
> "Ethy H. Brito" <ethy.brito@inexo.com.br> wrote:
> 
> > On Tue, 22 Jun 2021 17:22:16 -0300
> > "Ethy"  wrote:
> >   
> > > Hi
> > > 
> > > Runing "# ./xdp_iphash_to_cpu --dev eth1 --all-cpus --wan" returns 2 loading errors:
> > > 
> > > ----------------------8<----------------------------------
> > > libbpf: Error loading BTF: Invalid argument(22)
> > > libbpf: magic: 0xeb9f
> > > version: 1
> > > flags: 0x0
> > > hdr_len: 24
> > > type_off: 0
> > > type_len: 492
> > > str_off: 492
> > > str_len: 1724
> > > btf_total_size: 2240
> > > [1] PTR (anon) type_id=2
> > > [2] STRUCT xdp_md size=20 vlen=5
> > > 	data type_id=3 bits_offset=0
> > > 	data_end type_id=3 bits_offset=32
> > > 	data_meta type_id=3 bits_offset=64
> > > 	ingress_ifindex type_id=3 bits_offset=96
> > > 	rx_queue_index type_id=3 bits_offset=128
> > > [3] TYPEDEF __u32 type_id=4
> > > [4] INT unsigned int size=4 bits_offset=0 nr_bits=32 encoding=(none)
> > > [5] FUNC_PROTO (anon) return=6 args=(1 ctx)
> > > [6] INT int size=4 bits_offset=0 nr_bits=32 encoding=SIGNED
> > > [7] FUNC xdp_program type_id=5 vlen != 0
> > > 
> > > libbpf: Error loading ELF section .BTF: -22. Ignored and continue.  
> 
> NOTICE above: "Ignored and continue".
> 
> Thus, this is not a real *error*, instead this is just a (very) verbose
> *warning* that your system (kernel version? / compile options?) doesn't
> support BTF correctly.  The BTF info is not really needed by
> xdp-cpuma-tc, so you can safely ignore this "warning".

Noticed!

> 
> 
> > > TC workaround for mapname: map_ifindex_type map_fd:5
> > > Added CPU:0 queue_size:192
> > > Added CPU:1 queue_size:192
> > > Added CPU:2 queue_size:192
> > > Added CPU:3 queue_size:192
> > > Added CPU:4 queue_size:192
> > > Added CPU:5 queue_size:192
> > > Added CPU:6 queue_size:192
> > > Added CPU:7 queue_size:192
> > > Added CPU:8 queue_size:192
> > > Added CPU:9 queue_size:192
> > > Added CPU:10 queue_size:192
> > > Added CPU:11 queue_size:192
> > > Added CPU:12 queue_size:192
> > > Added CPU:13 queue_size:192
> > > Added CPU:14 queue_size:192
> > > Added CPU:15 queue_size:192
> > > Documentation:
> > >  XDP: Lookup IPv4 and redirect to CPU hash
> > > 
> > > This program loads the XDP eBPF program into the kernel.
> > > Use the cmdline tool for add/removing dest IPs to the hash
> > > 
> > >  - Attached to device:eth1 (ifindex:4) prog_id:73
> > > ----------------------------------8<-------------------------------
> > > 
> > > Since the XDP program was correctly listed by "bpftool net", I continued an added just ONE ip
> > > 
> > > # ./xdp_iphash_to_cpu_cmdline --add --ip 1.2.3.4 --classid 1:1234 --cpu 1
> > > key: 0x4030201
> > > WARN: Looks like map_txq_config --base-setup is missing
> > > WARN: Fixing, doing map_txq_config --base-setup
> > > iphash_modify() IP:1.2.3.4 key:0x4030201 TC-handle:0x11234
> > > 
> > > listed the map
> > > 
> > > # ./xdp_iphash_to_cpu_cmdline --list
> > > {
> > > "1.2.3.4" : { "cpu" : 1, "tc_maj" : "1" , "tc_min" : "1234" }}
> > > 
> > > 
> > > And right after I tried to remove the program from XDP:
> > > 
> > > # ./xdp_iphash_to_cpu --dev eth1 --remove
> > > Removing XDP program on ifindex:4 device:eth1
> > > ERR: Clear ifindex type failed 
> > > INFO: not cleanup pinned map file:/sys/fs/bpf/tc/globals/map_ip_hash (use 'rm')


What about this one?
"--revove" could not do its job.
None of the three map files were erased. 
And not only this one that the error message states.
Won't this cause a memory leak??


> > > 
> > > All seems to be rigth despite the errors.
> > > Are all these errors and warnings serious problems? Should I worry? 
> > > 
> > > Hints??    
> > 
> > Hi.
> > 
> > Doesn't anybody know how to answer these?  
> 
> This is just a warning you can ignore.

Ignoring them.

> It looks like it works...


Thank you for your reply. Ignoring them.

I am running tests at a test environment. Things seem to be Ok.
Just some doubts of a fist time user. 

Regards

Ethy