On Thu, Jul 02, 2020 at 06:05:29PM -0700, Linus Torvalds wrote: > On Thu, Jul 2, 2020 at 1:03 PM Alexei Starovoitov > <alexei.starovoitov@xxxxxxxxx> wrote: > > > > The BPF program dump_bpf_prog() in iterators.bpf.c is printing this data about > > all BPF programs currently loaded in the system. This information is unstable > > and will change from kernel to kernel. > > If so, it should probably be in debugfs, not in /sys/fs/ /sys/fs/bpf/ is just a historic location where we chose to mount bpffs. iirc iproute2 picked that location and systemd followed. bpffs itself is not a single mount and not sharing anything with sysfs or debugfs. By default it's not mounted anywhere. Every instance is independent and can contain only pinned BPF objects: progs, maps, links. Folks are using bpffs to store BPF objects that need to survive the life time of the process that created those objects. Another use is to share that BPF object with another process. Like firewall service can load bpf prog attach to netdev and pin bpf map with firewall rules in some /var/my_firewall/bpf/my_fw_rules. Then another process can do map lookup/delete on that map if it can access the path. I've seen such use case in production. As far as preloading "progs" and "maps" iterators into bpffs the following works just as well: $ mkdir /aa $ mount bpffs /aa -t bpf $ ll /aa total 4 drwxrwxrwt 2 root root 0 Jul 2 00:27 . drwxr-xr-x 19 root root 4096 Jul 2 00:09 .. -rw------- 1 root root 0 Jul 2 00:27 maps -rw------- 1 root root 0 Jul 2 00:27 progs $ cat /aa/progs id name pages attached 17 dump_bpf_map 1 bpf_iter_bpf_map 18 dump_bpf_prog 1 bpf_iter_bpf_prog May be I misunderstood what you meant?
