When creating map-in-maps, the outer map can be prepopulated using the
inner_idx field of inner maps. That field defines the index of the inner
map in the outer map. It is ignored if set to -1.
Commit 6d61a2b557 ("lib: add libbpf support") however started using
that field to identify inner maps. While iterating over all maps looking
for inner maps, maps with inner_idx set to -1 are erroneously skipped.
As a result, trying to create a map-in-map with prepopulation disabled
fails because the inner_id of the outer map is not correctly set.
This bug can be observed with strace -ebpf (notice the zero inner_map_fd
for the outer map creation):
bpf(BPF_MAP_CREATE, {map_type=BPF_MAP_TYPE_ARRAY, key_size=4, value_size=130996, max_entries=1, map_flags=0, inner_map_fd=0, map_name="maglev_inner", map_ifindex=0, btf_fd=0, btf_key_type_id=0, btf_value_type_id=0}, 128) = 32
bpf(BPF_MAP_CREATE, {map_type=BPF_MAP_TYPE_HASH_OF_MAPS, key_size=2, value_size=4, max_entries=65536, map_flags=BPF_F_NO_PREALLOC, inner_map_fd=0, map_name="maglev_outer", map_ifindex=0, btf_fd=0, btf_key_type_id=0, btf_value_type_id=0}, 128) = -1 EINVAL (Invalid argument)
Fixes: 6d61a2b557 ("lib: add libbpf support")
Signed-off-by: Paul Chaignon <paul@isovalent.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
bpf selftests using iproute2 fails with:
$ ip link set dev veth0 xdp object ../bpf/xdp_dummy.o section xdp_dummy
Continuing without mounted eBPF fs. Too old kernel?
mkdir (null)/globals failed: No such file or directory
Unable to load program
This happens when the /sys/fs/bpf directory exists. In this case, mkdir
in bpf_mnt_check_target() fails with errno == EEXIST, and the function
returns -1. Thus bpf_get_work_dir() does not call bpf_mnt_fs() and the
bpffs is not mounted.
Fix this in bpf_mnt_check_target(), returning 0 when the mountpoint
exists.
Fixes: d4fcdbbec9 ("lib/bpf: Fix and simplify bpf_mnt_check_target()")
Reported-by: Mingyu Shi <mshi@redhat.com>
Reported-by: Jiri Benc <jbenc@redhat.com>
Suggested-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
The BPF program name is included when dumping the BPF program info and the
kernel only stores the first (BPF_PROG_NAME_LEN - 1) bytes for the program
name.
$ sudo ip link show dev docker0
4: docker0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 xdpgeneric qdisc noqueue state UP mode DEFAULT group default
link/ether 02:42:4c:df:a4:54 brd ff:ff:ff:ff:ff:ff
prog/xdp id 789 name xdp_drop_func tag 57cd311f2e27366b jited
The BPF program load time (ns since boottime), UID of the user who loaded
the program and the BTF ID are also included when dumping the BPF program
information when the user expects a detailed ip link info output.
$ sudo ip -details link show dev docker0
4: docker0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 xdpgeneric qdisc noqueue state UP mode DEFAULT group default
link/ether 02:42:4c:df:a4:54 brd ff:ff:ff:ff:ff:ff promiscuity 0 minmtu 68 maxmtu 65535
bridge forward_delay 1500 hello_time 200 max_age 2000 ageing_time 30000 stp_state 0 priority 32768 vlan_filt
ering 0 vlan_protocol 802.1Q bridge_id 8000.2:42:4c:df:a4:54 designated_root 8000.2:42:4c:df:a4:54 root_port 0 r
oot_path_cost 0 topology_change 0 topology_change_detected 0 hello_timer 0.00 tcn_timer 0.00 topology_chan
ge_timer 0.00 gc_timer 265.36 vlan_default_pvid 1 vlan_stats_enabled 0 vlan_stats_per_port 0 group_fwd_mask
0 group_address 01:80:c2:00:00:00 mcast_snooping 1 mcast_router 1 mcast_query_use_ifaddr 0 mcast_querier 0 mcast
_hash_elasticity 16 mcast_hash_max 4096 mcast_last_member_count 2 mcast_startup_query_count 2 mcast_last_member_
interval 100 mcast_membership_interval 26000 mcast_querier_interval 25500 mcast_query_interval 12500 mcast_query
_response_interval 1000 mcast_startup_query_interval 3124 mcast_stats_enabled 0 mcast_igmp_version 2 mcast_mld_v
ersion 1 nf_call_iptables 0 nf_call_ip6tables 0 nf_call_arptables 0 addrgenmode eui64 numtxqueues 1 numrxqueues
1 gso_max_size 65536 gso_max_segs 65535
prog/xdp id 789 name xdp_drop_func tag 57cd311f2e27366b jited load_time 2676682607316255 created_by_uid 0 btf_id 708
Signed-off-by: Gokul Sivakumar <gokulkumar792@gmail.com>
Signed-off-by: David Ahern <dsahern@kernel.org>
If bpf_map_fetch_name() returns NULL, strlen() hits a NULL-pointer
dereference on outer_map_name.
Fix this checking outer_map_name value, and returning false when NULL,
as already done for inner_map_name before.
Fixes: 6d61a2b557 ("lib: add libbpf support")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Use tc with no verbose, when bpf_btf_attach fail,
the conditions:
"if (fd < 0 && (errno == ENOSPC || !ctx->log_size))"
will make ctx->log_size != 0. And then, bpf_prog_attach,
ctx->log_size != 0. so enable debug log.
The verifier log sometimes is so chatty on larger programs.
bpf_prog_attach is failed.
"Log buffer too small to dump verifier log 16777215 bytes (9 tries)!"
BTF load failure does not affect prog load. prog still work.
So when BTF/PROG load fail, enlarge log_size and re-fail with
having verbose.
Signed-off-by: Feng Zhou <zhoufeng.zf@bytedance.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
In function bpf_obj_open, if bpf_fetch_prog_arg() return an error, we
end up in the out: path with a negative value for fd, and pass it to
close.
Avoid this checking for fd to be positive.
Fixes: 32e93fb7f6 ("{f,m}_bpf: allow for sharing maps")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: David Ahern <dsahern@kernel.org>
In functions bpf_{send,recv}_map_fds(), when connect fails after a
socket is successfully opened, we return with error missing a close on
the socket.
Fix this closing the socket if opened and using a single return point
for both the functions.
Fixes: 6256f8c9e4 ("tc, bpf: finalize eBPF support for cls and act front-end")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As stated in the man page(), open returns a non-negative integer as a
file descriptor. Hence, when checking for its return value to be ok, we
should include 0 as a valid value.
This fixes a covscan warning about a missing close() in this function.
Fixes: ecb05c0f99 ("bpf: improve error reporting around tail calls")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
As stated in commit ac3415f5c1 ("lib/fs: Fix and simplify make_path()"),
calling stat() before mkdir() is racey, because the entry might change in
between.
As the call to stat() seems to only check for target existence, we can
simply call mkdir() unconditionally and catch all errors but EEXIST.
Fixes: 95ae9a4870 ("bpf: fix mnt path when from env")
Signed-off-by: Andrea Claudi <aclaudi@redhat.com>
This patch converts iproute2 to use libbpf for loading and attaching
BPF programs when it is available, which is started by Toke's
implementation[1]. With libbpf iproute2 could correctly process BTF
information and support the new-style BTF-defined maps, while keeping
compatibility with the old internal map definition syntax.
The old iproute2 bpf code is kept and will be used if no suitable libbpf
is available. When using libbpf, wrapper code in bpf_legacy.c ensures that
iproute2 will still understand the old map definition format, including
populating map-in-map and tail call maps before load.
In bpf_libbpf.c, we init iproute2 ctx and elf info first to check the
legacy bytes. When handling the legacy maps, for map-in-maps, we create
them manually and re-use the fd as they are associated with id/inner_id.
For pin maps, we only set the pin path and let libbp load to handle it.
For tail calls, we find it first and update the element after prog load.
Other maps/progs will be loaded by libbpf directly.
[1] https://lore.kernel.org/bpf/20190820114706.18546-1-toke@redhat.com/
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Hangbin Liu <haliu@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
There are directly calls in libbpf for bpf program load/attach.
So we could just use two wrapper functions for ipvrf and convert
them with libbpf support.
Function bpf_prog_load() is removed as it's conflict with libbpf
function name.
bpf.c is moved to bpf_legacy.c for later main libbpf support in
iproute2.
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Signed-off-by: Hangbin Liu <haliu@redhat.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
David Ahern
Paul Chaignon
Andrea Claudi
Gokul Sivakumar
Feng Zhou
Hangbin Liu