samples/bpf: modernize BPF functionality test programs
Subsystems (4)
cover samples/bpf: modernize BPF functionality test programs
Currently, there are many programs under samples/bpf to test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.
Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
- bpf_trace_printk() always append newline at the end [2]
- deprecated styled BPF section header (bpf_load style) [3]
- urandom_read tracepoint is removed (used for testing overhead) [4]
- ping sends packet with SOCK_DGRAM instead of SOCK_RAW [5]*
- use "vmlinux.h" instead of including individual headers
In addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from xdp_sample and bunch of constants)
[1]: https://github.com/libbpf/libbpf/issues/282
[2]: commit ac5a72ea5c89 ("bpf: Use dedicated bpf_trace_printk event instead of trace_printk()") [3]: commit ceb5dea56543 ("samples: bpf: Remove bpf_load loader completely") [4]: commit 14c174633f34 ("random: remove unused tracepoints")
[5]: https://lwn.net/Articles/422330/
*: This is quite old, but I'm not sure why the code was initially
developed to filter only SOCK_RAW.
Daniel T. Lee (10):
samples/bpf: ensure ipv6 is enabled before running tests samples/bpf: refactor BPF functionality testing scripts samples/bpf: fix broken lightweight tunnel testing samples/bpf: fix broken cgroup socket testing samples/bpf: replace broken overhead microbenchmark with
fib_table_lookup
samples/bpf: replace legacy map with the BTF-defined map samples/bpf: split common macros to net_shared.h samples/bpf: replace BPF programs header with net_shared.h samples/bpf: use vmlinux.h instead of implicit headers in BPF test
program
samples/bpf: change _kern suffix to .bpf with BPF test programs
samples/bpf/Makefile | 14 +++---
...lwt_len_hist_kern.c => lwt_len_hist.bpf.c} | 29 +++--------
samples/bpf/lwt_len_hist.sh | 4 +-
samples/bpf/net_shared.h | 32 ++++++++++++
.../{sock_flags_kern.c => sock_flags.bpf.c} | 24 ++++-----
samples/bpf/tc_l2_redirect.sh | 3 ++
samples/bpf/test_cgrp2_sock.sh | 16 +++---
samples/bpf/test_cgrp2_sock2.sh | 9 +++-
...st_cgrp2_tc_kern.c => test_cgrp2_tc.bpf.c} | 34 ++++---------
samples/bpf/test_cgrp2_tc.sh | 8 +--
samples/bpf/test_lwt_bpf.c | 50 ++++++++-----------
samples/bpf/test_lwt_bpf.sh | 19 ++++---
...ap_in_map_kern.c => test_map_in_map.bpf.c} | 7 +--
samples/bpf/test_map_in_map_user.c | 2 +-
...robe_kern.c => test_overhead_kprobe.bpf.c} | 6 +--
...w_tp_kern.c => test_overhead_raw_tp.bpf.c} | 4 +-
...rhead_tp_kern.c => test_overhead_tp.bpf.c} | 29 +++++++----
samples/bpf/test_overhead_user.c | 34 ++++++++-----
samples/bpf/xdp_sample.bpf.h | 22 +-------
19 files changed, 179 insertions(+), 167 deletions(-)
rename samples/bpf/{lwt_len_hist_kern.c => lwt_len_hist.bpf.c} (75%)
create mode 100644 samples/bpf/net_shared.h
rename samples/bpf/{sock_flags_kern.c => sock_flags.bpf.c} (66%)
rename samples/bpf/{test_cgrp2_tc_kern.c => test_cgrp2_tc.bpf.c} (70%)
rename samples/bpf/{test_map_in_map_kern.c => test_map_in_map.bpf.c} (97%)
rename samples/bpf/{test_overhead_kprobe_kern.c => test_overhead_kprobe.bpf.c} (92%)
rename samples/bpf/{test_overhead_raw_tp_kern.c => test_overhead_raw_tp.bpf.c} (82%)
rename samples/bpf/{test_overhead_tp_kern.c => test_overhead_tp.bpf.c} (61%)Signature · 1 line
2.34.1
PATCH 1/10 samples/bpf: ensure ipv6 is enabled before running tests
Currently, a few of BPF tests use ipv6 functionality. The problem here is that if ipv6 is disabled, these tests will fail, and even if the test fails, it will not tell you why it failed.
$ sudo ./test_cgrp2_sock2.sh
RTNETLINK answers: Permission deniedIn order to fix this, this commit ensures ipv6 is enabled prior to running tests.
--- samples/bpf/tc_l2_redirect.sh | 3 +++ samples/bpf/test_cgrp2_sock2.sh | 4 +++- samples/bpf/test_cgrp2_tc.sh | 2 ++ 3 files changed, 8 insertions(+), 1 deletion(-)
samples/bpf/tc_l2_redirect.sh +3 -0
@@ -8,6 +8,7 @@ REDIRECT_USER='./tc_l2_redirect' REDIRECT_BPF='./tc_l2_redirect_kern.o' RP_FILTER=$(< /proc/sys/net/ipv4/conf/all/rp_filter) +IPV6_DISABLED=$(< /proc/sys/net/ipv6/conf/all/disable_ipv6) IPV6_FORWARDING=$(< /proc/sys/net/ipv6/conf/all/forwarding) function config_common {
@@ -64,6 +65,7 @@ function config_common { sysctl -q -w net.ipv4.conf.all.rp_filter=0 sysctl -q -w net.ipv6.conf.all.forwarding=1 + sysctl -q -w net.ipv6.conf.all.disable_ipv6=0 } function cleanup {
@@ -77,6 +79,7 @@ function cleanup { $IP link del ip6t >& /dev/null sysctl -q -w net.ipv4.conf.all.rp_filter=$RP_FILTER sysctl -q -w net.ipv6.conf.all.forwarding=$IPV6_FORWARDING + sysctl -q -w net.ipv6.conf.all.disable_ipv6=$IPV6_DISABLED rm -f /sys/fs/bpf/tc/globals/tun_iface [[ -z $DEBUG ]] || set -x set -e
samples/bpf/test_cgrp2_sock2.sh +3 -1
@@ -7,13 +7,15 @@ LINK_PIN=$BPFFS/test_cgrp2_sock2 function config_device { ip netns add at_ns0 ip link add veth0 type veth peer name veth0b - ip link set veth0b up ip link set veth0 netns at_ns0 + ip netns exec at_ns0 sysctl -q net.ipv6.conf.veth0.disable_ipv6=0 ip netns exec at_ns0 ip addr add 172.16.1.100/24 dev veth0 ip netns exec at_ns0 ip addr add 2401:db00::1/64 dev veth0 nodad ip netns exec at_ns0 ip link set dev veth0 up + sysctl -q net.ipv6.conf.veth0b.disable_ipv6=0 ip addr add 172.16.1.101/24 dev veth0b ip addr add 2401:db00::2/64 dev veth0b nodad + ip link set veth0b up } function config_cgroup {
samples/bpf/test_cgrp2_tc.sh +2 -0
@@ -73,11 +73,13 @@ setup_net() { start) $IP link add $HOST_IFC type veth peer name $NS_IFC || return $? $IP link set dev $HOST_IFC up || return $? + sysctl -q net.ipv6.conf.$HOST_IFC.disable_ipv6=0 sysctl -q net.ipv6.conf.$HOST_IFC.accept_dad=0 $IP netns add ns || return $? $IP link set dev $NS_IFC netns ns || return $? $IP -n $NS link set dev $NS_IFC up || return $? + $IP netns exec $NS sysctl -q net.ipv6.conf.$NS_IFC.disable_ipv6=0 $IP netns exec $NS sysctl -q net.ipv6.conf.$NS_IFC.accept_dad=0 $TC qdisc add dev $HOST_IFC clsact || return $? $TC filter add dev $HOST_IFC egress bpf da obj $BPF_PROG sec $BPF_SECTION || return $?
Signature · 1 line
2.34.1
PATCH 2/10 samples/bpf: refactor BPF functionality testing scripts
In reply to Daniel T. LeeCurrently, some test scripts are experiencing minor errors related to executing tests.
$ sudo ./test_cgrp2_sock.sh
./test_cgrp2_sock.sh: 22: test_cgrp2_sock: not foundThis problem occurs because the path to the execution target is not properly specified. Therefore, this commit solves this problem by specifying a relative path to its executables. This commit also makes a concise refactoring of hard-coded BPF program names.
--- samples/bpf/lwt_len_hist.sh | 4 ++-- samples/bpf/test_cgrp2_sock.sh | 16 +++++++++------- samples/bpf/test_cgrp2_sock2.sh | 5 ++++- samples/bpf/test_cgrp2_tc.sh | 4 ++-- samples/bpf/test_lwt_bpf.sh | 8 +++++--- 5 files changed, 22 insertions(+), 15 deletions(-)
samples/bpf/lwt_len_hist.sh +2 -2
@@ -4,7 +4,7 @@ NS1=lwt_ns1 VETH0=tst_lwt1a VETH1=tst_lwt1b - +BPF_PROG=lwt_len_hist_kern.o TRACE_ROOT=/sys/kernel/debug/tracing function cleanup {
@@ -30,7 +30,7 @@ ip netns exec $NS1 netserver echo 1 > ${TRACE_ROOT}/tracing_on cp /dev/null ${TRACE_ROOT}/trace -ip route add 192.168.253.2/32 encap bpf out obj lwt_len_hist_kern.o section len_hist dev $VETH0 +ip route add 192.168.253.2/32 encap bpf out obj $BPF_PROG section len_hist dev $VETH0 netperf -H 192.168.253.2 -t TCP_STREAM cat ${TRACE_ROOT}/trace | grep -v '^#' ./lwt_len_hist
samples/bpf/test_cgrp2_sock.sh +9 -7
@@ -3,6 +3,8 @@ # Test various socket options that can be set by attaching programs to cgroups. +MY_DIR=$(dirname $0) +TEST=$MY_DIR/test_cgrp2_sock CGRP_MNT="/tmp/cgroupv2-test_cgrp2_sock" ################################################################################
@@ -19,7 +21,7 @@ print_result() check_sock() { - out=$(test_cgrp2_sock) + out=$($TEST) echo $out | grep -q "$1" if [ $? -ne 0 ]; then print_result 1 "IPv4: $2"
@@ -33,7 +35,7 @@ check_sock() check_sock6() { - out=$(test_cgrp2_sock -6) + out=$($TEST -6) echo $out | grep -q "$1" if [ $? -ne 0 ]; then print_result 1 "IPv6: $2"
@@ -61,7 +63,7 @@ cleanup_and_exit() [ -n "$msg" ] && echo "ERROR: $msg" - test_cgrp2_sock -d ${CGRP_MNT}/sockopts + $TEST -d ${CGRP_MNT}/sockopts ip li del cgrp2_sock umount ${CGRP_MNT}
@@ -98,7 +100,7 @@ check_sock6 "dev , mark 0, priority 0" "No programs attached" # verify device is set # -test_cgrp2_sock -b cgrp2_sock ${CGRP_MNT}/sockopts +$TEST -b cgrp2_sock ${CGRP_MNT}/sockopts if [ $? -ne 0 ]; then cleanup_and_exit 1 "Failed to install program to set device" fi
@@ -107,7 +109,7 @@ check_sock6 "dev cgrp2_sock, mark 0, priority 0" "Device set" # verify mark is set # -test_cgrp2_sock -m 666 ${CGRP_MNT}/sockopts +$TEST -m 666 ${CGRP_MNT}/sockopts if [ $? -ne 0 ]; then cleanup_and_exit 1 "Failed to install program to set mark" fi
@@ -116,7 +118,7 @@ check_sock6 "dev , mark 666, priority 0" "Mark set" # verify priority is set # -test_cgrp2_sock -p 123 ${CGRP_MNT}/sockopts +$TEST -p 123 ${CGRP_MNT}/sockopts if [ $? -ne 0 ]; then cleanup_and_exit 1 "Failed to install program to set priority" fi
@@ -125,7 +127,7 @@ check_sock6 "dev , mark 0, priority 123" "Priority set" # all 3 at once # -test_cgrp2_sock -b cgrp2_sock -m 666 -p 123 ${CGRP_MNT}/sockopts +$TEST -b cgrp2_sock -m 666 -p 123 ${CGRP_MNT}/sockopts if [ $? -ne 0 ]; then cleanup_and_exit 1 "Failed to install program to set device, mark and priority" fi
samples/bpf/test_cgrp2_sock2.sh +4 -1
@@ -2,7 +2,10 @@ # SPDX-License-Identifier: GPL-2.0 BPFFS=/sys/fs/bpf +MY_DIR=$(dirname $0) +TEST=$MY_DIR/test_cgrp2_sock2 LINK_PIN=$BPFFS/test_cgrp2_sock2 +BPF_PROG=$MY_DIR/sock_flags_kern.o function config_device { ip netns add at_ns0
@@ -36,7 +39,7 @@ function config_bpffs { } function attach_bpf { - ./test_cgrp2_sock2 /tmp/cgroupv2/foo sock_flags_kern.o $1 + $TEST /tmp/cgroupv2/foo $BPF_PROG $1 [ $? -ne 0 ] && exit 1 }
samples/bpf/test_cgrp2_tc.sh +2 -2
@@ -76,8 +76,8 @@ setup_net() { sysctl -q net.ipv6.conf.$HOST_IFC.disable_ipv6=0 sysctl -q net.ipv6.conf.$HOST_IFC.accept_dad=0 - $IP netns add ns || return $? - $IP link set dev $NS_IFC netns ns || return $? + $IP netns add $NS || return $? + $IP link set dev $NS_IFC netns $NS || return $? $IP -n $NS link set dev $NS_IFC up || return $? $IP netns exec $NS sysctl -q net.ipv6.conf.$NS_IFC.disable_ipv6=0 $IP netns exec $NS sysctl -q net.ipv6.conf.$NS_IFC.accept_dad=0
samples/bpf/test_lwt_bpf.sh +5 -3
@@ -19,6 +19,8 @@ IPVETH3="192.168.111.2" IP_LOCAL="192.168.99.1" +PROG_SRC="test_lwt_bpf.c" +BPF_PROG="test_lwt_bpf.o" TRACE_ROOT=/sys/kernel/debug/tracing function lookup_mac()
@@ -36,7 +38,7 @@ function lookup_mac() function cleanup { set +ex - rm test_lwt_bpf.o 2> /dev/null + rm $BPF_PROG 2> /dev/null ip link del $VETH0 2> /dev/null ip link del $VETH1 2> /dev/null ip link del $VETH2 2> /dev/null
@@ -76,7 +78,7 @@ function install_test { cleanup_routes cp /dev/null ${TRACE_ROOT}/trace - OPTS="encap bpf headroom 14 $1 obj test_lwt_bpf.o section $2 $VERBOSE" + OPTS="encap bpf headroom 14 $1 obj $BPF_PROG section $2 $VERBOSE" if [ "$1" == "in" ]; then ip route add table local local ${IP_LOCAL}/32 $OPTS dev lo
@@ -374,7 +376,7 @@ DST_IFINDEX=$(cat /sys/class/net/$VETH0/ifindex) CLANG_OPTS="-O2 -target bpf -I ../include/" CLANG_OPTS+=" -DSRC_MAC=$SRC_MAC -DDST_MAC=$DST_MAC -DDST_IFINDEX=$DST_IFINDEX" -clang $CLANG_OPTS -c test_lwt_bpf.c -o test_lwt_bpf.o +clang $CLANG_OPTS -c $PROG_SRC -o $BPF_PROG test_ctx_xmit test_ctx_out
Signature · 1 line
2.34.1
PATCH 3/10 samples/bpf: fix broken lightweight tunnel testing
In reply to Daniel T. LeeThe test_lwt_bpf is a script that tests the functionality of BPF through the output of the ftrace with bpf_trace_printk. Currently, this program is not operating normally for several reasons.
First of all, this test script can't parse the ftrace results properly. GNU sed tries to be as greedy as possible when attempting pattern matching. Due to this, cutting metadata (such as timestamp) from the log entry of ftrace doesn't work properly, and also desired log isn't extracted properly. To make sed stripping clearer, 'nocontext-info' option with the ftrace has been used to remove metadata from the log. Also, instead of using unclear pattern matching, this commit specifies an explicit parse pattern.
Also, unlike before when this test was introduced, the way bpf_trace_printk behaves has changed[1]. The previous bpf_trace_printk had to always have '\n' in order to print newline, but now that the bpf_trace_printk call includes newline by default, so '\n' is no longer needed.
Lastly with the lwt ENCAP_BPF out, the context information with the sk_buff protocol is preserved. Therefore, this commit changes the previous test result from 'protocol 0' to 'protocol 8', which means ETH_P_IP.
[1]: commit ac5a72ea5c89 ("bpf: Use dedicated bpf_trace_printk event instead of trace_printk()")
--- samples/bpf/test_lwt_bpf.c | 36 ++++++++++++++++++------------------ samples/bpf/test_lwt_bpf.sh | 11 +++++++---- 2 files changed, 25 insertions(+), 22 deletions(-)
samples/bpf/test_lwt_bpf.c +18 -18
@@ -44,9 +44,9 @@ SEC("test_ctx")int do_test_ctx (struct __sk_buff *skb) { skb->cb[0 ] =CB_MAGIC ; -printk ("len %d hash %d protocol %d\n" , skb->len, skb->hash, +printk ("len %d hash %d protocol %d" , skb->len, skb->hash, skb->protocol); -printk ("cb %d ingress_ifindex %d ifindex %d\n" , skb->cb[0 ], +printk ("cb %d ingress_ifindex %d ifindex %d" , skb->cb[0 ], skb->ingress_ifindex, skb->ifindex);return BPF_OK ;
@@ -56,9 +56,9 @@ int do_test_ctx(struct __sk_buff *skb) SEC("test_cb")int do_test_cb (struct __sk_buff *skb) { -printk ("cb0: %x cb1: %x cb2: %x\n" , skb->cb[0 ], skb->cb[1 ], +printk ("cb0: %x cb1: %x cb2: %x" , skb->cb[0 ], skb->cb[1 ], skb->cb[2 ]); -printk ("cb3: %x cb4: %x\n" , skb->cb[3 ], skb->cb[4 ]); +printk ("cb3: %x cb4: %x" , skb->cb[3 ], skb->cb[4 ]);return BPF_OK ; }
@@ -72,11 +72,11 @@ int do_test_data(struct __sk_buff *skb)struct iphdr *iph = data;if (data +sizeof (*iph) > data_end) { -printk ("packet truncated\n" ); +printk ("packet truncated" );return BPF_DROP ; } -printk ("src: %x dst: %x\n" , iph->saddr, iph->daddr); +printk ("src: %x dst: %x" , iph->saddr, iph->daddr);return BPF_OK ; }
@@ -97,7 +97,7 @@ static inline int rewrite(struct __sk_buff *skb, uint32_t old_ip, ret =bpf_skb_load_bytes (skb,IP_PROTO_OFF , &proto,1 );if (ret <0 ) { -printk ("bpf_l4_csum_replace failed: %d\n" , ret); +printk ("bpf_l4_csum_replace failed: %d" , ret);return BPF_DROP ; }
@@ -120,14 +120,14 @@ static inline int rewrite(struct __sk_buff *skb, uint32_t old_ip, ret =bpf_l4_csum_replace (skb, off, old_ip, new_ip, flags |sizeof (new_ip));if (ret <0 ) { -printk ("bpf_l4_csum_replace failed: %d\n" ); +printk ("bpf_l4_csum_replace failed: %d" );return BPF_DROP ; } } ret =bpf_l3_csum_replace (skb,IP_CSUM_OFF , old_ip, new_ip,sizeof (new_ip));if (ret <0 ) { -printk ("bpf_l3_csum_replace failed: %d\n" , ret); +printk ("bpf_l3_csum_replace failed: %d" , ret);return BPF_DROP ; }
@@ -137,7 +137,7 @@ static inline int rewrite(struct __sk_buff *skb, uint32_t old_ip, ret =bpf_skb_store_bytes (skb,IP_SRC_OFF , &new_ip,sizeof (new_ip),0 );if (ret <0 ) { -printk ("bpf_skb_store_bytes() failed: %d\n" , ret); +printk ("bpf_skb_store_bytes() failed: %d" , ret);return BPF_DROP ; }
@@ -153,12 +153,12 @@ int do_test_rewrite(struct __sk_buff *skb) ret =bpf_skb_load_bytes (skb,IP_DST_OFF , &old_ip,4 );if (ret <0 ) { -printk ("bpf_skb_load_bytes failed: %d\n" , ret); +printk ("bpf_skb_load_bytes failed: %d" , ret);return BPF_DROP ; }if (old_ip ==0x2fea8c0 ) { -printk ("out: rewriting from %x to %x\n" , old_ip, new_ip); +printk ("out: rewriting from %x to %x" , old_ip, new_ip);return rewrite (skb, old_ip, new_ip,1 ); }
@@ -173,7 +173,7 @@ static inline int __do_push_ll_and_redirect(struct __sk_buff *skb) ret =bpf_skb_change_head (skb,14 ,0 );if (ret <0 ) { -printk ("skb_change_head() failed: %d\n" , ret); +printk ("skb_change_head() failed: %d" , ret); } ehdr.h_proto =__constant_htons (ETH_P_IP );
@@ -182,7 +182,7 @@ static inline int __do_push_ll_and_redirect(struct __sk_buff *skb) ret =bpf_skb_store_bytes (skb,0 , &ehdr,sizeof (ehdr),0 );if (ret <0 ) { -printk ("skb_store_bytes() failed: %d\n" , ret); +printk ("skb_store_bytes() failed: %d" , ret);return BPF_DROP ; }
@@ -202,7 +202,7 @@ int do_push_ll_and_redirect(struct __sk_buff *skb) ret =__do_push_ll_and_redirect (skb);if (ret >=0 ) -printk ("redirected to %d\n" , ifindex); +printk ("redirected to %d" , ifindex);return ret; }
@@ -229,7 +229,7 @@ SEC("fill_garbage")int do_fill_garbage (struct __sk_buff *skb) {__fill_garbage (skb); -printk ("Set initial 96 bytes of header to FF\n" ); +printk ("Set initial 96 bytes of header to FF" );return BPF_OK ; }
@@ -238,7 +238,7 @@ int do_fill_garbage_and_redirect(struct __sk_buff *skb) {int ifindex =DST_IFINDEX ;__fill_garbage (skb); -printk ("redirected to %d\n" , ifindex); +printk ("redirected to %d" , ifindex);return bpf_redirect (ifindex,0 ); }
@@ -246,7 +246,7 @@ int do_fill_garbage_and_redirect(struct __sk_buff *skb) SEC("drop_all")int do_drop_all (struct __sk_buff *skb) { -printk ("dropping with: %d\n" ,BPF_DROP ); +printk ("dropping with: %d" ,BPF_DROP );return BPF_DROP ; }
samples/bpf/test_lwt_bpf.sh +7 -4
@@ -22,6 +22,7 @@ IP_LOCAL="192.168.99.1" PROG_SRC="test_lwt_bpf.c" BPF_PROG="test_lwt_bpf.o" TRACE_ROOT=/sys/kernel/debug/tracing +CONTEXT_INFO=$(cat ${TRACE_ROOT}/trace_options | grep context) function lookup_mac() {
@@ -98,7 +99,7 @@ function remove_prog { function filter_trace { # Add newline to allow starting EXPECT= variables on newline NL=$'\n' - echo "${NL}$*" | sed -e 's/^.*: : //g' + echo "${NL}$*" | sed -e 's/bpf_trace_printk: //g' } function expect_fail {
@@ -162,11 +163,11 @@ function test_ctx_out { failure "test_ctx out: packets are dropped" } match_trace "$(get_trace)" " -len 84 hash 0 protocol 0 +len 84 hash 0 protocol 8 cb 1234 ingress_ifindex 0 ifindex 0 -len 84 hash 0 protocol 0 +len 84 hash 0 protocol 8 cb 1234 ingress_ifindex 0 ifindex 0 -len 84 hash 0 protocol 0 +len 84 hash 0 protocol 8 cb 1234 ingress_ifindex 0 ifindex 0" || exit 1 remove_prog out }
@@ -369,6 +370,7 @@ setup_one_veth $NS1 $VETH0 $VETH1 $IPVETH0 $IPVETH1 $IPVETH1b setup_one_veth $NS2 $VETH2 $VETH3 $IPVETH2 $IPVETH3 ip netns exec $NS1 netserver echo 1 > ${TRACE_ROOT}/tracing_on +echo nocontext-info > ${TRACE_ROOT}/trace_options DST_MAC=$(lookup_mac $VETH1 $NS1) SRC_MAC=$(lookup_mac $VETH0)
@@ -399,4 +401,5 @@ test_netperf_redirect cleanup echo 0 > ${TRACE_ROOT}/tracing_on +echo $CONTEXT_INFO > ${TRACE_ROOT}/trace_options exit 0
Signature · 1 line
2.34.1
PATCH 4/10 samples/bpf: fix broken cgroup socket testing
In reply to Daniel T. LeeCurrently, executing test_cgrp2_sock2 fails due to wrong section header. This 'cgroup/sock1' style section is previously used at 'samples/bpf_load' (deprecated) BPF loader. Because this style isn't supported in libbpf, this commit fixes this problem by correcting the section header.
$ sudo ./test_cgrp2_sock2.sh
libbpf: prog 'bpf_prog1': missing BPF prog type, check ELF section name 'cgroup/sock1'
libbpf: prog 'bpf_prog1': failed to load: -22
libbpf: failed to load object './sock_flags_kern.o'
ERROR: loading BPF object file failedIn addition, this BPF program filters ping packets by comparing whether the socket type uses SOCK_RAW. However, after the ICMP socket[1] was developed, ping sends ICMP packets using SOCK_DGRAM. Therefore, in this commit, the packet filtering is changed to use SOCK_DGRAM instead of SOCK_RAW.
$ strace --trace socket ping -6 -c1 -w1 ::1
socket(AF_INET6, SOCK_DGRAM, IPPROTO_ICMPV6) = 3[1]: https://lwn.net/Articles/422330/
--- samples/bpf/sock_flags_kern.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
samples/bpf/sock_flags_kern.c +6 -6
@@ -5,7 +5,7 @@#include <uapi/linux/in6.h> #include <bpf/bpf_helpers.h> -SEC("cgroup/sock1") +SEC("cgroup/sock")int bpf_prog1 (struct bpf_sock *sk) {char fmt[] ="socket: family %d type %d protocol %d\n" ;
@@ -17,29 +17,29 @@ int bpf_prog1(struct bpf_sock *sk)bpf_trace_printk (fmt,sizeof (fmt), sk->family, sk->type, sk->protocol);bpf_trace_printk (fmt2,sizeof (fmt2), uid, gid); -/* block PF_INET6, SOCK_RAW, IPPROTO_ICMPV6 sockets +/* block PF_INET6, SOCK_DGRAM, IPPROTO_ICMPV6 sockets * ie., make ping6 fail */ if (sk->family ==PF_INET6 && - sk->type ==SOCK_RAW && + sk->type ==SOCK_DGRAM && sk->protocol ==IPPROTO_ICMPV6 )return 0 ;return 1 ; } -SEC("cgroup/sock2") +SEC("cgroup/sock")int bpf_prog2 (struct bpf_sock *sk) {char fmt[] ="socket: family %d type %d protocol %d\n" ;bpf_trace_printk (fmt,sizeof (fmt), sk->family, sk->type, sk->protocol); -/* block PF_INET, SOCK_RAW, IPPROTO_ICMP sockets +/* block PF_INET, SOCK_DGRAM, IPPROTO_ICMP sockets * ie., make ping fail */ if (sk->family ==PF_INET && - sk->type ==SOCK_RAW && + sk->type ==SOCK_DGRAM && sk->protocol ==IPPROTO_ICMP )return 0 ;
Signature · 1 line
2.34.1
PATCH 5/10 samples/bpf: replace broken overhead microbenchmark with fib_table_lookup
In reply to Daniel T. LeeThe test_overhead bpf program is designed to compare performance between tracepoint and kprobe. Initially it used task_rename and urandom_read tracepoint.
However, commit 14c174633f34 ("random: remove unused tracepoints") removed urandom_read tracepoint, and for this reason the test_overhead got broken.
This commit introduces new microbenchmark using fib_table_lookup. This microbenchmark sends UDP packets to localhost in order to invoke fib_table_lookup.
In a nutshell:
fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
addr.sin_addr.s_addr = inet_addr(DUMMY_IP);
addr.sin_port = htons(DUMMY_PORT);
for() {
sendto(fd, buf, strlen(buf), 0,
(struct sockaddr *)&addr, sizeof(addr));
}on 4 cpus in parallel:
lookup per sec
base (no tracepoints, no kprobes) 381k with kprobe at fib_table_lookup() 325k with tracepoint at fib:fib_table_lookup 330k with raw_tracepoint at fib:fib_table_lookup 365k
--- samples/bpf/test_overhead_kprobe_kern.c | 2 +- samples/bpf/test_overhead_raw_tp_kern.c | 2 +- samples/bpf/test_overhead_tp_kern.c | 26 ++++++++++++++++------- samples/bpf/test_overhead_user.c | 28 +++++++++++++++++-------- 4 files changed, 40 insertions(+), 18 deletions(-)
samples/bpf/test_overhead_kprobe_kern.c +1 -1
@@ -39,7 +39,7 @@ int prog(struct pt_regs *ctx)return 0 ; } -SEC("kprobe/urandom_read") +SEC("kprobe/fib_table_lookup")int prog2 (struct pt_regs *ctx) {return 0 ;
samples/bpf/test_overhead_raw_tp_kern.c +1 -1
@@ -9,7 +9,7 @@ int prog(struct bpf_raw_tracepoint_args *ctx)return 0 ; } -SEC("raw_tracepoint/urandom_read") +SEC("raw_tracepoint/fib_table_lookup")int prog2 (struct bpf_raw_tracepoint_args *ctx) {return 0 ;
samples/bpf/test_overhead_tp_kern.c +19 -7
@@ -22,15 +22,27 @@ int prog(struct task_rename *ctx)return 0 ; } -/* from /sys/kernel/debug/tracing/events/random/urandom_read/format */ -struct urandom_read { +/* from /sys/kernel/debug/tracing/events/fib/fib_table_lookup/format */ +struct fib_table_lookup {__u64 pad; -int got_bits; -int pool_left; -int input_left; +__u32 tb_id; +int err; +int oif; +int iif; +__u8 proto; +__u8 tos; +__u8 scope; +__u8 flags; +__u8 src[4 ]; +__u8 dst[4 ]; +__u8 gw4[4 ]; +__u8 gw6[16 ]; +__u16 sport; +__u16 dport; +char name[16 ]; }; -SEC("tracepoint/random/urandom_read") -int prog2 (struct urandom_read *ctx) +SEC("tracepoint/fib/fib_table_lookup") +int prog2 (struct fib_table_lookup *ctx) {return 0 ; }
samples/bpf/test_overhead_user.c +19 -9
@@ -11,6 +11,8 @@#include <unistd.h> #include <assert.h> #include <sys/wait.h> +#include <sys/socket.h> +#include <arpa/inet.h> #include <stdlib.h> #include <signal.h> #include <linux/bpf.h>
@@ -20,6 +22,8 @@#include <bpf/libbpf.h> #define MAX_CNT 1000000 +#define DUMMY_IP "127.0.0.1" +#define DUMMY_PORT 80static struct bpf_link *links[2 ];static struct bpf_object *obj;
@@ -35,8 +39,8 @@ static __u64 time_get_ns(void)static void test_task_rename (int cpu) { -__u64 start_time;char buf[] ="test\n" ; +__u64 start_time;int i, fd; fd =open ("/proc/self/comm" ,O_WRONLY |O_TRUNC );
@@ -57,26 +61,32 @@ static void test_task_rename(int cpu)close (fd); } -static void test_urandom_read (int cpu) +static void test_fib_table_lookup (int cpu) { +struct sockaddr_in addr; +char buf[] ="test\n" ;__u64 start_time; -char buf[4 ];int i, fd; - fd =open ("/dev/urandom" ,O_RDONLY ); + fd =socket (AF_INET ,SOCK_DGRAM ,IPPROTO_UDP );if (fd <0 ) { -printf ("couldn't open /dev/urandom\n" ); +printf ("couldn't open socket\n" );exit (1 ); } +memset ((char *)&addr,0 ,sizeof (addr)); + addr.sin_addr.s_addr =inet_addr (DUMMY_IP ); + addr.sin_port =htons (DUMMY_PORT ); + addr.sin_family =AF_INET ; start_time =time_get_ns ();for (i =0 ; i <MAX_CNT ; i++) { -if (read (fd, buf,sizeof (buf)) <0 ) { -printf ("failed to read from /dev/urandom: %s\n" ,strerror (errno)); +if (sendto (fd, buf,strlen (buf),0 , + (struct sockaddr *)&addr,sizeof (addr)) <0 ) { +printf ("failed to start ping: %s\n" ,strerror (errno));close (fd);return ; } } -printf ("urandom_read:%d: %lld events per sec\n" , +printf ("fib_table_lookup:%d: %lld events per sec\n" , cpu,MAX_CNT *1000000000ll / (time_get_ns () - start_time));close (fd); }
@@ -92,7 +102,7 @@ static void loop(int cpu, int flags)if (flags &1 )test_task_rename (cpu);if (flags &2 ) -test_urandom_read (cpu); +test_fib_table_lookup (cpu); }static void run_perf_test (int tasks,int flags)
Signature · 1 line
2.34.1
PATCH 6/10 samples/bpf: replace legacy map with the BTF-defined map
In reply to Daniel T. LeeWith libbpf 1.0 release, support for legacy BPF map declaration syntax had been dropped. If you run a program using legacy BPF in the latest libbpf, the following error will be output.
libbpf: map 'lwt_len_hist_map' (legacy): legacy map definitions are deprecated, use BTF-defined maps instead
libbpf: Use of BPF_ANNOTATE_KV_PAIR is deprecated, use BTF-defined maps in .maps section instead
This commit replaces legacy map with the BTF-defined map.
--- samples/bpf/lwt_len_hist_kern.c | 24 +++++++----------------- samples/bpf/test_cgrp2_tc_kern.c | 25 +++++++------------------ 2 files changed, 14 insertions(+), 35 deletions(-)
samples/bpf/lwt_len_hist_kern.c +7 -17
@@ -16,23 +16,13 @@#include <uapi/linux/in.h> #include <bpf/bpf_helpers.h> -struct bpf_elf_map { -__u32 type; -__u32 size_key; -__u32 size_value; -__u32 max_elem; -__u32 flags; -__u32 id; -__u32 pinning; -}; - -struct bpf_elf_map SEC("maps") lwt_len_hist_map = { - .type = BPF_MAP_TYPE_PERCPU_HASH, - .size_key = sizeof(__u64), - .size_value = sizeof(__u64), - .pinning = 2, - .max_elem = 1024, -}; +struct { + __uint(type, BPF_MAP_TYPE_PERCPU_HASH); + __type(key, u64); + __type(value, u64); + __uint(pinning, LIBBPF_PIN_BY_NAME); + __uint(max_entries, 1024); +} lwt_len_hist_map SEC(".maps");static unsigned int log2 (unsigned int v) {
samples/bpf/test_cgrp2_tc_kern.c +7 -18
@@ -19,24 +19,13 @@ struct eth_hdr {unsigned short h_proto; }; -#define PIN_GLOBAL_NS 2 -struct bpf_elf_map { -__u32 type; -__u32 size_key; -__u32 size_value; -__u32 max_elem; -__u32 flags; -__u32 id; -__u32 pinning; -}; - -struct bpf_elf_map SEC("maps") test_cgrp2_array_pin = { - .type = BPF_MAP_TYPE_CGROUP_ARRAY, - .size_key = sizeof(uint32_t), - .size_value = sizeof(uint32_t), - .pinning = PIN_GLOBAL_NS, - .max_elem = 1, -}; +struct { + __uint(type, BPF_MAP_TYPE_CGROUP_ARRAY); + __type(key, u32); + __type(value, u32); + __uint(pinning, LIBBPF_PIN_BY_NAME); + __uint(max_entries, 1); +} test_cgrp2_array_pin SEC(".maps"); SEC("filter")int handle_egress (struct __sk_buff *skb)
Signature · 1 line
2.34.1
PATCH 7/10 samples/bpf: split common macros to net_shared.h
In reply to Daniel T. LeeCurrently, many programs under sample/bpf often include individual macros by directly including the header under "linux/" rather than using the "vmlinux.h" header.
However, there are some problems with migrating to "vmlinux.h" because there is no definition for utility functions such as endianness conversion (ntohs/htons). Fortunately, the xdp_sample program already has a function that can be replaced to solve this problem.
Therefore, this commit attempts to separate these functions into a file called net_shared.h to make them universally available. Additionally, this file includes network-related macros that are not defined in "vmlinux.h". (inspired by 'selftests' bpf_tracing_net.h)
--- samples/bpf/net_shared.h | 26 ++++++++++++++++++++++++++ samples/bpf/xdp_sample.bpf.h | 22 +--------------------- 2 files changed, 27 insertions(+), 21 deletions(-) create mode 100644 samples/bpf/net_shared.h
samples/bpf/net_shared.h new file +26 -0
@@ -0,0 +1,26 @@ +// SPDX-License-Identifier: GPL-2.0 +#ifndef _NET_SHARED_H +#define _NET_SHARED_H + +#define ETH_ALEN 6 +#define ETH_P_802_3_MIN 0x0600 +#define ETH_P_8021Q 0x8100 +#define ETH_P_8021AD 0x88A8 +#define ETH_P_IP 0x0800 +#define ETH_P_IPV6 0x86DD +#define ETH_P_ARP 0x0806 +#define IPPROTO_ICMPV6 58 + +#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && \ + __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ +#define bpf_ntohs (x) __builtin_bswap16(x) +#define bpf_htons (x) __builtin_bswap16(x) +#elif defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && \ + __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ +#define bpf_ntohs (x) (x) +#define bpf_htons (x) (x) +#else +# error "Endianness detection needs to be set up for your compiler?!" +#endif + +#endif
samples/bpf/xdp_sample.bpf.h +1 -21
@@ -7,17 +7,9 @@#include <bpf/bpf_core_read.h> #include <bpf/bpf_helpers.h> +#include "net_shared.h" #include "xdp_sample_shared.h" -#define ETH_ALEN 6 -#define ETH_P_802_3_MIN 0x0600 -#define ETH_P_8021Q 0x8100 -#define ETH_P_8021AD 0x88A8 -#define ETH_P_IP 0x0800 -#define ETH_P_IPV6 0x86DD -#define ETH_P_ARP 0x0806 -#define IPPROTO_ICMPV6 58 -#define EINVAL 22#define ENETDOWN 100#define EMSGSIZE 90
@@ -55,18 +47,6 @@ static __always_inline void swap_src_dst_mac(void *data) p[5 ] = dst[2 ]; } -#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && \ - __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ -#define bpf_ntohs (x) __builtin_bswap16(x) -#define bpf_htons (x) __builtin_bswap16(x) -#elif defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && \ - __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ -#define bpf_ntohs (x) (x) -#define bpf_htons (x) (x) -#else -# error "Endianness detection needs to be set up for your compiler?!" -#endif -/* * Note: including linux/compiler.h or linux/kernel.h for the macros below * conflicts with vmlinux.h include in BPF files, so we define them here.
Signature · 1 line
2.34.1
PATCH 8/10 samples/bpf: replace BPF programs header with net_shared.h
In reply to Daniel T. LeeThis commit applies "net_shared.h" to BPF programs to remove existing network related header dependencies. Also, this commit removes unnecessary headers before applying "vmlinux.h" to the BPF programs.
Mostly, endianness conversion function has been applied to the source. In addition, several macros have been defined to fulfill the INET, TC-related constants.
--- samples/bpf/net_shared.h | 6 ++++++ samples/bpf/sock_flags_kern.c | 10 +++++----- samples/bpf/test_cgrp2_tc_kern.c | 6 ++---- samples/bpf/test_lwt_bpf.c | 3 ++- 4 files changed, 15 insertions(+), 10 deletions(-)
samples/bpf/net_shared.h +6 -0
@@ -2,6 +2,9 @@#ifndef _NET_SHARED_H#define _NET_SHARED_H +#define AF_INET 2 +#define AF_INET6 10 +#define ETH_ALEN 6#define ETH_P_802_3_MIN 0x0600#define ETH_P_8021Q 0x8100
@@ -11,6 +14,9 @@#define ETH_P_ARP 0x0806#define IPPROTO_ICMPV6 58 +#define TC_ACT_OK 0 +#define TC_ACT_SHOT 2 +#if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && \ __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__#define bpf_ntohs (x) __builtin_bswap16(x)
samples/bpf/sock_flags_kern.c +5 -5
@@ -1,5 +1,5 @@ +#include "net_shared.h" #include <uapi/linux/bpf.h> -#include <linux/socket.h> #include <linux/net.h> #include <uapi/linux/in.h> #include <uapi/linux/in6.h>
@@ -17,10 +17,10 @@ int bpf_prog1(struct bpf_sock *sk)bpf_trace_printk (fmt,sizeof (fmt), sk->family, sk->type, sk->protocol);bpf_trace_printk (fmt2,sizeof (fmt2), uid, gid); -/* block PF_INET6, SOCK_DGRAM, IPPROTO_ICMPV6 sockets +/* block AF_INET6, SOCK_DGRAM, IPPROTO_ICMPV6 sockets * ie., make ping6 fail */ -if (sk->family ==PF_INET6 && +if (sk->family ==AF_INET6 && sk->type ==SOCK_DGRAM && sk->protocol ==IPPROTO_ICMPV6 )return 0 ;
@@ -35,10 +35,10 @@ int bpf_prog2(struct bpf_sock *sk)bpf_trace_printk (fmt,sizeof (fmt), sk->family, sk->type, sk->protocol); -/* block PF_INET, SOCK_DGRAM, IPPROTO_ICMP sockets +/* block AF_INET, SOCK_DGRAM, IPPROTO_ICMP sockets * ie., make ping fail */ -if (sk->family ==PF_INET && +if (sk->family ==AF_INET && sk->type ==SOCK_DGRAM && sk->protocol ==IPPROTO_ICMP )return 0 ;
samples/bpf/test_cgrp2_tc_kern.c +2 -4
@@ -5,10 +5,8 @@* License as published by the Free Software Foundation. */ #define KBUILD_MODNAME "foo" -#include <uapi/linux/if_ether.h> -#include <uapi/linux/in6.h> +#include "net_shared.h" #include <uapi/linux/ipv6.h> -#include <uapi/linux/pkt_cls.h> #include <uapi/linux/bpf.h> #include <bpf/bpf_helpers.h>
@@ -42,7 +40,7 @@ int handle_egress(struct __sk_buff *skb)if (data +sizeof (*eth) +sizeof (*ip6h) > data_end)return TC_ACT_OK ; -if (eth->h_proto !=htons (ETH_P_IPV6 ) || +if (eth->h_proto !=bpf_htons (ETH_P_IPV6 ) || ip6h->nexthdr !=IPPROTO_ICMPV6 ) {bpf_trace_printk (dont_care_msg,sizeof (dont_care_msg), eth->h_proto, ip6h->nexthdr);
samples/bpf/test_lwt_bpf.c +2 -1
@@ -10,6 +10,7 @@* General Public License for more details. */ +#include "net_shared.h" #include <stdint.h> #include <stddef.h> #include <linux/bpf.h>
@@ -176,7 +177,7 @@ static inline int __do_push_ll_and_redirect(struct __sk_buff *skb)printk ("skb_change_head() failed: %d" , ret); } - ehdr.h_proto =__constant_htons (ETH_P_IP ); + ehdr.h_proto =bpf_htons (ETH_P_IP );memcpy (&ehdr.h_source, &smac,6 );memcpy (&ehdr.h_dest, &dmac,6 );
Signature · 1 line
2.34.1
PATCH 9/10 samples/bpf: use vmlinux.h instead of implicit headers in BPF test program
In reply to Daniel T. LeeThis commit applies vmlinux.h to BPF functionality testing program. Macros that were not defined despite migration to "vmlinux.h" were defined separately in individual files.
--- samples/bpf/lwt_len_hist_kern.c | 5 +---- samples/bpf/sock_flags_kern.c | 6 ++---- samples/bpf/test_cgrp2_tc_kern.c | 3 +-- samples/bpf/test_lwt_bpf.c | 11 +---------- samples/bpf/test_map_in_map_kern.c | 7 ++++--- samples/bpf/test_overhead_kprobe_kern.c | 4 +--- samples/bpf/test_overhead_raw_tp_kern.c | 2 +- samples/bpf/test_overhead_tp_kern.c | 3 +-- 8 files changed, 12 insertions(+), 29 deletions(-)
samples/bpf/lwt_len_hist_kern.c +1 -4
@@ -10,10 +10,7 @@* General Public License for more details. */ -#include <uapi/linux/bpf.h> -#include <uapi/linux/if_ether.h> -#include <uapi/linux/ip.h> -#include <uapi/linux/in.h> +#include "vmlinux.h" #include <bpf/bpf_helpers.h> struct {
samples/bpf/sock_flags_kern.c +2 -4
@@ -1,8 +1,6 @@ +// SPDX-License-Identifier: GPL-2.0 +#include "vmlinux.h" #include "net_shared.h" -#include <uapi/linux/bpf.h> -#include <linux/net.h> -#include <uapi/linux/in.h> -#include <uapi/linux/in6.h> #include <bpf/bpf_helpers.h> SEC ("cgroup/sock" )
samples/bpf/test_cgrp2_tc_kern.c +1 -2
@@ -5,9 +5,8 @@* License as published by the Free Software Foundation. */ #define KBUILD_MODNAME "foo" +#include "vmlinux.h" #include "net_shared.h" -#include <uapi/linux/ipv6.h> -#include <uapi/linux/bpf.h> #include <bpf/bpf_helpers.h> /* copy of 'struct ethhdr' without __packed */
samples/bpf/test_lwt_bpf.c +1 -10
@@ -10,17 +10,8 @@* General Public License for more details. */ +#include "vmlinux.h" #include "net_shared.h" -#include <stdint.h> -#include <stddef.h> -#include <linux/bpf.h> -#include <linux/ip.h> -#include <linux/in.h> -#include <linux/in6.h> -#include <linux/tcp.h> -#include <linux/udp.h> -#include <linux/icmpv6.h> -#include <linux/if_ether.h> #include <bpf/bpf_helpers.h> #include <string.h>
samples/bpf/test_map_in_map_kern.c +4 -3
@@ -6,16 +6,17 @@* License as published by the Free Software Foundation. */ #define KBUILD_MODNAME "foo" -#include <linux/ptrace.h> +#include "vmlinux.h" #include <linux/version.h> -#include <uapi/linux/bpf.h> -#include <uapi/linux/in6.h> #include <bpf/bpf_helpers.h> #include <bpf/bpf_tracing.h> #include <bpf/bpf_core_read.h> #define MAX_NR_PORTS 65536 +#define EINVAL 22 +#define ENOENT 2 +/* map #0 */ struct inner_a { __uint(type, BPF_MAP_TYPE_ARRAY);
samples/bpf/test_overhead_kprobe_kern.c +1 -3
@@ -4,10 +4,8 @@* modify it under the terms of version 2 of the GNU General Public * License as published by the Free Software Foundation. */ +#include "vmlinux.h" #include <linux/version.h> -#include <linux/ptrace.h> -#include <linux/sched.h> -#include <uapi/linux/bpf.h> #include <bpf/bpf_helpers.h> #include <bpf/bpf_tracing.h>
samples/bpf/test_overhead_raw_tp_kern.c +1 -1
@@ -1,6 +1,6 @@// SPDX-License-Identifier: GPL-2.0 /* Copyright (c) 2018 Facebook */ -#include <uapi/linux/bpf.h> +#include "vmlinux.h" #include <bpf/bpf_helpers.h> SEC ("raw_tracepoint/task_rename" )
samples/bpf/test_overhead_tp_kern.c +1 -2
@@ -4,8 +4,7 @@* modify it under the terms of version 2 of the GNU General Public * License as published by the Free Software Foundation. */ -#include <linux/sched.h> -#include <uapi/linux/bpf.h> +#include "vmlinux.h" #include <bpf/bpf_helpers.h> /* from /sys/kernel/debug/tracing/events/task/task_rename/format */
Signature · 1 line
2.34.1
PATCH 10/10 samples/bpf: change _kern suffix to .bpf with BPF test programs
In reply to Daniel T. LeeThis commit changes the _kern suffix to .bpf with the BPF test programs. With this modification, test programs will inherit the benefit of the new CLANG-BPF compile target.
---
samples/bpf/Makefile | 14 +++++++-------
.../{lwt_len_hist_kern.c => lwt_len_hist.bpf.c} | 0
samples/bpf/lwt_len_hist.sh | 2 +-
.../bpf/{sock_flags_kern.c => sock_flags.bpf.c} | 0
samples/bpf/test_cgrp2_sock2.sh | 2 +-
.../{test_cgrp2_tc_kern.c => test_cgrp2_tc.bpf.c} | 0
samples/bpf/test_cgrp2_tc.sh | 2 +-
...est_map_in_map_kern.c => test_map_in_map.bpf.c} | 0
samples/bpf/test_map_in_map_user.c | 2 +-
...ad_kprobe_kern.c => test_overhead_kprobe.bpf.c} | 0
...ad_raw_tp_kern.c => test_overhead_raw_tp.bpf.c} | 0
...t_overhead_tp_kern.c => test_overhead_tp.bpf.c} | 0
samples/bpf/test_overhead_user.c | 6 +++---
13 files changed, 14 insertions(+), 14 deletions(-)
rename samples/bpf/{lwt_len_hist_kern.c => lwt_len_hist.bpf.c} (100%)
rename samples/bpf/{sock_flags_kern.c => sock_flags.bpf.c} (100%)
rename samples/bpf/{test_cgrp2_tc_kern.c => test_cgrp2_tc.bpf.c} (100%)
rename samples/bpf/{test_map_in_map_kern.c => test_map_in_map.bpf.c} (100%)
rename samples/bpf/{test_overhead_kprobe_kern.c => test_overhead_kprobe.bpf.c} (100%)
rename samples/bpf/{test_overhead_raw_tp_kern.c => test_overhead_raw_tp.bpf.c} (100%)
rename samples/bpf/{test_overhead_tp_kern.c => test_overhead_tp.bpf.c} (100%)samples/bpf/Makefile +7 -7
@@ -131,7 +131,7 @@ always-y += tracex4_kern.o always-y += tracex5_kern.o always-y += tracex6_kern.o always-y += tracex7_kern.o -always-y += sock_flags_kern.o +always-y += sock_flags.bpf.o always-y += test_probe_write_user.bpf.o always-y += trace_output.bpf.o always-y += tcbpf1_kern.o
@@ -140,19 +140,19 @@ always-y += lathist_kern.o always-y += offwaketime_kern.o always-y += spintest_kern.o always-y += map_perf_test.bpf.o -always-y += test_overhead_tp_kern.o -always-y += test_overhead_raw_tp_kern.o -always-y += test_overhead_kprobe_kern.o +always-y += test_overhead_tp.bpf.o +always-y += test_overhead_raw_tp.bpf.o +always-y += test_overhead_kprobe.bpf.o always-y += parse_varlen.o parse_simple.o parse_ldabs.o -always-y += test_cgrp2_tc_kern.o +always-y += test_cgrp2_tc.bpf.o always-y += xdp1_kern.o always-y += xdp2_kern.o always-y += test_current_task_under_cgroup.bpf.o always-y += trace_event_kern.o always-y += sampleip_kern.o -always-y += lwt_len_hist_kern.o +always-y += lwt_len_hist.bpf.o always-y += xdp_tx_iptunnel_kern.o -always-y += test_map_in_map_kern.o +always-y += test_map_in_map.bpf.o always-y += tcp_synrto_kern.o always-y += tcp_rwnd_kern.o always-y += tcp_bufs_kern.o
samples/bpf/lwt_len_hist.bpf.c renamed from samples/bpf/lwt_len_hist_kern.c +0 -0
samples/bpf/lwt_len_hist.sh +1 -1
@@ -4,7 +4,7 @@ NS1=lwt_ns1 VETH0=tst_lwt1a VETH1=tst_lwt1b -BPF_PROG=lwt_len_hist_kern.o +BPF_PROG=lwt_len_hist.bpf.o TRACE_ROOT=/sys/kernel/debug/tracing function cleanup {
samples/bpf/sock_flags.bpf.c renamed from samples/bpf/sock_flags_kern.c +0 -0
samples/bpf/test_cgrp2_sock2.sh +1 -1
@@ -5,7 +5,7 @@ BPFFS=/sys/fs/bpf MY_DIR=$(dirname $0) TEST=$MY_DIR/test_cgrp2_sock2 LINK_PIN=$BPFFS/test_cgrp2_sock2 -BPF_PROG=$MY_DIR/sock_flags_kern.o +BPF_PROG=$MY_DIR/sock_flags.bpf.o function config_device { ip netns add at_ns0
samples/bpf/test_cgrp2_tc.bpf.c renamed from samples/bpf/test_cgrp2_tc_kern.c +0 -0
samples/bpf/test_cgrp2_tc.sh +1 -1
@@ -4,7 +4,7 @@ MY_DIR=$(dirname $0) # Details on the bpf prog BPF_CGRP2_ARRAY_NAME='test_cgrp2_array_pin' -BPF_PROG="$MY_DIR/test_cgrp2_tc_kern.o" +BPF_PROG="$MY_DIR/test_cgrp2_tc.bpf.o" BPF_SECTION='filter' [ -z "$TC" ] && TC='tc'
samples/bpf/test_map_in_map.bpf.c renamed from samples/bpf/test_map_in_map_kern.c +0 -0
samples/bpf/test_map_in_map_user.c +1 -1
@@ -120,7 +120,7 @@ int main(int argc, char **argv)struct bpf_object *obj;char filename[256 ]; -snprintf (filename,sizeof (filename),"%s_kern.o" , argv[0 ]); +snprintf (filename,sizeof (filename),"%s.bpf.o" , argv[0 ]); obj =bpf_object__open_file (filename,NULL );if (libbpf_get_error (obj)) {fprintf (stderr,"ERROR: opening BPF object file failed\n" );
samples/bpf/test_overhead_kprobe.bpf.c renamed from samples/bpf/test_overhead_kprobe_kern.c +0 -0
samples/bpf/test_overhead_raw_tp.bpf.c renamed from samples/bpf/test_overhead_raw_tp_kern.c +0 -0
samples/bpf/test_overhead_tp.bpf.c renamed from samples/bpf/test_overhead_tp_kern.c +0 -0
samples/bpf/test_overhead_user.c +3 -3
@@ -189,7 +189,7 @@ int main(int argc, char **argv)if (test_flags &0xC ) {snprintf (filename,sizeof (filename), -"%s_kprobe_kern.o" , argv[0 ]); +"%s_kprobe.bpf.o" , argv[0 ]);printf ("w/KPROBE\n" ); err =load_progs (filename);
@@ -201,7 +201,7 @@ int main(int argc, char **argv)if (test_flags &0x30 ) {snprintf (filename,sizeof (filename), -"%s_tp_kern.o" , argv[0 ]); +"%s_tp.bpf.o" , argv[0 ]);printf ("w/TRACEPOINT\n" ); err =load_progs (filename);if (!err)
@@ -212,7 +212,7 @@ int main(int argc, char **argv)if (test_flags &0xC0 ) {snprintf (filename,sizeof (filename), -"%s_raw_tp_kern.o" , argv[0 ]); +"%s_raw_tp.bpf.o" , argv[0 ]);printf ("w/RAW_TRACEPOINT\n" ); err =load_progs (filename);if (!err)
Signature · 1 line
2.34.1
Quoted 17 lines · Daniel T. Lee
Currently, there are many programs under
samples/bpfto test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
-bpf_trace_printk()always append newline at the end [2]
- deprecated styled BPF section header (bpf_loadstyle) [3]
-urandom_readtracepoint is removed (used for testing overhead) [4]
- ping sends packet withSOCK_DGRAMinstead ofSOCK_RAW[5]*
- use "vmlinux.h" instead of including individual headers
Quoted 4 lines
In addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from
xdp_sampleand bunch of constants)
Nice set of cleanups. Applied.
As a follow up could you convert some of them to proper selftests/bpf ? Unfortunately samples/bpf will keep bit rotting despite your herculean efforts.
Quoted 19 lines · Alexei Starovoitov
Quoted 17 lines · Daniel T. Lee
Currently, there are many programs under
samples/bpfto test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
-bpf_trace_printk()always append newline at the end [2]
- deprecated styled BPF section header (bpf_loadstyle) [3]
-urandom_readtracepoint is removed (used for testing overhead) [4]
- ping sends packet withSOCK_DGRAMinstead ofSOCK_RAW[5]*
- use "vmlinux.h" instead of including individual headers
Quoted 8 lines
Quoted 4 lines
In addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from
xdp_sampleand bunch of constants)Nice set of cleanups. Applied.
As a follow up could you convert some of them to proper selftests/bpf ? Unfortunatelysamples/bpfwill keep bit rotting despite your herculean efforts.
I really appreciate for your compliment!
I'll try to convert the existing sample to selftest in the next patch.
Signature · 2 lines
Best, Daniel T. Lee
Quoted 26 lines · Daniel T. Lee
Quoted 24 lines · Alexei Starovoitov
Quoted 21 lines · Daniel T. Lee
Currently, there are many programs under
samples/bpfto test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
-bpf_trace_printk()always append newline at the end [2]
- deprecated styled BPF section header (bpf_loadstyle) [3]
-urandom_readtracepoint is removed (used for testing overhead) [4]
- ping sends packet withSOCK_DGRAMinstead ofSOCK_RAW[5]*
- use "vmlinux.h" instead of including individual headersIn addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from
xdp_sampleand bunch of constants)
Quoted 6 lines
Quoted 3 lines
Nice set of cleanups. Applied.
As a follow up could you convert some of them to proper selftests/bpf ? Unfortunatelysamples/bpfwill keep bit rotting despite your herculean efforts.I really appreciate for your compliment!
I'll try to convert the existing sample to selftest in the next patch.
Maybe this is a good time to mention that we recently ported some of the XDP utilities from samples/bpf to xdp-tools, in the form of the 'xdp-bench' utility:
https://github.com/xdp-project/xdp-tools/tree/master/xdp-bench
It's basically a combination of all the xdp_redirect* samples, but also includes the functionality from the xdp_rxq_info sample program (i.e., the ability to monitor RXQs and use other return codes).
I'm planning to submit a patch to remove those utilities from samples/bpf after we tag the next release of xdp-tools (have one or two outstanding issues to clear before we do that), but wanted to give you a head's up so you don't spend any time on those particular utilities when you're cleaning up samples :)
-Toke
Daniel Borkmann
maintainerQuoted 26 lines · Toke Høiland-Jørgensen
Quoted 25 lines · Daniel T. Lee
Quoted 23 lines · Alexei Starovoitov
Quoted 21 lines · Daniel T. Lee
Currently, there are many programs under
samples/bpfto test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
-bpf_trace_printk()always append newline at the end [2]
- deprecated styled BPF section header (bpf_loadstyle) [3]
-urandom_readtracepoint is removed (used for testing overhead) [4]
- ping sends packet withSOCK_DGRAMinstead ofSOCK_RAW[5]*
- use "vmlinux.h" instead of including individual headersIn addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from
xdp_sampleand bunch of constants)
Quoted 6 lines
Quoted 6 lines
Quoted 3 lines
Nice set of cleanups. Applied.
As a follow up could you convert some of them to proper selftests/bpf ? Unfortunatelysamples/bpfwill keep bit rotting despite your herculean efforts.I really appreciate for your compliment!
I'll try to convert the existing sample to selftest in the next patch.
This would be awesome, thanks a lot Daniel!
Quoted 9 lines
Maybe this is a good time to mention that we recently ported some of the XDP utilities from
samples/bpfto xdp-tools, in the form of the 'xdp-bench' utility:
https://github.com/xdp-project/xdp-tools/tree/master/xdp-benchIt's basically a combination of all the xdp_redirect* samples, but also includes the functionality from the
xdp_rxq_infosample program (i.e., the ability to monitor RXQs and use other return codes).
Quoted 5 lines
I'm planning to submit a patch to remove those utilities from
samples/bpfafter we tag the next release of xdp-tools (have one or two outstanding issues to clear before we do that), but wanted to give you a head's up so you don't spend any time on those particular utilities when you're cleaning up samples :)
Nice! Once we're through with most relevant ones from samples/bpf, it would be great to only have a readme in that dir (and that's really all) with pointers for developers on how to get started.. including BPF selftests, xdp tools, links to ebpf.io/applications and ebpf.io/infrastructure, etc where more resources can be found, essentially a small getting started doc for BPF devs.
Quoted 51 lines · Daniel Borkmann
Quoted 26 lines · Toke Høiland-Jørgensen
Quoted 25 lines · Daniel T. Lee
Quoted 23 lines · Alexei Starovoitov
Quoted 21 lines · Daniel T. Lee
Currently, there are many programs under
samples/bpfto test the various functionality of BPF that have been developed for a long time. However, the kernel (BPF) has changed a lot compared to the 2016 when some of these test programs were first introduced.Therefore, some of these programs use the deprecated function of BPF, and some programs no longer work normally due to changes in the API.
To list some of the kernel changes that this patch set is focusing on,
- legacy BPF map declaration syntax support had been dropped [1]
-bpf_trace_printk()always append newline at the end [2]
- deprecated styled BPF section header (bpf_loadstyle) [3]
-urandom_readtracepoint is removed (used for testing overhead) [4]
- ping sends packet withSOCK_DGRAMinstead ofSOCK_RAW[5]*
- use "vmlinux.h" instead of including individual headersIn addition to this, this patchset tries to modernize the existing testing scripts a bit. And for network-related testing programs, a separate header file was created and applied. (To use the Endianness conversion function from
xdp_sampleand bunch of constants)Quoted 6 lines
Quoted 6 lines
Quoted 3 lines
Nice set of cleanups. Applied.
As a follow up could you convert some of them to proper selftests/bpf ? Unfortunatelysamples/bpfwill keep bit rotting despite your herculean efforts.I really appreciate for your compliment!
I'll try to convert the existing sample to selftest in the next patch.This would be awesome, thanks a lot Daniel!
Quoted 9 lines
Maybe this is a good time to mention that we recently ported some of the XDP utilities from
samples/bpfto xdp-tools, in the form of the 'xdp-bench' utility:
https://github.com/xdp-project/xdp-tools/tree/master/xdp-benchIt's basically a combination of all the xdp_redirect* samples, but also includes the functionality from the
xdp_rxq_infosample program (i.e., the ability to monitor RXQs and use other return codes).Quoted 5 lines
I'm planning to submit a patch to remove those utilities from
samples/bpfafter we tag the next release of xdp-tools (have one or two outstanding issues to clear before we do that), but wanted to give you a head's up so you don't spend any time on those particular utilities when you're cleaning up samples :)
Quoted 5 lines
Nice! Once we're through with most relevant ones from
samples/bpf, it would be great to only have a readme in that dir (and that's really all) with pointers for developers on how to get started.. including BPF selftests, xdp tools, links to ebpf.io/applications and ebpf.io/infrastructure, etc where more resources can be found, essentially a small getting started doc for BPF devs.
Yeah, good point! Will keep that in mind :)
-Toke