monitor network traffic for flaky test cases
Subsystems (3)
v2 cover monitor network traffic for flaky test cases
Capture packets in the background for flaky test cases related to network features.
We have some flaky test cases that are difficult to debug without knowing what the traffic looks like. Capturing packets, the CI log and packet files may help developers to fix these flaky test cases.
This patch set monitors a few test cases. Recently, they have been showing flaky behavior.
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK
Packet file: packets-2172-86.log
#280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OK The above block is the log of a test case. It shows every packets of a connection. The captured packets are stored in the file called packets-2172-86.log.
The following block is an example that monitors the network traffic of a test case. This test is running in the network namespace "testns". You can pass NULL to traffic_monitor_start() if the entire test, from traffic_monitor_start() to traffic_monitor_stop(), is running in the same namespace.
struct tmonitor_ctx *tmon;
...
tmon = traffic_monitor_start("testns");
ASSERT_TRUE(tmon, "traffic_monitor_start");
... test ...
traffic_monitor_stop(tmon);traffic_monitor_start() may fail, but we just ignore it since the failure doesn't affect the following main test.
This feature is enabled only if BPF selftests are built with TRAFFIC_MONITOR variable being defined. For example,
make TRAFFIC_MONITOR=1 -C tools/testing/selftests/bpf
This command will enable traffic monitoring for BPF selftests. That means we have to turn it on to get the log at CI.
---
Changes from v1:
- Initialize log_fd in traffic_monitor_start().
- Remove redundant including.
v1: https://lore.kernel.org/all/[email protected]/
Kui-Feng Lee (4):
selftests/bpf: Add traffic monitor functions. selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime. selftests/bpf: Monitor traffic for sockmap_listen. selftests/bpf: Monitor traffic for select_reuseport.
tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + .../bpf/prog_tests/select_reuseport.c | 7 + .../selftests/bpf/prog_tests/sockmap_listen.c | 8 + .../selftests/bpf/prog_tests/tc_redirect.c | 5 + 6 files changed, 423 insertions(+)
Signature · 1 line
2.34.1
PATCH v2 1/4 selftests/bpf: Add traffic monitor functions.
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call traffic_monitor_start() to start a thread to capture packets in the background for a given namespace and call traffic_monitor_stop() to stop capturing.
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK
IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK
Packet file: packets-2172-86.log
#280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKThe above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory /tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.
This feature only works if TRAFFIC_MONITOR variable has been passed to build BPF selftests. For example,
make TRAFFIC_MONITOR=1 -C tools/testing/selftests/bpf
This command will build BPF selftests with this feature enabled.
--- tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + 3 files changed, 403 insertions(+)
tools/testing/selftests/bpf/Makefile +5 -0
@@ -41,6 +41,11 @@ CFLAGS += -g $(OPT_FLAGS) -rdynamic \ LDFLAGS += $(SAN_LDFLAGS) LDLIBS += $(LIBELF_LIBS) -lz -lrt -lpthread +ifneq ($(TRAFFIC_MONITOR),) +LDLIBS += -lpcap +CFLAGS += -DTRAFFIC_MONITOR=1 +endif + # The following tests perform type punning and they may break strict # aliasing rules, which are exploited by both GCC and clang by default # while optimizing. This can lead to broken programs.
tools/testing/selftests/bpf/network_helpers.c +382 -0
400 lines not on this page · raw message
tools/testing/selftests/bpf/network_helpers.h +16 -0
@@ -82,6 +82,22 @@ int get_socket_local_port(int sock_fd);int get_hw_ring_size (char *ifname,struct ethtool_ringparam *ring_param);int set_hw_ring_size (char *ifname,struct ethtool_ringparam *ring_param); +struct tmonitor_ctx ; + +#ifdef TRAFFIC_MONITOR +struct tmonitor_ctx *traffic_monitor_start (const char *netns); +void traffic_monitor_stop (struct tmonitor_ctx *ctx); +#else +static inline struct tmonitor_ctx *traffic_monitor_start (const char *netns) +{ +return (struct tmonitor_ctx *)-1 ; +} + +static inline void traffic_monitor_stop (struct tmonitor_ctx *ctx) +{ +} +#endif +struct nstoken ;/** * open_netns() - Switch to specified network namespace by name.
Signature · 1 line
2.34.1
Re: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
Quoted 62 lines · Kui-Feng Lee
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call
traffic_monitor_start()to start a thread to capture packets in the background for a given namespace and calltraffic_monitor_stop()to stop capturing.IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log #280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKThe above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory
/tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.This feature only works if
TRAFFIC_MONITORvariable has been passed to build BPF selftests. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpfThis command will build BPF selftests with this feature enabled.
Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + 3 files changed, 403 insertions(+)tools/testing/selftests/bpf/Makefile +5
-0@@ -41,6 +41,11 @@ CFLAGS += -g $(OPT_FLAGS) -rdynamic \ LDFLAGS += $(SAN_LDFLAGS) LDLIBS += $(LIBELF_LIBS) -lz -lrt -lpthread +ifneq ($(TRAFFIC_MONITOR),) +LDLIBS += -lpcap +CFLAGS += -DTRAFFIC_MONITOR=1 +endif + # The following tests perform type punning and they may break strict # aliasing rules, which are exploited by both GCC and clang by default # while optimizing. This can lead to broken programs.tools/testing/selftests/bpf/network_helpers.c +1
-0@@ -10,6 +10,7 @@#include <arpa/inet.h> #include <sys/mount.h> +#include <sys/select.h> #include <sys/stat.h> #include <sys/un.h>
tools/testing/selftests/bpf/network_helpers.c line 28
Quoted 10 lines
tools/testing/selftests/bpf/network_helpers.c +7
-0@@ -18,6 +19,14 @@#include <linux/in6.h> #include <linux/limits.h> +#include <netinet/udp.h> + +#include <pthread.h> +/* Prevent pcap.h from including pcap/bpf.h and causing conflicts */ +#define PCAP_DONT_INCLUDE_PCAP_BPF_H 1 +#include <pcap/pcap.h> +#include <pcap/dlt.h>
Just noticed that the above lines fails on an environment without pcap installed. It will be fixed in next version.
Quoted 411 lines
tools/testing/selftests/bpf/network_helpers.c +374
-0+#include "bpf_util.h" #include "network_helpers.h" #include "test_progs.h" @@ -575,6 +584,379 @@ int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param)return 0 ; } +#ifdef TRAFFIC_MONITOR +struct tmonitor_ctx { +pcap_t *pcap; +pcap_dumper_t *dumper; +pthread_t thread; +int wake_fd_r; +int wake_fd_w; + +bool done; +char pkt_fname[PATH_MAX ]; +int pcap_fd; +}; + +/* Is this packet captured with a Ethernet protocol type? */ +static bool is_ethernet (const u_char *packet) +{ +u16 arphdr_type; + +memcpy (&arphdr_type, packet +8 ,2 ); + arphdr_type =ntohs (arphdr_type); + +/* Except the following cases, the protocol type contains the +* Ethernet protocol type for the packet. +* +* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +*/ +switch (arphdr_type) { +case 770 :/* ARPHRD_FRAD */ +case 778 :/* ARPHDR_IPGRE */ +case 803 :/* ARPHRD_IEEE80211_RADIOTAP */ +return false; + } +return true; +} + +/* Show the information of the transport layer in the packet */ +static void show_transport (const u_char *packet,u16 len,u32 ifindex, +const char *src_addr,const char *dst_addr, +u16 proto,bool ipv6) +{ +struct udphdr *udp; +struct tcphdr *tcp; +u16 src_port, dst_port; +const char *transport_str; + +if (proto ==IPPROTO_UDP ) { + udp = (struct udphdr *)packet; + src_port =ntohs (udp->source); + dst_port =ntohs (udp->dest); + transport_str ="UDP" ; + }else if (proto ==IPPROTO_TCP ) { + tcp = (struct tcphdr *)packet; + src_port =ntohs (tcp->source); + dst_port =ntohs (tcp->dest); + transport_str ="TCP" +; + }else { +printf ("%s (proto %d): %s -> %s, ifindex %d\n" , + ipv6 ?"IPv6" :"IPv4" , proto, src_addr, dst_addr, ifindex); +return ; + } + +if (ipv6) +printf ("IPv6 %s packet: [%s]:%d -> [%s]:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); +else +printf ("IPv4 %s packet: %s:%d -> %s:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); + +if (proto ==IPPROTO_TCP ) { +if (tcp->fin) +printf (", FIN" ); +if (tcp->syn) +printf (", SYN" ); +if (tcp->rst) +printf (", RST" ); +if (tcp->ack) +printf (", ACK" ); + } + +printf ("\n" ); +} + +static void show_ipv6_packet (const u_char *packet,u32 ifindex) +{ +struct ipv6hdr *pkt = (struct ipv6hdr *)packet; +struct in6_addr src; +struct in6_addr dst; +char src_str[INET6_ADDRSTRLEN ], dst_str[INET6_ADDRSTRLEN ]; +u_char proto; + +memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET6 , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET6 , &dst, dst_str,sizeof (dst_str)); + proto = pkt->nexthdr; +show_transport (packet +sizeof (struct ipv6hdr ), +ntohs (pkt->payload_len), + ifindex, src_str, dst_str, proto, true); +} + +static void show_ipv4_packet (const u_char *packet,u32 ifindex) +{ +struct iphdr *pkt = (struct iphdr *)packet; +struct in_addr src; +struct in_addr dst; +u_char proto; +char src_str[INET_ADDRSTRLEN ], dst_str[INET_ADDRSTRLEN ]; + +memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET , &dst, dst_str,sizeof (dst_str)); + proto = pkt->protocol; +show_transport (packet +sizeof (struct iphdr ), +ntohs (pkt->tot_len), + ifindex, src_str, dst_str, proto, false); +} + +static void *traffic_monitor_thread (void *arg) +{ +const u_char *packet, *payload; +struct tmonitor_ctx *ctx = arg; +struct pcap_pkthdr header; +pcap_t *pcap = ctx->pcap; +pcap_dumper_t *dumper = ctx->dumper; +int fd = ctx->pcap_fd; +int wake_fd = ctx->wake_fd_r; +u16 proto; +u32 ifindex; +fd_set fds; +int nfds, r; + + nfds = (fd > wake_fd ? fd : wake_fd) +1 ; +FD_ZERO (&fds); + +while (!ctx->done) { +FD_SET (fd, &fds); +FD_SET (wake_fd, &fds); + r =select (nfds, &fds,NULL ,NULL ,NULL ); +if (!r) +continue ; +if (r <0 ) { +if (errno ==EINTR ) +continue ; +log_err ("Fail to select on pcap fd and wake fd: %s" ,strerror (errno)); +break ; + } + + packet =pcap_next (pcap, &header); +if (!packet) +continue ; + +/* According to the man page of pcap_dump(), first argument +* is the pcap_dumper_t pointer even it's argument type is +* u_char *. +*/ +pcap_dump ((u_char *)dumper, &header, packet); + +/* Not sure what other types of packets look like. Here, we +* parse only Ethernet and compatible packets. +*/ +if (!is_ethernet (packet)) { +printf ("Packet captured\n" ); +continue ; + } + +/* Skip SLL2 header +* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +* +* Although the document doesn't mention that, the payload +* doesn't include the Ethernet header. The payload starts +* from the first byte of the network layer header. +*/ + payload = packet +20 ; + +memcpy (&proto, packet,2 ); + proto =ntohs (proto); +memcpy (&ifindex, packet +4 ,4 ); + ifindex =ntohl (ifindex); + +if (proto ==ETH_P_IPV6 ) +show_ipv6_packet (payload, ifindex); +else if (proto ==ETH_P_IP ) +show_ipv4_packet (payload, ifindex); +else +printf ("Unknown network protocol type %x, ifindex %d\n" , proto, ifindex); + } + +return NULL ; +} + +/* Prepare the pcap handle to capture packets. +* +* This pcap is non-blocking and immediate mode is enabled to receive +* captured packets as soon as possible. The snaplen is set to 1024 bytes +* to limit the size of captured content. The format of the link-layer +* header is set to DLT_LINUX_SLL2 to enable handling various link-layer +* technologies. +*/ +static pcap_t *traffic_monitor_prepare_pcap (void ) +{ +char errbuf[PCAP_ERRBUF_SIZE ]; +pcap_t *pcap; +int r; + +/* Listen on all NICs in the namespace */ + pcap =pcap_create ("any" , errbuf); +if (!pcap) { +log_err ("Failed to open pcap: %s" , errbuf); +return NULL ; + } +/* Limit the size of the packet (first N bytes) */ + r =pcap_set_snaplen (pcap,1024 ); +if (r) { +log_err ("Failed to set snaplen: %s" ,pcap_geterr (pcap)); + goto error; + } +/* To receive packets as fast as possible */ + r =pcap_set_immediate_mode (pcap,1 ); +if (r) { +log_err ("Failed to set immediate mode: %s" ,pcap_geterr (pcap)); + goto error; + } + r =pcap_setnonblock (pcap,1 , errbuf); +if (r) { +log_err ("Failed to set nonblock: %s" , errbuf); + goto error; + } + r =pcap_activate (pcap); +if (r) { +log_err ("Failed to activate pcap: %s" ,pcap_geterr (pcap)); + goto error; + } +/* Determine the format of the link-layer header */ + r =pcap_set_datalink (pcap,DLT_LINUX_SLL2 ); +if (r) { +log_err ("Failed to set datalink: %s" ,pcap_geterr (pcap)); + goto error; + } + +return pcap; +error: +pcap_close (pcap); +return NULL ; +} + +#define PCAP_DIR "/tmp/tmon_pcap" + +/* Start to monitor the network traffic in the given network namespace. +* +* netns: the name of the network namespace to monitor. If NULL, the +* current network namespace is monitored. +* +* This function will start a thread to capture packets going through NICs +* in the give network namespace. +*/ +struct tmonitor_ctx *traffic_monitor_start (const char *netns) +{ +struct tmonitor_ctx *ctx =NULL ; +struct nstoken *nstoken =NULL ; +int pipefd[2 ] = {-1 ,-1 }; +static int tmon_seq; +int r; + +if (netns) { + nstoken =open_netns (netns); +if (!nstoken) +return NULL ; + } + ctx =malloc (sizeof (*ctx)); +if (!ctx) { +log_err ("Failed to malloc ctx" ); + goto fail_ctx; + } +memset (ctx,0 ,sizeof (*ctx)); + +snprintf (ctx->pkt_fname,sizeof (ctx->pkt_fname), +PCAP_DIR "/packets-%d-%d.log" ,getpid (), tmon_seq++); + + r =mkdir (PCAP_DIR ,0755 ); +if (r && errno !=EEXIST ) { +log_err ("Failed to create " PCAP_DIR ); + goto fail_pcap; + } + + ctx->pcap =traffic_monitor_prepare_pcap (); +if (!ctx->pcap) + goto fail_pcap; + ctx->pcap_fd =pcap_get_selectable_fd (ctx->pcap); +if (ctx->pcap_fd <0 ) { +log_err ("Failed to get pcap fd" ); + goto fail_dumper; + } + +/* Create a packet file */ + ctx->dumper =pcap_dump_open (ctx->pcap, ctx->pkt_fname); +if (!ctx->dumper) { +log_err ("Failed to open pcap dump" ); + goto fail_dumper; + } + +/* Create a pipe to wake up the monitor thread */ + r =pipe (pipefd); +if (r) { +log_err ("Failed to create pipe: %s" ,strerror (errno)); + goto fail; + } + ctx->wake_fd_r = pipefd[0 ]; + ctx->wake_fd_w = pipefd[1 ]; + + r =pthread_create (&ctx->thread,NULL , traffic_monitor_thread, ctx); +if (r) { +log_err ("Failed to create thread: %s" ,strerror (r)); + goto fail; + } + +close_netns (nstoken); + +return ctx; + +fail: +close (pipefd[0 ]); +close (pipefd[1 ]); + +pcap_dump_close (ctx->dumper); +unlink (ctx->pkt_fname); + +fail_dumper: +pcap_close (ctx->pcap); + +fail_pcap: +free (ctx); + +fail_ctx: +close_netns (nstoken); + +return NULL ; +} + +static void traffic_monitor_release (struct tmonitor_ctx *ctx) +{ +pcap_close (ctx->pcap); +pcap_dump_close (ctx->dumper); + +close (ctx->wake_fd_r); +close (ctx->wake_fd_w); + +free (ctx); +} + +/* Stop the network traffic monitor. +* +* ctx: the context returned by traffic_monitor_start() +*/ +void traffic_monitor_stop (struct tmonitor_ctx *ctx) +{ +if (!ctx) +return ; + +/* Stop the monitor thread */ + ctx->done = true; +write (ctx->wake_fd_w,"x" ,1 ); +pthread_join (ctx->thread,NULL ); + +printf ("Packet file: %s\n" ,strrchr (ctx->pkt_fname,'/' ) +1 ); + +traffic_monitor_release (ctx); +} +#endif /* TRAFFIC_MONITOR */ +struct send_recv_arg {int fd;uint32_t bytes;tools/testing/selftests/bpf/network_helpers.h +16
-0@@ -82,6 +82,22 @@ int get_socket_local_port(int sock_fd);int get_hw_ring_size (char *ifname,struct ethtool_ringparam *ring_param);int set_hw_ring_size (char *ifname,struct ethtool_ringparam *ring_param); +struct tmonitor_ctx ; + +#ifdef TRAFFIC_MONITOR +struct tmonitor_ctx *traffic_monitor_start (const char *netns); +void traffic_monitor_stop (struct tmonitor_ctx *ctx); +#else +static inline struct tmonitor_ctx *traffic_monitor_start (const char *netns) +{ +return (struct tmonitor_ctx *)-1 ; +} + +static inline void traffic_monitor_stop (struct tmonitor_ctx *ctx) +{ +} +#endif +struct nstoken ;/** * open_netns() - Switch to specified network namespace by name.
Re: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
In reply to Kui-Feng LeeQuoted 92 lines · Kui-Feng Lee
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call
traffic_monitor_start()to start a thread to capture packets in the background for a given namespace and calltraffic_monitor_stop()to stop capturing.IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log #280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKThe above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory
/tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.This feature only works if
TRAFFIC_MONITORvariable has been passed to build BPF selftests. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpfThis command will build BPF selftests with this feature enabled.
Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + 3 files changed, 403 insertions(+)tools/testing/selftests/bpf/Makefile +5
-0@@ -41,6 +41,11 @@ CFLAGS += -g $(OPT_FLAGS) -rdynamic \ LDFLAGS += $(SAN_LDFLAGS) LDLIBS += $(LIBELF_LIBS) -lz -lrt -lpthread +ifneq ($(TRAFFIC_MONITOR),) +LDLIBS += -lpcap +CFLAGS += -DTRAFFIC_MONITOR=1 +endif + # The following tests perform type punning and they may break strict # aliasing rules, which are exploited by both GCC and clang by default # while optimizing. This can lead to broken programs.tools/testing/selftests/bpf/network_helpers.c +21
-0@@ -10,6 +10,7 @@#include <arpa/inet.h> #include <sys/mount.h> +#include <sys/select.h> #include <sys/stat.h> #include <sys/un.h> @@ -18,6 +19,14 @@#include <linux/in6.h> #include <linux/limits.h> +#include <netinet/udp.h> + +#include <pthread.h> +/* Prevent pcap.h from including pcap/bpf.h and causing conflicts */ +#define PCAP_DONT_INCLUDE_PCAP_BPF_H 1 +#include <pcap/pcap.h> +#include <pcap/dlt.h> +#include "bpf_util.h" #include "network_helpers.h" #include "test_progs.h" @@ -575,6 +584,379 @@ int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param)return 0 ; } +#ifdef TRAFFIC_MONITOR +struct tmonitor_ctx { +pcap_t *pcap; +pcap_dumper_t *dumper; +pthread_t thread; +int wake_fd_r; +int wake_fd_w; + +bool done; +char pkt_fname[PATH_MAX ]; +int pcap_fd; +};
[..]
Quoted 11 lines
tools/testing/selftests/bpf/network_helpers.c +11
-0+/* Is this packet captured with a Ethernet protocol type? */ +static bool is_ethernet (const u_char *packet) +{ +u16 arphdr_type; + +memcpy (&arphdr_type, packet +8 ,2 ); + arphdr_type =ntohs (arphdr_type); + +/* Except the following cases, the protocol type contains the +* Ethernet protocol type for the packet. +*
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 620
Quoted 10 lines
tools/testing/selftests/bpf/network_helpers.c +10
-0+* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +*/ +switch (arphdr_type) { +case 770 :/* ARPHRD_FRAD */ +case 778 :/* ARPHDR_IPGRE */ +case 803 :/* ARPHRD_IEEE80211_RADIOTAP */ +return false; + } +return true; +}
Are we actually getting non-ethernet packets? Any idea why?
Quoted 57 lines
tools/testing/selftests/bpf/network_helpers.c +57
-0+/* Show the information of the transport layer in the packet */ +static void show_transport (const u_char *packet,u16 len,u32 ifindex, +const char *src_addr,const char *dst_addr, +u16 proto,bool ipv6) +{ +struct udphdr *udp; +struct tcphdr *tcp; +u16 src_port, dst_port; +const char *transport_str; + +if (proto ==IPPROTO_UDP ) { + udp = (struct udphdr *)packet; + src_port =ntohs (udp->source); + dst_port =ntohs (udp->dest); + transport_str ="UDP" ; + }else if (proto ==IPPROTO_TCP ) { + tcp = (struct tcphdr *)packet; + src_port =ntohs (tcp->source); + dst_port =ntohs (tcp->dest); + transport_str ="TCP" +; + }else { +printf ("%s (proto %d): %s -> %s, ifindex %d\n" , + ipv6 ?"IPv6" :"IPv4" , proto, src_addr, dst_addr, ifindex); +return ; + } + +if (ipv6) +printf ("IPv6 %s packet: [%s]:%d -> [%s]:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); +else +printf ("IPv4 %s packet: %s:%d -> %s:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); + +if (proto ==IPPROTO_TCP ) { +if (tcp->fin) +printf (", FIN" ); +if (tcp->syn) +printf (", SYN" ); +if (tcp->rst) +printf (", RST" ); +if (tcp->ack) +printf (", ACK" ); + } + +printf ("\n" ); +} + +static void show_ipv6_packet (const u_char *packet,u32 ifindex) +{ +struct ipv6hdr *pkt = (struct ipv6hdr *)packet; +struct in6_addr src; +struct in6_addr dst; +char src_str[INET6_ADDRSTRLEN ], dst_str[INET6_ADDRSTRLEN ]; +u_char proto;
[..]
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 683
Quoted 4 lines
tools/testing/selftests/bpf/network_helpers.c +4
-0+memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET6 , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET6 , &dst, dst_str,sizeof (dst_str));
nit: can probably inet_ntop(AF_INET6, &pkt->saddr, ...) directly? No need to copy. Same for ipv4.
Quoted 207 lines
tools/testing/selftests/bpf/network_helpers.c +207
-0+ proto = pkt->nexthdr; +show_transport (packet +sizeof (struct ipv6hdr ), +ntohs (pkt->payload_len), + ifindex, src_str, dst_str, proto, true); +} + +static void show_ipv4_packet (const u_char *packet,u32 ifindex) +{ +struct iphdr *pkt = (struct iphdr *)packet; +struct in_addr src; +struct in_addr dst; +u_char proto; +char src_str[INET_ADDRSTRLEN ], dst_str[INET_ADDRSTRLEN ]; + +memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET , &dst, dst_str,sizeof (dst_str)); + proto = pkt->protocol; +show_transport (packet +sizeof (struct iphdr ), +ntohs (pkt->tot_len), + ifindex, src_str, dst_str, proto, false); +} + +static void *traffic_monitor_thread (void *arg) +{ +const u_char *packet, *payload; +struct tmonitor_ctx *ctx = arg; +struct pcap_pkthdr header; +pcap_t *pcap = ctx->pcap; +pcap_dumper_t *dumper = ctx->dumper; +int fd = ctx->pcap_fd; +int wake_fd = ctx->wake_fd_r; +u16 proto; +u32 ifindex; +fd_set fds; +int nfds, r; + + nfds = (fd > wake_fd ? fd : wake_fd) +1 ; +FD_ZERO (&fds); + +while (!ctx->done) { +FD_SET (fd, &fds); +FD_SET (wake_fd, &fds); + r =select (nfds, &fds,NULL ,NULL ,NULL ); +if (!r) +continue ; +if (r <0 ) { +if (errno ==EINTR ) +continue ; +log_err ("Fail to select on pcap fd and wake fd: %s" ,strerror (errno)); +break ; + } + + packet =pcap_next (pcap, &header); +if (!packet) +continue ; + +/* According to the man page of pcap_dump(), first argument +* is the pcap_dumper_t pointer even it's argument type is +* u_char *. +*/ +pcap_dump ((u_char *)dumper, &header, packet); + +/* Not sure what other types of packets look like. Here, we +* parse only Ethernet and compatible packets. +*/ +if (!is_ethernet (packet)) { +printf ("Packet captured\n" ); +continue ; + } + +/* Skip SLL2 header +* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +* +* Although the document doesn't mention that, the payload +* doesn't include the Ethernet header. The payload starts +* from the first byte of the network layer header. +*/ + payload = packet +20 ; + +memcpy (&proto, packet,2 ); + proto =ntohs (proto); +memcpy (&ifindex, packet +4 ,4 ); + ifindex =ntohl (ifindex); + +if (proto ==ETH_P_IPV6 ) +show_ipv6_packet (payload, ifindex); +else if (proto ==ETH_P_IP ) +show_ipv4_packet (payload, ifindex); +else +printf ("Unknown network protocol type %x, ifindex %d\n" , proto, ifindex); + } + +return NULL ; +} + +/* Prepare the pcap handle to capture packets. +* +* This pcap is non-blocking and immediate mode is enabled to receive +* captured packets as soon as possible. The snaplen is set to 1024 bytes +* to limit the size of captured content. The format of the link-layer +* header is set to DLT_LINUX_SLL2 to enable handling various link-layer +* technologies. +*/ +static pcap_t *traffic_monitor_prepare_pcap (void ) +{ +char errbuf[PCAP_ERRBUF_SIZE ]; +pcap_t *pcap; +int r; + +/* Listen on all NICs in the namespace */ + pcap =pcap_create ("any" , errbuf); +if (!pcap) { +log_err ("Failed to open pcap: %s" , errbuf); +return NULL ; + } +/* Limit the size of the packet (first N bytes) */ + r =pcap_set_snaplen (pcap,1024 ); +if (r) { +log_err ("Failed to set snaplen: %s" ,pcap_geterr (pcap)); + goto error; + } +/* To receive packets as fast as possible */ + r =pcap_set_immediate_mode (pcap,1 ); +if (r) { +log_err ("Failed to set immediate mode: %s" ,pcap_geterr (pcap)); + goto error; + } + r =pcap_setnonblock (pcap,1 , errbuf); +if (r) { +log_err ("Failed to set nonblock: %s" , errbuf); + goto error; + } + r =pcap_activate (pcap); +if (r) { +log_err ("Failed to activate pcap: %s" ,pcap_geterr (pcap)); + goto error; + } +/* Determine the format of the link-layer header */ + r =pcap_set_datalink (pcap,DLT_LINUX_SLL2 ); +if (r) { +log_err ("Failed to set datalink: %s" ,pcap_geterr (pcap)); + goto error; + } + +return pcap; +error: +pcap_close (pcap); +return NULL ; +} + +#define PCAP_DIR "/tmp/tmon_pcap" + +/* Start to monitor the network traffic in the given network namespace. +* +* netns: the name of the network namespace to monitor. If NULL, the +* current network namespace is monitored. +* +* This function will start a thread to capture packets going through NICs +* in the give network namespace. +*/ +struct tmonitor_ctx *traffic_monitor_start (const char *netns) +{ +struct tmonitor_ctx *ctx =NULL ; +struct nstoken *nstoken =NULL ; +int pipefd[2 ] = {-1 ,-1 }; +static int tmon_seq; +int r; + +if (netns) { + nstoken =open_netns (netns); +if (!nstoken) +return NULL ; + } + ctx =malloc (sizeof (*ctx)); +if (!ctx) { +log_err ("Failed to malloc ctx" ); + goto fail_ctx; + } +memset (ctx,0 ,sizeof (*ctx)); + +snprintf (ctx->pkt_fname,sizeof (ctx->pkt_fname), +PCAP_DIR "/packets-%d-%d.log" ,getpid (), tmon_seq++); + + r =mkdir (PCAP_DIR ,0755 ); +if (r && errno !=EEXIST ) { +log_err ("Failed to create " PCAP_DIR ); + goto fail_pcap; + } + + ctx->pcap =traffic_monitor_prepare_pcap (); +if (!ctx->pcap) + goto fail_pcap; + ctx->pcap_fd =pcap_get_selectable_fd (ctx->pcap); +if (ctx->pcap_fd <0 ) { +log_err ("Failed to get pcap fd" ); + goto fail_dumper; + } + +/* Create a packet file */ + ctx->dumper =pcap_dump_open (ctx->pcap, ctx->pkt_fname); +if (!ctx->dumper) { +log_err ("Failed to open pcap dump" ); + goto fail_dumper; + } +
[..]
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 898
Quoted 8 lines
tools/testing/selftests/bpf/network_helpers.c +8
-0+/* Create a pipe to wake up the monitor thread */ + r =pipe (pipefd); +if (r) { +log_err ("Failed to create pipe: %s" ,strerror (errno)); + goto fail; + } + ctx->wake_fd_r = pipefd[0 ]; + ctx->wake_fd_w = pipefd[1 ];
eventfd might be a simpler way to handle this. Gives you one fd which is readable/writable. But probably ok to keep the pipe since you already have all the code written.
Re: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
Quoted 112 lines · Stanislav Fomichev
Quoted 92 lines · Kui-Feng Lee
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call
traffic_monitor_start()to start a thread to capture packets in the background for a given namespace and calltraffic_monitor_stop()to stop capturing.IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log #280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKThe above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory
/tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.This feature only works if
TRAFFIC_MONITORvariable has been passed to build BPF selftests. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpfThis command will build BPF selftests with this feature enabled.
Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + 3 files changed, 403 insertions(+)tools/testing/selftests/bpf/Makefile +5
-0@@ -41,6 +41,11 @@ CFLAGS += -g $(OPT_FLAGS) -rdynamic \ LDFLAGS += $(SAN_LDFLAGS) LDLIBS += $(LIBELF_LIBS) -lz -lrt -lpthread +ifneq ($(TRAFFIC_MONITOR),) +LDLIBS += -lpcap +CFLAGS += -DTRAFFIC_MONITOR=1 +endif + # The following tests perform type punning and they may break strict # aliasing rules, which are exploited by both GCC and clang by default # while optimizing. This can lead to broken programs.tools/testing/selftests/bpf/network_helpers.c +21
-0@@ -10,6 +10,7 @@#include <arpa/inet.h> #include <sys/mount.h> +#include <sys/select.h> #include <sys/stat.h> #include <sys/un.h> @@ -18,6 +19,14 @@#include <linux/in6.h> #include <linux/limits.h> +#include <netinet/udp.h> + +#include <pthread.h> +/* Prevent pcap.h from including pcap/bpf.h and causing conflicts */ +#define PCAP_DONT_INCLUDE_PCAP_BPF_H 1 +#include <pcap/pcap.h> +#include <pcap/dlt.h> +#include "bpf_util.h" #include "network_helpers.h" #include "test_progs.h" @@ -575,6 +584,379 @@ int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param)return 0 ; } +#ifdef TRAFFIC_MONITOR +struct tmonitor_ctx { +pcap_t *pcap; +pcap_dumper_t *dumper; +pthread_t thread; +int wake_fd_r; +int wake_fd_w; + +bool done; +char pkt_fname[PATH_MAX ]; +int pcap_fd; +};[..]
Quoted 15 lines
tools/testing/selftests/bpf/network_helpers.c +15
-0+/* Is this packet captured with a Ethernet protocol type? */ +static bool is_ethernet (const u_char *packet) +{ +u16 arphdr_type; + +memcpy (&arphdr_type, packet +8 ,2 ); + arphdr_type =ntohs (arphdr_type); + +/* Except the following cases, the protocol type contains the +* Ethernet protocol type for the packet. +* +* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +*/ +switch (arphdr_type) { +case 770 :/* ARPHRD_FRAD */
Quoted 8 lines
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 620
Quoted 6 lines
tools/testing/selftests/bpf/network_helpers.c +6
-0+case 778 :/* ARPHDR_IPGRE */ +case 803 :/* ARPHRD_IEEE80211_RADIOTAP */ +return false; + } +return true; +}Are we actually getting non-ethernet packets? Any idea why?
No, I haven't get any non-ethernet packets so far. This test is just for ensuring everything going as what we expected.
Quoted 61 lines
Quoted 58 lines
tools/testing/selftests/bpf/network_helpers.c +57
-0+/* Show the information of the transport layer in the packet */ +static void show_transport (const u_char *packet,u16 len,u32 ifindex, +const char *src_addr,const char *dst_addr, +u16 proto,bool ipv6) +{ +struct udphdr *udp; +struct tcphdr *tcp; +u16 src_port, dst_port; +const char *transport_str; + +if (proto ==IPPROTO_UDP ) { + udp = (struct udphdr *)packet; + src_port =ntohs (udp->source); + dst_port =ntohs (udp->dest); + transport_str ="UDP" ; + }else if (proto ==IPPROTO_TCP ) { + tcp = (struct tcphdr *)packet; + src_port =ntohs (tcp->source); + dst_port =ntohs (tcp->dest); + transport_str ="TCP" +; + }else { +printf ("%s (proto %d): %s -> %s, ifindex %d\n" , + ipv6 ?"IPv6" :"IPv4" , proto, src_addr, dst_addr, ifindex); +return ; + } + +if (ipv6) +printf ("IPv6 %s packet: [%s]:%d -> [%s]:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); +else +printf ("IPv4 %s packet: %s:%d -> %s:%d, len %d, ifindex %d" , + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); + +if (proto ==IPPROTO_TCP ) { +if (tcp->fin) +printf (", FIN" ); +if (tcp->syn) +printf (", SYN" ); +if (tcp->rst) +printf (", RST" ); +if (tcp->ack) +printf (", ACK" ); + } + +printf ("\n" ); +} + +static void show_ipv6_packet (const u_char *packet,u32 ifindex) +{ +struct ipv6hdr *pkt = (struct ipv6hdr *)packet; +struct in6_addr src; +struct in6_addr dst; +char src_str[INET6_ADDRSTRLEN ], dst_str[INET6_ADDRSTRLEN ]; +u_char proto;[..]
Quoted 7 lines
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 683
Quoted 4 lines
tools/testing/selftests/bpf/network_helpers.c +4
-0+memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET6 , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET6 , &dst, dst_str,sizeof (dst_str));nit: can probably
inet_ntop(AF_INET6, &pkt->saddr, ...)directly? No need to copy. Same for ipv4.
You are right.
Quoted 211 lines
Quoted 208 lines
tools/testing/selftests/bpf/network_helpers.c +207
-0+ proto = pkt->nexthdr; +show_transport (packet +sizeof (struct ipv6hdr ), +ntohs (pkt->payload_len), + ifindex, src_str, dst_str, proto, true); +} + +static void show_ipv4_packet (const u_char *packet,u32 ifindex) +{ +struct iphdr *pkt = (struct iphdr *)packet; +struct in_addr src; +struct in_addr dst; +u_char proto; +char src_str[INET_ADDRSTRLEN ], dst_str[INET_ADDRSTRLEN ]; + +memcpy (&src, &pkt->saddr,sizeof (src)); +memcpy (&dst, &pkt->daddr,sizeof (dst)); +inet_ntop (AF_INET , &src, src_str,sizeof (src_str)); +inet_ntop (AF_INET , &dst, dst_str,sizeof (dst_str)); + proto = pkt->protocol; +show_transport (packet +sizeof (struct iphdr ), +ntohs (pkt->tot_len), + ifindex, src_str, dst_str, proto, false); +} + +static void *traffic_monitor_thread (void *arg) +{ +const u_char *packet, *payload; +struct tmonitor_ctx *ctx = arg; +struct pcap_pkthdr header; +pcap_t *pcap = ctx->pcap; +pcap_dumper_t *dumper = ctx->dumper; +int fd = ctx->pcap_fd; +int wake_fd = ctx->wake_fd_r; +u16 proto; +u32 ifindex; +fd_set fds; +int nfds, r; + + nfds = (fd > wake_fd ? fd : wake_fd) +1 ; +FD_ZERO (&fds); + +while (!ctx->done) { +FD_SET (fd, &fds); +FD_SET (wake_fd, &fds); + r =select (nfds, &fds,NULL ,NULL ,NULL ); +if (!r) +continue ; +if (r <0 ) { +if (errno ==EINTR ) +continue ; +log_err ("Fail to select on pcap fd and wake fd: %s" ,strerror (errno)); +break ; + } + + packet =pcap_next (pcap, &header); +if (!packet) +continue ; + +/* According to the man page of pcap_dump(), first argument +* is the pcap_dumper_t pointer even it's argument type is +* u_char *. +*/ +pcap_dump ((u_char *)dumper, &header, packet); + +/* Not sure what other types of packets look like. Here, we +* parse only Ethernet and compatible packets. +*/ +if (!is_ethernet (packet)) { +printf ("Packet captured\n" ); +continue ; + } + +/* Skip SLL2 header +* https://www.tcpdump.org/linktypes/LINKTYPE_LINUX_SLL2.html +* +* Although the document doesn't mention that, the payload +* doesn't include the Ethernet header. The payload starts +* from the first byte of the network layer header. +*/ + payload = packet +20 ; + +memcpy (&proto, packet,2 ); + proto =ntohs (proto); +memcpy (&ifindex, packet +4 ,4 ); + ifindex =ntohl (ifindex); + +if (proto ==ETH_P_IPV6 ) +show_ipv6_packet (payload, ifindex); +else if (proto ==ETH_P_IP ) +show_ipv4_packet (payload, ifindex); +else +printf ("Unknown network protocol type %x, ifindex %d\n" , proto, ifindex); + } + +return NULL ; +} + +/* Prepare the pcap handle to capture packets. +* +* This pcap is non-blocking and immediate mode is enabled to receive +* captured packets as soon as possible. The snaplen is set to 1024 bytes +* to limit the size of captured content. The format of the link-layer +* header is set to DLT_LINUX_SLL2 to enable handling various link-layer +* technologies. +*/ +static pcap_t *traffic_monitor_prepare_pcap (void ) +{ +char errbuf[PCAP_ERRBUF_SIZE ]; +pcap_t *pcap; +int r; + +/* Listen on all NICs in the namespace */ + pcap =pcap_create ("any" , errbuf); +if (!pcap) { +log_err ("Failed to open pcap: %s" , errbuf); +return NULL ; + } +/* Limit the size of the packet (first N bytes) */ + r =pcap_set_snaplen (pcap,1024 ); +if (r) { +log_err ("Failed to set snaplen: %s" ,pcap_geterr (pcap)); + goto error; + } +/* To receive packets as fast as possible */ + r =pcap_set_immediate_mode (pcap,1 ); +if (r) { +log_err ("Failed to set immediate mode: %s" ,pcap_geterr (pcap)); + goto error; + } + r =pcap_setnonblock (pcap,1 , errbuf); +if (r) { +log_err ("Failed to set nonblock: %s" , errbuf); + goto error; + } + r =pcap_activate (pcap); +if (r) { +log_err ("Failed to activate pcap: %s" ,pcap_geterr (pcap)); + goto error; + } +/* Determine the format of the link-layer header */ + r =pcap_set_datalink (pcap,DLT_LINUX_SLL2 ); +if (r) { +log_err ("Failed to set datalink: %s" ,pcap_geterr (pcap)); + goto error; + } + +return pcap; +error: +pcap_close (pcap); +return NULL ; +} + +#define PCAP_DIR "/tmp/tmon_pcap" + +/* Start to monitor the network traffic in the given network namespace. +* +* netns: the name of the network namespace to monitor. If NULL, the +* current network namespace is monitored. +* +* This function will start a thread to capture packets going through NICs +* in the give network namespace. +*/ +struct tmonitor_ctx *traffic_monitor_start (const char *netns) +{ +struct tmonitor_ctx *ctx =NULL ; +struct nstoken *nstoken =NULL ; +int pipefd[2 ] = {-1 ,-1 }; +static int tmon_seq; +int r; + +if (netns) { + nstoken =open_netns (netns); +if (!nstoken) +return NULL ; + } + ctx =malloc (sizeof (*ctx)); +if (!ctx) { +log_err ("Failed to malloc ctx" ); + goto fail_ctx; + } +memset (ctx,0 ,sizeof (*ctx)); + +snprintf (ctx->pkt_fname,sizeof (ctx->pkt_fname), +PCAP_DIR "/packets-%d-%d.log" ,getpid (), tmon_seq++); + + r =mkdir (PCAP_DIR ,0755 ); +if (r && errno !=EEXIST ) { +log_err ("Failed to create " PCAP_DIR ); + goto fail_pcap; + } + + ctx->pcap =traffic_monitor_prepare_pcap (); +if (!ctx->pcap) + goto fail_pcap; + ctx->pcap_fd =pcap_get_selectable_fd (ctx->pcap); +if (ctx->pcap_fd <0 ) { +log_err ("Failed to get pcap fd" ); + goto fail_dumper; + } + +/* Create a packet file */ + ctx->dumper =pcap_dump_open (ctx->pcap, ctx->pkt_fname); +if (!ctx->dumper) { +log_err ("Failed to open pcap dump" ); + goto fail_dumper; + } +[..]
Quoted 12 lines
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 898
Quoted 8 lines
tools/testing/selftests/bpf/network_helpers.c +8
-0+/* Create a pipe to wake up the monitor thread */ + r =pipe (pipefd); +if (r) { +log_err ("Failed to create pipe: %s" ,strerror (errno)); + goto fail; + } + ctx->wake_fd_r = pipefd[0 ]; + ctx->wake_fd_w = pipefd[1 ];eventfd might be a simpler way to handle this. Gives you one fd which is readable/writable. But probably ok to keep the pipe since you already have all the code written.
Sure, I will move to eventfd. It is not a big deal.
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
In reply to Kui-Feng LeeQuoted 15 lines · Kui-Feng Lee
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call
traffic_monitor_start()to start a thread to capture packets in the background for a given namespace and calltraffic_monitor_stop()to stop capturing.IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log #280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OK
Quoted 8 lines
The above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory
/tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.This feature only works if
TRAFFIC_MONITORvariable has been passed to build BPF selftests. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpf
If bpf CI does not have libpcap, it is better to get bpf CI ready first/soon.
[ ... ]
Quoted 12 lines
tools/testing/selftests/bpf/network_helpers.c +12
-0+/* Show the information of the transport layer in the packet */ +static void show_transport (const u_char *packet,u16 len,u32 ifindex, +const char *src_addr,const char *dst_addr, +u16 proto,bool ipv6) +{ +struct udphdr *udp; +struct tcphdr *tcp; +u16 src_port, dst_port; +const char *transport_str; + +if (proto ==IPPROTO_UDP ) { + udp = (struct udphdr *)packet;
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 643
Quoted 10 lines
tools/testing/selftests/bpf/network_helpers.c +10
-0+ src_port =ntohs (udp->source); + dst_port =ntohs (udp->dest); + transport_str ="UDP" ; + }else if (proto ==IPPROTO_TCP ) { + tcp = (struct tcphdr *)packet; + src_port =ntohs (tcp->source); + dst_port =ntohs (tcp->dest); + transport_str ="TCP" +; + }else {
It will be useful to at least print the ICMP[46] also. Some tests use ping to test. For IPv6, printing the ICMPv6 messages will be useful for debugging, e.g. the neigh discovery. The icmp type (and code?) should be good enough.
That should be enough to begin with. The pcap dumped file can be used for the rest.
Thanks for switching to libpcap. It is easier to handle the captured packets in different ways.
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 650
Quoted 7 lines
tools/testing/selftests/bpf/network_helpers.c +7
-0+printf ("%s (proto %d): %s -> %s, ifindex %d\n" , + ipv6 ?"IPv6" :"IPv4" , proto, src_addr, dst_addr, ifindex); +return ; + } + +if (ipv6) + printf("IPv6 %s packet: [%s]:%d -> [%s]:%d, len %d, ifindex %d",
It will be useful to print the ifname also such that easier for human parsing. It should be possible by if_indextoname (cheap enough?) if libpcap doesn't have it. It could be something for a later followup though. Mostly nit here.
Quoted 20 lines
tools/testing/selftests/bpf/network_helpers.c +20
-0+ transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); + else + printf("IPv4 %s packet: %s:%d -> %s:%d, len %d, ifindex %d", + transport_str, src_addr, src_port, + dst_addr, dst_port, len, ifindex); + +if (proto ==IPPROTO_TCP ) { +if (tcp->fin) +printf (", FIN" ); +if (tcp->syn) +printf (", SYN" ); +if (tcp->rst) +printf (", RST" ); +if (tcp->ack) +printf (", ACK" ); + } + +printf ("\n" ); +}
[ ... ]
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 846
Quoted 9 lines
tools/testing/selftests/bpf/network_helpers.c +9
-0+/* Start to monitor the network traffic in the given network namespace. +* +* netns: the name of the network namespace to monitor. If NULL, the +* current network namespace is monitored. +* +* This function will start a thread to capture packets going through NICs +* in the give network namespace. +*/ +struct tmonitor_ctx *traffic_monitor_start (const char *netns)
There is opportunity to make the traffic monitoring easier for tests that create its own netns which I hope most of the networking tests fall into this bucket now. Especially for tests that create multiple netns such that the test does not have to start/stop for each individual netns.
May be adding an API like "struct nstoken *netns_new(const char *netns_name)". The netns_new() will create the netns and (optionally) start the monitoring thread also. It will need another "void netns_free(struct nstoken *nstoken)" to stop the thread and remove the netns. The "struct tmonitor_ctx" probably makes sense to be embedded into "struct nstoken" if we go with this new API.
This will need some changes to the tests creating netns but it probably should be obvious change considering most test do "ip netns add..." and then open_netns(). It can start with the flaky test at hand first like tc_redirect.
May be a little more changes for the test using "unshare(CLONE_NEWNET)" but should not be too bad either. This can be done only when we need to turn on libpcap to debug that test.
Also, when the test is flaky, make it easier for people not familiar with the codes of the networking test to turn on traffic monitoring without changing the test code. May be in a libpcap.list file (in parallel to the existing DENYLIST)?
For the tests without having its own netns, they can either move to netns (which I think it is a good thing to do) or use the traffic_monitor_start/stop() manually by changing the testing code, or a better way is to ask test_progs do it for the host netns (init_netns) automatically for all tests in the libpcap.list.
wdyt?
Quoted 11 lines
tools/testing/selftests/bpf/network_helpers.c +11
-0+{ +struct tmonitor_ctx *ctx =NULL ; +struct nstoken *nstoken =NULL ; +int pipefd[2 ] = {-1 ,-1 }; +static int tmon_seq; +int r; + +if (netns) { + nstoken =open_netns (netns); +if (!nstoken) +return NULL ;
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 867
Quoted 10 lines
tools/testing/selftests/bpf/network_helpers.c +10
-0+ } + ctx =malloc (sizeof (*ctx)); +if (!ctx) { +log_err ("Failed to malloc ctx" ); + goto fail_ctx; + } +memset (ctx,0 ,sizeof (*ctx)); + +snprintf (ctx->pkt_fname,sizeof (ctx->pkt_fname), +PCAP_DIR "/packets-%d-%d.log" ,getpid (), tmon_seq++);
nit. I wonder if it is useful to also have the netns name in the filename?
Not sure if it is more useful to have the test_num and subtest_num instead of pid. Probably doable from looking at test__start_subtest().
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
Quoted 2 lines · Martin KaFai Lau
For the tests without having its own netns, they can either move to netns (which I think it is a good thing to do) or use the traffic_monitor_start/stop() manually by changing the testing code, or a better way is to ask
test_progsdo it for the host netns (init_netns) automatically for all tests in the libpcap.list.
After another thought, probably scrape the init_netns auto capturing idea for now. I think it could be too noisy for the tests that do use netns.
Re: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
In reply to Martin KaFai LauQuoted 45 lines · Martin KaFai Lau
Quoted 15 lines · Kui-Feng Lee
Add functions that capture packets and print log in the background. They are supposed to be used for debugging flaky network test cases. A monitored test case should call
traffic_monitor_start()to start a thread to capture packets in the background for a given namespace and calltraffic_monitor_stop()to stop capturing.IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log
#280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKQuoted 8 lines
The above is the output of an example. It shows the packets of a connection and the name of the file that contains captured packets in the directory
/tmp/tmon_pcap. The file can be loaded by tcpdump or wireshark.This feature only works if
TRAFFIC_MONITORvariable has been passed to build BPF selftests. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpfIf bpf CI does not have libpcap, it is better to get bpf CI ready first/soon.
[ ... ]
Quoted 16 lines
tools/testing/selftests/bpf/network_helpers.c +16
-0+/* Show the information of the transport layer in the packet */ +static void show_transport (const u_char *packet,u16 len,u32 ifindex, + const char *src_addr, const char *dst_addr, + u16 proto, bool ipv6) +{ + struct udphdr *udp; + struct tcphdr *tcp; + u16 src_port, dst_port; + const char *transport_str; + + if (proto == IPPROTO_UDP) { + udp = (struct udphdr *)packet; + src_port = ntohs(udp->source); + dst_port = ntohs(udp->dest); + transport_str = "UDP"; + } else if (proto == IPPROTO_TCP) {
Quoted 8 lines
Quoted 6 lines
+ tcp = (struct tcphdr *)packet;
+src_port=ntohs(tcp->source);
+dst_port=ntohs(tcp->dest);
+transport_str= "TCP"
+;
+ } else {It will be useful to at least print the ICMP[46] also. Some tests use ping to test. For IPv6, printing the ICMPv6 messages will be useful for debugging, e.g. the neigh discovery. The icmp type (and code?) should be good enough.
Right, ICMP is something should be included.
Quoted 14 lines
That should be enough to begin with. The pcap dumped file can be used for the rest.
Thanks for switching to libpcap. It is easier to handle the captured packets in different ways.
Quoted 8 lines
+ printf("%s (proto %d): %s -> %s, ifindex %d\n",
+ ipv6 ? "IPv6" : "IPv4", proto,src_addr,dst_addr, ifindex);
+ return;
+ }
+
+ if (ipv6)
+ printf("IPv6 %s packet: [%s]:%d -> [%s]:%d, len %d, ifindex %d",
Quoted 1 line
It will be useful to print the ifname also such that easier for human parsing. It should be possible by
if_indextoname(cheap enough?) if libpcap doesn't have it. It could be something for a later followup though. Mostly nit here.
This is not a big work. I will include it in the next version.
Quoted 37 lines
Quoted 21 lines
+
transport_str,src_addr,src_port,
+dst_addr,dst_port, len, ifindex);
+ else
+ printf("IPv4 %s packet: %s:%d -> %s:%d, len %d, ifindex %d", +transport_str,src_addr,src_port,
+dst_addr,dst_port, len, ifindex);tools/testing/selftests/bpf/network_helpers.c +14
-0+ + if (proto == IPPROTO_TCP) { + if (tcp->fin) + printf(", FIN"); + if (tcp->syn) + printf(", SYN"); + if (tcp->rst) + printf(", RST"); + if (tcp->ack) + printf(", ACK"); + } + + printf("\n"); +}[ ... ]
Quoted 10 lines
+/* Start to monitor the network traffic in the given network namespace.
+ *
+ * netns: the name of the network namespace to monitor. If NULL, the
+ * current network namespace is monitored.
+ *
+ * This function will start a thread to capture packets going through NICs
+ * in the give network namespace.
+ */
+structtmonitor_ctx*traffic_monitor_start(const char *netns)There is opportunity to make the traffic monitoring easier for tests that create its own netns which I hope most of the networking tests fall into this bucket now. Especially for tests that create multiple netns such that the test does not have to start/stop for each individual netns.
Quoted 1 line
May be adding an API like "struct nstoken *netns_new(const char
*netns_name)". Thenetns_new()will create the netns and (optionally) start the monitoring thread also. It will need another "voidnetns_free(struct nstoken *nstoken)" to stop the thread and remove the netns. The "structtmonitor_ctx" probably makes sense to be embedded into "struct nstoken" if we go with this new API.
Agree! But, I think we need another type rather than to reuse "struct netns". People may accidentally call close_netns() on the nstoken returned by this function.
Quoted 9 lines
This will need some changes to the tests creating netns but it probably should be obvious change considering most test do "ip netns add..." and then
open_netns(). It can start with the flaky test at hand first liketc_redirect.May be a little more changes for the test using "
unshare(CLONE_NEWNET)" but should not be too bad either. This can be done only when we need to turn on libpcap to debug that test.Also, when the test is flaky, make it easier for people not familiar with the codes of the networking test to turn on traffic monitoring without changing the test code. May be in a libpcap.list file (in parallel to the existing DENYLIST)?
For the tests without having its own netns, they can either move to netns (which I think it is a good thing to do) or use the traffic_monitor_start/stop() manually by changing the testing code, or a better way is to ask
test_progsdo it for the host netns (init_netns) automatically for all tests in the libpcap.list.
Agree! I will start move some tests to netns, and use libpcap.list to enable them.
Quoted 27 lines
wdyt?
Quoted 11 lines
tools/testing/selftests/bpf/network_helpers.c +11
-0+{ + struct tmonitor_ctx *ctx = NULL; + struct nstoken *nstoken = NULL; + int pipefd[2] = {-1, -1}; + static int tmon_seq; + int r; + + if (netns) { + nstoken = open_netns(netns); + if (!nstoken) + return NULL;tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 867
Quoted 10 lines
tools/testing/selftests/bpf/network_helpers.c +10
-0+ } + ctx = malloc(sizeof(*ctx)); + if (!ctx) { + log_err("Failed to malloc ctx"); + goto fail_ctx; + } + memset(ctx, 0, sizeof(*ctx)); + + snprintf(ctx->pkt_fname, sizeof(ctx->pkt_fname), + PCAP_DIR "/packets-%d-%d.log", getpid(), tmon_seq++);nit. I wonder if it is useful to also have the netns name in the filename?
Quoted 1 line
Not sure if it is more useful to have the
test_numandsubtest_numinstead of pid. Probably doable from looking attest__start_subtest().
It should be doable. These information is available through test_env. Last time I tried to use env, but run into an error. I will try again.
Quoted 2 lines
Martin KaFai Lau
reviewerRe: [PATCH bpf-next v2 1/4] selftests/bpf: Add traffic monitor functions.
Quoted 12 lines · Kui-Feng Lee
Quoted 12 lines
tools/testing/selftests/bpf/network_helpers.c int set_hw_ring_size(char *ifname, struct ethtool_ringparam *ring_param) · line 846
Quoted 9 lines
tools/testing/selftests/bpf/network_helpers.c +9
-0+/* Start to monitor the network traffic in the given network namespace. +* +* netns: the name of the network namespace to monitor. If NULL, the +* current network namespace is monitored. +* +* This function will start a thread to capture packets going through NICs +* in the give network namespace. +*/ +struct tmonitor_ctx *traffic_monitor_start (const char *netns)There is opportunity to make the traffic monitoring easier for tests that create its own netns which I hope most of the networking tests fall into this bucket now. Especially for tests that create multiple netns such that the test does not have to start/stop for each individual netns.
Quoted 5 lines
Quoted 1 line
May be adding an API like "struct nstoken *netns_new(const char
*netns_name)". Thenetns_new()will create the netns and (optionally) start the monitoring thread also. It will need another "voidnetns_free(struct nstoken *nstoken)" to stop the thread and remove the netns. The "structtmonitor_ctx" probably makes sense to be embedded into "struct nstoken" if we go with this new API.Agree! But, I think we need another type rather than to reuse "struct netns". People may accidentally call
close_netns()on the nstoken returned by this function.
ah. Good point. close_netns() does free the nstoken also...
yep. probably make sense to have another type for netns create/destroy which start/stop the monitoring automatically based on the on/off in the libpcap.list.
Quoted 6 lines
Quoted 6 lines
This will need some changes to the tests creating netns but it probably should be obvious change considering most test do "ip netns add..." and then
open_netns(). It can start with the flaky test at hand first liketc_redirect.May be a little more changes for the test using "
unshare(CLONE_NEWNET)" but should not be too bad either. This can be done only when we need to turn on libpcap to debug that test.
Quoted 7 lines
Quoted 4 lines
Also, when the test is flaky, make it easier for people not familiar with the codes of the networking test to turn on traffic monitoring without changing the test code. May be in a libpcap.list file (in parallel to the existing DENYLIST)?
For the tests without having its own netns, they can either move to netns (which I think it is a good thing to do) or use the traffic_monitor_start/stop() manually by changing the testing code, or a better way is to ask
test_progsdo it for the host netns (init_netns) automatically for all tests in the libpcap.list.Agree! I will start move some tests to netns, and use libpcap.list to enable them.
The tc_redirect test should be in netns already. It seems the select_reuseport and the sockmap_listen test, that this patchset is touching, are not in netns. I hope the netns migration changes should be obvious for them. Other than those two flaky tests, I would separate other netns moving work to another effort.
PATCH v2 2/4 selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
In reply to Kui-Feng LeeEnable traffic monitoring for the test case tc_redirect/tc_redirect_dtime.
--- tools/testing/selftests/bpf/prog_tests/tc_redirect.c | 5 +++++ 1 file changed, 5 insertions(+)
tools/testing/selftests/bpf/prog_tests/tc_redirect.c +5 -0
@@ -900,6 +900,7 @@ static void test_udp_dtime(struct test_tc_dtime *skel, int family, bool bpf_fwd)static void test_tc_redirect_dtime (struct netns_setup_result *setup_result) {struct test_tc_dtime *skel; +struct tmonitor_ctx *tmon =NULL ;struct nstoken *nstoken;int hold_tstamp_fd, err;
@@ -934,6 +935,9 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result)if (!ASSERT_OK (err,"disable forwarding" )) goto done; + tmon =traffic_monitor_start (NS_DST ); +ASSERT_NEQ (tmon,NULL ,"traffic_monitor_start" ); +test_tcp_clear_dtime (skel);test_tcp_dtime (skel,AF_INET , true);
@@ -958,6 +962,7 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result)test_udp_dtime (skel,AF_INET6 , false); done: +traffic_monitor_stop (tmon);test_tc_dtime__destroy (skel);close (hold_tstamp_fd); }
Signature · 1 line
2.34.1
Re: [PATCH bpf-next v2 2/4] selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
Quoted 19 lines · Kui-Feng Lee
Enable traffic monitoring for the test case
tc_redirect/tc_redirect_dtime.Signed-off-by Kui-Feng Lee <[email protected]>---
tools/testing/selftests/bpf/prog_tests/tc_redirect.c | 5 +++++1 file changed, 5 insertions(+)tools/testing/selftests/bpf/prog_tests/tc_redirect.c +0
-0@@ -900,6 +900,7 @@ static void test_udp_dtime(struct test_tc_dtime *skel, int family, bool bpf_fwd)static void
test_tc_redirect_dtime(struct netns_setup_result *setup_result)
{
structtest_tc_dtime*skel;
tools/testing/selftests/bpf/prog_tests/tc_redirect.c static void test_udp_dtime(struct test_tc_dtime *skel, int family, bool bpf_fwd) · line 903
Quoted 1 line
tools/testing/selftests/bpf/prog_tests/tc_redirect.c +1
-0+struct tmonitor_ctx *tmon =NULL ;
nit: No need to set "tmon" to NULL here.
Quoted 23 lines
struct nstoken *nstoken;
inthold_tstamp_fd, err;tools/testing/selftests/bpf/prog_tests/tc_redirect.c +0
-0@@ -934,6 +935,9 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result) if (!ASSERT_OK(err, "disable forwarding")) goto done;tools/testing/selftests/bpf/prog_tests/tc_redirect.c +3
-0+ tmon =traffic_monitor_start (NS_DST ); +ASSERT_NEQ (tmon,NULL ,"traffic_monitor_start" ); +
test_tcp_clear_dtime(skel);
test_tcp_dtime(skel, AF_INET, true);tools/testing/selftests/bpf/prog_tests/tc_redirect.c +0
-0@@ -958,6 +962,7 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result) test_udp_dtime(skel, AF_INET6, false);
done:tools/testing/selftests/bpf/prog_tests/tc_redirect.c +1
-0+traffic_monitor_stop (tmon);
test_tc_dtime__destroy(skel);
close(hold_tstamp_fd);
}
Re: [PATCH bpf-next v2 2/4] selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
Quoted 9 lines · Geliang Tang
Quoted 8 lines · Kui-Feng Lee
Enable traffic monitoring for the test case
tc_redirect/tc_redirect_dtime.Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/prog_tests/tc_redirect.c | 5 +++++ 1 file changed, 5 insertions(+)
Quoted 14 lines
tools/testing/selftests/bpf/prog_tests/tc_redirect.c static void test_udp_dtime(struct test_tc_dtime · line 905
Quoted 12 lines
tools/testing/selftests/bpf/prog_tests/tc_redirect.c +1
-0@@ -900,6 +900,7 @@ static void test_udp_dtime(struct test_tc_dtime *skel, int family, bool bpf_fwd) static void test_tc_redirect_dtime(struct netns_setup_result *setup_result) { struct test_tc_dtime *skel; +struct tmonitor_ctx *tmon =NULL ;nit: No need to set "tmon" to NULL here.
Earlier "goto done" statements would need tmon to be initialized. Does that make sense?
Quoted 25 lines
Quoted 24 lines
tools/testing/selftests/bpf/prog_tests/tc_redirect.c +0
-0struct nstoken *nstoken;int
hold_tstamp_fd, err;tools/testing/selftests/bpf/prog_tests/tc_redirect.c +3
-0@@ -934,6 +935,9 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result) if (!ASSERT_OK(err, "disable forwarding")) goto done; + tmon =traffic_monitor_start (NS_DST ); +ASSERT_NEQ (tmon,NULL ,"traffic_monitor_start" ); + test_tcp_clear_dtime(skel);
test_tcp_dtime(skel, AF_INET, true);tools/testing/selftests/bpf/prog_tests/tc_redirect.c +1
-0@@ -958,6 +962,7 @@ static void test_tc_redirect_dtime(struct netns_setup_result *setup_result) test_udp_dtime(skel, AF_INET6, false); done: +traffic_monitor_stop (tmon); test_tc_dtime__destroy(skel); close(hold_tstamp_fd);}
Re: [PATCH bpf-next v2 2/4] selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
In reply to Kui-Feng LeeQuoted 1 line · Kui-Feng Lee
Enable traffic monitoring for the test case tc_redirect/tc_redirect_dtime.
Alternatively, we might extend test_progs to have some new generic arg to enable trafficmon for a given set of tests (and then pass this flag in the CI):
./test_progs --traffic_monitor=t1,t2,t3...
Might be useful in case we need to debug some other test in the future.
Re: [PATCH bpf-next v2 2/4] selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
Quoted 10 lines · Stanislav Fomichev
Quoted 1 line · Kui-Feng Lee
Enable traffic monitoring for the test case tc_redirect/tc_redirect_dtime.
Alternatively, we might extend
test_progsto have some new generic arg to enable trafficmon for a given set of tests (and then pass this flag in the CI):./test_progs --traffic_monitor=t1,t2,t3...
Might be useful in case we need to debug some other test in the future.
We run a few test cases with network namespaces. So we need to specify namespaces to monitor. And, these namespaces are not created yet when a test starts. To adapt this approach, these test cases should be changed to use a generic way that create network namespaces when a test starts.
Or, we just monitor default network namespace. For test cases with network namespaces, they need to call these functions.
WDYT?
Re: [PATCH bpf-next v2 2/4] selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime.
Quoted 20 lines · Kui-Feng Lee
Quoted 10 lines · Stanislav Fomichev
Quoted 1 line · Kui-Feng Lee
Enable traffic monitoring for the test case tc_redirect/tc_redirect_dtime.
Alternatively, we might extend
test_progsto have some new generic arg to enable trafficmon for a given set of tests (and then pass this flag in the CI):./test_progs --traffic_monitor=t1,t2,t3...
Might be useful in case we need to debug some other test in the future.
We run a few test cases with network namespaces. So we need to specify namespaces to monitor. And, these namespaces are not created yet when a test starts. To adapt this approach, these test cases should be changed to use a generic way that create network namespaces when a test starts.
Quoted 4 lines
Or, we just monitor default network namespace. For test cases with network namespaces, they need to call these functions.
WDYT?
Ah, true, in this case ignore me :-)
PATCH v2 3/4 selftests/bpf: Monitor traffic for sockmap_listen.
In reply to Kui-Feng LeeEnable traffic monitor for each subtest of sockmap_listen.
--- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 8 ++++++++ 1 file changed, 8 insertions(+)
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +8 -0
@@ -28,6 +28,7 @@#include "test_sockmap_listen.skel.h" #include "sockmap_helpers.h" +#include "network_helpers.h" static void test_insert_invalid (struct test_sockmap_listen *skel__always_unused ,int family,int sotype,int mapfd)
@@ -1893,14 +1894,21 @@ static void test_udp_unix_redir(struct test_sockmap_listen *skel, struct bpf_map {const char *family_name, *map_name;char s[MAX_TEST_NAME ]; +struct tmonitor_ctx *tmon; family_name =family_str (family); map_name =map_type_str (map);snprintf (s,sizeof (s),"%s %s %s" , map_name, family_name, __func__);if (!test__start_subtest (s))return ; + + tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" ); +inet_unix_skb_redir_to_connected (skel, map, family);unix_inet_skb_redir_to_connected (skel, map, family); + +traffic_monitor_stop (tmon); } static void run_tests(struct test_sockmap_listen *skel, struct bpf_map *map,
Signature · 1 line
2.34.1
Re: [PATCH bpf-next v2 3/4] selftests/bpf: Monitor traffic for sockmap_listen.
Quoted 34 lines · Kui-Feng Lee
Enable traffic monitor for each subtest of
sockmap_listen.Signed-off-by Kui-Feng Lee <[email protected]>---
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 8 ++++++++1 file changed, 8 insertions(+)tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +0
-0@@ -28,6 +28,7 @@ #include "test_sockmap_listen.skel.h"
#include "sockmap_helpers.h"
+#include "network_helpers.h"
static void test_insert_invalid(structtest_sockmap_listen*skel__always_unused,
int family, int sotype, int mapfd)tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +0
-0@@ -1893,14 +1894,21 @@ static void test_udp_unix_redir(struct
test_sockmap_listen*skel, structbpf_map
{
const char*family_name,*map_name;
char s[MAX_TEST_NAME];tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +1
-0+struct tmonitor_ctx *tmon;
family_name=family_str(family);
map_name=map_type_str(map);
snprintf(s, sizeof(s), "%s %s %s", map_name, family_name, __func__);
if (!test__start_subtest(s))
return;
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c static void test_udp_unix_redir(struct test_sockmap_listen *skel, struct bpf_map · line 1906
Quoted 3 lines
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +3
-0+ + tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" );
Using ASSERT_TRUE() on a pointer is a bit strange, it's better to use ASSERT_NEQ(NULL) like patch 2.
Quoted 9 lines
+
inet_unix_skb_redir_to_connected(skel, map, family);unix_inet_skb_redir_to_connected(skel, map, family);tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +2
-0+ +traffic_monitor_stop (tmon);}
static void run_tests(structtest_sockmap_listen*skel, structbpf_map*map,
Re: [PATCH bpf-next v2 3/4] selftests/bpf: Monitor traffic for sockmap_listen.
Quoted 29 lines · Geliang Tang
Quoted 28 lines · Kui-Feng Lee
Enable traffic monitor for each subtest of
sockmap_listen.Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 8 ++++++++ 1 file changed, 8 insertions(+)tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +1
-0@@ -28,6 +28,7 @@ #include "test_sockmap_listen.skel.h" #include "sockmap_helpers.h" +#include "network_helpers.h" static void test_insert_invalid(struct test_sockmap_listen *skel __always_unused,int family, int sotype, int mapfd)
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +0
-0@@ -1893,14 +1894,21 @@ static void test_udp_unix_redir(struct
test_sockmap_listen*skel, structbpf_map
{
const char*family_name,*map_name;
char s[MAX_TEST_NAME];tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +1
-0+struct tmonitor_ctx *tmon;
Quoted 12 lines
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c static void test_udp_unix_redir(struct test_sockmap_listen *skel, struct bpf_map · line 1906
Quoted 9 lines
family_name=family_str(family);
map_name=map_type_str(map);
snprintf(s, sizeof(s), "%s %s %s",map_name,family_name,__func__);
if (!test__start_subtest(s))
return;tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +3
-0+ + tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" );Using
ASSERT_TRUE()on a pointer is a bit strange, it's better to useASSERT_NEQ(NULL)like patch 2.
Sure!
Quoted 11 lines
Quoted 10 lines
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +3
-0+ inet_unix_skb_redir_to_connected(skel, map, family); unix_inet_skb_redir_to_connected(skel, map, family); + +traffic_monitor_stop (tmon); } static void run_tests(struct test_sockmap_listen *skel, struct
bpf_map*map,
Re: [PATCH bpf-next v2 3/4] selftests/bpf: Monitor traffic for sockmap_listen.
Quoted 46 lines · Kui-Feng Lee
Quoted 29 lines · Geliang Tang
Quoted 28 lines · Kui-Feng Lee
Enable traffic monitor for each subtest of
sockmap_listen.Signed-off-by Kui-Feng Lee <[email protected]>--- tools/testing/selftests/bpf/prog_tests/sockmap_listen.c | 8 ++++++++ 1 file changed, 8 insertions(+)tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +1
-0@@ -28,6 +28,7 @@#include "test_sockmap_listen.skel.h" #include "sockmap_helpers.h" +#include "network_helpers.h" static void test_insert_invalid(struct test_sockmap_listen *skel __always_unused,int family, int sotype, int mapfd)
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +0
-0@@ -1893,14 +1894,21 @@ static void test_udp_unix_redir(struct
test_sockmap_listen*skel, structbpf_map{ const char *family_name, *map_name; char s[MAX_TEST_NAME];+ struct
tmonitor_ctx*tmon;Quoted 12 lines
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c static void test_udp_unix_redir(struct test_sockmap_listen *skel, struct bpf_map · line 1906
Quoted 9 lines
family_name = family_str(family); map_name = map_type_str(map); snprintf(s, sizeof(s), "%s %s %s", map_name, family_name, __func__); if (!test__start_subtest(s)) return;tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +3
-0+ + tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" );Using
ASSERT_TRUE()on a pointer is a bit strange, it's better to useASSERT_NEQ(NULL)like patch 2.
Quoted 1 line
Sure!
we have ASSERT_OK_PTR() for pointers
Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
tools/testing/selftests/bpf/prog_tests/sockmap_listen.c +3
-0+inet_unix_skb_redir_to_connected (skel, map, family);unix_inet_skb_redir_to_connected (skel, map, family); + +traffic_monitor_stop (tmon); } static void run_tests(struct test_sockmap_listen *skel, struct
bpf_map*map,
PATCH v2 4/4 selftests/bpf: Monitor traffic for select_reuseport.
In reply to Kui-Feng LeeEnable traffic monitoring for the subtests of select_reuseport.
--- tools/testing/selftests/bpf/prog_tests/select_reuseport.c | 7 +++++++ 1 file changed, 7 insertions(+)
tools/testing/selftests/bpf/prog_tests/select_reuseport.c +7 -0
@@ -22,6 +22,7 @@#include "test_progs.h" #include "test_select_reuseport_common.h" +#include "network_helpers.h" #define MAX_TEST_NAME 80#define MIN_TCPHDR_LEN 20
@@ -795,6 +796,7 @@ static void test_config(int sotype, sa_family_t family, bool inany) };char s[MAX_TEST_NAME ];const struct test *t; +struct tmonitor_ctx *tmon;for (t = tests; t < tests +ARRAY_SIZE (tests); t++) {if (t->need_sotype && t->need_sotype != sotype)
@@ -808,9 +810,14 @@ static void test_config(int sotype, sa_family_t family, bool inany)if (!test__start_subtest (s))continue ; + tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" ); +setup_per_test (sotype, family, inany, t->no_inner_map); t->fn (sotype, family);cleanup_per_test (t->no_inner_map); + +traffic_monitor_stop (tmon); } }
Signature · 1 line
2.34.1
Re: [PATCH bpf-next v2 4/4] selftests/bpf: Monitor traffic for select_reuseport.
Quoted 36 lines · Kui-Feng Lee
Enable traffic monitoring for the subtests of
select_reuseport.Signed-off-by Kui-Feng Lee <[email protected]>---
tools/testing/selftests/bpf/prog_tests/select_reuseport.c | 7
+++++++1 file changed, 7 insertions(+)diff --git
a/tools/testing/selftests/bpf/prog_tests/select_reuseport.cb/tools/testing/selftests/bpf/prog_tests/select_reuseport.cindex64c5f5eb2994..d3039957ee94100644tools/testing/selftests/bpf/prog_tests/select_reuseport.c +0
-0@@ -22,6 +22,7 @@
#include "test_progs.h"
#include "test_select_reuseport_common.h"
+#include "network_helpers.h"
#defineMAX_TEST_NAME80
#defineMIN_TCPHDR_LEN20tools/testing/selftests/bpf/prog_tests/select_reuseport.c +1
-0@@ -795,6 +796,7 @@ static void test_config(int sotype, sa_family_t family, bool inany) }; char s[MAX_TEST_NAME]; const struct test *t; +struct tmonitor_ctx *tmon;
for (t = tests; t < tests +ARRAY_SIZE(tests); t++) { if (t->need_sotype&&t->need_sotype!= sotype)tools/testing/selftests/bpf/prog_tests/select_reuseport.c +0
-0@@ -808,9 +810,14 @@ static void test_config(int sotype, sa_family_t family, bool inany) if (!test__start_subtest(s)) continue;
tools/testing/selftests/bpf/prog_tests/select_reuseport.c static void test_config(int sotype, sa_family_t family, bool inany) · line 814
Quoted 2 lines
tools/testing/selftests/bpf/prog_tests/select_reuseport.c +2
-0+ tmon =traffic_monitor_start (NULL ); +ASSERT_TRUE (tmon,"traffic_monitor_start" );
The same here. It's better to use ASSERT_NEQ(NULL) like patch 2.
Quoted 10 lines
+
setup_per_test(sotype, family, inany, t-Quoted 1 line
no_inner_map);
t->fn(sotype, family);cleanup_per_test(t->no_inner_map);tools/testing/selftests/bpf/prog_tests/select_reuseport.c +2
-0+ +traffic_monitor_stop (tmon);}
}
Quoted 50 lines · Kui-Feng Lee
Capture packets in the background for flaky test cases related to network features.
We have some flaky test cases that are difficult to debug without knowing what the traffic looks like. Capturing packets, the CI log and packet files may help developers to fix these flaky test cases.
This patch set monitors a few test cases. Recently, they have been showing flaky behavior.
IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 68, ifindex 1, SYN IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 60, ifindex 1, SYN, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 60, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, ACK IPv4 TCP packet: 127.0.0.1:48165 -> 127.0.0.1:36707, len 52, ifindex 1, FIN, ACK IPv4 TCP packet: 127.0.0.1:36707 -> 127.0.0.1:48165, len 52, ifindex 1, RST, ACK Packet file: packets-2172-86.log #280/87 select_reuseport/sockhash IPv4/TCP LOOPBACK test_detach_bpf:OKThe above block is the log of a test case. It shows every packets of a connection. The captured packets are stored in the file called packets-2172-86.log.
The following block is an example that monitors the network traffic of a test case. This test is running in the network namespace "testns". You can pass NULL to
traffic_monitor_start()if the entire test, fromtraffic_monitor_start()totraffic_monitor_stop(), is running in the same namespace.struct tmonitor_ctx *tmon; ... tmon = traffic_monitor_start("testns"); ASSERT_TRUE(tmon, "traffic_monitor_start"); ... test ... traffic_monitor_stop(tmon);
traffic_monitor_start()may fail, but we just ignore it since the failure doesn't affect the following main test.This feature is enabled only if BPF selftests are built with
TRAFFIC_MONITORvariable being defined. For example,make TRAFFIC_MONITOR=1 -C
tools/testing/selftests/bpfThis command will enable traffic monitoring for BPF selftests. That means we have to turn it on to get the log at CI.
Quoted 7 lines
---
Changes from v1:
- Initialize
log_fdintraffic_monitor_start().- Remove redundant including.
Sorry for not updating changes correctly.
No more tcpdump, it moves to use call pcap directly
in a background thread. Packets are wrote to a packet file. In the log, it prints parsed information of TCP or UDP packets. For other packets, just print a string "Packet captured" to indicate a packet has been captured.
Quoted 17 lines
v1: https://lore.kernel.org/all/[email protected]/
Kui-Feng Lee (4):
selftests/bpf: Add traffic monitor functions. selftests/bpf: Monitor traffic for tc_redirect/tc_redirect_dtime. selftests/bpf: Monitor traffic forsockmap_listen. selftests/bpf: Monitor traffic forselect_reuseport.tools/testing/selftests/bpf/Makefile | 5 + tools/testing/selftests/bpf/network_helpers.c | 382 ++++++++++++++++++ tools/testing/selftests/bpf/network_helpers.h | 16 + .../bpf/prog_tests/select_reuseport.c | 7 + .../selftests/bpf/prog_tests/sockmap_listen.c | 8 + .../selftests/bpf/prog_tests/tc_redirect.c | 5 + 6 files changed, 423 insertions(+)