Message ID | 20220902141715.1038615-1-imagedong@tencent.com (mailing list archive) |
---|---|
State | Superseded |
Delegated to: | Netdev Maintainers |
Headers | show |
Series | [net-next,v2] net: skb: export skb drop reaons to user by TRACE_DEFINE_ENUM | expand |
On Fri, Sep 2, 2022 at 7:18 AM <menglong8.dong@gmail.com> wrote: > > From: Menglong Dong <imagedong@tencent.com> > > As Eric reported, the 'reason' field is not presented when trace the > kfree_skb event by perf: > > $ perf record -e skb:kfree_skb -a sleep 10 > $ perf script > ip_defrag 14605 [021] 221.614303: skb:kfree_skb: > skbaddr=0xffff9d2851242700 protocol=34525 location=0xffffffffa39346b1 > reason: > > The cause seems to be passing kernel address directly to TP_printk(), > which is not right. As the enum 'skb_drop_reason' is not exported to > user space through TRACE_DEFINE_ENUM(), perf can't get the drop reason > string from the 'reason' field, which is a number. > > Therefore, we introduce the macro DEFINE_DROP_REASON(), which is used > to define the trace enum by TRACE_DEFINE_ENUM(). With the help of > DEFINE_DROP_REASON(), now we can remove the auto-generate that we > introduced in the commit ec43908dd556 > ("net: skb: use auto-generation to convert skb drop reason to string"), > and define the string array 'drop_reasons'. > > Hmmmm...now we come back to the situation that have to maintain drop > reasons in both enum skb_drop_reason and DEFINE_DROP_REASON. But they > are both in dropreason.h, which makes it easier. > > After this commit, now the format of kfree_skb is like this: > > $ cat /tracing/events/skb/kfree_skb/format > name: kfree_skb > ID: 1524 > format: > field:unsigned short common_type; offset:0; size:2; signed:0; > field:unsigned char common_flags; offset:2; size:1; signed:0; > field:unsigned char common_preempt_count; offset:3; size:1; signed:0; > field:int common_pid; offset:4; size:4; signed:1; > > field:void * skbaddr; offset:8; size:8; signed:0; > field:void * location; offset:16; size:8; signed:0; > field:unsigned short protocol; offset:24; size:2; signed:0; > field:enum skb_drop_reason reason; offset:28; size:4; signed:0; > > print fmt: "skbaddr=%p protocol=%u location=%p reason: %s", REC->skbaddr, REC->protocol, REC->location, __print_symbolic(REC->reason, { 1, "NOT_SPECIFIED" }, { 2, "NO_SOCKET" } ...... > > Reported-by: Eric Dumazet <edumazet@google.com> Note that I also provided the sha1 of the faulty patch. You should add a corresponding Fixes: tag, to help both humans and bots. This would also hint that this patch should target net tree, not net-next ? > Signed-off-by: Menglong Dong <imagedong@tencent.com> > --- > v2: > - undef FN/FNe after use it (Jakub Kicinski) I would love some feedback from Steven :)
On Fri, Sep 2, 2022 at 11:43 PM Eric Dumazet <edumazet@google.com> wrote: > > On Fri, Sep 2, 2022 at 7:18 AM <menglong8.dong@gmail.com> wrote: > > > > From: Menglong Dong <imagedong@tencent.com> > > > > As Eric reported, the 'reason' field is not presented when trace the > > kfree_skb event by perf: > > > > $ perf record -e skb:kfree_skb -a sleep 10 > > $ perf script > > ip_defrag 14605 [021] 221.614303: skb:kfree_skb: > > skbaddr=0xffff9d2851242700 protocol=34525 location=0xffffffffa39346b1 > > reason: > > > > The cause seems to be passing kernel address directly to TP_printk(), > > which is not right. As the enum 'skb_drop_reason' is not exported to > > user space through TRACE_DEFINE_ENUM(), perf can't get the drop reason > > string from the 'reason' field, which is a number. > > > > Therefore, we introduce the macro DEFINE_DROP_REASON(), which is used > > to define the trace enum by TRACE_DEFINE_ENUM(). With the help of > > DEFINE_DROP_REASON(), now we can remove the auto-generate that we > > introduced in the commit ec43908dd556 > > ("net: skb: use auto-generation to convert skb drop reason to string"), > > and define the string array 'drop_reasons'. > > > > Hmmmm...now we come back to the situation that have to maintain drop > > reasons in both enum skb_drop_reason and DEFINE_DROP_REASON. But they > > are both in dropreason.h, which makes it easier. > > > > After this commit, now the format of kfree_skb is like this: > > > > $ cat /tracing/events/skb/kfree_skb/format > > name: kfree_skb > > ID: 1524 > > format: > > field:unsigned short common_type; offset:0; size:2; signed:0; > > field:unsigned char common_flags; offset:2; size:1; signed:0; > > field:unsigned char common_preempt_count; offset:3; size:1; signed:0; > > field:int common_pid; offset:4; size:4; signed:1; > > > > field:void * skbaddr; offset:8; size:8; signed:0; > > field:void * location; offset:16; size:8; signed:0; > > field:unsigned short protocol; offset:24; size:2; signed:0; > > field:enum skb_drop_reason reason; offset:28; size:4; signed:0; > > > > print fmt: "skbaddr=%p protocol=%u location=%p reason: %s", REC->skbaddr, REC->protocol, REC->location, __print_symbolic(REC->reason, { 1, "NOT_SPECIFIED" }, { 2, "NO_SOCKET" } ...... > > > > Reported-by: Eric Dumazet <edumazet@google.com> > > Note that I also provided the sha1 of the faulty patch. > > You should add a corresponding Fixes: tag, to help both humans and bots. > Okay, I'll add the Fixes tag......and the Link tag. > This would also hint that this patch should target net tree, not net-next ? > It seems that the net tree is more suitable. I'll change it. Thanks! > > Signed-off-by: Menglong Dong <imagedong@tencent.com> > > --- > > v2: > > - undef FN/FNe after use it (Jakub Kicinski) > > I would love some feedback from Steven :) Hi, Steven! Do you have any feedback? Yes, I add the TRACE_DEFINE_ENUM() back! (Hmm...I deleted it before, without knowing its function :/ ) Thanks! Menglong Dong
On Fri, 2 Sep 2022 08:43:07 -0700 Eric Dumazet <edumazet@google.com> wrote: > > --- > > v2: > > - undef FN/FNe after use it (Jakub Kicinski) > > I would love some feedback from Steven :) The undef should be fine. I usually do not, but that's more of a preference than a rule. As long as the undef is done after the C portion of where the macro is used: +#undef FN +#define FN(reason) TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason); +DEFINE_DROP_REASON(FN, FN) <<<--- C portion +#undef FN +#undef FNe +#define FN(reason) { SKB_DROP_REASON_##reason, #reason }, +#define FNe(reason) { SKB_DROP_REASON_##reason, #reason } + /* * Tracepoint for free an sk_buff: */ @@ -35,9 +44,13 @@ TRACE_EVENT(kfree_skb, TP_printk("skbaddr=%p protocol=%u location=%p reason: %s", __entry->skbaddr, __entry->protocol, __entry->location, - drop_reasons[__entry->reason]) + __print_symbolic(__entry->reason, + DEFINE_DROP_REASON(FN, FNe))) <<<--- C portion ); +#undef FN +#undef FNe So for this part: Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org> -- Steve
On Mon, Sep 5, 2022 at 7:37 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > On Fri, 2 Sep 2022 08:43:07 -0700 > Eric Dumazet <edumazet@google.com> wrote: > > > > --- > > > v2: > > > - undef FN/FNe after use it (Jakub Kicinski) > > > > I would love some feedback from Steven :) > > The undef should be fine. I usually do not, but that's more of a preference > than a rule. > > As long as the undef is done after the C portion of where the macro is used: > > +#undef FN > +#define FN(reason) TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason); > +DEFINE_DROP_REASON(FN, FN) <<<--- C portion > > +#undef FN > +#undef FNe > +#define FN(reason) { SKB_DROP_REASON_##reason, #reason }, > +#define FNe(reason) { SKB_DROP_REASON_##reason, #reason } > + > /* > * Tracepoint for free an sk_buff: > */ > @@ -35,9 +44,13 @@ TRACE_EVENT(kfree_skb, > > TP_printk("skbaddr=%p protocol=%u location=%p reason: %s", > __entry->skbaddr, __entry->protocol, __entry->location, > - drop_reasons[__entry->reason]) > + __print_symbolic(__entry->reason, > + DEFINE_DROP_REASON(FN, FNe))) <<<--- C portion > ); > > +#undef FN > +#undef FNe > > So for this part: Reviewed-by: Steven Rostedt (Google) <rostedt@goodmis.org> > > -- Steve This seems broken again (tried on latest net-next tree) perf script sshd 12192 [006] 200.090602: skb:kfree_skb: skbaddr=0x2fa000400000 protocol=7844 location=0x2fa000001ea4 reason: swapper 0 [030] 200.497468: skb:kfree_skb: skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 reason: kworker/30:1-ev 308 [030] 200.497476: skb:kfree_skb: skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 reason: swapper 0 [009] 200.957881: skb:kfree_skb: skbaddr=0x2fa400400000 protocol=12195 location=0x2fa400002fa3 reason: swapper 0 [026] 201.515769: skb:kfree_skb: skbaddr=0xb600400001 protocol=65535 location=0x9080c62800000000 reason: kworker/26:1-mm 276 [026] 201.515797: skb:kfree_skb: skbaddr=0xb600400001 protocol=65535 location=0x9080c62800000000 reason: kworker/26:1-mm 276 [026] 201.515802: skb:kfree_skb: skbaddr=0x2f9f00400000 protocol=12189 location=0x2f9f00002f9d reason: swapper 0 [030] 201.521484: skb:kfree_skb: skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 reason: kworker/30:1-ev 308 [030] 201.521491: skb:kfree_skb: skbaddr=0x2fa100400000 protocol=12192 location=0x2fa100002fa0 reason:
On Thu, 27 Oct 2022 08:32:02 -0700 Eric Dumazet <edumazet@google.com> wrote: > This seems broken again (tried on latest net-next tree) > > perf script Do you also have the latest perf and the latest libtraceevent installed? -- Steve > sshd 12192 [006] 200.090602: skb:kfree_skb: > skbaddr=0x2fa000400000 protocol=7844 location=0x2fa000001ea4 reason: > swapper 0 [030] 200.497468: skb:kfree_skb: > skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 > reason: > kworker/30:1-ev 308 [030] 200.497476: skb:kfree_skb: > skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 > reason: > swapper 0 [009] 200.957881: skb:kfree_skb: > skbaddr=0x2fa400400000 protocol=12195 location=0x2fa400002fa3 reason: > swapper 0 [026] 201.515769: skb:kfree_skb: > skbaddr=0xb600400001 protocol=65535 location=0x9080c62800000000 > reason: > kworker/26:1-mm 276 [026] 201.515797: skb:kfree_skb: > skbaddr=0xb600400001 protocol=65535 location=0x9080c62800000000 > reason: > kworker/26:1-mm 276 [026] 201.515802: skb:kfree_skb: > skbaddr=0x2f9f00400000 protocol=12189 location=0x2f9f00002f9d reason: > swapper 0 [030] 201.521484: skb:kfree_skb: > skbaddr=0xba00400001 protocol=65535 location=0x9080c62800000000 > reason: > kworker/30:1-ev 308 [030] 201.521491: skb:kfree_skb: > skbaddr=0x2fa100400000 protocol=12192 location=0x2fa100002fa0 reason:
On Thu, Oct 27, 2022 at 8:43 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > On Thu, 27 Oct 2022 08:32:02 -0700 > Eric Dumazet <edumazet@google.com> wrote: > > > This seems broken again (tried on latest net-next tree) > > > > perf script > > Do you also have the latest perf and the latest libtraceevent installed? > I tried a more recent perf binary we have, but it is also not right. I guess I will have to request a new perf binary at Google :/ perf5 script swapper 0 [030] 4147.704606: skb:kfree_skb: [UNKNOWN EVENT] kworker/30:1-ev 308 [030] 4147.704615: skb:kfree_skb: [UNKNOWN EVENT] swapper 0 [030] 4148.048173: skb:kfree_skb: [UNKNOWN EVENT] kworker/30:1-ev 308 [030] 4148.048179: skb:kfree_skb: [UNKNOWN EVENT] swapper 0 [008] 4148.048773: skb:kfree_skb: [UNKNOWN EVENT] swapper 0 [030] 4148.112271: skb:kfree_skb: [UNKNOWN EVENT] kworker/30:1-ev 308 [030] 4148.112280: skb:kfree_skb: [UNKNOWN EVENT] swapper 0 [030] 4148.720149: skb:kfree_skb: [UNKNOWN EVENT] kworker/30:1-ev 308 [030] 4148.720155: skb:kfree_skb: [UNKNOWN EVENT] swapper 0 [030] 4149.072141: skb:kfree_skb: [UNKNOWN EVENT] kworker/30:1-ev 308 [030] 4149.072149: skb:kfree_skb: [UNKNOWN EVENT]
On Thu, 27 Oct 2022 09:28:42 -0700 Eric Dumazet <edumazet@google.com> wrote: > I tried a more recent perf binary we have, but it is also not right. > > I guess I will have to request a new perf binary at Google :/ You could always use trace-cmd ;-) -- Steve
> On 2022/10/28 00:29,“Eric Dumazet”<edumazet@google.com> write: > > On Thu, Oct 27, 2022 at 8:43 AM Steven Rostedt <rostedt@goodmis.org> wrote: > > > > On Thu, 27 Oct 2022 08:32:02 -0700 > > Eric Dumazet <edumazet@google.com> wrote: > > > > > This seems broken again (tried on latest net-next tree) > > > > > > perf script > > > > Do you also have the latest perf and the latest libtraceevent installed? > > > > I tried a more recent perf binary we have, but it is also not right. > > I guess I will have to request a new perf binary at Google :/ > Ennn...I just have a try, and it seems fine: sudo perf script swapper 0 [049] 401460.138078: skb:kfree_skb: skbaddr=0xffff8881076d2400 protocol=2048 location=0xffffffff81d86546 reason: TCP_INVALID_SEQUENCE nc 429418 [006] 401465.652282: skb:kfree_skb: skbaddr=0xffff8881ea66cf00 protocol=0 location=0xffffffff81e0b48f reason: NOT_SPECIFIED nc 429418 [006] 401465.652293: skb:kfree_skb: skbaddr=0xffff8881ea66cf00 protocol=0 location=0xffffffff81e0b48f reason: NOT_SPECIFIED nc 429418 [006] 401465.652538: skb:kfree_skb: skbaddr=0xffff888133c014e8 protocol=2048 location=0xffffffff81d983a1 reason: NO_SOCKET The version of the perf I used is 6.0.3-1: sudo apt info linux-perf Package: linux-perf Version: 6.0.3-1 Thanks! Menglong Dong > perf5 script > swapper 0 [030] 4147.704606: skb:kfree_skb: [UNKNOWN EVENT] > kworker/30:1-ev 308 [030] 4147.704615: skb:kfree_skb: [UNKNOWN EVENT] > swapper 0 [030] 4148.048173: skb:kfree_skb: [UNKNOWN EVENT] > kworker/30:1-ev 308 [030] 4148.048179: skb:kfree_skb: [UNKNOWN EVENT] > swapper 0 [008] 4148.048773: skb:kfree_skb: [UNKNOWN EVENT] > swapper 0 [030] 4148.112271: skb:kfree_skb: [UNKNOWN EVENT] > kworker/30:1-ev 308 [030] 4148.112280: skb:kfree_skb: [UNKNOWN EVENT] > swapper 0 [030] 4148.720149: skb:kfree_skb: [UNKNOWN EVENT] > kworker/30:1-ev 308 [030] 4148.720155: skb:kfree_skb: [UNKNOWN EVENT] > swapper 0 [030] 4149.072141: skb:kfree_skb: [UNKNOWN EVENT] > kworker/30:1-ev 308 [030] 4149.072149: skb:kfree_skb: [UNKNOWN EVENT]
diff --git a/include/net/dropreason.h b/include/net/dropreason.h index fae9b40e54fa..c1cbcdbaf149 100644 --- a/include/net/dropreason.h +++ b/include/net/dropreason.h @@ -3,6 +3,73 @@ #ifndef _LINUX_DROPREASON_H #define _LINUX_DROPREASON_H +#define DEFINE_DROP_REASON(FN, FNe) \ + FN(NOT_SPECIFIED) \ + FN(NO_SOCKET) \ + FN(PKT_TOO_SMALL) \ + FN(TCP_CSUM) \ + FN(SOCKET_FILTER) \ + FN(UDP_CSUM) \ + FN(NETFILTER_DROP) \ + FN(OTHERHOST) \ + FN(IP_CSUM) \ + FN(IP_INHDR) \ + FN(IP_RPFILTER) \ + FN(UNICAST_IN_L2_MULTICAST) \ + FN(XFRM_POLICY) \ + FN(IP_NOPROTO) \ + FN(SOCKET_RCVBUFF) \ + FN(PROTO_MEM) \ + FN(TCP_MD5NOTFOUND) \ + FN(TCP_MD5UNEXPECTED) \ + FN(TCP_MD5FAILURE) \ + FN(SOCKET_BACKLOG) \ + FN(TCP_FLAGS) \ + FN(TCP_ZEROWINDOW) \ + FN(TCP_OLD_DATA) \ + FN(TCP_OVERWINDOW) \ + FN(TCP_OFOMERGE) \ + FN(TCP_RFC7323_PAWS) \ + FN(TCP_INVALID_SEQUENCE) \ + FN(TCP_RESET) \ + FN(TCP_INVALID_SYN) \ + FN(TCP_CLOSE) \ + FN(TCP_FASTOPEN) \ + FN(TCP_OLD_ACK) \ + FN(TCP_TOO_OLD_ACK) \ + FN(TCP_ACK_UNSENT_DATA) \ + FN(TCP_OFO_QUEUE_PRUNE) \ + FN(TCP_OFO_DROP) \ + FN(IP_OUTNOROUTES) \ + FN(BPF_CGROUP_EGRESS) \ + FN(IPV6DISABLED) \ + FN(NEIGH_CREATEFAIL) \ + FN(NEIGH_FAILED) \ + FN(NEIGH_QUEUEFULL) \ + FN(NEIGH_DEAD) \ + FN(TC_EGRESS) \ + FN(QDISC_DROP) \ + FN(CPU_BACKLOG) \ + FN(XDP) \ + FN(TC_INGRESS) \ + FN(UNHANDLED_PROTO) \ + FN(SKB_CSUM) \ + FN(SKB_GSO_SEG) \ + FN(SKB_UCOPY_FAULT) \ + FN(DEV_HDR) \ + FN(DEV_READY) \ + FN(FULL_RING) \ + FN(NOMEM) \ + FN(HDR_TRUNC) \ + FN(TAP_FILTER) \ + FN(TAP_TXFILTER) \ + FN(ICMP_CSUM) \ + FN(INVALID_PROTO) \ + FN(IP_INADDRERRORS) \ + FN(IP_INNOROUTES) \ + FN(PKT_TOO_BIG) \ + FNe(MAX) + /** * enum skb_drop_reason - the reasons of skb drops * diff --git a/include/trace/events/skb.h b/include/trace/events/skb.h index 45264e4bb254..50a974f7dfb4 100644 --- a/include/trace/events/skb.h +++ b/include/trace/events/skb.h @@ -9,6 +9,15 @@ #include <linux/netdevice.h> #include <linux/tracepoint.h> +#undef FN +#define FN(reason) TRACE_DEFINE_ENUM(SKB_DROP_REASON_##reason); +DEFINE_DROP_REASON(FN, FN) + +#undef FN +#undef FNe +#define FN(reason) { SKB_DROP_REASON_##reason, #reason }, +#define FNe(reason) { SKB_DROP_REASON_##reason, #reason } + /* * Tracepoint for free an sk_buff: */ @@ -35,9 +44,13 @@ TRACE_EVENT(kfree_skb, TP_printk("skbaddr=%p protocol=%u location=%p reason: %s", __entry->skbaddr, __entry->protocol, __entry->location, - drop_reasons[__entry->reason]) + __print_symbolic(__entry->reason, + DEFINE_DROP_REASON(FN, FNe))) ); +#undef FN +#undef FNe + TRACE_EVENT(consume_skb, TP_PROTO(struct sk_buff *skb), diff --git a/net/core/.gitignore b/net/core/.gitignore deleted file mode 100644 index df1e74372cce..000000000000 --- a/net/core/.gitignore +++ /dev/null @@ -1 +0,0 @@ -dropreason_str.c diff --git a/net/core/Makefile b/net/core/Makefile index e8ce3bd283a6..5857cec87b83 100644 --- a/net/core/Makefile +++ b/net/core/Makefile @@ -5,7 +5,7 @@ obj-y := sock.o request_sock.o skbuff.o datagram.o stream.o scm.o \ gen_stats.o gen_estimator.o net_namespace.o secure_seq.o \ - flow_dissector.o dropreason_str.o + flow_dissector.o obj-$(CONFIG_SYSCTL) += sysctl_net_core.o @@ -40,23 +40,3 @@ obj-$(CONFIG_NET_SOCK_MSG) += skmsg.o obj-$(CONFIG_BPF_SYSCALL) += sock_map.o obj-$(CONFIG_BPF_SYSCALL) += bpf_sk_storage.o obj-$(CONFIG_OF) += of_net.o - -clean-files := dropreason_str.c - -quiet_cmd_dropreason_str = GEN $@ -cmd_dropreason_str = awk -F ',' 'BEGIN{ print "\#include <net/dropreason.h>\n"; \ - print "const char * const drop_reasons[] = {" }\ - /^enum skb_drop/ { dr=1; }\ - /^\};/ { dr=0; }\ - /^\tSKB_DROP_REASON_/ {\ - if (dr) {\ - sub(/\tSKB_DROP_REASON_/, "", $$1);\ - printf "\t[SKB_DROP_REASON_%s] = \"%s\",\n", $$1, $$1;\ - }\ - }\ - END{ print "};" }' $< > $@ - -$(obj)/dropreason_str.c: $(srctree)/include/net/dropreason.h - $(call cmd,dropreason_str) - -$(obj)/dropreason_str.o: $(obj)/dropreason_str.c diff --git a/net/core/skbuff.c b/net/core/skbuff.c index 48ecfbf29174..f1b8b20fc20b 100644 --- a/net/core/skbuff.c +++ b/net/core/skbuff.c @@ -91,7 +91,11 @@ static struct kmem_cache *skbuff_ext_cache __ro_after_init; int sysctl_max_skb_frags __read_mostly = MAX_SKB_FRAGS; EXPORT_SYMBOL(sysctl_max_skb_frags); -/* The array 'drop_reasons' is auto-generated in dropreason_str.c */ +#undef FN +#define FN(reason) [SKB_DROP_REASON_##reason] = #reason, +const char * const drop_reasons[] = { + DEFINE_DROP_REASON(FN, FN) +}; EXPORT_SYMBOL(drop_reasons); /**