[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CADxym3an7YynHoEn6asiZ=V588BBp+3+ycg_zpexkbUmdcd+bg@mail.gmail.com>
Date: Mon, 13 Jun 2022 10:41:20 +0800
From: Menglong Dong <menglong8.dong@...il.com>
To: Eric Dumazet <edumazet@...gle.com>
Cc: Steven Rostedt <rostedt@...dmis.org>,
Ingo Molnar <mingo@...hat.com>,
David Miller <davem@...emloft.net>,
Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
David Ahern <dsahern@...nel.org>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
Menglong Dong <imagedong@...cent.com>,
Martin KaFai Lau <kafai@...com>,
Talal Ahmad <talalahmad@...gle.com>,
Kees Cook <keescook@...omium.org>,
Dongli Zhang <dongli.zhang@...cle.com>,
LKML <linux-kernel@...r.kernel.org>,
netdev <netdev@...r.kernel.org>,
Jiang Biao <benbjiang@...cent.com>,
Hao Peng <flyingpeng@...cent.com>
Subject: Re: [PATCH net-next v3 7/9] net: tcp: add skb drop reasons to tcp tw
code path
On Fri, Jun 10, 2022 at 5:04 PM Eric Dumazet <edumazet@...gle.com> wrote:
>
> On Thu, Jun 9, 2022 at 8:45 PM <menglong8.dong@...il.com> wrote:
> >
> > From: Menglong Dong <imagedong@...cent.com>
> >
> > In order to get the reasons of skb drops, add a function argument of
> > type 'enum skb_drop_reason *reason' to tcp_timewait_state_process().
> >
> > In the origin code, all packets to time-wait socket are treated as
> > dropping with kfree_skb(), which can make users confused. Therefore,
> > we use consume_skb() for the skbs that are 'good'. We can check the
> > value of 'reason' to decide use kfree_skb() or consume_skb().
> >
> > The new reason 'TIMEWAIT' is added for the case that the skb is dropped
> > as the socket in time-wait state.
> >
> > Reviewed-by: Jiang Biao <benbjiang@...cent.com>
> > Reviewed-by: Hao Peng <flyingpeng@...cent.com>
> > Signed-off-by: Menglong Dong <imagedong@...cent.com>
> > Reported-by: Eric Dumazet <edumazet@...gle.com>
>
> I have suggested that consume_skb() could be an alias of
> kfree_skb_reason(skb, SKB_NOT_DROPPED),
> or something like that.
>
> In order to avoid silly constructs all over the places :
>
> if (reason)
> kfree_skb_reason(skb, reason);
> else
> consume_skb(skb);
>
> ->
>
> kfree_skb_reason(skb, reason);
>
This seems to be a good idea in some cases. However, I
am a little worried about that SKB_NOT_DROPPED can
be passed to kfree_skb_reason() by mistake, which happened
before. (Maybe I'm just overly worried, as this can be avoided
during we coding)
> > ---
> > v2:
> > - skb is not freed on TCP_TW_ACK and 'ret' is not initizalized, fix
> > it (Eric Dumazet)
> > ---
> > include/net/dropreason.h | 6 ++++++
> > include/net/tcp.h | 7 ++++---
> > net/ipv4/tcp_ipv4.c | 9 ++++++++-
> > net/ipv4/tcp_minisocks.c | 24 ++++++++++++++++++++----
> > net/ipv6/tcp_ipv6.c | 8 +++++++-
> > 5 files changed, 45 insertions(+), 9 deletions(-)
> >
> > diff --git a/include/net/dropreason.h b/include/net/dropreason.h
> > index 86e89d3022b5..3c6f8e0f7f16 100644
> > --- a/include/net/dropreason.h
> > +++ b/include/net/dropreason.h
> > @@ -258,6 +258,12 @@ enum skb_drop_reason {
> > * socket is full, corresponding to LINUX_MIB_TCPREQQFULLDROP
> > */
> > SKB_DROP_REASON_TCP_REQQFULLDROP,
> > + /**
> > + * @SKB_DROP_REASON_TIMEWAIT: socket is in time-wait state and all
> > + * packet that received will be treated as 'drop', except a good
> > + * 'SYN' packet
> > + */
> > + SKB_DROP_REASON_TIMEWAIT,
> > /**
> > * @SKB_DROP_REASON_MAX: the maximum of drop reason, which shouldn't be
> > * used as a real 'reason'
> > diff --git a/include/net/tcp.h b/include/net/tcp.h
> > index 082dd0627e2e..88217b8d95ac 100644
> > --- a/include/net/tcp.h
> > +++ b/include/net/tcp.h
> > @@ -380,9 +380,10 @@ enum tcp_tw_status {
> > };
> >
> >
> > -enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
> > - struct sk_buff *skb,
> > - const struct tcphdr *th);
> > +enum tcp_tw_status
> > +tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > + const struct tcphdr *th,
> > + enum skb_drop_reason *reason);
> > struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
> > struct request_sock *req, bool fastopen,
> > bool *lost_race);
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index 804fc5e0203e..e7bd2f410a4a 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -2134,7 +2134,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
> > inet_twsk_put(inet_twsk(sk));
> > goto csum_error;
> > }
> > - switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
> > + switch (tcp_timewait_state_process(inet_twsk(sk), skb, th,
> > + &drop_reason)) {
> > case TCP_TW_SYN: {
> > struct sock *sk2 = inet_lookup_listener(dev_net(skb->dev),
> > &tcp_hashinfo, skb,
> > @@ -2150,11 +2151,17 @@ int tcp_v4_rcv(struct sk_buff *skb)
> > refcounted = false;
> > goto process;
> > }
> > + /* TCP_FLAGS or NO_SOCKET? */
> > + SKB_DR_SET(drop_reason, TCP_FLAGS);
> > }
> > /* to ACK */
> > fallthrough;
> > case TCP_TW_ACK:
> > tcp_v4_timewait_ack(sk, skb);
> > + if (!drop_reason) {
> > + consume_skb(skb);
> > + return 0;
> > + }
>
> Sorry, this is the kind of change which makes the whole exercise
> source of numerous problems later
> when backports are needed.
>
> You are sending patches today, but we are not sure you will be willing
> to spend days of work and tests to validate
> future backports with conflicts.
With making kfree_skb_reason(skb, SKB_NOT_DROPPED) consume_skb(skb),
it seems that the code here doesn't need to be changed anymore.
>
> > break;
> > case TCP_TW_RST:
> > tcp_v4_send_reset(sk, skb);
> > diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
> > index 1a21018f6f64..329724118b7f 100644
> > --- a/net/ipv4/tcp_minisocks.c
> > +++ b/net/ipv4/tcp_minisocks.c
> > @@ -83,13 +83,15 @@ tcp_timewait_check_oow_rate_limit(struct inet_timewait_sock *tw,
> > */
> > enum tcp_tw_status
> > tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > - const struct tcphdr *th)
> > + const struct tcphdr *th,
> > + enum skb_drop_reason *reason)
> > {
> > struct tcp_options_received tmp_opt;
> > struct tcp_timewait_sock *tcptw = tcp_twsk((struct sock *)tw);
> > bool paws_reject = false;
> >
> > tmp_opt.saw_tstamp = 0;
> > + *reason = SKB_DROP_REASON_NOT_SPECIFIED;
> > if (th->doff > (sizeof(*th) >> 2) && tcptw->tw_ts_recent_stamp) {
> > tcp_parse_options(twsk_net(tw), skb, &tmp_opt, 0, NULL);
> >
> > @@ -113,11 +115,16 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > return tcp_timewait_check_oow_rate_limit(
> > tw, skb, LINUX_MIB_TCPACKSKIPPEDFINWAIT2);
> >
> > - if (th->rst)
> > + if (th->rst) {
> > + SKB_DR_SET(*reason, TCP_RESET);
> > goto kill;
> > + }
> >
> > - if (th->syn && !before(TCP_SKB_CB(skb)->seq, tcptw->tw_rcv_nxt))
> > + if (th->syn && !before(TCP_SKB_CB(skb)->seq,
> > + tcptw->tw_rcv_nxt)) {
> > + SKB_DR_SET(*reason, TCP_FLAGS);
> > return TCP_TW_RST;
> > + }
> >
> > /* Dup ACK? */
> > if (!th->ack ||
> > @@ -143,6 +150,9 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > }
> >
> > inet_twsk_reschedule(tw, TCP_TIMEWAIT_LEN);
> > +
> > + /* skb should be free normally on this case. */
> > + *reason = SKB_NOT_DROPPED_YET;
> > return TCP_TW_ACK;
> > }
> >
> > @@ -174,6 +184,7 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > * protocol bug yet.
> > */
> > if (twsk_net(tw)->ipv4.sysctl_tcp_rfc1337 == 0) {
> > + SKB_DR_SET(*reason, TCP_RESET);
> > kill:
> > inet_twsk_deschedule_put(tw);
> > return TCP_TW_SUCCESS;
> > @@ -216,11 +227,14 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > if (isn == 0)
> > isn++;
> > TCP_SKB_CB(skb)->tcp_tw_isn = isn;
> > + *reason = SKB_NOT_DROPPED_YET;
> > return TCP_TW_SYN;
> > }
> >
> > - if (paws_reject)
> > + if (paws_reject) {
> > + SKB_DR_SET(*reason, TCP_RFC7323_PAWS);
> > __NET_INC_STATS(twsk_net(tw), LINUX_MIB_PAWSESTABREJECTED);
> > + }
> >
> > if (!th->rst) {
> > /* In this case we must reset the TIMEWAIT timer.
> > @@ -232,9 +246,11 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
> > if (paws_reject || th->ack)
> > inet_twsk_reschedule(tw, TCP_TIMEWAIT_LEN);
> >
> > + SKB_DR_OR(*reason, TIMEWAIT);
> > return tcp_timewait_check_oow_rate_limit(
> > tw, skb, LINUX_MIB_TCPACKSKIPPEDTIMEWAIT);
> > }
> > + SKB_DR_SET(*reason, TCP_RESET);
> > inet_twsk_put(tw);
> > return TCP_TW_SUCCESS;
> > }
> > diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> > index 0d2ba9d90529..41551a3b679b 100644
> > --- a/net/ipv6/tcp_ipv6.c
> > +++ b/net/ipv6/tcp_ipv6.c
> > @@ -1795,7 +1795,8 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
> > goto csum_error;
> > }
> >
> > - switch (tcp_timewait_state_process(inet_twsk(sk), skb, th)) {
> > + switch (tcp_timewait_state_process(inet_twsk(sk), skb, th,
> > + &drop_reason)) {
> > case TCP_TW_SYN:
> > {
> > struct sock *sk2;
> > @@ -1815,11 +1816,16 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
> > refcounted = false;
> > goto process;
> > }
> > + SKB_DR_SET(drop_reason, TCP_FLAGS);
> > }
> > /* to ACK */
> > fallthrough;
> > case TCP_TW_ACK:
> > tcp_v6_timewait_ack(sk, skb);
> > + if (!drop_reason) {
> > + consume_skb(skb);
> > + return 0;
> > + }
> > break;
> > case TCP_TW_RST:
> > tcp_v6_send_reset(sk, skb);
> > --
> > 2.36.1
> >
Powered by blists - more mailing lists