From: Jamal Hadi Salim <jhs@mojatatu.com>
To: Eric Dumazet <edumazet@google.com>
Cc: renmingshuai <renmingshuai@huawei.com>,
xiyou.wangcong@gmail.com, jiri@resnulli.us, davem@davemloft.net,
vladbu@nvidia.com, netdev@vger.kernel.org, yanan@huawei.com,
liaichun@huawei.com, caowangbao@huawei.com,
Eric Dumazet <eric.dumazet@gmail.com>,
Jakub Kicinski <kuba@kernel.org>,
Paolo Abeni <pabeni@redhat.com>,
Victor Nogueira <victor@mojatatu.com>,
Pedro Tammela <pctammela@mojatatu.com>,
Davide Caratti <dcaratti@redhat.com>
Subject: Re: [PATCH] net/sched: Forbid assigning mirred action to a filter attached to the egress
Date: Sun, 24 Mar 2024 11:27:42 -0400 [thread overview]
Message-ID: <CAM0EoM=28nsomF9PniYNtpjD4+=+hVi-dA2befgW0OCz+v0y3Q@mail.gmail.com> (raw)
In-Reply-To: <CAM0EoM=1DyZsgYnuTjXB88L=41g00pjat+Jq4jThpciXzcEKJQ@mail.gmail.com>
[-- Attachment #1: Type: text/plain, Size: 3419 bytes --]
On Wed, Mar 20, 2024 at 3:34 PM Jamal Hadi Salim <jhs@mojatatu.com> wrote:
>
> On Wed, Mar 20, 2024 at 2:26 PM Eric Dumazet <edumazet@google.com> wrote:
> >
> > On Wed, Mar 20, 2024 at 7:13 PM Eric Dumazet <edumazet@google.com> wrote:
> > >
> > > On Wed, Mar 20, 2024 at 6:50 PM Jamal Hadi Salim <jhs@mojatatu.com> wrote:
> >
> > > Nope, you just have to complete the patch, moving around
> > > dev_xmit_recursion_inc() and dev_xmit_recursion_dec()
> >
> > Untested part would be:
> >
> > diff --git a/net/core/dev.c b/net/core/dev.c
> > index 303a6ff46e4e16296e94ed6b726621abe093e567..dbeaf67282e8b6ec164d00d796c9fd8e4fd7c332
> > 100644
> > --- a/net/core/dev.c
> > +++ b/net/core/dev.c
> > @@ -4259,6 +4259,8 @@ int __dev_queue_xmit(struct sk_buff *skb, struct
> > net_device *sb_dev)
> > */
> > rcu_read_lock_bh();
> >
> > + dev_xmit_recursion_inc();
> > +
> > skb_update_prio(skb);
> >
> > qdisc_pkt_len_init(skb);
> > @@ -4331,9 +4333,7 @@ int __dev_queue_xmit(struct sk_buff *skb, struct
> > net_device *sb_dev)
> > HARD_TX_LOCK(dev, txq, cpu);
> >
> > if (!netif_xmit_stopped(txq)) {
> > - dev_xmit_recursion_inc();
> > skb = dev_hard_start_xmit(skb, dev, txq, &rc);
> > - dev_xmit_recursion_dec();
> > if (dev_xmit_complete(rc)) {
> > HARD_TX_UNLOCK(dev, txq);
> > goto out;
> > @@ -4353,12 +4353,14 @@ int __dev_queue_xmit(struct sk_buff *skb,
> > struct net_device *sb_dev)
> > }
> >
> > rc = -ENETDOWN;
> > + dev_xmit_recursion_dec();
> > rcu_read_unlock_bh();
> >
> > dev_core_stats_tx_dropped_inc(dev);
> > kfree_skb_list(skb);
> > return rc;
> > out:
> > + dev_xmit_recursion_dec();
> > rcu_read_unlock_bh();
> > return rc;
> > }
>
> This removed the deadlock but now every packet being redirected will
> be dropped. I was wrong earlier on the tc block because that only
> works on clsact and ingress which are fine not needing this lock.
> Here's a variation of the earlier patch that may work but comes at the
> cost of a new per-cpu increment on the qdisc.
>
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -3789,6 +3789,11 @@ static inline int __dev_xmit_skb(struct sk_buff
> *skb, struct Qdisc *q,
> if (unlikely(contended))
> spin_lock(&q->busylock);
>
> + if (__this_cpu_read(q->recursion_xmit) > 0) {
> + //drop
> + }
> +
> + __this_cpu_inc(q->recursion_xmit);
> spin_lock(root_lock);
> if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
> __qdisc_drop(skb, &to_free);
> @@ -3825,6 +3830,7 @@ static inline int __dev_xmit_skb(struct sk_buff
> *skb, struct Qdisc *q,
> }
> }
> spin_unlock(root_lock);
> + __this_cpu_dec(q->recursion_xmit);
> if (unlikely(to_free))
> kfree_skb_list_reason(to_free,
> tcf_get_drop_reason(to_free));
>
And here's a tested version(attached) that fixes both the A->A and A->B->A.
cheers,
jamal
[-- Attachment #2: qdisc-rec-xmit-var-fix.patch --]
[-- Type: text/x-patch, Size: 2619 bytes --]
diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index cefe0c4bdae3..f9f99df037ed 100644
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -125,6 +125,8 @@ struct Qdisc {
spinlock_t busylock ____cacheline_aligned_in_smp;
spinlock_t seqlock;
+ u16 __percpu *xmit_recursion;
+
struct rcu_head rcu;
netdevice_tracker dev_tracker;
/* private data */
diff --git a/net/core/dev.c b/net/core/dev.c
index 9a67003e49db..2b712388c06f 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3789,6 +3789,13 @@ static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
if (unlikely(contended))
spin_lock(&q->busylock);
+ if (__this_cpu_read(*q->xmit_recursion) > 0) {
+ __qdisc_drop(skb, &to_free);
+ rc = NET_XMIT_DROP;
+ goto free_skb_list;
+ }
+
+ __this_cpu_inc(*q->xmit_recursion);
spin_lock(root_lock);
if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
__qdisc_drop(skb, &to_free);
@@ -3825,7 +3832,9 @@ static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
}
}
spin_unlock(root_lock);
+ __this_cpu_dec(*q->xmit_recursion);
if (unlikely(to_free))
+free_skb_list:
kfree_skb_list_reason(to_free,
tcf_get_drop_reason(to_free));
if (unlikely(contended))
diff --git a/net/sched/sch_api.c b/net/sched/sch_api.c
index 65e05b0c98e4..6c3bc1aff89a 100644
--- a/net/sched/sch_api.c
+++ b/net/sched/sch_api.c
@@ -1260,6 +1260,7 @@ static struct Qdisc *qdisc_create(struct net_device *dev,
struct Qdisc *sch;
struct Qdisc_ops *ops;
struct qdisc_size_table *stab;
+ int cpu;
ops = qdisc_lookup_ops(kind);
#ifdef CONFIG_MODULES
@@ -1376,11 +1377,22 @@ static struct Qdisc *qdisc_create(struct net_device *dev,
}
}
+ sch->xmit_recursion = alloc_percpu(u16);
+ if (!sch->xmit_recursion) {
+ err = -ENOMEM;
+ goto err_out5;
+ }
+ for_each_possible_cpu(cpu)
+ (*per_cpu_ptr(sch->xmit_recursion, cpu)) = 0;
+
qdisc_hash_add(sch, false);
trace_qdisc_create(ops, dev, parent);
return sch;
+err_out5:
+ if (tca[TCA_RATE])
+ gen_kill_estimator(&sch->rate_est);
err_out4:
/* Even if ops->init() failed, we call ops->destroy()
* like qdisc_create_dflt().
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index ff5336493777..afbbd2e885a4 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -1070,6 +1070,8 @@ static void __qdisc_destroy(struct Qdisc *qdisc)
module_put(ops->owner);
netdev_put(dev, &qdisc->dev_tracker);
+ free_percpu(qdisc->xmit_recursion);
+
trace_qdisc_destroy(qdisc);
call_rcu(&qdisc->rcu, qdisc_free_cb);
next prev parent reply other threads:[~2024-03-24 15:27 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-03-14 11:17 [PATCH] net/sched: Forbid assigning mirred action to a filter attached to the egress renmingshuai
2024-03-14 11:43 ` Jiri Pirko
2024-03-14 14:04 ` renmingshuai
2024-03-14 14:47 ` Pedro Tammela
2024-03-14 17:14 ` Jamal Hadi Salim
2024-03-15 1:56 ` renmingshuai
2024-03-15 22:34 ` Jamal Hadi Salim
2024-03-17 16:10 ` Jamal Hadi Salim
2024-03-18 14:27 ` Jamal Hadi Salim
2024-03-18 15:46 ` Eric Dumazet
2024-03-18 17:36 ` Jamal Hadi Salim
2024-03-18 19:11 ` Eric Dumazet
2024-03-18 22:05 ` Jamal Hadi Salim
2024-03-19 9:38 ` Eric Dumazet
2024-03-19 20:54 ` Jamal Hadi Salim
2024-03-20 16:46 ` Jamal Hadi Salim
2024-03-20 16:57 ` Eric Dumazet
2024-03-20 17:31 ` Jamal Hadi Salim
2024-03-20 17:50 ` Jamal Hadi Salim
2024-03-20 18:13 ` Eric Dumazet
2024-03-20 18:25 ` Eric Dumazet
2024-03-20 19:34 ` Jamal Hadi Salim
2024-03-24 15:27 ` Jamal Hadi Salim [this message]
2024-03-26 23:18 ` Jamal Hadi Salim
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAM0EoM=28nsomF9PniYNtpjD4+=+hVi-dA2befgW0OCz+v0y3Q@mail.gmail.com' \
--to=jhs@mojatatu.com \
--cc=caowangbao@huawei.com \
--cc=davem@davemloft.net \
--cc=dcaratti@redhat.com \
--cc=edumazet@google.com \
--cc=eric.dumazet@gmail.com \
--cc=jiri@resnulli.us \
--cc=kuba@kernel.org \
--cc=liaichun@huawei.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=pctammela@mojatatu.com \
--cc=renmingshuai@huawei.com \
--cc=victor@mojatatu.com \
--cc=vladbu@nvidia.com \
--cc=xiyou.wangcong@gmail.com \
--cc=yanan@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).