Skip to contentSkip to conversation

net/sched: cls_api: Fix lockup on flushing explicitly created chain

PATCH Applied Vlad Buslov 4 participants 5 messages Jun 12, 2023 – Jun 15, 2023

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net/sched: cls_api: Fix lockup on flushing explicitly created chain

Mingshuai Ren reports:

When a new chain is added by using tc, one soft lockup alarm will be
generated after delete the prio 0 filter of the chain. To reproduce the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:

Fix the issue by accounting for additional reference to chains that are explicitly created by RTM_NEWCHAIN message as opposed to implicitly by RTM_NEWTFILTER message.

Fixes 726d061286ce ("net: sched: prevent insertion of new classifiers during chain flush")
---
 net/sched/cls_api.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)
net/sched/cls_api.c +7 -5
index 2621550bfddc..e4df96e133cd 100644
@@ -659,8 +659,8 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
 {
 	struct tcf_block *block = chain->block;
 	const struct tcf_proto_ops *tmplt_ops;
+	unsigned int refcnt, non_act_refcnt;
 	bool free_block = false;
-	unsigned int refcnt;
 	void *tmplt_priv;
 
 	mutex_lock(&block->lock);
@@ -680,13 +680,15 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
 	 * save these to temporary variables.
 	 */
 	refcnt = --chain->refcnt;
+	non_act_refcnt = refcnt - chain->action_refcnt;
 	tmplt_ops = chain->tmplt_ops;
 	tmplt_priv = chain->tmplt_priv;
 
-	/* The last dropped non-action reference will trigger notification. */
-	if (refcnt - chain->action_refcnt == 0 && !by_act) {
-		tc_chain_notify_delete(tmplt_ops, tmplt_priv, chain->index,
-				       block, NULL, 0, 0, false);
+	if (non_act_refcnt == chain->explicitly_created && !by_act) {
+		if (non_act_refcnt == 0)
+			tc_chain_notify_delete(tmplt_ops, tmplt_priv,
+					       chain->index, block, NULL, 0, 0,
+					       false);
 		/* Last reference to chain, no need to lock. */
 		chain->flushing = false;
 	}
Signature · 1 line
2.39.2
#1 lore raw
Quoted 14 lines · Vlad Buslov

Mingshuai Ren reports:

When a new chain is added by using tc, one soft lockup alarm will be
generated after delete the prio 0 filter of the chain. To reproduce the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:

Fix the issue by accounting for additional reference to chains that are explicitly created by RTM_NEWCHAIN message as opposed to implicitly by RTM_NEWTFILTER message.

Quoted 7 lines
Fixes 726d061286ce ("net: sched: prevent insertion of new classifiers during chain flush")
---
  net/sched/cls_api.c | 12 +++++++-----
  1 file changed, 7 insertions(+), 5 deletions(-)

Hi Vlad,

Thanks for taking a look.
Could you also carry over the tdc test or ask Ren to post in a separate patch?

Quoted 35 lines
net/sched/cls_api.c +7 -5
index 2621550bfddc..e4df96e133cd 100644
@@ -659,8 +659,8 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
  {
  	struct tcf_block *block = chain->block;
  	const struct tcf_proto_ops *tmplt_ops;
+	unsigned int refcnt, non_act_refcnt;
  	bool free_block = false;
-	unsigned int refcnt;
  	void *tmplt_priv;
 
  	mutex_lock(&block->lock);
@@ -680,13 +680,15 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
  	 * save these to temporary variables.
  	 */
  	refcnt = --chain->refcnt;
+	non_act_refcnt = refcnt - chain->action_refcnt;
  	tmplt_ops = chain->tmplt_ops;
  	tmplt_priv = chain->tmplt_priv;
 
-	/* The last dropped non-action reference will trigger notification. */
-	if (refcnt - chain->action_refcnt == 0 && !by_act) {
-		tc_chain_notify_delete(tmplt_ops, tmplt_priv, chain->index,
-				       block, NULL, 0, 0, false);
+	if (non_act_refcnt == chain->explicitly_created && !by_act) {
+		if (non_act_refcnt == 0)
+			tc_chain_notify_delete(tmplt_ops, tmplt_priv,
+					       chain->index, block, NULL, 0, 0,
+					       false);
  		/* Last reference to chain, no need to lock. */
  		chain->flushing = false;
  	}
#2 lore raw
Quoted 22 lines · Pedro Tammela
Quoted 11 lines · Vlad Buslov

Mingshuai Ren reports:
When a new chain is added by using tc, one soft lockup alarm will be
generated after delete the prio 0 filter of the chain. To reproduce the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:
Fix the issue by accounting for additional reference to chains that are explicitly created by RTM_NEWCHAIN message as opposed to implicitly by RTM_NEWTFILTER message.

Quoted 8 lines
Fixes 726d061286ce ("net: sched: prevent insertion of new classifiers during

chain flush")

---
  net/sched/cls_api.c | 12 +++++++-----
  1 file changed, 7 insertions(+), 5 deletions(-)
Quoted 4 lines

Hi Vlad,

Thanks for taking a look.
Could you also carry over the tdc test or ask Ren to post in a separate patch?

Sure. I was planning to ask Mingshuai Ren to submit the new test as standalone patch after my fix has been accepted since including his code with my fix would require explicit approval of the whole patch and his Signed-off-by clause AFAIK.

Quoted 33 lines
Quoted 33 lines
net/sched/cls_api.c +7 -4
index 2621550bfddc..e4df96e133cd 100644
@@ -659,8 +659,8 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
  {
  	struct tcf_block *block = chain->block;
  	const struct tcf_proto_ops *tmplt_ops;
+	unsigned int refcnt, non_act_refcnt;
  	bool free_block = false;
-	unsigned int refcnt;
  	void *tmplt_priv;
    	mutex_lock(&block->lock);
@@ -680,13 +680,15 @@ static void __tcf_chain_put(struct tcf_chain *chain, bool by_act,
  	 * save these to temporary variables.
  	 */
  	refcnt = --chain->refcnt;
+	non_act_refcnt = refcnt - chain->action_refcnt;
  	tmplt_ops = chain->tmplt_ops;
  	tmplt_priv = chain->tmplt_priv;
  -	/* The last dropped non-action reference will trigger notification. */
-	if (refcnt - chain->action_refcnt == 0 && !by_act) {
-		tc_chain_notify_delete(tmplt_ops, tmplt_priv, chain->index,
-				       block, NULL, 0, 0, false);
+	if (non_act_refcnt == chain->explicitly_created && !by_act) {
+		if (non_act_refcnt == 0)
+			tc_chain_notify_delete(tmplt_ops, tmplt_priv,
+					       chain->index, block, NULL, 0, 0,
+					       false);
  		/* Last reference to chain, no need to lock. */
  		chain->flushing = false;
  	}
#3 lore raw
Quoted 25 lines · Pedro Tammela
Quoted 24 lines · Vlad Buslov
Quoted 11 lines

Mingshuai Ren reports:
When a new chain is added by using tc, one soft lockup alarm will be
generated after delete the prio 0 filter of the chain. To reproduce the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:
Fix the issue by accounting for additional reference to chains that are explicitly created by RTM_NEWCHAIN message as opposed to implicitly by RTM_NEWTFILTER message.

Quoted 8 lines
Fixes 726d061286ce ("net: sched: prevent insertion of new classifiers during

chain flush")

---
  net/sched/cls_api.c | 12 +++++++-----
  1 file changed, 7 insertions(+), 5 deletions(-)

Hi Vlad,

Quoted 7 lines
Quoted 2 lines

Thanks for taking a look.
Could you also carry over the tdc test or ask Ren to post in a separate patch?

Sure. I was planning to ask Mingshuai Ren to submit the new test as standalone patch after my fix has been accepted since including his code with my fix would require explicit approval of the whole patch and his Signed-off-by clause AFAIK.

OK. I will submit the new test as standalone patch after your fix is been accepted.

#4 lore raw
In reply to Vlad Buslov
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 11 lines · On Mon, 12 Jun 2023 11:34:26 +0200 you wrote:

Mingshuai Ren reports:

When a new chain is added by using tc, one soft lockup alarm will be
generated after delete the prio 0 filter of the chain. To reproduce the problem, perform the following steps:
(1) tc qdisc add dev eth0 root handle 1: htb default 1
(2) tc chain add dev eth0
(3) tc filter del dev eth0 chain 0 parent 1: prio 0
(4) tc filter add dev eth0 chain 0 parent 1:

[...]

Here is the summary with links:
- [net] net/sched: cls_api: Fix lockup on flushing explicitly created chain

    https://git.kernel.org/netdev/net/c/c9a82bec02c3

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html