Skip to contentSkip to conversation

sch_htb: INIT_HLIST_NODE after hlist_del()

PATCH In review Jarek Poplawski 3 participants 6 messages Nov 6, 2006 – Nov 8, 2006

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

PATCH 1/1 sch_htb: INIT_HLIST_NODE after hlist_del()

After hlist_del() next and pprev pointers are not NULL so hlist_unhashed() doesn't work properly.

---

net/sched/sch_htb.c +6 -2
diff -Nurp linux-2.6.19-rc4-git10-/net/sched/sch_htb.c linux-2.6.19-rc4-git10/net/sched/sch_htb.c
@@ -1284,8 +1284,10 @@ static void htb_destroy_class(struct Qdi
 						  struct htb_class, sibling));
 
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}
 	list_del(&cl->sibling);
 
 	if (cl->prio_activity)
@@ -1333,8 +1335,10 @@ static int htb_delete(struct Qdisc *sch,
 	sch_tree_lock(sch);
 
 	/* delete from hash and active; remainder in destroy_class */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}
 
 	if (cl->prio_activity)
 		htb_deactivate(q, cl);
#1 lore raw
Quoted 12 lines · On Mon, 6 Nov 2006 12:33:53 +0100 Jarek Poplawski <[email protected]> wrote:

After hlist_del() next and pprev pointers are not NULL so hlist_unhashed() doesn't work properly.

---

diff -Nurp linux-2.6.19-rc4-git10-/net/sched/sch_htb.c linux-2.6.19-rc4-git10/net/sched/sch_htb.c

net/sched/sch_htb.c static void htb_destroy_class(struct Qdi · line 1290

Quoted 8 lines
net/sched/sch_htb.c +3 -1
@@ -1284,8 +1284,10 @@ static void htb_destroy_class(struct Qdi
 						  struct htb_class, sibling));
 
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}

why not use hlist_del_init?

Quoted 15 lines
net/sched/sch_htb.c +3 -1
 	list_del(&cl->sibling);
 
 	if (cl->prio_activity)
@@ -1333,8 +1335,10 @@ static int htb_delete(struct Qdisc *sch,
 	sch_tree_lock(sch);
 
 	/* delete from hash and active; remainder in destroy_class */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}
 
 	if (cl->prio_activity)
 		htb_deactivate(q, cl);
Signature · 1 line
Stephen Hemminger <[email protected]>
#2 lore raw
Quoted 17 lines · Stephen Hemminger
Quoted 14 lines · On Mon, 6 Nov 2006 12:33:53 +0100 Jarek Poplawski <[email protected]> wrote:

After hlist_del() next and pprev pointers are not NULL so hlist_unhashed() doesn't work properly.

---

diff -Nurp linux-2.6.19-rc4-git10-/net/sched/sch_htb.c linux-2.6.19-rc4-git10/net/sched/sch_htb.c

net/sched/sch_htb.c +0 -0
@@ -1284,8 +1284,10 @@ static void htb_destroy_class(struct Qdi
 						  struct htb_class, sibling));
 
Quoted 8 lines

net/sched/sch_htb.c static void htb_destroy_class(struct Qdi · line 1290

Quoted 6 lines
net/sched/sch_htb.c +3 -1
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}

why not use hlist_del_init?

Yes, this is the question!

As a matter of fact I expected another question. Yesterday I was short on time so I didn't describe the bug enough. I'm not sure if you know the problem, so here are more details (for me problem is 199% repeatable).

After something like this:

# tc qdisc add dev lo root handle 1: htb
# tc class add dev lo parent 1: classid 1:1 htb rate 200kbps
# tc class del dev lo classid 1:1

enter the BUG...

I've found the last command is the culprit and if you do:

# tc qdisc del dev lo root
there is no problem.

And probably it is enough to do the change only in htb_delete
- btw. is this hlist_del really needed there? and shouldn't all deletions be done after zeroing the refcount? - but you should know better.

Quoted 14 lines
Quoted 14 lines
net/sched/sch_htb.c +3 -1
 	list_del(&cl->sibling);
 
 	if (cl->prio_activity)
@@ -1333,8 +1335,10 @@ static int htb_delete(struct Qdisc *sch,
 	sch_tree_lock(sch);
 
 	/* delete from hash and active; remainder in destroy_class */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}
 
Quoted 2 lines

net/sched/sch_htb.c static int htb_delete(struct Qdisc *sch, · line 1344

Quoted 2 lines
net/sched/sch_htb.c +0 -0
 	if (cl->prio_activity)
 		htb_deactivate(q, cl);

Best regards,

Jarek P.

#3 lore raw
Quoted 18 lines · On Tue, 7 Nov 2006 07:49:43 +0100 Jarek Poplawski <[email protected]> wrote:
Quoted 17 lines · Stephen Hemminger
Quoted 14 lines · On Mon, 6 Nov 2006 12:33:53 +0100 Jarek Poplawski <[email protected]> wrote:

After hlist_del() next and pprev pointers are not NULL so hlist_unhashed() doesn't work properly.

---

diff -Nurp linux-2.6.19-rc4-git10-/net/sched/sch_htb.c linux-2.6.19-rc4-git10/net/sched/sch_htb.c

net/sched/sch_htb.c +0 -0
@@ -1284,8 +1284,10 @@ static void htb_destroy_class(struct Qdi
 						  struct htb_class, sibling));
 
Quoted 8 lines
Quoted 8 lines

net/sched/sch_htb.c static void htb_destroy_class(struct Qdi · line 1290

Quoted 6 lines
net/sched/sch_htb.c +3 -1
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
+	if (!hlist_unhashed(&cl->hlist)) {
 		hlist_del(&cl->hlist);
+		INIT_HLIST_NODE(&cl->hlist);
+	}

why not use hlist_del_init?

Your patch duplicated the code in hlist_del_init(). Why not do:

net/sched/sch_htb.c +2 -4
@@ -1284,8 +1284,7 @@
 						  struct htb_class, sibling));
 
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
-		hlist_del(&cl->hlist);
+	hlist_del_init(&cl->hlist);
 	list_del(&cl->sibling);
 
 	if (cl->prio_activity)
@@ -1333,8 +1332,7 @@
 	sch_tree_lock(sch);
 
 	/* delete from hash and active; remainder in destroy_class */
-	if (!hlist_unhashed(&cl->hlist))
-		hlist_del(&cl->hlist);
+	hlist_del_init(&cl->hlist);
 
 	if (cl->prio_activity)
 		htb_deactivate(q, cl);
maintainer
#4 lore raw

From: Stephen Hemminger <[email protected]>
Date: Tue, 7 Nov 2006 09:50:07 -0800

Quoted 1 line

Your patch duplicated the code in hlist_del_init(). Why not do:

Indeed, this is the patch I will apply.

Thanks Stephen.

#5 lore raw
In reply to Stephen Hemminger
Quoted 2 lines · Stephen Hemminger

On Tue, 7 Nov 2006 07:49:43 +0100
Jarek Poplawski <[email protected]> wrote:

...

Quoted 16 lines

Your patch duplicated the code in hlist_del_init(). Why not do:

net/sched/sch_htb.c +1 -2
@@ -1284,8 +1284,7 @@
 						  struct htb_class, sibling));
 
 	/* note: this delete may happen twice (see htb_delete) */
-	if (!hlist_unhashed(&cl->hlist))
-		hlist_del(&cl->hlist);
+	hlist_del_init(&cl->hlist);
 	list_del(&cl->sibling);
 
 	if (cl->prio_activity)

net/sched/sch_htb.c line 1338

Quoted 9 lines
net/sched/sch_htb.c +1 -2
@@ -1333,8 +1332,7 @@
 	sch_tree_lock(sch);
 
 	/* delete from hash and active; remainder in destroy_class */
-	if (!hlist_unhashed(&cl->hlist))
-		hlist_del(&cl->hlist);
+	hlist_del_init(&cl->hlist);
 
 	if (cl->prio_activity)
 		htb_deactivate(q, cl);

I've understood you first suggestion. But after sending my patch I've found it is also hiding a real problem of excessive deletion in one and possibly more places. So probably this should be done the right way and this hlist_unhashed testing left in BUG_ON only...

Cheers,
Jarek P.