Skip to contentSkip to conversation

Fixes slab corruption in cbq_destroy

PATCH In review Thomas Graf 3 participants 6 messages Sep 16, 2004

Subsystems (3)
Networking Network drivers Traffic control (tc)
lore
Share & tools
lore

Thomas Graf

author
#0 lore raw

PATCH 1/1 Fixes slab corruption in cbq_destroy

Fixes slab corruption in cbq_destroy. cbq_destroy_filters and qdisc_put_rtab(q->link.R_tab) are already called in cbq_destroy_class. The latter lead to a slab corruption due to repeated freeing of q->link.R_tab because q->link is part of q->classes. Problem introduced in 1.21.

Signed-off-by Thomas Graf <[email protected]>
net/sched/sch_cbq.c +0 -6
@@ -1770,10 +1770,6 @@
 #ifdef CONFIG_NET_CLS_POLICE
 	q->rx_class = NULL;
 #endif
-	for (h = 0; h < 16; h++) {
-		for (cl = q->classes[h]; cl; cl = cl->next)
-			cbq_destroy_filters(cl);
-	}
 
 	for (h = 0; h < 16; h++) {
 		struct cbq_class *next;

@@ -1783,8 +1779,6 @@
 			cbq_destroy_class(sch, cl);
 		}
 	}
-
-	qdisc_put_rtab(q->link.R_tab);
 }
 
 static void cbq_put(struct Qdisc *sch, unsigned long arg)
#1 lore raw
Quoted 7 lines · Thomas Graf

Fixes slab corruption in cbq_destroy. cbq_destroy_filters and qdisc_put_rtab(q->link.R_tab) are already called in cbq_destroy_class. The latter lead to a slab corruption due to repeated freeing of q->link.R_tab because q->link is part of q->classes. Problem introduced in 1.21.

I don't see how there can be slab corruption. qdisc_put_rtab only calls kfree if the table is found in qdisc_rtab_list, which only happens once. But the patch is still fine as cleanup :)

Regards
Patrick

Quoted 28 lines
Signed-off-by Thomas Graf <[email protected]>
net/sched/sch_cbq.c +0 -6
@@ -1770,10 +1770,6 @@
 #ifdef CONFIG_NET_CLS_POLICE
 	q->rx_class = NULL;
 #endif
-	for (h = 0; h < 16; h++) {
-		for (cl = q->classes[h]; cl; cl = cl->next)
-			cbq_destroy_filters(cl);
-	}
 
 	for (h = 0; h < 16; h++) {
 		struct cbq_class *next;

@@ -1783,8 +1779,6 @@
 			cbq_destroy_class(sch, cl);
 		}
 	}
-
-	qdisc_put_rtab(q->link.R_tab);
 }
 
 static void cbq_put(struct Qdisc *sch, unsigned long arg)
author
#2 lore raw

* Patrick McHardy <[email protected]> 2004-09-16 15:47

Quoted 9 lines
Quoted 7 lines · Thomas Graf

Fixes slab corruption in cbq_destroy. cbq_destroy_filters and qdisc_put_rtab(q->link.R_tab) are already called in cbq_destroy_class. The latter lead to a slab corruption due to repeated freeing of q->link.R_tab because q->link is part of q->classes. Problem introduced in 1.21.

Quoted 3 lines

I don't see how there can be slab corruption. qdisc_put_rtab only calls kfree if the table is found in qdisc_rtab_list, which only happens once. But the patch is still fine as cleanup :)

On second call to qdisc_put_rtab with tab pointing to an already freed qdisc_rate_table:

sch_api.c:271: if (!tab || --tab->refcnt)

Sep 16 01:52:51 axs kernel: Slab corruption: start=d8bd4f30, len=2048
Sep 16 01:52:51 axs kernel: Redzone: 0x5a2cf071/0x5a2cf071.
Sep 16 01:52:51 axs kernel: Last user: [<c036c9af>](cbq_destroy_class+0x36/0x60)
Sep 16 01:52:51 axs kernel: 410: 6a 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b 6b
                                 ^^
	I guess this is the --tab->refcnt
Sep 16 01:52:51 axs kernel: Prev obj: start=d8bd4724, len=2048
Sep 16 01:52:51 axs kernel: Redzone: 0x170fc2a5/0x170fc2a5.
Sep 16 01:52:51 axs kernel: Last user: [<c0364594>](qdisc_get_rtab+0x77/0xcd)
Sep 16 01:52:51 axs kernel: 000: 04 00 00 00 00 00 40 00 20 bc be 00 05 00 00 00
Sep 16 01:52:51 axs kernel: 010: 05 00 00 00 05 00 00 00 05 00 00 00 05 00 00 00
Sep 16 01:52:51 axs kernel: Next obj: start=d8bd573c, len=2048
Sep 16 01:52:51 axs kernel: Redzone: 0x170fc2a5/0x170fc2a5.
Sep 16 01:52:51 axs kernel: Last user: [<c03484ce>](alloc_skb+0x48/0xe1)
Sep 16 01:52:51 axs kernel: 000: 6c 04 00 00 28 00 05 06 d5 d5 48 41 00 00 00 00
Sep 16 01:52:51 axs kernel: 010: 00 00 00 00 04 00 00 00 12 00 10 00 00 00 10 00
#3 lore raw
Quoted 9 lines · Thomas Graf

* Patrick McHardy <[email protected]> 2004-09-16 15:47

Quoted 5 lines

I don't see how there can be slab corruption. qdisc_put_rtab only calls kfree if the table is found in qdisc_rtab_list, which only happens once. But the patch is still fine as cleanup :)

Quoted 6 lines

On second call to qdisc_put_rtab with tab pointing to an already freed qdisc_rate_table:

sch_api.c:271: if (!tab || --tab->refcnt)

You're right, no double free but accessing and modifying of freed memory.

Regards
Patrick

author
#4 lore raw

* Patrick McHardy <[email protected]> 2004-09-16 16:58

Quoted 11 lines
Quoted 9 lines · Thomas Graf

* Patrick McHardy <[email protected]> 2004-09-16 15:47

Quoted 5 lines

I don't see how there can be slab corruption. qdisc_put_rtab only calls kfree if the table is found in qdisc_rtab_list, which only happens once. But the patch is still fine as cleanup :)

Quoted 7 lines
Quoted 6 lines

On second call to qdisc_put_rtab with tab pointing to an already freed qdisc_rate_table:

sch_api.c:271: if (!tab || --tab->refcnt)

You're right, no double free but accessing and modifying of freed memory.

My patch description was misleading should have been something like this:

Fixes slab corruption in cbq_destroy. cbq_destroy_filters and qdisc_put_rtab(q->link.R_tab) are already called in cbq_destroy_class. The latter lead to a slab corruption due to use of q->link.R_tab after being freed by previous call to qdisc_put_rtab. Problem introduced in 1.21.

maintainer
#5 lore raw
In reply to Thomas Graf
Quoted 5 lines · On Thu, 16 Sep 2004 15:28:56 +0200 Thomas Graf <[email protected]> wrote:

Fixes slab corruption in cbq_destroy. cbq_destroy_filters and qdisc_put_rtab(q->link.R_tab) are already called in cbq_destroy_class. The latter lead to a slab corruption due to repeated freeing of q->link.R_tab because q->link is part of q->classes. Problem introduced in 1.21.

Good catch, applied.