Skip to contentSkip to conversation

: small read past the end of array in debug code

PATCH In review Dan Carpenter 3 participants 4 messages Apr 17, 2013 – Apr 22, 2013

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

PATCH 1/1 : small read past the end of array in debug code

The "reason" can come from skb->data[] and it hasn't been capped so it can be from 0-255 instead of just 0-6. For example in irlmp_state_dtr() the code does:

	reason = skb->data[3];
	...
	irlmp_disconnect_indication(self, reason, skb);

Also LMREASON has a couple other values which don't have entries in the irlmp_reasons[] array. And 0xff is a valid reason as well which means "unknown".

So far as I can see we don't actually care about "reason" except for in the debug code.

include/net/irda/irlmp.h +2 -1
index f741091..f132924 100644
@@ -256,7 +256,8 @@ static inline __u32 irlmp_get_daddr(const struct lsap_cb *self)
 	return (self && self->lap) ? self->lap->daddr : 0;
 }
 
-extern const char *irlmp_reasons[];
+const char *irlmp_reason_str(LM_REASON reason);
+
 extern int sysctl_discovery_timeout;
 extern int sysctl_discovery_slots;
 extern int sysctl_discovery;
net/irda/irlmp.c +9 -1
index 6115a44..1064621 100644
@@ -66,8 +66,15 @@ const char *irlmp_reasons[] = {
 	"LM_LAP_RESET",
 	"LM_INIT_DISCONNECT",
 	"ERROR, NOT USED",
+	"UNKNOWN",
 };
 
+const char *irlmp_reason_str(LM_REASON reason)
+{
+	reason = min_t(size_t, reason, ARRAY_SIZE(irlmp_reasons) - 1);
+	return irlmp_reasons[reason];
+}
+
 /*
  * Function irlmp_init (void)
  *
@@ -747,7 +754,8 @@ void irlmp_disconnect_indication(struct lsap_cb *self, LM_REASON reason,
 {
 	struct lsap_cb *lsap;
 
-	IRDA_DEBUG(1, "%s(), reason=%s\n", __func__, irlmp_reasons[reason]);
+	IRDA_DEBUG(1, "%s(), reason=%s [%d]\n", __func__,
+		   irlmp_reason_str(reason), reason);
 	IRDA_ASSERT(self != NULL, return;);
 	IRDA_ASSERT(self->magic = LMP_LSAP_MAGIC, return;);
 
net/irda/iriap.c +2 -1
index 29340a9..e1b37f5 100644
@@ -303,7 +303,8 @@ static void iriap_disconnect_indication(void *instance, void *sap,
 {
 	struct iriap_cb *self;
 
-	IRDA_DEBUG(4, "%s(), reason=%s\n", __func__, irlmp_reasons[reason]);
+	IRDA_DEBUG(4, "%s(), reason=%s [%d]\n", __func__,
+		   irlmp_reason_str(reason), reason);
 
 	self = instance;
 
maintainer
#1 lore raw

From: Dan Carpenter <[email protected]>
Date: Wed, 17 Apr 2013 10:10:38 +0300

Quoted 8 lines

The "reason" can come from skb->data[] and it hasn't been capped so it can be from 0-255 instead of just 0-6. For example in irlmp_state_dtr() the code does:

	reason = skb->data[3];
	...
	irlmp_disconnect_indication(self, reason, skb);
Quoted 8 lines

Also LMREASON has a couple other values which don't have entries in the irlmp_reasons[] array. And 0xff is a valid reason as well which means "unknown".

So far as I can see we don't actually care about "reason" except for in the debug code.

Applied but please put the subsystem prefix outside of [] which get stripped by git am and other tools. I fixed it up to be "irda: ".

Thanks.

#2 lore raw
Quoted 15 lines · David Miller

From: Dan Carpenter <[email protected]>
Date: Wed, 17 Apr 2013 10:10:38 +0300

Quoted 12 lines

The "reason" can come from skb->data[] and it hasn't been capped so it can be from 0-255 instead of just 0-6. For example in irlmp_state_dtr() the code does:

	reason = skb->data[3];
	...
	irlmp_disconnect_indication(self, reason, skb);

Also LMREASON has a couple other values which don't have entries in the irlmp_reasons[] array. And 0xff is a valid reason as well which means "unknown".

Quoted 8 lines
Quoted 4 lines

So far as I can see we don't actually care about "reason" except for in the debug code.

Applied but please put the subsystem prefix outside of [] which get stripped by git am and other tools. I fixed it up to be "irda: ".

Sorry about that. I won't do it again.

It's weird because the prefix was cut and pasted from an earlier commit. Also [media] and [SCSI] use square brackets in their subsystems. I tried some experiments with "git am" and you're right that "git am" always strips out the square bracket stuff. Odd.

regards,
dan carpenter

#3 lore raw
Quoted 25 lines · Dan Carpenter
Quoted 15 lines · David Miller

From: Dan Carpenter <[email protected]>
Date: Wed, 17 Apr 2013 10:10:38 +0300

Quoted 12 lines

The "reason" can come from skb->data[] and it hasn't been capped so it can be from 0-255 instead of just 0-6. For example in irlmp_state_dtr() the code does:

	reason = skb->data[3];
	...
	irlmp_disconnect_indication(self, reason, skb);

Also LMREASON has a couple other values which don't have entries in the irlmp_reasons[] array. And 0xff is a valid reason as well which means "unknown".

Quoted 8 lines
Quoted 4 lines

So far as I can see we don't actually care about "reason" except for in the debug code.

Applied but please put the subsystem prefix outside of [] which get stripped by git am and other tools. I fixed it up to be "irda: ".

Quoted 6 lines

Sorry about that. I won't do it again.

It's weird because the prefix was cut and pasted from an earlier commit. Also [media] and [SCSI] use square brackets in their subsystems. I tried some experiments with "git am" and you're right that "git am" always strips out the square bracket stuff. Odd.

The -k option suppresses this behaviour, but then it will also keep [PATCH]. It's a shame there's no option to strip just \[PATCH[^]]*\]

Ben.

Signature · 3 lines
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.