[patch 2.5] net: severe bug in icmp stats

Ivan Kokshaysky (ink@jurassic.park.msu.ru)
Tue, 1 Apr 2003 18:57:07 +0400


I believe many of those weird crash reports with recent 2.5
kernels can be explained by wrong pointer arithmetic in
ICMP_INC_STATS_xx_FIELD macros.

(*((long *)((void *)ptr) + offt))++

This actually adds "offt" _longs_, not _bytes_ to "ptr",
which causes increment of entirely unrelated kernel data.
Nasty thing is that a result of this corruption usually
shows up much later and depends on kernel layout.
I was "lucky" enough - on one of my boxes ICMP redirect
packets reproducibly caused increment of thread_info->task
pointer of kswapd, so finally I traced this back to icmp.h.

Ivan.

--- 2.5/include/net/icmp.h Tue Mar 25 01:01:22 2003
+++ linux/include/net/icmp.h Tue Apr 1 13:34:06 2003
@@ -39,15 +39,15 @@ DECLARE_SNMP_STAT(struct icmp_mib, icmp_
#define ICMP_INC_STATS_FIELD(offt) \
(*((unsigned long *) ((void *) \
per_cpu_ptr(icmp_statistics[!in_softirq()],\
- smp_processor_id())) + offt))++;
+ smp_processor_id()) + offt)))++
#define ICMP_INC_STATS_BH_FIELD(offt) \
(*((unsigned long *) ((void *) \
per_cpu_ptr(icmp_statistics[0], \
- smp_processor_id())) + offt))++;
+ smp_processor_id()) + offt)))++
#define ICMP_INC_STATS_USER_FIELD(offt) \
(*((unsigned long *) ((void *) \
per_cpu_ptr(icmp_statistics[1], \
- smp_processor_id())) + offt))++;
+ smp_processor_id()) + offt)))++

extern void icmp_send(struct sk_buff *skb_in, int type, int code, u32 info);
extern int icmp_rcv(struct sk_buff *skb);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/