* [PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum
@ 2007-11-16 9:31 Wang Chen
2007-11-16 10:59 ` Gerrit Renker
0 siblings, 1 reply; 2+ messages in thread
From: Wang Chen @ 2007-11-16 9:31 UTC (permalink / raw)
To: davem; +Cc: netdev, gerrit
Dave
If there is no socket filter, a bad checksum udp packet will be queued and
ready for user to read. At the same time the InDatagrams is increased.
So we need to decrement InDatagrams when recvmsg() or poll() notice the
bad checksum.
Seems Gerrit had fix it a year ago.(http://lkml.org/lkml/2006/8/28/218)
But it's not in kernel now. So I make the same patch for 2.6.24.rc2.
If this patch ok, I will make the same one for IPv6 UDP.
[PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum
Decrement UDP InDatagrams for bad checksum UDP packet in ready queue.
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
---
include/net/snmp.h | 2 ++
include/net/udp.h | 3 +++
net/ipv4/udp.c | 2 ++
3 files changed, 7 insertions(+)
diff -Nurp linux-2.6.24-rc2.org/include/net/snmp.h linux-2.6.24-rc2/include/net/snmp.h
--- linux-2.6.24-rc2.org/include/net/snmp.h 2007-11-09 16:37:08.000000000 +0800
+++ linux-2.6.24-rc2/include/net/snmp.h 2007-11-16 16:57:51.000000000 +0800
@@ -142,6 +142,8 @@ struct linux_mib {
(per_cpu_ptr(mib[!in_softirq()], raw_smp_processor_id())->mibs[field]++)
#define SNMP_DEC_STATS(mib, field) \
(per_cpu_ptr(mib[!in_softirq()], raw_smp_processor_id())->mibs[field]--)
+#define SNMP_DEC_STATS_BH(mib, field) \
+ (per_cpu_ptr(mib[0], raw_smp_processor_id())->mibs[field]--)
#define SNMP_ADD_STATS_BH(mib, field, addend) \
(per_cpu_ptr(mib[0], raw_smp_processor_id())->mibs[field] += addend)
#define SNMP_ADD_STATS_USER(mib, field, addend) \
diff -Nurp linux-2.6.24-rc2.org/include/net/udp.h linux-2.6.24-rc2/include/net/udp.h
--- linux-2.6.24-rc2.org/include/net/udp.h 2007-11-09 16:37:08.000000000 +0800
+++ linux-2.6.24-rc2/include/net/udp.h 2007-11-16 16:58:41.000000000 +0800
@@ -148,6 +148,9 @@ DECLARE_SNMP_STAT(struct udp_mib, udp_st
#define UDP_INC_STATS_BH(field, is_udplite) do { \
if (is_udplite) SNMP_INC_STATS_BH(udplite_statistics, field); \
else SNMP_INC_STATS_BH(udp_statistics, field); } while(0)
+#define UDP_DEC_STATS_BH(field, is_udplite) do { \
+ if (is_udplite) SNMP_DEC_STATS_BH(udplite_statistics, field); \
+ else SNMP_DEC_STATS_BH(udp_statistics, field); } while(0)
/* /proc */
struct udp_seq_afinfo {
diff -Nurp linux-2.6.24-rc2.org/net/ipv4/udp.c linux-2.6.24-rc2/net/ipv4/udp.c
--- linux-2.6.24-rc2.org/net/ipv4/udp.c 2007-11-09 16:37:57.000000000 +0800
+++ linux-2.6.24-rc2/net/ipv4/udp.c 2007-11-16 15:02:44.000000000 +0800
@@ -897,6 +897,7 @@ out:
csum_copy_err:
UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
+ UDP_DEC_STATS_BH(UDP_MIB_INDATAGRAMS, is_udplite);
skb_kill_datagram(sk, skb, flags);
@@ -1416,6 +1417,7 @@ unsigned int udp_poll(struct file *file,
while ((skb = skb_peek(rcvq)) != NULL &&
udp_lib_checksum_complete(skb)) {
UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
+ UDP_DEC_STATS_BH(UDP_MIB_INDATAGRAMS, is_lite);
__skb_unlink(skb, rcvq);
kfree_skb(skb);
}
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum
2007-11-16 9:31 [PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum Wang Chen
@ 2007-11-16 10:59 ` Gerrit Renker
0 siblings, 0 replies; 2+ messages in thread
From: Gerrit Renker @ 2007-11-16 10:59 UTC (permalink / raw)
To: Wang Chen; +Cc: davem, netdev
| If there is no socket filter, a bad checksum udp packet will be queued and
| ready for user to read. At the same time the InDatagrams is increased.
| So we need to decrement InDatagrams when recvmsg() or poll() notice the
| bad checksum.
|
| Seems Gerrit had fix it a year ago.(http://lkml.org/lkml/2006/8/28/218)
| But it's not in kernel now. So I make the same patch for 2.6.24.rc2.
|
The problem is known: packets with failed checksums will still increment
InDatagrams. The solution is also known (as acknowledged above). I
remember the discussion about this kind of solution: the problem is
apparently that these variables may sit on different CPUs, so that first
incrementing, and then later decrementing, would not have the desired
effect.
Desirable would be to only increment InDatagrams after passing all the
checksum tests (i.e. in udp_recvmsg). This solution was also discussed
on netdev, but people said that this would lead to trouble with the
data_ready handler (notably NFS).
| If this patch ok, I will make the same one for IPv6 UDP.
|
| [PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum
| Decrement UDP InDatagrams for bad checksum UDP packet in ready queue.
|
| Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
| ---
| include/net/snmp.h | 2 ++
| include/net/udp.h | 3 +++
| net/ipv4/udp.c | 2 ++
| 3 files changed, 7 insertions(+)
|
| diff -Nurp linux-2.6.24-rc2.org/include/net/snmp.h linux-2.6.24-rc2/include/net/snmp.h
| --- linux-2.6.24-rc2.org/include/net/snmp.h 2007-11-09 16:37:08.000000000 +0800
| +++ linux-2.6.24-rc2/include/net/snmp.h 2007-11-16 16:57:51.000000000 +0800
| @@ -142,6 +142,8 @@ struct linux_mib {
| (per_cpu_ptr(mib[!in_softirq()], raw_smp_processor_id())->mibs[field]++)
| #define SNMP_DEC_STATS(mib, field) \
| (per_cpu_ptr(mib[!in_softirq()], raw_smp_processor_id())->mibs[field]--)
| +#define SNMP_DEC_STATS_BH(mib, field) \
| + (per_cpu_ptr(mib[0], raw_smp_processor_id())->mibs[field]--)
| #define SNMP_ADD_STATS_BH(mib, field, addend) \
| (per_cpu_ptr(mib[0], raw_smp_processor_id())->mibs[field] += addend)
| #define SNMP_ADD_STATS_USER(mib, field, addend) \
| diff -Nurp linux-2.6.24-rc2.org/include/net/udp.h linux-2.6.24-rc2/include/net/udp.h
| --- linux-2.6.24-rc2.org/include/net/udp.h 2007-11-09 16:37:08.000000000 +0800
| +++ linux-2.6.24-rc2/include/net/udp.h 2007-11-16 16:58:41.000000000 +0800
| @@ -148,6 +148,9 @@ DECLARE_SNMP_STAT(struct udp_mib, udp_st
| #define UDP_INC_STATS_BH(field, is_udplite) do { \
| if (is_udplite) SNMP_INC_STATS_BH(udplite_statistics, field); \
| else SNMP_INC_STATS_BH(udp_statistics, field); } while(0)
| +#define UDP_DEC_STATS_BH(field, is_udplite) do { \
| + if (is_udplite) SNMP_DEC_STATS_BH(udplite_statistics, field); \
| + else SNMP_DEC_STATS_BH(udp_statistics, field); } while(0)
|
| /* /proc */
| struct udp_seq_afinfo {
| diff -Nurp linux-2.6.24-rc2.org/net/ipv4/udp.c linux-2.6.24-rc2/net/ipv4/udp.c
| --- linux-2.6.24-rc2.org/net/ipv4/udp.c 2007-11-09 16:37:57.000000000 +0800
| +++ linux-2.6.24-rc2/net/ipv4/udp.c 2007-11-16 15:02:44.000000000 +0800
| @@ -897,6 +897,7 @@ out:
|
| csum_copy_err:
| UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_udplite);
| + UDP_DEC_STATS_BH(UDP_MIB_INDATAGRAMS, is_udplite);
|
| skb_kill_datagram(sk, skb, flags);
|
| @@ -1416,6 +1417,7 @@ unsigned int udp_poll(struct file *file,
| while ((skb = skb_peek(rcvq)) != NULL &&
| udp_lib_checksum_complete(skb)) {
| UDP_INC_STATS_BH(UDP_MIB_INERRORS, is_lite);
| + UDP_DEC_STATS_BH(UDP_MIB_INDATAGRAMS, is_lite);
| __skb_unlink(skb, rcvq);
| kfree_skb(skb);
| }
|
|
--
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2007-11-16 10:56 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-11-16 9:31 [PATCH 1/2] [IPV4] SNMP: Decrement of UDP InDatagrams for bad checksum Wang Chen
2007-11-16 10:59 ` Gerrit Renker
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).