| To: | netdev@xxxxxxxxxxx |
|---|---|
| Subject: | netstat not tracking all packets |
| From: | Chris Friesen <cfriesen@xxxxxxxxxxxxxxxxxx> |
| Date: | Thu, 15 Jul 2004 11:17:59 -0400 |
| Sender: | netdev-bounce@xxxxxxxxxxx |
| User-agent: | Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040113 |
We just ran into an issue where incoming IP packets were being dropped due to socket buffer overflow, but there were no statistics showing it. Increasing the buffer size made the problems go away, but it was frustrating to not have accurate statistics.
After a bit of digging, I found the following path:
raw_v4_input
skb_clone
raw_rcv(sk, clone)
raw_rcv_skb
If the socket rx buffer is full raw_rcv_skb then drops the packet without
incrementing any counters. There's a FIXME comment in there that looks like its
been there for two years and counting.
Anyone have a fix for this, or know of other areas where the counters are not accurate? Chris |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: Fragment ID wrap workaround (read-only, untested)., David Stevens |
|---|---|
| Next by Date: | Re: [1/2] CARP implementation. HA master's failover., Evgeniy Polyakov |
| Previous by Thread: | [1/2] CARP implementation. HA master's failover., Evgeniy Polyakov |
| Next by Thread: | [PATCH Linux-2.6.8-rc1] pci.ids update for Prism GT devices, Margit Schubert-While |
| Indexes: | [Date] [Thread] [Top] [All Lists] |