| To: | Stephen Hemminger <shemminger@xxxxxxxx> |
|---|---|
| Subject: | Re: [RFC] Use RCU for fib_rules |
| From: | "David S. Miller" <davem@xxxxxxxxxx> |
| Date: | Wed, 31 Mar 2004 13:11:07 -0800 |
| Cc: | niv@xxxxxxxxxx, Paul.McKenney@xxxxxxxxxx, netdev@xxxxxxxxxxx |
| In-reply-to: | <20040331130541.4b1e8b0d@dell_ss3.pdx.osdl.net> |
| References: | <20040331112839.02e8d4c2@dell_ss3.pdx.osdl.net> <406B2828.7070800@us.ibm.com> <20040331130541.4b1e8b0d@dell_ss3.pdx.osdl.net> |
| Sender: | netdev-bounce@xxxxxxxxxxx |
On Wed, 31 Mar 2004 13:05:41 -0800 Stephen Hemminger <shemminger@xxxxxxxx> wrote: > No, don't have enough rules or load in this area to cause an impact. > But also, I don't have the test environment to really stress this > (ie 1000's of rules). You really don't need 1000's of rules. Every time the routing cache misses, it's going to dive into this code, so on SMP that would show up. You could use stream.c (http://www.securiteam.com/unixfocus/5YP0I000DG.html) as the DoS workload in order to thrash the routing cache and force the rule codepath. |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: [RFC] Use RCU for fib_rules, Stephen Hemminger |
|---|---|
| Next by Date: | [PATCH]Add IPv6 MIBs counters in MLD (mcast.c), Shirley Ma |
| Previous by Thread: | Re: [RFC] Use RCU for fib_rules, Stephen Hemminger |
| Next by Thread: | Re: [RFC] Use RCU for fib_rules, Paul McKenney |
| Indexes: | [Date] [Thread] [Top] [All Lists] |