| To: | Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> |
|---|---|
| Subject: | Re: [BUG] overflow in net/ipv4/route.c rt_check_expire() |
| From: | Robert Olsson <Robert.Olsson@xxxxxxxxxxx> |
| Date: | Mon, 4 Apr 2005 12:27:43 +0200 |
| Cc: | Robert Olsson <Robert.Olsson@xxxxxxxxxxx>, "David S. Miller" <davem@xxxxxxxxxxxxx>, dada1@xxxxxxxxxxxxx, netdev@xxxxxxxxxxx |
| In-reply-to: | <20050403214521.GB15901@xxxxxxxxxxxxxxxxxxx> |
| References: | <E1DHdsP-0003Lr-00@xxxxxxxxxxxxxxxxxxxxxxxx> <424E641A.1020609@xxxxxxxxxxxxx> <16974.41648.568927.54429@xxxxxxxxxxxx> <20050402193224.GA25157@xxxxxxxxxxxxxxxxxxx> <20050402115528.11f71a3c.davem@xxxxxxxxxxxxx> <20050403074337.GA8083@xxxxxxxxxxxxxxxxxxx> <16976.19092.562006.246545@xxxxxxxxxxxx> <20050403214521.GB15901@xxxxxxxxxxxxxxxxxxx> |
| Sender: | netdev-bounce@xxxxxxxxxxx |
Herbert Xu writes:
> The reason I'm suggesting the move to a kernel thread is because
> softirq context is not preemptible.
>
> So doing a large amount of work in it when your table is big means
> that a UP machine will freeze for a while.
The flush transient will happen also on UP... as I understand this
When we have changed the rt_hash_rnd and therefore invalidated all current
entries it would be best to blackhole *all* traffic until all old entries
are deleted this to avoid transients.
--ro
|
| Previous by Date: | Re: Linux support for RDMA (was: [Ksummit-2005-discuss] Summary of 2005 Kernel Summit Proposed Topics), David S. Miller |
|---|---|
| Next by Date: | Re: [BUG] overflow in net/ipv4/route.c rt_check_expire(), Robert Olsson |
| Previous by Thread: | Re: [BUG] overflow in net/ipv4/route.c rt_check_expire(), Herbert Xu |
| Next by Thread: | Re: [BUG] overflow in net/ipv4/route.c rt_check_expire(), Herbert Xu |
| Indexes: | [Date] [Thread] [Top] [All Lists] |