Re: [RFC PATCH 2/2] softirq: Per vector thread deferment

2018-01-12 Thread Frederic Weisbecker
On Fri, Jan 12, 2018 at 10:07:25AM +0100, Paolo Abeni wrote: > On Fri, 2018-01-12 at 06:35 +0100, Frederic Weisbecker wrote: > > Some softirq vectors can be more CPU hungry than others. Especially > > networking may sometimes deal with packet storm and need more CPU than > > IRQ tail can offer with

Re: [RFC PATCH 2/2] softirq: Per vector thread deferment

2018-01-12 Thread Paolo Abeni
On Fri, 2018-01-12 at 06:35 +0100, Frederic Weisbecker wrote: > Some softirq vectors can be more CPU hungry than others. Especially > networking may sometimes deal with packet storm and need more CPU than > IRQ tail can offer without inducing scheduler latencies. In this case > the current code def

Re: [RFC PATCH 2/2] softirq: Per vector thread deferment

2018-01-11 Thread Frederic Weisbecker
On Fri, Jan 12, 2018 at 06:35:54AM +0100, Frederic Weisbecker wrote: > Some softirq vectors can be more CPU hungry than others. Especially > networking may sometimes deal with packet storm and need more CPU than > IRQ tail can offer without inducing scheduler latencies. In this case > the current c

[RFC PATCH 2/2] softirq: Per vector thread deferment

2018-01-11 Thread Frederic Weisbecker
Some softirq vectors can be more CPU hungry than others. Especially networking may sometimes deal with packet storm and need more CPU than IRQ tail can offer without inducing scheduler latencies. In this case the current code defers to ksoftirqd that behaves nicer. Now this nice behaviour can be ba