Re: [PATCH 01/15] sched/isolation: Support dynamic allocation for housekeeping masks

Re: [PATCH 01/15] sched/isolation: Support dynamic allocation for housekeeping masks

[Qiliang Yuan]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=yes, Size: 882 bytes --]

Hi Peter,

On Wed, Mar 25, 2026 at 02:57:07PM +0100, Peter Zijlstra wrote:
> I think I asked this a while ago; why do we have more than one mask? 
> What is the actual purpose of being able to separate RCU from Timers?

That's a fair point. For the vast majority of use cases (like NOHZ_FULL), 
these masks are indeed identical and should be updated as a single unit. 

The original motivation for separation was to allow extreme fine-tuning in 
HFT environments—for example, offloading RCU callbacks to keep a core 
mostly clean but allowing pinned timers for specific localized 
telemetry/monitoring. 

However, I acknowledge this adds significant complexity. In V13, I will 
unify these into a single "Global Housekeeping Mask" by default to 
simplify the configuration space, while keeping the underlying notifier 
infrastructure flexible enough for future specialized needs.

Re: [PATCH 01/15] sched/isolation: Support dynamic allocation for housekeeping masks

[Frederic Weisbecker]

Le Mon, Mar 30, 2026 at 07:43:21PM +0800, Qiliang Yuan a écrit :
11;rgb:2e2e/3434/3636> Hi Peter,
> 
> On Wed, Mar 25, 2026 at 02:57:07PM +0100, Peter Zijlstra wrote:
> > I think I asked this a while ago; why do we have more than one mask? 
> > What is the actual purpose of being able to separate RCU from Timers?
> 
> That's a fair point. For the vast majority of use cases (like NOHZ_FULL), 
> these masks are indeed identical and should be updated as a single unit. 
> 
> The original motivation for separation was to allow extreme fine-tuning in 
> HFT environments—for example, offloading RCU callbacks to keep a core 
> mostly clean but allowing pinned timers for specific localized 
> telemetry/monitoring. 
> 
> However, I acknowledge this adds significant complexity. In V13, I will 
> unify these into a single "Global Housekeeping Mask" by default to 
> simplify the configuration space, while keeping the underlying notifier 
> infrastructure flexible enough for future specialized needs.

More precisely we only need four flags:

HK_TYPE_DOMAIN_BOOT (isolcpus=)
HK_TYPE_DOMAIN (cgroup v2 isolated partition + isolcpus=)
HK_TYPE_KERNEL_NOISE (nohz_full=)
HK_TYPE_MANAGED_IRQ (isolcpus=managed_irq)

All the aliases of HK_TYPE_KERNEL_NOISE need to be renamed to
HK_TYPE_KERNEL_NOISE.

Thanks.

-- 
Frederic Weisbecker
SUSE Labs