Message-ID: <Z7SWvr86RXlBbJlw@jlelli-thinkpadt14gen4.remote.csb>
Date: Tue, 18 Feb 2025 15:18:38 +0100
From: Juri Lelli <juri.lelli@...hat.com>
To: Dietmar Eggemann <dietmar.eggemann@....com>
Cc: Jon Hunter <jonathanh@...dia.com>,
	Christian Loehle <christian.loehle@....com>,
	Thierry Reding <treding@...dia.com>,
	Waiman Long <longman@...hat.com>, Tejun Heo <tj@...nel.org>,
	Johannes Weiner <hannes@...xchg.org>,
	Michal Koutny <mkoutny@...e.com>, Ingo Molnar <mingo@...hat.com>,
	Peter Zijlstra <peterz@...radead.org>,
	Vincent Guittot <vincent.guittot@...aro.org>,
	Steven Rostedt <rostedt@...dmis.org>,
	Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
	Valentin Schneider <vschneid@...hat.com>,
	Phil Auld <pauld@...hat.com>, Qais Yousef <qyousef@...alina.io>,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	"Joel Fernandes (Google)" <joel@...lfernandes.org>,
	Suleiman Souhlal <suleiman@...gle.com>,
	Aashish Sharma <shraash@...gle.com>,
	Shin Kawamura <kawasin@...gle.com>,
	Vineeth Remanan Pillai <vineeth@...byteword.org>,
	linux-kernel@...r.kernel.org, cgroups@...r.kernel.org,
	"linux-tegra@...r.kernel.org" <linux-tegra@...r.kernel.org>
Subject: Re: [PATCH v2 3/2] sched/deadline: Check bandwidth overflow earlier
 for hotplug

On 18/02/25 15:12, Dietmar Eggemann wrote:
> On 18/02/2025 10:58, Juri Lelli wrote:
> > Hi!
> > 
> > On 17/02/25 17:08, Juri Lelli wrote:
> >> On 14/02/25 10:05, Jon Hunter wrote:
> > 
> > ...
> > 
> >> At this point I believe you triggered suspend.
> >>
> >>> [   57.290150] Freezing remaining freezable tasks completed (elapsed 0.001 seconds)
> >>> [   57.335619] tegra-xusb 3530000.usb: Firmware timestamp: 2020-07-06 13:39:28 UTC
> >>> [   57.353364] dwc-eth-dwmac 2490000.ethernet eth0: Link is Down
> >>> [   57.397022] Disabling non-boot CPUs ...
> >>
> >> Offlining CPU5.
> >>
> >>> [   57.400904] dl_bw_manage: cpu=5 cap=3072 fair_server_bw=52428 total_bw=209712 dl_bw_cpus=4 type=DYN span=0,3-5
> >>> [   57.400949] CPU0 attaching NULL sched-domain.
> >>> [   57.415298] span=1-2
> >>> [   57.417483] __dl_sub: cpus=3 tsk_bw=52428 total_bw=157284 span=0,3-5 type=DYN
> >>> [   57.417487] __dl_server_detach_root: cpu=0 rd_span=0,3-5 total_bw=157284
> >>> [   57.417496] rq_attach_root: cpu=0 old_span=NULL new_span=1-2
> >>> [   57.417501] __dl_add: cpus=3 tsk_bw=52428 total_bw=157284 span=0-2 type=DEF
> >>> [   57.417504] __dl_server_attach_root: cpu=0 rd_span=0-2 total_bw=157284
> >>> [   57.417507] CPU3 attaching NULL sched-domain.
> >>> [   57.454804] span=0-2
> >>> [   57.456987] __dl_sub: cpus=2 tsk_bw=52428 total_bw=104856 span=3-5 type=DYN
> >>> [   57.456990] __dl_server_detach_root: cpu=3 rd_span=3-5 total_bw=104856
> >>> [   57.456998] rq_attach_root: cpu=3 old_span=NULL new_span=0-2
> >>> [   57.457000] __dl_add: cpus=4 tsk_bw=52428 total_bw=209712 span=0-3 type=DEF
> >>> [   57.457003] __dl_server_attach_root: cpu=3 rd_span=0-3 total_bw=209712
> >>> [   57.457006] CPU4 attaching NULL sched-domain.
> >>> [   57.493964] span=0-3
> >>> [   57.496152] __dl_sub: cpus=1 tsk_bw=52428 total_bw=52428 span=4-5 type=DYN
> >>> [   57.496156] __dl_server_detach_root: cpu=4 rd_span=4-5 total_bw=52428
> >>> [   57.496162] rq_attach_root: cpu=4 old_span=NULL new_span=0-3
> >>> [   57.496165] __dl_add: cpus=5 tsk_bw=52428 total_bw=262140 span=0-4 type=DEF
> >>> [   57.496168] __dl_server_attach_root: cpu=4 rd_span=0-4 total_bw=262140
> >>> [   57.496171] CPU5 attaching NULL sched-domain.
> >>> [   57.532952] span=0-4
> >>> [   57.535143] rq_attach_root: cpu=5 old_span= new_span=0-4
> >>> [   57.535147] __dl_add: cpus=5 tsk_bw=52428 total_bw=314568 span=0-5 type=DEF
> >>
> >> Maybe we shouldn't add the dl_server contribution of a CPU that is going
> >> to be offline.
> > 
> > I tried to implement this idea and ended up with the following. As usual
> > also pushed it to the branch on github. Could you please update and
> > re-test?
> > 
> > Another thing that I noticed is that in my case an hotplug operation
> > generating a sched/root domain rebuild ends up calling dl_rebuild_
> > rd_accounting() (from partition_and_rebuild_sched_domains()) which
> > resets accounting for def and dyn domains. In your case (looking again
> > at the last dmesg you shared) I don't see this call, so I wonder if for
> > some reason related to your setup we do the rebuild by calling partition_
> > sched_domains() (instead of partition_and_rebuild_) and this doesn't
> > call dl_rebuild_rd_accounting() after partition_sched_domains_locked() -
> > maybe it should? Dietmar, Christian, Peter, what do you think?
> 
> Yeah, looks like suspend/resume behaves differently compared to CPU hotplug.
> 
> On my Juno [L b b L L L]
>                 ^^^
>                 isolcpus=[2,3]
> 
> # ps2 | grep DLN
>    98    98 S 140      0   - DLN sugov:0
>    99    99 S 140      0   - DLN sugov:1
> 
> # taskset -p 98; taskset -p 99
> pid 98's current affinity mask: 39
> pid 99's current affinity mask: 6
> 
> 
> [   87.679282] partition_sched_domains() called
> ...
> [   87.684013] partition_sched_domains() called
> ...
> [   87.687961] partition_sched_domains() called
> ...
> [   87.689419] psci: CPU3 killed (polled 0 ms)
> [   87.689715] __dl_bw_capacity() mask=2-5 cap=1024
> [   87.689739] dl_bw_cpus() cpu=6 rd->span=2-5 cpu_active_mask=0-2 cpus=1
> [   87.689757] dl_bw_manage: cpu=2 cap=0 fair_server_bw=52428
> total_bw=209712 dl_bw_cpus=1 type=DEF span=2-5
> [   87.689775] dl_bw_cpus() cpu=6 rd->span=2-5 cpu_active_mask=0-2 cpus=1
> [   87.689789] dl_bw_manage() cpu=2 cap=0 overflow=1 return=-16
> [   87.689864] Error taking CPU2 down: -16                       <-- !!!
> ...
> [   87.690674] partition_sched_domains() called
> ...
> [   87.691496] partition_sched_domains() called
> ...
> [   87.693702] partition_sched_domains() called
> ...
> [   87.695819] partition_and_rebuild_sched_domains() called
> 

Ah, OK. Did you try with my last proposed change?

Powered by blists - more mailing lists