linux-kernel - Re: [PATCH v9 3/9] x86/resctrl: Modify update_mba_bw() to use per ctrl

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <156beac1-4596-4e39-bb04-d3f508bbb552@intel.com>
Date: Tue, 19 Nov 2024 19:39:27 -0800
From: Reinette Chatre <reinette.chatre@...el.com>
To: Tony Luck <tony.luck@...el.com>, Fenghua Yu <fenghua.yu@...el.com>, "Peter
 Newman" <peternewman@...gle.com>, Jonathan Corbet <corbet@....net>,
	<x86@...nel.org>
CC: James Morse <james.morse@....com>, Jamie Iles <quic_jiles@...cinc.com>,
	Babu Moger <babu.moger@....com>, Randy Dunlap <rdunlap@...radead.org>,
	"Shaopeng Tan (Fujitsu)" <tan.shaopeng@...itsu.com>,
	<linux-kernel@...r.kernel.org>, <linux-doc@...r.kernel.org>,
	<patches@...ts.linux.dev>
Subject: Re: [PATCH v9 3/9] x86/resctrl: Modify update_mba_bw() to use per
 ctrl_mon group event

Hi Tony,

On 11/13/24 4:17 PM, Tony Luck wrote:
> Instead of hard-coding the memory bandwidth local event as the
> input to the mba_sc feedback look, use the event that the user

"feedback look" -> "feedback loop"

> configured for each ctrl_mon group.

>From "Changelog" in Documentation/process/maintainer-tip.rst:
	"It's also useful to structure the changelog into several paragraphs and not     
	 lump everything together into a single one. A good structure is to explain      
	 the context, the problem and the solution in separate paragraphs and this       
	 order."

I do not find there to be a context nor problem description in the
changelog. Do you believe this changelog is appropriate for tip? Am I missing
something?

> 
> Signed-off-by: Tony Luck <tony.luck@...el.com>
> ---
>  arch/x86/kernel/cpu/resctrl/monitor.c | 23 ++++++++++++++++++-----
>  1 file changed, 18 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/resctrl/monitor.c b/arch/x86/kernel/cpu/resctrl/monitor.c
> index 7ef1a293cc13..2176e355e864 100644
> --- a/arch/x86/kernel/cpu/resctrl/monitor.c
> +++ b/arch/x86/kernel/cpu/resctrl/monitor.c
> @@ -752,20 +752,31 @@ static void update_mba_bw(struct rdtgroup *rgrp, struct rdt_mon_domain *dom_mbm)
>  	u32 closid, rmid, cur_msr_val, new_msr_val;
>  	struct mbm_state *pmbm_data, *cmbm_data;
>  	struct rdt_ctrl_domain *dom_mba;
> +	enum resctrl_event_id evt_id;
>  	struct rdt_resource *r_mba;
> -	u32 cur_bw, user_bw, idx;
>  	struct list_head *head;
>  	struct rdtgroup *entry;
> +	u32 cur_bw, user_bw;
>  
> -	if (!is_mbm_local_enabled())
> +	if (!is_mbm_enabled())

This change in the check is unexpected because at this point the event is still enforced to be
local MBM. This change is also undocumented so difficult to reason about.

>  		return;
>  
>  	r_mba = &rdt_resources_all[RDT_RESOURCE_MBA].r_resctrl;
> +	evt_id = rgrp->mba_mbps_event;

(To also answer the question in https://lore.kernel.org/all/Zzvtj8n1_ukhnRWT@agluck-desk3/ )

One key point from previous patch is that there is a new "contract" that rgrp->mba_mbps_event
is valid if mba_sc is enabled. If that contract is respected with appropriate initialization
and change of rgrp->mba_mbps_event then I do not believe that the three checks below
nor the "is_mbm_enabled()" above in reader of rgrp->mba_mbps_event are necessary since the
access of rgrp->mba_mbps_event is within "contract".
Note that caller does the checking if mba_sc is enabled:
	if (is_mba_sc(NULL))
		update_mba_bw(prgrp, d);

Thus doing same check within update_mba_bw() should not be necessary. 

It does take a lot of digging to understand so it would really be helpful to document these types
of design decisions and reinforce them through the series.

> +
> +	if (WARN_ON_ONCE(!is_mbm_event(evt_id)))
> +		return;
> +	if (WARN_ON_ONCE(evt_id == QOS_L3_MBM_LOCAL_EVENT_ID && !is_mbm_local_enabled()))
> +		return;
> +	if (WARN_ON_ONCE(evt_id == QOS_L3_MBM_TOTAL_EVENT_ID && !is_mbm_total_enabled()))
> +		return;
> +
>  
>  	closid = rgrp->closid;
>  	rmid = rgrp->mon.rmid;
> -	idx = resctrl_arch_rmid_idx_encode(closid, rmid);
> -	pmbm_data = &dom_mbm->mbm_local[idx];
> +	pmbm_data = get_mbm_state(dom_mbm, closid, rmid, evt_id);
> +	if (WARN_ON_ONCE(!pmbm_data))
> +		return;
>  
>  	dom_mba = get_ctrl_domain_from_cpu(smp_processor_id(), r_mba);
>  	if (!dom_mba) {
> @@ -784,7 +795,9 @@ static void update_mba_bw(struct rdtgroup *rgrp, struct rdt_mon_domain *dom_mbm)
>  	 */
>  	head = &rgrp->mon.crdtgrp_list;
>  	list_for_each_entry(entry, head, mon.crdtgrp_list) {
> -		cmbm_data = &dom_mbm->mbm_local[entry->mon.rmid];
> +		cmbm_data = get_mbm_state(dom_mbm, entry->closid, entry->mon.rmid, evt_id);
> +		if (WARN_ON_ONCE(!cmbm_data))
> +			return;
>  		cur_bw += cmbm_data->prev_bw;
>  	}
>  

Reinette