[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CALPaoCh0dzx2T8-u5ZQXXM0XqaZgJbUAGnRVBmbzHmQiHjvWTA@mail.gmail.com>
Date: Fri, 29 Sep 2023 16:21:54 +0200
From: Peter Newman <peternewman@...gle.com>
To: Tony Luck <tony.luck@...el.com>
Cc: Fenghua Yu <fenghua.yu@...el.com>,
Reinette Chatre <reinette.chatre@...el.com>,
Jonathan Corbet <corbet@....net>,
Shuah Khan <skhan@...uxfoundation.org>, x86@...nel.org,
Shaopeng Tan <tan.shaopeng@...itsu.com>,
James Morse <james.morse@....com>,
Jamie Iles <quic_jiles@...cinc.com>,
Babu Moger <babu.moger@....com>,
Randy Dunlap <rdunlap@...radead.org>,
linux-kernel@...r.kernel.org, linux-doc@...r.kernel.org,
patches@...ts.linux.dev
Subject: Re: [PATCH v6 6/8] x86/resctrl: Introduce snc_nodes_per_l3_cache
Hi Tony,
On Thu, Sep 28, 2023 at 9:14 PM Tony Luck <tony.luck@...el.com> wrote:
>
> Intel Sub-NUMA Cluster (SNC) is a feature that subdivides the CPU cores
> and memory controllers on a socket into two or more groups. These are
> presented to the operating system as NUMA nodes.
>
> This may enable some workloads to have slightly lower latency to memory
> as the memory controller(s) in an SNC node are electrically closer to the
> CPU cores on that SNC node. This cost may be offset by lower bandwidth
> since the memory accesses for each core can only be interleaved between
> the memory controllers on the same SNC node.
>
> Resctrl monitoring on Intel system depends upon attaching RMIDs to tasks
> to track L3 cache occupancy and memory bandwidth. There is an MSR that
> controls how the RMIDs are shared between SNC nodes.
>
> The default mode divides them numerically. E.g. when there are two SNC
> nodes on a socket the lower number half of the RMIDs are given to the
> first node, the remainder to the second node. This would be difficult
> to use with the Linux resctrl interface as specific RMID values assigned
> to resctrl groups are not visible to users.
>
> The other mode divides the RMIDs and renumbers the ones on the second
> SNC node to start from zero.
>
> Even with this redumbering SNC mode requires several changes in resctrl
> behavior for correct operation.
redumbering? Harsh.
> diff --git a/arch/x86/kernel/cpu/resctrl/rdtgroup.c b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> index b0901fb95aa9..a5404c412f53 100644
> --- a/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> +++ b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> @@ -1357,7 +1357,7 @@ unsigned int rdtgroup_cbm_to_size(struct rdt_resource *r,
> }
> }
>
> - return size;
> + return size / snc_nodes_per_l3_cache;
To confirm, the size represented by a bit goes down rather than the
CBM mask shrinking in each sub-NUMA domain?
I would maybe have expected the CBM mask to already be allocating at
the smallest granularity the hardware supports.
> }
>
> /**
> @@ -2590,7 +2590,7 @@ static int rdt_parse_param(struct fs_context *fc, struct fs_parameter *param)
> ctx->enable_cdpl2 = true;
> return 0;
> case Opt_mba_mbps:
> - if (!supports_mba_mbps())
> + if (!supports_mba_mbps() || snc_nodes_per_l3_cache > 1)
Factor into supports_mba_mbps()?
> return -EINVAL;
> ctx->enable_mba_mbps = true;
> return 0;
> --
> 2.41.0
>
Reviewed-by: Peter Newman <peternewman@...gle.com>
Powered by blists - more mailing lists