[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <SL2PR06MB3082F70F1F4CE661288103C1BDE69@SL2PR06MB3082.apcprd06.prod.outlook.com>
Date: Thu, 7 Apr 2022 03:11:51 +0000
From: 王擎 <wangqing@...o.com>
To: Sudeep Holla <sudeep.holla@....com>
CC: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
"Rafael J. Wysocki" <rafael@...nel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
wangqing <11112896@...tel.com>
Subject: [PATCH] arch_topology: support parsing cache topology from DT
>> From: wangqing <11112896@...tel.com>
>>
>> When ACPI is not enabled, we can parse cache topolopy from DT:
>> * cpu0: cpu@000 {
>> * next-level-cache = <&L2_1>;
>> * L2_1: l2-cache {
>> * compatible = "cache";
>> * next-level-cache = <&L3_1>;
>> * };
>> * L3_1: l3-cache {
>> * compatible = "cache";
>> * };
>> * };
>> *
>> * cpu1: cpu@001 {
>> * next-level-cache = <&L2_1>;
>> * };
>> * cpu2: cpu@002 {
>> * L2_2: l2-cache {
>> * compatible = "cache";
>> * next-level-cache = <&L3_1>;
>> * };
>> * };
>> *
>> * cpu3: cpu@003 {
>> * next-level-cache = <&L2_2>;
>> * };
>> cache_topology hold the pointer describing "next-level-cache",
>> it can describe the cache topology of every level.
>>
>> Expand the use of llc_sibling when ACPI is not enabled.
>>
>
>You seem to have posted this patch as part of the series first. One patch
>was rejected and then you post this without any history. It confuses if you
>don't provide all the background/history.
Yes, the series contains several parts, the sched_domain part was rejected
temporary. But it has nothing to do with this patch, that's why I took it apart.
The background doesn't matter, let's focus on this patch itself.
>
>Having said that, NACK for this patch as it stands. We have
>drivers/base/cacheinfo.c which has all the parsing of cache information.
>IIRC we already consider LLC but highlight if anything is particularly
>missing. I am unable to follow/understand with you commit message.
cacheinfo.c just describes the properties of the cache, It can't describe
the cache topology, some like cpuinfo and cpu topology.
llc_sibling is not used at all if ACPI is not enabled, because llc_id
always be -1, and cpu_coregroup_mask() always return the core_sibling.
Why not get the cache topology from DT if the arch support GENERIC_ARCH_TOPOLOGY.
Thanks,
Wang
>
>--
>Regards,
>Sudeep
Powered by blists - more mailing lists