linux-kernel - Re: [RFC PATCH v4 0/9] mm: Hot page tracking and promotion infrastructure

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: <b9aafed7-f250-4ce9-b8c4-ec9053f30348@amd.com>
Date: Tue, 13 Jan 2026 20:08:34 +0530
From: Bharata B Rao <bharata@....com>
To: Gregory Price <gourry@...rry.net>
CC: <linux-kernel@...r.kernel.org>, <linux-mm@...ck.org>,
	<Jonathan.Cameron@...wei.com>, <dave.hansen@...el.com>,
	<mgorman@...hsingularity.net>, <mingo@...hat.com>, <peterz@...radead.org>,
	<raghavendra.kt@....com>, <riel@...riel.com>, <rientjes@...gle.com>,
	<sj@...nel.org>, <weixugc@...gle.com>, <willy@...radead.org>,
	<ying.huang@...ux.alibaba.com>, <ziy@...dia.com>, <dave@...olabs.net>,
	<nifan.cxl@...il.com>, <xuezhengchu@...wei.com>, <yiannis@...corp.com>,
	<akpm@...ux-foundation.org>, <david@...hat.com>, <byungchul@...com>,
	<kinseyho@...gle.com>, <joshua.hahnjy@...il.com>, <yuanchu@...gle.com>,
	<balbirs@...dia.com>, <alok.rathore@...sung.com>, <shivankg@....com>
Subject: Re: [RFC PATCH v4 0/9] mm: Hot page tracking and promotion
 infrastructure

On 13-Jan-26 7:54 PM, Gregory Price wrote:
> On Sat, Dec 06, 2025 at 03:44:14PM +0530, Bharata B Rao wrote:
>> Hi,
>>
>> This is v4 of page hotness tracking and promotion sub-system pghot.
>>
> 
> Just fyi, we haven't started real workload testing, but we're setting
> up the process soon.  Just wanted to let you know it's on the radar.
> 
> We want to test this at the same time with some fixes we've been
> working on for multi-tenant fairness in the demotion/reclaim path.

Thanks Gregory for taking this up for testing.

I am working on v5 with following enhancements:

1. Default mode: 1 byte per PFN for storing hotness info. NID isn't stored but
promotion will be done to a default toptier node(specifiable via sysfs/debugs).
Frequency and time are stored but due to paucity of bits, I will have to use
some sort of bucketing scheme for storing time like how current hot page
promotion logic does.

2. Precision mode: 4 bytes or 32bits per PFN where NID is stored in addition to
time and frequency.

2 is completed and 1 is in the works. Once I get it working correctly, I will
post v5.

Since there were a few bugs in v4, I think it makes sense to start testing on v5
itself.

Regards,
Bharata.