linux-kernel - min_filelist_kbytes vs file_is

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>] [day] [month] [year] [list]

Message-ID: <20201224111327.knhqdz3hloxfcksv@spock.localdomain>
Date:   Thu, 24 Dec 2020 12:13:27 +0100
From:   Oleksandr Natalenko <oleksandr@...alenko.name>
To:     linux-kernel@...r.kernel.org
Cc:     Mandeep Singh Baines <mandeep.baines@...il.com>,
        Guenter Roeck <groeck@...omium.org>,
        Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org
Subject: min_filelist_kbytes vs file_is_tiny

Hello, Mandeep, Guenter et al.

I came across the out-of-tree patch [1] that apparently is still alive
after 10 years of residing in the Chromium OS tree, and I have a couple
of questions if you don't mind spending your time answering them.

1. is this knob really necessary given there's an explicit bailout
mechanism relying on `file_is_tiny` which depends on the sum of high
watermarks across the zones? Wouldn't increasing `vm.min_free_kbytes`
achieve basically the same?

2. if `vm.min_free_kbytes` is not an option, would setting
`file_is_tiny` based on your `min_filelist_kbytes` knob achieve the
same?

3. if not, is `memory.min` cgroup2 knob supposed to work in a similar
manner? it looks to be unavailable for the root cgroup, though.

What I'm looking for, basically, is to achieve the effect of the
mentioned patch using mechanisms that are already available in the
upstream kernel.

Thank you.

[1] https://chromium.googlesource.com/chromiumos/third_party/kernel-next/+/545e2917dbd863760a51379de8c26631e667c563^!/

-- 
  Oleksandr Natalenko (post-factum)