linux-kernel - Re: [PATCH] arm64: add early fixmap initialization flag

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

Message-ID: <ZdcpJ2zoyHJuNkSD@FVFF77S0Q05N>
Date: Thu, 22 Feb 2024 10:59:51 +0000
From: Mark Rutland <mark.rutland@....com>
To: Itaru Kitayama <itaru.kitayama@...ux.dev>
Cc: skseofh@...il.com, catalin.marinas@....com, will@...nel.org,
	ryan.roberts@....com, linux-arm-kernel@...ts.infradead.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] arm64: add early fixmap initialization flag

On Wed, Feb 21, 2024 at 08:14:00AM +0900, Itaru Kitayama wrote:
> On Tue, Feb 20, 2024 at 11:55:30AM +0000, Mark Rutland wrote:
> > From 5f07f9c1d352f760fa7aba97f1b4f42d9cb99496 Mon Sep 17 00:00:00 2001
> > From: Mark Rutland <mark.rutland@....com>
> > Date: Tue, 20 Feb 2024 11:09:17 +0000
> > Subject: [PATCH] arm64: clean up fixmap initalization
> > 
> > Currently we have redundant initialization of the fixmap, first in
> > early_fdt_map(), and then again in setup_arch() before we call
> > early_ioremap_init(). This redundant initialization isn't harmful, as
> > early_fixmap_init() happens to be idempotent, but it's redundant and
> > potentially confusing.
> > 
> > We need to call early_fixmap_init() before we map the FDT and before we
> > call early_ioremap_init(). Ideally we'd place early_fixmap_init() and
> > early_ioremap_init() in the same caller as early_ioremap_init() is
> > somewhat coupled with the fixmap code.
> > 
> > Clean this up by moving the calls to early_fixmap_init() and
> > early_ioremap_init() into a new early_mappings_init() function, and
> > calling this once in __primary_switched() before we call
> > early_fdt_map(). This means we initialize the fixmap once, and keep
> > early_fixmap_init() and early_ioremap_init() together.

> Thanks for this. Makes sense to me; would you post a proper patch
> so I can build and do a boot test, just to make sure?

I took a look, and Ard's recent changes to the boot code have removed the
duplicate call to early_fixmap_init() by other means, so we don't need to do
anything, and can forget about this patch. :)

Mark.