linux-kernel - Machine specific static mappings iotable

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]

Date:   Tue, 6 Aug 2019 18:49:52 +0530
From:   Suniel Mahesh <sunil.m@...hveda.org>
To:     linux-arm-kernel@...ts.infradead.org, linux@...linux.org.uk,
        Linux kernel mailing list <linux-kernel@...r.kernel.org>
Cc:     thomas.petazzoni@...e-electrons.com
Subject: Machine specific static mappings iotable_init(), are they required ?

Hi,

I am trying to port a machine based on arm926 with MMU, having 64MB of RAM.  

I am trying to understand the difference between: 

machine specific static I/O mappings which are done via iotable_init()
(done via callback .map_io in DT_MACHINE_START) and 
dynamic I/O mappings done via ioremap()

In the kernel docs/mailing list, I have encountered a statement which states:

"with machine specific static I/O mappings which are done via iotable_init(), 
registers can be mapped at the upper end of vmalloc area so that one can use as
little of the VA space as possible so vmalloc and friends have a better chance of 
getting memory"

I am writing board initialization C file and got stuck at .map_io callback function,
whether to define it or not. If yes, under what scenario should I do it

now-a-days I think less boards are using iotable_init(). (is this defunct) ? 
I might be wrong here

Can't I use ioremap and do dynamic mappings when ever required via device tree ?
If I do so will I encounter any problems with vmalloc area.

Thanks & Regards
-- 
Suniel Mahesh