linux-kernel - Re: setup_boot_APIC_clock() NULL dereference during early boot on reduced hardware platforms

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: <CAD8Lp452GdoL-Bt7rSP=u3RKEZ2H3qm3LvKfe=cCsjP0biG_sQ@mail.gmail.com>
Date:   Thu, 1 Aug 2019 17:00:23 +0800
From:   Daniel Drake <drake@...lessm.com>
To:     Aubrey Li <aubrey.intel@...il.com>
Cc:     Thomas Gleixner <tglx@...utronix.de>, x86@...nel.org,
        "Li, Aubrey" <aubrey.li@...ux.intel.com>,
        Ingo Molnar <mingo@...hat.com>,
        "H . Peter Anvin" <hpa@...or.com>,
        Linux Kernel <linux-kernel@...r.kernel.org>,
        Endless Linux Upstreaming Team <linux@...lessm.com>
Subject: Re: setup_boot_APIC_clock() NULL dereference during early boot on
 reduced hardware platforms

On Thu, Aug 1, 2019 at 3:16 PM Aubrey Li <aubrey.intel@...il.com> wrote:
> No, the platform needs a global clock event, can you turn on some other
> clock source on your platform, like HPET?

Thanks Audrey and Thomas for the quick hints!

I double checked under Windows - it seems to be using a HPET there.
Also there is the HPET ACPI table. So I think this is the right angle
to look at.

Under Linux, hpet_legacy_clockevent_register() is the function where
global_clock_event can be set to HPET.

However, the only way this can be called is from hpet_enable().

hpet_enable() is called from 2 places:
 1. From hpet_time_init(). This is the default x86 timer_init that
acpi_generic_reduced_hw_init() took out of action here.
 2. From hpet_late_init(). However that function is only called late,
after calibrate_APIC_clock() has already crashed the kernel. Also,
even if moved earlier it would also not call hpet_enable() here
because the ACPI HPET table parsing has already populated
hpet_address.

I tried slotting in a call to hpet_enable() at an earlier point
regardless, but I still end up with the kernel hanging later during
boot, probably because irq0 fails to be setup and this error is hit:
    if (setup_irq(0, &irq0))
        pr_info("Failed to register legacy timer interrupt\n");

I'll go deeper into that; further hints welcome too.

Thanks
Daniel