linux-kernel - [patch 0/2] timekeeping: NMI safe timekeeper enhancements

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-next>] [day] [month] [year] [list]

Date:   Fri, 14 Aug 2020 12:19:33 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     LKML <linux-kernel@...r.kernel.org>
Cc:     Petr Mladek <pmladek@...e.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Orson Zhai <orsonzhai@...il.com>,
        Prarit Bhargava <prarit@...hat.com>,
        Dave Young <dyoung@...hat.com>, Baoquan He <bhe@...hat.com>,
        Vivek Goyal <vgoyal@...hat.com>,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
        Steven Rostedt <rostedt@...dmis.org>,
        John Stultz <john.stultz@...aro.org>,
        Stephen Boyd <sboyd@...nel.org>, kexec@...ts.infradead.org
Subject: [patch 0/2] timekeeping: NMI safe timekeeper enhancements

printk intends to store various timestamps (MONOTONIC, REALTIME, BOOTTIME)
to make correlation of dmesg accross different machines easier.

The NMI safe timekeeper allows to retrieve these timestamps from any
context, but it lacks a few things:

  1) The nmi safe accessors are not providing time stamps until timekeeping
     is initialized during early boot.

     This can be mitigated by using sched clock up to the point where time-
     keeping becomes available. This has no side effects because clock
     monotonic takes sched clock into account at initialization time
     anyway. So no random time jumps are possible.

     If early sched clock is not available then there is no difference
     either, obviously. Both return 0.

  2) It requires a new accessor which allows to retrieve all three clock
     timestamps in one go.

     Trivial excercise. But there are a few twists:

     A) Access to boot time can be racy if the sleep time offset on resume
     	is injected after timekeeping resume. That's the case when the RTC
     	or whatever is used to calculate sleep time is not availble when
	the timekeeping core is resumed.

     B) Timestamps are frozen accross the very inner low level
     	suspend/resume path. Not a big problem, but might affect the
     	developer debug printks.

     A detailed description of these two points is in the changelog of
     patch 2.

Thanks,

	tglx