linux-kernel - question on isolcpus and CONFIG

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: <CAC8NTUX15L9-WL0Gc7Knh092tAxG9V8ExEeRWJ4W=GZuWKg0pw@mail.gmail.com>
Date:   Tue, 11 Jul 2017 14:09:09 -0700
From:   Radha Mohan <mohun106@...il.com>
To:     linux-kernel@...r.kernel.org
Subject: question on isolcpus and CONFIG_PREEMPT behavior

Hi,
I got a question on the expected behavior with isolcpus and CONFIG_PREEMPT=y.

On a 10 core arm64 system and kernel config having CONFIG_PREEMPT=y,
I passed "isolcpus=1-9". The system boots fine and all the user
processes have affinity of 1 which is good. I have shell script that
prints some string in an infinite loop every 2 seconds - i run this in
background.

Now I crash the kernel by "echo c > /proc/sysrq-trigger". There's a
NULL pointer dereference crash followed by the backtrace. But I can
still see the output from the shell script I am running.

If I run the kernel with either CONFIG_PREEMPT_VOLUNTARY=y or
CONFIG_PREEMPT_NONE=y then I do not see this behaviour.

The user space processes still running which makes the system alive.
This also means the watchdog daemon can update the hardware watchdog
timer and the system can never reset.

Any pointers ?

- Radha