linux-kernel - Fwd: Kernel Power management ?? turning off used devices that are in use.

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, 04 Nov 2011 14:56:25 +0800
From:	hagar <hagar@...et.net.au>
To:	lkml <linux-kernel@...r.kernel.org>
CC:	Rusty Russell <rusty@...tcorp.com.au>,
	Srivatsa Vaddagiri <vatsa@...ibm.com>,
	Joel Schopp <jschopp@...tin.ibm.com>,
	Ashok Raj <ashok.raj@...el.com>,
	Heiko Carstens <heiko.carstens@...ibm.com>
Subject: Fwd: Kernel Power management ?? turning off used devices that are
 in use.

I have traced this behaviour to the cpu hotplug code.

I need to be able to either switch off cpu hotpluging - or at least 
prevent hotplugging certain cpu's

there should really be a mincpu's option or better a 
nohotplug="0,1,2,5,7,etc" option.

It would appear that cpu1 is the default cpu for those interrupts - and 
they are not being transferred to cpu0 properly.

I have tried the kernel parameters noapic nosmp maxcpus acpi=off 
possible_cpus=2 etc - with no success

I did find a flaw in "possible_cpus=0" - this prevents the kernel 
booting (no cpus available )
This option should be set to a minimum of 1

I also tried to switch off acpi in the bios - no effect.

the nosmp option still used SMP code
It actually switched back to SMP code to offline cpu1
which kinda defeated the purpose of the option.

If there is streaming video when the cpu gos offline - then the stream 
is cut off and frozen.

This behaviour is becoming intolerable.

-------- Original Message --------

Every time my computer is idle for about 30 sec's  the USB subsystem
completely switches off
and the eth0 goes down if there are no active computers.

This has become a major problem because the dhcp server can no longer
listen and the OpenVPN server crashes on that interface
Resulting in an unusable network.

The USB subsystem powers off every USB device and resets them on activity.

cat /sys/module/usbcore/parameters/autosuspend
-1

Not all devices behave well when doing this.
The mouse will not wake up the system when this happens.

The USB problem did exist in the 2.6 kernels - but on a longer delay -
several minutes.
But I am sure that the network problem did not exist ( At least it did
not show up )

The syslog trace always starts with "CPU 1 is now offline"

I have a 64bit Dual core 2.13 GHz system with 8G of memory, USB and 2
Intel Corporation 82573E/L Gigabit Ethernet Controllers on board.
I use a PCI Radeon HD 5000 Series with 3 screens at 1280x1024 - to a
Desktop of 3840x1024 ( OpenSource Radeon driver )
a PCIe Creative Labs [SB X-Fi Xtreme Audio]
and a PCIe Hauppauge WinTV-HVR2200

I have had to create a script -
/root/RestartNetwork
#!/bin/bash
/usr/bin/logger -p daemon.err -t "Network Check" "Eth0 is down -
Restarting Network"
/etc/init.d/networking stop
/etc/init.d/networking start
/etc/init.d/openvpn restart
/etc/init.d/isc-dhcp-server restart
/sbin/route | /bin/grep -q eth0 || /usr/bin/logger -p daemon.err -t
"Network Check" "Eth0 is still down"

and run it in the crontab -
*/1 * * * * /sbin/route | /bin/grep -q eth0 || /root/RestartNetwork

just to keep the network up.

View attachment "KernelBug.txt" of type "text/plain" (100601 bytes)