On Thu, Jun 05, 2014 at 07:05:24PM +0100, Greg Kroah-Hartman wrote:

On Thu, Jun 05, 2014 at 05:05:00PM +0100, Mark Rutland wrote:

quoted

Hi Greg,

On Thu, Jun 05, 2014 at 12:24:28AM +0100, Greg Kroah-Hartman wrote:

quoted

3.10-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Thomas Gleixner <redacted>

commit ffde1de64012c406dfdda8690918248b472f24e4 upstream.

To support the affinity setting of per cpu timers in the early startup
of a not yet online cpu, implement the force logic, which disables the
cpu online check.

Tagged for stable to allow a simple fix of the affected SoC clock
event drivers.

This patch alone has the possiblity of breaking CPU hotplug on arm and
arm64 (specifically it breaks hot unplugging CPU0 where interrupts may
be left targetting the offline CPU).

For arm64 [1] that's fixed by commit 601c942176d8 (arm64: use
cpu_online_mask when using forced irq_set_affinity).

Thanks, I'll apply that to the 3.14-stable tree, but, it doesn't seem
relevant at all for 3.10-stable as arch/arm64/kernel/irq.c doesn't do
anything with cpumask or set_affinity.  If it's relevant for 3.10,
please provide a backported version.

Sorry, my bad. It's not relevant for v3.10, I'd gotten confused over
when I added CPU hotplug.

Thanks,
Mark.

quoted

Unfortunately there is not an equivalent fix for arm - Russell objected
to the approach [2,3,4], and that's not yet settled.

Ok, be sure to cc: stable on the patch when it goes into the tree and
I'll pick it up then.

thanks,

greg k-h