lkml.org 
[lkml]   [2013]   [Feb]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH] ACPI / scan: Make it clear that acpi_bus_trim() cannot fail
Date
On Monday, February 18, 2013 01:25:49 PM Vasilis Liaskovitis wrote:
> Hi,
>
> On Thu, Jan 24, 2013 at 11:56:50PM +0100, Rafael J. Wysocki wrote:
> > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> >
> > Since acpi_bus_trim() cannot fail, change its definition to a void
> > function, so that its callers don't check the return value in vain
> > and update the callers.
>
> I have missed a few patchsets/discussions in the last month and wanted to
> ask a question related to this: Does the new always-succeed 2-pass
> trim_device design guarantee safe memory hot-remove operations?

I doesn't by itself. Nor it really can, because the .remove() callbacks of
device drivers are not allowed to fail.

> Afaict if memory offline fails now, the device is ejected (_EJ0) anyways
> causing a panic. Tested in a VM with linux-next-20130207 and
> linux-next-20130218 by doing an SCI-eject request on a hot-plugged dimm.
>
> Are there more patches in development for safe memory hot-remove?

Yes, there are. I sent a patch series yesterday introducing some safety
measures (you can disable memory hotplug from user space or disable
automatic ejection). There's more to come still.

The plan is to introduce offline/online operations for memory modules (in
analogy with CPU core online/offline) that can be started by user space and
memory eject will only be possible after offline (i.e. when the memory module
is known to be not in use).

Thanks,
Rafael


--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.


\
 
 \ /
  Last update: 2013-02-18 15:01    [W:2.570 / U:0.036 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site