Re: Cannot hot remove a memory device (patch, updated)

[Rafael J. Wysocki]

On Tuesday, August 06, 2013 11:12:51 AM Yasuaki Ishimatsu wrote:
> (2013/08/06 9:15), Rafael J. Wysocki wrote:
> > On Monday, August 05, 2013 05:19:56 PM Toshi Kani wrote:
> >> On Mon, 2013-08-05 at 15:14 +0200, Rafael J. Wysocki wrote:
> >>    :
> >>> Can you please test the appended patch?  I tested it somewhat, but since the
> >>> greatest number of physical nodes per ACPI device object I can get on my test
> >>> machines is 2 (and even that after hacking the kernel somewhat), that was kind
> >>> of unconclusive.
> >>>
> >>> Thanks,
> >>> Rafael
> >>>
> >>>
> >>> ---
> >>> From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> >>> Subject: ACPI: Drop physical_node_id_bitmap from struct acpi_device
> >>>
> >>> The physical_node_id_bitmap in struct acpi_device is only used for
> >>> looking up the first currently unused phyiscal dependent node ID
> >>> by acpi_bind_one().  It is not really necessary, however, because
> >>> acpi_bind_one() walks the entire physical_node_list of the given
> >>> device object for sanity checking anyway and if that list is always
> >>> sorted by node_id, it is straightforward to find the first gap
> >>> between the currently used node IDs and use that number as the ID
> >>> of the new list node.
> >>>
> >>> This also removes the artificial limit of the maximum number of
> >>> dependent physical devices per ACPI device object, which now depends
> >>> only on the capacity of unsigend int.
> >>>
> >>> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> >>
> >> I like the change. Much better :-)
> >>
> >> Acked-by: Toshi Kani <toshi.kani@xxxxxx>
> >
> > However, it introduces a bug in acpi_unbind_one(), because the size of the name
> > array in there has to be increased too.  Updated patch follows.
> >
> > Thanks,
> > Rafael
> >
> >
> > ---
> > From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> > Subject: ACPI: Drop physical_node_id_bitmap from struct acpi_device
> >
> > The physical_node_id_bitmap in struct acpi_device is only used for
> > looking up the first currently unused dependent phyiscal node ID
> > by acpi_bind_one().  It is not really necessary, however, because
> > acpi_bind_one() walks the entire physical_node_list of the given
> > device object for sanity checking anyway and if that list is always
> > sorted by node_id, it is straightforward to find the first gap
> > between the currently used node IDs and use that number as the ID
> > of the new list node.
> >
> > This also removes the artificial limit of the maximum number of
> > dependent physical devices per ACPI device object, which now depends
> > only on the capacity of unsigend int.
> >
> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> 
> Reviewed-by: Yasuaki Ishimatsu <isimatu.yasuaki@xxxxxxxxxxxxxx>
> Tested-by: Yasuaki Ishimatsu <isimatu.yasuaki@xxxxxxxxxxxxxx>
> 
> I confirmed that I can add and remove a memory device correctly.

Great, thanks for the confirmation!

Rafael

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/