Re: [PATCH] acpi, nfit: fix acpi event notifications for nfit

[Dan Williams]

On Thu, Aug 18, 2016 at 11:48 AM, Vishal Verma <vishal....@intel.com> wrote:
> The nfit driver had an acpi event notification handler, but it never
> would've worked because we weren't setting the
> ACPI_DRIVER_ALL_NOTIFY_EVENTS flag in acpi_driver.

Let's update the changelog to be helpful for someone implementing a
backport or taking this back to a -stable branch. Something like:

Subject: acpi, nfit: fix event notifications

Commit 209851649dc4 "acpi: nfit: Add support for hot-add" added
support for _FIT notifications, but it neglected to set the
ACPI_DRIVER_ALL_NOTIFY_EVENTS flag that acpi_bus_notify() uses to gate
notification delivery.

Fixes: 209851649dc4 ("acpi: nfit: Add support for hot-add")

[Vishal Verma]

The nfit driver had an acpi event notification handler, but it never
would've worked because we weren't setting the
ACPI_DRIVER_ALL_NOTIFY_EVENTS flag in acpi_driver.

Cc: <sta...@vger.kernel.org>
Cc: <linux...@vger.kernel.org>
Cc: Dan Williams <dan.j.w...@intel.com>
Cc: Xiao Guangrong <guangro...@linux.intel.com>
Signed-off-by: Vishal Verma <vishal....@intel.com>
---
drivers/acpi/nfit/core.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 80cc7c0..5d33d81 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -2731,6 +2731,7 @@ MODULE_DEVICE_TABLE(acpi, acpi_nfit_ids);
static struct acpi_driver acpi_nfit_driver = {
.name = KBUILD_MODNAME,
.ids = acpi_nfit_ids,
+ .flags = ACPI_DRIVER_ALL_NOTIFY_EVENTS,
.ops = {
.add = acpi_nfit_add,
.remove = acpi_nfit_remove,
--
2.7.4

[Dan Williams]

On Thu, Aug 18, 2016 at 12:52 PM, Linda Knippers <linda.k...@hpe.com> wrote:

> While we're at it, should we update the notifier function to explicitly check
> for event 0x80 before re-evaluating the _FIT? I'm thinking about some time
> in the future when there might be more than one event.

Yes, good idea.

[Dan Williams]

On Thu, Aug 18, 2016 at 1:22 PM, Verma, Vishal L
<vishal....@intel.com> wrote:
> On Thu, 2016-08-18 at 12:54 -0700, Dan Williams wrote:
>> On Thu, Aug 18, 2016 at 12:52 PM, Linda Knippers <linda.knippers@hpe.c

>> om> wrote:
>> >
>> >
>> >
>> > On 8/18/2016 3:48 PM, Dan Williams wrote:
>> > >

>> > > On Thu, Aug 18, 2016 at 11:48 AM, Vishal Verma <vishal.l.verma@int

>> > > el.com> wrote:
>> > > >
>> > > > The nfit driver had an acpi event notification handler, but it
>> > > > never
>> > > > would've worked because we weren't setting the
>> > > > ACPI_DRIVER_ALL_NOTIFY_EVENTS flag in acpi_driver.
>> > >
>> > > Let's update the changelog to be helpful for someone implementing
>> > > a
>> > > backport or taking this back to a -stable branch. Something like:
>> > >
>> > > Subject: acpi, nfit: fix event notifications
>> > >
>> > > Commit 209851649dc4 "acpi: nfit: Add support for hot-add" added
>> > > support for _FIT notifications, but it neglected to set the
>> > > ACPI_DRIVER_ALL_NOTIFY_EVENTS flag that acpi_bus_notify() uses to
>> > > gate
>> > > notification delivery.
>> >
>> > While we're at it, should we update the notifier function to
>> > explicitly check
>> > for event 0x80 before re-evaluating the _FIT? I'm thinking about
>> > some time
>> > in the future when there might be more than one event.
>>
>> Yes, good idea.
>

> Sounds good, part of the same patch or separately?

Separate.

In general, if you would write the word "also" in a patch description,
that's an indication to make a separate patch.

[Verma, Vishal L]

On Thu, 2016-08-18 at 12:54 -0700, Dan Williams wrote:
> On Thu, Aug 18, 2016 at 12:52 PM, Linda Knippers <linda.knippers@hpe.c

> om> wrote:
> >
> >
> >
> > On 8/18/2016 3:48 PM, Dan Williams wrote:
> > >

> > > On Thu, Aug 18, 2016 at 11:48 AM, Vishal Verma <vishal.l.verma@int

> > > el.com> wrote:
> > > >
> > > > The nfit driver had an acpi event notification handler, but it
> > > > never
> > > > would've worked because we weren't setting the
> > > > ACPI_DRIVER_ALL_NOTIFY_EVENTS flag in acpi_driver.
> > >
> > > Let's update the changelog to be helpful for someone implementing
> > > a
> > > backport or taking this back to a -stable branch.  Something like:
> > >
> > > Subject: acpi, nfit: fix event notifications
> > >
> > > Commit 209851649dc4 "acpi: nfit: Add support for hot-add" added
> > > support for _FIT notifications, but it neglected to set the
> > > ACPI_DRIVER_ALL_NOTIFY_EVENTS flag that acpi_bus_notify() uses to
> > > gate
> > > notification delivery.
> >
> > While we're at it, should we update the notifier function to
> > explicitly check
> > for event 0x80 before re-evaluating the _FIT?  I'm thinking about
> > some time
> > in the future when there might be more than one event.
>
> Yes, good idea.