Re: [PATCH 1/6] ACPI: Separate adding ACPI device objects from probing ACPI drivers

From: "Rafael J. Wysocki" <rjw@sisk.pl>
To: Bjorn Helgaas <bhelgaas@google.com>
Cc: LKML <linux-kernel@vger.kernel.org>,
	ACPI Devel Maling List <linux-acpi@vger.kernel.org>,
	linux-pci@vger.kernel.org, Yinghai Lu <yinghai@kernel.org>,
	Toshi Kani <toshi.kani@hp.com>,
	Myron Stowe <myron.stowe@redhat.com>
Subject: Re: [PATCH 1/6] ACPI: Separate adding ACPI device objects from probing ACPI drivers
Date: Thu, 13 Dec 2012 12:41:38 +0100	[thread overview]
Message-ID: <9837718.ktuaxVB488@vostro.rjw.lan> (raw)
In-Reply-To: <CAErSpo7=2WUbNgtFHT6Sm6h6OKfXkSH99qSoD0n73Tga0=XpTA@mail.gmail.com>

On Wednesday, December 12, 2012 06:00:03 PM Bjorn Helgaas wrote:
> On Sun, Dec 9, 2012 at 4:00 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> >
> > Currently, as soon as an ACPI device node object (struct acpi_device)
> > is created, the driver core attempts to probe ACPI drivers against
> > it.  That leads to some unpleasant side effects, like the fact that
> > the boot code path for ACPI namespace scanning is different from the
> > analogous hot-plug code path (during boot ACPI drivers are not
> > present when ACPI device node objects are registered, so they are
> > guaranteed not to be probed, which is not the case during hot-plug).
> > That, in turn, leads to unnecessary complications in the PCI
> > enumeration algorithm.
> 
> Can you elaborate a bit on the complications in PCI enumeration?
> Hopefully this will lead to some simplification in PCI enumeration,
> but these patches don't go that far (yet).

For example, in the hotplug case it is not guaranteed that companion
struct acpi_device objects have been created already when struct pci_dev
objects are created during acpi_pci_root_add().  This forces us to use
acpi_get_pci_dev() in acpi_pci_bind(), which we could avoid otherwise.

Moreover, if we knew that the companion struct acpi_device objects existed as
soon as in pci_scan_device(), we could just populate the ACPI handle of the
struct pci_dev in there and that would allow acpi_platform_notify() later on
to skip the bus browsing (due to recent changes just merged in).  I suppose
we could even get rid of acpi_pci_bind()/acpi_pci_unbind() by integrating
that code directly into the initialization/removal code paths.

> > Reduce the differences between the boot and hot-plug cases by
> > splitting the ACPI namespace scanning for devices into two passes,
> > such that struct acpi_device objects are registerd in the first
> > patch without probing ACPI drivers and the drivers are probed
> > against them directly in the second pass.  This way ACPI drivers
> > can assume that all of the ACPI device node objects in the given
> > scope will be registered when their .add() routines run and the
> > hot-plug case becomes the same as the boot case from their
> > perspective.
> 
> If I understand correctly, you're talking about a hierarchical
> topology where a device, e.g., a bridge, can have subordinate devices
> below it.
> 
> In general terms, if a driver claims X, and the driver's add() method
> can assume that every device below X has already been registered, I
> guess that means any bridge drivers must be integrated into the core,
> so the core can enumerate things below the bridge without the driver's
> help.

Well, I don't quite understand.  Do you have an example?

We're registering only struct acpi_device things kind of in advance here.
I would assume that device drivers would operate on device structures of
other types (like struct pci_dev) and those struct acpi_device objects
would only be companions of the "physical" device nodes.

IOW what we do here is to extract some information from ACPI tables and
put it into a number of objects in the device hierarchy (the fact that they
are in the device hierarchy allows us to export some of their properties via
sysfs easily), but that's about it.  How buses/drivers use that information
is up to them.

And by the way, I think that binding drivers to struct acpi_device objects
directly is too much of a shortcut and it leads to more problems than it
solves.  I'd like to get rid of that eventually.

> The PCI core effectively has the P2P bridge driver integrated
> into it, so that's the way it is already for PCI.  I can imagine
> scenarios, e.g., NTBs, where it's not practical to integrate the
> bridge driver, so I'm not 100% comfortable with this assumption.  But
> I admit that's a totally hypothetical situation, and there are lots of
> other obstacles to non-integrated bridge drivers, like the fact that
> we can't do dynamic resource assignment.

Well, precisely. :-)

> s/registerd/registered/ above

Yup, thanks!

Thanks a lot for having a look. :-)

Rafael

> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > ---
> >  drivers/acpi/scan.c     |   96 +++++++++++++++++++++++++++++++++---------------
> >  include/acpi/acpi_bus.h |    1
> >  2 files changed, 68 insertions(+), 29 deletions(-)
> >
> > Index: linux/include/acpi/acpi_bus.h
> > ===================================================================
> > --- linux.orig/include/acpi/acpi_bus.h
> > +++ linux/include/acpi/acpi_bus.h
> > @@ -98,6 +98,7 @@ typedef void (*acpi_op_notify) (struct a
> >  struct acpi_bus_ops {
> >         u32 acpi_op_add:1;
> >         u32 acpi_op_start:1;
> > +       u32 acpi_op_match:1;
> >  };
> >
> >  struct acpi_device_ops {
> > Index: linux/drivers/acpi/scan.c
> > ===================================================================
> > --- linux.orig/drivers/acpi/scan.c
> > +++ linux/drivers/acpi/scan.c
> > @@ -494,7 +494,8 @@ static int acpi_bus_match(struct device
> >         struct acpi_device *acpi_dev = to_acpi_device(dev);
> >         struct acpi_driver *acpi_drv = to_acpi_driver(drv);
> >
> > -       return !acpi_match_device_ids(acpi_dev, acpi_drv->ids);
> > +       return acpi_dev->bus_ops.acpi_op_match
> > +               && !acpi_match_device_ids(acpi_dev, acpi_drv->ids);
> >  }
> >
> >  static int acpi_device_uevent(struct device *dev, struct kobj_uevent_env *env)
> > @@ -1418,6 +1419,17 @@ static int acpi_bus_remove(struct acpi_d
> >         return 0;
> >  }
> >
> > +/*
> > + * acpi_hot_add_bind - Bind _ADR-based devices on hot-add.
> > + * @device: ACPI device node to bind.
> > + */
> > +static void acpi_hot_add_bind(struct acpi_device *device)
> > +{
> > +       if (device->flags.bus_address
> > +           && device->parent && device->parent->ops.bind)
> > +               device->parent->ops.bind(device);
> > +}
> > +
> >  static int acpi_add_single_object(struct acpi_device **child,
> >                                   acpi_handle handle, int type,
> >                                   unsigned long long sta,
> > @@ -1490,13 +1502,8 @@ static int acpi_add_single_object(struct
> >
> >         result = acpi_device_register(device);
> >
> > -       /*
> > -        * Bind _ADR-Based Devices when hot add
> > -        */
> > -       if (device->flags.bus_address) {
> > -               if (device->parent && device->parent->ops.bind)
> > -                       device->parent->ops.bind(device);
> > -       }
> > +       if (device->bus_ops.acpi_op_match)
> > +               acpi_hot_add_bind(device);
> >
> >  end:
> >         if (!result) {
> > @@ -1522,6 +1529,7 @@ static void acpi_bus_add_power_resource(
> >         struct acpi_bus_ops ops = {
> >                 .acpi_op_add = 1,
> >                 .acpi_op_start = 1,
> > +               .acpi_op_match = 1,
> >         };
> >         struct acpi_device *device = NULL;
> >
> > @@ -1574,9 +1582,9 @@ static acpi_status acpi_bus_check_add(ac
> >                                       void *context, void **return_value)
> >  {
> >         struct acpi_bus_ops *ops = context;
> > +       struct acpi_device *device = NULL;
> >         int type;
> >         unsigned long long sta;
> > -       struct acpi_device *device;
> >         acpi_status status;
> >         int result;
> >
> > @@ -1600,48 +1608,77 @@ static acpi_status acpi_bus_check_add(ac
> >          * We may already have an acpi_device from a previous enumeration.  If
> >          * so, we needn't add it again, but we may still have to start it.
> >          */
> > -       device = NULL;
> >         acpi_bus_get_device(handle, &device);
> >         if (ops->acpi_op_add && !device) {
> > -               acpi_add_single_object(&device, handle, type, sta, ops);
> > -               /* Is the device a known good platform device? */
> > -               if (device
> > -                   && !acpi_match_device_ids(device, acpi_platform_device_ids))
> > -                       acpi_create_platform_device(device);
> > -       }
> > +               struct acpi_bus_ops add_ops = *ops;
> >
> > -       if (!device)
> > -               return AE_CTRL_DEPTH;
> > -
> > -       if (ops->acpi_op_start && !(ops->acpi_op_add)) {
> > -               status = acpi_start_single_object(device);
> > -               if (ACPI_FAILURE(status))
> > +               add_ops.acpi_op_match = 0;
> > +               acpi_add_single_object(&device, handle, type, sta, &add_ops);
> > +               if (!device)
> >                         return AE_CTRL_DEPTH;
> > +
> > +               device->bus_ops.acpi_op_match = 1;
> >         }
> >
> >         if (!*return_value)
> >                 *return_value = device;
> > +
> >         return AE_OK;
> >  }
> >
> > +static acpi_status acpi_bus_probe_start(acpi_handle handle, u32 lvl,
> > +                                       void *context, void **not_used)
> > +{
> > +       struct acpi_bus_ops *ops = context;
> > +       struct acpi_device *device;
> > +       acpi_status status = AE_OK;
> > +
> > +       if (acpi_bus_get_device(handle, &device))
> > +               return AE_CTRL_DEPTH;
> > +
> > +       if (ops->acpi_op_add) {
> > +               if (!acpi_match_device_ids(device, acpi_platform_device_ids)) {
> > +                       /* This is a known good platform device. */
> > +                       acpi_create_platform_device(device);
> > +               } else {
> > +                       int ret = device_attach(&device->dev);
> > +                       acpi_hot_add_bind(device);
> > +                       if (ret)
> > +                               status = AE_CTRL_DEPTH;
> > +               }
> > +       } else if (ops->acpi_op_start) {
> > +               if (ACPI_FAILURE(acpi_start_single_object(device)))
> > +                       status = AE_CTRL_DEPTH;
> > +       }
> > +       return status;
> > +}
> > +
> >  static int acpi_bus_scan(acpi_handle handle, struct acpi_bus_ops *ops,
> >                          struct acpi_device **child)
> >  {
> > -       acpi_status status;
> >         void *device = NULL;
> > +       acpi_status status;
> > +       int ret = 0;
> >
> >         status = acpi_bus_check_add(handle, 0, ops, &device);
> > -       if (ACPI_SUCCESS(status))
> > +       if (ACPI_FAILURE(status)) {
> > +               ret = -ENODEV;
> > +               goto out;
> > +       }
> > +
> > +       acpi_walk_namespace(ACPI_TYPE_ANY, handle, ACPI_UINT32_MAX,
> > +                           acpi_bus_check_add, NULL, ops, &device);
> > +       if (device)
> >                 acpi_walk_namespace(ACPI_TYPE_ANY, handle, ACPI_UINT32_MAX,
> > -                                   acpi_bus_check_add, NULL, ops, &device);
> > +                                   acpi_bus_probe_start, NULL, ops, NULL);
> > +       else
> > +               ret = -ENODEV;
> >
> > + out:
> >         if (child)
> >                 *child = device;
> >
> > -       if (device)
> > -               return 0;
> > -       else
> > -               return -ENODEV;
> > +       return ret;
> >  }
> >
> >  /*
> > @@ -1752,6 +1789,7 @@ static int acpi_bus_scan_fixed(void)
> >         memset(&ops, 0, sizeof(ops));
> >         ops.acpi_op_add = 1;
> >         ops.acpi_op_start = 1;
> > +       ops.acpi_op_match = 1;
> >
> >         /*
> >          * Enumerate all fixed-feature devices.
> >
-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.