The driver core debugging messages are a mess.  This provides a unified
message that makes them actually useful.

The format for new kobject debug messages should be:
	driver/bus/class: 'OBJECT_NAME': FUNCTION_NAME: message.\n

Note, the class code is not changed in this patch due to pending patches
in my queue that this would conflict with.  A later patch will clean
them up.

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch removes the kobject, and a few other driver-core-only fields
out of struct driver and into the driver core only.  Now drivers can be
safely create on the stack or statically (like they currently are.)

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

struct bus_type is static everywhere in the kernel.  This moves the
kobject in the structure out of it, and a bunch of other private only to
the driver core fields are now moved to a private structure.  This lets
us dynamically create the backing kobject properly and gives us the
chance to be able to document to users exactly how to use the struct
bus_type as there are no fields they can improperly access.

Thanks to Kay for the build fixes on this patch.

Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Signed-off-by: NStefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

  CC      drivers/base/dd.o
drivers/base/dd.c:211: warning: =E2=80=98device_probe_drivers=E2=80=99 defi=
ned but not used

Looks like the following is dead.
Signed-off-by: NStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch removes the PCI_MULTITHREAD_PROBE option that had already 
been marked as broken.
Signed-off-by: NAdrian Bunk <bunk@stusta.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Don't fail bus_attach_device() if the device cannot be bound.

If dev->driver has been specified, reset it to NULL if device_bind_driver()
failed and add the device as an unbound device.  As a result,
bus_attach_device() now cannot fail, and we can remove some checking from
device_add().

Also remove an unneeded check in bus_rescan_devices_helper().
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make multithreaded probing work per subsystem instead of per driver.

It doesn't make much sense to probe the same device for multiple drivers in
parallel (after all, only one driver can bind to the device).  Instead, create
a probing thread for each device that probes the drivers one after another. 
Also make the decision to use multi-threaded probe per bus instead of per
device and adapt the pci code.
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Implement device resource management, in short, devres.  A device
driver can allocate arbirary size of devres data which is associated
with a release function.  On driver detach, release function is
invoked on the devres data, then, devres data is freed.

devreses are typed by associated release functions.  Some devreses are
better represented by single instance of the type while others need
multiple instances sharing the same release function.  Both usages are
supported.

devreses can be grouped using devres group such that a device driver
can easily release acquired resources halfway through initialization
or selectively release resources (e.g. resources for port 1 out of 4
ports).

This patch adds devres core including documentation and the following
managed interfaces.

* alloc/free	: devm_kzalloc(), devm_kzfree()
* IO region	: devm_request_region(), devm_release_region()
* IRQ		: devm_request_irq(), devm_free_irq()
* DMA		: dmam_alloc_coherent(), dmam_free_coherent(),
		  dmam_declare_coherent_memory(), dmam_pool_create(),
		  dmam_pool_destroy()
* PCI		: pcim_enable_device(), pcim_pin_device(), pci_is_managed()
* iomap		: devm_ioport_map(), devm_ioport_unmap(), devm_ioremap(),
		  devm_ioremap_nocache(), devm_iounmap(), pcim_iomap_table(),
		  pcim_iomap(), pcim_iounmap()
Signed-off-by: NTejun Heo <htejun@gmail.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

Change function call order in device_bind_driver().

If we create symlinks (which might fail) before adding the device to the list
we don't have to clean up afterwards (which we didn't).
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Don't stop on the first ->probe error that is not -ENODEV/-ENXIO.

There might be a driver registered returning an unresonable return code, and
this stops probing completely even though it may make sense to try the next
possible driver. At worst, we may end up with an unbound device.
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Create the "driver" link before the child device may be created by
the probing logic. This makes it possible for userspace (udev), to
determine the driver property of the parent device, at the time the
child device is created.
Signed-off-by: NKay Sievers <kay.sievers@novell.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

I finally did as you suggested and added the notifier to the struct
bus_type itself. There are still problems to be expected is something
attaches to a bus type where the code can hook in different struct
device sub-classes (which is imho a big bogosity but I won't even try to
argue that case now) but it will solve nicely a number of issues I've
had so far.

That also means that clients interested in registering for such
notifications have to do it before devices are added and after bus types
are registered. Fortunately, most bus types that matter for the various
usage scenarios I have in mind are registerd at postcore_initcall time,
which means I have a really nice spot at arch_initcall time to add my
notifiers.

There are 4 notifications provided. Device being added (before hooked to
the bus) and removed (failure of previous case or after being unhooked
from the bus), along with driver being bound to a device and about to be
unbound.

The usage I have for these are:

 - The 2 first ones are used to maintain a struct device_ext that is
hooked to struct device.firmware_data. This structure contains for now a
pointer to the Open Firmware node related to the device (if any), the
NUMA node ID (for quick access to it) and the DMA operations pointers &
iommu table instance for DMA to/from this device. For bus types I own
(like IBM VIO or EBUS), I just maintain that structure directly from the
bus code when creating the devices. But for bus types managed by generic
code like PCI or platform (actually, of_platform which is a variation of
platform linked to Open Firmware device-tree), I need this notifier.

 - The other two ones have a completely different usage scenario. I have
cases where multiple devices and their drivers depend on each other. For
example, the IBM EMAC network driver needs to attach to a MAL DMA engine
which is a separate device, and a PHY interface which is also a separate
device. They are all of_platform_device's (well, about to be with my
upcoming patches) but there is no say in what precise order the core
will "probe" them and instanciate the various modules. The solution I
found for that is to have the drivers for emac to use multithread_probe,
and wait for a driver to be bound to the target MAL and PHY control
devices (the device-tree contains reference to the MAL and PHY interface
nodes, which I can then match to of_platform_devices). Right now, I've
been polling, but with that notifier, I can more cleanly wait (with a
timeout of course).
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The multithreaded-probing code has a problem: after one initcall level (eg,
core_initcall) has been processed, we will then start processing the next
level (postcore_initcall) while the kernel threads which are handling
core_initcall are still executing.  This breaks the guarantees which the
layered initcalls previously gave us.

IOW, we want to be multithreaded _within_ an initcall level, but not between
different levels.

Fix that up by causing the probing code to wait for all outstanding probes at
one level to complete before we start processing the next level.

Cc: Greg KH <greg@kroah.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

driver_probe_device() is missing kmalloc() failure check.
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Make sure data is freed if the kthread fails to start.
Signed-off-by: NDuncan Sands <baldrick@free.fr>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

There is a potential deadlock in the driver core.  It boils down to
the fact that bus_remove_device() calls klist_remove() instead of
klist_del(), thereby waiting until the reference count of the
klist_node in the bus's klist of devices drops to 0.  The refcount
can't reach 0 so long as a modprobe process is trying to bind a new
driver to the device being removed, by calling __driver_attach().  The
problem is that __driver_attach() tries to acquire the device's
parent's semaphore, but the caller of bus_remove_device() is quite
likely to own that semaphore already.

It isn't sufficient just to replace klist_remove() with klist_del().
Doing so runs the risk that the device would remain on the bus's klist
of devices for some time, and so could be bound to another driver even
after it was unregistered.  What's needed is a new way to distinguish
whether or not a device is registered, based on a criterion other than
whether its klist_node is linked into the bus's klist of devices.  That
way driver binding can fail when the device is unregistered, even if
it is still linked into the klist.

This patch (as782) implements the solution, by adding a new bitflag to
indiate when a struct device is registered, by testing the flag before
allowing a driver to bind a device, and by changing the definition of
the device_is_registered() inline.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This adds the infrastructure for drivers to do a threaded probe, and
waits at init time for all currently outstanding probes to complete.

A new kernel thread will be created when the probe() function for the
driver is called, if the multithread_probe bit is set in the driver
saying it can support this kind of operation.

I have tested this with USB and PCI, and it works, and shaves off a lot
of time in the boot process, but there are issues with finding root boot
disks, and some USB drivers assume that this can never happen, so it is
currently not enabled for any bus type.  Individual drivers can enable
this right now if they wish, and bus authors can selectivly turn it on
as well, once they determine that their subsystem will work properly
with it.
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Add lots of return-value checking.

<pcornelia.huck@de.ibm.com>: fix bus_rescan_devices()]
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: NCornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch (as667) changes the __device_release_driver() routine to
prevent it from crashing when it runs across a device not on any bus.
This seems logical, inasmuch as the corresponding bus_add_device()
routine has an explicit check allowing it to accept such devices.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

Add bus_type probe, remove and shutdown methods to replace the
corresponding methods in struct device_driver.  This matches
the way we handle the suspend/resume methods.

Since the bus methods override the device_driver methods, warn
if a device driver is registered whose methods will not be
called.

The long-term idea is to remove the device_driver methods entirely.
Signed-off-by: NRussell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch (as604) makes the driver core hold a device's parent's lock
as well as the device's lock during calls to the probe and remove
methods in a driver.  This facility is needed by USB device drivers,
owing to the peculiar way USB devices work:

	A device provides multiple interfaces, and drivers are bound
	to interfaces rather than to devices;

	Nevertheless a reset, reset-configuration, suspend, or resume
	affects the entire device and requires the caller to hold the
	lock for the device, not just a lock for one of the interfaces.

Since a USB driver's probe method is always called with the interface
lock held, the locking order rules (always lock parent before child)
prevent these methods from acquiring the device lock.  The solution
provided here is to call all probe and remove methods, for all devices
(not just USB), with the parent lock already acquired.

Although currently only the USB subsystem requires these changes, people
have mentioned in prior discussion that the overhead of acquiring an
extra semaphore in all the prove/remove sequences is not overly large.

Up to now, the USB core has been using its own set of private
semaphores.  A followup patch will remove them, relying entirely on the
device semaphores provided by the driver core.

The code paths affected by this patch are:

	device_add and device_del: The USB core already holds the parent
	lock, so no actual change is needed.

	driver_register and driver_unregister: The driver core will now
	lock both the parent and the device before probing or removing.

	driver_bind and driver_unbind (in sysfs): These routines will
	now lock both the parent and the device before binding or
	unbinding.

	bus_rescan_devices: The helper routine will lock the parent
	before probing a device.

I have not tested this patch for conflicts with other subsystems.  As
far as I can see, the only possibility of conflict would lie in the
bus_rescan_devices pathway, and it seems pretty remote.  Nevertheless,
it would be good for this to get a lot of testing in -mm.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch (as603) makes a few small fixes to the driver core:

Change spin_lock_irq for a klist lock to spin_lock;

Fix reference count leaks;

Minor spelling and formatting changes.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Acked-by Patrick Mochel <mochel@digitalimplant.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

bus_rescan_devices_helper() does not hold the dev->sem when it checks for
!dev->driver().  device_attach() holds the sem, but calls again
device_bind_driver() even when dev->driver is set.

What happens is that a first device_attach() call (module insertion time)
is on the way binding the device to a driver.  Another thread calls
bus_rescan_devices().  Now when bus_rescan_devices_helper() checks for
dev->driver it is still NULL 'cos the the prior device_attach() is not yet
finished.  But as soon as the first one releases the dev->sem the second
device_attach() tries to rebind the already bound device again.
device_bind_driver() does this blindly which leads to a corrupt
driver->klist_devices list (the device links itself, the head points to the
device).  Later a call to device_release_driver() sets dev->driver to NULL
and breaks the link it has to itself on knode_driver.  Rmmoding the driver
later calls driver_detach() which leads to an endless loop 'cos the list
head in klist_devices still points to the device.  And since dev->driver is
NULL it's stuck with the same device forever.  Boom.  And rmmod hangs.

Very easy to reproduce with new-style pcmcia and a 16bit card.  Just loop
modprobe <pcmcia-modules> ;cardctl eject; rmmod <card driver, pcmcia
modules>.

Easiest fix is to check if the device is already bound to a driver in
device_bind_driver().  This avoids the double binding.
Signed-off-by: NDaniel Ritz <daniel.ritz@gmx.ch>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

at the moment, the list_head semantics are

list_add(node, head)

whereas current klist semantics are

klist_add(head, node)

This is bound to cause confusion, and since klist is the newcomer, it
should follow the list_head semantics.

I also added missing include guards to klist.h
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This adds a single file, "bind", to the sysfs directory of every driver
registered with the driver core.  To bind a device to a driver, write
the bus id of the device you wish to bind to that specific driver to the
"bind" file (remember to not add a trailing \n).  If that bus id matches
a device on that bus, and it does not currently have a driver bound to
it, the probe sequence will be initiated with that driver and device.

Note, this requires that the driver itself be willing and able to accept
that device (usually through a device id type table).  This patch does
not make it possible to override the driver's id table.
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

The error handling in bus_add_device() and device_attach() is simply
non-existing. This patch propagates any error from device_attach to
the upper layers to allow for a proper recovery.

From: Hannes Reinecke <hare@suse.de>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This patch is intended for your "driver" tree.  It fixes several subtle
races in driver_detach() and device_release_driver() in the driver-model
core.

The major change is to use klist_remove() rather than klist_del() when
taking a device off its driver's list.  There's no other way to guarantee
that the list pointers will be updated before some other driver binds to
the device.  For this to work driver_detach() can't use a klist iterator,
so the loop over the devices must be written out in full.  In addition the
patch protects against the possibility that, when a driver and a device
are unregistered at the same time, one may be unloaded from memory before
the other is finished using it.
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

There's no check to see if the device is already bound to a driver, which
could do bad things.  The first thing to go wrong is that it will try to match
a driver with a device already bound to one.  In some cases (it appears with
USB with drivers/usb/core/usb.c::usb_match_id()), some drivers will match a
device based on the class type, so it would be common (especially for HID
devices) to match a device that is already bound.

The fun comes when ->probe() is called, it fails, then
driver_probe_device() does this:

	dev->driver = NULL;

Later on, that pointer could be be dereferenced without checking and cause
hell to break loose.

This problem could be nasty. It's very hardware dependent, since some
devices could have a different set of matching qualifiers than others.

Now, I don't quite see exactly where/how you were getting that crash.
You're dereferencing bad memory, but I'm not sure which pointer was bad
and where it came from, but it could have come from a couple of different
places.

The patch below will hopefully fix it all up for you. It's against
2.6.12-rc2-mm1, and does the following:

- Move logic to driver_probe_device() and comments uncommon returns:
  1 - If device is bound
  0 - If device not bound, and no error
  error - If there was an error.

- Move locking to caller of that function, since we want to lock a
  device for the entire time we're trying to bind it to a driver (to
  prevent against a driver being loaded at the same time).

- Update __device_attach() and __driver_attach() to do that locking.

- Check if device is already bound in __driver_attach()

- Update the converse device_release_driver() so it locks the device
  around all of the operations.

- Mark driver_probe_device() as static and remove export. It's an
  internal function, it should stay that way, and there are no other
  callers. If there is ever a need to export it, we can audit it as
  necessary.
Signed-off-by: NAndrew Morton <akpm@osdl.org>

Also stops looping over the lists when a match is found.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de

- Can't wait on removing the current item in the list (the positive refcount *because*
  we are using it causes it to deadlock).
Signed-off-by: NPatrick Mochel <mochel@digitalimplant.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

- Now possible, since the lists are locked using the klist lock and not the
  global rwsem.
Signed-off-by: NPatrick Mochel <mochel@digitalimplant.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

- Use it in driver_for_each_device() instead of the regular list_head and stop using
  the bus's rwsem for protection.
- Use driver_for_each_device() in driver_detach() so we don't deadlock on the
  bus's rwsem.
- Remove ->devices.
- Move klist access and sysfs link access out from under device's semaphore, since
  they're synchronized through other means.
Signed-off-by: NPatrick Mochel <mochel@digitalimplant.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>

This relocates the driver binding/unbinding code to drivers/base/dd.c. This is done
for two reasons: One, it's not code related to the bus_type itself; it uses some from
that, some from devices, and some from drivers. And Two, it will make it easier to do
some of the upcoming lock removal on that code..
Signed-off-by: NPatrick Mochel <mochel@digitalimplant.org>
Signed-off-by: NGreg Kroah-Hartman <gregkh@suse.de>