GitHub/exynos8895/android_kernel_samsung_universal8895.git
15 years agoPCI PM: Register power state of devices during initialization
Rafael J. Wysocki [Wed, 7 Jan 2009 12:10:35 +0000 (13:10 +0100)]
PCI PM: Register power state of devices during initialization

Use the observation that the power state of a PCI device can be
loaded into its pci_dev structure as soon as pci_pm_init() is run for
it and make that happen.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Call pci_fixup_device from legacy routines
Rafael J. Wysocki [Wed, 7 Jan 2009 12:09:37 +0000 (13:09 +0100)]
PCI PM: Call pci_fixup_device from legacy routines

The size of drivers/pci/pci-driver.c can be reduced quite a bit
if pci_fixup_device() is called from the legacy PM callbacks, so make
it happen.

No functional changes should result from this.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Rearrange code in pci-driver.c
Rafael J. Wysocki [Wed, 7 Jan 2009 13:15:17 +0000 (14:15 +0100)]
PCI PM: Rearrange code in pci-driver.c

Rename two functions and rearrange code in drivers/pci/pci-driver.c
so that it's easier to follow.  In particular, separate invocations
of the legacy callbacks from the rest of the new callbacks' code.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Avoid touching devices behind bridges in unknown state
Rafael J. Wysocki [Wed, 7 Jan 2009 12:07:15 +0000 (13:07 +0100)]
PCI PM: Avoid touching devices behind bridges in unknown state

It generally is better to avoid accessing devices behind bridges that
may not be in the D0 power state, because in that case the bridges'
secondary buses may not be accessible.  For this reason, during the
early phase of resume (ie. with interrupts disabled), before
restoring the standard config registers of a device, check the power
state of the bridge the device is behind and postpone the restoration
of the device's config space, as well as any other operations that
would involve accessing the device, if that state is not D0.

In such cases the restoration of the device's config space will be
retried during the "normal" phase of resume (ie. with interrupts
enabled), so that the bridge can be put into D0 before that happens.

Also, save standard configuration registers of PCI devices during the
"normal" phase of suspend (ie. with interrupts enabled), so that the
bridges the devices are behind can be put into low power states (we
don't put bridges into low power states at the moment, but we may
want to do it in the future and it seems reasonable to design for
that).

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Move pci_has_legacy_pm_support
Rafael J. Wysocki [Wed, 7 Jan 2009 12:06:10 +0000 (13:06 +0100)]
PCI PM: Move pci_has_legacy_pm_support

Move pci_has_legacy_pm_support() closer to the functions that
call it.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Power-manage devices without drivers during suspend-resume
Rafael J. Wysocki [Wed, 7 Jan 2009 12:05:05 +0000 (13:05 +0100)]
PCI PM: Power-manage devices without drivers during suspend-resume

PCI devices without drivers can be put into low power states during
suspend with the help of pci_prepare_to_sleep() and prevented from
generating wake-up events during resume with the help of
pci_enable_wake().  However, it's better not to put bridges into
low power states during suspend, because that might result in entire
bus segments being powered off.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Add suspend counterpart of pci_reenable_device
Rafael J. Wysocki [Wed, 7 Jan 2009 12:03:42 +0000 (13:03 +0100)]
PCI PM: Add suspend counterpart of pci_reenable_device

PCI devices without drivers are not disabled during suspend and
hibernation, but they are enabled during resume, with the help of
pci_reenable_device(), so there is an unbalanced execution of
pcibios_enable_device() in the resume code path.

To correct this introduce function pci_disable_enabled_device()
that will disable the argument device, if it is enabled when the
function is being run, without updating the device's pci_dev
structure and use it in the suspend code path to balance the
pci_reenable_device() executed during resume.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Fix poweroff and restore callbacks
Rafael J. Wysocki [Wed, 7 Jan 2009 12:02:36 +0000 (13:02 +0100)]
PCI PM: Fix poweroff and restore callbacks

pci_fixup_device() is called too early in pci_pm_poweroff() and too
late in pci_pm_restore().  Moreover, pci_pm_restore_noirq() calls
pci_fixup_device() twice and in a wrong way.  Fix that.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Use msleep instead of cpu_relax during ASPM link retraining
Andrew Patterson [Mon, 5 Jan 2009 23:21:04 +0000 (16:21 -0700)]
PCI: Use msleep instead of cpu_relax during ASPM link retraining

The cpu_relax() function can be a noop on certain architectures like
IA-64 when CPU threads are disabled, so use msleep instead during link
retraining busy/wait loop.

Introduce define LINK_RETRAIN_TIMEOUT instead of hard-coding timeout in
pcie_aspm_configure_common_clock.

Use time_after() to avoid jiffy wraparound when checking for expired
timeout.

After timeout expires, recheck link status register link training bit
instead of checking for expired timeout to avoid possible false
positive.

Note that Matthew Wilcox came up with the first rough version of this
patch.

Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: PCIe portdrv: Add kerneldoc comments to remining core funtions
Rafael J. Wysocki [Thu, 1 Jan 2009 18:53:32 +0000 (19:53 +0100)]
PCI: PCIe portdrv: Add kerneldoc comments to remining core funtions

Add kerneldoc comments to the reamining functions in
drivers/pci/pcie/portdrv_core.c .

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: PCIe portdrv: Rearrange code so that related things are together
Rafael J. Wysocki [Thu, 1 Jan 2009 18:52:12 +0000 (19:52 +0100)]
PCI: PCIe portdrv: Rearrange code so that related things are together

Rearrange code in drivers/pci/pcie/portdrv_bus.c and
drivers/pci/pcie/portdrv_core.c so that related functions and data
structures are closer together.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: PCIe portdrv: Fix suspend and resume of PCI Express port services
Rafael J. Wysocki [Thu, 1 Jan 2009 18:51:15 +0000 (19:51 +0100)]
PCI: PCIe portdrv: Fix suspend and resume of PCI Express port services

There is a problem with the suspend and resume of PCI Express port
service devices that the ->suspend() and ->resume() routines of each
service device are called twice in each suspend-resume cycle, which
is obviously wrong.

The scenario is that first, the PCI Express port driver calls
suspend and resume routines of each port service driver from its
pcie_portdrv_suspend() and pcie_portdrv_resume() callbacks,
respectively (which is correct), and second, the pcie_port_bus_type
driver calls them from its ->suspend() and ->resume() callbacks
(which is not correct, because it doesn't happen at the right time).

The solution is to remove the ->suspend() and ->resume() callbacks
from pcie_port_bus_type and the associated functions.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: PCIe portdrv: Add kerneldoc comments to some core functions
Rafael J. Wysocki [Thu, 1 Jan 2009 18:48:55 +0000 (19:48 +0100)]
PCI: PCIe portdrv: Add kerneldoc comments to some core functions

Add kerneldoc comments to some functions in
drivers/pci/pcie/portdrv_core.c, since the code in there is not
easy to follow without any additional description.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agox86/PCI: Do not use interrupt links for devices using MSI-X
Rafael J. Wysocki [Mon, 5 Jan 2009 13:50:27 +0000 (14:50 +0100)]
x86/PCI: Do not use interrupt links for devices using MSI-X

pcibios_enable_device() and pcibios_disable_device() don't handle
IRQs for devices that have MSI enabled and it should treat the
devices with MSI-X enabled in the same way.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agonet: sfc: Use pci_clear_master() to disable bus mastering
Ben Hutchings [Tue, 23 Dec 2008 03:09:53 +0000 (03:09 +0000)]
net: sfc: Use pci_clear_master() to disable bus mastering

pci_disable_device() disables many features, like MSI-X, which we
never reenable in efx_reset().  Further, calls to pci_enable_device()
and pci_disable_device() must be matched since the nesting count was
introduced, so switch to using pci_clear_master() instead.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Add pci_clear_master() as opposite of pci_set_master()
Ben Hutchings [Tue, 23 Dec 2008 03:08:29 +0000 (03:08 +0000)]
PCI: Add pci_clear_master() as opposite of pci_set_master()

During an online device reset it may be useful to disable bus-mastering.
pci_disable_device() does that, and far more besides, so is not suitable
for an online reset.

Add pci_clear_master() which does just this.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: remove redundant test in cpq hotplug
Julia Lawall [Sun, 21 Dec 2008 15:39:37 +0000 (16:39 +0100)]
PCI hotplug: remove redundant test in cpq hotplug

func is checked not to be NULL a few lines before.

A simplified version of the semantic patch that makes this change is as
follows: (http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@r exists@
local idexpression x;
expression E;
position p1,p2;
@@

if (x@p1 == NULL || ...) { ... when forall
   return ...; }
... when != \(x=E\|x--\|x++\|--x\|++x\|x-=E\|x+=E\|x|=E\|x&=E\|&x\)
(
x@p2 == NULL
|
x@p2 != NULL
)

// another path to the test that is not through p1?
@s exists@
local idexpression r.x;
position r.p1,r.p2;
@@

... when != x@p1
(
x@p2 == NULL
|
x@p2 != NULL
)

@fix depends on !s@
position r.p1,r.p2;
expression x,E;
statement S1,S2;
@@

(
- if ((x@p2 != NULL) || ...)
  S1
|
- if ((x@p2 == NULL) && ...) S1
|
- BUG_ON(x@p2 == NULL);
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pciehp: cleanup register and field definitions
Kenji Kaneshige [Fri, 19 Dec 2008 06:19:02 +0000 (15:19 +0900)]
PCI: pciehp: cleanup register and field definitions

Clean up register definitions related to PCI Express Hot plug.

  - Add register definitions into include/linux/pci_regs.h, and use
    them instead of pciehp's locally definied register definitions.
  - Remove pciehp's locally defined register definitions
  - Remove unused register definitions in pciehp.
  - Some minor cleanups.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pciehp: ignore undefined bit in link status register
Kenji Kaneshige [Fri, 19 Dec 2008 06:18:10 +0000 (15:18 +0900)]
PCI: pciehp: ignore undefined bit in link status register

Bit 10 in Link Status register used to be defined as Training Error in
the PCI Express 1.0a specification. But it was removed by Training Error
ECN and is no longer defined. So pciehp must ignore the value read from
it.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: keep ASPM link state consistent throughout PCIe hierarchy
Shaohua Li [Fri, 19 Dec 2008 01:27:42 +0000 (09:27 +0800)]
PCI: keep ASPM link state consistent throughout PCIe hierarchy

In a PCIe hierarchy with a switch present, if the link state of an
endpoint device is changed, we must check the whole hierarchy from the
endpoint device to root port, and for each link in the hierarchy, the new
link state should be configured. Previously, the implementation checked
the state but forgot to configure the links between root port to switch.
Fixes Novell bz #448987.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agox86/PCI: use dev_printk for PCI bus locality messages
Bjorn Helgaas [Thu, 18 Dec 2008 23:34:51 +0000 (16:34 -0700)]
x86/PCI: use dev_printk for PCI bus locality messages

Since pci_bus has a struct device, use dev_printk directly instead
of faking it by hand.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: use dev_printk for PCI bus resource mssages
Bjorn Helgaas [Thu, 18 Dec 2008 23:34:19 +0000 (16:34 -0700)]
PCI: use dev_printk for PCI bus resource mssages

Since pci_bus has a struct device, use dev_printk directly instead
of faking it by hand.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: add interface to set visible size of VPD
Stephen Hemminger [Thu, 18 Dec 2008 17:17:16 +0000 (09:17 -0800)]
PCI: add interface to set visible size of VPD

The VPD on all devices may not be 32K. Unfortunately, there is no
generic way to find the size, so this adds a simple API hook
to reset it.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: revise VPD access interface
Stephen Hemminger [Thu, 18 Dec 2008 17:17:16 +0000 (09:17 -0800)]
PCI: revise VPD access interface

Change PCI VPD API which was only used by sysfs to something usable
in drivers.
   * move iteration over multiple words to the low level
   * use conventional types for arguments
   * add exportable wrapper

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: handle long delays in VPD access
Stephen Hemminger [Thu, 18 Dec 2008 17:17:16 +0000 (09:17 -0800)]
PCI: handle long delays in VPD access

Accessing the VPD area can take a long time.  The existing
VPD access code fails consistently on my hardware. There are comments
in the SysKonnect vendor driver that it can take up to 13ms per word.

Change the access routines to:
  * use a mutex rather than spinning with IRQ's disabled and lock held
  * have a much longer timeout
  * call cond_resched while spinning

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: x86/visws: use generic INTx swizzle from PCI core
Bjorn Helgaas [Wed, 17 Dec 2008 04:37:20 +0000 (21:37 -0700)]
PCI: x86/visws: use generic INTx swizzle from PCI core

Use the generic pci_common_swizzle() instead of arch-specific code.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: sh: use generic INTx swizzle from PCI core
Bjorn Helgaas [Wed, 17 Dec 2008 04:37:15 +0000 (21:37 -0700)]
PCI: sh: use generic INTx swizzle from PCI core

Use the generic pci_common_swizzle() instead of arch-specific code.

Note that pci_common_swizzle() loops based on dev->bus->self, not
dev->bus->parent as the sh simple_swizzle() did.  I think they
are equivalent for this purpose.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: mips: use generic INTx swizzle from PCI core
Bjorn Helgaas [Wed, 17 Dec 2008 04:37:10 +0000 (21:37 -0700)]
PCI: mips: use generic INTx swizzle from PCI core

Use the generic pci_common_swizzle() instead of arch-specific code.

Note that pci_common_swizzle() loops based on dev->bus->self, not
dev->bus->parent as the mips common_swizzle() did.  I think they
are equivalent for this purpose.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: arm: use generic INTx swizzle from PCI core
Bjorn Helgaas [Wed, 17 Dec 2008 04:37:05 +0000 (21:37 -0700)]
PCI: arm: use generic INTx swizzle from PCI core

Use the generic pci_common_swizzle() instead of arch-specific code.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: alpha: use generic INTx swizzle from PCI core
Bjorn Helgaas [Wed, 17 Dec 2008 04:37:00 +0000 (21:37 -0700)]
PCI: alpha: use generic INTx swizzle from PCI core

Use the generic pci_common_swizzle() instead of arch-specific code.

Note that pci_common_swizzle() loops based on dev->bus->self, not
dev->bus->parent as the alpha common_swizzle() did.  I think they
are equivalent for this purpose.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: add pci_common_swizzle() for INTx swizzling
Bjorn Helgaas [Wed, 17 Dec 2008 04:36:55 +0000 (21:36 -0700)]
PCI: add pci_common_swizzle() for INTx swizzling

This patch adds pci_common_swizzle(), which swizzles INTx values all the
way up to a root bridge.

This common implementation can replace several architecture-specific
ones.  This should someday be combined with pci_get_interrupt_pin(),
but I left it separate for now to make reviewing easier.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: introduce functions for ACPI slot detection
Kenji Kaneshige [Wed, 17 Dec 2008 03:09:12 +0000 (12:09 +0900)]
PCI hotplug: introduce functions for ACPI slot detection

Some ACPI related PCI hotplug code can be shared among PCI hotplug
drivers. This patch introduces the following functions in
drivers/pci/hotplug/acpi_pcihp.c to share the code, and changes
acpiphp and pciehp to use them.

- int acpi_pci_detect_ejectable(struct pci_bus *pbus)
  This checks if the specified PCI bus has ejectable slots.

- int acpi_pci_check_ejectable(struct pci_bus *pbus, acpi_handle handle)
  This checks if the specified handle is ejectable ACPI PCI slot. The
  'pbus' parameter is needed to check if 'handle' is PCI related ACPI
  object.

This patch also introduces the following inline function in
include/linux/pci-acpi.h, which is useful to get ACPI handle of the
PCI bridge from struct pci_bus of the bridge's secondary bus.

- static inline acpi_handle acpi_pci_get_bridge_handle(struct pci_bus *pbus)
  This returns ACPI handle of the PCI bridge which generates PCI bus
  specified by 'pbus'.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pciehp: add auto option to pciehp_detect_mode
Kenji Kaneshige [Wed, 17 Dec 2008 03:08:15 +0000 (12:08 +0900)]
PCI: pciehp: add auto option to pciehp_detect_mode

ACPI based hot-pluggable PCIe slot detection logic was added to
prevent the problem non hot-pluggable PCIe slot was detected as
hot-pluggable. The slot detection logic can be selected through
'pciehp_detect_mode', but it would be better if it is selected
automatically.

This patch adds 'auto' option for 'pciehp_detect_mode'. When it is
specified, pciehp judges which 'acpi' or 'pcie' should be used. It
seems that the physical slot number is duplicated among some slots on
most of the platforms with the above-mentioned problem. So 'auto' mode
uses this information to judge which 'acpi' or 'pcie' should be
used. That is, if duplicated physical slot numbers are detected,
'acpi' mode is used. This method is not perfect, but it's realistic.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pciehp: add ACPI based slot detection
Kenji Kaneshige [Wed, 17 Dec 2008 03:07:38 +0000 (12:07 +0900)]
PCI: pciehp: add ACPI based slot detection

There is a problem that some non hot-pluggable PCIe slots are detected
as hot-pluggable by pciehp on some platforms. The immediate cause of
this problem is that hot-plug capable bit in the Slot Capabilities
register is set even for non hot-pluggable slots on those platforms.
It seems a BIOS/hardware problem, but we need workaround about that.

Some of those platforms define hot-pluggable PCIe slots on ACPI
namespace properly, while hot-plug capable bit in the Slot
Capabilities register is set improperly. So using ACPI namespace
information in pciehp to detect PCIe hot-pluggable slots would be a
workaround.

This patch adds 'pciehp_detect_mode' module option. When 'acpi' is
specified, pciehp uses ACPI namespace information to detect PCIe
hot-pluggable slots.

Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: work_on_cpu: use in drivers/pci/pci-driver.c
Rusty Russell [Wed, 31 Dec 2008 13:24:56 +0000 (23:54 +1030)]
PCI: work_on_cpu: use in drivers/pci/pci-driver.c

This uses work_on_cpu(), rather than altering the cpumask of the
thread which we happen to be.

Note the cleanups:

1) I've removed the CONFIG_NUMA test, since dev_to_node() returns -1
   for !CONFIG_NUMA anyway and the compiler will eliminate it.

2) No need to reset mempolicy to default (a bad idea anyway) since
   work_on_cpu is run from a workqueue.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Split PCI Express port suspend-resume
Rafael J. Wysocki [Sat, 27 Dec 2008 15:28:58 +0000 (16:28 +0100)]
PCI PM: Split PCI Express port suspend-resume

Suspend-resume of PCI Express ports has recently been moved into
_suspend_late() and _resume_early() callbacks, but some functions
executed from there should not be called with interrupts disabled,
eg. pci_enable_device().  For this reason, split the suspend-resume
of PCI Express ports into parts to be executed with interrupts
disabled and with interrupts enabled.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI PM: Fix pci_update_current_state
Rafael J. Wysocki [Sat, 27 Dec 2008 15:30:52 +0000 (16:30 +0100)]
PCI PM: Fix pci_update_current_state

Currently, PCI devices without the PM capability that are power
manageable by the platform (eg. ACPI) are not handled correctly
by pci_set_power_state(), because their current_state field is not
updated to reflect the new power state of the device.  Fix this by
making pci_update_current_state() accept additional argument
representing the power state of the device as set by the platform.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: set device wakeup capable flag if platform support is present
Jesse Barnes [Wed, 17 Dec 2008 20:10:05 +0000 (12:10 -0800)]
PCI: set device wakeup capable flag if platform support is present

When PCI devices are initialized, we check whether they support PCI PM
caps and set the device can_wakeup flag if so.  However, some devices
may have platform provided wakeup events rather than PCI PME signals, so
we need to set can_wakeup in that case too.  Doing so should allow
wakeups from many more devices, especially on cost constrained systems.

Reported-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Joseph Chan <JosephChan@via.com.tw>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: factor pci_bus_add_child() from pci_bus_add_devices()
Yu Zhao [Fri, 21 Nov 2008 18:42:35 +0000 (02:42 +0800)]
PCI: factor pci_bus_add_child() from pci_bus_add_devices()

This patch splits a new function, pci_bus_add_child(), from
pci_bus_add_devices(). The new function can be used to register PCI
buses to the device core.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: cleanup pci_bus_add_devices()
Yu Zhao [Fri, 21 Nov 2008 18:41:45 +0000 (02:41 +0800)]
PCI: cleanup pci_bus_add_devices()

Cleanup pci_bus_add_devices() by negating the conditional and
continuing, rather than having a single conditional take up the whole
body.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: add a new function to map BAR offsets
Yu Zhao [Fri, 21 Nov 2008 18:41:27 +0000 (02:41 +0800)]
PCI: add a new function to map BAR offsets

Add a function to map a given resource number to a corresponding
register so drivers can get the offset and type of device specific BARs.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: allow pci_alloc_child_bus() to handle a NULL bridge
Yu Zhao [Fri, 21 Nov 2008 18:41:07 +0000 (02:41 +0800)]
PCI: allow pci_alloc_child_bus() to handle a NULL bridge

Allow pci_alloc_child_bus() to allocate buses without bridge devices.
Some SR-IOV devices can occupy more than one bus number, but there is no
explicit bridges because that have internal routing mechanism.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: export __pci_read_base()
Yu Zhao [Fri, 21 Nov 2008 18:40:40 +0000 (02:40 +0800)]
PCI: export __pci_read_base()

Export __pci_read_base() so it can be used by whole PCI subsystem.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: remove unnecessary condition check in pci_restore_bars()
Yu Zhao [Fri, 21 Nov 2008 18:40:00 +0000 (02:40 +0800)]
PCI: remove unnecessary condition check in pci_restore_bars()

Remove the unnecessary number of resources condition checks because
the pci_update_resource() will check availability of the resources.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: define PCI resource names in an 'enum'
Yu Zhao [Fri, 21 Nov 2008 18:39:32 +0000 (02:39 +0800)]
PCI: define PCI resource names in an 'enum'

This patch moves all definitions of the PCI resource names to an 'enum',
and also replaces some hard-coded resource variables with symbol
names. This change eases introduction of device specific resources.

Reviewed-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: remove unnecessary arg of pci_update_resource()
Yu Zhao [Fri, 21 Nov 2008 18:38:52 +0000 (02:38 +0800)]
PCI: remove unnecessary arg of pci_update_resource()

This cleanup removes unnecessary argument 'struct resource *res' in
pci_update_resource(), so it takes same arguments as other companion
functions (pci_assign_resource(), etc.).

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: enhance pci_ari_enabled()
Yu Zhao [Fri, 21 Nov 2008 18:38:21 +0000 (02:38 +0800)]
PCI: enhance pci_ari_enabled()

Change parameter of pci_ari_enabled() from 'pci_dev' to 'pci_bus'.

ARI forwarding on the bridge mostly concerns the subordinate devices
rather than the bridge itself. So this change will make the function
easier to use.

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI quirks: piix3: warn softer about enabling passive release
Adam Jackson [Mon, 1 Dec 2008 22:30:29 +0000 (14:30 -0800)]
PCI quirks: piix3: warn softer about enabling passive release

All the other quirks are dev_info() not dev_err(), this one isn't special.
This makes 'quiet' boot in qemu really quiet.

Signed-off-by: Adam Jackson <ajax@redhat.com>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Tejun Heo <htejun@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Make settable sysfs attributes more consistent
Trent Piepho [Mon, 1 Dec 2008 01:10:12 +0000 (17:10 -0800)]
PCI: Make settable sysfs attributes more consistent

PCI devices have three settable boolean attributes, enable,
broken_parity_status, and msi_bus.

The store functions for these would silently interpret "0x01" as false,
"1llogical" as true, and "true" would be (silently!) ignored and do
nothing.

This is inconsistent with typical sysfs handling of settable attributes,
and just plain doesn't make much sense.

So, use strict_strtoul(), which was created for this purpose.  The store
functions will treat a value of 0 as false, non-zero as true, and return
-EINVAL for a parse failure.

Additionally, is_enabled_store() and msi_bus_store() return -EPERM if
CAP_SYS_ADMIN is lacking, rather than silently doing nothing.  This is more
typical behavior for sysfs attributes that need a capability.

And msi_bus_store() will only print the "forced subordinate bus ..."
warning if the MSI flag was actually forced to a different value.

Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: uninline pci_ioremap_bar()
Andrew Morton [Mon, 1 Dec 2008 22:30:30 +0000 (14:30 -0800)]
PCI: uninline pci_ioremap_bar()

It's too large to be inlined.

Acked-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: fix incorrect error return in pci_enable_wake
Alan Stern [Tue, 16 Dec 2008 19:06:58 +0000 (14:06 -0500)]
PCI: fix incorrect error return in pci_enable_wake

This patch (as1186) fixes a minor mistake in pci_enable_wake().  When
the routine is asked to disable remote wakeup, it should not return an
error merely because the device is not allowed to do wakeups!

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: parisc: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:22 +0000 (16:12 -0700)]
PCI: parisc: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Acked-by: Kyle McMartin <kyle@infradead.org>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: mips: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:17 +0000 (16:12 -0700)]
PCI: mips: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: x86: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:37 +0000 (16:12 -0700)]
PCI: x86: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: hpa@zytor.com
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: alpha: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:07 +0000 (16:12 -0700)]
PCI: alpha: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: powerpc: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:27 +0000 (16:12 -0700)]
PCI: powerpc: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: sh: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:32 +0000 (16:12 -0700)]
PCI: sh: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Acked-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: arm: use generic pci_swizzle_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:12:12 +0000 (16:12 -0700)]
PCI: arm: use generic pci_swizzle_interrupt_pin()

Use the generic pci_swizzle_interrupt_pin() instead of arch-specific code.

Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: add pci_swizzle_interrupt_pin()
Bjorn Helgaas [Thu, 11 Dec 2008 18:24:23 +0000 (11:24 -0700)]
PCI: add pci_swizzle_interrupt_pin()

This patch adds pci_swizzle_interrupt_pin(), which implements the
INTx swizzling algorithm specified in Table 9-1 of the "PCI-to-PCI
Bridge Architecture Specification," revision 1.2.

There are many architecture-specific implementations of this
swizzle that can be replaced by this common one.

Reviewed-by: David Howells <dhowells@redhat.com>
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agox86/PCI: minor logic simplications
Bjorn Helgaas [Tue, 9 Dec 2008 23:11:57 +0000 (16:11 -0700)]
x86/PCI: minor logic simplications

Test "pin" immediately to simplify the subsequent code.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: hpa@zytor.com
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agox86/PCI: use config space encoding for interrupt pins
Bjorn Helgaas [Tue, 9 Dec 2008 23:11:51 +0000 (16:11 -0700)]
x86/PCI: use config space encoding for interrupt pins

Keep "pin" encoded as it is in the "Interrupt Pin" value in PCI config
space, i.e., 0=device doesn't use interrupts, 1=INTA, ..., 4=INTD.

This makes the bridge INTx swizzle match other architectures.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: hpa@zytor.com
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: use config space encoding in pci_get_interrupt_pin()
Bjorn Helgaas [Tue, 9 Dec 2008 23:11:46 +0000 (16:11 -0700)]
PCI: use config space encoding in pci_get_interrupt_pin()

This patch makes pci_get_interrupt_pin() return values encoded
the same way as the "Interrupt Pin" value in PCI config space,
i.e., 1=INTA, ..., 4=INTD.

pirq_bios_set() is the only in-tree caller of pci_get_interrupt_pin()
and pci_get_interrupt_pin() is not exported.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: hpa@zytor.com
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: cpqphp: use config space PCI interrupt pin encoding
Bjorn Helgaas [Tue, 9 Dec 2008 23:11:41 +0000 (16:11 -0700)]
PCI hotplug: cpqphp: use config space PCI interrupt pin encoding

This patch changes cpqphp to use interrupt pin values just as they
come from PCI config space, i.e., 1=INTA, ..., 4=INTD.

pcibios_set_irq_routing() takes pin arguments in the range 0=INTA, ...,
3=INTD, so we'll adjust the pin just before calling it.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Acked-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: avoid early PCI mmconfig init if pci=noearly is given in cmdline
Jacob Pan [Mon, 8 Dec 2008 17:44:16 +0000 (09:44 -0800)]
PCI: avoid early PCI mmconfig init if pci=noearly is given in cmdline

Early type 1 accesses can cause problems on some platforms, and
pci=noearly is supposed to prevent them from occurring.  However, early
mcfg probing code uses type 1 and  isn't protected by a check for
noearly.  This patch fixes that problem.

Signed-off-by: Jacob Pan <jacob.jun.pan@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pcie port driver: remove extra printks
Bjorn Helgaas [Wed, 10 Dec 2008 20:00:21 +0000 (13:00 -0700)]
PCI: pcie port driver: remove extra printks

These printks don't contain enough information to be useful.  I think it
would be more useful to have a message when a service driver binds to a
root port.  That could contain the service type, the interrupt mode and
IRQ, etc.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agox86/PCI: make PCI bus locality messages more meaningful
Bjorn Helgaas [Wed, 10 Dec 2008 20:12:00 +0000 (13:12 -0700)]
x86/PCI: make PCI bus locality messages more meaningful

Change PCI bus locality messages so they have a bit more context
and look like the rest of PCI, e.g.,

    - bus 01 -> node 0
    - bus 04 -> node 0
    + pci 0000:01: bus on NUMA node 0
    + pci 0000:04: bus on NUMA node 0

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: make PCI bus resource messages more meaningful
Bjorn Helgaas [Wed, 10 Dec 2008 20:02:18 +0000 (13:02 -0700)]
PCI: make PCI bus resource messages more meaningful

Change PCI bus resource messages so they have a bit more context
and look like the rest of PCI, e.g.,

    - bus: 00 index 0 io port: [0x00-0xffff]
    - bus: 00 index 1 mmio: [0x000000-0xffffffff]
    + pci 0000:00: bus resource 0 io : [0x00-0xffff]
    + pci 0000:00: bus resource 1 mem: [0x000000-0xffffffff]

This also changes them from KERN_INFO to KERN_DEBUG.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: acpiphp whitespace cleanup
Alex Chiang [Thu, 11 Dec 2008 18:17:55 +0000 (11:17 -0700)]
PCI hotplug: acpiphp whitespace cleanup

Clean up whitespace.

Setting 'let c_space_errors=1' in .vimrc shows all sorts of
ugliness. ;)

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Don't carp about BAR allocation failures in quiet boot
Ingo Molnar [Fri, 12 Dec 2008 05:49:40 +0000 (06:49 +0100)]
PCI: Don't carp about BAR allocation failures in quiet boot

These are easy to trigger (more or less harmlessly) with multiple video
cards, since the ROM BAR will typically not be given any space by the
BIOS bridge setup.  No reason to punish quiet boot for this.

Signed-off-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: aerdrv: fix a typo in error message
Hidetoshi Seto [Fri, 5 Dec 2008 05:05:23 +0000 (14:05 +0900)]
PCI hotplug: aerdrv: fix a typo in error message

"TLP" is an acronym for "Transaction Layer Packet."

Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: handle PCI state saving with interrupts disabled
Rafael J. Wysocki [Sun, 7 Dec 2008 21:02:58 +0000 (22:02 +0100)]
PCI: handle PCI state saving with interrupts disabled

Since interrupts will soon be disabled at PCI resume time, we need to
pre-allocate memory to save/restore PCI config space (or use GFP_ATOMIC,
but this is safer).

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: extend on the ICH motherboard IO decode quirk list
Linus Torvalds [Sat, 6 Dec 2008 18:10:10 +0000 (10:10 -0800)]
PCI: extend on the ICH motherboard IO decode quirk list

This adds more LPC controller IO range decode quirks for the Intel ICH
family of chipsets. They differ a bit between the older ICH6 chipset and
the more modern layout of the ICH7-ICH10 chipsets.

This patch just prints out the IO decode information found by the quirks,
but eventually we may want to add them to the resource tree, in order to
know to avoid allocating things over them.

That's especially true if it turns out that any firmware ends up putting
the magic motherboard resources in an address range that we use for
dynamic allocations (ie above PCIBIOS_MIN_IO, which is 0x1000 on x86).

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Greg KH <greg@kroah.com>
Cc: Frans Pop <elendil@planet.nl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Suspend and resume PCI Express ports with interrupts disabled
Rafael J. Wysocki [Sat, 6 Dec 2008 14:07:59 +0000 (15:07 +0100)]
PCI: Suspend and resume PCI Express ports with interrupts disabled

I don't see why the suspend and resume of PCI Express ports should be
handled with interrupts enabled and it may even lead to problems in
some situations.  For this reason, move the suspending and resuming
of PCI Express ports into ->suspend_late() and ->resume_early()
callbacks executed with interrupts disabled.

This patch addresses the regression from 2.6.26 tracked as
http://bugzilla.kernel.org/show_bug.cgi?id=12121 .

Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: pci-stub module to reserve pci device
Chris Wright [Wed, 26 Nov 2008 05:17:13 +0000 (21:17 -0800)]
PCI: pci-stub module to reserve pci device

When doing device assignment with KVM there's currently nothing to
protect the device from having a driver in the host as well as the guest.
This trivial module just binds the pci device on the host to a stub
driver so that a real host driver can't bind to the device.  It has no
pci id table, it supports only dynamic ids.

 # echo "8086 10f5" > /sys/bus/pci/drivers/pci-stub/new_id
 # echo -n 0000:00:19.0 > /sys/bus/pci/drivers/e1000e/unbind
 # echo -n 0000:00:19.0 > /sys/bus/pci/drivers/pci-stub/bind
 # ls -l /sys/bus/pci/devices/0000:00:19.0/driver
 lrwxrwxrwx 1 root root 0 2008-11-25 19:10 /sys/bus/pci/devices/0000:00:19.0/driver -> ../../../bus/pci/drivers/pci-stub

Cc: "Kay, Allen M" <allen.m.kay@intel.com>
Cc: "Nakajima, Jun" <jun.nakajima@intel.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: allow pci driver to support only dynids
Chris Wright [Wed, 26 Nov 2008 03:36:10 +0000 (19:36 -0800)]
PCI: allow pci driver to support only dynids

commit b41d6cf38e27 (PCI: Check dynids driver_data value for validity)
requires all drivers to include an id table to try and match
driver_data.  Before validating driver_data check driver has an id
table.

Acked-by: Jean Delvare <khali@linux-fr.org>
Cc: Milton Miller <miltonm@bga.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI/ACPI: acpiphp: Identify more removable slots
Matthew Garrett [Tue, 25 Nov 2008 21:48:14 +0000 (21:48 +0000)]
PCI/ACPI:  acpiphp: Identify more removable slots

According to section 6.3.6 of the ACPI spec, the presence of an _RMV
method that evaluates to 1 is sufficient to indicate that a slot is
removable without needing an eject method. This patch refactors the
ejectable slot detection code a little in order to flag these slots as
ejectable and register them. Acpihp then binds to the expresscard slot
on my HP test machine.

Acked-by: Kristen Carlson Accardi <kristen.c.accardi@intel.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agopci-acpi: Cleanup _OSC evaluation code
Taku Izumi [Thu, 20 Nov 2008 06:22:39 +0000 (15:22 +0900)]
pci-acpi: Cleanup _OSC evaluation code

Cleanup _OSC evaluation code.  Some whitespace changes and a few other
minor cleanups.

Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI/ACPI: Remove unnecessary _OSC evaluation for control request
Taku Izumi [Thu, 20 Nov 2008 06:22:37 +0000 (15:22 +0900)]
PCI/ACPI: Remove unnecessary _OSC evaluation for control request

If a control had already been granted, we don't need to re-evaluate
_OSC for it because firmware may not reject control of any feature it
has previously granted control to.

Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: revert additional _OSC evaluation
Taku Izumi [Thu, 20 Nov 2008 06:22:32 +0000 (15:22 +0900)]
PCI: revert additional _OSC evaluation

Reverts adf411b819adc9fa96e9b3e638c7480d5e71d270.

The commit adf411b819adc9fa96e9b3e638c7480d5e71d270 was based on the
improper assumption that queried result was not updated when _OSC
support field was changed. But, in fact, queried result is updated
whenever _OSC support field was changed through __acpi_query_osc().
As a result, the commit adf411b819adc9fa96e9b3e638c7480d5e71d270 only
introduced unnecessary additional _OSC evaluation...

Tested-by: Andrew Patterson <andrew.patterson@hp.com>
Reviewed-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Taku Izumi <izumi.taku@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agodrivers/pci/hotplug: Add missing pci_dev_get
Julia Lawall [Sun, 23 Nov 2008 08:08:28 +0000 (09:08 +0100)]
drivers/pci/hotplug: Add missing pci_dev_get

pci_get_slot does a pci_dev_get, so pci_dev_put needs to be called in an
error case.

An alterative would be to move the test_and_set_bit before the call to
pci_get_slot.

The problem was fixed using the following semantic patch.
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
local idexpression *n;
statement S1,S2;
expression E,E1;
expression *ptr != NULL;
type T,T1;
@@

(
if (!(n = pci_get_slot(...))) S1
|
n = pci_get_slot(...)
)
<... when != pci_dev_put(n)
    when != if (...) { <+... pci_dev_put(n) ...+> }
    when != true !n  || ...
    when != n = (T)E
    when != E = n
if (!n || ...) S2
...>
(
  return \(0\|<+...n...+>\|ptr\);
|
+ pci_dev_put(n);
return ...;
|
pci_dev_put(n);
|
n = (T1)E1
|
E1 = n
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoresource: allow MMIO exclusivity for device drivers
Arjan van de Ven [Thu, 23 Oct 2008 02:55:31 +0000 (19:55 -0700)]
resource: allow MMIO exclusivity for device drivers

Device drivers that use pci_request_regions() (and similar APIs) have a
reasonable expectation that they are the only ones accessing their device.
As part of the e1000e hunt, we were afraid that some userland (X or some
bootsplash stuff) was mapping the MMIO region that the driver thought it
had exclusively via /dev/mem or via various sysfs resource mappings.

This patch adds the option for device drivers to cause their reserved
regions to the "banned from /dev/mem use" list, so now both kernel memory
and device-exclusive MMIO regions are banned.
NOTE: This is only active when CONFIG_STRICT_DEVMEM is set.

In addition to the config option, a kernel parameter iomem=relaxed is
provided for the cases where developers want to diagnose, in the field,
drivers issues from userspace.

Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: remove obsolete _OSC capability support functions
Andrew Patterson [Mon, 10 Nov 2008 22:31:10 +0000 (15:31 -0700)]
ACPI/PCI: remove obsolete _OSC capability support functions

The acpi_query_osc, __pci_osc_support_set, pci_osc_support_set, and
pcie_osc_support_set functions have been obsoleted in favor of setting
these capabilities during root bridge discovery with
pci_acpi_osc_support.  There are no longer any callers of these
functions, so remove them.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: PCI MSI _OSC support capabilities called when root bridge added
Andrew Patterson [Mon, 10 Nov 2008 22:31:05 +0000 (15:31 -0700)]
ACPI/PCI: PCI MSI _OSC support capabilities called when root bridge added

The _OSC capability OSC_MSI_SUPPORT is set when the root bridge is added
with pci_acpi_osc_support(), so we no longer need to do it in the PCI
MSI driver.  Also adds the function pci_msi_enabled, which returns true
if pci=nomsi is not on the kernel command-line.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: PCIe AER _OSC support capabilities called when root bridge added
Andrew Patterson [Mon, 10 Nov 2008 22:31:00 +0000 (15:31 -0700)]
ACPI/PCI: PCIe AER _OSC support capabilities called when root bridge added

The _OSC capability OSC_EXT_PCI_CONFIG_SUPPORT is set when the root
bridge is added with pci_acpi_osc_support(), so we no longer need to do
it in the PCIe AER driver.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: PCIe ASPM _OSC support capabilities called when root bridge added
Andrew Patterson [Mon, 10 Nov 2008 22:30:55 +0000 (15:30 -0700)]
ACPI/PCI: PCIe ASPM _OSC support capabilities called when root bridge added

The _OSC capabilities OSC_ACTIVE_STATE_PWR_SUPPORT and
OSC_CLOCK_PWR_CAPABILITY_SUPPORT are set when the root bridge is added
with pci_acpi_osc_support(), so we no longer need to do it in the ASPM
driver.  Also add the function pcie_aspm_enabled, which returns true if
pcie_aspm=off is not on the kernel command-line.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: PCI extended config _OSC support called when root bridge added
Andrew Patterson [Mon, 10 Nov 2008 22:30:50 +0000 (15:30 -0700)]
ACPI/PCI: PCI extended config _OSC support called when root bridge added

The _OSC capability OSC_EXT_PCI_CONFIG_SUPPORT is set when the root
bridge is added with pci_acpi_osc_support() if we can access PCI
extended config space.

This adds the function pci_ext_cfg_avail which returns true if we can
access PCI extended config space (offset greater than 0xff). It
currently only returns false if arch=x86 and raw_pci_ext_ops is not set
(which might happen if pci=nommcfg is set on the kernel command-line).

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: call _OSC support during root bridge discovery
Andrew Patterson [Mon, 10 Nov 2008 22:30:45 +0000 (15:30 -0700)]
ACPI/PCI: call _OSC support during root bridge discovery

Add pci_acpi_osc_support() and call it when a PCI bridge is added.  This
allows us to avoid having every individual PCI root bridge driver call
_OSC support for every root bridge in their probe functions, a
significant savings in boot time.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoACPI/PCI: include missing acpi.h file in pci-acpi.h.
Andrew Patterson [Mon, 10 Nov 2008 22:30:40 +0000 (15:30 -0700)]
ACPI/PCI: include missing acpi.h file in pci-acpi.h.

The pci-acpi.h file will not compile without including linux/acpi.h.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Add legacy_io/mem to all busses
Benjamin Herrenschmidt [Wed, 12 Nov 2008 03:38:53 +0000 (14:38 +1100)]
PCI: Add legacy_io/mem to all busses

Currently, only PHBs get the legacy_* files, which makes it tricky for
userland to get access to the legacy space.  This commit exposes them in
every bus, since even child buses may forward legacy cycles if
configured properly.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Extend pci_reset_function() to support PCI Advanced Features
Sheng Yang [Tue, 11 Nov 2008 09:17:48 +0000 (17:17 +0800)]
PCI: Extend pci_reset_function() to support PCI Advanced Features

Some PCI devices implement PCI Advanced Features, which means they
support Function Level Reset(FLR).  Implement support for that in
pci_reset_function.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: add PCI Advanced Feature Capability defines
Sheng Yang [Tue, 11 Nov 2008 09:17:46 +0000 (17:17 +0800)]
PCI: add PCI Advanced Feature Capability defines

PCI Advanced Features Capability is introduced by "Conventional PCI
Advanced Caps ECN" (can be downloaded in pcisig.com).  Add defines for
the various AF capabilities, including function level reset (FLR).

Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: Refactor pci_reset_function()
Sheng Yang [Tue, 11 Nov 2008 09:17:47 +0000 (17:17 +0800)]
PCI: Refactor pci_reset_function()

Separate out function level reset so that pci_reset_function can be more
easily extended.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: struct device - replace bus_id with dev_name(), dev_set_name()
Kay Sievers [Thu, 30 Oct 2008 01:17:49 +0000 (02:17 +0100)]
PCI: struct device - replace bus_id with dev_name(), dev_set_name()

This patch is part of a larger patch series which will remove
the "char bus_id[20]" name string from struct device. The device
name is managed in the kobject anyway, and without any size
limitation, and just needlessly copied into "struct device".

To set and read the device name dev_name(dev) and dev_set_name(dev)
must be used. If your code uses static kobjects, which it shouldn't
do, "const char *init_name" can be used to statically provide the
name the registered device should have. At registration time, the
init_name field is cleared, to enforce the use of dev_name(dev) to
access the device name at a later time.

We need to get rid of all occurrences of bus_id in the entire tree
to be able to enable the new interface. Please apply this patch,
and possibly convert any remaining remaining occurrences of bus_id.

Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-Off-By: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: fixup whitespace in quirks.c
Jesse Barnes [Tue, 28 Oct 2008 00:50:21 +0000 (17:50 -0700)]
PCI: fixup whitespace in quirks.c

Had a space before tab in do_fixups, prototype wasn't wrapped properly either.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI hotplug: pciehp: remove unnecessary wait after turning power off
Kenji Kaneshige [Fri, 24 Oct 2008 05:26:35 +0000 (14:26 +0900)]
PCI hotplug: pciehp: remove unnecessary wait after turning power off

The pciehp driver waits for 1000 msec after turning power off to make
sure the power has been completely removed. But this 1000 msec wait is
not needed if a slot doesn't implement power control because software
cannot control the power. Power will be automatically removed at adapter
removal time on such a slot

Tested-by: "Phil Endecott" <phil_pibbu_endecott@chezphil.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoPCI: check mmap range of /proc/bus/pci files too
Jesse Barnes [Fri, 24 Oct 2008 17:32:33 +0000 (10:32 -0700)]
PCI: check mmap range of /proc/bus/pci files too

/proc/bus/pci allows you to mmap resource ranges too, so we should probably be
checking to make sure the mapping is somewhat valid.  Uses the same code as the recent sysfs mmap range checking patch from Linus.

Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
15 years agoFix up 64-bit byte swaps for most 32-bit architectures
Linus Torvalds [Wed, 7 Jan 2009 05:17:57 +0000 (21:17 -0800)]
Fix up 64-bit byte swaps for most 32-bit architectures

The __SWAB_64_THRU_32__ case of a 64-bit byte swap was depending on the
no-longer-existant ___swab32() method (three underscores).  We got rid
of some of the worst indirection and complexity, and now it should just
use the 32-bit swab function that was defined right above it.

Reported-and-tested-by: Nicolas Pitre <nico@cam.org>
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agobyteorder: remove the now unused byteorder.h
Harvey Harrison [Tue, 6 Jan 2009 21:30:58 +0000 (13:30 -0800)]
byteorder: remove the now unused byteorder.h

This implementation caused problems in userspace which can, and does
define _both_ __LITTLE_ENDIAN and __BIG_ENDIAN.

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agomn10300: introduce asm/swab.h
Harvey Harrison [Tue, 6 Jan 2009 22:56:34 +0000 (14:56 -0800)]
mn10300: introduce asm/swab.h

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofrv: introduce asm/swab.h
Harvey Harrison [Tue, 6 Jan 2009 22:56:33 +0000 (14:56 -0800)]
frv: introduce asm/swab.h

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>