10 years agoPCI: PCIE AER: Document for PCIE AER software error injection
Huang Ying [Fri, 24 Apr 2009 02:45:31 +0000]
PCI: PCIE AER: Document for PCIE AER software error injection

This patch adds a minimal HOWTO for PCIE AER software error injection
in Documentation/PCI/pcieaer-howto.txt.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: PCIE AER: export aer_irq
Huang Ying [Fri, 24 Apr 2009 02:45:23 +0000]
PCI: PCIE AER: export aer_irq

This is used by PCIE AER error injection to fake an PCI AER interrupt.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: Add pci_bus_set_ops
Huang Ying [Fri, 24 Apr 2009 02:45:17 +0000]
PCI: Add pci_bus_set_ops

pci_bus_set_ops changes pci_ops associated with a pci_bus. This can be
used by debug tools such as PCIE AER error injection to fake some PCI
configuration registers.

Acked-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in pci_common_swizzle()
Kenji Kaneshige [Tue, 26 May 2009 07:08:36 +0000]
PCI: use pci_is_root_bus() in pci_common_swizzle()

Use pci_is_root_bus() in pci_common_swizzle() for checking if the pci
bus is root, for code consistency.

Reviewed-by: Alex Chiang <achiang@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in pci_get_interrupt_pin()
Kenji Kaneshige [Tue, 26 May 2009 07:07:33 +0000]
PCI: use pci_is_root_bus() in pci_get_interrupt_pin()

Use pci_is_root_bus() in pci_get_interrupt_pin() for checking if the
pci bus is root, for code consistency.

Reviewed-by: Alex Chiang <achiang@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in pci_read_bridge_bases()
Kenji Kaneshige [Tue, 26 May 2009 07:06:48 +0000]
PCI: use pci_is_root_bus() in pci_read_bridge_bases()

Use pci_is_root_bus() in pci_read_bridge_bases() to check if the pci
bus is root, for code consistency.

Reviewed-by: Alex Chiang <achiang@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in pci_find_upstream_pcie_bridge()
Kenji Kaneshige [Tue, 26 May 2009 07:06:10 +0000]
PCI: use pci_is_root_bus() in pci_find_upstream_pcie_bridge()

Use pci_is_root_bus() in pci_find_upstream_pcie_bridge() to check if
the pci bus is root, for code consistency.

Reviewed-by: Alex Chiang <achiang@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in acpi_find_root_bridge_handle()
Kenji Kaneshige [Tue, 26 May 2009 07:05:33 +0000]
PCI: use pci_is_root_bus() in acpi_find_root_bridge_handle()

Use pci_is_root_bus() in acpi_find_root_bridge_handle() to check if
the pci bus is root, for code consistency.

Reviewed-by: Alex Chiang <achiang@hp.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use pci_is_root_bus() in acpi_pci_get_bridge_handle()
Kenji Kaneshige [Tue, 26 May 2009 07:05:06 +0000]
PCI: use pci_is_root_bus() in acpi_pci_get_bridge_handle()

Use pci_is_root_bus() in acpi_pci_get_bridge_handle() to check if the
pci bus is root, for code consistency.

Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Reviewed-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI quirk: HP hides SMBus controller in Compaq nx9500 laptops
Michal Miroslaw [Tue, 12 May 2009 20:49:25 +0000]
PCI quirk: HP hides SMBus controller in Compaq nx9500 laptops

I found no references to SMBus in ACPI DSDT disassembly on my laptop
so this should be safe.

Signed-off-by: Michal Miroslaw <mirq-linux@rere.qmqm.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI quirk: unhide 'Overflow' device on i828{6,7}5P/PE chipsets
Michal Miroslaw [Tue, 12 May 2009 20:49:26 +0000]
PCI quirk: unhide 'Overflow' device on i828{6,7}5P/PE chipsets

Some BIOSes hide 'overflow' device (dev #6) for i82875P/PE chipsets.
The same happens for i82865P/PE. Add a quirk to enable this device.
This allows i82875 EDAC driver to bind to chipset's dev #6 and not
dev #0 as the latter is used by AGP driver.

On my laptop (i82865P based) ACPI code is disabling this device
again in \_SB.PCI0._CRS method (called at least at PNP init time).
This can be easily worked around by patching DSDT.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Michal Miroslaw <mirq-linux@rere.qmqm.pl>
Acked-by: Doug Thompson <norsk5@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: expose SMBus on Asus notebook A6L
Mats Erik Andersson [Tue, 12 May 2009 10:05:23 +0000]
PCI: expose SMBus on Asus notebook A6L

Addition of one unknown subsystem identifier to the quirks handler for
chipset i82855GM_HB on notebook Asus A6L. This exposes the otherwise
hidden SMBus controller within the south bridge ICH4-M.

Signed-off-by: Mats Erik Andersson <mats.andersson@gisladisker.se>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: Add support for turning PCIe ECRC on or off
Andrew Patterson [Wed, 22 Apr 2009 22:52:09 +0000]
PCI: Add support for turning PCIe ECRC on or off

Adds support for PCI Express transaction layer end-to-end CRC checking
(ECRC).  This patch will enable/disable ECRC checking by setting/clearing
the ECRC Check Enable and/or ECRC Generation Enable bits for devices that
support ECRC.

The ECRC setting is controlled by the "pci=ecrc=<policy>" command-line
option. If this option is not set or is set to 'bios", the enable and
generation bits are left in whatever state that firmware/BIOS set them to.
The "off" setting turns them off, and the "on" option turns them on (if the
device supports it).

Turning ECRC on or off can be a data integrity versus performance
tradeoff.  In theory, turning it on will catch more data errors, turning
it off means possibly better performance since CRC does not need to be
calculated by the PCIe hardware and packet sizes are reduced.

Signed-off-by: Andrew Patterson <andrew.patterson@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI PM: Follow PCI_PM_CTRL_NO_SOFT_RESET during transitions from D3
Rafael J. Wysocki [Mon, 18 May 2009 20:51:12 +0000]
PCI PM: Follow PCI_PM_CTRL_NO_SOFT_RESET during transitions from D3

According to the PCI PM specification (PCI Bus Power Management
Interface Specification, Rev. 1.2, Section 5.4.1) we are supposed to
reinitialize devices that have PCI_PM_CTRL_NO_SOFT_RESET clear during
all transitions from PCI_D3hot to PCI_D0, but we only do it if the
device's current_state field is equal to PCI_UNKNOWN.

This may lead to problems if a device with PCI_PM_CTRL_NO_SOFT_RESET
unset is put into PCI_D3hot at run time by its driver and
pci_set_power_state() is used to put it back into PCI_D0, because in
that case the device will remain uninitialized after
pci_set_power_state() has returned.  Prevent that from happening by
modifying pci_raw_set_power_state() to reinitialize devices with
PCI_PM_CTRL_NO_SOFT_RESET unset during all transitions from D3 to D0.

Cc: stable@kernel.org
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: fix SR-IOV function dependency link problem
Yu Zhao [Wed, 20 May 2009 09:11:57 +0000]
PCI: fix SR-IOV function dependency link problem

PCIe root complex integrated endpoint does not implement ARI, so this
kind of endpoint uses 3-bit function number. The function dependency
link of the integrated endpoint should be calculated using the device
number plus the value from function dependency link register.

Normal endpoint always implements ARI and the function dependency link
register contains 8-bit function number (i.e. `devfn' from software's
perspective).

Signed-off-by: Yu Zhao <yu.zhao@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: eliminate redundant pci_stop_dev() call from pci_destroy_dev()
Alex Chiang [Tue, 19 May 2009 01:02:38 +0000]
PCI: eliminate redundant pci_stop_dev() call from pci_destroy_dev()

We always call pci_stop_bus_device before calling pci_destroy_dev.

Since pci_stop_bus_device calls pci_stop_dev, there is no need
for pci_destroy_dev to repeat the call.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI MSI: let drivers retry when not enough vectors
Michael S. Tsirkin [Thu, 7 May 2009 08:28:41 +0000]
PCI MSI: let drivers retry when not enough vectors

pci_enable_msix currently returns -EINVAL if you ask
for more vectors than supported by the device, which would
typically cause fallback to regular interrupts.

It's better to return the table size, making the driver retry
MSI-X with less vectors.

Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: disable ASPM on VIA root-port-under-bridge configurations
Shaohua Li [Mon, 8 Jun 2009 01:27:25 +0000]
PCI: disable ASPM on VIA root-port-under-bridge configurations

VIA has a strange chipset, it has root port under a bridge. Disable ASPM
for such strange chipset.

Cc: stable@kernel.org
Tested-by: Wolfgang Denk <wd@denx.de>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: use ACPI _CRS data by default
Jesse Barnes [Thu, 11 Jun 2009 17:58:28 +0000]
PCI: use ACPI _CRS data by default

At this point, it seems to solve more problems than it causes, so let's try using it by default.  It's an easy revert if it ends up causing trouble.

Reviewed-by: Yinghai Lu <yhlu.kernel@gmail.com>
Acked-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: remove deprecated pci_find_slot() interface
Alex Chiang [Tue, 31 Mar 2009 15:24:22 +0000]
PCI: remove deprecated pci_find_slot() interface

The last in-tree caller of pci_find_slot has been converted, so
let's get rid of this deprecated interface.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: don't use pci_find_slot()
Alex Chiang [Tue, 31 Mar 2009 15:24:17 +0000]
PCI Hotplug: cpqphp: don't use pci_find_slot()

Convert uses of pci_find_slot to modern API.

In the conversion sites, we end up calling pci_dev_put() right away.

This may seem like it misses the entire point of doing something like
pci_get_bus_and_slot(), since we drop the reference so soon, but it turns
out we don't actually do much with the returned pci_dev.

I plan on untangling cpqphp further, but clearly cpqphp never worried too
much about a properly refcounted pci_dev anyway. For now, this conversion
seems reasonable, as it gets rid of the last in-tree caller of pci_find_slot.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: constify slot_name()
Alex Chiang [Tue, 31 Mar 2009 15:24:12 +0000]
PCI Hotplug: cpqphp: constify slot_name()

Eliminate this warning:
warning: return discards qualifiers from pointer target type

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: eliminate dead code - PCI_ScanBusNonBridge
Alex Chiang [Tue, 31 Mar 2009 15:24:07 +0000]
PCI Hotplug: cpqphp: eliminate dead code - PCI_ScanBusNonBridge

I have no clue what the original intent here was, but the code as
written is useless.

The old dbg() statement above the old callsite might lead one to think
that at one point, there was supposed to be some recursion, but any
sense of sanity here has been lost to the ravages of time.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: clean up accesses to pcibios_get_irq_routing_table()
Alex Chiang [Tue, 31 Mar 2009 15:24:02 +0000]
PCI Hotplug: cpqphp: clean up accesses to pcibios_get_irq_routing_table()

Instead of making multiple calls to pcibios_get_irq_routing_table, let's
just do it once and save the answer.

The reason we were making multiple calls is because we liked to calculate
its length and perform some loop over it. Instead of open-coding the length
calculation every time, provide it in an inline helper function.

Finally, since pci_print_IRQ_route() is used only for debug, let's only
do it when cpqhp_debug is set.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: refactor cpqhp_save_config
Alex Chiang [Tue, 31 Mar 2009 15:23:57 +0000]
PCI Hotplug: cpqphp: refactor cpqhp_save_config

Handle an empty slot at the top of the loop, and continue early.

This allows us to un-indent the rest of the function by one level.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: style cleanups
Alex Chiang [Tue, 31 Mar 2009 15:23:52 +0000]
PCI Hotplug: cpqphp: style cleanups

Clean up style, whitespace in cpqphp_pci.c

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: refactor cpqphp_save_slot_config
Alex Chiang [Tue, 31 Mar 2009 15:23:46 +0000]
PCI Hotplug: cpqphp: refactor cpqphp_save_slot_config

Check for an empty slot, and return early if so.

This allows us to un-indent the rest of the function by one level.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: clean up cpqphp_ctrl.c
Alex Chiang [Tue, 31 Mar 2009 15:23:41 +0000]
PCI Hotplug: cpqphp: clean up cpqphp_ctrl.c

Style and whitespace cleanups, no functional change.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: refactor cpqhp_probe
Alex Chiang [Tue, 31 Mar 2009 15:23:36 +0000]
PCI Hotplug: cpqphp: refactor cpqhp_probe

Apply DeMorgan's theorem:

if ((pdev->revision > 2) || (vendor_id == PCI_VENDOR_ID_INTEL))

turns into

if ((pdev->revision <= 2) && (vendor_id != PCI_VENDOR_ID_INTEL))

Now we can bail out early from the function if the controller is not
supported.

This allows us to un-indent the remainder of the function quite a bit and
make it much more readable.

Fix up some extra braces, and un-indent the 'case' labels in the switch
statement as per CodingStyle.

No functional change.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: eliminate stray braces
Alex Chiang [Tue, 31 Mar 2009 15:23:31 +0000]
PCI Hotplug: cpqphp: eliminate stray braces

Clean up style and eliminate superfluous braces and parens.

No functional change.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: remove useless prototypes in cpqphp_core.c
Alex Chiang [Tue, 31 Mar 2009 15:23:26 +0000]
PCI Hotplug: cpqphp: remove useless prototypes in cpqphp_core.c

Impact: refactor

Refactor code to follow convention more closely and eliminate the need
for some useless prototypes.

No functional change.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: obey 80 column convention in cpqphp.h
Alex Chiang [Tue, 31 Mar 2009 15:23:21 +0000]
PCI Hotplug: cpqphp: obey 80 column convention in cpqphp.h

Clean up cpqphp.h to follow 80 column convention.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: fix comment style
Alex Chiang [Tue, 31 Mar 2009 15:23:16 +0000]
PCI Hotplug: cpqphp: fix comment style

Fix up comments from C++ to C-style, wrapping if necessary, etc.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI Hotplug: cpqphp: stray whitespace cleanups
Alex Chiang [Tue, 31 Mar 2009 15:23:11 +0000]
PCI Hotplug: cpqphp: stray whitespace cleanups

Clean up all stray whitespace issues, such as trailing whitespace,
spaces before tabs, etc. and whatever else vim's c_space_errors
highlights in red.

Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: improve resource allocation under transparent bridges
Yinghai Lu [Fri, 24 Apr 2009 03:49:25 +0000]
PCI: improve resource allocation under transparent bridges

We could run out of space under under 4g, but devices under transparent
bridges can use 64bit resources, so keep trying on the parent bus until
we hit a non-transparent bridge.

Impact: better support for assigning unassigned resources

Reviewed-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI/x86: don't assume prefetchable ranges are 64bit
Yinghai Lu [Fri, 24 Apr 2009 03:48:32 +0000]
PCI/x86: don't assume prefetchable ranges are 64bit

We should not assign 64bit ranges to PCI devices that only take 32bit
prefetchable addresses.

Try to set IORESOURCE_MEM_64 in 64bit resource of pci_device/pci_bridge
and make the bus resource only have that bit set when all devices under
it support 64bit prefetchable memory.  Use that flag to allocate
resources from that range.

Reported-by: Yannick <yannick.roehlly@free.fr>
Reviewed-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI MSI: Define PCI_MSI_MASK_32/64
Hidetoshi Seto [Mon, 20 Apr 2009 01:54:59 +0000]
PCI MSI: Define PCI_MSI_MASK_32/64

Impact: cleanup, improve readability

Define PCI_MSI_MASK_32/64 for 32/64bit devices, instead of using
implicit offset (-4), "PCI_MSI_MASK_BIT - 4" and "PCI_MSI_MASK_BIT".

Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI: ibmphp_core.c: fix warning due to missing module_exit()
akpm@linux-foundation.org [Wed, 15 Apr 2009 21:24:08 +0000]
PCI: ibmphp_core.c: fix warning due to missing module_exit()

drivers/pci/hotplug/ibmphp_core.c:1414: warning: `ibmphp_exit' defined but not used

Signed-off-by: Zhenwen Xu <helight.xu@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCIE: remove driver_data direct access of struct device
Greg Kroah-Hartman [Thu, 30 Apr 2009 22:20:11 +0000]
PCIE: remove driver_data direct access of struct device

In the near future, the driver core is going to not allow direct access
to the driver_data pointer in struct device.  Instead, the functions
dev_get_drvdata() and dev_set_drvdata() should be used.  These functions
have been around since the beginning, so are backwards compatible with
all older kernel versions.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoPCI MSI: Remove unused/obsolete macros and definitions
Hidetoshi Seto [Mon, 20 Apr 2009 01:54:52 +0000]
PCI MSI: Remove unused/obsolete macros and definitions

Impact: cleanup, spec compliance

This patch does:

 - Remove unused msi/msix_enable/disable macros.
   User should use msi/msix_set_enable() functions instead.

 - Remove unused msix_mask/unmask/pending macros.
   These macros are useless because they are not based on any of
   the PCI Local Bus Specifications properly.
   It seems that they were written based on a draft of PCI spec,
   and that the draft was the MSI-X ECN that underwent membership
   review in September 2002.
   (* In the draft, the size of a entry in MSI-X table was 64bit,
      containing 32bit message data and DWORD aligned lower address
      plus a pending bit and a mask bit.(30+1+1bit)  The higher
      address was placed in MSI-X capability structure and shared
      by all entries.)

 - Remove PCI_MSIX_FLAGS_BITMASK.
   This definition also come from the draft ECN.

Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Reviewed-by: Matthew Wilcox <willy@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

10 years agoLinux 2.6.30
Linus Torvalds [Wed, 10 Jun 2009 03:05:27 +0000]
Linux 2.6.30

10 years agochar: mxser, fix ISA board lookup
Peter Botha [Wed, 10 Jun 2009 00:16:32 +0000]
char: mxser, fix ISA board lookup

There's a bug in the mxser kernel module that still appears in the
2.6.29.4 kernel.

mxser_get_ISA_conf takes a ioaddress as its first argument, by passing the
not of the ioaddr, you're effectively passing 0 which means it won't be
able to talk to an ISA card.  I have tested this, and removing the !
fixes the problem.

Cc: "Peter Botha" <peterb@goldcircle.co.za>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agojbd: fix race in buffer processing in commit code
Jan Kara [Tue, 9 Jun 2009 23:26:26 +0000]
jbd: fix race in buffer processing in commit code

In commit code, we scan buffers attached to a transaction.  During this
scan, we sometimes have to drop j_list_lock and then we recheck whether
the journal buffer head didn't get freed by journal_try_to_free_buffers().
 But checking for buffer_jbd(bh) isn't enough because a new journal head
could get attached to our buffer head.  So add a check whether the journal
head remained the same and whether it's still at the same transaction and
list.

This is a nasty bug and can cause problems like memory corruption (use after
free) or trigger various assertions in JBD code (observed).

Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <stable@kernel.org>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoautofs4: remove hashed check in validate_wait()
Ian Kent [Tue, 9 Jun 2009 23:26:24 +0000]
autofs4: remove hashed check in validate_wait()

The recent ->lookup() deadlock correction required the directory inode
mutex to be dropped while waiting for expire completion.  We were
concerned about side effects from this change and one has been identified.

I saw several error messages.

They cause autofs to become quite confused and don't really point to the
actual problem.

Things like:

handle_packet_missing_direct:1376: can't find map entry for (43,1827932)

which is usually totally fatal (although in this case it wouldn't be
except that I treat is as such because it normally is).

do_mount_direct: direct trigger not valid or already mounted
/test/nested/g3c/s1/ss1

which is recoverable, however if this problem is at play it can cause
autofs to become quite confused as to the dependencies in the mount tree
because mount triggers end up mounted multiple times.  It's hard to
accurately check for this over mounting case and automount shouldn't need
to if the kernel module is doing its job.

There was one other message, similar in consequence of this last one but I
can't locate a log example just now.

When checking if a mount has already completed prior to adding a new mount
request to the wait queue we check if the dentry is hashed and, if so, if
it is a mount point.  But, if a mount successfully completed while we
slept on the wait queue mutex the dentry must exist for the mount to have
completed so the test is not really needed.

Mounts can also be done on top of a global root dentry, so for the above
case, where a mount request completes and the wait queue entry has already
been removed, the hashed test returning false can cause an incorrect
callback to the daemon.  Also, d_mountpoint() is not sufficient to check
if a mount has completed for the multi-mount case when we don't have a
real mount at the base of the tree.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoshm: fix unused warnings on nommu
Mike Frysinger [Tue, 9 Jun 2009 23:26:23 +0000]
shm: fix unused warnings on nommu

The massive nommu update (8feae131) resulted in these warnings:
ipc/shm.c: In function `sys_shmdt':
ipc/shm.c:974: warning: unused variable `size'
ipc/shm.c:972: warning: unused variable `next'

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
Linus Torvalds [Tue, 9 Jun 2009 15:48:32 +0000]
Merge git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  kvm: fix kvm reboot crash when MAXSMP is used
  cpumask: alloc zeroed cpumask for static cpumask_var_ts
  cpumask: introduce zalloc_cpumask_var

10 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Tue, 9 Jun 2009 15:47:43 +0000]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  bsg: setting rq->bio to NULL

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Tue, 9 Jun 2009 15:47:27 +0000]
Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
  cls_cgroup: Fix oops when user send improperly 'tc filter add' request
  r8169: fix crash when large packets are received

10 years agoMerge branch 'for-linus' of git://neil.brown.name/md
Linus Torvalds [Tue, 9 Jun 2009 15:41:22 +0000]
Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md/raid5: fix bug in reshape code when chunk_size decreases.
  md/raid5 - avoid deadlocks in get_active_stripe during reshape
  md/raid5: use conf->raid_disks in preference to mddev->raid_disk

10 years agobsg: setting rq->bio to NULL
FUJITA Tomonori [Tue, 9 Jun 2009 13:17:37 +0000]
bsg: setting rq->bio to NULL

Due to commit 1cd96c242a829d52f7a5ae98f554ca9775429685 ("block: WARN
in __blk_put_request() for potential bio leak"), BSG SMP requests get
the false warnings:

WARNING: at block/blk-core.c:1068 __blk_put_request+0x52/0xc0()

This sets rq->bio to NULL to avoid that false warnings.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

10 years agokvm: fix kvm reboot crash when MAXSMP is used
Avi Kivity [Sat, 6 Jun 2009 21:52:35 +0000]
kvm: fix kvm reboot crash when MAXSMP is used

one system was found there is crash during reboot then kvm/MAXSMP
Sending all processes the KILL signal...                              done
Please stand by while rebooting the system...
[ 1721.856538] md: stopping all md devices.
[ 1722.852139] kvm: exiting hardware virtualization
[ 1722.854601] BUG: unable to handle kernel NULL pointer dereference at (null)
[ 1722.872219] IP: [<ffffffff8102c6b6>] hardware_disable+0x4c/0xb4
[ 1722.877955] PGD 0
[ 1722.880042] Oops: 0000 [#1] SMP
[ 1722.892548] last sysfs file: /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/host0/target0:2:0/0:2:0:0/vendor
[ 1722.900977] CPU 9
[ 1722.912606] Modules linked in:
[ 1722.914226] Pid: 0, comm: swapper Not tainted 2.6.30-rc7-tip-01843-g2305324-dirty #299 ...
[ 1722.932589] RIP: 0010:[<ffffffff8102c6b6>]  [<ffffffff8102c6b6>] hardware_disable+0x4c/0xb4
[ 1722.942709] RSP: 0018:ffffc900010b6ed8  EFLAGS: 00010046
[ 1722.956121] RAX: 0000000000000000 RBX: ffffc9000e253140 RCX: 0000000000000009
[ 1722.972202] RDX: 000000000000b020 RSI: ffffc900010c3220 RDI: ffffffffffffd790
[ 1722.977399] RBP: ffffc900010b6f08 R08: 0000000000000000 R09: 0000000000000000
[ 1722.995149] R10: 00000000000004b8 R11: 966912b6c78fddbd R12: 0000000000000009
[ 1723.011551] R13: 000000000000b020 R14: 0000000000000009 R15: 0000000000000000
[ 1723.019898] FS:  0000000000000000(0000) GS:ffffc900010b3000(0000) knlGS:0000000000000000
[ 1723.034389] CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
[ 1723.041164] CR2: 0000000000000000 CR3: 0000000001001000 CR4: 00000000000006e0
[ 1723.056192] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 1723.072546] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[ 1723.080562] Process swapper (pid: 0, threadinfo ffff88107e464000, task ffff88047e5a2550)
[ 1723.096144] Stack:
[ 1723.099071]  0000000000000046 ffffc9000e253168 966912b6c78fddbd ffffc9000e253140
[ 1723.115471]  ffff880c7d4304d0 ffffc9000e253168 ffffc900010b6f28 ffffffff81011022
[ 1723.132428]  ffffc900010b6f48 966912b6c78fddbd ffffc900010b6f48 ffffffff8100b83b
[ 1723.141973] Call Trace:
[ 1723.142981]  <IRQ> <0> [<ffffffff81011022>] kvm_arch_hardware_disable+0x26/0x3c
[ 1723.158153]  [<ffffffff8100b83b>] hardware_disable+0x3f/0x55
[ 1723.172168]  [<ffffffff810b95f6>] generic_smp_call_function_interrupt+0x76/0x13c
[ 1723.178836]  [<ffffffff8104cbea>] smp_call_function_interrupt+0x3a/0x5e
[ 1723.194689]  [<ffffffff81035bf3>] call_function_interrupt+0x13/0x20
[ 1723.199750]  <EOI> <0> [<ffffffff814ad3b4>] ? acpi_idle_enter_c1+0xd3/0xf4
[ 1723.217508]  [<ffffffff814ad3ae>] ? acpi_idle_enter_c1+0xcd/0xf4
[ 1723.232172]  [<ffffffff814ad4bc>] ? acpi_idle_enter_bm+0xe7/0x2ce
[ 1723.235141]  [<ffffffff81a8d93f>] ? __atomic_notifier_call_chain+0x0/0xac
[ 1723.253381]  [<ffffffff818c3dff>] ? menu_select+0x58/0xd2
[ 1723.258179]  [<ffffffff818c2c9d>] ? cpuidle_idle_call+0xa4/0xf3
[ 1723.272828]  [<ffffffff81034085>] ? cpu_idle+0xb8/0x101
[ 1723.277085]  [<ffffffff81a80163>] ? start_secondary+0x1bc/0x1d7
[ 1723.293708] Code: b0 00 00 65 48 8b 04 25 28 00 00 00 48 89 45 e0 31 c0 48 8b 04 cd 30 ee 27 82 49 89 cc 49 89 d5 48 8b 04 10 48 8d b8 90 d7 ff ff <48> 8b 87 70 28 00 00 48 8d 98 90 d7 ff ff eb 16 e8 e9 fe ff ff
[ 1723.335524] RIP  [<ffffffff8102c6b6>] hardware_disable+0x4c/0xb4
[ 1723.342076]  RSP <ffffc900010b6ed8>
[ 1723.352021] CR2: 0000000000000000
[ 1723.354348] ---[ end trace e2aec53dae150aa1 ]---

it turns out that we need clear cpus_hardware_enabled in that case.

Reported-and-tested-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

10 years agocpumask: alloc zeroed cpumask for static cpumask_var_ts
Yinghai Lu [Sat, 6 Jun 2009 21:51:36 +0000]
cpumask: alloc zeroed cpumask for static cpumask_var_ts

These are defined as static cpumask_var_t so if MAXSMP is not used,
they are cleared already.  Avoid surprises when MAXSMP is enabled.

Signed-off-by: Yinghai Lu <yinghai.lu@kernel.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

10 years agocpumask: introduce zalloc_cpumask_var
Yinghai Lu [Sat, 6 Jun 2009 21:50:36 +0000]
cpumask: introduce zalloc_cpumask_var

So can get cpumask_var with cpumask_clear

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

10 years agocls_cgroup: Fix oops when user send improperly 'tc filter add' request
Minoru Usui [Tue, 9 Jun 2009 11:03:09 +0000]
cls_cgroup: Fix oops when user send improperly 'tc filter add' request

I found a bug in cls_cgroup_change() in cls_cgroup.c.
cls_cgroup_change() expected tca[TCA_OPTIONS] was set from user space properly,
but tc in iproute2-2.6.29-1 (which I used) didn't set it.

In the current source code of tc in git, it set tca[TCA_OPTIONS].

  git://git.kernel.org/pub/scm/linux/kernel/git/shemminger/iproute2.git

If we always use a newest iproute2 in git when we use cls_cgroup,
we don't face this oops probably.
But I think, kernel shouldn't panic regardless of use program's behaviour.

Signed-off-by: Minoru Usui <usui@mxm.nes.nec.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>

10 years agor8169: fix crash when large packets are received
Eric Dumazet [Tue, 9 Jun 2009 11:01:02 +0000]
r8169: fix crash when large packets are received

Michael Tokarev reported receiving a large packet could crash
a machine with RTL8169 NIC.
( original thread at http://lkml.org/lkml/2009/6/8/192 )

Problem is this driver tells that NIC frames up to 16383 bytes
can be received but provides skb to rx ring allocated with
smaller sizes (1536 bytes in case standard 1500 bytes MTU is used)

When a frame larger than what was allocated by driver is received,
dma transfert can occurs past the end of buffer and corrupt
kernel memory.

Fix is to tell to NIC what is the maximum size a frame can be.

This bug is very old, (before git introduction, linux-2.6.10), and
should be backported to stable versions.

Reported-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

10 years agomd/raid5: fix bug in reshape code when chunk_size decreases.
NeilBrown [Tue, 9 Jun 2009 06:32:22 +0000]
md/raid5: fix bug in reshape code when chunk_size decreases.

Now that we support changing the chunksize, we calculate
"reshape_sectors" to be the max of number of sectors in old
and new chunk size.
However there is one please where we still use 'chunksize'
rather than 'reshape_sectors'.
This causes a reshape that reduces the size of chunks to freeze.

Signed-off-by: NeilBrown <neilb@suse.de>

10 years agomd/raid5 - avoid deadlocks in get_active_stripe during reshape
NeilBrown [Tue, 9 Jun 2009 04:39:59 +0000]
md/raid5 - avoid deadlocks in get_active_stripe during reshape

md has functionality to 'quiesce' and array so that all pending
IO completed and no new IO starts.  This is used to achieve a
stable state before making internal changes.

Currently this quiescing applies equally to normal IO, resync
IO, and reshape IO.
However there is a problem with applying it to reshape IO.
Reshape can have multiple 'stripe_heads' that must be active together.
If the quiesce come between allocating the first and the last of
such a collection, then we deadlock, as the last will not be allocated
until the quiesce is lifted, the quiesce will not be lifted until the
first (which has been allocated) gets used, and that first cannot be
used until the last is allocated.

It is not necessary to inhibit reshape IO when a quiesce is
requested.  Those places in the code that require a full quiesce will
ensure the reshape thread is not running at all.

So allow reshape requests to get access to new stripe_heads without
being blocked by a 'quiesce'.

This only affects in-place reshapes (i.e. where the array does not
grow or shrink) and these are only newly supported.  So this patch is
not needed in earlier kernels.

Signed-off-by: NeilBrown <neilb@suse.de>

10 years agomd/raid5: use conf->raid_disks in preference to mddev->raid_disk
NeilBrown [Tue, 9 Jun 2009 04:30:31 +0000]
md/raid5: use conf->raid_disks in preference to mddev->raid_disk

mddev->raid_disks can be changed and any time by a request from
user-space.  It is a suggestion as to what number of raid_disks is
desired.

conf->raid_disks can only be changed by the raid5 module with suitable
locks in place.  It is a statement as to the current number of
raid_disks.

There are two places where the latter should be used, but the former
is used.  This can lead to a crash when reshaping an array.

This patch changes to mddev-> to conf->

Signed-off-by: NeilBrown <neilb@suse.de>

10 years agoasync: Fix lack of boot-time console due to insufficient synchronization
Linus Torvalds [Mon, 8 Jun 2009 19:31:53 +0000]
async: Fix lack of boot-time console due to insufficient synchronization

Our async work synchronization was broken by "async: make sure
independent async domains can't accidentally entangle" (commit
d5a877e8dd409d8c702986d06485c374b705d340), because it would report
the wrong lowest active async ID when there was both running and
pending async work.

This caused things like no being able to read the root filesystem,
resulting in missing console devices and inability to run 'init',
causing a boot-time panic.

This fixes it by properly returning the lowest pending async ID: if
there is any running async work, that will have a lower ID than any
pending work, and we should _not_ look at the pending work list.

There were alternative patches from Jaswinder and James, but this one
also cleans up the code by removing the pointless 'ret' variable and
the unnecesary testing for an empty list around 'for_each_entry()' (if
the list is empty, the for_each_entry() thing just won't execute).

Fixes-bug: http://bugzilla.kernel.org/show_bug.cgi?id=13474
Reported-and-tested-by: Chris Clayton <chris2553@googlemail.com>
Cc: Jaswinder Singh Rajput <jaswinder@kernel.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Mon, 8 Jun 2009 16:22:53 +0000]
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus:
  MIPS: Outline udelay and fix a few issues.
  MIPS: ioctl.h: Fix headers_check warnings
  MIPS: Cobalt: PCI bus is always required to obtain the board ID
  MIPS: Kconfig: Remove "Support for" from Cavium system type
  MIPS: Sibyte: Honor CONFIG_CMDLINE
  SSB: BCM47xx: Export ssb_watchdog_timer_set

10 years agopata_netcell: Fix typo
Alan Cox [Mon, 8 Jun 2009 11:31:00 +0000]
pata_netcell: Fix typo

The previous patch submission had a I typo I didn't catch but Bartlomiej
noted. Guess this proves the point about any patch being risky late in an rc

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge branch 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Mon, 8 Jun 2009 16:05:48 +0000]
Merge branch 'kvm-updates/2.6.30' of git://git./virt/kvm/kvm

* 'kvm-updates/2.6.30' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Explicity initialize cpus_hardware_enabled

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6
Linus Torvalds [Mon, 8 Jun 2009 16:04:55 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/bart/ide-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6:
  pdc202xx_old: fix resetproc() method
  pdc202xx_old: fix 'pdc20246_dma_ops'

10 years agoMIPS: Outline udelay and fix a few issues.
Ralf Baechle [Sat, 28 Feb 2009 09:44:28 +0000]
MIPS: Outline udelay and fix a few issues.

Outlining fixes the issue were on certain CPUs such as the R10000 family
the delay loop would need an extra cycle if it overlaps a cacheline
boundary.

The rewrite also fixes build errors with GCC 4.4 which was changed in
way incompatible with the kernel's inline assembly.

Relying on pure C for computation of the delay value removes the need for
explicit.  The price we pay is a slight slowdown of the computation - to
be fixed on another day.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoMIPS: ioctl.h: Fix headers_check warnings
Jaswinder Singh Rajput [Thu, 4 Jun 2009 12:35:49 +0000]
MIPS: ioctl.h: Fix headers_check warnings

Make ioctl.h compatible with asm-generic/ioctl.h and userspace

fix the following 'make headers_check' warning:

  usr/include/asm-mips/ioctl.h:64: extern's make no sense in userspace

Signed-off-by: Jaswinder Singh Rajput <jaswinderrajput@gmail.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoMIPS: Cobalt: PCI bus is always required to obtain the board ID
Yoichi Yuasa [Tue, 2 Jun 2009 14:17:07 +0000]
MIPS: Cobalt: PCI bus is always required to obtain the board ID

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoMIPS: Kconfig: Remove "Support for" from Cavium system type
Yoichi Yuasa [Tue, 2 Jun 2009 14:15:10 +0000]
MIPS: Kconfig: Remove "Support for" from Cavium system type

Signed-off-by: Yoichi Yuasa <yoichi_yuasa@tripeaks.co.jp>
Acked-by: David Daney <ddaney@caviumnetworks.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoMIPS: Sibyte: Honor CONFIG_CMDLINE
Ralf Baechle [Tue, 2 Jun 2009 18:05:28 +0000]
MIPS: Sibyte: Honor CONFIG_CMDLINE

Original patch by Imre Kaloz <kaloz@openwrt.org>.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoSSB: BCM47xx: Export ssb_watchdog_timer_set
Matthieu Castet [Fri, 22 May 2009 20:25:04 +0000]
SSB: BCM47xx: Export ssb_watchdog_timer_set

this patch export ssb_watchdog_timer_set to allow to use it in a Linux
watchdog driver.

Signed-off-by: Matthieu CASTET <castet.matthieu@free.fr>
Acked-by : Michael Buesch <mb@bu3sch.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

10 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Mon, 8 Jun 2009 15:29:31 +0000]
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 5543/1: arm: serial amba: add missing declaration in serial.h
  [ARM] pxa: fix pxa27x_udc default pullup GPIO
  [ARM] pxa/imote2: fix UCAM sensor board ADC model number
  mx[23]: don't put clock lookups in __initdata
  fix oops when using console=ttymxcN with N > 0
  [ARM] ARMv7 errata: only apply fixes when running on applicable CPU
  [ARM] 5534/1: kmalloc must return a cache line aligned buffer

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Mon, 8 Jun 2009 14:53:59 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  sdhci-of: Fix the wrong accessor to HOSTVER register
  mvsdio: fix config failure with some high speed SDHC cards
  mvsdio: ignore high speed timing requests from the core
  mmc/omap: Use disable_irq_nosync() from within irq handlers.
  sdhci-of: Add fsl,esdhc as a valid compatible to bind against
  mvsdio: allow automatic loading when modular
  mxcmmc: Fix missing return value checking in DMA setup code.
  mxcmmc : Reset the SDHC hardware if software timeout occurs.
  omap_hsmmc: Trivial fix for a typo in comment
  mxcmmc: decrease minimum frequency to make MMC cards work

10 years agoKVM: Explicity initialize cpus_hardware_enabled
Avi Kivity [Sat, 6 Jun 2009 09:34:39 +0000]
KVM: Explicity initialize cpus_hardware_enabled

Under CONFIG_MAXSMP, cpus_hardware_enabled is allocated from the heap and
not statically initialized.  This causes a crash on reboot when kvm thinks
vmx is enabled on random nonexistent cpus and accesses nonexistent percpu
lists.

Fix by explicitly clearing the variable.

Cc: stable@kernel.org
Reported-and-tested-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Avi Kivity <avi@redhat.com>

10 years ago[ARM] 5543/1: arm: serial amba: add missing declaration in serial.h
Alessandro Rubini [Sat, 6 Jun 2009 09:17:57 +0000]
[ARM] 5543/1: arm: serial amba: add missing declaration in serial.h

This header is sometimes included in the uncompress stage to get
register values, but no <linux/amba/bus.h> can be included there.
So declare "struct amba_device" here before using it in a prototype.

Signed-off-by: Alessandro Rubini <rubini@unipv.it>
Acked-by: Andrea Gallo <andrea.gallo@stericsson.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

10 years agopdc202xx_old: fix resetproc() method
Sergei Shtylyov [Sun, 7 Jun 2009 11:52:50 +0000]
pdc202xx_old: fix resetproc() method

pdc202xx_reset() calls pdc202xx_reset_host() twice, for both channels, while
that function actually twiddles the single, shared software reset bit -- the
net effect is a duplicated reset and horrendous 4 second delay happening not
only on a channel reset but also when dma_lost_irq() and dma_clear() methods
are called.  Fold pdc202xx_reset_host() into pdc202xx_reset(), fix printk(),
and move it before the actual reset...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

10 years agopdc202xx_old: fix 'pdc20246_dma_ops'
Sergei Shtylyov [Sun, 7 Jun 2009 11:52:50 +0000]
pdc202xx_old: fix 'pdc20246_dma_ops'

Commit ac95beedf8bc97b24f9540d4da9952f07221c023 (ide: add struct ide_port_ops
(take 2)) erroneously converted the driver's dma_timeout() and dma_lost_irq()
methods to call the driver's resetproc() method regardless of whether it was
defined for this specific controller while it hadn't been defined and hence
called for PDC20246. So the dma_clear() method, the successor of dma_timeout(),
shouldn't exist and the dma_lost_irq() method should be standard for PDC20246.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Sat, 6 Jun 2009 21:33:54 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  x86/pci: fix mmconfig detection with 32bit near 4g
  PCI: use fixed-up device class when configuring device

10 years agointegrity: fix IMA inode leak
Hugh Dickins [Sat, 6 Jun 2009 20:18:09 +0000]
integrity: fix IMA inode leak

CONFIG_IMA=y inode activity leaks iint_cache and radix_tree_node objects
until the system runs out of memory.  Nowhere is calling ima_inode_free()
a.k.a. ima_iint_delete().  Fix that by calling it from destroy_inode().

Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Sat, 6 Jun 2009 19:18:14 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  ext3/4 with synchronous writes gets wedged by Postfix
  Fix nobh_truncate_page() to not pass stack garbage to get_block()

10 years agoMerge branch 'upstream-linus2' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarz...
Linus Torvalds [Sat, 6 Jun 2009 19:17:03 +0000]
Merge branch 'upstream-linus2' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus2' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  [libata] pata_ali: Use IGN_SIMPLEX

10 years agoext3/4 with synchronous writes gets wedged by Postfix
Al Viro [Wed, 13 May 2009 18:13:40 +0000]
ext3/4 with synchronous writes gets wedged by Postfix

OK, that's probably the easiest way to do that, as much as I don't like it...
Since iget() et.al. will not accept I_FREEING (will wait to go away
and restart), and since we'd better have serialization between new/free
on fs data structures anyway, we can afford simply skipping I_FREEING
et.al. in insert_inode_locked().

We do that from new_inode, so it won't race with free_inode in any interesting
ways and it won't race with iget (of any origin; nfsd or in case of fs
corruption a lookup) since both still will wait for I_LOCK.

Reviewed-by: "Theodore Ts'o" <tytso@mit.edu>
Acked-by: Jan Kara <jack@suse.cz>
Tested-by: David Watson <dbwatson@ukfsn.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

10 years agoFix nobh_truncate_page() to not pass stack garbage to get_block()
Theodore Ts'o [Tue, 12 May 2009 11:37:56 +0000]
Fix nobh_truncate_page() to not pass stack garbage to get_block()

The nobh_truncate_page() function is used by ext2, exofs, and jfs.  Of
these three, only ext2 and jfs's get_block() function pays attention
to bh->b_size --- which is normally always the filesystem blocksize
except when the get_block() function is called by either
mpage_readpage(), mpage_readpages(), or the direct I/O routines in
fs/direct_io.c.

Unfortunately, nobh_truncate_page() does not initialize map_bh before
calling the filesystem-supplied get_block() function.  So ext2 and jfs
will try to calculate the number of blocks to map by taking stack
garbage and shifting it left by inode->i_blkbits.  This should be
*mostly* harmless (except the filesystem will do some unnneeded work)
unless the stack garbage is less than filesystem's blocksize, in which
case maxblocks will be zero, and the attempt to find out whether or
not the filesystem has a hole at a given logical block will fail, and
the page cache entry might not get zero'ed out.

Also if the stack garbage in in map_bh->state happens to have the
BH_Mapped bit set, there could be an attempt to call readpage() on a
non-existent page, which could cause nobh_truncate_page() to return an
error when it should not.

Fix this by initializing map_bh->state and map_bh->size.

Fortunately, it's probably fairly unlikely that ext2 and jfs users
mount with nobh these days.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

10 years ago[libata] pata_ali: Use IGN_SIMPLEX
Alan Cox [Wed, 13 May 2009 14:02:27 +0000]
[libata] pata_ali: Use IGN_SIMPLEX

Some ALi devices report simplex if they have been disabled and re-enabled, and
restoring the byte does not work. Ignore it - the needed supporting logic is
already present for the SATA ULi ports.

Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable
Linus Torvalds [Fri, 5 Jun 2009 18:54:28 +0000]
Merge git://git./linux/kernel/git/mason/btrfs-unstable

* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
  Btrfs: Fix oops and use after free during space balancing
  Btrfs: set device->total_disk_bytes when adding new device

10 years agomtd: davinci nand: update clock naming
Kevin Hilman [Fri, 5 Jun 2009 17:48:08 +0000]
mtd: davinci nand: update clock naming

DaVinci clock support has been updated in mainline.
Update clock names accordingly.

Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Fri, 5 Jun 2009 18:53:44 +0000]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ata_piix: Add HP Compaq nc6000 to the broken poweroff list
  ahci: add warning messages for hp laptops with broken suspend
  pata_efar: fix PIO2 underclocking
  pata_legacy: wait for async probing

10 years agoata_piix: Add HP Compaq nc6000 to the broken poweroff list
Ville Syrjala [Mon, 18 May 2009 22:37:44 +0000]
ata_piix: Add HP Compaq nc6000 to the broken poweroff list

HP Compaq nc6000 suffers from the double disk spindown issue.
Add it to the broken poweroff DMI list.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

10 years agoahci: add warning messages for hp laptops with broken suspend
Tejun Heo [Sat, 30 May 2009 11:50:12 +0000]
ahci: add warning messages for hp laptops with broken suspend

Harddisks on HP dv[4-6] and HDX18 fail to come online after resume on
earlier BIOSen.  Fortunately, HP recently released BIOS updates for
all machines to fix the issue.  Detect old BIOSen, warn the user to
update BIOS on boot and suspend attempts and fail suspend.

Kudos to all the bug reporters.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: kernel.org@epperson.homelinux.net
Cc: emisca@gmail.com
Cc: Gadi Cohen <dragon@wastelands.net>
Cc: Paul Swanson <paul@procursa.com>
Cc: s@ourada.org
Cc: Trevor Davenport <trevor.davenport@gmail.com>
Cc: corruptor1972 <steven_tierney@yahoo.co.uk>
Cc: Victoria Wilson <mail@vwilson.co.uk>
Cc: khiraly <khiraly.list@gmail.com>
Cc: Sean <wollombi@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

10 years agopata_efar: fix PIO2 underclocking
Sergei Shtylyov [Mon, 1 Jun 2009 19:42:10 +0000]
pata_efar: fix PIO2 underclocking

Fix the PIO mode 2 using mode 0 timings -- this driver should enable the
fast timing bank starting with PIO2, just like the PIIX/ICH drivers do.
Also, fix/rephrase some comments while at it.

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

10 years agopata_legacy: wait for async probing
James Bottomley [Fri, 5 Jun 2009 14:41:39 +0000]
pata_legacy: wait for async probing

The basic problem here that pata_legacy attaches the host, sees if it found
any devices and detaches it if none were found.  With async probing, it's not
waiting until discovery is finished before deciding it has no devices and
trying the detach leading to this warning:

ata1: PATA max PIO4 cmd 0x1f0 ctl 0x3f6 irq 14
------------[ cut here ]------------
WARNING: at drivers/ata/libata-core.c:6222 ata_host_detach+0x75/0x90()
Modules linked in:
Pid: 1, comm: swapper Not tainted 2.6.30-rc7 #1
Call Trace:
 [<c01fbb05>] ? ata_host_detach+0x75/0x90
 [<c01fbb05>] ? ata_host_detach+0x75/0x90
 [<c01139b5>] ? warn_slowpath_common+0x45/0x80
 [<c01139fa>] ? warn_slowpath_null+0xa/0x10
 [<c01fbb05>] ? ata_host_detach+0x75/0x90
 [<c02f40e0>] ? legacy_init+0x44e/0x87f
 [<c02f3c92>] ? legacy_init+0x0/0x87f
 [<c0101021>] ? _stext+0x21/0x140
 [<c01890ff>] ? proc_register+0x2f/0x190
 [<c018938c>] ? create_proc_entry+0x5c/0xc0
 [<c0135ebe>] ? register_irq_proc+0x6e/0x90
 [<c02e6484>] ? kernel_init+0x6e/0xbf
 [<c02e6416>] ? kernel_init+0x0/0xbf
 [<c01031d7>] ? kernel_thread_helper+0x7/0x10
---[ end trace ef1ee36e873ae3a0 ]---

Because it detaches before the probe is complete.

One way to fix it would be to put an async_synchronize_full() before looking
for devices, which this patch does.  A better way might be to separate libata
into its own domain and only wait for that.

Reported-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

10 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Fri, 5 Jun 2009 17:46:48 +0000]
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] powernow-k8: check space_id of _PCT registers to be FFH

10 years ago[CPUFREQ] powernow-k8: check space_id of _PCT registers to be FFH
Dave Jones [Fri, 5 Jun 2009 16:37:07 +0000]
[CPUFREQ] powernow-k8: check space_id of _PCT registers to be FFH

The powernow-k8 driver checks to see that the Performance Control/Status
Registers are declared as FFH (functional fixed hardware) by the BIOS.
However, this check got broken in the commit:
 0e64a0c982c06a6b8f5e2a7f29eb108fdf257b2f
 [CPUFREQ] checkpatch cleanups for powernow-k8

Fix based on an original patch from Naga Chumbalkar.

Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Cc: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>

10 years agoRevert "drm: don't associate _DRM_DRIVER maps with a master"
Linus Torvalds [Fri, 5 Jun 2009 17:21:52 +0000]
Revert "drm: don't associate _DRM_DRIVER maps with a master"

This reverts commit 6c51d1cfa0a370b48a157163340190cf5fd2346b, which
apparently causes DRI initialization failures on Radeons.

Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Requested-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoivtv: Fix PCI DMA direction
Alan Cox [Fri, 5 Jun 2009 10:56:18 +0000]
ivtv: Fix PCI DMA direction

The ivtv stream buffers may be for receive or for send but the attached
sg handle is always destined cpu->device.  We flush it correctly but the
allocation is wrongly done with the same type as the buffers.

See bug: http://bugzilla.kernel.org/show_bug.cgi?id=13385

(Note this doesn't close the bug - it fixes the ivtv part and in turn
the logging next shows up some rather alarming DMA sg list warnings in
libata)

Signed-off-by: Alan Cox <alan@linux.intel.com>
Acked-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoptrace: revert "ptrace_detach: the wrong wakeup breaks the ERESTARTxxx logic"
Oleg Nesterov [Thu, 4 Jun 2009 23:29:09 +0000]
ptrace: revert "ptrace_detach: the wrong wakeup breaks the ERESTARTxxx logic"

Commit 95a3540da9c81a5987be810e1d9a83640a366bd5 ("ptrace_detach: the wrong
wakeup breaks the ERESTARTxxx logic") removed the "extra"
wake_up_process() from ptrace_detach(), but as Jan pointed out this breaks
the compatibility.

I believe the changelog is right and this wake_up() is wrong in many
ways, but GDB assumes that ptrace(PTRACE_DETACH, child, 0, 0) always
wakes up the tracee.

Despite the fact this breaks SIGNAL_STOP_STOPPED/group_stop_count logic,
and despite the fact this wake_up_process() can break another
assumption: PTRACE_DETACH with SIGSTOP should leave the tracee in
TASK_STOPPED case.  Because the untraced child can dequeue SIGSTOP and
call do_signal_stop() before ptrace_detach() calls wake_up_process().

Revert this change for now.  We need some fixes even if we we want to keep
the current behaviour, but these fixes are not for 2.6.30.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agokbuild: fix detection of CONFIG_FRAME_WARN=0
Mike Frysinger [Thu, 4 Jun 2009 23:29:08 +0000]
kbuild: fix detection of CONFIG_FRAME_WARN=0

The checking of CONFIG_FRAME_WARN in the top level Makefile forgot to
actually derefence the variable thus leading to an always true check.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoptrace: tracehook_report_clone: fix false positives
Oleg Nesterov [Thu, 4 Jun 2009 23:29:07 +0000]
ptrace: tracehook_report_clone: fix false positives

The "trace || CLONE_PTRACE" check in tracehook_report_clone() is not right,

- If the untraced task does clone(CLONE_PTRACE) the new child is not traced,
  we must not queue SIGSTOP.

- If we forked the traced task, but the tracer exits and untraces both the
  forking task and the new child (after copy_process() drops tasklist_lock),
  we should not queue SIGSTOP too.

Change the code to check task_ptrace() != 0 instead. This is still racy, but
the race is harmless.

We can race with another tracer attaching to this child, or the tracer can
exit and detach in parallel. But giwen that we didn't do wake_up_new_task()
yet, the child must have the pending SIGSTOP anyway.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt...
Linus Torvalds [Thu, 4 Jun 2009 22:23:51 +0000]
Merge branch 'for-linus' of git://git./linux/kernel/git/anholt/drm-intel

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel:
  drm/i915: Remove a bad BUG_ON in the fence management code.

10 years agoMerge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Thu, 4 Jun 2009 22:23:39 +0000]
Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm: ignore EDID with really tiny modes.
  drm: don't associate _DRM_DRIVER maps with a master
  drm/i915: intel_lvds.c fix section mismatch
  drm: Hook up DPMS property handling in drm_crtc.c. Add drm_helper_connector_dpms.
  drm: set permissions on edid file to 0444
  drm: add newlines to text sysfs files
  drm/radeon: fix ring free alignment calculations
  drm: fix irq naming for kms drivers.

10 years agodrivers/char/mem.c: avoid OOM lockup during large reads from /dev/zero
Salman Qazi [Thu, 4 Jun 2009 22:20:39 +0000]
drivers/char/mem.c: avoid OOM lockup during large reads from /dev/zero

While running 20 parallel instances of dd as follows:

  #!/bin/bash
  for i in `seq 1 20`; do
           dd if=/dev/zero of=/export/hda3/dd_$i bs=1073741824 count=1 &
  done
  wait

on a 16G machine, we noticed that rather than just killing the processes,
the entire kernel went down.  Stracing dd reveals that it first does an
mmap2, which makes 1GB worth of zero page mappings.  Then it performs a
read on those pages from /dev/zero, and finally it performs a write.

The machine died during the reads.  Looking at the code, it was noticed
that /dev/zero's read operation had been changed by
557ed1fa2620dc119adb86b34c614e152a629a80 ("remove ZERO_PAGE") from giving
zero page mappings to actually zeroing the page.

The zeroing of the pages causes physical pages to be allocated to the
process.  But, when the process exhausts all the memory that it can, the
kernel cannot kill it, as it is still in the kernel mode allocating more
memory.  Consequently, the kernel eventually crashes.

To fix this, I propose that when a fatal signal is pending during
/dev/zero read operation, we simply return and let the user process die.

Signed-off-by: Salman Qazi <sqazi@google.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Modified error return and comment trivially.  - Linus]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

10 years agoBtrfs: Fix oops and use after free during space balancing
Chris Mason [Thu, 4 Jun 2009 19:34:51 +0000]
Btrfs: Fix oops and use after free during space balancing

The btrfs allocator uses list_for_each to walk the available block
groups when searching for free blocks.  It starts off with a hint
to help find the best block group for a given allocation.

The hint is resolved into a block group, but we don't properly check
to make sure the block group we find isn't in the middle of being
freed due to filesystem shrinking or balancing.  If it is being
freed, the list pointers in it are bogus and can't be trusted.  But,
the code happily goes along and uses them in the list_for_each loop,
leading to all kinds of fun.

The fix used here is to check to make sure the block group we find really
is on the list before we use it.  list_del_init is used when removing
it from the list, so we can do a proper check.

The allocation clustering code has a similar bug where it will trust
the block group in the current free space cluster.  If our allocation
flags have changed (going from single spindle dup to raid1 for example)
because the drives in the FS have changed, we're not allowed to use
the old block group any more.

The fix used here is to check the current cluster against the
current allocation flags.

Signed-off-by: Chris Mason <chris.mason@oracle.com>