GitHub/exynos8895/android_kernel_samsung_universal8895.git
13 years agoMerge remote branch 'kumar/next' into next
Benjamin Herrenschmidt [Fri, 20 May 2011 03:43:47 +0000 (13:43 +1000)]
Merge remote branch 'kumar/next' into next

13 years agopowerpc/kvm: Fix the build for 32-bit Book 3S (classic) processors
Paul Mackerras [Wed, 11 May 2011 00:39:50 +0000 (00:39 +0000)]
powerpc/kvm: Fix the build for 32-bit Book 3S (classic) processors

Commits a5d4f3ad3a ("powerpc: Base support for exceptions using
HSRR0/1") and 673b189a2e ("powerpc: Always use SPRN_SPRG_HSCRATCH0
when running in HV mode") cause compile and link errors for 32-bit
classic Book 3S processors when KVM is enabled.  This fixes these
errors.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/kvm: Fix kvmppc_core_pending_dec
Paul Mackerras [Wed, 11 May 2011 00:38:50 +0000 (00:38 +0000)]
powerpc/kvm: Fix kvmppc_core_pending_dec

The vcpu->arch.pending_exceptions field is a bitfield indexed by
interrupt priority number as returned by kvmppc_book3s_vec2irqprio.
However, kvmppc_core_pending_dec was using an interrupt vector shifted
by 7 as the bit index.  Fix it to use the irqprio value for the
decrementer interrupt instead.  This problem was found by code
inspection.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove last piece of GEMINI
Sebastian Siewior [Mon, 16 May 2011 08:58:13 +0000 (08:58 +0000)]
powerpc: Remove last piece of GEMINI

It seems that Adrian is getting old. He removed almost everything of
GEMINI in commit c53653130 ("[POWERPC] Remove the broken Gemini
support") except this piece.

Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Fix for Pegasos keyboard and mouse
Gabriel Paubert [Fri, 13 May 2011 01:03:13 +0000 (01:03 +0000)]
powerpc: Fix for Pegasos keyboard and mouse

[See http://lists.ozlabs.org/pipermail/linuxppc-dev/2010-October/086424.html
and followups. Part of the commit message is directly copied from that.]

Commit 540c6c392f01887dcc96bef0a41e63e6c1334f01 tries to find i8042 IRQs in
the device-tree but doesn't fall back to the old hardcoded 1 and 12 in all
failure cases.

Specifically, the case where the device-tree contains nothing matching
pnpPNP,303 or pnpPNP,f03 doesn't seem to be handled well. It sort of falls
through to the old code, but leaves the IRQs set to 0.

Signed-off-by: Gabriel Paubert <paubert@iram.es>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Make early memory scan more resilient to out of order nodes
Benjamin Herrenschmidt [Wed, 11 May 2011 20:58:18 +0000 (20:58 +0000)]
powerpc: Make early memory scan more resilient to out of order nodes

We keep track of the size of the lowest block of memory and call
setup_initial_memory_limit() only after we've parsed them all

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Milton Miller <miltonm@bga.com>
13 years agopowerpc/pseries/iommu: Cleanup ddw naming
Milton Miller [Wed, 11 May 2011 12:25:00 +0000 (12:25 +0000)]
powerpc/pseries/iommu: Cleanup ddw naming

When using a property refering to the availibily of dynamic dma windows
call it ddw_avail not ddr_avail.

dupe_ddw_if_already_created does not dupilcate anything, it only finds
and reuses the windows we already created, so rename it to
find_existing_ddw.  Also, it does not need the pci device node, so
remove that argument.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries/iommu: Find windows after kexec during boot
Milton Miller [Wed, 11 May 2011 12:24:59 +0000 (12:24 +0000)]
powerpc/pseries/iommu: Find windows after kexec during boot

Move the discovery of windows previously setup from when the pci driver
calls set_dma_mask to an arch_initcall.

When kexecing into a kernel with dynamic dma windows allocated, we need
to find the windows early so that memory hot remove will be able to
delete the tces mapping the to be removed memory and memory hotplug add
will map the new memory into the window.  We should not wait for the
driver to be loaded and the device to be probed.  The iommu init hooks
are before kmalloc is setup, so defer to arch_initcall.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries/iommu: Remove ddw property when destroying window
Milton Miller [Wed, 11 May 2011 12:24:58 +0000 (12:24 +0000)]
powerpc/pseries/iommu: Remove ddw property when destroying window

If we destroy the window, we need to remove the property recording that
we setup the window.  Otherwise the next kernel we kexec will be
confused.

Also we should remove the property if even if we don't find the
ibm,ddw-applicable window or if one of the property sizes is unexpected;
presumably these came from a prior kernel via kexec, and we will not be
maintaining the window with respect to memory hotplug.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries/iommu: Add additional checks when changing iommu mask
Milton Miller [Wed, 11 May 2011 12:24:57 +0000 (12:24 +0000)]
powerpc/pseries/iommu: Add additional checks when changing iommu mask

Do not check dma supported until we have chosen the right dma ops.
Check that the device is pci before treating it as such.

Check the mask is supported by the selected dma ops before
committing it.

We only need to set iommu ops if it is not the current ops; this
avoids searching the tree for the iommu table unnecessarily.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries/iommu: Use correct return type in dupe_ddw_if_already_created
Nishanth Aravamudan [Wed, 11 May 2011 11:07:51 +0000 (11:07 +0000)]
powerpc/pseries/iommu: Use correct return type in dupe_ddw_if_already_created

Otherwise we get silent truncations.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Milton Miller <miltonm@bga.com>
Cc: linuxppc-dev@ozlabs.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agoMerge branch 'merge' into next
Benjamin Herrenschmidt [Thu, 19 May 2011 07:00:06 +0000 (17:00 +1000)]
Merge branch 'merge' into next

13 years agopowerpc: Remove unused/obsolete CONFIG_XICS
Benjamin Herrenschmidt [Thu, 19 May 2011 04:44:31 +0000 (14:44 +1000)]
powerpc: Remove unused/obsolete CONFIG_XICS

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agomisc: Add CARMA DATA-FPGA Programmer support
Ira Snyder [Fri, 11 Feb 2011 13:34:30 +0000 (13:34 +0000)]
misc: Add CARMA DATA-FPGA Programmer support

This adds support for programming the data processing FPGAs on the OVRO
CARMA board. These FPGAs have a special programming sequence that
requires that we program the Freescale DMA engine, which is only
available inside the kernel.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agomisc: Add CARMA DATA-FPGA Access Driver
Ira Snyder [Fri, 11 Feb 2011 13:34:29 +0000 (13:34 +0000)]
misc: Add CARMA DATA-FPGA Access Driver

This driver allows userspace to access the data processing FPGAs on the
OVRO CARMA board. It has two modes of operation:

1) random access

This allows users to poke any DATA-FPGA registers by using mmap to map
the address region directly into their memory map.

2) correlation dumping

When correlating, the DATA-FPGA's have special requirements for getting
the data out of their memory before the next correlation. This nominally
happens at 64Hz (every 15.625ms). If the data is not dumped before the
next correlation, data is lost.

The data dumping driver handles buffering up to 1 second worth of
correlation data from the FPGAs. This lowers the realtime scheduling
requirements for the userspace process reading the device.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Make IRQ_NOREQUEST last to clear, first to set
Milton Miller [Tue, 10 May 2011 19:30:44 +0000 (19:30 +0000)]
powerpc: Make IRQ_NOREQUEST last to clear, first to set

When creating an irq, don't allow a concurent driver request until
we have caled map, which will likley call set_chip_and_handler to
change the irq_chip and its operations.

Similarly, when tearing down an IRQ, make sure no new uses come
along while we change the irq back to the nop chip and then reset
the descriptor to freed status.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Integrated Flash controller device tree bindings
Dipen Dudhat [Thu, 19 May 2011 06:38:23 +0000 (12:08 +0530)]
powerpc: Integrated Flash controller device tree bindings

Signed-off-by: Dipen Dudhat <Dipen.Dudhat@freescale.com>
Acked-By: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/85xx: Create dts of each core in CAMP mode for P1020RDB
Prabhakar Kushwaha [Thu, 28 Apr 2011 07:00:00 +0000 (12:30 +0530)]
powerpc/85xx: Create dts of each core in CAMP mode for P1020RDB

Create the dts files for each core and splits the devices between the two
cores for P1020RDB.

Core0 has core0 to have memory, l2, i2c, spi, gpio, tdm, dma, usb, eth1,
eth2, sdhc, crypto, global-util, message, pci0, pci1, msi.
Core1 has l2, eth0, crypto.

MPIC is shared between two cores but each core will protect its interrupts
from other core by using "protected-sources" of mpic.

Fix compatible property for global-util node of P1020si.dtsi.

Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/85xx: Fix PCIe IDSEL for Px020RDB
Prabhakar Kushwaha [Wed, 20 Apr 2011 04:12:44 +0000 (09:42 +0530)]
powerpc/85xx: Fix PCIe IDSEL for Px020RDB

PCIe device in legacy mode can trigger interrupts using the wires #INTA,
#INTB ,#INTC and #INTD. PCI devices are obligated to use #INTx for
interrupts under legacy mode.  Each PCI slot or device is typically wired
to different inputs on the interrupt controller.

So, Define interrupt-map and interrupt-map-mask properties for device tree
to of map each PCI interrupt signal to the inputs of the interrupt
controller.

Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/85xx: P2020 DTS: re-organize dts files
Prabhakar Kushwaha [Fri, 8 Apr 2011 12:27:05 +0000 (17:57 +0530)]
powerpc/85xx: P2020 DTS: re-organize dts files

Creates P2020si.dtsi, containing information for P2020 SoC. Modifies dts
files for P2020 based systems to use dtsi file.

Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/85xx: P1020 DTS : re-organize dts files
Prabhakar Kushwaha [Thu, 7 Apr 2011 09:10:55 +0000 (14:40 +0530)]
powerpc/85xx: P1020 DTS : re-organize dts files

Creates P1020si.dtsi, containing information for the P1020 SoC. Modifies dts
files for P1020 based systems to use dtsi file

Signed-off-by: Prabhakar Kushwaha <prabhakar@freescale.com>
Acked-by: Grant Likely <grant.likelY@secretlab.ca>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc: Adding bindings for flexcan controller
Bhaskar Upadhaya [Tue, 19 Apr 2011 13:58:31 +0000 (19:28 +0530)]
powerpc: Adding bindings for flexcan controller

Signed-off-by: Bhaskar Upadhaya <bhaskar.upadhaya@freescale.com>
Acked-By: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/fsl: enable verbose bug output
Scott Wood [Tue, 10 May 2011 18:02:06 +0000 (13:02 -0500)]
powerpc/fsl: enable verbose bug output

This debug option has no overhead other than a slight increase in
kernel size, and makes bug reports more useful.  While some end users
may prefer to save the space, as a default on a kernel config aimed
primarily at development on reference boards, it should be enabled.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/e5500: add networking to defconfig
Scott Wood [Tue, 10 May 2011 18:01:47 +0000 (13:01 -0500)]
powerpc/e5500: add networking to defconfig

Even though support for the p5020's on-chip ethernet is not yet upstream,
it is not appropriate to disable all networking support (including
loopback, unix domain sockets, external ethernet devices, etc) in the
defconfig.  The networking settings are taken from mpc85xx_smp_defconfig,
minus the drivers for ethernet devices not found on any current e5500
chip.

The other changes are the result of running "make savedefconfig".

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/mpic: add the mpic global timer support
Scott Wood [Thu, 24 Mar 2011 21:43:55 +0000 (16:43 -0500)]
powerpc/mpic: add the mpic global timer support

Add support for MPIC timers as requestable interrupt sources.

Based on http://patchwork.ozlabs.org/patch/20941/ by Dave Liu.

Signed-off-by: Dave Liu <daveliu@freescale.com>
Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/mpic: parse 4-cell intspec types other than zero
Scott Wood [Thu, 24 Mar 2011 21:43:54 +0000 (16:43 -0500)]
powerpc/mpic: parse 4-cell intspec types other than zero

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/p1022ds: fix broken mpic timer node
Scott Wood [Thu, 24 Mar 2011 21:43:52 +0000 (16:43 -0500)]
powerpc/p1022ds: fix broken mpic timer node

There is no hardware interrupt 0xf7.  But now we can express the timer
interrupt using 4-cell interrupts.  This requires converting all of the
other interrupt specifiers in the tree as well.

Also add the second timer group, and fix the reg property to only
describe the timer registers.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc: Add fsl mpic timer binding
Scott Wood [Thu, 24 Mar 2011 21:43:15 +0000 (16:43 -0500)]
powerpc: Add fsl mpic timer binding

Update the existing example in the general mpic binding to have a
separate TCRx region.  Currently the example doesn't describe TCRx at
all.  The one upstream device tree with an mpic timer node (p1022ds)
uses one large reg region to describe both, even though there are other
unrelated registers in between.  That device tree also contains a bogus
interrupt specifier, and there's no upstream software that uses this yet,
so changing this shouldn't be a problem.

Add a full binding for the MPIC timer node, not just an example of
4-cell interrupts in the MPIC binding.

Add fsl,available-ranges, similar to msi-available-ranges.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/86xx: don't pretend that we support 8-bit pixels on the MPC8610 HPCD
Timur Tabi [Mon, 9 May 2011 19:29:40 +0000 (14:29 -0500)]
powerpc/86xx: don't pretend that we support 8-bit pixels on the MPC8610 HPCD

If the video mode is set to 16-, 24-, or 32-bit pixels, then the pixel data
contains actual levels of red, blue, and green.  However, if the video mode
is set to 8-bit pixels, then the 8-bit value represents an index into color
table.  This is called "palette mode" on the Freescale DIU video controller.

The DIU driver does not currently support palette mode, but the MPC8610 HPCD
board file returned a non-zero (although incorrect) pixel format value for
8-bit mode.

Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/mpc8610_hpcd: Do not use "/" in interrupt names
Geert Uytterhoeven [Wed, 4 May 2011 14:29:31 +0000 (16:29 +0200)]
powerpc/mpc8610_hpcd: Do not use "/" in interrupt names

It may trigger a warning in fs/proc/generic.c:__xlate_proc_name() when
trying to add an entry for the interrupt handler to sysfs.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/e5500: set non-base IVORs
Scott Wood [Mon, 9 May 2011 21:26:00 +0000 (16:26 -0500)]
powerpc/e5500: set non-base IVORs

Without this, we attempt to use doorbells for IPIs, and end up
branching to some bad address.  Plus, even for the exceptions
we don't implement, it's good to handle it and get a message out.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agopowerpc/fsl-booke64: Add support for Debug Level exception handler
Kumar Gala [Wed, 6 Apr 2011 05:18:48 +0000 (00:18 -0500)]
powerpc/fsl-booke64: Add support for Debug Level exception handler

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
13 years agoMerge remote branch 'benh/merge' into benh-next
Kumar Gala [Thu, 19 May 2011 05:36:21 +0000 (00:36 -0500)]
Merge remote branch 'benh/merge' into benh-next

13 years agopowerpc: Remove virq_to_host
Milton Miller [Tue, 10 May 2011 19:30:40 +0000 (19:30 +0000)]
powerpc: Remove virq_to_host

The only references to the irq_map[].host field are internal to
arch/powerpc/kernel/irq.c

Signed-off-by: Milton Miller <miltonm@bga.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Add virq_is_host to reduce virq_to_host usage
Milton Miller [Tue, 10 May 2011 19:30:36 +0000 (19:30 +0000)]
powerpc: Add virq_is_host to reduce virq_to_host usage

Some irq_host implementations are using virq_to_host to check if
they are the irq_host for a virtual irq.  To allow us to make space
versus time tradeoffs, replace this usage with an assertive
virq_is_host that confirms or denies the irq is associated with the
given irq_host.

Signed-off-by: Milton Miller <miltonm@bga.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/axon_msi: Validate msi irq via chip_data
Milton Miller [Tue, 10 May 2011 19:30:33 +0000 (19:30 +0000)]
powerpc/axon_msi: Validate msi irq via chip_data

Instead of checking for rogue msi numbers via the irq_map host field
set the chip_data to h.host_data (which is the msic struct pointer)
at map and compare it in get_irq.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/spider-pic: Get pic from chip_data instead of irq_map
Milton Miller [Tue, 10 May 2011 19:30:29 +0000 (19:30 +0000)]
powerpc/spider-pic: Get pic from chip_data instead of irq_map

Building on Grant's efforts to remove the irq_map array, this patch
moves spider-pics use of virq_to_host() to use irq_data_get_chip_data
and sets the irq chip data in the map call, like most other interrupt
controllers in powerpc.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove irq_host_ops->remap hook
Milton Miller [Tue, 10 May 2011 19:30:26 +0000 (19:30 +0000)]
powerpc: Remove irq_host_ops->remap hook

It was called from irq_create_mapping if that was called for a host
and hwirq that was previously mapped, "to update the flags".  But the
only implementation was in beat_interrupt and all it did was repeat a
hypervisor call without error checking that was performed with error
checking at the beginning of the map hook.  In addition, the comment on
the beat remap hook says it will only called once for a given mapping,
which would apply to map not remap.

All flags should be known by the time the match hook is called, before
we call the map hook.  Removing this mostly unused hook will simpify
the requirements of irq_domain concept.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/psurge: Create a irq_host for secondary cpus
Milton Miller [Tue, 10 May 2011 19:30:22 +0000 (19:30 +0000)]
powerpc/psurge: Create a irq_host for secondary cpus

Create a dummy irq_host using the generic dummy irq chip for the secondary
cpus to use.  Create a direct irq mapping for the ipi and register the
ipi action handler against it.  If for some unlikely reason part of this
fails then don't detect the secondary cpus.

This removes another instance of NO_IRQ_IGNORE, records the ipi stats
for the secondary cpus, and runs the ipi on the interrupt stack.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mpc62xx_pic: Fix get_irq handling of NO_IRQ
Milton Miller [Tue, 10 May 2011 19:30:18 +0000 (19:30 +0000)]
powerpc/mpc62xx_pic: Fix get_irq handling of NO_IRQ

If none of irq category bits were set mpc52xx_get_irq() would pass
NO_IRQ_IGNORE (-1) to irq_linear_revmap, which does an unsigned compare
and declares the interrupt above the linear map range.  It then punts
to irq_find_mapping, which performs a linear search of all irqs,
which will likely miss and only then return NO_IRQ.

If no status bit is set, then we should return NO_IRQ directly.
The interrupt should not be suppressed from spurious counting, in fact
that is the definition of supurious.

Signed-off-by: Milton Miller <miltonm@bga.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mpc5121_ads_cpld: Remove use of NO_IRQ_IGNORE
Milton Miller [Tue, 10 May 2011 19:30:15 +0000 (19:30 +0000)]
powerpc/mpc5121_ads_cpld: Remove use of NO_IRQ_IGNORE

As NO_IRQ_IGNORE is only used between the static function cpld_pic_get_irq
and its caller cpld_pic_cascade, and cpld_pic_cascade only uses it to
suppress calling handle_generic_irq, we can change these uses to NO_IRQ
and remove the extra tests and pathlength in cpld_pic_cascade.

Signed-off-by: Milton Miller <miltonm@bga.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/fsl_msi: Use chip_data not handler_data
Milton Miller [Tue, 10 May 2011 19:30:11 +0000 (19:30 +0000)]
powerpc/fsl_msi: Use chip_data not handler_data

handler_data should be reserved for flow handlers on the dependent
irq, not consumed by the parent irq code that is part of the irq_chip
code.  The msi_data pointer was already set in msidesc->irqhost->hostdata
and being copied to irq_data->chipdata in the msidesc->irqhost->map()
method called via create_irq_mapping, so we can obtain the pointer
from there and free the instance it in teardown_msi_irqs.

Also remove the unnecessary cast of irq_get_handler_data in the
cascade handler, which is the demux flow handler of the parent
msi interrupt.  (This is the expected usage for handler_data).

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/fsl_msi: Don't abuse platform_data for driver_data
Milton Miller [Tue, 10 May 2011 19:30:07 +0000 (19:30 +0000)]
powerpc/fsl_msi: Don't abuse platform_data for driver_data

The msi platform device driver was abusing dev.platform_data for its
platform_driver_data.  Use the correct pointer for storage.

Platform_data is supposed to be for platforms to communicate to drivers
parameters that are not otherwise discoverable.  Its lifetime matches
the platform_device not the platform device driver.  It is generally
not needed for drivers that only support systems with device trees.

Signed-off-by: Milton Miller <miltonm@bga.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove i8259 irq_host_ops->unmap
Milton Miller [Tue, 10 May 2011 19:30:04 +0000 (19:30 +0000)]
powerpc: Remove i8259 irq_host_ops->unmap

It was never called because the host is always IRQ_HOST_MAP_LEGACY.

And what it purported to do was mask the interrupt (which will already
have happend if we shutdown the interrupt), then synchronise_irq and
clear the chip pointer, both of which will have been be done by the
caller were we to call unmap on a legacy irq.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove trival irq_host_ops.unmap
Milton Miller [Tue, 10 May 2011 19:30:00 +0000 (19:30 +0000)]
powerpc: Remove trival irq_host_ops.unmap

These all just clear chip or chipdata fields, which will be done
by the generic code when we call irq_free_descs.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Return early if irq_host lookup type is wrong
Milton Miller [Tue, 10 May 2011 19:29:57 +0000 (19:29 +0000)]
powerpc: Return early if irq_host lookup type is wrong

If for some reason the code incrorectly calls the wrong function to
manage the revmap, not only should we warn, we should take action.
However, in the paths we expect to be taken every delivered interrupt
change to WARN_ON_ONCE.  Use the if (WARN_ON(x)) format to get the
unlikely for free.

Signed-off-by: Milton Miller <miltonm@bga.com>
Reviewed-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Radix trees are available before init_IRQ
Milton Miller [Tue, 10 May 2011 19:29:53 +0000 (19:29 +0000)]
powerpc: Radix trees are available before init_IRQ

Since the generic irq code uses a radix tree for sparse interrupts,
the initcall ordering has been changed to initialize radix trees before
irqs.   We no longer need to defer creating revmap radix trees to the
arch_initcall irq_late_init.

Also, the kmem caches are allocated so we don't need to use
zalloc_maybe_bootmem.

Signed-off-by: Milton Miller <miltonm@bga.com>
Reviewed-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/xics: Cleanup xics_host_map and ipi
Milton Miller [Tue, 10 May 2011 19:29:49 +0000 (19:29 +0000)]
powerpc/xics: Cleanup xics_host_map and ipi

Since we already have a special case in map to set the ipi handler, use
the desired flow.

If we don't find an ics to handle the interrupt complain instead of
returning 0 without having set a chip or handler.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Use bytes instead of bitops in smp ipi multiplexing
Milton Miller [Tue, 10 May 2011 19:29:46 +0000 (19:29 +0000)]
powerpc: Use bytes instead of bitops in smp ipi multiplexing

Since there are only 4 messages, we can replace the atomic bit set
(which uses atomic load reserve and store conditional sequence) with
a byte stores to seperate bytes.  We still have to perform a load
reserve and store conditional sequence to avoid loosing messages on
reception but we can do that with a single call to xchg.

The do {} while and __BIG_ENDIAN specific mask testing was chosen by
looking at the generated asm code.  On gcc-4.4, the bit masking becomes
a simple bit mask and test of the register returned from xchg without
storing and loading the value to the stack like attempts with a union
of bytes and an int (or worse, loading single bit constants from the
constant pool into non-voliatle registers that had to be preseved on
the stack).  The do {} while avoids an unconditional branch to the
end of the loop to test the entry / repeat condition of a while loop
and instead optimises for the expected single iteration of the loop.

We have a full mb() at the beginning to cover ordering between send,
ipi, and receive so we can use xchg_local and forgo the further
acquire and release barriers of xchg.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Add kconfig for muxed smp ipi support
Milton Miller [Tue, 10 May 2011 19:29:42 +0000 (19:29 +0000)]
powerpc: Add kconfig for muxed smp ipi support

Compile the new smp ipi mux and demux code only if a platform
will make use of it.  The new config is selected as required.

The new cause_ipi smp op is only available conditionally to point out
configs where the select is required; this makes setting the op an
immediate fail instead of a deferred unresolved symbol at link.

This also creates a new config for power surge powermac upgrade support
that can be disabled in expert mode but is default on.

I also removed the depends / default y on CONFIG_XICS since it is selected
by PSERIES.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Consolidate ipi message mux and demux
Milton Miller [Tue, 10 May 2011 19:29:39 +0000 (19:29 +0000)]
powerpc: Consolidate ipi message mux and demux

Consolidate the mux and demux of ipi messages into smp.c and call
a new smp_ops callback to actually trigger the ipi.

The powerpc architecture code is optimised for having 4 distinct
ipi triggers, which are mapped to 4 distinct messages (ipi many, ipi
single, scheduler ipi, and enter debugger).  However, several interrupt
controllers only provide a single software triggered interrupt that
can be delivered to each cpu.  To resolve this limitation, each smp_ops
implementation created a per-cpu variable that is manipulated with atomic
bitops.  Since these lines will be contended they are optimialy marked as
shared_aligned and take a full cache line for each cpu.  Distro kernels
may have 2 or 3 of these in their config, each taking per-cpu space
even though at most one will be in use.

This consolidation removes smp_message_recv and replaces the single call
actions cases with direct calls from the common message recognition loop.
The complicated debugger ipi case with its muxed crash handling code is
moved to debug_ipi_action which is now called from the demux code (instead
of the multi-message action calling smp_message_recv).

I put a call to reschedule_action to increase the likelyhood of correctly
merging the anticipated scheduler_ipi() hook coming from the scheduler
tree; that single required call can be inlined later.

The actual message decode is a copy of the old pseries xics code with its
memory barriers and cache line spacing, augmented with a per-cpu unsigned
long based on the book-e doorbell code.  The optional data is set via a
callback from the implementation and is passed to the new cause-ipi hook
along with the logical cpu number.  While currently only the doorbell
implemntation uses this data it should be almost zero cost to retrieve and
pass it -- it adds a single register load for the argument from the same
cache line to which we just completed a store and the register is dead
on return from the call.  I extended the data element from unsigned int
to unsigned long in case some other code wanted to associate a pointer.

The doorbell check_self is replaced by a call to smp_muxed_ipi_resend,
conditioned on the CPU_DBELL feature.  The ifdef guard could be relaxed
to CONFIG_SMP but I left it with BOOKE for now.

Also, the doorbell interrupt vector for book-e was not calling irq_enter
and irq_exit, which throws off cpu accounting and causes code to not
realize it is running in interrupt context.  Add the missing calls.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Move smp_ops_t from machdep.h to smp.h
Milton Miller [Tue, 10 May 2011 19:29:35 +0000 (19:29 +0000)]
powerpc: Move smp_ops_t from machdep.h to smp.h

I can't see any reason these functions are needed by machdep.h
and they are all hidden by CONFIG_SMP with no UP alternative.

Also move the declarations for the fallback timebase ops, which
are used to fill in the smp ops.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove stubbed beat smp support
Milton Miller [Tue, 10 May 2011 19:29:28 +0000 (19:29 +0000)]
powerpc: Remove stubbed beat smp support

I have no idea if the beat hypervisor supports multiple cpus in
a partition, but the code has not been touched since these stubs
were added in February of 2007 except to move them in April of 2008.
These are stubs: start_cpu always returns fail (which is dropped),
the message passing and reciving are empty functions, and the top
of file comment says "Incomplete".

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove alloc_maybe_bootmem for zalloc version
Milton Miller [Tue, 10 May 2011 19:29:24 +0000 (19:29 +0000)]
powerpc: Remove alloc_maybe_bootmem for zalloc version

Replace all remaining callers of alloc_maybe_bootmem with
zalloc_maybe_bootmem.   The callsite in pci_dn is followed with a
memset to clear the memory, and not zeroing at the other callsites
in the celleb fake pci code could lead to following uninitialized
memory as pointers or even freeing said pointers on error paths.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove powermac/pic.h
Milton Miller [Tue, 10 May 2011 19:29:20 +0000 (19:29 +0000)]
powerpc: Remove powermac/pic.h

Its unused, and of the three declarations, one is duplicated in pmac.h,
the second is static and the third is renamed and static.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mpic: Simplify ipi cpu mask handling
Milton Miller [Tue, 10 May 2011 19:29:17 +0000 (19:29 +0000)]
powerpc/mpic: Simplify ipi cpu mask handling

Now that MSG_ALL and MSG_ALL_BUT_SELF have been eliminated,
smp_mpic_mesage_pass no longer needs to lookup the cpumask just to
have mpic_send_ipi extract part of it and recode it in a NR_CPUS loop
by mpic_physmask.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove checks for MSG_ALL and MSG_ALL_BUT_SELF
Milton Miller [Tue, 10 May 2011 19:29:10 +0000 (19:29 +0000)]
powerpc: Remove checks for MSG_ALL and MSG_ALL_BUT_SELF

Now that smp_ops->smp_message_pass is always called with an (online) cpu
number for the target remove the checks for MSG_ALL and MSG_ALL_BUT_SELF.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove call sites of MSG_ALL_BUT_SELF
Milton Miller [Tue, 10 May 2011 19:29:06 +0000 (19:29 +0000)]
powerpc: Remove call sites of MSG_ALL_BUT_SELF

The only user of MSG_ALL_BUT_SELF in the whole kernel tree is powerpc,
and it only uses it to start the debugger. Both debuggers always call
smp_send_debugger_break with MSG_ALL_BUT_SELF, and only mpic can do
anything more optimal than a loop over all online cpus, but all message
passing implementations have to code for this special delivery target.

Convert smp_send_debugger_break to take void and loop calling the smp_ops
message_pass function for each of the other cpus in the online cpumask.

Use raw_smp_processor_id() because we are either entering the debugger
or trying to start kdump and the additional warning it not useful were
it to trigger.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mpic: Break cpumask abstraction earlier
Milton Miller [Tue, 10 May 2011 19:29:02 +0000 (19:29 +0000)]
powerpc/mpic: Break cpumask abstraction earlier

mpic_set_affinity is allocating and freeing a cpumask var even though
it was breaking the cpumask abstraction when passing the mask to
mpic_physmask.  It also didn't have any check for allocatin failure.

Break the cpumask abstraction earlier and use simple bitwise and of the
bits from the mask with the bits of cpu_online_mask.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mpic: Limit NR_CPUS loop to 32 bit
Milton Miller [Tue, 10 May 2011 19:28:59 +0000 (19:28 +0000)]
powerpc/mpic: Limit NR_CPUS loop to 32 bit

mpic_physmask was looping NR_CPUS times over a mask that was passed as
a u32. Since mpic is architecturaly limited to 32 physical cpus, clamp
the logical cpus to 32 when compiling (we could also clamp at runtime
to nr_cpu_ids).

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Call no-longer static setup_nr_cpu_ids instead of replicating it
Milton Miller [Tue, 10 May 2011 19:28:55 +0000 (19:28 +0000)]
powerpc: Call no-longer static setup_nr_cpu_ids instead of replicating it

c1854e00727f50f7ac99e98d26ece04c087ef785 (powerpc: Set nr_cpu_ids early
and use it to free PACAs) copied the formerly static setup_nr_cpu_ids
from init/main.c but 34db18a054c600b6f81787165669dc572fe4de25 (smp:
move smp setup functions to kernel/smp.c) moved it to kernel/smp.c
with a declaration in include/linux/smp.h, so we can call it instead of
replicating it.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Use nr_cpu_ids in initial paca allocation
Milton Miller [Tue, 10 May 2011 19:28:52 +0000 (19:28 +0000)]
powerpc: Use nr_cpu_ids in initial paca allocation

Now that we never set a cpu above nr_cpu_ids possible we can
limit our initial paca allocation to nr_cpu_ids.  We can then
clamp the number of cpus in platforms/iseries/setup.c.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Respect nr_cpu_ids when calling set_cpu_possible and set_cpu_present
Milton Miller [Tue, 10 May 2011 19:28:48 +0000 (19:28 +0000)]
powerpc: Respect nr_cpu_ids when calling set_cpu_possible and set_cpu_present

We should not set cpus above nr_cpu_ids to possible.  While we
will trigger a warning with CONFIG_CPUMASK_DEBUG, even then the mask
initializers will set the bits beyond what the iterators check and cause
nr_cpu_ids to increase.

Respecting nr_cpu_ids during setup will allow us to use it in our initial
paca allocation.  It can be reduced from NR_CPUS by the existing early param
nr_cpus=, which was added in 2b633e3fac5efada088b57d31e65401f22bcc18f (smp:
Use nr_cpus= to set nr_cpu_ids early).  We already call parse_early_parms
between finding the command line and allocating the pacas.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/iseries: Cleanup and fix secondary startup
Milton Miller [Tue, 10 May 2011 19:28:44 +0000 (19:28 +0000)]
powerpc/iseries: Cleanup and fix secondary startup

9cb82f2f4692293a27c578c3038518ce4477de72 (Make iSeries spin on
__secondary_hold_spinloop, like pSeries) added a load of current_set
but this load was repeated later and we don't even have the paca yet.
It also checked __secondary_hold_spinloop with a 32 bit compare instead
of a 64 bit compare.

b6f6b98a4e91fcf31db7de54c3aa86252fc6fb5f (Don't spin on sync instruction
at boot time) missed the copy of the startup code in iseries.

1426d5a3bd07589534286375998c0c8c6fdc5260 (Dynamically allocate pacas)
doesn't allow for pacas to be less than lppacas and recalculated the paca
location from the cpu id in r0 every time through the secondary loop.

Various revisions over time made the comments on conditional branches
confusing with respect to being a hold loop or forward progress

Mostly in-order description of the changes:

Replicate the few lines of code saved by the ugly scoped ifdef CONFIG_SMP
in the secondary loop between yielding on UP and marking time with the
hypervisor on SMP.  Always compile the iseries_secondary_yield loop and
use it if the cpu id is above nr_cpu_ids.  Change all forward progress
paths to be forward branches to the next numerical label.  Assign a
label to all loops.  Move all sync instructions from the loops to the
forward progress path.  Wait to load current_set until paca is set to go.
Move the iseries_secondary_smp_loop label to cover the whole spin loop.
Add HMT_MEDIUM when we make forward progress.

Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/kdump64: Don't reference freed memory as pacas
Milton Miller [Tue, 10 May 2011 19:28:41 +0000 (19:28 +0000)]
powerpc/kdump64: Don't reference freed memory as pacas

Starting with 1426d5a3bd07589534286375998c0c8c6fdc5260 (powerpc:
Dynamically allocate pacas) the space for pacas beyond cpu_possible
is freed, but we failed to update the loop in crash.c.

Since c1854e00727f50f7ac99e98d26ece04c087ef785 (powerpc: Set nr_cpu_ids
early and use it to free PACAs) the number of pacas allocated is
always nr_cpu_ids.

Signed-off-by: Milton Miller <miltonm@bga.com>
Cc: <stable@kernel.org> # .34.x
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Don't search for paca in freed memory
Milton Miller [Tue, 10 May 2011 19:28:37 +0000 (19:28 +0000)]
powerpc: Don't search for paca in freed memory

Starting with 1426d5a3bd07589534286375998c0c8c6fdc5260 (powerpc:
Dynamically allocate pacas) we free the memory for pacas beyond
cpu_possible, but we failed to update the loop the secondary cpus use
to find their paca.  If the system has running cpu threads for which
the kernel did not allocate a paca for they will search the memory that
was freed.  For instance this could happen when the device tree for
a kdump kernel was not updated after a cpu hotplug, or the kernel is
running with more cpus than the kernel was configured.

Since c1854e00727f50f7ac99e98d26ece04c087ef785 (powerpc: Set nr_cpu_ids
early and use it to free PACAs) we set nr_cpu_ids before telling the
cpus to advance, so use that to limit the search.

We can't reference nr_cpu_ids without CONFIG_SMP because it is defined
as 1 instead of a memory location, but any extra threads should be sent
to kexec_wait in that case anyways, so make that explicit and remove
the search loop for UP.

Note to stable: The fix also requires
c1854e00727f50f7ac99e98d26ece04c087ef785 (powerpc: Set
nr_cpu_ids early and use it to free PACAs) to function.  Also
9d07bc841c9779b4d7902e417f4e509996ce805d (Properly handshake CPUs going
out of boot spin loop) affects the second chunk, specifically the branch
target was 3b before and is 4b after that patch, and there was a blank
line before the #ifdef CONFIG_SMP that was removed

Cc: <stable@kernel.org> # .34.x: c1854e0072 powerpc: Set nr_cpu_ids early
Cc: <stable@kernel.org> # .34.x
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/kexec: Fix memory corruption from unallocated slaves
Milton Miller [Tue, 10 May 2011 19:28:33 +0000 (19:28 +0000)]
powerpc/kexec: Fix memory corruption from unallocated slaves

Commit 1fc711f7ffb01089efc58042cfdbac8573d1b59a (powerpc/kexec: Fix race
in kexec shutdown) moved the write to signal the cpu had exited the kernel
from before the transition to real mode in kexec_smp_wait to kexec_wait.

Unfornately it missed that kexec_wait is used both by cpus leaving the
kernel and by secondary slave cpus that were not allocated a paca for
what ever reason -- they could be beyond nr_cpus or not described in
the current device tree for whatever reason (for example, kexec-load
was not refreshed after a cpu hotplug operation).  Cpus coming through
that path they will write to paca[NR_CPUS] which is beyond the space
allocated for the paca data and overwrite memory not allocated to pacas
but very likely still real mode accessable).

Move the write back to kexec_smp_wait, which is used only by cpus that
found their paca, but after the transition to real mode.

Signed-off-by: Milton Miller <miltonm@bga.com>
Cc: <stable@kernel.org> # (1fc711f was backported to 2.6.32)
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries: Print corrupt r3 in FWNMI code
Anton Blanchard [Tue, 10 May 2011 13:34:03 +0000 (13:34 +0000)]
powerpc/pseries: Print corrupt r3 in FWNMI code

I have a report of an FWNMI with an r3 value that we think is
corrupt, but since we don't print r3 we have no idea what was
wrong with it.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopseries/iommu: Restore iommu table pointer when restoring iommu ops
Nishanth Aravamudan [Mon, 9 May 2011 12:58:03 +0000 (12:58 +0000)]
pseries/iommu: Restore iommu table pointer when restoring iommu ops

When we swtich to direct dma ops, we set the dma data union to have the
dma offset.  When we switch back to iommu table ops because of a later
dma_set_mask, we need to restore the iommu table pointer. Without this
change, crashes have been observed on kexec where (for reasons still
being investigated) we fall back to a 32-bit dma mask on a particular
device and then panic because the table pointer is not valid.

The easiset way to find this value is to call
pci_dma_dev_setup_pSeriesLP which will search up the pci tree until it
finds the node with the table.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Milton Miller <miltonm@bga.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove ioremap_flags
Anton Blanchard [Sun, 8 May 2011 21:43:47 +0000 (21:43 +0000)]
powerpc: Remove ioremap_flags

We have a confusing number of ioremap functions. Make things just a
bit simpler by merging ioremap_flags and ioremap_prot.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Add ioremap_wc
Anton Blanchard [Sun, 8 May 2011 21:41:59 +0000 (21:41 +0000)]
powerpc: Add ioremap_wc

Add ioremap_wc so drivers can request write combining on kernel
mappings.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Improve scheduling of system call entry instructions
Anton Blanchard [Sun, 8 May 2011 21:36:44 +0000 (21:36 +0000)]
powerpc: Improve scheduling of system call entry instructions

After looking at our system call path, Mary Brown suggested that we
should put all mfspr SRR* instructions before any mtspr SRR*.

To test this I used a very simple null syscall (actually getppid)
testcase at http://ozlabs.org/~anton/junkcode/null_syscall.c

I tested with the following changes against the pseries_defconfig:

CONFIG_VIRT_CPU_ACCOUNTING=n
CONFIG_AUDIT=n

to remove the overhead of virtual CPU accounting and syscall
auditing.

POWER6:
baseline:       mean = 757.2 cycles       sd = 2.108
modified:       mean = 759.1 cycles       sd = 2.020

POWER7:
baseline:       mean = 411.4 cycles       sd = 0.138
modified:       mean = 404.1 cycles       sd = 0.109

So we have 1.77% improvement on POWER7 which looks significant. The
POWER6 suggest a 0.25% slowdown, but the results are within 1
standard deviation and may be in the noise.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove static branch hint in giveup_altivec
Anton Blanchard [Sun, 8 May 2011 21:20:19 +0000 (21:20 +0000)]
powerpc: Remove static branch hint in giveup_altivec

A static branch hint will override dynamic branch prediction on
recent POWER CPUs. Since we are about to use more altivec in the
kernel remove the static hint in giveup_altivec that assumes
a userspace task is using altivec.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Simplify 4k/64k copy_page logic
Anton Blanchard [Sun, 8 May 2011 21:18:38 +0000 (21:18 +0000)]
powerpc: Simplify 4k/64k copy_page logic

To make it easier to add optimised versions of copy_page, remove
the 4kB loop for 64kB pages and just do all the work in copy_page.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries: Enable iSCSI support for a number of cards
Anton Blanchard [Sun, 8 May 2011 13:19:30 +0000 (13:19 +0000)]
powerpc/pseries: Enable iSCSI support for a number of cards

Enable iSCSI support for a number of cards. We had the base
networking devices enabled but forgot to enable iSCSI.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/pseries: Enable Emulex and Qlogic 10Gbit cards
Anton Blanchard [Sun, 8 May 2011 13:18:27 +0000 (13:18 +0000)]
powerpc/pseries: Enable Emulex and Qlogic 10Gbit cards

Enable the Qlogic and Emulex 10Gbit adapters.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/mm: Fix compiler warning in pgtable-ppc64.h [-Wunused-but-set-variable]
Stratos Psomadakis [Sat, 7 May 2011 04:11:31 +0000 (04:11 +0000)]
powerpc/mm: Fix compiler warning in pgtable-ppc64.h [-Wunused-but-set-variable]

The variable 'old' is set but not used in the wrprotect functions in
arch/powerpc/include/asm/pgtable-ppc64.h, which can trigger a compiler warning.

Remove the variable, since it's not used anyway.

Signed-off-by: Stratos Psomadakis <psomas@ece.ntua.gr>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Ensure dtl buffers do not cross 4k boundary
Nishanth Aravamudan [Wed, 4 May 2011 12:54:16 +0000 (12:54 +0000)]
powerpc: Ensure dtl buffers do not cross 4k boundary

Future releases of fimrware will enforce a requirement that DTL buffers
do not cross a 4k boundary. Commit
127493d5dc73589cbe00ea5ec8357cc2a4c0d82a satisfies this requirement for
CONFIG_VIRT_CPU_ACCOUNTING=y kernels, but if !CONFIG_VIRT_CPU_ACCOUNTING
&& CONFIG_DTL=y, the current code will fail at dtl registration time.
Fix this by making the kmem cache from
127493d5dc73589cbe00ea5ec8357cc2a4c0d82a visible outside of setup.c and
using the same cache in both dtl.c and setup.c. This requires a bit of
reorganization to ensure ordering of the kmem cache and buffer
allocations.

Note: Since firmware now limits the size of the buffer, I made
dtl_buf_entries read-only in debugfs.

Tested with upcoming firmware with the 4 combinations of
CONFIG_VIRT_CPU_ACCOUNTING and CONFIG_DTL.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Fix kexec with dynamic dma windows
Nishanth Aravamudan [Fri, 6 May 2011 13:27:30 +0000 (13:27 +0000)]
powerpc: Fix kexec with dynamic dma windows

When we kexec we look for a particular property added by the first
kernel, "linux,direct64-ddr-window-info", per-device where we already
have set up dynamic dma windows. The current code, though, wasn't
initializing the size of this property and thus when we kexec'd, we
would find the property but read uninitialized memory resulting in
garbage ddw values for the kexec'd kernel and panics. Fix this by
setting the size at enable_ddw() time and ensuring that the size of the
found property is valid at dupe_ddw_if_kexec() time.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Use the deterministic mode of ar
Michal Marek [Thu, 5 May 2011 05:22:55 +0000 (05:22 +0000)]
powerpc: Use the deterministic mode of ar

Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Remove unused config in the Makefile
Justin Mattock [Tue, 5 Apr 2011 06:58:22 +0000 (06:58 +0000)]
powerpc: Remove unused config in the Makefile

The patch below removes an unused config variable found by using a kernel
cleanup script.
Note: I did try to cross compile these but hit erros while doing so..
(gcc is not setup to cross compile) and am unsure if anymore needs to be done.
Please have a look if/when anybody has free time.

Signed-off-by: Justin P. Mattock <justinmattock@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc: Call gzip with -n
Michal Marek [Tue, 5 Apr 2011 04:58:50 +0000 (04:58 +0000)]
powerpc: Call gzip with -n

The timestamps recorded in the .gz files add no value.

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Michal Marek <mmarek@suse.cz>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/4xx: Fix regression in SMP on 476
kerstin jonsson [Tue, 17 May 2011 23:57:11 +0000 (23:57 +0000)]
powerpc/4xx: Fix regression in SMP on 476

commit c56e58537d504706954a06570b4034c04e5b7500 breaks SMP support in PPC_47x chip.
 secondary_ti must be set to current thread info before callin kick_cpu or else
 start_secondary_47x will jump into void when trying to return to c-code.
 In the current setup secondary_ti is initialized before the CPU idle task is started
 and only the boot core will start. I am not sure this is the correct solution, but it
 makes SMP possible in my chip.
 Note! The HOTPLUG support probably need some fixing to, There is no trampoline code
 available in head_44x.S - start_secondary_resume?

Signed-off-by: Kerstin Jonsson <kerstin.jonsson@ericsson.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/kexec: Fix build failure on 32-bit SMP
Ben Hutchings [Sun, 24 Apr 2011 15:04:31 +0000 (15:04 +0000)]
powerpc/kexec: Fix build failure on 32-bit SMP

Commit b987812b3fcaf70fdf0037589e5d2f5f2453e6ce left
crash_kexec_wait_realmode() undefined for UP.

Commit 7c7a81b53e581d727d069cc45df5510516faac31 defined it for UP but
left it undefined for 32-bit SMP.

Seems like people are getting confused by nested #ifdef's, so move the
definitions of crash_kexec_wait_realmode() after the #ifdef CONFIG_SMP
section.

Compile-tested with 32-bit UP, 32-bit SMP and 64-bit SMP configurations.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Tested-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agopowerpc/smp: Make start_secondary_resume available to all CPU variants
Benjamin Herrenschmidt [Thu, 19 May 2011 03:07:12 +0000 (13:07 +1000)]
powerpc/smp: Make start_secondary_resume available to all CPU variants

This should fix SMP & Hotplug builds on FSL BookE and 476

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
13 years agoMerge branch 'devicetree/merge' of git://git.secretlab.ca/git/linux-2.6
Linus Torvalds [Wed, 18 May 2011 20:25:57 +0000 (13:25 -0700)]
Merge branch 'devicetree/merge' of git://git.secretlab.ca/git/linux-2.6

* 'devicetree/merge' of git://git.secretlab.ca/git/linux-2.6:
  drivercore: revert addition of of_match to struct device
  of: fix race when matching drivers

13 years agoMerge branch 'upstream' of git://git.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Wed, 18 May 2011 20:21:43 +0000 (13:21 -0700)]
Merge branch 'upstream' of git://git.linux-mips.org/upstream-linus

* 'upstream' of git://git.linux-mips.org/pub/scm/upstream-linus:
  MIPS: Kludge IP27 build for 2.6.39.
  MIPS: AR7: Fix GPIO register size for Titan variant.
  MIPS: Fix duplicate invocation of notify_die.
  MIPS: RB532: Fix iomap resource size miscalculation.

13 years agodrivercore: revert addition of of_match to struct device
Grant Likely [Wed, 18 May 2011 17:19:24 +0000 (11:19 -0600)]
drivercore: revert addition of of_match to struct device

Commit b826291c, "drivercore/dt: add a match table pointer to struct
device" added an of_match pointer to struct device to cache the
of_match_table entry discovered at driver match time.  This was unsafe
because matching is not an atomic operation with probing a driver.  If
two or more drivers are attempted to be matched to a driver at the
same time, then the cached matching entry pointer could get
overwritten.

This patch reverts the of_match cache pointer and reworks all users to
call of_match_device() directly instead.

Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
13 years agoof: fix race when matching drivers
Milton Miller [Wed, 18 May 2011 15:27:39 +0000 (10:27 -0500)]
of: fix race when matching drivers

If two drivers are probing devices at the same time, both will write
their match table result to the dev->of_match cache at the same time.

Only write the result if the device matches.

In a thread titled "SBus devices sometimes detected, sometimes not",
Meelis reported his SBus hme was not detected about 50% of the time.
From the debug suggested by Grant it was obvious another driver matched
some devices between the call to match the hme and the hme discovery
failling.

Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Milton Miller <miltonm@bga.com>
[grant.likely: modified to only call of_match_device() once]
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
13 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Wed, 18 May 2011 13:49:02 +0000 (06:49 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: don't delay blk_run_queue_async
  scsi: remove performance regression due to async queue run
  blk-throttle: Use task_subsys_state() to determine a task's blkio_cgroup
  block: rescan partitions on invalidated devices on -ENOMEDIA too
  cdrom: always check_disk_change() on open
  block: unexport DISK_EVENT_MEDIA_CHANGE for legacy/fringe drivers

13 years agoMIPS: Kludge IP27 build for 2.6.39.
Ralf Baechle [Wed, 18 May 2011 12:14:36 +0000 (13:14 +0100)]
MIPS: Kludge IP27 build for 2.6.39.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
13 years agoMIPS: AR7: Fix GPIO register size for Titan variant.
Florian Fainelli [Fri, 13 May 2011 15:41:21 +0000 (17:41 +0200)]
MIPS: AR7: Fix GPIO register size for Titan variant.

The 'size' variable contains the correct register size for both AR7
and Titan, but we never used it to ioremap the correct register size.
This problem only shows up on Titan.

[ralf@linux-mips.org: Fixed the fix.  The original patch as in patchwork
recognizes the problem correctly then fails to fix it ...]

Reported-by: Alexander Clouter <alex@digriz.org.uk>
Signed-off-by: Florian Fainelli <florian@openwrt.org>
Patchwork: https://patchwork.linux-mips.org/patch/2380/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
13 years agoMIPS: Fix duplicate invocation of notify_die.
Ralf Baechle [Fri, 13 May 2011 09:33:28 +0000 (10:33 +0100)]
MIPS: Fix duplicate invocation of notify_die.

Initial patch by Yury Polyanskiy <ypolyans@princeton.edu>.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Patchwork: https://patchwork.linux-mips.org/patch/2373/

13 years agoMIPS: RB532: Fix iomap resource size miscalculation.
Ralf Baechle [Thu, 12 May 2011 12:55:48 +0000 (13:55 +0100)]
MIPS: RB532: Fix iomap resource size miscalculation.

This is the MIPS portion of Joe Perches <joe@perches.com>'s
https://patchwork.linux-mips.org/patch/2172/ which seems to have been
lost in time and space.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
13 years agoblock: don't delay blk_run_queue_async
Shaohua Li [Wed, 18 May 2011 09:22:43 +0000 (11:22 +0200)]
block: don't delay blk_run_queue_async

Let's check a scenario:
1. blk_delay_queue(q, SCSI_QUEUE_DELAY);
2. blk_run_queue_async();
the second one will became a noop, because q->delay_work already has
WORK_STRUCT_PENDING_BIT set, so the delayed work will still run after
SCSI_QUEUE_DELAY. But blk_run_queue_async actually hopes the delayed
work runs immediately.

Fix this by doing a cancel on potentially pending delayed work
before queuing an immediate run of the workqueue.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
13 years agoMerge branch 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Wed, 18 May 2011 10:16:38 +0000 (03:16 -0700)]
Merge branch 'v4l_for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
  [media] V4L: soc-camera: regression fix: calculate .sizeimage in soc_camera.c
  [media] v4l2-subdev: fix broken subdev control enumeration
  [media] Fix cx88 remote control input
  [media] v4l: Release module if subdev registration fails

13 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 18 May 2011 10:14:34 +0000 (03:14 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, AMD: Fix ARAT feature setting again
  Revert "x86, AMD: Fix APIC timer erratum 400 affecting K8 Rev.A-E processors"
  x86, apic: Fix spurious error interrupts triggering on all non-boot APs
  x86, mce, AMD: Fix leaving freed data in a list
  x86: Fix UV BAU for non-consecutive nasids
  x86, UV: Fix NMI handler for UV platforms

13 years agoMerge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 18 May 2011 10:13:46 +0000 (03:13 -0700)]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  perf evlist: Fix per thread mmap setup
  perf tools: Honour the cpu list parameter when also monitoring a thread list
  kprobes, x86: Disable irqs during optimized callback

13 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Wed, 18 May 2011 10:13:11 +0000 (03:13 -0700)]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  cifs: fix cifsConvertToUCS() for the mapchars case
  cifs: add fallback in is_path_accessible for old servers

13 years agoprocfs: add stub for proc_mkdir_mode()
Randy Dunlap [Tue, 17 May 2011 22:44:12 +0000 (15:44 -0700)]
procfs: add stub for proc_mkdir_mode()

Provide a stub for proc_mkdir_mode() when CONFIG_PROC_FS is not
enabled, just like the stub for proc_mkdir().

Fixes this linux-next build error:

  drivers/net/wireless/airo.c:4504: error: implicit declaration of function 'proc_mkdir_mode'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "John W. Linville" <linville@tuxdriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>