gart alloc_coherent need to do virtual mapppings only when an
allocated buffer is not DMA-capable for a device.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86's common alloc_coherent (dma_alloc_coherent in dma-mapping.h) sets
up the gfp flag according to the device dma_mask but Calgary doesn't
need it because of virtual mappings. This patch avoids unnecessary low
zone allocation.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Currently, GART IOMMU ingores device's dma_mask when it does virtual
mappings. So it could give a device a virtual address that the device
can't access to.

This patch fixes the above problem.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

As well as discard fake accessed bit and dirty bit of EPT.

Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>

Accesses to CR4 are intercepted even with Nested Paging enabled. But the code
does not check if the guest wants to do a global TLB flush. So this flush gets
lost. This patch adds the check and the flush to svm_set_cr4.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>

This patch introduces a guest TLB flush on every NPF exit in KVM. This fixes
random segfaults and #UD exceptions in the guest seen under some workloads
(e.g. long running compile workloads or tbench). A kernbench run with and
without that fix showed that it has a slowdown lower than 0.5%

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@qumranet.com>

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

When using kdump modifying the e820 map is yielding strange results.

For example starting with

 BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000100 - 0000000000093400 (usable)
 BIOS-e820: 0000000000093400 - 00000000000a0000 (reserved)
 BIOS-e820: 0000000000100000 - 000000003fee0000 (usable)
 BIOS-e820: 000000003fee0000 - 000000003fef3000 (ACPI data)
 BIOS-e820: 000000003fef3000 - 000000003ff80000 (ACPI NVS)
 BIOS-e820: 000000003ff80000 - 0000000040000000 (reserved)
 BIOS-e820: 00000000e0000000 - 00000000f0000000 (reserved)
 BIOS-e820: 00000000fec00000 - 00000000fec10000 (reserved)
 BIOS-e820: 00000000fee00000 - 00000000fee01000 (reserved)
 BIOS-e820: 00000000ff000000 - 0000000100000000 (reserved)

and booting with args

memmap=exactmap memmap=640K@0K memmap=5228K@16384K memmap=125188K@22252K memmap=76K#1047424K memmap=564K#1047500K

resulted in:

 user-defined physical RAM map:
 user: 0000000000000000 - 0000000000093400 (usable)
 user: 0000000000093400 - 00000000000a0000 (reserved)
 user: 0000000000100000 - 000000003fee0000 (usable)
 user: 000000003fee0000 - 000000003fef3000 (ACPI data)
 user: 000000003fef3000 - 000000003ff80000 (ACPI NVS)
 user: 000000003ff80000 - 0000000040000000 (reserved)
 user: 00000000e0000000 - 00000000f0000000 (reserved)
 user: 00000000fec00000 - 00000000fec10000 (reserved)
 user: 00000000fee00000 - 00000000fee01000 (reserved)
 user: 00000000ff000000 - 0000000100000000 (reserved)

But should have resulted in:

 user-defined physical RAM map:
 user: 0000000000000000 - 00000000000a0000 (usable)
 user: 0000000001000000 - 000000000151b000 (usable)
 user: 00000000015bb000 - 0000000008ffc000 (usable)
 user: 000000003fee0000 - 000000003ff80000 (ACPI data)

This is happening because of an improper usage of strcmp() in the
e820 parsing code.  The strcmp() always returns !0 and never resets the
value for e820.nr_map and returns an incorrect user-defined map.

This patch fixes the problem.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

On 32-bit, at least the generic nops are fairly reasonable, but the
default nops for 64-bit really look pretty sad, and the P6 nops really do
look better.

So I would suggest perhaps moving the static P6 nop selection into the
CONFIG_X86_64 thing.

The alternative is to just get rid of that static nop selection, and just
have two cases: 32-bit and 64-bit, and just pick obviously safe cases for
them.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Non real IOMMU implemenations (which doesn't do virtual mappings,
e.g. swiotlb, pci-nommu, etc) need to use proper gfp flags and
dma_mask to allocate pages in their own dma_alloc_coherent()
(allocated page need to be suitable for device's coherent_dma_mask).

This patch makes dma_alloc_coherent do this job so that IOMMUs don't
need to take care of it any more.

Real IOMMU implemenataions can simply ignore the gfp flags.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

We need to use __GFP_DMA for NULL device argument (fallback_dev) with
pci-nommu. It's a hack for ISA (and some old code) so we need to use
GFP_DMA.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

The check to see if dev->dma_mask is NULL in pci-nommu is more
appropriate for dma_alloc_coherent().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Exception stacks are allocated each time a CPU is set online.
But the allocated space is never freed. Thus with one CPU hotplug
offline/online cycle there is a memory leak of 24K (6 pages) for
a CPU.

Fix is to allocate exception stacks only once -- when the CPU is
set online for the first time.

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: akpm@linux-foundation.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

pda->irqstackptr is allocated whenever a CPU is set online.
But it is never freed. This results in a memory leak of 16K
for each CPU offline/online cycle.

Fix is to allocate pda->irqstackptr only once.

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: akpm@linux-foundation.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Using native_pte_val triggers the BUG_ON() in the paravirt_ops
version of pte_flags().

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Acked-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Krzysztof Helt found MTRR is not detected on k6-2

root cause:
	we moved mtrr_bp_init() early for mtrr trimming,
and in early_detect we only read the CPU capability from cpuid,
so some cpu doesn't have that bit in cpuid.

So we need to add early_init_xxxx to preset those bit before mtrr_bp_init
for those earlier cpus.

this patch is for v2.6.27

Reported-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Move early cpu initialization after cpu early get cap so the
early cpu initialization can fix up cpu caps.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

After fixing the u32 thinko I sill had occasional hickups on ATI chipsets
with small deltas. There seems to be a delay between writing the compare
register and the transffer to the internal register which triggers the
interrupt. Reading back the value makes sure, that it hit the internal
match register befor we compare against the counter value.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

We use the HPET only in 32bit mode because:
1) some HPETs are 32bit only
2) on i386 there is no way to read/write the HPET atomic 64bit wide

The HPET code unification done by the "moron of the year" did
not take into account that unsigned long is different on 32 and
64 bit.

This thinko results in a possible endless loop in the clockevents
code, when the return comparison fails due to the 64bit/332bit
unawareness. 

unsigned long cnt = (u32) hpet_read() + delta can wrap over 32bit.
but the final compare will fail and return -ETIME causing endless
loops.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

Use X86_FEATURE_NOPL to determine if it is safe to use P6 NOPs in
alternatives.  Also, replace table and loop with simple if statement.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

The long noops ("NOPL") are supposed to be detected by family >= 6.
Unfortunately, several non-Intel x86 implementations, both hardware
and software, don't obey this dictum.  Instead, probe for NOPL
directly by executing a NOPL instruction and see if we get #UD.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

The CPU feature detection code in the boot code is somewhat minimal,
and doesn't include all possible CPUID words.  In particular, it
doesn't contain the code for CPU feature words 2 (Transmeta),
3 (Linux-specific), 5 (VIA), or 7 (scattered).  Zero them out, so we
can still set those bits as known at compile time; in particular, this
allows creating a Linux-specific NOPL flag and have it required (and
therefore resolvable at compile time) in 64-bit mode.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Some BIOSes (the Intel DG33BU, for example) wrongly claim to have DMAR
when they don't. Avoid the resulting crashes when it doesn't work as
expected.

I'd still be grateful if someone could test it on a DG33BU with the old
BIOS though, since I've killed mine. I tested the DMI version, but not
this one.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

This patch adds the detection of the northbridges in the AMD family 0x11
processors. It also fixes the magic numbers there while changing this code.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

asm/dma-mapping.h guarantees that gart alloc_coherent doesn't get NULL
device argument.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

The minimum reprogramming delta was hardcoded in HPET ticks,
which is stupid as it does not work with faster running HPETs.
The C1E idle patches made this prominent on AMD/RS690 chipsets,
where the HPET runs with 25MHz. Set it to 5us which seems to be
a reasonable value and fixes the problems on the bug reporters
machines. We have a further sanity check now in the clock events,
which increases the delta when it is not sufficient.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br>
Tested-by: Dmitry Nezhevenko <dion@inhex.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

When calibration against PIT fails, the warning that we print is misleading.
In a virtualized environment the VM may get descheduled while calibration
or, the check in PIT calibration may fail due to other virtualization
overheads.

The warning message explicitly assumes that calibration failed due to SMI's
which may not be the case. Change that to something proper.

Signed-off-by: Alok N Kataria <akataria@vmware.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Manually adding "io_delay=0xed" fixes system lockups in ioapic
mode on this machine.

System Information
	Manufacturer: Hewlett-Packard
	Product Name: Presario F700 (KA695EA#ABF)

Base Board Information
	Manufacturer: Quanta
	Product Name: 30D3

Reference:
https://bugzilla.redhat.com/show_bug.cgi?id=459546



Signed-off-by: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>

The TSC calibration function is still very complicated, but this makes
it at least a little bit less so by moving the PIT part out into a
helper function of its own.

Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-of-by: Linus Torvalds <torvalds@linux-foundation.org>

Larry Finger reported at http://lkml.org/lkml/2008/9/1/90

:
An ancient laptop of mine started throwing errors from b43legacy when
I started using 2.6.27 on it. This has been bisected to commit bfc0f594
"x86: merge tsc calibration".

The unification of the TSC code adopted mostly the 64bit code, which
prefers PMTIMER/HPET over the PIT calibration.

Larrys system has an AMD K6 CPU. Such systems are known to have
PMTIMER incarnations which run at double speed. This results in a
miscalibration of the TSC by factor 0.5. So the resulting calibrated
CPU/TSC speed is half of the real CPU speed, which means that the TSC
based delay loop will run half the time it should run. That might
explain why the b43legacy driver went berserk.

On the other hand we know about systems, where the PIT based
calibration results in random crap due to heavy SMI/SMM
disturbance. On those systems the PMTIMER/HPET based calibration logic
with SMI detection shows better results.

According to Alok also virtualized systems suffer from the PIT
calibration method.

The solution is to use a more wreckage aware aproach than the current
either/or decision.

1) reimplement the retry loop which was dropped from the 32bit code
during the merge. It repeats the calibration and selects the lowest
frequency value as this is probably the closest estimate to the real
frequency

2) Monitor the delta of the TSC values in the delay loop which waits
for the PIT counter to reach zero. If the maximum value is
significantly different from the minimum, then we have a pretty safe
indicator that the loop was disturbed by an SMI.

3) keep the pmtimer/hpet reference as a backup solution for systems
where the SMI disturbance is a permanent point of failure for PIT
based calibration

4) do the loop iteration for both methods, record the lowest value and
decide after all iterations finished.

5) Set a clear preference to PIT based calibration when the result
makes sense.

The implementation does the reference calibration based on
HPET/PMTIMER around the delay, which is necessary for the PIT anyway,
but keeps separate TSC values to ensure the "independency" of the
resulting calibration values.

Tested on various 32bit/64bit machines including Geode 266Mhz, AMD K6
(affected machine with a double speed pmtimer which I grabbed out of
the dump), Pentium class machines and AMD/Intel 64 bit boxen.

Bisected-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Breaking lines due to some imaginary problem with a long line length is
often stupid and wrong, but never more so when it splits a string that
is printed out into multiple lines.  This really ended up making it much
harder to find where some error strings were printed out, because a
simple 'grep' didn't work.

I'm sure there is tons more of this particular idiocy hiding in other
places, but this particular case hit me once more last week. So fix it.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This reverts commit a2bd7274.

It wasn't really right to begin with (there's a better fix for the
problem with e820 reservations clashing with PCI BAR's pending), but it
also actually causes more regressions, so it should be reverted even
before the better fix is finalized.

Rafael reports that this commit broke AHCI detection, and thus causes
the kernel to not boot on his quad core test box.

Reported-and-bisected-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Yinghai Lu <yhlu.kernel@gmail.com>
Cc: David Witbrodt <dawitbro@sbcglobal.net>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Enable some option commonly used by testers in defconfig, including
some very common device drivers and network boot support.  defconfig
is still not meant to be a kitchen-sink configuration.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Return the correct return value when the CPUID driver partially
completes a request (we should return the number of bytes actually
read or written, instead of the error code.)

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Return the correct return value when the MSR driver partially
completes a request (we should return the number of bytes actually
read or written, instead of the error code.)

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Propagate error (-ENXIO) from smp_call_function_single() in the CPUID
driver.  This can happen when a CPU is unplugged while the CPUID
driver is open.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

Propagate error (-ENXIO) from smp_call_function_single().  These
errors can happen when a CPU is unplugged while the MSR driver is
open.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>

This fixes a regression that was indirectly caused by commit
1184dc2f ("x86: modify Kconfig to allow
up to 4096 cpus").

Allowing 4k CPU's is not practical at this time, because we still have a
number of places that have several 'cpumask_t's on the stack, and a
4k-bit cpumask is 512 bytes of stack-space for each such variable.  This
literally caused functions like 'smp_call_function_mask' to have a 2.5kB
stack frame, and several functions to have 2kB stackframes.

With an 8kB stack total, smashing the stack was simply much too likely.
At least bugzilla entry

	http://bugzilla.kernel.org/show_bug.cgi?id=11342



was due to this.

The earlier commit to not inline load_module() into sys_init_module()
fixed the particular symptoms of this that Alan Brunelle saw in that
bugzilla entry, but the huge stack waste by cpumask_t's was the more
direct cause.

Some day we'll have allocation helpers that allocate large CPU masks
dynamically, but in the meantime we simply cannot allow cpumasks this
large.

Cc: Alan D. Brunelle <Alan.Brunelle@hp.com>
Cc: Mike Travis <travis@sgi.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The shadow code assigns a pte directly in one place, which is nonatomic on
i386 can can cause random memory references.  Fix by using an atomic setter.

Signed-off-by: Avi Kivity <avi@qumranet.com>