free_cache_attributes() kfree's:

   per_cpu(ici_cpuid4_info, cpu)->l3

which is a pointer to memory which was allocated as a block in
amd_init_l3_cache(). l3 of a particular cpu points to a part of this
memory blob. The part and the rest of the blob are still referenced by
other cpus.

As far as I can tell from the git history this is a leftover from the
conversion from per cpu to node data with commit ba06edb6(x86,
cacheinfo: Make L3 cache info per node) and the following commit
f658bcfb(x86, cacheinfo: Cleanup L3 cache index disable support)
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Cc: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Mike Travis <travis@sgi.com>
Link: http://lkml.kernel.org/r/20110723212626.550539989@linutronix.deSigned-off-by: NIngo Molnar <mingo@elte.hu>

When the moduleu.h splitting tree is merged to the latest
tip:x86/cpu tree, the x86_64 allmodconfig build fails like this:

 arch/x86/kernel/cpu/amd.c: In function 'bsp_init_amd':
 arch/x86/kernel/cpu/amd.c:437:3: error: 'va_align' undeclared (first use in this function)
 arch/x86/kernel/cpu/amd.c:438:23: error: 'ALIGN_VA_32' undeclared (first use in this function)
 arch/x86/kernel/cpu/amd.c:438:37: error: 'ALIGN_VA_64' undeclared (first use in this function)

This is caused by the module.h split up intreacting with commit
dfb09f9b ("x86, amd: Avoid cache aliasing penalties on AMD
family 15h") from the tip:x86/cpu tree.

I have added the following patch for today (this, or something
similar, could be applied to the tip tree directly - the
export.h include below was added by the module.h splitup).

So include elf.h to use va_align and remove this implicit
dependency on module.h doing it for us.
Signed-off-by: NStephen Rothwell <sfr@canb.auug.org.au>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Link: http://lkml.kernel.org/r/20110810114956.238d66772883636e3040d29f@canb.auug.org.auSigned-off-by: NIngo Molnar <mingo@elte.hu>

Move code which is run once on the BSP during boot into the cpu_dev
helper.

[ hpa: removed bogus cpu_has -> static_cpu_has conversion ]
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/20110805180409.GC26217@aftabSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Add a function ptr to struct cpu_dev which is destined to be run only
once on the BSP during boot.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/20110805180116.GB26217@aftabSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This patch provides performance tuning for the "Bulldozer" CPU. With its
shared instruction cache there is a chance of generating an excessive
number of cache cross-invalidates when running specific workloads on the
cores of a compute module.

This excessive amount of cross-invalidations can be observed if cache
lines backed by shared physical memory alias in bits [14:12] of their
virtual addresses, as those bits are used for the index generation.

This patch addresses the issue by clearing all the bits in the [14:12]
slice of the file mapping's virtual address at generation time, thus
forcing those bits the same for all mappings of a single shared library
across processes and, in doing so, avoids instruction cache aliases.

It also adds the command line option "align_va_addr=(32|64|on|off)" with
which virtual address alignment can be enabled for 32-bit or 64-bit x86
individually, or both, or be completely disabled.

This change leaves virtual region address allocation on other families
and/or vendors unaffected.
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1312550110-24160-2-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Fix the printk_once() so that it actually prints (didn't print before
due to a stray comma.)

[ hpa: changed to an incremental patch and adjusted the description
  accordingly. ]
Signed-off-by: NLen Brown <len.brown@intel.com>
Link: http://lkml.kernel.org/r/alpine.LFD.2.02.1107151732480.18606@x980
Cc: <table@kernel.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Since 2.6.36 (23016bf0), Linux prints the existence of "epb" in /proc/cpuinfo,
Since 2.6.38 (d5532ee7), the x86_energy_perf_policy(8) utility has
been available in-tree to update MSR_IA32_ENERGY_PERF_BIAS.

However, the typical BIOS fails to initialize the MSR, presumably
because this is handled by high-volume shrink-wrap operating systems...

Linux distros, on the other hand, do not yet invoke x86_energy_perf_policy(8).
As a result, WSM-EP, SNB, and later hardware from Intel will run in its
default hardware power-on state (performance), which assumes that users
care for performance at all costs and not for energy efficiency.
While that is fine for performance benchmarks, the hardware's intended default
operating point is "normal" mode...

Initialize the MSR to the "normal" by default during kernel boot.

x86_energy_perf_policy(8) is available to change the default after boot,
should the user have a different preference.
Signed-off-by: NLen Brown <len.brown@intel.com>
Link: http://lkml.kernel.org/r/alpine.LFD.2.02.1107140051020.18606@x980Acked-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@kernel.org>

We'd rather that modern machines not check if HLT works on
every entry into idle, for the benefit of machines that had
marginal electricals 15-years ago.  If those machines are still running
the upstream kernel, they can use "idle=poll".  The only difference
will be that they'll now invoke HLT in machine_hlt().

cc: x86@kernel.org # .39.x
Signed-off-by: NLen Brown <len.brown@intel.com>

The workaround for AMD erratum 400 uses the term "c1e" falsely suggesting:
1. Intel C1E is somehow involved
2. All AMD processors with C1E are involved

Use the string "amd_c1e" instead of simply "c1e" to clarify that
this workaround is specific to AMD's version of C1E.
Use the string "e400" to clarify that the workaround is specific
to AMD processors with Erratum 400.

This patch is text-substitution only, with no functional change.

cc: x86@kernel.org
Acked-by: NBorislav Petkov <borislav.petkov@amd.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Commit b87cf80a added support for
ARAT (Always Running APIC timer) on AMD processors that are not
affected by erratum 400. This erratum is present on certain processor
families and prevents APIC timer from waking up the CPU when it
is in a deep C state, including C1E state.

Determining whether a processor is affected by this erratum may
have some corner cases and handling these cases is somewhat
complicated. In the interest of simplicity we won't claim ARAT
support on processor families below 0x12 and will go back to
broadcasting timer when going idle.
Signed-off-by: NBoris Ostrovsky <ostr@amd64.org>
Link: http://lkml.kernel.org/r/1306423192-19774-1-git-send-email-ostr@amd64.orgTested-by: NBoris Petkov <borislav.petkov@amd.com>
Cc: Hans Rosenfeld <Hans.Rosenfeld@amd.com>
Cc: Andreas Herrmann <Andreas.Herrmann3@amd.com>
Cc: Chuck Ebbert <cebbert@redhat.com>
Cc: stable@kernel.org # 32.x, 38.x, 39.x
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This patch removes a check that causes incorrect scheduler
domain setup (SMP instead of SMT) and bootlog warning messages
when cpuid extensions for topology enumeration are not supported
and the number of processors reported to the OS is smaller than
smp_num_siblings.
Acked-by: NSuresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: NNikhil P Rao <nikhil.rao@intel.com>
Link: http://lkml.kernel.org/r/1306343921.19325.1.camel@fedora13Signed-off-by: NIngo Molnar <mingo@elte.hu>

The setup_smep function gets calle at resume time too, and is thus not a
pure __init function.  When marked as __init, it gets thrown out after
the kernel has initialized, and when the kernel is suspended and
resumed, the code will no longer be around, and we'll get a nice "kernel
tried to execute NX-protected page" oops because the page is no longer
marked executable.
Reported-and-tested-by: NParag Warudkar <parag.lkml@gmail.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: "H. Peter Anvin" <hpa@linux.intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix these kernel compilation warnings:

 WARNING: arch/x86/built-in.o(.cpuinit.text+0x1e07): Section mismatch ...
 WARNING: arch/x86/built-in.o(.cpuinit.text+0x1b10): Section mismatch ...

introduced by:

  de5397ad: x86, cpu: Enable/disable Supervisor Mode Execution Protection

Change disable_smep from __initdata to __cpuinitdata.
Change setup_smep() from __init to __cpuinit.
Reported-by: NYinghai Lu <yinghai@kernel.org>
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
Cc: Asit K Mallick <asit.k.mallick@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1305930797-11409-1-git-send-email-fenghua.yu@intel.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

The workaround for Bugzilla:

	https://bugzilla.kernel.org/show_bug.cgi?id=33012

introduced a read and a write to the MC4 mask msr.

Unfortunatly this MSR is not emulated by the KVM hypervisor
so that the kernel will get a #GP and crashes when applying
this workaround when running inside KVM.

This issue was reported as:

	https://bugzilla.kernel.org/show_bug.cgi?id=35132

and is fixed with this patch. The change just let the kernel
ignore any #GP it gets while accessing this MSR by using the
_safe msr access methods.
Reported-by: NTörök Edwin <edwintorok@gmail.com>
Signed-off-by: NJoerg Roedel <joerg.roedel@amd.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Maciej Rutecki <maciej.rutecki@gmail.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: <stable@kernel.org> # .39.x
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Signed-off-by: NDave Jones <davej@redhat.com>

Enable/disable newly documented SMEP (Supervisor Mode Execution Protection) CPU
feature in kernel. CR4.SMEP (bit 20) is 0 at power-on. If the feature is
supported by CPU (X86_FEATURE_SMEP), enable SMEP by setting CR4.SMEP. New kernel
option nosmep disables the feature even if the feature is supported by CPU.

[ hpa: moved the call to setup_smep() until after the vendor-specific
  initialization; that ensures that CPUID features are unmasked.  We
  will still run it before we have userspace (never mind uncontrolled
  userspace). ]
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
LKML-Reference: <1305157865-31727-1-git-send-email-fenghua.yu@intel.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

If kernel intends to use enhanced REP MOVSB/STOSB, it must ensure
IA32_MISC_ENABLE.Fast_String_Enable (bit 0) is set and CPUID.(EAX=07H, ECX=0H):
EBX[bit 9] also reports 1.
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-3-git-send-email-fenghua.yu@intel.comSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

CPUID leaf 7, subleaf 0 returns the maximum subleaf in EAX, not the
number of subleaves.  Since so far only subleaf 0 is defined (and only
the EBX bitfield) we do not need to qualify the test.
Signed-off-by: NFenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305660806-17519-1-git-send-email-fenghua.yu@intel.comSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@kernel.org> 2.6.36..39

Trying to enable the local APIC timer on early K8 revisions
uncovers a number of other issues with it, in conjunction with
the C1E enter path on AMD. Fixing those causes much more churn
and troubles than the benefit of using that timer brings so
don't enable it on K8 at all, falling back to the original
functionality the kernel had wrt to that.
Reported-and-bisected-by: NNick Bowler <nbowler@elliptictech.com>
Cc: Boris Ostrovsky <Boris.Ostrovsky@amd.com>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Greg Kroah-Hartman <greg@kroah.com>
Cc: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Nick Bowler <nbowler@elliptictech.com>
Cc: Joerg-Volker-Peetz <jvpeetz@web.de>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1305636919-31165-3-git-send-email-bp@amd64.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

This reverts commit e20a2d20, as it crashes
certain boxes with specific AMD CPU models.

Moving the lower endpoint of the Erratum 400 check to accomodate
earlier K8 revisions (A-E) opens a can of worms which is simply
not worth to fix properly by tweaking the errata checking
framework:

* missing IntPenging MSR on revisions < CG cause #GP:

http://marc.info/?l=linux-kernel&m=130541471818831

* makes earlier revisions use the LAPIC timer instead of the C1E
idle routine which switches to HPET, thus not waking up in
deeper C-states:

http://lkml.org/lkml/2011/4/24/20

Therefore, leave the original boundary starting with K8-revF.
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Steppings A1 and B0 of Celeron Covington are currently misdetected as
Pentium II (Dixon). Fix it by removing the stepping check.

[ hpa: this fixes this specific bug... the CPUID documentation
  specifies that the L2 cache size can disambiguate additional CPUs;
  this patch does not fix that. ]
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Link: http://lkml.kernel.org/r/201105162138.15416.linux@rainbow-software.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

We provide two slots to disable cache indices, and have a check to
prevent both slots to be used for the same index.

If the user disables the same index on different subcaches, both slots
will hold the same index, e.g.

  $ echo 2047 > /sys/devices/system/cpu/cpu0/cache/index3/cache_disable_0
  $ cat /sys/devices/system/cpu/cpu0/cache/index3/cache_disable_0
  2047
  $ echo 1050623 > /sys/devices/system/cpu/cpu0/cache/index3/cache_disable_1
  $ cat /sys/devices/system/cpu/cpu0/cache/index3/cache_disable_1
  2047

due to the fact that the check was looking only at index bits [11:0]
and was ignoring writes to bits outside that range. The more correct
fix is to simply check whether the index is within the bounds of
[0..l3->indices].

While at it, cleanup comments and drop now-unused local macros.
Signed-off-by: NFrank Arnold <frank.arnold@amd.com>
Link: http://lkml.kernel.org/r/1305553188-21061-3-git-send-email-bp@amd64.orgSigned-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

732eacc0 converted code around the
kernel using nested max() macros to use the new max3 macro but forgot to
remove the old line in intel_cacheinfo.c. Fix it.

Cc: Hagen Paul Pfeifer <hagen@jauu.net>
Cc: Frank Arnold <farnold@amd64.org>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Link: http://lkml.kernel.org/r/1305553188-21061-2-git-send-email-bp@amd64.orgSigned-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This patch fixes a bug reported by a customer, who found
that many unreasonable error interrupts reported on all
non-boot CPUs (APs) during the system boot stage.

According to Chapter 10 of Intel Software Developer Manual
Volume 3A, Local APIC may signal an illegal vector error when
an LVT entry is set as an illegal vector value (0~15) under
FIXED delivery mode (bits 8-11 is 0), regardless of whether
the mask bit is set or an interrupt actually happen. These
errors are seen as error interrupts.

The initial value of thermal LVT entries on all APs always reads
0x10000 because APs are woken up by BSP issuing INIT-SIPI-SIPI
sequence to them and LVT registers are reset to 0s except for
the mask bits which are set to 1s when APs receive INIT IPI.

When the BIOS takes over the thermal throttling interrupt,
the LVT thermal deliver mode should be SMI and it is required
from the kernel to keep AP's LVT thermal monitoring register
programmed as such as well.

This issue happens when BIOS does not take over thermal throttling
interrupt, AP's LVT thermal monitor register will be restored to
0x10000 which means vector 0 and fixed deliver mode, so all APs will
signal illegal vector error interrupts.

This patch check if interrupt delivery mode is not fixed mode before
restoring AP's LVT thermal monitor register.
Signed-off-by: NYouquan Song <youquan.song@intel.com>
Acked-by: NSuresh Siddha <suresh.b.siddha@intel.com>
Acked-by: NYong Wang <yong.y.wang@intel.com>
Cc: hpa@linux.intel.com
Cc: joe@perches.com
Cc: jbaron@redhat.com
Cc: trenn@suse.de
Cc: kent.liu@intel.com
Cc: chaohong.guo@intel.com
Cc: <stable@kernel.org> # As far back as possible
Link: http://lkml.kernel.org/r/1303402963-17738-1-git-send-email-youquan.song@intel.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

b may be added to a list, but is not removed before being freed
in the case of an error.  This is done in the corresponding
deallocation function, so the code here has been changed to
follow that.

The sematic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression E,E1,E2;
identifier l;
@@

*list_add(&E->l,E1);
... when != E1
    when != list_del(&E->l)
    when != list_del_init(&E->l)
    when != E = E2
*kfree(E);// </smpl>
Signed-off-by: NJulia Lawall <julia@diku.dk>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: <stable@kernel.org>
Link: http://lkml.kernel.org/r/1305294731-12127-1-git-send-email-julia@diku.dkSigned-off-by: NIngo Molnar <mingo@elte.hu>

Both warning and warning_symbol are nowhere used.
Let's get rid of them.
Signed-off-by: NRichard Weinberger <richard@nod.at>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Soeren Sandmann Pedersen <ssp@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: x86 <x86@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Link: http://lkml.kernel.org/r/1305205872-10321-2-git-send-email-richard@nod.atSigned-off-by: NFrederic Weisbecker <fweisbec@gmail.com>

The Intel Nehalem offcore bits implemented in:

  e994d7d2: perf: Fix LLC-* events on Intel Nehalem/Westmere

... are wrong: they implemented _ACCESS as _HIT and counted OTHER_CORE_HIT* as
MISS even though its clearly documented as an L3 hit ...

Fix them and the Westmere definitions as well.

Cc: Andi Kleen <ak@linux.intel.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1299119690-13991-3-git-send-email-ming.m.lin@intel.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

Extend the Intel SandyBridge PMU driver with definitions
for generic front-end and back-end stall events.

( As commit 30112039 "perf events, x86: Add Westmere stalled-cycles-frontend/backend
  events" says, these are only approximations. )
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1304666042-17577-1-git-send-email-ming.m.lin@intel.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

With dynamic debug having gained the capability to report debug messages
also during the boot process, it offers a far superior interface for
debug messages than the custom cpufreq infrastructure. As a first step,
remove the old cpufreq_debug_printk() function and replace it with a call
to the generic pr_debug() function.

How can dynamic debug be used on cpufreq? You need a kernel which has
CONFIG_DYNAMIC_DEBUG enabled.

To enabled debugging during runtime, mount debugfs and

$ echo -n 'module cpufreq +p' > /sys/kernel/debug/dynamic_debug/control

for debugging the complete "cpufreq" module. To achieve the same goal during
boot, append

	ddebug_query="module cpufreq +p"

as a boot parameter to the kernel of your choice.

For more detailled instructions, please see
Documentation/dynamic-debug-howto.txt
Signed-off-by: NDominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: NDave Jones <davej@redhat.com>

UUID needs to be written out the way it is described in
Sec 18.5.124 of ACPI 4.0a Specification.

Platform firmware's use of this UUID/_OSC is optional, which is
why we didn't notice this bug earlier.
Signed-off-by: NNaga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: NDave Jones <davej@redhat.com>
Cc: stable@kernel.org

Older AMD K8 processors (Revisions A-E) are affected by erratum
400 (APIC timer interrupts don't occur in C states greater than
C1). This, for example, means that X86_FEATURE_ARAT flag should
not be set for these parts.

This addresses regression introduced by commit
b87cf80a ("x86, AMD: Set ARAT
feature on AMD processors") where the system may become
unresponsive until external interrupt (such as keyboard input)
occurs. This results, for example, in time not being reported
correctly, lack of progress on the system and other lockups.
Reported-by: NJoerg-Volker Peetz <jvpeetz@web.de>
Tested-by: NJoerg-Volker Peetz <jvpeetz@web.de>
Acked-by: NBorislav Petkov <borislav.petkov@amd.com>
Signed-off-by: NBoris Ostrovsky <Boris.Ostrovsky@amd.com>
Cc: stable@kernel.org
Link: http://lkml.kernel.org/r/1304113663-6586-1-git-send-email-ostr@amd64.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

Extend the Intel Westmere PMU driver with definitions for generic front-end and
back-end stall events.

( These are only approximations. )
Reported-by: NDavid Ahern <dsahern@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-7y40wib8n008io7hjpn1dsrm@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

Extend the Intel and AMD event definitions with generic front-end and
back-end stall events.

( These are only approximations - suggestions are welcome for better events. )

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-7y40wib8n001io7hjpn1dsrm@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

Add two generic hardware events: front-end and back-end stalled cycles.

These events measure conditions when the CPU is executing code but its
capabilities are not fully utilized. Understanding such situations and
analyzing them is an important sub-task of code optimization workflows.

Both events limit performance: most front end stalls tend to be caused
by branch misprediction or instruction fetch cachemisses, backend
stalls can be caused by various resource shortages or inefficient
instruction scheduling.

Front-end stalls are the more important ones: code cannot run fast
if the instruction stream is not being kept up.

An over-utilized back-end can cause front-end stalls and thus
has to be kept an eye on as well.

The exact composition is very program logic and instruction mix
dependent.

We use the terms 'stall', 'front-end' and 'back-end' loosely and
try to use the best available events from specific CPUs that
approximate these concepts.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-7y40wib8n000io7hjpn1dsrm@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

Use the UOPS_EXECUTED.*,c=1,i=1 event on Intel CPUs - it is a rather
good indicator of CPU execution stalls, more sensitive and more inclusive
than the 0xa2 resource stalls event (which does not count nearly as many
stall types).

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-7y40wib8n1eqio7hjpn2dsrm@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

It was noticed that P4 machines were generating double NMIs for
each perf event.  These extra NMIs lead to 'Dazed and confused'
messages on the screen.

I tracked this down to a P4 quirk that said the overflow bit had
to be cleared before re-enabling the apic LVT mask.  My first
attempt was to move the un-masking inside the perf nmi handler
from before the chipset NMI handler to after.

This broke Nehalem boxes that seem to like the unmasking before
the counters themselves are re-enabled.

In order to keep this change simple for 2.6.39, I decided to
just simply move the apic LVT un-masking to the beginning of all
the chipset NMI handlers, with the exception of Pentium4's to
fix the double NMI issue.

Later on we can move the un-masking to later in the handlers to
save a number of 'extra' NMIs on those particular chipsets.

I tested this change on a P4 machine, an AMD machine, a Nehalem
box, and a core2quad box.  'perf top' worked correctly along
with various other small 'perf record' runs.  Anything high
stress breaks all the machines but that is a different problem.

Thanks to various people for testing different versions of this
patch.
Reported-and-tested-by: NShaun Ruffell <sruffell@digium.com>
Signed-off-by: NDon Zickus <dzickus@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Link: http://lkml.kernel.org/r/1303900353-10242-1-git-send-email-dzickus@redhat.comSigned-off-by: NIngo Molnar <mingo@elte.hu>
CC: Cyrill Gorcunov <gorcunov@gmail.com>

Various constraint tables were not marked read-mostly.
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-wpqwwvmhxucy5e718wnamjiv@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

The new PERF_COUNT_HW_STALLED_CYCLES event tries to approximate
cycles the CPU does nothing useful, because it is stalled on a
cache-miss or some other condition.
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NArnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-fue11vymwqsoo5to72jxxjyl@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

On Nehalem CPUs the retired branch-misses event can be completely bogus,
when there are no branch-misses occuring. When there are a lot of branch
misses then the count is pretty accurate. Still, this leaves us with an
event that over-counts a lot.

Detect this erratum and work it around by using BR_MISP_EXEC.ANY events.
These will also count speculated branches but still it's a lot more
precise in practice than the architectural event.
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-yyfg0bxo9jsqxd6a0ovfny27@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>

Currently the x86 backend incorrectly assumes that any BRANCH_INSN
with sample_period==1 is a BTS request. This is not true when we do
frequency driven profiling such as 'perf record -e branches'.

Solves this error:

  $ perf record -e branches ./array
  Error: sys_perf_event_open() syscall returned with 95 (Operation not supported).
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Reported-by: NIngo Molnar <mingo@elte.hu>
Cc: "Metzger, Markus T" <markus.t.metzger@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-rd2y4ct71hjawzz6fpvsy9hg@git.kernel.orgSigned-off-by: NIngo Molnar <mingo@elte.hu>