Add more details to the dynamic function tracing design implementation.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
LKML-Reference: <1279610015-10250-1-git-send-email-vapier@gentoo.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

Documentation/trace/ftrace.txt says

  buffer_size_kb:

        This sets or displays the number of kilobytes each CPU
        buffer can hold. The tracer buffers are the same size
        for each CPU. The displayed number is the size of the
        CPU buffer and not total size of all buffers. The
        trace buffers are allocated in pages (blocks of memory
        that the kernel uses for allocation, usually 4 KB in size).
        If the last page allocated has room for more bytes
        than requested, the rest of the page will be used,
        making the actual allocation bigger than requested.
        ( Note, the size may not be a multiple of the page size
          due to buffer management overhead. )

        This can only be updated when the current_tracer
        is set to "nop".

But it's incorrect. currently total memory consumption is
'buffer_size_kb x CPUs x 2'.

Why two times difference is there? because ftrace implicitly allocate
the buffer for max latency too.

That makes sad result when admin want to use large buffer. (If admin
want full logging and makes detail analysis). example, If admin
have 24 CPUs machine and write 200MB to buffer_size_kb, the system
consume ~10GB memory (200MB x 24 x 2). umm.. 5GB memory waste is
usually unacceptable.

Fortunatelly, almost all users don't use max latency feature.
The max latency buffer can be disabled easily.

This patch shrink buffer size of the max latency buffer if
unnecessary.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
LKML-Reference: <20100701104554.DA2D.A69D9226@jp.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

__print_flags() and __print_symbolic() use percpu trace_seq:

1) Its memory is allocated at compile time, it wastes memory if we don't use tracing.
2) It is percpu data and it wastes more memory for multi-cpus system.
3) It disables preemption when it executes its core routine
   "trace_seq_printf(s, "%s: ", #call);" and introduces latency.

So we move this trace_seq to struct trace_iterator.

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
LKML-Reference: <4C078350.7090106@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

Reorder structure to remove 8 bytes of padding on 64 bit builds.
This shrinks the size to 128 bytes so allowing allocation from a smaller
slab & needed one fewer cache lines.

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
LKML-Reference: <1269516456.2054.8.camel@localhost>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

We found that even enabling a single trace event that will rarely be
triggered can add big overhead to context switch.

(lmbench context switch test)
 -------------------------------------------------
 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
 ctxsw  ctxsw  ctxsw ctxsw  ctxsw   ctxsw   ctxsw
------ ------ ------ ------ ------ ------- -------
  2.19   2.3   2.21   2.56   2.13     2.54    2.07
  2.39   2.51  2.35   2.75   2.27     2.81    2.24

The overhead is 6% ~ 11%.

It's because when a trace event is enabled 3 tracepoints (sched_switch,
sched_wakeup, sched_wakeup_new) will be activated to map pid to cmdname.

We'd like to avoid this overhead, so add a trace option '(no)record-cmd'
to allow to disable cmdline recording.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <4C2D57F4.2050204@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

Introducing hists__remove_entry_filter.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Right now ENTER doesn't always exits the newt tree widget, as it is used
for expanding/collapsing branches, but with the new tree widget being
developed we need to regain control to handle it, expanding/collapsing
branches.

In fact its really up to the ui_browser user to state what extra keys
should stop ui_browser__run, and it should handle just the ones needed
for basic browsing.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing into perf/core

Invert the return value of die_compare_name(), because it returns a 'bool'
result which should be expeced true if the die's name is same as compared
string.

LKML-Reference: <4C36EBED.1000006@hitachi.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Gcc generates DW_AT_comp_dir and stores relative source path if building kernel
without O= option. In that case, perf probe --line sometimes doesn't work
without --source option, because it tries to access relative source path.

This adds DW_AT_comp_dir support to perf probe for finding an absolute source
path when no --source option.

LKML-Reference: <4C36EBE7.3060802@hitachi.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Perf probe -L shows incorrect error message (Dwarf error) if it fails to find
source file. This can confuse users.

# ./perf probe -s /nowhere -L vfs_read
Debuginfo analysis failed. (-2)
  Error: Failed to show lines. (-2)

With this patch, it shows correct message.

# ./perf probe -s /nowhere -L vfs_read
Failed to find source file. (-2)
  Error: Failed to show lines. (-2)

LKML-Reference: <4C36EBDB.4020308@hitachi.com>
Cc: Chase Douglas <chase.douglas@canonical.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Markers have been removed, but we forgot to remove their
section.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

The ksym (breakpoint) ftrace plugin has been superseded by perf
tools that are much more poweful to use the cpu breakpoints.
This tracer doesn't bring more feature. It has been deprecated
for a while now, lets remove it.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Prasad <prasad@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>

ftrace and perf events now use the same development branch.
Don't show a stale branch to developers.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>

Implement get_arch_regstr() for SH so that, given a DWARF register number, the
corresponding symbolic name of that register can be looked up.

Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ian Munsie <imunsie@au.ibm.com>
LKML-Reference: <e55812819ad18c2ceca5651ac7698a2af46180d7.1278774279.git.matt@console-pimps.org>
Signed-off-by: Matt Fleming <matt@console-pimps.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge branch 'perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 into perf/core

Add static and global variables support to perf probe.
This allows user to trace non-local variables (and
structure members) at probe points.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100519195749.2885.17451.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Add array-entry tracing support to perf probe. This enables to trace an entry
of array which is indexed by constant value, e.g. array[0].

For example:

  $ perf probe -a 'bio_split bi->bi_io_vec[0]'

Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100519195742.2885.5344.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Support string type casting to event argument. If perf-probe finds an argument
casted as string, it ensures the target variable is "(unsigned/signed) char
*(or []). perf-probe also adds dereference if the target is a pointer.

So, both of 'char buf[10];' and 'char *buf;' can be accessed by 'buf:string'

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100519195734.2885.1666.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Support string type tracing and printing in kprobe-tracer.

This allows user to trace string data in kernel including __user data. Note
that sometimes __user data may not be accessed if it is paged-out (sorry, but
kprobes operation should be done in atomic, we can not wait for page-in).

Commiter note: Fixed up conflicts with b7e2ecef.

Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20100519195724.2885.18788.stgit@localhost6.localdomain6>
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

To support cache events we have reserved the low 6 bits in
hw_perf_event::config (which is a part of CCCR register
configuration actually).

These bits represent Replay Event mertic enumerated in
enum P4_PEBS_METRIC. The caller should not care about
which exact bits should be set and how -- the caller
just chooses one P4_PEBS_METRIC entity and puts it into
the config. The kernel will track it and set appropriate
additional MSR registers (metrics) when needed.

The reason for this redesign was the PEBS enable bit, which
should not be set until DS (and PEBS sampling) support will
be implemented properly.

TODO
====

 - PEBS sampling (note it's tricky and works with _one_ counter only
   so for HT machines it will be not that easy to handle both threads)

 - tracking of PEBS registers state, a user might need to turn
   PEBS off completely (ie no PEBS enable, no UOP_tag) but some
   other event may need it, such events clashes and should not
   run simultaneously, at moment we just don't support such events

 - eventually export user space bits in separate header which will
   allow user apps to configure raw events more conveniently.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1278295769.9540.15.camel@minggr.sh.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge reason: Pick up the latest perf fixes

Signed-off-by: Ingo Molnar <mingo@elte.hu>

* master.kernel.org:/home/rmk/linux-2.6-arm:
  ARM: 6205/1: perf: ensure counter delta is treated as unsigned
  ARM: 6202/1: Do not ARM_DMA_MEM_BUFFERABLE on RealView boards with L210/L220
  ARM: 6201/1: RealView: Do not use outer_sync() on ARM11MPCore boards with L220
  ARM: 6195/1: OMAP3: pmu: make CPU_HAS_PMU dependent on OMAP3_EMU
  ARM: 6194/1: change definition of cpu_relax() for ARM11MPCore
  ARM: 6193/1: RealView: Align the machine_desc.phys_io to 1MB section
  ARM: 6192/1: VExpress: Align the machine_desc.phys_io to 1MB section
  ARM: 6188/1: Add a config option for the ARM11MPCore DMA cache maintenance workaround
  ARM: 6187/1: The v6_dma_inv_range() function must preserve data on SMP
  ARM: 6186/1: Avoid the CONSISTENT_DMA_SIZE warning on noMMU builds
  ARM: mx3: mx31lilly: fix build error for !CONFIG_USB_ULPI
  [ARM] mmp: fix build failure due to IRQ_PMU depends on ARCH_PXA
  [ARM] pxa/mioa701: fix camera regression
  [ARM] pxa/z2: fix flash layout to final version
  [ARM] pxa/z2: fix missing include in battery driver
  [ARM] pxa: fix incorrect gpio type in udc_pxa2xx.h

Merge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/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, x86: Fix incorrect branches event on AMD CPUs
  perf tools: Fix find tids routine by excluding "." and ".."
  x86: Send a SIGTRAP for user icebp traps

We should initialize the module dynamic debug datastructures
only after determining that the module is not loaded yet. This
fixes a bug that introduced in 2.6.35-rc2, where when a trying
to load a module twice, we also load it's dynamic printing data
twice which causes all sorts of nasty issues. Also handle
the dynamic debug cleanup later on failure.

Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (removed a #ifdef)
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: remove block number from inode lookup code
  xfs: rename XFS_IGET_BULKSTAT to XFS_IGET_UNTRUSTED
  xfs: validate untrusted inode numbers during lookup
  xfs: always use iget in bulkstat
  xfs: prevent swapext from operating on write-only files

* 'merge-devicetree' of git://git.secretlab.ca/git/linux-2.6:
  of/dma: fix build breakage in ppc4xx adma driver

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/i7core:
  MAINTAINERS: Add an entry for i7core_edac
  i7core_edac: Avoid doing multiple probes for the same card
  i7core_edac: Properly discover the first QPI device

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild-2.6:
  kbuild: Propagate LOCALVERSION= down to scripts/setlocalversion
  kbuild: Clean up and speed up the localversion logic

Hardware performance counters on ARM are 32-bits wide but atomic64_t
variables are used to represent counter data in the hw_perf_event structure.

The armpmu_event_update function right-shifts a signed 64-bit delta variable
and adds the result to the event count. This can lead to shifting in sign-bits
if the MSB of the 32-bit counter value is set. This results in perf output
such as:

 Performance counter stats for 'sleep 20':

 18446744073460670464  cycles             <-- 0xFFFFFFFFF12A6000
        7783773  instructions             #      0.000 IPC
            465  context-switches
            161  page-faults
        1172393  branches

   20.154242147  seconds time elapsed

This patch ensures that the delta value is treated as unsigned so that the
right shift sets the upper bits to zero.

Cc: <stable@kernel.org>
Acked-by: Jamie Iles <jamie.iles@picochip.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

While doing some performance counter validation tests on some
assembly language programs I noticed that the "branches:u"
count was very wrong on AMD machines.

It looks like the wrong event was selected.

Signed-off-by: Vince Weaver <vweaver1@eecs.utk.edu>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: <stable@kernel.org>
LKML-Reference: <alpine.DEB.2.00.1007011526010.23160@cl320.eecs.utk.edu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Convert ppc4xx adma driver to use new node pointer location

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Anatolij Gustschin <agust@denx.de>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>

While here, fixes the mailing list for i5400_edac

Acked-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

As Nehalem/Nehalem-EP/Westmere devices uses several devices for the same
functionality (memory controller), the default way of proping devices doesn't
work. So, instead of a per-device probe, all devices should be probed at once.

This means that we should block any new attempt of probe, otherwise, it will
try to register the same device several times.

Acked-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

On Nehalem/Nehalem-EP/Westmere, the first QPI device is the last PCI bus.
The last bus is generally at 0x3f or 0xff, but there are also other systems
using different setups. For example, HP Z800 has 0x7f as the last bus.

This patch adds a logic to discover the last bus, dynamically detecting it
at runtime.

Acked-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

Merge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: Cure nr_iowait_cpu() users
  init: Fix comment
  init, sched: Fix race between init and kthreadd

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp:
  amd64_edac: Fix syndrome calculation on K8

When calculating the DCT channel from the syndrome we need to know the
syndrome type (x4 vs x8). On F10h, this is read out from extended PCI
cfg space register F3x180 while on K8 we only support x4 syndromes and
don't have extended PCI config space anyway.

Make the code accessing F3x180 F10h only and fall back to x4 syndromes
on everything else.

Cc: <stable@kernel.org> # .33.x .34.x
Reported-by: Jeffrey Merkey <jeffmerkey@gmail.com>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>