Currently, the function graph tracer depends on the trace_printk
to record the depth. All the information is already there in the trace
to calculate function depth, with the exception of having the printk
be the first item. But as soon as a entry or exit is reached, then
we know the depth.

This patch changes the iter->private data from recording a per cpu
last_pid, to a structure that holds both the last_pid and the current
depth. This data is used to determine the function depth for the
printks.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

This patch makes print_printk_msg_only and print_bprintk_msg_only
global for other functions to use. It also renames them by adding
a "trace_" to the beginning to avoid namespace collisions.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: clean up

The prologue of the function graph entry, return and comments all
start out pretty much the same. Each of these duplicate code and
do so slightly differently.

This patch consolidates the printing of the pid, absolute time,
cpu and proc (and for entry, the interrupt).
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

When I review the sensitive code ftrace_nmi_enter(), I found
the atomic variable nmi_running does protect NMI VS do_ftrace_mod_code(),
but it can not protects NMI(entered nmi) VS NMI(ftrace_nmi_enter()).

cpu#1                   | cpu#2                 | cpu#3
ftrace_nmi_enter()      | do_ftrace_mod_code()  |
  not modify            |                       |
------------------------|-----------------------|--
executing               | set mod_code_write = 1|
executing             --|-----------------------|--------------------
executing               |                       | ftrace_nmi_enter()
executing               |                       |    do modify
------------------------|-----------------------|-----------------
ftrace_nmi_exit()       |                       |

cpu#3 may be being modified the code which is still being executed on cpu#1,
it will have undefined results and possibly take a GPF, this patch
prevents it occurred.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
LKML-Reference: <49C0B411.30003@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

There is currently no easy way to clear the trace buffer. Currently
the only way is to change the current tracer.

This patch lets the user clear the trace buffer by simply writing
into the trace files.

 echo > /debug/tracing/trace

or to clear a single cpu (i.e. for CPU 1):

 echo > /debug/tracing/per_cpu/cpu1/trace
Requested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: fix a selftest warning

In some cases, it's possible to see the following warning on irqsoff
tracer selftest:

[    4.640003] Testing tracer irqsoff: <4>------------[ cut here ]------------
[    4.653562] WARNING: at kernel/trace/trace.c:458 update_max_tr_single+0x9a/0xc4()
[    4.660000] Hardware name: System Product Name
[    4.660000] Modules linked in:
[    4.660000] Pid: 301, comm: kstop/1 Not tainted 2.6.29-rc8-tip #35837
[    4.660000] Call Trace:
[    4.660000]  [<4014b588>] warn_slowpath+0x79/0x8f
[    4.660000]  [<402d6949>] ? put_dec+0x64/0x6b
[    4.660000]  [<40162b56>] ? getnstimeofday+0x58/0xdd
[    4.660000]  [<40162210>] ? clocksource_read+0x3/0xf
[    4.660000]  [<4015eb44>] ? ktime_set+0x8/0x34
[    4.660000]  [<4014101a>] ? balance_runtime+0x8/0x56
[    4.660000]  [<405f6f11>] ? _spin_lock+0x3/0x10
[    4.660000]  [<4011f643>] ? ftrace_call+0x5/0x8
[    4.660000]  [<4015d0f1>] ? task_cputime_zero+0x3/0x27
[    4.660000]  [<40190ee7>] ? cpupri_set+0x90/0xcb
[    4.660000]  [<405f7208>] ? _spin_lock_irqsave+0x22/0x34
[    4.660000]  [<40190f12>] ? cpupri_set+0xbb/0xcb
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<4018493f>] ? ring_buffer_reset_cpu+0x27/0x51
[    4.660000]  [<405f7208>] ? _spin_lock_irqsave+0x22/0x34
[    4.660000]  [<40184962>] ? ring_buffer_reset_cpu+0x4a/0x51
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<4018cc29>] ? trace_hardirqs_off+0x1a/0x1c
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<40184962>] ? ring_buffer_reset_cpu+0x4a/0x51
[    4.660000]  [<401850f3>] ? cpumask_next+0x15/0x18
[    4.660000]  [<4018a41f>] update_max_tr_single+0x9a/0xc4
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4018cd13>] check_critical_timing+0xcc/0x11e
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4018cdf1>] stop_critical_timing+0x8c/0x9f
[    4.660000]  [<4014e5c4>] ? forget_original_parent+0xac/0xd0
[    4.660000]  [<4018ce3a>] trace_hardirqs_on+0x1a/0x1c
[    4.660000]  [<4014e5c4>] forget_original_parent+0xac/0xd0
[    4.660000]  [<4014e5fe>] exit_notify+0x16/0xf2
[    4.660000]  [<4014e8a5>] do_exit+0x1cb/0x225
[    4.660000]  [<4015c72b>] ? kthread+0x0/0x69
[    4.660000]  [<4011f61d>] kernel_thread_helper+0xd/0x10
[    4.660000] ---[ end trace a7919e7f17c0a725 ]---
[    4.660164] .. no entries found ..FAILED!

During the selftest of irqsoff tracer, we do that:

	/* disable interrupts for a bit */
	local_irq_disable();
	udelay(100);
	local_irq_enable();
	/* stop the tracing. */
	tracing_stop();
	/* check both trace buffers */
	ret = trace_test_buffer(tr, NULL);

If a callsite performs a new max delay with irqs off just after
tracing_stop, update_max_tr_single() -> ring_buffer_swap_cpu()
will be called with the buffers disabled by tracing_stop(), hence
the warning, then ring_buffer_swap_cpu() return -EAGAIN and
update_max_tr_single() complains.

Fix it by also stopping the tracer before stopping the tracing globally.
A similar situation can happen with preemptoff and preemptirqsoff tracers
where we apply the same fix.
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1237325938-5240-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace into tracing/ftrace

The start/stop methods of a tracer should be able to be executed
in all contexts. This patch converts the power tracer to do so.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

The stopping and starting of a tracer should be light weight and
be able to be called in all contexts. The sched_switch grabbed
mutexes in the start/stop functions. This patch changes it to a
simple variable, on/off.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: better performance for if branch tracer

Use an array to count the hit and misses of a conditional instead
of using another conditional. This cuts down on saturation of branch
predictions and increases performance of modern pipelined architectures.
Signed-off-by: Witold Baryluk <baryluk@smp.if.uj.edu.pl>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: feature to allow better serialized clock

This patch adds an option called "global-clock" that will allow
the tracer to switch to a slower but more accurate (across CPUs)
clock.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

This patch adds a new function called ring_buffer_set_clock that
allows a tracer to assign its own clock source to the buffer.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Commit ee6f779b ("filp->f_pos not
correctly updated in proc_task_readdir") changed the proc code to use
filp->f_pos directly, rather than through a temporary variable.  In the
process, that caused the operations to be done on the full 64 bits, even
though the offset is never that big.

That's all fine and dandy per se, but for some unfathomable reason gcc
generates absolutely horrid code when using 64-bit values in switch()
statements.  To the point of actually calling out to gcc helper
functions like __cmpdi2 rather than just doing the trivial comparisons
directly the way gcc does for normal compares.  At which point we get
link failures, because we really don't want to support that kind of
crazy code.

Fix this by just casting the f_pos value to "unsigned long", which
is plenty big enough for /proc, and avoids the gcc code generation issue.
Reported-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Zhang Le <r0bertz@gentoo.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Don't boost at the addresses which are listed on exception tables,
because major page fault will occur on those addresses.  In that case,
kprobes can not ensure that when instruction buffer can be freed since
some processes will sleep on the buffer.

kprobes-ia64 already has same check.
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm:
  dm crypt: wait for endio to complete before destruction
  dm crypt: fix kcryptd_async_done parameter
  dm io: respect BIO_MAX_PAGES limit
  dm table: rework reference counting fix
  dm ioctl: validate name length when renaming

In order for ntpd to correctly synchronize the clocks, the frequency of
the system clock must not be off by more than 500 ppm (or, put another
way, 1:2000), or ntpd will end up giving up on trying to synchronize
properly, and ends up reseting the clock in jumps instead.

The fast TSC PIT calibration sometimes failed this test - it was
assuming that the PIT reads always took about one microsecond each (2us
for the two reads to get a 16-bit timer), and that calibrating TSC to
the PIT over 15ms should thus be sufficient to get much closer than
500ppm (max 2us error on both sides giving 4us over 15ms: a 270 ppm
error value).

However, that assumption does not always hold: apparently some hardware
is either very much slower at reading the PIT registers, or there was
other noise causing at least one machine to get 700+ ppm errors.

So instead of using a fixed 15ms timing loop, this changes the fast PIT
calibration to read the TSC delta over the individual PIT timer reads,
and use the result to calculate the error bars on the PIT read timing
properly.  We then successfully calibrate the TSC only if the maximum
error bars fall below 500ppm.

In the process, we also relax the timing to allow up to 25ms for the
calibration, although it can happen much faster depending on hardware.
Reported-and-tested-by: Jesper Krogh <jesper@krogh.cc>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

During bootup, when we reprogram the PIT (programmable interval timer)
to start counting down from 0xffff in order to use it for the fast TSC
calibration, we should also make sure to delay a bit afterwards to allow
the PIT hardware to actually start counting with the new value.

That will happens at the next CLK pulse (1.193182 MHz), so the easiest
way to do that is to just wait at least one microsecond after
programming the new PIT counter value.  We do that by just reading the
counter value back once - which will take about 2us on PC hardware.
Reported-and-tested-by: john stultz <johnstul@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace into tracing/ftrace

Impact: fix memory leak

If event_format_read() exits early due to nonzero ppos, the
previous kmalloc doesn't get freed - might as well do the
check before the kmalloc and avoid the problem.
Signed-off-by: Tom Zanussi <tzanussi@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: =?ISO-8859-1?Q?Fr=E9d=E9ric?= Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237270859.8033.141.camel@charm-linux>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix for losing comms in trace

The command lines of tasks are cached at sched switch to not need
to record them at every trace point.  Disabling the tracing on stops
the recording of traces, but does not stop the caching of command lines.
When the tracing is off the cache may overflow and cause the tracing
to show incorrect tasks matching the PIDs.

This patch disables prevents updates to the comm cache when the ring buffer
is off.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: fix to one cause of incorrect comm outputs in trace

The spinlock only protected the creation of a comm <=> pid pair.
But it was possible that a reader could look up a pid, and get the
wrong comm because it had no locking.

This also required changing trace_find_cmdline to copy the comm cache
and not just send back a pointer to it.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

Impact: fix a dynamic tracing failure

Recently, the function and function graph tracers failed to use dynamic
tracing after the following commit:

fa9d13cf
(ftrace: don't try to __ftrace_replace_code on !FTRACE_FL_CONVERTED rec)

The patch is right except a mistake on the check for the FTRACE_FL_CONVERTED
flag. The code patching is aborted in case of successfully nopped sites.
What we want is the opposite: ignore the callsites that haven't been nopped.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Steven Rostedt <srostedt@redhat.com>

* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6:
  acpi-wmi: unsigned cannot be less than 0
  thinkpad-acpi: fix module autoloading for older models
  acer-wmi: Unmark as 'experimental'
  acpi-wmi: Unmark as 'experimental'
  acer-wmi: double free in acer_rfkill_exit()
  platform/x86: depends instead of select for laptop platform drivers
  asus-laptop: use select instead of depends on
  eeepc-laptop: restore acpi_generate_proc_event()
  asus-laptop: restore acpi_generate_proc_event()
  acpi: check for pxm_to_node_map overflow
  ACPI: remove doubled status checking
  ACPI suspend: Blacklist Toshiba Satellite L300 that requires to set SCI_EN directly on resume
  Revert "ACPI: make some IO ports off-limits to AML"
  suspend: switch the Asus Pundit P1-AH2 to old ACPI sleep ordering

The following oops has been reported when dm-crypt runs over a loop device.

...
[   70.381058] Process loop0 (pid: 4268, ti=cf3b2000 task=cf1cc1f0 task.ti=cf3b2000)
...
[   70.381058] Call Trace:
[   70.381058]  [<d0d76601>] ? crypt_dec_pending+0x5e/0x62 [dm_crypt]
[   70.381058]  [<d0d767b8>] ? crypt_endio+0xa2/0xaa [dm_crypt]
[   70.381058]  [<d0d76716>] ? crypt_endio+0x0/0xaa [dm_crypt]
[   70.381058]  [<c01a2f24>] ? bio_endio+0x2b/0x2e
[   70.381058]  [<d0806530>] ? dec_pending+0x224/0x23b [dm_mod]
[   70.381058]  [<d08066e4>] ? clone_endio+0x79/0xa4 [dm_mod]
[   70.381058]  [<d080666b>] ? clone_endio+0x0/0xa4 [dm_mod]
[   70.381058]  [<c01a2f24>] ? bio_endio+0x2b/0x2e
[   70.381058]  [<c02bad86>] ? loop_thread+0x380/0x3b7
[   70.381058]  [<c02ba8a1>] ? do_lo_send_aops+0x0/0x165
[   70.381058]  [<c013754f>] ? autoremove_wake_function+0x0/0x33
[   70.381058]  [<c02baa06>] ? loop_thread+0x0/0x3b7

When a table is being replaced, it waits for I/O to complete
before destroying the mempool, but the endio function doesn't
call mempool_free() until after completing the bio.

Fix it by swapping the order of those two operations.

The same problem occurs in dm.c with md referenced after dec_pending.
Again, we swap the order.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

In the async encryption-complete function (kcryptd_async_done), the
crypto_async_request passed in may be different from the one passed to
crypto_ablkcipher_encrypt/decrypt.  Only crypto_async_request->data is
guaranteed to be same as the one passed in.  The current
kcryptd_async_done uses the passed-in crypto_async_request directly
which may cause the AES-NI-based AES algorithm implementation to panic.

This patch fixes this bug by only using crypto_async_request->data,
which points to dm_crypt_request, the crypto_async_request passed in.
The original data (convert_context) is gotten from dm_crypt_request.

[mbroz@redhat.com: reworked]
Cc: stable@kernel.org
Signed-off-by: Huang Ying <ying.huang@intel.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

dm-io calls bio_get_nr_vecs to get the maximum number of pages to use
for a given device.  It allocates one additional bio_vec to use
internally but failed to respect BIO_MAX_PAGES, so fix this.

This was the likely cause of:
  https://bugzilla.redhat.com/show_bug.cgi?id=173153

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Fix an error introduced in dm-table-rework-reference-counting.patch.

When there is failure after table initialization, we need to use
dm_table_destroy, not dm_table_put, to free the table.

dm_table_put may be used only after dm_table_get.

Cc: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

When renaming a mapped device validate the length of the new name.

The rename ioctl accepted any correctly-terminated string enclosed
within the data passed from userspace.  The other ioctls enforce a
size limit of DM_NAME_LEN.  If the name is changed and becomes longer
than that, the device can no longer be addressed by name.

Fix it by properly checking for device name length (including
terminating zero).

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Reviewed-by: Jonathan Brassow <jbrassow@redhat.com>
Reviewed-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (21 commits)
  r8169: revert "r8169: read MAC address from EEPROM on init (2nd attempt)"
  r8169: use hardware auto-padding.
  igb: remove ASPM L0s workaround
  netxen: remove old flash check.
  mv643xx_eth: fix unicast address filter corruption on mtu change
  xfrm: Fix xfrm_state_find() wrt. wildcard source address.
  emac: Fix clock control for 405EX and 405EXr chips
  ixgbe: fix multiple unicast address support
  via-velocity: Fix DMA mapping length errors on transmit.
  qlge: bugfix: Pad outbound frames smaller than 60 bytes.
  qlge: bugfix: Move netif_napi_del() to common call point.
  qlge: bugfix: Tell hw to strip vlan header.
  qlge: bugfix: Increase filter on inbound csum.
  dnet: replace obsolete *netif_rx_* functions with *napi_*
  net: Add be2net driver.
  dnet: Fix warnings on 64-bit.
  dnet: Dave DNET ethernet controller driver (updated)
  ipv6:  Fix BUG when disabled ipv6 module is unloaded
  bnx2x: Using DMAE to initialize the chip
  bnx2x: Casting page alignment
  ...

Impact: help prevent extinction of species

The Tasmanian Devil is a shy iconic Australian creature named for its
spine-chilling screech.  It is threatened with extinction due to a
scientifically interesting but horrific transmissible facial cancer.

This one is standing in for Tux for one release using the far less-known
Devil Facial Tux Disguise.

	Save The Tasmanian Devil http://tassiedevil.com.auSigned-off-by: Linux.conf.au Hobart Team <contact@marchsouth.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

filp->f_pos only get updated at the end of the function. Thus d_off of those
dirents who are in the middle will be 0, and this will cause a problem in
glibc's readdir implementation, specifically endless loop. Because when overflow
occurs, f_pos will be set to next dirent to read, however it will be 0, unless
the next one is the last one. So it will start over again and again.

There is a sample program in man 2 gendents. This is the output of the program
running on a multithread program's task dir before this patch is applied:

  $ ./a.out /proc/3807/task
  --------------- nread=128 ---------------
  i-node#  file type  d_reclen  d_off   d_name
    506442  directory    16          1  .
    506441  directory    16          0  ..
    506443  directory    16          0  3807
    506444  directory    16          0  3809
    506445  directory    16          0  3812
    506446  directory    16          0  3861
    506447  directory    16          0  3862
    506448  directory    16          8  3863

This is the output after this patch is applied

  $ ./a.out /proc/3807/task
  --------------- nread=128 ---------------
  i-node#  file type  d_reclen  d_off   d_name
    506442  directory    16          1  .
    506441  directory    16          2  ..
    506443  directory    16          3  3807
    506444  directory    16          4  3809
    506445  directory    16          5  3812
    506446  directory    16          6  3861
    506447  directory    16          7  3862
    506448  directory    16          8  3863
Signed-off-by: Zhang Le <r0bertz@gentoo.org>
Acked-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Impact: fix crashes when tracing cpumasks

While ring-buffer allocation, the cpumasks are allocated too,
including the tracing cpumask and the per-cpu file mask handler.
But these cpumasks are freed accidentally just after.
Fix it.
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237164303-11476-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix a warning during preemptirqsoff selftests

When the preemptirqsoff selftest fails, we see the following
warning:

[    6.050000] Testing tracer preemptirqsoff: .. no entries found ..
------------[ cut here ]------------
[    6.060000] WARNING: at kernel/trace/trace.c:688 tracing_start+0x67/0xd3()
[    6.060000] Modules linked in:
[    6.060000] Pid: 1, comm: swapper Tainted: G
[    6.060000] Call Trace:
[    6.060000]  [<ffffffff802460ff>] warn_slowpath+0xb1/0x100
[    6.060000]  [<ffffffff802a8f5b>] ? trace_preempt_on+0x35/0x4b
[    6.060000]  [<ffffffff802a37fb>] ? tracing_start+0x31/0xd3
[    6.060000]  [<ffffffff802a37fb>] ? tracing_start+0x31/0xd3
[    6.060000]  [<ffffffff80271e0b>] ? __lock_acquired+0xe6/0x1f2
[    6.060000]  [<ffffffff802a37fb>] ? tracing_start+0x31/0xd3
[    6.060000]  [<ffffffff802a3831>] tracing_start+0x67/0xd3
[    6.060000]  [<ffffffff802a8ace>] ? irqsoff_tracer_reset+0x2d/0x57
[    6.060000]  [<ffffffff802a4d1c>] trace_selftest_startup_preemptirqsoff+0x1c8/0x1f1
[    6.060000]  [<ffffffff802a4798>] register_tracer+0x12f/0x241
[    6.060000]  [<ffffffff810250d0>] ? init_irqsoff_tracer+0x0/0x53
[    6.060000]  [<ffffffff8102510b>] init_irqsoff_tracer+0x3b/0x53

This is because in fail case, the preemptirqsoff tracer selftest calls twice
the tracing_start() function:

int
trace_selftest_startup_preemptirqsoff(struct tracer *trace, struct trace_array *tr)
{
        if (!ret && !count) {
                printk(KERN_CONT ".. no entries found ..");
                ret = -1;
                tracing_start(); <-----
                goto out;
        }
        [...]
out:
        trace->reset(tr);
        tracing_start(); <------
        tracing_max_latency = save_max;

        return ret;
}

Since it is well handled in the out path, we don't need the conditional one.
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237159961-7447-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix possible locking imbalance

In case of ring buffer resize failure, tracing_set_tracer forgot to
release trace_types_lock. Fix it.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237151439-6755-5-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Syscall tracing must select kallsysms.

The arch code builds a table to find the syscall metadata by syscall
number. It needs the syscalls names resolution from the symbol table
to know which name found on the syscalls metadatas match a function
pointer from the arch sys_call_table.
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237151439-6755-4-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix syscall tracer enable/disable race

The current thread flag toggling is racy as shown in the following
scenario:

- task A is the last user of syscall tracing, it releases the
  TIF_SYSCALL_FTRACE on each tasks

- at the same time task B start syscall tracing. refcount == 0 so
  it sets up TIF_SYSCALL_FTRACE on each tasks.

The effect of the mixup is unpredictable.
So this fix adds a mutex on {start,stop}_syscall_tracing().
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <1237151439-6755-3-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: cleanup

- Drop unused cpu variable
- Fix some errors on comments
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237151439-6755-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Impact: fix 'stuck' syscall tracer

The syscall tracer uses a refcounter to enable several users
simultaneously.

But the refcounter did not behave correctly and always restored
its value to 0 after calling start_syscall_tracing(). Therefore,
stop_syscall_tracing() couldn't release correctly the tasks from
tracing.

Also the tracer forgot to reset the buffer when it is released.

Drop the pointless refcount decrement on start_syscall_tracing()
and reset the buffer when we release the tracer.

This fixes two reported issue:

- when we switch from syscall tracer to another tracer, syscall
  tracing continued.

- incorrect use of the refcount.
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1237151439-6755-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>