Masami Hiramatsu [Thu, 11 Aug 2011 11:02:47 +0000 (20:02 +0900)]
perf probe: Fix to walk all inline instances
Fix line-range collector to walk all instances of inlined function,
because some execution paths can be optimized out depending on the
function argument of instances.
E.g.)
inline_func (arg) {
if (arg)
do_something;
else
do_another;
}
func_A() {
inline_func(1)
}
func_B() {
inline_func(0)
}
In this case, func_A may have only do_something code and func_B may have
only do_another.
Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <masami.hiramatsu@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pekka Enberg <penberg@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: yrl.pp-manager.tt@hitachi.com Link: http://lkml.kernel.org/r/20110811110247.19900.93702.stgit@fedora15 Signed-off-by: Masami Hiramatsu <masami.hiramatsu@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Masami Hiramatsu [Thu, 11 Aug 2011 11:02:41 +0000 (20:02 +0900)]
perf probe: Fix to search nested inlined functions in CU
Fix perf probe to walk through the lines of all nested inlined function
call sites and declared lines when a whole CU is passed to the line
walker.
The die_walk_lines() can have two different type of DIEs, subprogram (or
inlined-subroutine) DIE and CU DIE.
If a caller passes a subprogram DIE, this means that the walker walk on
lines of given subprogram. In this case, it just needs to search on
direct children of DIE tree for finding call-site information of inlined
function which directly called from given subprogram.
On the other hand, if a caller passes a CU DIE to the walker, this means
that the walker have to walk on all lines in the source files included
in given CU DIE. In this case, it has to search whole DIE trees of all
subprograms to find the call-site information of all nested inlined
functions.
Jiri Olsa [Wed, 1 Jun 2011 19:43:46 +0000 (21:43 +0200)]
perf report: Use properly build_id kernel binaries
If we bring the recorded perf data together with kernel binary from another
machine using:
on server A:
perf archive
on server B:
tar xjvf perf.data.tar.bz2 -C ~/.debug
the build_id kernel dso is not properly recognized during the "perf report"
command on server B.
The reason is, that build_id dsos are added during the session initialization,
while the kernel maps are created during the sample event processing.
The machine__create_kernel_maps functions ends up creating new dso object for
kernel, but it does not check if we already have one added by build_id
processing.
Also the build_id reading ABI quirk added in commit:
populates the "struct build_id_event::pid" with 0, which
is later interpreted as DEFAULT_GUEST_KERNEL_ID.
This is not always correct, so it's better to guess the pid
value based on the "struct build_id_event::header::misc" value.
- Tested with data generated on x86 kernel version v2.6.34
and reported back on x86_64 current kernel.
- Not tested for guest kernel case.
Note the problem stays for PERF_RECORD_MMAP events recorded by perf that
does not use proper pid (HOST_KERNEL_ID/DEFAULT_GUEST_KERNEL_ID). They are
misinterpreted within the current perf code. Probably there's not much we
can do about that.
Cc: Avi Kivity <avi@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Yanmin Zhang <yanmin_zhang@linux.intel.com> Link: http://lkml.kernel.org/r/20110601194346.GB1934@jolsa.brq.redhat.com Signed-off-by: Jiri Olsa <jolsa@redhat.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Jiri Olsa [Mon, 8 Aug 2011 21:03:34 +0000 (23:03 +0200)]
perf sched: Do not delete session object prematurely
The session object is released prematurely when processing events for
latency command. The session's thread objects are used within the
output_lat_thread function.
Runnning following commands:
# perf sched record
# perf sched latency
the latter displays incorrect data and might cause access violation.
Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1312837414-3819-1-git-send-email-jolsa@redhat.com Signed-off-by: Jiri Olsa <jolsa@redhat.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Jiri Olsa [Fri, 22 Jul 2011 11:33:07 +0000 (13:33 +0200)]
perf tools: Add support to install perf python extension
Adding install-python_ext target to install python extension related
files. Installation directory is governed by python distutils package
and follows the DESTDIR variable settings.
Also moving python extension build output into '$(O)python_ext_build'
directory and making it configurable via PYTHON_EXTBUILD variable.
Keeping the '$(O)python/perf.so' file, so it could be used for testing
as of until now.
Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20110722113307.GA1931@jolsa.brq.redhat.com Signed-off-by: Jiri Olsa <jolsa@redhat.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Jonathan Nieder [Fri, 5 Aug 2011 16:58:38 +0000 (18:58 +0200)]
perf tools: do not look at ./config for configuration
In addition to /etc/perfconfig and $HOME/.perfconfig, perf looks for
configuration in the file ./config, imitating git which looks at
$GIT_DIR/config. If ./config is not a perf configuration file, it
fails, or worse, treats it as a configuration file and changes behavior
in some unexpected way.
"config" is not an unusual name for a file to be lying around and perf
does not have a private directory dedicated for its own use, so let's
just stop looking for configuration in the cwd. Callers needing
context-sensitive configuration can use the PERF_CONFIG environment
variable.
Requested-by: Christian Ohm <chr.ohm@gmx.net> Cc: 632923@bugs.debian.org Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Christian Ohm <chr.ohm@gmx.net> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20110805165838.GA7237@elie.gateway.2wire.net Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
So those friggin "spurious" PERF_RECORD_MMAP events were actually a
brain fart copy'n'paste error in the python binding, doh. I.e. they
weren't MMAPs, just SAMPLEs.
Fix it by providing routines for these events instead of using the MMAP
ones.
Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-b0rc8y5jd03f9f11kftodvkm@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Robert Richter [Tue, 7 Jun 2011 09:49:55 +0000 (11:49 +0200)]
x86, perf: Make copy_from_user_nmi() a library function
copy_from_user_nmi() is used in oprofile and perf. Moving it to other
library functions like copy_from_user(). As this is x86 code for 32
and 64 bits, create a new file usercopy.c for unified code.
x86, perf: P4 PMU - Fix typos in comments and style cleanup
This patch:
- fixes typos in comments and clarifies the text
- renames obscure p4_event_alias::original and ::alter members to
::original and ::alternative as appropriate
- drops parenthesis from the return of p4_get_alias_event()
Jiri Olsa [Thu, 14 Jul 2011 09:25:32 +0000 (11:25 +0200)]
perf tools: De-opt the parse_events function
Moving out the option parameter from parse_events function,
and adding new parse_events_option function instead.
The option parameter is used only to carry "struct perf_evlist"
pointer for chaining new events. Putting it away, enable us
to call parse_events from other places without using the
option parameter.
The perf_event_attr struct has two __u32's at the top and
they need to be swapped individually.
With this change I was able to analyze a perf.data collected in a
32-bit PPC VM on an x86 system. I tested both 32-bit and 64-bit
binaries for the Intel analysis side; both read the PPC perf.data
file correctly.
-v2:
- changed the existing perf_event__attr_swap() to swap only elements
of perf_event_attr and exported it for use in swapping the
attributes in the file header
- updated swap_ops used for processing events
Jiri Olsa [Wed, 13 Jul 2011 20:58:18 +0000 (22:58 +0200)]
perf tools: Add missing 'node' alias to the hw_cache[] array
Add "node" as a simple alias for NODE cache events.
The addition of NODE cache events broke the parse_alias
function, so any mismatched event caused the segfault, like:
# ./perf stat -e krava ls
The hw_cache/hw_cache_op/hw_cache_result arrays needs to follow
PERF_COUNT_HW_CACHE_*MAX enums. Adding those MAXs to be size
of those arrays, so possible ommision in future wil not lead to
segfault.
Adding read/write/prefetch as allowed operations for node cache
event.
Masami Hiramatsu [Mon, 27 Jun 2011 07:27:51 +0000 (16:27 +0900)]
perf probe: Support adding probes on offline kernel modules
Support adding probes on offline kernel modules. This enables
perf-probe to trace kernel-module init functions via perf-probe.
If user gives the path of module with -m option, perf-probe
expects the module is offline.
This feature works with --add, --funcs, and --vars.
E.g)
# perf probe -m /lib/modules/`uname -r`/kernel/fs/btrfs/btrfs.ko \
-a "extent_io_init:5 extent_state_cache"
Add new events:
probe:extent_io_init (on extent_io_init:5 with extent_state_cache)
probe:extent_io_init_1 (on extent_io_init:5 with extent_state_cache)
You can now use it on all perf tools, such as:
perf record -e probe:extent_io_init_1 -aR sleep 1
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Link: http://lkml.kernel.org/r/20110627072751.6528.10230.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:27:45 +0000 (16:27 +0900)]
perf probe: Add probed module in front of function
Add probed module name and ":" in front of function name
if -m module option is given. In the result, the symbol
name passed to kprobe-tracer becomes MODULE:FUNCTION,
so that kallsyms can solve it as a symbol in the module
correctly.
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Link: http://lkml.kernel.org/r/20110627072745.6528.26416.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:27:27 +0000 (16:27 +0900)]
perf-probe: Move dwarf library routines to dwarf-aux.{c, h}
Move dwarf library related routines to dwarf-aux.{c,h}.
This includes several minor changes.
- Add simple documents for each API.
- Rename die_find_real_subprogram() to die_find_realfunc()
- Rename line_walk_handler_t to line_walk_callback_t.
- Minor cleanups.
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Link: http://lkml.kernel.org/r/20110627072727.6528.57647.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:27:21 +0000 (16:27 +0900)]
perf probe: Remove redundant dwarf functions
Since there are dwarf_bitsize, dwarf_bitoffset and dwarf_bytesize
defined in libdw, we don't need die_get_bit_size, die_get_bit_offset
and die_get_byte_size anymore.
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Link: http://lkml.kernel.org/r/20110627072721.6528.2747.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:27:03 +0000 (16:27 +0900)]
tracing/kprobe: Update symbol reference when loading module
Since the address of a module-local variable can only be
solved after the target module is loaded, the symbol
fetch-argument should be updated when loading target
module.
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Link: http://lkml.kernel.org/r/20110627072703.6528.75042.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:26:56 +0000 (16:26 +0900)]
tracing/kprobes: Support module init function probing
To support probing module init functions, kprobe-tracer allows
user to define a probe on non-existed function when it is given
with a module name. This also enables user to set a probe on
a function on a specific module, even if a same name (but different)
function is locally defined in another module.
The module name must be in the front of function name and separated
by a ':'. e.g. btrfs:btrfs_init_sysfs
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Link: http://lkml.kernel.org/r/20110627072656.6528.89970.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:26:50 +0000 (16:26 +0900)]
kprobes: Return -ENOENT if probe point doesn't exist
Return -ENOENT if probe point doesn't exist, but still returns
-EINVAL if both of kprobe->addr and kprobe->symbol_name are
specified or both are not specified.
Acked-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> Cc: "David S. Miller" <davem@davemloft.net> Link: http://lkml.kernel.org/r/20110627072650.6528.67329.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Fri, 15 Jul 2011 03:02:27 +0000 (23:02 -0400)]
ftrace: Fix regression where ftrace breaks when modules are loaded
Enabling function tracer to trace all functions, then load a module and
then disable function tracing will cause ftrace to fail.
This can also happen by enabling function tracing on the command line:
ftrace=function
and during boot up, modules are loaded, then you disable function tracing
with 'echo nop > current_tracer' you will trigger a bug in ftrace that
will shut itself down.
The reason is, the new ftrace code keeps ref counts of all ftrace_ops that
are registered for tracing. When one or more ftrace_ops are registered,
all the records that represent the functions that the ftrace_ops will
trace have a ref count incremented. If this ref count is not zero,
when the code modification runs, that function will be enabled for tracing.
If the ref count is zero, that function will be disabled from tracing.
To make sure the accounting was working, FTRACE_WARN_ON()s were added
to updating of the ref counts.
If the ref count hits its max (> 2^30 ftrace_ops added), or if
the ref count goes below zero, a FTRACE_WARN_ON() is triggered which
disables all modification of code.
Since it is common for ftrace_ops to trace all functions in the kernel,
instead of creating > 20,000 hash items for the ftrace_ops, the hash
count is just set to zero, and it represents that the ftrace_ops is
to trace all functions. This is where the issues arrise.
If you enable function tracing to trace all functions, and then add
a module, the modules function records do not get the ref count updated.
When the function tracer is disabled, all function records ref counts
are subtracted. Since the modules never had their ref counts incremented,
they go below zero and the FTRACE_WARN_ON() is triggered.
The solution to this is rather simple. When modules are loaded, and
their functions are added to the the ftrace pool, look to see if any
ftrace_ops are registered that trace all functions. And for those,
update the ref count for the module function records.
Reported-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Masami Hiramatsu [Mon, 27 Jun 2011 07:26:36 +0000 (16:26 +0900)]
tracing/kprobes: Rename probe_* to trace_probe_*
Rename probe_* to trace_probe_* for avoiding namespace
confliction. This also fixes improper names of find_probe_event()
and cleanup_all_probes() to find_trace_probe() and
release_all_trace_probes() respectively.
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/20110627072636.6528.60374.stgit@fedora15 Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Instead of hw_nmi_watchdog_set_attr() weak function
and appropriate x86_pmu::hw_watchdog_set_attr() call
we introduce even alias mechanism which allow us
to drop this routines completely and isolate quirks
of Netburst architecture inside P4 PMU code only.
The main idea remains the same though -- to allow
nmi-watchdog and perf top run simultaneously.
Note the aliasing mechanism applies to generic
PERF_COUNT_HW_CPU_CYCLES event only because arbitrary
event (say passed as RAW initially) might have some
additional bits set inside ESCR register changing
the behaviour of event and we can't guarantee anymore
that alias event will give the same result.
P.S. Thanks a huge to Don and Steven for for testing
and early review.
Acked-by: Don Zickus <dzickus@redhat.com> Tested-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> CC: Ingo Molnar <mingo@elte.hu> CC: Peter Zijlstra <a.p.zijlstra@chello.nl> CC: Stephane Eranian <eranian@google.com> CC: Lin Ming <ming.m.lin@intel.com> CC: Arnaldo Carvalho de Melo <acme@redhat.com> CC: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/20110708201712.GS23657@sun Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Thu, 14 Jul 2011 20:36:53 +0000 (16:36 -0400)]
tracing: Have dynamic size event stack traces
Currently the stack trace per event in ftace is only 8 frames.
This can be quite limiting and sometimes useless. Especially when
the "ignore frames" is wrong and we also use up stack frames for
the event processing itself.
Change this to be dynamic by adding a percpu buffer that we can
write a large stack frame into and then copy into the ring buffer.
For interrupts and NMIs that come in while another event is being
process, will only get to use the 8 frame stack. That should be enough
as the task that it interrupted will have the full stack frame anyway.
Requested-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
While attempting to create a timechart of boot up I found perf didn't
tolerate modules being loaded/unloaded. This patch fixes this by
reading the file once and then writing the size read at the correct
point in the file. It also simplifies the code somewhat.
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Signed-off-by: Sonny Rao <sonnyrao@chromium.org> Signed-off-by: Michael Neuling <mikey@neuling.org> Link: http://lkml.kernel.org/r/10011.1310614483@neuling.org Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Wed, 13 Jul 2011 19:11:02 +0000 (15:11 -0400)]
ftrace: Fix dynamic selftest failure on some archs
Archs that do not implement CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST, will
fail the dynamic ftrace selftest.
The function tracer has a quick 'off' variable that will prevent
the call back functions from being called. This variable is called
function_trace_stop. In x86, this is implemented directly in the mcount
assembly, but for other archs, an intermediate function is used called
ftrace_test_stop_func().
In dynamic ftrace, the function pointer variable ftrace_trace_function is
used to update the caller code in the mcount caller. But for archs that
do not have CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST set, it only calls
ftrace_test_stop_func() instead, which in turn calls __ftrace_trace_function.
When more than one ftrace_ops is registered, the function it calls is
ftrace_ops_list_func(), which will iterate over all registered ftrace_ops
and call the callbacks that have their hash matching.
The issue happens when two ftrace_ops are registered for different functions
and one is then unregistered. The __ftrace_trace_function is then pointed
to the remaining ftrace_ops callback function directly. This mean it will
be called for all functions that were registered to trace by both ftrace_ops
that were registered.
This is not an issue for archs with CONFIG_HAVE_FUNCTION_TRACE_MCOUNT_TEST,
because the update of ftrace_trace_function doesn't happen until after all
functions have been updated, and then the mcount caller is updated. But
for those archs that do use the ftrace_test_stop_func(), the update is
immediate.
The dynamic selftest fails because it hits this situation, and the
ftrace_ops that it registers fails to only trace what it was suppose to
and instead traces all other functions.
The solution is to delay the setting of __ftrace_trace_function until
after all the functions have been updated according to the registered
ftrace_ops. Also, function_trace_stop is set during the update to prevent
function tracing from calling code that is caused by the function tracer
itself.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Wed, 13 Jul 2011 19:08:31 +0000 (15:08 -0400)]
ftrace: Update filter when tracing enabled in set_ftrace_filter()
Currently, if set_ftrace_filter() is called when the ftrace_ops is
active, the function filters will not be updated. They will only be updated
when tracing is disabled and re-enabled.
Update the functions immediately during set_ftrace_filter().
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Wed, 13 Jul 2011 19:03:44 +0000 (15:03 -0400)]
ftrace: Balance records when updating the hash
Whenever the hash of the ftrace_ops is updated, the record counts
must be balance. This requires disabling the records that are set
in the original hash, and then enabling the records that are set
in the updated hash.
Moving the update into ftrace_hash_move() removes the bug where the
hash was updated but the records were not, which results in ftrace
triggering a warning and disabling itself because the ftrace_ops filter
is updated while the ftrace_ops was registered, and then the failure
happens when the ftrace_ops is unregistered.
The current code will not trigger this bug, but new code will.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Sat, 25 Jun 2011 03:28:13 +0000 (23:28 -0400)]
ftrace: Do not disable interrupts for modules in mcount update
When I mounted an NFS directory, it caused several modules to be loaded. At the
time I was running the preemptirqsoff tracer, and it showed the following
output:
That's over 1ms that interrupts are disabled on a Real-Time kernel!
Looking at the cause (being the ftrace author helped), I found that the
interrupts are disabled before the code modification of mcounts into nops. The
interrupts only need to be disabled on start up around this code, not when
modules are being loaded.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Tue, 14 Jun 2011 23:02:29 +0000 (19:02 -0400)]
tracing: Still trace filtered irq functions when irq trace is disabled
If a function is set to be traced by the set_graph_function, but the
option funcgraph-irqs is zero, and the traced function happens to be
called from a interrupt, it will not be traced.
The point of funcgraph-irqs is to not trace interrupts when we are
preempted by an irq, not to not trace functions we want to trace that
happen to be *in* a irq.
Luckily the current->trace_recursion element is perfect to add a flag
to help us be able to trace functions within an interrupt even when
we are not tracing interrupts that preempt the trace.
Steven Rostedt [Sat, 2 Jul 2011 03:04:36 +0000 (23:04 -0400)]
tracing, x86/irq: Do not trace arch_local_{*,irq_*}() functions
I triggered a triple fault with gcc 4.5.1 because it did not
honor the inline annotation to arch_local_save_flags() function
and that function was added to the pool of functions traced by
the function tracer.
When preempt_schedule() called arch_local_save_flags() (called
by irqs_disabled()), it was traced, but the first thing the
function tracer does is disable preemption. When it enables
preemption, the NEED_RESCHED flag will not have been cleared and
the preemption check will trigger the call to preempt_schedule()
again.
Although the dynamic function tracer crashed immediately, the
static version of the function tracer (CONFIG_DYNAMIC_FTRACE is
not set) actually was able to show where the problem was.
It went on for a while before it triple faulted with a corrupted
stack.
The arch_local_save_flags and arch_local_irq_* functions should
not be traced. Even though they are marked as inline, gcc may
still make them a function and enable tracing of them.
The simple solution is to just mark them as notrace. I had to
add the <linux/types.h> for this file to include the notrace
tag.
Steven Rostedt [Thu, 7 Jul 2011 15:09:22 +0000 (11:09 -0400)]
ftrace: Fix regression of :mod:module function enabling
The new code that allows different utilities to pick and choose
what functions they trace broke the :mod: hook that allows users
to trace only functions of a particular module.
The reason is that the :mod: hook bypasses the hash that is setup
to allow individual users to trace their own functions and uses
the global hash directly. But if the global hash has not been
set up, it will cause a bug:
Steven Rostedt [Tue, 5 Jul 2011 18:32:51 +0000 (14:32 -0400)]
tracing: Have "enable" file use refcounts like the "filter" file
The "enable" file for the event system can be removed when a module
is unloaded and the event system only has events from that module.
As the event system nr_events count goes to zero, it may be freed
if its ref_count is also set to zero.
Like the "filter" file, the "enable" file may be opened by a task and
referenced later, after a module has been unloaded and the events for
that event system have been removed.
Although the "filter" file referenced the event system structure,
the "enable" file only references a pointer to the event system
name. Since the name is freed when the event system is removed,
it is possible that an access to the "enable" file may reference
a freed pointer.
Update the "enable" file to use the subsystem_open() routine that
the "filter" file uses, to keep a reference to the event system
structure while the "enable" file is opened.
Cc: <stable@kernel.org> Reported-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Steven Rostedt [Tue, 5 Jul 2011 15:36:06 +0000 (11:36 -0400)]
tracing: Fix bug when reading system filters on module removal
The event system is freed when its nr_events is set to zero. This happens
when a module created an event system and then later the module is
removed. Modules may share systems, so the system is allocated when
it is created and freed when the modules are unloaded and all the
events under the system are removed (nr_events set to zero).
The problem arises when a task opened the "filter" file for the
system. If the module is unloaded and it removed the last event for
that system, the system structure is freed. If the task that opened
the filter file accesses the "filter" file after the system has
been freed, the system will access an invalid pointer.
By adding a ref_count, and using it to keep track of what
is using the event system, we can free it after all users
are finished with the event system.
Cc: <stable@kernel.org> Reported-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Anton Blanchard [Mon, 4 Jul 2011 11:57:50 +0000 (21:57 +1000)]
perf report/annotate/script: Add option to specify a CPU range
Add an option to perf report/annotate/script to specify which
CPUs to operate on. This enables us to take a single system wide
profile and analyse each CPU (or group of CPUs) in isolation.
This was useful when profiling a multiprocess workload where the
bottleneck was on one CPU but this was hidden in the overall
profile. Per process and per thread breakdowns didn't help
because multiple processes were running on each CPU and no
single process consumed an entire CPU.
The patch converts the list of CPUs returned by cpu_map__new
into a bitmap for fast lookup. I wanted to use -C to be
consistent with perf top/record/stat, but unfortunately perf
report already uses -C <comms>.
v2: Incorporate suggestions from David Ahern:
- Added -c to perf script
- Check that SAMPLE_CPU is set when -c is used
- Update documentation
v3: Create perf_session__cpu_bitmap()
Signed-off-by: Anton Blanchard <anton@samba.org> Acked-by: David Ahern <dsahern@gmail.com> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Link: http://lkml.kernel.org/r/20110704215750.11647eb9@kryten Signed-off-by: Ingo Molnar <mingo@elte.hu>
x86: Don't use frame pointer to save old stack on irq entry
rbp is used in SAVE_ARGS_IRQ to save the old stack pointer
in order to restore it later in ret_from_intr.
It is convenient because we save its value in the irq regs
and it's easily restored using the leave instruction.
However this is a kind of abuse of the frame pointer which
role is to help unwinding the kernel by chaining frames
together, each node following the return address to the
previous frame.
But although we are breaking the frame by changing the stack
pointer, there is no preceding return address before the new
frame. Hence using the frame pointer to link the two stacks
breaks the stack unwinders that find a random value instead of
a return address here.
There is no workaround that can work in every case. We are using
the fixup_bp_irq_link() function to dereference that abused frame
pointer in the case of non nesting interrupt (which means stack
changed).
But that doesn't fix the case of interrupts that don't change the
stack (but we still have the unconditional frame link), which is
the case of hardirq interrupting softirq. We have no way to detect
this transition so the frame irq link is considered as a real frame
pointer and the return address is dereferenced but it is still a
spurious one.
There are two possible results of this: either the spurious return
address, a random stack value, luckily belongs to the kernel text
and then the unwinding can continue and we just have a weird entry
in the stack trace. Or it doesn't belong to the kernel text and
unwinding stops there.
This is the reason why stacktraces (including perf callchains) on
irqs that interrupted softirqs don't work very well.
To solve this, we don't save the old stack pointer on rbp anymore
but we save it to a scratch register that we push on the new
stack and that we pop back later on irq return.
This preserves the whole frame chain without spurious return addresses
in the middle and drops the need for the horrid fixup_bp_irq_link()
workaround.
And finally irqs that interrupt softirq are sanely unwinded.
The save_regs function that saves the regs on low level
irq entry is complicated because of the fact it changes
its stack in the middle and also because it manipulates
data allocated in the caller frame and accesses there
are directly calculated from callee rsp value with the
return address in the middle of the way.
This complicates the static stack offsets calculation and
require more dynamic ones. It also needs a save/restore
of the function's return address.
To simplify and optimize this, turn save_regs() into a
macro.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jan Beulich <JBeulich@novell.com>
x86: Fetch stack from regs when possible in dump_trace()
When regs are passed to dump_stack(), we fetch the frame
pointer from the regs but the stack pointer is taken from
the current frame.
Thus the frame and stack pointers may not come from the same
context. For example this can result in the unwinder to
think the context is in irq, due to the current value of
the stack, but the frame pointer coming from the regs points
to a frame from another place. It then tries to fix up
the irq link but ends up dereferencing a random frame
pointer that doesn't belong to the irq stack:
In order to prepare for fetching the stack pointer from the
regs when possible in dump_trace() instead of taking the
local one, save the current stack pointer in perf live regs saving.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Peter Zijlstra [Fri, 1 Jul 2011 13:24:53 +0000 (15:24 +0200)]
perf, powerpc: Fix build borkage
The patch a8b0ca17b80e ("perf: Remove the nmi parameter from the swevent
and overflow interface") missed a spot in the ppc hw_breakpoint code,
fix this up so things compile again.
Reported-by: Ingo Molnar <mingo@elte.hu> Cc: Anton Blanchard <anton@samba.org> Cc: Eric B Munson <emunson@mgebm.net> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/n/tip-09pfip95g88s70iwkxu6nnbt@git.kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
Avi Kivity [Wed, 29 Jun 2011 15:42:36 +0000 (18:42 +0300)]
x86, perf: Add constraints for architectural PMU
The v1 PMU does not have any fixed counters. Using the v2 constraints,
which do have fixed counters, causes an additional choice to be present
in the weight calculation, but not when actually scheduling the event,
leading to an event being not scheduled at all.
Avi Kivity [Wed, 29 Jun 2011 15:42:35 +0000 (18:42 +0300)]
perf: Add context field to perf_event
The perf_event overflow handler does not receive any caller-derived
argument, so many callers need to resort to looking up the perf_event
in their local data structure. This is ugly and doesn't scale if a
single callback services many perf_events.
Fix by adding a context parameter to perf_event_create_kernel_counter()
(and derived hardware breakpoints APIs) and storing it in the perf_event.
The field can be accessed from the callback as event->overflow_handler_context.
All callers are updated.
Peter Zijlstra [Fri, 22 Apr 2011 21:37:06 +0000 (23:37 +0200)]
perf, arch: Add generic NODE cache events
Add a NODE level to the generic cache events which is used to measure
local vs remote memory accesses. Like all other cache events, an
ACCESS is HIT+MISS, if there is no way to distinguish between reads
and writes do reads only etc..
The below needs filling out for !x86 (which I filled out with
unsupported events).
I'm fairly sure ARM can leave it like that since it doesn't strike me as
an architecture that even has NUMA support. SH might have something since
it does appear to have some NUMA bits.
Sparc64, PowerPC and MIPS certainly want a good look there since they
clearly are NUMA capable.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: David Miller <davem@davemloft.net> Cc: Anton Blanchard <anton@samba.org> Cc: David Daney <ddaney@caviumnetworks.com> Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Will Deacon <will.deacon@arm.com> Cc: Robert Richter <robert.richter@amd.com> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/r/1303508226.4865.8.camel@laptop Signed-off-by: Ingo Molnar <mingo@elte.hu>
perf_events: Add Intel Sandy Bridge offcore_response low-level support
This patch adds Intel Sandy Bridge offcore_response support by
providing the low-level constraint table for those events.
On Sandy Bridge, there are two offcore_response events. Each uses
its own dedictated extra register. But those registers are NOT shared
between sibling CPUs when HT is on unlike Nehalem/Westmere. They are
always private to each CPU. But they still need to be controlled within
an event group. All events within an event group must use the same
value for the extra MSR. That's not controlled by the second patch in
this series.
Furthermore on Sandy Bridge, the offcore_response events have NO
counter constraints contrary to what the official documentation
indicates, so drop the events from the contraint table.
perf_events: Fix validation of events using an extra reg
The validate_group() function needs to validate events with
extra shared regs. Within an event group, only events with
the same value for the extra reg can co-exist. This was not
checked by validate_group() because it was missing the
shared_regs logic.
This patch changes the allocation of the fake cpuc used for
validation to also point to a fake shared_regs structure such
that group events be properly testing.
It modifies __intel_shared_reg_get_constraints() to use
spin_lock_irqsave() to avoid lockdep issues.
perf_events: Update Intel extra regs shared constraints management
This patch improves the code managing the extra shared registers
used for offcore_response events on Intel Nehalem/Westmere. The
idea is to use static allocation instead of dynamic allocation.
This simplifies greatly the get and put constraint routines for
those events.
The patch also renames per_core to shared_regs because the same
data structure gets used whether or not HT is on. When HT is
off, those events still need to coordination because they use
a extra MSR that has to be shared within an event group.
Peter Zijlstra [Mon, 27 Jun 2011 14:47:16 +0000 (16:47 +0200)]
perf: Remove the perf_output_begin(.sample) argument
Since only samples call perf_output_sample() its much saner (and more
correct) to put the sample logic in there than in the
perf_output_begin()/perf_output_end() pair.
Saves a useless argument, reduces conditionals and shrinks
struct perf_output_handle, win!
Peter Zijlstra [Mon, 27 Jun 2011 12:41:57 +0000 (14:41 +0200)]
perf: Remove the nmi parameter from the swevent and overflow interface
The nmi parameter indicated if we could do wakeups from the current
context, if not, we would set some state and self-IPI and let the
resulting interrupt do the wakeup.
For the various event classes:
- hardware: nmi=0; PMI is in fact an NMI or we run irq_work_run from
the PMI-tail (ARM etc.)
- tracepoint: nmi=0; since tracepoint could be from NMI context.
- software: nmi=[0,1]; some, like the schedule thing cannot
perform wakeups, and hence need 0.
As one can see, there is very little nmi=1 usage, and the down-side of
not using it is that on some platforms some software events can have a
jiffy delay in wakeup (when arch_irq_work_raise isn't implemented).
The up-side however is that we can remove the nmi parameter and save a
bunch of conditionals in fast paths.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Michael Cree <mcree@orcon.net.nz> Cc: Will Deacon <will.deacon@arm.com> Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com> Cc: Anton Blanchard <anton@samba.org> Cc: Eric B Munson <emunson@mgebm.net> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: David S. Miller <davem@davemloft.net> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jason Wessel <jason.wessel@windriver.com> Cc: Don Zickus <dzickus@redhat.com> Link: http://lkml.kernel.org/n/tip-agjev8eu666tvknpb3iaj0fg@git.kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cyrill Gorcunov [Thu, 23 Jun 2011 12:49:18 +0000 (16:49 +0400)]
perf, x86: Add hw_watchdog_set_attr() in a sake of nmi-watchdog on P4
Due to restriction and specifics of Netburst PMU we need a separated
event for NMI watchdog. In particular every Netburst event
consumes not just a counter and a config register, but also an
additional ESCR register.
Since ESCR registers are grouped upon counters (i.e. if ESCR is occupied
for some event there is no room for another event to enter until its
released) we need to pick up the "least" used ESCR (or the most available
one) for nmi-watchdog purposes -- so MSR_P4_CRU_ESCR2/3 was chosen.
With this patch nmi-watchdog and perf top should be able to run simultaneously.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org> CC: Lin Ming <ming.m.lin@intel.com> CC: Arnaldo Carvalho de Melo <acme@redhat.com> CC: Frederic Weisbecker <fweisbec@gmail.com> Tested-and-reviewed-by: Don Zickus <dzickus@redhat.com> Tested-and-reviewed-by: Stephane Eranian <eranian@google.com> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20110623124918.GC13050@sun Signed-off-by: Ingo Molnar <mingo@elte.hu>
Eric B Munson [Fri, 24 Jun 2011 16:26:26 +0000 (12:26 -0400)]
events: Ensure that timers are updated without requiring read() call
The event tracing infrastructure exposes two timers which should be updated
each time the value of the counter is updated. Currently, these counters are
only updated when userspace calls read() on the fd associated with an event.
This means that counters which are read via the mmap'd page exclusively never
have their timers updated. This patch adds ensures that the timers are updated
each time the values in the mmap'd page are updated.
Richard Kennedy [Tue, 7 Jun 2011 15:33:38 +0000 (16:33 +0100)]
perf: Remove 64-bit alignment padding from perf_event_context
Reorder perf_event_context to remove 8 bytes of 64 bit alignment padding
shrinking its size to 192 bytes, allowing it to fit into a smaller slab
and use one fewer cache lines.
Vince Weaver [Wed, 1 Jun 2011 19:15:36 +0000 (15:15 -0400)]
perf_events: Fix perf buffer watermark setting
Since 2.6.36 (specifically commit d57e34fdd60b ("perf: Simplify the
ring-buffer logic: make perf_buffer_alloc() do everything needed"),
the perf_buffer_init_code() has been mis-setting the buffer watermark
if perf_event_attr.wakeup_events has a non-zero value.
This is because perf_event_attr.wakeup_events is a union with
perf_event_attr.wakeup_watermark.
This commit re-enables the check for perf_event_attr.watermark being
set before continuing with setting a non-default watermark.
This bug is most noticable when you are trying to use PERF_IOC_REFRESH
with a value larger than one and perf_event_attr.wakeup_events is set to
one. In this case the buffer watermark will be set to 1 and you will
get extraneous POLL_IN overflows rather than POLL_HUP as expected.
[ avoid using attr.wakeup_events when attr.watermark is set ]
Peter Zijlstra [Tue, 28 Jun 2011 10:15:51 +0000 (12:15 +0200)]
irq_work, alpha: Fix up arch hooks
Commit e360adbe29 ("irq_work: Add generic hardirq context
callbacks") fouled up the Alpha bit, not properly naming the
arch specific function that raises the 'self-IPI'.
Peter Zijlstra [Mon, 27 Jun 2011 15:22:43 +0000 (17:22 +0200)]
irq_work, ppc: Fix up arch hooks
Commit e360adbe29 ("irq_work: Add generic hardirq context
callbacks") fouled up the ppc bit, not properly naming the
arch specific function that raises the 'self-IPI'.
Cc: Huang Ying <ying.huang@intel.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Anton Blanchard <anton@samba.org> Cc: Eric B Munson <emunson@mgebm.net> Cc: stable@kernel.org # 37+ Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/n/tip-eg0aqien8p1aqvzu9dft6dtv@git.kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
perf tools: Allow sort dimensions to be registered more than once
So that the parent sort dimension can be registered twice: once
if we add it as an explicit sort dimension (-s parent) and twice
if we request a parent filter (-p foo).
We'll have only one parent sort dimension in the end but this
allows to override the default parent filter with we gave in "-p"
option. The goal of this is to prepare to allow the use of
"-s parent" and "-p foo" at the same time, ie: sort by filtered
parent.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: David Ahern <dsahern@gmail.com> Cc: Sam Liao <phyomh@gmail.com>
perf tools: Don't display ignored entries on stdio ui
As for newt ui, don't display entries that have been marked
as ignored.
The practical current effect of this is to make parent
filtering really working. Before, entries that were ignored
were given a null parent but were still displayed. This
resulted in some weird effects:
Add "caller/callee" option to support inverted butterfly report,
in the inverted report (with caller option), the call graph start
from the callee's ancestor. Users can use such view to catch system's
performance bottleneck from a sysprof like view. Using this option
with specified sort order like pid gives us high level view of call
graph statistics.
Also add "-G" alias for inverted call graph.
Signed-off-by: Sam Liao <phyomh@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Stephane Eranian <eranian@google.com> Cc: David Ahern <dsahern@gmail.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Hugh Dickins [Mon, 27 Jun 2011 23:18:20 +0000 (16:18 -0700)]
drm/i915: more struct_mutex locking
When auditing the locking in i915_gem.c (for a prospective change which
I then abandoned), I noticed two places where struct_mutex is not held
across GEM object manipulations that would usually require it.
Since one is in initial setup and the other in driver unload, I'm
guessing the mutex is not required for either; but post a patch in case
it is.
Signed-off-by: Hugh Dickins <hughd@google.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Keith Packard <keithp@keithp.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>