The callchains are fed with an array of a fixed size.
As a result we iterate over each callchains three times:
- 1st to resolve symbols
- 2nd to filter out context boundaries
- 3rd for the insertion into the tree
This also involves some pairs of memory allocation/deallocation
everytime we insert a callchain, for the filtered out array of
addresses and for the array of symbols that comes along.
Instead, feed the callchains through a linked list with persistent
allocations. It brings several pros like:
- Merge the 1st and 2nd iterations in one. That was possible before
but in a way that would involve allocating an array slightly taller
than necessary because we don't know in advance the number of context
boundaries to filter out.
- Much lesser allocations/deallocations. The linked list keeps
persistent empty entries for the next usages and is extendable at
will.
- Makes it easier for multiple sources of callchains to feed a
stacktrace together. This is deemed to pave the way for cfi based
callchains wherein traditional frame pointer based kernel
stacktraces will precede cfi based user ones, producing an overall
callchain which size is hardly predictable. This requirement
makes the static array obsolete and makes a linked list based
iterator a much more flexible fit.
Basic testing on a big perf file containing callchains (~ 176 MB)
has shown a throughput gain of about 11% with perf report.
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1294977121-5700-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Each histogram entry has a callchain root that stores the
callchain samples. However we forgot to initialize the
tracking of children hits of these roots, which then got
random values on their creation.
The root children hits is multiplied by the minimum percentage
of hits provided by the user, and the result becomes the minimum
hits expected from children branches. If the random value due
to the uninitialization is big enough, then this minimum number
of hits can be huge and eventually filter every children branches.
The end result was invisible callchains. All we need to
fix this is to initialize the children hits of the root.
Reported-by: Christoph Hellwig <hch@infradead.org>
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: Paul Mackerras <paulus@samba.org>
Cc: 2.6.32.x-2.6.35.y <stable@kernel.org>
If we sort the histograms by comm, which is the default,
we need to merge some of them, typically different thread
histograms of a same process, or just same comm. But during
this merge, we forgot to merge callchains.
So imagine we have three threads (tids: 1000, 1001, 1002) that
belong to comm "foo".
tid 1000 got 100 events
tid 1001 got 10 events
tid 1002 got 3 events
Once we merge these histograms to get a per comm result, we'll
finally get:
"foo" got 113 events
The problem is if we merge 1000 and 1001 histograms into 1002, then
the end merge result, wrt callchains, will be only callchains that
belong to 1002.
This is because we haven't handled callchains in the merge. Only those
from one of the threads inside a common comm survive.
It means during this merge, we can lose a lot of callchains.
Fix this by implementing callchains merge and apply it on histograms
that collapse.
Reported-by: Christoph Hellwig <hch@infradead.org>
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: Paul Mackerras <paulus@samba.org>
Do that to start a consistant callchain API namespace.
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: Paul Mackerras <paulus@samba.org>
Cc: Christoph Hellwig <hch@infradead.org>
In order to implement callchains collapsing, we need to keep
track of the maximum depth in a histogram tree of callchains.
This way we'll avoid allocating an arbitrary temporary buffer
size on callchain merge time.
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: Paul Mackerras <paulus@samba.org>
Cc: Christoph Hellwig <hch@infradead.org>
Hists have their hits increased by the event period. And this
period based counting is the foundation of all the stats in
perf report.
But callchains still use the raw number of hits, without taking
the period into account. So when we compute the percentage,
absolute based percentages are totally broken, and relative ones
too in the first parent level. Because we pass the number of events
muliplied by their period as the total number of hits to the
callchain filtering, while callchains expect this number to be
the number of raw hits.
perf report -g graph was simply not working, showing no graph unless
the min percent was zero. And even there the percentage of the
branches was always 0. And may be fractal filtering was broken on
the first branch level too.
flat also was broken, but it was hidden because of other breakages.
Anyway fix this by counting using periods on callchains.
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: Paul Mackerras <paulus@samba.org>
Initialize the callchain radix tree root correctly.
When we walk through the parents, we must stop after the root, but
since it wasn't well initialized, its parent pointer was random.
Also the number of hits was random because uninitialized, hence it
was part of the callchain while the root doesn't contain anything.
This fixes segfaults and percentages followed by empty callchains
while running:
perf report -g flat
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: 2.6.31.x-2.6.34.x <stable@kernel.org>
Simplifying the tools that were using both in sequence and allowing
upcoming simplifications, such as Arun's patch to sort by cpus.
Cc: David S. Miller <davem@davemloft.net>
Cc: Frédéric 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>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
We were still using the pathname found on the MMAP event, that could not
be the one we used when recording, so use the build-id cache for that,
only falling back to use the pathname in the MMAP event if no build-ids
are available.
With this we now also are able to do secure, seamless offline annotation.
Example:
[root@doppio linux-2.6-tip]# perf report -g none -v 2> /dev/null | head -10
8.12% Xorg /usr/lib64/libpixman-1.so.0.14.0 0x0000000000026d02 B [.] pixman_rasterize_edges
4.68% firefox /usr/lib64/xulrunner-1.9.1/libxul.so 0x00000000005dbdba B [.] 0x000000005dbdba
3.70% swapper /lib/modules/2.6.34-rc6/build/vmlinux 0xffffffff81022cea ! [k] read_hpet
2.96% init /lib/modules/2.6.34-rc6/build/vmlinux 0xffffffff81022cea ! [k] read_hpet
2.73% swapper /lib/modules/2.6.34-rc6/build/vmlinux 0xffffffff8100a738 ! [k] mwait_idle_with_hints
[root@doppio linux-2.6-tip]# perf annotate -v pixman_rasterize_edges 2>&1 | grep Executing
Executing: objdump --start-address=0x000000371ce26670 --stop-address=0x000000371ce2709f -dS /root/.debug/.build-id/bd/6ac5199137aaeb279f864717d8d061477466c1|grep -v /root/.debug/.build-id/bd/6ac5199137aaeb279f864717d8d061477466c1|expand
[root@doppio linux-2.6-tip]# perf buildid-list | grep libpixman-1.so.0.14.0
bd6ac5199137aaeb279f864717d8d061477466c1 /usr/lib64/libpixman-1.so.0.14.0
[root@doppio linux-2.6-tip]#
Reported-by: Stephane Eranian <eranian@google.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Works by adding a third parameter to the '-g' argument, after the graph
type and minimum percentage, for example:
[root@doppio linux-2.6-tip]# perf report -g fractal,0.5,2
Will show only the first two symbols where at least 0.5% of the samples
took place.
All the other symbols that don't fall outside these constraints will be
put together in the last entry, prefixed with "[...]" and the total
percentage for them.
Suggested-by: Arjan van de Ven <arjan@linux.intel.com>
Acked-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
We need this to know where a symbol in a callchain came from,
for various reasons, among them precise annotation from a
TUI/GUI tool.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1269459619-982-5-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Callchains have markers inside their capture to tell we
enter a context (kernel, user, ...).
Those are not displayed in the callchains but they are
incidentally an active part of the radix tree where
callchains are stored, just like any other address.
If we have the two following callchains:
addr1 -> addr2 -> user context -> addr3
addr1 -> addr2 -> user context -> addr4
addr1 -> addr2 -> addr 5
This is pretty common if addr1 and addr2 are part of an
interrupt path, addr3 and addr4 are user addresses and
addr5 is a kernel non interrupt path.
This will be stored as follows in the tree:
addr1
addr2
/ \
/ addr5
user context
/ \
addr3 addr4
But we ignore the context markers in the report, hence
the addr3 and addr4 will appear as orphan branches:
|--28.30%-- hrtimer_interrupt
| smp_apic_timer_interrupt
| apic_timer_interrupt
| | <------------- here, no parent!
| | |
| | |--11.11%-- 0x7fae7bccb875
| | |
| | |--11.11%-- 0xffffffffff60013b
| | |
| | |--11.11%-- __pthread_mutex_lock_internal
| | |
| | |--11.11%-- __errno_location
Fix this by removing the context markers when we process the
callchains to the tree.
Reported-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1269274173-20328-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
There was a colorful mix of header guards - standardize them.
Signed-off-by: John Kacur <jkacur@redhat.com>
LKML-Reference: <alpine.LFD.2.00.0909241756530.11383@localhost.localdomain>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Factorize the dso mapping helpers into a single purpose common file
"util/map.c"
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Brice Goglin <Brice.Goglin@inria.fr>
If we recorded with -g option to record the callchain, right now
we require a -g option to perf report as well - and people reported
this as unnecessary complication: the user already specified -g
once, no need to require it a second time.
So if the recording includes call-chains, display the callchain by
default from perf report.
( The user can override this default using "-g none" option from
perf report. )
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <1249690585-9145-3-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The callchain fractal mode builds each new total hits in a new
branch of profiling by using the parent's hits of the current
branch plus the hits of the children.
This is wrong, the total hits of a branch should be made of the
sum of every children hits, we must ignore the parent hits in
this scope.
This patch also fixes another mistake with the hit counting.
Now the rates are correct.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The current callchain displays the overhead rates as absolute:
relative to the total overhead.
This patch provides relative overhead percentage, in which each
branch of the callchain tree is a independant instrumentated object.
This provides a 'fractal' view of the call-chain profile: each
sub-graph looks like a profile in itself - relative to its parent.
You can produce such output by using the "fractal" mode
that you can abbreviate via f, fr, fra, frac, etc...
./perf report -s sym -c fractal
Example:
8.46% [k] copy_user_generic_string
|
|--52.01%-- generic_file_aio_read
| do_sync_read
| vfs_read
| |
| |--97.20%-- sys_pread64
| | system_call_fastpath
| | pread64
| |
| --2.81%-- sys_read
| system_call_fastpath
| __read
|
|--39.85%-- generic_file_buffered_write
| __generic_file_aio_write_nolock
| generic_file_aio_write
| do_sync_write
| reiserfs_file_write
| vfs_write
| |
| |--97.05%-- sys_pwrite64
| | system_call_fastpath
| | __pwrite64
| |
| --2.95%-- sys_write
| system_call_fastpath
| __write_nocancel
[...]
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1246772361-9960-5-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Callchains output may become a burden on a trace because even
rarely hit site are exposed. This can be too much information.
Let the user set a threshold as a minimum percent of hits using
the new pattern for the -c option:
-c mode,min_percent
Example:
$ perf report -s sym -c flat,4
8.25% [k] copy_user_generic_string
4.19%
copy_user_generic_string
generic_file_aio_read
do_sync_read
vfs_read
sys_pread64
system_call_fastpath
pread64
5.39% [k] search_by_key
4.63% 0x00000000009e0a
2.36% [k] memcpy_c
[...]
$ perf report -s sym -c graph,2
8.25% [k] copy_user_generic_string
|
|--4.31%-- generic_file_aio_read
| do_sync_read
| vfs_read
| |
| --4.19%-- sys_pread64
| system_call_fastpath
| pread64
|
--3.24%-- generic_file_buffered_write
__generic_file_aio_write_nolock
generic_file_aio_write
do_sync_write
reiserfs_file_write
vfs_write
|
--3.14%-- sys_pwrite64
system_call_fastpath
__pwrite64
5.39% [k] search_by_key
|
--2.23%-- reiserfs_update_sd_size
4.63% 0x00000000009e0a
2.36% [k] memcpy_c
[...]
You can also omit it and it will default to 0.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1246558475-10624-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Currently, the printing of callchains is done in a single
vertical level, this is the "flat" mode:
8.25% [k] copy_user_generic_string
4.19%
copy_user_generic_string
generic_file_aio_read
do_sync_read
vfs_read
sys_pread64
system_call_fastpath
pread64
This patch introduces a new "graph" mode which provides a
hierarchical output of factorized paths recursively sorted:
8.25% [k] copy_user_generic_string
|
|--4.31%-- generic_file_aio_read
| do_sync_read
| vfs_read
| |
| |--4.19%-- sys_pread64
| | system_call_fastpath
| | pread64
| |
| --0.12%-- sys_read
| system_call_fastpath
| __read
|
|--3.24%-- generic_file_buffered_write
| __generic_file_aio_write_nolock
| generic_file_aio_write
| do_sync_write
| reiserfs_file_write
| vfs_write
| |
| |--3.14%-- sys_pwrite64
| | system_call_fastpath
| | __pwrite64
| |
| --0.10%-- sys_write
[...]
The command line has then changed.
By providing the -c option, the callchain will output in the
flat mode by default.
But you can override it:
perf report -c graph
or
perf report -c flat
You can also pass the abreviated mode:
perf report -c g
or
perf report -c gra
will both make use of the graph mode.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <1246550301-8954-3-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The copy we were using came from another copy I did for the dwarves
(pahole) package, that came from the kernel years ago.
The only function that is used by the perf tools and that isn't in the
kernel is list_del_range, that I'm leaving in the perf tools only for
now.
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: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20090701174608.GA5823@ghostprotocols.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The tools/perf/util/rbtree.c copy already drifted by three
csets:
4b324126e04c6011781116c047add3
So remove the copy and use the lib/rbtree.c directly, sharing
the source code while still generating a separate object file,
since tools/perf uses a far more agressive -O6 switch.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <20090701152837.GG15682@ghostprotocols.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Enable -Wextra. This found a few real bugs plus a number
of signed/unsigned type mismatches/uncleanlinesses. It
also required a few annotations
All things considered it was still worth it so lets try with
this enabled for now.
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
We plan to display the callchains depending on some user-configurable
parameters.
To gather the callchains stats from the recorded stream in a fast way,
this patch introduces an ad hoc radix tree adapted for callchains and also
a rbtree to sort these callchains once we have gathered every events
from the stream.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1246026481-8314-2-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
