We were always looking at the running machine /proc/modules,
even when processing a perf.data file, which only makes sense
when we're doing 'perf record' and 'perf report' on the same
machine, and in close sucession, or if we don't use modules at
all, right Peter? ;-)
Now, at 'perf record' time we read /proc/modules, find the long
path for modules, and put them as PERF_MMAP events, just like we
did to encode the reloc reference symbol for vmlinux. Talking
about that now it is encoded in .pgoff, so that we can use
.{start,len} to store the address boundaries for the kernel so
that when we reconstruct the kmaps tree we can do lookups right
away, without having to fixup the end of the kernel maps like we
did in the past (and now only in perf record).
One more step in the 'perf archive' direction when we'll finally
be able to collect data in one machine and analyse in another.
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: <1263396139-4798-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Add --line option to support showing probable source-code lines.
perf probe --line SRC:LN[-LN|+NUM]
or
perf probe --line FUNC[:LN[-LN|+NUM]]
This option shows source-code with line number if the line can
be probed. Lines without line number (and blue color) means that
the line can not be probed, because debuginfo doesn't have the
information of those lines.
The argument specifies the range of lines, "source.c:100-120"
shows lines between 100th to l20th in source.c file. And
"func:10+20" shows 20 lines from 10th line of func function.
e.g.
# ./perf probe --line kernel/sched.c:1080
<kernel/sched.c:1080>
*
* called with rq->lock held and irqs disabled
*/
static void hrtick_start(struct rq *rq, u64 delay)
{
struct hrtimer *timer = &rq->hrtick_timer;
1086 ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
hrtimer_set_expires(timer, time);
1090 if (rq == this_rq()) {
1091 hrtimer_restart(timer);
1092 } else if (!rq->hrtick_csd_pending) {
1093 __smp_call_function_single(cpu_of(rq), &rq->hrtick_csd,
1094 rq->hrtick_csd_pending = 1;
If you specifying function name, this shows function-relative
line number.
# ./perf probe --line schedule
<schedule:0>
asmlinkage void __sched schedule(void)
1 {
struct task_struct *prev, *next;
unsigned long *switch_count;
struct rq *rq;
int cpu;
need_resched:
preempt_disable();
9 cpu = smp_processor_id();
10 rq = cpu_rq(cpu);
11 rcu_sched_qs(cpu);
12 prev = rq->curr;
13 switch_count = &prev->nivcsw;
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: systemtap <systemtap@sources.redhat.com>
Cc: DLE <dle-develop@lists.sourceforge.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20100106144534.27218.77939.stgit@dhcp-100-2-132.bos.redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
DSOs don't have this problem because the kernel emits a
PERF_MMAP for each new executable mapping it performs on
monitored threads.
To fix the kernel case we simulate the same behaviour, by having
'perf record' to synthesize a PERF_MMAP for the kernel, encoded
like this:
[root@doppio ~]# perf record -a -f sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ]
[root@doppio ~]# perf report -D | head -10
0xd0 [0x40]: event: 1
.
. ... raw event: size 64 bytes
. 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........
. 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ...............
. 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel
. 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text]
. 0xd0
[0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text]
I.e. we identify such event as having:
.pid = 0
.filename = [kernel.kallsyms.REFNAME]
.start = REFNAME addr in /proc/kallsyms at 'perf record' time
and use now a hardcoded value of '.text' for REFNAME.
Then, later, in 'perf report', if there are any kernel hits and
thus we need to resolve kernel symbols, we search for REFNAME
and if its address changed, relocation happened and we thus must
change the kernel mapping routines to one that uses .pgoff as
the relocation to apply.
This way we use the same mechanism used for the other DSOs and
don't have to do a two pass in all the kernel symbols.
Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
To avoid the funny:
[root@doppio ~]# perf record -a -f sleep 2s
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.334 MB perf.data (~14572 samples) ]
[root@doppio ~]# perf report --no-call-graph
selected -g but no callchain data. Did you call perf record without -g?
And fix the bug reported by peterz when we do indeed record with
callchains and then ask for a report without:
[root@doppio ~]# perf record -a -g -f sleep 2s
[root@doppio ~]# perf report --no-call-graph
Segmentation fault
[root@doppio ~]#
Reported-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
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: <1262699685-27820-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
At least on Debian PARISC64, using:
acme@parisc:~/git/linux-2.6-tip$ gcc -v
Using built-in specs.
Target: hppa-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian
4.3.4-6' --with-bugurl=file:///usr/share/doc/gcc-4.3/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr
--enable-shared --enable-multiarch --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --enable-nls --with-gxx-include-dir=/usr/include/c++/4.3 --program-suffix=-4.3 --enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc --enable-mpfr --disable-libssp --enable-checking=release --build=hppa-linux-gnu --host=hppa-linux-gnu --target=hppa-linux-gnu Thread model: posix gcc version 4.3.4 (Debian 4.3.4-6)
there are issues about using 'gcc -o /dev/null':
/usr/bin/ld: final link failed: File truncated
collect2: ld returned 1 exit status
So we test that and use /dev/null in environments where it
works, while using an .INTERMEDIATE file on those where it can't
be used, so that the .perf.dev.null file can be used instead and
then deleted when make exits.
Researched-with: Kyle McMartin <kyle@mcmartin.ca>
Researched-with: Mauro Carvalho Chehab <mchehab@redhat.com>
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: <1263293910-8484-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When we finish creating the hist_entries we _already_ have them
sorted "by name", in fact by what is in --sort, that is exactly
how we can find the pairs in perf_session__match_hists as
'comm', 'dso' & 'symbol' all are strings we need to find the
matches in the baseline session.
So only do the sort by hits followed by a resort by --sort if we
need to find the position for shwowing the --displacement of
hist entries.
Now all these modes work correctly:
Example is a simple 'perf record -f find / > /dev/null' ran
twice then followed by the following commands:
$ perf diff -f --sort comm
# Baseline Delta Command
# ........ .......... .......
#
0.00% +100.00% find
$ perf diff -f --sort dso
# Baseline Delta Shared Object
# ........ .......... ..................
#
59.97% -0.44% [kernel]
21.17% +0.28% libc-2.5.so
18.49% +0.16% [ext3]
0.37% find
$ perf diff -f --sort symbol | head -8
# Baseline Delta Symbol
# ........ .......... ......
#
6.21% +0.36% [k] ext3fs_dirhash
3.43% +0.41% [.] __GI_strlen
3.53% +0.16% [k] __kmalloc
3.17% +0.49% [k] system_call
3.06% +0.37% [k] ext3_htree_store_dirent
$ perf diff -f --sort dso,symbol | head -8
# Baseline Delta Shared Object Symbol
# ........ .......... .................. ......
#
6.21% +0.36% [ext3] [k] ext3fs_dirhash
3.43% +0.41% libc-2.5.so [.] __GI_strlen
3.53% +0.16% [kernel] [k] __kmalloc
3.17% +0.49% [kernel] [k] system_call
3.06% +0.37% [ext3] [k] ext3_htree_store_dirent
$
And we don't have to do two expensive resorts in the common, non
--displacement case.
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: <1262047716-23171-5-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
