- 29 Jan, 2010 5 commits
-
-
Stephane Eranian authored
This patch improves event scheduling by maximizing the use of PMU registers regardless of the order in which events are created in a group. The algorithm takes into account the list of counter constraints for each event. It assigns events to counters from the most constrained, i.e., works on only one counter, to the least constrained, i.e., works on any counter. Intel Fixed counter events and the BTS special event are also handled via this algorithm which is designed to be fairly generic. The patch also updates the validation of an event to use the scheduling algorithm. This will cause early failure in perf_event_open(). The 2nd version of this patch follows the model used by PPC, by running the scheduling algorithm and the actual assignment separately. Actual assignment takes place in hw_perf_enable() whereas scheduling is implemented in hw_perf_group_sched_in() and x86_pmu_enable(). Signed-off-by:
Stephane Eranian <eranian@google.com> [ fixup whitespace and style nits as well as adding is_x86_event() ] Signed-off-by:
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: <4b5430c6.0f975e0a.1bf9.ffff85fe@mx.google.com> Signed-off-by:
Ingo Molnar <mingo@elte.hu>
-
K.Prasad authored
Processing of debug exceptions in do_debug() can stop if it originated from a hw-breakpoint exception by returning NOTIFY_STOP in most cases. But for certain cases such as: a) user-space breakpoints with pending SIGTRAP signal delivery (as in the case of ptrace induced breakpoints). b) exceptions due to other causes than breakpoints We will continue to process the exception by returning NOTIFY_DONE. Signed-off-by:
K.Prasad <prasad@linux.vnet.ibm.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Roland McGrath <roland@redhat.com> Cc: Alan Stern <stern@rowland.harvard.edu> Cc: Jan Kiszka <jan.kiszka@siemens.com> LKML-Reference: <20100128111415.GC13935@in.ibm.com> Signed-off-by:
Frederic Weisbecker <fweisbec@gmail.com>
-
K.Prasad authored
Clear the reserved bits from the stored copy of debug status register (DR6). This will help easy bitwise operations such as quick testing of a debug event origin. Signed-off-by:
-
Xiao Guangrong authored
The return values of the kprobe's tracing functions are meaningless, lets remove these. Signed-off-by:
Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Acked-by:
Masami Hiramatsu <mhiramat@redhat.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Jason Baron <jbaron@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> LKML-Reference: <4B60E9A3.2040505@cn.fujitsu.com> [fweisbec@gmail: whitespace fixes, drop useless void returns in end of functions] Signed-off-by:
-
Xiao Guangrong authored
Introduce ftrace_perf_buf_prepare() and ftrace_perf_buf_submit() to gather the common code that operates on raw events sampling buffer. This cleans up redundant code between regular trace events, syscall events and kprobe events. Changelog v1->v2: - Rename function name as per Masami and Frederic's suggestion - Add __kprobes for ftrace_perf_buf_prepare() and make ftrace_perf_buf_submit() inline as per Masami's suggestion - Export ftrace_perf_buf_prepare since modules will use it Signed-off-by:
-
- 28 Jan, 2010 1 commit
-
-
Anton Blanchard authored
When running perf across all cpus with backtracing (-a -g), sometimes we get samples without associated backtraces: 23.44% init [kernel] [k] restore 11.46% init eeba0c [k] 0x00000000eeba0c 6.77% swapper [kernel] [k] .perf_ctx_adjust_freq 5.73% init [kernel] [k] .__trace_hcall_entry 4.69% perf libc-2.9.so [.] 0x0000000006bb8c | |--11.11%-- 0xfffa941bbbc It turns out the backtrace code has a check for the idle task and the IP sampling does not. This creates problems when profiling an interrupt heavy workload (in my case 10Gbit ethernet) since we get no backtraces for interrupts received while idle (ie most of the workload). Right now x86 and sh check that current is not NULL, which should never happen so remove that too. Idle task's exclusion must be performed from the core code, on top of perf_event_attr:exclude_idle. Signed-off-by:Anton Blanchard <anton@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mundt <lethal@linux-sh.org> LKML-Reference: <20100118054707.GT12666@kryten> Signed-off-by:
-
- 27 Jan, 2010 6 commits
-
-
Hitoshi Mitake authored
perf trace lacks -i option for choosing input file. This patch adds it to perf trace. Signed-off-by:
Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> LKML-Reference: <1264167929-6741-1-git-send-email-mitake@dcl.info.waseda.ac.jp> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Probably this wasn't noticed when testing this on my parisc machine because I must have copied manually to its cache the vmlinux file used in the x86_64 machine, now that I tried looking on a x86-32 machine with a fresh cache, kernel symbols weren't being resolved even with the right kallsyms copy on its cache, duh. 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: <1264178102-4203-2-git-send-email-acme@infradead.org> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Only if we parsed /proc/kallsyms (or a copy found in the buildid cache) we should set the dso long name to "[kernel.kallsyms]". Reported-by:
Mike Galbraith <efault@gmx.de> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
As noticed by Mike, symbols in new tasks were not being processed as we weren't processing these events. Reported-by:
-
Arnaldo Carvalho de Melo authored
Broken since "5b2bb75a perf top: Support userspace symbols too". Reported-by:
-
Peter Zijlstra authored
There was a bug in the old period code that caused intel_pmu_enable_all() or native_write_msr_safe() to show up quite high in the profiles. In staring at that code it made my head hurt, so I rewrote it in a hopefully simpler fashion. Its now fully symetric between tick and overflow driven adjustments and uses less data to boot. The only complication is that it basically wants to do a u128 division. The code approximates that in a rather simple truncate until it fits fashion, taking care to balance the terms while truncating. This version does not generate that sampling artefact. Signed-off-by:
-
- 21 Jan, 2010 1 commit
-
-
Arnaldo Carvalho de Melo authored
For now it just has operations to examine a given file, find its build-id and add or remove it to/from the cache. Useful, for instance, when adding binaries sent together with a perf.data file, so that we can add them to the cache and have the tools find it when resolving symbols. It'll also manage the size of the cache like 'ccache' does. Signed-off-by:
-
- 20 Jan, 2010 6 commits
-
-
Arnaldo Carvalho de Melo authored
Found while analysing a perf.data file collected on an ARM machine where an explicitely specified vmlinux was being disregarded. Reported-by:
Jamie Iles <jamie.iles@picochip.com> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Because it may be possible that there was no buildid section, where we would set this to 1. Found while analysing a perf.data file collected on an ARM machine where an explicitely specified vmlinux was being disregarded. Reported-by:
-
Arnaldo Carvalho de Melo authored
Fix these warning: acme@parisc:~/git/linux-2.6-tip$ make -C tools/perf/ install make: Entering directory `/home/acme/git/linux-2.6-tip/tools/perf' Makefile:833: warning: overriding commands for target `perf-archive' Makefile:822: warning: ignoring old commands for target `perf-archive' Reported-by:
-
Arnaldo Carvalho de Melo authored
This also makes it appear on the 'perf --help' output, i.e. util/generate-cmdlist.sh now takes it into account. Signed-off-by:
-
Motohiro KOSAKI authored
Latest kprobetrace can remove probe points selectively, thus the documentation should be updated too. Signed-off-by:
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com> Signed-off-by:
-
Motohiro KOSAKI authored
Shell interprets $val as shell variable, thus we need quote if we use the echo command. Signed-off-by:
-
- 18 Jan, 2010 1 commit
-
-
Ingo Molnar authored
Merge branch 'perf/scheduling' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/random-tracing into perf/core
-
- 17 Jan, 2010 7 commits
-
-
Frederic Weisbecker authored
When a task gets scheduled in. We don't touch the cpu bound events so the priority order becomes: cpu pinned, cpu flexible, task pinned, task flexible. So schedule out cpu flexibles when a new task context gets in and correctly order the groups to schedule in: task pinned, cpu flexible, task flexible. Cpu pinned groups don't need to be touched at this time. Signed-off-by: -
Frederic Weisbecker authored
We don't need to schedule in/out pinned events on task tick, now that pinned and flexible groups can be scheduled separately. Signed-off-by: -
Frederic Weisbecker authored
Tune the scheduling helpers so that we can choose to schedule either pinned and/or flexible groups from a context. And while at it, refactor a bit the naming of these helpers to make these more consistent and flexible. There is no (intended) change in scheduling behaviour in this patch. Signed-off-by: -
Frederic Weisbecker authored
__perf_event_sched_out doesn't need to be globally available, make it static. Signed-off-by: -
Masami Hiramatsu authored
Update kprobe tracing self test for new syntax (it supports deleting individual probes, and drops $argN support) and behavior change (new probes are disabled in default). This selftest includes the following checks: - Adding function-entry probe and return probe with arguments. - Enabling these probes. - Deleting it individually. Signed-off-by:
-
Hitoshi Mitake authored
I got this build error when building tip tree: | cc1: warnings being treated as errors | builtin-probe.c:123: error: 'opt_show_lines' defined but not used This error is caused by: | #ifndef NO_LIBDWARF | OPT_CALLBACK('L', "line", NULL, | "FUNC[:RLN[+NUM|:RLN2]]|SRC:ALN[+NUM|:ALN2]", | "Show source code lines.", opt_show_lines), | #endif My environment defines NO_LIBDWARF, so gcc treated opt_show_lines() as garbage. So I moved opt_show_lines() into #ifndef NO_LIBDWARF ... #endif block. Signed-off-by: -
Frederic Weisbecker authored
getline() is considered as undeclared in util/util.c because it includes string.h, that in turn includes stdio.h, without having defined _GNU_SOURCE. But util.c also includes util.h that handles the _GNU_SOURCE and all the needed inclusions already. Let's include only util.h and sys/mman.h which is the only one header not handled by util.h This fixes the following build error: util/util.c: In function 'slow_copyfile': util/util.c:49: erreur: implicit declaration of function 'getline' util/util.c:49: erreur: nested extern declaration of 'getline' Signed-off-by:
-
- 16 Jan, 2010 13 commits
-
-
Frederic Weisbecker authored
Before scheduling an event group, we first check if a group can go on. We first check if the group is made of software only events first, in which case it is enough to know if the group can be scheduled in. For that purpose, we iterate through the whole group, which is wasteful as we could do this check when we add/delete an event to a group. So we create a group_flags field in perf event that can host characteristics from a group of events, starting with a first PERF_GROUP_SOFTWARE flag that reduces the check on the fast path. Signed-off-by:
Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Ingo Molnar <mingo@elte.hu> Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
-
Frederic Weisbecker authored
This is more proper that doing it through a list_for_each_entry() that breaks after the first entry. v2: Don't rotate pinned groups as its not needed to time share them. Signed-off-by:
-
Frederic Weisbecker authored
Bring a new list_rotate_left() helper that rotates a list to the left. This is useful for codes that need to round roubin elements which queue priority increases from tail to head. Signed-off-by:
-
Frederic Weisbecker authored
Split-up struct perf_event_context::group_list into pinned_groups and flexible_groups (non-pinned). This first appears to be useless as it duplicates various loops around the group list handlings. But it scales better in the fast-path in perf_sched_in(). We don't anymore iterate twice through the entire list to separate pinned and non-pinned scheduling. Instead we interate through two distinct lists. The another desired effect is that it makes easier to define distinct scheduling rules on both. Changes in v2: - Respectively rename pinned_grp_list and volatile_grp_list into pinned_groups and flexible_groups as per Ingo suggestion. - Various cleanups Signed-off-by:
-
Arnaldo Carvalho de Melo authored
As it is in PARISC64: parisc:~# uname -a Linux parisc 2.6.33-rc4-tip+ #1 SMP Thu Jan 14 13:33:34 BRST 2010 parisc64 GNU/Linux parisc:~# grep -w _text /proc/kallsyms 0000000040100000 A _text parisc:~# grep 0000000040100000 /proc/kallsyms 0000000040100000 T stext 0000000040100000 T _stext 0000000040100000 A _text parisc:~# Signed-off-by:
-
Arnaldo Carvalho de Melo authored
The event interception we need to do in 'perf record' to create a list of all DSOs in PERF_RECORD_MMAP events wasn't seeing all events, make sure that happens by checking size agains event_t->header.size. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
It uses 'perf buildid-list --with-hits' to create a tarball with what is needed to have in the destination machine ~/.debug hierarchy to properly decode the perf.data file specified. Here is an example where a perf.data file collected on a x86-64 machine running Fedora 12 is used and then the data is packaged, transferred and decoded on a PARISC64 machine running Debian Testing, 32-bit userspace: [root@doppio linux-2.6-tip]# uname -a Linux doppio.ghostprotocols.net 2.6.33-rc4-tip+ #3 SMP Wed Jan 13 11:58:15 BRST 2010 x86_64 x86_64 x86_64 GNU/Linux [root@doppio linux-2.6-tip]# perf archive [root@doppio linux-2.6-tip]# ls -la perf.data* -rw------- 1 root root 737696 2010-01-14 23:36 perf.data -rw-r--r-- 1 root root 8840025 2010-01-15 12:27 perf.data.tar.bz2 [root@doppio linux-2.6-tip]# scp perf.data.* parisc64:. Password: perf.data.tar.bz2 100% 8633KB 1.4MB/s 00:06 [root@doppio linux-2.6-tip]# ssh parisc64 Password: Linux parisc 2.6.19-g2bbf29ac-dirty #1 Sun Dec 3 17:24:04 BRST 2006 parisc64 The programs included with the Debian GNU/Linux system are free software; the exact distribution terms for each program are described in the individual files in /usr/share/doc/*/copyright. Debian GNU/Linux comes with ABSOLUTELY NO WARRANTY, to the extent permitted by applicable law. Last login: Thu Jan 14 11:23:24 2010 from d parisc:~# uname -a Linux parisc 2.6.19-g2bbf29ac-dirty #1 Sun Dec 3 17:24:04 BRST 2006 parisc64 GNU/Linux parisc:~# mkdir .debug parisc:~# tar xvf perf.data.tar.bz2 -C ~/.debug tar: Record size = 8 blocks .build-id/74/f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms]/74f9930ee94475b6b3238caf3725a50d59cb994b .build-id/9f/fdcac0a7935922d1f04b6cc9029dfef0f066ef lib/modules/2.6.33-rc4-tip+/kernel/arch/x86/crypto/aes-x86_64.ko/9ffdcac0a7935922d1f04b6cc9029dfef0f066ef .build-id/3a/af89c32ebfc438ff546c93597d41788e3e65f3 lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwl3945.ko/3aaf89c32ebfc438ff546c93597d41788e3e65f3 .build-id/19/f46033f73e1ec612937189bb118c5daba5a0c8 lib/modules/2.6.33-rc4-tip+/kernel/net/mac80211/mac80211.ko/19f46033f73e1ec612937189bb118c5daba5a0c8 .build-id/17/72f014a7a7272859655acb0c64a20ab20b75ee lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko/1772f014a7a7272859655acb0c64a20ab20b75ee .build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 .build-id/5c/68f7afeb33309c78037e374b0deee84dd441f6 lib64/libpthread-2.10.2.so/5c68f7afeb33309c78037e374b0deee84dd441f6 .build-id/e9/c9ad5c138ef882e4507d2605645b597da43873 bin/dbus-daemon/e9c9ad5c138ef882e4507d2605645b597da43873 .build-id/bc/da7d09eb6c9ee380dae0ed3d591d4311decc31 lib64/libdbus-1.so.3.4.0/bcda7d09eb6c9ee380dae0ed3d591d4311decc31 .build-id/7c/c449a77f48b85d6088114000e970ced613bed8 usr/lib64/libcrypto.so.0.9.8k/7cc449a77f48b85d6088114000e970ced613bed8 .build-id/fd/d1ccd1ff7917ab020653147ab3bacf0a85b5b9 lib64/libglib-2.0.so.0.2000.5/fdd1ccd1ff7917ab020653147ab3bacf0a85b5b9 .build-id/e4/417ebb8762e5f2eee93c8011a71115ff5edad8 lib64/libgobject-2.0.so.0.2000.5/e4417ebb8762e5f2eee93c8011a71115ff5edad8 .build-id/93/1e49461f6df99104f0febcc52f6fed5e2efce6 usr/sbin/sshd/931e49461f6df99104f0febcc52f6fed5e2efce6 .build-id/da/b5f724c088f89fbd8304da553ed6cb30bbec96 usr/lib64/libgdk-x11-2.0.so.0.1600.6/dab5f724c088f89fbd8304da553ed6cb30bbec96 .build-id/f2/037a091ef36b591187a858d75e203690ea9409 usr/sbin/openvpn/f2037a091ef36b591187a858d75e203690ea9409 .build-id/a8/e4f743b40fb1fd8b85e2f9b88d93b661472b8f bin/find/a8e4f743b40fb1fd8b85e2f9b88d93b661472b8f .build-id/81/120aada06e68b1e85882925a0fc6d7345ef59a home/acme/bin/perf/81120aada06e68b1e85882925a0fc6d7345ef59a parisc:~# perf report 2> /dev/null | head -25 9.07% find find [.] 0x0000000000fb0e 3.29% perf libc-2.10.2.so [.] __GI_strcmp 3.19% find [kernel.kallsyms] [k] _raw_spin_unlock_irqrestore 2.70% find libc-2.10.2.so [.] __GI_memmove 2.62% perf [kernel.kallsyms] [k] vsnprintf 2.03% find libc-2.10.2.so [.] _int_malloc 2.02% perf [kernel.kallsyms] [k] format_decode 1.70% find [kernel.kallsyms] [k] n_tty_write 1.70% find [kernel.kallsyms] [k] half_md4_transform 1.67% find libc-2.10.2.so [.] _IO_vfprintf_internal 1.66% perf [kernel.kallsyms] [k] audit_free_aux 1.62% swapper [kernel.kallsyms] [k] mwait_idle_with_hints 1.58% find [kernel.kallsyms] [k] __kmalloc 1.35% find [kernel.kallsyms] [k] sched_clock_local 1.35% find [kernel.kallsyms] [k] ext4_check_dir_entry 1.35% find [kernel.kallsyms] [k] ext4_htree_store_dirent 1.35% find [kernel.kallsyms] [k] sys_write 1.35% find [e1000e] [k] e1000_clean 1.35% find [kernel.kallsyms] [k] _atomic_dec_and_lock 1.34% find [kernel.kallsyms] [k] __d_lookup parisc:~# Probably the next step is to have 'perf report' notice that there is a perf.data.tar.bz2 file in the same directory and look if it was already added to ~/.debug/. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Since we use ->long_name in dsos__find now. Now 'perf buildid_list' is not duplicating those and managing to show the proper build-ids for the DSOs with hits: [root@doppio linux-2.6-tip]# perf buildid-list -H 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 9ffdcac0a7935922d1f04b6cc9029dfef0f066ef /lib/modules/2.6.33-rc4-tip+/kernel/arch/x86/crypto/aes-x86_64.ko 3aaf89c32ebfc438ff546c93597d41788e3e65f3 /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwl3945.ko 19f46033f73e1ec612937189bb118c5daba5a0c8 /lib/modules/2.6.33-rc4-tip+/kernel/net/mac80211/mac80211.ko 1772f014a7a7272859655acb0c64a20ab20b75ee /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 /lib64/libc-2.10.2.so 5c68f7afeb33309c78037e374b0deee84dd441f6 /lib64/libpthread-2.10.2.so e9c9ad5c138ef882e4507d2605645b597da43873 /bin/dbus-daemon bcda7d09eb6c9ee380dae0ed3d591d4311decc31 /lib64/libdbus-1.so.3.4.0 7cc449a77f48b85d6088114000e970ced613bed8 /usr/lib64/libcrypto.so.0.9.8k fdd1ccd1ff7917ab020653147ab3bacf0a85b5b9 /lib64/libglib-2.0.so.0.2000.5 e4417ebb8762e5f2eee93c8011a71115ff5edad8 /lib64/libgobject-2.0.so.0.2000.5 931e49461f6df99104f0febcc52f6fed5e2efce6 /usr/sbin/sshd dab5f724c088f89fbd8304da553ed6cb30bbec96 /usr/lib64/libgdk-x11-2.0.so.0.1600.6 f2037a091ef36b591187a858d75e203690ea9409 /usr/sbin/openvpn a8e4f743b40fb1fd8b85e2f9b88d93b661472b8f /bin/find 81120aada06e68b1e85882925a0fc6d7345ef59a /home/acme/bin/perf [root@doppio linux-2.6-tip]# Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Using this option 'perf buildid-list' will process all samples, marking the DSOs that had some hits to list just them. This in turn will be used by a new porcelain, 'perf archive', that will be just a shell script to create a tarball from the 'perf buildid-list --with-hits' output and the files cached by 'perf record' in ~/.debug. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
Because some tools will only want to know with maps had hits, not needing the full symbol resolution done by thread__find_addr_location. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
In the past 'perf record' had to process only userspace MMAP events, the ones generated in the kernel, but after we reused the MMAP events to encode the module mapings we ended up adding them first to the list of userspace DSOs (dsos__user) and to the kernel one (dsos__kernel). Fix this by encoding the header.misc field and then using it, like other parts to decide the right DSOs list to insert/find. The gotcha here is that since the kernel puts zero in .misc, which isn't PERF_RECORD_MISC_KERNEL (1 << 1), to differentiate, we put 1 in .misc. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
If not we end up duplicating the module DSOs because first we insert them using the short name found in /proc/modules, then, when processing synthesized MMAP events we add them again. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
So that when we don't have a vmlinux handy we can store the kallsyms for later use by 'perf report'. Signed-off-by:
-
