aboutsummaryrefslogtreecommitdiff
path: root/tools
AgeCommit message (Collapse)AuthorFilesLines
2016-07-25Merge branch 'x86-mm-for-linus' of ↵Linus Torvalds7-2/+2342
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip Pull x86 mm updates from Ingo Molnar: "Various x86 low level modifications: - preparatory work to support virtually mapped kernel stacks (Andy Lutomirski) - support for 64-bit __get_user() on 32-bit kernels (Benjamin LaHaise) - (involved) workaround for Knights Landing CPU erratum (Dave Hansen) - MPX enhancements (Dave Hansen) - mremap() extension to allow remapping of the special VDSO vma, for purposes of user level context save/restore (Dmitry Safonov) - hweight and entry code cleanups (Borislav Petkov) - bitops code generation optimizations and cleanups with modern GCC (H. Peter Anvin) - syscall entry code optimizations (Paolo Bonzini)" * 'x86-mm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (43 commits) x86/mm/cpa: Add missing comment in populate_pdg() x86/mm/cpa: Fix populate_pgd(): Stop trying to deallocate failed PUDs x86/syscalls: Add compat_sys_preadv64v2/compat_sys_pwritev64v2 x86/smp: Remove unnecessary initialization of thread_info::cpu x86/smp: Remove stack_smp_processor_id() x86/uaccess: Move thread_info::addr_limit to thread_struct x86/dumpstack: Rename thread_struct::sig_on_uaccess_error to sig_on_uaccess_err x86/uaccess: Move thread_info::uaccess_err and thread_info::sig_on_uaccess_err to thread_struct x86/dumpstack: When OOPSing, rewind the stack before do_exit() x86/mm/64: In vmalloc_fault(), use CR3 instead of current->active_mm x86/dumpstack/64: Handle faults when printing the "Stack: " part of an OOPS x86/dumpstack: Try harder to get a call trace on stack overflow x86/mm: Remove kernel_unmap_pages_in_pgd() and efi_cleanup_page_tables() x86/mm/cpa: In populate_pgd(), don't set the PGD entry until it's populated x86/mm/hotplug: Don't remove PGD entries in remove_pagetable() x86/mm: Use pte_none() to test for empty PTE x86/mm: Disallow running with 32-bit PTEs to work around erratum x86/mm: Ignore A/D bits in pte/pmd/pud_none() x86/mm: Move swap offset/type up in PTE to work around erratum x86/entry: Inline enter_from_user_mode() ...
2016-07-25Merge branch 'perf-core-for-linus' of ↵Linus Torvalds303-2309/+13229
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip Pull perf updates from Ingo Molnar: "With over 300 commits it's been a busy cycle - with most of the work concentrated on the tooling side (as it should). The main kernel side enhancements were: - Add per event callchain limit: Recently we introduced a sysctl to tune the max-stack for all events for which callchains were requested: $ sysctl kernel.perf_event_max_stack kernel.perf_event_max_stack = 127 Now this patch introduces a way to configure this per event, i.e. this becomes possible: $ perf record -e sched:*/max-stack=2/ -e block:*/max-stack=10/ -a allowing finer tuning of how much buffer space callchains use. This uses an u16 from the reserved space at the end, leaving another u16 for future use. There has been interest in even finer tuning, namely to control the max stack for kernel and userspace callchains separately. Further discussion is needed, we may for instance use the remaining u16 for that and when it is present, assume that the sample_max_stack introduced in this patch applies for the kernel, and the u16 left is used for limiting the userspace callchain (Arnaldo Carvalho de Melo) - Optimize AUX event (hardware assisted side-band event) delivery (Kan Liang) - Rework Intel family name macro usage (this is partially x86 arch work) (Dave Hansen) - Refine and fix Intel LBR support (David Carrillo-Cisneros) - Add support for Intel 'TopDown' events (Andi Kleen) - Intel uncore PMU driver fixes and enhancements (Kan Liang) - ... other misc changes. Here's an incomplete list of the tooling enhancements (but there's much more, see the shortlog and the git log for details): - Support cross unwinding, i.e. collecting '--call-graph dwarf' perf.data files in one machine and then doing analysis in another machine of a different hardware architecture. This enables, for instance, to do: $ perf record -a --call-graph dwarf on a x86-32 or aarch64 system and then do 'perf report' on it on a x86_64 workstation (He Kuang) - Allow reading from a backward ring buffer (one setup via sys_perf_event_open() with perf_event_attr.write_backward = 1) (Wang Nan) - Finish merging initial SDT (Statically Defined Traces) support, see cset comments for details about how it all works (Masami Hiramatsu) - Support attaching eBPF programs to tracepoints (Wang Nan) - Add demangling of symbols in programs written in the Rust language (David Tolnay) - Add support for tracepoints in the python binding, including an example, that sets up and parses sched:sched_switch events, tools/perf/python/tracepoint.py (Jiri Olsa) - Introduce --stdio-color to set up the color output mode selection in 'annotate' and 'report', allowing emit color escape sequences when redirecting the output of these tools (Arnaldo Carvalho de Melo) - Add 'callindent' option to 'perf script -F', to indent the Intel PT call stack, making this output more ftrace-like (Adrian Hunter, Andi Kleen) - Allow dumping the object files generated by llvm when processing eBPF scriptlet events (Wang Nan) - Add stackcollapse.py script to help generating flame graphs (Paolo Bonzini) - Add --ldlat option to 'perf mem' to specify load latency for loads event (e.g. cpu/mem-loads/ ) (Jiri Olsa) - Tooling support for Intel TopDown counters, recently added to the kernel (Andi Kleen)" * 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (303 commits) perf tests: Add is_printable_array test perf tools: Make is_printable_array global perf script python: Fix string vs byte array resolving perf probe: Warn unmatched function filter correctly perf cpu_map: Add more helpers perf stat: Balance opening and reading events tools: Copy linux/{hash,poison}.h and check for drift perf tools: Remove include/linux/list.h from perf's MANIFEST tools: Copy the bitops files accessed from the kernel and check for drift Remove: kernel unistd*h files from perf's MANIFEST, not used perf tools: Remove tools/perf/util/include/linux/const.h perf tools: Remove tools/perf/util/include/asm/byteorder.h perf tools: Add missing linux/compiler.h include to perf-sys.h perf jit: Remove some no-op error handling perf jit: Add missing curly braces objtool: Initialize variable to silence old compiler objtool: Add -I$(srctree)/tools/arch/$(ARCH)/include/uapi perf record: Add --tail-synthesize option perf session: Don't warn about out of order event if write_backward is used perf tools: Enable overwrite settings ...
2016-07-25Merge branch 'core-rcu-for-linus' of ↵Linus Torvalds5-17/+60
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip Pull RCU updates from Ingo Molnar: "The main changes in this cycle were: - documentation updates - miscellaneous fixes - minor reorganization of code - torture-test updates" * 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (30 commits) rcu: Correctly handle sparse possible cpus rcu: sysctl: Panic on RCU Stall rcu: Fix a typo in a comment rcu: Make call_rcu_tasks() tolerate first call with irqs disabled rcu: Disable TASKS_RCU for usermode Linux rcu: No ordering for rcu_assign_pointer() of NULL rcutorture: Fix error return code in rcu_perf_init() torture: Inflict default jitter rcuperf: Don't treat gp_exp mis-setting as a WARN rcutorture: Drop "-soundhw pcspkr" from x86 boot arguments rcutorture: Don't specify the cpu type of QEMU on PPC rcutorture: Make -soundhw a x86 specific option rcutorture: Use vmlinux as the fallback kernel image rcutorture/doc: Create initrd using dracut torture: Stop onoff task if there is only one cpu torture: Add starvation events to error summary torture: Break online and offline functions out of torture_onoff() torture: Forgive lengthy trace dumps and preemption torture: Remove CONFIG_RCU_TORTURE_TEST_RUNNABLE, simplify code torture: Simplify code, eliminate RCU_PERF_TEST_RUNNABLE ...
2016-07-24Merge tag 'staging-4.8-rc1' of ↵Linus Torvalds2-83/+199
git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging Pull staging and IIO driver updates from Greg KH: "Here is the big Staging and IIO driver update for 4.8-rc1. We ended up adding more code than removing, again, but it's not all that bad. Lots of cleanups all over the staging tree, and new IIO drivers, full details in the shortlog. All of these have been in linux-next for a while with no reported issues" * tag 'staging-4.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging: (417 commits) drivers:iio:accel:mma8452: removed unwanted return statements drivers:iio:accel:mma8452: added cleanup provision in case of failure. iio: Add iio.git tree to MAINTAINERS iio:st_pressure: clean useless static channel initializers iio:st_pressure:lps22hb: temperature support iio:st_pressure:lps22hb: open drain support iio:st_pressure: temperature triggered buffering iio:st_pressure: document sampling gains iio:st_pressure: align storagebits on power of 2 iio:st_sensors: align on storagebits boundaries staging:iio:lis3l02dq drop separate driver iio: accel: st_accel: Add lis3l02dq support iio: adc: add missing of_node references to iio_dev iio: adc: ti-ads1015: add indio_dev->dev.of_node reference iio: potentiometer: Fix typo in Kconfig iio: potentiometer: mcp4531: Add device tree binding iio: potentiometer: mcp4531: Add device tree binding documentation iio: potentiometer: mcp4531: Add support for MCP454x, MCP456x, MCP464x and MCP466x iio:imu:mpu6050: icm20608 initial support iio: adc: max1363: Add device tree binding ...
2016-07-23tools/vm/slabinfo: fix an unintentional printfDan Carpenter1-1/+2
The curly braces are missing here so we print stuff unintentionally. Fixes: 9da4714a2d44 ('slub: slabinfo update for cmpxchg handling') Link: http://lkml.kernel.org/r/20160715211243.GE19522@mwanda Signed-off-by: Dan Carpenter <[email protected]> Acked-by: Christoph Lameter <[email protected]> Cc: Sergey Senozhatsky <[email protected]> Cc: Colin Ian King <[email protected]> Cc: Laura Abbott <[email protected]> Signed-off-by: Andrew Morton <[email protected]> Signed-off-by: Linus Torvalds <[email protected]>
2016-07-23testing/radix-tree: fix a macro expansion bugDan Carpenter1-1/+1
There are no parentheses around this macro and it causes a problem when we do: index = rand() % THRASH_SIZE; Link: http://lkml.kernel.org/r/20160715210953.GC19522@mwanda Signed-off-by: Dan Carpenter <[email protected]> Acked-by: Ross Zwisler <[email protected]> Signed-off-by: Andrew Morton <[email protected]> Signed-off-by: Linus Torvalds <[email protected]>
2016-07-18perf tests: Add is_printable_array testJiri Olsa4-0/+42
Add automated test for is_printable_array function. Signed-off-by: Jiri Olsa <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Pirko <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Steven Rostedt <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf tools: Make is_printable_array globalJiri Olsa4-28/+17
It's used from 2 objects in perf, so it's better to keep just one copy. Signed-off-by: Jiri Olsa <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Pirko <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Steven Rostedt <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf script python: Fix string vs byte array resolvingJiri Olsa1-6/+33
Jirka reported that python code returns all arrays as strings. This makes impossible to get all items for byte array tracepoint field containing 0x00 value item. Fixing this by scanning full length of the array and returning it as PyByteArray object in case non printable byte is found. Signed-off-by: Jiri Olsa <[email protected]> Reported-and-Tested-by: Jiri Pirko <[email protected]> Cc: David Ahern <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Steven Rostedt <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf probe: Warn unmatched function filter correctlyMasami Hiramatsu2-2/+13
Warn unmatched function filter correctly instead of warning "symbol-loading error", since that can be a filter issue. From the technical point of view, this adds a filter chech in map__load and if there is a filter, it returns -2 (filter-out), instead of -1 (error), and perf-probe checks it and change message. E.g. without this fix: # perf probe -F rt_sp* no symbols found in [kernel.kallsyms], maybe install a debug package? Failed to load symbols in kernel With this fix: # perf probe -F rt_sp* no symbols passed the given filter. Failed to find symbols matched to "rt_sp*" Error: Failed to show functions. Reported-and-Tested-by: Arnaldo Carvalho de Melo <[email protected]> Signed-off-by: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Link: http://lkml.kernel.org/r/146885835596.16106.2293540792775552481.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf cpu_map: Add more helpersMark Rutland2-2/+14
In some cases it's necessry to figure out the map-local index of a given Linux logical CPU ID. Add a new helper, cpu_map__idx, to acquire this. As the logic is largely the same as the existing cpu_map__has, this is rewritten in terms of the new helper. At the same time, add the inverse operation, cpu_map__cpu, which yields the logical CPU id for a map-local index. While this can be performed manually, wrapping this in a helper can make code more legible. Signed-off-by: Mark Rutland <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: Adrian Hunter <[email protected]> Cc: Alexander Shishkin <[email protected]> Cc: He Kuang <[email protected]> Cc: Kan Liang <[email protected]> Cc: Mark Rutland <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf stat: Balance opening and reading eventsMark Rutland1-2/+6
In create_perf_stat_counter, when a target CPU has not been provided, we call __perf_evsel__open with empty_cpu_map, and open a single FD per thread. However, in read_counter we assume that we opened events for the product of threads and CPUs described in the evsel's cpu_map. Thus, if an evsel has a cpu_map with more than one entry, we will attempt to access FDs that we didn't open. This could result in a number of problems (e.g. blocking while reading from STDIN if the fd memory happened to be initialised to zero). This is problematic for systems were a logical CPU PMU covers some arbitrary subset of CPUs. The cpu_map of any evsel for that PMU will be initialised based on the cpumask exposed through sysfs, even if the user requests per-thread events. Signed-off-by: Mark Rutland <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: Adrian Hunter <[email protected]> Cc: Alexander Shishkin <[email protected]> Cc: He Kuang <[email protected]> Cc: Kan Liang <[email protected]> Cc: Mark Rutland <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18tools: Copy linux/{hash,poison}.h and check for driftArnaldo Carvalho de Melo4-6/+195
We were also using this directly from the kernel sources, the two last cases, fix it. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf tools: Remove include/linux/list.h from perf's MANIFESTArnaldo Carvalho de Melo1-1/+0
It hasn't been used since we made tools/ self sufficiente wrt list.h. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Josh Poimboeuf <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Fixes: d1b39d41ebec ("tools: Make list.h self-sufficient") Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18tools: Copy the bitops files accessed from the kernel and check for driftArnaldo Carvalho de Melo7-10/+203
copy some more kernel files accessed from tools/, check for drift. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18Remove: kernel unistd*h files from perf's MANIFEST, not usedArnaldo Carvalho de Melo1-2/+0
No need to copy it to a detached tarball as they aren't used anymore Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf tools: Remove tools/perf/util/include/linux/const.hArnaldo Carvalho de Melo2-2/+0
Not used anymore, remove one more file referencing kernel sources, i.e. outside of tools/ Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf tools: Remove tools/perf/util/include/asm/byteorder.hArnaldo Carvalho de Melo2-4/+0
Not used anymore. This also stops include linux/swab.h directly from the kernel sources, remove that reference from the MANIFEST. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf tools: Add missing linux/compiler.h include to perf-sys.hArnaldo Carvalho de Melo1-0/+1
It uses the likely/unlikely macros, so need to include <linux/compiler.h>. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf jit: Remove some no-op error handlingDan Carpenter1-6/+1
The 'info.e_machine' struct member is an uint16_t so 'm' is never less than zero. It looks like this was maybe left over code from earlier versions so I've just removed it. Signed-off-by: Dan Carpenter <[email protected]> Cc: Adrian Hunter <[email protected]> Cc: Alexander Shishkin <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Stephane Eranian <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/20160715210836.GB19522@mwanda Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-18perf jit: Add missing curly bracesDan Carpenter1-1/+2
It doesn't change the runtime behavior, but my static checker complains that curly braces were intended. Signed-off-by: Dan Carpenter <[email protected]> Cc: Adrian Hunter <[email protected]> Cc: Alexander Shishkin <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Stephane Eranian <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/20160715210712.GA19522@mwanda Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15objtool: Initialize variable to silence old compilerArnaldo Carvalho de Melo1-1/+1
gcc version 4.1.2 20080704 (Red Hat 4.1.2-55) barfs with: CC /tmp/build/objtool/builtin-check.o cc1: warnings being treated as errors builtin-check.c: In function 'cmd_check': builtin-check.c:667: warning: 'prev_rela' may be used uninitialized in this function mv: cannot stat `/tmp/build/objtool/.builtin-check.o.tmp': No such file or directory make[1]: *** [/tmp/build/objtool/builtin-check.o] Error 1 Init it to NULL to silence it. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Josh Poimboeuf <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15objtool: Add -I$(srctree)/tools/arch/$(ARCH)/include/uapiArnaldo Carvalho de Melo1-1/+1
So that it can find asm/bitsperlong.h to get the __BITS_PER_LONG definition. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Josh Poimboeuf <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf record: Add --tail-synthesize optionWang Nan3-6/+34
When working with overwritable ring buffer there's a inconvenience problem: if perf dumps data after a long period after it starts, non-sample events may lost, which makes following 'perf report' unable to identify proc name and mmap layout. For example: # perf record -m 4 -e raw_syscalls:* -g --overwrite --switch-output \ dd if=/dev/zero of=/dev/null send SIGUSR2 after dd runs long enough. The resuling perf.data lost correct comm and mmap events: # perf script -i perf.data.2016061522374354 perf 24478 [004] 2581325.601789: raw_syscalls:sys_exit: NR 0 = 512 ^^^^ Should be 'dd' 27b2e8 syscall_slow_exit_work+0xfe2000e3 (/lib/modules/4.6.0-rc3+/build/vmlinux) 203cc7 do_syscall_64+0xfe200117 (/lib/modules/4.6.0-rc3+/build/vmlinux) b18d83 return_from_SYSCALL_64+0xfe200000 (/lib/modules/4.6.0-rc3+/build/vmlinux) 7f47c417edf0 [unknown] ([unknown]) ^^^^^^^^^^^^ Fail to unwind This patch provides a '--tail-synthesize' option, allows perf to collect system status when finalizing output file. In resuling output file, the non-sample events reflect system status when dumping data. After this patch: # perf record -m 4 -e raw_syscalls:* -g --overwrite --switch-output --tail-synthesize \ dd if=/dev/zero of=/dev/null # perf script -i perf.data.2016061600544998 dd 27364 [004] 2583244.994464: raw_syscalls:sys_enter: NR 1 (1, ... ^^ Correct comm 203a18 syscall_trace_enter_phase2+0xfe2001a8 ([kernel.kallsyms]) 203aa5 syscall_trace_enter+0xfe200055 ([kernel.kallsyms]) 203caa do_syscall_64+0xfe2000fa ([kernel.kallsyms]) b18d83 return_from_SYSCALL_64+0xfe200000 ([kernel.kallsyms]) d8e50 __GI___libc_write+0xffff01d9639f4010 (/tmp/oxygen_root-w00229757/lib64/libc-2.18.so) ^^^^^ Correct unwind This option doesn't aim to solve this problem completely. If a process terminates before SIGUSR2, we still lost its COMM and MMAP events. For example, we can't unwind correctly from the final perf.data we get from the previous example, because when perf collects the final output file (when we press C-c), 'dd' has been terminated so its '/proc/<pid>/mmap' becomes empty. However, this is a cheaper choice. To completely solve this problem we need to continously output non-sample events. To satisify the requirement of daemonization, we need to merge them periodically. It is possible but requires much more code and cycles. Automatically select --tail-synthesize when --overwrite is provided. Signed-off-by: Wang Nan <[email protected]> Cc: He Kuang <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf session: Don't warn about out of order event if write_backward is usedWang Nan1-3/+19
If write_backward attribute is set, records are written into kernel ring buffer from end to beginning, but read from beginning to end. To avoid 'XX out of order events recorded' warning message (timestamps of records is in reverse order when using write_backward), suppress the warning message if write_backward is selected by at lease one event. Result: Before this patch: # perf record -m 1 -e raw_syscalls:sys_exit/overwrite/ \ -e raw_syscalls:sys_enter \ dd if=/dev/zero of=/dev/null count=300 300+0 records in 300+0 records out 153600 bytes (154 kB) copied, 0.000601617 s, 255 MB/s [ perf record: Woken up 5 times to write data ] Warning: 40 out of order events recorded. [ perf record: Captured and wrote 0.096 MB perf.data (696 samples) ] After this patch: # perf record -m 1 -e raw_syscalls:sys_exit/overwrite/ \ -e raw_syscalls:sys_enter \ dd if=/dev/zero of=/dev/null count=300 300+0 records in 300+0 records out 153600 bytes (154 kB) copied, 0.000644873 s, 238 MB/s [ perf record: Woken up 5 times to write data ] [ perf record: Captured and wrote 0.096 MB perf.data (696 samples) ] Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: He Kuang <[email protected]> Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf tools: Enable overwrite settingsWang Nan9-7/+49
This patch allows following config terms and option: Globally setting events to overwrite; # perf record --overwrite ... Set specific events to be overwrite or no-overwrite. # perf record --event cycles/overwrite/ ... # perf record --event cycles/no-overwrite/ ... Add missing config terms and update the config term array size because the longest string length has changed. For overwritable events, it automatically selects attr.write_backward since perf requires it to be backward for reading. Test result: # perf record --overwrite -e syscalls:*enter_nanosleep* usleep 1 [ perf record: Woken up 2 times to write data ] [ perf record: Captured and wrote 0.011 MB perf.data (1 samples) ] # perf evlist -v syscalls:sys_enter_nanosleep: type: 2, size: 112, config: 0x134, { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|CPU|PERIOD|RAW, disabled: 1, inherit: 1, mmap: 1, comm: 1, enable_on_exec: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, write_backward: 1 # Tip: use 'perf evlist --trace-fields' to show fields for tracepoint events Signed-off-by: Wang Nan <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: He Kuang <[email protected]> Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Make {pause,resume} internal helpersWang Nan2-4/+2
There's no user of these two function outside evlist.c. Remove them from public namespace. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf record: Read from overwritable ring bufferWang Nan1-1/+30
Drive the evlist->bkw_mmap_state state machine during draining and when SIGUSR2 is received. Read the backward ring buffer in record__mmap_read_all. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: He Kuang <[email protected]> Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Setup backward mmap state machineWang Nan3-0/+98
Introduce a bkw_mmap_state state machine to evlist: .________________(forbid)_____________. | V NOTREADY --(0)--> RUNNING --(1)--> DATA_PENDING --(2)--> EMPTY ^ ^ | ^ | | |__(forbid)____/ |___(forbid)___/| | | \_________________(3)_______________/ NOTREADY : Backward ring buffers are not ready RUNNING : Backward ring buffers are recording DATA_PENDING : We are required to collect data from backward ring buffers EMPTY : We have collected data from backward ring buffers. (0): Setup backward ring buffer (1): Pause ring buffers for reading (2): Read from ring buffers (3): Resume ring buffers for recording We can't avoid this complexity. Since we deliberately drop records from overwritable ring buffer, there's no way for us to check remaining from ring buffer itself (by checking head and old pointers). Therefore, we need DATA_PENDING and EMPTY state to help us recording what we have done to the ring buffer. In record__mmap_read_evlist(), drive this state machine from DATA_PENDING to EMPTY. In perf_evlist__mmap_per_evsel(), drive this state machine from NOTREADY to RUNNING when creating backward mmap. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Drop evlist->backwardWang Nan3-6/+1
Now there's no real user of evlist->backward. Drop it. We are going to use evlist->backward_mmap as a container for backward ring buffer. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Map backward events to backward_mmapWang Nan2-29/+29
In perf_evlist__mmap_per_evsel(), select backward_mmap for backward events. Utilize new perf_mmap APIs. Dynamically alloc backward_mmap. Remove useless functions. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Introduce backward_mmap array for evlistWang Nan3-7/+16
Add backward_mmap to evlist, free it together with normal mmap. Improve perf_evlist__pick_pc(), search backward_mmap if evlist->mmap is not available. This patch doesn't alloc this array. It will be allocated conditionally in the following commits. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Extract common code in mmap failure processingWang Nan1-12/+8
In perf_evlist__mmap_per_cpu() and perf_evlist__mmap_per_thread(), in case of mmap failure, successfully created maps should be cleared. Current code uses two loops calling __perf_evlist__munmap() for each function. This patch extracts common code to perf_evlist__munmap_nofree() and use previous introduced decoupled API perf_mmap__munmap(). Now __perf_evlist__munmap() can be removed because of no user. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Record mmap cookie into fdarray private fieldWang Nan1-6/+9
Insetad of saving a index into fdarray entries private field, save the corresponding 'struct perf_mmap' pointer, and release them directly using perf_mmap__put(). Following commits introduce multiple mmap arrays to evlist. Without this patch, perf_evlist__munmap_filtered() is unable to retrive correct 'struct perf_mmap' pointer. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf record: Decouple record__mmap_read() and evlist.Wang Nan1-15/+22
Perf evlist will have multiple mmap arrays. Update record__mmap_read(): it should read from 'struct perf_mmap' directly. Also, make record__mmap_read() ready to read from backward ring buffer. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Update mmap related APIs and helpersWang Nan2-43/+108
Currently, the evlist mmap related helpers and APIs accept evlist and idx, and dereference 'struct perf_mmap' by evlist->mmap[idx]. This is unnecessary, and force each evlist contains only one mmap array. Following commits are going to introduce multiple mmap arrays to a evlist. This patch refators these APIs and helpers, introduces functions accept perf_mmap pointer directly. New helpers and APIs are decoupled with perf_evlist, and become perf_mmap functions (so they have perf_mmap prefix). Old functions are reimplemented with new functions. Some of them will be removed in following commits. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15tools lib fd array: Allow associating a pointer cookie with each entryWang Nan1-0/+1
Add a 'ptr' field to fdarray->priv array. This feature will be used by following commits, which introduce muiltiple 'struct perf_mmap' arrays for different types of mapping. Because of this, during fdarray__filter(), a simple 'idx' is not enough. Add a pointer cookie that allows to directly associate a 'struct perf_mmap' pointer to an fdarray entry. Signed-off-by: Wang Nan <[email protected]> Acked-by: Jiri Olsa <[email protected]> Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: He Kuang <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15tools: Simplify BITS_PER_LONG definePeter Zijlstra1-21/+2
Do it using (__CHAR_BIT__ * __SIZEOF_LONG__), simpler, works everywhere, reduces the complexity by ditching CONFIG_64BIT, that was being synthesized from yet another set of defines, which proved fragile, breaking the build on linux-next for no obvious reasons. Committer Note: Except on: gcc version 4.1.2 20080704 (Red Hat 4.1.2-55) Fallback to __WORDSIZE in that case... Reported-by: Stephen Rothwell <[email protected]> Signed-off-by: Peter Zijlstra <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Cc: Andy Lutomirski <[email protected]> Cc: H. Peter Anvin <[email protected]> Cc: Thomas Gleixner <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf evlist: Drop redundant evsel->overwrite indicatorArnaldo Carvalho de Melo4-10/+8
evsel->overwrite indicator means an event should be put into overwritable ring buffer. In current implementation, it equals to evsel->attr.write_backward. To reduce compliexity, remove evsel->overwrite, use evsel->attr.write_backward instead. In addition, in __perf_evsel__open(), if kernel doesn't support write_backward and user explicitly set it in evsel, don't fallback like other missing feature, since it is meaningless to fall back to a forward ring buffer in this case: we are unable to stably read from an forward overwritable ring buffer. Cc: He Kuang <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Masami Hiramatsu <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Nilay Vaish <[email protected]> Cc: Wang Nan <[email protected]> Cc: Zefan Li <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Wang Nan <[email protected]> Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15tools lib api fs: Use base 0 in filename__read_ullJiri Olsa1-1/+6
By using 0 for base, the strtoull() detects the base automatically (see 'man strtoull'). ATM we have just one user of this function, the cpu__get_max_freq function reading the "cpuinfo_max_freq" sysfs file. It should not get affected by this change. Committer note: This change seems motivated by this discussion: "[PATCH] [RFC V1]s390/perf: fix 'start' address of module's map" http://lkml.kernel.org/r/20160711120155.GA29929@krava I.e. this patches paves the way for filename__read_ull() to be used in a S/390 related fix. Signed-off-by: Jiri Olsa <[email protected]> Cc: David Ahern <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Songshan Gong <[email protected]> Link: http://lkml.kernel.org/r/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf tools: Bail out at "--sort dcacheline" and cacheline_size not knownArnaldo Carvalho de Melo1-1/+7
There are cases where further work would be needed to overcome the fact that neither sysconf(_SC_LEVEL1_DCACHE_LINESIZE) nor /sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size are available in some systems (Android, for instance), so bail out when such a situation takes place. Cc: Adrian Hunter <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15perf tools: Just pr_debug() about not being able to read cacheline_sizeArnaldo Carvalho de Melo1-1/+1
So far the cacheline_size is only useful for the "dcacheline" --sort order, i.e. if that is not used, which is the norm, then the user shouldn't care that he is running this, say, on an Android system where sysconf(_SC_LEVEL1_DCACHE_LINESIZE) and the /sys/devices/system/cpu/cpu0/cache/index0/coherency_line_size sysfs file isn't available. An upcoming patch will emit an warning only for "--sort ...,dcacheline,...". Cc: Adrian Hunter <[email protected]> Cc: Chris Phlipot <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-15Merge branch 'x86/asm' into x86/mm, to resolve conflictsIngo Molnar6-1/+2230
Conflicts: tools/testing/selftests/x86/Makefile Signed-off-by: Ingo Molnar <[email protected]>
2016-07-14tools: Make "__always_inline" just "inline" on AndroidArnaldo Carvalho de Melo1-0/+11
As the gcc there is producing tons of: "warning: always_inline function might not be inlinable" At least on android-ndk-r12/platforms/android-24/arch-arm, so, for the time being, use this big hammer. Cc: Adrian Hunter <[email protected]> Cc: Chris Phlipot <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-14perf tools: Do not provide dup sched_getcpu() prototype on AndroidArnaldo Carvalho de Melo1-1/+1
The Bionic libc has this definition, so don't duplicate it. Cc: Adrian Hunter <[email protected]> Cc: Chris Phlipot <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-14tools lib traceevent: Add correct header for ipv6 definitionsArnaldo Carvalho de Melo1-1/+2
We need to include netinet/in.h to get the in6_addr struct definition, needed to build it on the Android NDK: In file included from event-parse.c:36:0: /home/acme/android/android-ndk-r12/platforms/android-24/arch-arm/usr/include/netinet/ip6.h:82:18: error: field 'ip6_src' has incomplete type struct in6_addr ip6_src; /* source address */ And it is the canonical way of getting IPv6 definitions, as described, for instance, in Linux's 'man ipv6' Doing that uncovers another problem: this source file uses PRIu64 but doesn't include it, depending on it being included by chance via the now replaced header (netinet/ip6.h), fix it. Cc: Adrian Hunter <[email protected]> Cc: Chris Phlipot <[email protected]> Cc: David Ahern <[email protected]> Cc: Jiri Olsa <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Steven Rostedt <[email protected]> Cc: Wang Nan <[email protected]> Link: http://lkml.kernel.org/n/[email protected] Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-13perf test: Add a test case for SDT eventMasami Hiramatsu4-0/+121
Add a basic test case for SDT event support. This test scans an SDT event in perftools and check whether the SDT event is correctly stored into the buildid cache. Here is an example: ---- $ perf test sdt -v 47: Test SDT event probing : --- start --- test child forked, pid 20732 Found 72 SDTs in /home/mhiramat/ksrc/linux/tools/perf/perf Writing cache: %sdt_perf:test_target=test_target Cache committed: 0 symbol:test_target file:(null) line:0 offset:0 return:0 lazy:(null) test child finished with 0 ---- end ---- Test SDT event probing: Ok ---- Signed-off-by: Masami Hiramatsu <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Cc: Ananth N Mavinakayanahalli <[email protected]> Cc: Brendan Gregg <[email protected]> Cc: Hemant Kumar <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Link: http://lkml.kernel.org/r/146831796546.17065.1502584370844087537.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-13perf build: Add sdt feature detectionMasami Hiramatsu7-3/+34
This checks whether sys/sdt.h is available or not, which is required for DTRACE_PROBE(). We can disable this feature by passing NO_SDT=1 when building. This flag will be used for SDT test case and further SDT events in perftools. Signed-off-by: Masami Hiramatsu <[email protected]> Cc: Ananth N Mavinakayanahalli <[email protected]> Cc: Brendan Gregg <[email protected]> Cc: Hemant Kumar <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Link: http://lkml.kernel.org/r/146831795615.17065.17513820540591053933.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-13perf probe: Support a special SDT probe formatMasami Hiramatsu2-3/+13
Support a special SDT probe format which can omit the '%' prefix only if the SDT group name starts with "sdt_". So, for example both of "%sdt_libc:setjump" and "sdt_libc:setjump" are acceptable for perf probe --add. E.g. without this: # perf probe -a sdt_libc:setjmp Semantic error :There is non-digit char in line number. ... With this: # perf probe -a sdt_libc:setjmp Added new event: sdt_libc:setjmp (on %setjmp in /usr/lib64/libc-2.20.so) You can now use it in all perf tools, such as: perf record -e sdt_libc:setjmp -aR sleep 1 Suggested-by: Brendan Gregg <[email protected]> Signed-off-by: Masami Hiramatsu <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Cc: Ananth N Mavinakayanahalli <[email protected]> Cc: Brendan Gregg <[email protected]> Cc: Hemant Kumar <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Link: http://lkml.kernel.org/r/146831794674.17065.13359473252168740430.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
2016-07-13perf probe: Support @BUILDID or @FILE suffix for SDT eventsMasami Hiramatsu3-2/+59
Support @BUILDID or @FILE suffix for SDT events. This allows perf to add probes on SDTs/pre-cached events on given FILE or the file which has given BUILDID (also, this complements BUILDID.) For example, both gcc and libstdc++ has same SDTs as below. If you would like to add a probe on sdt_libstdcxx:catch on gcc, you can do as below. ---- # perf list sdt | tail -n 6 sdt_libstdcxx:catch@/usr/bin/gcc(0cc207fc4b27) [SDT event] sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49) sdt_libstdcxx:rethrow@/usr/bin/gcc(0cc207fc4b27) [SDT event] sdt_libstdcxx:rethrow@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49) sdt_libstdcxx:throw@/usr/bin/gcc(0cc207fc4b27) [SDT event] sdt_libstdcxx:throw@/usr/lib64/libstdc++.so.6.0.20(91c7a88fdf49) # perf probe -a %sdt_libstdcxx:catch@0cc Added new event: sdt_libstdcxx:catch (on %catch in /usr/bin/gcc) You can now use it in all perf tools, such as: perf record -e sdt_libstdcxx:catch -aR sleep 1 ---- Committer note: Doing the full sequence of steps to get the results above: With a clean build-id cache: [root@jouet ~]# rm -rf ~/.debug/ [root@jouet ~]# perf list sdt List of pre-defined events (to be used in -e): [root@jouet ~]# No events whatsoever, then, we can add all events in gcc to the build-id cache, doing a --add + --dry-run: [root@jouet ~]# perf probe --dry-run --cache -x /usr/bin/gcc --add %sdt_libstdcxx:\* Added new events: sdt_libstdcxx:throw (on %* in /usr/bin/gcc) sdt_libstdcxx:rethrow (on %* in /usr/bin/gcc) sdt_libstdcxx:catch (on %* in /usr/bin/gcc) You can now use it in all perf tools, such as: perf record -e sdt_libstdcxx:catch -aR sleep 1 [root@jouet ~]# It really didn't add any events, it just cached them: [root@jouet ~]# perf probe -l [root@jouet ~]# We can see that it was cached as: [root@jouet ~]# ls -la ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/ total 976 drwxr-xr-x. 2 root root 4096 Jul 13 21:47 . drwxr-xr-x. 3 root root 4096 Jul 13 21:47 .. -rwxr-xr-x. 4 root root 985912 Jun 22 18:52 elf -rw-r--r--. 1 root root 303 Jul 13 21:47 probes [root@jouet ~]# file ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/elf /root/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/elf: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked, interpreter /lib64/ld-linux-x86-64.so.2, for GNU/Linux 2.6.32, BuildID[sha1]=9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2, stripped [root@jouet ~]# cat ~/.debug/usr/bin/gcc/9a0730e2bcc6d2a2003d21ac46807e8ee6bcb7c2/probes %sdt_libstdcxx:throw=throw p:sdt_libstdcxx/throw /usr/bin/gcc:0x71ffd %sdt_libstdcxx:rethrow=rethrow p:sdt_libstdcxx/rethrow /usr/bin/gcc:0x720b8 %sdt_libstdcxx:catch=catch p:sdt_libstdcxx/catch /usr/bin/gcc:0x7307f %sdt_libgcc:unwind=unwind p:sdt_libgcc/unwind /usr/bin/gcc:0x7eec0 #sdt_libstdcxx:*=%* [root@jouet ~]# Ok, now we can use 'perf probe' to refer to those cached entries as: Humm, nope, doing as above we end up with: [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch Semantic error :* is bad for event name -it must follow C symbol-naming rule. Error: Failed to add events. [root@jouet ~]# But it worked at some point, lets try not using --dry-run: Resetting everything: # rm -rf ~/.debug/ # perf probe -d *:* # perf probe -l # perf list sdt List of pre-defined events (to be used in -e): # Ok, now it cached everything, even things we haven't asked it to (sdt_libgcc:unwind): [root@jouet ~]# perf probe -x /usr/bin/gcc --add %sdt_libstdcxx:\* Added new events: sdt_libstdcxx:throw (on %* in /usr/bin/gcc) sdt_libstdcxx:rethrow (on %* in /usr/bin/gcc) sdt_libstdcxx:catch (on %* in /usr/bin/gcc) You can now use it in all perf tools, such as: perf record -e sdt_libstdcxx:catch -aR sleep 1 [root@jouet ~]# perf list sdt List of pre-defined events (to be used in -e): sdt_libgcc:unwind [SDT event] sdt_libstdcxx:catch [SDT event] sdt_libstdcxx:rethrow [SDT event] sdt_libstdcxx:throw [SDT event] [root@jouet ~]# And we have the events in place: [root@jouet ~]# perf probe -l sdt_libstdcxx:catch (on execute_cfa_program+1551@../../../libgcc/unwind-dw2.c in /usr/bin/gcc) sdt_libstdcxx:rethrow (on d_print_subexpr+280@libsupc++/cp-demangle.c in /usr/bin/gcc) sdt_libstdcxx:throw (on d_print_subexpr+93@libsupc++/cp-demangle.c in /usr/bin/gcc) [root@jouet ~]# And trying to use them at least has 'perf trace --event sdt*:*' working. Then, if we try to add the ones in libstdc++: [root@jouet ~]# perf probe -x /usr/lib64/libstdc++.so.6 -a %sdt_libstdcxx:\* Error: event "catch" already exists. Hint: Remove existing event by 'perf probe -d' or force duplicates by 'perf probe -f' or set 'force=yes' in BPF source. Error: Failed to add events. [root@jouet ~]# Doesn't work, dups, but at least this served to, unbeknownst to the user, add the SDT probes in /usr/lib64/libstdc++.so.6! [root@jouet ~]# perf list sdt List of pre-defined events (to be used in -e): sdt_libgcc:unwind [SDT event] sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6) [SDT event] sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event] sdt_libstdcxx:rethrow@/usr/bin/gcc(9a0730e2bcc6) [SDT event] sdt_libstdcxx:rethrow@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event] sdt_libstdcxx:throw@/usr/bin/gcc(9a0730e2bcc6) [SDT event] sdt_libstdcxx:throw@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event] [root@jouet ~]# Now we should be able to get to the original cset comment, if we remove all SDTs events in place, not from the cache, from the kernel, where it was set up as: [root@jouet ~]# ls -la /sys/kernel/debug/tracing/events/sdt_libstdcxx/ total 0 drwxr-xr-x. 5 root root 0 Jul 13 22:00 . drwxr-xr-x. 80 root root 0 Jul 13 21:56 .. drwxr-xr-x. 2 root root 0 Jul 13 22:00 catch -rw-r--r--. 1 root root 0 Jul 13 22:00 enable -rw-r--r--. 1 root root 0 Jul 13 22:00 filter drwxr-xr-x. 2 root root 0 Jul 13 22:00 rethrow drwxr-xr-x. 2 root root 0 Jul 13 22:00 throw [root@jouet ~]# [root@jouet ~]# head -2 /sys/kernel/debug/tracing/events/sdt_libstdcxx/throw/format name: throw ID: 2059 [root@jouet ~]# Now to remove it: [root@jouet ~]# perf probe -d sdt_libstdc*:* Removed event: sdt_libstdcxx:catch Removed event: sdt_libstdcxx:rethrow Removed event: sdt_libstdcxx:throw [root@jouet ~]# Which caused: [root@jouet ~]# ls -la /sys/kernel/debug/tracing/events/sdt_libstdcxx/ ls: cannot access '/sys/kernel/debug/tracing/events/sdt_libstdcxx/': No such file or directory [root@jouet ~]# Ok, now we can do: [root@jouet ~]# perf list sdt_libstdcxx:catch List of pre-defined events (to be used in -e): sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6) [SDT event] sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event] [root@jouet ~]# So, these are not really 'pre-defined events', i.e. we can't use them with 'perf record --event': [root@jouet ~]# perf record --event sdt_libstdcxx:catch* event syntax error: 'sdt_libstdcxx:catch*' \___ unknown tracepoint Error: File /sys/kernel/debug/tracing/events/sdt_libstdcxx/catch* not found. Hint: Perhaps this kernel misses some CONFIG_ setting to enable this feature?. <SNIP> [root@jouet ~]# To have it really pre-defined we must use perf probe to get its definition from the cache and set it up in the kernel, creating the tracepoint to _then_ use it with 'perf record --event': [root@jouet ~]# perf probe -a sdt_libstdcxx:catch Semantic error :There is non-digit char in line number. <SNIP> Oops, there is another gotcha here, we need that pesky '%' character: [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch Added new events: sdt_libstdcxx:catch (on %catch in /usr/bin/gcc) sdt_libstdcxx:catch_1 (on %catch in /usr/lib64/libstdc++.so.6.0.22) You can now use it in all perf tools, such as: perf record -e sdt_libstdcxx:catch_1 -aR sleep 1 [root@jouet ~]# But then we added _two_ events, one with the name we expected, the other one with a _ added, when doing the analysis we need to pay attention to who maps to who. And here is where we get to the point of this patch, which is to be able to disambiguate those definitions for 'catch' in the build-id cache, but first we need remove those events we just added: [root@jouet ~]# perf probe -d %sdt_libstdcxx:catch Oops, that didn't remove anything, we need to _remove_ that % char in this case: [root@jouet ~]# perf probe -d sdt_libstdcxx:catch Removed event: sdt_libstdcxx:catch And we need to remove the other event added, i.e. I forgot to add a * at the end: [root@jouet ~]# perf probe -d sdt_libstdcxx:catch* Removed event: sdt_libstdcxx:catch_1 [root@jouet ~]# Ok, disambiguating it using what is in this patch: [root@jouet ~]# perf list sdt_libstdcxx:catch List of pre-defined events (to be used in -e): sdt_libstdcxx:catch@/usr/bin/gcc(9a0730e2bcc6) [SDT event] sdt_libstdcxx:catch@/usr/lib64/libstdc++.so.6.0.22(ef2b7066559a) [SDT event] [root@jouet ~]# [root@jouet ~]# perf probe -a %sdt_libstdcxx:catch@9a07 Added new event: sdt_libstdcxx:catch (on %catch in /usr/bin/gcc) You can now use it in all perf tools, such as: perf record -e sdt_libstdcxx:catch -aR sleep 1 [root@jouet ~]# perf probe -l sdt_libstdcxx:catch (on execute_cfa_program+1551@../../../libgcc/unwind-dw2.c in /usr/bin/gcc) [root@jouet ~]# Yeah, it works! But we need to try and simplify this :-) Update: Some aspects of this simplification take place in the following patches. Signed-off-by: Masami Hiramatsu <[email protected]> Tested-by: Arnaldo Carvalho de Melo <[email protected]> Cc: Ananth N Mavinakayanahalli <[email protected]> Cc: Brendan Gregg <[email protected]> Cc: Hemant Kumar <[email protected]> Cc: Namhyung Kim <[email protected]> Cc: Peter Zijlstra <[email protected]> Link: http://lkml.kernel.org/r/146831793746.17065.13065062753978236612.stgit@devbox Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
Powered by cgit, Gruvboxed by Blaster4385.