From 9886167d20c0720dcfb01e62cdff4d906b226f43 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 3 Oct 2013 16:02:23 +0200 Subject: [PATCH 1/8] perf: Fix perf_pmu_migrate_context While auditing the list_entry usage due to a trinity bug I found that perf_pmu_migrate_context violates the rules for perf_event::event_entry. The problem is that perf_event::event_entry is a RCU list element, and hence we must wait for a full RCU grace period before re-using the element after deletion. Therefore the usage in perf_pmu_migrate_context() which re-uses the entry immediately is broken. For now introduce another list_head into perf_event for this specific usage. This doesn't actually fix the trinity report because that never goes through this code. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-mkj72lxagw1z8fvjm648iznw@git.kernel.org Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 24 +++++++++++++++++++++++- kernel/events/core.c | 6 +++--- 2 files changed, 26 insertions(+), 4 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 866e85c5eb94..c8ba627c1d60 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -294,9 +294,31 @@ struct ring_buffer; */ struct perf_event { #ifdef CONFIG_PERF_EVENTS - struct list_head group_entry; + /* + * entry onto perf_event_context::event_list; + * modifications require ctx->lock + * RCU safe iterations. + */ struct list_head event_entry; + + /* + * XXX: group_entry and sibling_list should be mutually exclusive; + * either you're a sibling on a group, or you're the group leader. + * Rework the code to always use the same list element. + * + * Locked for modification by both ctx->mutex and ctx->lock; holding + * either sufficies for read. + */ + struct list_head group_entry; struct list_head sibling_list; + + /* + * We need storage to track the entries in perf_pmu_migrate_context; we + * cannot use the event_entry because of RCU and we want to keep the + * group in tact which avoids us using the other two entries. + */ + struct list_head migrate_entry; + struct hlist_node hlist_entry; int nr_siblings; int group_flags; diff --git a/kernel/events/core.c b/kernel/events/core.c index cb4238e85b38..d49a9d29334c 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -7234,15 +7234,15 @@ void perf_pmu_migrate_context(struct pmu *pmu, int src_cpu, int dst_cpu) perf_remove_from_context(event); unaccount_event_cpu(event, src_cpu); put_ctx(src_ctx); - list_add(&event->event_entry, &events); + list_add(&event->migrate_entry, &events); } mutex_unlock(&src_ctx->mutex); synchronize_rcu(); mutex_lock(&dst_ctx->mutex); - list_for_each_entry_safe(event, tmp, &events, event_entry) { - list_del(&event->event_entry); + list_for_each_entry_safe(event, tmp, &events, migrate_entry) { + list_del(&event->migrate_entry); if (event->state >= PERF_EVENT_STATE_OFF) event->state = PERF_EVENT_STATE_INACTIVE; account_event_cpu(event, dst_cpu); From d8b11a0cbd1c66ce283eb9dabe0498dfa6483f32 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 3 Oct 2013 16:00:14 +0200 Subject: [PATCH 2/8] perf/x86: Clean up cap_user_time* setting Currently the cap_user_time_zero capability has different tests than cap_user_time; even though they expose the exact same data. Switch from CONSTANT && NONSTOP to sched_clock_stable to also deal with multi cabinet machines and drop the tsc_disabled() check.. non of this will work sanely without tsc anyway. Signed-off-by: Peter Zijlstra Link: http://lkml.kernel.org/n/tip-nmgn0j0muo1r4c94vlfh23xy@git.kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c index 897783b3302a..9d8449158cf9 100644 --- a/arch/x86/kernel/cpu/perf_event.c +++ b/arch/x86/kernel/cpu/perf_event.c @@ -1888,10 +1888,7 @@ void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now) userpg->cap_user_rdpmc = x86_pmu.attr_rdpmc; userpg->pmc_width = x86_pmu.cntval_bits; - if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) - return; - - if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) + if (!sched_clock_stable) return; userpg->cap_user_time = 1; @@ -1899,10 +1896,8 @@ void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now) userpg->time_shift = CYC2NS_SCALE_FACTOR; userpg->time_offset = this_cpu_read(cyc2ns_offset) - now; - if (sched_clock_stable && !check_tsc_disabled()) { - userpg->cap_user_time_zero = 1; - userpg->time_zero = this_cpu_read(cyc2ns_offset); - } + userpg->cap_user_time_zero = 1; + userpg->time_zero = this_cpu_read(cyc2ns_offset); } /* From 6adb0b0ae26fcc35cfec068d71f13863faac5b44 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Sun, 22 Sep 2013 19:44:59 -0600 Subject: [PATCH 3/8] perf tools: Add default handler for mmap2 events Commands that do not implement an mmap2 handler should at least not die with a segfault when processing files with MMAP2 events. Signed-off-by: David Ahern Link: http://lkml.kernel.org/r/1379900700-5186-5-git-send-email-dsahern@gmail.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/session.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 70ffa41518f3..37c4718f6be1 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -256,6 +256,8 @@ void perf_tool__fill_defaults(struct perf_tool *tool) tool->sample = process_event_sample_stub; if (tool->mmap == NULL) tool->mmap = process_event_stub; + if (tool->mmap2 == NULL) + tool->mmap2 = process_event_stub; if (tool->comm == NULL) tool->comm = process_event_stub; if (tool->fork == NULL) From d20a47e70b57a77f4d1b0dba98a668e14c2f9bf9 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Mon, 30 Sep 2013 18:01:11 +0900 Subject: [PATCH 4/8] perf stat: Set child_pid after perf_evlist__prepare_workload() The commit acf2892270dc ("perf stat: Use perf_evlist__prepare/ start_workload()") converted to use the function but forgot to update child_pid. Fix it. Signed-off-by: Namhyung Kim Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1380531671-28076-1-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-stat.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index f686d5ff594e..5098f144b92d 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -457,6 +457,7 @@ static int __run_perf_stat(int argc, const char **argv) perror("failed to prepare workload"); return -1; } + child_pid = evsel_list->workload.pid; } if (group) From 47a92b828639dcb1a6033cd0a441099df828b7a9 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Wed, 2 Oct 2013 18:25:26 -0300 Subject: [PATCH 5/8] perf tools: Fix libaudit test MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit In ubuntu systems the libaudit test was always failing due to the newline in the printf call not being escaped, which somehow didn't prevented the test from working as expected on other systems, such as fedora18. Fix it by removing the newline, as this is just a test, that program is just a compile test. The error messages, obtained using 'make V=1': CHK libaudit : In function ‘main’: :5:9: error: missing terminating " character [-Werror] :5:2: error: missing terminating " character :6:1: error: missing terminating " character [-Werror] :6:1: error: missing terminating " character :7:2: error: expected expression before ‘return’ :8:1: error: expected ‘;’ before ‘}’ token cc1: all warnings being treated as errors config/Makefile:241: No libaudit.h found, disables 'trace' tool, please install audit-libs-devel or libaudit-dev After this change the test works as expected in all systems tested and the 'trace' tool is built when the needed devel packages are installed. Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-0trw8qs9hafeopc0vj1sicay@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/config/feature-tests.mak | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/config/feature-tests.mak b/tools/perf/config/feature-tests.mak index d5a8dd44945f..f79305739ecc 100644 --- a/tools/perf/config/feature-tests.mak +++ b/tools/perf/config/feature-tests.mak @@ -219,7 +219,7 @@ define SOURCE_LIBAUDIT int main(void) { - printf(\"error message: %s\n\", audit_errno_to_name(0)); + printf(\"error message: %s\", audit_errno_to_name(0)); return audit_open(); } endef From e08cfd4bda7683cdbe6971c26cf23e2afdb1e7a8 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Mon, 30 Sep 2013 18:21:44 +0900 Subject: [PATCH 6/8] perf probe: Fix to find line information for probe list Fix to find the correct (as much as possible) line information for listing probes. Without this fix, perf probe --list action will show incorrect line information as below; probe:getname_flags (on getname_flags@ksrc/linux-3/fs/namei.c) probe:getname_flags_1 (on getname:-89@x86/include/asm/current.h) probe:getname_flags_2 (on user_path_at_empty:-2054@x86/include/asm/current.h) The minus line number is obviously wrong, and current.h is not related to the probe point. Deeper investigation discovered that there were 2 issues related to this bug, and minor typos too. The 1st issue is the rack of considering about nested inlined functions, which causes the wrong (relative) line number. The 2nd issue is that the dwarf line info is not correct at those points. It points 14th line of current.h. Since it seems that the line info includes somewhat unreliable information, this fixes perf to try to find correct line information from both of debuginfo and line info as below. 1) Probe address is the entry of a function instance In this case, the line is set as the function declared line. 2) Probe address is the entry of an expanded inline function block In this case, the line is set as the function call-site line. This means that the line number is relative from the entry line of caller function (which can be an inlined function if nested) 3) Probe address is inside a function instance or an expanded inline function block In this case, perf probe queries the line number from lineinfo and verify the function declared file is same as the file name queried from lineinfo. If the file name is different, it is a failure case. The probe address is shown as symbol+offset. 4) Probe address is not in the any function instance This is a failure case, the probe address is shown as symbol+offset. With this fix, perf probe -l shows correct probe lines as below; probe:getname_flags (on getname_flags@ksrc/linux-3/fs/namei.c) probe:getname_flags_1 (on getname:2@ksrc/linux-3/fs/namei.c) probe:getname_flags_2 (on user_path_at_empty:4@ksrc/linux-3/fs/namei.c) Changes at v2: - Fix typos in the function comments. (Thanks to Namhyung Kim) - Use die_find_top_inlinefunc instead of die_find_inlinefunc_next. Signed-off-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20130930092144.1693.11058.stgit@udc4-manage.rcp.hitachi.co.jp Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/dwarf-aux.c | 27 +++++++++++++++---- tools/perf/util/dwarf-aux.h | 6 ++++- tools/perf/util/probe-finder.c | 49 +++++++++++++++++++++++----------- 3 files changed, 60 insertions(+), 22 deletions(-) diff --git a/tools/perf/util/dwarf-aux.c b/tools/perf/util/dwarf-aux.c index e23bde19d590..7defd77105d0 100644 --- a/tools/perf/util/dwarf-aux.c +++ b/tools/perf/util/dwarf-aux.c @@ -426,7 +426,7 @@ static int __die_search_func_cb(Dwarf_Die *fn_die, void *data) * @die_mem: a buffer for result DIE * * Search a non-inlined function DIE which includes @addr. Stores the - * DIE to @die_mem and returns it if found. Returns NULl if failed. + * DIE to @die_mem and returns it if found. Returns NULL if failed. */ Dwarf_Die *die_find_realfunc(Dwarf_Die *cu_die, Dwarf_Addr addr, Dwarf_Die *die_mem) @@ -454,15 +454,32 @@ static int __die_find_inline_cb(Dwarf_Die *die_mem, void *data) } /** - * die_find_inlinefunc - Search an inlined function at given address - * @cu_die: a CU DIE which including @addr + * die_find_top_inlinefunc - Search the top inlined function at given address + * @sp_die: a subprogram DIE which including @addr * @addr: target address * @die_mem: a buffer for result DIE * * Search an inlined function DIE which includes @addr. Stores the - * DIE to @die_mem and returns it if found. Returns NULl if failed. + * DIE to @die_mem and returns it if found. Returns NULL if failed. + * Even if several inlined functions are expanded recursively, this + * doesn't trace it down, and returns the topmost one. + */ +Dwarf_Die *die_find_top_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr, + Dwarf_Die *die_mem) +{ + return die_find_child(sp_die, __die_find_inline_cb, &addr, die_mem); +} + +/** + * die_find_inlinefunc - Search an inlined function at given address + * @sp_die: a subprogram DIE which including @addr + * @addr: target address + * @die_mem: a buffer for result DIE + * + * Search an inlined function DIE which includes @addr. Stores the + * DIE to @die_mem and returns it if found. Returns NULL if failed. * If several inlined functions are expanded recursively, this trace - * it and returns deepest one. + * it down and returns deepest one. */ Dwarf_Die *die_find_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr, Dwarf_Die *die_mem) diff --git a/tools/perf/util/dwarf-aux.h b/tools/perf/util/dwarf-aux.h index 8658d41697d2..b4fe90c6cb2d 100644 --- a/tools/perf/util/dwarf-aux.h +++ b/tools/perf/util/dwarf-aux.h @@ -79,7 +79,11 @@ extern Dwarf_Die *die_find_child(Dwarf_Die *rt_die, extern Dwarf_Die *die_find_realfunc(Dwarf_Die *cu_die, Dwarf_Addr addr, Dwarf_Die *die_mem); -/* Search an inlined function including given address */ +/* Search the top inlined function including given address */ +extern Dwarf_Die *die_find_top_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr, + Dwarf_Die *die_mem); + +/* Search the deepest inlined function including given address */ extern Dwarf_Die *die_find_inlinefunc(Dwarf_Die *sp_die, Dwarf_Addr addr, Dwarf_Die *die_mem); diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 371476cb8ddc..c09e0a9fdf4c 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -1327,8 +1327,8 @@ int debuginfo__find_probe_point(struct debuginfo *self, unsigned long addr, struct perf_probe_point *ppt) { Dwarf_Die cudie, spdie, indie; - Dwarf_Addr _addr, baseaddr; - const char *fname = NULL, *func = NULL, *tmp; + Dwarf_Addr _addr = 0, baseaddr = 0; + const char *fname = NULL, *func = NULL, *basefunc = NULL, *tmp; int baseline = 0, lineno = 0, ret = 0; /* Adjust address with bias */ @@ -1349,27 +1349,36 @@ int debuginfo__find_probe_point(struct debuginfo *self, unsigned long addr, /* Find a corresponding function (name, baseline and baseaddr) */ if (die_find_realfunc(&cudie, (Dwarf_Addr)addr, &spdie)) { /* Get function entry information */ - tmp = dwarf_diename(&spdie); - if (!tmp || + func = basefunc = dwarf_diename(&spdie); + if (!func || dwarf_entrypc(&spdie, &baseaddr) != 0 || - dwarf_decl_line(&spdie, &baseline) != 0) + dwarf_decl_line(&spdie, &baseline) != 0) { + lineno = 0; goto post; - func = tmp; + } - if (addr == (unsigned long)baseaddr) + if (addr == (unsigned long)baseaddr) { /* Function entry - Relative line number is 0 */ lineno = baseline; - else if (die_find_inlinefunc(&spdie, (Dwarf_Addr)addr, - &indie)) { + fname = dwarf_decl_file(&spdie); + goto post; + } + + /* Track down the inline functions step by step */ + while (die_find_top_inlinefunc(&spdie, (Dwarf_Addr)addr, + &indie)) { + /* There is an inline function */ if (dwarf_entrypc(&indie, &_addr) == 0 && - _addr == addr) + _addr == addr) { /* * addr is at an inline function entry. * In this case, lineno should be the call-site - * line number. + * line number. (overwrite lineinfo) */ lineno = die_get_call_lineno(&indie); - else { + fname = die_get_call_file(&indie); + break; + } else { /* * addr is in an inline function body. * Since lineno points one of the lines @@ -1377,19 +1386,27 @@ int debuginfo__find_probe_point(struct debuginfo *self, unsigned long addr, * be the entry line of the inline function. */ tmp = dwarf_diename(&indie); - if (tmp && - dwarf_decl_line(&spdie, &baseline) == 0) - func = tmp; + if (!tmp || + dwarf_decl_line(&indie, &baseline) != 0) + break; + func = tmp; + spdie = indie; } } + /* Verify the lineno and baseline are in a same file */ + tmp = dwarf_decl_file(&spdie); + if (!tmp || strcmp(tmp, fname) != 0) + lineno = 0; } post: /* Make a relative line number or an offset */ if (lineno) ppt->line = lineno - baseline; - else if (func) + else if (basefunc) { ppt->offset = addr - (unsigned long)baseaddr; + func = basefunc; + } /* Duplicate strings */ if (func) { From 027a7e86a17fb2d172f4eb17cd6b0e9f139224fd Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 4 Oct 2013 10:24:17 -0300 Subject: [PATCH 7/8] perf tools: Fix installation of libexec components Doing a fresh install on a user home directory needs to first make sure that the ~/libexec/perf-core/ directory is present so that 'perf-archive' like scripts, 'perf test' attr config files and 'perf script' scripts can be installed. Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-z7ryi3r1b9dn9smbfnab0fdc@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Makefile | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/perf/Makefile b/tools/perf/Makefile index 3a0ff7fb71b6..64c043b7a438 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -770,6 +770,7 @@ check: $(OUTPUT)common-cmds.h install-bin: all $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(bindir_SQ)' $(INSTALL) $(OUTPUT)perf '$(DESTDIR_SQ)$(bindir_SQ)' + $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' $(INSTALL) $(OUTPUT)perf-archive -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' ifndef NO_LIBPERL $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace' From b314e5cfd11fd78545ce6c2be42646254390c1aa Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Mon, 30 Sep 2013 17:19:48 +0900 Subject: [PATCH 8/8] perf session: Fix infinite loop on invalid perf.data file perf-record updates the header in the perf.data file at termination. Without this update perf-report (and other processing built-ins) it caused an infinite loop when perf report (or something like) called. This is because the algorithm in __perf_session__process_events() depends on the data_size which is read from file header. Use file size directly instead in this case to do the best-effort processing. Signed-off-by: Namhyung Kim Tested-by: David Ahern Tested-by: Sonny Rao Acked-by: Ingo Molnar Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Sonny Rao Link: http://lkml.kernel.org/r/1380529188-27193-1-git-send-email-namhyung@kernel.org Signed-off-by: David Ahern [ Reworded warning as per Ingo Molnar suggestion, replaces 'perf.data' with session->filename, to precisely identify the data file involved ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/header.c | 12 ++++++++++++ tools/perf/util/session.c | 2 +- 2 files changed, 13 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index ce69901176d8..c3e5a3b817ab 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -2768,6 +2768,18 @@ int perf_session__read_header(struct perf_session *session) if (perf_file_header__read(&f_header, header, fd) < 0) return -EINVAL; + /* + * Sanity check that perf.data was written cleanly; data size is + * initialized to 0 and updated only if the on_exit function is run. + * If data size is still 0 then the file contains only partial + * information. Just warn user and process it as much as it can. + */ + if (f_header.data.size == 0) { + pr_warning("WARNING: The %s file's data size field is 0 which is unexpected.\n" + "Was the 'perf record' command properly terminated?\n", + session->filename); + } + nr_attrs = f_header.attrs.size / f_header.attr_size; lseek(fd, f_header.attrs.offset, SEEK_SET); diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 37c4718f6be1..568b750c01f6 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -1312,7 +1312,7 @@ int __perf_session__process_events(struct perf_session *session, file_offset = page_offset; head = data_offset - page_offset; - if (data_offset + data_size < file_size) + if (data_size && (data_offset + data_size < file_size)) file_size = data_offset + data_size; progress_next = file_size / 16;