From 914a76ca5eedc9f286a36f61c4eaa95b451ba3e6 Mon Sep 17 00:00:00 2001 From: Andi Kleen Date: Wed, 16 Mar 2011 15:44:33 -0400 Subject: [PATCH 1/5] oprofile, x86: Allow setting EDGE/INV/CMASK for counter events For some performance events it's useful to set the EDGE and INV bits and the CMASK mask in the counter control register. The list of predefined events Intel releases for each CPU has some events which require these settings to get more "natural" to use higher level events. oprofile currently doesn't allow this. This patch adds new extra configuration fields for them, so that they can be specified in oprofilefs. An updated oprofile daemon can then make use of this to set them. v2: Write back masked extra value to variable. Signed-off-by: Andi Kleen Signed-off-by: Robert Richter --- arch/x86/oprofile/nmi_int.c | 5 +++++ arch/x86/oprofile/op_counter.h | 1 + 2 files changed, 6 insertions(+) diff --git a/arch/x86/oprofile/nmi_int.c b/arch/x86/oprofile/nmi_int.c index e2b7b0c06cdf..ee165f1a1f37 100644 --- a/arch/x86/oprofile/nmi_int.c +++ b/arch/x86/oprofile/nmi_int.c @@ -49,6 +49,10 @@ u64 op_x86_get_ctrl(struct op_x86_model_spec const *model, val |= counter_config->user ? ARCH_PERFMON_EVENTSEL_USR : 0; val |= counter_config->kernel ? ARCH_PERFMON_EVENTSEL_OS : 0; val |= (counter_config->unit_mask & 0xFF) << 8; + counter_config->extra &= (ARCH_PERFMON_EVENTSEL_INV | + ARCH_PERFMON_EVENTSEL_EDGE | + ARCH_PERFMON_EVENTSEL_CMASK); + val |= counter_config->extra; event &= model->event_mask ? model->event_mask : 0xFF; val |= event & 0xFF; val |= (event & 0x0F00) << 24; @@ -440,6 +444,7 @@ static int nmi_create_files(struct super_block *sb, struct dentry *root) oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask); oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel); oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user); + oprofilefs_create_ulong(sb, dir, "extra", &counter_config[i].extra); } return 0; diff --git a/arch/x86/oprofile/op_counter.h b/arch/x86/oprofile/op_counter.h index e28398df0df2..0b7b7b179cbe 100644 --- a/arch/x86/oprofile/op_counter.h +++ b/arch/x86/oprofile/op_counter.h @@ -22,6 +22,7 @@ struct op_counter_config { unsigned long kernel; unsigned long user; unsigned long unit_mask; + unsigned long extra; }; extern struct op_counter_config counter_config[]; From 6c6804fb2cef2d6aceb38b0eb0803210d77ff390 Mon Sep 17 00:00:00 2001 From: Andrew Lutomirski Date: Thu, 24 Mar 2011 00:36:56 -0400 Subject: [PATCH 2/5] perf symbols: Fix vsyscall symbol lookup Perf can't currently trace into the vsyscall page. It looks like it was meant to work. Tested on 2.6.38 and today's -git. The bug is easy to reproduce. Compile this: int main() { int i; struct timespec t; for(i = 0; i < 10000000; i++) clock_gettime(CLOCK_MONOTONIC, &t); return 0; } and run it through perf record; perf report. The top entry shows "[unknown]" and you can't zoom in. It looks like there are two issues. The first is a that a test for user mode executing in kernel space is backwards. (That's the first hunk below). The second (I think) is that something's wrong with the code that generates lots of little struct dso objects for different sections -- when it runs on vmlinux it results in bogus long_name values which cause objdump to fail. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra LPU-Reference: Signed-off-by: Andy Lutomirski Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 2 +- tools/perf/util/symbol.c | 3 +++ 2 files changed, 4 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 2b15c362ef56..1023f67633a4 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -710,7 +710,7 @@ try_again: * in the whole kernel symbol list. */ if ((long long)al->addr < 0 && - cpumode == PERF_RECORD_MISC_KERNEL && + cpumode == PERF_RECORD_MISC_USER && machine && mg != &machine->kmaps) { mg = &machine->kmaps; goto try_again; diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 17df793c8924..8f73907a959e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1196,6 +1196,8 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name, if (curr_dso == NULL) goto out_elf_end; curr_dso->kernel = self->kernel; + curr_dso->long_name = self->long_name; + curr_dso->long_name_len = self->long_name_len; curr_map = map__new2(start, curr_dso, map->type); if (curr_map == NULL) { @@ -1842,6 +1844,7 @@ int dso__load_vmlinux(struct dso *self, struct map *map, if (fd < 0) return -1; + dso__set_long_name(self, (char *)vmlinux); dso__set_loaded(self, map->type); err = dso__load_sym(self, map, symfs_vmlinux, fd, filter, 0, 0); close(fd); From c286c419c784c238cd699be37fec7a9acc30d89f Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 28 Mar 2011 09:50:11 -0300 Subject: [PATCH 3/5] perf tools: Fixup exit path when not able to open events We have to deal with the TUI mode in perf top, so that we don't end up with a garbled screen when, say, a non root user on a machine with a paranoid setting (the default) tries to use 'perf top'. Introduce a ui__warning_paranoid() routine shared by top and record that tells the user the valid values for /proc/sys/kernel/perf_event_paranoid. Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 9 ++++----- tools/perf/builtin-top.c | 38 +++++++++++++++++++++++-------------- tools/perf/util/debug.c | 10 ++++++++++ tools/perf/util/debug.h | 1 + 4 files changed, 39 insertions(+), 19 deletions(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 6febcc168a8c..623695e18254 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -275,11 +275,10 @@ try_again: !no_inherit) < 0) { int err = errno; - if (err == EPERM || err == EACCES) - die("Permission error - are you root?\n" - "\t Consider tweaking" - " /proc/sys/kernel/perf_event_paranoid.\n"); - else if (err == ENODEV && cpu_list) { + if (err == EPERM || err == EACCES) { + ui__warning_paranoid(); + exit(EXIT_FAILURE); + } else if (err == ENODEV && cpu_list) { die("No such device - did you specify" " an out-of-range profile CPU?\n"); } else if (err == EINVAL && sample_id_all_avail) { diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 676b4fb0070f..935fc4fd878e 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -850,10 +850,10 @@ try_again: top.evlist->threads, group, inherit) < 0) { int err = errno; - if (err == EPERM || err == EACCES) - die("Permission error - are you root?\n" - "\t Consider tweaking" - " /proc/sys/kernel/perf_event_paranoid.\n"); + if (err == EPERM || err == EACCES) { + ui__warning_paranoid(); + goto out_err; + } /* * If it's cycles then fall back to hrtimer * based cpu-clock-tick sw counter, which @@ -861,25 +861,35 @@ try_again: */ if (attr->type == PERF_TYPE_HARDWARE && attr->config == PERF_COUNT_HW_CPU_CYCLES) { - if (verbose) - warning(" ... trying to fall back to cpu-clock-ticks\n"); + ui__warning("Cycles event not supported,\n" + "trying to fall back to cpu-clock-ticks\n"); attr->type = PERF_TYPE_SOFTWARE; attr->config = PERF_COUNT_SW_CPU_CLOCK; goto try_again; } - printf("\n"); - error("sys_perf_event_open() syscall returned with %d " - "(%s). /bin/dmesg may provide additional information.\n", - err, strerror(err)); - die("No CONFIG_PERF_EVENTS=y kernel support configured?\n"); - exit(-1); + + ui__warning("The sys_perf_event_open() syscall " + "returned with %d (%s). /bin/dmesg " + "may provide additional information.\n" + "No CONFIG_PERF_EVENTS=y kernel support " + "configured?\n", err, strerror(err)); + goto out_err; } } - if (perf_evlist__mmap(evlist, mmap_pages, false) < 0) - die("failed to mmap with %d (%s)\n", errno, strerror(errno)); + if (perf_evlist__mmap(evlist, mmap_pages, false) < 0) { + ui__warning("Failed to mmap with %d (%s)\n", + errno, strerror(errno)); + goto out_err; + } + + return; + +out_err: + exit_browser(0); + exit(0); } static int __cmd_top(void) diff --git a/tools/perf/util/debug.c b/tools/perf/util/debug.c index d4536a9e0d8c..155749d74350 100644 --- a/tools/perf/util/debug.c +++ b/tools/perf/util/debug.c @@ -57,6 +57,16 @@ void ui__warning(const char *format, ...) } #endif +void ui__warning_paranoid(void) +{ + ui__warning("Permission error - are you root?\n" + "Consider tweaking /proc/sys/kernel/perf_event_paranoid:\n" + " -1 - Not paranoid at all\n" + " 0 - Disallow raw tracepoint access for unpriv\n" + " 1 - Disallow cpu events for unpriv\n" + " 2 - Disallow kernel profiling for unpriv\n"); +} + void trace_event(union perf_event *event) { unsigned char *raw_event = (void *)event; diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h index 93516cf4682c..fd53db47e3de 100644 --- a/tools/perf/util/debug.h +++ b/tools/perf/util/debug.h @@ -36,5 +36,6 @@ int ui_helpline__show_help(const char *format, va_list ap); #endif void ui__warning(const char *format, ...) __attribute__((format(printf, 1, 2))); +void ui__warning_paranoid(void); #endif /* __PERF_DEBUG_H */ From ca6a42586fae639ff9e5285d9bdc550fcb1b8d41 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Fri, 25 Mar 2011 13:11:11 -0600 Subject: [PATCH 4/5] perf tools: Emit clearer message for sys_perf_event_open ENOENT return Resend of patch sent back in January 2011 in light of recent confusion around unsupported events for a given platform. Improve sys_perf_event_open ENOENT return handling in top and record, just like 5a3446b does for stat. Retry of Arnaldo's patch using ui_warning instead of die which allows the fallback from hardware cycles to software clock. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org LKML-Reference: <1301080271-20945-1-git-send-email-daahern@cisco.com> Signed-off-by: David Ahern [ committer note: Some adjustments to make it apply to newer codebase ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 10 +++++++++- tools/perf/builtin-top.c | 6 ++++++ 2 files changed, 15 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 623695e18254..db6adecf46f1 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -301,11 +301,19 @@ try_again: && attr->config == PERF_COUNT_HW_CPU_CYCLES) { if (verbose) - warning(" ... trying to fall back to cpu-clock-ticks\n"); + ui__warning("The cycles event is not supported, " + "trying to fall back to cpu-clock-ticks\n"); attr->type = PERF_TYPE_SOFTWARE; attr->config = PERF_COUNT_SW_CPU_CLOCK; goto try_again; } + + if (err == ENOENT) { + ui__warning("The %s event is not supported.\n", + event_name(pos)); + exit(EXIT_FAILURE); + } + printf("\n"); error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n", err, strerror(err)); diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 935fc4fd878e..fc1273e976c5 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -870,6 +870,12 @@ try_again: goto try_again; } + if (err == ENOENT) { + ui__warning("The %s event is not supported.\n", + event_name(counter)); + goto out_err; + } + ui__warning("The sys_perf_event_open() syscall " "returned with %d (%s). /bin/dmesg " "may provide additional information.\n" From cd25f8bc2696664877b21d33b7994e12fa570919 Mon Sep 17 00:00:00 2001 From: Lin Ming Date: Fri, 25 Mar 2011 16:27:48 +0800 Subject: [PATCH 5/5] perf probe: Add fastpath to do lookup by function name v3 -> v2: - Make pubname_search_cb more generic - Add fastpath to find_probes also v2 -> v1: - Don't compare file names with cu_find_realpath(...), instead, compare them with the name returned by dwarf_decl_file(sp_die) The vmlinux file may have thousands of CUs. We can lookup function name from .debug_pubnames section to avoid the slow loop on CUs. 1. Improvement data for find_line_range ./perf stat -e cycles -r 10 -- ./perf probe -k /home/mlin/vmlinux \ -s /home/mlin/linux-2.6 \ --line csum_partial_copy_to_user > tmp.log before patch applied ===================== 847,988,276 cycles 0.355075856 seconds time elapsed after patch applied ===================== 206,102,622 cycles 0.086883555 seconds time elapsed 2. Improvement data for find_probes ./perf stat -e cycles -r 10 -- ./perf probe -k /home/mlin/vmlinux \ -s /home/mlin/linux-2.6 \ --vars csum_partial_copy_to_user > tmp.log before patch applied ===================== 848,490,844 cycles 0.355307901 seconds time elapsed after patch applied ===================== 205,684,469 cycles 0.086694010 seconds time elapsed Acked-by: Masami Hiramatsu Cc: Ingo Molnar Cc: Masami Hiramatsu Cc: Peter Zijlstra Cc: linux-kernel LKML-Reference: <1301041668.14111.52.camel@minggr.sh.intel.com> Signed-off-by: Lin Ming Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-finder.c | 72 ++++++++++++++++++++++++++++++++++ tools/perf/util/probe-finder.h | 2 + 2 files changed, 74 insertions(+) diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 194f9e2a3285..ff416b85f7e8 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -1435,6 +1435,38 @@ static int find_probe_point_by_func(struct probe_finder *pf) return _param.retval; } +struct pubname_callback_param { + char *function; + char *file; + Dwarf_Die *cu_die; + Dwarf_Die *sp_die; + int found; +}; + +static int pubname_search_cb(Dwarf *dbg, Dwarf_Global *gl, void *data) +{ + struct pubname_callback_param *param = data; + + if (dwarf_offdie(dbg, gl->die_offset, param->sp_die)) { + if (dwarf_tag(param->sp_die) != DW_TAG_subprogram) + return DWARF_CB_OK; + + if (die_compare_name(param->sp_die, param->function)) { + if (!dwarf_offdie(dbg, gl->cu_offset, param->cu_die)) + return DWARF_CB_OK; + + if (param->file && + strtailcmp(param->file, dwarf_decl_file(param->sp_die))) + return DWARF_CB_OK; + + param->found = 1; + return DWARF_CB_ABORT; + } + } + + return DWARF_CB_OK; +} + /* Find probe points from debuginfo */ static int find_probes(int fd, struct probe_finder *pf) { @@ -1461,6 +1493,27 @@ static int find_probes(int fd, struct probe_finder *pf) off = 0; line_list__init(&pf->lcache); + + /* Fastpath: lookup by function name from .debug_pubnames section */ + if (pp->function) { + struct pubname_callback_param pubname_param = { + .function = pp->function, + .file = pp->file, + .cu_die = &pf->cu_die, + .sp_die = &pf->sp_die, + }; + struct dwarf_callback_param probe_param = { + .data = pf, + }; + + dwarf_getpubnames(dbg, pubname_search_cb, &pubname_param, 0); + if (pubname_param.found) { + ret = probe_point_search_cb(&pf->sp_die, &probe_param); + if (ret) + goto found; + } + } + /* Loop on CUs (Compilation Unit) */ while (!dwarf_nextcu(dbg, off, &noff, &cuhl, NULL, NULL, NULL)) { /* Get the DIE(Debugging Information Entry) of this CU */ @@ -1488,6 +1541,8 @@ static int find_probes(int fd, struct probe_finder *pf) } off = noff; } + +found: line_list__free(&pf->lcache); if (dwfl) dwfl_end(dwfl); @@ -1895,6 +1950,22 @@ int find_line_range(int fd, struct line_range *lr) return -EBADF; } + /* Fastpath: lookup by function name from .debug_pubnames section */ + if (lr->function) { + struct pubname_callback_param pubname_param = { + .function = lr->function, .file = lr->file, + .cu_die = &lf.cu_die, .sp_die = &lf.sp_die, .found = 0}; + struct dwarf_callback_param line_range_param = { + .data = (void *)&lf, .retval = 0}; + + dwarf_getpubnames(dbg, pubname_search_cb, &pubname_param, 0); + if (pubname_param.found) { + line_range_search_cb(&lf.sp_die, &line_range_param); + if (lf.found) + goto found; + } + } + /* Loop on CUs (Compilation Unit) */ while (!lf.found && ret >= 0) { if (dwarf_nextcu(dbg, off, &noff, &cuhl, NULL, NULL, NULL) != 0) @@ -1923,6 +1994,7 @@ int find_line_range(int fd, struct line_range *lr) off = noff; } +found: /* Store comp_dir */ if (lf.found) { comp_dir = cu_get_comp_dir(&lf.cu_die); diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h index beaefc3c1223..605730a366db 100644 --- a/tools/perf/util/probe-finder.h +++ b/tools/perf/util/probe-finder.h @@ -49,6 +49,7 @@ struct probe_finder { Dwarf_Addr addr; /* Address */ const char *fname; /* Real file name */ Dwarf_Die cu_die; /* Current CU */ + Dwarf_Die sp_die; struct list_head lcache; /* Line cache for lazy match */ /* For variable searching */ @@ -83,6 +84,7 @@ struct line_finder { int lno_s; /* Start line number */ int lno_e; /* End line number */ Dwarf_Die cu_die; /* Current CU */ + Dwarf_Die sp_die; int found; };