From 267b677eec06de178558337f0074e30ec3d11d9b Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 19 Jan 2015 11:54:06 -0300 Subject: perf evlist: Remove extraneous 'was' on error message [acme@mica ~]$ trace -p 3330 Error: Unable to find debugfs Hint: Was your kernel was compiled with debugfs support? ^^^ ^^^ Hint: Is the debugfs filesystem mounted? Hint: Try 'sudo mount -t debugfs nodev /sys/kernel/debug' Fix it. Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Don Zickus Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-kb9s0xy5z8i51abdu4bgm3rv@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index cbab1fb..2e507b5 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -1445,7 +1445,7 @@ int perf_evlist__strerror_tp(struct perf_evlist *evlist __maybe_unused, case ENOENT: scnprintf(buf, size, "%s", "Error:\tUnable to find debugfs\n" - "Hint:\tWas your kernel was compiled with debugfs support?\n" + "Hint:\tWas your kernel compiled with debugfs support?\n" "Hint:\tIs the debugfs filesystem mounted?\n" "Hint:\tTry 'sudo mount -t debugfs nodev /sys/kernel/debug'"); break; -- cgit v1.1 From 75e0b5f010837f29b5773653a248852a7317562b Mon Sep 17 00:00:00 2001 From: Wang Nan Date: Mon, 19 Jan 2015 20:43:17 +0800 Subject: perf scripting perl: Force to use stdbool When building perf for arm64 I hit a warning (and be treated as an error) like below: aarch64-oe-linux-gcc -o .../scripts/perl/Perf-Trace-Util/Context.o -c -Wbad-function-cast \ ... scripts/perl/Perf-Trace-Util/Context.c In file included from .../usr/lib64/perl/5.14.3/CORE/perl.h:2464:0, from Context.xs:23: /.../usr/lib64/perl/5.14.3/CORE/handy.h:108:0: error: "bool" redefined [-Werror] # define bool char ^ In file included from /.../usr/src/kernel/tools/include/linux/types.h:4:0, from /.../usr/src/kernel/arch/arm64/include/uapi/asm/sigcontext.h:19, from /.../usr/include/bits/sigcontext.h:27, from /.../usr/include/signal.h:340, from /.../usr/include/sys/param.h:28, from /.../usr/lib64/perl/5.14.3/CORE/perl.h:678, from Context.xs:23: /.../usr/lib/aarch64-oe-linux/gcc/aarch64-oe-linux/4.9.2/include/stdbool.h:33:0: note: this is the location of the previous definition #define bool _Bool Looks like the failure is caused by arm64 uapi/asm/sigcontext.h, which includes linux/types.h while other archs not. Current perl consider this problem: http://perl5.git.perl.org/perl.git/commit/bd31be4baa3ee68abdb92c0db3200efe0fad903b However there are users which use old version of perl. This patch includes stdbool.h before Context.xs and define HAS_BOOL to prevent perl'e headers define its own 'bool'. Code is learn from perl's git tree. Signed-off-by: Wang Nan Acked-by: Jiri Olsa Acked-by: Namhyung Kim Cc: Li Zefan Link: http://lkml.kernel.org/r/1421671397-4659-1-git-send-email-wangnan0@huawei.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/scripts/perl/Perf-Trace-Util/Context.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c index 790ceba..28431d1 100644 --- a/tools/perf/scripts/perl/Perf-Trace-Util/Context.c +++ b/tools/perf/scripts/perl/Perf-Trace-Util/Context.c @@ -5,7 +5,10 @@ * ANY CHANGES MADE HERE WILL BE LOST! * */ - +#include +#ifndef HAS_BOOL +# define HAS_BOOL 1 +#endif #line 1 "Context.xs" /* * Context.xs. XS interfaces for perf script. -- cgit v1.1 From be81908c2289f405df75d2511ccf5da785945400 Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Sun, 18 Jan 2015 20:00:20 +0100 Subject: perf annotate: Handle ins parsing failures Don't use the ins's ->sncprintf() if the parsing failed. For example, this fixes the display of "imul %edx". Without this patch: | imul (null),(null) After this patch: | imul %edx Signed-off-by: Rabin Vincent Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1421607621-15005-1-git-send-email-rabin@rab.in Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 79999ce..d5da1b8 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -183,8 +183,9 @@ static int lock__parse(struct ins_operands *ops) if (!ops->locked.ins->ops) return 0; - if (ops->locked.ins->ops->parse) - ops->locked.ins->ops->parse(ops->locked.ops); + if (ops->locked.ins->ops->parse && + ops->locked.ins->ops->parse(ops->locked.ops) < 0) + goto out_free_ops; return 0; @@ -531,8 +532,8 @@ static void disasm_line__init_ins(struct disasm_line *dl) if (!dl->ins->ops) return; - if (dl->ins->ops->parse) - dl->ins->ops->parse(&dl->ops); + if (dl->ins->ops->parse && dl->ins->ops->parse(&dl->ops) < 0) + dl->ins = NULL; } static int disasm_line__parse(char *line, char **namep, char **rawp) -- cgit v1.1 From 0fb9f2aab738eec9dd9b929ed7d37bf744d2ac77 Mon Sep 17 00:00:00 2001 From: Rabin Vincent Date: Sun, 18 Jan 2015 20:00:21 +0100 Subject: perf annotate: Fix memory leaks in LOCK handling The lock prefix handling fails to free the strdup()'d name as well as the fields allocated by the instruction parsing. Signed-off-by: Rabin Vincent Cc: Paul Mackerras Link: http://lkml.kernel.org/r/1421607621-15005-2-git-send-email-rabin@rab.in Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index d5da1b8..01bc4e2 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -177,6 +177,8 @@ static int lock__parse(struct ins_operands *ops) goto out_free_ops; ops->locked.ins = ins__find(name); + free(name); + if (ops->locked.ins == NULL) goto out_free_ops; @@ -209,6 +211,13 @@ static int lock__scnprintf(struct ins *ins, char *bf, size_t size, static void lock__delete(struct ins_operands *ops) { + struct ins *ins = ops->locked.ins; + + if (ins && ins->ops->free) + ins->ops->free(ops->locked.ops); + else + ins__delete(ops->locked.ops); + zfree(&ops->locked.ops); zfree(&ops->target.raw); zfree(&ops->target.name); -- cgit v1.1 From de4809999dac44c51e1f9ad892deb0336296dc4e Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Fri, 16 Jan 2015 15:31:28 -0300 Subject: perf symbols: Return the first entry with a given name in find_by_name method When a dso contains multiple symbols which have same name, current dso__find_symbol_by_name() only finds an one of them and there's no way to get the all symbols without going through the rbtree. So make symbols__find_by_name() return the first entry with the given name and the next patch in this series will provide a way to iterate from there, by the name ordered rb_tree, till a suitable symbol is found. Signed-off-by: Namhyung Kim Cc: David Ahern Cc: Jiri Olsa Cc: Masami Hiramatsu Link: http://lkml.kernel.org/r/1421234288-22758-2-git-send-email-namhyung@kernel.org [ Yanked this independent hunk, without changes, from a larger patch ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index c24c5b8..3cb928e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -396,6 +396,7 @@ static struct symbol *symbols__find_by_name(struct rb_root *symbols, const char *name) { struct rb_node *n; + struct symbol_name_rb_node *s; if (symbols == NULL) return NULL; @@ -403,7 +404,6 @@ static struct symbol *symbols__find_by_name(struct rb_root *symbols, n = symbols->rb_node; while (n) { - struct symbol_name_rb_node *s; int cmp; s = rb_entry(n, struct symbol_name_rb_node, rb_node); @@ -414,10 +414,24 @@ static struct symbol *symbols__find_by_name(struct rb_root *symbols, else if (cmp > 0) n = n->rb_right; else - return &s->sym; + break; } - return NULL; + if (n == NULL) + return NULL; + + /* return first symbol that has same name (if any) */ + for (n = rb_prev(n); n; n = rb_prev(n)) { + struct symbol_name_rb_node *tmp; + + tmp = rb_entry(n, struct symbol_name_rb_node, rb_node); + if (strcmp(tmp->sym.name, s->sym.name)) + break; + + s = tmp; + } + + return &s->sym; } struct symbol *dso__find_symbol(struct dso *dso, -- cgit v1.1 From 18bd726418f2a2e12f85104f0f4b5d5d077e1752 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 16 Jan 2015 15:39:53 -0300 Subject: perf symbols: Introduce method to iterate symbols ordered by name Given a symbol, go to the next entry in a rbtree sorted by symbol name. Cc: David Ahern Cc: Jiri Olsa Cc: Masami Hiramatsu Link: http://lkml.kernel.org/n/tip-aq210drxprnu2so4dye5xa3j@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 11 +++++++++++ tools/perf/util/symbol.h | 1 + 2 files changed, 12 insertions(+) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 3cb928e..a194702 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -450,6 +450,17 @@ struct symbol *dso__next_symbol(struct symbol *sym) return symbols__next(sym); } +struct symbol *symbol__next_by_name(struct symbol *sym) +{ + struct symbol_name_rb_node *s = container_of(sym, struct symbol_name_rb_node, sym); + struct rb_node *n = rb_next(&s->rb_node); + + return n ? &rb_entry(n, struct symbol_name_rb_node, rb_node)->sym : NULL; +} + + /* + * Teturns first symbol that matched with @name. + */ struct symbol *dso__find_symbol_by_name(struct dso *dso, enum map_type type, const char *name) { diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h index 9d602e9..1650dcb 100644 --- a/tools/perf/util/symbol.h +++ b/tools/perf/util/symbol.h @@ -231,6 +231,7 @@ struct symbol *dso__find_symbol(struct dso *dso, enum map_type type, u64 addr); struct symbol *dso__find_symbol_by_name(struct dso *dso, enum map_type type, const char *name); +struct symbol *symbol__next_by_name(struct symbol *sym); struct symbol *dso__first_symbol(struct dso *dso, enum map_type type); struct symbol *dso__next_symbol(struct symbol *sym); -- cgit v1.1 From 564c62a4d7f9a4b60920ee0b02cfe890471d87f4 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 14 Jan 2015 20:18:07 +0900 Subject: perf probe: Do not rely on map__load() filter to find symbols The find_probe_trace_events_from_map() searches matching symbol from a map (so from a backing dso). For uprobes, it'll create a new map (and dso) and loads it using a filter. It's a little bit inefficient in that it'll read out the symbol table everytime but works well anyway. For kprobes however, it'll reuse existing kernel map which might be loaded before. In this case map__load() just returns with no result. It makes kprobes always failed to find symbol even if it exists in the map (dso). To fix it, use map__find_symbol_by_name() instead. It'll load a map with full symbols and sorts them by name. It needs to search sibing nodes since there can be multiple (local) symbols with same name. Signed-off-by: Namhyung Kim Cc: David Ahern Cc: Jiri Olsa Cc: Masami Hiramatsu Link: http://lkml.kernel.org/r/1421234288-22758-3-git-send-email-namhyung@kernel.org [ Use symbol__next_by_name ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-event.c | 38 ++++++++++++++++++++++---------------- 1 file changed, 22 insertions(+), 16 deletions(-) diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 94a717b..b24482e 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -2193,18 +2193,20 @@ static int __add_probe_trace_events(struct perf_probe_event *pev, return ret; } -static char *looking_function_name; -static int num_matched_functions; - -static int probe_function_filter(struct map *map __maybe_unused, - struct symbol *sym) +static int find_probe_functions(struct map *map, char *name) { - if ((sym->binding == STB_GLOBAL || sym->binding == STB_LOCAL) && - strcmp(looking_function_name, sym->name) == 0) { - num_matched_functions++; - return 0; + int found = 0; + struct symbol *sym = map__find_symbol_by_name(map, name, NULL); + + while (sym != NULL) { + if (sym->binding == STB_GLOBAL || sym->binding == STB_LOCAL) + found++; + sym = symbol__next_by_name(sym); + if (sym == NULL || strcmp(sym->name, name)) + break; } - return 1; + + return found; } #define strdup_or_goto(str, label) \ @@ -2222,10 +2224,10 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, struct kmap *kmap = NULL; struct ref_reloc_sym *reloc_sym = NULL; struct symbol *sym; - struct rb_node *nd; struct probe_trace_event *tev; struct perf_probe_point *pp = &pev->point; struct probe_trace_point *tp; + int num_matched_functions; int ret, i; /* Init maps of given executable or kernel */ @@ -2242,10 +2244,8 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, * Load matched symbols: Since the different local symbols may have * same name but different addresses, this lists all the symbols. */ - num_matched_functions = 0; - looking_function_name = pp->function; - ret = map__load(map, probe_function_filter); - if (ret || num_matched_functions == 0) { + num_matched_functions = find_probe_functions(map, pp->function); + if (num_matched_functions == 0) { pr_err("Failed to find symbol %s in %s\n", pp->function, target ? : "kernel"); ret = -ENOENT; @@ -2275,7 +2275,9 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, } ret = 0; - map__for_each_symbol(map, sym, nd) { + sym = map__find_symbol_by_name(map, pp->function, NULL); + + while (sym != NULL) { tev = (*tevs) + ret; tp = &tev->point; if (ret == num_matched_functions) { @@ -2323,6 +2325,10 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, strdup_or_goto(pev->args[i].type, nomem_out); } + + sym = symbol__next_by_name(sym); + if (sym == NULL || strcmp(sym->name, pp->function)) + break; } out: -- cgit v1.1 From 0a3873a8e2709ebc759361e2d872d9f71541806c Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 16 Jan 2015 16:40:25 -0300 Subject: perf symbols: Introduce 'for' method to iterate over the symbols with a given name Removing boilerplate from two places, where one would have to find the first entry, then iterate using symbol__next_by_name + strcmp to see if the next member had the same name. Cc: David Ahern Cc: Jiri Olsa Cc: Masami Hiramatsu Link: http://lkml.kernel.org/n/tip-eh73z8gthv20yowirmx2yk38@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/map.h | 16 ++++++++++++++++ tools/perf/util/probe-event.c | 14 +++----------- 2 files changed, 19 insertions(+), 11 deletions(-) diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h index 6951a9d..0e42438 100644 --- a/tools/perf/util/map.h +++ b/tools/perf/util/map.h @@ -116,6 +116,22 @@ struct thread; #define map__for_each_symbol(map, pos, n) \ dso__for_each_symbol(map->dso, pos, n, map->type) +/* map__for_each_symbol_with_name - iterate over the symbols in the given map + * that have the given name + * + * @map: the 'struct map *' in which symbols itereated + * @sym_name: the symbol name + * @pos: the 'struct symbol *' to use as a loop cursor + * @filter: to use when loading the DSO + */ +#define __map__for_each_symbol_by_name(map, sym_name, pos, filter) \ + for (pos = map__find_symbol_by_name(map, sym_name, filter); \ + pos && strcmp(pos->name, sym_name) == 0; \ + pos = symbol__next_by_name(pos)) + +#define map__for_each_symbol_by_name(map, sym_name, pos) \ + __map__for_each_symbol_by_name(map, sym_name, (pos), NULL) + typedef int (*symbol_filter_t)(struct map *map, struct symbol *sym); void map__init(struct map *map, enum map_type type, diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index b24482e..7cc89b1 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -2196,14 +2196,11 @@ static int __add_probe_trace_events(struct perf_probe_event *pev, static int find_probe_functions(struct map *map, char *name) { int found = 0; - struct symbol *sym = map__find_symbol_by_name(map, name, NULL); + struct symbol *sym; - while (sym != NULL) { + map__for_each_symbol_by_name(map, name, sym) { if (sym->binding == STB_GLOBAL || sym->binding == STB_LOCAL) found++; - sym = symbol__next_by_name(sym); - if (sym == NULL || strcmp(sym->name, name)) - break; } return found; @@ -2275,9 +2272,8 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, } ret = 0; - sym = map__find_symbol_by_name(map, pp->function, NULL); - while (sym != NULL) { + map__for_each_symbol_by_name(map, pp->function, sym) { tev = (*tevs) + ret; tp = &tev->point; if (ret == num_matched_functions) { @@ -2325,10 +2321,6 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, strdup_or_goto(pev->args[i].type, nomem_out); } - - sym = symbol__next_by_name(sym); - if (sym == NULL || strcmp(sym->name, pp->function)) - break; } out: -- cgit v1.1 From 25dd9171f51c482eb7c4dc8618766ae733756e2d Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Wed, 14 Jan 2015 20:18:08 +0900 Subject: perf probe: Fix probing kretprobes The commit dfef99cd0b2c ("perf probe: Use ref_reloc_sym based address instead of the symbol name") converts kprobes to use ref_reloc_sym (i.e. _stext) and offset instead of using symbol's name directly. So on my system, adding do_fork ends up with like below: $ sudo perf probe -v --add do_fork%return probe-definition(0): do_fork%return symbol:do_fork file:(null) line:0 offset:0 return:1 lazy:(null) 0 arguments Looking at the vmlinux_path (7 entries long) Using /lib/modules/3.17.6-1-ARCH/build/vmlinux for symbols Could not open debuginfo. Try to use symbols. Opening /sys/kernel/debug/tracing/kprobe_events write=1 Added new event: Writing event: r:probe/do_fork _stext+456136 Failed to write event: Invalid argument Error: Failed to add events. Reason: Operation not permitted (Code: -1) As you can see, the do_fork was translated to _stext+456136. This was because to support (local) symbols that have same name. But the problem is that kretprobe requires to be inserted at function start point so it simply checks whether it's called with offset 0. And if not, it'll return with -EINVAL. You can see it with dmesg. $ dmesg | tail -1 [125621.764103] Return probe must be used without offset. So we need to use the symbol name instead of ref_reloc_sym in case of return probes. Reported-by: Jiri Olsa Signed-off-by: Namhyung Kim Tested-by: Jiri Olsa Acked-by: Masami Hiramatsu Cc: David Ahern Link: http://lkml.kernel.org/r/1421234288-22758-4-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-event.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index 7cc89b1..919937e 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -446,7 +446,7 @@ static int post_process_probe_trace_events(struct probe_trace_event *tevs, } for (i = 0; i < ntevs; i++) { - if (tevs[i].point.address) { + if (tevs[i].point.address && !tevs[i].point.retprobe) { tmp = strdup(reloc_sym->name); if (!tmp) return -ENOMEM; @@ -2254,7 +2254,7 @@ static int find_probe_trace_events_from_map(struct perf_probe_event *pev, goto out; } - if (!pev->uprobes) { + if (!pev->uprobes && !pp->retprobe) { kmap = map__kmap(map); reloc_sym = kmap->ref_reloc_sym; if (!reloc_sym) { -- cgit v1.1 From c05199e5a57a579fea1e8fa65e2b511ceb524ffc Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Tue, 20 Jan 2015 04:54:25 +0000 Subject: perf/x86/intel/uncore: Move uncore_box_init() out of driver initialization There were some issues about the uncore driver tried to access non-existing boxes, which caused boot crashes. These issues have been all fixed. But we should avoid boot failures if that ever happens again. This patch intends to prevent this kind of potential issues. It moves uncore_box_init out of driver initialization. The box will be initialized when it's first enabled. Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Link: http://lkml.kernel.org/r/1421729665-5912-1-git-send-email-kan.liang@intel.com Cc: Arnaldo Carvalho de Melo Cc: Stephane Eranian Cc: Yan, Zheng Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel_uncore.c | 9 ++------- arch/x86/kernel/cpu/perf_event_intel_uncore.h | 18 ++++++++++-------- 2 files changed, 12 insertions(+), 15 deletions(-) diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.c b/arch/x86/kernel/cpu/perf_event_intel_uncore.c index 10b8d3e..c635b8b 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_uncore.c +++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.c @@ -840,7 +840,6 @@ static int uncore_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id box->phys_id = phys_id; box->pci_dev = pdev; box->pmu = pmu; - uncore_box_init(box); pci_set_drvdata(pdev, box); raw_spin_lock(&uncore_box_lock); @@ -1004,10 +1003,8 @@ static int uncore_cpu_starting(int cpu) pmu = &type->pmus[j]; box = *per_cpu_ptr(pmu->box, cpu); /* called by uncore_cpu_init? */ - if (box && box->phys_id >= 0) { - uncore_box_init(box); + if (box && box->phys_id >= 0) continue; - } for_each_online_cpu(k) { exist = *per_cpu_ptr(pmu->box, k); @@ -1023,10 +1020,8 @@ static int uncore_cpu_starting(int cpu) } } - if (box) { + if (box) box->phys_id = phys_id; - uncore_box_init(box); - } } } return 0; diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.h b/arch/x86/kernel/cpu/perf_event_intel_uncore.h index 863d9b0..6c8c1e7 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_uncore.h +++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.h @@ -257,6 +257,14 @@ static inline int uncore_num_counters(struct intel_uncore_box *box) return box->pmu->type->num_counters; } +static inline void uncore_box_init(struct intel_uncore_box *box) +{ + if (!test_and_set_bit(UNCORE_BOX_FLAG_INITIATED, &box->flags)) { + if (box->pmu->type->ops->init_box) + box->pmu->type->ops->init_box(box); + } +} + static inline void uncore_disable_box(struct intel_uncore_box *box) { if (box->pmu->type->ops->disable_box) @@ -265,6 +273,8 @@ static inline void uncore_disable_box(struct intel_uncore_box *box) static inline void uncore_enable_box(struct intel_uncore_box *box) { + uncore_box_init(box); + if (box->pmu->type->ops->enable_box) box->pmu->type->ops->enable_box(box); } @@ -287,14 +297,6 @@ static inline u64 uncore_read_counter(struct intel_uncore_box *box, return box->pmu->type->ops->read_counter(box, event); } -static inline void uncore_box_init(struct intel_uncore_box *box) -{ - if (!test_and_set_bit(UNCORE_BOX_FLAG_INITIATED, &box->flags)) { - if (box->pmu->type->ops->init_box) - box->pmu->type->ops->init_box(box); - } -} - static inline bool uncore_box_is_fake(struct intel_uncore_box *box) { return (box->phys_id < 0); -- cgit v1.1 From 98b008dff8452653909d9263efda925873e8d8bb Mon Sep 17 00:00:00 2001 From: Stephane Eranian Date: Thu, 22 Jan 2015 21:38:34 +0100 Subject: perf/rapl: Fix crash in rapl_scale() This patch fixes a systematic crash in rapl_scale() due to an invalid pointer. The bug was introduced by commit: 89cbc76768c2 ("x86: Replace __get_cpu_var uses") The fix is simple. Just put the parenthesis where it needs to be, i.e., around rapl_pmu. To my surprise, the compiler was not complaining about passing an integer instead of a pointer. Reported-by: Vince Weaver Tested-by: Vince Weaver Fixes: 89cbc76768c2 ("x86: Replace __get_cpu_var uses") Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: cl@linux.com Cc: Linus Torvalds Link: http://lkml.kernel.org/r/20150122203834.GA10228@thinkpad Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel_rapl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/kernel/cpu/perf_event_intel_rapl.c b/arch/x86/kernel/cpu/perf_event_intel_rapl.c index 6e434f8..c4bb8b8 100644 --- a/arch/x86/kernel/cpu/perf_event_intel_rapl.c +++ b/arch/x86/kernel/cpu/perf_event_intel_rapl.c @@ -142,7 +142,7 @@ static inline u64 rapl_scale(u64 v) * or use ldexp(count, -32). * Watts = Joules/Time delta */ - return v << (32 - __this_cpu_read(rapl_pmu->hw_unit)); + return v << (32 - __this_cpu_read(rapl_pmu)->hw_unit); } static u64 rapl_event_update(struct perf_event *event) -- cgit v1.1 From ef454caeb740ee4e1b89aeb7f7692d5ddffb6830 Mon Sep 17 00:00:00 2001 From: Kan Liang Date: Thu, 22 Jan 2015 07:50:53 +0000 Subject: perf/x86/intel: Add model number for Airmont Intel Airmont supports the same architectural and non-architectural performance monitoring events as Silvermont. Signed-off-by: Kan Liang Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: Linus Torvalds Link: http://lkml.kernel.org/r/1421913053-99803-1-git-send-email-kan.liang@intel.com Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_event_intel.c | 1 + 1 file changed, 1 insertion(+) diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c index 944bf01..498b6d9 100644 --- a/arch/x86/kernel/cpu/perf_event_intel.c +++ b/arch/x86/kernel/cpu/perf_event_intel.c @@ -2431,6 +2431,7 @@ __init int intel_pmu_init(void) break; case 55: /* 22nm Atom "Silvermont" */ + case 76: /* 14nm Atom "Airmont" */ case 77: /* 22nm Atom "Silvermont Avoton/Rangely" */ memcpy(hw_cache_event_ids, slm_hw_cache_event_ids, sizeof(hw_cache_event_ids)); -- cgit v1.1 From c3c87e770458aa004bd7ed3f29945ff436fd6511 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 23 Jan 2015 11:19:48 +0100 Subject: perf: Tighten (and fix) the grouping condition The fix from 9fc81d87420d ("perf: Fix events installation during moving group") was incomplete in that it failed to recognise that creating a group with events for different CPUs is semantically broken -- they cannot be co-scheduled. Furthermore, it leads to real breakage where, when we create an event for CPU Y and then migrate it to form a group on CPU X, the code gets confused where the counter is programmed -- triggered in practice as well by me via the perf fuzzer. Fix this by tightening the rules for creating groups. Only allow grouping of counters that can be co-scheduled in the same context. This means for the same task and/or the same cpu. Fixes: 9fc81d87420d ("perf: Fix events installation during moving group") Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Link: http://lkml.kernel.org/r/20150123125834.090683288@infradead.org Signed-off-by: Ingo Molnar --- include/linux/perf_event.h | 6 ------ kernel/events/core.c | 15 +++++++++++++-- 2 files changed, 13 insertions(+), 8 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 4f7a61c..664de5a 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -450,11 +450,6 @@ struct perf_event { #endif /* CONFIG_PERF_EVENTS */ }; -enum perf_event_context_type { - task_context, - cpu_context, -}; - /** * struct perf_event_context - event context structure * @@ -462,7 +457,6 @@ enum perf_event_context_type { */ struct perf_event_context { struct pmu *pmu; - enum perf_event_context_type type; /* * Protect the states of the events in the list, * nr_active, and the list: diff --git a/kernel/events/core.c b/kernel/events/core.c index 882f835..19efcf1 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -6776,7 +6776,6 @@ skip_type: __perf_event_init_context(&cpuctx->ctx); lockdep_set_class(&cpuctx->ctx.mutex, &cpuctx_mutex); lockdep_set_class(&cpuctx->ctx.lock, &cpuctx_lock); - cpuctx->ctx.type = cpu_context; cpuctx->ctx.pmu = pmu; __perf_cpu_hrtimer_init(cpuctx, cpu); @@ -7420,7 +7419,19 @@ SYSCALL_DEFINE5(perf_event_open, * task or CPU context: */ if (move_group) { - if (group_leader->ctx->type != ctx->type) + /* + * Make sure we're both on the same task, or both + * per-cpu events. + */ + if (group_leader->ctx->task != ctx->task) + goto err_context; + + /* + * Make sure we're both events for the same CPU; + * grouping events for different CPUs is broken; since + * you can never concurrently schedule them anyhow. + */ + if (group_leader->cpu != event->cpu) goto err_context; } else { if (group_leader->ctx != ctx) -- cgit v1.1