perf: Implement lbr-as-callgraph v2
This patchkit implements lbr-as-callgraphs in per freport, as an alternative way to present LBR information.
Current perf report does a histogram over the branch edges, which is useful to look at basic blocks, but doesn't tell you anything about the larger control flow.
This patchkit adds a new option --branch-history that adds the branch paths to the callgraph history instead.
This allows to reason about individual branch paths leading to specific samples.
Updates to v1:
- rebased on perf/core
- fix various issues
- rename the option to --branch-history
- various fixes to display the information more concise
Example output:
% perf record -b -g ./tsrc/tcall
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.044 MB perf.data (~1923 samples) ] % perf report --branch-history
...
54.91% tcall.c:6 [.] f2 tcall
|
|--66.53%-- f2 tcall.c:5
| |
| |--70.83%-- f1 tcall.c:11
| | f1 tcall.c:10
| | main tcall.c:18
| | main tcall.c:18
| | main tcall.c:17
| | main tcall.c:17
| | f1 tcall.c:13
| | f1 tcall.c:13
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:12
| | f1 tcall.c:12
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:11
PATCH 1/9 perf, tools: fix BFD detection on opensuse
From: Andi Kleen <[email protected]>
opensuse libbfd requires -lz -liberty to build. Add those to the BFD feature detection.
--- tools/perf/config/Makefile | 2 +- tools/perf/config/feature-checks/Makefile | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
tools/perf/config/Makefile +1 -1
@@ -478,7 +478,7 @@ else endif ifeq ($(feature-libbfd), 1) - EXTLIBS += -lbfd + EXTLIBS += -lbfd -lz -liberty endif ifdef NO_DEMANGLE
tools/perf/config/feature-checks/Makefile +1 -1
@@ -120,7 +120,7 @@ test-libpython-version.bin: $(BUILD) $(FLAGS_PYTHON_EMBED) test-libbfd.bin: - $(BUILD) -DPACKAGE='"perf"' -lbfd -ldl + $(BUILD) -DPACKAGE='"perf"' -lbfd -lz -liberty -ldl test-liberty.bin: $(CC) -o $(OUTPUT)$@ test-libbfd.c -DPACKAGE='"perf"' -lbfd -ldl -liberty
Signature · 1 line
1.8.3.1
Re: [PATCH 1/9] perf, tools: fix BFD detection on opensuse
Quoted 16 lines · Andi Kleen
From: Andi Kleen <[email protected]>
opensuse libbfd requires -lz -liberty to build. Add those to the BFD feature detection.
Signed-off-by Andi Kleen <[email protected]>--- tools/perf/config/Makefile | 2 +- tools/perf/config/feature-checks/Makefile | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)tools/perf/config/Makefile +0
-0
tools/perf/config/Makefile else · line 482
Quoted 6 lines
tools/perf/config/Makefile +1
-1@@ -478,7 +478,7 @@ else endif ifeq ($(feature-libbfd), 1) - EXTLIBS += -lbfd + EXTLIBS += -lbfd -lz -liberty endif
ok, Fedora is using linker script in place of libbfd.so which adds those anyway..
---
[jolsa@krava perf]$ cat /usr/lib64/libbfd.so
/* GNU ld script */
/* Ensure this .so library will not be used by a link for a different format
on a multi-architecture system. */OUTPUT_FORMAT(elf64-x86-64)
/* The libz dependency is unexpected by legacy build scripts. */ /* The libdl dependency is for plugin support. (BZ 889134) */ INPUT ( /usr/lib64/libbfd.a -liberty -lz -ldl )
---
we also need to check and probably get rid of follow up settings of EXTLIBS which seems useless now:
...
ifneq ($(feature-libbfd), 1)
$(call feature_check,liberty)
ifeq ($(feature-liberty), 1)
EXTLIBS += -lbfd -liberty
else
$(call feature_check,liberty-z)
ifeq ($(feature-liberty-z), 1)
EXTLIBS += -lbfd -liberty -lz...
jirka
Namhyung Kim
maintainerRe: [PATCH 1/9] perf, tools: fix BFD detection on opensuse
Quoted 52 lines · Jiri Olsa
Quoted 16 lines · Andi Kleen
From: Andi Kleen <[email protected]>
opensuse libbfd requires -lz -liberty to build. Add those to the BFD feature detection.
Signed-off-by Andi Kleen <[email protected]>--- tools/perf/config/Makefile | 2 +- tools/perf/config/feature-checks/Makefile | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)tools/perf/config/Makefile +0
-0tools/perf/config/Makefile else · line 482
Quoted 6 lines
tools/perf/config/Makefile +1
-1@@ -478,7 +478,7 @@ else endif ifeq ($(feature-libbfd), 1) - EXTLIBS += -lbfd + EXTLIBS += -lbfd -lz -liberty endifok, Fedora is using linker script in place of libbfd.so which adds those anyway..
---
[jolsa@krava perf]$ cat/usr/lib64/libbfd.so
/* GNU ld script *//* Ensure this .so library will not be used by a link for a different format
on a multi-architecture system. */OUTPUT_FORMAT(elf64-x86-64)/* The libz dependency is unexpected by legacy build scripts. */ /* The libdl dependency is for plugin support. (BZ 889134) */ INPUT (
/usr/lib64/libbfd.a-liberty -lz -ldl )
---we also need to check and probably get rid of follow up settings of EXTLIBS which seems useless now:
...
ifneq ($(feature-libbfd), 1) $(call feature_check,liberty) ifeq ($(feature-liberty), 1) EXTLIBS += -lbfd -liberty else $(call feature_check,liberty-z) ifeq ($(feature-liberty-z), 1) EXTLIBS += -lbfd -liberty -lz
Quoted 1 line
...
Agreed. I think it's only for keeping dependency minimal. But no need to do it if it's always called with -liberty and -lz.
Thanks,
Namhyung
Re: [PATCH 1/9] perf, tools: fix BFD detection on opensuse
In reply to Andi KleenQuoted 5 lines · Andi Kleen
From: Andi Kleen<[email protected]>
opensuse libbfd requires -lz -liberty to build. Add those to the BFD feature detection.
Fixes static builds on Fedora as well.
David
tip-bot for Andi Kleen
merge bot[tip:perf/urgent] perf tools: fix BFD detection on opensuse
In reply to Andi Kleenperf tools: fix BFD detection on opensuse
Commit-ID: 280e7c48c3b873e4987a63da276ecab25383f494 Gitweb: http://git.kernel.org/tip/280e7c48c3b873e4987a63da276ecab25383f494 Author: Andi Kleen <[email protected]> AuthorDate: Sat, 11 Jan 2014 11:42:51 -0800 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Thu, 27 Feb 2014 18:29:08 -0300
perf tools: fix BFD detection on opensuse
opensuse libbfd requires -lz -liberty to build. Add those to the BFD feature detection.
--- tools/perf/config/Makefile | 2 +- tools/perf/config/feature-checks/Makefile | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
tools/perf/config/Makefile +1 -1
@@ -478,7 +478,7 @@ else endif ifeq ($(feature-libbfd), 1) - EXTLIBS += -lbfd + EXTLIBS += -lbfd -lz -liberty endif ifdef NO_DEMANGLE
tools/perf/config/feature-checks/Makefile +1 -1
@@ -121,7 +121,7 @@ test-libpython-version.bin: $(BUILD) $(FLAGS_PYTHON_EMBED) test-libbfd.bin: - $(BUILD) -DPACKAGE='"perf"' -lbfd -ldl + $(BUILD) -DPACKAGE='"perf"' -lbfd -lz -liberty -ldl test-liberty.bin: $(CC) -o $(OUTPUT)$@ test-libbfd.c -DPACKAGE='"perf"' -lbfd -ldl -liberty
PATCH 2/9 perf, tools: Support handling complete branch stacks as histograms
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Currently branch stacks can be only shown as edge histograms for individual branches. I never found this display particularly useful.
This implements an alternative mode that creates histograms over complete branch traces, instead of individual branches, similar to how normal callgraphs are handled. This is done by putting it in
front of the normal callgraph and then using the normal callgraph histogram infrastructure to unify them.
This way in complex functions we can understand the control flow that lead to a particular sample, and may even see some control flow in the caller for short functions.
Example (simplified, of course for such simple code this is usually not needed):
tcall.c:
volatile a = 10000, b = 100000, c;
__attribute__((noinline)) f2()
{
c = a / b;
}__attribute__((noinline)) f1()
{
f2();
f2();
}
main()
{
int i;
for (i = 0; i < 1000000; i++)
f1();
}% perf record -b -g ./tsrc/tcall
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.044 MB perf.data (~1923 samples) ] % perf report --branch-history
...
54.91% tcall.c:6 [.] f2 tcall
|
|--65.53%-- f2 tcall.c:5
| |
| |--70.83%-- f1 tcall.c:11
| | f1 tcall.c:10
| | main tcall.c:18
| | main tcall.c:18
| | main tcall.c:17
| | main tcall.c:17
| | f1 tcall.c:13
| | f1 tcall.c:13
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:12
| | f1 tcall.c:12
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:11
| |
| --29.17%-- f1 tcall.c:12
| f1 tcall.c:12
| f2 tcall.c:7
| f2 tcall.c:5
| f1 tcall.c:11
| f1 tcall.c:10
| main tcall.c:18
| main tcall.c:18
| main tcall.c:17
| main tcall.c:17
| f1 tcall.c:13
| f1 tcall.c:13
| f2 tcall.c:7
| f2 tcall.c:5
| f1 tcall.c:12The default output is unchanged.
This is only implemented in perf report, no change to record or anywhere else.
This adds the basic code to report:
- add a new "branch" option to the -g option parser to enable this mode
- when the flag is set include the LBR into the callstack in machine.c. The rest of the history code is unchanged and doesn't know the difference between LBR entry and normal call entry.
Current limitations:
- The LBR flags (mispredict etc.) are not shown in the history and LBR entries have no special marker.
- It would be nice if annotate marked the LBR entries somehow (e.g. with arrows)
v2: Various fixes.
--- tools/perf/builtin-report.c | 15 ++++-- tools/perf/util/callchain.h | 1 + tools/perf/util/machine.c | 113 ++++++++++++++++++++++++++++++++++++-------- tools/perf/util/symbol.h | 3 +- 4 files changed, 106 insertions(+), 26 deletions(-)
tools/perf/builtin-report.c +11 -4
@@ -658,7 +658,7 @@ parse_callchain_opt(const struct option *opt, const char *arg, int unset) callchain_param.order =ORDER_CALLER ; else if (!strncmp(tok2, "callee", strlen("callee"))) callchain_param.order =ORDER_CALLEE ; - else + else if (tok2[0] != 0) return -1;/* Get the sort key */
@@ -669,8 +669,15 @@ parse_callchain_opt(const struct option *opt, const char *arg, int unset) callchain_param.key =CCKEY_FUNCTION ; else if (!strncmp(tok2, "address", strlen("address"))) callchain_param.key =CCKEY_ADDRESS ; - else + else if (tok2[0] != 0) return -1; + + tok2 =strtok (NULL ,"," ); +if (!tok2) + goto setup; +if (!strncmp (tok2,"branch" ,6 )) + callchain_param.branch_callstack =1 ; + setup:if (callchain_register_param (&callchain_param) <0 ) {pr_err ("Can't register callchain params\n" );
@@ -786,8 +793,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused) "regex filter to identify parent, see: '--sort parent'"),OPT_BOOLEAN ('x' ,"exclude-other" , &symbol_conf.exclude_other,"Only display entries with parent-match" ), -OPT_CALLBACK_DEFAULT ('g' ,"call-graph" , &report,"output_type,min_percent[,print_limit],call_order" , -"Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold, optional print limit, callchain order, key (function or address). " +OPT_CALLBACK_DEFAULT ('g' ,"call-graph" , &report,"output_type,min_percent[,print_limit],call_order[,branch]" , +"Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold, optional print limit, callchain order, key (function or address), add branches. " "Default: fractal,0.5,callee,function" , &parse_callchain_opt, callchain_default_opt),OPT_INTEGER (0 ,"max-stack" , &report.max_stack, "Set the maximum stack depth when parsing the callchain, "
tools/perf/util/callchain.h +1 -0
@@ -53,6 +53,7 @@ struct callchain_param {sort_chain_func_t sort;enum chain_order order;enum chain_key key; +bool branch_callstack; };struct callchain_list {
tools/perf/util/machine.c +92 -21
147 lines not on this page · raw message
tools/perf/util/symbol.h +2 -1
@@ -99,7 +99,8 @@ struct symbol_conf { annotate_asm_raw, annotate_src, event_group, - demangle; + demangle, + branch_callstack;const char *vmlinux_name, *kallsyms_name, *source_prefix,
Signature · 1 line
1.8.3.1
PATCH 3/9 perf, tools: Add --branch-history option to report v2
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add a --branch-history option to perf report that changes all the settings necessary for using the branches in callstacks.
This is just a short cut to make this nicer to use, it does not enable any functionality by itself.
v2: Change sort order. Rename option to --branch-history to be less confusing.
--- tools/perf/Documentation/perf-report.txt | 5 +++++ tools/perf/builtin-report.c | 27 ++++++++++++++++++++++++--- 2 files changed, 29 insertions(+), 3 deletions(-)
tools/perf/Documentation/perf-report.txt +5 -0
@@ -223,6 +223,11 @@ OPTIONS branch stacks and it will automatically switch to the branch view mode, unless --no-branch-stack is used. +--branch-history:: + Add the addresses of sampled taken branches to the callstack. + This allows to examine the path the program took to each sample. + The data collection must have used -b (or -j) and -g. + --objdump=<path>:: Path to objdump binary.
tools/perf/builtin-report.c +24 -3
@@ -715,6 +715,16 @@ parse_branch_mode(const struct option *opt __maybe_unused, }static int +parse_branch_call_mode (const struct option *opt__maybe_unused , +const char *str__maybe_unused ,int unset) +{ +int *branch_mode = opt->value; + + *branch_mode = !unset; +return 0 ; +} + +static int parse_percent_limit (const struct option *opt,const char *str,int unset__maybe_unused ) {
@@ -729,7 +739,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)struct perf_session *session;struct stat st;bool has_br_stack = false; -int branch_mode =-1 ; +int branch_mode =-1 , branch_call_mode =-1 ;int ret =-1 ;char callchain_default_opt[] ="fractal,0.5,callee" ;const char *const report_usage[] = {
@@ -838,7 +848,10 @@ int cmd_report(int argc, const char **argv, const char *prefix __maybe_unused)OPT_BOOLEAN (0 ,"group" , &symbol_conf.event_group,"Show event group information together" ),OPT_CALLBACK_NOOPT ('b' ,"branch-stack" , &branch_mode,"" , -"use branch records for histogram filling" , parse_branch_mode), +"use branch records for per branch histogram filling" , parse_branch_mode), +OPT_CALLBACK_NOOPT (0 ,"branch-history" , &branch_call_mode,"" , +"add last branch records to call history" , + parse_branch_call_mode),OPT_STRING (0 ,"objdump" , &objdump_path,"path" ,"objdump binary to use for disassembly and annotations" ), OPT_BOOLEAN(0, "demangle", &symbol_conf.demangle,
@@ -886,8 +899,16 @@ repeat: has_br_stack =perf_header__has_feat (&session->header,HEADER_BRANCH_STACK ); -if (branch_mode ==-1 && has_br_stack) +if (branch_mode ==-1 && has_br_stack && branch_call_mode ==-1 ) sort__mode =SORT_MODE__BRANCH ; +if (branch_call_mode !=-1 ) { + callchain_param.branch_callstack =1 ; + callchain_param.key =CCKEY_ADDRESS ; + symbol_conf.use_callchain = true; +callchain_register_param (&callchain_param); +if (sort_order == default_sort_order) + sort_order ="srcline,symbol,dso" ; + }/* sort__mode could be NORMAL if --no-branch-stack */ if (sort__mode ==SORT_MODE__BRANCH ) {
Signature · 1 line
1.8.3.1
PATCH 4/9 perf, tools: Filter out small loops from LBR-as-call-stack
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Small loops can cause unnecessary duplication in the LBR-as-callstack, because the loop body appears multiple times. Filter out duplications from the LBR before unifying it into the histories. This way the same loop body only appears once.
This uses a simple hash based cycle detector. It takes some short cuts (not handling hash collisions) so in rare cases duplicates may be missed.
--- tools/perf/util/machine.c | 73 ++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 62 insertions(+), 11 deletions(-)
tools/perf/util/machine.c +62 -11
@@ -11,6 +11,7 @@#include <stdbool.h> #include <symbol/kallsyms.h> #include "unwind.h" +#include "linux/hash.h" int machine__init (struct machine *machine,const char *root_dir,pid_t pid) {
@@ -1296,6 +1297,46 @@ static int add_callchain_ip(struct machine *machine,return callchain_cursor_append (&callchain_cursor, ip, al.map, al.sym); } +#define CHASHSZ 127 +#define CHASHBITS 7 +#define NO_ENTRY 0xff + +#define PERF_MAX_BRANCH_DEPTH 127 + +/* Remove loops. */ +static int remove_loops (struct branch_entry *l,int nr) +{ +int i, j, off; +unsigned char chash[CHASHSZ ]; +memset (chash,-1 ,sizeof (chash)); + +BUG_ON (nr >=256 ); +for (i =0 ; i < nr; i++) { +int h =hash_64 (l[i].from,CHASHBITS ) %CHASHSZ ; + +/* no collision handling for now */ +if (chash[h] ==NO_ENTRY ) { + chash[h] = i; + }else if (l[chash[h]].from == l[i].from) { +bool is_loop = true; +/* check if it is a real loop */ + off =0 ; +for (j = chash[h]; j < i && i + off < nr; j++, off++) +if (l[j].from != l[i + off].from) { + is_loop = false; +break ; + } +if (is_loop) { +memmove (l + i, l + i + off, + (nr - (i + off)) + *sizeof (struct branch_entry )); + nr -= off; + } + } + } +return nr; +} + static int machine__resolve_callchain_sample(struct machine *machine, struct thread *thread, struct ip_callchain *chain,
@@ -1322,29 +1363,39 @@ static int machine__resolve_callchain_sample(struct machine *machine,* - No extra filters * - No annotations (should annotate somehow) * - When the sample is near the beginning of the function -* we may overlap with the real callstack. Could handle this -* case later, by checking against the last ip. +* we may overlap with the real callstack. */ +if (branch->nr >PERF_MAX_BRANCH_DEPTH ) { +pr_warning ("corrupted branch chain. skipping...\n" ); +return 0 ; + } +if (callchain_param.branch_callstack) { -for (i =0 ; i < branch->nr; i++) { -struct branch_entry *b; +int nr = branch->nr; +struct branch_entry be[nr]; +for (i =0 ; i < nr; i++) {if (callchain_param.order ==ORDER_CALLEE ) - b = &branch->entries[i]; + be[i] = branch->entries[i];else - b = &branch->entries[branch->nr - i -1 ]; + be[i] = branch->entries[branch->nr - i -1 ]; + } - err =add_callchain_ip (machine, thread, parent, root_al, --1 , b->to); + nr =remove_loops (be, nr); + +for (i =0 ; i < nr; i++) { + err =add_callchain_ip (machine, thread, parent, + root_al, +-1 , be[i].to);if (!err) - err =add_callchain_ip (machine, thread, parent, root_al, --1 , b->from); + err =add_callchain_ip (machine, thread, + parent, root_al, +-1 , be[i].from);if (err == -EINVAL )break ;if (err)return err; - } }
Signature · 1 line
1.8.3.1
PATCH 5/9 perf, tools: Enable printing the srcline in the history
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
For lbr-as-callgraph we need to see the line number in the history, because many LBR entries can be in a single function, and just showing the same function name many times is not useful.
When the history code is configured to sort by address, also try to resolve the address to a file:srcline and display this in the browser. If that doesn't work still display the address.
This can be also useful without LBRs for understanding which call in a large function (or in which inlined function) called something else.
Contains fixes from Namhyung Kim
--- tools/perf/ui/browsers/hists.c | 15 ++++++++++++--- tools/perf/ui/stdio/hist.c | 16 +++++++++++++--- tools/perf/util/callchain.h | 1 + tools/perf/util/machine.c | 2 +- tools/perf/util/srcline.c | 6 ++++-- 5 files changed, 31 insertions(+), 9 deletions(-)
tools/perf/ui/browsers/hists.c +12 -3
@@ -399,9 +399,18 @@ static char *callchain_list__sym_name(struct callchain_list *cl, {int printed; -if (cl->ms.sym) - printed =scnprintf (bf, bfsize,"%s" , cl->ms.sym->name); -else +if (cl->ms.sym) { +if (callchain_param.key ==CCKEY_ADDRESS && + cl->ms.map && !cl->srcline) + cl->srcline =get_srcline (cl->ms.map->dso, +map__rip_2objdump (cl->ms.map, + cl->ip)); +if (cl->srcline) + printed =scnprintf (bf, bfsize,"%s %s" , + cl->ms.sym->name, cl->srcline); +else + printed =scnprintf (bf, bfsize,"%s" , cl->ms.sym->name); + }else printed =scnprintf (bf, bfsize,"%#" PRIx64, cl->ip);if (show_dso)
tools/perf/ui/stdio/hist.c +13 -3
@@ -56,9 +56,19 @@ static size_t ipchain__fprintf_graph(FILE *fp, struct callchain_list *chain, }else ret +=fprintf (fp,"%s" ," " ); } -if (chain->ms.sym) - ret +=fprintf (fp,"%s\n" , chain->ms.sym->name); -else +if (chain->ms.sym) { +if (callchain_param.key ==CCKEY_ADDRESS && + chain->ms.map) + chain->srcline =get_srcline (chain->ms.map->dso, +map__rip_2objdump ( + chain->ms.map, + chain->ip)); +if (chain->srcline) + ret +=fprintf (fp,"%s %s\n" , + chain->ms.sym->name, chain->srcline); +else + ret +=fprintf (fp,"%s\n" , chain->ms.sym->name); + }else ret +=fprintf (fp,"0x%0" PRIx64"\n" , chain->ip);return ret;
tools/perf/util/callchain.h +1 -0
@@ -59,6 +59,7 @@ struct callchain_param {struct callchain_list {u64 ip;struct map_symbol ms; +char *srcline;struct list_head list; };
tools/perf/util/machine.c +1 -1
@@ -1294,7 +1294,7 @@ static int add_callchain_ip(struct machine *machine,return -EINVAL ; } -return callchain_cursor_append (&callchain_cursor, ip, al.map, al.sym); +return callchain_cursor_append (&callchain_cursor, al.addr, al.map, al.sym); }#define CHASHSZ 127
tools/perf/util/srcline.c +4 -2
@@ -258,7 +258,7 @@ char *get_srcline(struct dso *dso, unsigned long addr)const char *dso_name;if (!dso->has_srcline) -return SRCLINE_UNKNOWN ; + goto out;if (dso->symsrc_filename) dso_name = dso->symsrc_filename;
@@ -289,7 +289,9 @@ out: dso->has_srcline =0 ;dso__free_a2l (dso); } -return SRCLINE_UNKNOWN ; +if (asprintf (&srcline,"%s[%lx]" , dso->short_name, addr) <0 ) +return SRCLINE_UNKNOWN ; +return srcline; }void free_srcline (char *srcline)
Signature · 1 line
1.8.3.1
PATCH 6/9 perf, tools: Fix max stack handling with lbr-as-callgraph
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
The original lbr-as-callstack code ignored the maxium callgraph length set by the user. Fix this.
--- tools/perf/util/machine.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
tools/perf/util/machine.c +2 -1
@@ -1372,7 +1372,7 @@ static int machine__resolve_callchain_sample(struct machine *machine, }if (callchain_param.branch_callstack) { -int nr = branch->nr; +int nr =min (max_stack, branch->nr);struct branch_entry be[nr];for (i =0 ; i < nr; i++) {
@@ -1397,6 +1397,7 @@ static int machine__resolve_callchain_sample(struct machine *machine,if (err)return err; } + chain_nr -= nr; }if (chain->nr >PERF_MAX_STACK_DEPTH ) {
Signature · 1 line
1.8.3.1
PATCH 7/9 perf, tools: Add overlap detection for report branch-call-stack mode
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add a simple heuristic to detect overlap of LBR entries and the call stack when in lbr-as-callgraph mode. The return address in the normal callstack is one off compared to the from entry in the branch stack. Handle this with a simple "assume call instruction is not longer than 8 bytes" heuristic. With that we can remove any redundant call in the callstack that is already in the branch stack.
--- tools/perf/util/machine.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-)
tools/perf/util/machine.c +15 -6
@@ -1349,6 +1349,7 @@ static int machine__resolve_callchain_sample(struct machine *machine,int chain_nr =min (max_stack, (int )chain->nr);int i;int err; +int first_call =0 ;callchain_cursor_reset (&callchain_cursor);
@@ -1362,8 +1363,6 @@ static int machine__resolve_callchain_sample(struct machine *machine,* Limitations for now: * - No extra filters * - No annotations (should annotate somehow) -* - When the sample is near the beginning of the function -* we may overlap with the real callstack. */ if (branch->nr >PERF_MAX_BRANCH_DEPTH ) {
@@ -1372,13 +1371,23 @@ static int machine__resolve_callchain_sample(struct machine *machine, }if (callchain_param.branch_callstack) { -int nr =min (max_stack, branch->nr); +int nr =min (max_stack, (int )branch->nr);struct branch_entry be[nr];for (i =0 ; i < nr; i++) { -if (callchain_param.order ==ORDER_CALLEE ) +if (callchain_param.order ==ORDER_CALLEE ) { be[i] = branch->entries[i]; -else +/* +* Check for overlap into the callchain. +* The return address is one off compared to +* the branch entry. To adjust for this +* assume the calling instruction is not longer +* than 8 bytes. +*/ +if (be[i].from < chain->ips[first_call] && + be[i].from >= chain->ips[first_call] -8 ) + first_call++; + }else be[i] = branch->entries[branch->nr - i -1 ]; }
@@ -1405,7 +1414,7 @@ static int machine__resolve_callchain_sample(struct machine *machine,return 0 ; } -for (i =0 ; i < chain_nr; i++) { +for (i = first_call; i < chain_nr; i++) {u64 ip;if (callchain_param.order ==ORDER_CALLEE )
Signature · 1 line
1.8.3.1
PATCH 8/9 perf, tools: Only print base source file for srcline
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
For perf report with --sort srcline only print the base source file name. This makes the results generally fit much better to the screen. The path is usually not that useful anyways because it is often from different systems.
--- tools/perf/util/srcline.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/srcline.c +1 -1
@@ -274,7 +274,7 @@ char *get_srcline(struct dso *dso, unsigned long addr)if (!addr2line (dso_name, addr, &file, &line, dso)) goto out; -if (asprintf (&srcline,"%s:%u" , file, line) <0 ) { +if (asprintf (&srcline,"%s:%u" ,basename (file), line) <0 ) {free (file); goto out; }
Signature · 1 line
1.8.3.1
PATCH 9/9 perf, tools: Support source line numbers in annotate
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
With srcline key/sort'ing it's useful to have line numbers in the annotate window. This patch implements this.
Use objdump -l to request the line numbers and save them in the line structure. Then the browser displays them for source lines.
The line numbers are not displayed by default, but can be toggled on with 'k'
There is one unfortunate problem with this setup. For lines not containing source and which are outside functions objdump -l reports the wrong line numbers (it always reports the first line number in the next function even for lines that are outside the function)
I haven't find a nice way to detect/correct this. Probably objdump has to be fixed.
See https://sourceware.org/bugzilla/show_bug.cgi?id=16433
The line numbers are still useful even with these problems, as most are correct.
--- tools/perf/ui/browsers/annotate.c | 13 ++++++++++++- tools/perf/util/annotate.c | 30 +++++++++++++++++++++++++----- tools/perf/util/annotate.h | 1 + 3 files changed, 38 insertions(+), 6 deletions(-)
tools/perf/ui/browsers/annotate.c +12 -1
@@ -27,6 +27,7 @@ static struct annotate_browser_opt {bool hide_src_code, use_offset, jump_arrows, + show_linenr, show_nr_jumps; } annotate_browser__opts = { .use_offset = true,
@@ -128,7 +129,11 @@ static void annotate_browser__write(struct ui_browser *browser, void *entry, intif (!*dl->line)slsmg_write_nstring (" " , width - pcnt_width);else if (dl->offset ==-1 ) { - printed =scnprintf (bf,sizeof (bf),"%*s " , +if (dl->line_nr && annotate_browser__opts.show_linenr) + printed =scnprintf (bf,sizeof (bf),"%*s %-5d " , + ab->addr_width," " , dl->line_nr); +else + printed =scnprintf (bf,sizeof (bf),"%*s " , ab->addr_width," " );slsmg_write_nstring (bf, printed);slsmg_write_nstring (dl->line, width - printed - pcnt_width +1 );
@@ -733,6 +738,7 @@ static int annotate_browser__run(struct annotate_browser *browser,"o Toggle disassembler output/simplified view\n" "s Toggle source code view\n" "/ Search string\n" +"k Toggle line numbers\n" "r Run available scripts\n" "? Search string backwards\n");continue ;
@@ -741,6 +747,10 @@ static int annotate_browser__run(struct annotate_browser *browser,script_browse (NULL );continue ; } +case 'k' : + annotate_browser__opts.show_linenr = + !annotate_browser__opts.show_linenr; +break ;case 'H' : nd = browser->curr_hot;break ;
@@ -984,6 +994,7 @@ static struct annotate_config { } annotate__configs[] = {ANNOTATE_CFG (hide_src_code),ANNOTATE_CFG (jump_arrows), +ANNOTATE_CFG (show_linenr),ANNOTATE_CFG (show_nr_jumps), ANNOTATE_CFG(use_offset), };
tools/perf/util/annotate.c +25 -5
@@ -15,11 +15,13 @@#include "debug.h" #include "annotate.h" #include "evsel.h" +#include <regex.h> #include <pthread.h> #include <linux/bitops.h> const char *disassembler_style;const char *objdump_path; +static regex_t file_lineno;static struct ins *ins__find (const char *name);static int disasm_line__parse (char *line,char **namep,char **rawp);
@@ -562,13 +564,15 @@ out_free_name:return -1 ; } -static struct disasm_line *disasm_line__new (s64 offset,char *line,size_t privsize) +static struct disasm_line *disasm_line__new (s64 offset,char *line, +size_t privsize,int line_nr) {struct disasm_line *dl =zalloc (sizeof (*dl) + privsize);if (dl !=NULL ) { dl->offset = offset; dl->line =strdup (line); + dl->line_nr = line_nr;if (dl->line ==NULL ) goto out_delete;
@@ -780,13 +784,15 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st* The ops.raw part will be parsed further according to type of the instruction. */ static int symbol__parse_objdump_line (struct symbol *sym,struct map *map, -FILE *file,size_t privsize) +FILE *file,size_t privsize, +int *line_nr) {struct annotation *notes =symbol__annotation (sym);struct disasm_line *dl;char *line =NULL , *parsed_line, *tmp, *tmp2, *c;size_t line_len;s64 line_ip, offset =-1 ; +regmatch_t match[2 ];if (getline (&line, &line_len, file) <0 )return -1 ;
@@ -804,6 +810,12 @@ static int symbol__parse_objdump_line(struct symbol *sym, struct map *map, line_ip =-1 ; parsed_line = line; +/* /filename:linenr ? Save line number and ignore. */ +if (regexec (&file_lineno, line,2 , match,0 ) ==0 ) { + *line_nr =atoi (line + match[1 ].rm_so); +return 0 ; + } +/* * Strip leading spaces: */
@@ -834,8 +846,9 @@ static int symbol__parse_objdump_line(struct symbol *sym, struct map *map, parsed_line = tmp2 +1 ; } - dl =disasm_line__new (offset, parsed_line, privsize); + dl =disasm_line__new (offset, parsed_line, privsize, *line_nr);free (line); + (*line_nr)++;if (dl ==NULL )return -1 ;
@@ -861,6 +874,11 @@ static int symbol__parse_objdump_line(struct symbol *sym, struct map *map,return 0 ; } +static __attribute__((constructor))void symbol__init_regexpr (void ) +{ +regcomp (&file_lineno,"^/[^:]+:([0-9]+)$" ,REG_EXTENDED ); +} +static void delete_last_nop (struct symbol *sym) {struct annotation *notes =symbol__annotation (sym);
@@ -896,6 +914,7 @@ int symbol__annotate(struct symbol *sym, struct map *map, size_t privsize)char symfs_filename[PATH_MAX ];struct kcore_extract kce;bool delete_extract = false; +int lineno =0 ;if (filename) { snprintf(symfs_filename, sizeof(symfs_filename), "%s%s",
@@ -977,7 +996,7 @@ fallback: snprintf(command, sizeof(command), "%s %s%s --start-address=0x%016" PRIx64 " --stop-address=0x%016" PRIx64 - " -d %s %s -C %s 2>/dev/null|grep -v %s|expand", + " -l -d %s %s -C %s 2>/dev/null|grep -v %s|expand", objdump_path ? objdump_path : "objdump", disassembler_style ? "-M " : "", disassembler_style ? disassembler_style : "",
@@ -994,7 +1013,8 @@ fallback: goto out_free_filename;while (!feof (file)) -if (symbol__parse_objdump_line (sym, map, file, privsize) <0 ) +if (symbol__parse_objdump_line (sym, map, file, privsize, + &lineno) <0 )break ;/*
tools/perf/util/annotate.h +1 -0
@@ -58,6 +58,7 @@ struct disasm_line {char *line;char *name;struct ins *ins; +int line_nr;struct ins_operands ops; };
Signature · 1 line
1.8.3.1
Quoted 37 lines · Andi Kleen
This patchkit implements lbr-as-callgraphs in per freport, as an alternative way to present LBR information.
Current perf report does a histogram over the branch edges, which is useful to look at basic blocks, but doesn't tell you anything about the larger control flow.
This patchkit adds a new option --branch-history that adds the branch paths to the callgraph history instead.
This allows to reason about individual branch paths leading to specific samples.
Updates to v1:
- rebased on perf/core
- fix various issues
- rename the option to --branch-history
- various fixes to display the information more conciseExample output:
% perf record -b -g ./tsrc/tcall
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.044 MB perf.data (~1923 samples) ] % perf report --branch-history
...
54.91% tcall.c:6 [.] f2 tcall
|
|--66.53%-- f2 tcall.c:5
| |
| |--70.83%-- f1 tcall.c:11
| | f1 tcall.c:10
| | main tcall.c:18
| | main tcall.c:18
| | main tcall.c:17
| | main tcall.c:17
| | f1 tcall.c:13
Quoted 9 lines
| | f1 tcall.c:13
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:12
| | f1 tcall.c:12
| | f2 tcall.c:7
| | f2 tcall.c:5
| | f1 tcall.c:11
got some whitespace issues:
Applying: perf, tools: fix BFD detection on opensuse
Applying: perf, tools: Support handling complete branch stacks as histograms /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:85: space before tab in indent.
* We cannot use the header.misc hint to determine whether a/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:86: space before tab in indent.
* branch stack address is user, kernel, guest, hypervisor./home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:87: space before tab in indent.
* Branches may straddle the kernel/user/hypervisor boundaries./home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:88: space before tab in indent.
* Thus, we have to try consecutively until we find a match/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:89: space before tab in indent.
* or else, the symbol is unknown
warning: squelched 8 whitespace errors
warning: 13 lines add whitespace errors.
Applying: perf, tools: Add --branch-history option to report v2
Applying: perf, tools: Filter out small loops from LBR-as-call-stack /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:51: trailing whitespace.
memmove(l + i, l + i + off,/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:70: trailing whitespace, space before tab in indent.
* we may overlap with the real callstack./home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:84: trailing whitespace.
for (i = 0; i < nr; i++) { /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:98: trailing whitespace.
err = add_callchain_ip(machine, thread, parent,/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:104: trailing whitespace.
err = add_callchain_ip(machine, thread,
warning: 5 lines add whitespace errors.
Applying: perf, tools: Enable printing the srcline in the history /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:21: trailing whitespace.
if (callchain_param.key == CCKEY_ADDRESS &&
/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:27: trailing whitespace.
printed = scnprintf(bf, bfsize, "%s %s",
/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:47: trailing whitespace.
if (callchain_param.key == CCKEY_ADDRESS &&
/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:54: trailing whitespace.
ret += fprintf(fp, "%s %s\n", warning: 4 lines add whitespace errors.
Applying: perf, tools: Fix max stack handling with lbr-as-callgraph
Applying: perf, tools: Add overlap detection for report branch-call-stack mode /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:39: trailing whitespace.
/*
/home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:42: trailing whitespace.
* the branch entry. To adjust for this
warning: 2 lines add whitespace errors.
Applying: perf, tools: Only print base source file for srcline
Applying: perf, tools: Support source line numbers in annotate /home/jolsa/kernel.org/linux-perf/.git/rebase-apply/patch:24: trailing whitespace.
if (dl->line_nr && annotate_browser__opts.show_linenr)
warning: 1 line adds whitespace errors.