Skip to contentSkip to conversation

perf callchain: Pass relative address to hist entry

PATCH In review Namhyung Kim 2 participants 3 messages Oct 25, 2017 perf events

lore
Share & tools
lore

Namhyung Kim

author maintainer
#0 lore raw

PATCH 1/1 perf callchain: Pass relative address to hist entry

The addr_location->addr should have relative address to be passed to hist entry so that it can compare with others which might come from a different address space.

The callchain_cursor_node->ip already has a relative address thus it should not map it again. But I found a bug handling cumulative (children) entries. For chilren entries that has no self period, the al->addr (so he->ip) ends up having an doubly-mapped address.

It seems to be there from the beginning but only affects entries that have no srclines - finding srcline itself is done using a different address but it will show the invalid address if no srcline was found.

Fixes c7405d85d7a3 ("perf tools: Update cpumode for each cumulative entry")
---
 tools/perf/util/callchain.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)
tools/perf/util/callchain.c +1 -4
index 3a3916934a92..837012147c7b 100644
@@ -1091,10 +1091,7 @@ int fill_callchain_info(struct addr_location *al, struct callchain_cursor_node *
 	al->map = node->map;
 	al->sym = node->sym;
 	al->srcline = node->srcline;
-	if (node->map)
-		al->addr = node->map->map_ip(node->map, node->ip);
-	else
-		al->addr = node->ip;
+	al->addr = node->ip;
 
 	if (al->sym == NULL) {
 		if (hide_unresolved)
Signature · 1 line
2.14.2
#1 lore raw
Quoted 9 lines · Namhyung Kim

The addr_location->addr should have relative address to be passed to hist entry so that it can compare with others which might come from a different address space.

The callchain_cursor_node->ip already has a relative address thus it should not map it again. But I found a bug handling cumulative (children) entries. For chilren entries that has no self period, the al->addr (so he->ip) ends up having an doubly-mapped address.

Quoted 7 lines

It seems to be there from the beginning but only affects entries that have no srclines - finding srcline itself is done using a different address but it will show the invalid address if no srcline was found.

Fixes c7405d85d7a3 ("perf tools: Update cpumode for each cumulative entry")

hum, I'm getting differences in hist entries for:

 $ perf record --all-user -ga sleep 1
 $ perf perf report --stdio > ...
 ...
 $ diff -puw  /tmp/o.old /tmp/o.new

jirka

---
@@ -51,7 +51,7 @@

                           |
                            --0.94%--0x3
-     3.38%     0.00%  Web Content      libxul.so                    [.] 0xffff8087f8e046a5
+     3.38%     0.00%  Web Content      libxul.so                    [.] 0x00000000031c16a5
             |
             ---0x31c16a5
                |
@@ -91,7 +91,7 @@
                |
                 --1.62%--malloc
-     2.47%     0.00%  Compositor       libxul.so                    [.] 0xffff80e6f2807790
+     2.47%     0.00%  Compositor       libxul.so                    [.] 0x0000000000b9b790
             |
             ---0xb9b790
                |
@@ -163,7 +163,7 @@
             |
             ---0x72be62f5fdc55d00
-     1.05%     0.00%  Compositor       libxul.so                    [.] 0xffff80e6f280f100
+     1.05%     0.00%  Compositor       libxul.so                    [.] 0x0000000000ba3100
             |
             ---0xba3100
                |
author maintainer
#2 lore raw

Hi Jiri,

Quoted 18 lines · Jiri Olsa
Quoted 9 lines · Namhyung Kim

The addr_location->addr should have relative address to be passed to hist entry so that it can compare with others which might come from a different address space.

The callchain_cursor_node->ip already has a relative address thus it should not map it again. But I found a bug handling cumulative (children) entries. For chilren entries that has no self period, the al->addr (so he->ip) ends up having an doubly-mapped address.

Quoted 7 lines

It seems to be there from the beginning but only affects entries that have no srclines - finding srcline itself is done using a different address but it will show the invalid address if no srcline was found.

Fixes c7405d85d7a3 ("perf tools: Update cpumode for each cumulative entry")
Quoted 6 lines

hum, I'm getting differences in hist entries for:

 $ perf record --all-user -ga sleep 1
 $ perf perf report --stdio > ...
 ...
 $ diff -puw  /tmp/o.old /tmp/o.new

I'd like to say that the old address is invalid since user space function address should not be started with 0xffff...

The last paragraph of the commit message should say that it also affects entries don't have symbols.

Thanks,
Namhyung

Quoted 34 lines

jirka

---

(quoted diff) +3 -3
@@ -51,7 +51,7 @@
                           |
                            --0.94%--0x3
 
-     3.38%     0.00%  Web Content      libxul.so                    [.] 0xffff8087f8e046a5
+     3.38%     0.00%  Web Content      libxul.so                    [.] 0x00000000031c16a5
             |
             ---0x31c16a5
                |
@@ -91,7 +91,7 @@
                |
                 --1.62%--malloc
 
-     2.47%     0.00%  Compositor       libxul.so                    [.] 0xffff80e6f2807790
+     2.47%     0.00%  Compositor       libxul.so                    [.] 0x0000000000b9b790
             |
             ---0xb9b790
                |
@@ -163,7 +163,7 @@
             |
             ---0x72be62f5fdc55d00
 
-     1.05%     0.00%  Compositor       libxul.so                    [.] 0xffff80e6f280f100
+     1.05%     0.00%  Compositor       libxul.so                    [.] 0x0000000000ba3100
             |
             ---0xba3100
                |