[PATCH 2/2] perf symbols: work around incorrect ET_EXEC symbol adjustment

From: Dave Martin
Date: Tue Aug 03 2010 - 07:49:07 EST

Workaround to adjust .text symbols from ET_EXEC images correctly.

This is not a complete fix, but addresses the common case and does not
create additional assumptions beyond those which perf already makes.
ELF images with a weird link map may still be processed incorrectly
(as at present) -- that will require a separate fix.

The workaround allows correct symbol offsets to be generated for
ET_EXEC executables on systems with separated debug info (where
section sh_offset fields from the debug image may bear no relation to
the layout of the executable image).

The cover note accompanying this patch contains a more detailed

Signed-off-by: Dave Martin <dave.martin@xxxxxxxxxx>
tools/perf/util/symbol.c | 10 +++++++++-
1 files changed, 9 insertions(+), 1 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 3b8c005..3930398 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -1135,7 +1135,15 @@ static int dso__load_sym(struct dso *self, struct map *map, const char *name,
"sh_addr: %#Lx sh_offset: %#Lx\n", __func__,
(u64)sym.st_value, (u64)shdr.sh_addr,
- sym.st_value -= shdr.sh_addr - shdr.sh_offset;
+ /* Assumptions:
+ * a) shdr.sh_addr - shdr.sh_offset ==
+ * map->start - map->pgoff
+ * b) map->pgoff == 0
+ * These are true iff we are looking at a function
+ * symbol in the main executable segment _and_
+ * the main executable segment starts at the start of
+ * the ELF image (normally true). */
+ sym.st_value -= map->start;
* We need to figure out if the object was created from C++ sources

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/