lkml.org 
[lkml]   [2011]   [Jul]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH -tip 4/5] [BUGFIX] perf probe: Fix the order of searching scopes for variables
Hi Pekka,

(2011/07/10 20:05), Pekka Enberg wrote:
> On Sun, Jul 10, 2011 at 1:00 PM, Masami Hiramatsu
> <masami.hiramatsu.pt@hitachi.com> wrote:
>> Fix variable searching routine to search from inner scope
>> to outer scope.
>
> Admittedly, I don't really know this code at all. However, the
> changelog is somewhat vague. Why do we need this change? What problem
> does it fix? Why does the order in which we iterate 'scopes' matter?

Hmm, I see.

Debuginfo expresses functions and nested inlined functions as
DIE(Dwarf Information Entry) tree, and "scope" means each
level of the DIE.

Outer
^
|
CU DIE [A](compile unit: it's an object file)
+-global variable DIEs
+-inline function declaration DIEs
+-function [B] DIE
+-local variable DIEs
+-inline function instance [C] DIE
+-inline local variable DIEs
+-(nested)inline function instance [D] DIE
+...
|
v
Inner

Imagine that if we are at [C],

nscopes = dwarf_getscopes_die([C], &scopes)

this returns scopes[] DIEs containing [C] DIE, and nscopes = 3
On that array, scopes[0] is [C], scopes[1] is [B], and scopes[2]
is [A].

And, the below loop

while (nscopes-- > 1)
die_find_variable_at(&scopes[nscopes], ...

starts to find variables from the outermost scope [A] and
find a global variable first. This changes searching order
from inner scope ([B]) to outer ([A]) to search global vars
last.


Thank you!


>
>> Signed-off-by: Masami Hiramatsu <masami.hiramatsu@gmail.com>
>> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
>> Cc: Paul Mackerras <paulus@samba.org>
>> Cc: Ingo Molnar <mingo@elte.hu>
>> Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
>> ---
>>
>> tools/perf/util/probe-finder.c | 14 +++++++-------
>> 1 files changed, 7 insertions(+), 7 deletions(-)
>>
>> diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c
>> index 222ba66..cf7d48d 100644
>> --- a/tools/perf/util/probe-finder.c
>> +++ b/tools/perf/util/probe-finder.c
>> @@ -1135,7 +1135,7 @@ static int find_variable(Dwarf_Die *sp_die, struct probe_finder *pf)
>> {
>> Dwarf_Die vr_die, *scopes;
>> char buf[32], *ptr;
>> - int ret, nscopes;
>> + int ret, nscopes, i;
>>
>> if (!is_c_varname(pf->pvar->var)) {
>> /* Copy raw parameters */
>> @@ -1179,11 +1179,11 @@ static int find_variable(Dwarf_Die *sp_die, struct probe_finder *pf)
>> /* Search upper class */
>> nscopes = dwarf_getscopes_die(sp_die, &scopes);
>> ret = -ENOENT;
>> - while (nscopes-- > 1) {
>> + for (i = 1; i < nscopes; i++) { /* scopes[0] is sp_die */
>> pr_debug("Searching variables in %s\n",
>> - dwarf_diename(&scopes[nscopes]));
>> + dwarf_diename(&scopes[i]));
>> /* We should check this scope, so give dummy address */
>> - if (die_find_variable_at(&scopes[nscopes],
>> + if (die_find_variable_at(&scopes[i],
>> pf->pvar->var, 0,
>> &vr_die)) {
>> ret = convert_variable(&vr_die, pf);
>> @@ -1693,7 +1693,7 @@ static int add_available_vars(Dwarf_Die *sp_die, struct probe_finder *pf)
>> container_of(pf, struct available_var_finder, pf);
>> struct variable_list *vl;
>> Dwarf_Die die_mem, *scopes = NULL;
>> - int ret, nscopes;
>> + int ret, nscopes, i;
>>
>> /* Check number of tevs */
>> if (af->nvls == af->max_vls) {
>> @@ -1723,8 +1723,8 @@ static int add_available_vars(Dwarf_Die *sp_die, struct probe_finder *pf)
>> /* Don't need to search child DIE for externs. */
>> af->child = false;
>> nscopes = dwarf_getscopes_die(sp_die, &scopes);
>> - while (nscopes-- > 1)
>> - die_find_child(&scopes[nscopes], collect_variables_cb,
>> + for (i = 1; i < nscopes; i++) /* scopes[0] is sp_die */
>> + die_find_child(&scopes[i], collect_variables_cb,
>> (void *)af, &die_mem);
>> if (scopes)
>> free(scopes);
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>> Please read the FAQ at http://www.tux.org/lkml/
>>
>


--
Masami HIRAMATSU
Software Platform Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@hitachi.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2011-07-10 14:11    [W:0.051 / U:4.828 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site