[alsa-devel] [PATCH] ASoC: soc-core: fix an uninitialized use
Fixed the uninitialized use of a signed integer variable ret in soc_probe_component when all its definitions are not executed. This caused -ftrivial-auto-var-init=pattern to initialize the variable to repeated 0xAA (i.e. a negative value) and triggered the following code unintentionally.
err_probe: if (ret < 0) soc_cleanup_component(component);
Signed-off-by: Jian Cai caij2003@gmail.com --- sound/soc/soc-core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c index 068d809c349a..bfb813ba34f3 100644 --- a/sound/soc/soc-core.c +++ b/sound/soc/soc-core.c @@ -1180,7 +1180,7 @@ static int soc_probe_component(struct snd_soc_card *card, snd_soc_component_get_dapm(component); struct snd_soc_dai *dai; int probed = 0; - int ret; + int ret = 0;
if (!strcmp(component->name, "snd-soc-dummy")) return 0;
Fixed the uninitialized use of a signed integer variable ret in soc_probe_component when all its definitions are not executed. This caused -ftrivial-auto-var-init=pattern to initialize the variable to repeated 0xAA (i.e. a negative value) and triggered the following code unintentionally.
Signed-off-by: Jian Cai caij2003@gmail.com
Hi Jian, I don't quite follow; it looks like `ret` is assigned to multiple times in `soc_probe_component`. Are one of the return values of one of the functions that are called then assigned to `ret` undefined? What control flow path leaves `ret` unitialized?
Hi Nick,
'ret' is only defined in if branches and for loops (e.g. for_each_component_dais). If none of these branches or loops get executed, then eventually we end up having
int ret;
err_probe: if (ret < 0) soc_cleanup_component(component);
With -ftrivial-auto-var-init=pattern, this code becomes
int ret;
err_probe: ret = 0xAAAAAAAA; if (ret < 0) soc_cleanup_component(component);
So soc_cleanup_component gets called unintentionally this case, which causes the built kernel to miss some files.
On Thu, Feb 6, 2020 at 3:28 PM Nick Desaulniers ndesaulniers@google.com wrote:
Fixed the uninitialized use of a signed integer variable ret in soc_probe_component when all its definitions are not executed. This caused -ftrivial-auto-var-init=pattern to initialize the variable to repeated 0xAA (i.e. a negative value) and triggered the following code unintentionally.
Signed-off-by: Jian Cai caij2003@gmail.com
Hi Jian, I don't quite follow; it looks like `ret` is assigned to multiple times in `soc_probe_component`. Are one of the return values of one of the functions that are called then assigned to `ret` undefined? What control flow path leaves `ret` unitialized?
On Fri, Feb 7, 2020 at 12:55 AM Jian Cai caij2003@gmail.com wrote:
Hi Nick,
'ret' is only defined in if branches and for loops (e.g. for_each_component_dais). If none of these branches or loops get executed, then eventually we end up having
https://elixir.bootlin.com/linux/latest/source/sound/soc/soc-core.c#L1276 and https://elixir.bootlin.com/linux/latest/source/sound/soc/soc-core.c#L1287 both assign to `ret` before any `goto` is taken. Are you perhaps looking at an older branch of the LTS tree, but not the master branch of the mainline tree? (Or it's possible that it's 1am here in Zurich, and I should go to bed).
int ret;
err_probe: if (ret < 0) soc_cleanup_component(component);
With -ftrivial-auto-var-init=pattern, this code becomes
int ret;
err_probe: ret = 0xAAAAAAAA; if (ret < 0) soc_cleanup_component(component);
So soc_cleanup_component gets called unintentionally this case, which causes the built kernel to miss some files.
On Thu, Feb 6, 2020 at 3:28 PM Nick Desaulniers ndesaulniers@google.com wrote:
Fixed the uninitialized use of a signed integer variable ret in soc_probe_component when all its definitions are not executed. This caused -ftrivial-auto-var-init=pattern to initialize the variable to repeated 0xAA (i.e. a negative value) and triggered the following code unintentionally.
Signed-off-by: Jian Cai caij2003@gmail.com
Hi Jian, I don't quite follow; it looks like `ret` is assigned to multiple times in `soc_probe_component`. Are one of the return values of one of the functions that are called then assigned to `ret` undefined? What control flow path leaves `ret` unitialized?
Thanks for the pointers. You are absolutely right (despite working late), this is not an issue upstream anymore. I was looking at 4.14 and 4.19 on ChromeOS. I did double check the upstream code but stopped right after seeing 'ret' was still uninitialized. Thanks again for the information.
On Thu, Feb 6, 2020 at 4:04 PM Nick Desaulniers ndesaulniers@google.com wrote:
On Fri, Feb 7, 2020 at 12:55 AM Jian Cai caij2003@gmail.com wrote:
Hi Nick,
'ret' is only defined in if branches and for loops (e.g.
for_each_component_dais). If none of these branches or loops get executed, then eventually we end up having
https://elixir.bootlin.com/linux/latest/source/sound/soc/soc-core.c#L1276 and https://elixir.bootlin.com/linux/latest/source/sound/soc/soc-core.c#L1287 both assign to `ret` before any `goto` is taken. Are you perhaps looking at an older branch of the LTS tree, but not the master branch of the mainline tree? (Or it's possible that it's 1am here in Zurich, and I should go to bed).
int ret;
err_probe: if (ret < 0) soc_cleanup_component(component);
With -ftrivial-auto-var-init=pattern, this code becomes
int ret;
err_probe: ret = 0xAAAAAAAA; if (ret < 0) soc_cleanup_component(component);
So soc_cleanup_component gets called unintentionally this case, which
causes the built kernel to miss some files.
On Thu, Feb 6, 2020 at 3:28 PM Nick Desaulniers ndesaulniers@google.com
wrote:
Fixed the uninitialized use of a signed integer variable ret in soc_probe_component when all its definitions are not executed. This caused -ftrivial-auto-var-init=pattern to initialize the variable to repeated 0xAA (i.e. a negative value) and triggered the following code unintentionally.
Signed-off-by: Jian Cai caij2003@gmail.com
Hi Jian, I don't quite follow; it looks like `ret` is assigned to multiple times
in
`soc_probe_component`. Are one of the return values of one of the
functions
that are called then assigned to `ret` undefined? What control flow
path leaves
`ret` unitialized?
-- Thanks, ~Nick Desaulniers
participants (2)
-
Jian Cai -
Nick Desaulniers