gbaudio_dapm_free_controls() iterates over widgets using list_for_each_entry_safe(), which leaves the loop cursor pointing to a meaningless structure if it completes a traversal of the list. The cursor was set to NULL at the end of the loop body, but would be overwritten by the final loop cursor update.
Because of this behavior, the widget could be non-null after the loop even if the widget wasn't found, and the cleanup logic would treat the pointer as a valid widget to free.
To fix this, introduce a temporary variable to act as the loop cursor and copy it to a variable that can be accessed after the loop finishes.
This was detected with the help of Coccinelle.
Signed-off-by: Jared Kangas kangas.jd@gmail.com --- drivers/staging/greybus/audio_helper.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/drivers/staging/greybus/audio_helper.c b/drivers/staging/greybus/audio_helper.c index 843760675876..07461a5d97c7 100644 --- a/drivers/staging/greybus/audio_helper.c +++ b/drivers/staging/greybus/audio_helper.c @@ -115,7 +115,7 @@ int gbaudio_dapm_free_controls(struct snd_soc_dapm_context *dapm, int num) { int i; - struct snd_soc_dapm_widget *w, *next_w; + struct snd_soc_dapm_widget *w, *next_w, *tmp_w; #ifdef CONFIG_DEBUG_FS struct dentry *parent = dapm->debugfs_dapm; struct dentry *debugfs_w = NULL; @@ -124,13 +124,14 @@ int gbaudio_dapm_free_controls(struct snd_soc_dapm_context *dapm, mutex_lock(&dapm->card->dapm_mutex); for (i = 0; i < num; i++) { /* below logic can be optimized to identify widget pointer */ - list_for_each_entry_safe(w, next_w, &dapm->card->widgets, + w = NULL; + list_for_each_entry_safe(tmp_w, next_w, &dapm->card->widgets, list) { - if (w->dapm != dapm) - continue; - if (!strcmp(w->name, widget->name)) + if (tmp_w->dapm == dapm && + !strcmp(tmp_w->name, widget->name)) { + w = tmp_w; break; - w = NULL; + } } if (!w) { dev_err(dapm->dev, "%s: widget not found\n",
On Sun, Jun 05, 2022 at 04:18:06PM -0700, Jared Kangas wrote:
gbaudio_dapm_free_controls() iterates over widgets using list_for_each_entry_safe(), which leaves the loop cursor pointing to a meaningless structure if it completes a traversal of the list. The cursor was set to NULL at the end of the loop body, but would be overwritten by the final loop cursor update.
Because of this behavior, the widget could be non-null after the loop even if the widget wasn't found, and the cleanup logic would treat the pointer as a valid widget to free.
To fix this, introduce a temporary variable to act as the loop cursor and copy it to a variable that can be accessed after the loop finishes.
This was detected with the help of Coccinelle.
This needs a Fixes tag.
Fixes: 510e340efe0c ("staging: greybus: audio: Add helper APIs for dynamic audio modules")
Otherwise, looks good!
Reviewed-by: Dan Carpenter dan.carpenter@oracle.com
regards, dan carpenter
Thanks for catching that! Is there anything I need to do to add the 'Fixes:' tag to the patch? From my understanding, adding tags is done by a maintainer later on, but I'm new to the patch submission process so I want to make sure I'm not missing anything.
Jared
On Tue, Jun 07, 2022 at 09:40:50AM -0700, Jared Kangas wrote:
Thanks for catching that! Is there anything I need to do to add the 'Fixes:' tag to the patch? From my understanding, adding tags is done by a maintainer later on, but I'm new to the patch submission process so I want to make sure I'm not missing anything.
No, it's up to the person sending the patch to add the fixes tag.
regards, dan carpenter
On 6/7/22 1:39 PM, Dan Carpenter wrote:
On Tue, Jun 07, 2022 at 09:40:50AM -0700, Jared Kangas wrote:
Thanks for catching that! Is there anything I need to do to add the 'Fixes:' tag to the patch? From my understanding, adding tags is done by a maintainer later on, but I'm new to the patch submission process so I want to make sure I'm not missing anything.
No, it's up to the person sending the patch to add the fixes tag.
And FYI, "git blame" can be your friend here.
git blame drivers/staging/greybus/audio_helper.c
It looks like commit 510e340efe0cbd is a possible candidate, but I'll leave it up to you to determine that.
-Alex
regards, dan carpenter
On Sun, Jun 05, 2022 at 04:18:06PM -0700, Jared Kangas wrote:
gbaudio_dapm_free_controls() iterates over widgets using list_for_each_entry_safe(), which leaves the loop cursor pointing to a meaningless structure if it completes a traversal of the list. The cursor was set to NULL at the end of the loop body, but would be overwritten by the final loop cursor update.
Because of this behavior, the widget could be non-null after the loop even if the widget wasn't found, and the cleanup logic would treat the pointer as a valid widget to free.
To fix this, introduce a temporary variable to act as the loop cursor and copy it to a variable that can be accessed after the loop finishes.
This was detected with the help of Coccinelle.
Please add the missing Fixes tag and a CC stable tag here as Dan mentioned.
Signed-off-by: Jared Kangas kangas.jd@gmail.com
drivers/staging/greybus/audio_helper.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/drivers/staging/greybus/audio_helper.c b/drivers/staging/greybus/audio_helper.c index 843760675876..07461a5d97c7 100644 --- a/drivers/staging/greybus/audio_helper.c +++ b/drivers/staging/greybus/audio_helper.c @@ -115,7 +115,7 @@ int gbaudio_dapm_free_controls(struct snd_soc_dapm_context *dapm, int num) { int i;
- struct snd_soc_dapm_widget *w, *next_w;
- struct snd_soc_dapm_widget *w, *next_w, *tmp_w;
#ifdef CONFIG_DEBUG_FS struct dentry *parent = dapm->debugfs_dapm; struct dentry *debugfs_w = NULL; @@ -124,13 +124,14 @@ int gbaudio_dapm_free_controls(struct snd_soc_dapm_context *dapm, mutex_lock(&dapm->card->dapm_mutex); for (i = 0; i < num; i++) { /* below logic can be optimized to identify widget pointer */
list_for_each_entry_safe(w, next_w, &dapm->card->widgets,
w = NULL;list_for_each_entry_safe(tmp_w, next_w, &dapm->card->widgets,
This should be list_for_each_entry() as w is not unlinked in this loop.
list) {
if (w->dapm != dapm)continue;if (!strcmp(w->name, widget->name))
if (tmp_w->dapm == dapm &&!strcmp(tmp_w->name, widget->name)) {w = tmp_w; break;
w = NULL;
}
Looks good otherwise:
Reviewed-by: Johan Hovold johan@kernel.org
-
Alex Elder -
Dan Carpenter -
Jared Kangas -
Johan Hovold