[patch] [ALSA] seq_oss_event: missing range checks

[patch] [ALSA] seq_oss_event: missing range checks

[Dan Carpenter]

The "dev" variable could be out of bounds.  Calling
snd_seq_oss_synth_is_valid() checks that it is is a valid device
which has been opened.  We check this inside set_note_event() so
this function can't succeed without a valid "dev".  But we need to
do the check earlier to prevent invalid dereferences and memory
corruption.

One call tree where "dev" could be out of bounds is:
-> snd_seq_oss_oob_user()
   -> snd_seq_oss_process_event()
      -> extended_event()
         -> note_on_event()

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

diff --git a/sound/core/seq/oss/seq_oss_event.c b/sound/core/seq/oss/seq_oss_event.c
index 066f5f3..c390886 100644
--- a/sound/core/seq/oss/seq_oss_event.c
+++ b/sound/core/seq/oss/seq_oss_event.c
@@ -285,7 +285,12 @@ local_event(struct seq_oss_devinfo *dp, union evrec *q, struct snd_seq_event *ev
 static int
 note_on_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, struct snd_seq_event *ev)
 {
-	struct seq_oss_synthinfo *info = &dp->synths[dev];
+	struct seq_oss_synthinfo *info;
+
+	if (!snd_seq_oss_synth_is_valid(dp, dev))
+		return -ENXIO;
+
+	info = &dp->synths[dev];
 	switch (info->arg.event_passing) {
 	case SNDRV_SEQ_OSS_PROCESS_EVENTS:
 		if (! info->ch || ch < 0 || ch >= info->nr_voices) {
@@ -340,7 +345,12 @@ note_on_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, st
 static int
 note_off_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, struct snd_seq_event *ev)
 {
-	struct seq_oss_synthinfo *info = &dp->synths[dev];
+	struct seq_oss_synthinfo *info;
+
+	if (!snd_seq_oss_synth_is_valid(dp, dev))
+		return -ENXIO;
+
+	info = &dp->synths[dev];
 	switch (info->arg.event_passing) {
 	case SNDRV_SEQ_OSS_PROCESS_EVENTS:
 		if (! info->ch || ch < 0 || ch >= info->nr_voices) {

Re: [alsa-devel] [patch] [ALSA] seq_oss_event: missing range checks

[Takashi Iwai]

At Mon, 4 Mar 2013 15:19:18 +0300,
Dan Carpenter wrote:
> 
> The "dev" variable could be out of bounds.  Calling
> snd_seq_oss_synth_is_valid() checks that it is is a valid device
> which has been opened.  We check this inside set_note_event() so
> this function can't succeed without a valid "dev".  But we need to
> do the check earlier to prevent invalid dereferences and memory
> corruption.
> 
> One call tree where "dev" could be out of bounds is:
> -> snd_seq_oss_oob_user()
>    -> snd_seq_oss_process_event()
>       -> extended_event()
>          -> note_on_event()
> 
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

Thanks, applied.


Takashi

> 
> diff --git a/sound/core/seq/oss/seq_oss_event.c b/sound/core/seq/oss/seq_oss_event.c
> index 066f5f3..c390886 100644
> --- a/sound/core/seq/oss/seq_oss_event.c
> +++ b/sound/core/seq/oss/seq_oss_event.c
> @@ -285,7 +285,12 @@ local_event(struct seq_oss_devinfo *dp, union evrec *q, struct snd_seq_event *ev
>  static int
>  note_on_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, struct snd_seq_event *ev)
>  {
> -	struct seq_oss_synthinfo *info = &dp->synths[dev];
> +	struct seq_oss_synthinfo *info;
> +
> +	if (!snd_seq_oss_synth_is_valid(dp, dev))
> +		return -ENXIO;
> +
> +	info = &dp->synths[dev];
>  	switch (info->arg.event_passing) {
>  	case SNDRV_SEQ_OSS_PROCESS_EVENTS:
>  		if (! info->ch || ch < 0 || ch >= info->nr_voices) {
> @@ -340,7 +345,12 @@ note_on_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, st
>  static int
>  note_off_event(struct seq_oss_devinfo *dp, int dev, int ch, int note, int vel, struct snd_seq_event *ev)
>  {
> -	struct seq_oss_synthinfo *info = &dp->synths[dev];
> +	struct seq_oss_synthinfo *info;
> +
> +	if (!snd_seq_oss_synth_is_valid(dp, dev))
> +		return -ENXIO;
> +
> +	info = &dp->synths[dev];
>  	switch (info->arg.event_passing) {
>  	case SNDRV_SEQ_OSS_PROCESS_EVENTS:
>  		if (! info->ch || ch < 0 || ch >= info->nr_voices) {
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@alsa-project.org
> http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
>