Skip to contentSkip to conversation

sb16 - info leak in snd_sb_csp_ioctl()

PATCH In review Dan Carpenter 2 participants 4 messages Nov 7, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 sb16 - info leak in snd_sb_csp_ioctl()

There is a 2 byte hole after "info.func_nr" so we could leak unitialized stack information to userspace.

Fixes 1da177e4c3f4 ('Linux-2.6.12-rc2')
sound/isa/sb/sb16_csp.c +1 -0
index c1aa21e..48da227 100644
@@ -208,6 +208,7 @@ static int snd_sb_csp_ioctl(struct snd_hwdep * hw, struct file *file, unsigned i
 	switch (cmd) {
 		/* get information */
 	case SNDRV_SB_CSP_IOCTL_INFO:
+		memset(&info, 0, sizeof(info));
 		*info.codec_name = *p->codec_name;
 		info.func_nr = p->func_nr;
 		info.acc_format = p->acc_format;
#1 lore raw

At Thu, 7 Nov 2013 11:09:54 +0300,

Quoted 5 lines · Dan Carpenter

There is a 2 byte hole after "info.func_nr" so we could leak unitialized stack information to userspace.

Fixes 1da177e4c3f4 ('Linux-2.6.12-rc2')

Does this help at all? It means that the bug has been there even before moving to git. I think it's better to be removed for avoid confusion.

thanks,

Takashi

Quoted 15 lines
sound/isa/sb/sb16_csp.c +1 -0
index c1aa21e..48da227 100644
@@ -208,6 +208,7 @@ static int snd_sb_csp_ioctl(struct snd_hwdep * hw, struct file *file, unsigned i
 	switch (cmd) {
 		/* get information */
 	case SNDRV_SB_CSP_IOCTL_INFO:
+		memset(&info, 0, sizeof(info));
 		*info.codec_name = *p->codec_name;
 		info.func_nr = p->func_nr;
 		info.acc_format = p->acc_format;
#2 lore raw
Quoted 11 lines · Takashi Iwai

At Thu, 7 Nov 2013 11:09:54 +0300,

Quoted 5 lines · Dan Carpenter

There is a 2 byte hole after "info.func_nr" so we could leak unitialized stack information to userspace.

Fixes 1da177e4c3f4 ('Linux-2.6.12-rc2')

Does this help at all? It means that the bug has been there even before moving to git. I think it's better to be removed for avoid confusion.

I think if you are back porting it then you know it goes back all the way. That seems useful.

The Fixes tag is still new so it's not totally clear what the rules are. I don't have strong feelings about this either way.

regards,
dan carpenter

#3 lore raw

At Thu, 7 Nov 2013 12:09:47 +0300,

Quoted 8 lines · Dan Carpenter
Quoted 6 lines · Takashi Iwai

At Thu, 7 Nov 2013 11:09:54 +0300,

Quoted 4 lines · Dan Carpenter

There is a 2 byte hole after "info.func_nr" so we could leak unitialized stack information to userspace.

Quoted 8 lines
Quoted 5 lines
Quoted 1 line
Fixes 1da177e4c3f4 ('Linux-2.6.12-rc2')

Does this help at all? It means that the bug has been there even before moving to git. I think it's better to be removed for avoid confusion.

I think if you are back porting it then you know it goes back all the way. That seems useful.

Yeah, I understand the usefulness of the tag. But my understanding is that this is used for pointing a regression point. However, in this particular case, the commit you pointed there isn't the actual commit introducing the bug. It's the genesis commit containing everything.

Quoted 2 lines

The Fixes tag is still new so it's not totally clear what the rules are. I don't have strong feelings about this either way.

OK, then let me drop that tag in this case.

thanks,

Takashi