Skip to contentSkip to conversation

Fix NULL dereference when suspending snd_opl3sa2

PATCH In review Lubomir Rintel 3 participants 5 messages Mar 16, 2009 – Mar 17, 2009 Sound

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix NULL dereference when suspending snd_opl3sa2

This should fix the following OOPS:
http://www.kerneloops.org/raw.php?rawid€591&msgid

---
 sound/isa/opl3sa2.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
sound/isa/opl3sa2.c +0 -0
index 58c972b..483c284 100644
@@ -553,7 +553,8 @@ static int snd_opl3sa2_suspend(struct snd_card

*card, pm_message_t state)
struct snd_opl3sa2 *chip = card->private_data;

snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
- chip->wss->suspend(chip->wss);
+ if (chip->wss->suspend)
+ chip->wss->suspend(chip->wss);
/* power down */
snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D3);

Signature · 1 line
1.5.5.6
#1 lore raw
Quoted 2 lines · Lubomir Rintel

This should fix the following OOPS:
http://www.kerneloops.org/raw.php?rawid€591&msgid

Wrong, sorry, the disassembly from kerneloops was a bit confusing to me. Looking at the disassembly of the actual module:

Review static int snd_opl3sa2_suspend(struct snd_card · line 554

Quoted 3 lines
(quoted diff) +0 -0
@@ -553,7 +553,8 @@ static int snd_opl3sa2_suspend(struct snd_card
 *card, pm_message_t state)
 	struct snd_opl3sa2 *chip = card->private_data;

card must somehow be NULL here:

00000270 <snd_opl3sa2_suspend>:

 270:   53                      push   %ebx
 271:   8b 98 38 01 00 00       mov    0x138(%eax),%ebx
 277:   b9 01 00 00 00          mov    $0x1,%ecx
 27c:   ba 03 00 00 00          mov    $0x3,%edx
 281:   c7 80 a8 01 00 00 00    movl   $0x300,0x1a8(%eax)
 288:   03 00 00 
 28b:   05 bc 01 00 00          add    $0x1bc,%eax
 290:   6a 00                   push   $0x0
 292:   e8 fc ff ff ff          call   293 <snd_opl3sa2_suspend+0x23>
 297:   8b 53 24                mov    0x24(%ebx),%edx
 29a:   89 d0                   mov    %edx,%eax
 29c:   ff 92 d4 00 00 00       call   *0xd4(%edx)
 2a2:   89 d8                   mov    %ebx,%eax
 2a4:   b9 27 00 00 00          mov    $0x27,%ecx
 2a9:   ba 01 00 00 00          mov    $0x1,%edx
 2ae:   e8 72 fe ff ff          call   125 <snd_opl3sa2_write>
 2b3:   58                      pop    %eax
 2b4:   31 c0                   xor    %eax,%eax
 2b6:   5b                      pop    %ebx
 2b7:   c3                      ret    

This is the same issue, with 2.6.28:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bugQ9939

#3 lore raw
In reply to Lubomir Rintel
Quoted 9 lines · On Mon, 16 Mar 2009 15:16:08 +0100 Lubomir Rintel <[email protected]> wrote:
Quoted 2 lines · Lubomir Rintel

This should fix the following OOPS:
http://www.kerneloops.org/raw.php?rawid=80591&msgid=

This is the same issue, with 2.6.28:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=519939

One should no force loading of the snd-opl3sa2 driver in the above cases. It was forced as the opl3sa2 is a PnP chip and the drivers were loaded as plain ISA devices.

However, this is not a cause of the OOPS. The cause is that despite the driver returns -ENODEV, the higher layer hides it and the driver is bound to the bus. I suppose, the suspend and resume functions are called on it.

Look at the ISA probe function:

static int __devinit snd_opl3sa2_isa_probe(struct device *pdev,
unsigned int dev)

{
    struct snd_card *card;
    int err;

    card = snd_opl3sa2_card_new(dev);
    if (! card)
        return -ENOMEM;
    snd_card_set_dev(card, pdev);
    if ((err = snd_opl3sa2_probe(card, dev)) < 0) {
        snd_card_free(card);
        return err;
    }
    dev_set_drvdata(pdev, card);
    return 0;
}

If the probe function fails, the drvdata is
never set to the snd_card structure pointer.

The patch below should fix suspend and resume
functions.

Regards,
Krzysztof
PS. It is the same situation for PnP and
BIOS PnP probing in this driver.
---

Fix the OOPS during a opl3sa2 card suspend
and resume if the driver is loaded but the card
is not found.

sound/isa/opl3sa2.c renamed from sound/isa/opl3sa2.c~ +12 -6
@@ -550,21 +550,27 @@ static int __devinit snd_opl3sa2_mixer(s
 #ifdef CONFIG_PM
 static int snd_opl3sa2_suspend(struct snd_card *card, pm_message_t state)
 {
-	struct snd_opl3sa2 *chip = card->private_data;
+	if (card) {
+		struct snd_opl3sa2 *chip = card->private_data;
 
-	snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
-	chip->wss->suspend(chip->wss);
-	/* power down */
-	snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D3);
+		snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
+		chip->wss->suspend(chip->wss);
+		/* power down */
+		snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D3);
+	}
 
 	return 0;
 }
 
 static int snd_opl3sa2_resume(struct snd_card *card)
 {
-	struct snd_opl3sa2 *chip = card->private_data;
+	struct snd_opl3sa2 *chip;
 	int i;
 
+	if (!card)
+		return 0;
+
+	chip = card->private_data;
 	/* power up */
 	snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D0);
 
#4 lore raw

At Mon, 16 Mar 2009 21:32:25 +0100,

Quoted 54 lines · Krzysztof Helt
Quoted 8 lines · On Mon, 16 Mar 2009 15:16:08 +0100 Lubomir Rintel <[email protected]> wrote:
Quoted 2 lines · Lubomir Rintel

This should fix the following OOPS:
http://www.kerneloops.org/raw.php?rawid€591&msgid> >

This is the same issue, with 2.6.28:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bugQ9939

One should no force loading of the snd-opl3sa2 driver in the above cases. It was forced as the opl3sa2 is a PnP chip and the drivers were loaded as plain ISA devices.

However, this is not a cause of the OOPS. The cause is that despite the driver returns -ENODEV, the higher layer hides it and the driver is bound to the bus. I suppose, the suspend and resume functions are called on it.

Look at the ISA probe function:

static int __devinit snd_opl3sa2_isa_probe(struct device *pdev,
unsigned int dev)

{
    struct snd_card *card;
    int err;

    card = snd_opl3sa2_card_new(dev);
    if (! card)
        return -ENOMEM;
    snd_card_set_dev(card, pdev);
    if ((err = snd_opl3sa2_probe(card, dev)) < 0) {
        snd_card_free(card);
        return err;
    }
    dev_set_drvdata(pdev, card);
    return 0;
}

If the probe function fails, the drvdata is
never set to the snd_card structure pointer.

The patch below should fix suspend and resume
functions.

Regards,
Krzysztof
PS. It is the same situation for PnP and
BIOS PnP probing in this driver.
---

Quoted 5 lines

Fix the OOPS during a opl3sa2 card suspend
and resume if the driver is loaded but the card
is not found.

This patch looks good. I applied it to sound git tree now.

Thanks,

Takashi

Quoted 43 lines
sound/isa/opl3sa2.c renamed from sound/isa/opl3sa2.c~ +12 -7
@@ -550,21 +550,27 @@ static int __devinit snd_opl3sa2_mixer(s
 #ifdef CONFIG_PM
 static int snd_opl3sa2_suspend(struct snd_card *card, pm_message_t state)
 {
-	struct snd_opl3sa2 *chip = card->private_data;
+	if (card) {
+		struct snd_opl3sa2 *chip = card->private_data;
 
-	snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
-	chip->wss->suspend(chip->wss);
-	/* power down */
-	snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D3);
+		snd_power_change_state(card, SNDRV_CTL_POWER_D3hot);
+		chip->wss->suspend(chip->wss);
+		/* power down */
+		snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D3);
+	}
 
 	return 0;
 }
 
 static int snd_opl3sa2_resume(struct snd_card *card)
 {
-	struct snd_opl3sa2 *chip = card->private_data;
+	struct snd_opl3sa2 *chip;
 	int i;
 
+	if (!card)
+		return 0;
+
+	chip = card->private_data;
 	/* power up */
 	snd_opl3sa2_write(chip, OPL3SA2_PM_CTRL, OPL3SA2_PM_D0);
 
 
--

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/