[PATCH RESEND] dm9601: don't do usb transfers of data on stack

[PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Peter Korsgaard]

dm_{read,write}_shared_word() were doing USB transfers of data on stack,
which isn't allowed. Fix it by using the usbnet->data area like in
dm9601_set_multicast().

Signed-off-by: Peter Korsgaard <jacmet-OfajU3CKLf1/SzgSGea1oA@public.gmane.org>
---
 Resending since no feedback (27/06).

 drivers/net/usb/dm9601.c |   14 ++++++++++++--
 1 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/drivers/net/usb/dm9601.c b/drivers/net/usb/dm9601.c
index f7319d3..1a5e649 100644
--- a/drivers/net/usb/dm9601.c
+++ b/drivers/net/usb/dm9601.c
@@ -158,6 +158,10 @@ static void dm_write_reg_async(struct usbnet *dev, u8 reg, u8 value)
 static int dm_read_shared_word(struct usbnet *dev, int phy, u8 reg, __le16 *value)
 {
 	int ret, i;
+	/* USB transfers of data on stack is not allowed  - We use
+	   the 20 byte dev->data for value to avoid allocating memory.
+	   Notice that dm9601_set_multicast() uses the lower 8 bytes */
+	void *data = &dev->data[4];
 
 	mutex_lock(&dev->phy_mutex);
 
@@ -184,7 +188,8 @@ static int dm_read_shared_word(struct usbnet *dev, int phy, u8 reg, __le16 *valu
 	}
 
 	dm_write_reg(dev, DM_SHARED_CTRL, 0x0);
-	ret = dm_read(dev, DM_SHARED_DATA, 2, value);
+	ret = dm_read(dev, DM_SHARED_DATA, 2, data);
+	memcpy(value, data, 2);
 
 	devdbg(dev, "read shared %d 0x%02x returned 0x%04x, %d",
 	       phy, reg, *value, ret);
@@ -197,10 +202,15 @@ static int dm_read_shared_word(struct usbnet *dev, int phy, u8 reg, __le16 *valu
 static int dm_write_shared_word(struct usbnet *dev, int phy, u8 reg, __le16 value)
 {
 	int ret, i;
+	/* USB transfers of data on stack is not allowed  - We use
+	   the 20 byte dev->data for value to avoid allocating memory.
+	   Notice that dm9601_set_multicast() uses the lower 8 bytes */
+	void *data = &dev->data[4];
 
 	mutex_lock(&dev->phy_mutex);
 
-	ret = dm_write(dev, DM_SHARED_DATA, 2, &value);
+	memcpy(data, &value, sizeof(value));
+	ret = dm_write(dev, DM_SHARED_DATA, 2, data);
 	if (ret < 0)
 		goto out;
 
-- 
1.5.5.1


-- 
Bye, Peter Korsgaard
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Oliver Neukum]

Am Freitag 11 Juli 2008 13:37:39 schrieb Peter Korsgaard:
> dm_{read,write}_shared_word() were doing USB transfers of data on stack,
> which isn't allowed. Fix it by using the usbnet->data area like in
> dm9601_set_multicast().

You cannot do this. You've identified the bug correctly, but the fix
is incorrect. Doing DMA on non coherent architectures means that
any cacheline the buffer shares must not be touched while DMA
may be running. If you embed the buffer into struct usbnet, how could
you meet the guarantee?

	Regards
		Oliver
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Peter Korsgaard]

>>>>> "Oliver" == Oliver Neukum <oliver-GvhC2dPhHPQdnm+yROfE0A@public.gmane.org> writes:

 >> dm_{read,write}_shared_word() were doing USB transfers of data on stack,
 >> which isn't allowed. Fix it by using the usbnet->data area like in
 >> dm9601_set_multicast().

 Oliver> You cannot do this. You've identified the bug correctly, but
 Oliver> the fix is incorrect. Doing DMA on non coherent architectures
 Oliver> means that any cacheline the buffer shares must not be
 Oliver> touched while DMA may be running. If you embed the buffer
 Oliver> into struct usbnet, how could you meet the guarantee?

Ahh. So kmalloc() / kfree() around the usb_control_msg isn dm_read /
dm_write is the way to go?

-- 
Bye, Peter Korsgaard
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Oliver Neukum]

Am Freitag 11 Juli 2008 14:17:58 schrieb Peter Korsgaard:
> >>>>> "Oliver" == Oliver Neukum <oliver@neukum.org> writes:
> 
>  >> dm_{read,write}_shared_word() were doing USB transfers of data on stack,
>  >> which isn't allowed. Fix it by using the usbnet->data area like in
>  >> dm9601_set_multicast().
> 
>  Oliver> You cannot do this. You've identified the bug correctly, but
>  Oliver> the fix is incorrect. Doing DMA on non coherent architectures
>  Oliver> means that any cacheline the buffer shares must not be
>  Oliver> touched while DMA may be running. If you embed the buffer
>  Oliver> into struct usbnet, how could you meet the guarantee?
> 
> Ahh. So kmalloc() / kfree() around the usb_control_msg isn dm_read /
> dm_write is the way to go?

The buffer should be allocated with kmalloc. That doesn't mean it has to
be allocated each time the function is called. If several drivers do this, the
best fix is to modify struct usbnet so that it contains only a pointer to the
buffer, not the buffer itself and modify drivers accordingly.

	Regards
		Oliver

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Peter Korsgaard]

>>>>> "Oliver" == Oliver Neukum <oliver-GvhC2dPhHPQdnm+yROfE0A@public.gmane.org> writes:

Hi,

 >> Ahh. So kmalloc() / kfree() around the usb_control_msg isn dm_read /
 >> dm_write is the way to go?

 Oliver> The buffer should be allocated with kmalloc. That doesn't
 Oliver> mean it has to be allocated each time the function is
 Oliver> called. If several drivers do this, the best fix is to modify
 Oliver> struct usbnet so that it contains only a pointer to the
 Oliver> buffer, not the buffer itself and modify drivers accordingly.

Ok, but this is just for a few config registers - Nothing in the fast
path, so I'll keep it simple.

-- 
Bye, Peter Korsgaard
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Peter Korsgaard]

dm_{read,write}() were doing USB transfers of data on stack, which isn't
allowed. Fix it by kmalloc'ing a temporary buffer.
Clean up the error handling for short transfers while we're at it.

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
---
 drivers/net/usb/dm9601.c |   52 +++++++++++++++++++++++++++++++++++++--------
 1 files changed, 42 insertions(+), 10 deletions(-)

diff --git a/drivers/net/usb/dm9601.c b/drivers/net/usb/dm9601.c
index f7319d3..78df2be 100644
--- a/drivers/net/usb/dm9601.c
+++ b/drivers/net/usb/dm9601.c
@@ -55,12 +55,28 @@
 
 static int dm_read(struct usbnet *dev, u8 reg, u16 length, void *data)
 {
+	void *buf;
+	int err = -ENOMEM;
+
 	devdbg(dev, "dm_read() reg=0x%02x length=%d", reg, length);
-	return usb_control_msg(dev->udev,
-			       usb_rcvctrlpipe(dev->udev, 0),
-			       DM_READ_REGS,
-			       USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
-			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
+
+	buf = kmalloc(length, GFP_KERNEL);
+	if (!buf)
+		goto out;
+
+	err = usb_control_msg(dev->udev,
+			      usb_rcvctrlpipe(dev->udev, 0),
+			      DM_READ_REGS,
+			      USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
+			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
+	if (err == length)
+		memcpy(data, buf, length);
+	else if (err >= 0)
+		err = -EINVAL;
+	kfree(buf);
+
+ out:
+	return err;
 }
 
 static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
@@ -70,12 +86,28 @@ static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
 
 static int dm_write(struct usbnet *dev, u8 reg, u16 length, void *data)
 {
+	void *buf = NULL;
+	int err = -ENOMEM;
+
 	devdbg(dev, "dm_write() reg=0x%02x, length=%d", reg, length);
-	return usb_control_msg(dev->udev,
-			       usb_sndctrlpipe(dev->udev, 0),
-			       DM_WRITE_REGS,
-			       USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
-			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
+
+	if (data) {
+		buf = kmalloc(length, GFP_KERNEL);
+		if (!buf)
+			goto out;
+		memcpy(buf, data, length);
+	}
+
+	err = usb_control_msg(dev->udev,
+			      usb_sndctrlpipe(dev->udev, 0),
+			      DM_WRITE_REGS,
+			      USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
+			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
+	kfree(buf);
+	if (err >= 0 && err < length)
+		err = -EINVAL;
+ out:
+	return err;
 }
 
 static int dm_write_reg(struct usbnet *dev, u8 reg, u8 value)
-- 
1.5.6.2


-- 
Bye, Peter Korsgaard

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[David Brownell]

On Friday 11 July 2008, Peter Korsgaard wrote:
> dm_{read,write}() were doing USB transfers of data on stack, which isn't
> allowed. Fix it by kmalloc'ing a temporary buffer.
> Clean up the error handling for short transfers while we're at it.
> 
> Signed-off-by: Peter Korsgaard <jacmet-OfajU3CKLf1/SzgSGea1oA@public.gmane.org>

Acked-by: David Brownell <dbrownell-Rn4VEauK+AKRv+LV9MX5uipxlwaOVQ5f@public.gmane.org>

> ---
>  drivers/net/usb/dm9601.c |   52 +++++++++++++++++++++++++++++++++++++--------
>  1 files changed, 42 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/usb/dm9601.c b/drivers/net/usb/dm9601.c
> index f7319d3..78df2be 100644
> --- a/drivers/net/usb/dm9601.c
> +++ b/drivers/net/usb/dm9601.c
> @@ -55,12 +55,28 @@
>  
>  static int dm_read(struct usbnet *dev, u8 reg, u16 length, void *data)
>  {
> +	void *buf;
> +	int err = -ENOMEM;
> +
>  	devdbg(dev, "dm_read() reg=0x%02x length=%d", reg, length);
> -	return usb_control_msg(dev->udev,
> -			       usb_rcvctrlpipe(dev->udev, 0),
> -			       DM_READ_REGS,
> -			       USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
> -			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
> +
> +	buf = kmalloc(length, GFP_KERNEL);
> +	if (!buf)
> +		goto out;
> +
> +	err = usb_control_msg(dev->udev,
> +			      usb_rcvctrlpipe(dev->udev, 0),
> +			      DM_READ_REGS,
> +			      USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
> +			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
> +	if (err == length)
> +		memcpy(data, buf, length);
> +	else if (err >= 0)
> +		err = -EINVAL;
> +	kfree(buf);
> +
> + out:
> +	return err;
>  }
>  
>  static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
> @@ -70,12 +86,28 @@ static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
>  
>  static int dm_write(struct usbnet *dev, u8 reg, u16 length, void *data)
>  {
> +	void *buf = NULL;
> +	int err = -ENOMEM;
> +
>  	devdbg(dev, "dm_write() reg=0x%02x, length=%d", reg, length);
> -	return usb_control_msg(dev->udev,
> -			       usb_sndctrlpipe(dev->udev, 0),
> -			       DM_WRITE_REGS,
> -			       USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
> -			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
> +
> +	if (data) {
> +		buf = kmalloc(length, GFP_KERNEL);
> +		if (!buf)
> +			goto out;
> +		memcpy(buf, data, length);
> +	}
> +
> +	err = usb_control_msg(dev->udev,
> +			      usb_sndctrlpipe(dev->udev, 0),
> +			      DM_WRITE_REGS,
> +			      USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
> +			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
> +	kfree(buf);
> +	if (err >= 0 && err < length)
> +		err = -EINVAL;
> + out:
> +	return err;
>  }
>  
>  static int dm_write_reg(struct usbnet *dev, u8 reg, u8 value)
> -- 
> 1.5.6.2
> 
> 
> -- 
> Bye, Peter Korsgaard
> --
> To unsubscribe from this list: send the line "unsubscribe linux-usb" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Peter Korsgaard]

dm_{read,write}() were doing USB transfers of data on stack, which isn't
allowed. Fix it by kmalloc'ing a temporary buffer.
Clean up the error handling for short transfers while we're at it.

Signed-off-by: Peter Korsgaard <jacmet-OfajU3CKLf1/SzgSGea1oA@public.gmane.org>
Acked-by: David Brownell <dbrownell-Rn4VEauK+AKRv+LV9MX5uipxlwaOVQ5f@public.gmane.org>
---

Jeff, please enqueue for 2.6.27.

 drivers/net/usb/dm9601.c |   52 +++++++++++++++++++++++++++++++++++++--------
 1 files changed, 42 insertions(+), 10 deletions(-)

diff --git a/drivers/net/usb/dm9601.c b/drivers/net/usb/dm9601.c
index f7319d3..78df2be 100644
--- a/drivers/net/usb/dm9601.c
+++ b/drivers/net/usb/dm9601.c
@@ -55,12 +55,28 @@
 
 static int dm_read(struct usbnet *dev, u8 reg, u16 length, void *data)
 {
+	void *buf;
+	int err = -ENOMEM;
+
 	devdbg(dev, "dm_read() reg=0x%02x length=%d", reg, length);
-	return usb_control_msg(dev->udev,
-			       usb_rcvctrlpipe(dev->udev, 0),
-			       DM_READ_REGS,
-			       USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
-			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
+
+	buf = kmalloc(length, GFP_KERNEL);
+	if (!buf)
+		goto out;
+
+	err = usb_control_msg(dev->udev,
+			      usb_rcvctrlpipe(dev->udev, 0),
+			      DM_READ_REGS,
+			      USB_DIR_IN | USB_TYPE_VENDOR | USB_RECIP_DEVICE,
+			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
+	if (err == length)
+		memcpy(data, buf, length);
+	else if (err >= 0)
+		err = -EINVAL;
+	kfree(buf);
+
+ out:
+	return err;
 }
 
 static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
@@ -70,12 +86,28 @@ static int dm_read_reg(struct usbnet *dev, u8 reg, u8 *value)
 
 static int dm_write(struct usbnet *dev, u8 reg, u16 length, void *data)
 {
+	void *buf = NULL;
+	int err = -ENOMEM;
+
 	devdbg(dev, "dm_write() reg=0x%02x, length=%d", reg, length);
-	return usb_control_msg(dev->udev,
-			       usb_sndctrlpipe(dev->udev, 0),
-			       DM_WRITE_REGS,
-			       USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
-			       0, reg, data, length, USB_CTRL_SET_TIMEOUT);
+
+	if (data) {
+		buf = kmalloc(length, GFP_KERNEL);
+		if (!buf)
+			goto out;
+		memcpy(buf, data, length);
+	}
+
+	err = usb_control_msg(dev->udev,
+			      usb_sndctrlpipe(dev->udev, 0),
+			      DM_WRITE_REGS,
+			      USB_DIR_OUT | USB_TYPE_VENDOR |USB_RECIP_DEVICE,
+			      0, reg, buf, length, USB_CTRL_SET_TIMEOUT);
+	kfree(buf);
+	if (err >= 0 && err < length)
+		err = -EINVAL;
+ out:
+	return err;
 }
 
 static int dm_write_reg(struct usbnet *dev, u8 reg, u8 value)
-- 
1.5.6.2


-- 
Bye, Peter Korsgaard
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH RESEND] dm9601: don't do usb transfers of data on stack

[Jeff Garzik]

Peter Korsgaard wrote:
> dm_{read,write}() were doing USB transfers of data on stack, which isn't
> allowed. Fix it by kmalloc'ing a temporary buffer.
> Clean up the error handling for short transfers while we're at it.
> 
> Signed-off-by: Peter Korsgaard <jacmet-OfajU3CKLf1/SzgSGea1oA@public.gmane.org>
> Acked-by: David Brownell <dbrownell-Rn4VEauK+AKRv+LV9MX5uipxlwaOVQ5f@public.gmane.org>
> ---
> 
> Jeff, please enqueue for 2.6.27.

applied


--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html