Skip to contentSkip to conversation

net: fix uninitialised access in mii_nway_restart()

PATCH Qasim Ijaz 2 participants 4 messages Feb 18, 2025 – Mar 7, 2025

Subsystems (4)
lore
Share & tools
lore

Qasim Ijaz

author
#0 lore raw
Quoted 27 lines · Andrew Lunn
Quoted 26 lines · Qasim Ijaz

In mii_nway_restart() during the line:

	bmcr = mii->mdio_read(mii->dev, mii->phy_id, MII_BMCR);

The code attempts to call mii->mdio_read which is ch9200_mdio_read().

ch9200_mdio_read() utilises a local buffer, which is initialised with control_read():

	unsigned char buff[2];

However buff is conditionally initialised inside control_read():

	if (err == size) {
		memcpy(data, buf, size);
	}

If the condition of "err == size" is not met, then buff remains uninitialised. Once this happens the uninitialised buff is accessed and returned during ch9200_mdio_read():

	return (buff[0] | buff[1] << 8);

The problem stems from the fact that ch9200_mdio_read() ignores the return value of control_read(), leading to uinit-access of buff.

Quoted 8 lines
Quoted 2 lines

To fix this we should check the return value of control_read() and return early on error.

What about get_mac_address()?

If you find a bug, it is a good idea to look around and see if there are any more instances of the same bug. I could be wrong, but it seems like get_mac_address() suffers from the same problem?

Thank you for the feedback Andrew. I checked get_mac_address() before sending this patch and to me it looks like it does check the return value of control_read(). It accumulates the return value of each control_read() call into rd_mac_len and then checks if it not equal to what is expected (ETH_ALEN which is 6), I believe each call should return 2.

Quoted 2 lines

Andrew

#1 lore raw
Quoted 37 lines · Qasim Ijaz
Quoted 27 lines · Andrew Lunn
Quoted 26 lines · Qasim Ijaz

In mii_nway_restart() during the line:

	bmcr = mii->mdio_read(mii->dev, mii->phy_id, MII_BMCR);

The code attempts to call mii->mdio_read which is ch9200_mdio_read().

ch9200_mdio_read() utilises a local buffer, which is initialised with control_read():

	unsigned char buff[2];

However buff is conditionally initialised inside control_read():

	if (err == size) {
		memcpy(data, buf, size);
	}

If the condition of "err == size" is not met, then buff remains uninitialised. Once this happens the uninitialised buff is accessed and returned during ch9200_mdio_read():

	return (buff[0] | buff[1] << 8);

The problem stems from the fact that ch9200_mdio_read() ignores the return value of control_read(), leading to uinit-access of buff.

Quoted 8 lines
Quoted 2 lines

To fix this we should check the return value of control_read() and return early on error.

What about get_mac_address()?

If you find a bug, it is a good idea to look around and see if there are any more instances of the same bug. I could be wrong, but it seems like get_mac_address() suffers from the same problem?

Quoted 5 lines

Thank you for the feedback Andrew. I checked get_mac_address() before sending this patch and to me it looks like it does check the return value of control_read(). It accumulates the return value of each control_read() call into rd_mac_len and then checks if it not equal to what is expected (ETH_ALEN which is 6), I believe each call should return 2.

It is unlikely a real device could trigger an issue, but a USB Rubber Ducky might be able to. So the question is, are you interested in protecting against malicious devices, or just making a static analyser happy? Feel free to submit the patch as is.

Andrew

author
#2 lore raw
Quoted 44 lines · Andrew Lunn
Quoted 37 lines · Qasim Ijaz
Quoted 27 lines · Andrew Lunn
Quoted 26 lines · Qasim Ijaz

In mii_nway_restart() during the line:

	bmcr = mii->mdio_read(mii->dev, mii->phy_id, MII_BMCR);

The code attempts to call mii->mdio_read which is ch9200_mdio_read().

ch9200_mdio_read() utilises a local buffer, which is initialised with control_read():

	unsigned char buff[2];

However buff is conditionally initialised inside control_read():

	if (err == size) {
		memcpy(data, buf, size);
	}

If the condition of "err == size" is not met, then buff remains uninitialised. Once this happens the uninitialised buff is accessed and returned during ch9200_mdio_read():

	return (buff[0] | buff[1] << 8);

The problem stems from the fact that ch9200_mdio_read() ignores the return value of control_read(), leading to uinit-access of buff.

Quoted 8 lines
Quoted 2 lines

To fix this we should check the return value of control_read() and return early on error.

What about get_mac_address()?

If you find a bug, it is a good idea to look around and see if there are any more instances of the same bug. I could be wrong, but it seems like get_mac_address() suffers from the same problem?

Quoted 5 lines

Thank you for the feedback Andrew. I checked get_mac_address() before sending this patch and to me it looks like it does check the return value of control_read(). It accumulates the return value of each control_read() call into rd_mac_len and then checks if it not equal to what is expected (ETH_ALEN which is 6), I believe each call should return 2.

Quoted 6 lines

It is unlikely a real device could trigger an issue, but a USB Rubber Ducky might be able to. So the question is, are you interested in protecting against malicious devices, or just making a static analyser happy? Feel free to submit the patch as is.

Andrew

Hi Andrew,

Just following up on my patch regarding the uninitialized access fix in mii_nway_restart(). As I mentioned in my previous message, how about an approach similar to the patch for ch9200_mdio_read() for get_mac_address() where we immediately check the return value of each control_read() call and return an error if any call fails? This way we don't continue if failure occurs. If you're good with this approach, should I submit a patch v2?

Thanks,
Qasim

Quoted 1 line
#3 lore raw
Quoted 7 lines

Hi Andrew,

Just following up on my patch regarding the uninitialized access fix in mii_nway_restart(). As I mentioned in my previous message, how about an approach similar to the patch for ch9200_mdio_read() for get_mac_address() where we immediately check the return value of each control_read() call and return an error if any call fails? This way we don't continue if failure occurs. If you're good with this approach, should I submit a patch v2?

Yes, this would be good.

Thanks
Andrew