Skip to contentSkip to conversation

cpqphp: NULL ptr deref in cpqhpc_probe

DISCUSSION Jiri Slaby 5 participants 7 messages Jun 3, 2010 – Jun 10, 2010

lore
Share & tools
lore

Jiri Slaby

author
#0 lore raw

Hi,

we have a system where there is a pci hotplug class device to be handled by cpqphp, but it is not a bridge. But in cpqhpc_probe there is:
struct pci_bus *bus;
...
bus = pdev->subordinate;
...
bus->max_bus_speed = PCI_SPEED_66MHz_PCIX;

But as it is not a bridge, subordinate is NULL and the kernel crashes.

Any idea what would be a correct fix here?

The bugzilla entry is at:
https://bugzilla.novell.com/show_bug.cgi?id=609338

thanks,

Signature · 2 lines
js
suse labs
#1 lore raw
Quoted 8 lines · Jiri Slaby

we have a system where there is a pci hotplug class device to be handled by cpqphp, but it is not a bridge. But in cpqhpc_probe there is:
struct pci_bus *bus;
...
bus = pdev->subordinate;
...
bus->max_bus_speed = PCI_SPEED_66MHz_PCIX;

Quoted 6 lines

But as it is not a bridge, subordinate is NULL and the kernel crashes.

Any idea what would be a correct fix here?

The bugzilla entry is at:
https://bugzilla.novell.com/show_bug.cgi?id=609338

I forgot to attach the device information:
00:0b.0 PCI Hot-plug controller: Compaq Computer Corporation PCI Hotplug Controller (rev 11)
Subsystem: Compaq Computer Corporation Device a2f8
Control: I/O- Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr+
Stepping- SERR+ FastB2B- DisINTx-
Status: Cap- 66MHz+ UDF- FastB2B- ParErr- DEVSEL=slow >TAbort- <TAbort-
<MAbort- >SERR- <PERR- INTx-

	Interrupt: pin A routed to IRQ 15
	Region 0: Memory at f7ad0000 (32-bit, non-prefetchable) [size=256]
#2 lore raw
In reply to Jiri Slaby
Quoted 10 lines · On Thu, 03 Jun 2010 11:24:22 +0200 Jiri Slaby <[email protected]> wrote:

Hi,

we have a system where there is a pci hotplug class device to be handled by cpqphp, but it is not a bridge. But in cpqhpc_probe there is:
struct pci_bus *bus;
...
bus = pdev->subordinate;
...
bus->max_bus_speed = PCI_SPEED_66MHz_PCIX;

Quoted 6 lines

But as it is not a bridge, subordinate is NULL and the kernel crashes.

Any idea what would be a correct fix here?

The bugzilla entry is at:
https://bugzilla.novell.com/show_bug.cgi?id=609338

I don't think we have anyone actively working on CPQHPC these days. Seems like the simple patch would be to check whether pdev->subordinate or bus exists before using it... Have you poked around for specs on this at all?

Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center
#3 lore raw
Quoted 17 lines · Jesse Barnes
Quoted 14 lines · On Thu, 03 Jun 2010 11:24:22 +0200 Jiri Slaby <[email protected]> wrote:

Hi,

we have a system where there is a pci hotplug class device to be handled by cpqphp, but it is not a bridge. But in cpqhpc_probe there is:
struct pci_bus *bus;
...
bus = pdev->subordinate;
...
bus->max_bus_speed = PCI_SPEED_66MHz_PCIX;

But as it is not a bridge, subordinate is NULL and the kernel crashes.

Any idea what would be a correct fix here?

Quoted 7 lines
Quoted 2 lines

The bugzilla entry is at:
https://bugzilla.novell.com/show_bug.cgi?id=609338

I don't think we have anyone actively working on CPQHPC these days. Seems like the simple patch would be to check whether pdev->subordinate or bus exists before using it... Have you poked around for specs on this at all?

I don't think I even have the specs anymore. I ported the code from 2.4 back in 2001 or so and got it merged. I didn't think this device was still even around in systems.

a simple check would be best.

thanks,

greg k-h

#4 lore raw
In reply to Jesse Barnes

Hi,

My hp.com address is dead. This is my new address.

Luckily, I just happened to be scanning LKML and thought "who would be crazy enough to be touching cpqphp?" :)

* Jesse Barnes <[email protected]>:

Quoted 15 lines
Quoted 14 lines · Jiri Slaby

Hi,

we have a system where there is a pci hotplug class device to be handled by cpqphp, but it is not a bridge. But in cpqhpc_probe there is:
struct pci_bus *bus;
...
bus = pdev->subordinate;
...
bus->max_bus_speed = PCI_SPEED_66MHz_PCIX;

But as it is not a bridge, subordinate is NULL and the kernel crashes.

Any idea what would be a correct fix here?

Quoted 7 lines
Quoted 2 lines

The bugzilla entry is at:
https://bugzilla.novell.com/show_bug.cgi?id=609338

I don't think we have anyone actively working on CPQHPC these days. Seems like the simple patch would be to check whether pdev->subordinate or bus exists before using it... Have you poked around for specs on this at all?

I think Greg/Jesse's suggestion is correct - just return if it's not a bridge.

I managed to find an ancient machine that actually had this hardware but never had time to work on it, and then switched jobs. As far as I know, the hardware is still there, with my old group.

But I seriously think it's just time to rip this driver out of the tree. If anyone is actually using this driver to do hotplug, I will personally buy that person a full half-barrel of Colorado's finest microbrew of choice in order to numb the pain of making poor career decisions. Hell, we can drink it together, during which time I'll hack on the driver. The code surely can't get any worse.

The occasional bug report shows that it's just more of a maintenance burden than anything else.

/ac

#5 lore raw
Quoted 7 lines · Alex Chiang

But I seriously think it's just time to rip this driver out of the tree. If anyone is actually using this driver to do hotplug, I will personally buy that person a full half-barrel of Colorado's finest microbrew of choice in order to numb the pain of making poor career decisions. Hell, we can drink it together, during which time I'll hack on the driver. The code surely can't get any worse.

A real user found this problem on their machine, so you might have to go buy that barrel. Jiri has more details on that :)

Me, I gave the machine that had this hardware in it to osuosl.org a long time ago, odds are, it's still being used by some group.

thanks,

greg k-h

#6 lore raw

* Greg KH <[email protected]>:

Quoted 11 lines · Alex Chiang
Quoted 7 lines

But I seriously think it's just time to rip this driver out of the tree. If anyone is actually using this driver to do hotplug, I will personally buy that person a full half-barrel of Colorado's finest microbrew of choice in order to numb the pain of making poor career decisions. Hell, we can drink it together, during which time I'll hack on the driver. The code surely can't get any worse.

A real user found this problem on their machine, so you might have to go buy that barrel. Jiri has more details on that :)

Well, I went and read the bz before responding, and the user did say that he wan't interested in the actual hotplug aspect of it. ;)

Quoted 3 lines

Me, I gave the machine that had this hardware in it to osuosl.org a long time ago, odds are, it's still being used by some group.

Oh, I don't doubt that someone is using the hardware to do something useful. But I do doubt that anyone's using the hotplug driver. It's been in various stages of brokenness for several years now.

/ac