On Wed, Aug 21, 2019 at 04:26:53PM +1000, Oliver O'Halloran wrote:
> When disabling virtual functions on an SR-IOV adapter we currently do not
> correctly remove the EEH state for the now-dead virtual functions. When
> removing the pci_dn that was created for the VF when SR-IOV was enabled
> we free the corresponding eeh_dev without removing it from the child device
> list of the eeh_pe that contained it. This can result in crashes due to the
> use-after-free.
> 
> Signed-off-by: Oliver O'Halloran <oohall@gmail.com>
> ---
> No Fixes: here since I'm not sure if the commit that added this actually
> introduced the bug. EEH is amazing.

Yep.

> I suspect backporting this would cause more problems than it solves since
> reliably replicating the crash required enabling memory poisoning and
> hacking a device driver to remove the PCI error handling callbacks so
> the EEH fallback path (which removes and re-probes PCI devices)
> would be used.

I gave this a quick test with some added instrumentation, and I can see
that the new code is used during VF removal and it doesn't cause any new
problems. I agree that even if it's difficult to trigger, it was
definitely a bug.

Reviewed-by: Sam Bobroff <sbobroff@linux.ibm.com>
Tested-by: Sam Bobroff <sbobroff@linux.ibm.com>
> ---
>  arch/powerpc/kernel/pci_dn.c | 15 ++++++++++++++-
>  1 file changed, 14 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/powerpc/kernel/pci_dn.c b/arch/powerpc/kernel/pci_dn.c
> index 6556b57..795c4e3 100644
> --- a/arch/powerpc/kernel/pci_dn.c
> +++ b/arch/powerpc/kernel/pci_dn.c
> @@ -244,9 +244,22 @@ void remove_dev_pci_data(struct pci_dev *pdev)
>  				continue;
>  
>  #ifdef CONFIG_EEH
> -			/* Release EEH device for the VF */
> +			/*
> +			 * Release EEH state for this VF. The PCI core
> +			 * has already torn down the pci_dev for this VF, but
> +			 * we're responsible to removing the eeh_dev since it
> +			 * has the same lifetime as the pci_dn that spawned it.
> +			 */
>  			edev = pdn_to_eeh_dev(pdn);
>  			if (edev) {
> +				/*
> +				 * We allocate pci_dn's for the totalvfs count,
> +				 * but only only the vfs that were activated
> +				 * have a configured PE.
> +				 */
> +				if (edev->pe)
> +					eeh_rmv_from_parent_pe(edev);
> +
>  				pdn->edev = NULL;
>  				kfree(edev);
>  			}
> -- 
> 2.9.5
>