Re: [RFC PATCH] PCI/DPC: Fix info->id initialization in dpc_process_error()

[Bjorn Helgaas <helgaas@kernel.org>]

On Wed, Nov 04, 2020 at 07:06:44PM -0800, Kuppuswamy, Sathyanarayanan wrote:
> On 10/31/20 3:01 AM, Saheed Olayemi Bolarinwa wrote:
> > From: "Saheed O. Bolarinwa" <refactormyself@gmail.com>
> > 
> > In the dpc_process_error() path, the error source ID is obtained
> > but not stored inside the aer_err_info object. So aer_print_error()
> > is not aware of the error source if it gets called.
> > 
> > Use the obtained valued to initialise info->id

> Is it useful for DPC case ? I don't think we set info->error_dev_num for
> DPC case right ?
> 
> if (info->id && info->error_dev_num > 1 && info->id == id)
>  726                 pci_err(dev, "  Error of this Agent is reported first\n");

It's true that we only assign to info->error_dev_num in aer.c.  That
looks like another defect.

We're passing "info" to aer_print_error() when both info->id and
info->error_dev_num are stack junk.

> > Signed-off-by: Saheed O. Bolarinwa <refactormyself@gmail.com>
> > ---
> >   drivers/pci/pcie/dpc.c | 1 +
> >   1 file changed, 1 insertion(+)
> > 
> > diff --git a/drivers/pci/pcie/dpc.c b/drivers/pci/pcie/dpc.c
> > index e05aba86a317..9f8698812939 100644
> > --- a/drivers/pci/pcie/dpc.c
> > +++ b/drivers/pci/pcie/dpc.c
> > @@ -223,6 +223,7 @@ void dpc_process_error(struct pci_dev *pdev)
> >   	else if (reason == 0 &&
> >   		 dpc_get_aer_uncorrect_severity(pdev, &info) &&
> >   		 aer_get_device_error_info(pdev, &info)) {
> > +		info.id = source;
> >   		aer_print_error(pdev, &info);
> >   		pci_aer_clear_nonfatal_status(pdev);
> >   		pci_aer_clear_fatal_status(pdev);
> > 
> 
> -- 
> Sathyanarayanan Kuppuswamy
> Linux Kernel Developer