Re: [PATCH v7 7/9] PCI/IDE: Add IDE establishment helpers

[<dan.j.williams@intel.com>]

Aneesh Kumar K.V wrote:
> Dan Williams <dan.j.williams@intel.com> writes:
> ...
> 
> > +DEFINE_FREE(free_stream, struct stream_index *, if (_T) free_stream_index(_T))
> > +static struct stream_index *alloc_stream_index(struct ida *ida, u8 max,
> > +					       struct stream_index *stream)
> > +{
> > +	int id;
> > +
> > +	if (!max)
> > +		return NULL;
> > +
> > +	id = ida_alloc_range(ida, 0, max - 1, GFP_KERNEL);
> > +	if (id < 0)
> > +		return NULL;
> > +
> > +	*stream = (struct stream_index) {
> > +		.ida = ida,
> > +		.stream_index = id,
> > +	};
> > +	return stream;
> > +}
> > +
> 
> We do
> 
> struct stream_index *hb_stream __free(free_stream) = alloc_stream_index(
> 		&hb->ide_stream_ida, hb->nr_ide_streams, &__stream[PCI_IDE_HB]);
> 
> and the default value for hb->nr_ide_streams is 256
> 
> void pci_ide_init_host_bridge(struct pci_host_bridge *hb)
> {
> 	hb->nr_ide_streams = 256;
> }
> 
> That overflows the u8 max argument for alloc_stream_index and results in
> a NULL return.

Ah, I missed this straggling u8. I missed this failure because the
samples/devsec/ implementation sets the host bridge pool to something
less than 256. Will fix.