On Tue, 2018-11-13 at 22:57 +0000, Marc Zyngier wrote:
> Bizarrely, there is no lock taken in the irq_ack helper. This
> puts the ack callback provided by a specific platform in a awkward
> situation where there is no synchronization that would be expected
> on other callback.
>
> Introduce the required lock, giving some level of uniformity among
> callbacks.
The locking strategy here does not entirely make sense to me.
The only call path I see that can reach this function is via the
generic_handle_irq() call in dw_handle_msi_irq(). Is there another
path I missed?
If that is the only path, then how could this method be called
reentrantly on the same port? It would require that
dw_handle_msi_irq() was called reentrantly. But look at that function,
it's clearly not reentrant. It will race with itself trying to handle
the same status bits.
> @@ static void dw_pci_bottom_ack(struct irq_data *d)
> {
> struct msi_desc *msi = irq_data_get_msi_desc(d);
> struct pcie_port *pp;
> + unsigned long flags;
>
> pp = msi_desc_to_pci_sysdata(msi);
>
> + raw_spin_lock_irqsave(&pp->lock, flags);
> +
> if (pp->ops->msi_irq_ack)
> pp->ops->msi_irq_ack(d->hwirq, pp);
> +
> + raw_spin_unlock_irqrestore(&pp->lock, flags);
>
