Hi,
The kbuild test robot has reported a build error on some configs,
pointing out that another #ifdef should be used. I've checked, the
DEBUG_LOCK_ALLOC seems to suit better here:
On Fri, 2020-12-18 at 20:39 +0300, Sergei Miroshnichenko wrote:
> This is a yet another approach to fix an old [1-2] concurrency issue,
> when:
> - two or more devices are being hot-added into a bridge which was
> initially empty;
> - a bridge with two or more devices is being hot-added;
> - the BIOS/bootloader/firmware doesn't pre-enable bridges during
> boot;
>
> ...
>
> +#ifdef CONFIG_PROVE_LOCKING
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
> +static int pci_bridge_depth(struct pci_dev *dev)
> +{
> + struct pci_dev *bridge = pci_upstream_bridge(dev);
> +
> + if (!bridge)
> + return 0;
> +
> + return 1 + pci_bridge_depth(bridge);
> +}
> +#endif /* CONFIG_PROVE_LOCKING */
+#endif /* CONFIG_DEBUG_LOCK_ALLOC */
> +
> static void pci_enable_bridge(struct pci_dev *dev)
> {
> struct pci_dev *bridge;
> int retval;
>
> + mutex_lock_nested(&dev->enable_mutex, pci_bridge_depth(dev));
> +
> bridge = pci_upstream_bridge(dev);
> if (bridge)
> pci_enable_bridge(bridge);
Is there a proper way to send a "hotfix" for a single patch of the
series of 26, without resending them all?
Best regards,
Serge
