[PATCH v2] intel-iommu: Quiesce devices before disabling IOMMU

[Joerg Roedel]

On Wed, Sep 18, 2013 at 03:09:01PM +0900, Takao Indoh wrote:
> +	/*
> +	 *  In the case of kdump, ioremap is needed because root-entry table
> +	 *  exists in first kernel's memory area which is not mapped in second
> +	 *  kernel
> +	 */
> +	root = (struct root_entry *)ioremap(addr, PAGE_SIZE);
> +	if (!root)
> +		return;
> +
> +	for (bus = 0; bus < ROOT_ENTRY_NR; bus++) {
> +		if (!root_present(&root[bus]))
> +			continue;
> +
> +		context = (struct context_entry *)ioremap(
> +			root[bus].val & VTD_PAGE_MASK, PAGE_SIZE);
> +		if (!context)
> +			continue;
> +
> +		for (devfn = 0; devfn < CONTEXT_ENTRY_NR; devfn++) {
> +			if (!context_present(&context[devfn]))
> +				continue;
> +
> +			dev = pci_get_domain_bus_and_slot(segment, bus, devfn);
> +			if (!dev)
> +				continue;
> +
> +			if (!pci_reset_bus(dev->bus)) /* go to next bus */
> +				break;
> +			else /* Try per-function reset */
> +				pci_reset_function(dev);
> +
> +		}
> +		iounmap(context);
> +	}
> +	iounmap(root);

I am not convinced that this is the right approach. If a device wasn't
translated by VT-d in the old kernel doesn't mean it will not be
translated in the new kernel. How about unconditionally resetting all
PCI busses and/or functions here before IOMMU initialization proceeds?


	Joerg