Re: [v4 0/8] Add VT-d Posted-Interrupts support - IOMMU part

["Li, ZhenHua" <zhen-hual-VXdhtT5mjnY@public.gmane.org>]

Hi Feng Wu,
In my patchset, I created a new member ir_table->base_old_phys; In the 
normal kernel, everything is the same. In kdump kernel, ir_table->base 
is used for a buffer, and  ir_table->base_old_phys is the physical 
address of the tables used by the old kernel, also being used by the 
current kernel.

I did this in modify_irte():

         set_64bit(&irte->high, irte_modified->high);
+
+#ifdef CONFIG_CRASH_DUMP
+       if (is_kdump_kernel())
+               __iommu_update_old_irte(iommu, index);
+#endif
         __iommu_flush_cache(iommu, irte, sizeof(*irte));


Here the irte tables are stored in two places:
iommu->ir_table->base : It is a buffer in kdump kernel, which is the 
running kernel;
iommu->ir_table->base_old_phys : It is the irte used by the old kernel;

And function __iommu_update_old_irte is used to save the content of 
iommu->ir_table->base  to iommu->ir_table->base_old_phys. Because in 
kdump kernel, the vt-d is using ir_table->base_old_phys, not 
ir_table->base, so we need to copy the updated ir_table->base to 
ir_table->base_old_phys .


Thanks
Zhenhua

On 04/02/2015 07:28 PM, Joerg Roedel wrote:
> On Mon, Feb 02, 2015 at 04:06:56PM +0800, Feng Wu wrote:
>> VT-d Posted-Interrupts is an enhancement to CPU side Posted-Interrupt.
>> With VT-d Posted-Interrupts enabled, external interrupts from
>> direct-assigned devices can be delivered to guests without VMM
>> intervention when guest is running in non-root mode.
>>
>> You can find the VT-d Posted-Interrtups Spec. in the following URL:
>> http://www.intel.com/content/www/us/en/intelligent-systems/intel-technology/vt-directed-io-spec.html
>>
>> This series was part of http://thread.gmane.org/gmane.linux.kernel.iommu/7708. To make things clear, send out IOMMU part here.
>
> Besides the modify_irte() changes I asked for the patch-set looks good.
> I just have some concerns what these changes mean for the VT-d kdump
> improvements Zhen-Hua Li is working on. Can you please discuss the
> implications of having both patch-sets applied with him and make sure
> they work together? I think in its current form your patch-set breaks
> the kdump support patches. I added Zhen-Hua to Cc.
>
> Thanks,
>
> 	Joerg
>