[RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA

[Balbir Singh]

On 31/08/16 14:14, Alexey Kardashevskiy wrote:
> On 14/07/16 14:25, Balbir Singh wrote:
>>
>> From: Balbir Singh <bsingharora at gmail.com>
>> Subject: [RESEND][v2][PATCH] KVM: PPC: Book3S HV: Migrate pinned pages out of CMA
>>
>> When PCI Device pass-through is enabled via VFIO, KVM-PPC will
>> pin pages using get_user_pages_fast(). One of the downsides of
>> the pinning is that the page could be in CMA region. The CMA
>> region is used for other allocations like the hash page table.
>> Ideally we want the pinned pages to be from non CMA region.
>>
>> This patch (currently only for KVM PPC with VFIO) forcefully
>> migrates the pages out (huge pages are omitted for the moment).
>> There are more efficient ways of doing this, but that might
>> be elaborate and might impact a larger audience beyond just
>> the kvm ppc implementation.
>>
>> The magic is in new_iommu_non_cma_page() which allocates the
>> new page from a non CMA region.
>>
>> I've tested the patches lightly at my end, but there might be bugs
>> For example if after lru_add_drain(), the page is not isolated
>> is this a BUG?
>>
>> Previous discussion was at
>> http://permalink.gmane.org/gmane.linux.kernel.mm/136738
>>
>> Cc: Benjamin Herrenschmidt <benh at kernel.crashing.org>
>> Cc: Michael Ellerman <mpe at ellerman.id.au>
>> Cc: Paul Mackerras <paulus at ozlabs.org>
>> Cc: Alexey Kardashevskiy <aik at ozlabs.ru>
>>
>> Signed-off-by: Balbir Singh <bsingharora at gmail.com>
> 
> 
> 
> Acked-by: Alexey Kardashevskiy <aik at ozlabs.ru>
> 

Thanks! I tested this patch against latest mainline and here are the test results


System RAM - 64GB

VM instance 1 - size 55GB

Before patch - nr_free_cma after launch 8900
After patch - nr_free_cma after launch 39500

Balbir Singh.