On 07/02/2013 06:28 PM, Aneesh Kumar K.V wrote:
Alexander Graf<agraf@xxxxxxx> writes:
On 07/02/2013 05:29 PM, Aneesh Kumar K.V wrote:
Alexander Graf<agraf@xxxxxxx> writes:
On 07/02/2013 07:45 AM, Aneesh Kumar K.V wrote:
From: "Aneesh Kumar K.V"<aneesh.kumar@xxxxxxxxxxxxxxxxxx>
Older version of power architecture use Real Mode Offset register and Real Mode Limit
Selector for mapping guest Real Mode Area. The guest RMA should be physically
contigous since we use the range when address translation is not enabled.
This patch switch RMA allocation code to use contigous memory allocator. The patch
also remove the the linear allocator which not used any more
Acked-by: Paul Mackerras<paulus@xxxxxxxxx>
Signed-off-by: Aneesh Kumar K.V<aneesh.kumar@xxxxxxxxxxxxxxxxxx>
---
.... snip ....
diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c
index 550f592..55c8519 100644
--- a/arch/powerpc/kvm/book3s_hv.c
+++ b/arch/powerpc/kvm/book3s_hv.c
@@ -1511,10 +1511,10 @@ static inline int lpcr_rmls(unsigned long rma_size)
static int kvm_rma_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- struct kvmppc_linear_info *ri = vma->vm_file->private_data;
struct page *page;
+ struct kvm_rma_info *ri = vma->vm_file->private_data;
- if (vmf->pgoff>= ri->npages)
+ if (vmf->pgoff>= kvm_rma_pages)
return VM_FAULT_SIGBUS;
page = pfn_to_page(ri->base_pfn + vmf->pgoff);
@@ -1536,7 +1536,7 @@ static int kvm_rma_mmap(struct file *file, struct vm_area_struct *vma)
static int kvm_rma_release(struct inode *inode, struct file *filp)
{
- struct kvmppc_linear_info *ri = filp->private_data;
+ struct kvm_rma_info *ri = filp->private_data;
kvm_release_rma(ri);
return 0;
@@ -1549,8 +1549,17 @@ static const struct file_operations kvm_rma_fops = {
long kvm_vm_ioctl_allocate_rma(struct kvm *kvm, struct kvm_allocate_rma *ret)
{
- struct kvmppc_linear_info *ri;
long fd;
+ struct kvm_rma_info *ri;
+ /*
+ * Only do this on PPC970 in HV mode
+ */
+ if (!cpu_has_feature(CPU_FTR_HVMODE) ||
+ !cpu_has_feature(CPU_FTR_ARCH_201))
+ return -EINVAL;
Is this really what we want? User space may want to use an RMA on POWER7
systems, no?
IIUC they will use virtual real mode area (VRMA) and not RMA
Then I suppose we should at least update the comment a bit further down
the patch that indicates that on POWER7 systems we do support a real
RMA. I can't really think of any reason why user space would want to use
RMA over VRMA.
where ? We have comments like
/* On POWER7, use VRMA; on PPC970, give up */
/*
- * This maintains a list of RMAs (real mode areas) for KVM guests to use.
+ * We allocate RMAs (real mode areas) for KVM guests from the KVM CMA area.
* Each RMA has to be physically contiguous and of a size that the
* hardware supports. PPC970 and POWER7 support 64MB, 128MB and 256MB,
* and other larger sizes. Since we are unlikely to be allocate that
* much physically contiguous memory after the system is up and running,
- * we preallocate a set of RMAs in early boot for KVM to use.
+ * we preallocate a set of RMAs in early boot using CMA.
+ * should be power of 2.
*/
This could be falsely interpreted as "POWER7 can use an RMA".
Alex
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
