2011-06-29 00:22:05 +00:00
|
|
|
/*
|
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License, version 2, as
|
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* Copyright 2010-2011 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/types.h>
|
|
|
|
#include <linux/string.h>
|
|
|
|
#include <linux/kvm.h>
|
|
|
|
#include <linux/kvm_host.h>
|
|
|
|
#include <linux/hugetlb.h>
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
#include <linux/module.h>
|
2011-06-29 00:22:05 +00:00
|
|
|
|
|
|
|
#include <asm/tlbflush.h>
|
|
|
|
#include <asm/kvm_ppc.h>
|
|
|
|
#include <asm/kvm_book3s.h>
|
|
|
|
#include <asm/mmu-hash64.h>
|
|
|
|
#include <asm/hvcall.h>
|
|
|
|
#include <asm/synch.h>
|
|
|
|
#include <asm/ppc-opcode.h>
|
|
|
|
|
2011-12-12 12:27:39 +00:00
|
|
|
/* Translate address of a vmalloc'd thing to a linear map address */
|
|
|
|
static void *real_vmalloc_addr(void *x)
|
|
|
|
{
|
|
|
|
unsigned long addr = (unsigned long) x;
|
|
|
|
pte_t *p;
|
|
|
|
|
|
|
|
p = find_linux_pte(swapper_pg_dir, addr);
|
|
|
|
if (!p || !pte_present(*p))
|
|
|
|
return NULL;
|
|
|
|
/* assume we don't have huge pages in vmalloc space... */
|
|
|
|
addr = (pte_pfn(*p) << PAGE_SHIFT) | (addr & ~PAGE_MASK);
|
|
|
|
return __va(addr);
|
|
|
|
}
|
2011-06-29 00:22:05 +00:00
|
|
|
|
2011-12-12 12:33:07 +00:00
|
|
|
/*
|
|
|
|
* Add this HPTE into the chain for the real page.
|
|
|
|
* Must be called with the chain locked; it unlocks the chain.
|
|
|
|
*/
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
void kvmppc_add_revmap_chain(struct kvm *kvm, struct revmap_entry *rev,
|
2011-12-12 12:33:07 +00:00
|
|
|
unsigned long *rmap, long pte_index, int realmode)
|
|
|
|
{
|
|
|
|
struct revmap_entry *head, *tail;
|
|
|
|
unsigned long i;
|
|
|
|
|
|
|
|
if (*rmap & KVMPPC_RMAP_PRESENT) {
|
|
|
|
i = *rmap & KVMPPC_RMAP_INDEX;
|
|
|
|
head = &kvm->arch.revmap[i];
|
|
|
|
if (realmode)
|
|
|
|
head = real_vmalloc_addr(head);
|
|
|
|
tail = &kvm->arch.revmap[head->back];
|
|
|
|
if (realmode)
|
|
|
|
tail = real_vmalloc_addr(tail);
|
|
|
|
rev->forw = i;
|
|
|
|
rev->back = head->back;
|
|
|
|
tail->forw = pte_index;
|
|
|
|
head->back = pte_index;
|
|
|
|
} else {
|
|
|
|
rev->forw = rev->back = pte_index;
|
|
|
|
i = pte_index;
|
|
|
|
}
|
|
|
|
smp_wmb();
|
|
|
|
*rmap = i | KVMPPC_RMAP_REFERENCED | KVMPPC_RMAP_PRESENT; /* unlock */
|
|
|
|
}
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
EXPORT_SYMBOL_GPL(kvmppc_add_revmap_chain);
|
2011-12-12 12:33:07 +00:00
|
|
|
|
|
|
|
/* Remove this HPTE from the chain for a real page */
|
|
|
|
static void remove_revmap_chain(struct kvm *kvm, long pte_index,
|
2011-12-15 02:02:02 +00:00
|
|
|
struct revmap_entry *rev,
|
|
|
|
unsigned long hpte_v, unsigned long hpte_r)
|
2011-12-12 12:33:07 +00:00
|
|
|
{
|
2011-12-15 02:02:02 +00:00
|
|
|
struct revmap_entry *next, *prev;
|
2011-12-12 12:33:07 +00:00
|
|
|
unsigned long gfn, ptel, head;
|
|
|
|
struct kvm_memory_slot *memslot;
|
|
|
|
unsigned long *rmap;
|
2011-12-15 02:02:02 +00:00
|
|
|
unsigned long rcbits;
|
2011-12-12 12:33:07 +00:00
|
|
|
|
2011-12-15 02:02:02 +00:00
|
|
|
rcbits = hpte_r & (HPTE_R_R | HPTE_R_C);
|
|
|
|
ptel = rev->guest_rpte |= rcbits;
|
2011-12-12 12:33:07 +00:00
|
|
|
gfn = hpte_rpn(ptel, hpte_page_size(hpte_v, ptel));
|
2012-01-12 20:09:51 +00:00
|
|
|
memslot = __gfn_to_memslot(kvm_memslots(kvm), gfn);
|
2011-12-12 12:33:07 +00:00
|
|
|
if (!memslot || (memslot->flags & KVM_MEMSLOT_INVALID))
|
|
|
|
return;
|
|
|
|
|
2012-08-01 09:03:28 +00:00
|
|
|
rmap = real_vmalloc_addr(&memslot->arch.rmap[gfn - memslot->base_gfn]);
|
2011-12-12 12:33:07 +00:00
|
|
|
lock_rmap(rmap);
|
|
|
|
|
|
|
|
head = *rmap & KVMPPC_RMAP_INDEX;
|
|
|
|
next = real_vmalloc_addr(&kvm->arch.revmap[rev->forw]);
|
|
|
|
prev = real_vmalloc_addr(&kvm->arch.revmap[rev->back]);
|
|
|
|
next->back = rev->back;
|
|
|
|
prev->forw = rev->forw;
|
|
|
|
if (head == pte_index) {
|
|
|
|
head = rev->forw;
|
|
|
|
if (head == pte_index)
|
|
|
|
*rmap &= ~(KVMPPC_RMAP_PRESENT | KVMPPC_RMAP_INDEX);
|
|
|
|
else
|
|
|
|
*rmap = (*rmap & ~KVMPPC_RMAP_INDEX) | head;
|
|
|
|
}
|
2011-12-15 02:02:02 +00:00
|
|
|
*rmap |= rcbits << KVMPPC_RMAP_RC_SHIFT;
|
2011-12-12 12:33:07 +00:00
|
|
|
unlock_rmap(rmap);
|
|
|
|
}
|
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
static pte_t lookup_linux_pte(struct kvm_vcpu *vcpu, unsigned long hva,
|
2011-12-12 12:38:51 +00:00
|
|
|
int writing, unsigned long *pte_sizep)
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
{
|
|
|
|
pte_t *ptep;
|
|
|
|
unsigned long ps = *pte_sizep;
|
|
|
|
unsigned int shift;
|
|
|
|
|
|
|
|
ptep = find_linux_pte_or_hugepte(vcpu->arch.pgdir, hva, &shift);
|
|
|
|
if (!ptep)
|
|
|
|
return __pte(0);
|
|
|
|
if (shift)
|
|
|
|
*pte_sizep = 1ul << shift;
|
|
|
|
else
|
|
|
|
*pte_sizep = PAGE_SIZE;
|
|
|
|
if (ps > *pte_sizep)
|
|
|
|
return __pte(0);
|
|
|
|
if (!pte_present(*ptep))
|
|
|
|
return __pte(0);
|
2011-12-12 12:38:51 +00:00
|
|
|
return kvmppc_read_update_linux_pte(ptep, writing);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
}
|
|
|
|
|
2011-12-15 02:01:10 +00:00
|
|
|
static inline void unlock_hpte(unsigned long *hpte, unsigned long hpte_v)
|
|
|
|
{
|
|
|
|
asm volatile(PPC_RELEASE_BARRIER "" : : : "memory");
|
|
|
|
hpte[0] = hpte_v;
|
|
|
|
}
|
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
long kvmppc_h_enter(struct kvm_vcpu *vcpu, unsigned long flags,
|
|
|
|
long pte_index, unsigned long pteh, unsigned long ptel)
|
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
unsigned long i, pa, gpa, gfn, psize;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
unsigned long slot_fn, hva;
|
2011-06-29 00:22:05 +00:00
|
|
|
unsigned long *hpte;
|
2011-12-12 12:27:39 +00:00
|
|
|
struct revmap_entry *rev;
|
|
|
|
unsigned long g_ptel = ptel;
|
2011-12-12 12:28:21 +00:00
|
|
|
struct kvm_memory_slot *memslot;
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
unsigned long *physp, pte_size;
|
2011-12-12 12:32:27 +00:00
|
|
|
unsigned long is_io;
|
2011-12-12 12:33:07 +00:00
|
|
|
unsigned long *rmap;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
pte_t pte;
|
2011-12-12 12:38:51 +00:00
|
|
|
unsigned int writing;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
unsigned long mmu_seq;
|
2011-12-15 02:02:02 +00:00
|
|
|
unsigned long rcbits;
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
bool realmode = vcpu->arch.vcore->vcore_state == VCORE_RUNNING;
|
|
|
|
|
|
|
|
psize = hpte_page_size(pteh, ptel);
|
|
|
|
if (!psize)
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_PARAMETER;
|
2011-12-12 12:38:51 +00:00
|
|
|
writing = hpte_is_writable(ptel);
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
pteh &= ~(HPTE_V_HVLOCK | HPTE_V_ABSENT | HPTE_V_VALID);
|
2011-12-12 12:28:21 +00:00
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
/* used later to detect if we might have been invalidated */
|
|
|
|
mmu_seq = kvm->mmu_notifier_seq;
|
|
|
|
smp_rmb();
|
|
|
|
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
/* Find the memslot (if any) for this address */
|
|
|
|
gpa = (ptel & HPTE_R_RPN) & ~(psize - 1);
|
|
|
|
gfn = gpa >> PAGE_SHIFT;
|
2012-01-12 20:09:51 +00:00
|
|
|
memslot = __gfn_to_memslot(kvm_memslots(kvm), gfn);
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
pa = 0;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
is_io = ~0ul;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
rmap = NULL;
|
|
|
|
if (!(memslot && !(memslot->flags & KVM_MEMSLOT_INVALID))) {
|
|
|
|
/* PPC970 can't do emulated MMIO */
|
|
|
|
if (!cpu_has_feature(CPU_FTR_ARCH_206))
|
|
|
|
return H_PARAMETER;
|
|
|
|
/* Emulated MMIO - mark this with key=31 */
|
|
|
|
pteh |= HPTE_V_ABSENT;
|
|
|
|
ptel |= HPTE_R_KEY_HI | HPTE_R_KEY_LO;
|
|
|
|
goto do_insert;
|
|
|
|
}
|
2011-12-12 12:31:41 +00:00
|
|
|
|
|
|
|
/* Check if the requested page fits entirely in the memslot. */
|
|
|
|
if (!slot_is_aligned(memslot, psize))
|
|
|
|
return H_PARAMETER;
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
slot_fn = gfn - memslot->base_gfn;
|
2012-08-01 09:03:28 +00:00
|
|
|
rmap = &memslot->arch.rmap[slot_fn];
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
if (!kvm->arch.using_mmu_notifiers) {
|
|
|
|
physp = kvm->arch.slot_phys[memslot->id];
|
|
|
|
if (!physp)
|
|
|
|
return H_PARAMETER;
|
|
|
|
physp += slot_fn;
|
|
|
|
if (realmode)
|
|
|
|
physp = real_vmalloc_addr(physp);
|
|
|
|
pa = *physp;
|
|
|
|
if (!pa)
|
|
|
|
return H_TOO_HARD;
|
|
|
|
is_io = pa & (HPTE_R_I | HPTE_R_W);
|
|
|
|
pte_size = PAGE_SIZE << (pa & KVMPPC_PAGE_ORDER_MASK);
|
|
|
|
pa &= PAGE_MASK;
|
|
|
|
} else {
|
|
|
|
/* Translate to host virtual address */
|
2012-08-24 08:50:28 +00:00
|
|
|
hva = __gfn_to_hva_memslot(memslot, gfn);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
|
|
|
|
/* Look up the Linux PTE for the backing page */
|
|
|
|
pte_size = psize;
|
2011-12-12 12:38:51 +00:00
|
|
|
pte = lookup_linux_pte(vcpu, hva, writing, &pte_size);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
if (pte_present(pte)) {
|
2011-12-12 12:38:51 +00:00
|
|
|
if (writing && !pte_write(pte))
|
|
|
|
/* make the actual HPTE be read-only */
|
|
|
|
ptel = hpte_make_readonly(ptel);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
is_io = hpte_cache_bits(pte_val(pte));
|
|
|
|
pa = pte_pfn(pte) << PAGE_SHIFT;
|
|
|
|
}
|
|
|
|
}
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
if (pte_size < psize)
|
|
|
|
return H_PARAMETER;
|
|
|
|
if (pa && pte_size > psize)
|
|
|
|
pa |= gpa & (pte_size - 1);
|
|
|
|
|
|
|
|
ptel &= ~(HPTE_R_PP0 - psize);
|
|
|
|
ptel |= pa;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
|
|
|
|
if (pa)
|
|
|
|
pteh |= HPTE_V_VALID;
|
|
|
|
else
|
|
|
|
pteh |= HPTE_V_ABSENT;
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
/* Check WIMG */
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
if (is_io != ~0ul && !hpte_cache_flags_ok(ptel, is_io)) {
|
2011-12-12 12:32:27 +00:00
|
|
|
if (is_io)
|
|
|
|
return H_PARAMETER;
|
|
|
|
/*
|
|
|
|
* Allow guest to map emulated device memory as
|
|
|
|
* uncacheable, but actually make it cacheable.
|
|
|
|
*/
|
|
|
|
ptel &= ~(HPTE_R_W|HPTE_R_I|HPTE_R_G);
|
|
|
|
ptel |= HPTE_R_M;
|
|
|
|
}
|
2011-12-12 12:30:16 +00:00
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
/* Find and lock the HPTEG slot to use */
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
do_insert:
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
if (pte_index >= kvm->arch.hpt_npte)
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_PARAMETER;
|
|
|
|
if (likely((flags & H_EXACT) == 0)) {
|
|
|
|
pte_index &= ~7UL;
|
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
|
2011-12-12 12:30:16 +00:00
|
|
|
for (i = 0; i < 8; ++i) {
|
2011-06-29 00:22:05 +00:00
|
|
|
if ((*hpte & HPTE_V_VALID) == 0 &&
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
try_lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID |
|
|
|
|
HPTE_V_ABSENT))
|
2011-06-29 00:22:05 +00:00
|
|
|
break;
|
|
|
|
hpte += 2;
|
|
|
|
}
|
2011-12-12 12:30:16 +00:00
|
|
|
if (i == 8) {
|
|
|
|
/*
|
|
|
|
* Since try_lock_hpte doesn't retry (not even stdcx.
|
|
|
|
* failures), it could be that there is a free slot
|
|
|
|
* but we transiently failed to lock it. Try again,
|
|
|
|
* actually locking each slot and checking it.
|
|
|
|
*/
|
|
|
|
hpte -= 16;
|
|
|
|
for (i = 0; i < 8; ++i) {
|
|
|
|
while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
|
|
|
|
cpu_relax();
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (!(*hpte & (HPTE_V_VALID | HPTE_V_ABSENT)))
|
2011-12-12 12:30:16 +00:00
|
|
|
break;
|
|
|
|
*hpte &= ~HPTE_V_HVLOCK;
|
|
|
|
hpte += 2;
|
|
|
|
}
|
|
|
|
if (i == 8)
|
|
|
|
return H_PTEG_FULL;
|
|
|
|
}
|
2011-12-12 12:27:39 +00:00
|
|
|
pte_index += i;
|
2011-06-29 00:22:05 +00:00
|
|
|
} else {
|
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (!try_lock_hpte(hpte, HPTE_V_HVLOCK | HPTE_V_VALID |
|
|
|
|
HPTE_V_ABSENT)) {
|
2011-12-12 12:30:16 +00:00
|
|
|
/* Lock the slot and check again */
|
|
|
|
while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
|
|
|
|
cpu_relax();
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (*hpte & (HPTE_V_VALID | HPTE_V_ABSENT)) {
|
2011-12-12 12:30:16 +00:00
|
|
|
*hpte &= ~HPTE_V_HVLOCK;
|
|
|
|
return H_PTEG_FULL;
|
|
|
|
}
|
|
|
|
}
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-12 12:27:39 +00:00
|
|
|
|
|
|
|
/* Save away the guest's idea of the second HPTE dword */
|
2011-12-12 12:33:07 +00:00
|
|
|
rev = &kvm->arch.revmap[pte_index];
|
|
|
|
if (realmode)
|
|
|
|
rev = real_vmalloc_addr(rev);
|
2011-12-12 12:27:39 +00:00
|
|
|
if (rev)
|
|
|
|
rev->guest_rpte = g_ptel;
|
2011-12-12 12:33:07 +00:00
|
|
|
|
|
|
|
/* Link HPTE into reverse-map chain */
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (pteh & HPTE_V_VALID) {
|
|
|
|
if (realmode)
|
|
|
|
rmap = real_vmalloc_addr(rmap);
|
|
|
|
lock_rmap(rmap);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
/* Check for pending invalidations under the rmap chain lock */
|
|
|
|
if (kvm->arch.using_mmu_notifiers &&
|
|
|
|
mmu_notifier_retry(vcpu, mmu_seq)) {
|
|
|
|
/* inval in progress, write a non-present HPTE */
|
|
|
|
pteh |= HPTE_V_ABSENT;
|
|
|
|
pteh &= ~HPTE_V_VALID;
|
|
|
|
unlock_rmap(rmap);
|
|
|
|
} else {
|
|
|
|
kvmppc_add_revmap_chain(kvm, rev, rmap, pte_index,
|
|
|
|
realmode);
|
2011-12-15 02:02:02 +00:00
|
|
|
/* Only set R/C in real HPTE if already set in *rmap */
|
|
|
|
rcbits = *rmap >> KVMPPC_RMAP_RC_SHIFT;
|
|
|
|
ptel &= rcbits | ~(HPTE_R_R | HPTE_R_C);
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
}
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
}
|
2011-12-12 12:33:07 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
hpte[1] = ptel;
|
2011-12-12 12:33:07 +00:00
|
|
|
|
|
|
|
/* Write the first HPTE dword, unlocking the HPTE and making it valid */
|
2011-06-29 00:22:05 +00:00
|
|
|
eieio();
|
|
|
|
hpte[0] = pteh;
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
2011-12-12 12:33:07 +00:00
|
|
|
|
2011-12-12 12:27:39 +00:00
|
|
|
vcpu->arch.gpr[4] = pte_index;
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_SUCCESS;
|
|
|
|
}
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
EXPORT_SYMBOL_GPL(kvmppc_h_enter);
|
2011-06-29 00:22:05 +00:00
|
|
|
|
|
|
|
#define LOCK_TOKEN (*(u32 *)(&get_paca()->lock_token))
|
|
|
|
|
|
|
|
static inline int try_lock_tlbie(unsigned int *lock)
|
|
|
|
{
|
|
|
|
unsigned int tmp, old;
|
|
|
|
unsigned int token = LOCK_TOKEN;
|
|
|
|
|
|
|
|
asm volatile("1:lwarx %1,0,%2\n"
|
|
|
|
" cmpwi cr0,%1,0\n"
|
|
|
|
" bne 2f\n"
|
|
|
|
" stwcx. %3,0,%2\n"
|
|
|
|
" bne- 1b\n"
|
|
|
|
" isync\n"
|
|
|
|
"2:"
|
|
|
|
: "=&r" (tmp), "=&r" (old)
|
|
|
|
: "r" (lock), "r" (token)
|
|
|
|
: "cc", "memory");
|
|
|
|
return old == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
long kvmppc_h_remove(struct kvm_vcpu *vcpu, unsigned long flags,
|
|
|
|
unsigned long pte_index, unsigned long avpn,
|
|
|
|
unsigned long va)
|
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
|
|
|
unsigned long *hpte;
|
|
|
|
unsigned long v, r, rb;
|
2011-12-15 02:01:10 +00:00
|
|
|
struct revmap_entry *rev;
|
2011-06-29 00:22:05 +00:00
|
|
|
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
if (pte_index >= kvm->arch.hpt_npte)
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_PARAMETER;
|
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
|
2011-12-12 12:30:16 +00:00
|
|
|
while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
|
2011-06-29 00:22:05 +00:00
|
|
|
cpu_relax();
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if ((hpte[0] & (HPTE_V_ABSENT | HPTE_V_VALID)) == 0 ||
|
2011-06-29 00:22:05 +00:00
|
|
|
((flags & H_AVPN) && (hpte[0] & ~0x7fUL) != avpn) ||
|
|
|
|
((flags & H_ANDCOND) && (hpte[0] & avpn) != 0)) {
|
|
|
|
hpte[0] &= ~HPTE_V_HVLOCK;
|
|
|
|
return H_NOT_FOUND;
|
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
|
|
|
rev = real_vmalloc_addr(&kvm->arch.revmap[pte_index]);
|
|
|
|
v = hpte[0] & ~HPTE_V_HVLOCK;
|
|
|
|
if (v & HPTE_V_VALID) {
|
|
|
|
hpte[0] &= ~HPTE_V_VALID;
|
|
|
|
rb = compute_tlbie_rb(v, hpte[1], pte_index);
|
|
|
|
if (!(flags & H_LOCAL) && atomic_read(&kvm->online_vcpus) > 1) {
|
|
|
|
while (!try_lock_tlbie(&kvm->arch.tlbie_lock))
|
|
|
|
cpu_relax();
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
asm volatile(PPC_TLBIE(%1,%0)"; eieio; tlbsync"
|
|
|
|
: : "r" (rb), "r" (kvm->arch.lpid));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
kvm->arch.tlbie_lock = 0;
|
|
|
|
} else {
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
asm volatile("tlbiel %0" : : "r" (rb));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
}
|
2011-12-15 02:02:02 +00:00
|
|
|
/* Read PTE low word after tlbie to get final R/C values */
|
|
|
|
remove_revmap_chain(kvm, pte_index, rev, v, hpte[1]);
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
r = rev->guest_rpte;
|
|
|
|
unlock_hpte(hpte, 0);
|
|
|
|
|
|
|
|
vcpu->arch.gpr[4] = v;
|
|
|
|
vcpu->arch.gpr[5] = r;
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
long kvmppc_h_bulk_remove(struct kvm_vcpu *vcpu)
|
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
|
|
|
unsigned long *args = &vcpu->arch.gpr[4];
|
2011-12-15 02:01:10 +00:00
|
|
|
unsigned long *hp, *hptes[4], tlbrb[4];
|
|
|
|
long int i, j, k, n, found, indexes[4];
|
|
|
|
unsigned long flags, req, pte_index, rcbits;
|
2011-06-29 00:22:05 +00:00
|
|
|
long int local = 0;
|
|
|
|
long int ret = H_SUCCESS;
|
2011-12-15 02:01:10 +00:00
|
|
|
struct revmap_entry *rev, *revs[4];
|
2011-06-29 00:22:05 +00:00
|
|
|
|
|
|
|
if (atomic_read(&kvm->online_vcpus) == 1)
|
|
|
|
local = 1;
|
2011-12-15 02:01:10 +00:00
|
|
|
for (i = 0; i < 4 && ret == H_SUCCESS; ) {
|
|
|
|
n = 0;
|
|
|
|
for (; i < 4; ++i) {
|
|
|
|
j = i * 2;
|
|
|
|
pte_index = args[j];
|
|
|
|
flags = pte_index >> 56;
|
|
|
|
pte_index &= ((1ul << 56) - 1);
|
|
|
|
req = flags >> 6;
|
|
|
|
flags &= 3;
|
|
|
|
if (req == 3) { /* no more requests */
|
|
|
|
i = 4;
|
2011-06-29 00:22:05 +00:00
|
|
|
break;
|
2011-12-15 02:01:10 +00:00
|
|
|
}
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
if (req != 1 || flags == 3 ||
|
|
|
|
pte_index >= kvm->arch.hpt_npte) {
|
2011-12-15 02:01:10 +00:00
|
|
|
/* parameter error */
|
|
|
|
args[j] = ((0xa0 | flags) << 56) + pte_index;
|
|
|
|
ret = H_PARAMETER;
|
2011-06-29 00:22:05 +00:00
|
|
|
break;
|
2011-12-15 02:01:10 +00:00
|
|
|
}
|
|
|
|
hp = (unsigned long *)
|
|
|
|
(kvm->arch.hpt_virt + (pte_index << 4));
|
|
|
|
/* to avoid deadlock, don't spin except for first */
|
|
|
|
if (!try_lock_hpte(hp, HPTE_V_HVLOCK)) {
|
|
|
|
if (n)
|
|
|
|
break;
|
|
|
|
while (!try_lock_hpte(hp, HPTE_V_HVLOCK))
|
|
|
|
cpu_relax();
|
|
|
|
}
|
|
|
|
found = 0;
|
|
|
|
if (hp[0] & (HPTE_V_ABSENT | HPTE_V_VALID)) {
|
|
|
|
switch (flags & 3) {
|
|
|
|
case 0: /* absolute */
|
2011-06-29 00:22:05 +00:00
|
|
|
found = 1;
|
2011-12-15 02:01:10 +00:00
|
|
|
break;
|
|
|
|
case 1: /* andcond */
|
|
|
|
if (!(hp[0] & args[j + 1]))
|
|
|
|
found = 1;
|
|
|
|
break;
|
|
|
|
case 2: /* AVPN */
|
|
|
|
if ((hp[0] & ~0x7fUL) == args[j + 1])
|
|
|
|
found = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!found) {
|
|
|
|
hp[0] &= ~HPTE_V_HVLOCK;
|
|
|
|
args[j] = ((0x90 | flags) << 56) + pte_index;
|
|
|
|
continue;
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
|
|
|
args[j] = ((0x80 | flags) << 56) + pte_index;
|
|
|
|
rev = real_vmalloc_addr(&kvm->arch.revmap[pte_index]);
|
|
|
|
|
2011-12-15 02:02:02 +00:00
|
|
|
if (!(hp[0] & HPTE_V_VALID)) {
|
|
|
|
/* insert R and C bits from PTE */
|
|
|
|
rcbits = rev->guest_rpte & (HPTE_R_R|HPTE_R_C);
|
|
|
|
args[j] |= rcbits << (56 - 5);
|
2012-05-09 23:49:24 +00:00
|
|
|
hp[0] = 0;
|
2011-12-15 02:01:10 +00:00
|
|
|
continue;
|
2011-12-15 02:02:02 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
|
|
|
hp[0] &= ~HPTE_V_VALID; /* leave it locked */
|
|
|
|
tlbrb[n] = compute_tlbie_rb(hp[0], hp[1], pte_index);
|
|
|
|
indexes[n] = j;
|
|
|
|
hptes[n] = hp;
|
|
|
|
revs[n] = rev;
|
|
|
|
++n;
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
|
|
|
if (!n)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/* Now that we've collected a batch, do the tlbies */
|
|
|
|
if (!local) {
|
|
|
|
while(!try_lock_tlbie(&kvm->arch.tlbie_lock))
|
|
|
|
cpu_relax();
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
for (k = 0; k < n; ++k)
|
|
|
|
asm volatile(PPC_TLBIE(%1,%0) : :
|
|
|
|
"r" (tlbrb[k]),
|
|
|
|
"r" (kvm->arch.lpid));
|
|
|
|
asm volatile("eieio; tlbsync; ptesync" : : : "memory");
|
|
|
|
kvm->arch.tlbie_lock = 0;
|
|
|
|
} else {
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
for (k = 0; k < n; ++k)
|
|
|
|
asm volatile("tlbiel %0" : : "r" (tlbrb[k]));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
2011-12-15 02:02:02 +00:00
|
|
|
/* Read PTE low words after tlbie to get final R/C values */
|
2011-12-15 02:01:10 +00:00
|
|
|
for (k = 0; k < n; ++k) {
|
|
|
|
j = indexes[k];
|
|
|
|
pte_index = args[j] & ((1ul << 56) - 1);
|
|
|
|
hp = hptes[k];
|
|
|
|
rev = revs[k];
|
2011-12-15 02:02:02 +00:00
|
|
|
remove_revmap_chain(kvm, pte_index, rev, hp[0], hp[1]);
|
|
|
|
rcbits = rev->guest_rpte & (HPTE_R_R|HPTE_R_C);
|
|
|
|
args[j] |= rcbits << (56 - 5);
|
|
|
|
hp[0] = 0;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
}
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
2011-12-15 02:01:10 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
long kvmppc_h_protect(struct kvm_vcpu *vcpu, unsigned long flags,
|
|
|
|
unsigned long pte_index, unsigned long avpn,
|
|
|
|
unsigned long va)
|
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
|
|
|
unsigned long *hpte;
|
2011-12-12 12:27:39 +00:00
|
|
|
struct revmap_entry *rev;
|
|
|
|
unsigned long v, r, rb, mask, bits;
|
2011-06-29 00:22:05 +00:00
|
|
|
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
if (pte_index >= kvm->arch.hpt_npte)
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_PARAMETER;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
|
2011-12-12 12:30:16 +00:00
|
|
|
while (!try_lock_hpte(hpte, HPTE_V_HVLOCK))
|
2011-06-29 00:22:05 +00:00
|
|
|
cpu_relax();
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if ((hpte[0] & (HPTE_V_ABSENT | HPTE_V_VALID)) == 0 ||
|
2011-06-29 00:22:05 +00:00
|
|
|
((flags & H_AVPN) && (hpte[0] & ~0x7fUL) != avpn)) {
|
|
|
|
hpte[0] &= ~HPTE_V_HVLOCK;
|
|
|
|
return H_NOT_FOUND;
|
|
|
|
}
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
if (atomic_read(&kvm->online_vcpus) == 1)
|
|
|
|
flags |= H_LOCAL;
|
|
|
|
v = hpte[0];
|
2011-12-12 12:27:39 +00:00
|
|
|
bits = (flags << 55) & HPTE_R_PP0;
|
|
|
|
bits |= (flags << 48) & HPTE_R_KEY_HI;
|
|
|
|
bits |= flags & (HPTE_R_PP | HPTE_R_N | HPTE_R_KEY_LO);
|
|
|
|
|
|
|
|
/* Update guest view of 2nd HPTE dword */
|
|
|
|
mask = HPTE_R_PP0 | HPTE_R_PP | HPTE_R_N |
|
|
|
|
HPTE_R_KEY_HI | HPTE_R_KEY_LO;
|
|
|
|
rev = real_vmalloc_addr(&kvm->arch.revmap[pte_index]);
|
|
|
|
if (rev) {
|
|
|
|
r = (rev->guest_rpte & ~mask) | bits;
|
|
|
|
rev->guest_rpte = r;
|
|
|
|
}
|
|
|
|
r = (hpte[1] & ~mask) | bits;
|
|
|
|
|
|
|
|
/* Update HPTE */
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (v & HPTE_V_VALID) {
|
|
|
|
rb = compute_tlbie_rb(v, r, pte_index);
|
|
|
|
hpte[0] = v & ~HPTE_V_VALID;
|
|
|
|
if (!(flags & H_LOCAL)) {
|
|
|
|
while(!try_lock_tlbie(&kvm->arch.tlbie_lock))
|
|
|
|
cpu_relax();
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
asm volatile(PPC_TLBIE(%1,%0)"; eieio; tlbsync"
|
|
|
|
: : "r" (rb), "r" (kvm->arch.lpid));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
kvm->arch.tlbie_lock = 0;
|
|
|
|
} else {
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
asm volatile("tlbiel %0" : : "r" (rb));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
}
|
2011-06-29 00:22:05 +00:00
|
|
|
}
|
|
|
|
hpte[1] = r;
|
|
|
|
eieio();
|
|
|
|
hpte[0] = v & ~HPTE_V_HVLOCK;
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
return H_SUCCESS;
|
|
|
|
}
|
|
|
|
|
|
|
|
long kvmppc_h_read(struct kvm_vcpu *vcpu, unsigned long flags,
|
|
|
|
unsigned long pte_index)
|
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
unsigned long *hpte, v, r;
|
2011-06-29 00:22:05 +00:00
|
|
|
int i, n = 1;
|
2011-12-12 12:27:39 +00:00
|
|
|
struct revmap_entry *rev = NULL;
|
2011-06-29 00:22:05 +00:00
|
|
|
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
if (pte_index >= kvm->arch.hpt_npte)
|
2011-06-29 00:22:05 +00:00
|
|
|
return H_PARAMETER;
|
|
|
|
if (flags & H_READ_4) {
|
|
|
|
pte_index &= ~3;
|
|
|
|
n = 4;
|
|
|
|
}
|
2011-12-15 02:02:02 +00:00
|
|
|
rev = real_vmalloc_addr(&kvm->arch.revmap[pte_index]);
|
2011-06-29 00:22:05 +00:00
|
|
|
for (i = 0; i < n; ++i, ++pte_index) {
|
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (pte_index << 4));
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
v = hpte[0] & ~HPTE_V_HVLOCK;
|
2011-06-29 00:22:05 +00:00
|
|
|
r = hpte[1];
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
if (v & HPTE_V_ABSENT) {
|
|
|
|
v &= ~HPTE_V_ABSENT;
|
|
|
|
v |= HPTE_V_VALID;
|
|
|
|
}
|
2011-12-15 02:02:02 +00:00
|
|
|
if (v & HPTE_V_VALID)
|
|
|
|
r = rev[i].guest_rpte | (r & (HPTE_R_R | HPTE_R_C));
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
vcpu->arch.gpr[4 + i * 2] = v;
|
2011-06-29 00:22:05 +00:00
|
|
|
vcpu->arch.gpr[5 + i * 2] = r;
|
|
|
|
}
|
|
|
|
return H_SUCCESS;
|
|
|
|
}
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
void kvmppc_invalidate_hpte(struct kvm *kvm, unsigned long *hptep,
|
|
|
|
unsigned long pte_index)
|
|
|
|
{
|
|
|
|
unsigned long rb;
|
|
|
|
|
|
|
|
hptep[0] &= ~HPTE_V_VALID;
|
|
|
|
rb = compute_tlbie_rb(hptep[0], hptep[1], pte_index);
|
|
|
|
while (!try_lock_tlbie(&kvm->arch.tlbie_lock))
|
|
|
|
cpu_relax();
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
asm volatile(PPC_TLBIE(%1,%0)"; eieio; tlbsync"
|
|
|
|
: : "r" (rb), "r" (kvm->arch.lpid));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
kvm->arch.tlbie_lock = 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kvmppc_invalidate_hpte);
|
|
|
|
|
2011-12-15 02:02:47 +00:00
|
|
|
void kvmppc_clear_ref_hpte(struct kvm *kvm, unsigned long *hptep,
|
|
|
|
unsigned long pte_index)
|
|
|
|
{
|
|
|
|
unsigned long rb;
|
|
|
|
unsigned char rbyte;
|
|
|
|
|
|
|
|
rb = compute_tlbie_rb(hptep[0], hptep[1], pte_index);
|
|
|
|
rbyte = (hptep[1] & ~HPTE_R_R) >> 8;
|
|
|
|
/* modify only the second-last byte, which contains the ref bit */
|
|
|
|
*((char *)hptep + 14) = rbyte;
|
|
|
|
while (!try_lock_tlbie(&kvm->arch.tlbie_lock))
|
|
|
|
cpu_relax();
|
|
|
|
asm volatile(PPC_TLBIE(%1,%0)"; eieio; tlbsync"
|
|
|
|
: : "r" (rb), "r" (kvm->arch.lpid));
|
|
|
|
asm volatile("ptesync" : : : "memory");
|
|
|
|
kvm->arch.tlbie_lock = 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kvmppc_clear_ref_hpte);
|
|
|
|
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
static int slb_base_page_shift[4] = {
|
|
|
|
24, /* 16M */
|
|
|
|
16, /* 64k */
|
|
|
|
34, /* 16G */
|
|
|
|
20, /* 1M, unsupported */
|
|
|
|
};
|
|
|
|
|
|
|
|
long kvmppc_hv_find_lock_hpte(struct kvm *kvm, gva_t eaddr, unsigned long slb_v,
|
|
|
|
unsigned long valid)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
|
|
|
unsigned int pshift;
|
|
|
|
unsigned long somask;
|
|
|
|
unsigned long vsid, hash;
|
|
|
|
unsigned long avpn;
|
|
|
|
unsigned long *hpte;
|
|
|
|
unsigned long mask, val;
|
|
|
|
unsigned long v, r;
|
|
|
|
|
|
|
|
/* Get page shift, work out hash and AVPN etc. */
|
|
|
|
mask = SLB_VSID_B | HPTE_V_AVPN | HPTE_V_SECONDARY;
|
|
|
|
val = 0;
|
|
|
|
pshift = 12;
|
|
|
|
if (slb_v & SLB_VSID_L) {
|
|
|
|
mask |= HPTE_V_LARGE;
|
|
|
|
val |= HPTE_V_LARGE;
|
|
|
|
pshift = slb_base_page_shift[(slb_v & SLB_VSID_LP) >> 4];
|
|
|
|
}
|
|
|
|
if (slb_v & SLB_VSID_B_1T) {
|
|
|
|
somask = (1UL << 40) - 1;
|
|
|
|
vsid = (slb_v & ~SLB_VSID_B) >> SLB_VSID_SHIFT_1T;
|
|
|
|
vsid ^= vsid << 25;
|
|
|
|
} else {
|
|
|
|
somask = (1UL << 28) - 1;
|
|
|
|
vsid = (slb_v & ~SLB_VSID_B) >> SLB_VSID_SHIFT;
|
|
|
|
}
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
hash = (vsid ^ ((eaddr & somask) >> pshift)) & kvm->arch.hpt_mask;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
avpn = slb_v & ~(somask >> 16); /* also includes B */
|
|
|
|
avpn |= (eaddr & somask) >> 16;
|
|
|
|
|
|
|
|
if (pshift >= 24)
|
|
|
|
avpn &= ~((1UL << (pshift - 16)) - 1);
|
|
|
|
else
|
|
|
|
avpn &= ~0x7fUL;
|
|
|
|
val |= avpn;
|
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (hash << 7));
|
|
|
|
|
|
|
|
for (i = 0; i < 16; i += 2) {
|
|
|
|
/* Read the PTE racily */
|
|
|
|
v = hpte[i] & ~HPTE_V_HVLOCK;
|
|
|
|
|
|
|
|
/* Check valid/absent, hash, segment size and AVPN */
|
|
|
|
if (!(v & valid) || (v & mask) != val)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/* Lock the PTE and read it under the lock */
|
|
|
|
while (!try_lock_hpte(&hpte[i], HPTE_V_HVLOCK))
|
|
|
|
cpu_relax();
|
|
|
|
v = hpte[i] & ~HPTE_V_HVLOCK;
|
|
|
|
r = hpte[i+1];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check the HPTE again, including large page size
|
|
|
|
* Since we don't currently allow any MPSS (mixed
|
|
|
|
* page-size segment) page sizes, it is sufficient
|
|
|
|
* to check against the actual page size.
|
|
|
|
*/
|
|
|
|
if ((v & valid) && (v & mask) == val &&
|
|
|
|
hpte_page_size(v, r) == (1ul << pshift))
|
|
|
|
/* Return with the HPTE still locked */
|
|
|
|
return (hash << 3) + (i >> 1);
|
|
|
|
|
|
|
|
/* Unlock and move on */
|
|
|
|
hpte[i] = v;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (val & HPTE_V_SECONDARY)
|
|
|
|
break;
|
|
|
|
val |= HPTE_V_SECONDARY;
|
KVM: PPC: Book3S HV: Make the guest hash table size configurable
This adds a new ioctl to enable userspace to control the size of the guest
hashed page table (HPT) and to clear it out when resetting the guest.
The KVM_PPC_ALLOCATE_HTAB ioctl is a VM ioctl and takes as its parameter
a pointer to a u32 containing the desired order of the HPT (log base 2
of the size in bytes), which is updated on successful return to the
actual order of the HPT which was allocated.
There must be no vcpus running at the time of this ioctl. To enforce
this, we now keep a count of the number of vcpus running in
kvm->arch.vcpus_running.
If the ioctl is called when a HPT has already been allocated, we don't
reallocate the HPT but just clear it out. We first clear the
kvm->arch.rma_setup_done flag, which has two effects: (a) since we hold
the kvm->lock mutex, it will prevent any vcpus from starting to run until
we're done, and (b) it means that the first vcpu to run after we're done
will re-establish the VRMA if necessary.
If userspace doesn't call this ioctl before running the first vcpu, the
kernel will allocate a default-sized HPT at that point. We do it then
rather than when creating the VM, as the code did previously, so that
userspace has a chance to do the ioctl if it wants.
When allocating the HPT, we can allocate either from the kernel page
allocator, or from the preallocated pool. If userspace is asking for
a different size from the preallocated HPTs, we first try to allocate
using the kernel page allocator. Then we try to allocate from the
preallocated pool, and then if that fails, we try allocating decreasing
sizes from the kernel page allocator, down to the minimum size allowed
(256kB). Note that the kernel page allocator limits allocations to
1 << CONFIG_FORCE_MAX_ZONEORDER pages, which by default corresponds to
16MB (on 64-bit powerpc, at least).
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix module compilation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-05-04 02:32:53 +00:00
|
|
|
hash = hash ^ kvm->arch.hpt_mask;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(kvmppc_hv_find_lock_hpte);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Called in real mode to check whether an HPTE not found fault
|
2011-12-12 12:38:51 +00:00
|
|
|
* is due to accessing a paged-out page or an emulated MMIO page,
|
|
|
|
* or if a protection fault is due to accessing a page that the
|
|
|
|
* guest wanted read/write access to but which we made read-only.
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
* Returns a possibly modified status (DSISR) value if not
|
|
|
|
* (i.e. pass the interrupt to the guest),
|
|
|
|
* -1 to pass the fault up to host kernel mode code, -2 to do that
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
* and also load the instruction word (for MMIO emulation),
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
* or 0 if we should make the guest retry the access.
|
|
|
|
*/
|
|
|
|
long kvmppc_hpte_hv_fault(struct kvm_vcpu *vcpu, unsigned long addr,
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
unsigned long slb_v, unsigned int status, bool data)
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
{
|
|
|
|
struct kvm *kvm = vcpu->kvm;
|
|
|
|
long int index;
|
|
|
|
unsigned long v, r, gr;
|
|
|
|
unsigned long *hpte;
|
|
|
|
unsigned long valid;
|
|
|
|
struct revmap_entry *rev;
|
|
|
|
unsigned long pp, key;
|
|
|
|
|
2011-12-12 12:38:51 +00:00
|
|
|
/* For protection fault, expect to find a valid HPTE */
|
|
|
|
valid = HPTE_V_VALID;
|
|
|
|
if (status & DSISR_NOHPTE)
|
|
|
|
valid |= HPTE_V_ABSENT;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
index = kvmppc_hv_find_lock_hpte(kvm, addr, slb_v, valid);
|
2011-12-12 12:38:51 +00:00
|
|
|
if (index < 0) {
|
|
|
|
if (status & DSISR_NOHPTE)
|
|
|
|
return status; /* there really was no HPTE */
|
|
|
|
return 0; /* for prot fault, HPTE disappeared */
|
|
|
|
}
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
hpte = (unsigned long *)(kvm->arch.hpt_virt + (index << 4));
|
|
|
|
v = hpte[0] & ~HPTE_V_HVLOCK;
|
|
|
|
r = hpte[1];
|
|
|
|
rev = real_vmalloc_addr(&kvm->arch.revmap[index]);
|
|
|
|
gr = rev->guest_rpte;
|
|
|
|
|
2011-12-15 02:01:10 +00:00
|
|
|
unlock_hpte(hpte, v);
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
|
2011-12-12 12:38:51 +00:00
|
|
|
/* For not found, if the HPTE is valid by now, retry the instruction */
|
|
|
|
if ((status & DSISR_NOHPTE) && (v & HPTE_V_VALID))
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* Check access permissions to the page */
|
|
|
|
pp = gr & (HPTE_R_PP0 | HPTE_R_PP);
|
|
|
|
key = (vcpu->arch.shregs.msr & MSR_PR) ? SLB_VSID_KP : SLB_VSID_KS;
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
status &= ~DSISR_NOHPTE; /* DSISR_NOHPTE == SRR1_ISI_NOPT */
|
|
|
|
if (!data) {
|
|
|
|
if (gr & (HPTE_R_N | HPTE_R_G))
|
|
|
|
return status | SRR1_ISI_N_OR_G;
|
|
|
|
if (!hpte_read_permission(pp, slb_v & key))
|
|
|
|
return status | SRR1_ISI_PROT;
|
|
|
|
} else if (status & DSISR_ISSTORE) {
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
/* check write permission */
|
|
|
|
if (!hpte_write_permission(pp, slb_v & key))
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
return status | DSISR_PROTFAULT;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
} else {
|
|
|
|
if (!hpte_read_permission(pp, slb_v & key))
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
return status | DSISR_PROTFAULT;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Check storage key, if applicable */
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
if (data && (vcpu->arch.shregs.msr & MSR_DR)) {
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
unsigned int perm = hpte_get_skey_perm(gr, vcpu->arch.amr);
|
|
|
|
if (status & DSISR_ISSTORE)
|
|
|
|
perm >>= 1;
|
|
|
|
if (perm & 1)
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
return status | DSISR_KEYFAULT;
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Save HPTE info for virtual-mode handler */
|
|
|
|
vcpu->arch.pgfault_addr = addr;
|
|
|
|
vcpu->arch.pgfault_index = index;
|
|
|
|
vcpu->arch.pgfault_hpte[0] = v;
|
|
|
|
vcpu->arch.pgfault_hpte[1] = r;
|
|
|
|
|
KVM: PPC: Implement MMU notifiers for Book3S HV guests
This adds the infrastructure to enable us to page out pages underneath
a Book3S HV guest, on processors that support virtualized partition
memory, that is, POWER7. Instead of pinning all the guest's pages,
we now look in the host userspace Linux page tables to find the
mapping for a given guest page. Then, if the userspace Linux PTE
gets invalidated, kvm_unmap_hva() gets called for that address, and
we replace all the guest HPTEs that refer to that page with absent
HPTEs, i.e. ones with the valid bit clear and the HPTE_V_ABSENT bit
set, which will cause an HDSI when the guest tries to access them.
Finally, the page fault handler is extended to reinstantiate the
guest HPTE when the guest tries to access a page which has been paged
out.
Since we can't intercept the guest DSI and ISI interrupts on PPC970,
we still have to pin all the guest pages on PPC970. We have a new flag,
kvm->arch.using_mmu_notifiers, that indicates whether we can page
guest pages out. If it is not set, the MMU notifier callbacks do
nothing and everything operates as before.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:38:05 +00:00
|
|
|
/* Check the storage key to see if it is possibly emulated MMIO */
|
|
|
|
if (data && (vcpu->arch.shregs.msr & MSR_IR) &&
|
|
|
|
(r & (HPTE_R_KEY_HI | HPTE_R_KEY_LO)) ==
|
|
|
|
(HPTE_R_KEY_HI | HPTE_R_KEY_LO))
|
KVM: PPC: Implement MMIO emulation support for Book3S HV guests
This provides the low-level support for MMIO emulation in Book3S HV
guests. When the guest tries to map a page which is not covered by
any memslot, that page is taken to be an MMIO emulation page. Instead
of inserting a valid HPTE, we insert an HPTE that has the valid bit
clear but another hypervisor software-use bit set, which we call
HPTE_V_ABSENT, to indicate that this is an absent page. An
absent page is treated much like a valid page as far as guest hcalls
(H_ENTER, H_REMOVE, H_READ etc.) are concerned, except of course that
an absent HPTE doesn't need to be invalidated with tlbie since it
was never valid as far as the hardware is concerned.
When the guest accesses a page for which there is an absent HPTE, it
will take a hypervisor data storage interrupt (HDSI) since we now set
the VPM1 bit in the LPCR. Our HDSI handler for HPTE-not-present faults
looks up the hash table and if it finds an absent HPTE mapping the
requested virtual address, will switch to kernel mode and handle the
fault in kvmppc_book3s_hv_page_fault(), which at present just calls
kvmppc_hv_emulate_mmio() to set up the MMIO emulation.
This is based on an earlier patch by Benjamin Herrenschmidt, but since
heavily reworked.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:36:37 +00:00
|
|
|
return -2; /* MMIO emulation - load instr word */
|
|
|
|
|
|
|
|
return -1; /* send fault up to host kernel mode */
|
|
|
|
}
|